1 tbd Sun Microsystems, Inc.
4 2.6.9-67.0.4.EL (RHEL 4),
5 2.6.16.54-0.2.3 (SLES 10),
6 2.6.18-53.1.6.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).
18 Description: Fix replay if there is an un-replied request and open
19 Details : In some cases, older replay request will revert the
20 mcd->mcd_last_xid on MDS which is used to record the client's
23 Severity : enhancement
25 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
27 Severity : enhancement
29 Description: Add rhel5 support to HEAD.
31 Serverity : enhancement
33 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
38 Description: Don't allow skipping OSTs if index has been specified.
39 Details : Don't allow skipping OSTs if index has been specified, make locking
40 in internal create lots better.
44 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
45 Details : access to bitfield in structure is always rounded to long
46 and this produce problem with not atomic change any bit.
50 Description: Lustre make rpms failed.
51 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
52 builds Lustre rpms from tarball.
55 Frequency : rare on shutdown ost
57 Description: If llog cancel was not send before clean_exports phase, this can
58 produce deadlock in llog code.
59 Details : If llog thread has last reference to obd and call class_import_put
60 this produce deadlock because llog_cleanup_commit_master wait when
61 last llog_commit_thread exited, but this never success because was
62 called from llog_commit_thread.
66 Description: allow userland application know is lost one of stripes.
67 Details : fill lvb_blocks with error code on ost and return it to
68 application if error flag found.
72 Description: NULL lov_tgts causing MDS oops
73 Details : more safe checks for NULL lov_tgts for avoid oops.
75 Severity : enhancement
77 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
81 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
82 Details : If an OST is down the MDS will hang indefinitely in
83 obd_statfs_rqset() waiting for the statfs data. While for
84 MDS QOS usage of statfs, it should not stuck in waiting.
86 Severity : enhancement
88 Description: remote_acl support
89 Details : Support ACL-based permission check for remote user.
90 Support setfacl/getfacl for remote user with the utils
91 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
92 the system "{s,g}etfacl" utils.
94 Severity : enhancement
96 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
98 Severity : enhancement
100 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
104 Description: make mgs_setparam() handle fsname containing dash
105 Details : fsname containing a dash does not work with lctl conf_param
107 Severity : enhancement
109 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
111 Severity : enhancement
113 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
117 Frequency : rare, at shutdown
118 Description: access already free / zero obd_namespace.
119 Details : if client_disconnect_export was called without force flag set,
120 and exist connect request in flight, this can produce access to
121 NULL pointer (or already free pointer) when connect_interpret
122 store ocd flags in obd_namespace.
126 Frequency : only at startup
127 Description: not alloc memory with spinlock held.
128 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
129 if any spinlock held.
131 Severity : enhancement
133 Description: make lustre randomly fail allocating memory
134 Details : Make lustre randomly failed allocating memory for testing purpose.
136 Severity : enhancement
138 Description: lost problems with lov objid file
139 Details : Fixes some scability and access to not inited memory problems
140 in work with lov objdid file.
145 Description: lfs find does not continue on file error
146 Details : Continue other files processing when a file/dir is absent.
150 Description: Inconsistent usage of lustre_pack_reply()
151 Details : Standardize the usage of lustre_pack_reply() such that it
152 always generate a CERROR on failure.
155 Frequency : occasional
157 Description: MDS hang or stay in waiting lock
158 Details : If client receive lock with CBPENDING flag ldlm need send lock
159 cancel as separate rpc, to avoid situation when cancel request
160 can't processed due all i/o threads stay in wait lock.
164 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
168 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
172 Description: Update to SLES9 kernel-2.6.5-7.287.3.
176 Description: Update to RHEL4 latest kernel.
178 Severity : enhancement
180 Description: Build SLES10 patchless client fails
181 Details : The configure was broken by run ./configure with
182 --with-linux-obj=.... argument for patchless client. When the
183 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
184 can't search header adequately. Use absolute path such as
185 -I($LINUX)/include instead.
189 Description: interrupt oig_wait produce painc on resend.
190 Details : brw_redo_request can be used for resend requests from ptlrpcd and
191 private set, and this produce situation when rq_ptlrpcd_data not
192 copyed to new allocated request and triggered LBUG on assert
193 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
194 ptlrpcd set if request is changed and can be safety changed to use
199 Description: LASSERT_{REQ,REP}SWAB macros are buggy
200 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
201 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
202 these macros and replace them with their difinitions instead.
206 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
207 Details : Update which_patch & target file for SLES10 latest kernel.
210 Description: Remove client patches from SLES 10 kernel.
211 Details : This causes SLES 10 clients to behave as patchless clients
212 even on a Lustre-patched (server) kernel.
214 Severity : enhancement
216 Description: self-adjustable client's lru lists
217 Details : use adaptive algorithm for managing client cached locks lru
218 lists according to current server load, other client's work
219 pattern, memory activities, etc. Both, server and client
220 side namespaces provide number of proc tunables for controlling
223 Severity : enhancement
225 Description: light-weight GSS support
226 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
227 and reduce performance overhead.
229 Severity : enhancement
231 Description: Linux keyring support
232 Details : Support using service of Linux keyring for Lustre GSS internal
233 context refresh/cache mechanism.
237 Description: Fix errors in lfs documentation
238 Details : Fixes man pages
242 Description: don't use GFP_* in generic Lustre code.
243 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
247 Description: obdclass is limited by single OBD_ALLOC(idarray)
248 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
252 Description: warnings with build patchless client with vanila 2.6.19 and up
253 Details : change old ctl_table style and replace ctl_table/ctl_table_header
254 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
258 Frequency : Only for SLES
259 Description: sanity_quota fail test_1
260 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
261 file, which makes TSTID[2] inunique.
265 Frequency : Only for SLES9
266 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
267 Details : Update target/ChangeLog/which_patch .
271 Frequency : Only for test_10 in sanity-quota.sh
272 Description: a bug in quota test code
273 Details : checking a test flag in a wrong place
277 Frequency : Only for test_13 in sanity-quota.sh
278 Description: a bug in quota test script
279 Details : shouldn't check overfull allocated quota
283 Description: jbd statistics
284 Details : Port older jbd statistics patch for sles10
286 Severity : enhancement
288 Description: Add printing inode info into message about error in writepage.
293 Description: reply_lock_interpret crash due to race with it and lock cancel.
294 Details : Do not replay locks that are being cancelled. Do not reference
295 locks by their address during replay, just by their handle.
299 Frequency : When flocks are used.
300 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
301 Details : Flock locks might destroy just granted lock if it could be merged
302 with another existing flock, this is done in completion handler,
303 so teach ldlm_cli_enquque_fini that this is a valid case for
309 Description: Oops in read and write path when failing to allocate lock.
310 Details : Check if lock allocation failed and return error back.
314 Description: lstripe command fails for valid OST index
315 Details : The stripe offset is compared to lov->desc.ld_tgt_count
316 instead of lov->desc.ld_active_tgt_count.
320 Description: sanity.sh failed test 103
321 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
326 Description: df doesn't work properly if diskfs blocksize != 4K
327 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
331 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
332 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
336 Description: obd_health_check_timeout too short
337 Details : set obd_health_check_timeout as 1.5x of obd_timeout
341 Description: llapi_file_create() does not allow some changes
342 Details : add llapi_file_open() that allows specifying the mode and
343 open flags, and also returns an open file handle.
347 Description: merge and cleanup kernel patches.
348 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
352 Description: Add journal checksum support.(Kernel part)
353 Details : The journal checksum feature adds two new flags i.e
354 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
355 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
356 indicates that the commit block contains the checksum for
357 the blocks described by the descriptor blocks. Now commit
358 record can be sent to disk without waiting for descriptor
359 blocks to be written to disk. This behavior is controlled
360 using JBD2_FEATURE_ASYNC_COMMIT flag.
364 Description: OSS needs mutliple precreate threads
365 Details : Add ability to start more than one create thread per OSS.
369 Description: Sanity.sh test_65k failed: setstripe should have succeeded
370 Details : execute lfs setstripe on client
374 Description: mds_obd_create error creating tmp object
375 Details : When the user sets quota on root, llog will be affected and can't
376 create files and write files.
379 Frequency : Always on ia64 patchless client, and possibly others.
381 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
382 Details : This allows the patchless client to be loaded on architectures
387 Description: RedHat Update kernel for RHEL5
388 Details : Add the kernel config file for RHEL5.
392 Description: RedHat Update kernel for RHEL5
393 Details : Modify the kernel config file more closer RHEL5.
397 Description: Build failure against Centos5 (RHEL5)
398 Details : Define PAGE_SIZE when it isn't present.
402 Description: test_7 on sanity-quota.sh can't be used on separate machines
403 Details : add facet to handle it
407 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
408 Details : with NFS, the anon dentry's parent was set to itself in
409 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
410 corresponding dentry other than use rec->ur_name.
412 Severity : enhancement
414 Description: lfs setstripe enhancement
415 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
419 Description: enable data checksumming by default
420 Details : enable checksum by default, allow --disable-checksum
421 configure option and "-o nochecksum" mount option. Checksums
422 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
426 Description: lustre support for RHEL5
427 Details : Add support for RHEL5.
431 Description: Client eviction due to failover config
432 Details : after a connection loss, the lustre client should attempt to
433 reconnect to the last active server first before trying the
434 other potential connections.
438 Description: when mds and osts use different quota unit(32bit and 64bit),
439 quota will be released repeatly.
440 Details : void sending multiple quota reqs to mds, which will keep the
441 status between the reqs.
445 Description: osts not allocated evenly to files
446 Details : change the condition to increase offset_idx
449 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
451 Description: Data corruption for OSTs that are formatted larger than 2TB
453 Details : When generating the bio request for lustre file writes the
454 sector number would overflow a temporary variable before being
455 used for the IO. The data reads correctly from Lustre (which
456 will overflow in a similar manner) but other file data or
457 filesystem metadata may be corrupted in some cases.
461 Description: Tune the kernel for good SCSI performance.
462 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
463 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
468 Description: rewrite ext2-derived code in obdclass/uuid.c
469 Details : rewrite inherited code (uuid parsing code from ext2 utils)
470 from scratch preserving functionality.
474 Description: "lfs find -obd UUID" prints directories
475 Details : "lfs find -obd UUID" will return all directory names instead
476 of just file names. It is incorrect because the directories
477 do not reside on the OSTs.
481 Description: use i_size_read and i_size_write in 2.6 port
482 Details : replace inode->i_size access with i_size_read/write()
485 Frequency : when removing large files
487 Description: scheduling issue during removal of large Lustre files
488 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
489 It causes scheduling issues when removing large files (17TB in the
493 Frequency : only with liblustre clients on XT3
495 Description: evictions taking too long
496 Details : allow llrd to evict clients directly on OSTs
499 Frequency : only on ppc
501 Description: /proc/fs/lustre/devices broken on ppc
502 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
503 We should make sure the type of 'index' is loff_t and then cast
504 explicitly as needed below. Do not assign an explicitly cast
508 Frequency : only for rhel5
510 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
511 Details : Modify the target file & which_kernel.
513 Severity : enhancement
515 Description: omit set fsid for export NFS
516 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
517 as FS_REQUIRES_DEV for avoid problems with generate fsid.
521 Frequency : Always, for kernels after 2.6.16
522 Description: Fix warning idr_remove called for id=.. which is not allocated.
523 Details : Last kernels save old s_dev before kill super and not allow
524 to restore from callback - restore it before call kill_anon_super.
528 Description: buffer overruns could theoretically occur
529 Details : llapi_semantic_traverse() modifies the "path" argument by
530 appending values to the end of the origin string, and a
531 overrun may occur. Adding buffer overrun check in liblustreapi.
535 Description: Fix error on 'ls .' at the top of the Lustre mount.
536 Details : Don't revalidate dentry if it is a root dentry.
540 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
541 Details : Modify vm-tunables-rhel4.patch.
545 Description: Kernel config for 2.6.18-vanilla.
546 Details : Modify targets/2.6-vanilla.target.in.
547 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
548 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
549 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
550 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
554 Description: improve handling recoverable errors
555 Details : if request processig with error which can be recoverable on server
556 request should be resend, otherwise page released from cache and
561 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
562 Details : Modify target file & which_patch.
563 A flaw was found in the IA32 system call emulation provided
564 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
565 value could be stored in the %RAX register, which could trigger an
566 out-of-bounds system call table access. An untrusted local user
567 could exploit this flaw to run code in the kernel
568 (ie a root privilege escalation). (CVE-2007-4573).
572 Description: O_DIRECT bypasses client statistics.
573 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
574 read_ahead_stats were updated. copied the stats section from
575 osc_send_oap_rpc() into async_internal().
579 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
583 Description: change order of libsysio includes
584 Details : '#include sysio.h' should always come before '#include xtio.h'
588 Description: handle "serious error: objid * already exists" more gracefully
589 Details : If LAST_ID value on disk is smaller than the objects existing in
590 the O/0/d* directories, it indicates disk corruption and causes an
591 LBUG(). If the object is 0-length, then we should use the existing
592 object. This will help to avoid a full fsck in most cases.
594 Severity : enhancement
596 Description: adapt the lustre_config script to support the upgrade case
597 Details : Add "-u" option for lustre_config script to support upgrading 1.4
598 server targets to 1.6 in parallel.
602 Description: To avoid grant space > avaible space when the disk is almost
603 full. Without this patch you might see the error "grant XXXX >
604 available" or some LBUG about grant, when the disk is almost
606 Details : In filter_check_grant, for non_grant cache write, we should
607 check the left space by if (*left > ungranted + bytes), instead
608 of (*left > ungranted), because only we are sure the left space
609 is enough for another "bytes", then the ungrant space should be
610 increase. In client, we should update cl_avail_grant only there
611 is OBD_MD_FLGRANT in the reply.
615 Description: Update RHEL 4 kernel to fix local root privilege escalation.
616 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
617 described in CVE-2007-4573. This problem could allow untrusted
618 local users to gain root access.
621 Frequency : when using O_DIRECT and quotas
623 Description: Incorrect file ownership on O_DIRECT output files
624 Details : block usage reported by 'lfs quota' does not take into account
625 files that have been written with O_DIRECT.
630 Description: touch file failed when fs is not full
631 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
632 otherwise we can get ENOSP while fs is not full.
636 Description: parallel lock callbacks
637 Details : Instead of sending blocking and completion callbacks as separated
638 requests, adding them to a set and sending in parallel.
641 Frequency : only for Cray XT3
642 Bugzilla : 12829/13455
643 Description: Changing primary group doesn't change the group lustre assigns to
645 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
646 overridden with the primary group provided by the group upcall,
647 whereas the supplied fsgid can be trusted if it is in the list of
648 supplementary groups returned by the group upcall.
650 Severity : enhancement
652 Description: Allow masking D_WARNING, D_ERROR messages from console
653 Details : Console messages can now be disabled via lnet.printk.
657 Description: User code with malformed file open parameter crashes client node
658 Details : Before packing join_file req, all the related reference should be
659 checked carefully in case some malformed flags cause fake
660 join_file req on client.
664 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
665 Details : ldlm_completion_ast() assumes that a lock is granted when the req
666 mode is equal to the granted mode. However, it should also check
667 that LDLM_FL_CP_REQD is not set.
671 Description: Heavy nfs access might result in deadlocks
672 Details : After ELC code landed, it is now improper to enqueue any mds
673 locks under och_sem, because enqueue might want to decide to
674 cancel open locks for same inode we are holding och_sem for.
678 Description: Client eviction while running blogbench
679 Details : A lot of unlink operations with concurrent I/O can lead to a
680 deadlock causing evictions. To address the problem, the number of
681 oustanding OST_DESTROY requests is now throttled to
682 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
683 callbacks are processed in priority.
687 Description: enable ACLs on MDS by default
688 Details : ACLs must be enabled on MDS by default.
690 --------------------------------------------------------------------------------
692 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
694 * Support for kernels:
695 2.6.5-7.283 (SLES 9),
696 2.6.9-55.EL (RHEL 4),
697 2.6.16.46-0.14 (SLES 10),
698 2.6.18.8 vanilla (kernel.org)
699 * Client support for unpatched kernels:
700 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
701 2.6.16 - 2.6.22 vanilla (kernel.org)
702 * Due to recently discovered recovery problems, we do not recommend
703 using patchless RHEL 4 clients with this or any earlier release.
704 * Recommended e2fsprogs version: 1.39.cfs8
705 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
706 * Starting with this release, the ldiskfs backing filesystem required
707 by Lustre is now in its own package, lustre-ldiskfs. This package
708 should be installed. It is versioned separately from Lustre and
709 may be released separately in future.
713 Description: block reactivating mgc import until all deactivates complete
714 Details : Fix race when failing back MDT/MGS to itself (testing)
717 Frequency : at statup only
719 Description: mds_lov_synchronize race leads to various problems
720 Details : simultaneous MDT->OST connections at startup can cause the
721 sync to abort, leaving the OSC in a bad state.
723 Severity : enhancement
725 Description: add optional extra BUILD_VERSION info
726 Details : add a new environment variable (namely LUSTRE_VERS) which allows
727 to override the lustre version.
730 Frequency : 2.6.18 servers only
732 Description: ll_kern_mount() doesn't release the module reference
733 Details : The ldiskfs module reference count never drops down to 0
734 because ll_kern_mount() doesn't release the module reference.
739 Description: server LBUG when using old ost_num_threads parameter
740 Details : Accept the old ost_num_threads parameter but warn that it
741 is deprecated, and fix an off-by-one error that caused an LBUG.
746 Description: Transient SCSI error results in persistent IO issue
747 Details : iobuf->dr_error is not reinitialized to 0 between two
751 Frequency : sometimes when underlying device returns I/O errors
753 Description: OSTs not going read-only during write failures
754 Details : OSTs are not remounted read-only when the journal commit threads
755 get I/O errors because fsfilt_ext3 calls journal_start/stop()
756 instead of the ext3 wrappers.
760 Description: poor connect scaling with increasing client count
761 Details : Don't run filter_grant_sanity_check for more than 100 exports
762 to improve scaling for large numbers of clients.
765 Frequency : SLES10 only
767 Description: sanity-quota.sh quotacheck failed: rc = -22
768 Details : Quotas cannot be enabled on SLES10.
771 Frequency : liblustre clients only
773 Description: getdirentries does not give error when run on compute nodes
774 Details : getdirentries does not fail when the size specified as an argument
775 is too small to contain at least one entry
777 Severity : enhancement
779 Description: Add LNET router traceability for debug purposes
780 Details : If a checksum failure occurs with a router as part of the
781 IO path, the NID of the last router that forwarded the bulk data
782 is printed so it can be identified.
787 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
788 Details : Due to a race condition, liblustre clients were occasionally
791 Severity : enhancement
793 Description: lfs setstripe use optional parameters instead of postional
796 Severity : enhancement
798 Description: Nanosecond timestamp support for ldiskfs
799 Details : The on-disk ldiskfs filesystem has added support for nanosecond
800 resolution timestamps. There is not yet support for this at
801 the Lustre filesystem level.
804 Frequency : during server recovery
806 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
807 Details : request with rq_no_resend flag not awake l_wait_event if they get a
811 Frequency : nfs export on patchless client
813 Description: connectathon hang when test nfs export over patchless client
814 Details : Disconnected dentry cannot be found with lookup, so we do not need
815 to unhash it or make it invalid
818 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
819 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
820 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
825 Description: open req refcounting wrong on reconnect
826 Details : If reconnect happened between getting open reply from server and
827 call to mdc_set_replay_data in ll_file_open, we will schedule
828 replay for unreferenced request that we are about to free.
829 Subsequent close will crash in variety of ways.
830 Check that request is still eligible for replay in
831 mdc_set_replay_data().
836 Description: disable writes to filesystem when reading health_check file
837 Details : the default for reading the health_check proc file has changed
838 to NOT do a journal transaction and write to disk, because this
839 can cause reads of the /proc file to hang and block HA state
840 checking on a healthy but otherwise heavily loaded system. It
841 is possible to return to the previous behaviour during configure
842 with --enable-health-write.
844 Severity : enhancement
846 Description: 64-bit inode version
847 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
848 to the inode. This will be required for version-based recovery.
853 Description: MDS fails to start if a duplicate client export is detected
854 Details : in some rare cases it was possible for a client to connect to
855 an MDS multiple times. Upon recovery the MDS would detect this
856 and fail during startup. Handle this more gracefully.
858 Severity : enhancement
860 Description: Add -o localflock option to simulate old noflock
862 Details : This will achieve local-only flock/fcntl locks
868 Description: log_commit_thread vs filter_destroy race leads to crash
869 Details : Take import reference before releasing llog record semaphore
874 Description: Wrong request locking in request set processing
875 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
876 imp_delayed_list, in this place should be used imp_lock.
879 Frequency : when reconnection
881 Description: Grant Leak when osc reconnect to OST
882 Details : When osc reconnect ost, OST(filter) should check whether it
883 should grant more space to client by comparing fed_grant and
884 cl_avail_grant, and return the granted space to client instead
885 of "new granted" space, because client will call osc_init_grant
886 to update the client grant space info.
889 Frequency : when client reconnect to OST
891 Description: Grant Leak when osc do resend and replay bulk write
892 Details : When osc reconnect to OST, OST(filter)should clear grant info of
893 bulk write request, because the grant info while be sync between
894 OSC and OST when reconnect, and we should ignore the grant info
895 these of resend/replay write req.
900 Description: Grant space more than avaiable left space sometimes.
901 Details : When then OST is about to be full, if two bulk writing from
902 different clients came to OST. Accord the avaliable space of the
903 OST, the first req should be permitted, and the second one
904 should be denied by ENOSPC. But if the seconde arrived before
905 the first one is commited. The OST might wrongly permit second
906 writing, which will cause grant space > avaiable space.
909 Frequency : when client is evicted
911 Description: Grant might be wrongly erased when osc is evicted by OST
912 Details : when the import is evicted by server, it will fork another
913 thread ptlrpc_invalidate_import_thread to invalidate the
914 import, where the grant will be set to 0. While the original
915 thread will update the grant it got when connecting. So if
916 the former happened latter, the grant will be wrongly errased
917 because of this race.
922 Description: Checking Stale with correct fid
923 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
924 to check whether it is stale, because sometimes, we want the
925 enqueue happened anyway, and op_data.fid2 will not be initialized.
927 Severity : enhancement
929 Description: update patchless client
930 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
933 Frequency : only with 2.4 kernel
935 Description: random memory corruption
936 Details : size of struct ll_inode_info is to big for union inode.u and this
937 can be cause of random memory corruption.
942 Description: Memory leak in recovery
943 Details : Lov_mds_md was not free in an error handler in mds_create_object.
944 It should also check obd_fail before fsfilt_start, otherwise if
945 fsfilt_start return -EROFS,(failover mds during mds recovery).
946 then the req will return with repmsg->transno = 0 and rc = EROFS.
947 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
948 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
949 be freed no matter whether fsfilt_commit success or not.
952 Frequency : only with huge count clients
954 Description: Prevents from taking the superblock lock in llap_from_page for
956 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
957 for llap_from_page call in ll_removepage prevents from taking the
958 superblock lock for a soon died page.
963 Description: Not check open intent error before release open handle
964 Details : in some rare cases, the open intent error is not checked before
965 release open handle, which may cause
966 ASSERTION(open_req->rq_transno != 0), because it tries to release
967 the failed open handle.
972 Description: Set cat log bitmap only after create log success.
973 Details : in some rare cases, the cat log bitmap is set too early. and it
974 should be set only after create log success.
978 Description: Accessing a block bevice can re-enable I/O when Lustre is
979 tearing down a device.
980 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
985 Description: service threads may hog cpus when there are a lot of requests
986 Details : Insert cond_resched to give other threads a chance to use some CPU
991 Description: the cat log was not initialized in recovery
992 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
993 unlink log on mds will not be initialized until mds post
994 recovery. And also in mds post recovery, the unlink log will
995 initialization will be done asynchronausly, so there will be race
996 between add unlink log and unlink log initialization.
1000 Description: brw_stats were being printed incorrectly
1001 Details : brw_stats were being printed as log2 but all of them were not
1002 recorded as log2. Also remove some code duplication arising from
1003 filter_tally_{read,write}.
1007 Frequency : rare, only in recovery.
1008 Description: ASSERTION(req->rq_type != LI_POISON) failed
1009 Details : imp_lock should be held while iterating over imp_sending_list for
1010 prevent destroy request after get timeout in ptlrpc_queue_wait.
1014 Description: replay-single.sh test 52 fails
1015 Details : A lock's skiplist need to be cleanup when it being unlinked
1016 from its resource list.
1020 Description: Short directio read returns full requested size rather than
1022 Details : Direct I/O operations should return actual amount of bytes
1023 transferred rather than requested size.
1025 Severity : enhancement
1027 Description: metadata RPC reduction (e.g. for rm performance)
1028 Details : decrease the amount of synchronous RPC between clients and servers
1029 by canceling conflicing lock before the operation on the client
1030 and packing thier handles into the main operation RPC to server.
1032 Severity : enhancement
1034 Description: Async OSC create to avoid the blocking unnecessarily.
1035 Details : If a OST has no remain object, system will block on the creating
1036 when need to create a new object on this OST. Now, ways use
1037 pre-created objects when available, instead of blocking on an
1038 empty osc while others are not empty. If we must block, we block
1039 for the shortest possible period of time.
1041 Severity : enhancement
1043 Description: refine locking for avoid write wrong info into lov_objid file
1044 Details : fix possible races with add new target and write/update data in
1047 --------------------------------------------------------------------------------
1049 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1054 Frequency : on some architectures
1056 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1057 Details : Uninitialized flags sometimes cause configuration commands to
1061 Frequency : patchless clients only
1063 Description: missing __iget() symbol export
1064 Details : The __iget() symbol export is missing. To avoid the need for
1065 this on patchless clients the deathrow inode reaper is turned
1066 off, and we depend on the VM to clean up old inodes. This
1067 dependency was during via the fix for bug 12181.
1069 --------------------------------------------------------------------------------
1071 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1073 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1074 INTEROPERATE with older versions automatically. In many cases a
1075 special upgrade step is needed. Please read the
1076 user documentation before upgrading any part of a live system.
1077 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1078 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1079 * WARNING: Lustre configuration and startup changes are required with
1080 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1086 Severity : enhancement
1088 Description: MountConf
1089 Details : Lustre configuration is now managed via mkfs and mount
1090 commands instead of lmc and lconf. New obd types (MGS, MGC)
1091 are added for dynamic configuration management. See
1092 https://mail.clusterfs.com/wikis/lustre/MountConf for
1095 Severity : enhancement
1097 Description: dynamic OST addition
1098 Details : OSTs can now be added to a live filesystem
1100 Severity : enhancement
1102 Description: startup order invariance
1103 Details : MDTs and OSTs can be started in any order. Clients only
1104 require the MDT to complete startup.
1106 Severity : enhancement
1108 Description: parallel, asynchronous orphan cleanup
1109 Details : orphan cleanup is now performed in separate threads for each
1110 OST, allowing parallel non-blocking operation.
1112 Severity : enhancement
1114 Description: optimized stripe assignment
1115 Details : stripe assignments are now made based on ost space available,
1116 ost previous usage, and OSS previous usage, in order to try
1117 to optimize storage space and networking resources.
1119 Severity : enhancement
1121 Description: Permanently set tunables
1122 Details : All writable /proc/fs/lustre tunables can now be permanently
1123 set on a per-server basis, at mkfs time or on a live system.
1125 Severity : enhancement
1127 Description: Lustre message v2
1128 Details : Add lustre message format v2.
1130 Severity : enhancement
1132 Description: client OST exclusion list
1133 Details : Clients can be started with a list of OSTs that should be
1134 declared "inactive" for known non-responsive OSTs.
1138 Description: SPEC SFS validation failure on NFS v2 over lustre.
1139 Details : Changes the blocksize for regular files to be 2x RPC size,
1140 and not depend on stripe size.
1142 Severity : enhancement
1144 Description: Multiple MD RPCs in flight.
1145 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1146 To avoid overly-overloading MDS, introduce a limit on number of
1147 MDS RPCs in flight for a single client and add /proc controls
1148 to adjust this limit.
1150 Severity : enhancement
1152 Description: client read/write statistics
1153 Details : Add client read/write call usage stats for performance
1154 analysis of user processes.
1155 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1156 file access. extents_stats shows chunk size distribution.
1157 extents_stats_per_process show chunk size distribution per
1160 Severity : enhancement
1162 Description: mds statistics
1163 Details : Add detailed mds operations statistics in
1164 /proc/fs/lustre/mds/*/stats.
1168 Description: Failure of copying files with lustre special EAs.
1169 Details : Client side always return success for setxattr call for lustre
1170 special xattr (currently only "trusted.lov").
1175 Description: Refcount LNET uuids
1176 Details : The global LNET uuid list grew linearly with every startup;
1177 refcount repeated list entries instead of always adding to
1180 Severity : enhancement
1182 Description: Dynamic service threads
1183 Details : Within a small range, start extra service threads
1184 automatically when the request queue builds up.
1187 Frequency : mixed-endian client/server environments
1189 Description: mixed-endian crashes
1190 Details : The new msg_v2 system had some failures in mixed-endian
1193 Severity : enhancement
1195 Description: Easy OST removal
1196 Details : OSTs can be permanently deactivated with e.g. 'lctl
1197 conf_param lustre-OST0001.osc.active=0'
1199 Severity : enhancement
1201 Description: MGS proc entries
1202 Details : Added basic proc entries for the MGS showing what filesystems
1205 Severity : enhancement
1207 Description: provide MGS failover
1208 Details : Added config lock reacquisition after MGS server failover.
1210 Severity : enhancement
1212 Description: add Linux 2.4 support
1213 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1217 Description: a large application tries to do I/O to the same resource and dies
1218 in the middle of it.
1219 Details : Check the req->rq_arrival time after the call to
1220 ost_brw_lock_get(), but before we do anything about
1221 processing it & sending the BULK transfer request. This
1222 should help move old stale pending locks off the queue as
1223 quickly as obd_timeout.
1226 Frequency : when an incorrect nid is specified during startup
1228 Description: ptlrpc connect to non-existant node causes kernel crash
1229 Details : LNET can't be re-entered from an event callback, which
1230 happened when we expire a message after the export has been
1231 cleaned up. Instead, hand the zombie cleanup off to another
1234 Severity : enhancement
1236 Description: plain/inodebits lock performance improvement
1237 Details : Grouping plain/inodebits in granted list by their request modes
1238 and bits policy, thus improving the performance of search through
1242 Frequency : only if OST filesystem is corrupted
1244 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1245 Details : for a short time RPCs with bulk IO are in the replay list,
1246 but replay of bulk IOs is unimplemented. If the OST filesystem
1247 is corrupted due to disk cache incoherency and then replay is
1248 started it is possible to trip an assertion. Avoid putting
1249 committed RPCs into the replay list at all to avoid this issue.
1252 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1254 Description: System hang on startup
1255 Details : This bug allowed the liblustre (e.g. catamount) client to
1256 return to the app before handling all startup RPCs. This
1257 could leave the node unresponsive to lustre network traffic
1258 and manifested as a server ptllnd timeout.
1260 Severity : enhancement
1262 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1263 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1264 boolean to control whether to print peer debug info when a
1265 client's RPC times out.
1268 Frequency : only for kernels with patches from Lustre below 1.4.3
1270 Description: Remove old rdonly API
1271 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1274 Frequency : only for devices with external journals
1276 Description: Set external device read-only also
1277 Details : During a commanded failover stop, we set the disk device
1278 read-only while the server shuts down. We now also set any
1279 external journal device read-only at the same time.
1282 Frequency : when upgrading from 1.4 while trying to change parameters
1284 Description: The wrong (new) MDC name was used when setting parameters for
1285 upgraded MDT's. Also allows changing of OSC (and MDC)
1286 parameters if --writeconf is specified at tunefs upgrade time.
1289 Frequency : when setting specific ost indicies
1291 Description: QOS code breaks on skipped indicies
1292 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1293 created with --index need not be sequential.
1295 Severity : enhancement
1297 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1298 Details : The uninit_groups feature works in conjunction with the kernel
1299 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1300 up the pass1 processing of e2fsck. This is a read-only feature
1301 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1302 mount filesystems that have had this feature enabled.
1304 Severity : enhancement
1306 Description: Improve multi-block allocation algorithm to avoid fragmentation
1307 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1308 allocation locality and avoid filesystem fragmentation.
1310 ------------------------------------------------------------------------------
1312 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1314 * Support for kernels:
1315 2.4.21-47.0.1.EL (RHEL 3)
1316 2.6.5-7.283 (SLES 9)
1317 2.6.9-42.0.10.EL (RHEL 4)
1318 2.6.12.6 vanilla (kernel.org)
1319 2.6.16.27-0.9 (SLES 10)
1320 * Recommended e2fsprogs version: 1.39.cfs5
1322 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1326 Frequency : occasional, depends on client load and configuration
1327 Bugzilla : 12181, 12203
1328 Description: data loss for recently-modified files
1330 Details : In some cases it is possible that recently written or created
1331 files may not be written to disk in a timely manner (this should
1332 normally be within 30s unless client IO load is very high).
1333 The problem appears as zero-length files or files that are a
1334 multiple of 1MB in size after a client crash or client eviction
1335 that are missing data at the end of the file.
1337 This problem is more likely to be hit on clients where files are
1338 repeatedly created and unlinked in the same directory, clients
1339 have a large amount of RAM, have many CPUs, the filesystem has
1340 many OSTs, the clients are rebooted frequently, and/or the files
1341 are not accessed by other nodes after being written.
1343 The presence of the problem can be detected by looking at
1344 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1345 smaller than the second (nr_unused) then dirty files will not
1346 be flushed automatically to disk. "sync; sleep 10" should be
1347 run several times on the node before unmounting it to update
1348 Lustre (this is also safe to run on nodes without this problem).
1350 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1351 that can cause this same problem, so customers using that kernel
1352 also need to update the kernel in addition to Lustre. In order
1353 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1355 It is normal that files written just before a client crash (less
1356 than 30s) may not yet have been flushed to disk, even for local
1360 Frequency : frequent on thin XT3 nodes
1362 Description: UUID collision on thin XT3 Linux nodes
1363 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1364 randomly, since we relied on an insufficiently-seeded PRNG.
1369 Description: OSS hangs after "All ost request buffers busy"
1370 Details : A deadlock between quota and journal operations caused OSS
1371 hangs after printing "All ost request buffers busy."
1374 Frequency : always on liblustre builds
1376 Description: Cleanup compiler warnings on liblustre
1379 Frequency : always on liblustre builds on XT3
1381 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1385 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1386 > 32000 subdirectories
1387 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1388 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1389 subdirectory count crosses 32000. This will aid e2fsck to
1390 correctly handle more than 32000 subdirectories.
1393 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1395 Description: System hang on startup
1396 Details : This bug allowed the liblustre (e.g. catamount) client to
1397 return to the app before handling all startup RPCs. This
1398 could leave the node unresponsive to lustre network traffic
1399 and manifested as a server ptllnd timeout.
1401 Severity : enhancement
1403 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1404 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1405 boolean to control whether to print peer debug info when a
1406 client's RPC times out.
1411 Description: make O_SYNC working on 2.6 kernels
1412 Details : 2.6 kernels use different method for mark pages for write,
1413 so need add a code to lustre for O_SYNC work.
1418 Description: Failure to close file and release space on NFS
1419 Details : Put inode details into lock acquired in ll_intent_file_open.
1420 Use mdc_intent_lock in ll_intent_open to properly
1421 detect all kind of errors unhandled by mdc_enqueue.
1426 Description: proc file read during shutdown sometimes raced obd removal,
1428 Details : Add lock to prevent obd access after proc file removal.
1431 Frequency : Only for files larger than 4GB on 32-bit clients.
1433 Description: improperly doing page alignment of locks
1434 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1435 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1440 Description: RPCs being resent when they shouldn't be
1441 Details : Some RPCs that should not be resent are being resent. This
1442 can cause inconsistencies in the RPC state machine. Do not
1443 resend such requests.
1446 Frequency : rare, only with NFS export
1448 Description: Crash on NFS re-export node
1449 Details : under very unusual load conditions an assertion is hit in
1450 ll_intent_file_open()
1453 Frequency : only if OST filesystem is corrupted
1455 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1456 Details : for a short time RPCs with bulk IO are in the replay list,
1457 but replay of bulk IOs is unimplemented. If the OST filesystem
1458 is corrupted due to disk cache incoherency and then replay is
1459 started it is possible to trip an assertion. Avoid putting
1460 committed RPCs into the replay list at all to avoid this issue.
1465 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1466 Details : Large single O_DIRECT read and write calls can fail to allocate
1467 a sufficiently large buffer to process the request. In case of
1468 allocation failure the allocation is retried with a smaller
1469 buffer and broken into smaller requests.
1471 Severity : enhancement
1473 Description: Add -o localflock option to simulate old noflock behaviour.
1474 Details : This will achieve local-only flock/fcntl locks coherentness.
1479 Description: versioning check is incomplete
1480 Details : Checking the version difference of client vs. server, report
1481 error if the gap is too big.
1486 Description: add support PG_writeback bit
1487 Details : add support for PG_writeback bit for Lustre, for more carefull
1488 work with page cache in 2.6 kernel. This also fix some deadlocks
1489 and remove hack for work O_SYNC with 2.6 kernel.
1491 Severity : enhancement
1493 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1494 Details : The uninit_groups feature works in conjunction with the kernel
1495 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1496 up the pass1 processing of e2fsck. This is a read-only feature
1497 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1498 mount filesystems that have had this feature enabled.
1500 Severity : enhancement
1502 Description: Improve multi-block allocation algorithm to avoid fragmentation
1503 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1504 allocation locality and avoid filesystem fragmentation.
1506 ------------------------------------------------------------------------------
1508 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1510 * Support for kernels:
1511 2.6.9-42.0.3.EL (RHEL 4)
1512 2.6.5-7.276 (SLES 9)
1513 2.4.21-47.0.1.EL (RHEL 3)
1514 2.6.12.6 vanilla (kernel.org)
1515 2.6.16.21-0.8 (SLES10)
1516 * Recommended e2fsprogs version: 1.39.cfs2-0
1518 * The backwards-compatible /proc/sys/portals symlink has been removed
1519 in this release. Before upgrading, please ensure that you change
1520 any configuration scripts or /etc/sysctl.conf files that access
1521 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1522 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1523 in advance of the upgrade on any system running Lustre 1.4.6 or
1524 newer, since /proc/sys/lnet was added in that version.
1525 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1529 Frequency : only when quota is used
1531 Description: avoid scanning export list for quota master
1532 Details : Change the algorithms to avoid scanning export list in order
1533 to improve the efficiency.
1536 Frequency : MDS failover only, very rarely
1538 Description: "went back in time" messages on mds failover
1539 Details : The greatest transno may be lost when the current operation
1540 finishes with an error (transno==0) and the client's last_rcvd
1541 record is over-written. Save the greatest transno in the
1542 mds_last_transno for this case.
1545 Frequency : always for specific kernels and striping counts
1547 Description: client may get "Matching packet too big" without ACL support
1548 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1549 when trying to access files in certain configurations. The
1550 clients should in fact be denied when mounting because they do
1551 not understand ACLs.
1554 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1556 Description: many clients connecting with IO in progress causes connect timeouts
1557 Details : Avoid synchronous journal commits to avoid delays caused by many
1558 clients connecting/disconnecting when bulk IO is in progress.
1559 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1560 OST_IO_PORTAL to avoid delays behind potentially many pending
1564 Frequency : occasionally with multiple writers to a single file
1566 Description: shared writes to file may result in wrong size reported by stat()
1567 Details : Allow growing of kms when extent lock is cancelled
1570 Frequency : always with random mmap IO to multi-striped file
1572 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1573 Details : Only if the hole is at the end of OST object so that kms is too
1574 small. Fix is to increase kms accordingly in ll_nopage.
1577 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1579 Description: writes to a missing object would leak memory on the OST
1580 Details : If there is an inconsistency between the MDS and OST filesystems,
1581 such that the MDS references an object that doesn't exist, writes
1582 to that object will leak memory due to incorrect cleanup in the
1583 error handling path, eventually running out of memory on the OST.
1588 Description: Creating too long symlink causes lustre errors
1589 Details : Check symlink and name lengths before sending requests to MDS.
1592 Frequency : only if flock is enabled (not on by default)
1594 Description: posix locks not released on fd closure on 2.6.9+
1595 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1596 kernels, this caused such locks not to be released on fd close and
1597 then assertions on fs unmount about still used locks.
1600 Frequency : MDS failover only, very rarely
1602 Description: clients may get ASSERTION(granted_lock != NULL)
1603 Details : When request was taking a long time, and a client was resending
1604 a getattr by name lock request. The were multiple lock requests
1605 with the same client lock handle and
1606 mds_getattr_name->fixup_handle_for_resent_request found one of the
1607 lock handles but later failed with ASSERTION(granted_lock != NULL).
1612 Description: handle->h_buffer_credits > 0, assertion failure
1613 Details : h_buffer_credits is zero after truncate, causing assertion
1614 failure. This patch extends the transaction or creates a new
1618 Frequency : NFS re-export or patchless client
1619 Bugzilla : 11179, 10796
1620 Description: Crash on NFS re-export node (__d_move)
1621 Details : We do not want to hash the dentry if we don't have a lock.
1622 But if this dentry is later used in d_move, we'd hit uninitialised
1623 list head d_hash, so we just do this to init d_hash field but
1624 leave dentry unhashed.
1627 Frequency : NFS re-export or patchless client
1629 Description: NFS exports has problem with symbolic link
1630 Details : lustre client didn't properly install dentry when re-exported
1631 to NFS or running patchless client.
1634 Frequency : NFS re-export or patchless client
1636 Description: Various nfs/patchless fixes.
1637 Details : fixes reuse disconected alias for lookup process - this fixes
1638 warning "find_exported_dentry: npd != pd",
1639 fix permission error with open files at nfs.
1640 fix apply umask when do revalidate.
1643 Frequency : occasional
1645 Description: Crash on NFS re-export node
1646 Details : calling clear_page() on the wrong pointer triggered oops in
1647 generic_mapping_read().
1650 Frequency : rarely, using O_DIRECT IO
1652 Description: unaligned directio crashes client with LASSERT
1653 Details : check for unaligned buffers before trying any requests.
1656 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1658 Description: stale data returned from RAID cache
1659 Details : If only a small amount of IO is done to the RAID device before
1660 reading it again it is possible to get stale data from the RAID
1661 cache instead of reading it from disk.
1664 Frequency : always for sles10 kernel
1666 Description: sles10 support
1667 Details : ll_follow_link: compile fixes and using of nd_set_link
1668 under newer kernels.
1671 Frequency : depends on arch, kernel and compiler version, always on sles10
1674 Description: recursive or deep enough symlinks cause stack overflow
1675 Details : getting rid of large stack-allocated variable in
1679 Frequency : depends on hardware
1681 Description: lustre write performance loss in the SLES10 kernel
1682 Details : the performance loss is caused by using of write barriers in the
1683 ext3 code. The SLES10 kernel turns barrier support on by
1684 default. The fix is to undo that change for ldiskfs.
1686 ------------------------------------------------------------------------------
1688 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1690 * Support for kernels:
1691 2.6.9-42.0.3EL (RHEL 4)
1692 2.6.5-7.276 (SLES 9)
1693 2.4.21-47.0.1.EL (RHEL 3)
1694 2.6.12.6 vanilla (kernel.org)
1698 Frequency : quota enabled and large files being deleted
1700 Description: releasing more than 4GB of quota at once hangs OST
1701 Details : If a user deletes more than 4GB of files on a single OST it
1702 will cause the OST to spin in an infinite loop. Release
1703 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1708 Description: statfs data retrieved from /proc may be stale or zero
1709 Details : When reading per-device statfs data from /proc, in the
1710 {kbytes,files}_{total,free,avail} files, it may appear
1711 as zero or be out of date.
1714 Frequency : systems with MD RAID1 external journal devices
1716 Description: lconf's call to blkid is confused by RAID1 journal devices
1717 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1718 of returning all block devices that match the journal UUID.
1721 Frequency : always, for aggregate stripe size over 4GB
1723 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1724 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1725 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1728 Frequency : always if "lfs find" used on a local file/directory
1730 Description: "lfs find" segfaults if used on a local file/directory
1731 Details : The case where a directory component was not specified wasn't
1732 handled correctly. Handle this properly.
1735 Frequency : always on ppc64
1737 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1738 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1739 perform endianness conversion.
1742 Frequency : rarely (truncate to non-zero file size after write under load)
1743 Bugzilla : 10730, 10687
1744 Description: Files padded with zeros to next 4K multiple
1745 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1746 it is possible that files that are truncated to a non-zero size
1747 immediately after being written are filled with zero bytes beyond
1748 the truncated size. No file data is lost.
1750 Severity : enhancement
1752 Description: Allow recovery/failover for liblustre clients.
1753 Details : liblustre clients were unaware of failover configurations until
1756 Severity : enhancement
1758 Description: user file locks should fail when not mounting with flock option
1759 Details : Set up an error-returning stub in ll_file_operations.lock field
1760 to prevent incorrect behaviour when client is mounted without
1761 flock option. Also, set up properly f_op->flock field for
1765 Frequency : always on ia64
1767 Description: "lfs df" loops on printing out MDS statfs information
1768 Details : The obd_ioctl_data was not initialized and in some systems
1769 this caused a failure during the ioctl that did not return
1770 an error. Initialize the struct and return an error on failure.
1773 Frequency : SLES 9 only
1775 Description: Error of copying files with lustre special EAs as root
1776 Details : Client side always return success for setxattr call for lustre
1777 special xattr (currently only "trusted.lov").
1780 Frequency : rarely on clusters with both ia64+i386 clients
1782 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1783 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1784 clients are doing concurrent writes to the same file it is
1785 possible that the ia64 clients may LASSERT because the OST
1786 extent locks are not PAGE_SIZE aligned. Ensure that grown
1787 locks are always aligned on the request boundary.
1790 Frequency : specific use, occasional
1792 Description: Overwriting in use executable truncates on-disk binary image
1793 Details : If one node attempts to overwrite an executable in use by
1794 another node, we now correctly return ETXTBSY instead of
1795 truncating the file.
1797 Severity : enhancement
1799 Description: Async OSC create to avoid the blocking unnecessarily.
1800 Details : If a OST has no remain object, system will block on the creating
1801 when need to create a new object on this OST. Now, ways use
1802 pre-created objects when available, instead of blocking on an
1803 empty osc while others are not empty. If we must block, we block
1804 for the shortest possible period of time.
1809 Description: chmod on Lustre root is propagated to other clients
1810 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1811 invalid by the follow_mount time.
1814 Frequency : liblustre clients only
1816 Description: Race in 'instant cancel' lock handling could lead to such locks
1817 never to be granted in case of SMP MDS
1818 Details : Do not destroy not yet granted but cbpending locks in
1822 Frequency : replay/resend of open
1824 Description: non null lock assetion failure in mds_intent_policy
1825 Details : Trying to replay/resend lockless open requests resulted in
1826 mds_open() returning 0 with no lock. Now it sets a flag if
1827 a lock is going to be returned.
1829 Severity : enhancement
1831 Description: Checksum enhancements
1832 Details : New checksum enhancements allow for resending RPCs that failed
1835 Severity : enhancement
1837 Description: Tunables on number of dirty pages in cacche
1838 Details : Allow to set limit on number of dirty pages cached.
1843 Description: client crash on unmount - lock still has references
1844 Details : In some error handling cases it was possible to leak a lock
1845 reference on a client while accessing a file. This was not
1846 harmful to the client during operation, but would cause the
1847 client to crash when the filesystem is unmounted.
1850 Frequency : specific case, rare
1852 Description: ETXTBSY on mds though file not in use by client
1853 Details : ETXTBSY is no longer incorrectly returned when attempting to
1854 chmod or chown a directory that the user previously tried to
1855 execute or a currently-executing binary.
1858 Frequency : extremely rare except on liblustre-based clients
1860 Description: Lustre space not freed when files are deleted
1861 Details : Clean up open-unlinked files after client eviction. Previously
1862 the unlink was skipped and the files remained as orphans.
1867 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1868 Details : In some cases it was possible to send a blocking callback to
1869 a client doing a glimpse, even though that client didn't get
1870 a lock granted. When the glimpse lock is cancelled on the OST
1871 the freed lock is left on the waiting list and corrupted the list.
1874 Frequency : all core dumps
1876 Description: Broke core dumps to lustre
1877 Details : Negative dentry may be unhashed if parent does not have UPDATE
1878 lock, but some callers, e.g. do_coredump, expect dentry to be
1879 hashed after successful create, hash it in ll_create_it.
1881 ------------------------------------------------------------------------------
1883 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1885 * Support for kernels:
1886 2.6.9-42.0.2.EL (RHEL 4)
1887 2.6.5-7.276 (SLES 9)
1888 2.4.21-40.EL (RHEL 3)
1889 2.6.12.6 vanilla (kernel.org)
1893 Frequency : always on RHEL 3
1895 Description: Number of open files grows over time
1896 Details : The number of open files grows over time, whether or not
1897 Lustre is started. This was due to a filp leak introduced
1898 by one of our kernel patches.
1900 ------------------------------------------------------------------------------
1902 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1904 * Support for kernels:
1905 2.6.9-42.EL (RHEL 4)
1906 2.6.5-7.276 (SLES 9)
1907 2.4.21-40.EL (RHEL 3)
1908 2.6.12.6 vanilla (kernel.org)
1913 Bugzilla : 5719, 9635, 9792, 9684
1914 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1915 Details : If a server is under heavy load and cannot reply to new
1916 connection requests before the client resends the (re)connect,
1917 the connection handling code can behave badly if two service
1918 threads are concurrently handing separate (re)connections from
1919 the same client. Add better locking to the connection handling
1920 code, and ensure that only a single connection will be processed
1921 for a given client UUID, even if the lock is dropped.
1923 Severity : enhancement
1925 Description: add TCP zero-copy support to kernel
1926 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1927 sends if the hardware supports scatter-gather and checksumming.
1928 This allows socklnd to do client-write and server-read more
1929 efficiently and reduce CPU utilization from skbuf copying.
1932 Frequency : only if NFS exporting from client
1934 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1935 Details : If setting attributes on a file created under NFS that had
1936 never been opened it would be possible to oops the client
1937 if the file had no objects.
1940 Frequency : always for liblustre
1942 Description: liblustre client does MDS+OSTs setattr RPC for each write
1943 Details : When doing a write from a liblustre client, the client
1944 incorrectly issued an RPC to the MDS and each OST the file was
1945 striped over in order to update the timestamps. When writing
1946 with small chunks and many clients this could overwhelm the MDS
1947 with RPCs. In all cases it would slow down the write because
1948 these RPCs are unnecessary.
1950 Severity : enhancement
1952 Description: allow number of MDS service threads to be changed at module load
1953 Details : It is now possible to change the number of MDS service threads
1954 running. Adding "options mds mds_num_threads={N}" to the MDS's
1955 /etc/modprobe.conf will set the number of threads for the next
1956 time Lustre is restarted (assuming the "mds" module is also
1957 reloaded at that time). The default number of threads will
1958 stay the same, 32 for most systems.
1963 Description: OST crash if filesystem is unformatted or corrupt
1964 Details : If an OST is started on a device that has never been formatted
1965 or if the filesystem is corrupt and cannot even mount then the
1966 error handling cleanup routines would dereference a NULL pointer.
1971 Description: NULL pointer deref in llap_from_page.
1972 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1973 incorrect) mapping if the page was truncated/reclaimed while it was
1974 searched for. Check for this condition and skip such pages when
1975 doing readahead. Introduce extra check to llap_from_page() to
1976 verify page->mapping->host is non-NULL (so page is not anonymous).
1979 Frequency : Sometimes when using sys_sendfile
1981 Description: "page not covered by a lock" warnings from ll_readpage
1982 Details : sendfile called ll_readpage without right page locks present.
1983 Now we introduced ll_file_sendfile that does necessary locking
1984 around call to generic_file_sendfile() much like we do in
1988 Frequency : with certain MDS communication failures at client mount time
1990 Description: NULL pointer deref after failed client mount
1991 Details : a client connection request may delayed by the network layer
1992 and not be sent until after the PTLRPC layer has timed out the
1993 request. If the client fails the mount immediately it will try
1994 to clean up before the network times out the request. Add a
1995 reference from the request import to the obd device and delay
1996 the cleanup until the network drops the request.
1999 Frequency : occasionally during client (re)connect
2001 Description: assertion failure during client (re)connect
2002 Details : processing a client connection request may be delayed by the
2003 client or server longer than the client connect timeout. This
2004 causes the client to resend the connection request. If the
2005 original connection request is replied in this interval, the
2006 client may trip an assertion failure in ptlrpc_connect_interpret()
2007 which thought it would be the only running connect process.
2010 Frequency : only with obd_echo servers and clients that are rebooted
2012 Description: kernel BUG accessing uninitialized data structure
2013 Details : When running an obd_echo server it did not start the ping_evictor
2014 thread, and when a client was evicted an uninitialized data
2015 structure was accessed. Start the ping_evictor in the RPC
2016 service startup instead of the OBD startup.
2018 Severity : enhancement
2019 Bugzilla : 10193 (patchless)
2020 Description: Remove dependency on various unexported kernel interfaces.
2021 Details : No longer need reparent_to_init, exit_mm, exit_files,
2022 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2025 Frequency : rare (only users of deprecated and unsupported LDAP config)
2027 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2028 Details : LDAP apparently contains 'lustreName' attributes instead of
2029 'name'. A simple remapping of the name is sufficient.
2032 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2034 Description: waiting_locks_callback trips kernel BUG if client is evicted
2035 Details : Running with the dump_on_timeout debug flag turned on makes
2036 it possible that the waiting_locks_callback() can try to dump
2037 the Lustre kernel debug logs from an interrupt handler. Defer
2038 this log dumping to the expired_lock_main() thread.
2040 Severity : enhancement
2042 Description: Support NFS exporting on 2.6 kernels.
2043 Details : Implement non-rawops metadata methods for NFS server to use without
2044 changing NFS server code.
2047 Frequency : very rare (synthetic metadata workload only)
2049 Description: two racing renames might cause an MDS thread to deadlock
2050 Details : Running the "racer" program may cause one MDS thread to rename
2051 a file from being the source of a rename to being the target of
2052 a rename at exactly the same time that another thread is doing
2053 so, and the second thread has already enqueued these locks after
2054 doing a lookup of the target and is trying to relock them in
2055 order. Ensure that we don't try to re-lock the same resource.
2058 Frequency : only very large systems with liblustre clients
2060 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2061 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2062 all OSTs in parallel. This allows the eviction of stale liblustre
2063 clients to proceed much faster than if they were done in series,
2064 and also offers similar improvements for other set_info RPCs.
2069 Description: excessive CPU usage during initial read phase on client
2070 Details : During the initial read phase on a client, it would agressively
2071 retry readahead on the file, consuming too much CPU and impacting
2072 performance (since 1.4.5.8). Improve the readahead algorithm
2073 to avoid this, and also improve some other common cases (read
2074 of small files in particular, where "small" is files smaller than
2075 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2080 Description: MDS crash when receiving packet with unknown intent.
2081 Details : Do not LBUG in unknown intent case, just return -EFAULT
2083 Severity : enhancement
2084 Bugzilla : 9293, 9385
2085 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2086 Details : Do not serialize getattr (non-intent version) and statfs.
2089 Frequency : occasional, when OST network is overloaded/intermittent
2091 Description: client evicted by OST after bulk IO timeout
2092 Details : If a client sends a bulk IO request (read or write) the OST
2093 may evict the client if it is unresposive to its data GET/PUT
2094 request. This is incorrect if the network is overloaded (takes
2095 too long to transfer the RPC data) or dropped the OST GET/PUT
2096 request. There is no need to evict the client at all, since
2097 the pinger and/or lock callbacks will handle this, and the
2098 client can restart the bulk request.
2101 Frequency : Always when mmapping file with no objects
2103 Description: client crashes when mmapping file with no objects
2104 Details : Check that we actually have objects in a file before doing any
2105 operations on objects in ll_vm_open, ll_vm_close and
2111 Description: Request leak when working with deleted CWD
2112 Details : Introduce advanced request refcount tracking for requests
2113 referenced from lustre intent.
2115 Severity : Enhancement
2117 Description: Cache open file handles on client.
2118 Details : MDS now will return special lock along with openhandle, if
2119 requested and client is allowed to hold openhandle, even if unused,
2120 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2121 closing files for every read/write openration.
2123 Severity : Enhancement
2125 Description: Cache open negative dentries on client when possible.
2126 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2127 negative dentries on lock revocation.
2132 Description: Remounting a client read-only wasn't possible with a zconf mount
2133 Details : It wasn't possible to remount a client read-only with llmount.
2135 Severity : enhancement
2136 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2137 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2138 MPICH is included to add Lustre-specific ADIO interfaces.
2139 This is based closely on the UFS ADIO layer and only differs
2140 in file creation, in order to allow the OST striping to be set.
2141 This is user-contributed code and not supported by CFS.
2146 Description: extended inode attributes (immutable, append-only) work improperly
2147 when 2.4 and 2.6 kernels are used on client/server or vice versa
2148 Details : Introduce kernel-independent values for these flags.
2150 Severity : enhancement
2153 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2154 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2155 granularity. Now, Fractional MB granularity is be supported,
2156 this is very useful for low memory system.
2158 Severity : enhancement
2160 Description: Getattr by fid
2161 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2162 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2163 have locks and avoids getattr rpc requests for it on every stat.
2166 Frequency : Always, for filesystems larger than 2TB
2168 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2169 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2170 filesystems larger than 2TB to be mounted with mballoc enabled.
2173 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2175 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2176 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2177 filesystems larger than 2TB is mounted then the kernel will
2178 silently corrupt the start of the filesystem. CONFIG_LBD is
2179 enabled for all CFS-supported kernels, but the possibility of
2180 this happening with a modified kernel config exists.
2182 Severity : enhancement
2184 Description: add client O_DIRECT support for 2.6 kernels
2185 Details : It is now possible to do O_DIRECT reads and writes to files
2186 in the Lustre client mountpoint on 2.6 kernel clients.
2188 Severity : enhancement
2190 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2191 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2192 parallel, not waiting for response from every OST before sending
2193 a rpc to the next OST.
2198 Description: setattr vs write race when updating file timestamps
2199 Details : Client processes that update a file timestamp into the past
2200 right after writing to the file (e.g. tar) it is possible that
2201 the updated file modification time can be reset to the current
2202 time due to a race between processing the setattr and write RPC.
2204 Severity : enhancement
2206 Description: Bring 'lfs find' closer in line with regular Linux find.
2207 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2208 -print0 options and obtains all the needed info through the lustre
2211 Severity : enhancement
2213 Description: support up to 1024 configured devices on one node
2214 Details : change obd_dev array from statically allocated to dynamically
2215 allocated structs as they are first used to reduce memory usage
2220 Description: Flush dirty partially truncated pages during truncate
2221 Details : Immediatelly flush partially truncated pages in filter_setattr,
2222 this way we completely avoid having any pages in page cache on OST
2223 and can retire ugly workarounds during writes to flush such pages.
2228 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2229 Details : protect inode from truncation within vfs_unlink() context
2230 just take a reference before calling vfs_unlink() and release it
2231 when parent's i_sem is free.
2236 Description: last_id value checked outside lock on OST caused LASSERT failure
2237 Details : If there were multiple MDS->OST object precreate requests in
2238 flight, it was possible that the OST's last object id was checked
2239 outside a lock and incorrectly tripped an assertion. Move checks
2240 inside locks, and discard old precreate requests.
2243 Frequency : always, if extents are used on OSTs
2245 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2246 Details : OSTs using the extents format would not zero the high 16 bits of
2247 the index physical block number. This is not a problem for any
2248 OST filesystems smaller than 16TB, and no kernels support ext3
2249 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2250 new/modified files) and can be fixed for existing filesystems
2251 with e2fsprogs-1.39-cfs1.
2256 Description: import connection selection may be incorrect if timer wraps
2257 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2258 import connections to be ignored if the 32-bit jiffies counter
2259 wraps. Use a 64-bit jiffies counter.
2262 Frequency : very large clusters immediately after boot
2264 Description: LNET request buffers exhausted under heavy short-term load
2265 Details : If a large number of client requests are generated on a service
2266 that has previously never seen so many requests it is possible
2267 that the request buffer growth cannot keep up with the spike in
2268 demand. Instead of dropping incoming requests, they are held in
2269 the LND until the RPC service can accept more requests.
2272 Frequency : Sometimes during replay
2274 Description: Assertion failure in ll_local_open after replay.
2275 Details : If replay happened on an open request reply before we were able
2276 to set replay handler, reply will become not swabbed tripping the
2277 assertion in ll_local_open. Now we set the handler right after
2278 recognising of open request
2281 Frequency : very rare
2283 Description: kernel reports "badness in vsnprintf"
2284 Details : Reading from the "recovery_status" /proc file in small chunks
2285 may cause a negative length in lprocfs_obd_rd_recovery_status()
2286 call to vsnprintf() (which is otherwise harmless). Exit early
2287 if there is no more space in the output buffer.
2289 Severity : enhancement
2291 Description: clear OBD RPC statistics by writing to them
2292 Details : It is now possible to clear the OBD RPC statistics by writing
2293 to the "stats" file.
2298 Description: Client mtime is not the same on different clients after utimes
2299 Details : In some cases, the client was using the utimes() syscall on
2300 a file cached on another node. The clients now validate the
2301 ctime from the MDS + OSTs to determine which one is right.
2306 Description: Inability to activate failout mode
2307 Details : lconf script incorrectly assumed that in pythong string's numeric
2308 value is used in comparisons.
2311 Frequency : always with multiple stripes per file
2313 Description: Inefficient object allocation for mutli-stripe files
2314 Details : When selecting which OSTs to stripe files over, for files with
2315 a stripe count that divides evenly into the number of OSTs,
2316 the MDS is always picking the same starting OST for each file.
2317 Return the OST selection heuristic to the original design.
2322 Description: mount failures may take full timeout to return an error
2323 Details : Under some heavy load conditions it is possible that a
2324 failed mount can wait for the full obd_timeout interval,
2325 possibly several minutes, before reporting an error.
2326 Instead return an error as soon as the status is known.
2328 Frequency : quota enabled and large files being deleted
2330 Description: releasing more than 4GB of quota at once hangs OST
2331 Details : If a user deletes more than 4GB of files on a single OST it
2332 will cause the OST to spin in an infinite loop. Release
2333 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2338 Description: statfs data retrieved from /proc may be stale or zero
2339 Details : When reading per-device statfs data from /proc, in the
2340 {kbytes,files}_{total,free,avail} files, it may appear
2341 as zero or be out of date.
2344 Frequency : systems with MD RAID1 external journal devices
2346 Description: lconf's call to blkid is confused by RAID1 journal devices
2347 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2348 of returning all block devices that match the journal UUID.
2351 Frequency : always, for aggregate stripe size over 4GB
2353 Description: assertion fails when trying to use 4GB stripe size
2354 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2355 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2358 Frequency : always on ppc64
2360 Description: the first write on an ext3 filesystem with mballoc got stuck
2361 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2362 perform endianness conversion.
2364 ------------------------------------------------------------------------------
2366 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2368 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2369 INTEROPERATE with older versions automatically. Please read the
2370 user documentation before upgrading any part of a live system.
2371 * WARNING: Lustre networking configuration changes are required with
2372 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2375 * Support for kernels:
2376 2.6.9-22.0.2.EL (RHEL 4)
2377 2.6.5-7.244 (SLES 9)
2378 2.6.12.6 vanilla (kernel.org)
2381 Severity : enhancement
2382 Bugzilla : 7981/8208
2383 Description: Introduced Lustre Networking (LNET)
2384 Details : LNET is new networking infrastructure for Lustre, it includes
2385 a reorganized network configuration mode (see the user
2386 documentation for full details) as well as support for routing
2387 between different network fabrics. Lustre Networking Devices
2388 (LNDs) for the supported network fabrics have also been
2389 created for this new infrastructure.
2391 Severity : enhancement
2392 Description: Introduced Access control lists
2393 Details : clients can set ACLs on files and directories in order to have
2394 more fine-grained permissions than the standard Unix UGO+RWX.
2395 The MDS must be started with the "-o acl" mount option.
2397 Severity : enhancement
2398 Description: Introduced filesystem quotas
2399 Details : Administrators may now establish per-user quotas on the
2402 Severity : enhancement
2404 Description: Configuration change for the XT3
2405 The PTLLND is now used to run Lustre over Portals on the XT3
2406 The configure option(s) --with-cray-portals are no longer used.
2407 Rather --with-portals=<path-to-portals-includes> is used to
2408 enable building on the XT3. In addition to enable XT3 specific
2409 features the option --enable-cray-xt3 must be used.
2414 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2415 Details : A race between allocating a new llcd and re-getting the llcd_lock
2416 allowed another thread to grab newly-allocated llcd.
2418 Severity : enhancement
2420 Description: 2.6 OST async journal commit and locking fix to improve performance
2421 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2422 2.6 kernels are now async as they already were in 2.4 kernels so
2423 that they can commit concurrently with the network bulk transfer.
2424 For block-allocated files the filter allocation semaphore is held
2425 to avoid filesystem fragmentation during allocation. BKL lock
2426 removed for 2.6 xattr operations where it is no longer needed.
2431 Description: lconf incorrectly determined whether two IP networks could talk
2432 Details : In some more complicated routing and multiple-network
2433 configurations, lconf will avoid trying to make a network
2434 connection to a disjoint part of the IP space. It was doing the
2435 math incorrectly for one set of cases.
2440 Description: Fix for potential infinite loop processing records in an llog.
2441 Details : If an llog record is corrupted/zeroed, it is possible to loop
2442 forever in llog_process(). Validate the llog record length
2443 and skip the remainder of the block on error.
2446 Frequency : occasional (liblustre only)
2448 Description: liblustre could not open files whose last component is a symlink
2449 Details : sysio_path_walk() would incorrectly pass the open intent to
2450 intermediate path components.
2453 Frequency : rare (liblustre only with non-standard tuning)
2454 Bugzilla : 7201 (7350)
2455 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2456 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2457 after it has been released, only do this for referenced locks
2459 Severity : enhancement
2461 Description: specifying an (invalid) directory default stripe_size of -1
2462 would reset the directory default striping
2463 Details : stripe_size -1 was used internally to signal directory stripe
2464 removal, now use "all default" to signal dir stripe removal
2465 as a directory striping of "all default" is not useful
2468 Frequency : common for large clusters running liblustre clients
2470 Description: doing an ls when liblustre clients are running is slow
2471 Details : sending a glimpse AST to a liblustre client waits for every AST
2472 to time out, as liblustre clients will not respond. Since they
2473 cannot cache data we refresh the OST lock LVB from disk instead.
2475 Severity : enhancement
2477 Description: doing an ls at the same time as file IO can be slow
2478 Details : enqueue and other "small" requests can be blocked behind many
2479 large IO requests. Create a new OST IO portal for non-IO
2480 requests so they can be processed faster.
2483 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2485 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2486 created file with mode 000
2487 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2488 allow an ftruncate() as a non-root user to a file with mode 000.
2489 Lustre now respects this flag to disable mode checks when
2490 truncating a file owned by the user
2493 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2496 Description: Revoking locks from clients that went dead or catatonic might take
2498 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2499 cancellation of such locks instant on servers without waiting for
2500 any reply from clients. Clients drops these locks when cancel
2501 notification from server is received without replying.
2504 Frequency : liblustre-only, when liblustre client dies or becomes busy
2506 Description: Doing ls on Linux clients can take a long time with active
2508 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2509 granting such locks to it in the first place if possible. Locks
2510 are taken by proxy on the OST during the read or write and
2511 dropped immediately afterward. Add connect flags handling, do
2512 not grant locks to liblustre clients for glimpse ASTs.
2514 Severity : enhancement
2516 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2517 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2518 sequential reads and behaves badly with applications doing random
2519 reads. Improve it by reading ahead at least read region, and
2520 avoiding excessive large RPC for small reads.
2522 Severity : enhancement
2524 Description: Creating more than 1000 files for a single job may cause a load
2525 imbalance on the OSTs if there are also a large number of OSTs.
2526 Details : qos_prep_create() uses an OST index reseed value that is an
2527 even multiple of the number of available OSTs so that if the
2528 reseed happens in the middle of the object allocation it will
2529 still utilize the OSTs as uniformly as possible.
2534 Description: OST or MDS may oops in ping_evictor_main()
2535 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2536 but doesn't restart at beginning of obd_exports_timed list
2539 Severity : enhancement
2541 Description: improve by-nid export eviction on the MDS and OST
2542 Details : allow multiple exports with the same NID to be evicted at one
2543 time without re-searching the exports list.
2546 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2548 Description: MDS may oops in groups_free()
2549 Details : in rare race conditions a newly allocated group_info struct is
2550 freed again, and this can be NULL. The 2.4 compatibility code
2551 for groups_free() checked for a NULL pointer, but 2.6 did not.
2554 Frequency : common for liblustre clients doing little filesystem IO
2555 Bugzilla : 9352, 7313
2556 Description: server may evict liblustre clients accessing contended locks
2557 Details : if a client is granted a lock or receives a completion AST
2558 with a blocking AST already set it would not reply to the AST
2559 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2562 Frequency : lfs setstripe, only systems with more than 160 OSTs
2564 Description: unable to set striping with a starting offset beyond OST 160
2565 Details : llapi_create_file() incorrectly limited the starting stripe
2566 index to the maximum single-file stripe count.
2569 Frequency : LDAP users only
2571 Description: lconf did not handle in-kernel recovery with LDAP properly
2572 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2574 Severity : enhancement
2576 Description: bind OST threads to NUMA nodes to improve performance
2577 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2578 node and do their allocations there to localize memory access
2580 Severity : enhancement
2582 Description: llmount can determine client NID directly from Myrinet (GM)
2583 Details : the client NID code from gmnalnid was moved directly into
2584 llmount, removing the need to use this or specifying the
2585 client NID explicitly when mounting GM clients with zeroconf
2588 Frequency : if client is started with down MDS
2590 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2591 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2592 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2593 ptlrpc_import_delayed_req() because we didn't check if the
2594 request was interrupted, and we also didn't break out of the
2595 event loop if there was no timeout
2600 Description: data loss during non-page-aligned writes to a single file from
2601 both multiple nodes and multiple threads on one node at same time
2602 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2603 inconsistency led to false short-reads, that were cached and later
2604 used by ->prepare_write() to fill in partially written page,
2605 leading to data loss.
2608 Frequency : always, if lconf --abort_recovery used
2610 Description: lconf --abort_recovery fails with 'Operation not supported'
2611 Details : lconf was attempting to abort recovery on the MDT device and not
2614 Severity : enhancement
2616 Description: remove cleanup logs
2617 Details : replace lconf-generated cleanup logs with lustre internal
2618 cleanup routines. Eliminates the need for client-cleanup and
2621 Severity : enhancement
2623 Description: add support for EAs (user and system) on lustre filesystems
2624 Details : it is now possible to store extended attributes in the Lustre
2625 client filesystem, and with the user_xattr mount option it
2626 is possible to allow users to store EAs on their files also
2628 Severity : enhancement
2630 Description: Add possibility (config option) to show minimal available OST free
2632 Details : When compiled with --enable-mindf configure option, statfs(2)
2633 (and so, df) will return least minimal free space available from
2634 all OSTs as amount of free space on FS, instead of summary of
2635 free spaces of all OSTs.
2637 Severity : enhancement
2639 Description: do not expand extent locks acquired on OST-side
2640 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2641 by server: they are not cached anyway.
2644 Frequency : when mmap is used/binaries executed from Lustre
2646 Description: Unmmap pages before throwing them away from read cache.
2647 Details : llap_shrink cache now attempts to unmap pages before discarding
2648 them (if unmapping failed - do not discard). SLES9 kernel has
2649 extra checks that trigger if this unmapping is not done first.
2654 Description: lconf didn't resolve symlinks before checking to see whether a
2655 given mountpoint was already in use
2658 Frequency : when migrating failover services
2659 Bugzilla : 6395, 9514
2660 Description: When migrating a subset of services from a node (e.g. failback
2661 from a failover service node) the remaining services would
2662 time out and evict clients.
2663 Details : lconf --force (implied by --failover) sets the global obd_timeout
2664 to 5 seconds in order to quickly disconnect, but this caused
2665 other RPCs to time out too quickly. Do not change the global
2666 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2668 Severity : enhancement
2669 Frequency : if MDS is started with down OST
2670 Bugzilla : 9439,5706
2671 Description: Allow startup/shutdown of an MDS without depending on the
2672 availability of the OSTs.
2673 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2674 Add appropriate locking and lov-osc refcounts for safe
2675 cleaning. Add osc abort_inflight calls in case the
2676 synchronize never started.
2679 Frequency : occasional (Cray XT3 only)
2681 Description: root not authorized to access files in CRAY_PORTALS environment
2682 Details : The client process capabilities were not honoured on the MDS in
2683 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2684 been accessed by an authorized user then root was able to access
2685 the file on the local client also. The root user capabilities
2686 are now allowed on the MDS, as this environment has secure UID.
2689 Frequency : occasional
2691 Description: ldiskfs "too long searching" message happens too often
2692 Details : A debugging message (otherwise harmless) prints too often on
2693 the OST console. This has been reduced to only happen when
2694 there are fragmentation problems on the filesystem.
2699 Description: Division by zero in statfs when all OSCs are inactive
2700 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2701 lov_statfs divides by value returned by lov_get_stripecnt().
2705 Bugzilla : 9489, 3273
2706 Description: First write from each client to each OST was only 4kB in size,
2707 to initialize client writeback cache, which caused sub-optimal
2708 RPCs and poor layout on disk for the first writen file.
2709 Details : Clients now request an initial cache grant at (re)connect time
2710 and so that they can start streaming writes to the cache right
2711 away and always do full-sized RPCs if there is enough data.
2712 If the OST is rebooted the client also re-establishes its grant
2713 so that client cached writes will be honoured under the grant.
2718 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2719 Details : Now I/O RPCs go to different portal number and (presumably) fast
2720 lock requests (and glimses) and other RPCs get their own service
2721 threads pool that should be able to service those RPCs
2724 Severity : enhancement
2726 Description: Ability to exchange lustre version between client and servers and
2727 issue warnings at client side if client is too old. Also for
2728 liblustre clients there is ability to refuse connection of too old
2730 Details : New 'version' field is added to connect data structure that is
2731 filled with version info. That info is later checked by server and
2735 Frequency : rare, liblustre only.
2736 Bugzilla : 9296, 9581
2737 Description: Two simultaneous writes from liblustre at offset within same page
2738 might proceed at the same time overwriting eachother with stale
2740 Details : I/O lock withing llu_file_prwv was released too early, before data
2741 actually was hitting the wire. Extended lock-holding time until
2742 server acknowledges receiving data.
2745 Frequency : extremely rare. Never observed in practice.
2747 Description: avoid generating lustre_handle cookie of 0.
2748 Details : class_handle_hash() generates handle cookies by incrementing
2749 global counter, and can hit 0 occasionaly (this is unlikely, but
2750 not impossible, because initial value of cookie counter is
2751 selected randonly). Value of 0 is used as a sentinel meaning
2752 "unassigned handle" --- avoid it. Also coalesce two critical
2753 sections in this function into one.
2755 Severity : enhancement
2757 Description: allow liblustre clients to delegate truncate locking to OST
2758 Details : To avoid overhead of locking, liblustre client instructs OST to
2759 take extent lock in ost_punch() on client's behalf. New connection
2760 flag is added to handle backward compatibility.
2762 Severity : enhancement
2763 Bugzilla : 4928, 7341, 9758
2764 Description: allow number of OST service threads to be specified
2765 Details : a module parameter allows the number of OST service threads
2766 to be specified via "options ost ost_num_threads={N}" in the
2767 OSS's /etc/modules.conf or /etc/modprobe.conf.
2771 Bugzilla : 6146, 9635, 9895
2772 Description: servers crash with bad pointer in target_handle_connect()
2773 Details : In rare cases when a client is reconnecting it was possible that
2774 the connection request was the last reference for that export.
2775 We would temporarily drop the export reference and get a new
2776 one, but this may have been the last reference and the export
2777 was just destroyed. Get new reference before dropping old one.
2779 Severity : enhancement
2780 Frequency : if client is started with failover MDS
2782 Description: Allow multiple MDS hostnames in the mount command
2783 Details : Try to read the configuration from all specified MDS
2784 hostnames during a client mount in case the "primary"
2787 Severity : enhancement
2789 Description: Stop sending data to evicted clients as soon as possible.
2790 Details : Check if the client we are about to send or are sending data to
2791 was evicted already. (Check is done every second of waiting,
2792 for which l_wait_event interface was extended to allow checking
2793 of exit condition at specified intervals).
2796 Frequency : rare, normally only when NFS exporting is done from client
2798 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2800 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2803 Frequency : always, if srand() is called before liblustre initialization
2805 Description: Liblustre uses system PRNG disturbing its usage by user application
2806 Details : Introduce internal to lustre fast and high-quality PRNG for
2807 lustre usage and make liblustre and some other places in generic
2808 lustre code to use it.
2810 Severity : enhancement
2811 Bugzilla : 9477, 9557, 9870
2812 Description: Verify that the MDS configuration logs are updated when xml is
2813 Details : Check if the .xml configuration logs are newer than the config
2814 logs stored on the MDS and report an error if this is the case.
2815 Request --write-conf, or allow starting with --old_conf.
2817 Severity : enhancement
2819 Description: Handle symlinks in the path when checking if Lustre is mounted.
2820 Details : Resolve intermediate symlinks when checking if a client has
2821 mounted a filesystem to avoid duplicate client mounts.
2826 Description: lconf can hit an error exception but still return success.
2827 Details : The lconf command catches the Command error exception at the top
2828 level script context and will exit with the associated exit
2829 status, but doesn't ensure that this exit status is non-zero.
2834 Description: failure of ptlrpc thread startup can cause oops
2835 Details : Starting a ptlrpc service thread can fail if there are a large
2836 number of threads or the server memory is very fragmented.
2837 Handle this without oopsing.
2840 Frequency : always, only if liblustre and non-default acceptor port was used
2842 Description: liblustre cannot connect to servers with non-default acceptor port
2843 Details : tcpnal_set_default_params() was not called and was therefore
2844 ignoring the environment varaible TCPNAL_PORT, as well as other
2845 TCPNAL_ environment variables
2850 Description: two objects could be created on the same OST for a single file
2851 Details : If an OST is down, in some cases it was possible to create two
2852 objects on a single OST for a single file. No problems other
2853 than potential performance impact and spurious error messages.
2857 Bugzilla : 5681, 9562
2858 Description: Client may oops in ll_unhash_aliases
2859 Details : Client dcache may become inconsistent in race condition.
2860 In some cases "getcwd" can fail if the current directory is
2866 Description: Inode refcounting problems in NFS export code
2867 Details : link_raw functions used to call d_instantiate without obtaining
2868 extra inode reference first.
2872 Bugzilla : 9942, 9903
2873 Description: Referencing freed requests leading to crash, memleaks with NFS.
2874 Details : We used to require that call to ll_revalidate_it was always
2875 followed by ll_lookup_it. Also with revalidate_special() it is
2876 possible to call ll_revalidate_it() twice for the same dentry
2877 even if first occurence returned success. This fix changes semantic
2878 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2879 reference on a request referred from the intent.
2880 ll_intent_release() then releases such a request.
2883 Frequency : rare, normally benchmark loads only
2885 Description: unlinked inodes were kept in memory on the client
2886 Details : If a client is repeatedly creating and unlinking files it
2887 can accumulate a lot of stale inodes in the inode slab cache.
2888 If there is no other client load running this can cause the
2889 client node to run out of memory. Instead flush old inodes
2890 from client cache that have the same inode number as a new inode.
2893 Frequency : SLES9 2.6.5 kernel and long filenames only
2894 Bugzilla : 9969, 10379
2895 Description: utime reports stale NFS file handle
2896 Details : SLES9 uses out-of-dentry names in some cases, which confused
2897 the lustre dentry revalidation. Change it to always use the
2901 Frequency : rare, unless heavy write-truncate concurrency is continuous
2902 Bugzilla : 4180, 6984, 7171, 9963, 9331
2903 Description: OST becomes very slow and/or deadlocked during object unlink
2904 Details : filter_destroy() was holding onto the parent directory lock
2905 while truncating+unlinking objects. For very large objects this
2906 may block other threads for a long time and slow overall OST
2907 responsiveness. It may also be possible to get a lock ordering
2908 deadlock in this case, or run out of journal credits because of
2909 the combined truncate+unlink. Solution is to do object truncate
2910 first in one transaction without parent lock, and then do the
2911 final unlink in a new transaction with the parent lock. This
2912 reduces the lock hold time dramatically.
2915 Frequency : rare, 2.4 kernels only
2917 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2918 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2919 before unmounting their filesystem in order to avoid deadlock.
2920 The kernel_locked() function in 2.4 kernels only checks whether
2921 the kernel lock is held, not whether it is this process that is
2922 holding it as 2.6 kernels do.
2927 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2928 Details : The client ptlrpc code may be trying to reconnect to a down
2929 server before a previous connection attempt has timed out.
2930 Increase the reconnect interval to be longer than the connection
2931 timeout interval to avoid sending duplicate connections to
2935 Frequency : echo_client brw_test command
2937 Description: fix echo_client to work with OST preallocated code
2938 Details : OST preallocation code (5137) didn't take echo_client IO path
2939 into account: echo_client calls filter methods outside of any
2940 OST thread and, hence, there is no per-thread preallocated
2941 pages and buffers to use. Solution: hijack pga pages for IO. As
2942 a byproduct, this avoids unnecessary data copying.
2946 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2947 Description: Client can oops in mdc_commit_close() after open replay
2948 Details : It was possible for the MDS to return an open request with no
2949 transaction number in mds_finish_transno() if the client was
2950 evicted, but without actually returning an error. Clients
2951 would later try to replay that open and may trip an assertion
2952 Simplify the client close codepath, and always return an error
2953 from the MDS in case the open is not successful.
2956 Frequency : rare, 2.6 OSTs only
2958 Description: OST may deadlock under high load on fragmented files
2959 Details : If there was a heavy load and highly-fragmented OST filesystems
2960 it was possible to have all the OST threads deadlock waiting on
2961 allocation of biovecs, because the biovecs were not released
2962 until the entire RPC IO was completed. Instead, release biovecs
2963 as soon as they are complete to ensure forward IO progress.
2965 Severity : enhancement
2967 Description: Support for specifying external journal device at mount
2968 Details : If an OST or MDS device is formatted with an external journal
2969 device, this device major/minor is stored in the ext3 superblock
2970 and may not be valid for failover. Allow detecting and
2971 specifying the external journal at mount time.
2976 Description: Mounting an MDS with pending unlinked files may cause oops
2977 Details : target_finish_recovery() calls mds_postrecov() which returned
2978 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2979 considers this an error and immediately begins cleaning up the
2980 lov, just after starting the mds_lov process
2982 Severity : enhancement
2984 Description: Implement 'lfs df' to report actual free space on per-OST basis
2985 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2986 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2987 usage of objects; '-h' to report in human readable format.
2989 ------------------------------------------------------------------------------
2991 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2998 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2999 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3000 at startup during orphan unlinking. Instead, skip buddy update
3001 before setup, as it will be regenerated after recovery is complete.
3006 Description: If an OST is inactive, its locks might reference stale inodes.
3007 Details : lov_change_cbdata() must iterate over all namespaces, even if
3008 they are inactive to clear inode references from the lock.
3010 Severity : enhancement
3011 Frequency : occasional, if non-standard max_dirty_mb used
3013 Description: Client will block write RPCs if not enough grant
3014 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3015 then the client will block writes while waiting for another RPC
3016 to complete instead of consuming its dirty limit. With change
3017 we get improved performance when max_dirty_mb is small.
3019 Severity : enhancement
3020 Bugzilla : 3389, 6253
3021 Description: Add support for supplementary groups on the MDS.
3022 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3023 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3024 lookups for user supplementary groups into a cache.
3028 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3029 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3030 for liblustre O_EXCL is enforced by MDS.
3034 Description: readdir never returns NULL in liblustre.
3035 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3036 dentry in struct dirent.
3040 Frequency : liblustre only, depends on application IO pattern
3041 Description: liblustre clients evicted if not contacting servers
3042 Details : Don't put liblustre clients into the ping_evictor list, so
3043 they will not be evicted by the pinger ever.
3045 Severity : enhancement
3047 Description: Add ability to evict clients by NID from MDS.
3048 Details : By echoing "nid:$NID" string into
3049 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3050 $NID would be instantly evicted from this MDS and from all active
3051 OSTs connected to it.
3055 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3056 Details : lookup_it_finish() used to query file size from OSTs that was not
3061 Description: service threads change working directory to that of init
3062 Details : Starting lustre service threads may pin the working directory
3063 of the parent thread, making that filesystem busy. Threads
3064 now change to the working directory of init to avoid this.
3068 Frequency : during shutdown only
3069 Description: shutdown with a failed MDS or OST can cause unmount to hang
3070 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3075 Frequency : 2.6 only
3076 Description: chmod/chown may include an extra supplementary group
3077 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3078 supplementary group and if none is specified this is used.
3081 Bugzilla : 5479 (6816)
3083 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3084 Details : If lookup is in progress on a file that is unlinked we might try
3085 to revalidate the inode and fail in revalidate after lookup is
3086 complete and ll_file_open() enqueues the open again but
3087 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3090 Frequency : always, if lconf --abort_recovery used
3092 Description: lconf --abort_recovery fails with 'Operation not supported'
3093 Details : lconf was attempting to abort recovery on the MDT device and not
3096 ------------------------------------------------------------------------------
3098 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3103 Frequency : rare (only unsupported configurations with a node running as an
3105 Bugzilla : 6514, 5137
3106 Description: Mounting a Lustre file system on a node running as an OST could
3108 Details : OSTs now preallocates memory needed to write out data at
3109 startup, instead of when needed, to avoid having to
3110 allocate memory in possibly low memory situations.
3111 Specifically, if the file system is mounted on on OST,
3112 memory pressure could force it to try to write out data,
3113 which it needed to allocate memory to do. Due to the low
3114 memory, it would be unable to do so and the node would
3115 become unresponsive.
3117 Severity : enhancement
3119 Description: Addition of lconf --service command line option
3120 Details : lconf now accepts a '--service <arg>' option, which is
3121 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3123 Severity : enhancement
3125 Description: Failover mode is now the default for OSTs.
3126 Details : By default, OSTs will now run in failover mode. To return to
3127 the old behaviour, add '--failout' to the lmc line for OSTs.
3129 Severity : enhancement
3131 Description: Health checks are now provided for MDS and OSTs
3132 Details : Additional detailed health check information on MSD and OSTs
3133 is now provided through the procfs health_check value.
3136 Frequency : occasional, depends on IO load
3138 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3139 after numerous create/delete cycles
3140 Details : The ext3 inode allocation policy would not allocate new inodes
3141 very well on the OSTs because there are no new directories
3142 being created. Instead we look for groups with free space if
3143 the parent directories are nearly full.
3148 Description: Network or server problems during mount may cause partially
3149 mounted clients instead of returning an error.
3150 Details : The config llog parsing code may overwrite the error return
3151 code during mount error handling, returning success instead
3157 Description: MDS can fail to allocate large reply buffers
3158 Details : After long uptimes the MDS can fail to allocate large reply
3159 buffers (e.g. zconf client mount config records) due to memory
3160 fragmentation or consumption by the buffer cache. Preallocate
3161 some large reply buffers so that these replies can be sent even
3162 under memory pressure.
3166 Frequency : rare (liblustre)
3167 Description: fsx running with liblustre complained that using truncate() to
3168 extend the file doesn't work. This patch corrects that issue.
3169 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3170 ATTR_SIZE and lsm use in llu_setattr_raw.
3174 Frequency : rare, only 2.6 kernels
3175 Description: Unusual file access patterns on the MDS may result in inode
3176 data being lost in very rare circumstances.
3177 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3178 code in the 2.6 kernel caused the inode and/or EA data not to
3179 be read from disk, causing single-file corruption.
3183 Frequency : rare, only 2.6 filesystems using extents
3184 Description: Heavy concurrent write and delete load may cause data corruption.
3185 Details : It was possible under high-load situations to have an extent
3186 metadata block in the block device cache from a just-unlinked
3187 file overwrite a newly-allocated data block. We now unmap any
3188 metadata buffers that alias just-allocated data blocks.
3192 Frequency : filesystems with default stripe_count larger than 77
3193 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3194 Details : lconf specifies an inode size of 4096 bytes when the default
3195 stripe_count is larger than 77. This conflicts with the default
3196 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3197 this case to avoid pinning too much memory for large EAs.
3199 ------------------------------------------------------------------------------
3201 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3206 Frequency : rare (extremely heavy IO load with hundreds of clients)
3208 Description: Client is evicted, gets IO error writing to file
3209 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3210 caused clients to be evicted for AST timeouts. The fixes in
3211 bug 5192 mean we no longer need to have such short AST timeouts
3212 so ldlm_timeout has been increased.
3215 Frequency : occasional during --force or --failover shutdown under load
3216 Bugzilla : 5949, 4834
3217 Description: Server oops/LBUG if stopped with --force or --failover under load
3218 Details : a collection of import/export refcount and cleanup ordering
3219 issues fixed for safer force cleanup
3222 Frequency : only filesystems larger than 120 OSTs
3223 Bugzilla : 5990, 6223
3224 Description: lfs getstripe would oops on a very large filesystem
3225 Details : lov_getconfig used kfree on vmalloc'd memory
3228 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3229 Bugzilla : 6242, 6243
3230 Description: reading from files that had been truncated to a non-zero size
3231 but never opened returned no data
3232 Details : ll_file_read() reads zeros from no-object files to EOF
3237 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3238 Details : There's little harm in aborting MDS/OSS recovery and letting it
3239 try again, so I removed the LASSERT and return an error instead.
3241 Severity : enhancement
3243 Description: New debugging infrastructure for tracking down data corruption
3244 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3245 (b) cover more of the client-side code path, and (c) try to narrow
3246 down where problems occurred
3250 Bugzilla : 3819, 4364, 4397, 6313
3251 Description: Racing close and eviction MDS could cause assertion in mds_close
3252 Details : It was possible to get multiple mfd references during close and
3253 client eviction, leading to one thread referencing a freed mfd.
3255 Severity: : enhancement
3256 Bugzilla : 3262, 6359
3257 Description: Attempts to reconnect to servers are now more aggressive.
3258 Details : This builds on the enhanced upcall-less recovery that was added
3259 in 1.4.2. When trying to reconnect to servers, clients will
3260 now try each server in the failover group every 10 seconds. By
3261 default, clients would previously try one server every 25 seconds.
3266 Description: After recovery, certain operations trigger a failed
3267 assertion on a client.
3268 Details : Failing over an mds, using lconf -d --failover, while a
3269 client was doing a readdir() call would cause the client to
3270 LBUG after recovery completed and the readdir() was resent.
3272 Severity : enhancement
3274 Description: Default groups are now added by lconf
3275 Details : You can now run lconf --group <servicename> without having to
3276 manually add groups with lmc.
3279 Frequency : occasional
3281 Description: Nodes with an elan id of 0 trigger a failed assertion
3284 Frequency : always when accessing e.g. tty/console device nodes
3286 Description: tty and some other devices nodes cannot be used on lustre
3287 Details : file's private_data field is used by device data and lustre
3288 values in there got lost. New field was added to struct file to
3289 store fs-specific private data.
3292 Frequency : when exporting Lustre via NFS
3294 Description: NFSD failed occasionally when looking up a path component
3295 Details : NFSD is looking up ".." which was broken in ext3 directories
3296 that had grown large enough to become hashed.
3299 Frequency : Clusters with multiple interfaces not on the same subnet
3301 Description: Nodes will repeatedly try to reconnect to an interface which it
3302 cannot reach and report an error to the log.
3303 Details : Extra peer list entries will be created by lconf with some peers
3304 unreachable. lconf now validates the peer before adding it.
3307 Frequency : Only if a default stripe is set on the filesystem root.
3309 Description: Setting a default stripe on the filesystem root prevented the
3310 filesystem from being remounted.
3311 Details : The client was sending extra request flags in the root getattr
3312 request and did not allocate a reply buffer for the dir EA.
3315 Frequency : occasional, higher if lots of files are accessed by one client
3316 Bugzilla : 6159, 6097
3317 Description: Client trips assertion regarding lsm mismatch/magic
3318 Details : While revalidating inodes the VFS looks up inodes with ifind()
3319 and in rare cases can find an inode that is being freed.
3320 The ll_test_inode() code will free the lsm during ifind()
3321 when it finds an existing inode and then the VFS later attaches
3322 this free lsm to a new inode.
3326 Bugzilla : 6422, 7030
3327 Description: MDS deadlock between mkdir and client eviction
3328 Details : Creating a new file via mkdir or mknod (starting a transaction
3329 and getting the ns lock) can deadlock with client eviction
3330 (gets ns lock and trying to finish a synchronous transaction).
3333 Frequency : occasional
3334 Description: While starting a server, the fsfilt_ext3 module could not be
3336 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3337 kernels. Previously, lconf would still use the ext3 name
3338 when trying to load modules. Now, it will correctly use
3339 ext3 on 2.4 and ldiskfs on 2.6.
3341 Severity : enhancement
3342 Description: The default stripe count has been changed to 1
3343 Details : The interpretation of the default stripe count (0, to lfs
3344 or lmc) has been changed to mean striping across a single
3345 OST, rather than all available. For general usage we have
3346 found a stripe count of 1 or 2 works best.
3348 Severity : enhancement
3349 Description: Add support for compiling against Cray portals.
3350 Details : Conditional compiling for some areas that are different
3354 Frequency : occasional
3355 Bugzilla : 6409, 6834
3356 Description: Creating files with an explicit stripe count may lead to
3357 a failed assertion on the MDS
3358 Details : If some OSTs are full or unavailable, creating files may
3359 trigger a failed assertion on the MDS. Now, Lustre will
3360 try to use other servers or return an error to the
3364 Frequency : occasional
3366 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3367 caused lock timeout during readahead (since 1.4.2).
3368 Details : Processes doing readahead might match a lock that hasn't been
3369 granted yet if there are overlapping and conflicting lock
3370 requests. The readahead process waits on ungranted lock
3371 (original lock is CBPENDING), while OST waits for that process
3372 to cancel CBPENDING read lock and eventually evicts client.
3374 Severity : enhancement
3376 Description: Initial enabling of flock support for clients
3377 Details : Implements fcntl advisory locking and file status functions.
3378 This feature is provided as an optional mount flag (default
3379 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3380 locking are implemented yet, and those that are are not guaranteed
3381 to be completely correct in production environments.
3382 mount -t lustre -o [flock|noflock] ...
3385 Frequency : occasional
3387 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3388 assertion in the ext3 JBD (journaling) layer.
3389 Details : The b_committed_data struct is protected by the big kernel lock
3390 in 2.4 kernels, serializing journal_commit_transaction() and
3391 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3392 access to this struct. In 2.6 kernels there is finer grained
3393 locking to improve SMP performance of the JBD layer.
3397 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3402 Description: merge and cleanup kernel patches.
3405 -----------------------------------------------------------------------------
3407 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3409 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3410 versions of Lustre. Please update all servers and clients to
3411 version 1.4.2 or later at the same time. You must also run
3412 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3413 to update the configuration logs.
3415 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3416 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3417 - swab llog records of type '0' so we get proper header size/idx (5861)
3418 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3419 - fix rename of one directory over another leaking an inode (5953)
3420 - avoid SetPageDirty on 2.6 (5981)
3421 - don't re-add just-being-destroyed locks to the waiting list (5653)
3422 - when creating new directories, inherit the parent's custom
3423 striping settings if present parent (3048)
3424 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3425 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3426 - don't hold client locks on temporary worklist from l_lru (5666)
3427 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3428 - automatically evict dead clients (5921)
3429 - Update file size properly in create+truncate+fstat case (6196)
3430 - Do not unhash mountpoint dentries, do not allow removal of
3432 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3433 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3434 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3435 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3436 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3437 - Correctly handle reads of files with no objects (6243)
3438 - lctl recover will also mark a device active if deactivate used (5933)
3440 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3441 - allow --write-conf on an MDS with different nettype than client (5619)
3442 - don't write config llogs to MDS for mounts not from that MDS (5617)
3443 - lconf should create multiple TCP connections from a client (5201)
3444 - init scripts are now turned off by default; run chkconfig --on
3445 lustre and chkconfig --on lustrefs to use them
3446 - upcalls are no longer needed for clients to recover to failover
3448 - add --abort-recovery option to lconf to abort recovery on device
3450 - add support for an arbitrary number of OSTs (3026)
3451 - Quota support protocol changes.
3452 - forward compatibility changes to wire structs (6007)
3453 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3454 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3455 - improved "lustre status" script
3456 - initialize blocksize for non-regular files (6062)
3457 - added --disable-server and --disable-client configure options (5782)
3458 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3459 - Vanilla 2.4.29 support
3460 - increase maximum number of obd devices to 520 (6242)
3461 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3462 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3463 SLES 9 2.6.5 kernels
3464 - limit stripes per file to 160 (the maximum EA size) (6093)
3466 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3469 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3470 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3471 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3472 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3473 - small ext3 extents cleanups and fixes (5733)
3474 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3475 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3476 - avoid needless client->OST connect, fix handle mismatch (5317)
3477 - fix DLM error path that led to out-of-sync client, long delays (5779)
3478 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3479 - fix several locking issues related to i_size (5492,5624,5654,5672)
3480 - don't move pending lock onto export if it is already evicted (5683)
3481 - fix kernel oops when creating .foo in unlinked directory (5548)
3482 - fix deadlock in obdfilter statistics vs. object create (5811)
3483 - use time_{before,after} to avoid timer jiffies wrap (5882)
3484 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3485 - Do not leak request if server was not able to process it (5154)
3486 - If mds_open unable to find parent dir, make that negative lookup(5154)
3487 - don't create new directories with extent-mapping (5909, 5936)
3489 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3490 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3491 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3492 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3493 - support for RHEL4 kernel on i686 (5773)
3494 - provide error messages when incompatible logs are encountered (5898)
3496 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3497 * version 1.4.0.10 (1.4.1 release candidate 1)
3499 - don't keep a lock reference when lock is not granted (4238)
3500 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3501 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3502 - drop import inflight refcount on signal_completed_replay error (5255)
3503 - unlock page after async write error during send (3677)
3504 - handle missing objects in filter_preprw_read properly (5265)
3505 - no transno return for symlink open, don't save no-trasno open (3440)
3506 - don't try to complete elan receive that already failed (4012)
3507 - free RPC server reply state on error (5406)
3508 - clean up thread from ptlrpc_start_thread() on error (5160)
3509 - readahead could read extra page into cache that wasn't ejected (5388)
3510 - prevent races in class_attach/setup/cleanup/detach (5260)
3511 - don't dereference de->d_inode after l_dput of de (5458)
3512 - use "int" for stripe value returned from lock_to_stripe (5544)
3513 - mballoc allocation and error-checking fixes in 2.6 (5504)
3514 - block device patches to fix I/O request sizes in 2.6 (5482)
3515 - look up hostnames for IB nals (5602)
3516 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3517 - don't start multiple acceptors for the same port (5277)
3518 - fix incorrect LASSERT in mds_getattr_name (5635)
3519 - export a proc file for general "ping" checking (5628)
3520 - fix "lfs check" to not block when the MDS is down (5628)
3522 - service request history (4965)
3523 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3524 - create an "evict_client" /proc entry on OSTs, like the MDS has
3525 - fix mount usage message, return errors per mount(8) (5168)
3526 - change grep [] to grep "[]" in tests so they work in more UMLs
3527 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3528 - remove ancient LOV_MAGIC_V0 EA support (5047)
3529 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3530 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3531 - allow readahead allocations to fail when low on memory (5383)
3532 - mmap locking landed again, after considerable improvement (2828)
3533 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3535 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3538 - send OST transaction number in read/write reply to free req (4966)
3539 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3540 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3541 - the watchdog thread now runs as interruptible (5246)
3542 - flock/lockf fixes (but it's still disabled, pending 5135)
3543 - don't use EXT3 constants in llite code (5094)
3544 - memory shortage at startup could cause assertion (5176)
3546 - reorganization of lov code
3547 - single portals codebase
3549 - add extents/mballoc support (5025)
3550 - direct I/O reads in the obdfilter (4048)
3551 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3553 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3556 - send OST transaction number in read/write reply to free req (4966)
3557 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3558 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3559 - the watchdog thread now runs as interruptible (5246)
3560 - handle missing objects in filter_preprw_read properly (5265)
3561 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3562 - drop import inflight refcount on signal_completed_replay error (5255)
3563 - unlock page after async write error during send (3677)
3564 - return original error code on reconstructed replies (3761)
3565 - no transno return for symlink open, don't save no-trasno open (3440)
3567 - add pid to ldlm debugging output (4922)
3568 - bump the watchdog timeouts -- we can't handle 30sec yet
3569 - extra debugging for orphan dentry/inode bug (5259)
3571 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3574 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3575 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3576 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3577 - fix NULL dereference in /proc/sys/portals/routes (4827)
3578 - allow failed mdc_close() operations to be interrupted (4561)
3579 - stop precreate on OST before MDS would time out on it (4778)
3580 - don't send partial-page writes before EOF from client (4410)
3581 - discard client grant for sub-page writes on large-page clients (4520)
3582 - don't free dentries not owned by NFS code, check generation (4806)
3583 - fix lsm leak if mds_create_objects() fails (4801)
3584 - limit debug_daemon file size, always print CERROR messages (4789)
3585 - use transno after validating reply (3892)
3586 - process timed out requests if import state changes (3754)
3587 - update mtime on OST during writes, return in glimpse (4829)
3588 - add mkfsoptions to LDAP (4679)
3589 - use ->max_readahead method instead of zapping global ra (5039)
3590 - don't interrupt __l_wait_event() during strace
3592 - add software watchdogs to catch hung threads quickly (4941)
3593 - make lustrefs init script start after nfs is mounted
3594 - fix CWARN/ERROR duplication (4930)
3595 - return async write errors to application if possible (2248)
3596 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3597 - print NAL number in %x format (4645)
3598 - update barely-supported suse-2.4.21-171 series (4842)
3599 - support for sles 9 %post scripts
3600 - support for building 2.6 kernel-source packages
3601 - support for sles km_* packages
3603 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3606 - ignore -ENOENT errors in osc_destroy (3639)
3607 - notify osc create thread that OSC is being cleaned up (4600)
3608 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3609 - reconstruct ost_handle() like mds_handle() (4657)
3610 - create a new thread to do import eviction to avoid deadlock (3969)
3611 - let lconf resolve symlinked-to devices (4629)
3612 - don't unlink "objects" from directory with default EA (4554)
3613 - hold socknal file ref over connect in case target is down (4394)
3614 - allow more than 32000 subdirectories in a single directory (3244)
3615 - fix blocks count for O_DIRECT writes (3751)
3616 - OST returns ENOSPC from object create when no space left (4539)
3617 - don't send truncate RPC if file size isn't changing (4410)
3618 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3619 - bind to privileged port in socknal and tcpnal (3689)
3621 - rate limit CERROR/CWARN console message to avoid overload (4519)
3622 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3623 - basic mmap support (3918)
3624 - kernel patch series update from b1_4 (4711)
3626 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3629 - avoid crash during MDS cleanup with OST shut down (2775)
3630 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3631 - don't use bad inodes on the MDS (3744)
3632 - dynamic object preallocation to improve recovery speed (4236)
3633 - don't hold spinlock over lock dumping or change debug flags (4401)
3634 - don't zero obd_dev when it is force cleaned (3651)
3635 - print grants to console if they go negative (4431)
3636 - "lctl deactivate" will stop automatic recovery attempts (3406)
3637 - look for existing locks in ldlm_handle_enqueue() (3764)
3638 - don't resolve lock handle twice in recovery avoiding race (4401)
3639 - revalidate should check working dir is a directory (4134)
3641 - don't always mark "slow" obdfilter messages as errors (4418)
3643 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3646 - don't close LustreDB during write_conf until it is done (3860)
3647 - fix typo in lconf for_each_profile (3821)
3648 - allow dumping logs from multiple threads at one time (3820)
3649 - don't allow multiple threads in OSC recovery (3812)
3650 - fix debug_size parameters (3864)
3651 - fix mds_postrecov to initialize import for llog ctxt (3121)
3652 - replace config semaphore with spinlock (3306)
3653 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3654 - don't allow enqueue to complete on a destroyed export (3822)
3655 - down write_lock before checking llog header bitmap (3825)
3656 - recover from lock replay timeout (3764)
3657 - up llog sem before sending rpc (3652)
3658 - reduce ns lock hold times when setting kms (3267)
3659 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3660 - fix NULL deref and obd_dev leak on setup error (3312)
3661 - replace some LBUG about llog ops with error handling (3841)
3662 - don't match INVALID dentries from d_lookup and spin (3784)
3663 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3664 - fix invalid assertion in ptlrpc_set_wait (3880)
3666 - add libwrap support for the TCP acceptor (3996)
3667 - add /proc/sys/portals/routes for non-root route listing (3994)
3668 - allow setting MDS UUID in .xml (2580)
3669 - print the stack of a process that LBUGs (4228)
3671 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3674 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3675 - make sure to unset replay flag from failed open requests (3440)
3676 - if default stripe count is 0, use OST count for inode size (3636)
3677 - update parent mtime/ctime on client for create/unlink (2611)
3678 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3679 - free recovery state on server during a forced cleanup (3571)
3680 - unregister_reply for resent reqs (3063)
3681 - loop back devices mounting and status check on 2.6 (3563)
3682 - fix resource-creation race that can provoke i_size == 0 (3513)
3683 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3684 - more debugging for page-accounting assertion (3746)
3685 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3686 - avoid deadlock after precreation failure (3758)
3687 - fix race and lock order deadlock in orphan handling (3450, 3750)
3688 - add validity checks when grabbing inodes from l_ast_data (3599)
3690 - add /proc/.../recovery_status to obdfilter (3428)
3691 - lightweight CDEBUG infrastructure, debug daemon (3668)
3692 - change default OSC RPC parameters to be better on small clusters
3693 - turn off OST read cache for files smaller than 32MB
3694 - install man pages and include them in rpms (3100)
3695 - add new init script for (un)mounting lustre filesystems (2593)
3696 - run chkconfig in %post for init scripts (3701)
3697 - drop scimac NAL (unmaintained)
3699 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3702 - clean kiobufs before and after use (3485)
3703 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3704 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3705 - change init script to fail more clearly if not run as root (1528)
3706 - allow clients to reconnect during replay (1742)
3707 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3708 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3709 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3710 - fix cond_resched (3554)
3711 - only evict unfinished clients after recovery (3515)
3712 - allow bulk resend, prevent data loss (3570)
3713 - dynamic ptlrpc request buffer allocation (2102)
3714 - don't allow unlinking open directory if it isn't empty (2904)
3715 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3716 - remove extraneous obd dereference causing LASSERT failure (3334)
3717 - don't use get_cycles() when creating temp. files on the mds (3156)
3718 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3719 - handle EEXIST for set-stripe, set proper directory name (3336)
3721 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3722 - fix ksocknal_fmb_callback() error messages (2918)
3724 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3727 - don't copy lvb into (possibly NULL) reply on error (2983)
3728 - don't deref dentry after dput, don't free lvb on error (2922)
3729 - use the kms to determine writeback rpc length (2947)
3730 - increment oti_logcookies when osc is inactive (2948)
3731 - update client's i_blocks count via lvb messages (2543)
3732 - handle intent open/close of special files properly (1557)
3733 - mount MDS with errors=remount-ro, like obdfilter (2009)
3734 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3735 - don't use cancelling-locks' kms values (2947)
3736 - use highest lock extent for kms, not last one (2925)
3737 - don't dereference ERR_PTR() dentry in error handling path (3107)
3738 - fix thread race in portals_debug_dumplog() (3122)
3739 - create lprocfs device entries at setup instead of at attach (1519)
3740 - common AST error handler, don't evict client on completion race (3145)
3741 - zero nameidata in detach_mnt in 2.6 (3118)
3742 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3743 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3744 - initialize RPC timeout timer earlier for 2.6 (3219)
3745 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3746 - print nal/nid for unknown nid (3258)
3747 - additional checks for oscc recovery before doing precreate (3284)
3748 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3749 - don't crash in mdc_close for bad permissions on open (3285)
3750 - zero i_rdev for non-device files (3147)
3751 - clear page->private before handing to FS, better assertion (3119)
3752 - tune the read pipeline (3236)
3753 - fix incorrect decref of invalidated dentry (2350)
3754 - provide read-ahead stats and refine rpc in flight stats (3328)
3755 - don't hold journal transaction open across create RPC (3313)
3756 - update atime on MDS at close time (3265)
3757 - close LDAP connection when recovering to avoid server load (3315)
3758 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3759 - don't leak open file on MDS after open resend (3325)
3760 - serialize filter_precreate and filter_destroy_precreated (3329)
3761 - loop device shouldn't call sync_dev() for nul device (3092)
3762 - clear page cache after eviction (2766)
3763 - resynchronize MDS->OST in background (2824)
3764 - refuse to mount the same filesystem twice on same mountpoint (3394)
3765 - allow llmount to create routes for mounting behind routers (3320)
3766 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3767 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3768 - fix rare problem with rename on htree directories (3417)
3770 - allow default OST striping configuration per directory (1414)
3771 - fix compilation for qswnal for 2.6 kernels (3125)
3772 - increase maximum number of MDS request buffers for large systems
3773 - change liblustreapi to be useful for external progs like lfsck (3098)
3774 - increase local configuration timeout for slow disks (3353)
3775 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3777 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3780 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3781 - don't overwrite extent policy data in reply if lock was blocked (2901)
3782 - drop filter export grants atomically with removal from device (2663)
3783 - del obd_self_export from work_list in class_disconnect_exports (2908)
3784 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3785 - swab reply message in mdc_close, other PPC fixes (2464)
3786 - fix destroying of named logs (2325)
3787 - overwrite old logs when running lconf --write_conf (2264)
3788 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3789 - fix race in target_handle_connect (2898)
3790 - mds_reint_create() should take same inode create lock (2926)
3791 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3792 - don't close files for self_export to avoid uninitialized obd (2936)
3793 - allow MDS with the same name as client node (2939)
3794 - hold dentry reference for closed log files for unlink (2325)
3795 - reserve space for all logs during transactions (2059)
3796 - don't evict page beyond end of stripe extent (2925)
3797 - don't oops on a deleted current working directory (2399)
3798 - handle hard links to targets without a parent properly (2517)
3799 - don't dereference NULL lock when racing during eviction (2867)
3800 - don't grow lock extents when lots of conflicting locks (2919)
3802 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3805 - account for cache space usage on clients to avoid data loss (974)
3806 - lfsck support in lustre kernel code (2349)
3807 - reduce journal credits needed for BRW writes (2370)
3808 - orphan handling to avoid losing space on client/server crashes
3809 - ptlrpcd can be blocked, stopping ALL progress (2477)
3810 - use lock value blocks to assist in proper KMS, faster stat (1021)
3811 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3812 - recovery for initial connections (2355)
3813 - fixes for mds_cleanup_orphans (1934)
3814 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3815 - block all file creations until orphan recovery completes (1901)
3816 - client remove rq_connection from request struct (2423)
3817 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3818 - recovery timer race (2670)
3819 - mdc_close recovey bug (2532)
3820 - ptlrpc cleanup bug (2710)
3821 - mds timeout on local locks (2588)
3822 - namespace lock held during RPCs (2431)
3823 - handle interrupted sync write properly (2503)
3824 - don't try to handle a message that hasn't been replied to (2699)
3825 - client assert failure during cleanup after abort recovery (2701)
3826 - leak mdc device after failed mount (2712)
3827 - ptlrpc_check_set allows timedout requests to complete (2714)
3828 - wait for inflight reqs when ptlrpcd finishes (2710)
3829 - make sure unregistered services are removed from the srv_list
3830 - reset bulk XID's when resending them (caught by 1138 test)
3831 - unregister_bulk after timeout
3832 - fix lconf error (2694)
3833 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3834 - readahead locks pages, leaves pending causing memory pressure (2673)
3835 - increase OST request buffers to 4096 on large machines (2729)
3836 - fix up permission of existing directories in simple_mkdir (2661)
3837 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3838 - don't assign transno to errored transactions (2742)
3839 - don't delete objects on OST if given a bogus objid from MDS (2751)
3840 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3841 - if rq_no_resend, then timeout request after recovery (2432)
3842 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3843 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3844 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3845 - MDS recovery shouldn't delete objects that it has given out (2730)
3846 - if enqueue arrives after completion, don't clobber LVB (2819)
3847 - don't unlock pages twice when trigger_group_io returns error (2814)
3848 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3849 - don't write pages to disk if there was an error (1450)
3850 - don't ping imports that have recovery disabled (2676)
3851 - take buffered bytes into account when balancing socknal conn (2817)
3852 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3853 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3854 - remove little-endian swabbing of llog records (1987)
3855 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3856 - retry reposting request buffers if they fail (1191)
3857 - grow extent at grant time to avoid granting a revoked lock (2809)
3858 - lock revoke doesn't evict page if covered by a second lock (2765)
3859 - disable VM readahead to avoid reading outside lock extents (2805)
3861 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3862 - updated kernel patches for hp-2.4.20 kernel (2681)
3864 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3867 - fix truncated write corruption (2366)
3868 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3870 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3871 - ensure len(uuid) < 37 in lmc (1171)
3872 - fix ia64 OOPS in llog_test (2255)
3873 - zero end of page at obdfilter for partial page writes (2648)
3874 - don't leave stale dentries around after renames (bug 2428)
3875 - fix timeouts when evicting a client with a single lock held (2642)
3876 - set deadline for the initial HELLO message to drain (2634)
3877 - print out dotted-quad IP addresses in the socknal (2302)
3879 - additional debugging for MDS client eviction problem (2443)
3880 - fix mkfsoptions support for osts (2603, 2604)
3882 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3885 - add series for the vanilla 2.6.0 kernel
3886 - add series for the vanilla 2.4.24 kernel
3887 - add series for a cray x86/64 UL kernel drop
3888 - fix xattr patches for the vanilla 2.4.19 series
3890 - generate true UUIDs in lmc (1171)
3891 - have portals stack dumping break in UML (2466)
3892 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3893 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3894 - wake callback waiting threads on client eviction (2460)
3895 - Add --ptldebug and --subsystem to lmc (1719)
3896 - update assertion to allow safe interrupt allocation
3897 - set rq_no_resend for cancel requests (2432)
3898 - recalculate ptlrpcd timeout after resend (2494)
3899 - call vfs_rmdir when removing pending directories (2368)
3900 - fix renaming a file to itself (2429)
3901 - lmc creates a default one-stripe lov (2454)
3902 - expand procfs space to handle large clusters (2326)
3903 - increase UML stack to avoid overflow
3904 - update lconf's list of debug and subsystem masks
3905 - fix lfs find --obd (2510)
3906 - /proc tunable for disabling filter read caching (2591)
3907 - stop rpm packages from altering slapd.conf (2301)
3908 - disable nagle in the socknal under 0conf (2578)
3909 - choose mds inode size based on stripe count (2572)
3910 - fix kernel-source rpm problems (2516)
3912 - add --disable-doc to avoid pdf generation (2421)
3913 - update documentation, tests, type-os, comments
3914 - avoid format warnings on ia64
3915 - remove the TOE NAL
3916 - tiny code cleanups by removing unused fields
3918 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3921 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3922 - remove the most blatant lies from BUILDING file (2371)
3923 - change default debug level to reasonable production setting
3924 - reduce client side cache size to reduce cache flush time
3925 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3926 - make TCP zerocopy and pinger support enabled by default (2476)
3927 - sync writes completed after process exits caused crashes (2319)
3928 - maintain correct mount count on the MDS (2356)
3929 - backout 1557, because 2316 wasn't really fixed
3930 - better file I/O statistics gathering in /proc
3931 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3932 - another kernel patch to fix zero-copy TCP function export
3933 - don't take duplicate lock when processing re-sent getattr (2420)
3934 - lctl uses obd_self_export instead of creating new conn (2353)
3935 - MDS/OST recovery case which requires object creation asserted (2425)
3936 - move lfs from /usr/sbin to /usr/bin in packages
3937 - fix race between mds_client_add and mds_client_free (2417)
3938 - use kmalloc instead of slabs in portals (2430)
3939 - don't create duplicate records when a failover MDS is present (2442)
3940 - remove unnecessary mount age check (2332)
3941 - don't remove directory inodes from locks prematurely (2451)
3942 - don't break if MDS service name is the same as hostname (2103)
3943 - fix races in client write RPC generation when cache full (2482)
3945 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3948 - remove now-unused request->rq_obd (278)
3949 - if an allocation fails, print out how much memory we've used (1933)
3950 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3951 - add the "configurable stack size" patch to most series files (1256)
3952 - ability to write large log records, for 100+ OST configs (2306)
3953 - fix NULL deref when filter_prep fails (2314)
3954 - fix operator precedence error in filter_sync
3955 - dynamic allocation of socknal TX descriptors (2315)
3956 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3957 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3958 - ensure that CWARN messages in Portals always get to the syslog
3959 - __init/__exit are not for prototype decls (ldlm_init/exit)
3960 - x86-64 compile warning fixes
3961 - fix gateway LMC keyword conflict (2318)
3962 - fix MDS lock inversions in getattr/reint paths (1844)
3963 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3964 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3965 - DLM race condition prevented some lock evictions (2328)
3966 - ENOMEM detection and retry on socknal sends (2230)
3967 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3968 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3969 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3970 - better error messages when a client is rejected during recovery (1505)
3971 - avoid cancelling locks which were never granted, after failure (2330)
3972 - fix i_sem/journal inversion in mds_client_add (2333)
3973 - fix truncate/getattr lock cycle deadlock (2334)
3974 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3975 - fix two rare exit paths which could leak an l_lock() ref (2321)
3976 - fencepost error in MDS/OST orphan recovery (2226)
3977 - make log record alignment 8 bytes (1988)
3978 - lstripe now fails when requested offset > ost_count (2237)
3979 - ensure that all kernel series have a complete list.h (1607)
3980 - fix crashes in special-file operations (2316)
3981 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3983 - allow configurable automake binary, for testing new versions
3984 - small update to the lfs documentation
3986 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3988 * fix negative export reference count in fsfilt_sync (2312)
3990 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3991 * release candidate 0.9.1
3993 - orphans are moved into the PENDING directory for possible recovery
3994 - replayed opens now open by fid for orphan/rename safety (1042)
3995 - last close of an orphan inode generates a transno (683)
3996 - chdir() and mount() now pin the directory entry (1020)
3997 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3998 - discard very old requests without processing them (1502)
3999 - remove some common, well-understood CERRORs (1505)
4000 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4001 - clear "grant" flags in OST replies until OST grant code lands (1644)
4002 - fix read performance by not clobbering i_blksize on client (1598)
4003 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4004 - make LRU size a /proc tunable, clears locks when reduced (707)
4005 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4006 - server locks take references on exports now (1558)
4007 - build fixes for 2.4.20-rh trees (1663)
4008 - return an error from lov_create if all OSCs are inactive (1751)
4009 - fix import levels when a reconnect happens without a timeout (1597)
4010 - exit early from mds_open if we get a lookup error (1749)
4011 - partial page read at EOF wouldn't wait for disk before sending (1642)
4012 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4013 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4014 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4015 - write/truncate lock inversion (1639)
4016 - Don't auto-load obdclass, portals modules during cleanup (1495)
4017 - fix timestamps from jumping to "now" (1763)
4018 - extra journal assertions (1648)
4019 - add an extra multiunlink test (1771)
4020 - fix read_record/write_record API (1776)
4021 - fix leak of offset_extent, possible incorrect i_size later (1772)
4022 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4023 - Debugging for the kqswnal_get_idle_tx problems (1820)
4024 - Allow recovery to be attempted multiple times (1536)
4025 - Write out MDS last_rcvd file after it is first created (1600)
4026 - Fix tx_descriptor leak in failed transmit situations (1827)
4027 - ext3 journaling fixes for assertion failure after IO error (1871)
4028 - class_export_put() on freed export after completion AST error (1896)
4029 - Fix revalidate looping in VFS (1322)
4030 - Don't access a freed export during MDS_REINT timeout (1521)
4031 - Add open-unlink recovery support on the MDS (1673,1764)
4032 - Return an error if no MDS data was read from last_rcvd (1946)
4033 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4034 - Don't setup a disk device that doesn't match exported UUID (317)
4035 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4036 - avoid committing NULL handle in force close
4037 - local.sh is now a one-stripe LOV configuration
4038 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4039 - fix ext3 htree duplicate directory entry corruption (1516)
4040 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4041 - update to new LOV EA format (2097)
4042 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4044 2003-06-15 Phil Schwan <phil@clusterfs.com>
4047 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4048 - per-import/export recovery handling (958, 931, 959)
4049 - multiple last-rcvd slots, for serving multiple FSes (949)
4050 - connections are again shared between multiple imp/exports (963, 964)
4051 - "umount -f" would hang if any requests needed to be sent (393, 978)
4052 - avoid pinning large req buffer by copying for queued messages (989)
4053 - add "uuid" to "lctl device" command to help upcalls (991)
4054 - "open" RPCs with transnos would confuse recovery counters (1037)
4055 - do proper endian conversion of all wire messages (288, 340, 891)
4056 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4057 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4058 - fix ptlrpc_connection leak in target_handle_connect (1174)
4059 - fix import refcounting bug in OST and MDS cleanup (1134)
4060 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4061 - fix very unlikely obd_types race condition (501)
4062 - remove osc_open hack for echo_client (1187)
4063 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4064 - a failure in read_inode2 leads to deadlock (1139)
4065 - cancel ack-locks as soon as transaction is committed (1072)
4066 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4067 - make sure to commitrw after any preprw to avoid deadlock (1162)
4068 - failing to execute a file in a lustre FS would lock inode (1203)
4069 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4070 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4071 - obd_uuid2tgt would walk off the end of the list (1255)
4072 - on IA64 the osc would give portals incorrect bulk size (1258)
4073 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4074 - fix lock inversion caused by new llite matching code (1282)
4075 - limit the number of dirty pages on a client to 10MB (1286)
4076 - timed out locks were not being corrected cancelled (1289)
4077 - fix O_DIRECT above 4GB on IA-32 (1292)
4078 * major user-visible changes
4079 - fail out/fail over policy now controlled by the upcall (993)
4081 - add OBD_PING to check server availability and failure (954)
4082 - lustre messages are now sent in sending host order (288, 340, 891)
4083 - add eadatalen to MDS getattr reply (340)
4084 - OST read replies may contain second buffer, with per-page status (593)
4086 2003-03-11 Phil Schwan <phil@clusterfs.com>
4089 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4090 - failed open()s could cause deadlock; fixed (867, 869)
4091 - stop cancelling OST locks when files are closed (481)
4092 - overlapping XID spaces caused network corruption (851, 853)
4093 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4094 - setattr_raw added, to do single-RPC, server-side setattrs
4095 - lmc/lconf syntax change for OST UUIDs
4096 - fix crashy race condition between ptlrpc_free_req and osc_close
4097 - don't use request in mdc_enqueue if we hit a timeout (889)
4098 - don't set the inode i_size for regular files from the MDS (896)
4099 - handle out of order completion AST (842)
4100 - don't LBUG if a lock request times out after receiving AST (913)
4101 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4102 - if a bad lock AST arrives, send an error instead of dropping entirely
4103 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4104 - fix leak in bulk IO when only partially completed (899, 900, 926)
4105 - fix O_DIRECT for ia64 (55)
4106 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4107 - C-z after timeout could hang a process forever; fixed (977)
4109 - client-side I/O cache (678, 924, 929, 941, 970)
4111 - READPAGE and SETATTRs which don't take server-side locks get
4114 2003-02-11 Phil Schwan <phil@clusterfs.com>
4117 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4118 - Fix fsfilt_extN_readpage() to read a full page of directory
4119 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4120 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4121 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4122 - Most intent operations were removed, in favour of a new RPC mode
4123 that does a single RPC to the server and bypasses most of the VFS
4124 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4125 - Aggressively cancel local locks on DLM servers
4126 - mds_reint_unlink sends EA to the client if it's the last nlink.
4127 client uses that EA to unlink OST objects.
4128 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4129 - recursive symlinks were fixed (439)
4130 - fixed NULL deref in DEBUG_REQ
4131 - filter_update_lastobjid no longer calls sync, which annoyed extN
4132 - fixed multi-client small-writes to a single file problem (445)
4133 - fixed mtime updates during file writes (607)
4134 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4135 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4136 - fixed Linux OST size reporting problem (444, 656)
4137 - OST now updates object mtime with writes or setattr (607, 619)
4138 - client verifies file size before zeroing page past EOF (445)
4139 - OST now writes last allocated objid to disk with allocation (108)
4140 - LOV on echo now works (409)
4142 - mds_reint_unlink sends a new buffer, with the EA included. this
4143 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4144 set if a regular file was being unlinked, and it was the last link
4145 - use PtlGet from the target for bulk writes (315)
4146 - OST now updates object mtime with writes or setattr (607, 619)
4147 - LDLM now has a grant-time callback to revalidate locked items, if
4149 - Many MDS operations were reorganized to combat race conditions
4151 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4152 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4154 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4157 - Fully reactivate OST imports after reconnection (512, others)
4158 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4159 - More graceful error handling for truncating on dead OST (515)
4160 - Don't error out unless we're actually accessing dead stripes (474)
4161 - Fix garbage sizes when stripes are missing (410)
4162 - LRU counters were broken, causing constant lock purge (433, 432)
4163 - garbage on read from stripes with failed OSTs (441)
4164 - mark OSCs as active before reconnecting during recovery (438)
4165 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4166 - lfind did not preserve OST order in output (443)
4167 - symlinks cause hung clients, incorrect data (439)
4168 - stop dereferencing request after dropping refcount (457)
4169 - don't LASSERT(spin_is_locked) on non-SMP (455)
4170 - fixes for many rename() bugs
4171 - fstat didn't correctly synchronize attributes (399)
4172 - server must handle lock cancellation during blocking AST prep (487)
4173 - bulk descriptors were free()d too soon (511)
4174 - fix paths in lconf, which would load incorrect modules (451, 507)
4175 - fix confusing lconf 'host not found' error message (386)
4176 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4177 - fix race condition in mdc_blocking_ast() for inode access (526)
4178 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4179 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4180 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4181 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4182 - fix lstripe in several ways (kernel side) (527)
4183 - fix request leak in ldlm_cli_enqueue (262)
4184 - incorrect OSC was marked inactive after OST failure
4185 - call mds_fs_cleanup before unmounting filesystem (524)
4186 - fix races between taking ns_lock and ldlm_lock_change_resource
4187 - fix races updating LOV export open file list
4188 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4189 - fix recovery NULL deref in ldlm_cli_cancel_unused
4190 - fix some DLM races by using new hash table for lock handles (419)
4191 - permit the client to specify desired inodes, at replay
4192 - duplicate requests when we queue them for replay reintegration
4193 - fix last_rcvd offset calculation
4194 - sync after each recovered transaction, so we always make progress
4195 - never, not always, ERESTART requests without transnos
4196 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4198 - skip replay if the MDS says that the client is already connected
4199 - don't check for a recovery-enabled export to match lctl's UUID
4200 - don't INC_USE_COUNT for phantom exports
4201 - don't crash when cleaning up phantom exports (567)
4202 - don't double-finish or set replay data for errored mdc_open requests
4203 - abort requests when they time out, so we don't get old replies
4204 - send/receive replies for AST messages again
4205 - if the client says that it doesn't have the lock, cancel it on the
4207 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4208 mark it as destroyed, it will all work out when decref is called
4209 - fix module use counts (22, 581)
4211 - ASTs now expect a reply (server cancels lock on error reply)
4213 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4216 - fix many simultaneous client startup (392)
4217 - fix dentry->d_it clobbering
4218 - credentials weren't being shipped for readdir/getattr operations
4219 - remove invalid assertions triggered during some concurrent MD
4221 - proper Lustre versions added (336, 389)
4222 - fix memory leak for create error case (398)
4223 - fix LOV locking bug that would get cli/srv out of sync
4224 - fix echo client over LOV (409)
4225 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4226 - fix double-O_EXCL intent crash (424)
4227 - avoid sending multiple lock CANCELs (352)
4229 - MDS can do multi-client recovery (modulo bugs in new code)
4231 - many updates, edits, cleanups
4233 2002-11-18 Phil Schwan <phil@clusterfs.com>
4236 - fix null d_it dereference (346)
4237 - fix full OST/dbench hang (333)
4238 - fix permission problem with file removal (286)
4239 - fix removal of OSCs from LOV when they fail
4240 - fix NULL deref during bulk timeout (214)
4241 - fix problems related to multiple filesystems on one MDS (241)
4242 - fixed serious subtle metadata locking bugs
4243 - free locks on clients when inodes are removed due to memory
4245 - fix inode pointer in lock data (285)
4246 - partial support for multiple MDS on a single host (241)
4247 - data locks weren't cancelled at clear_inode time (290, 311)
4248 - intent locks could lead to unbounded lock growth (205)
4249 - added a maximum lock count, an LRU list, and a flusher
4250 - fix multiple rename (365)
4251 - properly abstracted the echo client
4252 - OSC locked 1 byte too many; fixed
4253 - rewrote brw callback code:
4254 - fixed recovery bugs related to LOVs (306)
4255 - fixed too-many-pages-in-one-write crash (191)
4256 - fixed (again) crash in sync_io_timeout (214)
4257 - probably fixed callback-related race (385)
4259 - Add capability to MDS protocol
4260 - LDLM cancellations and callbacks on different portals
4262 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4265 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4266 - defer open object destruction to close time (601981, 138)
4267 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4268 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4269 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4270 - don't call obd_setattr in ll_file_release for destroyed objects
4271 * protocol change to lustre_msg: move |version| and add |flags|
4272 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4273 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4274 * added replay of create, unlink, link and rename operations during
4275 MDS failover; recovery should be much more robust now
4276 * remove failed OSCs from LOVs (only lov_create uses this so far)
4277 * the lustre-HOWTO was brought (more) up to date (582544)
4279 2002-10-23 Phil Schwan <phil@clusterfs.com>
4282 - in-use dentries weren't being reused properly (617851)
4283 - prevent multiple LDLM setup (599178)
4284 - fix LOV size calculations for truncate (617853)
4285 - fix client handling of MDS intent errors (POSIX)
4286 - fix permission bug in lovstripe.c test (624321)
4287 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4288 - truncate past end of file could corrupt data
4289 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4290 - a race in recovery could return ETIMEDOUT to apps (623947)
4291 - building outside the source directory was fixed
4292 * the lustre-HOWTO was brought (more) up to date (582544)
4293 * major progress was made on recovery functionality
4295 2002-10-10 Phil Schwan <phil@clusterfs.com>
4298 - recovery deadlock fix
4299 - rm -rf causes LBUG fix (617817)
4300 - file open by multiple tasks fix (618962)
4301 - directory permissions bugs (602707 and 620007)
4302 - journal_stop fixed with locking (611313)
4303 - O_APPEND failures resolved (618273, perhaps 614459)
4304 - lconf PATH fix (619770)
4305 - IA64 build fix (621450)
4306 - RPC buffer sizes scale with amount of memory
4308 2002-10-01 Phil Schwan <phil@clusterfs.com>
4311 - locks would be cancelled without throwing away data pages,
4312 resulting in inconsistent data (605627)
4313 - inode attributes were not always being refreshed (605627, 612449)
4314 - lconf now continues to cleanup after lctl reports an error
4315 - MDS now enforces user permissions (602707)
4316 - lprocfs cleanup fixed, but not yet enabled (614157)
4317 - fixed infinite server hang, should a client not respond to an AST
4318 - avoid going into recovery if user calls readlink() with a buffer
4319 that's too small (613941)
4320 - AST RPCs no longer require replies (614867) -- this may be changed
4321 - don't crash server if client sends an IOV that's too big (611336)
4322 - fixed lock conversion deadlock (611892)
4323 - fixed the following of symlinks (614622)
4324 * recovery: the server can remove locks from a client that dies, other
4325 clients can make progress
4326 * more extN patch fixes
4327 * compile-time configurable ptlrpc buffer allocations
4329 - collaborative read cache document
4330 - Lustre Lite Performance CDR document-in-progress
4332 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4335 - fix typo in patch-2.4.18
4337 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4340 - clear ptlrpc request each time in handle_incoming_request()
4341 - unlink of files now destroys the object on the OST
4343 2002-09-19 Peter Braam <braam@clusterfs.com>
4345 * add hard link support
4346 * change obdfile creation method
4347 * kernel patch changed
4349 2002-09-19 Peter Braam <braam@clusterfs.com>
4352 - stack overflow bug in extN fixed
4354 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4356 * documentation updates
4357 - add man pages for config tools
4358 - update tests/README to describe testing with new config tools
4359 - finish metadata API descriptions
4360 * bug fixes and cleanups
4361 - statfs workaround for 16TB limit
4362 - LOV stripe allocation improved, can stripe on subset of OSTs
4363 - LOV file size/IO offset was wrong for files > 4GB in size
4364 - object EA data was being dropped, caused files to be unreadable
4365 - memory overflow with non-LOV OST caused memory corruption
4366 - fixed regression tests to work with new config tools, obdfilter
4367 - fixed bug when directory size became larger than 1 block
4368 - fixed bug (for single client case) when PWD was deleted
4369 - invalidate local directory pages when doing intent-based ops
4370 - avoid LDLM oops when lock callback contained bad data
4372 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4374 * documentation updates
4375 * bug fixes and cleanups
4376 - configuration tools
4379 - 64-bit compile warnings
4380 - 64-bit internal statfs data
4382 * test_brw on persistent OST devices
4384 * lprocfs (disabled)
4386 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4388 * documentation updates
4389 * bug fixes and cleanups
4390 * configuration tools
4392 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4395 * many small fixes to 0_5_4
4396 * io/network handling
4397 * thinkos in MDS operations
4399 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4402 * crucial basic fixes to 0.5.3
4403 * IOR, Iozone work over Elan
4406 2002-08-07 Phil Schwan <phil@clusterfs.com>
4407 * version 0_5_3, our first alpha
4408 * we use the new Portals iovs
4409 * documentation updates
4410 * bug fixes and cleanups
4411 * small changes in the DLM wire protocol
4413 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4414 * version 0_5_1 with some initial stability,
4415 * locking on MD and file I/O.
4416 * documentation updates
4417 * several bug fixes since 0.5.0
4418 * small changes in wire protocol
4420 2002-07-18 Phil Schwan <phil@clusterfs.com>
4422 * delivered as Lustre Light Alpha
4423 * fixed a crash after handling invalid MDS requests
4424 * fixed directory pages for architectures with non-4k pages sizes
4426 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4427 * release version v0_4_4
4428 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4430 * Fixes a number of interruption problems with OST operations.
4431 * Update documentation for portals header changes
4432 * Move all wire protocol structs/defines to lustre_idl.h
4433 * Fixes symlink length bug.
4434 * Add tcpdump to repository.
4436 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4437 * release version v0_4_3
4438 * Fixes statfs for inodes on extN.
4439 * Fixes bug in runtests which would delete /etc/hosts.
4440 * Use 64-bit object IDs wherever possible (not into VFS though)
4441 Remove ost_get_info, which is unused by lustre, and out of date.
4443 2002-07-03 Peter Braam <braam@clusterfs.com>
4444 * release version v0_4_2 Fixes a lookup error (type not passed)
4445 * move forward to head of Portals
4446 * move forward to latest Lustre kernel
4448 2002-06-25 Peter Braam <braam@clusterfs.com>
4449 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob