4 2.6.9-55.0.9.EL (RHEL 4),
5 2.6.16.53-0.8 (SLES 10),
6 2.6.18-8.1.14.EL5 (RHEL 5).
7 2.6.18.8 vanilla (kernel.org)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.9-42.0.10.EL (RHEL 4),
11 2.6.16 - 2.6.21 vanilla (kernel.org)
12 * Recommended e2fsprogs version: 1.40.2-cfs1
13 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
17 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
18 Details : Update which_patch & target file for SLES10 latest kernel.
21 Description: Remove client patches from SLES 10 kernel.
22 Details : This causes SLES 10 clients to behave as patchless clients
23 even on a Lustre-patched (server) kernel.
25 Severity : enhancement
27 Description: self-adjustable client's lru lists
28 Details : use adaptive algorithm for managing client cached locks lru
29 lists according to current server load, other client's work
30 pattern, memory activities, etc. Both, server and client
31 side namespaces provide number of proc tunables for controlling
34 Severity : enhancement
36 Description: light-weight GSS support
37 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
38 and reduce performance overhead.
40 Severity : enhancement
42 Description: Linux keyring support
43 Details : Support using service of Linux keyring for Lustre GSS internal
44 context refresh/cache mechanism.
48 Description: Fix errors in lfs documentation
49 Details : Fixes man pages
53 Description: don't use GFP_* in generic Lustre code.
54 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
58 Description: obdclass is limited by single OBD_ALLOC(idarray)
59 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
63 Description: warnings with build patchless client with vanila 2.6.19 and up
64 Details : change old ctl_table style and replace ctl_table/ctl_table_header
65 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
69 Frequency : Only for SLES
70 Description: sanity_quota fail test_1
71 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
72 file, which makes TSTID[2] inunique.
76 Frequency : Only for SLES9
77 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
78 Details : Update target/ChangeLog/which_patch .
82 Frequency : Only for test_10 in sanity-quota.sh
83 Description: a bug in quota test code
84 Details : checking a test flag in a wrong place
88 Frequency : Only for test_13 in sanity-quota.sh
89 Description: a bug in quota test script
90 Details : shouldn't check overfull allocated quota
94 Description: jbd statistics
95 Details : Port older jbd statistics patch for sles10
97 Severity : enhancement
99 Description: Add printing inode info into message about error in writepage.
104 Description: reply_lock_interpret crash due to race with it and lock cancel.
105 Details : Do not replay locks that are being cancelled. Do not reference
106 locks by their address during replay, just by their handle.
110 Frequency : When flocks are used.
111 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
112 Details : Flock locks might destroy just granted lock if it could be merged
113 with another existing flock, this is done in completion handler,
114 so teach ldlm_cli_enquque_fini that this is a valid case for
120 Description: Oops in read and write path when failing to allocate lock.
121 Details : Check if lock allocation failed and return error back.
125 Description: lstripe command fails for valid OST index
126 Details : The stripe offset is compared to lov->desc.ld_tgt_count
127 instead of lov->desc.ld_active_tgt_count.
131 Description: sanity.sh failed test 103
132 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
137 Description: df doesn't work properly if diskfs blocksize != 4K
138 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
142 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
143 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
147 Description: obd_health_check_timeout too short
148 Details : set obd_health_check_timeout as 1.5x of obd_timeout
152 Description: llapi_file_create() does not allow some changes
153 Details : add llapi_file_open() that allows specifying the mode and
154 open flags, and also returns an open file handle.
158 Description: merge and cleanup kernel patches.
159 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
163 Description: Add journal checksum support.(Kernel part)
164 Details : The journal checksum feature adds two new flags i.e
165 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
166 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
167 indicates that the commit block contains the checksum for
168 the blocks described by the descriptor blocks. Now commit
169 record can be sent to disk without waiting for descriptor
170 blocks to be written to disk. This behavior is controlled
171 using JBD2_FEATURE_ASYNC_COMMIT flag.
175 Description: OSS needs mutliple precreate threads
176 Details : Add ability to start more than one create thread per OSS.
180 Description: Sanity.sh test_65k failed: setstripe should have succeeded
181 Details : execute lfs setstripe on client
185 Description: mds_obd_create error creating tmp object
186 Details : When the user sets quota on root, llog will be affected and can't
187 create files and write files.
190 Frequency : Always on ia64 patchless client, and possibly others.
192 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
193 Details : This allows the patchless client to be loaded on architectures
198 Description: RedHat Update kernel for RHEL5
199 Details : Add the kernel config file for RHEL5.
203 Description: RedHat Update kernel for RHEL5
204 Details : Modify the kernel config file more closer RHEL5.
208 Description: Build failure against Centos5 (RHEL5)
209 Details : Define PAGE_SIZE when it isn't present.
213 Description: test_7 on sanity-quota.sh can't be used on separate machines
214 Details : add facet to handle it
218 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
219 Details : with NFS, the anon dentry's parent was set to itself in
220 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
221 corresponding dentry other than use rec->ur_name.
223 Severity : enhancement
225 Description: lfs setstripe enhancement
226 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
230 Description: testing performance impact of enabling checksumming
231 Details : enable checksum by default, allow --disable-checksum
232 configure option and "-o nochecksum" mount option
236 Description: lustre support for RHEL5
237 Details : Add support for RHEL5.
241 Description: Client eviction due to failover config
242 Details : after a connection loss, the lustre client should attempt to
243 reconnect to the last active server first before trying the
244 other potential connections.
248 Description: when mds and osts use different quota unit(32bit and 64bit),
249 quota will be released repeatly.
250 Details : void sending multiple quota reqs to mds, which will keep the
251 status between the reqs.
255 Description: osts not allocated evenly to files
256 Details : change the condition to increase offset_idx
259 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
261 Description: Data corruption for OSTs that are formatted larger than 2TB
263 Details : When generating the bio request for lustre file writes the
264 sector number would overflow a temporary variable before being
265 used for the IO. The data reads correctly from Lustre (which
266 will overflow in a similar manner) but other file data or
267 filesystem metadata may be corrupted in some cases.
271 Description: Tune the kernel for good SCSI performance.
272 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
273 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
278 Description: rewrite ext2-derived code in obdclass/uuid.c
279 Details : rewrite inherited code (uuid parsing code from ext2 utils)
280 from scratch preserving functionality.
284 Description: "lfs find -obd UUID" prints directories
285 Details : "lfs find -obd UUID" will return all directory names instead
286 of just file names. It is incorrect because the directories
287 do not reside on the OSTs.
291 Description: use i_size_read and i_size_write in 2.6 port
292 Details : replace inode->i_size access with i_size_read/write()
295 Frequency : when removing large files
297 Description: scheduling issue during removal of large Lustre files
298 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
299 It causes scheduling issues when removing large files (17TB in the
303 Frequency : only with liblustre clients on XT3
305 Description: evictions taking too long
306 Details : allow llrd to evict clients directly on OSTs
309 Frequency : only on ppc
311 Description: /proc/fs/lustre/devices broken on ppc
312 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
313 We should make sure the type of 'index' is loff_t and then cast
314 explicitly as needed below. Do not assign an explicitly cast
318 Frequency : only for rhel5
320 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
321 Details : Modify the target file & which_kernel.
323 Severity : enhancement
325 Description: omit set fsid for export NFS
326 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
327 as FS_REQUIRES_DEV for avoid problems with generate fsid.
331 Frequency : Always, for kernels after 2.6.16
332 Description: Fix warning idr_remove called for id=.. which is not
334 Details : Last kernels save old s_dev before kill super and not allow
335 to restore from callback - restore it before call kill_anon_super.
339 Description: buffer overruns could theoretically occur
340 Details : llapi_semantic_traverse() modifies the "path" argument by
341 appending values to the end of the origin string, and a
342 overrun may occur. Adding buffer overrun check in liblustreapi.
346 Description: Fix error on 'ls .' at the top of the Lustre mount.
347 Details : Don't revalidate dentry if it is a root dentry.
351 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
352 Details : Modify vm-tunables-rhel4.patch.
356 Description: Kernel config for 2.6.18-vanilla.
357 Details : Modify targets/2.6-vanilla.target.in.
358 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
359 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
360 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
361 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
365 Description: improve handling recoverable errors
366 Details : if request processig with error which can be recoverable on server
367 request should be resend, otherwise page released from cache and
372 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
373 Details : Modify target file & which_patch.
374 A flaw was found in the IA32 system call emulation provided
375 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
376 value could be stored in the %RAX register, which could trigger an
377 out-of-bounds system call table access. An untrusted local user
378 could exploit this flaw to run code in the kernel
379 (ie a root privilege escalation). (CVE-2007-4573).
383 Description: Remove 2.4 kernel checks lustre 1.8
384 Details : Remove 2.5.0 check from quota_check.c, quota_ctl.c,
385 quota_interface.c in quota subsystem.
386 Remove 2.4 kernel check from obd_mount.c, linux/linux-sysctl.c in
388 Remove lock_24kernel from fsfilt_ext3.c fsfilt_reiserfs.c
389 lvfs_linux.c upcall_cache.c in lvfs subsystem.
393 Description: O_DIRECT bypasses client statistics.
394 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
395 read_ahead_stats were updated. copied the stats section from
396 osc_send_oap_rpc() into async_internal().
400 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
404 Description: change order of libsysio includes
405 Details : '#include sysio.h' should always come before '#include xtio.h'
409 Description: handle "serious error: objid * already exists" more gracefully
410 Details : If LAST_ID value on disk is smaller than the objects existing in
411 the O/0/d* directories, it indicates disk corruption and causes an
412 LBUG(). If the object is 0-length, then we should use the existing
413 object. This will help to avoid a full fsck in most cases.
415 Severity : enhancement
417 Description: adapt the lustre_config script to support the upgrade case
418 Details : Add "-u" option for lustre_config script to support upgrading 1.4
419 server targets to 1.6 in parallel.
423 Description: To avoid grant space > avaible space when the disk is almost
424 full. Without this patch you might see the error "grant XXXX >
425 available" or some LBUG about grant, when the disk is almost
427 Details : In filter_check_grant, for non_grant cache write, we should
428 check the left space by if (*left > ungranted + bytes), instead
429 of (*left > ungranted), because only we are sure the left space
430 is enough for another "bytes", then the ungrant space should be
431 increase. In client, we should update cl_avail_grant only there
432 is OBD_MD_FLGRANT in the reply.
436 Description: Update RHEL 4 kernel to fix local root privilege escalation.
437 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
438 described in CVE-2007-4573. This problem could allow untrusted
439 local users to gain root access.
443 Description: Remove 2.4 kernel checks lustre 1.8
444 Details : Remove 2.5.0 check from dcache.c dir.c file.c llite_internal.h
445 llite_lib.c llite_mmap.c lproc_llite.c namei.c rw.c symlink.c
447 Remove 2.5.0 check from handler.c lproc_mds.c mds_fs.c
448 mds_internal.h mds_lib.c mds_open.c mds_reint.c mds_xattr.c
451 --------------------------------------------------------------------------------
453 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
455 * Support for kernels:
456 2.6.5-7.283 (SLES 9),
457 2.6.9-55.EL (RHEL 4),
458 2.6.16.46-0.14 (SLES 10),
459 2.6.18.8 vanilla (kernel.org)
460 * Client support for unpatched kernels:
461 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
462 2.6.16 - 2.6.22 vanilla (kernel.org)
463 * Due to recently discovered recovery problems, we do not recommend
464 using patchless RHEL 4 clients with this or any earlier release.
465 * Recommended e2fsprogs version: 1.39.cfs8
466 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
467 * Starting with this release, the ldiskfs backing filesystem required
468 by Lustre is now in its own package, lustre-ldiskfs. This package
469 should be installed. It is versioned separately from Lustre and
470 may be released separately in future.
474 Description: block reactivating mgc import until all deactivates complete
475 Details : Fix race when failing back MDT/MGS to itself (testing)
477 Severity : enhancement
479 Description: add optional extra BUILD_VERSION info
480 Details : add a new environment variable (namely LUSTRE_VERS) which allows
481 to override the lustre version.
484 Frequency : 2.6.18 servers only
486 Description: ll_kern_mount() doesn't release the module reference
487 Details : The ldiskfs module reference count never drops down to 0
488 because ll_kern_mount() doesn't release the module reference.
493 Description: server LBUG when using old ost_num_threads parameter
494 Details : Accept the old ost_num_threads parameter but warn that it
495 is deprecated, and fix an off-by-one error that caused an LBUG.
500 Description: Transient SCSI error results in persistent IO issue
501 Details : iobuf->dr_error is not reinitialized to 0 between two
505 Frequency : sometimes when underlying device returns I/O errors
507 Description: OSTs not going read-only during write failures
508 Details : OSTs are not remounted read-only when the journal commit threads
509 get I/O errors because fsfilt_ext3 calls journal_start/stop()
510 instead of the ext3 wrappers.
514 Description: poor connect scaling with increasing client count
515 Details : Don't run filter_grant_sanity_check for more than 100 exports
516 to improve scaling for large numbers of clients.
519 Frequency : SLES10 only
521 Description: sanity-quota.sh quotacheck failed: rc = -22
522 Details : Quotas cannot be enabled on SLES10.
525 Frequency : liblustre clients only
527 Description: getdirentries does not give error when run on compute nodes
528 Details : getdirentries does not fail when the size specified as an argument
529 is too small to contain at least one entry
531 Severity : enhancement
533 Description: Add LNET router traceability for debug purposes
534 Details : If a checksum failure occurs with a router as part of the
535 IO path, the NID of the last router that forwarded the bulk data
536 is printed so it can be identified.
541 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
542 Details : Due to a race condition, liblustre clients were occasionally
545 Severity : enhancement
547 Description: lfs setstripe use optional parameters instead of postional
550 Severity : enhancement
552 Description: Nanosecond timestamp support for ldiskfs
553 Details : The on-disk ldiskfs filesystem has added support for nanosecond
554 resolution timestamps. There is not yet support for this at
555 the Lustre filesystem level.
558 Frequency : during server recovery
560 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
561 Details : request with rq_no_resend flag not awake l_wait_event if they get a
565 Frequency : nfs export on patchless client
567 Description: connectathon hang when test nfs export over patchless client
568 Details : Disconnected dentry cannot be found with lookup, so we do not need
569 to unhash it or make it invalid
572 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
573 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
574 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
579 Description: open req refcounting wrong on reconnect
580 Details : If reconnect happened between getting open reply from server and
581 call to mdc_set_replay_data in ll_file_open, we will schedule
582 replay for unreferenced request that we are about to free.
583 Subsequent close will crash in variety of ways.
584 Check that request is still eligible for replay in
585 mdc_set_replay_data().
590 Description: disable writes to filesystem when reading health_check file
591 Details : the default for reading the health_check proc file has changed
592 to NOT do a journal transaction and write to disk, because this
593 can cause reads of the /proc file to hang and block HA state
594 checking on a healthy but otherwise heavily loaded system. It
595 is possible to return to the previous behaviour during configure
596 with --enable-health-write.
598 Severity : enhancement
600 Description: 64-bit inode version
601 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
602 to the inode. This will be required for version-based recovery.
607 Description: MDS fails to start if a duplicate client export is detected
608 Details : in some rare cases it was possible for a client to connect to
609 an MDS multiple times. Upon recovery the MDS would detect this
610 and fail during startup. Handle this more gracefully.
612 Severity : enhancement
614 Description: Add -o localflock option to simulate old noflock
616 Details : This will achieve local-only flock/fcntl locks
622 Description: log_commit_thread vs filter_destroy race leads to crash
623 Details : Take import reference before releasing llog record semaphore
628 Description: Wrong request locking in request set processing
629 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
630 imp_delayed_list, in this place should be used imp_lock.
633 Frequency : when reconnection
635 Description: Grant Leak when osc reconnect to OST
636 Details : When osc reconnect ost, OST(filter) should check whether it
637 should grant more space to client by comparing fed_grant and
638 cl_avail_grant, and return the granted space to client instead
639 of "new granted" space, because client will call osc_init_grant
640 to update the client grant space info.
643 Frequency : when client reconnect to OST
645 Description: Grant Leak when osc do resend and replay bulk write
646 Details : When osc reconnect to OST, OST(filter)should clear grant info of
647 bulk write request, because the grant info while be sync between
648 OSC and OST when reconnect, and we should ignore the grant info
649 these of resend/replay write req.
654 Description: Grant space more than avaiable left space sometimes.
655 Details : When then OST is about to be full, if two bulk writing from
656 different clients came to OST. Accord the avaliable space of the
657 OST, the first req should be permitted, and the second one
658 should be denied by ENOSPC. But if the seconde arrived before
659 the first one is commited. The OST might wrongly permit second
660 writing, which will cause grant space > avaiable space.
663 Frequency : when client is evicted
665 Description: Grant might be wrongly erased when osc is evicted by OST
666 Details : when the import is evicted by server, it will fork another
667 thread ptlrpc_invalidate_import_thread to invalidate the
668 import, where the grant will be set to 0. While the original
669 thread will update the grant it got when connecting. So if
670 the former happened latter, the grant will be wrongly errased
671 because of this race.
676 Description: Checking Stale with correct fid
677 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
678 to check whether it is stale, because sometimes, we want the
679 enqueue happened anyway, and op_data.fid2 will not be initialized.
681 Severity : enhancement
683 Description: update patchless client
684 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
687 Frequency : only with 2.4 kernel
689 Description: random memory corruption
690 Details : size of struct ll_inode_info is to big for union inode.u and this
691 can be cause of random memory corruption.
696 Description: Memory leak in recovery
697 Details : Lov_mds_md was not free in an error handler in mds_create_object.
698 It should also check obd_fail before fsfilt_start, otherwise if
699 fsfilt_start return -EROFS,(failover mds during mds recovery).
700 then the req will return with repmsg->transno = 0 and rc = EROFS.
701 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
702 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
703 be freed no matter whether fsfilt_commit success or not.
706 Frequency : only with huge count clients
708 Description: Prevents from taking the superblock lock in llap_from_page for
710 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
711 for llap_from_page call in ll_removepage prevents from taking the
712 superblock lock for a soon died page.
717 Description: Not check open intent error before release open handle
718 Details : in some rare cases, the open intent error is not checked before
719 release open handle, which may cause
720 ASSERTION(open_req->rq_transno != 0), because it tries to release
721 the failed open handle.
726 Description: Set cat log bitmap only after create log success.
727 Details : in some rare cases, the cat log bitmap is set too early. and it
728 should be set only after create log success.
732 Description: Accessing a block bevice can re-enable I/O when Lustre is
733 tearing down a device.
734 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
739 Description: service threads may hog cpus when there are a lot of requests
741 Details : Insert cond_resched to give other threads a chance to use some of
747 Description: the cat log was not initialized in recovery
748 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
749 unlink log on mds will not be initialized until mds post
750 recovery. And also in mds post recovery, the unlink log will
751 initialization will be done asynchronausly, so there will be race
752 between add unlink log and unlink log initialization.
756 Description: brw_stats were being printed incorrectly
757 Details : brw_stats were being printed as log2 but all of them were not
758 recorded as log2. Also remove some code duplication arising from
759 filter_tally_{read,write}.
763 Frequency : rare, only in recovery.
764 Description: ASSERTION(req->rq_type != LI_POISON) failed
765 Details : imp_lock should be held while iterating over imp_sending_list for
766 prevent destroy request after get timeout in ptlrpc_queue_wait.
770 Description: replay-single.sh test 52 fails
771 Details : A lock's skiplist need to be cleanup when it being unlinked
772 from its resource list.
776 Description: Short directio read returns full requested size rather than
778 Details : Direct I/O operations should return actual amount of bytes
779 transferred rather than requested size.
781 Severity : enhancement
783 Description: metadata RPC reduction (e.g. for rm performance)
784 Details : decrease the amount of synchronous RPC between clients and servers
785 by canceling conflicing lock before the operation on the client side
786 and packing thier handles into the main operation RPC to server.
788 Severity : enhancement
790 Description: Async OSC create to avoid the blocking unnecessarily.
791 Details : If a OST has no remain object, system will block on the creating
792 when need to create a new object on this OST. Now, ways use
793 pre-created objects when available, instead of blocking on an
794 empty osc while others are not empty. If we must block, we block
795 for the shortest possible period of time.
799 Description: improve handling recoverable errors
800 Details : if request processig with error which can be recoverable on server
801 request should be resend, otherwise page released from cache and
804 Severity : enhancement
806 Description: refine locking for avoid write wrong info into lov_objid file
807 Details : fix possible races with add new target and write/update data in
810 --------------------------------------------------------------------------------
812 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
817 Frequency : on some architectures
819 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
820 Details : Uninitialized flags sometimes cause configuration commands to
824 Frequency : patchless clients only
826 Description: missing __iget() symbol export
827 Details : The __iget() symbol export is missing. To avoid the need for
828 this on patchless clients the deathrow inode reaper is turned
829 off, and we depend on the VM to clean up old inodes. This
830 dependency was during via the fix for bug 12181.
832 --------------------------------------------------------------------------------
834 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
836 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
837 INTEROPERATE with older versions automatically. In many cases a
838 special upgrade step is needed. Please read the
839 user documentation before upgrading any part of a live system.
840 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
841 version will not interoperate with 1.6 betas before beta5 (1.5.95).
842 * WARNING: Lustre configuration and startup changes are required with
843 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
849 Severity : enhancement
851 Description: MountConf
852 Details : Lustre configuration is now managed via mkfs and mount
853 commands instead of lmc and lconf. New obd types (MGS, MGC)
854 are added for dynamic configuration management. See
855 https://mail.clusterfs.com/wikis/lustre/MountConf for
858 Severity : enhancement
860 Description: dynamic OST addition
861 Details : OSTs can now be added to a live filesystem
863 Severity : enhancement
865 Description: startup order invariance
866 Details : MDTs and OSTs can be started in any order. Clients only
867 require the MDT to complete startup.
869 Severity : enhancement
871 Description: parallel, asynchronous orphan cleanup
872 Details : orphan cleanup is now performed in separate threads for each
873 OST, allowing parallel non-blocking operation.
875 Severity : enhancement
877 Description: optimized stripe assignment
878 Details : stripe assignments are now made based on ost space available,
879 ost previous usage, and OSS previous usage, in order to try
880 to optimize storage space and networking resources.
882 Severity : enhancement
884 Description: Permanently set tunables
885 Details : All writable /proc/fs/lustre tunables can now be permanently
886 set on a per-server basis, at mkfs time or on a live
889 Severity : enhancement
891 Description: Lustre message v2
892 Details : Add lustre message format v2.
894 Severity : enhancement
896 Description: client OST exclusion list
897 Details : Clients can be started with a list of OSTs that should be
898 declared "inactive" for known non-responsive OSTs.
902 Description: SPEC SFS validation failure on NFS v2 over lustre.
903 Details : Changes the blocksize for regular files to be 2x RPC size,
904 and not depend on stripe size.
906 Severity : enhancement
908 Description: Multiple MD RPCs in flight.
909 Details : Further unserialise some read-only MDS RPCs - learn about intents.
910 To avoid overly-overloading MDS, introduce a limit on number of
911 MDS RPCs in flight for a single client and add /proc controls
912 to adjust this limit.
914 Severity : enhancement
916 Description: client read/write statistics
917 Details : Add client read/write call usage stats for performance
918 analysis of user processes.
919 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
920 file access. extents_stats shows chunk size distribution.
921 extents_stats_per_process show chunk size distribution per
924 Severity : enhancement
926 Description: mds statistics
927 Details : Add detailed mds operations statistics in
928 /proc/fs/lustre/mds/*/stats.
932 Description: Failure of copying files with lustre special EAs.
933 Details : Client side always return success for setxattr call for lustre
934 special xattr (currently only "trusted.lov").
939 Description: Refcount LNET uuids
940 Details : The global LNET uuid list grew linearly with every startup;
941 refcount repeated list entries instead of always adding to
944 Severity : enhancement
946 Description: Dynamic service threads
947 Details : Within a small range, start extra service threads
948 automatically when the request queue builds up.
951 Frequency : mixed-endian client/server environments
953 Description: mixed-endian crashes
954 Details : The new msg_v2 system had some failures in mixed-endian
957 Severity : enhancement
959 Description: Easy OST removal
960 Details : OSTs can be permanently deactivated with e.g. 'lctl
961 conf_param lustre-OST0001.osc.active=0'
963 Severity : enhancement
965 Description: MGS proc entries
966 Details : Added basic proc entries for the MGS showing what filesystems
969 Severity : enhancement
971 Description: provide MGS failover
972 Details : Added config lock reacquisition after MGS server failover.
974 Severity : enhancement
976 Description: add Linux 2.4 support
977 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
981 Description: a large application tries to do I/O to the same resource and dies
983 Details : Check the req->rq_arrival time after the call to
984 ost_brw_lock_get(), but before we do anything about
985 processing it & sending the BULK transfer request. This
986 should help move old stale pending locks off the queue as
987 quickly as obd_timeout.
990 Frequency : when an incorrect nid is specified during startup
992 Description: ptlrpc connect to non-existant node causes kernel crash
993 Details : LNET can't be re-entered from an event callback, which
994 happened when we expire a message after the export has been
995 cleaned up. Instead, hand the zombie cleanup off to another
998 Severity : enhancement
1000 Description: plain/inodebits lock performance improvement
1001 Details : Grouping plain/inodebits in granted list by their request modes
1002 and bits policy, thus improving the performance of search through
1006 Frequency : only if OST filesystem is corrupted
1008 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1009 Details : for a short time RPCs with bulk IO are in the replay list,
1010 but replay of bulk IOs is unimplemented. If the OST filesystem
1011 is corrupted due to disk cache incoherency and then replay is
1012 started it is possible to trip an assertion. Avoid putting
1013 committed RPCs into the replay list at all to avoid this issue.
1016 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1018 Description: System hang on startup
1019 Details : This bug allowed the liblustre (e.g. catamount) client to
1020 return to the app before handling all startup RPCs. This
1021 could leave the node unresponsive to lustre network traffic
1022 and manifested as a server ptllnd timeout.
1024 Severity : enhancement
1026 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1027 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1028 boolean to control whether to print peer debug info when a
1029 client's RPC times out.
1032 Frequency : only for kernels with patches from Lustre below 1.4.3
1034 Description: Remove old rdonly API
1035 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1038 Frequency : only for devices with external journals
1040 Description: Set external device read-only also
1041 Details : During a commanded failover stop, we set the disk device
1042 read-only while the server shuts down. We now also set any
1043 external journal device read-only at the same time.
1046 Frequency : when upgrading from 1.4 while trying to change parameters
1048 Description: The wrong (new) MDC name was used when setting parameters for
1049 upgraded MDT's. Also allows changing of OSC (and MDC)
1050 parameters if --writeconf is specified at tunefs upgrade time.
1053 Frequency : when setting specific ost indicies
1055 Description: QOS code breaks on skipped indicies
1056 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1057 created with --index need not be sequential.
1059 Severity : enhancement
1061 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1062 Details : The uninit_groups feature works in conjunction with the kernel
1063 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1064 up the pass1 processing of e2fsck. This is a read-only feature
1065 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1066 mount filesystems that have had this feature enabled.
1068 Severity : enhancement
1070 Description: Improve multi-block allocation algorithm to avoid fragmentation
1071 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1072 allocation locality and avoid filesystem fragmentation.
1074 ------------------------------------------------------------------------------
1076 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1078 * Support for kernels:
1079 2.4.21-47.0.1.EL (RHEL 3)
1080 2.6.5-7.283 (SLES 9)
1081 2.6.9-42.0.10.EL (RHEL 4)
1082 2.6.12.6 vanilla (kernel.org)
1083 2.6.16.27-0.9 (SLES 10)
1084 * Recommended e2fsprogs version: 1.39.cfs5
1086 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1090 Frequency : occasional, depends on client load and configuration
1091 Bugzilla : 12181, 12203
1092 Description: data loss for recently-modified files
1094 Details : In some cases it is possible that recently written or created
1095 files may not be written to disk in a timely manner (this should
1096 normally be within 30s unless client IO load is very high).
1097 The problem appears as zero-length files or files that are a
1098 multiple of 1MB in size after a client crash or client eviction
1099 that are missing data at the end of the file.
1101 This problem is more likely to be hit on clients where files are
1102 repeatedly created and unlinked in the same directory, clients
1103 have a large amount of RAM, have many CPUs, the filesystem has
1104 many OSTs, the clients are rebooted frequently, and/or the files
1105 are not accessed by other nodes after being written.
1107 The presence of the problem can be detected by looking at
1108 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1109 smaller than the second (nr_unused) then dirty files will not
1110 be flushed automatically to disk. "sync; sleep 10" should be
1111 run several times on the node before unmounting it to update
1112 Lustre (this is also safe to run on nodes without this problem).
1114 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1115 that can cause this same problem, so customers using that kernel
1116 also need to update the kernel in addition to Lustre. In order
1117 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1119 It is normal that files written just before a client crash (less
1120 than 30s) may not yet have been flushed to disk, even for local
1124 Frequency : frequent on thin XT3 nodes
1126 Description: UUID collision on thin XT3 Linux nodes
1127 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1128 randomly, since we relied on an insufficiently-seeded PRNG.
1133 Description: OSS hangs after "All ost request buffers busy"
1134 Details : A deadlock between quota and journal operations caused OSS
1135 hangs after printing "All ost request buffers busy."
1138 Frequency : always on liblustre builds
1140 Description: Cleanup compiler warnings on liblustre
1143 Frequency : always on liblustre builds on XT3
1145 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1149 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1150 > 32000 subdirectories
1151 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1152 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1153 subdirectory count crosses 32000. This will aid e2fsck to
1154 correctly handle more than 32000 subdirectories.
1157 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1159 Description: System hang on startup
1160 Details : This bug allowed the liblustre (e.g. catamount) client to
1161 return to the app before handling all startup RPCs. This
1162 could leave the node unresponsive to lustre network traffic
1163 and manifested as a server ptllnd timeout.
1165 Severity : enhancement
1167 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1168 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1169 boolean to control whether to print peer debug info when a
1170 client's RPC times out.
1175 Description: make O_SYNC working on 2.6 kernels
1176 Details : 2.6 kernels use different method for mark pages for write,
1177 so need add a code to lustre for O_SYNC work.
1182 Description: Failure to close file and release space on NFS
1183 Details : Put inode details into lock acquired in ll_intent_file_open.
1184 Use mdc_intent_lock in ll_intent_open to properly
1185 detect all kind of errors unhandled by mdc_enqueue.
1190 Description: proc file read during shutdown sometimes raced obd removal,
1192 Details : Add lock to prevent obd access after proc file removal.
1195 Frequency : Only for files larger than 4GB on 32-bit clients.
1197 Description: improperly doing page alignment of locks
1198 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1199 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1204 Description: RPCs being resent when they shouldn't be
1205 Details : Some RPCs that should not be resent are being resent. This
1206 can cause inconsistencies in the RPC state machine. Do not
1207 resend such requests.
1210 Frequency : rare, only with NFS export
1212 Description: Crash on NFS re-export node
1213 Details : under very unusual load conditions an assertion is hit in
1214 ll_intent_file_open()
1217 Frequency : only if OST filesystem is corrupted
1219 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1220 Details : for a short time RPCs with bulk IO are in the replay list,
1221 but replay of bulk IOs is unimplemented. If the OST filesystem
1222 is corrupted due to disk cache incoherency and then replay is
1223 started it is possible to trip an assertion. Avoid putting
1224 committed RPCs into the replay list at all to avoid this issue.
1229 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1230 Details : Large single O_DIRECT read and write calls can fail to allocate
1231 a sufficiently large buffer to process the request. In case of
1232 allocation failure the allocation is retried with a smaller
1233 buffer and broken into smaller requests.
1235 Severity : enhancement
1237 Description: Add -o localflock option to simulate old noflock behaviour.
1238 Details : This will achieve local-only flock/fcntl locks coherentness.
1243 Description: versioning check is incomplete
1244 Details : Checking the version difference of client vs. server, report
1245 error if the gap is too big.
1250 Description: add support PG_writeback bit
1251 Details : add support for PG_writeback bit for Lustre, for more carefull
1252 work with page cache in 2.6 kernel. This also fix some deadlocks
1253 and remove hack for work O_SYNC with 2.6 kernel.
1255 Severity : enhancement
1257 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1258 Details : The uninit_groups feature works in conjunction with the kernel
1259 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1260 up the pass1 processing of e2fsck. This is a read-only feature
1261 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1262 mount filesystems that have had this feature enabled.
1264 Severity : enhancement
1266 Description: Improve multi-block allocation algorithm to avoid fragmentation
1267 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1268 allocation locality and avoid filesystem fragmentation.
1270 ------------------------------------------------------------------------------
1272 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1274 * Support for kernels:
1275 2.6.9-42.0.3.EL (RHEL 4)
1276 2.6.5-7.276 (SLES 9)
1277 2.4.21-47.0.1.EL (RHEL 3)
1278 2.6.12.6 vanilla (kernel.org)
1279 2.6.16.21-0.8 (SLES10)
1280 * Recommended e2fsprogs version: 1.39.cfs2-0
1282 * The backwards-compatible /proc/sys/portals symlink has been removed
1283 in this release. Before upgrading, please ensure that you change
1284 any configuration scripts or /etc/sysctl.conf files that access
1285 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1286 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1287 in advance of the upgrade on any system running Lustre 1.4.6 or
1288 newer, since /proc/sys/lnet was added in that version.
1289 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1293 Frequency : only when quota is used
1295 Description: avoid scanning export list for quota master
1296 Details : Change the algorithms to avoid scanning export list in order
1297 to improve the efficiency.
1300 Frequency : MDS failover only, very rarely
1302 Description: "went back in time" messages on mds failover
1303 Details : The greatest transno may be lost when the current operation
1304 finishes with an error (transno==0) and the client's last_rcvd
1305 record is over-written. Save the greatest transno in the
1306 mds_last_transno for this case.
1309 Frequency : always for specific kernels and striping counts
1311 Description: client may get "Matching packet too big" without ACL support
1312 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1313 when trying to access files in certain configurations. The
1314 clients should in fact be denied when mounting because they do
1315 not understand ACLs.
1318 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1320 Description: many clients connecting with IO in progress causes connect timeouts
1321 Details : Avoid synchronous journal commits to avoid delays caused by many
1322 clients connecting/disconnecting when bulk IO is in progress.
1323 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1324 OST_IO_PORTAL to avoid delays behind potentially many pending
1328 Frequency : occasionally with multiple writers to a single file
1330 Description: shared writes to file may result in wrong size reported by stat()
1331 Details : Allow growing of kms when extent lock is cancelled
1334 Frequency : always with random mmap IO to multi-striped file
1336 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1337 Details : Only if the hole is at the end of OST object so that kms is too
1338 small. Fix is to increase kms accordingly in ll_nopage.
1341 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1343 Description: writes to a missing object would leak memory on the OST
1344 Details : If there is an inconsistency between the MDS and OST filesystems,
1345 such that the MDS references an object that doesn't exist, writes
1346 to that object will leak memory due to incorrect cleanup in the
1347 error handling path, eventually running out of memory on the OST.
1352 Description: Creating too long symlink causes lustre errors
1353 Details : Check symlink and name lengths before sending requests to MDS.
1356 Frequency : only if flock is enabled (not on by default)
1358 Description: posix locks not released on fd closure on 2.6.9+
1359 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1360 kernels, this caused such locks not to be released on fd close and
1361 then assertions on fs unmount about still used locks.
1364 Frequency : MDS failover only, very rarely
1366 Description: clients may get ASSERTION(granted_lock != NULL)
1367 Details : When request was taking a long time, and a client was resending
1368 a getattr by name lock request. The were multiple lock requests
1369 with the same client lock handle and
1370 mds_getattr_name->fixup_handle_for_resent_request found one of the
1371 lock handles but later failed with ASSERTION(granted_lock != NULL).
1376 Description: handle->h_buffer_credits > 0, assertion failure
1377 Details : h_buffer_credits is zero after truncate, causing assertion
1378 failure. This patch extends the transaction or creates a new
1382 Frequency : NFS re-export or patchless client
1383 Bugzilla : 11179, 10796
1384 Description: Crash on NFS re-export node (__d_move)
1385 Details : We do not want to hash the dentry if we don't have a lock.
1386 But if this dentry is later used in d_move, we'd hit uninitialised
1387 list head d_hash, so we just do this to init d_hash field but
1388 leave dentry unhashed.
1391 Frequency : NFS re-export or patchless client
1393 Description: NFS exports has problem with symbolic link
1394 Details : lustre client didn't properly install dentry when re-exported
1395 to NFS or running patchless client.
1398 Frequency : NFS re-export or patchless client
1400 Description: Various nfs/patchless fixes.
1401 Details : fixes reuse disconected alias for lookup process - this fixes
1402 warning "find_exported_dentry: npd != pd",
1403 fix permission error with open files at nfs.
1404 fix apply umask when do revalidate.
1407 Frequency : occasional
1409 Description: Crash on NFS re-export node
1410 Details : calling clear_page() on the wrong pointer triggered oops in
1411 generic_mapping_read().
1414 Frequency : rarely, using O_DIRECT IO
1416 Description: unaligned directio crashes client with LASSERT
1417 Details : check for unaligned buffers before trying any requests.
1420 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1422 Description: stale data returned from RAID cache
1423 Details : If only a small amount of IO is done to the RAID device before
1424 reading it again it is possible to get stale data from the RAID
1425 cache instead of reading it from disk.
1428 Frequency : always for sles10 kernel
1430 Description: sles10 support
1431 Details : ll_follow_link: compile fixes and using of nd_set_link
1432 under newer kernels.
1435 Frequency : depends on arch, kernel and compiler version, always on sles10
1438 Description: recursive or deep enough symlinks cause stack overflow
1439 Details : getting rid of large stack-allocated variable in
1443 Frequency : depends on hardware
1445 Description: lustre write performance loss in the SLES10 kernel
1446 Details : the performance loss is caused by using of write barriers in the
1447 ext3 code. The SLES10 kernel turns barrier support on by
1448 default. The fix is to undo that change for ldiskfs.
1450 ------------------------------------------------------------------------------
1452 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1454 * Support for kernels:
1455 2.6.9-42.0.3EL (RHEL 4)
1456 2.6.5-7.276 (SLES 9)
1457 2.4.21-47.0.1.EL (RHEL 3)
1458 2.6.12.6 vanilla (kernel.org)
1462 Frequency : quota enabled and large files being deleted
1464 Description: releasing more than 4GB of quota at once hangs OST
1465 Details : If a user deletes more than 4GB of files on a single OST it
1466 will cause the OST to spin in an infinite loop. Release
1467 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1472 Description: statfs data retrieved from /proc may be stale or zero
1473 Details : When reading per-device statfs data from /proc, in the
1474 {kbytes,files}_{total,free,avail} files, it may appear
1475 as zero or be out of date.
1478 Frequency : systems with MD RAID1 external journal devices
1480 Description: lconf's call to blkid is confused by RAID1 journal devices
1481 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1482 of returning all block devices that match the journal UUID.
1485 Frequency : always, for aggregate stripe size over 4GB
1487 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1488 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1489 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1492 Frequency : always if "lfs find" used on a local file/directory
1494 Description: "lfs find" segfaults if used on a local file/directory
1495 Details : The case where a directory component was not specified wasn't
1496 handled correctly. Handle this properly.
1499 Frequency : always on ppc64
1501 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1502 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1503 perform endianness conversion.
1506 Frequency : rarely (truncate to non-zero file size after write under load)
1507 Bugzilla : 10730, 10687
1508 Description: Files padded with zeros to next 4K multiple
1509 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1510 it is possible that files that are truncated to a non-zero size
1511 immediately after being written are filled with zero bytes beyond
1512 the truncated size. No file data is lost.
1514 Severity : enhancement
1516 Description: Allow recovery/failover for liblustre clients.
1517 Details : liblustre clients were unaware of failover configurations until
1520 Severity : enhancement
1522 Description: user file locks should fail when not mounting with flock option
1523 Details : Set up an error-returning stub in ll_file_operations.lock field
1524 to prevent incorrect behaviour when client is mounted without
1525 flock option. Also, set up properly f_op->flock field for
1529 Frequency : always on ia64
1531 Description: "lfs df" loops on printing out MDS statfs information
1532 Details : The obd_ioctl_data was not initialized and in some systems
1533 this caused a failure during the ioctl that did not return
1534 an error. Initialize the struct and return an error on failure.
1537 Frequency : SLES 9 only
1539 Description: Error of copying files with lustre special EAs as root
1540 Details : Client side always return success for setxattr call for lustre
1541 special xattr (currently only "trusted.lov").
1544 Frequency : rarely on clusters with both ia64+i386 clients
1546 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1547 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1548 clients are doing concurrent writes to the same file it is
1549 possible that the ia64 clients may LASSERT because the OST
1550 extent locks are not PAGE_SIZE aligned. Ensure that grown
1551 locks are always aligned on the request boundary.
1554 Frequency : specific use, occasional
1556 Description: Overwriting in use executable truncates on-disk binary image
1557 Details : If one node attempts to overwrite an executable in use by
1558 another node, we now correctly return ETXTBSY instead of
1559 truncating the file.
1561 Severity : enhancement
1563 Description: Async OSC create to avoid the blocking unnecessarily.
1564 Details : If a OST has no remain object, system will block on the creating
1565 when need to create a new object on this OST. Now, ways use
1566 pre-created objects when available, instead of blocking on an
1567 empty osc while others are not empty. If we must block, we block
1568 for the shortest possible period of time.
1573 Description: chmod on Lustre root is propagated to other clients
1574 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1575 invalid by the follow_mount time.
1578 Frequency : liblustre clients only
1580 Description: Race in 'instant cancel' lock handling could lead to such locks
1581 never to be granted in case of SMP MDS
1582 Details : Do not destroy not yet granted but cbpending locks in
1586 Frequency : replay/resend of open
1588 Description: non null lock assetion failure in mds_intent_policy
1589 Details : Trying to replay/resend lockless open requests resulted in
1590 mds_open() returning 0 with no lock. Now it sets a flag if
1591 a lock is going to be returned.
1593 Severity : enhancement
1595 Description: Checksum enhancements
1596 Details : New checksum enhancements allow for resending RPCs that failed
1599 Severity : enhancement
1601 Description: Tunables on number of dirty pages in cacche
1602 Details : Allow to set limit on number of dirty pages cached.
1607 Description: client crash on unmount - lock still has references
1608 Details : In some error handling cases it was possible to leak a lock
1609 reference on a client while accessing a file. This was not
1610 harmful to the client during operation, but would cause the
1611 client to crash when the filesystem is unmounted.
1614 Frequency : specific case, rare
1616 Description: ETXTBSY on mds though file not in use by client
1617 Details : ETXTBSY is no longer incorrectly returned when attempting to
1618 chmod or chown a directory that the user previously tried to
1619 execute or a currently-executing binary.
1622 Frequency : extremely rare except on liblustre-based clients
1624 Description: Lustre space not freed when files are deleted
1625 Details : Clean up open-unlinked files after client eviction. Previously
1626 the unlink was skipped and the files remained as orphans.
1631 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1632 Details : In some cases it was possible to send a blocking callback to
1633 a client doing a glimpse, even though that client didn't get
1634 a lock granted. When the glimpse lock is cancelled on the OST
1635 the freed lock is left on the waiting list and corrupted the list.
1638 Frequency : all core dumps
1640 Description: Broke core dumps to lustre
1641 Details : Negative dentry may be unhashed if parent does not have UPDATE
1642 lock, but some callers, e.g. do_coredump, expect dentry to be
1643 hashed after successful create, hash it in ll_create_it.
1645 ------------------------------------------------------------------------------
1647 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1649 * Support for kernels:
1650 2.6.9-42.0.2.EL (RHEL 4)
1651 2.6.5-7.276 (SLES 9)
1652 2.4.21-40.EL (RHEL 3)
1653 2.6.12.6 vanilla (kernel.org)
1657 Frequency : always on RHEL 3
1659 Description: Number of open files grows over time
1660 Details : The number of open files grows over time, whether or not
1661 Lustre is started. This was due to a filp leak introduced
1662 by one of our kernel patches.
1664 ------------------------------------------------------------------------------
1666 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1668 * Support for kernels:
1669 2.6.9-42.EL (RHEL 4)
1670 2.6.5-7.276 (SLES 9)
1671 2.4.21-40.EL (RHEL 3)
1672 2.6.12.6 vanilla (kernel.org)
1677 Bugzilla : 5719, 9635, 9792, 9684
1678 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1679 Details : If a server is under heavy load and cannot reply to new
1680 connection requests before the client resends the (re)connect,
1681 the connection handling code can behave badly if two service
1682 threads are concurrently handing separate (re)connections from
1683 the same client. Add better locking to the connection handling
1684 code, and ensure that only a single connection will be processed
1685 for a given client UUID, even if the lock is dropped.
1687 Severity : enhancement
1689 Description: add TCP zero-copy support to kernel
1690 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1691 sends if the hardware supports scatter-gather and checksumming.
1692 This allows socklnd to do client-write and server-read more
1693 efficiently and reduce CPU utilization from skbuf copying.
1696 Frequency : only if NFS exporting from client
1698 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1699 Details : If setting attributes on a file created under NFS that had
1700 never been opened it would be possible to oops the client
1701 if the file had no objects.
1704 Frequency : always for liblustre
1706 Description: liblustre client does MDS+OSTs setattr RPC for each write
1707 Details : When doing a write from a liblustre client, the client
1708 incorrectly issued an RPC to the MDS and each OST the file was
1709 striped over in order to update the timestamps. When writing
1710 with small chunks and many clients this could overwhelm the MDS
1711 with RPCs. In all cases it would slow down the write because
1712 these RPCs are unnecessary.
1714 Severity : enhancement
1716 Description: allow number of MDS service threads to be changed at module load
1717 Details : It is now possible to change the number of MDS service threads
1718 running. Adding "options mds mds_num_threads={N}" to the MDS's
1719 /etc/modprobe.conf will set the number of threads for the next
1720 time Lustre is restarted (assuming the "mds" module is also
1721 reloaded at that time). The default number of threads will
1722 stay the same, 32 for most systems.
1727 Description: OST crash if filesystem is unformatted or corrupt
1728 Details : If an OST is started on a device that has never been formatted
1729 or if the filesystem is corrupt and cannot even mount then the
1730 error handling cleanup routines would dereference a NULL pointer.
1735 Description: NULL pointer deref in llap_from_page.
1736 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1737 incorrect) mapping if the page was truncated/reclaimed while it was
1738 searched for. Check for this condition and skip such pages when
1739 doing readahead. Introduce extra check to llap_from_page() to
1740 verify page->mapping->host is non-NULL (so page is not anonymous).
1743 Frequency : Sometimes when using sys_sendfile
1745 Description: "page not covered by a lock" warnings from ll_readpage
1746 Details : sendfile called ll_readpage without right page locks present.
1747 Now we introduced ll_file_sendfile that does necessary locking
1748 around call to generic_file_sendfile() much like we do in
1752 Frequency : with certain MDS communication failures at client mount time
1754 Description: NULL pointer deref after failed client mount
1755 Details : a client connection request may delayed by the network layer
1756 and not be sent until after the PTLRPC layer has timed out the
1757 request. If the client fails the mount immediately it will try
1758 to clean up before the network times out the request. Add a
1759 reference from the request import to the obd device and delay
1760 the cleanup until the network drops the request.
1763 Frequency : occasionally during client (re)connect
1765 Description: assertion failure during client (re)connect
1766 Details : processing a client connection request may be delayed by the
1767 client or server longer than the client connect timeout. This
1768 causes the client to resend the connection request. If the
1769 original connection request is replied in this interval, the
1770 client may trip an assertion failure in ptlrpc_connect_interpret()
1771 which thought it would be the only running connect process.
1774 Frequency : only with obd_echo servers and clients that are rebooted
1776 Description: kernel BUG accessing uninitialized data structure
1777 Details : When running an obd_echo server it did not start the ping_evictor
1778 thread, and when a client was evicted an uninitialized data
1779 structure was accessed. Start the ping_evictor in the RPC
1780 service startup instead of the OBD startup.
1782 Severity : enhancement
1783 Bugzilla : 10193 (patchless)
1784 Description: Remove dependency on various unexported kernel interfaces.
1785 Details : No longer need reparent_to_init, exit_mm, exit_files,
1786 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1789 Frequency : rare (only users of deprecated and unsupported LDAP config)
1791 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1792 Details : LDAP apparently contains 'lustreName' attributes instead of
1793 'name'. A simple remapping of the name is sufficient.
1796 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1798 Description: waiting_locks_callback trips kernel BUG if client is evicted
1799 Details : Running with the dump_on_timeout debug flag turned on makes
1800 it possible that the waiting_locks_callback() can try to dump
1801 the Lustre kernel debug logs from an interrupt handler. Defer
1802 this log dumping to the expired_lock_main() thread.
1804 Severity : enhancement
1806 Description: Support NFS exporting on 2.6 kernels.
1807 Details : Implement non-rawops metadata methods for NFS server to use without
1808 changing NFS server code.
1811 Frequency : very rare (synthetic metadata workload only)
1813 Description: two racing renames might cause an MDS thread to deadlock
1814 Details : Running the "racer" program may cause one MDS thread to rename
1815 a file from being the source of a rename to being the target of
1816 a rename at exactly the same time that another thread is doing
1817 so, and the second thread has already enqueued these locks after
1818 doing a lookup of the target and is trying to relock them in
1819 order. Ensure that we don't try to re-lock the same resource.
1822 Frequency : only very large systems with liblustre clients
1824 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1825 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1826 all OSTs in parallel. This allows the eviction of stale liblustre
1827 clients to proceed much faster than if they were done in series,
1828 and also offers similar improvements for other set_info RPCs.
1833 Description: excessive CPU usage during initial read phase on client
1834 Details : During the initial read phase on a client, it would agressively
1835 retry readahead on the file, consuming too much CPU and impacting
1836 performance (since 1.4.5.8). Improve the readahead algorithm
1837 to avoid this, and also improve some other common cases (read
1838 of small files in particular, where "small" is files smaller than
1839 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1844 Description: MDS crash when receiving packet with unknown intent.
1845 Details : Do not LBUG in unknown intent case, just return -EFAULT
1847 Severity : enhancement
1848 Bugzilla : 9293, 9385
1849 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1850 Details : Do not serialize getattr (non-intent version) and statfs.
1853 Frequency : occasional, when OST network is overloaded/intermittent
1855 Description: client evicted by OST after bulk IO timeout
1856 Details : If a client sends a bulk IO request (read or write) the OST
1857 may evict the client if it is unresposive to its data GET/PUT
1858 request. This is incorrect if the network is overloaded (takes
1859 too long to transfer the RPC data) or dropped the OST GET/PUT
1860 request. There is no need to evict the client at all, since
1861 the pinger and/or lock callbacks will handle this, and the
1862 client can restart the bulk request.
1865 Frequency : Always when mmapping file with no objects
1867 Description: client crashes when mmapping file with no objects
1868 Details : Check that we actually have objects in a file before doing any
1869 operations on objects in ll_vm_open, ll_vm_close and
1875 Description: Request leak when working with deleted CWD
1876 Details : Introduce advanced request refcount tracking for requests
1877 referenced from lustre intent.
1879 Severity : Enhancement
1881 Description: Cache open file handles on client.
1882 Details : MDS now will return special lock along with openhandle, if
1883 requested and client is allowed to hold openhandle, even if unused,
1884 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1885 closing files for every read/write openration.
1887 Severity : Enhancement
1889 Description: Cache open negative dentries on client when possible.
1890 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1891 negative dentries on lock revocation.
1896 Description: Remounting a client read-only wasn't possible with a zconf mount
1897 Details : It wasn't possible to remount a client read-only with llmount.
1899 Severity : enhancement
1900 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1901 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1902 MPICH is included to add Lustre-specific ADIO interfaces.
1903 This is based closely on the UFS ADIO layer and only differs
1904 in file creation, in order to allow the OST striping to be set.
1905 This is user-contributed code and not supported by CFS.
1910 Description: extended inode attributes (immutable, append-only) work improperly
1911 when 2.4 and 2.6 kernels are used on client/server or vice versa
1912 Details : Introduce kernel-independent values for these flags.
1914 Severity : enhancement
1917 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1918 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1919 granularity. Now, Fractional MB granularity is be supported,
1920 this is very useful for low memory system.
1922 Severity : enhancement
1924 Description: Getattr by fid
1925 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1926 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1927 have locks and avoids getattr rpc requests for it on every stat.
1930 Frequency : Always, for filesystems larger than 2TB
1932 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1933 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1934 filesystems larger than 2TB to be mounted with mballoc enabled.
1937 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1939 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1940 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1941 filesystems larger than 2TB is mounted then the kernel will
1942 silently corrupt the start of the filesystem. CONFIG_LBD is
1943 enabled for all CFS-supported kernels, but the possibility of
1944 this happening with a modified kernel config exists.
1946 Severity : enhancement
1948 Description: add client O_DIRECT support for 2.6 kernels
1949 Details : It is now possible to do O_DIRECT reads and writes to files
1950 in the Lustre client mountpoint on 2.6 kernel clients.
1952 Severity : enhancement
1954 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1955 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1956 parallel, not waiting for response from every OST before sending
1957 a rpc to the next OST.
1962 Description: setattr vs write race when updating file timestamps
1963 Details : Client processes that update a file timestamp into the past
1964 right after writing to the file (e.g. tar) it is possible that
1965 the updated file modification time can be reset to the current
1966 time due to a race between processing the setattr and write RPC.
1968 Severity : enhancement
1970 Description: Bring 'lfs find' closer in line with regular Linux find.
1971 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
1972 -print0 options and obtains all the needed info through the lustre
1975 Severity : enhancement
1977 Description: support up to 1024 configured devices on one node
1978 Details : change obd_dev array from statically allocated to dynamically
1979 allocated structs as they are first used to reduce memory usage
1984 Description: Flush dirty partially truncated pages during truncate
1985 Details : Immediatelly flush partially truncated pages in filter_setattr,
1986 this way we completely avoid having any pages in page cache on OST
1987 and can retire ugly workarounds during writes to flush such pages.
1992 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
1993 Details : protect inode from truncation within vfs_unlink() context
1994 just take a reference before calling vfs_unlink() and release it
1995 when parent's i_sem is free.
2000 Description: last_id value checked outside lock on OST caused LASSERT failure
2001 Details : If there were multiple MDS->OST object precreate requests in
2002 flight, it was possible that the OST's last object id was checked
2003 outside a lock and incorrectly tripped an assertion. Move checks
2004 inside locks, and discard old precreate requests.
2007 Frequency : always, if extents are used on OSTs
2009 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2010 Details : OSTs using the extents format would not zero the high 16 bits of
2011 the index physical block number. This is not a problem for any
2012 OST filesystems smaller than 16TB, and no kernels support ext3
2013 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2014 new/modified files) and can be fixed for existing filesystems
2015 with e2fsprogs-1.39-cfs1.
2020 Description: import connection selection may be incorrect if timer wraps
2021 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2022 import connections to be ignored if the 32-bit jiffies counter
2023 wraps. Use a 64-bit jiffies counter.
2026 Frequency : very large clusters immediately after boot
2028 Description: LNET request buffers exhausted under heavy short-term load
2029 Details : If a large number of client requests are generated on a service
2030 that has previously never seen so many requests it is possible
2031 that the request buffer growth cannot keep up with the spike in
2032 demand. Instead of dropping incoming requests, they are held in
2033 the LND until the RPC service can accept more requests.
2036 Frequency : Sometimes during replay
2038 Description: Assertion failure in ll_local_open after replay.
2039 Details : If replay happened on an open request reply before we were able
2040 to set replay handler, reply will become not swabbed tripping the
2041 assertion in ll_local_open. Now we set the handler right after
2042 recognising of open request
2045 Frequency : very rare
2047 Description: kernel reports "badness in vsnprintf"
2048 Details : Reading from the "recovery_status" /proc file in small chunks
2049 may cause a negative length in lprocfs_obd_rd_recovery_status()
2050 call to vsnprintf() (which is otherwise harmless). Exit early
2051 if there is no more space in the output buffer.
2053 Severity : enhancement
2055 Description: clear OBD RPC statistics by writing to them
2056 Details : It is now possible to clear the OBD RPC statistics by writing
2057 to the "stats" file.
2062 Description: Client mtime is not the same on different clients after utimes
2063 Details : In some cases, the client was using the utimes() syscall on
2064 a file cached on another node. The clients now validate the
2065 ctime from the MDS + OSTs to determine which one is right.
2070 Description: Inability to activate failout mode
2071 Details : lconf script incorrectly assumed that in pythong string's numeric
2072 value is used in comparisons.
2075 Frequency : always with multiple stripes per file
2077 Description: Inefficient object allocation for mutli-stripe files
2078 Details : When selecting which OSTs to stripe files over, for files with
2079 a stripe count that divides evenly into the number of OSTs,
2080 the MDS is always picking the same starting OST for each file.
2081 Return the OST selection heuristic to the original design.
2086 Description: mount failures may take full timeout to return an error
2087 Details : Under some heavy load conditions it is possible that a
2088 failed mount can wait for the full obd_timeout interval,
2089 possibly several minutes, before reporting an error.
2090 Instead return an error as soon as the status is known.
2092 Frequency : quota enabled and large files being deleted
2094 Description: releasing more than 4GB of quota at once hangs OST
2095 Details : If a user deletes more than 4GB of files on a single OST it
2096 will cause the OST to spin in an infinite loop. Release
2097 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2102 Description: statfs data retrieved from /proc may be stale or zero
2103 Details : When reading per-device statfs data from /proc, in the
2104 {kbytes,files}_{total,free,avail} files, it may appear
2105 as zero or be out of date.
2108 Frequency : systems with MD RAID1 external journal devices
2110 Description: lconf's call to blkid is confused by RAID1 journal devices
2111 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2112 of returning all block devices that match the journal UUID.
2115 Frequency : always, for aggregate stripe size over 4GB
2117 Description: assertion fails when trying to use 4GB stripe size
2118 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2119 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2122 Frequency : always on ppc64
2124 Description: the first write on an ext3 filesystem with mballoc got stuck
2125 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2126 perform endianness conversion.
2128 ------------------------------------------------------------------------------
2130 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2132 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2133 INTEROPERATE with older versions automatically. Please read the
2134 user documentation before upgrading any part of a live system.
2135 * WARNING: Lustre networking configuration changes are required with
2136 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2139 * Support for kernels:
2140 2.6.9-22.0.2.EL (RHEL 4)
2141 2.6.5-7.244 (SLES 9)
2142 2.6.12.6 vanilla (kernel.org)
2145 Severity : enhancement
2146 Bugzilla : 7981/8208
2147 Description: Introduced Lustre Networking (LNET)
2148 Details : LNET is new networking infrastructure for Lustre, it includes
2149 a reorganized network configuration mode (see the user
2150 documentation for full details) as well as support for routing
2151 between different network fabrics. Lustre Networking Devices
2152 (LNDs) for the supported network fabrics have also been
2153 created for this new infrastructure.
2155 Severity : enhancement
2156 Description: Introduced Access control lists
2157 Details : clients can set ACLs on files and directories in order to have
2158 more fine-grained permissions than the standard Unix UGO+RWX.
2159 The MDS must be started with the "-o acl" mount option.
2161 Severity : enhancement
2162 Description: Introduced filesystem quotas
2163 Details : Administrators may now establish per-user quotas on the
2166 Severity : enhancement
2168 Description: Configuration change for the XT3
2169 The PTLLND is now used to run Lustre over Portals on the XT3
2170 The configure option(s) --with-cray-portals are no longer used.
2171 Rather --with-portals=<path-to-portals-includes> is used to
2172 enable building on the XT3. In addition to enable XT3 specific
2173 features the option --enable-cray-xt3 must be used.
2178 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2179 Details : A race between allocating a new llcd and re-getting the llcd_lock
2180 allowed another thread to grab newly-allocated llcd.
2182 Severity : enhancement
2184 Description: 2.6 OST async journal commit and locking fix to improve performance
2185 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2186 2.6 kernels are now async as they already were in 2.4 kernels so
2187 that they can commit concurrently with the network bulk transfer.
2188 For block-allocated files the filter allocation semaphore is held
2189 to avoid filesystem fragmentation during allocation. BKL lock
2190 removed for 2.6 xattr operations where it is no longer needed.
2195 Description: lconf incorrectly determined whether two IP networks could talk
2196 Details : In some more complicated routing and multiple-network
2197 configurations, lconf will avoid trying to make a network
2198 connection to a disjoint part of the IP space. It was doing the
2199 math incorrectly for one set of cases.
2204 Description: Fix for potential infinite loop processing records in an llog.
2205 Details : If an llog record is corrupted/zeroed, it is possible to loop
2206 forever in llog_process(). Validate the llog record length
2207 and skip the remainder of the block on error.
2210 Frequency : occasional (liblustre only)
2212 Description: liblustre could not open files whose last component is a symlink
2213 Details : sysio_path_walk() would incorrectly pass the open intent to
2214 intermediate path components.
2217 Frequency : rare (liblustre only with non-standard tuning)
2218 Bugzilla : 7201 (7350)
2219 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2220 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2221 after it has been released, only do this for referenced locks
2223 Severity : enhancement
2225 Description: specifying an (invalid) directory default stripe_size of -1
2226 would reset the directory default striping
2227 Details : stripe_size -1 was used internally to signal directory stripe
2228 removal, now use "all default" to signal dir stripe removal
2229 as a directory striping of "all default" is not useful
2232 Frequency : common for large clusters running liblustre clients
2234 Description: doing an ls when liblustre clients are running is slow
2235 Details : sending a glimpse AST to a liblustre client waits for every AST
2236 to time out, as liblustre clients will not respond. Since they
2237 cannot cache data we refresh the OST lock LVB from disk instead.
2239 Severity : enhancement
2241 Description: doing an ls at the same time as file IO can be slow
2242 Details : enqueue and other "small" requests can be blocked behind many
2243 large IO requests. Create a new OST IO portal for non-IO
2244 requests so they can be processed faster.
2247 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2249 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2250 created file with mode 000
2251 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2252 allow an ftruncate() as a non-root user to a file with mode 000.
2253 Lustre now respects this flag to disable mode checks when
2254 truncating a file owned by the user
2257 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2260 Description: Revoking locks from clients that went dead or catatonic might take
2262 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2263 cancellation of such locks instant on servers without waiting for
2264 any reply from clients. Clients drops these locks when cancel
2265 notification from server is received without replying.
2268 Frequency : liblustre-only, when liblustre client dies or becomes busy
2270 Description: Doing ls on Linux clients can take a long time with active
2272 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2273 granting such locks to it in the first place if possible. Locks
2274 are taken by proxy on the OST during the read or write and
2275 dropped immediately afterward. Add connect flags handling, do
2276 not grant locks to liblustre clients for glimpse ASTs.
2278 Severity : enhancement
2280 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2281 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2282 sequential reads and behaves badly with applications doing random
2283 reads. Improve it by reading ahead at least read region, and
2284 avoiding excessive large RPC for small reads.
2286 Severity : enhancement
2288 Description: Creating more than 1000 files for a single job may cause a load
2289 imbalance on the OSTs if there are also a large number of OSTs.
2290 Details : qos_prep_create() uses an OST index reseed value that is an
2291 even multiple of the number of available OSTs so that if the
2292 reseed happens in the middle of the object allocation it will
2293 still utilize the OSTs as uniformly as possible.
2298 Description: OST or MDS may oops in ping_evictor_main()
2299 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2300 but doesn't restart at beginning of obd_exports_timed list
2303 Severity : enhancement
2305 Description: improve by-nid export eviction on the MDS and OST
2306 Details : allow multiple exports with the same NID to be evicted at one
2307 time without re-searching the exports list.
2310 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2312 Description: MDS may oops in groups_free()
2313 Details : in rare race conditions a newly allocated group_info struct is
2314 freed again, and this can be NULL. The 2.4 compatibility code
2315 for groups_free() checked for a NULL pointer, but 2.6 did not.
2318 Frequency : common for liblustre clients doing little filesystem IO
2319 Bugzilla : 9352, 7313
2320 Description: server may evict liblustre clients accessing contended locks
2321 Details : if a client is granted a lock or receives a completion AST
2322 with a blocking AST already set it would not reply to the AST
2323 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2326 Frequency : lfs setstripe, only systems with more than 160 OSTs
2328 Description: unable to set striping with a starting offset beyond OST 160
2329 Details : llapi_create_file() incorrectly limited the starting stripe
2330 index to the maximum single-file stripe count.
2333 Frequency : LDAP users only
2335 Description: lconf did not handle in-kernel recovery with LDAP properly
2336 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2338 Severity : enhancement
2340 Description: bind OST threads to NUMA nodes to improve performance
2341 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2342 node and do their allocations there to localize memory access
2344 Severity : enhancement
2346 Description: llmount can determine client NID directly from Myrinet (GM)
2347 Details : the client NID code from gmnalnid was moved directly into
2348 llmount, removing the need to use this or specifying the
2349 client NID explicitly when mounting GM clients with zeroconf
2352 Frequency : if client is started with down MDS
2354 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2355 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2356 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2357 ptlrpc_import_delayed_req() because we didn't check if the
2358 request was interrupted, and we also didn't break out of the
2359 event loop if there was no timeout
2364 Description: data loss during non-page-aligned writes to a single file from
2365 both multiple nodes and multiple threads on one node at same time
2366 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2367 inconsistency led to false short-reads, that were cached and later
2368 used by ->prepare_write() to fill in partially written page,
2369 leading to data loss.
2372 Frequency : always, if lconf --abort_recovery used
2374 Description: lconf --abort_recovery fails with 'Operation not supported'
2375 Details : lconf was attempting to abort recovery on the MDT device and not
2378 Severity : enhancement
2380 Description: remove cleanup logs
2381 Details : replace lconf-generated cleanup logs with lustre internal
2382 cleanup routines. Eliminates the need for client-cleanup and
2385 Severity : enhancement
2387 Description: add support for EAs (user and system) on lustre filesystems
2388 Details : it is now possible to store extended attributes in the Lustre
2389 client filesystem, and with the user_xattr mount option it
2390 is possible to allow users to store EAs on their files also
2392 Severity : enhancement
2394 Description: Add possibility (config option) to show minimal available OST free
2396 Details : When compiled with --enable-mindf configure option, statfs(2)
2397 (and so, df) will return least minimal free space available from
2398 all OSTs as amount of free space on FS, instead of summary of
2399 free spaces of all OSTs.
2401 Severity : enhancement
2403 Description: do not expand extent locks acquired on OST-side
2404 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2405 by server: they are not cached anyway.
2408 Frequency : when mmap is used/binaries executed from Lustre
2410 Description: Unmmap pages before throwing them away from read cache.
2411 Details : llap_shrink cache now attempts to unmap pages before discarding
2412 them (if unmapping failed - do not discard). SLES9 kernel has
2413 extra checks that trigger if this unmapping is not done first.
2418 Description: lconf didn't resolve symlinks before checking to see whether a
2419 given mountpoint was already in use
2422 Frequency : when migrating failover services
2423 Bugzilla : 6395, 9514
2424 Description: When migrating a subset of services from a node (e.g. failback
2425 from a failover service node) the remaining services would
2426 time out and evict clients.
2427 Details : lconf --force (implied by --failover) sets the global obd_timeout
2428 to 5 seconds in order to quickly disconnect, but this caused
2429 other RPCs to time out too quickly. Do not change the global
2430 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2432 Severity : enhancement
2433 Frequency : if MDS is started with down OST
2434 Bugzilla : 9439,5706
2435 Description: Allow startup/shutdown of an MDS without depending on the
2436 availability of the OSTs.
2437 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2438 Add appropriate locking and lov-osc refcounts for safe
2439 cleaning. Add osc abort_inflight calls in case the
2440 synchronize never started.
2443 Frequency : occasional (Cray XT3 only)
2445 Description: root not authorized to access files in CRAY_PORTALS environment
2446 Details : The client process capabilities were not honoured on the MDS in
2447 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2448 been accessed by an authorized user then root was able to access
2449 the file on the local client also. The root user capabilities
2450 are now allowed on the MDS, as this environment has secure UID.
2453 Frequency : occasional
2455 Description: ldiskfs "too long searching" message happens too often
2456 Details : A debugging message (otherwise harmless) prints too often on
2457 the OST console. This has been reduced to only happen when
2458 there are fragmentation problems on the filesystem.
2463 Description: Division by zero in statfs when all OSCs are inactive
2464 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2465 lov_statfs divides by value returned by lov_get_stripecnt().
2469 Bugzilla : 9489, 3273
2470 Description: First write from each client to each OST was only 4kB in size,
2471 to initialize client writeback cache, which caused sub-optimal
2472 RPCs and poor layout on disk for the first writen file.
2473 Details : Clients now request an initial cache grant at (re)connect time
2474 and so that they can start streaming writes to the cache right
2475 away and always do full-sized RPCs if there is enough data.
2476 If the OST is rebooted the client also re-establishes its grant
2477 so that client cached writes will be honoured under the grant.
2482 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2483 Details : Now I/O RPCs go to different portal number and (presumably) fast
2484 lock requests (and glimses) and other RPCs get their own service
2485 threads pool that should be able to service those RPCs
2488 Severity : enhancement
2490 Description: Ability to exchange lustre version between client and servers and
2491 issue warnings at client side if client is too old. Also for
2492 liblustre clients there is ability to refuse connection of too old
2494 Details : New 'version' field is added to connect data structure that is
2495 filled with version info. That info is later checked by server and
2499 Frequency : rare, liblustre only.
2500 Bugzilla : 9296, 9581
2501 Description: Two simultaneous writes from liblustre at offset within same page
2502 might proceed at the same time overwriting eachother with stale
2504 Details : I/O lock withing llu_file_prwv was released too early, before data
2505 actually was hitting the wire. Extended lock-holding time until
2506 server acknowledges receiving data.
2509 Frequency : extremely rare. Never observed in practice.
2511 Description: avoid generating lustre_handle cookie of 0.
2512 Details : class_handle_hash() generates handle cookies by incrementing
2513 global counter, and can hit 0 occasionaly (this is unlikely, but
2514 not impossible, because initial value of cookie counter is
2515 selected randonly). Value of 0 is used as a sentinel meaning
2516 "unassigned handle" --- avoid it. Also coalesce two critical
2517 sections in this function into one.
2519 Severity : enhancement
2521 Description: allow liblustre clients to delegate truncate locking to OST
2522 Details : To avoid overhead of locking, liblustre client instructs OST to
2523 take extent lock in ost_punch() on client's behalf. New connection
2524 flag is added to handle backward compatibility.
2526 Severity : enhancement
2527 Bugzilla : 4928, 7341, 9758
2528 Description: allow number of OST service threads to be specified
2529 Details : a module parameter allows the number of OST service threads
2530 to be specified via "options ost ost_num_threads={N}" in the
2531 OSS's /etc/modules.conf or /etc/modprobe.conf.
2535 Bugzilla : 6146, 9635, 9895
2536 Description: servers crash with bad pointer in target_handle_connect()
2537 Details : In rare cases when a client is reconnecting it was possible that
2538 the connection request was the last reference for that export.
2539 We would temporarily drop the export reference and get a new
2540 one, but this may have been the last reference and the export
2541 was just destroyed. Get new reference before dropping old one.
2543 Severity : enhancement
2544 Frequency : if client is started with failover MDS
2546 Description: Allow multiple MDS hostnames in the mount command
2547 Details : Try to read the configuration from all specified MDS
2548 hostnames during a client mount in case the "primary"
2551 Severity : enhancement
2553 Description: Stop sending data to evicted clients as soon as possible.
2554 Details : Check if the client we are about to send or are sending data to
2555 was evicted already. (Check is done every second of waiting,
2556 for which l_wait_event interface was extended to allow checking
2557 of exit condition at specified intervals).
2560 Frequency : rare, normally only when NFS exporting is done from client
2562 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2564 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2567 Frequency : always, if srand() is called before liblustre initialization
2569 Description: Liblustre uses system PRNG disturbing its usage by user application
2570 Details : Introduce internal to lustre fast and high-quality PRNG for
2571 lustre usage and make liblustre and some other places in generic
2572 lustre code to use it.
2574 Severity : enhancement
2575 Bugzilla : 9477, 9557, 9870
2576 Description: Verify that the MDS configuration logs are updated when xml is
2577 Details : Check if the .xml configuration logs are newer than the config
2578 logs stored on the MDS and report an error if this is the case.
2579 Request --write-conf, or allow starting with --old_conf.
2581 Severity : enhancement
2583 Description: Handle symlinks in the path when checking if Lustre is mounted.
2584 Details : Resolve intermediate symlinks when checking if a client has
2585 mounted a filesystem to avoid duplicate client mounts.
2590 Description: lconf can hit an error exception but still return success.
2591 Details : The lconf command catches the Command error exception at the top
2592 level script context and will exit with the associated exit
2593 status, but doesn't ensure that this exit status is non-zero.
2598 Description: failure of ptlrpc thread startup can cause oops
2599 Details : Starting a ptlrpc service thread can fail if there are a large
2600 number of threads or the server memory is very fragmented.
2601 Handle this without oopsing.
2604 Frequency : always, only if liblustre and non-default acceptor port was used
2606 Description: liblustre cannot connect to servers with non-default acceptor port
2607 Details : tcpnal_set_default_params() was not called and was therefore
2608 ignoring the environment varaible TCPNAL_PORT, as well as other
2609 TCPNAL_ environment variables
2614 Description: two objects could be created on the same OST for a single file
2615 Details : If an OST is down, in some cases it was possible to create two
2616 objects on a single OST for a single file. No problems other
2617 than potential performance impact and spurious error messages.
2621 Bugzilla : 5681, 9562
2622 Description: Client may oops in ll_unhash_aliases
2623 Details : Client dcache may become inconsistent in race condition.
2624 In some cases "getcwd" can fail if the current directory is
2630 Description: Inode refcounting problems in NFS export code
2631 Details : link_raw functions used to call d_instantiate without obtaining
2632 extra inode reference first.
2636 Bugzilla : 9942, 9903
2637 Description: Referencing freed requests leading to crash, memleaks with NFS.
2638 Details : We used to require that call to ll_revalidate_it was always
2639 followed by ll_lookup_it. Also with revalidate_special() it is
2640 possible to call ll_revalidate_it() twice for the same dentry
2641 even if first occurence returned success. This fix changes semantic
2642 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2643 reference on a request referred from the intent.
2644 ll_intent_release() then releases such a request.
2647 Frequency : rare, normally benchmark loads only
2649 Description: unlinked inodes were kept in memory on the client
2650 Details : If a client is repeatedly creating and unlinking files it
2651 can accumulate a lot of stale inodes in the inode slab cache.
2652 If there is no other client load running this can cause the
2653 client node to run out of memory. Instead flush old inodes
2654 from client cache that have the same inode number as a new inode.
2657 Frequency : SLES9 2.6.5 kernel and long filenames only
2658 Bugzilla : 9969, 10379
2659 Description: utime reports stale NFS file handle
2660 Details : SLES9 uses out-of-dentry names in some cases, which confused
2661 the lustre dentry revalidation. Change it to always use the
2665 Frequency : rare, unless heavy write-truncate concurrency is continuous
2666 Bugzilla : 4180, 6984, 7171, 9963, 9331
2667 Description: OST becomes very slow and/or deadlocked during object unlink
2668 Details : filter_destroy() was holding onto the parent directory lock
2669 while truncating+unlinking objects. For very large objects this
2670 may block other threads for a long time and slow overall OST
2671 responsiveness. It may also be possible to get a lock ordering
2672 deadlock in this case, or run out of journal credits because of
2673 the combined truncate+unlink. Solution is to do object truncate
2674 first in one transaction without parent lock, and then do the
2675 final unlink in a new transaction with the parent lock. This
2676 reduces the lock hold time dramatically.
2679 Frequency : rare, 2.4 kernels only
2681 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2682 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2683 before unmounting their filesystem in order to avoid deadlock.
2684 The kernel_locked() function in 2.4 kernels only checks whether
2685 the kernel lock is held, not whether it is this process that is
2686 holding it as 2.6 kernels do.
2691 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2692 Details : The client ptlrpc code may be trying to reconnect to a down
2693 server before a previous connection attempt has timed out.
2694 Increase the reconnect interval to be longer than the connection
2695 timeout interval to avoid sending duplicate connections to
2699 Frequency : echo_client brw_test command
2701 Description: fix echo_client to work with OST preallocated code
2702 Details : OST preallocation code (5137) didn't take echo_client IO path
2703 into account: echo_client calls filter methods outside of any
2704 OST thread and, hence, there is no per-thread preallocated
2705 pages and buffers to use. Solution: hijack pga pages for IO. As
2706 a byproduct, this avoids unnecessary data copying.
2710 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2711 Description: Client can oops in mdc_commit_close() after open replay
2712 Details : It was possible for the MDS to return an open request with no
2713 transaction number in mds_finish_transno() if the client was
2714 evicted, but without actually returning an error. Clients
2715 would later try to replay that open and may trip an assertion
2716 Simplify the client close codepath, and always return an error
2717 from the MDS in case the open is not successful.
2720 Frequency : rare, 2.6 OSTs only
2722 Description: OST may deadlock under high load on fragmented files
2723 Details : If there was a heavy load and highly-fragmented OST filesystems
2724 it was possible to have all the OST threads deadlock waiting on
2725 allocation of biovecs, because the biovecs were not released
2726 until the entire RPC IO was completed. Instead, release biovecs
2727 as soon as they are complete to ensure forward IO progress.
2729 Severity : enhancement
2731 Description: Support for specifying external journal device at mount
2732 Details : If an OST or MDS device is formatted with an external journal
2733 device, this device major/minor is stored in the ext3 superblock
2734 and may not be valid for failover. Allow detecting and
2735 specifying the external journal at mount time.
2740 Description: Mounting an MDS with pending unlinked files may cause oops
2741 Details : target_finish_recovery() calls mds_postrecov() which returned
2742 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2743 considers this an error and immediately begins cleaning up the
2744 lov, just after starting the mds_lov process
2746 Severity : enhancement
2748 Description: Implement 'lfs df' to report actual free space on per-OST basis
2749 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2750 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2751 usage of objects; '-h' to report in human readable format.
2753 ------------------------------------------------------------------------------
2755 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2762 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2763 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2764 at startup during orphan unlinking. Instead, skip buddy update
2765 before setup, as it will be regenerated after recovery is complete.
2770 Description: If an OST is inactive, its locks might reference stale inodes.
2771 Details : lov_change_cbdata() must iterate over all namespaces, even if
2772 they are inactive to clear inode references from the lock.
2774 Severity : enhancement
2775 Frequency : occasional, if non-standard max_dirty_mb used
2777 Description: Client will block write RPCs if not enough grant
2778 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2779 then the client will block writes while waiting for another RPC
2780 to complete instead of consuming its dirty limit. With change
2781 we get improved performance when max_dirty_mb is small.
2783 Severity : enhancement
2784 Bugzilla : 3389, 6253
2785 Description: Add support for supplementary groups on the MDS.
2786 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2787 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2788 lookups for user supplementary groups into a cache.
2792 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2793 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2794 for liblustre O_EXCL is enforced by MDS.
2798 Description: readdir never returns NULL in liblustre.
2799 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2800 dentry in struct dirent.
2804 Frequency : liblustre only, depends on application IO pattern
2805 Description: liblustre clients evicted if not contacting servers
2806 Details : Don't put liblustre clients into the ping_evictor list, so
2807 they will not be evicted by the pinger ever.
2809 Severity : enhancement
2811 Description: Add ability to evict clients by NID from MDS.
2812 Details : By echoing "nid:$NID" string into
2813 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2814 $NID would be instantly evicted from this MDS and from all active
2815 OSTs connected to it.
2819 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2820 Details : lookup_it_finish() used to query file size from OSTs that was not
2825 Description: service threads change working directory to that of init
2826 Details : Starting lustre service threads may pin the working directory
2827 of the parent thread, making that filesystem busy. Threads
2828 now change to the working directory of init to avoid this.
2832 Frequency : during shutdown only
2833 Description: shutdown with a failed MDS or OST can cause unmount to hang
2834 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2839 Frequency : 2.6 only
2840 Description: chmod/chown may include an extra supplementary group
2841 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2842 supplementary group and if none is specified this is used.
2845 Bugzilla : 5479 (6816)
2847 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2848 Details : If lookup is in progress on a file that is unlinked we might try
2849 to revalidate the inode and fail in revalidate after lookup is
2850 complete and ll_file_open() enqueues the open again but
2851 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2854 Frequency : always, if lconf --abort_recovery used
2856 Description: lconf --abort_recovery fails with 'Operation not supported'
2857 Details : lconf was attempting to abort recovery on the MDT device and not
2860 ------------------------------------------------------------------------------
2862 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2867 Frequency : rare (only unsupported configurations with a node running as an
2869 Bugzilla : 6514, 5137
2870 Description: Mounting a Lustre file system on a node running as an OST could
2872 Details : OSTs now preallocates memory needed to write out data at
2873 startup, instead of when needed, to avoid having to
2874 allocate memory in possibly low memory situations.
2875 Specifically, if the file system is mounted on on OST,
2876 memory pressure could force it to try to write out data,
2877 which it needed to allocate memory to do. Due to the low
2878 memory, it would be unable to do so and the node would
2879 become unresponsive.
2881 Severity : enhancement
2883 Description: Addition of lconf --service command line option
2884 Details : lconf now accepts a '--service <arg>' option, which is
2885 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2887 Severity : enhancement
2889 Description: Failover mode is now the default for OSTs.
2890 Details : By default, OSTs will now run in failover mode. To return to
2891 the old behaviour, add '--failout' to the lmc line for OSTs.
2893 Severity : enhancement
2895 Description: Health checks are now provided for MDS and OSTs
2896 Details : Additional detailed health check information on MSD and OSTs
2897 is now provided through the procfs health_check value.
2900 Frequency : occasional, depends on IO load
2902 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2903 after numerous create/delete cycles
2904 Details : The ext3 inode allocation policy would not allocate new inodes
2905 very well on the OSTs because there are no new directories
2906 being created. Instead we look for groups with free space if
2907 the parent directories are nearly full.
2912 Description: Network or server problems during mount may cause partially
2913 mounted clients instead of returning an error.
2914 Details : The config llog parsing code may overwrite the error return
2915 code during mount error handling, returning success instead
2921 Description: MDS can fail to allocate large reply buffers
2922 Details : After long uptimes the MDS can fail to allocate large reply
2923 buffers (e.g. zconf client mount config records) due to memory
2924 fragmentation or consumption by the buffer cache. Preallocate
2925 some large reply buffers so that these replies can be sent even
2926 under memory pressure.
2930 Frequency : rare (liblustre)
2931 Description: fsx running with liblustre complained that using truncate() to
2932 extend the file doesn't work. This patch corrects that issue.
2933 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2934 ATTR_SIZE and lsm use in llu_setattr_raw.
2938 Frequency : rare, only 2.6 kernels
2939 Description: Unusual file access patterns on the MDS may result in inode
2940 data being lost in very rare circumstances.
2941 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2942 code in the 2.6 kernel caused the inode and/or EA data not to
2943 be read from disk, causing single-file corruption.
2947 Frequency : rare, only 2.6 filesystems using extents
2948 Description: Heavy concurrent write and delete load may cause data corruption.
2949 Details : It was possible under high-load situations to have an extent
2950 metadata block in the block device cache from a just-unlinked
2951 file overwrite a newly-allocated data block. We now unmap any
2952 metadata buffers that alias just-allocated data blocks.
2956 Frequency : filesystems with default stripe_count larger than 77
2957 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2958 Details : lconf specifies an inode size of 4096 bytes when the default
2959 stripe_count is larger than 77. This conflicts with the default
2960 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2961 this case to avoid pinning too much memory for large EAs.
2963 ------------------------------------------------------------------------------
2965 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2970 Frequency : rare (extremely heavy IO load with hundreds of clients)
2972 Description: Client is evicted, gets IO error writing to file
2973 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
2974 caused clients to be evicted for AST timeouts. The fixes in
2975 bug 5192 mean we no longer need to have such short AST timeouts
2976 so ldlm_timeout has been increased.
2979 Frequency : occasional during --force or --failover shutdown under load
2980 Bugzilla : 5949, 4834
2981 Description: Server oops/LBUG if stopped with --force or --failover under load
2982 Details : a collection of import/export refcount and cleanup ordering
2983 issues fixed for safer force cleanup
2986 Frequency : only filesystems larger than 120 OSTs
2987 Bugzilla : 5990, 6223
2988 Description: lfs getstripe would oops on a very large filesystem
2989 Details : lov_getconfig used kfree on vmalloc'd memory
2992 Frequency : only filesystems exporting via NFS to Solaris 10 clients
2993 Bugzilla : 6242, 6243
2994 Description: reading from files that had been truncated to a non-zero size
2995 but never opened returned no data
2996 Details : ll_file_read() reads zeros from no-object files to EOF
3001 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3002 Details : There's little harm in aborting MDS/OSS recovery and letting it
3003 try again, so I removed the LASSERT and return an error instead.
3005 Severity : enhancement
3007 Description: New debugging infrastructure for tracking down data corruption
3008 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3009 (b) cover more of the client-side code path, and (c) try to narrow
3010 down where problems occurred
3014 Bugzilla : 3819, 4364, 4397, 6313
3015 Description: Racing close and eviction MDS could cause assertion in mds_close
3016 Details : It was possible to get multiple mfd references during close and
3017 client eviction, leading to one thread referencing a freed mfd.
3019 Severity: : enhancement
3020 Bugzilla : 3262, 6359
3021 Description: Attempts to reconnect to servers are now more aggressive.
3022 Details : This builds on the enhanced upcall-less recovery that was added
3023 in 1.4.2. When trying to reconnect to servers, clients will
3024 now try each server in the failover group every 10 seconds. By
3025 default, clients would previously try one server every 25 seconds.
3030 Description: After recovery, certain operations trigger a failed
3031 assertion on a client.
3032 Details : Failing over an mds, using lconf -d --failover, while a
3033 client was doing a readdir() call would cause the client to
3034 LBUG after recovery completed and the readdir() was resent.
3036 Severity : enhancement
3038 Description: Default groups are now added by lconf
3039 Details : You can now run lconf --group <servicename> without having to
3040 manually add groups with lmc.
3043 Frequency : occasional
3045 Description: Nodes with an elan id of 0 trigger a failed assertion
3048 Frequency : always when accessing e.g. tty/console device nodes
3050 Description: tty and some other devices nodes cannot be used on lustre
3051 Details : file's private_data field is used by device data and lustre
3052 values in there got lost. New field was added to struct file to
3053 store fs-specific private data.
3056 Frequency : when exporting Lustre via NFS
3058 Description: NFSD failed occasionally when looking up a path component
3059 Details : NFSD is looking up ".." which was broken in ext3 directories
3060 that had grown large enough to become hashed.
3063 Frequency : Clusters with multiple interfaces not on the same subnet
3065 Description: Nodes will repeatedly try to reconnect to an interface which it
3066 cannot reach and report an error to the log.
3067 Details : Extra peer list entries will be created by lconf with some peers
3068 unreachable. lconf now validates the peer before adding it.
3071 Frequency : Only if a default stripe is set on the filesystem root.
3073 Description: Setting a default stripe on the filesystem root prevented the
3074 filesystem from being remounted.
3075 Details : The client was sending extra request flags in the root getattr
3076 request and did not allocate a reply buffer for the dir EA.
3079 Frequency : occasional, higher if lots of files are accessed by one client
3080 Bugzilla : 6159, 6097
3081 Description: Client trips assertion regarding lsm mismatch/magic
3082 Details : While revalidating inodes the VFS looks up inodes with ifind()
3083 and in rare cases can find an inode that is being freed.
3084 The ll_test_inode() code will free the lsm during ifind()
3085 when it finds an existing inode and then the VFS later attaches
3086 this free lsm to a new inode.
3090 Bugzilla : 6422, 7030
3091 Description: MDS deadlock between mkdir and client eviction
3092 Details : Creating a new file via mkdir or mknod (starting a transaction
3093 and getting the ns lock) can deadlock with client eviction
3094 (gets ns lock and trying to finish a synchronous transaction).
3097 Frequency : occasional
3098 Description: While starting a server, the fsfilt_ext3 module could not be
3100 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3101 kernels. Previously, lconf would still use the ext3 name
3102 when trying to load modules. Now, it will correctly use
3103 ext3 on 2.4 and ldiskfs on 2.6.
3105 Severity : enhancement
3106 Description: The default stripe count has been changed to 1
3107 Details : The interpretation of the default stripe count (0, to lfs
3108 or lmc) has been changed to mean striping across a single
3109 OST, rather than all available. For general usage we have
3110 found a stripe count of 1 or 2 works best.
3112 Severity : enhancement
3113 Description: Add support for compiling against Cray portals.
3114 Details : Conditional compiling for some areas that are different
3118 Frequency : occasional
3119 Bugzilla : 6409, 6834
3120 Description: Creating files with an explicit stripe count may lead to
3121 a failed assertion on the MDS
3122 Details : If some OSTs are full or unavailable, creating files may
3123 trigger a failed assertion on the MDS. Now, Lustre will
3124 try to use other servers or return an error to the
3128 Frequency : occasional
3130 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3131 caused lock timeout during readahead (since 1.4.2).
3132 Details : Processes doing readahead might match a lock that hasn't been
3133 granted yet if there are overlapping and conflicting lock
3134 requests. The readahead process waits on ungranted lock
3135 (original lock is CBPENDING), while OST waits for that process
3136 to cancel CBPENDING read lock and eventually evicts client.
3138 Severity : enhancement
3140 Description: Initial enabling of flock support for clients
3141 Details : Implements fcntl advisory locking and file status functions.
3142 This feature is provided as an optional mount flag (default
3143 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3144 locking are implemented yet, and those that are are not guaranteed
3145 to be completely correct in production environments.
3146 mount -t lustre -o [flock|noflock] ...
3149 Frequency : occasional
3151 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3152 assertion in the ext3 JBD (journaling) layer.
3153 Details : The b_committed_data struct is protected by the big kernel lock
3154 in 2.4 kernels, serializing journal_commit_transaction() and
3155 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3156 access to this struct. In 2.6 kernels there is finer grained
3157 locking to improve SMP performance of the JBD layer.
3161 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3166 Description: merge and cleanup kernel patches.
3169 -----------------------------------------------------------------------------
3171 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3173 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3174 versions of Lustre. Please update all servers and clients to
3175 version 1.4.2 or later at the same time. You must also run
3176 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3177 to update the configuration logs.
3179 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3180 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3181 - swab llog records of type '0' so we get proper header size/idx (5861)
3182 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3183 - fix rename of one directory over another leaking an inode (5953)
3184 - avoid SetPageDirty on 2.6 (5981)
3185 - don't re-add just-being-destroyed locks to the waiting list (5653)
3186 - when creating new directories, inherit the parent's custom
3187 striping settings if present parent (3048)
3188 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3189 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3190 - don't hold client locks on temporary worklist from l_lru (5666)
3191 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3192 - automatically evict dead clients (5921)
3193 - Update file size properly in create+truncate+fstat case (6196)
3194 - Do not unhash mountpoint dentries, do not allow removal of
3196 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3197 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3198 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3199 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3200 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3201 - Correctly handle reads of files with no objects (6243)
3202 - lctl recover will also mark a device active if deactivate used (5933)
3204 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3205 - allow --write-conf on an MDS with different nettype than client (5619)
3206 - don't write config llogs to MDS for mounts not from that MDS (5617)
3207 - lconf should create multiple TCP connections from a client (5201)
3208 - init scripts are now turned off by default; run chkconfig --on
3209 lustre and chkconfig --on lustrefs to use them
3210 - upcalls are no longer needed for clients to recover to failover
3212 - add --abort-recovery option to lconf to abort recovery on device
3214 - add support for an arbitrary number of OSTs (3026)
3215 - Quota support protocol changes.
3216 - forward compatibility changes to wire structs (6007)
3217 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3218 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3219 - improved "lustre status" script
3220 - initialize blocksize for non-regular files (6062)
3221 - added --disable-server and --disable-client configure options (5782)
3222 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3223 - Vanilla 2.4.29 support
3224 - increase maximum number of obd devices to 520 (6242)
3225 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3226 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3227 SLES 9 2.6.5 kernels
3228 - limit stripes per file to 160 (the maximum EA size) (6093)
3230 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3233 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3234 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3235 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3236 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3237 - small ext3 extents cleanups and fixes (5733)
3238 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3239 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3240 - avoid needless client->OST connect, fix handle mismatch (5317)
3241 - fix DLM error path that led to out-of-sync client, long delays (5779)
3242 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3243 - fix several locking issues related to i_size (5492,5624,5654,5672)
3244 - don't move pending lock onto export if it is already evicted (5683)
3245 - fix kernel oops when creating .foo in unlinked directory (5548)
3246 - fix deadlock in obdfilter statistics vs. object create (5811)
3247 - use time_{before,after} to avoid timer jiffies wrap (5882)
3248 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3249 - Do not leak request if server was not able to process it (5154)
3250 - If mds_open unable to find parent dir, make that negative lookup(5154)
3251 - don't create new directories with extent-mapping (5909, 5936)
3253 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3254 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3255 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3256 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3257 - support for RHEL4 kernel on i686 (5773)
3258 - provide error messages when incompatible logs are encountered (5898)
3260 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3261 * version 1.4.0.10 (1.4.1 release candidate 1)
3263 - don't keep a lock reference when lock is not granted (4238)
3264 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3265 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3266 - drop import inflight refcount on signal_completed_replay error (5255)
3267 - unlock page after async write error during send (3677)
3268 - handle missing objects in filter_preprw_read properly (5265)
3269 - no transno return for symlink open, don't save no-trasno open (3440)
3270 - don't try to complete elan receive that already failed (4012)
3271 - free RPC server reply state on error (5406)
3272 - clean up thread from ptlrpc_start_thread() on error (5160)
3273 - readahead could read extra page into cache that wasn't ejected (5388)
3274 - prevent races in class_attach/setup/cleanup/detach (5260)
3275 - don't dereference de->d_inode after l_dput of de (5458)
3276 - use "int" for stripe value returned from lock_to_stripe (5544)
3277 - mballoc allocation and error-checking fixes in 2.6 (5504)
3278 - block device patches to fix I/O request sizes in 2.6 (5482)
3279 - look up hostnames for IB nals (5602)
3280 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3281 - don't start multiple acceptors for the same port (5277)
3282 - fix incorrect LASSERT in mds_getattr_name (5635)
3283 - export a proc file for general "ping" checking (5628)
3284 - fix "lfs check" to not block when the MDS is down (5628)
3286 - service request history (4965)
3287 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3288 - create an "evict_client" /proc entry on OSTs, like the MDS has
3289 - fix mount usage message, return errors per mount(8) (5168)
3290 - change grep [] to grep "[]" in tests so they work in more UMLs
3291 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3292 - remove ancient LOV_MAGIC_V0 EA support (5047)
3293 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3294 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3295 - allow readahead allocations to fail when low on memory (5383)
3296 - mmap locking landed again, after considerable improvement (2828)
3297 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3299 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3302 - send OST transaction number in read/write reply to free req (4966)
3303 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3304 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3305 - the watchdog thread now runs as interruptible (5246)
3306 - flock/lockf fixes (but it's still disabled, pending 5135)
3307 - don't use EXT3 constants in llite code (5094)
3308 - memory shortage at startup could cause assertion (5176)
3310 - reorganization of lov code
3311 - single portals codebase
3313 - add extents/mballoc support (5025)
3314 - direct I/O reads in the obdfilter (4048)
3315 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3317 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3320 - send OST transaction number in read/write reply to free req (4966)
3321 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3322 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3323 - the watchdog thread now runs as interruptible (5246)
3324 - handle missing objects in filter_preprw_read properly (5265)
3325 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3326 - drop import inflight refcount on signal_completed_replay error (5255)
3327 - unlock page after async write error during send (3677)
3328 - return original error code on reconstructed replies (3761)
3329 - no transno return for symlink open, don't save no-trasno open (3440)
3331 - add pid to ldlm debugging output (4922)
3332 - bump the watchdog timeouts -- we can't handle 30sec yet
3333 - extra debugging for orphan dentry/inode bug (5259)
3335 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3338 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3339 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3340 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3341 - fix NULL dereference in /proc/sys/portals/routes (4827)
3342 - allow failed mdc_close() operations to be interrupted (4561)
3343 - stop precreate on OST before MDS would time out on it (4778)
3344 - don't send partial-page writes before EOF from client (4410)
3345 - discard client grant for sub-page writes on large-page clients (4520)
3346 - don't free dentries not owned by NFS code, check generation (4806)
3347 - fix lsm leak if mds_create_objects() fails (4801)
3348 - limit debug_daemon file size, always print CERROR messages (4789)
3349 - use transno after validating reply (3892)
3350 - process timed out requests if import state changes (3754)
3351 - update mtime on OST during writes, return in glimpse (4829)
3352 - add mkfsoptions to LDAP (4679)
3353 - use ->max_readahead method instead of zapping global ra (5039)
3354 - don't interrupt __l_wait_event() during strace
3356 - add software watchdogs to catch hung threads quickly (4941)
3357 - make lustrefs init script start after nfs is mounted
3358 - fix CWARN/ERROR duplication (4930)
3359 - return async write errors to application if possible (2248)
3360 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3361 - print NAL number in %x format (4645)
3362 - update barely-supported suse-2.4.21-171 series (4842)
3363 - support for sles 9 %post scripts
3364 - support for building 2.6 kernel-source packages
3365 - support for sles km_* packages
3367 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3370 - ignore -ENOENT errors in osc_destroy (3639)
3371 - notify osc create thread that OSC is being cleaned up (4600)
3372 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3373 - reconstruct ost_handle() like mds_handle() (4657)
3374 - create a new thread to do import eviction to avoid deadlock (3969)
3375 - let lconf resolve symlinked-to devices (4629)
3376 - don't unlink "objects" from directory with default EA (4554)
3377 - hold socknal file ref over connect in case target is down (4394)
3378 - allow more than 32000 subdirectories in a single directory (3244)
3379 - fix blocks count for O_DIRECT writes (3751)
3380 - OST returns ENOSPC from object create when no space left (4539)
3381 - don't send truncate RPC if file size isn't changing (4410)
3382 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3383 - bind to privileged port in socknal and tcpnal (3689)
3385 - rate limit CERROR/CWARN console message to avoid overload (4519)
3386 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3387 - basic mmap support (3918)
3388 - kernel patch series update from b1_4 (4711)
3390 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3393 - avoid crash during MDS cleanup with OST shut down (2775)
3394 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3395 - don't use bad inodes on the MDS (3744)
3396 - dynamic object preallocation to improve recovery speed (4236)
3397 - don't hold spinlock over lock dumping or change debug flags (4401)
3398 - don't zero obd_dev when it is force cleaned (3651)
3399 - print grants to console if they go negative (4431)
3400 - "lctl deactivate" will stop automatic recovery attempts (3406)
3401 - look for existing locks in ldlm_handle_enqueue() (3764)
3402 - don't resolve lock handle twice in recovery avoiding race (4401)
3403 - revalidate should check working dir is a directory (4134)
3405 - don't always mark "slow" obdfilter messages as errors (4418)
3407 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3410 - don't close LustreDB during write_conf until it is done (3860)
3411 - fix typo in lconf for_each_profile (3821)
3412 - allow dumping logs from multiple threads at one time (3820)
3413 - don't allow multiple threads in OSC recovery (3812)
3414 - fix debug_size parameters (3864)
3415 - fix mds_postrecov to initialize import for llog ctxt (3121)
3416 - replace config semaphore with spinlock (3306)
3417 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3418 - don't allow enqueue to complete on a destroyed export (3822)
3419 - down write_lock before checking llog header bitmap (3825)
3420 - recover from lock replay timeout (3764)
3421 - up llog sem before sending rpc (3652)
3422 - reduce ns lock hold times when setting kms (3267)
3423 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3424 - fix NULL deref and obd_dev leak on setup error (3312)
3425 - replace some LBUG about llog ops with error handling (3841)
3426 - don't match INVALID dentries from d_lookup and spin (3784)
3427 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3428 - fix invalid assertion in ptlrpc_set_wait (3880)
3430 - add libwrap support for the TCP acceptor (3996)
3431 - add /proc/sys/portals/routes for non-root route listing (3994)
3432 - allow setting MDS UUID in .xml (2580)
3433 - print the stack of a process that LBUGs (4228)
3435 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3438 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3439 - make sure to unset replay flag from failed open requests (3440)
3440 - if default stripe count is 0, use OST count for inode size (3636)
3441 - update parent mtime/ctime on client for create/unlink (2611)
3442 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3443 - free recovery state on server during a forced cleanup (3571)
3444 - unregister_reply for resent reqs (3063)
3445 - loop back devices mounting and status check on 2.6 (3563)
3446 - fix resource-creation race that can provoke i_size == 0 (3513)
3447 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3448 - more debugging for page-accounting assertion (3746)
3449 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3450 - avoid deadlock after precreation failure (3758)
3451 - fix race and lock order deadlock in orphan handling (3450, 3750)
3452 - add validity checks when grabbing inodes from l_ast_data (3599)
3454 - add /proc/.../recovery_status to obdfilter (3428)
3455 - lightweight CDEBUG infrastructure, debug daemon (3668)
3456 - change default OSC RPC parameters to be better on small clusters
3457 - turn off OST read cache for files smaller than 32MB
3458 - install man pages and include them in rpms (3100)
3459 - add new init script for (un)mounting lustre filesystems (2593)
3460 - run chkconfig in %post for init scripts (3701)
3461 - drop scimac NAL (unmaintained)
3463 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3466 - clean kiobufs before and after use (3485)
3467 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3468 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3469 - change init script to fail more clearly if not run as root (1528)
3470 - allow clients to reconnect during replay (1742)
3471 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3472 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3473 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3474 - fix cond_resched (3554)
3475 - only evict unfinished clients after recovery (3515)
3476 - allow bulk resend, prevent data loss (3570)
3477 - dynamic ptlrpc request buffer allocation (2102)
3478 - don't allow unlinking open directory if it isn't empty (2904)
3479 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3480 - remove extraneous obd dereference causing LASSERT failure (3334)
3481 - don't use get_cycles() when creating temp. files on the mds (3156)
3482 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3483 - handle EEXIST for set-stripe, set proper directory name (3336)
3485 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3486 - fix ksocknal_fmb_callback() error messages (2918)
3488 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3491 - don't copy lvb into (possibly NULL) reply on error (2983)
3492 - don't deref dentry after dput, don't free lvb on error (2922)
3493 - use the kms to determine writeback rpc length (2947)
3494 - increment oti_logcookies when osc is inactive (2948)
3495 - update client's i_blocks count via lvb messages (2543)
3496 - handle intent open/close of special files properly (1557)
3497 - mount MDS with errors=remount-ro, like obdfilter (2009)
3498 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3499 - don't use cancelling-locks' kms values (2947)
3500 - use highest lock extent for kms, not last one (2925)
3501 - don't dereference ERR_PTR() dentry in error handling path (3107)
3502 - fix thread race in portals_debug_dumplog() (3122)
3503 - create lprocfs device entries at setup instead of at attach (1519)
3504 - common AST error handler, don't evict client on completion race (3145)
3505 - zero nameidata in detach_mnt in 2.6 (3118)
3506 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3507 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3508 - initialize RPC timeout timer earlier for 2.6 (3219)
3509 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3510 - print nal/nid for unknown nid (3258)
3511 - additional checks for oscc recovery before doing precreate (3284)
3512 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3513 - don't crash in mdc_close for bad permissions on open (3285)
3514 - zero i_rdev for non-device files (3147)
3515 - clear page->private before handing to FS, better assertion (3119)
3516 - tune the read pipeline (3236)
3517 - fix incorrect decref of invalidated dentry (2350)
3518 - provide read-ahead stats and refine rpc in flight stats (3328)
3519 - don't hold journal transaction open across create RPC (3313)
3520 - update atime on MDS at close time (3265)
3521 - close LDAP connection when recovering to avoid server load (3315)
3522 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3523 - don't leak open file on MDS after open resend (3325)
3524 - serialize filter_precreate and filter_destroy_precreated (3329)
3525 - loop device shouldn't call sync_dev() for nul device (3092)
3526 - clear page cache after eviction (2766)
3527 - resynchronize MDS->OST in background (2824)
3528 - refuse to mount the same filesystem twice on same mountpoint (3394)
3529 - allow llmount to create routes for mounting behind routers (3320)
3530 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3531 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3532 - fix rare problem with rename on htree directories (3417)
3534 - allow default OST striping configuration per directory (1414)
3535 - fix compilation for qswnal for 2.6 kernels (3125)
3536 - increase maximum number of MDS request buffers for large systems
3537 - change liblustreapi to be useful for external progs like lfsck (3098)
3538 - increase local configuration timeout for slow disks (3353)
3539 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3541 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3544 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3545 - don't overwrite extent policy data in reply if lock was blocked (2901)
3546 - drop filter export grants atomically with removal from device (2663)
3547 - del obd_self_export from work_list in class_disconnect_exports (2908)
3548 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3549 - swab reply message in mdc_close, other PPC fixes (2464)
3550 - fix destroying of named logs (2325)
3551 - overwrite old logs when running lconf --write_conf (2264)
3552 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3553 - fix race in target_handle_connect (2898)
3554 - mds_reint_create() should take same inode create lock (2926)
3555 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3556 - don't close files for self_export to avoid uninitialized obd (2936)
3557 - allow MDS with the same name as client node (2939)
3558 - hold dentry reference for closed log files for unlink (2325)
3559 - reserve space for all logs during transactions (2059)
3560 - don't evict page beyond end of stripe extent (2925)
3561 - don't oops on a deleted current working directory (2399)
3562 - handle hard links to targets without a parent properly (2517)
3563 - don't dereference NULL lock when racing during eviction (2867)
3564 - don't grow lock extents when lots of conflicting locks (2919)
3566 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3569 - account for cache space usage on clients to avoid data loss (974)
3570 - lfsck support in lustre kernel code (2349)
3571 - reduce journal credits needed for BRW writes (2370)
3572 - orphan handling to avoid losing space on client/server crashes
3573 - ptlrpcd can be blocked, stopping ALL progress (2477)
3574 - use lock value blocks to assist in proper KMS, faster stat (1021)
3575 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3576 - recovery for initial connections (2355)
3577 - fixes for mds_cleanup_orphans (1934)
3578 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3579 - block all file creations until orphan recovery completes (1901)
3580 - client remove rq_connection from request struct (2423)
3581 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3582 - recovery timer race (2670)
3583 - mdc_close recovey bug (2532)
3584 - ptlrpc cleanup bug (2710)
3585 - mds timeout on local locks (2588)
3586 - namespace lock held during RPCs (2431)
3587 - handle interrupted sync write properly (2503)
3588 - don't try to handle a message that hasn't been replied to (2699)
3589 - client assert failure during cleanup after abort recovery (2701)
3590 - leak mdc device after failed mount (2712)
3591 - ptlrpc_check_set allows timedout requests to complete (2714)
3592 - wait for inflight reqs when ptlrpcd finishes (2710)
3593 - make sure unregistered services are removed from the srv_list
3594 - reset bulk XID's when resending them (caught by 1138 test)
3595 - unregister_bulk after timeout
3596 - fix lconf error (2694)
3597 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3598 - readahead locks pages, leaves pending causing memory pressure (2673)
3599 - increase OST request buffers to 4096 on large machines (2729)
3600 - fix up permission of existing directories in simple_mkdir (2661)
3601 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3602 - don't assign transno to errored transactions (2742)
3603 - don't delete objects on OST if given a bogus objid from MDS (2751)
3604 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3605 - if rq_no_resend, then timeout request after recovery (2432)
3606 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3607 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3608 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3609 - MDS recovery shouldn't delete objects that it has given out (2730)
3610 - if enqueue arrives after completion, don't clobber LVB (2819)
3611 - don't unlock pages twice when trigger_group_io returns error (2814)
3612 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3613 - don't write pages to disk if there was an error (1450)
3614 - don't ping imports that have recovery disabled (2676)
3615 - take buffered bytes into account when balancing socknal conn (2817)
3616 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3617 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3618 - remove little-endian swabbing of llog records (1987)
3619 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3620 - retry reposting request buffers if they fail (1191)
3621 - grow extent at grant time to avoid granting a revoked lock (2809)
3622 - lock revoke doesn't evict page if covered by a second lock (2765)
3623 - disable VM readahead to avoid reading outside lock extents (2805)
3625 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3626 - updated kernel patches for hp-2.4.20 kernel (2681)
3628 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3631 - fix truncated write corruption (2366)
3632 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3634 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3635 - ensure len(uuid) < 37 in lmc (1171)
3636 - fix ia64 OOPS in llog_test (2255)
3637 - zero end of page at obdfilter for partial page writes (2648)
3638 - don't leave stale dentries around after renames (bug 2428)
3639 - fix timeouts when evicting a client with a single lock held (2642)
3640 - set deadline for the initial HELLO message to drain (2634)
3641 - print out dotted-quad IP addresses in the socknal (2302)
3643 - additional debugging for MDS client eviction problem (2443)
3644 - fix mkfsoptions support for osts (2603, 2604)
3646 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3649 - add series for the vanilla 2.6.0 kernel
3650 - add series for the vanilla 2.4.24 kernel
3651 - add series for a cray x86/64 UL kernel drop
3652 - fix xattr patches for the vanilla 2.4.19 series
3654 - generate true UUIDs in lmc (1171)
3655 - have portals stack dumping break in UML (2466)
3656 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3657 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3658 - wake callback waiting threads on client eviction (2460)
3659 - Add --ptldebug and --subsystem to lmc (1719)
3660 - update assertion to allow safe interrupt allocation
3661 - set rq_no_resend for cancel requests (2432)
3662 - recalculate ptlrpcd timeout after resend (2494)
3663 - call vfs_rmdir when removing pending directories (2368)
3664 - fix renaming a file to itself (2429)
3665 - lmc creates a default one-stripe lov (2454)
3666 - expand procfs space to handle large clusters (2326)
3667 - increase UML stack to avoid overflow
3668 - update lconf's list of debug and subsystem masks
3669 - fix lfs find --obd (2510)
3670 - /proc tunable for disabling filter read caching (2591)
3671 - stop rpm packages from altering slapd.conf (2301)
3672 - disable nagle in the socknal under 0conf (2578)
3673 - choose mds inode size based on stripe count (2572)
3674 - fix kernel-source rpm problems (2516)
3676 - add --disable-doc to avoid pdf generation (2421)
3677 - update documentation, tests, type-os, comments
3678 - avoid format warnings on ia64
3679 - remove the TOE NAL
3680 - tiny code cleanups by removing unused fields
3682 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3685 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3686 - remove the most blatant lies from BUILDING file (2371)
3687 - change default debug level to reasonable production setting
3688 - reduce client side cache size to reduce cache flush time
3689 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3690 - make TCP zerocopy and pinger support enabled by default (2476)
3691 - sync writes completed after process exits caused crashes (2319)
3692 - maintain correct mount count on the MDS (2356)
3693 - backout 1557, because 2316 wasn't really fixed
3694 - better file I/O statistics gathering in /proc
3695 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3696 - another kernel patch to fix zero-copy TCP function export
3697 - don't take duplicate lock when processing re-sent getattr (2420)
3698 - lctl uses obd_self_export instead of creating new conn (2353)
3699 - MDS/OST recovery case which requires object creation asserted (2425)
3700 - move lfs from /usr/sbin to /usr/bin in packages
3701 - fix race between mds_client_add and mds_client_free (2417)
3702 - use kmalloc instead of slabs in portals (2430)
3703 - don't create duplicate records when a failover MDS is present (2442)
3704 - remove unnecessary mount age check (2332)
3705 - don't remove directory inodes from locks prematurely (2451)
3706 - don't break if MDS service name is the same as hostname (2103)
3707 - fix races in client write RPC generation when cache full (2482)
3709 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3712 - remove now-unused request->rq_obd (278)
3713 - if an allocation fails, print out how much memory we've used (1933)
3714 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3715 - add the "configurable stack size" patch to most series files (1256)
3716 - ability to write large log records, for 100+ OST configs (2306)
3717 - fix NULL deref when filter_prep fails (2314)
3718 - fix operator precedence error in filter_sync
3719 - dynamic allocation of socknal TX descriptors (2315)
3720 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3721 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3722 - ensure that CWARN messages in Portals always get to the syslog
3723 - __init/__exit are not for prototype decls (ldlm_init/exit)
3724 - x86-64 compile warning fixes
3725 - fix gateway LMC keyword conflict (2318)
3726 - fix MDS lock inversions in getattr/reint paths (1844)
3727 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3728 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3729 - DLM race condition prevented some lock evictions (2328)
3730 - ENOMEM detection and retry on socknal sends (2230)
3731 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3732 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3733 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3734 - better error messages when a client is rejected during recovery (1505)
3735 - avoid cancelling locks which were never granted, after failure (2330)
3736 - fix i_sem/journal inversion in mds_client_add (2333)
3737 - fix truncate/getattr lock cycle deadlock (2334)
3738 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3739 - fix two rare exit paths which could leak an l_lock() ref (2321)
3740 - fencepost error in MDS/OST orphan recovery (2226)
3741 - make log record alignment 8 bytes (1988)
3742 - lstripe now fails when requested offset > ost_count (2237)
3743 - ensure that all kernel series have a complete list.h (1607)
3744 - fix crashes in special-file operations (2316)
3745 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3747 - allow configurable automake binary, for testing new versions
3748 - small update to the lfs documentation
3750 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3752 * fix negative export reference count in fsfilt_sync (2312)
3754 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3755 * release candidate 0.9.1
3757 - orphans are moved into the PENDING directory for possible recovery
3758 - replayed opens now open by fid for orphan/rename safety (1042)
3759 - last close of an orphan inode generates a transno (683)
3760 - chdir() and mount() now pin the directory entry (1020)
3761 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3762 - discard very old requests without processing them (1502)
3763 - remove some common, well-understood CERRORs (1505)
3764 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3765 - clear "grant" flags in OST replies until OST grant code lands (1644)
3766 - fix read performance by not clobbering i_blksize on client (1598)
3767 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3768 - make LRU size a /proc tunable, clears locks when reduced (707)
3769 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3770 - server locks take references on exports now (1558)
3771 - build fixes for 2.4.20-rh trees (1663)
3772 - return an error from lov_create if all OSCs are inactive (1751)
3773 - fix import levels when a reconnect happens without a timeout (1597)
3774 - exit early from mds_open if we get a lookup error (1749)
3775 - partial page read at EOF wouldn't wait for disk before sending (1642)
3776 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3777 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3778 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3779 - write/truncate lock inversion (1639)
3780 - Don't auto-load obdclass, portals modules during cleanup (1495)
3781 - fix timestamps from jumping to "now" (1763)
3782 - extra journal assertions (1648)
3783 - add an extra multiunlink test (1771)
3784 - fix read_record/write_record API (1776)
3785 - fix leak of offset_extent, possible incorrect i_size later (1772)
3786 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3787 - Debugging for the kqswnal_get_idle_tx problems (1820)
3788 - Allow recovery to be attempted multiple times (1536)
3789 - Write out MDS last_rcvd file after it is first created (1600)
3790 - Fix tx_descriptor leak in failed transmit situations (1827)
3791 - ext3 journaling fixes for assertion failure after IO error (1871)
3792 - class_export_put() on freed export after completion AST error (1896)
3793 - Fix revalidate looping in VFS (1322)
3794 - Don't access a freed export during MDS_REINT timeout (1521)
3795 - Add open-unlink recovery support on the MDS (1673,1764)
3796 - Return an error if no MDS data was read from last_rcvd (1946)
3797 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3798 - Don't setup a disk device that doesn't match exported UUID (317)
3799 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3800 - avoid committing NULL handle in force close
3801 - local.sh is now a one-stripe LOV configuration
3802 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3803 - fix ext3 htree duplicate directory entry corruption (1516)
3804 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3805 - update to new LOV EA format (2097)
3806 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3808 2003-06-15 Phil Schwan <phil@clusterfs.com>
3811 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3812 - per-import/export recovery handling (958, 931, 959)
3813 - multiple last-rcvd slots, for serving multiple FSes (949)
3814 - connections are again shared between multiple imp/exports (963, 964)
3815 - "umount -f" would hang if any requests needed to be sent (393, 978)
3816 - avoid pinning large req buffer by copying for queued messages (989)
3817 - add "uuid" to "lctl device" command to help upcalls (991)
3818 - "open" RPCs with transnos would confuse recovery counters (1037)
3819 - do proper endian conversion of all wire messages (288, 340, 891)
3820 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3821 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3822 - fix ptlrpc_connection leak in target_handle_connect (1174)
3823 - fix import refcounting bug in OST and MDS cleanup (1134)
3824 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3825 - fix very unlikely obd_types race condition (501)
3826 - remove osc_open hack for echo_client (1187)
3827 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3828 - a failure in read_inode2 leads to deadlock (1139)
3829 - cancel ack-locks as soon as transaction is committed (1072)
3830 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3831 - make sure to commitrw after any preprw to avoid deadlock (1162)
3832 - failing to execute a file in a lustre FS would lock inode (1203)
3833 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3834 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3835 - obd_uuid2tgt would walk off the end of the list (1255)
3836 - on IA64 the osc would give portals incorrect bulk size (1258)
3837 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3838 - fix lock inversion caused by new llite matching code (1282)
3839 - limit the number of dirty pages on a client to 10MB (1286)
3840 - timed out locks were not being corrected cancelled (1289)
3841 - fix O_DIRECT above 4GB on IA-32 (1292)
3842 * major user-visible changes
3843 - fail out/fail over policy now controlled by the upcall (993)
3845 - add OBD_PING to check server availability and failure (954)
3846 - lustre messages are now sent in sending host order (288, 340, 891)
3847 - add eadatalen to MDS getattr reply (340)
3848 - OST read replies may contain second buffer, with per-page status (593)
3850 2003-03-11 Phil Schwan <phil@clusterfs.com>
3853 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3854 - failed open()s could cause deadlock; fixed (867, 869)
3855 - stop cancelling OST locks when files are closed (481)
3856 - overlapping XID spaces caused network corruption (851, 853)
3857 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3858 - setattr_raw added, to do single-RPC, server-side setattrs
3859 - lmc/lconf syntax change for OST UUIDs
3860 - fix crashy race condition between ptlrpc_free_req and osc_close
3861 - don't use request in mdc_enqueue if we hit a timeout (889)
3862 - don't set the inode i_size for regular files from the MDS (896)
3863 - handle out of order completion AST (842)
3864 - don't LBUG if a lock request times out after receiving AST (913)
3865 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3866 - if a bad lock AST arrives, send an error instead of dropping entirely
3867 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3868 - fix leak in bulk IO when only partially completed (899, 900, 926)
3869 - fix O_DIRECT for ia64 (55)
3870 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3871 - C-z after timeout could hang a process forever; fixed (977)
3873 - client-side I/O cache (678, 924, 929, 941, 970)
3875 - READPAGE and SETATTRs which don't take server-side locks get
3878 2003-02-11 Phil Schwan <phil@clusterfs.com>
3881 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3882 - Fix fsfilt_extN_readpage() to read a full page of directory
3883 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3884 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3885 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3886 - Most intent operations were removed, in favour of a new RPC mode
3887 that does a single RPC to the server and bypasses most of the VFS
3888 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3889 - Aggressively cancel local locks on DLM servers
3890 - mds_reint_unlink sends EA to the client if it's the last nlink.
3891 client uses that EA to unlink OST objects.
3892 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3893 - recursive symlinks were fixed (439)
3894 - fixed NULL deref in DEBUG_REQ
3895 - filter_update_lastobjid no longer calls sync, which annoyed extN
3896 - fixed multi-client small-writes to a single file problem (445)
3897 - fixed mtime updates during file writes (607)
3898 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3899 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3900 - fixed Linux OST size reporting problem (444, 656)
3901 - OST now updates object mtime with writes or setattr (607, 619)
3902 - client verifies file size before zeroing page past EOF (445)
3903 - OST now writes last allocated objid to disk with allocation (108)
3904 - LOV on echo now works (409)
3906 - mds_reint_unlink sends a new buffer, with the EA included. this
3907 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3908 set if a regular file was being unlinked, and it was the last link
3909 - use PtlGet from the target for bulk writes (315)
3910 - OST now updates object mtime with writes or setattr (607, 619)
3911 - LDLM now has a grant-time callback to revalidate locked items, if
3913 - Many MDS operations were reorganized to combat race conditions
3915 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3916 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3918 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3921 - Fully reactivate OST imports after reconnection (512, others)
3922 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3923 - More graceful error handling for truncating on dead OST (515)
3924 - Don't error out unless we're actually accessing dead stripes (474)
3925 - Fix garbage sizes when stripes are missing (410)
3926 - LRU counters were broken, causing constant lock purge (433, 432)
3927 - garbage on read from stripes with failed OSTs (441)
3928 - mark OSCs as active before reconnecting during recovery (438)
3929 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3930 - lfind did not preserve OST order in output (443)
3931 - symlinks cause hung clients, incorrect data (439)
3932 - stop dereferencing request after dropping refcount (457)
3933 - don't LASSERT(spin_is_locked) on non-SMP (455)
3934 - fixes for many rename() bugs
3935 - fstat didn't correctly synchronize attributes (399)
3936 - server must handle lock cancellation during blocking AST prep (487)
3937 - bulk descriptors were free()d too soon (511)
3938 - fix paths in lconf, which would load incorrect modules (451, 507)
3939 - fix confusing lconf 'host not found' error message (386)
3940 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3941 - fix race condition in mdc_blocking_ast() for inode access (526)
3942 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3943 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3944 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3945 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3946 - fix lstripe in several ways (kernel side) (527)
3947 - fix request leak in ldlm_cli_enqueue (262)
3948 - incorrect OSC was marked inactive after OST failure
3949 - call mds_fs_cleanup before unmounting filesystem (524)
3950 - fix races between taking ns_lock and ldlm_lock_change_resource
3951 - fix races updating LOV export open file list
3952 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3953 - fix recovery NULL deref in ldlm_cli_cancel_unused
3954 - fix some DLM races by using new hash table for lock handles (419)
3955 - permit the client to specify desired inodes, at replay
3956 - duplicate requests when we queue them for replay reintegration
3957 - fix last_rcvd offset calculation
3958 - sync after each recovered transaction, so we always make progress
3959 - never, not always, ERESTART requests without transnos
3960 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3962 - skip replay if the MDS says that the client is already connected
3963 - don't check for a recovery-enabled export to match lctl's UUID
3964 - don't INC_USE_COUNT for phantom exports
3965 - don't crash when cleaning up phantom exports (567)
3966 - don't double-finish or set replay data for errored mdc_open requests
3967 - abort requests when they time out, so we don't get old replies
3968 - send/receive replies for AST messages again
3969 - if the client says that it doesn't have the lock, cancel it on the
3971 - if we timeout during I/O, don't try to cancel an in-use lock; instead
3972 mark it as destroyed, it will all work out when decref is called
3973 - fix module use counts (22, 581)
3975 - ASTs now expect a reply (server cancels lock on error reply)
3977 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
3980 - fix many simultaneous client startup (392)
3981 - fix dentry->d_it clobbering
3982 - credentials weren't being shipped for readdir/getattr operations
3983 - remove invalid assertions triggered during some concurrent MD
3985 - proper Lustre versions added (336, 389)
3986 - fix memory leak for create error case (398)
3987 - fix LOV locking bug that would get cli/srv out of sync
3988 - fix echo client over LOV (409)
3989 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
3990 - fix double-O_EXCL intent crash (424)
3991 - avoid sending multiple lock CANCELs (352)
3993 - MDS can do multi-client recovery (modulo bugs in new code)
3995 - many updates, edits, cleanups
3997 2002-11-18 Phil Schwan <phil@clusterfs.com>
4000 - fix null d_it dereference (346)
4001 - fix full OST/dbench hang (333)
4002 - fix permission problem with file removal (286)
4003 - fix removal of OSCs from LOV when they fail
4004 - fix NULL deref during bulk timeout (214)
4005 - fix problems related to multiple filesystems on one MDS (241)
4006 - fixed serious subtle metadata locking bugs
4007 - free locks on clients when inodes are removed due to memory
4009 - fix inode pointer in lock data (285)
4010 - partial support for multiple MDS on a single host (241)
4011 - data locks weren't cancelled at clear_inode time (290, 311)
4012 - intent locks could lead to unbounded lock growth (205)
4013 - added a maximum lock count, an LRU list, and a flusher
4014 - fix multiple rename (365)
4015 - properly abstracted the echo client
4016 - OSC locked 1 byte too many; fixed
4017 - rewrote brw callback code:
4018 - fixed recovery bugs related to LOVs (306)
4019 - fixed too-many-pages-in-one-write crash (191)
4020 - fixed (again) crash in sync_io_timeout (214)
4021 - probably fixed callback-related race (385)
4023 - Add capability to MDS protocol
4024 - LDLM cancellations and callbacks on different portals
4026 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4029 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4030 - defer open object destruction to close time (601981, 138)
4031 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4032 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4033 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4034 - don't call obd_setattr in ll_file_release for destroyed objects
4035 * protocol change to lustre_msg: move |version| and add |flags|
4036 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4037 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4038 * added replay of create, unlink, link and rename operations during
4039 MDS failover; recovery should be much more robust now
4040 * remove failed OSCs from LOVs (only lov_create uses this so far)
4041 * the lustre-HOWTO was brought (more) up to date (582544)
4043 2002-10-23 Phil Schwan <phil@clusterfs.com>
4046 - in-use dentries weren't being reused properly (617851)
4047 - prevent multiple LDLM setup (599178)
4048 - fix LOV size calculations for truncate (617853)
4049 - fix client handling of MDS intent errors (POSIX)
4050 - fix permission bug in lovstripe.c test (624321)
4051 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4052 - truncate past end of file could corrupt data
4053 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4054 - a race in recovery could return ETIMEDOUT to apps (623947)
4055 - building outside the source directory was fixed
4056 * the lustre-HOWTO was brought (more) up to date (582544)
4057 * major progress was made on recovery functionality
4059 2002-10-10 Phil Schwan <phil@clusterfs.com>
4062 - recovery deadlock fix
4063 - rm -rf causes LBUG fix (617817)
4064 - file open by multiple tasks fix (618962)
4065 - directory permissions bugs (602707 and 620007)
4066 - journal_stop fixed with locking (611313)
4067 - O_APPEND failures resolved (618273, perhaps 614459)
4068 - lconf PATH fix (619770)
4069 - IA64 build fix (621450)
4070 - RPC buffer sizes scale with amount of memory
4072 2002-10-01 Phil Schwan <phil@clusterfs.com>
4075 - locks would be cancelled without throwing away data pages,
4076 resulting in inconsistent data (605627)
4077 - inode attributes were not always being refreshed (605627, 612449)
4078 - lconf now continues to cleanup after lctl reports an error
4079 - MDS now enforces user permissions (602707)
4080 - lprocfs cleanup fixed, but not yet enabled (614157)
4081 - fixed infinite server hang, should a client not respond to an AST
4082 - avoid going into recovery if user calls readlink() with a buffer
4083 that's too small (613941)
4084 - AST RPCs no longer require replies (614867) -- this may be changed
4085 - don't crash server if client sends an IOV that's too big (611336)
4086 - fixed lock conversion deadlock (611892)
4087 - fixed the following of symlinks (614622)
4088 * recovery: the server can remove locks from a client that dies, other
4089 clients can make progress
4090 * more extN patch fixes
4091 * compile-time configurable ptlrpc buffer allocations
4093 - collaborative read cache document
4094 - Lustre Lite Performance CDR document-in-progress
4096 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4099 - fix typo in patch-2.4.18
4101 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4104 - clear ptlrpc request each time in handle_incoming_request()
4105 - unlink of files now destroys the object on the OST
4107 2002-09-19 Peter Braam <braam@clusterfs.com>
4109 * add hard link support
4110 * change obdfile creation method
4111 * kernel patch changed
4113 2002-09-19 Peter Braam <braam@clusterfs.com>
4116 - stack overflow bug in extN fixed
4118 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4120 * documentation updates
4121 - add man pages for config tools
4122 - update tests/README to describe testing with new config tools
4123 - finish metadata API descriptions
4124 * bug fixes and cleanups
4125 - statfs workaround for 16TB limit
4126 - LOV stripe allocation improved, can stripe on subset of OSTs
4127 - LOV file size/IO offset was wrong for files > 4GB in size
4128 - object EA data was being dropped, caused files to be unreadable
4129 - memory overflow with non-LOV OST caused memory corruption
4130 - fixed regression tests to work with new config tools, obdfilter
4131 - fixed bug when directory size became larger than 1 block
4132 - fixed bug (for single client case) when PWD was deleted
4133 - invalidate local directory pages when doing intent-based ops
4134 - avoid LDLM oops when lock callback contained bad data
4136 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4138 * documentation updates
4139 * bug fixes and cleanups
4140 - configuration tools
4143 - 64-bit compile warnings
4144 - 64-bit internal statfs data
4146 * test_brw on persistent OST devices
4148 * lprocfs (disabled)
4150 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4152 * documentation updates
4153 * bug fixes and cleanups
4154 * configuration tools
4156 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4159 * many small fixes to 0_5_4
4160 * io/network handling
4161 * thinkos in MDS operations
4163 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4166 * crucial basic fixes to 0.5.3
4167 * IOR, Iozone work over Elan
4170 2002-08-07 Phil Schwan <phil@clusterfs.com>
4171 * version 0_5_3, our first alpha
4172 * we use the new Portals iovs
4173 * documentation updates
4174 * bug fixes and cleanups
4175 * small changes in the DLM wire protocol
4177 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4178 * version 0_5_1 with some initial stability,
4179 * locking on MD and file I/O.
4180 * documentation updates
4181 * several bug fixes since 0.5.0
4182 * small changes in wire protocol
4184 2002-07-18 Phil Schwan <phil@clusterfs.com>
4186 * delivered as Lustre Light Alpha
4187 * fixed a crash after handling invalid MDS requests
4188 * fixed directory pages for architectures with non-4k pages sizes
4190 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4191 * release version v0_4_4
4192 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4194 * Fixes a number of interruption problems with OST operations.
4195 * Update documentation for portals header changes
4196 * Move all wire protocol structs/defines to lustre_idl.h
4197 * Fixes symlink length bug.
4198 * Add tcpdump to repository.
4200 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4201 * release version v0_4_3
4202 * Fixes statfs for inodes on extN.
4203 * Fixes bug in runtests which would delete /etc/hosts.
4204 * Use 64-bit object IDs wherever possible (not into VFS though)
4205 Remove ost_get_info, which is unused by lustre, and out of date.
4207 2002-07-03 Peter Braam <braam@clusterfs.com>
4208 * release version v0_4_2 Fixes a lookup error (type not passed)
4209 * move forward to head of Portals
4210 * move forward to latest Lustre kernel
4212 2002-06-25 Peter Braam <braam@clusterfs.com>
4213 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob