1 tbd Sun Microsystems, Inc.
5 2.6.16.54-0.2.3 (SLES 10),
6 2.6.18-53.1.4.el5 (RHEL 5).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.40.4-cfs1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
15 Severity : enhancement
17 Description: Add rhel5 support to HEAD.
21 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
22 Details : If an OST is down the MDS will hang indefinitely in
23 obd_statfs_rqset() waiting for the statfs data. While for
24 MDS QOS usage of statfs, it should not stuck in waiting.
26 Severity : enhancement
28 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
30 Severity : enhancement
32 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
36 Description: make mgs_setparam() handle fsname containing dash
37 Details : fsname containing a dash does not work with lctl conf_param
39 Severity : enhancement
41 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
43 Severity : enhancement
45 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
49 Frequency : rare, at shutdown
50 Description: access already free / zero obd_namespace.
51 Details : if client_disconnect_export was called without force flag set,
52 and exist connect request in flight, this can produce access to
53 NULL pointer (or already free pointer) when connect_interpret
54 store ocd flags in obd_namespace.
58 Frequency : only at startup
59 Description: not alloc memory with spinlock held.
60 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
63 Severity : enhancement
65 Description: make lustre randomly fail allocating memory
66 Details : Make lustre randomly failed allocating memory for testing purpose.
68 Severity : enhancement
70 Description: lost problems with lov objid file
71 Details : Fixes some scability and access to not inited memory problems
72 in work with lov objdid file.
77 Description: lfs find does not continue on file error
78 Details : Continue other files processing when a file/dir is absent.
82 Description: Inconsistent usage of lustre_pack_reply()
83 Details : Standardize the usage of lustre_pack_reply() such that it
84 always generate a CERROR on failure.
87 Frequency : occasional
89 Description: MDS hang or stay in waiting lock
90 Details : If client receive lock with CBPENDING flag ldlm need send lock
91 cancel as separate rpc, to avoid situation when cancel request
92 can't processed due all i/o threads stay in wait lock.
96 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
100 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
104 Description: Update to SLES9 kernel-2.6.5-7.287.3.
108 Description: Update to RHEL4 latest kernel.
110 Severity : enhancement
112 Description: Build SLES10 patchless client fails
113 Details : The configure was broken by run ./configure with
114 --with-linux-obj=.... argument for patchless client. When the
115 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
116 can't search header adequately. Use absolute path such as
117 -I($LINUX)/include instead.
121 Description: interrupt oig_wait produce painc on resend.
122 Details : brw_redo_request can be used for resend requests from ptlrpcd and
123 private set, and this produce situation when rq_ptlrpcd_data not
124 copyed to new allocated request and triggered LBUG on assert
125 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
126 ptlrpcd set if request is changed and can be safety changed to use
131 Description: LASSERT_{REQ,REP}SWAB macros are buggy
132 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
133 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
134 these macros and replace them with their difinitions instead.
138 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
139 Details : Update which_patch & target file for SLES10 latest kernel.
142 Description: Remove client patches from SLES 10 kernel.
143 Details : This causes SLES 10 clients to behave as patchless clients
144 even on a Lustre-patched (server) kernel.
146 Severity : enhancement
148 Description: self-adjustable client's lru lists
149 Details : use adaptive algorithm for managing client cached locks lru
150 lists according to current server load, other client's work
151 pattern, memory activities, etc. Both, server and client
152 side namespaces provide number of proc tunables for controlling
155 Severity : enhancement
157 Description: light-weight GSS support
158 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
159 and reduce performance overhead.
161 Severity : enhancement
163 Description: Linux keyring support
164 Details : Support using service of Linux keyring for Lustre GSS internal
165 context refresh/cache mechanism.
169 Description: Fix errors in lfs documentation
170 Details : Fixes man pages
174 Description: don't use GFP_* in generic Lustre code.
175 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
179 Description: obdclass is limited by single OBD_ALLOC(idarray)
180 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
184 Description: warnings with build patchless client with vanila 2.6.19 and up
185 Details : change old ctl_table style and replace ctl_table/ctl_table_header
186 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
190 Frequency : Only for SLES
191 Description: sanity_quota fail test_1
192 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
193 file, which makes TSTID[2] inunique.
197 Frequency : Only for SLES9
198 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
199 Details : Update target/ChangeLog/which_patch .
203 Frequency : Only for test_10 in sanity-quota.sh
204 Description: a bug in quota test code
205 Details : checking a test flag in a wrong place
209 Frequency : Only for test_13 in sanity-quota.sh
210 Description: a bug in quota test script
211 Details : shouldn't check overfull allocated quota
215 Description: jbd statistics
216 Details : Port older jbd statistics patch for sles10
218 Severity : enhancement
220 Description: Add printing inode info into message about error in writepage.
225 Description: reply_lock_interpret crash due to race with it and lock cancel.
226 Details : Do not replay locks that are being cancelled. Do not reference
227 locks by their address during replay, just by their handle.
231 Frequency : When flocks are used.
232 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
233 Details : Flock locks might destroy just granted lock if it could be merged
234 with another existing flock, this is done in completion handler,
235 so teach ldlm_cli_enquque_fini that this is a valid case for
241 Description: Oops in read and write path when failing to allocate lock.
242 Details : Check if lock allocation failed and return error back.
246 Description: lstripe command fails for valid OST index
247 Details : The stripe offset is compared to lov->desc.ld_tgt_count
248 instead of lov->desc.ld_active_tgt_count.
252 Description: sanity.sh failed test 103
253 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
258 Description: df doesn't work properly if diskfs blocksize != 4K
259 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
263 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
264 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
268 Description: obd_health_check_timeout too short
269 Details : set obd_health_check_timeout as 1.5x of obd_timeout
273 Description: llapi_file_create() does not allow some changes
274 Details : add llapi_file_open() that allows specifying the mode and
275 open flags, and also returns an open file handle.
279 Description: merge and cleanup kernel patches.
280 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
284 Description: Add journal checksum support.(Kernel part)
285 Details : The journal checksum feature adds two new flags i.e
286 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
287 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
288 indicates that the commit block contains the checksum for
289 the blocks described by the descriptor blocks. Now commit
290 record can be sent to disk without waiting for descriptor
291 blocks to be written to disk. This behavior is controlled
292 using JBD2_FEATURE_ASYNC_COMMIT flag.
296 Description: OSS needs mutliple precreate threads
297 Details : Add ability to start more than one create thread per OSS.
301 Description: Sanity.sh test_65k failed: setstripe should have succeeded
302 Details : execute lfs setstripe on client
306 Description: mds_obd_create error creating tmp object
307 Details : When the user sets quota on root, llog will be affected and can't
308 create files and write files.
311 Frequency : Always on ia64 patchless client, and possibly others.
313 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
314 Details : This allows the patchless client to be loaded on architectures
319 Description: RedHat Update kernel for RHEL5
320 Details : Add the kernel config file for RHEL5.
324 Description: RedHat Update kernel for RHEL5
325 Details : Modify the kernel config file more closer RHEL5.
329 Description: Build failure against Centos5 (RHEL5)
330 Details : Define PAGE_SIZE when it isn't present.
334 Description: test_7 on sanity-quota.sh can't be used on separate machines
335 Details : add facet to handle it
339 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
340 Details : with NFS, the anon dentry's parent was set to itself in
341 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
342 corresponding dentry other than use rec->ur_name.
344 Severity : enhancement
346 Description: lfs setstripe enhancement
347 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
351 Description: enable data checksumming by default
352 Details : enable checksum by default, allow --disable-checksum
353 configure option and "-o nochecksum" mount option. Checksums
354 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
358 Description: lustre support for RHEL5
359 Details : Add support for RHEL5.
363 Description: Client eviction due to failover config
364 Details : after a connection loss, the lustre client should attempt to
365 reconnect to the last active server first before trying the
366 other potential connections.
370 Description: when mds and osts use different quota unit(32bit and 64bit),
371 quota will be released repeatly.
372 Details : void sending multiple quota reqs to mds, which will keep the
373 status between the reqs.
377 Description: osts not allocated evenly to files
378 Details : change the condition to increase offset_idx
381 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
383 Description: Data corruption for OSTs that are formatted larger than 2TB
385 Details : When generating the bio request for lustre file writes the
386 sector number would overflow a temporary variable before being
387 used for the IO. The data reads correctly from Lustre (which
388 will overflow in a similar manner) but other file data or
389 filesystem metadata may be corrupted in some cases.
393 Description: Tune the kernel for good SCSI performance.
394 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
395 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
400 Description: rewrite ext2-derived code in obdclass/uuid.c
401 Details : rewrite inherited code (uuid parsing code from ext2 utils)
402 from scratch preserving functionality.
406 Description: "lfs find -obd UUID" prints directories
407 Details : "lfs find -obd UUID" will return all directory names instead
408 of just file names. It is incorrect because the directories
409 do not reside on the OSTs.
413 Description: use i_size_read and i_size_write in 2.6 port
414 Details : replace inode->i_size access with i_size_read/write()
417 Frequency : when removing large files
419 Description: scheduling issue during removal of large Lustre files
420 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
421 It causes scheduling issues when removing large files (17TB in the
425 Frequency : only with liblustre clients on XT3
427 Description: evictions taking too long
428 Details : allow llrd to evict clients directly on OSTs
431 Frequency : only on ppc
433 Description: /proc/fs/lustre/devices broken on ppc
434 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
435 We should make sure the type of 'index' is loff_t and then cast
436 explicitly as needed below. Do not assign an explicitly cast
440 Frequency : only for rhel5
442 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
443 Details : Modify the target file & which_kernel.
445 Severity : enhancement
447 Description: omit set fsid for export NFS
448 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
449 as FS_REQUIRES_DEV for avoid problems with generate fsid.
453 Frequency : Always, for kernels after 2.6.16
454 Description: Fix warning idr_remove called for id=.. which is not allocated.
455 Details : Last kernels save old s_dev before kill super and not allow
456 to restore from callback - restore it before call kill_anon_super.
460 Description: buffer overruns could theoretically occur
461 Details : llapi_semantic_traverse() modifies the "path" argument by
462 appending values to the end of the origin string, and a
463 overrun may occur. Adding buffer overrun check in liblustreapi.
467 Description: Fix error on 'ls .' at the top of the Lustre mount.
468 Details : Don't revalidate dentry if it is a root dentry.
472 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
473 Details : Modify vm-tunables-rhel4.patch.
477 Description: Kernel config for 2.6.18-vanilla.
478 Details : Modify targets/2.6-vanilla.target.in.
479 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
480 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
481 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
482 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
486 Description: improve handling recoverable errors
487 Details : if request processig with error which can be recoverable on server
488 request should be resend, otherwise page released from cache and
493 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
494 Details : Modify target file & which_patch.
495 A flaw was found in the IA32 system call emulation provided
496 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
497 value could be stored in the %RAX register, which could trigger an
498 out-of-bounds system call table access. An untrusted local user
499 could exploit this flaw to run code in the kernel
500 (ie a root privilege escalation). (CVE-2007-4573).
504 Description: O_DIRECT bypasses client statistics.
505 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
506 read_ahead_stats were updated. copied the stats section from
507 osc_send_oap_rpc() into async_internal().
511 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
515 Description: change order of libsysio includes
516 Details : '#include sysio.h' should always come before '#include xtio.h'
520 Description: handle "serious error: objid * already exists" more gracefully
521 Details : If LAST_ID value on disk is smaller than the objects existing in
522 the O/0/d* directories, it indicates disk corruption and causes an
523 LBUG(). If the object is 0-length, then we should use the existing
524 object. This will help to avoid a full fsck in most cases.
526 Severity : enhancement
528 Description: adapt the lustre_config script to support the upgrade case
529 Details : Add "-u" option for lustre_config script to support upgrading 1.4
530 server targets to 1.6 in parallel.
534 Description: To avoid grant space > avaible space when the disk is almost
535 full. Without this patch you might see the error "grant XXXX >
536 available" or some LBUG about grant, when the disk is almost
538 Details : In filter_check_grant, for non_grant cache write, we should
539 check the left space by if (*left > ungranted + bytes), instead
540 of (*left > ungranted), because only we are sure the left space
541 is enough for another "bytes", then the ungrant space should be
542 increase. In client, we should update cl_avail_grant only there
543 is OBD_MD_FLGRANT in the reply.
547 Description: Update RHEL 4 kernel to fix local root privilege escalation.
548 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
549 described in CVE-2007-4573. This problem could allow untrusted
550 local users to gain root access.
553 Frequency : when using O_DIRECT and quotas
555 Description: Incorrect file ownership on O_DIRECT output files
556 Details : block usage reported by 'lfs quota' does not take into account
557 files that have been written with O_DIRECT.
562 Description: touch file failed when fs is not full
563 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
564 otherwise we can get ENOSP while fs is not full.
568 Description: parallel lock callbacks
569 Details : Instead of sending blocking and completion callbacks as separated
570 requests, adding them to a set and sending in parallel.
573 Frequency : only for Cray XT3
574 Bugzilla : 12829/13455
575 Description: Changing primary group doesn't change the group lustre assigns to
577 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
578 overridden with the primary group provided by the group upcall,
579 whereas the supplied fsgid can be trusted if it is in the list of
580 supplementary groups returned by the group upcall.
582 Severity : enhancement
584 Description: Allow masking D_WARNING, D_ERROR messages from console
585 Details : Console messages can now be disabled via lnet.printk.
589 Description: User code with malformed file open parameter crashes client node
590 Details : Before packing join_file req, all the related reference should be
591 checked carefully in case some malformed flags cause fake
592 join_file req on client.
594 --------------------------------------------------------------------------------
596 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
598 * Support for kernels:
599 2.6.5-7.283 (SLES 9),
600 2.6.9-55.EL (RHEL 4),
601 2.6.16.46-0.14 (SLES 10),
602 2.6.18.8 vanilla (kernel.org)
603 * Client support for unpatched kernels:
604 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
605 2.6.16 - 2.6.22 vanilla (kernel.org)
606 * Due to recently discovered recovery problems, we do not recommend
607 using patchless RHEL 4 clients with this or any earlier release.
608 * Recommended e2fsprogs version: 1.39.cfs8
609 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
610 * Starting with this release, the ldiskfs backing filesystem required
611 by Lustre is now in its own package, lustre-ldiskfs. This package
612 should be installed. It is versioned separately from Lustre and
613 may be released separately in future.
617 Description: block reactivating mgc import until all deactivates complete
618 Details : Fix race when failing back MDT/MGS to itself (testing)
621 Frequency : at statup only
623 Description: mds_lov_synchronize race leads to various problems
624 Details : simultaneous MDT->OST connections at startup can cause the
625 sync to abort, leaving the OSC in a bad state.
627 Severity : enhancement
629 Description: add optional extra BUILD_VERSION info
630 Details : add a new environment variable (namely LUSTRE_VERS) which allows
631 to override the lustre version.
634 Frequency : 2.6.18 servers only
636 Description: ll_kern_mount() doesn't release the module reference
637 Details : The ldiskfs module reference count never drops down to 0
638 because ll_kern_mount() doesn't release the module reference.
643 Description: server LBUG when using old ost_num_threads parameter
644 Details : Accept the old ost_num_threads parameter but warn that it
645 is deprecated, and fix an off-by-one error that caused an LBUG.
650 Description: Transient SCSI error results in persistent IO issue
651 Details : iobuf->dr_error is not reinitialized to 0 between two
655 Frequency : sometimes when underlying device returns I/O errors
657 Description: OSTs not going read-only during write failures
658 Details : OSTs are not remounted read-only when the journal commit threads
659 get I/O errors because fsfilt_ext3 calls journal_start/stop()
660 instead of the ext3 wrappers.
664 Description: poor connect scaling with increasing client count
665 Details : Don't run filter_grant_sanity_check for more than 100 exports
666 to improve scaling for large numbers of clients.
669 Frequency : SLES10 only
671 Description: sanity-quota.sh quotacheck failed: rc = -22
672 Details : Quotas cannot be enabled on SLES10.
675 Frequency : liblustre clients only
677 Description: getdirentries does not give error when run on compute nodes
678 Details : getdirentries does not fail when the size specified as an argument
679 is too small to contain at least one entry
681 Severity : enhancement
683 Description: Add LNET router traceability for debug purposes
684 Details : If a checksum failure occurs with a router as part of the
685 IO path, the NID of the last router that forwarded the bulk data
686 is printed so it can be identified.
691 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
692 Details : Due to a race condition, liblustre clients were occasionally
695 Severity : enhancement
697 Description: lfs setstripe use optional parameters instead of postional
700 Severity : enhancement
702 Description: Nanosecond timestamp support for ldiskfs
703 Details : The on-disk ldiskfs filesystem has added support for nanosecond
704 resolution timestamps. There is not yet support for this at
705 the Lustre filesystem level.
708 Frequency : during server recovery
710 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
711 Details : request with rq_no_resend flag not awake l_wait_event if they get a
715 Frequency : nfs export on patchless client
717 Description: connectathon hang when test nfs export over patchless client
718 Details : Disconnected dentry cannot be found with lookup, so we do not need
719 to unhash it or make it invalid
722 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
723 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
724 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
729 Description: open req refcounting wrong on reconnect
730 Details : If reconnect happened between getting open reply from server and
731 call to mdc_set_replay_data in ll_file_open, we will schedule
732 replay for unreferenced request that we are about to free.
733 Subsequent close will crash in variety of ways.
734 Check that request is still eligible for replay in
735 mdc_set_replay_data().
740 Description: disable writes to filesystem when reading health_check file
741 Details : the default for reading the health_check proc file has changed
742 to NOT do a journal transaction and write to disk, because this
743 can cause reads of the /proc file to hang and block HA state
744 checking on a healthy but otherwise heavily loaded system. It
745 is possible to return to the previous behaviour during configure
746 with --enable-health-write.
748 Severity : enhancement
750 Description: 64-bit inode version
751 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
752 to the inode. This will be required for version-based recovery.
757 Description: MDS fails to start if a duplicate client export is detected
758 Details : in some rare cases it was possible for a client to connect to
759 an MDS multiple times. Upon recovery the MDS would detect this
760 and fail during startup. Handle this more gracefully.
762 Severity : enhancement
764 Description: Add -o localflock option to simulate old noflock
766 Details : This will achieve local-only flock/fcntl locks
772 Description: log_commit_thread vs filter_destroy race leads to crash
773 Details : Take import reference before releasing llog record semaphore
778 Description: Wrong request locking in request set processing
779 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
780 imp_delayed_list, in this place should be used imp_lock.
783 Frequency : when reconnection
785 Description: Grant Leak when osc reconnect to OST
786 Details : When osc reconnect ost, OST(filter) should check whether it
787 should grant more space to client by comparing fed_grant and
788 cl_avail_grant, and return the granted space to client instead
789 of "new granted" space, because client will call osc_init_grant
790 to update the client grant space info.
793 Frequency : when client reconnect to OST
795 Description: Grant Leak when osc do resend and replay bulk write
796 Details : When osc reconnect to OST, OST(filter)should clear grant info of
797 bulk write request, because the grant info while be sync between
798 OSC and OST when reconnect, and we should ignore the grant info
799 these of resend/replay write req.
804 Description: Grant space more than avaiable left space sometimes.
805 Details : When then OST is about to be full, if two bulk writing from
806 different clients came to OST. Accord the avaliable space of the
807 OST, the first req should be permitted, and the second one
808 should be denied by ENOSPC. But if the seconde arrived before
809 the first one is commited. The OST might wrongly permit second
810 writing, which will cause grant space > avaiable space.
813 Frequency : when client is evicted
815 Description: Grant might be wrongly erased when osc is evicted by OST
816 Details : when the import is evicted by server, it will fork another
817 thread ptlrpc_invalidate_import_thread to invalidate the
818 import, where the grant will be set to 0. While the original
819 thread will update the grant it got when connecting. So if
820 the former happened latter, the grant will be wrongly errased
821 because of this race.
826 Description: Checking Stale with correct fid
827 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
828 to check whether it is stale, because sometimes, we want the
829 enqueue happened anyway, and op_data.fid2 will not be initialized.
831 Severity : enhancement
833 Description: update patchless client
834 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
837 Frequency : only with 2.4 kernel
839 Description: random memory corruption
840 Details : size of struct ll_inode_info is to big for union inode.u and this
841 can be cause of random memory corruption.
846 Description: Memory leak in recovery
847 Details : Lov_mds_md was not free in an error handler in mds_create_object.
848 It should also check obd_fail before fsfilt_start, otherwise if
849 fsfilt_start return -EROFS,(failover mds during mds recovery).
850 then the req will return with repmsg->transno = 0 and rc = EROFS.
851 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
852 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
853 be freed no matter whether fsfilt_commit success or not.
856 Frequency : only with huge count clients
858 Description: Prevents from taking the superblock lock in llap_from_page for
860 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
861 for llap_from_page call in ll_removepage prevents from taking the
862 superblock lock for a soon died page.
867 Description: Not check open intent error before release open handle
868 Details : in some rare cases, the open intent error is not checked before
869 release open handle, which may cause
870 ASSERTION(open_req->rq_transno != 0), because it tries to release
871 the failed open handle.
876 Description: Set cat log bitmap only after create log success.
877 Details : in some rare cases, the cat log bitmap is set too early. and it
878 should be set only after create log success.
882 Description: Accessing a block bevice can re-enable I/O when Lustre is
883 tearing down a device.
884 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
889 Description: service threads may hog cpus when there are a lot of requests
890 Details : Insert cond_resched to give other threads a chance to use some CPU
895 Description: the cat log was not initialized in recovery
896 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
897 unlink log on mds will not be initialized until mds post
898 recovery. And also in mds post recovery, the unlink log will
899 initialization will be done asynchronausly, so there will be race
900 between add unlink log and unlink log initialization.
904 Description: brw_stats were being printed incorrectly
905 Details : brw_stats were being printed as log2 but all of them were not
906 recorded as log2. Also remove some code duplication arising from
907 filter_tally_{read,write}.
911 Frequency : rare, only in recovery.
912 Description: ASSERTION(req->rq_type != LI_POISON) failed
913 Details : imp_lock should be held while iterating over imp_sending_list for
914 prevent destroy request after get timeout in ptlrpc_queue_wait.
918 Description: replay-single.sh test 52 fails
919 Details : A lock's skiplist need to be cleanup when it being unlinked
920 from its resource list.
924 Description: Short directio read returns full requested size rather than
926 Details : Direct I/O operations should return actual amount of bytes
927 transferred rather than requested size.
929 Severity : enhancement
931 Description: metadata RPC reduction (e.g. for rm performance)
932 Details : decrease the amount of synchronous RPC between clients and servers
933 by canceling conflicing lock before the operation on the client
934 and packing thier handles into the main operation RPC to server.
936 Severity : enhancement
938 Description: Async OSC create to avoid the blocking unnecessarily.
939 Details : If a OST has no remain object, system will block on the creating
940 when need to create a new object on this OST. Now, ways use
941 pre-created objects when available, instead of blocking on an
942 empty osc while others are not empty. If we must block, we block
943 for the shortest possible period of time.
945 Severity : enhancement
947 Description: refine locking for avoid write wrong info into lov_objid file
948 Details : fix possible races with add new target and write/update data in
951 --------------------------------------------------------------------------------
953 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
958 Frequency : on some architectures
960 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
961 Details : Uninitialized flags sometimes cause configuration commands to
965 Frequency : patchless clients only
967 Description: missing __iget() symbol export
968 Details : The __iget() symbol export is missing. To avoid the need for
969 this on patchless clients the deathrow inode reaper is turned
970 off, and we depend on the VM to clean up old inodes. This
971 dependency was during via the fix for bug 12181.
973 --------------------------------------------------------------------------------
975 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
977 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
978 INTEROPERATE with older versions automatically. In many cases a
979 special upgrade step is needed. Please read the
980 user documentation before upgrading any part of a live system.
981 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
982 version will not interoperate with 1.6 betas before beta5 (1.5.95).
983 * WARNING: Lustre configuration and startup changes are required with
984 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
990 Severity : enhancement
992 Description: MountConf
993 Details : Lustre configuration is now managed via mkfs and mount
994 commands instead of lmc and lconf. New obd types (MGS, MGC)
995 are added for dynamic configuration management. See
996 https://mail.clusterfs.com/wikis/lustre/MountConf for
999 Severity : enhancement
1001 Description: dynamic OST addition
1002 Details : OSTs can now be added to a live filesystem
1004 Severity : enhancement
1006 Description: startup order invariance
1007 Details : MDTs and OSTs can be started in any order. Clients only
1008 require the MDT to complete startup.
1010 Severity : enhancement
1012 Description: parallel, asynchronous orphan cleanup
1013 Details : orphan cleanup is now performed in separate threads for each
1014 OST, allowing parallel non-blocking operation.
1016 Severity : enhancement
1018 Description: optimized stripe assignment
1019 Details : stripe assignments are now made based on ost space available,
1020 ost previous usage, and OSS previous usage, in order to try
1021 to optimize storage space and networking resources.
1023 Severity : enhancement
1025 Description: Permanently set tunables
1026 Details : All writable /proc/fs/lustre tunables can now be permanently
1027 set on a per-server basis, at mkfs time or on a live system.
1029 Severity : enhancement
1031 Description: Lustre message v2
1032 Details : Add lustre message format v2.
1034 Severity : enhancement
1036 Description: client OST exclusion list
1037 Details : Clients can be started with a list of OSTs that should be
1038 declared "inactive" for known non-responsive OSTs.
1042 Description: SPEC SFS validation failure on NFS v2 over lustre.
1043 Details : Changes the blocksize for regular files to be 2x RPC size,
1044 and not depend on stripe size.
1046 Severity : enhancement
1048 Description: Multiple MD RPCs in flight.
1049 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1050 To avoid overly-overloading MDS, introduce a limit on number of
1051 MDS RPCs in flight for a single client and add /proc controls
1052 to adjust this limit.
1054 Severity : enhancement
1056 Description: client read/write statistics
1057 Details : Add client read/write call usage stats for performance
1058 analysis of user processes.
1059 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1060 file access. extents_stats shows chunk size distribution.
1061 extents_stats_per_process show chunk size distribution per
1064 Severity : enhancement
1066 Description: mds statistics
1067 Details : Add detailed mds operations statistics in
1068 /proc/fs/lustre/mds/*/stats.
1072 Description: Failure of copying files with lustre special EAs.
1073 Details : Client side always return success for setxattr call for lustre
1074 special xattr (currently only "trusted.lov").
1079 Description: Refcount LNET uuids
1080 Details : The global LNET uuid list grew linearly with every startup;
1081 refcount repeated list entries instead of always adding to
1084 Severity : enhancement
1086 Description: Dynamic service threads
1087 Details : Within a small range, start extra service threads
1088 automatically when the request queue builds up.
1091 Frequency : mixed-endian client/server environments
1093 Description: mixed-endian crashes
1094 Details : The new msg_v2 system had some failures in mixed-endian
1097 Severity : enhancement
1099 Description: Easy OST removal
1100 Details : OSTs can be permanently deactivated with e.g. 'lctl
1101 conf_param lustre-OST0001.osc.active=0'
1103 Severity : enhancement
1105 Description: MGS proc entries
1106 Details : Added basic proc entries for the MGS showing what filesystems
1109 Severity : enhancement
1111 Description: provide MGS failover
1112 Details : Added config lock reacquisition after MGS server failover.
1114 Severity : enhancement
1116 Description: add Linux 2.4 support
1117 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1121 Description: a large application tries to do I/O to the same resource and dies
1122 in the middle of it.
1123 Details : Check the req->rq_arrival time after the call to
1124 ost_brw_lock_get(), but before we do anything about
1125 processing it & sending the BULK transfer request. This
1126 should help move old stale pending locks off the queue as
1127 quickly as obd_timeout.
1130 Frequency : when an incorrect nid is specified during startup
1132 Description: ptlrpc connect to non-existant node causes kernel crash
1133 Details : LNET can't be re-entered from an event callback, which
1134 happened when we expire a message after the export has been
1135 cleaned up. Instead, hand the zombie cleanup off to another
1138 Severity : enhancement
1140 Description: plain/inodebits lock performance improvement
1141 Details : Grouping plain/inodebits in granted list by their request modes
1142 and bits policy, thus improving the performance of search through
1146 Frequency : only if OST filesystem is corrupted
1148 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1149 Details : for a short time RPCs with bulk IO are in the replay list,
1150 but replay of bulk IOs is unimplemented. If the OST filesystem
1151 is corrupted due to disk cache incoherency and then replay is
1152 started it is possible to trip an assertion. Avoid putting
1153 committed RPCs into the replay list at all to avoid this issue.
1156 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1158 Description: System hang on startup
1159 Details : This bug allowed the liblustre (e.g. catamount) client to
1160 return to the app before handling all startup RPCs. This
1161 could leave the node unresponsive to lustre network traffic
1162 and manifested as a server ptllnd timeout.
1164 Severity : enhancement
1166 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1167 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1168 boolean to control whether to print peer debug info when a
1169 client's RPC times out.
1172 Frequency : only for kernels with patches from Lustre below 1.4.3
1174 Description: Remove old rdonly API
1175 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1178 Frequency : only for devices with external journals
1180 Description: Set external device read-only also
1181 Details : During a commanded failover stop, we set the disk device
1182 read-only while the server shuts down. We now also set any
1183 external journal device read-only at the same time.
1186 Frequency : when upgrading from 1.4 while trying to change parameters
1188 Description: The wrong (new) MDC name was used when setting parameters for
1189 upgraded MDT's. Also allows changing of OSC (and MDC)
1190 parameters if --writeconf is specified at tunefs upgrade time.
1193 Frequency : when setting specific ost indicies
1195 Description: QOS code breaks on skipped indicies
1196 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1197 created with --index need not be sequential.
1199 Severity : enhancement
1201 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1202 Details : The uninit_groups feature works in conjunction with the kernel
1203 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1204 up the pass1 processing of e2fsck. This is a read-only feature
1205 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1206 mount filesystems that have had this feature enabled.
1208 Severity : enhancement
1210 Description: Improve multi-block allocation algorithm to avoid fragmentation
1211 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1212 allocation locality and avoid filesystem fragmentation.
1214 ------------------------------------------------------------------------------
1216 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1218 * Support for kernels:
1219 2.4.21-47.0.1.EL (RHEL 3)
1220 2.6.5-7.283 (SLES 9)
1221 2.6.9-42.0.10.EL (RHEL 4)
1222 2.6.12.6 vanilla (kernel.org)
1223 2.6.16.27-0.9 (SLES 10)
1224 * Recommended e2fsprogs version: 1.39.cfs5
1226 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1230 Frequency : occasional, depends on client load and configuration
1231 Bugzilla : 12181, 12203
1232 Description: data loss for recently-modified files
1234 Details : In some cases it is possible that recently written or created
1235 files may not be written to disk in a timely manner (this should
1236 normally be within 30s unless client IO load is very high).
1237 The problem appears as zero-length files or files that are a
1238 multiple of 1MB in size after a client crash or client eviction
1239 that are missing data at the end of the file.
1241 This problem is more likely to be hit on clients where files are
1242 repeatedly created and unlinked in the same directory, clients
1243 have a large amount of RAM, have many CPUs, the filesystem has
1244 many OSTs, the clients are rebooted frequently, and/or the files
1245 are not accessed by other nodes after being written.
1247 The presence of the problem can be detected by looking at
1248 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1249 smaller than the second (nr_unused) then dirty files will not
1250 be flushed automatically to disk. "sync; sleep 10" should be
1251 run several times on the node before unmounting it to update
1252 Lustre (this is also safe to run on nodes without this problem).
1254 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1255 that can cause this same problem, so customers using that kernel
1256 also need to update the kernel in addition to Lustre. In order
1257 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1259 It is normal that files written just before a client crash (less
1260 than 30s) may not yet have been flushed to disk, even for local
1264 Frequency : frequent on thin XT3 nodes
1266 Description: UUID collision on thin XT3 Linux nodes
1267 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1268 randomly, since we relied on an insufficiently-seeded PRNG.
1273 Description: OSS hangs after "All ost request buffers busy"
1274 Details : A deadlock between quota and journal operations caused OSS
1275 hangs after printing "All ost request buffers busy."
1278 Frequency : always on liblustre builds
1280 Description: Cleanup compiler warnings on liblustre
1283 Frequency : always on liblustre builds on XT3
1285 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1289 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1290 > 32000 subdirectories
1291 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1292 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1293 subdirectory count crosses 32000. This will aid e2fsck to
1294 correctly handle more than 32000 subdirectories.
1297 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1299 Description: System hang on startup
1300 Details : This bug allowed the liblustre (e.g. catamount) client to
1301 return to the app before handling all startup RPCs. This
1302 could leave the node unresponsive to lustre network traffic
1303 and manifested as a server ptllnd timeout.
1305 Severity : enhancement
1307 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1308 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1309 boolean to control whether to print peer debug info when a
1310 client's RPC times out.
1315 Description: make O_SYNC working on 2.6 kernels
1316 Details : 2.6 kernels use different method for mark pages for write,
1317 so need add a code to lustre for O_SYNC work.
1322 Description: Failure to close file and release space on NFS
1323 Details : Put inode details into lock acquired in ll_intent_file_open.
1324 Use mdc_intent_lock in ll_intent_open to properly
1325 detect all kind of errors unhandled by mdc_enqueue.
1330 Description: proc file read during shutdown sometimes raced obd removal,
1332 Details : Add lock to prevent obd access after proc file removal.
1335 Frequency : Only for files larger than 4GB on 32-bit clients.
1337 Description: improperly doing page alignment of locks
1338 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1339 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1344 Description: RPCs being resent when they shouldn't be
1345 Details : Some RPCs that should not be resent are being resent. This
1346 can cause inconsistencies in the RPC state machine. Do not
1347 resend such requests.
1350 Frequency : rare, only with NFS export
1352 Description: Crash on NFS re-export node
1353 Details : under very unusual load conditions an assertion is hit in
1354 ll_intent_file_open()
1357 Frequency : only if OST filesystem is corrupted
1359 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1360 Details : for a short time RPCs with bulk IO are in the replay list,
1361 but replay of bulk IOs is unimplemented. If the OST filesystem
1362 is corrupted due to disk cache incoherency and then replay is
1363 started it is possible to trip an assertion. Avoid putting
1364 committed RPCs into the replay list at all to avoid this issue.
1369 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1370 Details : Large single O_DIRECT read and write calls can fail to allocate
1371 a sufficiently large buffer to process the request. In case of
1372 allocation failure the allocation is retried with a smaller
1373 buffer and broken into smaller requests.
1375 Severity : enhancement
1377 Description: Add -o localflock option to simulate old noflock behaviour.
1378 Details : This will achieve local-only flock/fcntl locks coherentness.
1383 Description: versioning check is incomplete
1384 Details : Checking the version difference of client vs. server, report
1385 error if the gap is too big.
1390 Description: add support PG_writeback bit
1391 Details : add support for PG_writeback bit for Lustre, for more carefull
1392 work with page cache in 2.6 kernel. This also fix some deadlocks
1393 and remove hack for work O_SYNC with 2.6 kernel.
1395 Severity : enhancement
1397 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1398 Details : The uninit_groups feature works in conjunction with the kernel
1399 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1400 up the pass1 processing of e2fsck. This is a read-only feature
1401 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1402 mount filesystems that have had this feature enabled.
1404 Severity : enhancement
1406 Description: Improve multi-block allocation algorithm to avoid fragmentation
1407 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1408 allocation locality and avoid filesystem fragmentation.
1410 ------------------------------------------------------------------------------
1412 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1414 * Support for kernels:
1415 2.6.9-42.0.3.EL (RHEL 4)
1416 2.6.5-7.276 (SLES 9)
1417 2.4.21-47.0.1.EL (RHEL 3)
1418 2.6.12.6 vanilla (kernel.org)
1419 2.6.16.21-0.8 (SLES10)
1420 * Recommended e2fsprogs version: 1.39.cfs2-0
1422 * The backwards-compatible /proc/sys/portals symlink has been removed
1423 in this release. Before upgrading, please ensure that you change
1424 any configuration scripts or /etc/sysctl.conf files that access
1425 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1426 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1427 in advance of the upgrade on any system running Lustre 1.4.6 or
1428 newer, since /proc/sys/lnet was added in that version.
1429 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1433 Frequency : only when quota is used
1435 Description: avoid scanning export list for quota master
1436 Details : Change the algorithms to avoid scanning export list in order
1437 to improve the efficiency.
1440 Frequency : MDS failover only, very rarely
1442 Description: "went back in time" messages on mds failover
1443 Details : The greatest transno may be lost when the current operation
1444 finishes with an error (transno==0) and the client's last_rcvd
1445 record is over-written. Save the greatest transno in the
1446 mds_last_transno for this case.
1449 Frequency : always for specific kernels and striping counts
1451 Description: client may get "Matching packet too big" without ACL support
1452 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1453 when trying to access files in certain configurations. The
1454 clients should in fact be denied when mounting because they do
1455 not understand ACLs.
1458 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1460 Description: many clients connecting with IO in progress causes connect timeouts
1461 Details : Avoid synchronous journal commits to avoid delays caused by many
1462 clients connecting/disconnecting when bulk IO is in progress.
1463 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1464 OST_IO_PORTAL to avoid delays behind potentially many pending
1468 Frequency : occasionally with multiple writers to a single file
1470 Description: shared writes to file may result in wrong size reported by stat()
1471 Details : Allow growing of kms when extent lock is cancelled
1474 Frequency : always with random mmap IO to multi-striped file
1476 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1477 Details : Only if the hole is at the end of OST object so that kms is too
1478 small. Fix is to increase kms accordingly in ll_nopage.
1481 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1483 Description: writes to a missing object would leak memory on the OST
1484 Details : If there is an inconsistency between the MDS and OST filesystems,
1485 such that the MDS references an object that doesn't exist, writes
1486 to that object will leak memory due to incorrect cleanup in the
1487 error handling path, eventually running out of memory on the OST.
1492 Description: Creating too long symlink causes lustre errors
1493 Details : Check symlink and name lengths before sending requests to MDS.
1496 Frequency : only if flock is enabled (not on by default)
1498 Description: posix locks not released on fd closure on 2.6.9+
1499 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1500 kernels, this caused such locks not to be released on fd close and
1501 then assertions on fs unmount about still used locks.
1504 Frequency : MDS failover only, very rarely
1506 Description: clients may get ASSERTION(granted_lock != NULL)
1507 Details : When request was taking a long time, and a client was resending
1508 a getattr by name lock request. The were multiple lock requests
1509 with the same client lock handle and
1510 mds_getattr_name->fixup_handle_for_resent_request found one of the
1511 lock handles but later failed with ASSERTION(granted_lock != NULL).
1516 Description: handle->h_buffer_credits > 0, assertion failure
1517 Details : h_buffer_credits is zero after truncate, causing assertion
1518 failure. This patch extends the transaction or creates a new
1522 Frequency : NFS re-export or patchless client
1523 Bugzilla : 11179, 10796
1524 Description: Crash on NFS re-export node (__d_move)
1525 Details : We do not want to hash the dentry if we don't have a lock.
1526 But if this dentry is later used in d_move, we'd hit uninitialised
1527 list head d_hash, so we just do this to init d_hash field but
1528 leave dentry unhashed.
1531 Frequency : NFS re-export or patchless client
1533 Description: NFS exports has problem with symbolic link
1534 Details : lustre client didn't properly install dentry when re-exported
1535 to NFS or running patchless client.
1538 Frequency : NFS re-export or patchless client
1540 Description: Various nfs/patchless fixes.
1541 Details : fixes reuse disconected alias for lookup process - this fixes
1542 warning "find_exported_dentry: npd != pd",
1543 fix permission error with open files at nfs.
1544 fix apply umask when do revalidate.
1547 Frequency : occasional
1549 Description: Crash on NFS re-export node
1550 Details : calling clear_page() on the wrong pointer triggered oops in
1551 generic_mapping_read().
1554 Frequency : rarely, using O_DIRECT IO
1556 Description: unaligned directio crashes client with LASSERT
1557 Details : check for unaligned buffers before trying any requests.
1560 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1562 Description: stale data returned from RAID cache
1563 Details : If only a small amount of IO is done to the RAID device before
1564 reading it again it is possible to get stale data from the RAID
1565 cache instead of reading it from disk.
1568 Frequency : always for sles10 kernel
1570 Description: sles10 support
1571 Details : ll_follow_link: compile fixes and using of nd_set_link
1572 under newer kernels.
1575 Frequency : depends on arch, kernel and compiler version, always on sles10
1578 Description: recursive or deep enough symlinks cause stack overflow
1579 Details : getting rid of large stack-allocated variable in
1583 Frequency : depends on hardware
1585 Description: lustre write performance loss in the SLES10 kernel
1586 Details : the performance loss is caused by using of write barriers in the
1587 ext3 code. The SLES10 kernel turns barrier support on by
1588 default. The fix is to undo that change for ldiskfs.
1590 ------------------------------------------------------------------------------
1592 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1594 * Support for kernels:
1595 2.6.9-42.0.3EL (RHEL 4)
1596 2.6.5-7.276 (SLES 9)
1597 2.4.21-47.0.1.EL (RHEL 3)
1598 2.6.12.6 vanilla (kernel.org)
1602 Frequency : quota enabled and large files being deleted
1604 Description: releasing more than 4GB of quota at once hangs OST
1605 Details : If a user deletes more than 4GB of files on a single OST it
1606 will cause the OST to spin in an infinite loop. Release
1607 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1612 Description: statfs data retrieved from /proc may be stale or zero
1613 Details : When reading per-device statfs data from /proc, in the
1614 {kbytes,files}_{total,free,avail} files, it may appear
1615 as zero or be out of date.
1618 Frequency : systems with MD RAID1 external journal devices
1620 Description: lconf's call to blkid is confused by RAID1 journal devices
1621 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1622 of returning all block devices that match the journal UUID.
1625 Frequency : always, for aggregate stripe size over 4GB
1627 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1628 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1629 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1632 Frequency : always if "lfs find" used on a local file/directory
1634 Description: "lfs find" segfaults if used on a local file/directory
1635 Details : The case where a directory component was not specified wasn't
1636 handled correctly. Handle this properly.
1639 Frequency : always on ppc64
1641 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1642 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1643 perform endianness conversion.
1646 Frequency : rarely (truncate to non-zero file size after write under load)
1647 Bugzilla : 10730, 10687
1648 Description: Files padded with zeros to next 4K multiple
1649 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1650 it is possible that files that are truncated to a non-zero size
1651 immediately after being written are filled with zero bytes beyond
1652 the truncated size. No file data is lost.
1654 Severity : enhancement
1656 Description: Allow recovery/failover for liblustre clients.
1657 Details : liblustre clients were unaware of failover configurations until
1660 Severity : enhancement
1662 Description: user file locks should fail when not mounting with flock option
1663 Details : Set up an error-returning stub in ll_file_operations.lock field
1664 to prevent incorrect behaviour when client is mounted without
1665 flock option. Also, set up properly f_op->flock field for
1669 Frequency : always on ia64
1671 Description: "lfs df" loops on printing out MDS statfs information
1672 Details : The obd_ioctl_data was not initialized and in some systems
1673 this caused a failure during the ioctl that did not return
1674 an error. Initialize the struct and return an error on failure.
1677 Frequency : SLES 9 only
1679 Description: Error of copying files with lustre special EAs as root
1680 Details : Client side always return success for setxattr call for lustre
1681 special xattr (currently only "trusted.lov").
1684 Frequency : rarely on clusters with both ia64+i386 clients
1686 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1687 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1688 clients are doing concurrent writes to the same file it is
1689 possible that the ia64 clients may LASSERT because the OST
1690 extent locks are not PAGE_SIZE aligned. Ensure that grown
1691 locks are always aligned on the request boundary.
1694 Frequency : specific use, occasional
1696 Description: Overwriting in use executable truncates on-disk binary image
1697 Details : If one node attempts to overwrite an executable in use by
1698 another node, we now correctly return ETXTBSY instead of
1699 truncating the file.
1701 Severity : enhancement
1703 Description: Async OSC create to avoid the blocking unnecessarily.
1704 Details : If a OST has no remain object, system will block on the creating
1705 when need to create a new object on this OST. Now, ways use
1706 pre-created objects when available, instead of blocking on an
1707 empty osc while others are not empty. If we must block, we block
1708 for the shortest possible period of time.
1713 Description: chmod on Lustre root is propagated to other clients
1714 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1715 invalid by the follow_mount time.
1718 Frequency : liblustre clients only
1720 Description: Race in 'instant cancel' lock handling could lead to such locks
1721 never to be granted in case of SMP MDS
1722 Details : Do not destroy not yet granted but cbpending locks in
1726 Frequency : replay/resend of open
1728 Description: non null lock assetion failure in mds_intent_policy
1729 Details : Trying to replay/resend lockless open requests resulted in
1730 mds_open() returning 0 with no lock. Now it sets a flag if
1731 a lock is going to be returned.
1733 Severity : enhancement
1735 Description: Checksum enhancements
1736 Details : New checksum enhancements allow for resending RPCs that failed
1739 Severity : enhancement
1741 Description: Tunables on number of dirty pages in cacche
1742 Details : Allow to set limit on number of dirty pages cached.
1747 Description: client crash on unmount - lock still has references
1748 Details : In some error handling cases it was possible to leak a lock
1749 reference on a client while accessing a file. This was not
1750 harmful to the client during operation, but would cause the
1751 client to crash when the filesystem is unmounted.
1754 Frequency : specific case, rare
1756 Description: ETXTBSY on mds though file not in use by client
1757 Details : ETXTBSY is no longer incorrectly returned when attempting to
1758 chmod or chown a directory that the user previously tried to
1759 execute or a currently-executing binary.
1762 Frequency : extremely rare except on liblustre-based clients
1764 Description: Lustre space not freed when files are deleted
1765 Details : Clean up open-unlinked files after client eviction. Previously
1766 the unlink was skipped and the files remained as orphans.
1771 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1772 Details : In some cases it was possible to send a blocking callback to
1773 a client doing a glimpse, even though that client didn't get
1774 a lock granted. When the glimpse lock is cancelled on the OST
1775 the freed lock is left on the waiting list and corrupted the list.
1778 Frequency : all core dumps
1780 Description: Broke core dumps to lustre
1781 Details : Negative dentry may be unhashed if parent does not have UPDATE
1782 lock, but some callers, e.g. do_coredump, expect dentry to be
1783 hashed after successful create, hash it in ll_create_it.
1785 ------------------------------------------------------------------------------
1787 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1789 * Support for kernels:
1790 2.6.9-42.0.2.EL (RHEL 4)
1791 2.6.5-7.276 (SLES 9)
1792 2.4.21-40.EL (RHEL 3)
1793 2.6.12.6 vanilla (kernel.org)
1797 Frequency : always on RHEL 3
1799 Description: Number of open files grows over time
1800 Details : The number of open files grows over time, whether or not
1801 Lustre is started. This was due to a filp leak introduced
1802 by one of our kernel patches.
1804 ------------------------------------------------------------------------------
1806 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1808 * Support for kernels:
1809 2.6.9-42.EL (RHEL 4)
1810 2.6.5-7.276 (SLES 9)
1811 2.4.21-40.EL (RHEL 3)
1812 2.6.12.6 vanilla (kernel.org)
1817 Bugzilla : 5719, 9635, 9792, 9684
1818 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1819 Details : If a server is under heavy load and cannot reply to new
1820 connection requests before the client resends the (re)connect,
1821 the connection handling code can behave badly if two service
1822 threads are concurrently handing separate (re)connections from
1823 the same client. Add better locking to the connection handling
1824 code, and ensure that only a single connection will be processed
1825 for a given client UUID, even if the lock is dropped.
1827 Severity : enhancement
1829 Description: add TCP zero-copy support to kernel
1830 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1831 sends if the hardware supports scatter-gather and checksumming.
1832 This allows socklnd to do client-write and server-read more
1833 efficiently and reduce CPU utilization from skbuf copying.
1836 Frequency : only if NFS exporting from client
1838 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1839 Details : If setting attributes on a file created under NFS that had
1840 never been opened it would be possible to oops the client
1841 if the file had no objects.
1844 Frequency : always for liblustre
1846 Description: liblustre client does MDS+OSTs setattr RPC for each write
1847 Details : When doing a write from a liblustre client, the client
1848 incorrectly issued an RPC to the MDS and each OST the file was
1849 striped over in order to update the timestamps. When writing
1850 with small chunks and many clients this could overwhelm the MDS
1851 with RPCs. In all cases it would slow down the write because
1852 these RPCs are unnecessary.
1854 Severity : enhancement
1856 Description: allow number of MDS service threads to be changed at module load
1857 Details : It is now possible to change the number of MDS service threads
1858 running. Adding "options mds mds_num_threads={N}" to the MDS's
1859 /etc/modprobe.conf will set the number of threads for the next
1860 time Lustre is restarted (assuming the "mds" module is also
1861 reloaded at that time). The default number of threads will
1862 stay the same, 32 for most systems.
1867 Description: OST crash if filesystem is unformatted or corrupt
1868 Details : If an OST is started on a device that has never been formatted
1869 or if the filesystem is corrupt and cannot even mount then the
1870 error handling cleanup routines would dereference a NULL pointer.
1875 Description: NULL pointer deref in llap_from_page.
1876 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1877 incorrect) mapping if the page was truncated/reclaimed while it was
1878 searched for. Check for this condition and skip such pages when
1879 doing readahead. Introduce extra check to llap_from_page() to
1880 verify page->mapping->host is non-NULL (so page is not anonymous).
1883 Frequency : Sometimes when using sys_sendfile
1885 Description: "page not covered by a lock" warnings from ll_readpage
1886 Details : sendfile called ll_readpage without right page locks present.
1887 Now we introduced ll_file_sendfile that does necessary locking
1888 around call to generic_file_sendfile() much like we do in
1892 Frequency : with certain MDS communication failures at client mount time
1894 Description: NULL pointer deref after failed client mount
1895 Details : a client connection request may delayed by the network layer
1896 and not be sent until after the PTLRPC layer has timed out the
1897 request. If the client fails the mount immediately it will try
1898 to clean up before the network times out the request. Add a
1899 reference from the request import to the obd device and delay
1900 the cleanup until the network drops the request.
1903 Frequency : occasionally during client (re)connect
1905 Description: assertion failure during client (re)connect
1906 Details : processing a client connection request may be delayed by the
1907 client or server longer than the client connect timeout. This
1908 causes the client to resend the connection request. If the
1909 original connection request is replied in this interval, the
1910 client may trip an assertion failure in ptlrpc_connect_interpret()
1911 which thought it would be the only running connect process.
1914 Frequency : only with obd_echo servers and clients that are rebooted
1916 Description: kernel BUG accessing uninitialized data structure
1917 Details : When running an obd_echo server it did not start the ping_evictor
1918 thread, and when a client was evicted an uninitialized data
1919 structure was accessed. Start the ping_evictor in the RPC
1920 service startup instead of the OBD startup.
1922 Severity : enhancement
1923 Bugzilla : 10193 (patchless)
1924 Description: Remove dependency on various unexported kernel interfaces.
1925 Details : No longer need reparent_to_init, exit_mm, exit_files,
1926 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1929 Frequency : rare (only users of deprecated and unsupported LDAP config)
1931 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1932 Details : LDAP apparently contains 'lustreName' attributes instead of
1933 'name'. A simple remapping of the name is sufficient.
1936 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1938 Description: waiting_locks_callback trips kernel BUG if client is evicted
1939 Details : Running with the dump_on_timeout debug flag turned on makes
1940 it possible that the waiting_locks_callback() can try to dump
1941 the Lustre kernel debug logs from an interrupt handler. Defer
1942 this log dumping to the expired_lock_main() thread.
1944 Severity : enhancement
1946 Description: Support NFS exporting on 2.6 kernels.
1947 Details : Implement non-rawops metadata methods for NFS server to use without
1948 changing NFS server code.
1951 Frequency : very rare (synthetic metadata workload only)
1953 Description: two racing renames might cause an MDS thread to deadlock
1954 Details : Running the "racer" program may cause one MDS thread to rename
1955 a file from being the source of a rename to being the target of
1956 a rename at exactly the same time that another thread is doing
1957 so, and the second thread has already enqueued these locks after
1958 doing a lookup of the target and is trying to relock them in
1959 order. Ensure that we don't try to re-lock the same resource.
1962 Frequency : only very large systems with liblustre clients
1964 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1965 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1966 all OSTs in parallel. This allows the eviction of stale liblustre
1967 clients to proceed much faster than if they were done in series,
1968 and also offers similar improvements for other set_info RPCs.
1973 Description: excessive CPU usage during initial read phase on client
1974 Details : During the initial read phase on a client, it would agressively
1975 retry readahead on the file, consuming too much CPU and impacting
1976 performance (since 1.4.5.8). Improve the readahead algorithm
1977 to avoid this, and also improve some other common cases (read
1978 of small files in particular, where "small" is files smaller than
1979 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1984 Description: MDS crash when receiving packet with unknown intent.
1985 Details : Do not LBUG in unknown intent case, just return -EFAULT
1987 Severity : enhancement
1988 Bugzilla : 9293, 9385
1989 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1990 Details : Do not serialize getattr (non-intent version) and statfs.
1993 Frequency : occasional, when OST network is overloaded/intermittent
1995 Description: client evicted by OST after bulk IO timeout
1996 Details : If a client sends a bulk IO request (read or write) the OST
1997 may evict the client if it is unresposive to its data GET/PUT
1998 request. This is incorrect if the network is overloaded (takes
1999 too long to transfer the RPC data) or dropped the OST GET/PUT
2000 request. There is no need to evict the client at all, since
2001 the pinger and/or lock callbacks will handle this, and the
2002 client can restart the bulk request.
2005 Frequency : Always when mmapping file with no objects
2007 Description: client crashes when mmapping file with no objects
2008 Details : Check that we actually have objects in a file before doing any
2009 operations on objects in ll_vm_open, ll_vm_close and
2015 Description: Request leak when working with deleted CWD
2016 Details : Introduce advanced request refcount tracking for requests
2017 referenced from lustre intent.
2019 Severity : Enhancement
2021 Description: Cache open file handles on client.
2022 Details : MDS now will return special lock along with openhandle, if
2023 requested and client is allowed to hold openhandle, even if unused,
2024 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2025 closing files for every read/write openration.
2027 Severity : Enhancement
2029 Description: Cache open negative dentries on client when possible.
2030 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2031 negative dentries on lock revocation.
2036 Description: Remounting a client read-only wasn't possible with a zconf mount
2037 Details : It wasn't possible to remount a client read-only with llmount.
2039 Severity : enhancement
2040 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2041 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2042 MPICH is included to add Lustre-specific ADIO interfaces.
2043 This is based closely on the UFS ADIO layer and only differs
2044 in file creation, in order to allow the OST striping to be set.
2045 This is user-contributed code and not supported by CFS.
2050 Description: extended inode attributes (immutable, append-only) work improperly
2051 when 2.4 and 2.6 kernels are used on client/server or vice versa
2052 Details : Introduce kernel-independent values for these flags.
2054 Severity : enhancement
2057 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2058 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2059 granularity. Now, Fractional MB granularity is be supported,
2060 this is very useful for low memory system.
2062 Severity : enhancement
2064 Description: Getattr by fid
2065 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2066 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2067 have locks and avoids getattr rpc requests for it on every stat.
2070 Frequency : Always, for filesystems larger than 2TB
2072 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2073 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2074 filesystems larger than 2TB to be mounted with mballoc enabled.
2077 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2079 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2080 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2081 filesystems larger than 2TB is mounted then the kernel will
2082 silently corrupt the start of the filesystem. CONFIG_LBD is
2083 enabled for all CFS-supported kernels, but the possibility of
2084 this happening with a modified kernel config exists.
2086 Severity : enhancement
2088 Description: add client O_DIRECT support for 2.6 kernels
2089 Details : It is now possible to do O_DIRECT reads and writes to files
2090 in the Lustre client mountpoint on 2.6 kernel clients.
2092 Severity : enhancement
2094 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2095 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2096 parallel, not waiting for response from every OST before sending
2097 a rpc to the next OST.
2102 Description: setattr vs write race when updating file timestamps
2103 Details : Client processes that update a file timestamp into the past
2104 right after writing to the file (e.g. tar) it is possible that
2105 the updated file modification time can be reset to the current
2106 time due to a race between processing the setattr and write RPC.
2108 Severity : enhancement
2110 Description: Bring 'lfs find' closer in line with regular Linux find.
2111 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2112 -print0 options and obtains all the needed info through the lustre
2115 Severity : enhancement
2117 Description: support up to 1024 configured devices on one node
2118 Details : change obd_dev array from statically allocated to dynamically
2119 allocated structs as they are first used to reduce memory usage
2124 Description: Flush dirty partially truncated pages during truncate
2125 Details : Immediatelly flush partially truncated pages in filter_setattr,
2126 this way we completely avoid having any pages in page cache on OST
2127 and can retire ugly workarounds during writes to flush such pages.
2132 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2133 Details : protect inode from truncation within vfs_unlink() context
2134 just take a reference before calling vfs_unlink() and release it
2135 when parent's i_sem is free.
2140 Description: last_id value checked outside lock on OST caused LASSERT failure
2141 Details : If there were multiple MDS->OST object precreate requests in
2142 flight, it was possible that the OST's last object id was checked
2143 outside a lock and incorrectly tripped an assertion. Move checks
2144 inside locks, and discard old precreate requests.
2147 Frequency : always, if extents are used on OSTs
2149 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2150 Details : OSTs using the extents format would not zero the high 16 bits of
2151 the index physical block number. This is not a problem for any
2152 OST filesystems smaller than 16TB, and no kernels support ext3
2153 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2154 new/modified files) and can be fixed for existing filesystems
2155 with e2fsprogs-1.39-cfs1.
2160 Description: import connection selection may be incorrect if timer wraps
2161 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2162 import connections to be ignored if the 32-bit jiffies counter
2163 wraps. Use a 64-bit jiffies counter.
2166 Frequency : very large clusters immediately after boot
2168 Description: LNET request buffers exhausted under heavy short-term load
2169 Details : If a large number of client requests are generated on a service
2170 that has previously never seen so many requests it is possible
2171 that the request buffer growth cannot keep up with the spike in
2172 demand. Instead of dropping incoming requests, they are held in
2173 the LND until the RPC service can accept more requests.
2176 Frequency : Sometimes during replay
2178 Description: Assertion failure in ll_local_open after replay.
2179 Details : If replay happened on an open request reply before we were able
2180 to set replay handler, reply will become not swabbed tripping the
2181 assertion in ll_local_open. Now we set the handler right after
2182 recognising of open request
2185 Frequency : very rare
2187 Description: kernel reports "badness in vsnprintf"
2188 Details : Reading from the "recovery_status" /proc file in small chunks
2189 may cause a negative length in lprocfs_obd_rd_recovery_status()
2190 call to vsnprintf() (which is otherwise harmless). Exit early
2191 if there is no more space in the output buffer.
2193 Severity : enhancement
2195 Description: clear OBD RPC statistics by writing to them
2196 Details : It is now possible to clear the OBD RPC statistics by writing
2197 to the "stats" file.
2202 Description: Client mtime is not the same on different clients after utimes
2203 Details : In some cases, the client was using the utimes() syscall on
2204 a file cached on another node. The clients now validate the
2205 ctime from the MDS + OSTs to determine which one is right.
2210 Description: Inability to activate failout mode
2211 Details : lconf script incorrectly assumed that in pythong string's numeric
2212 value is used in comparisons.
2215 Frequency : always with multiple stripes per file
2217 Description: Inefficient object allocation for mutli-stripe files
2218 Details : When selecting which OSTs to stripe files over, for files with
2219 a stripe count that divides evenly into the number of OSTs,
2220 the MDS is always picking the same starting OST for each file.
2221 Return the OST selection heuristic to the original design.
2226 Description: mount failures may take full timeout to return an error
2227 Details : Under some heavy load conditions it is possible that a
2228 failed mount can wait for the full obd_timeout interval,
2229 possibly several minutes, before reporting an error.
2230 Instead return an error as soon as the status is known.
2232 Frequency : quota enabled and large files being deleted
2234 Description: releasing more than 4GB of quota at once hangs OST
2235 Details : If a user deletes more than 4GB of files on a single OST it
2236 will cause the OST to spin in an infinite loop. Release
2237 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2242 Description: statfs data retrieved from /proc may be stale or zero
2243 Details : When reading per-device statfs data from /proc, in the
2244 {kbytes,files}_{total,free,avail} files, it may appear
2245 as zero or be out of date.
2248 Frequency : systems with MD RAID1 external journal devices
2250 Description: lconf's call to blkid is confused by RAID1 journal devices
2251 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2252 of returning all block devices that match the journal UUID.
2255 Frequency : always, for aggregate stripe size over 4GB
2257 Description: assertion fails when trying to use 4GB stripe size
2258 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2259 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2262 Frequency : always on ppc64
2264 Description: the first write on an ext3 filesystem with mballoc got stuck
2265 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2266 perform endianness conversion.
2268 ------------------------------------------------------------------------------
2270 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2272 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2273 INTEROPERATE with older versions automatically. Please read the
2274 user documentation before upgrading any part of a live system.
2275 * WARNING: Lustre networking configuration changes are required with
2276 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2279 * Support for kernels:
2280 2.6.9-22.0.2.EL (RHEL 4)
2281 2.6.5-7.244 (SLES 9)
2282 2.6.12.6 vanilla (kernel.org)
2285 Severity : enhancement
2286 Bugzilla : 7981/8208
2287 Description: Introduced Lustre Networking (LNET)
2288 Details : LNET is new networking infrastructure for Lustre, it includes
2289 a reorganized network configuration mode (see the user
2290 documentation for full details) as well as support for routing
2291 between different network fabrics. Lustre Networking Devices
2292 (LNDs) for the supported network fabrics have also been
2293 created for this new infrastructure.
2295 Severity : enhancement
2296 Description: Introduced Access control lists
2297 Details : clients can set ACLs on files and directories in order to have
2298 more fine-grained permissions than the standard Unix UGO+RWX.
2299 The MDS must be started with the "-o acl" mount option.
2301 Severity : enhancement
2302 Description: Introduced filesystem quotas
2303 Details : Administrators may now establish per-user quotas on the
2306 Severity : enhancement
2308 Description: Configuration change for the XT3
2309 The PTLLND is now used to run Lustre over Portals on the XT3
2310 The configure option(s) --with-cray-portals are no longer used.
2311 Rather --with-portals=<path-to-portals-includes> is used to
2312 enable building on the XT3. In addition to enable XT3 specific
2313 features the option --enable-cray-xt3 must be used.
2318 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2319 Details : A race between allocating a new llcd and re-getting the llcd_lock
2320 allowed another thread to grab newly-allocated llcd.
2322 Severity : enhancement
2324 Description: 2.6 OST async journal commit and locking fix to improve performance
2325 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2326 2.6 kernels are now async as they already were in 2.4 kernels so
2327 that they can commit concurrently with the network bulk transfer.
2328 For block-allocated files the filter allocation semaphore is held
2329 to avoid filesystem fragmentation during allocation. BKL lock
2330 removed for 2.6 xattr operations where it is no longer needed.
2335 Description: lconf incorrectly determined whether two IP networks could talk
2336 Details : In some more complicated routing and multiple-network
2337 configurations, lconf will avoid trying to make a network
2338 connection to a disjoint part of the IP space. It was doing the
2339 math incorrectly for one set of cases.
2344 Description: Fix for potential infinite loop processing records in an llog.
2345 Details : If an llog record is corrupted/zeroed, it is possible to loop
2346 forever in llog_process(). Validate the llog record length
2347 and skip the remainder of the block on error.
2350 Frequency : occasional (liblustre only)
2352 Description: liblustre could not open files whose last component is a symlink
2353 Details : sysio_path_walk() would incorrectly pass the open intent to
2354 intermediate path components.
2357 Frequency : rare (liblustre only with non-standard tuning)
2358 Bugzilla : 7201 (7350)
2359 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2360 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2361 after it has been released, only do this for referenced locks
2363 Severity : enhancement
2365 Description: specifying an (invalid) directory default stripe_size of -1
2366 would reset the directory default striping
2367 Details : stripe_size -1 was used internally to signal directory stripe
2368 removal, now use "all default" to signal dir stripe removal
2369 as a directory striping of "all default" is not useful
2372 Frequency : common for large clusters running liblustre clients
2374 Description: doing an ls when liblustre clients are running is slow
2375 Details : sending a glimpse AST to a liblustre client waits for every AST
2376 to time out, as liblustre clients will not respond. Since they
2377 cannot cache data we refresh the OST lock LVB from disk instead.
2379 Severity : enhancement
2381 Description: doing an ls at the same time as file IO can be slow
2382 Details : enqueue and other "small" requests can be blocked behind many
2383 large IO requests. Create a new OST IO portal for non-IO
2384 requests so they can be processed faster.
2387 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2389 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2390 created file with mode 000
2391 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2392 allow an ftruncate() as a non-root user to a file with mode 000.
2393 Lustre now respects this flag to disable mode checks when
2394 truncating a file owned by the user
2397 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2400 Description: Revoking locks from clients that went dead or catatonic might take
2402 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2403 cancellation of such locks instant on servers without waiting for
2404 any reply from clients. Clients drops these locks when cancel
2405 notification from server is received without replying.
2408 Frequency : liblustre-only, when liblustre client dies or becomes busy
2410 Description: Doing ls on Linux clients can take a long time with active
2412 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2413 granting such locks to it in the first place if possible. Locks
2414 are taken by proxy on the OST during the read or write and
2415 dropped immediately afterward. Add connect flags handling, do
2416 not grant locks to liblustre clients for glimpse ASTs.
2418 Severity : enhancement
2420 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2421 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2422 sequential reads and behaves badly with applications doing random
2423 reads. Improve it by reading ahead at least read region, and
2424 avoiding excessive large RPC for small reads.
2426 Severity : enhancement
2428 Description: Creating more than 1000 files for a single job may cause a load
2429 imbalance on the OSTs if there are also a large number of OSTs.
2430 Details : qos_prep_create() uses an OST index reseed value that is an
2431 even multiple of the number of available OSTs so that if the
2432 reseed happens in the middle of the object allocation it will
2433 still utilize the OSTs as uniformly as possible.
2438 Description: OST or MDS may oops in ping_evictor_main()
2439 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2440 but doesn't restart at beginning of obd_exports_timed list
2443 Severity : enhancement
2445 Description: improve by-nid export eviction on the MDS and OST
2446 Details : allow multiple exports with the same NID to be evicted at one
2447 time without re-searching the exports list.
2450 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2452 Description: MDS may oops in groups_free()
2453 Details : in rare race conditions a newly allocated group_info struct is
2454 freed again, and this can be NULL. The 2.4 compatibility code
2455 for groups_free() checked for a NULL pointer, but 2.6 did not.
2458 Frequency : common for liblustre clients doing little filesystem IO
2459 Bugzilla : 9352, 7313
2460 Description: server may evict liblustre clients accessing contended locks
2461 Details : if a client is granted a lock or receives a completion AST
2462 with a blocking AST already set it would not reply to the AST
2463 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2466 Frequency : lfs setstripe, only systems with more than 160 OSTs
2468 Description: unable to set striping with a starting offset beyond OST 160
2469 Details : llapi_create_file() incorrectly limited the starting stripe
2470 index to the maximum single-file stripe count.
2473 Frequency : LDAP users only
2475 Description: lconf did not handle in-kernel recovery with LDAP properly
2476 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2478 Severity : enhancement
2480 Description: bind OST threads to NUMA nodes to improve performance
2481 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2482 node and do their allocations there to localize memory access
2484 Severity : enhancement
2486 Description: llmount can determine client NID directly from Myrinet (GM)
2487 Details : the client NID code from gmnalnid was moved directly into
2488 llmount, removing the need to use this or specifying the
2489 client NID explicitly when mounting GM clients with zeroconf
2492 Frequency : if client is started with down MDS
2494 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2495 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2496 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2497 ptlrpc_import_delayed_req() because we didn't check if the
2498 request was interrupted, and we also didn't break out of the
2499 event loop if there was no timeout
2504 Description: data loss during non-page-aligned writes to a single file from
2505 both multiple nodes and multiple threads on one node at same time
2506 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2507 inconsistency led to false short-reads, that were cached and later
2508 used by ->prepare_write() to fill in partially written page,
2509 leading to data loss.
2512 Frequency : always, if lconf --abort_recovery used
2514 Description: lconf --abort_recovery fails with 'Operation not supported'
2515 Details : lconf was attempting to abort recovery on the MDT device and not
2518 Severity : enhancement
2520 Description: remove cleanup logs
2521 Details : replace lconf-generated cleanup logs with lustre internal
2522 cleanup routines. Eliminates the need for client-cleanup and
2525 Severity : enhancement
2527 Description: add support for EAs (user and system) on lustre filesystems
2528 Details : it is now possible to store extended attributes in the Lustre
2529 client filesystem, and with the user_xattr mount option it
2530 is possible to allow users to store EAs on their files also
2532 Severity : enhancement
2534 Description: Add possibility (config option) to show minimal available OST free
2536 Details : When compiled with --enable-mindf configure option, statfs(2)
2537 (and so, df) will return least minimal free space available from
2538 all OSTs as amount of free space on FS, instead of summary of
2539 free spaces of all OSTs.
2541 Severity : enhancement
2543 Description: do not expand extent locks acquired on OST-side
2544 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2545 by server: they are not cached anyway.
2548 Frequency : when mmap is used/binaries executed from Lustre
2550 Description: Unmmap pages before throwing them away from read cache.
2551 Details : llap_shrink cache now attempts to unmap pages before discarding
2552 them (if unmapping failed - do not discard). SLES9 kernel has
2553 extra checks that trigger if this unmapping is not done first.
2558 Description: lconf didn't resolve symlinks before checking to see whether a
2559 given mountpoint was already in use
2562 Frequency : when migrating failover services
2563 Bugzilla : 6395, 9514
2564 Description: When migrating a subset of services from a node (e.g. failback
2565 from a failover service node) the remaining services would
2566 time out and evict clients.
2567 Details : lconf --force (implied by --failover) sets the global obd_timeout
2568 to 5 seconds in order to quickly disconnect, but this caused
2569 other RPCs to time out too quickly. Do not change the global
2570 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2572 Severity : enhancement
2573 Frequency : if MDS is started with down OST
2574 Bugzilla : 9439,5706
2575 Description: Allow startup/shutdown of an MDS without depending on the
2576 availability of the OSTs.
2577 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2578 Add appropriate locking and lov-osc refcounts for safe
2579 cleaning. Add osc abort_inflight calls in case the
2580 synchronize never started.
2583 Frequency : occasional (Cray XT3 only)
2585 Description: root not authorized to access files in CRAY_PORTALS environment
2586 Details : The client process capabilities were not honoured on the MDS in
2587 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2588 been accessed by an authorized user then root was able to access
2589 the file on the local client also. The root user capabilities
2590 are now allowed on the MDS, as this environment has secure UID.
2593 Frequency : occasional
2595 Description: ldiskfs "too long searching" message happens too often
2596 Details : A debugging message (otherwise harmless) prints too often on
2597 the OST console. This has been reduced to only happen when
2598 there are fragmentation problems on the filesystem.
2603 Description: Division by zero in statfs when all OSCs are inactive
2604 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2605 lov_statfs divides by value returned by lov_get_stripecnt().
2609 Bugzilla : 9489, 3273
2610 Description: First write from each client to each OST was only 4kB in size,
2611 to initialize client writeback cache, which caused sub-optimal
2612 RPCs and poor layout on disk for the first writen file.
2613 Details : Clients now request an initial cache grant at (re)connect time
2614 and so that they can start streaming writes to the cache right
2615 away and always do full-sized RPCs if there is enough data.
2616 If the OST is rebooted the client also re-establishes its grant
2617 so that client cached writes will be honoured under the grant.
2622 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2623 Details : Now I/O RPCs go to different portal number and (presumably) fast
2624 lock requests (and glimses) and other RPCs get their own service
2625 threads pool that should be able to service those RPCs
2628 Severity : enhancement
2630 Description: Ability to exchange lustre version between client and servers and
2631 issue warnings at client side if client is too old. Also for
2632 liblustre clients there is ability to refuse connection of too old
2634 Details : New 'version' field is added to connect data structure that is
2635 filled with version info. That info is later checked by server and
2639 Frequency : rare, liblustre only.
2640 Bugzilla : 9296, 9581
2641 Description: Two simultaneous writes from liblustre at offset within same page
2642 might proceed at the same time overwriting eachother with stale
2644 Details : I/O lock withing llu_file_prwv was released too early, before data
2645 actually was hitting the wire. Extended lock-holding time until
2646 server acknowledges receiving data.
2649 Frequency : extremely rare. Never observed in practice.
2651 Description: avoid generating lustre_handle cookie of 0.
2652 Details : class_handle_hash() generates handle cookies by incrementing
2653 global counter, and can hit 0 occasionaly (this is unlikely, but
2654 not impossible, because initial value of cookie counter is
2655 selected randonly). Value of 0 is used as a sentinel meaning
2656 "unassigned handle" --- avoid it. Also coalesce two critical
2657 sections in this function into one.
2659 Severity : enhancement
2661 Description: allow liblustre clients to delegate truncate locking to OST
2662 Details : To avoid overhead of locking, liblustre client instructs OST to
2663 take extent lock in ost_punch() on client's behalf. New connection
2664 flag is added to handle backward compatibility.
2666 Severity : enhancement
2667 Bugzilla : 4928, 7341, 9758
2668 Description: allow number of OST service threads to be specified
2669 Details : a module parameter allows the number of OST service threads
2670 to be specified via "options ost ost_num_threads={N}" in the
2671 OSS's /etc/modules.conf or /etc/modprobe.conf.
2675 Bugzilla : 6146, 9635, 9895
2676 Description: servers crash with bad pointer in target_handle_connect()
2677 Details : In rare cases when a client is reconnecting it was possible that
2678 the connection request was the last reference for that export.
2679 We would temporarily drop the export reference and get a new
2680 one, but this may have been the last reference and the export
2681 was just destroyed. Get new reference before dropping old one.
2683 Severity : enhancement
2684 Frequency : if client is started with failover MDS
2686 Description: Allow multiple MDS hostnames in the mount command
2687 Details : Try to read the configuration from all specified MDS
2688 hostnames during a client mount in case the "primary"
2691 Severity : enhancement
2693 Description: Stop sending data to evicted clients as soon as possible.
2694 Details : Check if the client we are about to send or are sending data to
2695 was evicted already. (Check is done every second of waiting,
2696 for which l_wait_event interface was extended to allow checking
2697 of exit condition at specified intervals).
2700 Frequency : rare, normally only when NFS exporting is done from client
2702 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2704 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2707 Frequency : always, if srand() is called before liblustre initialization
2709 Description: Liblustre uses system PRNG disturbing its usage by user application
2710 Details : Introduce internal to lustre fast and high-quality PRNG for
2711 lustre usage and make liblustre and some other places in generic
2712 lustre code to use it.
2714 Severity : enhancement
2715 Bugzilla : 9477, 9557, 9870
2716 Description: Verify that the MDS configuration logs are updated when xml is
2717 Details : Check if the .xml configuration logs are newer than the config
2718 logs stored on the MDS and report an error if this is the case.
2719 Request --write-conf, or allow starting with --old_conf.
2721 Severity : enhancement
2723 Description: Handle symlinks in the path when checking if Lustre is mounted.
2724 Details : Resolve intermediate symlinks when checking if a client has
2725 mounted a filesystem to avoid duplicate client mounts.
2730 Description: lconf can hit an error exception but still return success.
2731 Details : The lconf command catches the Command error exception at the top
2732 level script context and will exit with the associated exit
2733 status, but doesn't ensure that this exit status is non-zero.
2738 Description: failure of ptlrpc thread startup can cause oops
2739 Details : Starting a ptlrpc service thread can fail if there are a large
2740 number of threads or the server memory is very fragmented.
2741 Handle this without oopsing.
2744 Frequency : always, only if liblustre and non-default acceptor port was used
2746 Description: liblustre cannot connect to servers with non-default acceptor port
2747 Details : tcpnal_set_default_params() was not called and was therefore
2748 ignoring the environment varaible TCPNAL_PORT, as well as other
2749 TCPNAL_ environment variables
2754 Description: two objects could be created on the same OST for a single file
2755 Details : If an OST is down, in some cases it was possible to create two
2756 objects on a single OST for a single file. No problems other
2757 than potential performance impact and spurious error messages.
2761 Bugzilla : 5681, 9562
2762 Description: Client may oops in ll_unhash_aliases
2763 Details : Client dcache may become inconsistent in race condition.
2764 In some cases "getcwd" can fail if the current directory is
2770 Description: Inode refcounting problems in NFS export code
2771 Details : link_raw functions used to call d_instantiate without obtaining
2772 extra inode reference first.
2776 Bugzilla : 9942, 9903
2777 Description: Referencing freed requests leading to crash, memleaks with NFS.
2778 Details : We used to require that call to ll_revalidate_it was always
2779 followed by ll_lookup_it. Also with revalidate_special() it is
2780 possible to call ll_revalidate_it() twice for the same dentry
2781 even if first occurence returned success. This fix changes semantic
2782 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2783 reference on a request referred from the intent.
2784 ll_intent_release() then releases such a request.
2787 Frequency : rare, normally benchmark loads only
2789 Description: unlinked inodes were kept in memory on the client
2790 Details : If a client is repeatedly creating and unlinking files it
2791 can accumulate a lot of stale inodes in the inode slab cache.
2792 If there is no other client load running this can cause the
2793 client node to run out of memory. Instead flush old inodes
2794 from client cache that have the same inode number as a new inode.
2797 Frequency : SLES9 2.6.5 kernel and long filenames only
2798 Bugzilla : 9969, 10379
2799 Description: utime reports stale NFS file handle
2800 Details : SLES9 uses out-of-dentry names in some cases, which confused
2801 the lustre dentry revalidation. Change it to always use the
2805 Frequency : rare, unless heavy write-truncate concurrency is continuous
2806 Bugzilla : 4180, 6984, 7171, 9963, 9331
2807 Description: OST becomes very slow and/or deadlocked during object unlink
2808 Details : filter_destroy() was holding onto the parent directory lock
2809 while truncating+unlinking objects. For very large objects this
2810 may block other threads for a long time and slow overall OST
2811 responsiveness. It may also be possible to get a lock ordering
2812 deadlock in this case, or run out of journal credits because of
2813 the combined truncate+unlink. Solution is to do object truncate
2814 first in one transaction without parent lock, and then do the
2815 final unlink in a new transaction with the parent lock. This
2816 reduces the lock hold time dramatically.
2819 Frequency : rare, 2.4 kernels only
2821 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2822 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2823 before unmounting their filesystem in order to avoid deadlock.
2824 The kernel_locked() function in 2.4 kernels only checks whether
2825 the kernel lock is held, not whether it is this process that is
2826 holding it as 2.6 kernels do.
2831 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2832 Details : The client ptlrpc code may be trying to reconnect to a down
2833 server before a previous connection attempt has timed out.
2834 Increase the reconnect interval to be longer than the connection
2835 timeout interval to avoid sending duplicate connections to
2839 Frequency : echo_client brw_test command
2841 Description: fix echo_client to work with OST preallocated code
2842 Details : OST preallocation code (5137) didn't take echo_client IO path
2843 into account: echo_client calls filter methods outside of any
2844 OST thread and, hence, there is no per-thread preallocated
2845 pages and buffers to use. Solution: hijack pga pages for IO. As
2846 a byproduct, this avoids unnecessary data copying.
2850 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2851 Description: Client can oops in mdc_commit_close() after open replay
2852 Details : It was possible for the MDS to return an open request with no
2853 transaction number in mds_finish_transno() if the client was
2854 evicted, but without actually returning an error. Clients
2855 would later try to replay that open and may trip an assertion
2856 Simplify the client close codepath, and always return an error
2857 from the MDS in case the open is not successful.
2860 Frequency : rare, 2.6 OSTs only
2862 Description: OST may deadlock under high load on fragmented files
2863 Details : If there was a heavy load and highly-fragmented OST filesystems
2864 it was possible to have all the OST threads deadlock waiting on
2865 allocation of biovecs, because the biovecs were not released
2866 until the entire RPC IO was completed. Instead, release biovecs
2867 as soon as they are complete to ensure forward IO progress.
2869 Severity : enhancement
2871 Description: Support for specifying external journal device at mount
2872 Details : If an OST or MDS device is formatted with an external journal
2873 device, this device major/minor is stored in the ext3 superblock
2874 and may not be valid for failover. Allow detecting and
2875 specifying the external journal at mount time.
2880 Description: Mounting an MDS with pending unlinked files may cause oops
2881 Details : target_finish_recovery() calls mds_postrecov() which returned
2882 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2883 considers this an error and immediately begins cleaning up the
2884 lov, just after starting the mds_lov process
2886 Severity : enhancement
2888 Description: Implement 'lfs df' to report actual free space on per-OST basis
2889 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2890 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2891 usage of objects; '-h' to report in human readable format.
2893 ------------------------------------------------------------------------------
2895 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2902 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2903 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2904 at startup during orphan unlinking. Instead, skip buddy update
2905 before setup, as it will be regenerated after recovery is complete.
2910 Description: If an OST is inactive, its locks might reference stale inodes.
2911 Details : lov_change_cbdata() must iterate over all namespaces, even if
2912 they are inactive to clear inode references from the lock.
2914 Severity : enhancement
2915 Frequency : occasional, if non-standard max_dirty_mb used
2917 Description: Client will block write RPCs if not enough grant
2918 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2919 then the client will block writes while waiting for another RPC
2920 to complete instead of consuming its dirty limit. With change
2921 we get improved performance when max_dirty_mb is small.
2923 Severity : enhancement
2924 Bugzilla : 3389, 6253
2925 Description: Add support for supplementary groups on the MDS.
2926 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2927 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2928 lookups for user supplementary groups into a cache.
2932 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2933 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2934 for liblustre O_EXCL is enforced by MDS.
2938 Description: readdir never returns NULL in liblustre.
2939 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2940 dentry in struct dirent.
2944 Frequency : liblustre only, depends on application IO pattern
2945 Description: liblustre clients evicted if not contacting servers
2946 Details : Don't put liblustre clients into the ping_evictor list, so
2947 they will not be evicted by the pinger ever.
2949 Severity : enhancement
2951 Description: Add ability to evict clients by NID from MDS.
2952 Details : By echoing "nid:$NID" string into
2953 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2954 $NID would be instantly evicted from this MDS and from all active
2955 OSTs connected to it.
2959 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2960 Details : lookup_it_finish() used to query file size from OSTs that was not
2965 Description: service threads change working directory to that of init
2966 Details : Starting lustre service threads may pin the working directory
2967 of the parent thread, making that filesystem busy. Threads
2968 now change to the working directory of init to avoid this.
2972 Frequency : during shutdown only
2973 Description: shutdown with a failed MDS or OST can cause unmount to hang
2974 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2979 Frequency : 2.6 only
2980 Description: chmod/chown may include an extra supplementary group
2981 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2982 supplementary group and if none is specified this is used.
2985 Bugzilla : 5479 (6816)
2987 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2988 Details : If lookup is in progress on a file that is unlinked we might try
2989 to revalidate the inode and fail in revalidate after lookup is
2990 complete and ll_file_open() enqueues the open again but
2991 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2994 Frequency : always, if lconf --abort_recovery used
2996 Description: lconf --abort_recovery fails with 'Operation not supported'
2997 Details : lconf was attempting to abort recovery on the MDT device and not
3000 ------------------------------------------------------------------------------
3002 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3007 Frequency : rare (only unsupported configurations with a node running as an
3009 Bugzilla : 6514, 5137
3010 Description: Mounting a Lustre file system on a node running as an OST could
3012 Details : OSTs now preallocates memory needed to write out data at
3013 startup, instead of when needed, to avoid having to
3014 allocate memory in possibly low memory situations.
3015 Specifically, if the file system is mounted on on OST,
3016 memory pressure could force it to try to write out data,
3017 which it needed to allocate memory to do. Due to the low
3018 memory, it would be unable to do so and the node would
3019 become unresponsive.
3021 Severity : enhancement
3023 Description: Addition of lconf --service command line option
3024 Details : lconf now accepts a '--service <arg>' option, which is
3025 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3027 Severity : enhancement
3029 Description: Failover mode is now the default for OSTs.
3030 Details : By default, OSTs will now run in failover mode. To return to
3031 the old behaviour, add '--failout' to the lmc line for OSTs.
3033 Severity : enhancement
3035 Description: Health checks are now provided for MDS and OSTs
3036 Details : Additional detailed health check information on MSD and OSTs
3037 is now provided through the procfs health_check value.
3040 Frequency : occasional, depends on IO load
3042 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3043 after numerous create/delete cycles
3044 Details : The ext3 inode allocation policy would not allocate new inodes
3045 very well on the OSTs because there are no new directories
3046 being created. Instead we look for groups with free space if
3047 the parent directories are nearly full.
3052 Description: Network or server problems during mount may cause partially
3053 mounted clients instead of returning an error.
3054 Details : The config llog parsing code may overwrite the error return
3055 code during mount error handling, returning success instead
3061 Description: MDS can fail to allocate large reply buffers
3062 Details : After long uptimes the MDS can fail to allocate large reply
3063 buffers (e.g. zconf client mount config records) due to memory
3064 fragmentation or consumption by the buffer cache. Preallocate
3065 some large reply buffers so that these replies can be sent even
3066 under memory pressure.
3070 Frequency : rare (liblustre)
3071 Description: fsx running with liblustre complained that using truncate() to
3072 extend the file doesn't work. This patch corrects that issue.
3073 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3074 ATTR_SIZE and lsm use in llu_setattr_raw.
3078 Frequency : rare, only 2.6 kernels
3079 Description: Unusual file access patterns on the MDS may result in inode
3080 data being lost in very rare circumstances.
3081 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3082 code in the 2.6 kernel caused the inode and/or EA data not to
3083 be read from disk, causing single-file corruption.
3087 Frequency : rare, only 2.6 filesystems using extents
3088 Description: Heavy concurrent write and delete load may cause data corruption.
3089 Details : It was possible under high-load situations to have an extent
3090 metadata block in the block device cache from a just-unlinked
3091 file overwrite a newly-allocated data block. We now unmap any
3092 metadata buffers that alias just-allocated data blocks.
3096 Frequency : filesystems with default stripe_count larger than 77
3097 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3098 Details : lconf specifies an inode size of 4096 bytes when the default
3099 stripe_count is larger than 77. This conflicts with the default
3100 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3101 this case to avoid pinning too much memory for large EAs.
3103 ------------------------------------------------------------------------------
3105 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3110 Frequency : rare (extremely heavy IO load with hundreds of clients)
3112 Description: Client is evicted, gets IO error writing to file
3113 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3114 caused clients to be evicted for AST timeouts. The fixes in
3115 bug 5192 mean we no longer need to have such short AST timeouts
3116 so ldlm_timeout has been increased.
3119 Frequency : occasional during --force or --failover shutdown under load
3120 Bugzilla : 5949, 4834
3121 Description: Server oops/LBUG if stopped with --force or --failover under load
3122 Details : a collection of import/export refcount and cleanup ordering
3123 issues fixed for safer force cleanup
3126 Frequency : only filesystems larger than 120 OSTs
3127 Bugzilla : 5990, 6223
3128 Description: lfs getstripe would oops on a very large filesystem
3129 Details : lov_getconfig used kfree on vmalloc'd memory
3132 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3133 Bugzilla : 6242, 6243
3134 Description: reading from files that had been truncated to a non-zero size
3135 but never opened returned no data
3136 Details : ll_file_read() reads zeros from no-object files to EOF
3141 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3142 Details : There's little harm in aborting MDS/OSS recovery and letting it
3143 try again, so I removed the LASSERT and return an error instead.
3145 Severity : enhancement
3147 Description: New debugging infrastructure for tracking down data corruption
3148 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3149 (b) cover more of the client-side code path, and (c) try to narrow
3150 down where problems occurred
3154 Bugzilla : 3819, 4364, 4397, 6313
3155 Description: Racing close and eviction MDS could cause assertion in mds_close
3156 Details : It was possible to get multiple mfd references during close and
3157 client eviction, leading to one thread referencing a freed mfd.
3159 Severity: : enhancement
3160 Bugzilla : 3262, 6359
3161 Description: Attempts to reconnect to servers are now more aggressive.
3162 Details : This builds on the enhanced upcall-less recovery that was added
3163 in 1.4.2. When trying to reconnect to servers, clients will
3164 now try each server in the failover group every 10 seconds. By
3165 default, clients would previously try one server every 25 seconds.
3170 Description: After recovery, certain operations trigger a failed
3171 assertion on a client.
3172 Details : Failing over an mds, using lconf -d --failover, while a
3173 client was doing a readdir() call would cause the client to
3174 LBUG after recovery completed and the readdir() was resent.
3176 Severity : enhancement
3178 Description: Default groups are now added by lconf
3179 Details : You can now run lconf --group <servicename> without having to
3180 manually add groups with lmc.
3183 Frequency : occasional
3185 Description: Nodes with an elan id of 0 trigger a failed assertion
3188 Frequency : always when accessing e.g. tty/console device nodes
3190 Description: tty and some other devices nodes cannot be used on lustre
3191 Details : file's private_data field is used by device data and lustre
3192 values in there got lost. New field was added to struct file to
3193 store fs-specific private data.
3196 Frequency : when exporting Lustre via NFS
3198 Description: NFSD failed occasionally when looking up a path component
3199 Details : NFSD is looking up ".." which was broken in ext3 directories
3200 that had grown large enough to become hashed.
3203 Frequency : Clusters with multiple interfaces not on the same subnet
3205 Description: Nodes will repeatedly try to reconnect to an interface which it
3206 cannot reach and report an error to the log.
3207 Details : Extra peer list entries will be created by lconf with some peers
3208 unreachable. lconf now validates the peer before adding it.
3211 Frequency : Only if a default stripe is set on the filesystem root.
3213 Description: Setting a default stripe on the filesystem root prevented the
3214 filesystem from being remounted.
3215 Details : The client was sending extra request flags in the root getattr
3216 request and did not allocate a reply buffer for the dir EA.
3219 Frequency : occasional, higher if lots of files are accessed by one client
3220 Bugzilla : 6159, 6097
3221 Description: Client trips assertion regarding lsm mismatch/magic
3222 Details : While revalidating inodes the VFS looks up inodes with ifind()
3223 and in rare cases can find an inode that is being freed.
3224 The ll_test_inode() code will free the lsm during ifind()
3225 when it finds an existing inode and then the VFS later attaches
3226 this free lsm to a new inode.
3230 Bugzilla : 6422, 7030
3231 Description: MDS deadlock between mkdir and client eviction
3232 Details : Creating a new file via mkdir or mknod (starting a transaction
3233 and getting the ns lock) can deadlock with client eviction
3234 (gets ns lock and trying to finish a synchronous transaction).
3237 Frequency : occasional
3238 Description: While starting a server, the fsfilt_ext3 module could not be
3240 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3241 kernels. Previously, lconf would still use the ext3 name
3242 when trying to load modules. Now, it will correctly use
3243 ext3 on 2.4 and ldiskfs on 2.6.
3245 Severity : enhancement
3246 Description: The default stripe count has been changed to 1
3247 Details : The interpretation of the default stripe count (0, to lfs
3248 or lmc) has been changed to mean striping across a single
3249 OST, rather than all available. For general usage we have
3250 found a stripe count of 1 or 2 works best.
3252 Severity : enhancement
3253 Description: Add support for compiling against Cray portals.
3254 Details : Conditional compiling for some areas that are different
3258 Frequency : occasional
3259 Bugzilla : 6409, 6834
3260 Description: Creating files with an explicit stripe count may lead to
3261 a failed assertion on the MDS
3262 Details : If some OSTs are full or unavailable, creating files may
3263 trigger a failed assertion on the MDS. Now, Lustre will
3264 try to use other servers or return an error to the
3268 Frequency : occasional
3270 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3271 caused lock timeout during readahead (since 1.4.2).
3272 Details : Processes doing readahead might match a lock that hasn't been
3273 granted yet if there are overlapping and conflicting lock
3274 requests. The readahead process waits on ungranted lock
3275 (original lock is CBPENDING), while OST waits for that process
3276 to cancel CBPENDING read lock and eventually evicts client.
3278 Severity : enhancement
3280 Description: Initial enabling of flock support for clients
3281 Details : Implements fcntl advisory locking and file status functions.
3282 This feature is provided as an optional mount flag (default
3283 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3284 locking are implemented yet, and those that are are not guaranteed
3285 to be completely correct in production environments.
3286 mount -t lustre -o [flock|noflock] ...
3289 Frequency : occasional
3291 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3292 assertion in the ext3 JBD (journaling) layer.
3293 Details : The b_committed_data struct is protected by the big kernel lock
3294 in 2.4 kernels, serializing journal_commit_transaction() and
3295 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3296 access to this struct. In 2.6 kernels there is finer grained
3297 locking to improve SMP performance of the JBD layer.
3301 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3306 Description: merge and cleanup kernel patches.
3309 -----------------------------------------------------------------------------
3311 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3313 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3314 versions of Lustre. Please update all servers and clients to
3315 version 1.4.2 or later at the same time. You must also run
3316 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3317 to update the configuration logs.
3319 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3320 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3321 - swab llog records of type '0' so we get proper header size/idx (5861)
3322 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3323 - fix rename of one directory over another leaking an inode (5953)
3324 - avoid SetPageDirty on 2.6 (5981)
3325 - don't re-add just-being-destroyed locks to the waiting list (5653)
3326 - when creating new directories, inherit the parent's custom
3327 striping settings if present parent (3048)
3328 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3329 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3330 - don't hold client locks on temporary worklist from l_lru (5666)
3331 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3332 - automatically evict dead clients (5921)
3333 - Update file size properly in create+truncate+fstat case (6196)
3334 - Do not unhash mountpoint dentries, do not allow removal of
3336 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3337 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3338 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3339 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3340 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3341 - Correctly handle reads of files with no objects (6243)
3342 - lctl recover will also mark a device active if deactivate used (5933)
3344 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3345 - allow --write-conf on an MDS with different nettype than client (5619)
3346 - don't write config llogs to MDS for mounts not from that MDS (5617)
3347 - lconf should create multiple TCP connections from a client (5201)
3348 - init scripts are now turned off by default; run chkconfig --on
3349 lustre and chkconfig --on lustrefs to use them
3350 - upcalls are no longer needed for clients to recover to failover
3352 - add --abort-recovery option to lconf to abort recovery on device
3354 - add support for an arbitrary number of OSTs (3026)
3355 - Quota support protocol changes.
3356 - forward compatibility changes to wire structs (6007)
3357 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3358 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3359 - improved "lustre status" script
3360 - initialize blocksize for non-regular files (6062)
3361 - added --disable-server and --disable-client configure options (5782)
3362 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3363 - Vanilla 2.4.29 support
3364 - increase maximum number of obd devices to 520 (6242)
3365 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3366 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3367 SLES 9 2.6.5 kernels
3368 - limit stripes per file to 160 (the maximum EA size) (6093)
3370 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3373 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3374 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3375 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3376 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3377 - small ext3 extents cleanups and fixes (5733)
3378 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3379 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3380 - avoid needless client->OST connect, fix handle mismatch (5317)
3381 - fix DLM error path that led to out-of-sync client, long delays (5779)
3382 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3383 - fix several locking issues related to i_size (5492,5624,5654,5672)
3384 - don't move pending lock onto export if it is already evicted (5683)
3385 - fix kernel oops when creating .foo in unlinked directory (5548)
3386 - fix deadlock in obdfilter statistics vs. object create (5811)
3387 - use time_{before,after} to avoid timer jiffies wrap (5882)
3388 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3389 - Do not leak request if server was not able to process it (5154)
3390 - If mds_open unable to find parent dir, make that negative lookup(5154)
3391 - don't create new directories with extent-mapping (5909, 5936)
3393 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3394 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3395 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3396 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3397 - support for RHEL4 kernel on i686 (5773)
3398 - provide error messages when incompatible logs are encountered (5898)
3400 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3401 * version 1.4.0.10 (1.4.1 release candidate 1)
3403 - don't keep a lock reference when lock is not granted (4238)
3404 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3405 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3406 - drop import inflight refcount on signal_completed_replay error (5255)
3407 - unlock page after async write error during send (3677)
3408 - handle missing objects in filter_preprw_read properly (5265)
3409 - no transno return for symlink open, don't save no-trasno open (3440)
3410 - don't try to complete elan receive that already failed (4012)
3411 - free RPC server reply state on error (5406)
3412 - clean up thread from ptlrpc_start_thread() on error (5160)
3413 - readahead could read extra page into cache that wasn't ejected (5388)
3414 - prevent races in class_attach/setup/cleanup/detach (5260)
3415 - don't dereference de->d_inode after l_dput of de (5458)
3416 - use "int" for stripe value returned from lock_to_stripe (5544)
3417 - mballoc allocation and error-checking fixes in 2.6 (5504)
3418 - block device patches to fix I/O request sizes in 2.6 (5482)
3419 - look up hostnames for IB nals (5602)
3420 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3421 - don't start multiple acceptors for the same port (5277)
3422 - fix incorrect LASSERT in mds_getattr_name (5635)
3423 - export a proc file for general "ping" checking (5628)
3424 - fix "lfs check" to not block when the MDS is down (5628)
3426 - service request history (4965)
3427 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3428 - create an "evict_client" /proc entry on OSTs, like the MDS has
3429 - fix mount usage message, return errors per mount(8) (5168)
3430 - change grep [] to grep "[]" in tests so they work in more UMLs
3431 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3432 - remove ancient LOV_MAGIC_V0 EA support (5047)
3433 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3434 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3435 - allow readahead allocations to fail when low on memory (5383)
3436 - mmap locking landed again, after considerable improvement (2828)
3437 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3439 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3442 - send OST transaction number in read/write reply to free req (4966)
3443 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3444 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3445 - the watchdog thread now runs as interruptible (5246)
3446 - flock/lockf fixes (but it's still disabled, pending 5135)
3447 - don't use EXT3 constants in llite code (5094)
3448 - memory shortage at startup could cause assertion (5176)
3450 - reorganization of lov code
3451 - single portals codebase
3453 - add extents/mballoc support (5025)
3454 - direct I/O reads in the obdfilter (4048)
3455 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3457 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3460 - send OST transaction number in read/write reply to free req (4966)
3461 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3462 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3463 - the watchdog thread now runs as interruptible (5246)
3464 - handle missing objects in filter_preprw_read properly (5265)
3465 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3466 - drop import inflight refcount on signal_completed_replay error (5255)
3467 - unlock page after async write error during send (3677)
3468 - return original error code on reconstructed replies (3761)
3469 - no transno return for symlink open, don't save no-trasno open (3440)
3471 - add pid to ldlm debugging output (4922)
3472 - bump the watchdog timeouts -- we can't handle 30sec yet
3473 - extra debugging for orphan dentry/inode bug (5259)
3475 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3478 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3479 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3480 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3481 - fix NULL dereference in /proc/sys/portals/routes (4827)
3482 - allow failed mdc_close() operations to be interrupted (4561)
3483 - stop precreate on OST before MDS would time out on it (4778)
3484 - don't send partial-page writes before EOF from client (4410)
3485 - discard client grant for sub-page writes on large-page clients (4520)
3486 - don't free dentries not owned by NFS code, check generation (4806)
3487 - fix lsm leak if mds_create_objects() fails (4801)
3488 - limit debug_daemon file size, always print CERROR messages (4789)
3489 - use transno after validating reply (3892)
3490 - process timed out requests if import state changes (3754)
3491 - update mtime on OST during writes, return in glimpse (4829)
3492 - add mkfsoptions to LDAP (4679)
3493 - use ->max_readahead method instead of zapping global ra (5039)
3494 - don't interrupt __l_wait_event() during strace
3496 - add software watchdogs to catch hung threads quickly (4941)
3497 - make lustrefs init script start after nfs is mounted
3498 - fix CWARN/ERROR duplication (4930)
3499 - return async write errors to application if possible (2248)
3500 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3501 - print NAL number in %x format (4645)
3502 - update barely-supported suse-2.4.21-171 series (4842)
3503 - support for sles 9 %post scripts
3504 - support for building 2.6 kernel-source packages
3505 - support for sles km_* packages
3507 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3510 - ignore -ENOENT errors in osc_destroy (3639)
3511 - notify osc create thread that OSC is being cleaned up (4600)
3512 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3513 - reconstruct ost_handle() like mds_handle() (4657)
3514 - create a new thread to do import eviction to avoid deadlock (3969)
3515 - let lconf resolve symlinked-to devices (4629)
3516 - don't unlink "objects" from directory with default EA (4554)
3517 - hold socknal file ref over connect in case target is down (4394)
3518 - allow more than 32000 subdirectories in a single directory (3244)
3519 - fix blocks count for O_DIRECT writes (3751)
3520 - OST returns ENOSPC from object create when no space left (4539)
3521 - don't send truncate RPC if file size isn't changing (4410)
3522 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3523 - bind to privileged port in socknal and tcpnal (3689)
3525 - rate limit CERROR/CWARN console message to avoid overload (4519)
3526 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3527 - basic mmap support (3918)
3528 - kernel patch series update from b1_4 (4711)
3530 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3533 - avoid crash during MDS cleanup with OST shut down (2775)
3534 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3535 - don't use bad inodes on the MDS (3744)
3536 - dynamic object preallocation to improve recovery speed (4236)
3537 - don't hold spinlock over lock dumping or change debug flags (4401)
3538 - don't zero obd_dev when it is force cleaned (3651)
3539 - print grants to console if they go negative (4431)
3540 - "lctl deactivate" will stop automatic recovery attempts (3406)
3541 - look for existing locks in ldlm_handle_enqueue() (3764)
3542 - don't resolve lock handle twice in recovery avoiding race (4401)
3543 - revalidate should check working dir is a directory (4134)
3545 - don't always mark "slow" obdfilter messages as errors (4418)
3547 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3550 - don't close LustreDB during write_conf until it is done (3860)
3551 - fix typo in lconf for_each_profile (3821)
3552 - allow dumping logs from multiple threads at one time (3820)
3553 - don't allow multiple threads in OSC recovery (3812)
3554 - fix debug_size parameters (3864)
3555 - fix mds_postrecov to initialize import for llog ctxt (3121)
3556 - replace config semaphore with spinlock (3306)
3557 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3558 - don't allow enqueue to complete on a destroyed export (3822)
3559 - down write_lock before checking llog header bitmap (3825)
3560 - recover from lock replay timeout (3764)
3561 - up llog sem before sending rpc (3652)
3562 - reduce ns lock hold times when setting kms (3267)
3563 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3564 - fix NULL deref and obd_dev leak on setup error (3312)
3565 - replace some LBUG about llog ops with error handling (3841)
3566 - don't match INVALID dentries from d_lookup and spin (3784)
3567 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3568 - fix invalid assertion in ptlrpc_set_wait (3880)
3570 - add libwrap support for the TCP acceptor (3996)
3571 - add /proc/sys/portals/routes for non-root route listing (3994)
3572 - allow setting MDS UUID in .xml (2580)
3573 - print the stack of a process that LBUGs (4228)
3575 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3578 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3579 - make sure to unset replay flag from failed open requests (3440)
3580 - if default stripe count is 0, use OST count for inode size (3636)
3581 - update parent mtime/ctime on client for create/unlink (2611)
3582 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3583 - free recovery state on server during a forced cleanup (3571)
3584 - unregister_reply for resent reqs (3063)
3585 - loop back devices mounting and status check on 2.6 (3563)
3586 - fix resource-creation race that can provoke i_size == 0 (3513)
3587 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3588 - more debugging for page-accounting assertion (3746)
3589 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3590 - avoid deadlock after precreation failure (3758)
3591 - fix race and lock order deadlock in orphan handling (3450, 3750)
3592 - add validity checks when grabbing inodes from l_ast_data (3599)
3594 - add /proc/.../recovery_status to obdfilter (3428)
3595 - lightweight CDEBUG infrastructure, debug daemon (3668)
3596 - change default OSC RPC parameters to be better on small clusters
3597 - turn off OST read cache for files smaller than 32MB
3598 - install man pages and include them in rpms (3100)
3599 - add new init script for (un)mounting lustre filesystems (2593)
3600 - run chkconfig in %post for init scripts (3701)
3601 - drop scimac NAL (unmaintained)
3603 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3606 - clean kiobufs before and after use (3485)
3607 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3608 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3609 - change init script to fail more clearly if not run as root (1528)
3610 - allow clients to reconnect during replay (1742)
3611 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3612 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3613 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3614 - fix cond_resched (3554)
3615 - only evict unfinished clients after recovery (3515)
3616 - allow bulk resend, prevent data loss (3570)
3617 - dynamic ptlrpc request buffer allocation (2102)
3618 - don't allow unlinking open directory if it isn't empty (2904)
3619 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3620 - remove extraneous obd dereference causing LASSERT failure (3334)
3621 - don't use get_cycles() when creating temp. files on the mds (3156)
3622 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3623 - handle EEXIST for set-stripe, set proper directory name (3336)
3625 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3626 - fix ksocknal_fmb_callback() error messages (2918)
3628 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3631 - don't copy lvb into (possibly NULL) reply on error (2983)
3632 - don't deref dentry after dput, don't free lvb on error (2922)
3633 - use the kms to determine writeback rpc length (2947)
3634 - increment oti_logcookies when osc is inactive (2948)
3635 - update client's i_blocks count via lvb messages (2543)
3636 - handle intent open/close of special files properly (1557)
3637 - mount MDS with errors=remount-ro, like obdfilter (2009)
3638 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3639 - don't use cancelling-locks' kms values (2947)
3640 - use highest lock extent for kms, not last one (2925)
3641 - don't dereference ERR_PTR() dentry in error handling path (3107)
3642 - fix thread race in portals_debug_dumplog() (3122)
3643 - create lprocfs device entries at setup instead of at attach (1519)
3644 - common AST error handler, don't evict client on completion race (3145)
3645 - zero nameidata in detach_mnt in 2.6 (3118)
3646 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3647 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3648 - initialize RPC timeout timer earlier for 2.6 (3219)
3649 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3650 - print nal/nid for unknown nid (3258)
3651 - additional checks for oscc recovery before doing precreate (3284)
3652 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3653 - don't crash in mdc_close for bad permissions on open (3285)
3654 - zero i_rdev for non-device files (3147)
3655 - clear page->private before handing to FS, better assertion (3119)
3656 - tune the read pipeline (3236)
3657 - fix incorrect decref of invalidated dentry (2350)
3658 - provide read-ahead stats and refine rpc in flight stats (3328)
3659 - don't hold journal transaction open across create RPC (3313)
3660 - update atime on MDS at close time (3265)
3661 - close LDAP connection when recovering to avoid server load (3315)
3662 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3663 - don't leak open file on MDS after open resend (3325)
3664 - serialize filter_precreate and filter_destroy_precreated (3329)
3665 - loop device shouldn't call sync_dev() for nul device (3092)
3666 - clear page cache after eviction (2766)
3667 - resynchronize MDS->OST in background (2824)
3668 - refuse to mount the same filesystem twice on same mountpoint (3394)
3669 - allow llmount to create routes for mounting behind routers (3320)
3670 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3671 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3672 - fix rare problem with rename on htree directories (3417)
3674 - allow default OST striping configuration per directory (1414)
3675 - fix compilation for qswnal for 2.6 kernels (3125)
3676 - increase maximum number of MDS request buffers for large systems
3677 - change liblustreapi to be useful for external progs like lfsck (3098)
3678 - increase local configuration timeout for slow disks (3353)
3679 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3681 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3684 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3685 - don't overwrite extent policy data in reply if lock was blocked (2901)
3686 - drop filter export grants atomically with removal from device (2663)
3687 - del obd_self_export from work_list in class_disconnect_exports (2908)
3688 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3689 - swab reply message in mdc_close, other PPC fixes (2464)
3690 - fix destroying of named logs (2325)
3691 - overwrite old logs when running lconf --write_conf (2264)
3692 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3693 - fix race in target_handle_connect (2898)
3694 - mds_reint_create() should take same inode create lock (2926)
3695 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3696 - don't close files for self_export to avoid uninitialized obd (2936)
3697 - allow MDS with the same name as client node (2939)
3698 - hold dentry reference for closed log files for unlink (2325)
3699 - reserve space for all logs during transactions (2059)
3700 - don't evict page beyond end of stripe extent (2925)
3701 - don't oops on a deleted current working directory (2399)
3702 - handle hard links to targets without a parent properly (2517)
3703 - don't dereference NULL lock when racing during eviction (2867)
3704 - don't grow lock extents when lots of conflicting locks (2919)
3706 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3709 - account for cache space usage on clients to avoid data loss (974)
3710 - lfsck support in lustre kernel code (2349)
3711 - reduce journal credits needed for BRW writes (2370)
3712 - orphan handling to avoid losing space on client/server crashes
3713 - ptlrpcd can be blocked, stopping ALL progress (2477)
3714 - use lock value blocks to assist in proper KMS, faster stat (1021)
3715 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3716 - recovery for initial connections (2355)
3717 - fixes for mds_cleanup_orphans (1934)
3718 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3719 - block all file creations until orphan recovery completes (1901)
3720 - client remove rq_connection from request struct (2423)
3721 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3722 - recovery timer race (2670)
3723 - mdc_close recovey bug (2532)
3724 - ptlrpc cleanup bug (2710)
3725 - mds timeout on local locks (2588)
3726 - namespace lock held during RPCs (2431)
3727 - handle interrupted sync write properly (2503)
3728 - don't try to handle a message that hasn't been replied to (2699)
3729 - client assert failure during cleanup after abort recovery (2701)
3730 - leak mdc device after failed mount (2712)
3731 - ptlrpc_check_set allows timedout requests to complete (2714)
3732 - wait for inflight reqs when ptlrpcd finishes (2710)
3733 - make sure unregistered services are removed from the srv_list
3734 - reset bulk XID's when resending them (caught by 1138 test)
3735 - unregister_bulk after timeout
3736 - fix lconf error (2694)
3737 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3738 - readahead locks pages, leaves pending causing memory pressure (2673)
3739 - increase OST request buffers to 4096 on large machines (2729)
3740 - fix up permission of existing directories in simple_mkdir (2661)
3741 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3742 - don't assign transno to errored transactions (2742)
3743 - don't delete objects on OST if given a bogus objid from MDS (2751)
3744 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3745 - if rq_no_resend, then timeout request after recovery (2432)
3746 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3747 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3748 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3749 - MDS recovery shouldn't delete objects that it has given out (2730)
3750 - if enqueue arrives after completion, don't clobber LVB (2819)
3751 - don't unlock pages twice when trigger_group_io returns error (2814)
3752 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3753 - don't write pages to disk if there was an error (1450)
3754 - don't ping imports that have recovery disabled (2676)
3755 - take buffered bytes into account when balancing socknal conn (2817)
3756 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3757 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3758 - remove little-endian swabbing of llog records (1987)
3759 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3760 - retry reposting request buffers if they fail (1191)
3761 - grow extent at grant time to avoid granting a revoked lock (2809)
3762 - lock revoke doesn't evict page if covered by a second lock (2765)
3763 - disable VM readahead to avoid reading outside lock extents (2805)
3765 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3766 - updated kernel patches for hp-2.4.20 kernel (2681)
3768 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3771 - fix truncated write corruption (2366)
3772 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3774 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3775 - ensure len(uuid) < 37 in lmc (1171)
3776 - fix ia64 OOPS in llog_test (2255)
3777 - zero end of page at obdfilter for partial page writes (2648)
3778 - don't leave stale dentries around after renames (bug 2428)
3779 - fix timeouts when evicting a client with a single lock held (2642)
3780 - set deadline for the initial HELLO message to drain (2634)
3781 - print out dotted-quad IP addresses in the socknal (2302)
3783 - additional debugging for MDS client eviction problem (2443)
3784 - fix mkfsoptions support for osts (2603, 2604)
3786 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3789 - add series for the vanilla 2.6.0 kernel
3790 - add series for the vanilla 2.4.24 kernel
3791 - add series for a cray x86/64 UL kernel drop
3792 - fix xattr patches for the vanilla 2.4.19 series
3794 - generate true UUIDs in lmc (1171)
3795 - have portals stack dumping break in UML (2466)
3796 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3797 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3798 - wake callback waiting threads on client eviction (2460)
3799 - Add --ptldebug and --subsystem to lmc (1719)
3800 - update assertion to allow safe interrupt allocation
3801 - set rq_no_resend for cancel requests (2432)
3802 - recalculate ptlrpcd timeout after resend (2494)
3803 - call vfs_rmdir when removing pending directories (2368)
3804 - fix renaming a file to itself (2429)
3805 - lmc creates a default one-stripe lov (2454)
3806 - expand procfs space to handle large clusters (2326)
3807 - increase UML stack to avoid overflow
3808 - update lconf's list of debug and subsystem masks
3809 - fix lfs find --obd (2510)
3810 - /proc tunable for disabling filter read caching (2591)
3811 - stop rpm packages from altering slapd.conf (2301)
3812 - disable nagle in the socknal under 0conf (2578)
3813 - choose mds inode size based on stripe count (2572)
3814 - fix kernel-source rpm problems (2516)
3816 - add --disable-doc to avoid pdf generation (2421)
3817 - update documentation, tests, type-os, comments
3818 - avoid format warnings on ia64
3819 - remove the TOE NAL
3820 - tiny code cleanups by removing unused fields
3822 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3825 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3826 - remove the most blatant lies from BUILDING file (2371)
3827 - change default debug level to reasonable production setting
3828 - reduce client side cache size to reduce cache flush time
3829 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3830 - make TCP zerocopy and pinger support enabled by default (2476)
3831 - sync writes completed after process exits caused crashes (2319)
3832 - maintain correct mount count on the MDS (2356)
3833 - backout 1557, because 2316 wasn't really fixed
3834 - better file I/O statistics gathering in /proc
3835 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3836 - another kernel patch to fix zero-copy TCP function export
3837 - don't take duplicate lock when processing re-sent getattr (2420)
3838 - lctl uses obd_self_export instead of creating new conn (2353)
3839 - MDS/OST recovery case which requires object creation asserted (2425)
3840 - move lfs from /usr/sbin to /usr/bin in packages
3841 - fix race between mds_client_add and mds_client_free (2417)
3842 - use kmalloc instead of slabs in portals (2430)
3843 - don't create duplicate records when a failover MDS is present (2442)
3844 - remove unnecessary mount age check (2332)
3845 - don't remove directory inodes from locks prematurely (2451)
3846 - don't break if MDS service name is the same as hostname (2103)
3847 - fix races in client write RPC generation when cache full (2482)
3849 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3852 - remove now-unused request->rq_obd (278)
3853 - if an allocation fails, print out how much memory we've used (1933)
3854 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3855 - add the "configurable stack size" patch to most series files (1256)
3856 - ability to write large log records, for 100+ OST configs (2306)
3857 - fix NULL deref when filter_prep fails (2314)
3858 - fix operator precedence error in filter_sync
3859 - dynamic allocation of socknal TX descriptors (2315)
3860 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3861 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3862 - ensure that CWARN messages in Portals always get to the syslog
3863 - __init/__exit are not for prototype decls (ldlm_init/exit)
3864 - x86-64 compile warning fixes
3865 - fix gateway LMC keyword conflict (2318)
3866 - fix MDS lock inversions in getattr/reint paths (1844)
3867 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3868 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3869 - DLM race condition prevented some lock evictions (2328)
3870 - ENOMEM detection and retry on socknal sends (2230)
3871 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3872 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3873 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3874 - better error messages when a client is rejected during recovery (1505)
3875 - avoid cancelling locks which were never granted, after failure (2330)
3876 - fix i_sem/journal inversion in mds_client_add (2333)
3877 - fix truncate/getattr lock cycle deadlock (2334)
3878 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3879 - fix two rare exit paths which could leak an l_lock() ref (2321)
3880 - fencepost error in MDS/OST orphan recovery (2226)
3881 - make log record alignment 8 bytes (1988)
3882 - lstripe now fails when requested offset > ost_count (2237)
3883 - ensure that all kernel series have a complete list.h (1607)
3884 - fix crashes in special-file operations (2316)
3885 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3887 - allow configurable automake binary, for testing new versions
3888 - small update to the lfs documentation
3890 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3892 * fix negative export reference count in fsfilt_sync (2312)
3894 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3895 * release candidate 0.9.1
3897 - orphans are moved into the PENDING directory for possible recovery
3898 - replayed opens now open by fid for orphan/rename safety (1042)
3899 - last close of an orphan inode generates a transno (683)
3900 - chdir() and mount() now pin the directory entry (1020)
3901 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3902 - discard very old requests without processing them (1502)
3903 - remove some common, well-understood CERRORs (1505)
3904 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3905 - clear "grant" flags in OST replies until OST grant code lands (1644)
3906 - fix read performance by not clobbering i_blksize on client (1598)
3907 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3908 - make LRU size a /proc tunable, clears locks when reduced (707)
3909 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3910 - server locks take references on exports now (1558)
3911 - build fixes for 2.4.20-rh trees (1663)
3912 - return an error from lov_create if all OSCs are inactive (1751)
3913 - fix import levels when a reconnect happens without a timeout (1597)
3914 - exit early from mds_open if we get a lookup error (1749)
3915 - partial page read at EOF wouldn't wait for disk before sending (1642)
3916 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3917 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3918 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3919 - write/truncate lock inversion (1639)
3920 - Don't auto-load obdclass, portals modules during cleanup (1495)
3921 - fix timestamps from jumping to "now" (1763)
3922 - extra journal assertions (1648)
3923 - add an extra multiunlink test (1771)
3924 - fix read_record/write_record API (1776)
3925 - fix leak of offset_extent, possible incorrect i_size later (1772)
3926 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3927 - Debugging for the kqswnal_get_idle_tx problems (1820)
3928 - Allow recovery to be attempted multiple times (1536)
3929 - Write out MDS last_rcvd file after it is first created (1600)
3930 - Fix tx_descriptor leak in failed transmit situations (1827)
3931 - ext3 journaling fixes for assertion failure after IO error (1871)
3932 - class_export_put() on freed export after completion AST error (1896)
3933 - Fix revalidate looping in VFS (1322)
3934 - Don't access a freed export during MDS_REINT timeout (1521)
3935 - Add open-unlink recovery support on the MDS (1673,1764)
3936 - Return an error if no MDS data was read from last_rcvd (1946)
3937 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3938 - Don't setup a disk device that doesn't match exported UUID (317)
3939 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3940 - avoid committing NULL handle in force close
3941 - local.sh is now a one-stripe LOV configuration
3942 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3943 - fix ext3 htree duplicate directory entry corruption (1516)
3944 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3945 - update to new LOV EA format (2097)
3946 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3948 2003-06-15 Phil Schwan <phil@clusterfs.com>
3951 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3952 - per-import/export recovery handling (958, 931, 959)
3953 - multiple last-rcvd slots, for serving multiple FSes (949)
3954 - connections are again shared between multiple imp/exports (963, 964)
3955 - "umount -f" would hang if any requests needed to be sent (393, 978)
3956 - avoid pinning large req buffer by copying for queued messages (989)
3957 - add "uuid" to "lctl device" command to help upcalls (991)
3958 - "open" RPCs with transnos would confuse recovery counters (1037)
3959 - do proper endian conversion of all wire messages (288, 340, 891)
3960 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3961 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3962 - fix ptlrpc_connection leak in target_handle_connect (1174)
3963 - fix import refcounting bug in OST and MDS cleanup (1134)
3964 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3965 - fix very unlikely obd_types race condition (501)
3966 - remove osc_open hack for echo_client (1187)
3967 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3968 - a failure in read_inode2 leads to deadlock (1139)
3969 - cancel ack-locks as soon as transaction is committed (1072)
3970 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3971 - make sure to commitrw after any preprw to avoid deadlock (1162)
3972 - failing to execute a file in a lustre FS would lock inode (1203)
3973 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3974 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3975 - obd_uuid2tgt would walk off the end of the list (1255)
3976 - on IA64 the osc would give portals incorrect bulk size (1258)
3977 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3978 - fix lock inversion caused by new llite matching code (1282)
3979 - limit the number of dirty pages on a client to 10MB (1286)
3980 - timed out locks were not being corrected cancelled (1289)
3981 - fix O_DIRECT above 4GB on IA-32 (1292)
3982 * major user-visible changes
3983 - fail out/fail over policy now controlled by the upcall (993)
3985 - add OBD_PING to check server availability and failure (954)
3986 - lustre messages are now sent in sending host order (288, 340, 891)
3987 - add eadatalen to MDS getattr reply (340)
3988 - OST read replies may contain second buffer, with per-page status (593)
3990 2003-03-11 Phil Schwan <phil@clusterfs.com>
3993 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3994 - failed open()s could cause deadlock; fixed (867, 869)
3995 - stop cancelling OST locks when files are closed (481)
3996 - overlapping XID spaces caused network corruption (851, 853)
3997 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3998 - setattr_raw added, to do single-RPC, server-side setattrs
3999 - lmc/lconf syntax change for OST UUIDs
4000 - fix crashy race condition between ptlrpc_free_req and osc_close
4001 - don't use request in mdc_enqueue if we hit a timeout (889)
4002 - don't set the inode i_size for regular files from the MDS (896)
4003 - handle out of order completion AST (842)
4004 - don't LBUG if a lock request times out after receiving AST (913)
4005 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4006 - if a bad lock AST arrives, send an error instead of dropping entirely
4007 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4008 - fix leak in bulk IO when only partially completed (899, 900, 926)
4009 - fix O_DIRECT for ia64 (55)
4010 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4011 - C-z after timeout could hang a process forever; fixed (977)
4013 - client-side I/O cache (678, 924, 929, 941, 970)
4015 - READPAGE and SETATTRs which don't take server-side locks get
4018 2003-02-11 Phil Schwan <phil@clusterfs.com>
4021 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4022 - Fix fsfilt_extN_readpage() to read a full page of directory
4023 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4024 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4025 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4026 - Most intent operations were removed, in favour of a new RPC mode
4027 that does a single RPC to the server and bypasses most of the VFS
4028 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4029 - Aggressively cancel local locks on DLM servers
4030 - mds_reint_unlink sends EA to the client if it's the last nlink.
4031 client uses that EA to unlink OST objects.
4032 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4033 - recursive symlinks were fixed (439)
4034 - fixed NULL deref in DEBUG_REQ
4035 - filter_update_lastobjid no longer calls sync, which annoyed extN
4036 - fixed multi-client small-writes to a single file problem (445)
4037 - fixed mtime updates during file writes (607)
4038 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4039 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4040 - fixed Linux OST size reporting problem (444, 656)
4041 - OST now updates object mtime with writes or setattr (607, 619)
4042 - client verifies file size before zeroing page past EOF (445)
4043 - OST now writes last allocated objid to disk with allocation (108)
4044 - LOV on echo now works (409)
4046 - mds_reint_unlink sends a new buffer, with the EA included. this
4047 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4048 set if a regular file was being unlinked, and it was the last link
4049 - use PtlGet from the target for bulk writes (315)
4050 - OST now updates object mtime with writes or setattr (607, 619)
4051 - LDLM now has a grant-time callback to revalidate locked items, if
4053 - Many MDS operations were reorganized to combat race conditions
4055 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4056 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4058 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4061 - Fully reactivate OST imports after reconnection (512, others)
4062 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4063 - More graceful error handling for truncating on dead OST (515)
4064 - Don't error out unless we're actually accessing dead stripes (474)
4065 - Fix garbage sizes when stripes are missing (410)
4066 - LRU counters were broken, causing constant lock purge (433, 432)
4067 - garbage on read from stripes with failed OSTs (441)
4068 - mark OSCs as active before reconnecting during recovery (438)
4069 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4070 - lfind did not preserve OST order in output (443)
4071 - symlinks cause hung clients, incorrect data (439)
4072 - stop dereferencing request after dropping refcount (457)
4073 - don't LASSERT(spin_is_locked) on non-SMP (455)
4074 - fixes for many rename() bugs
4075 - fstat didn't correctly synchronize attributes (399)
4076 - server must handle lock cancellation during blocking AST prep (487)
4077 - bulk descriptors were free()d too soon (511)
4078 - fix paths in lconf, which would load incorrect modules (451, 507)
4079 - fix confusing lconf 'host not found' error message (386)
4080 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4081 - fix race condition in mdc_blocking_ast() for inode access (526)
4082 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4083 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4084 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4085 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4086 - fix lstripe in several ways (kernel side) (527)
4087 - fix request leak in ldlm_cli_enqueue (262)
4088 - incorrect OSC was marked inactive after OST failure
4089 - call mds_fs_cleanup before unmounting filesystem (524)
4090 - fix races between taking ns_lock and ldlm_lock_change_resource
4091 - fix races updating LOV export open file list
4092 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4093 - fix recovery NULL deref in ldlm_cli_cancel_unused
4094 - fix some DLM races by using new hash table for lock handles (419)
4095 - permit the client to specify desired inodes, at replay
4096 - duplicate requests when we queue them for replay reintegration
4097 - fix last_rcvd offset calculation
4098 - sync after each recovered transaction, so we always make progress
4099 - never, not always, ERESTART requests without transnos
4100 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4102 - skip replay if the MDS says that the client is already connected
4103 - don't check for a recovery-enabled export to match lctl's UUID
4104 - don't INC_USE_COUNT for phantom exports
4105 - don't crash when cleaning up phantom exports (567)
4106 - don't double-finish or set replay data for errored mdc_open requests
4107 - abort requests when they time out, so we don't get old replies
4108 - send/receive replies for AST messages again
4109 - if the client says that it doesn't have the lock, cancel it on the
4111 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4112 mark it as destroyed, it will all work out when decref is called
4113 - fix module use counts (22, 581)
4115 - ASTs now expect a reply (server cancels lock on error reply)
4117 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4120 - fix many simultaneous client startup (392)
4121 - fix dentry->d_it clobbering
4122 - credentials weren't being shipped for readdir/getattr operations
4123 - remove invalid assertions triggered during some concurrent MD
4125 - proper Lustre versions added (336, 389)
4126 - fix memory leak for create error case (398)
4127 - fix LOV locking bug that would get cli/srv out of sync
4128 - fix echo client over LOV (409)
4129 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4130 - fix double-O_EXCL intent crash (424)
4131 - avoid sending multiple lock CANCELs (352)
4133 - MDS can do multi-client recovery (modulo bugs in new code)
4135 - many updates, edits, cleanups
4137 2002-11-18 Phil Schwan <phil@clusterfs.com>
4140 - fix null d_it dereference (346)
4141 - fix full OST/dbench hang (333)
4142 - fix permission problem with file removal (286)
4143 - fix removal of OSCs from LOV when they fail
4144 - fix NULL deref during bulk timeout (214)
4145 - fix problems related to multiple filesystems on one MDS (241)
4146 - fixed serious subtle metadata locking bugs
4147 - free locks on clients when inodes are removed due to memory
4149 - fix inode pointer in lock data (285)
4150 - partial support for multiple MDS on a single host (241)
4151 - data locks weren't cancelled at clear_inode time (290, 311)
4152 - intent locks could lead to unbounded lock growth (205)
4153 - added a maximum lock count, an LRU list, and a flusher
4154 - fix multiple rename (365)
4155 - properly abstracted the echo client
4156 - OSC locked 1 byte too many; fixed
4157 - rewrote brw callback code:
4158 - fixed recovery bugs related to LOVs (306)
4159 - fixed too-many-pages-in-one-write crash (191)
4160 - fixed (again) crash in sync_io_timeout (214)
4161 - probably fixed callback-related race (385)
4163 - Add capability to MDS protocol
4164 - LDLM cancellations and callbacks on different portals
4166 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4169 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4170 - defer open object destruction to close time (601981, 138)
4171 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4172 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4173 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4174 - don't call obd_setattr in ll_file_release for destroyed objects
4175 * protocol change to lustre_msg: move |version| and add |flags|
4176 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4177 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4178 * added replay of create, unlink, link and rename operations during
4179 MDS failover; recovery should be much more robust now
4180 * remove failed OSCs from LOVs (only lov_create uses this so far)
4181 * the lustre-HOWTO was brought (more) up to date (582544)
4183 2002-10-23 Phil Schwan <phil@clusterfs.com>
4186 - in-use dentries weren't being reused properly (617851)
4187 - prevent multiple LDLM setup (599178)
4188 - fix LOV size calculations for truncate (617853)
4189 - fix client handling of MDS intent errors (POSIX)
4190 - fix permission bug in lovstripe.c test (624321)
4191 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4192 - truncate past end of file could corrupt data
4193 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4194 - a race in recovery could return ETIMEDOUT to apps (623947)
4195 - building outside the source directory was fixed
4196 * the lustre-HOWTO was brought (more) up to date (582544)
4197 * major progress was made on recovery functionality
4199 2002-10-10 Phil Schwan <phil@clusterfs.com>
4202 - recovery deadlock fix
4203 - rm -rf causes LBUG fix (617817)
4204 - file open by multiple tasks fix (618962)
4205 - directory permissions bugs (602707 and 620007)
4206 - journal_stop fixed with locking (611313)
4207 - O_APPEND failures resolved (618273, perhaps 614459)
4208 - lconf PATH fix (619770)
4209 - IA64 build fix (621450)
4210 - RPC buffer sizes scale with amount of memory
4212 2002-10-01 Phil Schwan <phil@clusterfs.com>
4215 - locks would be cancelled without throwing away data pages,
4216 resulting in inconsistent data (605627)
4217 - inode attributes were not always being refreshed (605627, 612449)
4218 - lconf now continues to cleanup after lctl reports an error
4219 - MDS now enforces user permissions (602707)
4220 - lprocfs cleanup fixed, but not yet enabled (614157)
4221 - fixed infinite server hang, should a client not respond to an AST
4222 - avoid going into recovery if user calls readlink() with a buffer
4223 that's too small (613941)
4224 - AST RPCs no longer require replies (614867) -- this may be changed
4225 - don't crash server if client sends an IOV that's too big (611336)
4226 - fixed lock conversion deadlock (611892)
4227 - fixed the following of symlinks (614622)
4228 * recovery: the server can remove locks from a client that dies, other
4229 clients can make progress
4230 * more extN patch fixes
4231 * compile-time configurable ptlrpc buffer allocations
4233 - collaborative read cache document
4234 - Lustre Lite Performance CDR document-in-progress
4236 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4239 - fix typo in patch-2.4.18
4241 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4244 - clear ptlrpc request each time in handle_incoming_request()
4245 - unlink of files now destroys the object on the OST
4247 2002-09-19 Peter Braam <braam@clusterfs.com>
4249 * add hard link support
4250 * change obdfile creation method
4251 * kernel patch changed
4253 2002-09-19 Peter Braam <braam@clusterfs.com>
4256 - stack overflow bug in extN fixed
4258 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4260 * documentation updates
4261 - add man pages for config tools
4262 - update tests/README to describe testing with new config tools
4263 - finish metadata API descriptions
4264 * bug fixes and cleanups
4265 - statfs workaround for 16TB limit
4266 - LOV stripe allocation improved, can stripe on subset of OSTs
4267 - LOV file size/IO offset was wrong for files > 4GB in size
4268 - object EA data was being dropped, caused files to be unreadable
4269 - memory overflow with non-LOV OST caused memory corruption
4270 - fixed regression tests to work with new config tools, obdfilter
4271 - fixed bug when directory size became larger than 1 block
4272 - fixed bug (for single client case) when PWD was deleted
4273 - invalidate local directory pages when doing intent-based ops
4274 - avoid LDLM oops when lock callback contained bad data
4276 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4278 * documentation updates
4279 * bug fixes and cleanups
4280 - configuration tools
4283 - 64-bit compile warnings
4284 - 64-bit internal statfs data
4286 * test_brw on persistent OST devices
4288 * lprocfs (disabled)
4290 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4292 * documentation updates
4293 * bug fixes and cleanups
4294 * configuration tools
4296 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4299 * many small fixes to 0_5_4
4300 * io/network handling
4301 * thinkos in MDS operations
4303 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4306 * crucial basic fixes to 0.5.3
4307 * IOR, Iozone work over Elan
4310 2002-08-07 Phil Schwan <phil@clusterfs.com>
4311 * version 0_5_3, our first alpha
4312 * we use the new Portals iovs
4313 * documentation updates
4314 * bug fixes and cleanups
4315 * small changes in the DLM wire protocol
4317 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4318 * version 0_5_1 with some initial stability,
4319 * locking on MD and file I/O.
4320 * documentation updates
4321 * several bug fixes since 0.5.0
4322 * small changes in wire protocol
4324 2002-07-18 Phil Schwan <phil@clusterfs.com>
4326 * delivered as Lustre Light Alpha
4327 * fixed a crash after handling invalid MDS requests
4328 * fixed directory pages for architectures with non-4k pages sizes
4330 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4331 * release version v0_4_4
4332 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4334 * Fixes a number of interruption problems with OST operations.
4335 * Update documentation for portals header changes
4336 * Move all wire protocol structs/defines to lustre_idl.h
4337 * Fixes symlink length bug.
4338 * Add tcpdump to repository.
4340 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4341 * release version v0_4_3
4342 * Fixes statfs for inodes on extN.
4343 * Fixes bug in runtests which would delete /etc/hosts.
4344 * Use 64-bit object IDs wherever possible (not into VFS though)
4345 Remove ost_get_info, which is unused by lustre, and out of date.
4347 2002-07-03 Peter Braam <braam@clusterfs.com>
4348 * release version v0_4_2 Fixes a lookup error (type not passed)
4349 * move forward to head of Portals
4350 * move forward to latest Lustre kernel
4352 2002-06-25 Peter Braam <braam@clusterfs.com>
4353 * release version v0_4_1. Hopefully stable on single node use.