3 2.6.5-7.287.3 (SLES 9),
4 2.6.9-55.0.12.EL (RHEL 4),
5 2.6.16.53-0.16 (SLES 10),
6 2.6.18-8.1.15.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-cfs4
13 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
15 Severity : enhancement
17 Description: make lustre randomly fail allocating memory
18 Details : Make lustre randomly failed allocating memory for testing purpose.
20 Severity : enhancement
22 Description: lost problems with lov objid file
23 Details : Fixes some scability and access to not inited memory problems
24 in work with lov objdid file.
29 Description: lfs find does not continue on file error
30 Details : Continue other files processing when a file/dir is absent.
34 Description: Inconsistent usage of lustre_pack_reply()
35 Details : Standardize the usage of lustre_pack_reply() such that it
36 always generate a CERROR on failure.
39 Frequency : occasional
41 Description: MDS hang or stay in waiting lock
42 Details : If client receive lock with CBPENDING flag ldlm need send lock
43 cancel as separate rpc, to avoid situation when cancel request
44 can't processed due all i/o threads stay in wait lock.
48 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
52 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
56 Description: Update to SLES9 kernel-2.6.5-7.287.3.
60 Description: Update to RHEL4 latest kernel.
62 Severity : enhancement
64 Description: Build SLES10 patchless client fails
65 Details : The configure was broken by run ./configure with
66 --with-linux-obj=.... argument for patchless client. When the
67 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
68 can't search header adequately. Use absolute path such as
69 -I($LINUX)/include instead.
73 Description: interrupt oig_wait produce painc on resend.
74 Details : brw_redo_request can be used for resend requests from ptlrpcd and
75 private set, and this produce situation when rq_ptlrpcd_data not
76 copyed to new allocated request and triggered LBUG on assert
77 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
78 ptlrpcd set if request is changed and can be safety changed to use
83 Description: LASSERT_{REQ,REP}SWAB macros are buggy
84 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
85 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
86 these macros and replace them with their difinitions instead.
90 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
91 Details : Update which_patch & target file for SLES10 latest kernel.
94 Description: Remove client patches from SLES 10 kernel.
95 Details : This causes SLES 10 clients to behave as patchless clients
96 even on a Lustre-patched (server) kernel.
98 Severity : enhancement
100 Description: self-adjustable client's lru lists
101 Details : use adaptive algorithm for managing client cached locks lru
102 lists according to current server load, other client's work
103 pattern, memory activities, etc. Both, server and client
104 side namespaces provide number of proc tunables for controlling
107 Severity : enhancement
109 Description: light-weight GSS support
110 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
111 and reduce performance overhead.
113 Severity : enhancement
115 Description: Linux keyring support
116 Details : Support using service of Linux keyring for Lustre GSS internal
117 context refresh/cache mechanism.
121 Description: Fix errors in lfs documentation
122 Details : Fixes man pages
126 Description: don't use GFP_* in generic Lustre code.
127 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
131 Description: obdclass is limited by single OBD_ALLOC(idarray)
132 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
136 Description: warnings with build patchless client with vanila 2.6.19 and up
137 Details : change old ctl_table style and replace ctl_table/ctl_table_header
138 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
142 Frequency : Only for SLES
143 Description: sanity_quota fail test_1
144 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
145 file, which makes TSTID[2] inunique.
149 Frequency : Only for SLES9
150 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
151 Details : Update target/ChangeLog/which_patch .
155 Frequency : Only for test_10 in sanity-quota.sh
156 Description: a bug in quota test code
157 Details : checking a test flag in a wrong place
161 Frequency : Only for test_13 in sanity-quota.sh
162 Description: a bug in quota test script
163 Details : shouldn't check overfull allocated quota
167 Description: jbd statistics
168 Details : Port older jbd statistics patch for sles10
170 Severity : enhancement
172 Description: Add printing inode info into message about error in writepage.
177 Description: reply_lock_interpret crash due to race with it and lock cancel.
178 Details : Do not replay locks that are being cancelled. Do not reference
179 locks by their address during replay, just by their handle.
183 Frequency : When flocks are used.
184 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
185 Details : Flock locks might destroy just granted lock if it could be merged
186 with another existing flock, this is done in completion handler,
187 so teach ldlm_cli_enquque_fini that this is a valid case for
193 Description: Oops in read and write path when failing to allocate lock.
194 Details : Check if lock allocation failed and return error back.
198 Description: lstripe command fails for valid OST index
199 Details : The stripe offset is compared to lov->desc.ld_tgt_count
200 instead of lov->desc.ld_active_tgt_count.
204 Description: sanity.sh failed test 103
205 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
210 Description: df doesn't work properly if diskfs blocksize != 4K
211 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
215 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
216 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
220 Description: obd_health_check_timeout too short
221 Details : set obd_health_check_timeout as 1.5x of obd_timeout
225 Description: llapi_file_create() does not allow some changes
226 Details : add llapi_file_open() that allows specifying the mode and
227 open flags, and also returns an open file handle.
231 Description: merge and cleanup kernel patches.
232 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
236 Description: Add journal checksum support.(Kernel part)
237 Details : The journal checksum feature adds two new flags i.e
238 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
239 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
240 indicates that the commit block contains the checksum for
241 the blocks described by the descriptor blocks. Now commit
242 record can be sent to disk without waiting for descriptor
243 blocks to be written to disk. This behavior is controlled
244 using JBD2_FEATURE_ASYNC_COMMIT flag.
248 Description: OSS needs mutliple precreate threads
249 Details : Add ability to start more than one create thread per OSS.
253 Description: Sanity.sh test_65k failed: setstripe should have succeeded
254 Details : execute lfs setstripe on client
258 Description: mds_obd_create error creating tmp object
259 Details : When the user sets quota on root, llog will be affected and can't
260 create files and write files.
263 Frequency : Always on ia64 patchless client, and possibly others.
265 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
266 Details : This allows the patchless client to be loaded on architectures
271 Description: RedHat Update kernel for RHEL5
272 Details : Add the kernel config file for RHEL5.
276 Description: RedHat Update kernel for RHEL5
277 Details : Modify the kernel config file more closer RHEL5.
281 Description: Build failure against Centos5 (RHEL5)
282 Details : Define PAGE_SIZE when it isn't present.
286 Description: test_7 on sanity-quota.sh can't be used on separate machines
287 Details : add facet to handle it
291 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
292 Details : with NFS, the anon dentry's parent was set to itself in
293 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
294 corresponding dentry other than use rec->ur_name.
296 Severity : enhancement
298 Description: lfs setstripe enhancement
299 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
303 Description: enable data checksumming by default
304 Details : enable checksum by default, allow --disable-checksum
305 configure option and "-o nochecksum" mount option. Checksums
306 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
310 Description: lustre support for RHEL5
311 Details : Add support for RHEL5.
315 Description: Client eviction due to failover config
316 Details : after a connection loss, the lustre client should attempt to
317 reconnect to the last active server first before trying the
318 other potential connections.
322 Description: when mds and osts use different quota unit(32bit and 64bit),
323 quota will be released repeatly.
324 Details : void sending multiple quota reqs to mds, which will keep the
325 status between the reqs.
329 Description: osts not allocated evenly to files
330 Details : change the condition to increase offset_idx
333 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
335 Description: Data corruption for OSTs that are formatted larger than 2TB
337 Details : When generating the bio request for lustre file writes the
338 sector number would overflow a temporary variable before being
339 used for the IO. The data reads correctly from Lustre (which
340 will overflow in a similar manner) but other file data or
341 filesystem metadata may be corrupted in some cases.
345 Description: Tune the kernel for good SCSI performance.
346 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
347 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
352 Description: rewrite ext2-derived code in obdclass/uuid.c
353 Details : rewrite inherited code (uuid parsing code from ext2 utils)
354 from scratch preserving functionality.
358 Description: "lfs find -obd UUID" prints directories
359 Details : "lfs find -obd UUID" will return all directory names instead
360 of just file names. It is incorrect because the directories
361 do not reside on the OSTs.
365 Description: use i_size_read and i_size_write in 2.6 port
366 Details : replace inode->i_size access with i_size_read/write()
369 Frequency : when removing large files
371 Description: scheduling issue during removal of large Lustre files
372 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
373 It causes scheduling issues when removing large files (17TB in the
377 Frequency : only with liblustre clients on XT3
379 Description: evictions taking too long
380 Details : allow llrd to evict clients directly on OSTs
383 Frequency : only on ppc
385 Description: /proc/fs/lustre/devices broken on ppc
386 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
387 We should make sure the type of 'index' is loff_t and then cast
388 explicitly as needed below. Do not assign an explicitly cast
392 Frequency : only for rhel5
394 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
395 Details : Modify the target file & which_kernel.
397 Severity : enhancement
399 Description: omit set fsid for export NFS
400 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
401 as FS_REQUIRES_DEV for avoid problems with generate fsid.
405 Frequency : Always, for kernels after 2.6.16
406 Description: Fix warning idr_remove called for id=.. which is not allocated.
407 Details : Last kernels save old s_dev before kill super and not allow
408 to restore from callback - restore it before call kill_anon_super.
412 Description: buffer overruns could theoretically occur
413 Details : llapi_semantic_traverse() modifies the "path" argument by
414 appending values to the end of the origin string, and a
415 overrun may occur. Adding buffer overrun check in liblustreapi.
419 Description: Fix error on 'ls .' at the top of the Lustre mount.
420 Details : Don't revalidate dentry if it is a root dentry.
424 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
425 Details : Modify vm-tunables-rhel4.patch.
429 Description: Kernel config for 2.6.18-vanilla.
430 Details : Modify targets/2.6-vanilla.target.in.
431 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
432 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
433 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
434 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
438 Description: improve handling recoverable errors
439 Details : if request processig with error which can be recoverable on server
440 request should be resend, otherwise page released from cache and
445 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
446 Details : Modify target file & which_patch.
447 A flaw was found in the IA32 system call emulation provided
448 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
449 value could be stored in the %RAX register, which could trigger an
450 out-of-bounds system call table access. An untrusted local user
451 could exploit this flaw to run code in the kernel
452 (ie a root privilege escalation). (CVE-2007-4573).
456 Description: O_DIRECT bypasses client statistics.
457 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
458 read_ahead_stats were updated. copied the stats section from
459 osc_send_oap_rpc() into async_internal().
463 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
467 Description: change order of libsysio includes
468 Details : '#include sysio.h' should always come before '#include xtio.h'
472 Description: handle "serious error: objid * already exists" more gracefully
473 Details : If LAST_ID value on disk is smaller than the objects existing in
474 the O/0/d* directories, it indicates disk corruption and causes an
475 LBUG(). If the object is 0-length, then we should use the existing
476 object. This will help to avoid a full fsck in most cases.
478 Severity : enhancement
480 Description: adapt the lustre_config script to support the upgrade case
481 Details : Add "-u" option for lustre_config script to support upgrading 1.4
482 server targets to 1.6 in parallel.
486 Description: To avoid grant space > avaible space when the disk is almost
487 full. Without this patch you might see the error "grant XXXX >
488 available" or some LBUG about grant, when the disk is almost
490 Details : In filter_check_grant, for non_grant cache write, we should
491 check the left space by if (*left > ungranted + bytes), instead
492 of (*left > ungranted), because only we are sure the left space
493 is enough for another "bytes", then the ungrant space should be
494 increase. In client, we should update cl_avail_grant only there
495 is OBD_MD_FLGRANT in the reply.
499 Description: Update RHEL 4 kernel to fix local root privilege escalation.
500 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
501 described in CVE-2007-4573. This problem could allow untrusted
502 local users to gain root access.
505 Frequency : when using O_DIRECT and quotas
507 Description: Incorrect file ownership on O_DIRECT output files
508 Details : block usage reported by 'lfs quota' does not take into account
509 files that have been written with O_DIRECT.
514 Description: touch file failed when fs is not full
515 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
516 otherwise we can get ENOSP while fs is not full.
520 Description: parallel lock callbacks
521 Details : Instead of sending blocking and completion callbacks as separated
522 requests, adding them to a set and sending in parallel.
525 Frequency : only for Cray XT3
526 Bugzilla : 12829/13455
527 Description: Changing primary group doesn't change the group lustre assigns to
529 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
530 overridden with the primary group provided by the group upcall,
531 whereas the supplied fsgid can be trusted if it is in the list of
532 supplementary groups returned by the group upcall.
534 --------------------------------------------------------------------------------
536 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
538 * Support for kernels:
539 2.6.5-7.283 (SLES 9),
540 2.6.9-55.EL (RHEL 4),
541 2.6.16.46-0.14 (SLES 10),
542 2.6.18.8 vanilla (kernel.org)
543 * Client support for unpatched kernels:
544 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
545 2.6.16 - 2.6.22 vanilla (kernel.org)
546 * Due to recently discovered recovery problems, we do not recommend
547 using patchless RHEL 4 clients with this or any earlier release.
548 * Recommended e2fsprogs version: 1.39.cfs8
549 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
550 * Starting with this release, the ldiskfs backing filesystem required
551 by Lustre is now in its own package, lustre-ldiskfs. This package
552 should be installed. It is versioned separately from Lustre and
553 may be released separately in future.
557 Description: block reactivating mgc import until all deactivates complete
558 Details : Fix race when failing back MDT/MGS to itself (testing)
561 Frequency : at statup only
563 Description: mds_lov_synchronize race leads to various problems
564 Details : simultaneous MDT->OST connections at startup can cause the
565 sync to abort, leaving the OSC in a bad state.
567 Severity : enhancement
569 Description: add optional extra BUILD_VERSION info
570 Details : add a new environment variable (namely LUSTRE_VERS) which allows
571 to override the lustre version.
574 Frequency : 2.6.18 servers only
576 Description: ll_kern_mount() doesn't release the module reference
577 Details : The ldiskfs module reference count never drops down to 0
578 because ll_kern_mount() doesn't release the module reference.
583 Description: server LBUG when using old ost_num_threads parameter
584 Details : Accept the old ost_num_threads parameter but warn that it
585 is deprecated, and fix an off-by-one error that caused an LBUG.
590 Description: Transient SCSI error results in persistent IO issue
591 Details : iobuf->dr_error is not reinitialized to 0 between two
595 Frequency : sometimes when underlying device returns I/O errors
597 Description: OSTs not going read-only during write failures
598 Details : OSTs are not remounted read-only when the journal commit threads
599 get I/O errors because fsfilt_ext3 calls journal_start/stop()
600 instead of the ext3 wrappers.
604 Description: poor connect scaling with increasing client count
605 Details : Don't run filter_grant_sanity_check for more than 100 exports
606 to improve scaling for large numbers of clients.
609 Frequency : SLES10 only
611 Description: sanity-quota.sh quotacheck failed: rc = -22
612 Details : Quotas cannot be enabled on SLES10.
615 Frequency : liblustre clients only
617 Description: getdirentries does not give error when run on compute nodes
618 Details : getdirentries does not fail when the size specified as an argument
619 is too small to contain at least one entry
621 Severity : enhancement
623 Description: Add LNET router traceability for debug purposes
624 Details : If a checksum failure occurs with a router as part of the
625 IO path, the NID of the last router that forwarded the bulk data
626 is printed so it can be identified.
631 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
632 Details : Due to a race condition, liblustre clients were occasionally
635 Severity : enhancement
637 Description: lfs setstripe use optional parameters instead of postional
640 Severity : enhancement
642 Description: Nanosecond timestamp support for ldiskfs
643 Details : The on-disk ldiskfs filesystem has added support for nanosecond
644 resolution timestamps. There is not yet support for this at
645 the Lustre filesystem level.
648 Frequency : during server recovery
650 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
651 Details : request with rq_no_resend flag not awake l_wait_event if they get a
655 Frequency : nfs export on patchless client
657 Description: connectathon hang when test nfs export over patchless client
658 Details : Disconnected dentry cannot be found with lookup, so we do not need
659 to unhash it or make it invalid
662 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
663 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
664 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
669 Description: open req refcounting wrong on reconnect
670 Details : If reconnect happened between getting open reply from server and
671 call to mdc_set_replay_data in ll_file_open, we will schedule
672 replay for unreferenced request that we are about to free.
673 Subsequent close will crash in variety of ways.
674 Check that request is still eligible for replay in
675 mdc_set_replay_data().
680 Description: disable writes to filesystem when reading health_check file
681 Details : the default for reading the health_check proc file has changed
682 to NOT do a journal transaction and write to disk, because this
683 can cause reads of the /proc file to hang and block HA state
684 checking on a healthy but otherwise heavily loaded system. It
685 is possible to return to the previous behaviour during configure
686 with --enable-health-write.
688 Severity : enhancement
690 Description: 64-bit inode version
691 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
692 to the inode. This will be required for version-based recovery.
697 Description: MDS fails to start if a duplicate client export is detected
698 Details : in some rare cases it was possible for a client to connect to
699 an MDS multiple times. Upon recovery the MDS would detect this
700 and fail during startup. Handle this more gracefully.
702 Severity : enhancement
704 Description: Add -o localflock option to simulate old noflock
706 Details : This will achieve local-only flock/fcntl locks
712 Description: log_commit_thread vs filter_destroy race leads to crash
713 Details : Take import reference before releasing llog record semaphore
718 Description: Wrong request locking in request set processing
719 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
720 imp_delayed_list, in this place should be used imp_lock.
723 Frequency : when reconnection
725 Description: Grant Leak when osc reconnect to OST
726 Details : When osc reconnect ost, OST(filter) should check whether it
727 should grant more space to client by comparing fed_grant and
728 cl_avail_grant, and return the granted space to client instead
729 of "new granted" space, because client will call osc_init_grant
730 to update the client grant space info.
733 Frequency : when client reconnect to OST
735 Description: Grant Leak when osc do resend and replay bulk write
736 Details : When osc reconnect to OST, OST(filter)should clear grant info of
737 bulk write request, because the grant info while be sync between
738 OSC and OST when reconnect, and we should ignore the grant info
739 these of resend/replay write req.
744 Description: Grant space more than avaiable left space sometimes.
745 Details : When then OST is about to be full, if two bulk writing from
746 different clients came to OST. Accord the avaliable space of the
747 OST, the first req should be permitted, and the second one
748 should be denied by ENOSPC. But if the seconde arrived before
749 the first one is commited. The OST might wrongly permit second
750 writing, which will cause grant space > avaiable space.
753 Frequency : when client is evicted
755 Description: Grant might be wrongly erased when osc is evicted by OST
756 Details : when the import is evicted by server, it will fork another
757 thread ptlrpc_invalidate_import_thread to invalidate the
758 import, where the grant will be set to 0. While the original
759 thread will update the grant it got when connecting. So if
760 the former happened latter, the grant will be wrongly errased
761 because of this race.
766 Description: Checking Stale with correct fid
767 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
768 to check whether it is stale, because sometimes, we want the
769 enqueue happened anyway, and op_data.fid2 will not be initialized.
771 Severity : enhancement
773 Description: update patchless client
774 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
777 Frequency : only with 2.4 kernel
779 Description: random memory corruption
780 Details : size of struct ll_inode_info is to big for union inode.u and this
781 can be cause of random memory corruption.
786 Description: Memory leak in recovery
787 Details : Lov_mds_md was not free in an error handler in mds_create_object.
788 It should also check obd_fail before fsfilt_start, otherwise if
789 fsfilt_start return -EROFS,(failover mds during mds recovery).
790 then the req will return with repmsg->transno = 0 and rc = EROFS.
791 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
792 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
793 be freed no matter whether fsfilt_commit success or not.
796 Frequency : only with huge count clients
798 Description: Prevents from taking the superblock lock in llap_from_page for
800 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
801 for llap_from_page call in ll_removepage prevents from taking the
802 superblock lock for a soon died page.
807 Description: Not check open intent error before release open handle
808 Details : in some rare cases, the open intent error is not checked before
809 release open handle, which may cause
810 ASSERTION(open_req->rq_transno != 0), because it tries to release
811 the failed open handle.
816 Description: Set cat log bitmap only after create log success.
817 Details : in some rare cases, the cat log bitmap is set too early. and it
818 should be set only after create log success.
822 Description: Accessing a block bevice can re-enable I/O when Lustre is
823 tearing down a device.
824 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
829 Description: service threads may hog cpus when there are a lot of requests
830 Details : Insert cond_resched to give other threads a chance to use some CPU
835 Description: the cat log was not initialized in recovery
836 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
837 unlink log on mds will not be initialized until mds post
838 recovery. And also in mds post recovery, the unlink log will
839 initialization will be done asynchronausly, so there will be race
840 between add unlink log and unlink log initialization.
844 Description: brw_stats were being printed incorrectly
845 Details : brw_stats were being printed as log2 but all of them were not
846 recorded as log2. Also remove some code duplication arising from
847 filter_tally_{read,write}.
851 Frequency : rare, only in recovery.
852 Description: ASSERTION(req->rq_type != LI_POISON) failed
853 Details : imp_lock should be held while iterating over imp_sending_list for
854 prevent destroy request after get timeout in ptlrpc_queue_wait.
858 Description: replay-single.sh test 52 fails
859 Details : A lock's skiplist need to be cleanup when it being unlinked
860 from its resource list.
864 Description: Short directio read returns full requested size rather than
866 Details : Direct I/O operations should return actual amount of bytes
867 transferred rather than requested size.
869 Severity : enhancement
871 Description: metadata RPC reduction (e.g. for rm performance)
872 Details : decrease the amount of synchronous RPC between clients and servers
873 by canceling conflicing lock before the operation on the client
874 and packing thier handles into the main operation RPC to server.
876 Severity : enhancement
878 Description: Async OSC create to avoid the blocking unnecessarily.
879 Details : If a OST has no remain object, system will block on the creating
880 when need to create a new object on this OST. Now, ways use
881 pre-created objects when available, instead of blocking on an
882 empty osc while others are not empty. If we must block, we block
883 for the shortest possible period of time.
885 Severity : enhancement
887 Description: refine locking for avoid write wrong info into lov_objid file
888 Details : fix possible races with add new target and write/update data in
891 --------------------------------------------------------------------------------
893 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
898 Frequency : on some architectures
900 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
901 Details : Uninitialized flags sometimes cause configuration commands to
905 Frequency : patchless clients only
907 Description: missing __iget() symbol export
908 Details : The __iget() symbol export is missing. To avoid the need for
909 this on patchless clients the deathrow inode reaper is turned
910 off, and we depend on the VM to clean up old inodes. This
911 dependency was during via the fix for bug 12181.
913 --------------------------------------------------------------------------------
915 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
917 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
918 INTEROPERATE with older versions automatically. In many cases a
919 special upgrade step is needed. Please read the
920 user documentation before upgrading any part of a live system.
921 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
922 version will not interoperate with 1.6 betas before beta5 (1.5.95).
923 * WARNING: Lustre configuration and startup changes are required with
924 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
930 Severity : enhancement
932 Description: MountConf
933 Details : Lustre configuration is now managed via mkfs and mount
934 commands instead of lmc and lconf. New obd types (MGS, MGC)
935 are added for dynamic configuration management. See
936 https://mail.clusterfs.com/wikis/lustre/MountConf for
939 Severity : enhancement
941 Description: dynamic OST addition
942 Details : OSTs can now be added to a live filesystem
944 Severity : enhancement
946 Description: startup order invariance
947 Details : MDTs and OSTs can be started in any order. Clients only
948 require the MDT to complete startup.
950 Severity : enhancement
952 Description: parallel, asynchronous orphan cleanup
953 Details : orphan cleanup is now performed in separate threads for each
954 OST, allowing parallel non-blocking operation.
956 Severity : enhancement
958 Description: optimized stripe assignment
959 Details : stripe assignments are now made based on ost space available,
960 ost previous usage, and OSS previous usage, in order to try
961 to optimize storage space and networking resources.
963 Severity : enhancement
965 Description: Permanently set tunables
966 Details : All writable /proc/fs/lustre tunables can now be permanently
967 set on a per-server basis, at mkfs time or on a live system.
969 Severity : enhancement
971 Description: Lustre message v2
972 Details : Add lustre message format v2.
974 Severity : enhancement
976 Description: client OST exclusion list
977 Details : Clients can be started with a list of OSTs that should be
978 declared "inactive" for known non-responsive OSTs.
982 Description: SPEC SFS validation failure on NFS v2 over lustre.
983 Details : Changes the blocksize for regular files to be 2x RPC size,
984 and not depend on stripe size.
986 Severity : enhancement
988 Description: Multiple MD RPCs in flight.
989 Details : Further unserialise some read-only MDS RPCs - learn about intents.
990 To avoid overly-overloading MDS, introduce a limit on number of
991 MDS RPCs in flight for a single client and add /proc controls
992 to adjust this limit.
994 Severity : enhancement
996 Description: client read/write statistics
997 Details : Add client read/write call usage stats for performance
998 analysis of user processes.
999 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1000 file access. extents_stats shows chunk size distribution.
1001 extents_stats_per_process show chunk size distribution per
1004 Severity : enhancement
1006 Description: mds statistics
1007 Details : Add detailed mds operations statistics in
1008 /proc/fs/lustre/mds/*/stats.
1012 Description: Failure of copying files with lustre special EAs.
1013 Details : Client side always return success for setxattr call for lustre
1014 special xattr (currently only "trusted.lov").
1019 Description: Refcount LNET uuids
1020 Details : The global LNET uuid list grew linearly with every startup;
1021 refcount repeated list entries instead of always adding to
1024 Severity : enhancement
1026 Description: Dynamic service threads
1027 Details : Within a small range, start extra service threads
1028 automatically when the request queue builds up.
1031 Frequency : mixed-endian client/server environments
1033 Description: mixed-endian crashes
1034 Details : The new msg_v2 system had some failures in mixed-endian
1037 Severity : enhancement
1039 Description: Easy OST removal
1040 Details : OSTs can be permanently deactivated with e.g. 'lctl
1041 conf_param lustre-OST0001.osc.active=0'
1043 Severity : enhancement
1045 Description: MGS proc entries
1046 Details : Added basic proc entries for the MGS showing what filesystems
1049 Severity : enhancement
1051 Description: provide MGS failover
1052 Details : Added config lock reacquisition after MGS server failover.
1054 Severity : enhancement
1056 Description: add Linux 2.4 support
1057 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1061 Description: a large application tries to do I/O to the same resource and dies
1062 in the middle of it.
1063 Details : Check the req->rq_arrival time after the call to
1064 ost_brw_lock_get(), but before we do anything about
1065 processing it & sending the BULK transfer request. This
1066 should help move old stale pending locks off the queue as
1067 quickly as obd_timeout.
1070 Frequency : when an incorrect nid is specified during startup
1072 Description: ptlrpc connect to non-existant node causes kernel crash
1073 Details : LNET can't be re-entered from an event callback, which
1074 happened when we expire a message after the export has been
1075 cleaned up. Instead, hand the zombie cleanup off to another
1078 Severity : enhancement
1080 Description: plain/inodebits lock performance improvement
1081 Details : Grouping plain/inodebits in granted list by their request modes
1082 and bits policy, thus improving the performance of search through
1086 Frequency : only if OST filesystem is corrupted
1088 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1089 Details : for a short time RPCs with bulk IO are in the replay list,
1090 but replay of bulk IOs is unimplemented. If the OST filesystem
1091 is corrupted due to disk cache incoherency and then replay is
1092 started it is possible to trip an assertion. Avoid putting
1093 committed RPCs into the replay list at all to avoid this issue.
1096 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1098 Description: System hang on startup
1099 Details : This bug allowed the liblustre (e.g. catamount) client to
1100 return to the app before handling all startup RPCs. This
1101 could leave the node unresponsive to lustre network traffic
1102 and manifested as a server ptllnd timeout.
1104 Severity : enhancement
1106 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1107 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1108 boolean to control whether to print peer debug info when a
1109 client's RPC times out.
1112 Frequency : only for kernels with patches from Lustre below 1.4.3
1114 Description: Remove old rdonly API
1115 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1118 Frequency : only for devices with external journals
1120 Description: Set external device read-only also
1121 Details : During a commanded failover stop, we set the disk device
1122 read-only while the server shuts down. We now also set any
1123 external journal device read-only at the same time.
1126 Frequency : when upgrading from 1.4 while trying to change parameters
1128 Description: The wrong (new) MDC name was used when setting parameters for
1129 upgraded MDT's. Also allows changing of OSC (and MDC)
1130 parameters if --writeconf is specified at tunefs upgrade time.
1133 Frequency : when setting specific ost indicies
1135 Description: QOS code breaks on skipped indicies
1136 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1137 created with --index need not be sequential.
1139 Severity : enhancement
1141 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1142 Details : The uninit_groups feature works in conjunction with the kernel
1143 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1144 up the pass1 processing of e2fsck. This is a read-only feature
1145 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1146 mount filesystems that have had this feature enabled.
1148 Severity : enhancement
1150 Description: Improve multi-block allocation algorithm to avoid fragmentation
1151 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1152 allocation locality and avoid filesystem fragmentation.
1154 ------------------------------------------------------------------------------
1156 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1158 * Support for kernels:
1159 2.4.21-47.0.1.EL (RHEL 3)
1160 2.6.5-7.283 (SLES 9)
1161 2.6.9-42.0.10.EL (RHEL 4)
1162 2.6.12.6 vanilla (kernel.org)
1163 2.6.16.27-0.9 (SLES 10)
1164 * Recommended e2fsprogs version: 1.39.cfs5
1166 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1170 Frequency : occasional, depends on client load and configuration
1171 Bugzilla : 12181, 12203
1172 Description: data loss for recently-modified files
1174 Details : In some cases it is possible that recently written or created
1175 files may not be written to disk in a timely manner (this should
1176 normally be within 30s unless client IO load is very high).
1177 The problem appears as zero-length files or files that are a
1178 multiple of 1MB in size after a client crash or client eviction
1179 that are missing data at the end of the file.
1181 This problem is more likely to be hit on clients where files are
1182 repeatedly created and unlinked in the same directory, clients
1183 have a large amount of RAM, have many CPUs, the filesystem has
1184 many OSTs, the clients are rebooted frequently, and/or the files
1185 are not accessed by other nodes after being written.
1187 The presence of the problem can be detected by looking at
1188 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1189 smaller than the second (nr_unused) then dirty files will not
1190 be flushed automatically to disk. "sync; sleep 10" should be
1191 run several times on the node before unmounting it to update
1192 Lustre (this is also safe to run on nodes without this problem).
1194 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1195 that can cause this same problem, so customers using that kernel
1196 also need to update the kernel in addition to Lustre. In order
1197 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1199 It is normal that files written just before a client crash (less
1200 than 30s) may not yet have been flushed to disk, even for local
1204 Frequency : frequent on thin XT3 nodes
1206 Description: UUID collision on thin XT3 Linux nodes
1207 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1208 randomly, since we relied on an insufficiently-seeded PRNG.
1213 Description: OSS hangs after "All ost request buffers busy"
1214 Details : A deadlock between quota and journal operations caused OSS
1215 hangs after printing "All ost request buffers busy."
1218 Frequency : always on liblustre builds
1220 Description: Cleanup compiler warnings on liblustre
1223 Frequency : always on liblustre builds on XT3
1225 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1229 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1230 > 32000 subdirectories
1231 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1232 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1233 subdirectory count crosses 32000. This will aid e2fsck to
1234 correctly handle more than 32000 subdirectories.
1237 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1239 Description: System hang on startup
1240 Details : This bug allowed the liblustre (e.g. catamount) client to
1241 return to the app before handling all startup RPCs. This
1242 could leave the node unresponsive to lustre network traffic
1243 and manifested as a server ptllnd timeout.
1245 Severity : enhancement
1247 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1248 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1249 boolean to control whether to print peer debug info when a
1250 client's RPC times out.
1255 Description: make O_SYNC working on 2.6 kernels
1256 Details : 2.6 kernels use different method for mark pages for write,
1257 so need add a code to lustre for O_SYNC work.
1262 Description: Failure to close file and release space on NFS
1263 Details : Put inode details into lock acquired in ll_intent_file_open.
1264 Use mdc_intent_lock in ll_intent_open to properly
1265 detect all kind of errors unhandled by mdc_enqueue.
1270 Description: proc file read during shutdown sometimes raced obd removal,
1272 Details : Add lock to prevent obd access after proc file removal.
1275 Frequency : Only for files larger than 4GB on 32-bit clients.
1277 Description: improperly doing page alignment of locks
1278 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1279 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1284 Description: RPCs being resent when they shouldn't be
1285 Details : Some RPCs that should not be resent are being resent. This
1286 can cause inconsistencies in the RPC state machine. Do not
1287 resend such requests.
1290 Frequency : rare, only with NFS export
1292 Description: Crash on NFS re-export node
1293 Details : under very unusual load conditions an assertion is hit in
1294 ll_intent_file_open()
1297 Frequency : only if OST filesystem is corrupted
1299 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1300 Details : for a short time RPCs with bulk IO are in the replay list,
1301 but replay of bulk IOs is unimplemented. If the OST filesystem
1302 is corrupted due to disk cache incoherency and then replay is
1303 started it is possible to trip an assertion. Avoid putting
1304 committed RPCs into the replay list at all to avoid this issue.
1309 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1310 Details : Large single O_DIRECT read and write calls can fail to allocate
1311 a sufficiently large buffer to process the request. In case of
1312 allocation failure the allocation is retried with a smaller
1313 buffer and broken into smaller requests.
1315 Severity : enhancement
1317 Description: Add -o localflock option to simulate old noflock behaviour.
1318 Details : This will achieve local-only flock/fcntl locks coherentness.
1323 Description: versioning check is incomplete
1324 Details : Checking the version difference of client vs. server, report
1325 error if the gap is too big.
1330 Description: add support PG_writeback bit
1331 Details : add support for PG_writeback bit for Lustre, for more carefull
1332 work with page cache in 2.6 kernel. This also fix some deadlocks
1333 and remove hack for work O_SYNC with 2.6 kernel.
1335 Severity : enhancement
1337 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1338 Details : The uninit_groups feature works in conjunction with the kernel
1339 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1340 up the pass1 processing of e2fsck. This is a read-only feature
1341 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1342 mount filesystems that have had this feature enabled.
1344 Severity : enhancement
1346 Description: Improve multi-block allocation algorithm to avoid fragmentation
1347 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1348 allocation locality and avoid filesystem fragmentation.
1350 ------------------------------------------------------------------------------
1352 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1354 * Support for kernels:
1355 2.6.9-42.0.3.EL (RHEL 4)
1356 2.6.5-7.276 (SLES 9)
1357 2.4.21-47.0.1.EL (RHEL 3)
1358 2.6.12.6 vanilla (kernel.org)
1359 2.6.16.21-0.8 (SLES10)
1360 * Recommended e2fsprogs version: 1.39.cfs2-0
1362 * The backwards-compatible /proc/sys/portals symlink has been removed
1363 in this release. Before upgrading, please ensure that you change
1364 any configuration scripts or /etc/sysctl.conf files that access
1365 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1366 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1367 in advance of the upgrade on any system running Lustre 1.4.6 or
1368 newer, since /proc/sys/lnet was added in that version.
1369 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1373 Frequency : only when quota is used
1375 Description: avoid scanning export list for quota master
1376 Details : Change the algorithms to avoid scanning export list in order
1377 to improve the efficiency.
1380 Frequency : MDS failover only, very rarely
1382 Description: "went back in time" messages on mds failover
1383 Details : The greatest transno may be lost when the current operation
1384 finishes with an error (transno==0) and the client's last_rcvd
1385 record is over-written. Save the greatest transno in the
1386 mds_last_transno for this case.
1389 Frequency : always for specific kernels and striping counts
1391 Description: client may get "Matching packet too big" without ACL support
1392 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1393 when trying to access files in certain configurations. The
1394 clients should in fact be denied when mounting because they do
1395 not understand ACLs.
1398 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1400 Description: many clients connecting with IO in progress causes connect timeouts
1401 Details : Avoid synchronous journal commits to avoid delays caused by many
1402 clients connecting/disconnecting when bulk IO is in progress.
1403 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1404 OST_IO_PORTAL to avoid delays behind potentially many pending
1408 Frequency : occasionally with multiple writers to a single file
1410 Description: shared writes to file may result in wrong size reported by stat()
1411 Details : Allow growing of kms when extent lock is cancelled
1414 Frequency : always with random mmap IO to multi-striped file
1416 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1417 Details : Only if the hole is at the end of OST object so that kms is too
1418 small. Fix is to increase kms accordingly in ll_nopage.
1421 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1423 Description: writes to a missing object would leak memory on the OST
1424 Details : If there is an inconsistency between the MDS and OST filesystems,
1425 such that the MDS references an object that doesn't exist, writes
1426 to that object will leak memory due to incorrect cleanup in the
1427 error handling path, eventually running out of memory on the OST.
1432 Description: Creating too long symlink causes lustre errors
1433 Details : Check symlink and name lengths before sending requests to MDS.
1436 Frequency : only if flock is enabled (not on by default)
1438 Description: posix locks not released on fd closure on 2.6.9+
1439 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1440 kernels, this caused such locks not to be released on fd close and
1441 then assertions on fs unmount about still used locks.
1444 Frequency : MDS failover only, very rarely
1446 Description: clients may get ASSERTION(granted_lock != NULL)
1447 Details : When request was taking a long time, and a client was resending
1448 a getattr by name lock request. The were multiple lock requests
1449 with the same client lock handle and
1450 mds_getattr_name->fixup_handle_for_resent_request found one of the
1451 lock handles but later failed with ASSERTION(granted_lock != NULL).
1456 Description: handle->h_buffer_credits > 0, assertion failure
1457 Details : h_buffer_credits is zero after truncate, causing assertion
1458 failure. This patch extends the transaction or creates a new
1462 Frequency : NFS re-export or patchless client
1463 Bugzilla : 11179, 10796
1464 Description: Crash on NFS re-export node (__d_move)
1465 Details : We do not want to hash the dentry if we don't have a lock.
1466 But if this dentry is later used in d_move, we'd hit uninitialised
1467 list head d_hash, so we just do this to init d_hash field but
1468 leave dentry unhashed.
1471 Frequency : NFS re-export or patchless client
1473 Description: NFS exports has problem with symbolic link
1474 Details : lustre client didn't properly install dentry when re-exported
1475 to NFS or running patchless client.
1478 Frequency : NFS re-export or patchless client
1480 Description: Various nfs/patchless fixes.
1481 Details : fixes reuse disconected alias for lookup process - this fixes
1482 warning "find_exported_dentry: npd != pd",
1483 fix permission error with open files at nfs.
1484 fix apply umask when do revalidate.
1487 Frequency : occasional
1489 Description: Crash on NFS re-export node
1490 Details : calling clear_page() on the wrong pointer triggered oops in
1491 generic_mapping_read().
1494 Frequency : rarely, using O_DIRECT IO
1496 Description: unaligned directio crashes client with LASSERT
1497 Details : check for unaligned buffers before trying any requests.
1500 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1502 Description: stale data returned from RAID cache
1503 Details : If only a small amount of IO is done to the RAID device before
1504 reading it again it is possible to get stale data from the RAID
1505 cache instead of reading it from disk.
1508 Frequency : always for sles10 kernel
1510 Description: sles10 support
1511 Details : ll_follow_link: compile fixes and using of nd_set_link
1512 under newer kernels.
1515 Frequency : depends on arch, kernel and compiler version, always on sles10
1518 Description: recursive or deep enough symlinks cause stack overflow
1519 Details : getting rid of large stack-allocated variable in
1523 Frequency : depends on hardware
1525 Description: lustre write performance loss in the SLES10 kernel
1526 Details : the performance loss is caused by using of write barriers in the
1527 ext3 code. The SLES10 kernel turns barrier support on by
1528 default. The fix is to undo that change for ldiskfs.
1530 ------------------------------------------------------------------------------
1532 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1534 * Support for kernels:
1535 2.6.9-42.0.3EL (RHEL 4)
1536 2.6.5-7.276 (SLES 9)
1537 2.4.21-47.0.1.EL (RHEL 3)
1538 2.6.12.6 vanilla (kernel.org)
1542 Frequency : quota enabled and large files being deleted
1544 Description: releasing more than 4GB of quota at once hangs OST
1545 Details : If a user deletes more than 4GB of files on a single OST it
1546 will cause the OST to spin in an infinite loop. Release
1547 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1552 Description: statfs data retrieved from /proc may be stale or zero
1553 Details : When reading per-device statfs data from /proc, in the
1554 {kbytes,files}_{total,free,avail} files, it may appear
1555 as zero or be out of date.
1558 Frequency : systems with MD RAID1 external journal devices
1560 Description: lconf's call to blkid is confused by RAID1 journal devices
1561 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1562 of returning all block devices that match the journal UUID.
1565 Frequency : always, for aggregate stripe size over 4GB
1567 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1568 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1569 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1572 Frequency : always if "lfs find" used on a local file/directory
1574 Description: "lfs find" segfaults if used on a local file/directory
1575 Details : The case where a directory component was not specified wasn't
1576 handled correctly. Handle this properly.
1579 Frequency : always on ppc64
1581 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1582 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1583 perform endianness conversion.
1586 Frequency : rarely (truncate to non-zero file size after write under load)
1587 Bugzilla : 10730, 10687
1588 Description: Files padded with zeros to next 4K multiple
1589 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1590 it is possible that files that are truncated to a non-zero size
1591 immediately after being written are filled with zero bytes beyond
1592 the truncated size. No file data is lost.
1594 Severity : enhancement
1596 Description: Allow recovery/failover for liblustre clients.
1597 Details : liblustre clients were unaware of failover configurations until
1600 Severity : enhancement
1602 Description: user file locks should fail when not mounting with flock option
1603 Details : Set up an error-returning stub in ll_file_operations.lock field
1604 to prevent incorrect behaviour when client is mounted without
1605 flock option. Also, set up properly f_op->flock field for
1609 Frequency : always on ia64
1611 Description: "lfs df" loops on printing out MDS statfs information
1612 Details : The obd_ioctl_data was not initialized and in some systems
1613 this caused a failure during the ioctl that did not return
1614 an error. Initialize the struct and return an error on failure.
1617 Frequency : SLES 9 only
1619 Description: Error of copying files with lustre special EAs as root
1620 Details : Client side always return success for setxattr call for lustre
1621 special xattr (currently only "trusted.lov").
1624 Frequency : rarely on clusters with both ia64+i386 clients
1626 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1627 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1628 clients are doing concurrent writes to the same file it is
1629 possible that the ia64 clients may LASSERT because the OST
1630 extent locks are not PAGE_SIZE aligned. Ensure that grown
1631 locks are always aligned on the request boundary.
1634 Frequency : specific use, occasional
1636 Description: Overwriting in use executable truncates on-disk binary image
1637 Details : If one node attempts to overwrite an executable in use by
1638 another node, we now correctly return ETXTBSY instead of
1639 truncating the file.
1641 Severity : enhancement
1643 Description: Async OSC create to avoid the blocking unnecessarily.
1644 Details : If a OST has no remain object, system will block on the creating
1645 when need to create a new object on this OST. Now, ways use
1646 pre-created objects when available, instead of blocking on an
1647 empty osc while others are not empty. If we must block, we block
1648 for the shortest possible period of time.
1653 Description: chmod on Lustre root is propagated to other clients
1654 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1655 invalid by the follow_mount time.
1658 Frequency : liblustre clients only
1660 Description: Race in 'instant cancel' lock handling could lead to such locks
1661 never to be granted in case of SMP MDS
1662 Details : Do not destroy not yet granted but cbpending locks in
1666 Frequency : replay/resend of open
1668 Description: non null lock assetion failure in mds_intent_policy
1669 Details : Trying to replay/resend lockless open requests resulted in
1670 mds_open() returning 0 with no lock. Now it sets a flag if
1671 a lock is going to be returned.
1673 Severity : enhancement
1675 Description: Checksum enhancements
1676 Details : New checksum enhancements allow for resending RPCs that failed
1679 Severity : enhancement
1681 Description: Tunables on number of dirty pages in cacche
1682 Details : Allow to set limit on number of dirty pages cached.
1687 Description: client crash on unmount - lock still has references
1688 Details : In some error handling cases it was possible to leak a lock
1689 reference on a client while accessing a file. This was not
1690 harmful to the client during operation, but would cause the
1691 client to crash when the filesystem is unmounted.
1694 Frequency : specific case, rare
1696 Description: ETXTBSY on mds though file not in use by client
1697 Details : ETXTBSY is no longer incorrectly returned when attempting to
1698 chmod or chown a directory that the user previously tried to
1699 execute or a currently-executing binary.
1702 Frequency : extremely rare except on liblustre-based clients
1704 Description: Lustre space not freed when files are deleted
1705 Details : Clean up open-unlinked files after client eviction. Previously
1706 the unlink was skipped and the files remained as orphans.
1711 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1712 Details : In some cases it was possible to send a blocking callback to
1713 a client doing a glimpse, even though that client didn't get
1714 a lock granted. When the glimpse lock is cancelled on the OST
1715 the freed lock is left on the waiting list and corrupted the list.
1718 Frequency : all core dumps
1720 Description: Broke core dumps to lustre
1721 Details : Negative dentry may be unhashed if parent does not have UPDATE
1722 lock, but some callers, e.g. do_coredump, expect dentry to be
1723 hashed after successful create, hash it in ll_create_it.
1725 ------------------------------------------------------------------------------
1727 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1729 * Support for kernels:
1730 2.6.9-42.0.2.EL (RHEL 4)
1731 2.6.5-7.276 (SLES 9)
1732 2.4.21-40.EL (RHEL 3)
1733 2.6.12.6 vanilla (kernel.org)
1737 Frequency : always on RHEL 3
1739 Description: Number of open files grows over time
1740 Details : The number of open files grows over time, whether or not
1741 Lustre is started. This was due to a filp leak introduced
1742 by one of our kernel patches.
1744 ------------------------------------------------------------------------------
1746 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1748 * Support for kernels:
1749 2.6.9-42.EL (RHEL 4)
1750 2.6.5-7.276 (SLES 9)
1751 2.4.21-40.EL (RHEL 3)
1752 2.6.12.6 vanilla (kernel.org)
1757 Bugzilla : 5719, 9635, 9792, 9684
1758 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1759 Details : If a server is under heavy load and cannot reply to new
1760 connection requests before the client resends the (re)connect,
1761 the connection handling code can behave badly if two service
1762 threads are concurrently handing separate (re)connections from
1763 the same client. Add better locking to the connection handling
1764 code, and ensure that only a single connection will be processed
1765 for a given client UUID, even if the lock is dropped.
1767 Severity : enhancement
1769 Description: add TCP zero-copy support to kernel
1770 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1771 sends if the hardware supports scatter-gather and checksumming.
1772 This allows socklnd to do client-write and server-read more
1773 efficiently and reduce CPU utilization from skbuf copying.
1776 Frequency : only if NFS exporting from client
1778 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1779 Details : If setting attributes on a file created under NFS that had
1780 never been opened it would be possible to oops the client
1781 if the file had no objects.
1784 Frequency : always for liblustre
1786 Description: liblustre client does MDS+OSTs setattr RPC for each write
1787 Details : When doing a write from a liblustre client, the client
1788 incorrectly issued an RPC to the MDS and each OST the file was
1789 striped over in order to update the timestamps. When writing
1790 with small chunks and many clients this could overwhelm the MDS
1791 with RPCs. In all cases it would slow down the write because
1792 these RPCs are unnecessary.
1794 Severity : enhancement
1796 Description: allow number of MDS service threads to be changed at module load
1797 Details : It is now possible to change the number of MDS service threads
1798 running. Adding "options mds mds_num_threads={N}" to the MDS's
1799 /etc/modprobe.conf will set the number of threads for the next
1800 time Lustre is restarted (assuming the "mds" module is also
1801 reloaded at that time). The default number of threads will
1802 stay the same, 32 for most systems.
1807 Description: OST crash if filesystem is unformatted or corrupt
1808 Details : If an OST is started on a device that has never been formatted
1809 or if the filesystem is corrupt and cannot even mount then the
1810 error handling cleanup routines would dereference a NULL pointer.
1815 Description: NULL pointer deref in llap_from_page.
1816 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1817 incorrect) mapping if the page was truncated/reclaimed while it was
1818 searched for. Check for this condition and skip such pages when
1819 doing readahead. Introduce extra check to llap_from_page() to
1820 verify page->mapping->host is non-NULL (so page is not anonymous).
1823 Frequency : Sometimes when using sys_sendfile
1825 Description: "page not covered by a lock" warnings from ll_readpage
1826 Details : sendfile called ll_readpage without right page locks present.
1827 Now we introduced ll_file_sendfile that does necessary locking
1828 around call to generic_file_sendfile() much like we do in
1832 Frequency : with certain MDS communication failures at client mount time
1834 Description: NULL pointer deref after failed client mount
1835 Details : a client connection request may delayed by the network layer
1836 and not be sent until after the PTLRPC layer has timed out the
1837 request. If the client fails the mount immediately it will try
1838 to clean up before the network times out the request. Add a
1839 reference from the request import to the obd device and delay
1840 the cleanup until the network drops the request.
1843 Frequency : occasionally during client (re)connect
1845 Description: assertion failure during client (re)connect
1846 Details : processing a client connection request may be delayed by the
1847 client or server longer than the client connect timeout. This
1848 causes the client to resend the connection request. If the
1849 original connection request is replied in this interval, the
1850 client may trip an assertion failure in ptlrpc_connect_interpret()
1851 which thought it would be the only running connect process.
1854 Frequency : only with obd_echo servers and clients that are rebooted
1856 Description: kernel BUG accessing uninitialized data structure
1857 Details : When running an obd_echo server it did not start the ping_evictor
1858 thread, and when a client was evicted an uninitialized data
1859 structure was accessed. Start the ping_evictor in the RPC
1860 service startup instead of the OBD startup.
1862 Severity : enhancement
1863 Bugzilla : 10193 (patchless)
1864 Description: Remove dependency on various unexported kernel interfaces.
1865 Details : No longer need reparent_to_init, exit_mm, exit_files,
1866 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1869 Frequency : rare (only users of deprecated and unsupported LDAP config)
1871 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1872 Details : LDAP apparently contains 'lustreName' attributes instead of
1873 'name'. A simple remapping of the name is sufficient.
1876 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1878 Description: waiting_locks_callback trips kernel BUG if client is evicted
1879 Details : Running with the dump_on_timeout debug flag turned on makes
1880 it possible that the waiting_locks_callback() can try to dump
1881 the Lustre kernel debug logs from an interrupt handler. Defer
1882 this log dumping to the expired_lock_main() thread.
1884 Severity : enhancement
1886 Description: Support NFS exporting on 2.6 kernels.
1887 Details : Implement non-rawops metadata methods for NFS server to use without
1888 changing NFS server code.
1891 Frequency : very rare (synthetic metadata workload only)
1893 Description: two racing renames might cause an MDS thread to deadlock
1894 Details : Running the "racer" program may cause one MDS thread to rename
1895 a file from being the source of a rename to being the target of
1896 a rename at exactly the same time that another thread is doing
1897 so, and the second thread has already enqueued these locks after
1898 doing a lookup of the target and is trying to relock them in
1899 order. Ensure that we don't try to re-lock the same resource.
1902 Frequency : only very large systems with liblustre clients
1904 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1905 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1906 all OSTs in parallel. This allows the eviction of stale liblustre
1907 clients to proceed much faster than if they were done in series,
1908 and also offers similar improvements for other set_info RPCs.
1913 Description: excessive CPU usage during initial read phase on client
1914 Details : During the initial read phase on a client, it would agressively
1915 retry readahead on the file, consuming too much CPU and impacting
1916 performance (since 1.4.5.8). Improve the readahead algorithm
1917 to avoid this, and also improve some other common cases (read
1918 of small files in particular, where "small" is files smaller than
1919 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1924 Description: MDS crash when receiving packet with unknown intent.
1925 Details : Do not LBUG in unknown intent case, just return -EFAULT
1927 Severity : enhancement
1928 Bugzilla : 9293, 9385
1929 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1930 Details : Do not serialize getattr (non-intent version) and statfs.
1933 Frequency : occasional, when OST network is overloaded/intermittent
1935 Description: client evicted by OST after bulk IO timeout
1936 Details : If a client sends a bulk IO request (read or write) the OST
1937 may evict the client if it is unresposive to its data GET/PUT
1938 request. This is incorrect if the network is overloaded (takes
1939 too long to transfer the RPC data) or dropped the OST GET/PUT
1940 request. There is no need to evict the client at all, since
1941 the pinger and/or lock callbacks will handle this, and the
1942 client can restart the bulk request.
1945 Frequency : Always when mmapping file with no objects
1947 Description: client crashes when mmapping file with no objects
1948 Details : Check that we actually have objects in a file before doing any
1949 operations on objects in ll_vm_open, ll_vm_close and
1955 Description: Request leak when working with deleted CWD
1956 Details : Introduce advanced request refcount tracking for requests
1957 referenced from lustre intent.
1959 Severity : Enhancement
1961 Description: Cache open file handles on client.
1962 Details : MDS now will return special lock along with openhandle, if
1963 requested and client is allowed to hold openhandle, even if unused,
1964 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1965 closing files for every read/write openration.
1967 Severity : Enhancement
1969 Description: Cache open negative dentries on client when possible.
1970 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1971 negative dentries on lock revocation.
1976 Description: Remounting a client read-only wasn't possible with a zconf mount
1977 Details : It wasn't possible to remount a client read-only with llmount.
1979 Severity : enhancement
1980 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1981 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1982 MPICH is included to add Lustre-specific ADIO interfaces.
1983 This is based closely on the UFS ADIO layer and only differs
1984 in file creation, in order to allow the OST striping to be set.
1985 This is user-contributed code and not supported by CFS.
1990 Description: extended inode attributes (immutable, append-only) work improperly
1991 when 2.4 and 2.6 kernels are used on client/server or vice versa
1992 Details : Introduce kernel-independent values for these flags.
1994 Severity : enhancement
1997 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1998 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1999 granularity. Now, Fractional MB granularity is be supported,
2000 this is very useful for low memory system.
2002 Severity : enhancement
2004 Description: Getattr by fid
2005 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2006 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2007 have locks and avoids getattr rpc requests for it on every stat.
2010 Frequency : Always, for filesystems larger than 2TB
2012 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2013 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2014 filesystems larger than 2TB to be mounted with mballoc enabled.
2017 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2019 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2020 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2021 filesystems larger than 2TB is mounted then the kernel will
2022 silently corrupt the start of the filesystem. CONFIG_LBD is
2023 enabled for all CFS-supported kernels, but the possibility of
2024 this happening with a modified kernel config exists.
2026 Severity : enhancement
2028 Description: add client O_DIRECT support for 2.6 kernels
2029 Details : It is now possible to do O_DIRECT reads and writes to files
2030 in the Lustre client mountpoint on 2.6 kernel clients.
2032 Severity : enhancement
2034 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2035 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2036 parallel, not waiting for response from every OST before sending
2037 a rpc to the next OST.
2042 Description: setattr vs write race when updating file timestamps
2043 Details : Client processes that update a file timestamp into the past
2044 right after writing to the file (e.g. tar) it is possible that
2045 the updated file modification time can be reset to the current
2046 time due to a race between processing the setattr and write RPC.
2048 Severity : enhancement
2050 Description: Bring 'lfs find' closer in line with regular Linux find.
2051 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2052 -print0 options and obtains all the needed info through the lustre
2055 Severity : enhancement
2057 Description: support up to 1024 configured devices on one node
2058 Details : change obd_dev array from statically allocated to dynamically
2059 allocated structs as they are first used to reduce memory usage
2064 Description: Flush dirty partially truncated pages during truncate
2065 Details : Immediatelly flush partially truncated pages in filter_setattr,
2066 this way we completely avoid having any pages in page cache on OST
2067 and can retire ugly workarounds during writes to flush such pages.
2072 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2073 Details : protect inode from truncation within vfs_unlink() context
2074 just take a reference before calling vfs_unlink() and release it
2075 when parent's i_sem is free.
2080 Description: last_id value checked outside lock on OST caused LASSERT failure
2081 Details : If there were multiple MDS->OST object precreate requests in
2082 flight, it was possible that the OST's last object id was checked
2083 outside a lock and incorrectly tripped an assertion. Move checks
2084 inside locks, and discard old precreate requests.
2087 Frequency : always, if extents are used on OSTs
2089 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2090 Details : OSTs using the extents format would not zero the high 16 bits of
2091 the index physical block number. This is not a problem for any
2092 OST filesystems smaller than 16TB, and no kernels support ext3
2093 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2094 new/modified files) and can be fixed for existing filesystems
2095 with e2fsprogs-1.39-cfs1.
2100 Description: import connection selection may be incorrect if timer wraps
2101 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2102 import connections to be ignored if the 32-bit jiffies counter
2103 wraps. Use a 64-bit jiffies counter.
2106 Frequency : very large clusters immediately after boot
2108 Description: LNET request buffers exhausted under heavy short-term load
2109 Details : If a large number of client requests are generated on a service
2110 that has previously never seen so many requests it is possible
2111 that the request buffer growth cannot keep up with the spike in
2112 demand. Instead of dropping incoming requests, they are held in
2113 the LND until the RPC service can accept more requests.
2116 Frequency : Sometimes during replay
2118 Description: Assertion failure in ll_local_open after replay.
2119 Details : If replay happened on an open request reply before we were able
2120 to set replay handler, reply will become not swabbed tripping the
2121 assertion in ll_local_open. Now we set the handler right after
2122 recognising of open request
2125 Frequency : very rare
2127 Description: kernel reports "badness in vsnprintf"
2128 Details : Reading from the "recovery_status" /proc file in small chunks
2129 may cause a negative length in lprocfs_obd_rd_recovery_status()
2130 call to vsnprintf() (which is otherwise harmless). Exit early
2131 if there is no more space in the output buffer.
2133 Severity : enhancement
2135 Description: clear OBD RPC statistics by writing to them
2136 Details : It is now possible to clear the OBD RPC statistics by writing
2137 to the "stats" file.
2142 Description: Client mtime is not the same on different clients after utimes
2143 Details : In some cases, the client was using the utimes() syscall on
2144 a file cached on another node. The clients now validate the
2145 ctime from the MDS + OSTs to determine which one is right.
2150 Description: Inability to activate failout mode
2151 Details : lconf script incorrectly assumed that in pythong string's numeric
2152 value is used in comparisons.
2155 Frequency : always with multiple stripes per file
2157 Description: Inefficient object allocation for mutli-stripe files
2158 Details : When selecting which OSTs to stripe files over, for files with
2159 a stripe count that divides evenly into the number of OSTs,
2160 the MDS is always picking the same starting OST for each file.
2161 Return the OST selection heuristic to the original design.
2166 Description: mount failures may take full timeout to return an error
2167 Details : Under some heavy load conditions it is possible that a
2168 failed mount can wait for the full obd_timeout interval,
2169 possibly several minutes, before reporting an error.
2170 Instead return an error as soon as the status is known.
2172 Frequency : quota enabled and large files being deleted
2174 Description: releasing more than 4GB of quota at once hangs OST
2175 Details : If a user deletes more than 4GB of files on a single OST it
2176 will cause the OST to spin in an infinite loop. Release
2177 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2182 Description: statfs data retrieved from /proc may be stale or zero
2183 Details : When reading per-device statfs data from /proc, in the
2184 {kbytes,files}_{total,free,avail} files, it may appear
2185 as zero or be out of date.
2188 Frequency : systems with MD RAID1 external journal devices
2190 Description: lconf's call to blkid is confused by RAID1 journal devices
2191 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2192 of returning all block devices that match the journal UUID.
2195 Frequency : always, for aggregate stripe size over 4GB
2197 Description: assertion fails when trying to use 4GB stripe size
2198 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2199 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2202 Frequency : always on ppc64
2204 Description: the first write on an ext3 filesystem with mballoc got stuck
2205 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2206 perform endianness conversion.
2208 ------------------------------------------------------------------------------
2210 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2212 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2213 INTEROPERATE with older versions automatically. Please read the
2214 user documentation before upgrading any part of a live system.
2215 * WARNING: Lustre networking configuration changes are required with
2216 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2219 * Support for kernels:
2220 2.6.9-22.0.2.EL (RHEL 4)
2221 2.6.5-7.244 (SLES 9)
2222 2.6.12.6 vanilla (kernel.org)
2225 Severity : enhancement
2226 Bugzilla : 7981/8208
2227 Description: Introduced Lustre Networking (LNET)
2228 Details : LNET is new networking infrastructure for Lustre, it includes
2229 a reorganized network configuration mode (see the user
2230 documentation for full details) as well as support for routing
2231 between different network fabrics. Lustre Networking Devices
2232 (LNDs) for the supported network fabrics have also been
2233 created for this new infrastructure.
2235 Severity : enhancement
2236 Description: Introduced Access control lists
2237 Details : clients can set ACLs on files and directories in order to have
2238 more fine-grained permissions than the standard Unix UGO+RWX.
2239 The MDS must be started with the "-o acl" mount option.
2241 Severity : enhancement
2242 Description: Introduced filesystem quotas
2243 Details : Administrators may now establish per-user quotas on the
2246 Severity : enhancement
2248 Description: Configuration change for the XT3
2249 The PTLLND is now used to run Lustre over Portals on the XT3
2250 The configure option(s) --with-cray-portals are no longer used.
2251 Rather --with-portals=<path-to-portals-includes> is used to
2252 enable building on the XT3. In addition to enable XT3 specific
2253 features the option --enable-cray-xt3 must be used.
2258 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2259 Details : A race between allocating a new llcd and re-getting the llcd_lock
2260 allowed another thread to grab newly-allocated llcd.
2262 Severity : enhancement
2264 Description: 2.6 OST async journal commit and locking fix to improve performance
2265 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2266 2.6 kernels are now async as they already were in 2.4 kernels so
2267 that they can commit concurrently with the network bulk transfer.
2268 For block-allocated files the filter allocation semaphore is held
2269 to avoid filesystem fragmentation during allocation. BKL lock
2270 removed for 2.6 xattr operations where it is no longer needed.
2275 Description: lconf incorrectly determined whether two IP networks could talk
2276 Details : In some more complicated routing and multiple-network
2277 configurations, lconf will avoid trying to make a network
2278 connection to a disjoint part of the IP space. It was doing the
2279 math incorrectly for one set of cases.
2284 Description: Fix for potential infinite loop processing records in an llog.
2285 Details : If an llog record is corrupted/zeroed, it is possible to loop
2286 forever in llog_process(). Validate the llog record length
2287 and skip the remainder of the block on error.
2290 Frequency : occasional (liblustre only)
2292 Description: liblustre could not open files whose last component is a symlink
2293 Details : sysio_path_walk() would incorrectly pass the open intent to
2294 intermediate path components.
2297 Frequency : rare (liblustre only with non-standard tuning)
2298 Bugzilla : 7201 (7350)
2299 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2300 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2301 after it has been released, only do this for referenced locks
2303 Severity : enhancement
2305 Description: specifying an (invalid) directory default stripe_size of -1
2306 would reset the directory default striping
2307 Details : stripe_size -1 was used internally to signal directory stripe
2308 removal, now use "all default" to signal dir stripe removal
2309 as a directory striping of "all default" is not useful
2312 Frequency : common for large clusters running liblustre clients
2314 Description: doing an ls when liblustre clients are running is slow
2315 Details : sending a glimpse AST to a liblustre client waits for every AST
2316 to time out, as liblustre clients will not respond. Since they
2317 cannot cache data we refresh the OST lock LVB from disk instead.
2319 Severity : enhancement
2321 Description: doing an ls at the same time as file IO can be slow
2322 Details : enqueue and other "small" requests can be blocked behind many
2323 large IO requests. Create a new OST IO portal for non-IO
2324 requests so they can be processed faster.
2327 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2329 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2330 created file with mode 000
2331 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2332 allow an ftruncate() as a non-root user to a file with mode 000.
2333 Lustre now respects this flag to disable mode checks when
2334 truncating a file owned by the user
2337 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2340 Description: Revoking locks from clients that went dead or catatonic might take
2342 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2343 cancellation of such locks instant on servers without waiting for
2344 any reply from clients. Clients drops these locks when cancel
2345 notification from server is received without replying.
2348 Frequency : liblustre-only, when liblustre client dies or becomes busy
2350 Description: Doing ls on Linux clients can take a long time with active
2352 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2353 granting such locks to it in the first place if possible. Locks
2354 are taken by proxy on the OST during the read or write and
2355 dropped immediately afterward. Add connect flags handling, do
2356 not grant locks to liblustre clients for glimpse ASTs.
2358 Severity : enhancement
2360 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2361 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2362 sequential reads and behaves badly with applications doing random
2363 reads. Improve it by reading ahead at least read region, and
2364 avoiding excessive large RPC for small reads.
2366 Severity : enhancement
2368 Description: Creating more than 1000 files for a single job may cause a load
2369 imbalance on the OSTs if there are also a large number of OSTs.
2370 Details : qos_prep_create() uses an OST index reseed value that is an
2371 even multiple of the number of available OSTs so that if the
2372 reseed happens in the middle of the object allocation it will
2373 still utilize the OSTs as uniformly as possible.
2378 Description: OST or MDS may oops in ping_evictor_main()
2379 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2380 but doesn't restart at beginning of obd_exports_timed list
2383 Severity : enhancement
2385 Description: improve by-nid export eviction on the MDS and OST
2386 Details : allow multiple exports with the same NID to be evicted at one
2387 time without re-searching the exports list.
2390 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2392 Description: MDS may oops in groups_free()
2393 Details : in rare race conditions a newly allocated group_info struct is
2394 freed again, and this can be NULL. The 2.4 compatibility code
2395 for groups_free() checked for a NULL pointer, but 2.6 did not.
2398 Frequency : common for liblustre clients doing little filesystem IO
2399 Bugzilla : 9352, 7313
2400 Description: server may evict liblustre clients accessing contended locks
2401 Details : if a client is granted a lock or receives a completion AST
2402 with a blocking AST already set it would not reply to the AST
2403 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2406 Frequency : lfs setstripe, only systems with more than 160 OSTs
2408 Description: unable to set striping with a starting offset beyond OST 160
2409 Details : llapi_create_file() incorrectly limited the starting stripe
2410 index to the maximum single-file stripe count.
2413 Frequency : LDAP users only
2415 Description: lconf did not handle in-kernel recovery with LDAP properly
2416 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2418 Severity : enhancement
2420 Description: bind OST threads to NUMA nodes to improve performance
2421 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2422 node and do their allocations there to localize memory access
2424 Severity : enhancement
2426 Description: llmount can determine client NID directly from Myrinet (GM)
2427 Details : the client NID code from gmnalnid was moved directly into
2428 llmount, removing the need to use this or specifying the
2429 client NID explicitly when mounting GM clients with zeroconf
2432 Frequency : if client is started with down MDS
2434 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2435 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2436 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2437 ptlrpc_import_delayed_req() because we didn't check if the
2438 request was interrupted, and we also didn't break out of the
2439 event loop if there was no timeout
2444 Description: data loss during non-page-aligned writes to a single file from
2445 both multiple nodes and multiple threads on one node at same time
2446 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2447 inconsistency led to false short-reads, that were cached and later
2448 used by ->prepare_write() to fill in partially written page,
2449 leading to data loss.
2452 Frequency : always, if lconf --abort_recovery used
2454 Description: lconf --abort_recovery fails with 'Operation not supported'
2455 Details : lconf was attempting to abort recovery on the MDT device and not
2458 Severity : enhancement
2460 Description: remove cleanup logs
2461 Details : replace lconf-generated cleanup logs with lustre internal
2462 cleanup routines. Eliminates the need for client-cleanup and
2465 Severity : enhancement
2467 Description: add support for EAs (user and system) on lustre filesystems
2468 Details : it is now possible to store extended attributes in the Lustre
2469 client filesystem, and with the user_xattr mount option it
2470 is possible to allow users to store EAs on their files also
2472 Severity : enhancement
2474 Description: Add possibility (config option) to show minimal available OST free
2476 Details : When compiled with --enable-mindf configure option, statfs(2)
2477 (and so, df) will return least minimal free space available from
2478 all OSTs as amount of free space on FS, instead of summary of
2479 free spaces of all OSTs.
2481 Severity : enhancement
2483 Description: do not expand extent locks acquired on OST-side
2484 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2485 by server: they are not cached anyway.
2488 Frequency : when mmap is used/binaries executed from Lustre
2490 Description: Unmmap pages before throwing them away from read cache.
2491 Details : llap_shrink cache now attempts to unmap pages before discarding
2492 them (if unmapping failed - do not discard). SLES9 kernel has
2493 extra checks that trigger if this unmapping is not done first.
2498 Description: lconf didn't resolve symlinks before checking to see whether a
2499 given mountpoint was already in use
2502 Frequency : when migrating failover services
2503 Bugzilla : 6395, 9514
2504 Description: When migrating a subset of services from a node (e.g. failback
2505 from a failover service node) the remaining services would
2506 time out and evict clients.
2507 Details : lconf --force (implied by --failover) sets the global obd_timeout
2508 to 5 seconds in order to quickly disconnect, but this caused
2509 other RPCs to time out too quickly. Do not change the global
2510 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2512 Severity : enhancement
2513 Frequency : if MDS is started with down OST
2514 Bugzilla : 9439,5706
2515 Description: Allow startup/shutdown of an MDS without depending on the
2516 availability of the OSTs.
2517 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2518 Add appropriate locking and lov-osc refcounts for safe
2519 cleaning. Add osc abort_inflight calls in case the
2520 synchronize never started.
2523 Frequency : occasional (Cray XT3 only)
2525 Description: root not authorized to access files in CRAY_PORTALS environment
2526 Details : The client process capabilities were not honoured on the MDS in
2527 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2528 been accessed by an authorized user then root was able to access
2529 the file on the local client also. The root user capabilities
2530 are now allowed on the MDS, as this environment has secure UID.
2533 Frequency : occasional
2535 Description: ldiskfs "too long searching" message happens too often
2536 Details : A debugging message (otherwise harmless) prints too often on
2537 the OST console. This has been reduced to only happen when
2538 there are fragmentation problems on the filesystem.
2543 Description: Division by zero in statfs when all OSCs are inactive
2544 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2545 lov_statfs divides by value returned by lov_get_stripecnt().
2549 Bugzilla : 9489, 3273
2550 Description: First write from each client to each OST was only 4kB in size,
2551 to initialize client writeback cache, which caused sub-optimal
2552 RPCs and poor layout on disk for the first writen file.
2553 Details : Clients now request an initial cache grant at (re)connect time
2554 and so that they can start streaming writes to the cache right
2555 away and always do full-sized RPCs if there is enough data.
2556 If the OST is rebooted the client also re-establishes its grant
2557 so that client cached writes will be honoured under the grant.
2562 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2563 Details : Now I/O RPCs go to different portal number and (presumably) fast
2564 lock requests (and glimses) and other RPCs get their own service
2565 threads pool that should be able to service those RPCs
2568 Severity : enhancement
2570 Description: Ability to exchange lustre version between client and servers and
2571 issue warnings at client side if client is too old. Also for
2572 liblustre clients there is ability to refuse connection of too old
2574 Details : New 'version' field is added to connect data structure that is
2575 filled with version info. That info is later checked by server and
2579 Frequency : rare, liblustre only.
2580 Bugzilla : 9296, 9581
2581 Description: Two simultaneous writes from liblustre at offset within same page
2582 might proceed at the same time overwriting eachother with stale
2584 Details : I/O lock withing llu_file_prwv was released too early, before data
2585 actually was hitting the wire. Extended lock-holding time until
2586 server acknowledges receiving data.
2589 Frequency : extremely rare. Never observed in practice.
2591 Description: avoid generating lustre_handle cookie of 0.
2592 Details : class_handle_hash() generates handle cookies by incrementing
2593 global counter, and can hit 0 occasionaly (this is unlikely, but
2594 not impossible, because initial value of cookie counter is
2595 selected randonly). Value of 0 is used as a sentinel meaning
2596 "unassigned handle" --- avoid it. Also coalesce two critical
2597 sections in this function into one.
2599 Severity : enhancement
2601 Description: allow liblustre clients to delegate truncate locking to OST
2602 Details : To avoid overhead of locking, liblustre client instructs OST to
2603 take extent lock in ost_punch() on client's behalf. New connection
2604 flag is added to handle backward compatibility.
2606 Severity : enhancement
2607 Bugzilla : 4928, 7341, 9758
2608 Description: allow number of OST service threads to be specified
2609 Details : a module parameter allows the number of OST service threads
2610 to be specified via "options ost ost_num_threads={N}" in the
2611 OSS's /etc/modules.conf or /etc/modprobe.conf.
2615 Bugzilla : 6146, 9635, 9895
2616 Description: servers crash with bad pointer in target_handle_connect()
2617 Details : In rare cases when a client is reconnecting it was possible that
2618 the connection request was the last reference for that export.
2619 We would temporarily drop the export reference and get a new
2620 one, but this may have been the last reference and the export
2621 was just destroyed. Get new reference before dropping old one.
2623 Severity : enhancement
2624 Frequency : if client is started with failover MDS
2626 Description: Allow multiple MDS hostnames in the mount command
2627 Details : Try to read the configuration from all specified MDS
2628 hostnames during a client mount in case the "primary"
2631 Severity : enhancement
2633 Description: Stop sending data to evicted clients as soon as possible.
2634 Details : Check if the client we are about to send or are sending data to
2635 was evicted already. (Check is done every second of waiting,
2636 for which l_wait_event interface was extended to allow checking
2637 of exit condition at specified intervals).
2640 Frequency : rare, normally only when NFS exporting is done from client
2642 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2644 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2647 Frequency : always, if srand() is called before liblustre initialization
2649 Description: Liblustre uses system PRNG disturbing its usage by user application
2650 Details : Introduce internal to lustre fast and high-quality PRNG for
2651 lustre usage and make liblustre and some other places in generic
2652 lustre code to use it.
2654 Severity : enhancement
2655 Bugzilla : 9477, 9557, 9870
2656 Description: Verify that the MDS configuration logs are updated when xml is
2657 Details : Check if the .xml configuration logs are newer than the config
2658 logs stored on the MDS and report an error if this is the case.
2659 Request --write-conf, or allow starting with --old_conf.
2661 Severity : enhancement
2663 Description: Handle symlinks in the path when checking if Lustre is mounted.
2664 Details : Resolve intermediate symlinks when checking if a client has
2665 mounted a filesystem to avoid duplicate client mounts.
2670 Description: lconf can hit an error exception but still return success.
2671 Details : The lconf command catches the Command error exception at the top
2672 level script context and will exit with the associated exit
2673 status, but doesn't ensure that this exit status is non-zero.
2678 Description: failure of ptlrpc thread startup can cause oops
2679 Details : Starting a ptlrpc service thread can fail if there are a large
2680 number of threads or the server memory is very fragmented.
2681 Handle this without oopsing.
2684 Frequency : always, only if liblustre and non-default acceptor port was used
2686 Description: liblustre cannot connect to servers with non-default acceptor port
2687 Details : tcpnal_set_default_params() was not called and was therefore
2688 ignoring the environment varaible TCPNAL_PORT, as well as other
2689 TCPNAL_ environment variables
2694 Description: two objects could be created on the same OST for a single file
2695 Details : If an OST is down, in some cases it was possible to create two
2696 objects on a single OST for a single file. No problems other
2697 than potential performance impact and spurious error messages.
2701 Bugzilla : 5681, 9562
2702 Description: Client may oops in ll_unhash_aliases
2703 Details : Client dcache may become inconsistent in race condition.
2704 In some cases "getcwd" can fail if the current directory is
2710 Description: Inode refcounting problems in NFS export code
2711 Details : link_raw functions used to call d_instantiate without obtaining
2712 extra inode reference first.
2716 Bugzilla : 9942, 9903
2717 Description: Referencing freed requests leading to crash, memleaks with NFS.
2718 Details : We used to require that call to ll_revalidate_it was always
2719 followed by ll_lookup_it. Also with revalidate_special() it is
2720 possible to call ll_revalidate_it() twice for the same dentry
2721 even if first occurence returned success. This fix changes semantic
2722 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2723 reference on a request referred from the intent.
2724 ll_intent_release() then releases such a request.
2727 Frequency : rare, normally benchmark loads only
2729 Description: unlinked inodes were kept in memory on the client
2730 Details : If a client is repeatedly creating and unlinking files it
2731 can accumulate a lot of stale inodes in the inode slab cache.
2732 If there is no other client load running this can cause the
2733 client node to run out of memory. Instead flush old inodes
2734 from client cache that have the same inode number as a new inode.
2737 Frequency : SLES9 2.6.5 kernel and long filenames only
2738 Bugzilla : 9969, 10379
2739 Description: utime reports stale NFS file handle
2740 Details : SLES9 uses out-of-dentry names in some cases, which confused
2741 the lustre dentry revalidation. Change it to always use the
2745 Frequency : rare, unless heavy write-truncate concurrency is continuous
2746 Bugzilla : 4180, 6984, 7171, 9963, 9331
2747 Description: OST becomes very slow and/or deadlocked during object unlink
2748 Details : filter_destroy() was holding onto the parent directory lock
2749 while truncating+unlinking objects. For very large objects this
2750 may block other threads for a long time and slow overall OST
2751 responsiveness. It may also be possible to get a lock ordering
2752 deadlock in this case, or run out of journal credits because of
2753 the combined truncate+unlink. Solution is to do object truncate
2754 first in one transaction without parent lock, and then do the
2755 final unlink in a new transaction with the parent lock. This
2756 reduces the lock hold time dramatically.
2759 Frequency : rare, 2.4 kernels only
2761 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2762 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2763 before unmounting their filesystem in order to avoid deadlock.
2764 The kernel_locked() function in 2.4 kernels only checks whether
2765 the kernel lock is held, not whether it is this process that is
2766 holding it as 2.6 kernels do.
2771 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2772 Details : The client ptlrpc code may be trying to reconnect to a down
2773 server before a previous connection attempt has timed out.
2774 Increase the reconnect interval to be longer than the connection
2775 timeout interval to avoid sending duplicate connections to
2779 Frequency : echo_client brw_test command
2781 Description: fix echo_client to work with OST preallocated code
2782 Details : OST preallocation code (5137) didn't take echo_client IO path
2783 into account: echo_client calls filter methods outside of any
2784 OST thread and, hence, there is no per-thread preallocated
2785 pages and buffers to use. Solution: hijack pga pages for IO. As
2786 a byproduct, this avoids unnecessary data copying.
2790 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2791 Description: Client can oops in mdc_commit_close() after open replay
2792 Details : It was possible for the MDS to return an open request with no
2793 transaction number in mds_finish_transno() if the client was
2794 evicted, but without actually returning an error. Clients
2795 would later try to replay that open and may trip an assertion
2796 Simplify the client close codepath, and always return an error
2797 from the MDS in case the open is not successful.
2800 Frequency : rare, 2.6 OSTs only
2802 Description: OST may deadlock under high load on fragmented files
2803 Details : If there was a heavy load and highly-fragmented OST filesystems
2804 it was possible to have all the OST threads deadlock waiting on
2805 allocation of biovecs, because the biovecs were not released
2806 until the entire RPC IO was completed. Instead, release biovecs
2807 as soon as they are complete to ensure forward IO progress.
2809 Severity : enhancement
2811 Description: Support for specifying external journal device at mount
2812 Details : If an OST or MDS device is formatted with an external journal
2813 device, this device major/minor is stored in the ext3 superblock
2814 and may not be valid for failover. Allow detecting and
2815 specifying the external journal at mount time.
2820 Description: Mounting an MDS with pending unlinked files may cause oops
2821 Details : target_finish_recovery() calls mds_postrecov() which returned
2822 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2823 considers this an error and immediately begins cleaning up the
2824 lov, just after starting the mds_lov process
2826 Severity : enhancement
2828 Description: Implement 'lfs df' to report actual free space on per-OST basis
2829 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2830 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2831 usage of objects; '-h' to report in human readable format.
2833 ------------------------------------------------------------------------------
2835 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2842 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2843 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2844 at startup during orphan unlinking. Instead, skip buddy update
2845 before setup, as it will be regenerated after recovery is complete.
2850 Description: If an OST is inactive, its locks might reference stale inodes.
2851 Details : lov_change_cbdata() must iterate over all namespaces, even if
2852 they are inactive to clear inode references from the lock.
2854 Severity : enhancement
2855 Frequency : occasional, if non-standard max_dirty_mb used
2857 Description: Client will block write RPCs if not enough grant
2858 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2859 then the client will block writes while waiting for another RPC
2860 to complete instead of consuming its dirty limit. With change
2861 we get improved performance when max_dirty_mb is small.
2863 Severity : enhancement
2864 Bugzilla : 3389, 6253
2865 Description: Add support for supplementary groups on the MDS.
2866 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2867 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2868 lookups for user supplementary groups into a cache.
2872 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2873 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2874 for liblustre O_EXCL is enforced by MDS.
2878 Description: readdir never returns NULL in liblustre.
2879 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2880 dentry in struct dirent.
2884 Frequency : liblustre only, depends on application IO pattern
2885 Description: liblustre clients evicted if not contacting servers
2886 Details : Don't put liblustre clients into the ping_evictor list, so
2887 they will not be evicted by the pinger ever.
2889 Severity : enhancement
2891 Description: Add ability to evict clients by NID from MDS.
2892 Details : By echoing "nid:$NID" string into
2893 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2894 $NID would be instantly evicted from this MDS and from all active
2895 OSTs connected to it.
2899 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2900 Details : lookup_it_finish() used to query file size from OSTs that was not
2905 Description: service threads change working directory to that of init
2906 Details : Starting lustre service threads may pin the working directory
2907 of the parent thread, making that filesystem busy. Threads
2908 now change to the working directory of init to avoid this.
2912 Frequency : during shutdown only
2913 Description: shutdown with a failed MDS or OST can cause unmount to hang
2914 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2919 Frequency : 2.6 only
2920 Description: chmod/chown may include an extra supplementary group
2921 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2922 supplementary group and if none is specified this is used.
2925 Bugzilla : 5479 (6816)
2927 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2928 Details : If lookup is in progress on a file that is unlinked we might try
2929 to revalidate the inode and fail in revalidate after lookup is
2930 complete and ll_file_open() enqueues the open again but
2931 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2934 Frequency : always, if lconf --abort_recovery used
2936 Description: lconf --abort_recovery fails with 'Operation not supported'
2937 Details : lconf was attempting to abort recovery on the MDT device and not
2940 ------------------------------------------------------------------------------
2942 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2947 Frequency : rare (only unsupported configurations with a node running as an
2949 Bugzilla : 6514, 5137
2950 Description: Mounting a Lustre file system on a node running as an OST could
2952 Details : OSTs now preallocates memory needed to write out data at
2953 startup, instead of when needed, to avoid having to
2954 allocate memory in possibly low memory situations.
2955 Specifically, if the file system is mounted on on OST,
2956 memory pressure could force it to try to write out data,
2957 which it needed to allocate memory to do. Due to the low
2958 memory, it would be unable to do so and the node would
2959 become unresponsive.
2961 Severity : enhancement
2963 Description: Addition of lconf --service command line option
2964 Details : lconf now accepts a '--service <arg>' option, which is
2965 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2967 Severity : enhancement
2969 Description: Failover mode is now the default for OSTs.
2970 Details : By default, OSTs will now run in failover mode. To return to
2971 the old behaviour, add '--failout' to the lmc line for OSTs.
2973 Severity : enhancement
2975 Description: Health checks are now provided for MDS and OSTs
2976 Details : Additional detailed health check information on MSD and OSTs
2977 is now provided through the procfs health_check value.
2980 Frequency : occasional, depends on IO load
2982 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2983 after numerous create/delete cycles
2984 Details : The ext3 inode allocation policy would not allocate new inodes
2985 very well on the OSTs because there are no new directories
2986 being created. Instead we look for groups with free space if
2987 the parent directories are nearly full.
2992 Description: Network or server problems during mount may cause partially
2993 mounted clients instead of returning an error.
2994 Details : The config llog parsing code may overwrite the error return
2995 code during mount error handling, returning success instead
3001 Description: MDS can fail to allocate large reply buffers
3002 Details : After long uptimes the MDS can fail to allocate large reply
3003 buffers (e.g. zconf client mount config records) due to memory
3004 fragmentation or consumption by the buffer cache. Preallocate
3005 some large reply buffers so that these replies can be sent even
3006 under memory pressure.
3010 Frequency : rare (liblustre)
3011 Description: fsx running with liblustre complained that using truncate() to
3012 extend the file doesn't work. This patch corrects that issue.
3013 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3014 ATTR_SIZE and lsm use in llu_setattr_raw.
3018 Frequency : rare, only 2.6 kernels
3019 Description: Unusual file access patterns on the MDS may result in inode
3020 data being lost in very rare circumstances.
3021 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3022 code in the 2.6 kernel caused the inode and/or EA data not to
3023 be read from disk, causing single-file corruption.
3027 Frequency : rare, only 2.6 filesystems using extents
3028 Description: Heavy concurrent write and delete load may cause data corruption.
3029 Details : It was possible under high-load situations to have an extent
3030 metadata block in the block device cache from a just-unlinked
3031 file overwrite a newly-allocated data block. We now unmap any
3032 metadata buffers that alias just-allocated data blocks.
3036 Frequency : filesystems with default stripe_count larger than 77
3037 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3038 Details : lconf specifies an inode size of 4096 bytes when the default
3039 stripe_count is larger than 77. This conflicts with the default
3040 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3041 this case to avoid pinning too much memory for large EAs.
3043 ------------------------------------------------------------------------------
3045 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3050 Frequency : rare (extremely heavy IO load with hundreds of clients)
3052 Description: Client is evicted, gets IO error writing to file
3053 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3054 caused clients to be evicted for AST timeouts. The fixes in
3055 bug 5192 mean we no longer need to have such short AST timeouts
3056 so ldlm_timeout has been increased.
3059 Frequency : occasional during --force or --failover shutdown under load
3060 Bugzilla : 5949, 4834
3061 Description: Server oops/LBUG if stopped with --force or --failover under load
3062 Details : a collection of import/export refcount and cleanup ordering
3063 issues fixed for safer force cleanup
3066 Frequency : only filesystems larger than 120 OSTs
3067 Bugzilla : 5990, 6223
3068 Description: lfs getstripe would oops on a very large filesystem
3069 Details : lov_getconfig used kfree on vmalloc'd memory
3072 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3073 Bugzilla : 6242, 6243
3074 Description: reading from files that had been truncated to a non-zero size
3075 but never opened returned no data
3076 Details : ll_file_read() reads zeros from no-object files to EOF
3081 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3082 Details : There's little harm in aborting MDS/OSS recovery and letting it
3083 try again, so I removed the LASSERT and return an error instead.
3085 Severity : enhancement
3087 Description: New debugging infrastructure for tracking down data corruption
3088 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3089 (b) cover more of the client-side code path, and (c) try to narrow
3090 down where problems occurred
3094 Bugzilla : 3819, 4364, 4397, 6313
3095 Description: Racing close and eviction MDS could cause assertion in mds_close
3096 Details : It was possible to get multiple mfd references during close and
3097 client eviction, leading to one thread referencing a freed mfd.
3099 Severity: : enhancement
3100 Bugzilla : 3262, 6359
3101 Description: Attempts to reconnect to servers are now more aggressive.
3102 Details : This builds on the enhanced upcall-less recovery that was added
3103 in 1.4.2. When trying to reconnect to servers, clients will
3104 now try each server in the failover group every 10 seconds. By
3105 default, clients would previously try one server every 25 seconds.
3110 Description: After recovery, certain operations trigger a failed
3111 assertion on a client.
3112 Details : Failing over an mds, using lconf -d --failover, while a
3113 client was doing a readdir() call would cause the client to
3114 LBUG after recovery completed and the readdir() was resent.
3116 Severity : enhancement
3118 Description: Default groups are now added by lconf
3119 Details : You can now run lconf --group <servicename> without having to
3120 manually add groups with lmc.
3123 Frequency : occasional
3125 Description: Nodes with an elan id of 0 trigger a failed assertion
3128 Frequency : always when accessing e.g. tty/console device nodes
3130 Description: tty and some other devices nodes cannot be used on lustre
3131 Details : file's private_data field is used by device data and lustre
3132 values in there got lost. New field was added to struct file to
3133 store fs-specific private data.
3136 Frequency : when exporting Lustre via NFS
3138 Description: NFSD failed occasionally when looking up a path component
3139 Details : NFSD is looking up ".." which was broken in ext3 directories
3140 that had grown large enough to become hashed.
3143 Frequency : Clusters with multiple interfaces not on the same subnet
3145 Description: Nodes will repeatedly try to reconnect to an interface which it
3146 cannot reach and report an error to the log.
3147 Details : Extra peer list entries will be created by lconf with some peers
3148 unreachable. lconf now validates the peer before adding it.
3151 Frequency : Only if a default stripe is set on the filesystem root.
3153 Description: Setting a default stripe on the filesystem root prevented the
3154 filesystem from being remounted.
3155 Details : The client was sending extra request flags in the root getattr
3156 request and did not allocate a reply buffer for the dir EA.
3159 Frequency : occasional, higher if lots of files are accessed by one client
3160 Bugzilla : 6159, 6097
3161 Description: Client trips assertion regarding lsm mismatch/magic
3162 Details : While revalidating inodes the VFS looks up inodes with ifind()
3163 and in rare cases can find an inode that is being freed.
3164 The ll_test_inode() code will free the lsm during ifind()
3165 when it finds an existing inode and then the VFS later attaches
3166 this free lsm to a new inode.
3170 Bugzilla : 6422, 7030
3171 Description: MDS deadlock between mkdir and client eviction
3172 Details : Creating a new file via mkdir or mknod (starting a transaction
3173 and getting the ns lock) can deadlock with client eviction
3174 (gets ns lock and trying to finish a synchronous transaction).
3177 Frequency : occasional
3178 Description: While starting a server, the fsfilt_ext3 module could not be
3180 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3181 kernels. Previously, lconf would still use the ext3 name
3182 when trying to load modules. Now, it will correctly use
3183 ext3 on 2.4 and ldiskfs on 2.6.
3185 Severity : enhancement
3186 Description: The default stripe count has been changed to 1
3187 Details : The interpretation of the default stripe count (0, to lfs
3188 or lmc) has been changed to mean striping across a single
3189 OST, rather than all available. For general usage we have
3190 found a stripe count of 1 or 2 works best.
3192 Severity : enhancement
3193 Description: Add support for compiling against Cray portals.
3194 Details : Conditional compiling for some areas that are different
3198 Frequency : occasional
3199 Bugzilla : 6409, 6834
3200 Description: Creating files with an explicit stripe count may lead to
3201 a failed assertion on the MDS
3202 Details : If some OSTs are full or unavailable, creating files may
3203 trigger a failed assertion on the MDS. Now, Lustre will
3204 try to use other servers or return an error to the
3208 Frequency : occasional
3210 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3211 caused lock timeout during readahead (since 1.4.2).
3212 Details : Processes doing readahead might match a lock that hasn't been
3213 granted yet if there are overlapping and conflicting lock
3214 requests. The readahead process waits on ungranted lock
3215 (original lock is CBPENDING), while OST waits for that process
3216 to cancel CBPENDING read lock and eventually evicts client.
3218 Severity : enhancement
3220 Description: Initial enabling of flock support for clients
3221 Details : Implements fcntl advisory locking and file status functions.
3222 This feature is provided as an optional mount flag (default
3223 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3224 locking are implemented yet, and those that are are not guaranteed
3225 to be completely correct in production environments.
3226 mount -t lustre -o [flock|noflock] ...
3229 Frequency : occasional
3231 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3232 assertion in the ext3 JBD (journaling) layer.
3233 Details : The b_committed_data struct is protected by the big kernel lock
3234 in 2.4 kernels, serializing journal_commit_transaction() and
3235 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3236 access to this struct. In 2.6 kernels there is finer grained
3237 locking to improve SMP performance of the JBD layer.
3241 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3246 Description: merge and cleanup kernel patches.
3249 -----------------------------------------------------------------------------
3251 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3253 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3254 versions of Lustre. Please update all servers and clients to
3255 version 1.4.2 or later at the same time. You must also run
3256 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3257 to update the configuration logs.
3259 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3260 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3261 - swab llog records of type '0' so we get proper header size/idx (5861)
3262 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3263 - fix rename of one directory over another leaking an inode (5953)
3264 - avoid SetPageDirty on 2.6 (5981)
3265 - don't re-add just-being-destroyed locks to the waiting list (5653)
3266 - when creating new directories, inherit the parent's custom
3267 striping settings if present parent (3048)
3268 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3269 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3270 - don't hold client locks on temporary worklist from l_lru (5666)
3271 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3272 - automatically evict dead clients (5921)
3273 - Update file size properly in create+truncate+fstat case (6196)
3274 - Do not unhash mountpoint dentries, do not allow removal of
3276 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3277 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3278 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3279 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3280 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3281 - Correctly handle reads of files with no objects (6243)
3282 - lctl recover will also mark a device active if deactivate used (5933)
3284 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3285 - allow --write-conf on an MDS with different nettype than client (5619)
3286 - don't write config llogs to MDS for mounts not from that MDS (5617)
3287 - lconf should create multiple TCP connections from a client (5201)
3288 - init scripts are now turned off by default; run chkconfig --on
3289 lustre and chkconfig --on lustrefs to use them
3290 - upcalls are no longer needed for clients to recover to failover
3292 - add --abort-recovery option to lconf to abort recovery on device
3294 - add support for an arbitrary number of OSTs (3026)
3295 - Quota support protocol changes.
3296 - forward compatibility changes to wire structs (6007)
3297 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3298 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3299 - improved "lustre status" script
3300 - initialize blocksize for non-regular files (6062)
3301 - added --disable-server and --disable-client configure options (5782)
3302 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3303 - Vanilla 2.4.29 support
3304 - increase maximum number of obd devices to 520 (6242)
3305 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3306 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3307 SLES 9 2.6.5 kernels
3308 - limit stripes per file to 160 (the maximum EA size) (6093)
3310 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3313 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3314 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3315 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3316 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3317 - small ext3 extents cleanups and fixes (5733)
3318 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3319 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3320 - avoid needless client->OST connect, fix handle mismatch (5317)
3321 - fix DLM error path that led to out-of-sync client, long delays (5779)
3322 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3323 - fix several locking issues related to i_size (5492,5624,5654,5672)
3324 - don't move pending lock onto export if it is already evicted (5683)
3325 - fix kernel oops when creating .foo in unlinked directory (5548)
3326 - fix deadlock in obdfilter statistics vs. object create (5811)
3327 - use time_{before,after} to avoid timer jiffies wrap (5882)
3328 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3329 - Do not leak request if server was not able to process it (5154)
3330 - If mds_open unable to find parent dir, make that negative lookup(5154)
3331 - don't create new directories with extent-mapping (5909, 5936)
3333 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3334 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3335 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3336 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3337 - support for RHEL4 kernel on i686 (5773)
3338 - provide error messages when incompatible logs are encountered (5898)
3340 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3341 * version 1.4.0.10 (1.4.1 release candidate 1)
3343 - don't keep a lock reference when lock is not granted (4238)
3344 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3345 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3346 - drop import inflight refcount on signal_completed_replay error (5255)
3347 - unlock page after async write error during send (3677)
3348 - handle missing objects in filter_preprw_read properly (5265)
3349 - no transno return for symlink open, don't save no-trasno open (3440)
3350 - don't try to complete elan receive that already failed (4012)
3351 - free RPC server reply state on error (5406)
3352 - clean up thread from ptlrpc_start_thread() on error (5160)
3353 - readahead could read extra page into cache that wasn't ejected (5388)
3354 - prevent races in class_attach/setup/cleanup/detach (5260)
3355 - don't dereference de->d_inode after l_dput of de (5458)
3356 - use "int" for stripe value returned from lock_to_stripe (5544)
3357 - mballoc allocation and error-checking fixes in 2.6 (5504)
3358 - block device patches to fix I/O request sizes in 2.6 (5482)
3359 - look up hostnames for IB nals (5602)
3360 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3361 - don't start multiple acceptors for the same port (5277)
3362 - fix incorrect LASSERT in mds_getattr_name (5635)
3363 - export a proc file for general "ping" checking (5628)
3364 - fix "lfs check" to not block when the MDS is down (5628)
3366 - service request history (4965)
3367 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3368 - create an "evict_client" /proc entry on OSTs, like the MDS has
3369 - fix mount usage message, return errors per mount(8) (5168)
3370 - change grep [] to grep "[]" in tests so they work in more UMLs
3371 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3372 - remove ancient LOV_MAGIC_V0 EA support (5047)
3373 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3374 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3375 - allow readahead allocations to fail when low on memory (5383)
3376 - mmap locking landed again, after considerable improvement (2828)
3377 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3379 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3382 - send OST transaction number in read/write reply to free req (4966)
3383 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3384 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3385 - the watchdog thread now runs as interruptible (5246)
3386 - flock/lockf fixes (but it's still disabled, pending 5135)
3387 - don't use EXT3 constants in llite code (5094)
3388 - memory shortage at startup could cause assertion (5176)
3390 - reorganization of lov code
3391 - single portals codebase
3393 - add extents/mballoc support (5025)
3394 - direct I/O reads in the obdfilter (4048)
3395 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3397 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3400 - send OST transaction number in read/write reply to free req (4966)
3401 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3402 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3403 - the watchdog thread now runs as interruptible (5246)
3404 - handle missing objects in filter_preprw_read properly (5265)
3405 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3406 - drop import inflight refcount on signal_completed_replay error (5255)
3407 - unlock page after async write error during send (3677)
3408 - return original error code on reconstructed replies (3761)
3409 - no transno return for symlink open, don't save no-trasno open (3440)
3411 - add pid to ldlm debugging output (4922)
3412 - bump the watchdog timeouts -- we can't handle 30sec yet
3413 - extra debugging for orphan dentry/inode bug (5259)
3415 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3418 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3419 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3420 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3421 - fix NULL dereference in /proc/sys/portals/routes (4827)
3422 - allow failed mdc_close() operations to be interrupted (4561)
3423 - stop precreate on OST before MDS would time out on it (4778)
3424 - don't send partial-page writes before EOF from client (4410)
3425 - discard client grant for sub-page writes on large-page clients (4520)
3426 - don't free dentries not owned by NFS code, check generation (4806)
3427 - fix lsm leak if mds_create_objects() fails (4801)
3428 - limit debug_daemon file size, always print CERROR messages (4789)
3429 - use transno after validating reply (3892)
3430 - process timed out requests if import state changes (3754)
3431 - update mtime on OST during writes, return in glimpse (4829)
3432 - add mkfsoptions to LDAP (4679)
3433 - use ->max_readahead method instead of zapping global ra (5039)
3434 - don't interrupt __l_wait_event() during strace
3436 - add software watchdogs to catch hung threads quickly (4941)
3437 - make lustrefs init script start after nfs is mounted
3438 - fix CWARN/ERROR duplication (4930)
3439 - return async write errors to application if possible (2248)
3440 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3441 - print NAL number in %x format (4645)
3442 - update barely-supported suse-2.4.21-171 series (4842)
3443 - support for sles 9 %post scripts
3444 - support for building 2.6 kernel-source packages
3445 - support for sles km_* packages
3447 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3450 - ignore -ENOENT errors in osc_destroy (3639)
3451 - notify osc create thread that OSC is being cleaned up (4600)
3452 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3453 - reconstruct ost_handle() like mds_handle() (4657)
3454 - create a new thread to do import eviction to avoid deadlock (3969)
3455 - let lconf resolve symlinked-to devices (4629)
3456 - don't unlink "objects" from directory with default EA (4554)
3457 - hold socknal file ref over connect in case target is down (4394)
3458 - allow more than 32000 subdirectories in a single directory (3244)
3459 - fix blocks count for O_DIRECT writes (3751)
3460 - OST returns ENOSPC from object create when no space left (4539)
3461 - don't send truncate RPC if file size isn't changing (4410)
3462 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3463 - bind to privileged port in socknal and tcpnal (3689)
3465 - rate limit CERROR/CWARN console message to avoid overload (4519)
3466 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3467 - basic mmap support (3918)
3468 - kernel patch series update from b1_4 (4711)
3470 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3473 - avoid crash during MDS cleanup with OST shut down (2775)
3474 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3475 - don't use bad inodes on the MDS (3744)
3476 - dynamic object preallocation to improve recovery speed (4236)
3477 - don't hold spinlock over lock dumping or change debug flags (4401)
3478 - don't zero obd_dev when it is force cleaned (3651)
3479 - print grants to console if they go negative (4431)
3480 - "lctl deactivate" will stop automatic recovery attempts (3406)
3481 - look for existing locks in ldlm_handle_enqueue() (3764)
3482 - don't resolve lock handle twice in recovery avoiding race (4401)
3483 - revalidate should check working dir is a directory (4134)
3485 - don't always mark "slow" obdfilter messages as errors (4418)
3487 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3490 - don't close LustreDB during write_conf until it is done (3860)
3491 - fix typo in lconf for_each_profile (3821)
3492 - allow dumping logs from multiple threads at one time (3820)
3493 - don't allow multiple threads in OSC recovery (3812)
3494 - fix debug_size parameters (3864)
3495 - fix mds_postrecov to initialize import for llog ctxt (3121)
3496 - replace config semaphore with spinlock (3306)
3497 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3498 - don't allow enqueue to complete on a destroyed export (3822)
3499 - down write_lock before checking llog header bitmap (3825)
3500 - recover from lock replay timeout (3764)
3501 - up llog sem before sending rpc (3652)
3502 - reduce ns lock hold times when setting kms (3267)
3503 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3504 - fix NULL deref and obd_dev leak on setup error (3312)
3505 - replace some LBUG about llog ops with error handling (3841)
3506 - don't match INVALID dentries from d_lookup and spin (3784)
3507 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3508 - fix invalid assertion in ptlrpc_set_wait (3880)
3510 - add libwrap support for the TCP acceptor (3996)
3511 - add /proc/sys/portals/routes for non-root route listing (3994)
3512 - allow setting MDS UUID in .xml (2580)
3513 - print the stack of a process that LBUGs (4228)
3515 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3518 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3519 - make sure to unset replay flag from failed open requests (3440)
3520 - if default stripe count is 0, use OST count for inode size (3636)
3521 - update parent mtime/ctime on client for create/unlink (2611)
3522 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3523 - free recovery state on server during a forced cleanup (3571)
3524 - unregister_reply for resent reqs (3063)
3525 - loop back devices mounting and status check on 2.6 (3563)
3526 - fix resource-creation race that can provoke i_size == 0 (3513)
3527 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3528 - more debugging for page-accounting assertion (3746)
3529 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3530 - avoid deadlock after precreation failure (3758)
3531 - fix race and lock order deadlock in orphan handling (3450, 3750)
3532 - add validity checks when grabbing inodes from l_ast_data (3599)
3534 - add /proc/.../recovery_status to obdfilter (3428)
3535 - lightweight CDEBUG infrastructure, debug daemon (3668)
3536 - change default OSC RPC parameters to be better on small clusters
3537 - turn off OST read cache for files smaller than 32MB
3538 - install man pages and include them in rpms (3100)
3539 - add new init script for (un)mounting lustre filesystems (2593)
3540 - run chkconfig in %post for init scripts (3701)
3541 - drop scimac NAL (unmaintained)
3543 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3546 - clean kiobufs before and after use (3485)
3547 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3548 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3549 - change init script to fail more clearly if not run as root (1528)
3550 - allow clients to reconnect during replay (1742)
3551 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3552 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3553 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3554 - fix cond_resched (3554)
3555 - only evict unfinished clients after recovery (3515)
3556 - allow bulk resend, prevent data loss (3570)
3557 - dynamic ptlrpc request buffer allocation (2102)
3558 - don't allow unlinking open directory if it isn't empty (2904)
3559 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3560 - remove extraneous obd dereference causing LASSERT failure (3334)
3561 - don't use get_cycles() when creating temp. files on the mds (3156)
3562 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3563 - handle EEXIST for set-stripe, set proper directory name (3336)
3565 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3566 - fix ksocknal_fmb_callback() error messages (2918)
3568 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3571 - don't copy lvb into (possibly NULL) reply on error (2983)
3572 - don't deref dentry after dput, don't free lvb on error (2922)
3573 - use the kms to determine writeback rpc length (2947)
3574 - increment oti_logcookies when osc is inactive (2948)
3575 - update client's i_blocks count via lvb messages (2543)
3576 - handle intent open/close of special files properly (1557)
3577 - mount MDS with errors=remount-ro, like obdfilter (2009)
3578 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3579 - don't use cancelling-locks' kms values (2947)
3580 - use highest lock extent for kms, not last one (2925)
3581 - don't dereference ERR_PTR() dentry in error handling path (3107)
3582 - fix thread race in portals_debug_dumplog() (3122)
3583 - create lprocfs device entries at setup instead of at attach (1519)
3584 - common AST error handler, don't evict client on completion race (3145)
3585 - zero nameidata in detach_mnt in 2.6 (3118)
3586 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3587 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3588 - initialize RPC timeout timer earlier for 2.6 (3219)
3589 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3590 - print nal/nid for unknown nid (3258)
3591 - additional checks for oscc recovery before doing precreate (3284)
3592 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3593 - don't crash in mdc_close for bad permissions on open (3285)
3594 - zero i_rdev for non-device files (3147)
3595 - clear page->private before handing to FS, better assertion (3119)
3596 - tune the read pipeline (3236)
3597 - fix incorrect decref of invalidated dentry (2350)
3598 - provide read-ahead stats and refine rpc in flight stats (3328)
3599 - don't hold journal transaction open across create RPC (3313)
3600 - update atime on MDS at close time (3265)
3601 - close LDAP connection when recovering to avoid server load (3315)
3602 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3603 - don't leak open file on MDS after open resend (3325)
3604 - serialize filter_precreate and filter_destroy_precreated (3329)
3605 - loop device shouldn't call sync_dev() for nul device (3092)
3606 - clear page cache after eviction (2766)
3607 - resynchronize MDS->OST in background (2824)
3608 - refuse to mount the same filesystem twice on same mountpoint (3394)
3609 - allow llmount to create routes for mounting behind routers (3320)
3610 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3611 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3612 - fix rare problem with rename on htree directories (3417)
3614 - allow default OST striping configuration per directory (1414)
3615 - fix compilation for qswnal for 2.6 kernels (3125)
3616 - increase maximum number of MDS request buffers for large systems
3617 - change liblustreapi to be useful for external progs like lfsck (3098)
3618 - increase local configuration timeout for slow disks (3353)
3619 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3621 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3624 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3625 - don't overwrite extent policy data in reply if lock was blocked (2901)
3626 - drop filter export grants atomically with removal from device (2663)
3627 - del obd_self_export from work_list in class_disconnect_exports (2908)
3628 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3629 - swab reply message in mdc_close, other PPC fixes (2464)
3630 - fix destroying of named logs (2325)
3631 - overwrite old logs when running lconf --write_conf (2264)
3632 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3633 - fix race in target_handle_connect (2898)
3634 - mds_reint_create() should take same inode create lock (2926)
3635 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3636 - don't close files for self_export to avoid uninitialized obd (2936)
3637 - allow MDS with the same name as client node (2939)
3638 - hold dentry reference for closed log files for unlink (2325)
3639 - reserve space for all logs during transactions (2059)
3640 - don't evict page beyond end of stripe extent (2925)
3641 - don't oops on a deleted current working directory (2399)
3642 - handle hard links to targets without a parent properly (2517)
3643 - don't dereference NULL lock when racing during eviction (2867)
3644 - don't grow lock extents when lots of conflicting locks (2919)
3646 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3649 - account for cache space usage on clients to avoid data loss (974)
3650 - lfsck support in lustre kernel code (2349)
3651 - reduce journal credits needed for BRW writes (2370)
3652 - orphan handling to avoid losing space on client/server crashes
3653 - ptlrpcd can be blocked, stopping ALL progress (2477)
3654 - use lock value blocks to assist in proper KMS, faster stat (1021)
3655 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3656 - recovery for initial connections (2355)
3657 - fixes for mds_cleanup_orphans (1934)
3658 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3659 - block all file creations until orphan recovery completes (1901)
3660 - client remove rq_connection from request struct (2423)
3661 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3662 - recovery timer race (2670)
3663 - mdc_close recovey bug (2532)
3664 - ptlrpc cleanup bug (2710)
3665 - mds timeout on local locks (2588)
3666 - namespace lock held during RPCs (2431)
3667 - handle interrupted sync write properly (2503)
3668 - don't try to handle a message that hasn't been replied to (2699)
3669 - client assert failure during cleanup after abort recovery (2701)
3670 - leak mdc device after failed mount (2712)
3671 - ptlrpc_check_set allows timedout requests to complete (2714)
3672 - wait for inflight reqs when ptlrpcd finishes (2710)
3673 - make sure unregistered services are removed from the srv_list
3674 - reset bulk XID's when resending them (caught by 1138 test)
3675 - unregister_bulk after timeout
3676 - fix lconf error (2694)
3677 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3678 - readahead locks pages, leaves pending causing memory pressure (2673)
3679 - increase OST request buffers to 4096 on large machines (2729)
3680 - fix up permission of existing directories in simple_mkdir (2661)
3681 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3682 - don't assign transno to errored transactions (2742)
3683 - don't delete objects on OST if given a bogus objid from MDS (2751)
3684 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3685 - if rq_no_resend, then timeout request after recovery (2432)
3686 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3687 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3688 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3689 - MDS recovery shouldn't delete objects that it has given out (2730)
3690 - if enqueue arrives after completion, don't clobber LVB (2819)
3691 - don't unlock pages twice when trigger_group_io returns error (2814)
3692 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3693 - don't write pages to disk if there was an error (1450)
3694 - don't ping imports that have recovery disabled (2676)
3695 - take buffered bytes into account when balancing socknal conn (2817)
3696 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3697 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3698 - remove little-endian swabbing of llog records (1987)
3699 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3700 - retry reposting request buffers if they fail (1191)
3701 - grow extent at grant time to avoid granting a revoked lock (2809)
3702 - lock revoke doesn't evict page if covered by a second lock (2765)
3703 - disable VM readahead to avoid reading outside lock extents (2805)
3705 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3706 - updated kernel patches for hp-2.4.20 kernel (2681)
3708 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3711 - fix truncated write corruption (2366)
3712 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3714 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3715 - ensure len(uuid) < 37 in lmc (1171)
3716 - fix ia64 OOPS in llog_test (2255)
3717 - zero end of page at obdfilter for partial page writes (2648)
3718 - don't leave stale dentries around after renames (bug 2428)
3719 - fix timeouts when evicting a client with a single lock held (2642)
3720 - set deadline for the initial HELLO message to drain (2634)
3721 - print out dotted-quad IP addresses in the socknal (2302)
3723 - additional debugging for MDS client eviction problem (2443)
3724 - fix mkfsoptions support for osts (2603, 2604)
3726 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3729 - add series for the vanilla 2.6.0 kernel
3730 - add series for the vanilla 2.4.24 kernel
3731 - add series for a cray x86/64 UL kernel drop
3732 - fix xattr patches for the vanilla 2.4.19 series
3734 - generate true UUIDs in lmc (1171)
3735 - have portals stack dumping break in UML (2466)
3736 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3737 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3738 - wake callback waiting threads on client eviction (2460)
3739 - Add --ptldebug and --subsystem to lmc (1719)
3740 - update assertion to allow safe interrupt allocation
3741 - set rq_no_resend for cancel requests (2432)
3742 - recalculate ptlrpcd timeout after resend (2494)
3743 - call vfs_rmdir when removing pending directories (2368)
3744 - fix renaming a file to itself (2429)
3745 - lmc creates a default one-stripe lov (2454)
3746 - expand procfs space to handle large clusters (2326)
3747 - increase UML stack to avoid overflow
3748 - update lconf's list of debug and subsystem masks
3749 - fix lfs find --obd (2510)
3750 - /proc tunable for disabling filter read caching (2591)
3751 - stop rpm packages from altering slapd.conf (2301)
3752 - disable nagle in the socknal under 0conf (2578)
3753 - choose mds inode size based on stripe count (2572)
3754 - fix kernel-source rpm problems (2516)
3756 - add --disable-doc to avoid pdf generation (2421)
3757 - update documentation, tests, type-os, comments
3758 - avoid format warnings on ia64
3759 - remove the TOE NAL
3760 - tiny code cleanups by removing unused fields
3762 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3765 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3766 - remove the most blatant lies from BUILDING file (2371)
3767 - change default debug level to reasonable production setting
3768 - reduce client side cache size to reduce cache flush time
3769 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3770 - make TCP zerocopy and pinger support enabled by default (2476)
3771 - sync writes completed after process exits caused crashes (2319)
3772 - maintain correct mount count on the MDS (2356)
3773 - backout 1557, because 2316 wasn't really fixed
3774 - better file I/O statistics gathering in /proc
3775 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3776 - another kernel patch to fix zero-copy TCP function export
3777 - don't take duplicate lock when processing re-sent getattr (2420)
3778 - lctl uses obd_self_export instead of creating new conn (2353)
3779 - MDS/OST recovery case which requires object creation asserted (2425)
3780 - move lfs from /usr/sbin to /usr/bin in packages
3781 - fix race between mds_client_add and mds_client_free (2417)
3782 - use kmalloc instead of slabs in portals (2430)
3783 - don't create duplicate records when a failover MDS is present (2442)
3784 - remove unnecessary mount age check (2332)
3785 - don't remove directory inodes from locks prematurely (2451)
3786 - don't break if MDS service name is the same as hostname (2103)
3787 - fix races in client write RPC generation when cache full (2482)
3789 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3792 - remove now-unused request->rq_obd (278)
3793 - if an allocation fails, print out how much memory we've used (1933)
3794 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3795 - add the "configurable stack size" patch to most series files (1256)
3796 - ability to write large log records, for 100+ OST configs (2306)
3797 - fix NULL deref when filter_prep fails (2314)
3798 - fix operator precedence error in filter_sync
3799 - dynamic allocation of socknal TX descriptors (2315)
3800 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3801 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3802 - ensure that CWARN messages in Portals always get to the syslog
3803 - __init/__exit are not for prototype decls (ldlm_init/exit)
3804 - x86-64 compile warning fixes
3805 - fix gateway LMC keyword conflict (2318)
3806 - fix MDS lock inversions in getattr/reint paths (1844)
3807 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3808 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3809 - DLM race condition prevented some lock evictions (2328)
3810 - ENOMEM detection and retry on socknal sends (2230)
3811 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3812 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3813 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3814 - better error messages when a client is rejected during recovery (1505)
3815 - avoid cancelling locks which were never granted, after failure (2330)
3816 - fix i_sem/journal inversion in mds_client_add (2333)
3817 - fix truncate/getattr lock cycle deadlock (2334)
3818 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3819 - fix two rare exit paths which could leak an l_lock() ref (2321)
3820 - fencepost error in MDS/OST orphan recovery (2226)
3821 - make log record alignment 8 bytes (1988)
3822 - lstripe now fails when requested offset > ost_count (2237)
3823 - ensure that all kernel series have a complete list.h (1607)
3824 - fix crashes in special-file operations (2316)
3825 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3827 - allow configurable automake binary, for testing new versions
3828 - small update to the lfs documentation
3830 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3832 * fix negative export reference count in fsfilt_sync (2312)
3834 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3835 * release candidate 0.9.1
3837 - orphans are moved into the PENDING directory for possible recovery
3838 - replayed opens now open by fid for orphan/rename safety (1042)
3839 - last close of an orphan inode generates a transno (683)
3840 - chdir() and mount() now pin the directory entry (1020)
3841 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3842 - discard very old requests without processing them (1502)
3843 - remove some common, well-understood CERRORs (1505)
3844 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3845 - clear "grant" flags in OST replies until OST grant code lands (1644)
3846 - fix read performance by not clobbering i_blksize on client (1598)
3847 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3848 - make LRU size a /proc tunable, clears locks when reduced (707)
3849 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3850 - server locks take references on exports now (1558)
3851 - build fixes for 2.4.20-rh trees (1663)
3852 - return an error from lov_create if all OSCs are inactive (1751)
3853 - fix import levels when a reconnect happens without a timeout (1597)
3854 - exit early from mds_open if we get a lookup error (1749)
3855 - partial page read at EOF wouldn't wait for disk before sending (1642)
3856 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3857 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3858 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3859 - write/truncate lock inversion (1639)
3860 - Don't auto-load obdclass, portals modules during cleanup (1495)
3861 - fix timestamps from jumping to "now" (1763)
3862 - extra journal assertions (1648)
3863 - add an extra multiunlink test (1771)
3864 - fix read_record/write_record API (1776)
3865 - fix leak of offset_extent, possible incorrect i_size later (1772)
3866 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3867 - Debugging for the kqswnal_get_idle_tx problems (1820)
3868 - Allow recovery to be attempted multiple times (1536)
3869 - Write out MDS last_rcvd file after it is first created (1600)
3870 - Fix tx_descriptor leak in failed transmit situations (1827)
3871 - ext3 journaling fixes for assertion failure after IO error (1871)
3872 - class_export_put() on freed export after completion AST error (1896)
3873 - Fix revalidate looping in VFS (1322)
3874 - Don't access a freed export during MDS_REINT timeout (1521)
3875 - Add open-unlink recovery support on the MDS (1673,1764)
3876 - Return an error if no MDS data was read from last_rcvd (1946)
3877 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3878 - Don't setup a disk device that doesn't match exported UUID (317)
3879 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3880 - avoid committing NULL handle in force close
3881 - local.sh is now a one-stripe LOV configuration
3882 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3883 - fix ext3 htree duplicate directory entry corruption (1516)
3884 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3885 - update to new LOV EA format (2097)
3886 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3888 2003-06-15 Phil Schwan <phil@clusterfs.com>
3891 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3892 - per-import/export recovery handling (958, 931, 959)
3893 - multiple last-rcvd slots, for serving multiple FSes (949)
3894 - connections are again shared between multiple imp/exports (963, 964)
3895 - "umount -f" would hang if any requests needed to be sent (393, 978)
3896 - avoid pinning large req buffer by copying for queued messages (989)
3897 - add "uuid" to "lctl device" command to help upcalls (991)
3898 - "open" RPCs with transnos would confuse recovery counters (1037)
3899 - do proper endian conversion of all wire messages (288, 340, 891)
3900 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3901 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3902 - fix ptlrpc_connection leak in target_handle_connect (1174)
3903 - fix import refcounting bug in OST and MDS cleanup (1134)
3904 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3905 - fix very unlikely obd_types race condition (501)
3906 - remove osc_open hack for echo_client (1187)
3907 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3908 - a failure in read_inode2 leads to deadlock (1139)
3909 - cancel ack-locks as soon as transaction is committed (1072)
3910 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3911 - make sure to commitrw after any preprw to avoid deadlock (1162)
3912 - failing to execute a file in a lustre FS would lock inode (1203)
3913 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3914 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3915 - obd_uuid2tgt would walk off the end of the list (1255)
3916 - on IA64 the osc would give portals incorrect bulk size (1258)
3917 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3918 - fix lock inversion caused by new llite matching code (1282)
3919 - limit the number of dirty pages on a client to 10MB (1286)
3920 - timed out locks were not being corrected cancelled (1289)
3921 - fix O_DIRECT above 4GB on IA-32 (1292)
3922 * major user-visible changes
3923 - fail out/fail over policy now controlled by the upcall (993)
3925 - add OBD_PING to check server availability and failure (954)
3926 - lustre messages are now sent in sending host order (288, 340, 891)
3927 - add eadatalen to MDS getattr reply (340)
3928 - OST read replies may contain second buffer, with per-page status (593)
3930 2003-03-11 Phil Schwan <phil@clusterfs.com>
3933 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3934 - failed open()s could cause deadlock; fixed (867, 869)
3935 - stop cancelling OST locks when files are closed (481)
3936 - overlapping XID spaces caused network corruption (851, 853)
3937 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3938 - setattr_raw added, to do single-RPC, server-side setattrs
3939 - lmc/lconf syntax change for OST UUIDs
3940 - fix crashy race condition between ptlrpc_free_req and osc_close
3941 - don't use request in mdc_enqueue if we hit a timeout (889)
3942 - don't set the inode i_size for regular files from the MDS (896)
3943 - handle out of order completion AST (842)
3944 - don't LBUG if a lock request times out after receiving AST (913)
3945 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3946 - if a bad lock AST arrives, send an error instead of dropping entirely
3947 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3948 - fix leak in bulk IO when only partially completed (899, 900, 926)
3949 - fix O_DIRECT for ia64 (55)
3950 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3951 - C-z after timeout could hang a process forever; fixed (977)
3953 - client-side I/O cache (678, 924, 929, 941, 970)
3955 - READPAGE and SETATTRs which don't take server-side locks get
3958 2003-02-11 Phil Schwan <phil@clusterfs.com>
3961 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3962 - Fix fsfilt_extN_readpage() to read a full page of directory
3963 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3964 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3965 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3966 - Most intent operations were removed, in favour of a new RPC mode
3967 that does a single RPC to the server and bypasses most of the VFS
3968 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3969 - Aggressively cancel local locks on DLM servers
3970 - mds_reint_unlink sends EA to the client if it's the last nlink.
3971 client uses that EA to unlink OST objects.
3972 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3973 - recursive symlinks were fixed (439)
3974 - fixed NULL deref in DEBUG_REQ
3975 - filter_update_lastobjid no longer calls sync, which annoyed extN
3976 - fixed multi-client small-writes to a single file problem (445)
3977 - fixed mtime updates during file writes (607)
3978 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3979 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3980 - fixed Linux OST size reporting problem (444, 656)
3981 - OST now updates object mtime with writes or setattr (607, 619)
3982 - client verifies file size before zeroing page past EOF (445)
3983 - OST now writes last allocated objid to disk with allocation (108)
3984 - LOV on echo now works (409)
3986 - mds_reint_unlink sends a new buffer, with the EA included. this
3987 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3988 set if a regular file was being unlinked, and it was the last link
3989 - use PtlGet from the target for bulk writes (315)
3990 - OST now updates object mtime with writes or setattr (607, 619)
3991 - LDLM now has a grant-time callback to revalidate locked items, if
3993 - Many MDS operations were reorganized to combat race conditions
3995 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3996 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3998 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4001 - Fully reactivate OST imports after reconnection (512, others)
4002 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4003 - More graceful error handling for truncating on dead OST (515)
4004 - Don't error out unless we're actually accessing dead stripes (474)
4005 - Fix garbage sizes when stripes are missing (410)
4006 - LRU counters were broken, causing constant lock purge (433, 432)
4007 - garbage on read from stripes with failed OSTs (441)
4008 - mark OSCs as active before reconnecting during recovery (438)
4009 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4010 - lfind did not preserve OST order in output (443)
4011 - symlinks cause hung clients, incorrect data (439)
4012 - stop dereferencing request after dropping refcount (457)
4013 - don't LASSERT(spin_is_locked) on non-SMP (455)
4014 - fixes for many rename() bugs
4015 - fstat didn't correctly synchronize attributes (399)
4016 - server must handle lock cancellation during blocking AST prep (487)
4017 - bulk descriptors were free()d too soon (511)
4018 - fix paths in lconf, which would load incorrect modules (451, 507)
4019 - fix confusing lconf 'host not found' error message (386)
4020 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4021 - fix race condition in mdc_blocking_ast() for inode access (526)
4022 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4023 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4024 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4025 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4026 - fix lstripe in several ways (kernel side) (527)
4027 - fix request leak in ldlm_cli_enqueue (262)
4028 - incorrect OSC was marked inactive after OST failure
4029 - call mds_fs_cleanup before unmounting filesystem (524)
4030 - fix races between taking ns_lock and ldlm_lock_change_resource
4031 - fix races updating LOV export open file list
4032 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4033 - fix recovery NULL deref in ldlm_cli_cancel_unused
4034 - fix some DLM races by using new hash table for lock handles (419)
4035 - permit the client to specify desired inodes, at replay
4036 - duplicate requests when we queue them for replay reintegration
4037 - fix last_rcvd offset calculation
4038 - sync after each recovered transaction, so we always make progress
4039 - never, not always, ERESTART requests without transnos
4040 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4042 - skip replay if the MDS says that the client is already connected
4043 - don't check for a recovery-enabled export to match lctl's UUID
4044 - don't INC_USE_COUNT for phantom exports
4045 - don't crash when cleaning up phantom exports (567)
4046 - don't double-finish or set replay data for errored mdc_open requests
4047 - abort requests when they time out, so we don't get old replies
4048 - send/receive replies for AST messages again
4049 - if the client says that it doesn't have the lock, cancel it on the
4051 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4052 mark it as destroyed, it will all work out when decref is called
4053 - fix module use counts (22, 581)
4055 - ASTs now expect a reply (server cancels lock on error reply)
4057 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4060 - fix many simultaneous client startup (392)
4061 - fix dentry->d_it clobbering
4062 - credentials weren't being shipped for readdir/getattr operations
4063 - remove invalid assertions triggered during some concurrent MD
4065 - proper Lustre versions added (336, 389)
4066 - fix memory leak for create error case (398)
4067 - fix LOV locking bug that would get cli/srv out of sync
4068 - fix echo client over LOV (409)
4069 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4070 - fix double-O_EXCL intent crash (424)
4071 - avoid sending multiple lock CANCELs (352)
4073 - MDS can do multi-client recovery (modulo bugs in new code)
4075 - many updates, edits, cleanups
4077 2002-11-18 Phil Schwan <phil@clusterfs.com>
4080 - fix null d_it dereference (346)
4081 - fix full OST/dbench hang (333)
4082 - fix permission problem with file removal (286)
4083 - fix removal of OSCs from LOV when they fail
4084 - fix NULL deref during bulk timeout (214)
4085 - fix problems related to multiple filesystems on one MDS (241)
4086 - fixed serious subtle metadata locking bugs
4087 - free locks on clients when inodes are removed due to memory
4089 - fix inode pointer in lock data (285)
4090 - partial support for multiple MDS on a single host (241)
4091 - data locks weren't cancelled at clear_inode time (290, 311)
4092 - intent locks could lead to unbounded lock growth (205)
4093 - added a maximum lock count, an LRU list, and a flusher
4094 - fix multiple rename (365)
4095 - properly abstracted the echo client
4096 - OSC locked 1 byte too many; fixed
4097 - rewrote brw callback code:
4098 - fixed recovery bugs related to LOVs (306)
4099 - fixed too-many-pages-in-one-write crash (191)
4100 - fixed (again) crash in sync_io_timeout (214)
4101 - probably fixed callback-related race (385)
4103 - Add capability to MDS protocol
4104 - LDLM cancellations and callbacks on different portals
4106 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4109 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4110 - defer open object destruction to close time (601981, 138)
4111 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4112 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4113 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4114 - don't call obd_setattr in ll_file_release for destroyed objects
4115 * protocol change to lustre_msg: move |version| and add |flags|
4116 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4117 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4118 * added replay of create, unlink, link and rename operations during
4119 MDS failover; recovery should be much more robust now
4120 * remove failed OSCs from LOVs (only lov_create uses this so far)
4121 * the lustre-HOWTO was brought (more) up to date (582544)
4123 2002-10-23 Phil Schwan <phil@clusterfs.com>
4126 - in-use dentries weren't being reused properly (617851)
4127 - prevent multiple LDLM setup (599178)
4128 - fix LOV size calculations for truncate (617853)
4129 - fix client handling of MDS intent errors (POSIX)
4130 - fix permission bug in lovstripe.c test (624321)
4131 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4132 - truncate past end of file could corrupt data
4133 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4134 - a race in recovery could return ETIMEDOUT to apps (623947)
4135 - building outside the source directory was fixed
4136 * the lustre-HOWTO was brought (more) up to date (582544)
4137 * major progress was made on recovery functionality
4139 2002-10-10 Phil Schwan <phil@clusterfs.com>
4142 - recovery deadlock fix
4143 - rm -rf causes LBUG fix (617817)
4144 - file open by multiple tasks fix (618962)
4145 - directory permissions bugs (602707 and 620007)
4146 - journal_stop fixed with locking (611313)
4147 - O_APPEND failures resolved (618273, perhaps 614459)
4148 - lconf PATH fix (619770)
4149 - IA64 build fix (621450)
4150 - RPC buffer sizes scale with amount of memory
4152 2002-10-01 Phil Schwan <phil@clusterfs.com>
4155 - locks would be cancelled without throwing away data pages,
4156 resulting in inconsistent data (605627)
4157 - inode attributes were not always being refreshed (605627, 612449)
4158 - lconf now continues to cleanup after lctl reports an error
4159 - MDS now enforces user permissions (602707)
4160 - lprocfs cleanup fixed, but not yet enabled (614157)
4161 - fixed infinite server hang, should a client not respond to an AST
4162 - avoid going into recovery if user calls readlink() with a buffer
4163 that's too small (613941)
4164 - AST RPCs no longer require replies (614867) -- this may be changed
4165 - don't crash server if client sends an IOV that's too big (611336)
4166 - fixed lock conversion deadlock (611892)
4167 - fixed the following of symlinks (614622)
4168 * recovery: the server can remove locks from a client that dies, other
4169 clients can make progress
4170 * more extN patch fixes
4171 * compile-time configurable ptlrpc buffer allocations
4173 - collaborative read cache document
4174 - Lustre Lite Performance CDR document-in-progress
4176 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4179 - fix typo in patch-2.4.18
4181 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4184 - clear ptlrpc request each time in handle_incoming_request()
4185 - unlink of files now destroys the object on the OST
4187 2002-09-19 Peter Braam <braam@clusterfs.com>
4189 * add hard link support
4190 * change obdfile creation method
4191 * kernel patch changed
4193 2002-09-19 Peter Braam <braam@clusterfs.com>
4196 - stack overflow bug in extN fixed
4198 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4200 * documentation updates
4201 - add man pages for config tools
4202 - update tests/README to describe testing with new config tools
4203 - finish metadata API descriptions
4204 * bug fixes and cleanups
4205 - statfs workaround for 16TB limit
4206 - LOV stripe allocation improved, can stripe on subset of OSTs
4207 - LOV file size/IO offset was wrong for files > 4GB in size
4208 - object EA data was being dropped, caused files to be unreadable
4209 - memory overflow with non-LOV OST caused memory corruption
4210 - fixed regression tests to work with new config tools, obdfilter
4211 - fixed bug when directory size became larger than 1 block
4212 - fixed bug (for single client case) when PWD was deleted
4213 - invalidate local directory pages when doing intent-based ops
4214 - avoid LDLM oops when lock callback contained bad data
4216 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4218 * documentation updates
4219 * bug fixes and cleanups
4220 - configuration tools
4223 - 64-bit compile warnings
4224 - 64-bit internal statfs data
4226 * test_brw on persistent OST devices
4228 * lprocfs (disabled)
4230 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4232 * documentation updates
4233 * bug fixes and cleanups
4234 * configuration tools
4236 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4239 * many small fixes to 0_5_4
4240 * io/network handling
4241 * thinkos in MDS operations
4243 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4246 * crucial basic fixes to 0.5.3
4247 * IOR, Iozone work over Elan
4250 2002-08-07 Phil Schwan <phil@clusterfs.com>
4251 * version 0_5_3, our first alpha
4252 * we use the new Portals iovs
4253 * documentation updates
4254 * bug fixes and cleanups
4255 * small changes in the DLM wire protocol
4257 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4258 * version 0_5_1 with some initial stability,
4259 * locking on MD and file I/O.
4260 * documentation updates
4261 * several bug fixes since 0.5.0
4262 * small changes in wire protocol
4264 2002-07-18 Phil Schwan <phil@clusterfs.com>
4266 * delivered as Lustre Light Alpha
4267 * fixed a crash after handling invalid MDS requests
4268 * fixed directory pages for architectures with non-4k pages sizes
4270 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4271 * release version v0_4_4
4272 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4274 * Fixes a number of interruption problems with OST operations.
4275 * Update documentation for portals header changes
4276 * Move all wire protocol structs/defines to lustre_idl.h
4277 * Fixes symlink length bug.
4278 * Add tcpdump to repository.
4280 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4281 * release version v0_4_3
4282 * Fixes statfs for inodes on extN.
4283 * Fixes bug in runtests which would delete /etc/hosts.
4284 * Use 64-bit object IDs wherever possible (not into VFS though)
4285 Remove ost_get_info, which is unused by lustre, and out of date.
4287 2002-07-03 Peter Braam <braam@clusterfs.com>
4288 * release version v0_4_2 Fixes a lookup error (type not passed)
4289 * move forward to head of Portals
4290 * move forward to latest Lustre kernel
4292 2002-06-25 Peter Braam <braam@clusterfs.com>
4293 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob