4 2.6.9-55.0.9.EL (RHEL 4),
5 2.6.16.46-0.14 (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.
16 Description: Remove client patches from SLES 10 kernel.
17 Details : This causes SLES 10 clients to behave as patchless clients
18 even on a Lustre-patched (server) kernel.
20 Severity : enhancement
22 Description: self-adjustable client's lru lists
23 Details : use adaptive algorithm for managing client cached locks lru
24 lists according to current server load, other client's work
25 pattern, memory activities, etc. Both, server and client
26 side namespaces provide number of proc tunables for controlling
29 Severity : enhancement
31 Description: light-weight GSS support
32 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
33 and reduce performance overhead.
35 Severity : enhancement
37 Description: Linux keyring support
38 Details : Support using service of Linux keyring for Lustre GSS internal
39 context refresh/cache mechanism.
43 Description: Fix errors in lfs documentation
44 Details : Fixes man pages
48 Description: don't use GFP_* in generic Lustre code.
49 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
53 Description: obdclass is limited by single OBD_ALLOC(idarray)
54 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
58 Description: warnings with build patchless client with vanila 2.6.19 and up
59 Details : change old ctl_table style and replace ctl_table/ctl_table_header
60 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
64 Frequency : Only for SLES
65 Description: sanity_quota fail test_1
66 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
67 file, which makes TSTID[2] inunique.
71 Frequency : Only for SLES9
72 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
73 Details : Update target/ChangeLog/which_patch .
77 Frequency : Only for test_10 in sanity-quota.sh
78 Description: a bug in quota test code
79 Details : checking a test flag in a wrong place
83 Frequency : Only for test_13 in sanity-quota.sh
84 Description: a bug in quota test script
85 Details : shouldn't check overfull allocated quota
89 Description: jbd statistics
90 Details : Port older jbd statistics patch for sles10
92 Severity : enhancement
94 Description: Add printing inode info into message about error in writepage.
99 Description: reply_lock_interpret crash due to race with it and lock cancel.
100 Details : Do not replay locks that are being cancelled. Do not reference
101 locks by their address during replay, just by their handle.
105 Frequency : When flocks are used.
106 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
107 Details : Flock locks might destroy just granted lock if it could be merged
108 with another existing flock, this is done in completion handler,
109 so teach ldlm_cli_enquque_fini that this is a valid case for
115 Description: Oops in read and write path when failing to allocate lock.
116 Details : Check if lock allocation failed and return error back.
120 Description: lstripe command fails for valid OST index
121 Details : The stripe offset is compared to lov->desc.ld_tgt_count
122 instead of lov->desc.ld_active_tgt_count.
126 Description: sanity.sh failed test 103
127 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
132 Description: df doesn't work properly if diskfs blocksize != 4K
133 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
137 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
138 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
142 Description: obd_health_check_timeout too short
143 Details : set obd_health_check_timeout as 1.5x of obd_timeout
147 Description: llapi_file_create() does not allow some changes
148 Details : add llapi_file_open() that allows specifying the mode and
149 open flags, and also returns an open file handle.
153 Description: merge and cleanup kernel patches.
154 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
158 Description: Add journal checksum support.(Kernel part)
159 Details : The journal checksum feature adds two new flags i.e
160 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
161 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
162 indicates that the commit block contains the checksum for
163 the blocks described by the descriptor blocks. Now commit
164 record can be sent to disk without waiting for descriptor
165 blocks to be written to disk. This behavior is controlled
166 using JBD2_FEATURE_ASYNC_COMMIT flag.
170 Description: OSS needs mutliple precreate threads
171 Details : Add ability to start more than one create thread per OSS.
175 Description: Sanity.sh test_65k failed: setstripe should have succeeded
176 Details : execute lfs setstripe on client
180 Description: mds_obd_create error creating tmp object
181 Details : When the user sets quota on root, llog will be affected and can't
182 create files and write files.
185 Frequency : Always on ia64 patchless client, and possibly others.
187 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
188 Details : This allows the patchless client to be loaded on architectures
193 Description: RedHat Update kernel for RHEL5
194 Details : Add the kernel config file for RHEL5.
198 Description: RedHat Update kernel for RHEL5
199 Details : Modify the kernel config file more closer RHEL5.
203 Description: Build failure against Centos5 (RHEL5)
204 Details : Define PAGE_SIZE when it isn't present.
208 Description: test_7 on sanity-quota.sh can't be used on separate machines
209 Details : add facet to handle it
213 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
214 Details : with NFS, the anon dentry's parent was set to itself in
215 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
216 corresponding dentry other than use rec->ur_name.
218 Severity : enhancement
220 Description: lfs setstripe enhancement
221 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
225 Description: testing performance impact of enabling checksumming
226 Details : enable checksum by default, allow --disable-checksum
227 configure option and "-o nochecksum" mount option
231 Description: lustre support for RHEL5
232 Details : Add support for RHEL5.
236 Description: Client eviction due to failover config
237 Details : after a connection loss, the lustre client should attempt to
238 reconnect to the last active server first before trying the
239 other potential connections.
243 Description: when mds and osts use different quota unit(32bit and 64bit),
244 quota will be released repeatly.
245 Details : void sending multiple quota reqs to mds, which will keep the
246 status between the reqs.
250 Description: osts not allocated evenly to files
251 Details : change the condition to increase offset_idx
254 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
256 Description: Data corruption for OSTs that are formatted larger than 2TB
258 Details : When generating the bio request for lustre file writes the
259 sector number would overflow a temporary variable before being
260 used for the IO. The data reads correctly from Lustre (which
261 will overflow in a similar manner) but other file data or
262 filesystem metadata may be corrupted in some cases.
266 Description: Tune the kernel for good SCSI performance.
267 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
268 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
273 Description: rewrite ext2-derived code in obdclass/uuid.c
274 Details : rewrite inherited code (uuid parsing code from ext2 utils)
275 from scratch preserving functionality.
279 Description: "lfs find -obd UUID" prints directories
280 Details : "lfs find -obd UUID" will return all directory names instead
281 of just file names. It is incorrect because the directories
282 do not reside on the OSTs.
286 Description: use i_size_read and i_size_write in 2.6 port
287 Details : replace inode->i_size access with i_size_read/write()
290 Frequency : when removing large files
292 Description: scheduling issue during removal of large Lustre files
293 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
294 It causes scheduling issues when removing large files (17TB in the
298 Frequency : only with liblustre clients on XT3
300 Description: evictions taking too long
301 Details : allow llrd to evict clients directly on OSTs
304 Frequency : only on ppc
306 Description: /proc/fs/lustre/devices broken on ppc
307 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
308 We should make sure the type of 'index' is loff_t and then cast
309 explicitly as needed below. Do not assign an explicitly cast
313 Frequency : only for rhel5
315 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
316 Details : Modify the target file & which_kernel.
318 Severity : enhancement
320 Description: omit set fsid for export NFS
321 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
322 as FS_REQUIRES_DEV for avoid problems with generate fsid.
326 Frequency : Always, for kernels after 2.6.16
327 Description: Fix warning idr_remove called for id=.. which is not
329 Details : Last kernels save old s_dev before kill super and not allow
330 to restore from callback - restore it before call kill_anon_super.
334 Description: buffer overruns could theoretically occur
335 Details : llapi_semantic_traverse() modifies the "path" argument by
336 appending values to the end of the origin string, and a
337 overrun may occur. Adding buffer overrun check in liblustreapi.
341 Description: client sometimes tried revalidate root dentry.
342 Details : not revalidate dentry if it root dentry.
346 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
347 Details : Modify vm-tunables-rhel4.patch.
351 Description: Kernel config for 2.6.18-vanilla.
352 Details : Modify targets/2.6-vanilla.target.in.
353 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
354 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
355 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
356 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
360 Description: improve handling recoverable errors
361 Details : if request processig with error which can be recoverable on server
362 request should be resend, otherwise page released from cache and
367 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
368 Details : Modify target file & which_patch.
369 A flaw was found in the IA32 system call emulation provided
370 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
371 value could be stored in the %RAX register, which could trigger an
372 out-of-bounds system call table access. An untrusted local user
373 could exploit this flaw to run code in the kernel
374 (ie a root privilege escalation). (CVE-2007-4573).
378 Description: Remove 2.4 kernel checks lustre 1.8
379 Details : Remove 2.5.0 check from quota_check.c, quota_ctl.c,
380 quota_interface.c in quota subsystem.
381 Remove 2.4 kernel check from obd_mount.c, linux/linux-sysctl.c in
383 Remove lock_24kernel from fsfilt_ext3.c fsfilt_reiserfs.c
384 lvfs_linux.c upcall_cache.c in lvfs subsystem.
388 Description: O_DIRECT bypasses client statistics.
389 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
390 read_ahead_stats were updated. copied the stats section from
391 osc_send_oap_rpc() into async_internal().
395 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
399 Description: change order of libsysio includes
400 Details : '#include sysio.h' should always come before '#include xtio.h'
404 Description: handle "serious error: objid * already exists" more gracefully
405 Details : If LAST_ID value on disk is smaller than the objects existing in
406 the O/0/d* directories, it indicates disk corruption and causes an
407 LBUG(). If the object is 0-length, then we should use the existing
408 object. This will help to avoid a full fsck in most cases.
410 Severity : enhancement
412 Description: adapt the lustre_config script to support the upgrade case
413 Details : Add "-u" option for lustre_config script to support upgrading 1.4
414 server targets to 1.6 in parallel.
418 Description: grant fixes to avoid grant_space > avail_space
419 Details : In filter_check_grant, for non_grant cache write, we should
420 check the left space by if (*left > ungranted + bytes), instead
421 of (*left > ungranted), because only we are sure the left space
422 is enough for another "bytes", then the ungrant space should be
423 increase. In client, we should update cl_avail_grant only there
424 is OBD_MD_FLGRANT in the reply.
428 Description: Update RHEL 4 kernel to fix local root privilege escalation.
429 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
430 described in CVE-2007-4573. This problem could allow untrusted
431 local users to gain root access.
435 Description: Remove 2.4 kernel checks lustre 1.8
436 Details : Remove 2.5.0 check from dcache.c dir.c file.c llite_internal.h
437 llite_lib.c llite_mmap.c lproc_llite.c namei.c rw.c symlink.c
439 Remove 2.5.0 check from handler.c lproc_mds.c mds_fs.c
440 mds_internal.h mds_lib.c mds_open.c mds_reint.c mds_xattr.c
443 --------------------------------------------------------------------------------
445 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
447 * Support for kernels:
448 2.6.5-7.283 (SLES 9),
449 2.6.9-55.EL (RHEL 4),
450 2.6.16.46-0.14 (SLES 10),
451 2.6.18.8 vanilla (kernel.org)
452 * Client support for unpatched kernels:
453 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
454 2.6.16 - 2.6.22 vanilla (kernel.org)
455 * Due to recently discovered recovery problems, we do not recommend
456 using patchless RHEL 4 clients with this or any earlier release.
457 * Recommended e2fsprogs version: 1.39.cfs8
458 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
459 * Starting with this release, the ldiskfs backing filesystem required
460 by Lustre is now in its own package, lustre-ldiskfs. This package
461 should be installed. It is versioned separately from Lustre and
462 may be released separately in future.
466 Description: block reactivating mgc import until all deactivates complete
467 Details : Fix race when failing back MDT/MGS to itself (testing)
469 Severity : enhancement
471 Description: add optional extra BUILD_VERSION info
472 Details : add a new environment variable (namely LUSTRE_VERS) which allows
473 to override the lustre version.
476 Frequency : 2.6.18 servers only
478 Description: ll_kern_mount() doesn't release the module reference
479 Details : The ldiskfs module reference count never drops down to 0
480 because ll_kern_mount() doesn't release the module reference.
485 Description: server LBUG when using old ost_num_threads parameter
486 Details : Accept the old ost_num_threads parameter but warn that it
487 is deprecated, and fix an off-by-one error that caused an LBUG.
492 Description: Transient SCSI error results in persistent IO issue
493 Details : iobuf->dr_error is not reinitialized to 0 between two
497 Frequency : sometimes when underlying device returns I/O errors
499 Description: OSTs not going read-only during write failures
500 Details : OSTs are not remounted read-only when the journal commit threads
501 get I/O errors because fsfilt_ext3 calls journal_start/stop()
502 instead of the ext3 wrappers.
506 Description: poor connect scaling with increasing client count
507 Details : Don't run filter_grant_sanity_check for more than 100 exports
508 to improve scaling for large numbers of clients.
511 Frequency : SLES10 only
513 Description: sanity-quota.sh quotacheck failed: rc = -22
514 Details : Quotas cannot be enabled on SLES10.
517 Frequency : liblustre clients only
519 Description: getdirentries does not give error when run on compute nodes
520 Details : getdirentries does not fail when the size specified as an argument
521 is too small to contain at least one entry
523 Severity : enhancement
525 Description: Add LNET router traceability for debug purposes
526 Details : If a checksum failure occurs with a router as part of the
527 IO path, the NID of the last router that forwarded the bulk data
528 is printed so it can be identified.
533 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
534 Details : Due to a race condition, liblustre clients were occasionally
537 Severity : enhancement
539 Description: lfs setstripe use optional parameters instead of postional
542 Severity : enhancement
544 Description: Nanosecond timestamp support for ldiskfs
545 Details : The on-disk ldiskfs filesystem has added support for nanosecond
546 resolution timestamps. There is not yet support for this at
547 the Lustre filesystem level.
550 Frequency : during server recovery
552 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
553 Details : request with rq_no_resend flag not awake l_wait_event if they get a
557 Frequency : nfs export on patchless client
559 Description: connectathon hang when test nfs export over patchless client
560 Details : Disconnected dentry cannot be found with lookup, so we do not need
561 to unhash it or make it invalid
564 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
565 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
566 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
571 Description: open req refcounting wrong on reconnect
572 Details : If reconnect happened between getting open reply from server and
573 call to mdc_set_replay_data in ll_file_open, we will schedule
574 replay for unreferenced request that we are about to free.
575 Subsequent close will crash in variety of ways.
576 Check that request is still eligible for replay in
577 mdc_set_replay_data().
582 Description: disable writes to filesystem when reading health_check file
583 Details : the default for reading the health_check proc file has changed
584 to NOT do a journal transaction and write to disk, because this
585 can cause reads of the /proc file to hang and block HA state
586 checking on a healthy but otherwise heavily loaded system. It
587 is possible to return to the previous behaviour during configure
588 with --enable-health-write.
590 Severity : enhancement
592 Description: 64-bit inode version
593 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
594 to the inode. This will be required for version-based recovery.
599 Description: MDS fails to start if a duplicate client export is detected
600 Details : in some rare cases it was possible for a client to connect to
601 an MDS multiple times. Upon recovery the MDS would detect this
602 and fail during startup. Handle this more gracefully.
604 Severity : enhancement
606 Description: Add -o localflock option to simulate old noflock
608 Details : This will achieve local-only flock/fcntl locks
614 Description: log_commit_thread vs filter_destroy race leads to crash
615 Details : Take import reference before releasing llog record semaphore
620 Description: Wrong request locking in request set processing
621 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
622 imp_delayed_list, in this place should be used imp_lock.
625 Frequency : when reconnection
627 Description: Grant Leak when osc reconnect to OST
628 Details : When osc reconnect ost, OST(filter) should check whether it
629 should grant more space to client by comparing fed_grant and
630 cl_avail_grant, and return the granted space to client instead
631 of "new granted" space, because client will call osc_init_grant
632 to update the client grant space info.
635 Frequency : when client reconnect to OST
637 Description: Grant Leak when osc do resend and replay bulk write
638 Details : When osc reconnect to OST, OST(filter)should clear grant info of
639 bulk write request, because the grant info while be sync between
640 OSC and OST when reconnect, and we should ignore the grant info
641 these of resend/replay write req.
646 Description: Grant space more than avaiable left space sometimes.
647 Details : When then OST is about to be full, if two bulk writing from
648 different clients came to OST. Accord the avaliable space of the
649 OST, the first req should be permitted, and the second one
650 should be denied by ENOSPC. But if the seconde arrived before
651 the first one is commited. The OST might wrongly permit second
652 writing, which will cause grant space > avaiable space.
655 Frequency : when client is evicted
657 Description: Grant might be wrongly erased when osc is evicted by OST
658 Details : when the import is evicted by server, it will fork another
659 thread ptlrpc_invalidate_import_thread to invalidate the
660 import, where the grant will be set to 0. While the original
661 thread will update the grant it got when connecting. So if
662 the former happened latter, the grant will be wrongly errased
663 because of this race.
668 Description: Checking Stale with correct fid
669 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
670 to check whether it is stale, because sometimes, we want the
671 enqueue happened anyway, and op_data.fid2 will not be initialized.
673 Severity : enhancement
675 Description: update patchless client
676 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
679 Frequency : only with 2.4 kernel
681 Description: random memory corruption
682 Details : size of struct ll_inode_info is to big for union inode.u and this
683 can be cause of random memory corruption.
688 Description: Memory leak in recovery
689 Details : Lov_mds_md was not free in an error handler in mds_create_object.
690 It should also check obd_fail before fsfilt_start, otherwise if
691 fsfilt_start return -EROFS,(failover mds during mds recovery).
692 then the req will return with repmsg->transno = 0 and rc = EROFS.
693 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
694 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
695 be freed no matter whether fsfilt_commit success or not.
698 Frequency : only with huge count clients
700 Description: Prevents from taking the superblock lock in llap_from_page for
702 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
703 for llap_from_page call in ll_removepage prevents from taking the
704 superblock lock for a soon died page.
709 Description: Not check open intent error before release open handle
710 Details : in some rare cases, the open intent error is not checked before
711 release open handle, which may cause
712 ASSERTION(open_req->rq_transno != 0), because it tries to release
713 the failed open handle.
718 Description: Set cat log bitmap only after create log success.
719 Details : in some rare cases, the cat log bitmap is set too early. and it
720 should be set only after create log success.
724 Description: Accessing a block bevice can re-enable I/O when Lustre is
725 tearing down a device.
726 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
731 Description: service threads may hog cpus when there are a lot of requests
733 Details : Insert cond_resched to give other threads a chance to use some of
739 Description: the cat log was not initialized in recovery
740 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
741 unlink log on mds will not be initialized until mds post
742 recovery. And also in mds post recovery, the unlink log will
743 initialization will be done asynchronausly, so there will be race
744 between add unlink log and unlink log initialization.
748 Description: brw_stats were being printed incorrectly
749 Details : brw_stats were being printed as log2 but all of them were not
750 recorded as log2. Also remove some code duplication arising from
751 filter_tally_{read,write}.
755 Frequency : rare, only in recovery.
756 Description: ASSERTION(req->rq_type != LI_POISON) failed
757 Details : imp_lock should be held while iterating over imp_sending_list for
758 prevent destroy request after get timeout in ptlrpc_queue_wait.
762 Description: replay-single.sh test 52 fails
763 Details : A lock's skiplist need to be cleanup when it being unlinked
764 from its resource list.
768 Description: Short directio read returns full requested size rather than
770 Details : Direct I/O operations should return actual amount of bytes
771 transferred rather than requested size.
773 Severity : enhancement
775 Description: metadata RPC reduction (e.g. for rm performance)
776 Details : decrease the amount of synchronous RPC between clients and servers
777 by canceling conflicing lock before the operation on the client side
778 and packing thier handles into the main operation RPC to server.
780 Severity : enhancement
782 Description: Async OSC create to avoid the blocking unnecessarily.
783 Details : If a OST has no remain object, system will block on the creating
784 when need to create a new object on this OST. Now, ways use
785 pre-created objects when available, instead of blocking on an
786 empty osc while others are not empty. If we must block, we block
787 for the shortest possible period of time.
791 Description: improve handling recoverable errors
792 Details : if request processig with error which can be recoverable on server
793 request should be resend, otherwise page released from cache and
796 Severity : enhancement
798 Description: refine locking for avoid write wrong info into lov_objid file
799 Details : fix possible races with add new target and write/update data in
802 --------------------------------------------------------------------------------
804 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
809 Frequency : on some architectures
811 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
812 Details : Uninitialized flags sometimes cause configuration commands to
816 Frequency : patchless clients only
818 Description: missing __iget() symbol export
819 Details : The __iget() symbol export is missing. To avoid the need for
820 this on patchless clients the deathrow inode reaper is turned
821 off, and we depend on the VM to clean up old inodes. This
822 dependency was during via the fix for bug 12181.
824 --------------------------------------------------------------------------------
826 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
828 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
829 INTEROPERATE with older versions automatically. In many cases a
830 special upgrade step is needed. Please read the
831 user documentation before upgrading any part of a live system.
832 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
833 version will not interoperate with 1.6 betas before beta5 (1.5.95).
834 * WARNING: Lustre configuration and startup changes are required with
835 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
841 Severity : enhancement
843 Description: MountConf
844 Details : Lustre configuration is now managed via mkfs and mount
845 commands instead of lmc and lconf. New obd types (MGS, MGC)
846 are added for dynamic configuration management. See
847 https://mail.clusterfs.com/wikis/lustre/MountConf for
850 Severity : enhancement
852 Description: dynamic OST addition
853 Details : OSTs can now be added to a live filesystem
855 Severity : enhancement
857 Description: startup order invariance
858 Details : MDTs and OSTs can be started in any order. Clients only
859 require the MDT to complete startup.
861 Severity : enhancement
863 Description: parallel, asynchronous orphan cleanup
864 Details : orphan cleanup is now performed in separate threads for each
865 OST, allowing parallel non-blocking operation.
867 Severity : enhancement
869 Description: optimized stripe assignment
870 Details : stripe assignments are now made based on ost space available,
871 ost previous usage, and OSS previous usage, in order to try
872 to optimize storage space and networking resources.
874 Severity : enhancement
876 Description: Permanently set tunables
877 Details : All writable /proc/fs/lustre tunables can now be permanently
878 set on a per-server basis, at mkfs time or on a live
881 Severity : enhancement
883 Description: Lustre message v2
884 Details : Add lustre message format v2.
886 Severity : enhancement
888 Description: client OST exclusion list
889 Details : Clients can be started with a list of OSTs that should be
890 declared "inactive" for known non-responsive OSTs.
894 Description: SPEC SFS validation failure on NFS v2 over lustre.
895 Details : Changes the blocksize for regular files to be 2x RPC size,
896 and not depend on stripe size.
898 Severity : enhancement
900 Description: Multiple MD RPCs in flight.
901 Details : Further unserialise some read-only MDS RPCs - learn about intents.
902 To avoid overly-overloading MDS, introduce a limit on number of
903 MDS RPCs in flight for a single client and add /proc controls
904 to adjust this limit.
906 Severity : enhancement
908 Description: client read/write statistics
909 Details : Add client read/write call usage stats for performance
910 analysis of user processes.
911 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
912 file access. extents_stats shows chunk size distribution.
913 extents_stats_per_process show chunk size distribution per
916 Severity : enhancement
918 Description: mds statistics
919 Details : Add detailed mds operations statistics in
920 /proc/fs/lustre/mds/*/stats.
924 Description: Failure of copying files with lustre special EAs.
925 Details : Client side always return success for setxattr call for lustre
926 special xattr (currently only "trusted.lov").
931 Description: Refcount LNET uuids
932 Details : The global LNET uuid list grew linearly with every startup;
933 refcount repeated list entries instead of always adding to
936 Severity : enhancement
938 Description: Dynamic service threads
939 Details : Within a small range, start extra service threads
940 automatically when the request queue builds up.
943 Frequency : mixed-endian client/server environments
945 Description: mixed-endian crashes
946 Details : The new msg_v2 system had some failures in mixed-endian
949 Severity : enhancement
951 Description: Easy OST removal
952 Details : OSTs can be permanently deactivated with e.g. 'lctl
953 conf_param lustre-OST0001.osc.active=0'
955 Severity : enhancement
957 Description: MGS proc entries
958 Details : Added basic proc entries for the MGS showing what filesystems
961 Severity : enhancement
963 Description: provide MGS failover
964 Details : Added config lock reacquisition after MGS server failover.
966 Severity : enhancement
968 Description: add Linux 2.4 support
969 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
973 Description: a large application tries to do I/O to the same resource and dies
975 Details : Check the req->rq_arrival time after the call to
976 ost_brw_lock_get(), but before we do anything about
977 processing it & sending the BULK transfer request. This
978 should help move old stale pending locks off the queue as
979 quickly as obd_timeout.
982 Frequency : when an incorrect nid is specified during startup
984 Description: ptlrpc connect to non-existant node causes kernel crash
985 Details : LNET can't be re-entered from an event callback, which
986 happened when we expire a message after the export has been
987 cleaned up. Instead, hand the zombie cleanup off to another
990 Severity : enhancement
992 Description: plain/inodebits lock performance improvement
993 Details : Grouping plain/inodebits in granted list by their request modes
994 and bits policy, thus improving the performance of search through
998 Frequency : only if OST filesystem is corrupted
1000 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1001 Details : for a short time RPCs with bulk IO are in the replay list,
1002 but replay of bulk IOs is unimplemented. If the OST filesystem
1003 is corrupted due to disk cache incoherency and then replay is
1004 started it is possible to trip an assertion. Avoid putting
1005 committed RPCs into the replay list at all to avoid this issue.
1008 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1010 Description: System hang on startup
1011 Details : This bug allowed the liblustre (e.g. catamount) client to
1012 return to the app before handling all startup RPCs. This
1013 could leave the node unresponsive to lustre network traffic
1014 and manifested as a server ptllnd timeout.
1016 Severity : enhancement
1018 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1019 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1020 boolean to control whether to print peer debug info when a
1021 client's RPC times out.
1024 Frequency : only for kernels with patches from Lustre below 1.4.3
1026 Description: Remove old rdonly API
1027 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1030 Frequency : only for devices with external journals
1032 Description: Set external device read-only also
1033 Details : During a commanded failover stop, we set the disk device
1034 read-only while the server shuts down. We now also set any
1035 external journal device read-only at the same time.
1038 Frequency : when upgrading from 1.4 while trying to change parameters
1040 Description: The wrong (new) MDC name was used when setting parameters for
1041 upgraded MDT's. Also allows changing of OSC (and MDC)
1042 parameters if --writeconf is specified at tunefs upgrade time.
1045 Frequency : when setting specific ost indicies
1047 Description: QOS code breaks on skipped indicies
1048 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1049 created with --index need not be sequential.
1051 Severity : enhancement
1053 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1054 Details : The uninit_groups feature works in conjunction with the kernel
1055 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1056 up the pass1 processing of e2fsck. This is a read-only feature
1057 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1058 mount filesystems that have had this feature enabled.
1060 Severity : enhancement
1062 Description: Improve multi-block allocation algorithm to avoid fragmentation
1063 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1064 allocation locality and avoid filesystem fragmentation.
1066 ------------------------------------------------------------------------------
1068 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1070 * Support for kernels:
1071 2.4.21-47.0.1.EL (RHEL 3)
1072 2.6.5-7.283 (SLES 9)
1073 2.6.9-42.0.10.EL (RHEL 4)
1074 2.6.12.6 vanilla (kernel.org)
1075 2.6.16.27-0.9 (SLES 10)
1076 * Recommended e2fsprogs version: 1.39.cfs5
1078 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1082 Frequency : occasional, depends on client load and configuration
1083 Bugzilla : 12181, 12203
1084 Description: data loss for recently-modified files
1086 Details : In some cases it is possible that recently written or created
1087 files may not be written to disk in a timely manner (this should
1088 normally be within 30s unless client IO load is very high).
1089 The problem appears as zero-length files or files that are a
1090 multiple of 1MB in size after a client crash or client eviction
1091 that are missing data at the end of the file.
1093 This problem is more likely to be hit on clients where files are
1094 repeatedly created and unlinked in the same directory, clients
1095 have a large amount of RAM, have many CPUs, the filesystem has
1096 many OSTs, the clients are rebooted frequently, and/or the files
1097 are not accessed by other nodes after being written.
1099 The presence of the problem can be detected by looking at
1100 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1101 smaller than the second (nr_unused) then dirty files will not
1102 be flushed automatically to disk. "sync; sleep 10" should be
1103 run several times on the node before unmounting it to update
1104 Lustre (this is also safe to run on nodes without this problem).
1106 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1107 that can cause this same problem, so customers using that kernel
1108 also need to update the kernel in addition to Lustre. In order
1109 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1111 It is normal that files written just before a client crash (less
1112 than 30s) may not yet have been flushed to disk, even for local
1116 Frequency : frequent on thin XT3 nodes
1118 Description: UUID collision on thin XT3 Linux nodes
1119 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1120 randomly, since we relied on an insufficiently-seeded PRNG.
1125 Description: OSS hangs after "All ost request buffers busy"
1126 Details : A deadlock between quota and journal operations caused OSS
1127 hangs after printing "All ost request buffers busy."
1130 Frequency : always on liblustre builds
1132 Description: Cleanup compiler warnings on liblustre
1135 Frequency : always on liblustre builds on XT3
1137 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1141 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1142 > 32000 subdirectories
1143 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1144 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1145 subdirectory count crosses 32000. This will aid e2fsck to
1146 correctly handle more than 32000 subdirectories.
1149 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1151 Description: System hang on startup
1152 Details : This bug allowed the liblustre (e.g. catamount) client to
1153 return to the app before handling all startup RPCs. This
1154 could leave the node unresponsive to lustre network traffic
1155 and manifested as a server ptllnd timeout.
1157 Severity : enhancement
1159 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1160 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1161 boolean to control whether to print peer debug info when a
1162 client's RPC times out.
1167 Description: make O_SYNC working on 2.6 kernels
1168 Details : 2.6 kernels use different method for mark pages for write,
1169 so need add a code to lustre for O_SYNC work.
1174 Description: Failure to close file and release space on NFS
1175 Details : Put inode details into lock acquired in ll_intent_file_open.
1176 Use mdc_intent_lock in ll_intent_open to properly
1177 detect all kind of errors unhandled by mdc_enqueue.
1182 Description: proc file read during shutdown sometimes raced obd removal,
1184 Details : Add lock to prevent obd access after proc file removal.
1187 Frequency : Only for files larger than 4GB on 32-bit clients.
1189 Description: improperly doing page alignment of locks
1190 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1191 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1196 Description: RPCs being resent when they shouldn't be
1197 Details : Some RPCs that should not be resent are being resent. This
1198 can cause inconsistencies in the RPC state machine. Do not
1199 resend such requests.
1202 Frequency : rare, only with NFS export
1204 Description: Crash on NFS re-export node
1205 Details : under very unusual load conditions an assertion is hit in
1206 ll_intent_file_open()
1209 Frequency : only if OST filesystem is corrupted
1211 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1212 Details : for a short time RPCs with bulk IO are in the replay list,
1213 but replay of bulk IOs is unimplemented. If the OST filesystem
1214 is corrupted due to disk cache incoherency and then replay is
1215 started it is possible to trip an assertion. Avoid putting
1216 committed RPCs into the replay list at all to avoid this issue.
1221 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1222 Details : Large single O_DIRECT read and write calls can fail to allocate
1223 a sufficiently large buffer to process the request. In case of
1224 allocation failure the allocation is retried with a smaller
1225 buffer and broken into smaller requests.
1227 Severity : enhancement
1229 Description: Add -o localflock option to simulate old noflock behaviour.
1230 Details : This will achieve local-only flock/fcntl locks coherentness.
1235 Description: versioning check is incomplete
1236 Details : Checking the version difference of client vs. server, report
1237 error if the gap is too big.
1242 Description: add support PG_writeback bit
1243 Details : add support for PG_writeback bit for Lustre, for more carefull
1244 work with page cache in 2.6 kernel. This also fix some deadlocks
1245 and remove hack for work O_SYNC with 2.6 kernel.
1247 Severity : enhancement
1249 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1250 Details : The uninit_groups feature works in conjunction with the kernel
1251 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1252 up the pass1 processing of e2fsck. This is a read-only feature
1253 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1254 mount filesystems that have had this feature enabled.
1256 Severity : enhancement
1258 Description: Improve multi-block allocation algorithm to avoid fragmentation
1259 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1260 allocation locality and avoid filesystem fragmentation.
1262 ------------------------------------------------------------------------------
1264 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1266 * Support for kernels:
1267 2.6.9-42.0.3.EL (RHEL 4)
1268 2.6.5-7.276 (SLES 9)
1269 2.4.21-47.0.1.EL (RHEL 3)
1270 2.6.12.6 vanilla (kernel.org)
1271 2.6.16.21-0.8 (SLES10)
1272 * Recommended e2fsprogs version: 1.39.cfs2-0
1274 * The backwards-compatible /proc/sys/portals symlink has been removed
1275 in this release. Before upgrading, please ensure that you change
1276 any configuration scripts or /etc/sysctl.conf files that access
1277 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1278 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1279 in advance of the upgrade on any system running Lustre 1.4.6 or
1280 newer, since /proc/sys/lnet was added in that version.
1281 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1285 Frequency : only when quota is used
1287 Description: avoid scanning export list for quota master
1288 Details : Change the algorithms to avoid scanning export list in order
1289 to improve the efficiency.
1292 Frequency : MDS failover only, very rarely
1294 Description: "went back in time" messages on mds failover
1295 Details : The greatest transno may be lost when the current operation
1296 finishes with an error (transno==0) and the client's last_rcvd
1297 record is over-written. Save the greatest transno in the
1298 mds_last_transno for this case.
1301 Frequency : always for specific kernels and striping counts
1303 Description: client may get "Matching packet too big" without ACL support
1304 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1305 when trying to access files in certain configurations. The
1306 clients should in fact be denied when mounting because they do
1307 not understand ACLs.
1310 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1312 Description: many clients connecting with IO in progress causes connect timeouts
1313 Details : Avoid synchronous journal commits to avoid delays caused by many
1314 clients connecting/disconnecting when bulk IO is in progress.
1315 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1316 OST_IO_PORTAL to avoid delays behind potentially many pending
1320 Frequency : occasionally with multiple writers to a single file
1322 Description: shared writes to file may result in wrong size reported by stat()
1323 Details : Allow growing of kms when extent lock is cancelled
1326 Frequency : always with random mmap IO to multi-striped file
1328 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1329 Details : Only if the hole is at the end of OST object so that kms is too
1330 small. Fix is to increase kms accordingly in ll_nopage.
1333 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1335 Description: writes to a missing object would leak memory on the OST
1336 Details : If there is an inconsistency between the MDS and OST filesystems,
1337 such that the MDS references an object that doesn't exist, writes
1338 to that object will leak memory due to incorrect cleanup in the
1339 error handling path, eventually running out of memory on the OST.
1344 Description: Creating too long symlink causes lustre errors
1345 Details : Check symlink and name lengths before sending requests to MDS.
1348 Frequency : only if flock is enabled (not on by default)
1350 Description: posix locks not released on fd closure on 2.6.9+
1351 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1352 kernels, this caused such locks not to be released on fd close and
1353 then assertions on fs unmount about still used locks.
1356 Frequency : MDS failover only, very rarely
1358 Description: clients may get ASSERTION(granted_lock != NULL)
1359 Details : When request was taking a long time, and a client was resending
1360 a getattr by name lock request. The were multiple lock requests
1361 with the same client lock handle and
1362 mds_getattr_name->fixup_handle_for_resent_request found one of the
1363 lock handles but later failed with ASSERTION(granted_lock != NULL).
1368 Description: handle->h_buffer_credits > 0, assertion failure
1369 Details : h_buffer_credits is zero after truncate, causing assertion
1370 failure. This patch extends the transaction or creates a new
1374 Frequency : NFS re-export or patchless client
1375 Bugzilla : 11179, 10796
1376 Description: Crash on NFS re-export node (__d_move)
1377 Details : We do not want to hash the dentry if we don't have a lock.
1378 But if this dentry is later used in d_move, we'd hit uninitialised
1379 list head d_hash, so we just do this to init d_hash field but
1380 leave dentry unhashed.
1383 Frequency : NFS re-export or patchless client
1385 Description: NFS exports has problem with symbolic link
1386 Details : lustre client didn't properly install dentry when re-exported
1387 to NFS or running patchless client.
1390 Frequency : NFS re-export or patchless client
1392 Description: Various nfs/patchless fixes.
1393 Details : fixes reuse disconected alias for lookup process - this fixes
1394 warning "find_exported_dentry: npd != pd",
1395 fix permission error with open files at nfs.
1396 fix apply umask when do revalidate.
1399 Frequency : occasional
1401 Description: Crash on NFS re-export node
1402 Details : calling clear_page() on the wrong pointer triggered oops in
1403 generic_mapping_read().
1406 Frequency : rarely, using O_DIRECT IO
1408 Description: unaligned directio crashes client with LASSERT
1409 Details : check for unaligned buffers before trying any requests.
1412 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1414 Description: stale data returned from RAID cache
1415 Details : If only a small amount of IO is done to the RAID device before
1416 reading it again it is possible to get stale data from the RAID
1417 cache instead of reading it from disk.
1420 Frequency : always for sles10 kernel
1422 Description: sles10 support
1423 Details : ll_follow_link: compile fixes and using of nd_set_link
1424 under newer kernels.
1427 Frequency : depends on arch, kernel and compiler version, always on sles10
1430 Description: recursive or deep enough symlinks cause stack overflow
1431 Details : getting rid of large stack-allocated variable in
1435 Frequency : depends on hardware
1437 Description: lustre write performance loss in the SLES10 kernel
1438 Details : the performance loss is caused by using of write barriers in the
1439 ext3 code. The SLES10 kernel turns barrier support on by
1440 default. The fix is to undo that change for ldiskfs.
1442 ------------------------------------------------------------------------------
1444 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1446 * Support for kernels:
1447 2.6.9-42.0.3EL (RHEL 4)
1448 2.6.5-7.276 (SLES 9)
1449 2.4.21-47.0.1.EL (RHEL 3)
1450 2.6.12.6 vanilla (kernel.org)
1454 Frequency : quota enabled and large files being deleted
1456 Description: releasing more than 4GB of quota at once hangs OST
1457 Details : If a user deletes more than 4GB of files on a single OST it
1458 will cause the OST to spin in an infinite loop. Release
1459 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1464 Description: statfs data retrieved from /proc may be stale or zero
1465 Details : When reading per-device statfs data from /proc, in the
1466 {kbytes,files}_{total,free,avail} files, it may appear
1467 as zero or be out of date.
1470 Frequency : systems with MD RAID1 external journal devices
1472 Description: lconf's call to blkid is confused by RAID1 journal devices
1473 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1474 of returning all block devices that match the journal UUID.
1477 Frequency : always, for aggregate stripe size over 4GB
1479 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1480 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1481 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1484 Frequency : always if "lfs find" used on a local file/directory
1486 Description: "lfs find" segfaults if used on a local file/directory
1487 Details : The case where a directory component was not specified wasn't
1488 handled correctly. Handle this properly.
1491 Frequency : always on ppc64
1493 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1494 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1495 perform endianness conversion.
1498 Frequency : rarely (truncate to non-zero file size after write under load)
1499 Bugzilla : 10730, 10687
1500 Description: Files padded with zeros to next 4K multiple
1501 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1502 it is possible that files that are truncated to a non-zero size
1503 immediately after being written are filled with zero bytes beyond
1504 the truncated size. No file data is lost.
1506 Severity : enhancement
1508 Description: Allow recovery/failover for liblustre clients.
1509 Details : liblustre clients were unaware of failover configurations until
1512 Severity : enhancement
1514 Description: user file locks should fail when not mounting with flock option
1515 Details : Set up an error-returning stub in ll_file_operations.lock field
1516 to prevent incorrect behaviour when client is mounted without
1517 flock option. Also, set up properly f_op->flock field for
1521 Frequency : always on ia64
1523 Description: "lfs df" loops on printing out MDS statfs information
1524 Details : The obd_ioctl_data was not initialized and in some systems
1525 this caused a failure during the ioctl that did not return
1526 an error. Initialize the struct and return an error on failure.
1529 Frequency : SLES 9 only
1531 Description: Error of copying files with lustre special EAs as root
1532 Details : Client side always return success for setxattr call for lustre
1533 special xattr (currently only "trusted.lov").
1536 Frequency : rarely on clusters with both ia64+i386 clients
1538 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1539 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1540 clients are doing concurrent writes to the same file it is
1541 possible that the ia64 clients may LASSERT because the OST
1542 extent locks are not PAGE_SIZE aligned. Ensure that grown
1543 locks are always aligned on the request boundary.
1546 Frequency : specific use, occasional
1548 Description: Overwriting in use executable truncates on-disk binary image
1549 Details : If one node attempts to overwrite an executable in use by
1550 another node, we now correctly return ETXTBSY instead of
1551 truncating the file.
1553 Severity : enhancement
1555 Description: Async OSC create to avoid the blocking unnecessarily.
1556 Details : If a OST has no remain object, system will block on the creating
1557 when need to create a new object on this OST. Now, ways use
1558 pre-created objects when available, instead of blocking on an
1559 empty osc while others are not empty. If we must block, we block
1560 for the shortest possible period of time.
1565 Description: chmod on Lustre root is propagated to other clients
1566 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1567 invalid by the follow_mount time.
1570 Frequency : liblustre clients only
1572 Description: Race in 'instant cancel' lock handling could lead to such locks
1573 never to be granted in case of SMP MDS
1574 Details : Do not destroy not yet granted but cbpending locks in
1578 Frequency : replay/resend of open
1580 Description: non null lock assetion failure in mds_intent_policy
1581 Details : Trying to replay/resend lockless open requests resulted in
1582 mds_open() returning 0 with no lock. Now it sets a flag if
1583 a lock is going to be returned.
1585 Severity : enhancement
1587 Description: Checksum enhancements
1588 Details : New checksum enhancements allow for resending RPCs that failed
1591 Severity : enhancement
1593 Description: Tunables on number of dirty pages in cacche
1594 Details : Allow to set limit on number of dirty pages cached.
1599 Description: client crash on unmount - lock still has references
1600 Details : In some error handling cases it was possible to leak a lock
1601 reference on a client while accessing a file. This was not
1602 harmful to the client during operation, but would cause the
1603 client to crash when the filesystem is unmounted.
1606 Frequency : specific case, rare
1608 Description: ETXTBSY on mds though file not in use by client
1609 Details : ETXTBSY is no longer incorrectly returned when attempting to
1610 chmod or chown a directory that the user previously tried to
1611 execute or a currently-executing binary.
1614 Frequency : extremely rare except on liblustre-based clients
1616 Description: Lustre space not freed when files are deleted
1617 Details : Clean up open-unlinked files after client eviction. Previously
1618 the unlink was skipped and the files remained as orphans.
1623 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1624 Details : In some cases it was possible to send a blocking callback to
1625 a client doing a glimpse, even though that client didn't get
1626 a lock granted. When the glimpse lock is cancelled on the OST
1627 the freed lock is left on the waiting list and corrupted the list.
1630 Frequency : all core dumps
1632 Description: Broke core dumps to lustre
1633 Details : Negative dentry may be unhashed if parent does not have UPDATE
1634 lock, but some callers, e.g. do_coredump, expect dentry to be
1635 hashed after successful create, hash it in ll_create_it.
1637 ------------------------------------------------------------------------------
1639 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1641 * Support for kernels:
1642 2.6.9-42.0.2.EL (RHEL 4)
1643 2.6.5-7.276 (SLES 9)
1644 2.4.21-40.EL (RHEL 3)
1645 2.6.12.6 vanilla (kernel.org)
1649 Frequency : always on RHEL 3
1651 Description: Number of open files grows over time
1652 Details : The number of open files grows over time, whether or not
1653 Lustre is started. This was due to a filp leak introduced
1654 by one of our kernel patches.
1656 ------------------------------------------------------------------------------
1658 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1660 * Support for kernels:
1661 2.6.9-42.EL (RHEL 4)
1662 2.6.5-7.276 (SLES 9)
1663 2.4.21-40.EL (RHEL 3)
1664 2.6.12.6 vanilla (kernel.org)
1669 Bugzilla : 5719, 9635, 9792, 9684
1670 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1671 Details : If a server is under heavy load and cannot reply to new
1672 connection requests before the client resends the (re)connect,
1673 the connection handling code can behave badly if two service
1674 threads are concurrently handing separate (re)connections from
1675 the same client. Add better locking to the connection handling
1676 code, and ensure that only a single connection will be processed
1677 for a given client UUID, even if the lock is dropped.
1679 Severity : enhancement
1681 Description: add TCP zero-copy support to kernel
1682 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1683 sends if the hardware supports scatter-gather and checksumming.
1684 This allows socklnd to do client-write and server-read more
1685 efficiently and reduce CPU utilization from skbuf copying.
1688 Frequency : only if NFS exporting from client
1690 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1691 Details : If setting attributes on a file created under NFS that had
1692 never been opened it would be possible to oops the client
1693 if the file had no objects.
1696 Frequency : always for liblustre
1698 Description: liblustre client does MDS+OSTs setattr RPC for each write
1699 Details : When doing a write from a liblustre client, the client
1700 incorrectly issued an RPC to the MDS and each OST the file was
1701 striped over in order to update the timestamps. When writing
1702 with small chunks and many clients this could overwhelm the MDS
1703 with RPCs. In all cases it would slow down the write because
1704 these RPCs are unnecessary.
1706 Severity : enhancement
1708 Description: allow number of MDS service threads to be changed at module load
1709 Details : It is now possible to change the number of MDS service threads
1710 running. Adding "options mds mds_num_threads={N}" to the MDS's
1711 /etc/modprobe.conf will set the number of threads for the next
1712 time Lustre is restarted (assuming the "mds" module is also
1713 reloaded at that time). The default number of threads will
1714 stay the same, 32 for most systems.
1719 Description: OST crash if filesystem is unformatted or corrupt
1720 Details : If an OST is started on a device that has never been formatted
1721 or if the filesystem is corrupt and cannot even mount then the
1722 error handling cleanup routines would dereference a NULL pointer.
1727 Description: NULL pointer deref in llap_from_page.
1728 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1729 incorrect) mapping if the page was truncated/reclaimed while it was
1730 searched for. Check for this condition and skip such pages when
1731 doing readahead. Introduce extra check to llap_from_page() to
1732 verify page->mapping->host is non-NULL (so page is not anonymous).
1735 Frequency : Sometimes when using sys_sendfile
1737 Description: "page not covered by a lock" warnings from ll_readpage
1738 Details : sendfile called ll_readpage without right page locks present.
1739 Now we introduced ll_file_sendfile that does necessary locking
1740 around call to generic_file_sendfile() much like we do in
1744 Frequency : with certain MDS communication failures at client mount time
1746 Description: NULL pointer deref after failed client mount
1747 Details : a client connection request may delayed by the network layer
1748 and not be sent until after the PTLRPC layer has timed out the
1749 request. If the client fails the mount immediately it will try
1750 to clean up before the network times out the request. Add a
1751 reference from the request import to the obd device and delay
1752 the cleanup until the network drops the request.
1755 Frequency : occasionally during client (re)connect
1757 Description: assertion failure during client (re)connect
1758 Details : processing a client connection request may be delayed by the
1759 client or server longer than the client connect timeout. This
1760 causes the client to resend the connection request. If the
1761 original connection request is replied in this interval, the
1762 client may trip an assertion failure in ptlrpc_connect_interpret()
1763 which thought it would be the only running connect process.
1766 Frequency : only with obd_echo servers and clients that are rebooted
1768 Description: kernel BUG accessing uninitialized data structure
1769 Details : When running an obd_echo server it did not start the ping_evictor
1770 thread, and when a client was evicted an uninitialized data
1771 structure was accessed. Start the ping_evictor in the RPC
1772 service startup instead of the OBD startup.
1774 Severity : enhancement
1775 Bugzilla : 10193 (patchless)
1776 Description: Remove dependency on various unexported kernel interfaces.
1777 Details : No longer need reparent_to_init, exit_mm, exit_files,
1778 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1781 Frequency : rare (only users of deprecated and unsupported LDAP config)
1783 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1784 Details : LDAP apparently contains 'lustreName' attributes instead of
1785 'name'. A simple remapping of the name is sufficient.
1788 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1790 Description: waiting_locks_callback trips kernel BUG if client is evicted
1791 Details : Running with the dump_on_timeout debug flag turned on makes
1792 it possible that the waiting_locks_callback() can try to dump
1793 the Lustre kernel debug logs from an interrupt handler. Defer
1794 this log dumping to the expired_lock_main() thread.
1796 Severity : enhancement
1798 Description: Support NFS exporting on 2.6 kernels.
1799 Details : Implement non-rawops metadata methods for NFS server to use without
1800 changing NFS server code.
1803 Frequency : very rare (synthetic metadata workload only)
1805 Description: two racing renames might cause an MDS thread to deadlock
1806 Details : Running the "racer" program may cause one MDS thread to rename
1807 a file from being the source of a rename to being the target of
1808 a rename at exactly the same time that another thread is doing
1809 so, and the second thread has already enqueued these locks after
1810 doing a lookup of the target and is trying to relock them in
1811 order. Ensure that we don't try to re-lock the same resource.
1814 Frequency : only very large systems with liblustre clients
1816 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1817 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1818 all OSTs in parallel. This allows the eviction of stale liblustre
1819 clients to proceed much faster than if they were done in series,
1820 and also offers similar improvements for other set_info RPCs.
1825 Description: excessive CPU usage during initial read phase on client
1826 Details : During the initial read phase on a client, it would agressively
1827 retry readahead on the file, consuming too much CPU and impacting
1828 performance (since 1.4.5.8). Improve the readahead algorithm
1829 to avoid this, and also improve some other common cases (read
1830 of small files in particular, where "small" is files smaller than
1831 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1836 Description: MDS crash when receiving packet with unknown intent.
1837 Details : Do not LBUG in unknown intent case, just return -EFAULT
1839 Severity : enhancement
1840 Bugzilla : 9293, 9385
1841 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1842 Details : Do not serialize getattr (non-intent version) and statfs.
1845 Frequency : occasional, when OST network is overloaded/intermittent
1847 Description: client evicted by OST after bulk IO timeout
1848 Details : If a client sends a bulk IO request (read or write) the OST
1849 may evict the client if it is unresposive to its data GET/PUT
1850 request. This is incorrect if the network is overloaded (takes
1851 too long to transfer the RPC data) or dropped the OST GET/PUT
1852 request. There is no need to evict the client at all, since
1853 the pinger and/or lock callbacks will handle this, and the
1854 client can restart the bulk request.
1857 Frequency : Always when mmapping file with no objects
1859 Description: client crashes when mmapping file with no objects
1860 Details : Check that we actually have objects in a file before doing any
1861 operations on objects in ll_vm_open, ll_vm_close and
1867 Description: Request leak when working with deleted CWD
1868 Details : Introduce advanced request refcount tracking for requests
1869 referenced from lustre intent.
1871 Severity : Enhancement
1873 Description: Cache open file handles on client.
1874 Details : MDS now will return special lock along with openhandle, if
1875 requested and client is allowed to hold openhandle, even if unused,
1876 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1877 closing files for every read/write openration.
1879 Severity : Enhancement
1881 Description: Cache open negative dentries on client when possible.
1882 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1883 negative dentries on lock revocation.
1888 Description: Remounting a client read-only wasn't possible with a zconf mount
1889 Details : It wasn't possible to remount a client read-only with llmount.
1891 Severity : enhancement
1892 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1893 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1894 MPICH is included to add Lustre-specific ADIO interfaces.
1895 This is based closely on the UFS ADIO layer and only differs
1896 in file creation, in order to allow the OST striping to be set.
1897 This is user-contributed code and not supported by CFS.
1902 Description: extended inode attributes (immutable, append-only) work improperly
1903 when 2.4 and 2.6 kernels are used on client/server or vice versa
1904 Details : Introduce kernel-independent values for these flags.
1906 Severity : enhancement
1909 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1910 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1911 granularity. Now, Fractional MB granularity is be supported,
1912 this is very useful for low memory system.
1914 Severity : enhancement
1916 Description: Getattr by fid
1917 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1918 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1919 have locks and avoids getattr rpc requests for it on every stat.
1922 Frequency : Always, for filesystems larger than 2TB
1924 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1925 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1926 filesystems larger than 2TB to be mounted with mballoc enabled.
1929 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1931 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1932 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1933 filesystems larger than 2TB is mounted then the kernel will
1934 silently corrupt the start of the filesystem. CONFIG_LBD is
1935 enabled for all CFS-supported kernels, but the possibility of
1936 this happening with a modified kernel config exists.
1938 Severity : enhancement
1940 Description: add client O_DIRECT support for 2.6 kernels
1941 Details : It is now possible to do O_DIRECT reads and writes to files
1942 in the Lustre client mountpoint on 2.6 kernel clients.
1944 Severity : enhancement
1946 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1947 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1948 parallel, not waiting for response from every OST before sending
1949 a rpc to the next OST.
1954 Description: setattr vs write race when updating file timestamps
1955 Details : Client processes that update a file timestamp into the past
1956 right after writing to the file (e.g. tar) it is possible that
1957 the updated file modification time can be reset to the current
1958 time due to a race between processing the setattr and write RPC.
1960 Severity : enhancement
1962 Description: Bring 'lfs find' closer in line with regular Linux find.
1963 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1964 -print0 options and obtains all the needed info through the lustre
1967 Severity : enhancement
1969 Description: support up to 1024 configured devices on one node
1970 Details : change obd_dev array from statically allocated to dynamically
1971 allocated structs as they are first used to reduce memory usage
1976 Description: Flush dirty partially truncated pages during truncate
1977 Details : Immediatelly flush partially truncated pages in filter_setattr,
1978 this way we completely avoid having any pages in page cache on OST
1979 and can retire ugly workarounds during writes to flush such pages.
1984 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1985 Details : protect inode from truncation within vfs_unlink() context
1986 just take a reference before calling vfs_unlink() and release it
1987 when parent's i_sem is free.
1992 Description: last_id value checked outside lock on OST caused LASSERT failure
1993 Details : If there were multiple MDS->OST object precreate requests in
1994 flight, it was possible that the OST's last object id was checked
1995 outside a lock and incorrectly tripped an assertion. Move checks
1996 inside locks, and discard old precreate requests.
1999 Frequency : always, if extents are used on OSTs
2001 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2002 Details : OSTs using the extents format would not zero the high 16 bits of
2003 the index physical block number. This is not a problem for any
2004 OST filesystems smaller than 16TB, and no kernels support ext3
2005 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2006 new/modified files) and can be fixed for existing filesystems
2007 with e2fsprogs-1.39-cfs1.
2012 Description: import connection selection may be incorrect if timer wraps
2013 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2014 import connections to be ignored if the 32-bit jiffies counter
2015 wraps. Use a 64-bit jiffies counter.
2018 Frequency : very large clusters immediately after boot
2020 Description: LNET request buffers exhausted under heavy short-term load
2021 Details : If a large number of client requests are generated on a service
2022 that has previously never seen so many requests it is possible
2023 that the request buffer growth cannot keep up with the spike in
2024 demand. Instead of dropping incoming requests, they are held in
2025 the LND until the RPC service can accept more requests.
2028 Frequency : Sometimes during replay
2030 Description: Assertion failure in ll_local_open after replay.
2031 Details : If replay happened on an open request reply before we were able
2032 to set replay handler, reply will become not swabbed tripping the
2033 assertion in ll_local_open. Now we set the handler right after
2034 recognising of open request
2037 Frequency : very rare
2039 Description: kernel reports "badness in vsnprintf"
2040 Details : Reading from the "recovery_status" /proc file in small chunks
2041 may cause a negative length in lprocfs_obd_rd_recovery_status()
2042 call to vsnprintf() (which is otherwise harmless). Exit early
2043 if there is no more space in the output buffer.
2045 Severity : enhancement
2047 Description: clear OBD RPC statistics by writing to them
2048 Details : It is now possible to clear the OBD RPC statistics by writing
2049 to the "stats" file.
2054 Description: Client mtime is not the same on different clients after utimes
2055 Details : In some cases, the client was using the utimes() syscall on
2056 a file cached on another node. The clients now validate the
2057 ctime from the MDS + OSTs to determine which one is right.
2062 Description: Inability to activate failout mode
2063 Details : lconf script incorrectly assumed that in pythong string's numeric
2064 value is used in comparisons.
2067 Frequency : always with multiple stripes per file
2069 Description: Inefficient object allocation for mutli-stripe files
2070 Details : When selecting which OSTs to stripe files over, for files with
2071 a stripe count that divides evenly into the number of OSTs,
2072 the MDS is always picking the same starting OST for each file.
2073 Return the OST selection heuristic to the original design.
2078 Description: mount failures may take full timeout to return an error
2079 Details : Under some heavy load conditions it is possible that a
2080 failed mount can wait for the full obd_timeout interval,
2081 possibly several minutes, before reporting an error.
2082 Instead return an error as soon as the status is known.
2084 Frequency : quota enabled and large files being deleted
2086 Description: releasing more than 4GB of quota at once hangs OST
2087 Details : If a user deletes more than 4GB of files on a single OST it
2088 will cause the OST to spin in an infinite loop. Release
2089 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2094 Description: statfs data retrieved from /proc may be stale or zero
2095 Details : When reading per-device statfs data from /proc, in the
2096 {kbytes,files}_{total,free,avail} files, it may appear
2097 as zero or be out of date.
2100 Frequency : systems with MD RAID1 external journal devices
2102 Description: lconf's call to blkid is confused by RAID1 journal devices
2103 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2104 of returning all block devices that match the journal UUID.
2107 Frequency : always, for aggregate stripe size over 4GB
2109 Description: assertion fails when trying to use 4GB stripe size
2110 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2111 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2114 Frequency : always on ppc64
2116 Description: the first write on an ext3 filesystem with mballoc got stuck
2117 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2118 perform endianness conversion.
2120 ------------------------------------------------------------------------------
2122 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2124 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2125 INTEROPERATE with older versions automatically. Please read the
2126 user documentation before upgrading any part of a live system.
2127 * WARNING: Lustre networking configuration changes are required with
2128 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2131 * Support for kernels:
2132 2.6.9-22.0.2.EL (RHEL 4)
2133 2.6.5-7.244 (SLES 9)
2134 2.6.12.6 vanilla (kernel.org)
2137 Severity : enhancement
2138 Bugzilla : 7981/8208
2139 Description: Introduced Lustre Networking (LNET)
2140 Details : LNET is new networking infrastructure for Lustre, it includes
2141 a reorganized network configuration mode (see the user
2142 documentation for full details) as well as support for routing
2143 between different network fabrics. Lustre Networking Devices
2144 (LNDs) for the supported network fabrics have also been
2145 created for this new infrastructure.
2147 Severity : enhancement
2148 Description: Introduced Access control lists
2149 Details : clients can set ACLs on files and directories in order to have
2150 more fine-grained permissions than the standard Unix UGO+RWX.
2151 The MDS must be started with the "-o acl" mount option.
2153 Severity : enhancement
2154 Description: Introduced filesystem quotas
2155 Details : Administrators may now establish per-user quotas on the
2158 Severity : enhancement
2160 Description: Configuration change for the XT3
2161 The PTLLND is now used to run Lustre over Portals on the XT3
2162 The configure option(s) --with-cray-portals are no longer used.
2163 Rather --with-portals=<path-to-portals-includes> is used to
2164 enable building on the XT3. In addition to enable XT3 specific
2165 features the option --enable-cray-xt3 must be used.
2170 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2171 Details : A race between allocating a new llcd and re-getting the llcd_lock
2172 allowed another thread to grab newly-allocated llcd.
2174 Severity : enhancement
2176 Description: 2.6 OST async journal commit and locking fix to improve performance
2177 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2178 2.6 kernels are now async as they already were in 2.4 kernels so
2179 that they can commit concurrently with the network bulk transfer.
2180 For block-allocated files the filter allocation semaphore is held
2181 to avoid filesystem fragmentation during allocation. BKL lock
2182 removed for 2.6 xattr operations where it is no longer needed.
2187 Description: lconf incorrectly determined whether two IP networks could talk
2188 Details : In some more complicated routing and multiple-network
2189 configurations, lconf will avoid trying to make a network
2190 connection to a disjoint part of the IP space. It was doing the
2191 math incorrectly for one set of cases.
2196 Description: Fix for potential infinite loop processing records in an llog.
2197 Details : If an llog record is corrupted/zeroed, it is possible to loop
2198 forever in llog_process(). Validate the llog record length
2199 and skip the remainder of the block on error.
2202 Frequency : occasional (liblustre only)
2204 Description: liblustre could not open files whose last component is a symlink
2205 Details : sysio_path_walk() would incorrectly pass the open intent to
2206 intermediate path components.
2209 Frequency : rare (liblustre only with non-standard tuning)
2210 Bugzilla : 7201 (7350)
2211 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2212 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2213 after it has been released, only do this for referenced locks
2215 Severity : enhancement
2217 Description: specifying an (invalid) directory default stripe_size of -1
2218 would reset the directory default striping
2219 Details : stripe_size -1 was used internally to signal directory stripe
2220 removal, now use "all default" to signal dir stripe removal
2221 as a directory striping of "all default" is not useful
2224 Frequency : common for large clusters running liblustre clients
2226 Description: doing an ls when liblustre clients are running is slow
2227 Details : sending a glimpse AST to a liblustre client waits for every AST
2228 to time out, as liblustre clients will not respond. Since they
2229 cannot cache data we refresh the OST lock LVB from disk instead.
2231 Severity : enhancement
2233 Description: doing an ls at the same time as file IO can be slow
2234 Details : enqueue and other "small" requests can be blocked behind many
2235 large IO requests. Create a new OST IO portal for non-IO
2236 requests so they can be processed faster.
2239 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2241 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2242 created file with mode 000
2243 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2244 allow an ftruncate() as a non-root user to a file with mode 000.
2245 Lustre now respects this flag to disable mode checks when
2246 truncating a file owned by the user
2249 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2252 Description: Revoking locks from clients that went dead or catatonic might take
2254 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2255 cancellation of such locks instant on servers without waiting for
2256 any reply from clients. Clients drops these locks when cancel
2257 notification from server is received without replying.
2260 Frequency : liblustre-only, when liblustre client dies or becomes busy
2262 Description: Doing ls on Linux clients can take a long time with active
2264 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2265 granting such locks to it in the first place if possible. Locks
2266 are taken by proxy on the OST during the read or write and
2267 dropped immediately afterward. Add connect flags handling, do
2268 not grant locks to liblustre clients for glimpse ASTs.
2270 Severity : enhancement
2272 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2273 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2274 sequential reads and behaves badly with applications doing random
2275 reads. Improve it by reading ahead at least read region, and
2276 avoiding excessive large RPC for small reads.
2278 Severity : enhancement
2280 Description: Creating more than 1000 files for a single job may cause a load
2281 imbalance on the OSTs if there are also a large number of OSTs.
2282 Details : qos_prep_create() uses an OST index reseed value that is an
2283 even multiple of the number of available OSTs so that if the
2284 reseed happens in the middle of the object allocation it will
2285 still utilize the OSTs as uniformly as possible.
2290 Description: OST or MDS may oops in ping_evictor_main()
2291 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2292 but doesn't restart at beginning of obd_exports_timed list
2295 Severity : enhancement
2297 Description: improve by-nid export eviction on the MDS and OST
2298 Details : allow multiple exports with the same NID to be evicted at one
2299 time without re-searching the exports list.
2302 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2304 Description: MDS may oops in groups_free()
2305 Details : in rare race conditions a newly allocated group_info struct is
2306 freed again, and this can be NULL. The 2.4 compatibility code
2307 for groups_free() checked for a NULL pointer, but 2.6 did not.
2310 Frequency : common for liblustre clients doing little filesystem IO
2311 Bugzilla : 9352, 7313
2312 Description: server may evict liblustre clients accessing contended locks
2313 Details : if a client is granted a lock or receives a completion AST
2314 with a blocking AST already set it would not reply to the AST
2315 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2318 Frequency : lfs setstripe, only systems with more than 160 OSTs
2320 Description: unable to set striping with a starting offset beyond OST 160
2321 Details : llapi_create_file() incorrectly limited the starting stripe
2322 index to the maximum single-file stripe count.
2325 Frequency : LDAP users only
2327 Description: lconf did not handle in-kernel recovery with LDAP properly
2328 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2330 Severity : enhancement
2332 Description: bind OST threads to NUMA nodes to improve performance
2333 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2334 node and do their allocations there to localize memory access
2336 Severity : enhancement
2338 Description: llmount can determine client NID directly from Myrinet (GM)
2339 Details : the client NID code from gmnalnid was moved directly into
2340 llmount, removing the need to use this or specifying the
2341 client NID explicitly when mounting GM clients with zeroconf
2344 Frequency : if client is started with down MDS
2346 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2347 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2348 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2349 ptlrpc_import_delayed_req() because we didn't check if the
2350 request was interrupted, and we also didn't break out of the
2351 event loop if there was no timeout
2356 Description: data loss during non-page-aligned writes to a single file from
2357 both multiple nodes and multiple threads on one node at same time
2358 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2359 inconsistency led to false short-reads, that were cached and later
2360 used by ->prepare_write() to fill in partially written page,
2361 leading to data loss.
2364 Frequency : always, if lconf --abort_recovery used
2366 Description: lconf --abort_recovery fails with 'Operation not supported'
2367 Details : lconf was attempting to abort recovery on the MDT device and not
2370 Severity : enhancement
2372 Description: remove cleanup logs
2373 Details : replace lconf-generated cleanup logs with lustre internal
2374 cleanup routines. Eliminates the need for client-cleanup and
2377 Severity : enhancement
2379 Description: add support for EAs (user and system) on lustre filesystems
2380 Details : it is now possible to store extended attributes in the Lustre
2381 client filesystem, and with the user_xattr mount option it
2382 is possible to allow users to store EAs on their files also
2384 Severity : enhancement
2386 Description: Add possibility (config option) to show minimal available OST free
2388 Details : When compiled with --enable-mindf configure option, statfs(2)
2389 (and so, df) will return least minimal free space available from
2390 all OSTs as amount of free space on FS, instead of summary of
2391 free spaces of all OSTs.
2393 Severity : enhancement
2395 Description: do not expand extent locks acquired on OST-side
2396 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2397 by server: they are not cached anyway.
2400 Frequency : when mmap is used/binaries executed from Lustre
2402 Description: Unmmap pages before throwing them away from read cache.
2403 Details : llap_shrink cache now attempts to unmap pages before discarding
2404 them (if unmapping failed - do not discard). SLES9 kernel has
2405 extra checks that trigger if this unmapping is not done first.
2410 Description: lconf didn't resolve symlinks before checking to see whether a
2411 given mountpoint was already in use
2414 Frequency : when migrating failover services
2415 Bugzilla : 6395, 9514
2416 Description: When migrating a subset of services from a node (e.g. failback
2417 from a failover service node) the remaining services would
2418 time out and evict clients.
2419 Details : lconf --force (implied by --failover) sets the global obd_timeout
2420 to 5 seconds in order to quickly disconnect, but this caused
2421 other RPCs to time out too quickly. Do not change the global
2422 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2424 Severity : enhancement
2425 Frequency : if MDS is started with down OST
2426 Bugzilla : 9439,5706
2427 Description: Allow startup/shutdown of an MDS without depending on the
2428 availability of the OSTs.
2429 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2430 Add appropriate locking and lov-osc refcounts for safe
2431 cleaning. Add osc abort_inflight calls in case the
2432 synchronize never started.
2435 Frequency : occasional (Cray XT3 only)
2437 Description: root not authorized to access files in CRAY_PORTALS environment
2438 Details : The client process capabilities were not honoured on the MDS in
2439 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2440 been accessed by an authorized user then root was able to access
2441 the file on the local client also. The root user capabilities
2442 are now allowed on the MDS, as this environment has secure UID.
2445 Frequency : occasional
2447 Description: ldiskfs "too long searching" message happens too often
2448 Details : A debugging message (otherwise harmless) prints too often on
2449 the OST console. This has been reduced to only happen when
2450 there are fragmentation problems on the filesystem.
2455 Description: Division by zero in statfs when all OSCs are inactive
2456 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2457 lov_statfs divides by value returned by lov_get_stripecnt().
2461 Bugzilla : 9489, 3273
2462 Description: First write from each client to each OST was only 4kB in size,
2463 to initialize client writeback cache, which caused sub-optimal
2464 RPCs and poor layout on disk for the first writen file.
2465 Details : Clients now request an initial cache grant at (re)connect time
2466 and so that they can start streaming writes to the cache right
2467 away and always do full-sized RPCs if there is enough data.
2468 If the OST is rebooted the client also re-establishes its grant
2469 so that client cached writes will be honoured under the grant.
2474 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2475 Details : Now I/O RPCs go to different portal number and (presumably) fast
2476 lock requests (and glimses) and other RPCs get their own service
2477 threads pool that should be able to service those RPCs
2480 Severity : enhancement
2482 Description: Ability to exchange lustre version between client and servers and
2483 issue warnings at client side if client is too old. Also for
2484 liblustre clients there is ability to refuse connection of too old
2486 Details : New 'version' field is added to connect data structure that is
2487 filled with version info. That info is later checked by server and
2491 Frequency : rare, liblustre only.
2492 Bugzilla : 9296, 9581
2493 Description: Two simultaneous writes from liblustre at offset within same page
2494 might proceed at the same time overwriting eachother with stale
2496 Details : I/O lock withing llu_file_prwv was released too early, before data
2497 actually was hitting the wire. Extended lock-holding time until
2498 server acknowledges receiving data.
2501 Frequency : extremely rare. Never observed in practice.
2503 Description: avoid generating lustre_handle cookie of 0.
2504 Details : class_handle_hash() generates handle cookies by incrementing
2505 global counter, and can hit 0 occasionaly (this is unlikely, but
2506 not impossible, because initial value of cookie counter is
2507 selected randonly). Value of 0 is used as a sentinel meaning
2508 "unassigned handle" --- avoid it. Also coalesce two critical
2509 sections in this function into one.
2511 Severity : enhancement
2513 Description: allow liblustre clients to delegate truncate locking to OST
2514 Details : To avoid overhead of locking, liblustre client instructs OST to
2515 take extent lock in ost_punch() on client's behalf. New connection
2516 flag is added to handle backward compatibility.
2518 Severity : enhancement
2519 Bugzilla : 4928, 7341, 9758
2520 Description: allow number of OST service threads to be specified
2521 Details : a module parameter allows the number of OST service threads
2522 to be specified via "options ost ost_num_threads={N}" in the
2523 OSS's /etc/modules.conf or /etc/modprobe.conf.
2527 Bugzilla : 6146, 9635, 9895
2528 Description: servers crash with bad pointer in target_handle_connect()
2529 Details : In rare cases when a client is reconnecting it was possible that
2530 the connection request was the last reference for that export.
2531 We would temporarily drop the export reference and get a new
2532 one, but this may have been the last reference and the export
2533 was just destroyed. Get new reference before dropping old one.
2535 Severity : enhancement
2536 Frequency : if client is started with failover MDS
2538 Description: Allow multiple MDS hostnames in the mount command
2539 Details : Try to read the configuration from all specified MDS
2540 hostnames during a client mount in case the "primary"
2543 Severity : enhancement
2545 Description: Stop sending data to evicted clients as soon as possible.
2546 Details : Check if the client we are about to send or are sending data to
2547 was evicted already. (Check is done every second of waiting,
2548 for which l_wait_event interface was extended to allow checking
2549 of exit condition at specified intervals).
2552 Frequency : rare, normally only when NFS exporting is done from client
2554 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2556 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2559 Frequency : always, if srand() is called before liblustre initialization
2561 Description: Liblustre uses system PRNG disturbing its usage by user application
2562 Details : Introduce internal to lustre fast and high-quality PRNG for
2563 lustre usage and make liblustre and some other places in generic
2564 lustre code to use it.
2566 Severity : enhancement
2567 Bugzilla : 9477, 9557, 9870
2568 Description: Verify that the MDS configuration logs are updated when xml is
2569 Details : Check if the .xml configuration logs are newer than the config
2570 logs stored on the MDS and report an error if this is the case.
2571 Request --write-conf, or allow starting with --old_conf.
2573 Severity : enhancement
2575 Description: Handle symlinks in the path when checking if Lustre is mounted.
2576 Details : Resolve intermediate symlinks when checking if a client has
2577 mounted a filesystem to avoid duplicate client mounts.
2582 Description: lconf can hit an error exception but still return success.
2583 Details : The lconf command catches the Command error exception at the top
2584 level script context and will exit with the associated exit
2585 status, but doesn't ensure that this exit status is non-zero.
2590 Description: failure of ptlrpc thread startup can cause oops
2591 Details : Starting a ptlrpc service thread can fail if there are a large
2592 number of threads or the server memory is very fragmented.
2593 Handle this without oopsing.
2596 Frequency : always, only if liblustre and non-default acceptor port was used
2598 Description: liblustre cannot connect to servers with non-default acceptor port
2599 Details : tcpnal_set_default_params() was not called and was therefore
2600 ignoring the environment varaible TCPNAL_PORT, as well as other
2601 TCPNAL_ environment variables
2606 Description: two objects could be created on the same OST for a single file
2607 Details : If an OST is down, in some cases it was possible to create two
2608 objects on a single OST for a single file. No problems other
2609 than potential performance impact and spurious error messages.
2613 Bugzilla : 5681, 9562
2614 Description: Client may oops in ll_unhash_aliases
2615 Details : Client dcache may become inconsistent in race condition.
2616 In some cases "getcwd" can fail if the current directory is
2622 Description: Inode refcounting problems in NFS export code
2623 Details : link_raw functions used to call d_instantiate without obtaining
2624 extra inode reference first.
2628 Bugzilla : 9942, 9903
2629 Description: Referencing freed requests leading to crash, memleaks with NFS.
2630 Details : We used to require that call to ll_revalidate_it was always
2631 followed by ll_lookup_it. Also with revalidate_special() it is
2632 possible to call ll_revalidate_it() twice for the same dentry
2633 even if first occurence returned success. This fix changes semantic
2634 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2635 reference on a request referred from the intent.
2636 ll_intent_release() then releases such a request.
2639 Frequency : rare, normally benchmark loads only
2641 Description: unlinked inodes were kept in memory on the client
2642 Details : If a client is repeatedly creating and unlinking files it
2643 can accumulate a lot of stale inodes in the inode slab cache.
2644 If there is no other client load running this can cause the
2645 client node to run out of memory. Instead flush old inodes
2646 from client cache that have the same inode number as a new inode.
2649 Frequency : SLES9 2.6.5 kernel and long filenames only
2650 Bugzilla : 9969, 10379
2651 Description: utime reports stale NFS file handle
2652 Details : SLES9 uses out-of-dentry names in some cases, which confused
2653 the lustre dentry revalidation. Change it to always use the
2657 Frequency : rare, unless heavy write-truncate concurrency is continuous
2658 Bugzilla : 4180, 6984, 7171, 9963, 9331
2659 Description: OST becomes very slow and/or deadlocked during object unlink
2660 Details : filter_destroy() was holding onto the parent directory lock
2661 while truncating+unlinking objects. For very large objects this
2662 may block other threads for a long time and slow overall OST
2663 responsiveness. It may also be possible to get a lock ordering
2664 deadlock in this case, or run out of journal credits because of
2665 the combined truncate+unlink. Solution is to do object truncate
2666 first in one transaction without parent lock, and then do the
2667 final unlink in a new transaction with the parent lock. This
2668 reduces the lock hold time dramatically.
2671 Frequency : rare, 2.4 kernels only
2673 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2674 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2675 before unmounting their filesystem in order to avoid deadlock.
2676 The kernel_locked() function in 2.4 kernels only checks whether
2677 the kernel lock is held, not whether it is this process that is
2678 holding it as 2.6 kernels do.
2683 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2684 Details : The client ptlrpc code may be trying to reconnect to a down
2685 server before a previous connection attempt has timed out.
2686 Increase the reconnect interval to be longer than the connection
2687 timeout interval to avoid sending duplicate connections to
2691 Frequency : echo_client brw_test command
2693 Description: fix echo_client to work with OST preallocated code
2694 Details : OST preallocation code (5137) didn't take echo_client IO path
2695 into account: echo_client calls filter methods outside of any
2696 OST thread and, hence, there is no per-thread preallocated
2697 pages and buffers to use. Solution: hijack pga pages for IO. As
2698 a byproduct, this avoids unnecessary data copying.
2702 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2703 Description: Client can oops in mdc_commit_close() after open replay
2704 Details : It was possible for the MDS to return an open request with no
2705 transaction number in mds_finish_transno() if the client was
2706 evicted, but without actually returning an error. Clients
2707 would later try to replay that open and may trip an assertion
2708 Simplify the client close codepath, and always return an error
2709 from the MDS in case the open is not successful.
2712 Frequency : rare, 2.6 OSTs only
2714 Description: OST may deadlock under high load on fragmented files
2715 Details : If there was a heavy load and highly-fragmented OST filesystems
2716 it was possible to have all the OST threads deadlock waiting on
2717 allocation of biovecs, because the biovecs were not released
2718 until the entire RPC IO was completed. Instead, release biovecs
2719 as soon as they are complete to ensure forward IO progress.
2721 Severity : enhancement
2723 Description: Support for specifying external journal device at mount
2724 Details : If an OST or MDS device is formatted with an external journal
2725 device, this device major/minor is stored in the ext3 superblock
2726 and may not be valid for failover. Allow detecting and
2727 specifying the external journal at mount time.
2732 Description: Mounting an MDS with pending unlinked files may cause oops
2733 Details : target_finish_recovery() calls mds_postrecov() which returned
2734 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2735 considers this an error and immediately begins cleaning up the
2736 lov, just after starting the mds_lov process
2738 Severity : enhancement
2740 Description: Implement 'lfs df' to report actual free space on per-OST basis
2741 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2742 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2743 usage of objects; '-h' to report in human readable format.
2745 ------------------------------------------------------------------------------
2747 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2754 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2755 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2756 at startup during orphan unlinking. Instead, skip buddy update
2757 before setup, as it will be regenerated after recovery is complete.
2762 Description: If an OST is inactive, its locks might reference stale inodes.
2763 Details : lov_change_cbdata() must iterate over all namespaces, even if
2764 they are inactive to clear inode references from the lock.
2766 Severity : enhancement
2767 Frequency : occasional, if non-standard max_dirty_mb used
2769 Description: Client will block write RPCs if not enough grant
2770 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2771 then the client will block writes while waiting for another RPC
2772 to complete instead of consuming its dirty limit. With change
2773 we get improved performance when max_dirty_mb is small.
2775 Severity : enhancement
2776 Bugzilla : 3389, 6253
2777 Description: Add support for supplementary groups on the MDS.
2778 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2779 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2780 lookups for user supplementary groups into a cache.
2784 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2785 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2786 for liblustre O_EXCL is enforced by MDS.
2790 Description: readdir never returns NULL in liblustre.
2791 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2792 dentry in struct dirent.
2796 Frequency : liblustre only, depends on application IO pattern
2797 Description: liblustre clients evicted if not contacting servers
2798 Details : Don't put liblustre clients into the ping_evictor list, so
2799 they will not be evicted by the pinger ever.
2801 Severity : enhancement
2803 Description: Add ability to evict clients by NID from MDS.
2804 Details : By echoing "nid:$NID" string into
2805 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2806 $NID would be instantly evicted from this MDS and from all active
2807 OSTs connected to it.
2811 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2812 Details : lookup_it_finish() used to query file size from OSTs that was not
2817 Description: service threads change working directory to that of init
2818 Details : Starting lustre service threads may pin the working directory
2819 of the parent thread, making that filesystem busy. Threads
2820 now change to the working directory of init to avoid this.
2824 Frequency : during shutdown only
2825 Description: shutdown with a failed MDS or OST can cause unmount to hang
2826 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2831 Frequency : 2.6 only
2832 Description: chmod/chown may include an extra supplementary group
2833 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2834 supplementary group and if none is specified this is used.
2837 Bugzilla : 5479 (6816)
2839 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2840 Details : If lookup is in progress on a file that is unlinked we might try
2841 to revalidate the inode and fail in revalidate after lookup is
2842 complete and ll_file_open() enqueues the open again but
2843 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2846 Frequency : always, if lconf --abort_recovery used
2848 Description: lconf --abort_recovery fails with 'Operation not supported'
2849 Details : lconf was attempting to abort recovery on the MDT device and not
2852 ------------------------------------------------------------------------------
2854 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2859 Frequency : rare (only unsupported configurations with a node running as an
2861 Bugzilla : 6514, 5137
2862 Description: Mounting a Lustre file system on a node running as an OST could
2864 Details : OSTs now preallocates memory needed to write out data at
2865 startup, instead of when needed, to avoid having to
2866 allocate memory in possibly low memory situations.
2867 Specifically, if the file system is mounted on on OST,
2868 memory pressure could force it to try to write out data,
2869 which it needed to allocate memory to do. Due to the low
2870 memory, it would be unable to do so and the node would
2871 become unresponsive.
2873 Severity : enhancement
2875 Description: Addition of lconf --service command line option
2876 Details : lconf now accepts a '--service <arg>' option, which is
2877 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2879 Severity : enhancement
2881 Description: Failover mode is now the default for OSTs.
2882 Details : By default, OSTs will now run in failover mode. To return to
2883 the old behaviour, add '--failout' to the lmc line for OSTs.
2885 Severity : enhancement
2887 Description: Health checks are now provided for MDS and OSTs
2888 Details : Additional detailed health check information on MSD and OSTs
2889 is now provided through the procfs health_check value.
2892 Frequency : occasional, depends on IO load
2894 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2895 after numerous create/delete cycles
2896 Details : The ext3 inode allocation policy would not allocate new inodes
2897 very well on the OSTs because there are no new directories
2898 being created. Instead we look for groups with free space if
2899 the parent directories are nearly full.
2904 Description: Network or server problems during mount may cause partially
2905 mounted clients instead of returning an error.
2906 Details : The config llog parsing code may overwrite the error return
2907 code during mount error handling, returning success instead
2913 Description: MDS can fail to allocate large reply buffers
2914 Details : After long uptimes the MDS can fail to allocate large reply
2915 buffers (e.g. zconf client mount config records) due to memory
2916 fragmentation or consumption by the buffer cache. Preallocate
2917 some large reply buffers so that these replies can be sent even
2918 under memory pressure.
2922 Frequency : rare (liblustre)
2923 Description: fsx running with liblustre complained that using truncate() to
2924 extend the file doesn't work. This patch corrects that issue.
2925 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2926 ATTR_SIZE and lsm use in llu_setattr_raw.
2930 Frequency : rare, only 2.6 kernels
2931 Description: Unusual file access patterns on the MDS may result in inode
2932 data being lost in very rare circumstances.
2933 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2934 code in the 2.6 kernel caused the inode and/or EA data not to
2935 be read from disk, causing single-file corruption.
2939 Frequency : rare, only 2.6 filesystems using extents
2940 Description: Heavy concurrent write and delete load may cause data corruption.
2941 Details : It was possible under high-load situations to have an extent
2942 metadata block in the block device cache from a just-unlinked
2943 file overwrite a newly-allocated data block. We now unmap any
2944 metadata buffers that alias just-allocated data blocks.
2948 Frequency : filesystems with default stripe_count larger than 77
2949 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2950 Details : lconf specifies an inode size of 4096 bytes when the default
2951 stripe_count is larger than 77. This conflicts with the default
2952 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2953 this case to avoid pinning too much memory for large EAs.
2955 ------------------------------------------------------------------------------
2957 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2962 Frequency : rare (extremely heavy IO load with hundreds of clients)
2964 Description: Client is evicted, gets IO error writing to file
2965 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2966 caused clients to be evicted for AST timeouts. The fixes in
2967 bug 5192 mean we no longer need to have such short AST timeouts
2968 so ldlm_timeout has been increased.
2971 Frequency : occasional during --force or --failover shutdown under load
2972 Bugzilla : 5949, 4834
2973 Description: Server oops/LBUG if stopped with --force or --failover under load
2974 Details : a collection of import/export refcount and cleanup ordering
2975 issues fixed for safer force cleanup
2978 Frequency : only filesystems larger than 120 OSTs
2979 Bugzilla : 5990, 6223
2980 Description: lfs getstripe would oops on a very large filesystem
2981 Details : lov_getconfig used kfree on vmalloc'd memory
2984 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2985 Bugzilla : 6242, 6243
2986 Description: reading from files that had been truncated to a non-zero size
2987 but never opened returned no data
2988 Details : ll_file_read() reads zeros from no-object files to EOF
2993 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
2994 Details : There's little harm in aborting MDS/OSS recovery and letting it
2995 try again, so I removed the LASSERT and return an error instead.
2997 Severity : enhancement
2999 Description: New debugging infrastructure for tracking down data corruption
3000 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3001 (b) cover more of the client-side code path, and (c) try to narrow
3002 down where problems occurred
3006 Bugzilla : 3819, 4364, 4397, 6313
3007 Description: Racing close and eviction MDS could cause assertion in mds_close
3008 Details : It was possible to get multiple mfd references during close and
3009 client eviction, leading to one thread referencing a freed mfd.
3011 Severity: : enhancement
3012 Bugzilla : 3262, 6359
3013 Description: Attempts to reconnect to servers are now more aggressive.
3014 Details : This builds on the enhanced upcall-less recovery that was added
3015 in 1.4.2. When trying to reconnect to servers, clients will
3016 now try each server in the failover group every 10 seconds. By
3017 default, clients would previously try one server every 25 seconds.
3022 Description: After recovery, certain operations trigger a failed
3023 assertion on a client.
3024 Details : Failing over an mds, using lconf -d --failover, while a
3025 client was doing a readdir() call would cause the client to
3026 LBUG after recovery completed and the readdir() was resent.
3028 Severity : enhancement
3030 Description: Default groups are now added by lconf
3031 Details : You can now run lconf --group <servicename> without having to
3032 manually add groups with lmc.
3035 Frequency : occasional
3037 Description: Nodes with an elan id of 0 trigger a failed assertion
3040 Frequency : always when accessing e.g. tty/console device nodes
3042 Description: tty and some other devices nodes cannot be used on lustre
3043 Details : file's private_data field is used by device data and lustre
3044 values in there got lost. New field was added to struct file to
3045 store fs-specific private data.
3048 Frequency : when exporting Lustre via NFS
3050 Description: NFSD failed occasionally when looking up a path component
3051 Details : NFSD is looking up ".." which was broken in ext3 directories
3052 that had grown large enough to become hashed.
3055 Frequency : Clusters with multiple interfaces not on the same subnet
3057 Description: Nodes will repeatedly try to reconnect to an interface which it
3058 cannot reach and report an error to the log.
3059 Details : Extra peer list entries will be created by lconf with some peers
3060 unreachable. lconf now validates the peer before adding it.
3063 Frequency : Only if a default stripe is set on the filesystem root.
3065 Description: Setting a default stripe on the filesystem root prevented the
3066 filesystem from being remounted.
3067 Details : The client was sending extra request flags in the root getattr
3068 request and did not allocate a reply buffer for the dir EA.
3071 Frequency : occasional, higher if lots of files are accessed by one client
3072 Bugzilla : 6159, 6097
3073 Description: Client trips assertion regarding lsm mismatch/magic
3074 Details : While revalidating inodes the VFS looks up inodes with ifind()
3075 and in rare cases can find an inode that is being freed.
3076 The ll_test_inode() code will free the lsm during ifind()
3077 when it finds an existing inode and then the VFS later attaches
3078 this free lsm to a new inode.
3082 Bugzilla : 6422, 7030
3083 Description: MDS deadlock between mkdir and client eviction
3084 Details : Creating a new file via mkdir or mknod (starting a transaction
3085 and getting the ns lock) can deadlock with client eviction
3086 (gets ns lock and trying to finish a synchronous transaction).
3089 Frequency : occasional
3090 Description: While starting a server, the fsfilt_ext3 module could not be
3092 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3093 kernels. Previously, lconf would still use the ext3 name
3094 when trying to load modules. Now, it will correctly use
3095 ext3 on 2.4 and ldiskfs on 2.6.
3097 Severity : enhancement
3098 Description: The default stripe count has been changed to 1
3099 Details : The interpretation of the default stripe count (0, to lfs
3100 or lmc) has been changed to mean striping across a single
3101 OST, rather than all available. For general usage we have
3102 found a stripe count of 1 or 2 works best.
3104 Severity : enhancement
3105 Description: Add support for compiling against Cray portals.
3106 Details : Conditional compiling for some areas that are different
3110 Frequency : occasional
3111 Bugzilla : 6409, 6834
3112 Description: Creating files with an explicit stripe count may lead to
3113 a failed assertion on the MDS
3114 Details : If some OSTs are full or unavailable, creating files may
3115 trigger a failed assertion on the MDS. Now, Lustre will
3116 try to use other servers or return an error to the
3120 Frequency : occasional
3122 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3123 caused lock timeout during readahead (since 1.4.2).
3124 Details : Processes doing readahead might match a lock that hasn't been
3125 granted yet if there are overlapping and conflicting lock
3126 requests. The readahead process waits on ungranted lock
3127 (original lock is CBPENDING), while OST waits for that process
3128 to cancel CBPENDING read lock and eventually evicts client.
3130 Severity : enhancement
3132 Description: Initial enabling of flock support for clients
3133 Details : Implements fcntl advisory locking and file status functions.
3134 This feature is provided as an optional mount flag (default
3135 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3136 locking are implemented yet, and those that are are not guaranteed
3137 to be completely correct in production environments.
3138 mount -t lustre -o [flock|noflock] ...
3141 Frequency : occasional
3143 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3144 assertion in the ext3 JBD (journaling) layer.
3145 Details : The b_committed_data struct is protected by the big kernel lock
3146 in 2.4 kernels, serializing journal_commit_transaction() and
3147 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3148 access to this struct. In 2.6 kernels there is finer grained
3149 locking to improve SMP performance of the JBD layer.
3153 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3158 Description: merge and cleanup kernel patches.
3161 -----------------------------------------------------------------------------
3163 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3165 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3166 versions of Lustre. Please update all servers and clients to
3167 version 1.4.2 or later at the same time. You must also run
3168 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3169 to update the configuration logs.
3171 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3172 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3173 - swab llog records of type '0' so we get proper header size/idx (5861)
3174 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3175 - fix rename of one directory over another leaking an inode (5953)
3176 - avoid SetPageDirty on 2.6 (5981)
3177 - don't re-add just-being-destroyed locks to the waiting list (5653)
3178 - when creating new directories, inherit the parent's custom
3179 striping settings if present parent (3048)
3180 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3181 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3182 - don't hold client locks on temporary worklist from l_lru (5666)
3183 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3184 - automatically evict dead clients (5921)
3185 - Update file size properly in create+truncate+fstat case (6196)
3186 - Do not unhash mountpoint dentries, do not allow removal of
3188 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3189 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3190 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3191 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3192 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3193 - Correctly handle reads of files with no objects (6243)
3194 - lctl recover will also mark a device active if deactivate used (5933)
3196 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3197 - allow --write-conf on an MDS with different nettype than client (5619)
3198 - don't write config llogs to MDS for mounts not from that MDS (5617)
3199 - lconf should create multiple TCP connections from a client (5201)
3200 - init scripts are now turned off by default; run chkconfig --on
3201 lustre and chkconfig --on lustrefs to use them
3202 - upcalls are no longer needed for clients to recover to failover
3204 - add --abort-recovery option to lconf to abort recovery on device
3206 - add support for an arbitrary number of OSTs (3026)
3207 - Quota support protocol changes.
3208 - forward compatibility changes to wire structs (6007)
3209 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3210 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3211 - improved "lustre status" script
3212 - initialize blocksize for non-regular files (6062)
3213 - added --disable-server and --disable-client configure options (5782)
3214 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3215 - Vanilla 2.4.29 support
3216 - increase maximum number of obd devices to 520 (6242)
3217 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3218 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3219 SLES 9 2.6.5 kernels
3220 - limit stripes per file to 160 (the maximum EA size) (6093)
3222 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3225 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3226 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3227 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3228 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3229 - small ext3 extents cleanups and fixes (5733)
3230 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3231 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3232 - avoid needless client->OST connect, fix handle mismatch (5317)
3233 - fix DLM error path that led to out-of-sync client, long delays (5779)
3234 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3235 - fix several locking issues related to i_size (5492,5624,5654,5672)
3236 - don't move pending lock onto export if it is already evicted (5683)
3237 - fix kernel oops when creating .foo in unlinked directory (5548)
3238 - fix deadlock in obdfilter statistics vs. object create (5811)
3239 - use time_{before,after} to avoid timer jiffies wrap (5882)
3240 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3241 - Do not leak request if server was not able to process it (5154)
3242 - If mds_open unable to find parent dir, make that negative lookup(5154)
3243 - don't create new directories with extent-mapping (5909, 5936)
3245 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3246 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3247 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3248 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3249 - support for RHEL4 kernel on i686 (5773)
3250 - provide error messages when incompatible logs are encountered (5898)
3252 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3253 * version 1.4.0.10 (1.4.1 release candidate 1)
3255 - don't keep a lock reference when lock is not granted (4238)
3256 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3257 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3258 - drop import inflight refcount on signal_completed_replay error (5255)
3259 - unlock page after async write error during send (3677)
3260 - handle missing objects in filter_preprw_read properly (5265)
3261 - no transno return for symlink open, don't save no-trasno open (3440)
3262 - don't try to complete elan receive that already failed (4012)
3263 - free RPC server reply state on error (5406)
3264 - clean up thread from ptlrpc_start_thread() on error (5160)
3265 - readahead could read extra page into cache that wasn't ejected (5388)
3266 - prevent races in class_attach/setup/cleanup/detach (5260)
3267 - don't dereference de->d_inode after l_dput of de (5458)
3268 - use "int" for stripe value returned from lock_to_stripe (5544)
3269 - mballoc allocation and error-checking fixes in 2.6 (5504)
3270 - block device patches to fix I/O request sizes in 2.6 (5482)
3271 - look up hostnames for IB nals (5602)
3272 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3273 - don't start multiple acceptors for the same port (5277)
3274 - fix incorrect LASSERT in mds_getattr_name (5635)
3275 - export a proc file for general "ping" checking (5628)
3276 - fix "lfs check" to not block when the MDS is down (5628)
3278 - service request history (4965)
3279 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3280 - create an "evict_client" /proc entry on OSTs, like the MDS has
3281 - fix mount usage message, return errors per mount(8) (5168)
3282 - change grep [] to grep "[]" in tests so they work in more UMLs
3283 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3284 - remove ancient LOV_MAGIC_V0 EA support (5047)
3285 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3286 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3287 - allow readahead allocations to fail when low on memory (5383)
3288 - mmap locking landed again, after considerable improvement (2828)
3289 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3291 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3294 - send OST transaction number in read/write reply to free req (4966)
3295 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3296 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3297 - the watchdog thread now runs as interruptible (5246)
3298 - flock/lockf fixes (but it's still disabled, pending 5135)
3299 - don't use EXT3 constants in llite code (5094)
3300 - memory shortage at startup could cause assertion (5176)
3302 - reorganization of lov code
3303 - single portals codebase
3305 - add extents/mballoc support (5025)
3306 - direct I/O reads in the obdfilter (4048)
3307 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3309 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3312 - send OST transaction number in read/write reply to free req (4966)
3313 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3314 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3315 - the watchdog thread now runs as interruptible (5246)
3316 - handle missing objects in filter_preprw_read properly (5265)
3317 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3318 - drop import inflight refcount on signal_completed_replay error (5255)
3319 - unlock page after async write error during send (3677)
3320 - return original error code on reconstructed replies (3761)
3321 - no transno return for symlink open, don't save no-trasno open (3440)
3323 - add pid to ldlm debugging output (4922)
3324 - bump the watchdog timeouts -- we can't handle 30sec yet
3325 - extra debugging for orphan dentry/inode bug (5259)
3327 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3330 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3331 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3332 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3333 - fix NULL dereference in /proc/sys/portals/routes (4827)
3334 - allow failed mdc_close() operations to be interrupted (4561)
3335 - stop precreate on OST before MDS would time out on it (4778)
3336 - don't send partial-page writes before EOF from client (4410)
3337 - discard client grant for sub-page writes on large-page clients (4520)
3338 - don't free dentries not owned by NFS code, check generation (4806)
3339 - fix lsm leak if mds_create_objects() fails (4801)
3340 - limit debug_daemon file size, always print CERROR messages (4789)
3341 - use transno after validating reply (3892)
3342 - process timed out requests if import state changes (3754)
3343 - update mtime on OST during writes, return in glimpse (4829)
3344 - add mkfsoptions to LDAP (4679)
3345 - use ->max_readahead method instead of zapping global ra (5039)
3346 - don't interrupt __l_wait_event() during strace
3348 - add software watchdogs to catch hung threads quickly (4941)
3349 - make lustrefs init script start after nfs is mounted
3350 - fix CWARN/ERROR duplication (4930)
3351 - return async write errors to application if possible (2248)
3352 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3353 - print NAL number in %x format (4645)
3354 - update barely-supported suse-2.4.21-171 series (4842)
3355 - support for sles 9 %post scripts
3356 - support for building 2.6 kernel-source packages
3357 - support for sles km_* packages
3359 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3362 - ignore -ENOENT errors in osc_destroy (3639)
3363 - notify osc create thread that OSC is being cleaned up (4600)
3364 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3365 - reconstruct ost_handle() like mds_handle() (4657)
3366 - create a new thread to do import eviction to avoid deadlock (3969)
3367 - let lconf resolve symlinked-to devices (4629)
3368 - don't unlink "objects" from directory with default EA (4554)
3369 - hold socknal file ref over connect in case target is down (4394)
3370 - allow more than 32000 subdirectories in a single directory (3244)
3371 - fix blocks count for O_DIRECT writes (3751)
3372 - OST returns ENOSPC from object create when no space left (4539)
3373 - don't send truncate RPC if file size isn't changing (4410)
3374 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3375 - bind to privileged port in socknal and tcpnal (3689)
3377 - rate limit CERROR/CWARN console message to avoid overload (4519)
3378 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3379 - basic mmap support (3918)
3380 - kernel patch series update from b1_4 (4711)
3382 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3385 - avoid crash during MDS cleanup with OST shut down (2775)
3386 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3387 - don't use bad inodes on the MDS (3744)
3388 - dynamic object preallocation to improve recovery speed (4236)
3389 - don't hold spinlock over lock dumping or change debug flags (4401)
3390 - don't zero obd_dev when it is force cleaned (3651)
3391 - print grants to console if they go negative (4431)
3392 - "lctl deactivate" will stop automatic recovery attempts (3406)
3393 - look for existing locks in ldlm_handle_enqueue() (3764)
3394 - don't resolve lock handle twice in recovery avoiding race (4401)
3395 - revalidate should check working dir is a directory (4134)
3397 - don't always mark "slow" obdfilter messages as errors (4418)
3399 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3402 - don't close LustreDB during write_conf until it is done (3860)
3403 - fix typo in lconf for_each_profile (3821)
3404 - allow dumping logs from multiple threads at one time (3820)
3405 - don't allow multiple threads in OSC recovery (3812)
3406 - fix debug_size parameters (3864)
3407 - fix mds_postrecov to initialize import for llog ctxt (3121)
3408 - replace config semaphore with spinlock (3306)
3409 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3410 - don't allow enqueue to complete on a destroyed export (3822)
3411 - down write_lock before checking llog header bitmap (3825)
3412 - recover from lock replay timeout (3764)
3413 - up llog sem before sending rpc (3652)
3414 - reduce ns lock hold times when setting kms (3267)
3415 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3416 - fix NULL deref and obd_dev leak on setup error (3312)
3417 - replace some LBUG about llog ops with error handling (3841)
3418 - don't match INVALID dentries from d_lookup and spin (3784)
3419 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3420 - fix invalid assertion in ptlrpc_set_wait (3880)
3422 - add libwrap support for the TCP acceptor (3996)
3423 - add /proc/sys/portals/routes for non-root route listing (3994)
3424 - allow setting MDS UUID in .xml (2580)
3425 - print the stack of a process that LBUGs (4228)
3427 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3430 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3431 - make sure to unset replay flag from failed open requests (3440)
3432 - if default stripe count is 0, use OST count for inode size (3636)
3433 - update parent mtime/ctime on client for create/unlink (2611)
3434 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3435 - free recovery state on server during a forced cleanup (3571)
3436 - unregister_reply for resent reqs (3063)
3437 - loop back devices mounting and status check on 2.6 (3563)
3438 - fix resource-creation race that can provoke i_size == 0 (3513)
3439 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3440 - more debugging for page-accounting assertion (3746)
3441 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3442 - avoid deadlock after precreation failure (3758)
3443 - fix race and lock order deadlock in orphan handling (3450, 3750)
3444 - add validity checks when grabbing inodes from l_ast_data (3599)
3446 - add /proc/.../recovery_status to obdfilter (3428)
3447 - lightweight CDEBUG infrastructure, debug daemon (3668)
3448 - change default OSC RPC parameters to be better on small clusters
3449 - turn off OST read cache for files smaller than 32MB
3450 - install man pages and include them in rpms (3100)
3451 - add new init script for (un)mounting lustre filesystems (2593)
3452 - run chkconfig in %post for init scripts (3701)
3453 - drop scimac NAL (unmaintained)
3455 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3458 - clean kiobufs before and after use (3485)
3459 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3460 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3461 - change init script to fail more clearly if not run as root (1528)
3462 - allow clients to reconnect during replay (1742)
3463 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3464 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3465 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3466 - fix cond_resched (3554)
3467 - only evict unfinished clients after recovery (3515)
3468 - allow bulk resend, prevent data loss (3570)
3469 - dynamic ptlrpc request buffer allocation (2102)
3470 - don't allow unlinking open directory if it isn't empty (2904)
3471 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3472 - remove extraneous obd dereference causing LASSERT failure (3334)
3473 - don't use get_cycles() when creating temp. files on the mds (3156)
3474 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3475 - handle EEXIST for set-stripe, set proper directory name (3336)
3477 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3478 - fix ksocknal_fmb_callback() error messages (2918)
3480 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3483 - don't copy lvb into (possibly NULL) reply on error (2983)
3484 - don't deref dentry after dput, don't free lvb on error (2922)
3485 - use the kms to determine writeback rpc length (2947)
3486 - increment oti_logcookies when osc is inactive (2948)
3487 - update client's i_blocks count via lvb messages (2543)
3488 - handle intent open/close of special files properly (1557)
3489 - mount MDS with errors=remount-ro, like obdfilter (2009)
3490 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3491 - don't use cancelling-locks' kms values (2947)
3492 - use highest lock extent for kms, not last one (2925)
3493 - don't dereference ERR_PTR() dentry in error handling path (3107)
3494 - fix thread race in portals_debug_dumplog() (3122)
3495 - create lprocfs device entries at setup instead of at attach (1519)
3496 - common AST error handler, don't evict client on completion race (3145)
3497 - zero nameidata in detach_mnt in 2.6 (3118)
3498 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3499 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3500 - initialize RPC timeout timer earlier for 2.6 (3219)
3501 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3502 - print nal/nid for unknown nid (3258)
3503 - additional checks for oscc recovery before doing precreate (3284)
3504 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3505 - don't crash in mdc_close for bad permissions on open (3285)
3506 - zero i_rdev for non-device files (3147)
3507 - clear page->private before handing to FS, better assertion (3119)
3508 - tune the read pipeline (3236)
3509 - fix incorrect decref of invalidated dentry (2350)
3510 - provide read-ahead stats and refine rpc in flight stats (3328)
3511 - don't hold journal transaction open across create RPC (3313)
3512 - update atime on MDS at close time (3265)
3513 - close LDAP connection when recovering to avoid server load (3315)
3514 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3515 - don't leak open file on MDS after open resend (3325)
3516 - serialize filter_precreate and filter_destroy_precreated (3329)
3517 - loop device shouldn't call sync_dev() for nul device (3092)
3518 - clear page cache after eviction (2766)
3519 - resynchronize MDS->OST in background (2824)
3520 - refuse to mount the same filesystem twice on same mountpoint (3394)
3521 - allow llmount to create routes for mounting behind routers (3320)
3522 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3523 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3524 - fix rare problem with rename on htree directories (3417)
3526 - allow default OST striping configuration per directory (1414)
3527 - fix compilation for qswnal for 2.6 kernels (3125)
3528 - increase maximum number of MDS request buffers for large systems
3529 - change liblustreapi to be useful for external progs like lfsck (3098)
3530 - increase local configuration timeout for slow disks (3353)
3531 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3533 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3536 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3537 - don't overwrite extent policy data in reply if lock was blocked (2901)
3538 - drop filter export grants atomically with removal from device (2663)
3539 - del obd_self_export from work_list in class_disconnect_exports (2908)
3540 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3541 - swab reply message in mdc_close, other PPC fixes (2464)
3542 - fix destroying of named logs (2325)
3543 - overwrite old logs when running lconf --write_conf (2264)
3544 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3545 - fix race in target_handle_connect (2898)
3546 - mds_reint_create() should take same inode create lock (2926)
3547 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3548 - don't close files for self_export to avoid uninitialized obd (2936)
3549 - allow MDS with the same name as client node (2939)
3550 - hold dentry reference for closed log files for unlink (2325)
3551 - reserve space for all logs during transactions (2059)
3552 - don't evict page beyond end of stripe extent (2925)
3553 - don't oops on a deleted current working directory (2399)
3554 - handle hard links to targets without a parent properly (2517)
3555 - don't dereference NULL lock when racing during eviction (2867)
3556 - don't grow lock extents when lots of conflicting locks (2919)
3558 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3561 - account for cache space usage on clients to avoid data loss (974)
3562 - lfsck support in lustre kernel code (2349)
3563 - reduce journal credits needed for BRW writes (2370)
3564 - orphan handling to avoid losing space on client/server crashes
3565 - ptlrpcd can be blocked, stopping ALL progress (2477)
3566 - use lock value blocks to assist in proper KMS, faster stat (1021)
3567 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3568 - recovery for initial connections (2355)
3569 - fixes for mds_cleanup_orphans (1934)
3570 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3571 - block all file creations until orphan recovery completes (1901)
3572 - client remove rq_connection from request struct (2423)
3573 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3574 - recovery timer race (2670)
3575 - mdc_close recovey bug (2532)
3576 - ptlrpc cleanup bug (2710)
3577 - mds timeout on local locks (2588)
3578 - namespace lock held during RPCs (2431)
3579 - handle interrupted sync write properly (2503)
3580 - don't try to handle a message that hasn't been replied to (2699)
3581 - client assert failure during cleanup after abort recovery (2701)
3582 - leak mdc device after failed mount (2712)
3583 - ptlrpc_check_set allows timedout requests to complete (2714)
3584 - wait for inflight reqs when ptlrpcd finishes (2710)
3585 - make sure unregistered services are removed from the srv_list
3586 - reset bulk XID's when resending them (caught by 1138 test)
3587 - unregister_bulk after timeout
3588 - fix lconf error (2694)
3589 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3590 - readahead locks pages, leaves pending causing memory pressure (2673)
3591 - increase OST request buffers to 4096 on large machines (2729)
3592 - fix up permission of existing directories in simple_mkdir (2661)
3593 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3594 - don't assign transno to errored transactions (2742)
3595 - don't delete objects on OST if given a bogus objid from MDS (2751)
3596 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3597 - if rq_no_resend, then timeout request after recovery (2432)
3598 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3599 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3600 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3601 - MDS recovery shouldn't delete objects that it has given out (2730)
3602 - if enqueue arrives after completion, don't clobber LVB (2819)
3603 - don't unlock pages twice when trigger_group_io returns error (2814)
3604 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3605 - don't write pages to disk if there was an error (1450)
3606 - don't ping imports that have recovery disabled (2676)
3607 - take buffered bytes into account when balancing socknal conn (2817)
3608 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3609 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3610 - remove little-endian swabbing of llog records (1987)
3611 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3612 - retry reposting request buffers if they fail (1191)
3613 - grow extent at grant time to avoid granting a revoked lock (2809)
3614 - lock revoke doesn't evict page if covered by a second lock (2765)
3615 - disable VM readahead to avoid reading outside lock extents (2805)
3617 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3618 - updated kernel patches for hp-2.4.20 kernel (2681)
3620 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3623 - fix truncated write corruption (2366)
3624 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3626 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3627 - ensure len(uuid) < 37 in lmc (1171)
3628 - fix ia64 OOPS in llog_test (2255)
3629 - zero end of page at obdfilter for partial page writes (2648)
3630 - don't leave stale dentries around after renames (bug 2428)
3631 - fix timeouts when evicting a client with a single lock held (2642)
3632 - set deadline for the initial HELLO message to drain (2634)
3633 - print out dotted-quad IP addresses in the socknal (2302)
3635 - additional debugging for MDS client eviction problem (2443)
3636 - fix mkfsoptions support for osts (2603, 2604)
3638 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3641 - add series for the vanilla 2.6.0 kernel
3642 - add series for the vanilla 2.4.24 kernel
3643 - add series for a cray x86/64 UL kernel drop
3644 - fix xattr patches for the vanilla 2.4.19 series
3646 - generate true UUIDs in lmc (1171)
3647 - have portals stack dumping break in UML (2466)
3648 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3649 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3650 - wake callback waiting threads on client eviction (2460)
3651 - Add --ptldebug and --subsystem to lmc (1719)
3652 - update assertion to allow safe interrupt allocation
3653 - set rq_no_resend for cancel requests (2432)
3654 - recalculate ptlrpcd timeout after resend (2494)
3655 - call vfs_rmdir when removing pending directories (2368)
3656 - fix renaming a file to itself (2429)
3657 - lmc creates a default one-stripe lov (2454)
3658 - expand procfs space to handle large clusters (2326)
3659 - increase UML stack to avoid overflow
3660 - update lconf's list of debug and subsystem masks
3661 - fix lfs find --obd (2510)
3662 - /proc tunable for disabling filter read caching (2591)
3663 - stop rpm packages from altering slapd.conf (2301)
3664 - disable nagle in the socknal under 0conf (2578)
3665 - choose mds inode size based on stripe count (2572)
3666 - fix kernel-source rpm problems (2516)
3668 - add --disable-doc to avoid pdf generation (2421)
3669 - update documentation, tests, type-os, comments
3670 - avoid format warnings on ia64
3671 - remove the TOE NAL
3672 - tiny code cleanups by removing unused fields
3674 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3677 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3678 - remove the most blatant lies from BUILDING file (2371)
3679 - change default debug level to reasonable production setting
3680 - reduce client side cache size to reduce cache flush time
3681 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3682 - make TCP zerocopy and pinger support enabled by default (2476)
3683 - sync writes completed after process exits caused crashes (2319)
3684 - maintain correct mount count on the MDS (2356)
3685 - backout 1557, because 2316 wasn't really fixed
3686 - better file I/O statistics gathering in /proc
3687 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3688 - another kernel patch to fix zero-copy TCP function export
3689 - don't take duplicate lock when processing re-sent getattr (2420)
3690 - lctl uses obd_self_export instead of creating new conn (2353)
3691 - MDS/OST recovery case which requires object creation asserted (2425)
3692 - move lfs from /usr/sbin to /usr/bin in packages
3693 - fix race between mds_client_add and mds_client_free (2417)
3694 - use kmalloc instead of slabs in portals (2430)
3695 - don't create duplicate records when a failover MDS is present (2442)
3696 - remove unnecessary mount age check (2332)
3697 - don't remove directory inodes from locks prematurely (2451)
3698 - don't break if MDS service name is the same as hostname (2103)
3699 - fix races in client write RPC generation when cache full (2482)
3701 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3704 - remove now-unused request->rq_obd (278)
3705 - if an allocation fails, print out how much memory we've used (1933)
3706 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3707 - add the "configurable stack size" patch to most series files (1256)
3708 - ability to write large log records, for 100+ OST configs (2306)
3709 - fix NULL deref when filter_prep fails (2314)
3710 - fix operator precedence error in filter_sync
3711 - dynamic allocation of socknal TX descriptors (2315)
3712 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3713 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3714 - ensure that CWARN messages in Portals always get to the syslog
3715 - __init/__exit are not for prototype decls (ldlm_init/exit)
3716 - x86-64 compile warning fixes
3717 - fix gateway LMC keyword conflict (2318)
3718 - fix MDS lock inversions in getattr/reint paths (1844)
3719 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3720 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3721 - DLM race condition prevented some lock evictions (2328)
3722 - ENOMEM detection and retry on socknal sends (2230)
3723 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3724 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3725 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3726 - better error messages when a client is rejected during recovery (1505)
3727 - avoid cancelling locks which were never granted, after failure (2330)
3728 - fix i_sem/journal inversion in mds_client_add (2333)
3729 - fix truncate/getattr lock cycle deadlock (2334)
3730 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3731 - fix two rare exit paths which could leak an l_lock() ref (2321)
3732 - fencepost error in MDS/OST orphan recovery (2226)
3733 - make log record alignment 8 bytes (1988)
3734 - lstripe now fails when requested offset > ost_count (2237)
3735 - ensure that all kernel series have a complete list.h (1607)
3736 - fix crashes in special-file operations (2316)
3737 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3739 - allow configurable automake binary, for testing new versions
3740 - small update to the lfs documentation
3742 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3744 * fix negative export reference count in fsfilt_sync (2312)
3746 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3747 * release candidate 0.9.1
3749 - orphans are moved into the PENDING directory for possible recovery
3750 - replayed opens now open by fid for orphan/rename safety (1042)
3751 - last close of an orphan inode generates a transno (683)
3752 - chdir() and mount() now pin the directory entry (1020)
3753 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3754 - discard very old requests without processing them (1502)
3755 - remove some common, well-understood CERRORs (1505)
3756 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3757 - clear "grant" flags in OST replies until OST grant code lands (1644)
3758 - fix read performance by not clobbering i_blksize on client (1598)
3759 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3760 - make LRU size a /proc tunable, clears locks when reduced (707)
3761 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3762 - server locks take references on exports now (1558)
3763 - build fixes for 2.4.20-rh trees (1663)
3764 - return an error from lov_create if all OSCs are inactive (1751)
3765 - fix import levels when a reconnect happens without a timeout (1597)
3766 - exit early from mds_open if we get a lookup error (1749)
3767 - partial page read at EOF wouldn't wait for disk before sending (1642)
3768 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3769 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3770 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3771 - write/truncate lock inversion (1639)
3772 - Don't auto-load obdclass, portals modules during cleanup (1495)
3773 - fix timestamps from jumping to "now" (1763)
3774 - extra journal assertions (1648)
3775 - add an extra multiunlink test (1771)
3776 - fix read_record/write_record API (1776)
3777 - fix leak of offset_extent, possible incorrect i_size later (1772)
3778 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3779 - Debugging for the kqswnal_get_idle_tx problems (1820)
3780 - Allow recovery to be attempted multiple times (1536)
3781 - Write out MDS last_rcvd file after it is first created (1600)
3782 - Fix tx_descriptor leak in failed transmit situations (1827)
3783 - ext3 journaling fixes for assertion failure after IO error (1871)
3784 - class_export_put() on freed export after completion AST error (1896)
3785 - Fix revalidate looping in VFS (1322)
3786 - Don't access a freed export during MDS_REINT timeout (1521)
3787 - Add open-unlink recovery support on the MDS (1673,1764)
3788 - Return an error if no MDS data was read from last_rcvd (1946)
3789 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3790 - Don't setup a disk device that doesn't match exported UUID (317)
3791 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3792 - avoid committing NULL handle in force close
3793 - local.sh is now a one-stripe LOV configuration
3794 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3795 - fix ext3 htree duplicate directory entry corruption (1516)
3796 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3797 - update to new LOV EA format (2097)
3798 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3800 2003-06-15 Phil Schwan <phil@clusterfs.com>
3803 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3804 - per-import/export recovery handling (958, 931, 959)
3805 - multiple last-rcvd slots, for serving multiple FSes (949)
3806 - connections are again shared between multiple imp/exports (963, 964)
3807 - "umount -f" would hang if any requests needed to be sent (393, 978)
3808 - avoid pinning large req buffer by copying for queued messages (989)
3809 - add "uuid" to "lctl device" command to help upcalls (991)
3810 - "open" RPCs with transnos would confuse recovery counters (1037)
3811 - do proper endian conversion of all wire messages (288, 340, 891)
3812 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3813 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3814 - fix ptlrpc_connection leak in target_handle_connect (1174)
3815 - fix import refcounting bug in OST and MDS cleanup (1134)
3816 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3817 - fix very unlikely obd_types race condition (501)
3818 - remove osc_open hack for echo_client (1187)
3819 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3820 - a failure in read_inode2 leads to deadlock (1139)
3821 - cancel ack-locks as soon as transaction is committed (1072)
3822 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3823 - make sure to commitrw after any preprw to avoid deadlock (1162)
3824 - failing to execute a file in a lustre FS would lock inode (1203)
3825 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3826 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3827 - obd_uuid2tgt would walk off the end of the list (1255)
3828 - on IA64 the osc would give portals incorrect bulk size (1258)
3829 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3830 - fix lock inversion caused by new llite matching code (1282)
3831 - limit the number of dirty pages on a client to 10MB (1286)
3832 - timed out locks were not being corrected cancelled (1289)
3833 - fix O_DIRECT above 4GB on IA-32 (1292)
3834 * major user-visible changes
3835 - fail out/fail over policy now controlled by the upcall (993)
3837 - add OBD_PING to check server availability and failure (954)
3838 - lustre messages are now sent in sending host order (288, 340, 891)
3839 - add eadatalen to MDS getattr reply (340)
3840 - OST read replies may contain second buffer, with per-page status (593)
3842 2003-03-11 Phil Schwan <phil@clusterfs.com>
3845 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3846 - failed open()s could cause deadlock; fixed (867, 869)
3847 - stop cancelling OST locks when files are closed (481)
3848 - overlapping XID spaces caused network corruption (851, 853)
3849 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3850 - setattr_raw added, to do single-RPC, server-side setattrs
3851 - lmc/lconf syntax change for OST UUIDs
3852 - fix crashy race condition between ptlrpc_free_req and osc_close
3853 - don't use request in mdc_enqueue if we hit a timeout (889)
3854 - don't set the inode i_size for regular files from the MDS (896)
3855 - handle out of order completion AST (842)
3856 - don't LBUG if a lock request times out after receiving AST (913)
3857 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3858 - if a bad lock AST arrives, send an error instead of dropping entirely
3859 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3860 - fix leak in bulk IO when only partially completed (899, 900, 926)
3861 - fix O_DIRECT for ia64 (55)
3862 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3863 - C-z after timeout could hang a process forever; fixed (977)
3865 - client-side I/O cache (678, 924, 929, 941, 970)
3867 - READPAGE and SETATTRs which don't take server-side locks get
3870 2003-02-11 Phil Schwan <phil@clusterfs.com>
3873 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3874 - Fix fsfilt_extN_readpage() to read a full page of directory
3875 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3876 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3877 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3878 - Most intent operations were removed, in favour of a new RPC mode
3879 that does a single RPC to the server and bypasses most of the VFS
3880 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3881 - Aggressively cancel local locks on DLM servers
3882 - mds_reint_unlink sends EA to the client if it's the last nlink.
3883 client uses that EA to unlink OST objects.
3884 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3885 - recursive symlinks were fixed (439)
3886 - fixed NULL deref in DEBUG_REQ
3887 - filter_update_lastobjid no longer calls sync, which annoyed extN
3888 - fixed multi-client small-writes to a single file problem (445)
3889 - fixed mtime updates during file writes (607)
3890 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3891 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3892 - fixed Linux OST size reporting problem (444, 656)
3893 - OST now updates object mtime with writes or setattr (607, 619)
3894 - client verifies file size before zeroing page past EOF (445)
3895 - OST now writes last allocated objid to disk with allocation (108)
3896 - LOV on echo now works (409)
3898 - mds_reint_unlink sends a new buffer, with the EA included. this
3899 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3900 set if a regular file was being unlinked, and it was the last link
3901 - use PtlGet from the target for bulk writes (315)
3902 - OST now updates object mtime with writes or setattr (607, 619)
3903 - LDLM now has a grant-time callback to revalidate locked items, if
3905 - Many MDS operations were reorganized to combat race conditions
3907 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3908 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3910 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3913 - Fully reactivate OST imports after reconnection (512, others)
3914 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3915 - More graceful error handling for truncating on dead OST (515)
3916 - Don't error out unless we're actually accessing dead stripes (474)
3917 - Fix garbage sizes when stripes are missing (410)
3918 - LRU counters were broken, causing constant lock purge (433, 432)
3919 - garbage on read from stripes with failed OSTs (441)
3920 - mark OSCs as active before reconnecting during recovery (438)
3921 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3922 - lfind did not preserve OST order in output (443)
3923 - symlinks cause hung clients, incorrect data (439)
3924 - stop dereferencing request after dropping refcount (457)
3925 - don't LASSERT(spin_is_locked) on non-SMP (455)
3926 - fixes for many rename() bugs
3927 - fstat didn't correctly synchronize attributes (399)
3928 - server must handle lock cancellation during blocking AST prep (487)
3929 - bulk descriptors were free()d too soon (511)
3930 - fix paths in lconf, which would load incorrect modules (451, 507)
3931 - fix confusing lconf 'host not found' error message (386)
3932 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3933 - fix race condition in mdc_blocking_ast() for inode access (526)
3934 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3935 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3936 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3937 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3938 - fix lstripe in several ways (kernel side) (527)
3939 - fix request leak in ldlm_cli_enqueue (262)
3940 - incorrect OSC was marked inactive after OST failure
3941 - call mds_fs_cleanup before unmounting filesystem (524)
3942 - fix races between taking ns_lock and ldlm_lock_change_resource
3943 - fix races updating LOV export open file list
3944 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3945 - fix recovery NULL deref in ldlm_cli_cancel_unused
3946 - fix some DLM races by using new hash table for lock handles (419)
3947 - permit the client to specify desired inodes, at replay
3948 - duplicate requests when we queue them for replay reintegration
3949 - fix last_rcvd offset calculation
3950 - sync after each recovered transaction, so we always make progress
3951 - never, not always, ERESTART requests without transnos
3952 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3954 - skip replay if the MDS says that the client is already connected
3955 - don't check for a recovery-enabled export to match lctl's UUID
3956 - don't INC_USE_COUNT for phantom exports
3957 - don't crash when cleaning up phantom exports (567)
3958 - don't double-finish or set replay data for errored mdc_open requests
3959 - abort requests when they time out, so we don't get old replies
3960 - send/receive replies for AST messages again
3961 - if the client says that it doesn't have the lock, cancel it on the
3963 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3964 mark it as destroyed, it will all work out when decref is called
3965 - fix module use counts (22, 581)
3967 - ASTs now expect a reply (server cancels lock on error reply)
3969 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3972 - fix many simultaneous client startup (392)
3973 - fix dentry->d_it clobbering
3974 - credentials weren't being shipped for readdir/getattr operations
3975 - remove invalid assertions triggered during some concurrent MD
3977 - proper Lustre versions added (336, 389)
3978 - fix memory leak for create error case (398)
3979 - fix LOV locking bug that would get cli/srv out of sync
3980 - fix echo client over LOV (409)
3981 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3982 - fix double-O_EXCL intent crash (424)
3983 - avoid sending multiple lock CANCELs (352)
3985 - MDS can do multi-client recovery (modulo bugs in new code)
3987 - many updates, edits, cleanups
3989 2002-11-18 Phil Schwan <phil@clusterfs.com>
3992 - fix null d_it dereference (346)
3993 - fix full OST/dbench hang (333)
3994 - fix permission problem with file removal (286)
3995 - fix removal of OSCs from LOV when they fail
3996 - fix NULL deref during bulk timeout (214)
3997 - fix problems related to multiple filesystems on one MDS (241)
3998 - fixed serious subtle metadata locking bugs
3999 - free locks on clients when inodes are removed due to memory
4001 - fix inode pointer in lock data (285)
4002 - partial support for multiple MDS on a single host (241)
4003 - data locks weren't cancelled at clear_inode time (290, 311)
4004 - intent locks could lead to unbounded lock growth (205)
4005 - added a maximum lock count, an LRU list, and a flusher
4006 - fix multiple rename (365)
4007 - properly abstracted the echo client
4008 - OSC locked 1 byte too many; fixed
4009 - rewrote brw callback code:
4010 - fixed recovery bugs related to LOVs (306)
4011 - fixed too-many-pages-in-one-write crash (191)
4012 - fixed (again) crash in sync_io_timeout (214)
4013 - probably fixed callback-related race (385)
4015 - Add capability to MDS protocol
4016 - LDLM cancellations and callbacks on different portals
4018 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4021 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4022 - defer open object destruction to close time (601981, 138)
4023 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4024 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4025 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4026 - don't call obd_setattr in ll_file_release for destroyed objects
4027 * protocol change to lustre_msg: move |version| and add |flags|
4028 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4029 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4030 * added replay of create, unlink, link and rename operations during
4031 MDS failover; recovery should be much more robust now
4032 * remove failed OSCs from LOVs (only lov_create uses this so far)
4033 * the lustre-HOWTO was brought (more) up to date (582544)
4035 2002-10-23 Phil Schwan <phil@clusterfs.com>
4038 - in-use dentries weren't being reused properly (617851)
4039 - prevent multiple LDLM setup (599178)
4040 - fix LOV size calculations for truncate (617853)
4041 - fix client handling of MDS intent errors (POSIX)
4042 - fix permission bug in lovstripe.c test (624321)
4043 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4044 - truncate past end of file could corrupt data
4045 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4046 - a race in recovery could return ETIMEDOUT to apps (623947)
4047 - building outside the source directory was fixed
4048 * the lustre-HOWTO was brought (more) up to date (582544)
4049 * major progress was made on recovery functionality
4051 2002-10-10 Phil Schwan <phil@clusterfs.com>
4054 - recovery deadlock fix
4055 - rm -rf causes LBUG fix (617817)
4056 - file open by multiple tasks fix (618962)
4057 - directory permissions bugs (602707 and 620007)
4058 - journal_stop fixed with locking (611313)
4059 - O_APPEND failures resolved (618273, perhaps 614459)
4060 - lconf PATH fix (619770)
4061 - IA64 build fix (621450)
4062 - RPC buffer sizes scale with amount of memory
4064 2002-10-01 Phil Schwan <phil@clusterfs.com>
4067 - locks would be cancelled without throwing away data pages,
4068 resulting in inconsistent data (605627)
4069 - inode attributes were not always being refreshed (605627, 612449)
4070 - lconf now continues to cleanup after lctl reports an error
4071 - MDS now enforces user permissions (602707)
4072 - lprocfs cleanup fixed, but not yet enabled (614157)
4073 - fixed infinite server hang, should a client not respond to an AST
4074 - avoid going into recovery if user calls readlink() with a buffer
4075 that's too small (613941)
4076 - AST RPCs no longer require replies (614867) -- this may be changed
4077 - don't crash server if client sends an IOV that's too big (611336)
4078 - fixed lock conversion deadlock (611892)
4079 - fixed the following of symlinks (614622)
4080 * recovery: the server can remove locks from a client that dies, other
4081 clients can make progress
4082 * more extN patch fixes
4083 * compile-time configurable ptlrpc buffer allocations
4085 - collaborative read cache document
4086 - Lustre Lite Performance CDR document-in-progress
4088 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4091 - fix typo in patch-2.4.18
4093 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4096 - clear ptlrpc request each time in handle_incoming_request()
4097 - unlink of files now destroys the object on the OST
4099 2002-09-19 Peter Braam <braam@clusterfs.com>
4101 * add hard link support
4102 * change obdfile creation method
4103 * kernel patch changed
4105 2002-09-19 Peter Braam <braam@clusterfs.com>
4108 - stack overflow bug in extN fixed
4110 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4112 * documentation updates
4113 - add man pages for config tools
4114 - update tests/README to describe testing with new config tools
4115 - finish metadata API descriptions
4116 * bug fixes and cleanups
4117 - statfs workaround for 16TB limit
4118 - LOV stripe allocation improved, can stripe on subset of OSTs
4119 - LOV file size/IO offset was wrong for files > 4GB in size
4120 - object EA data was being dropped, caused files to be unreadable
4121 - memory overflow with non-LOV OST caused memory corruption
4122 - fixed regression tests to work with new config tools, obdfilter
4123 - fixed bug when directory size became larger than 1 block
4124 - fixed bug (for single client case) when PWD was deleted
4125 - invalidate local directory pages when doing intent-based ops
4126 - avoid LDLM oops when lock callback contained bad data
4128 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4130 * documentation updates
4131 * bug fixes and cleanups
4132 - configuration tools
4135 - 64-bit compile warnings
4136 - 64-bit internal statfs data
4138 * test_brw on persistent OST devices
4140 * lprocfs (disabled)
4142 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4144 * documentation updates
4145 * bug fixes and cleanups
4146 * configuration tools
4148 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4151 * many small fixes to 0_5_4
4152 * io/network handling
4153 * thinkos in MDS operations
4155 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4158 * crucial basic fixes to 0.5.3
4159 * IOR, Iozone work over Elan
4162 2002-08-07 Phil Schwan <phil@clusterfs.com>
4163 * version 0_5_3, our first alpha
4164 * we use the new Portals iovs
4165 * documentation updates
4166 * bug fixes and cleanups
4167 * small changes in the DLM wire protocol
4169 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4170 * version 0_5_1 with some initial stability,
4171 * locking on MD and file I/O.
4172 * documentation updates
4173 * several bug fixes since 0.5.0
4174 * small changes in wire protocol
4176 2002-07-18 Phil Schwan <phil@clusterfs.com>
4178 * delivered as Lustre Light Alpha
4179 * fixed a crash after handling invalid MDS requests
4180 * fixed directory pages for architectures with non-4k pages sizes
4182 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4183 * release version v0_4_4
4184 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4186 * Fixes a number of interruption problems with OST operations.
4187 * Update documentation for portals header changes
4188 * Move all wire protocol structs/defines to lustre_idl.h
4189 * Fixes symlink length bug.
4190 * Add tcpdump to repository.
4192 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4193 * release version v0_4_3
4194 * Fixes statfs for inodes on extN.
4195 * Fixes bug in runtests which would delete /etc/hosts.
4196 * Use 64-bit object IDs wherever possible (not into VFS though)
4197 Remove ost_get_info, which is unused by lustre, and out of date.
4199 2002-07-03 Peter Braam <braam@clusterfs.com>
4200 * release version v0_4_2 Fixes a lookup error (type not passed)
4201 * move forward to head of Portals
4202 * move forward to latest Lustre kernel
4204 2002-06-25 Peter Braam <braam@clusterfs.com>
4205 * release version v0_4_1. Hopefully stable on single node use.