1 tbd Sun Microsystems, Inc.
4 2.6.16.54-0.2.5 (SLES 10),
5 2.6.18-53.1.14.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
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: 1.40.7-sun1
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).
17 Description: Don't update lov_desc members until making sure they are valid
18 Details : When updating lov_desc members via proc fs, need fix their
19 validities before doing the real update.
23 Description: don't put request into delay list while invalidate in flight.
24 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
25 import in flight. this produce timeout for invalidate and sometimes
30 Frequency : on ppc only
31 Description: not convert ost objects for directory because it's not exist.
32 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
36 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
38 Severity : enhancement
40 Description: Update to support 2.6.22.14 vanilla kernel.
44 Frequency : rare, on recovery
45 Description: read procfs can produce deadlock in some situation
46 Details : Holding lprocfs lock which send rpc can produce block for destroy
47 obd objects and this also block reconnect with -EALREADY. This isn't
48 fix all lprocfs bugs - but make it rare.
50 Severity : enhancement
52 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
55 Frequency : frequent on X2 node
57 Description: mdc_set_open_replay_data LBUG
58 Details : Set replay data for requests that are eligible for replay.
62 Description: lustre_mgs: operation 101 on unconnected MGS
63 Details : When MGC is disconnected from MGS long enough, MGS will evict the
64 MGC, and late on MGC cannot successfully connect to MGS and a lot
65 of the error messages complaining that MGS is not connected.
68 Frequency : on start mds
70 Description: Implement get_info(last_id) in obdfilter.
73 Frequency : occasional
75 Description: Correctly check stale fid, not start epoch if ost not support SOM
76 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
77 overwrited with new generated fid, but mds can anwer with one of these
78 two fids and both is not stale. setattr incorectly start epoch and
79 assume will be called done_writeting, but without SOM done_writing
83 Frequency : rare, depends on device drivers and load
85 Description: MDS or OSS nodes crash due to stack overflow
86 Details : Code changes in 1.8.0 increased the stack usage of some functions.
87 In some cases, in conjunction with device drivers that use a lot
88 of stack the MDS (or possibly OSS) service threads could overflow
89 the stack. One change which was identified to consume additional
90 stack has been reworked to avoid the extra stack usage.
93 Frequency : occasional
95 Description: Do not fail import if osc_interpret_create gets -EAGAIN
96 Details : If osc_interpret_create got -EAGAIN it immediately exits and
97 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
98 oscc_has_objects and see OSC has no objests and call
99 oscc_internal_create to resend create request.
101 Severity : enhancement
103 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
105 Severity : enhancement
107 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
110 Frequency : very rare
112 Description: Fix replay if there is an un-replied request and open
113 Details : In some cases, older replay request will revert the
114 mcd->mcd_last_xid on MDS which is used to record the client's
117 Severity : enhancement
119 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
121 Severity : enhancement
123 Description: Add rhel5 support to HEAD.
125 Serverity : enhancement
127 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
132 Description: Don't allow skipping OSTs if index has been specified.
133 Details : Don't allow skipping OSTs if index has been specified, make locking
134 in internal create lots better.
138 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
139 Details : access to bitfield in structure is always rounded to long
140 and this produce problem with not atomic change any bit.
144 Description: Lustre make rpms failed.
145 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
146 builds Lustre rpms from tarball.
149 Frequency : rare on shutdown ost
151 Description: If llog cancel was not send before clean_exports phase, this can
152 produce deadlock in llog code.
153 Details : If llog thread has last reference to obd and call class_import_put
154 this produce deadlock because llog_cleanup_commit_master wait when
155 last llog_commit_thread exited, but this never success because was
156 called from llog_commit_thread.
160 Description: allow userland application know is lost one of stripes.
161 Details : fill lvb_blocks with error code on ost and return it to
162 application if error flag found.
166 Description: NULL lov_tgts causing MDS oops
167 Details : more safe checks for NULL lov_tgts for avoid oops.
169 Severity : enhancement
171 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
175 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
176 Details : If an OST is down the MDS will hang indefinitely in
177 obd_statfs_rqset() waiting for the statfs data. While for
178 MDS QOS usage of statfs, it should not stuck in waiting.
180 Severity : enhancement
182 Description: remote_acl support
183 Details : Support ACL-based permission check for remote user.
184 Support setfacl/getfacl for remote user with the utils
185 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
186 the system "{s,g}etfacl" utils.
188 Severity : enhancement
190 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
192 Severity : enhancement
194 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
198 Description: make mgs_setparam() handle fsname containing dash
199 Details : fsname containing a dash does not work with lctl conf_param
201 Severity : enhancement
203 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
205 Severity : enhancement
207 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
211 Frequency : rare, at shutdown
212 Description: access already free / zero obd_namespace.
213 Details : if client_disconnect_export was called without force flag set,
214 and exist connect request in flight, this can produce access to
215 NULL pointer (or already free pointer) when connect_interpret
216 store ocd flags in obd_namespace.
220 Frequency : only at startup
221 Description: not alloc memory with spinlock held.
222 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
223 if any spinlock held.
225 Severity : enhancement
227 Description: make lustre randomly fail allocating memory
228 Details : Make lustre randomly failed allocating memory for testing purpose.
230 Severity : enhancement
232 Description: lost problems with lov objid file
233 Details : Fixes some scability and access to not inited memory problems
234 in work with lov objdid file.
239 Description: lfs find does not continue on file error
240 Details : Continue other files processing when a file/dir is absent.
244 Description: Inconsistent usage of lustre_pack_reply()
245 Details : Standardize the usage of lustre_pack_reply() such that it
246 always generate a CERROR on failure.
249 Frequency : occasional
251 Description: MDS hang or stay in waiting lock
252 Details : If client receive lock with CBPENDING flag ldlm need send lock
253 cancel as separate rpc, to avoid situation when cancel request
254 can't processed due all i/o threads stay in wait lock.
258 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
262 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
266 Description: Update to SLES9 kernel-2.6.5-7.287.3.
270 Description: Update to RHEL4 latest kernel.
272 Severity : enhancement
274 Description: Build SLES10 patchless client fails
275 Details : The configure was broken by run ./configure with
276 --with-linux-obj=.... argument for patchless client. When the
277 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
278 can't search header adequately. Use absolute path such as
279 -I($LINUX)/include instead.
283 Description: interrupt oig_wait produce painc on resend.
284 Details : brw_redo_request can be used for resend requests from ptlrpcd and
285 private set, and this produce situation when rq_ptlrpcd_data not
286 copyed to new allocated request and triggered LBUG on assert
287 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
288 ptlrpcd set if request is changed and can be safety changed to use
293 Description: LASSERT_{REQ,REP}SWAB macros are buggy
294 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
295 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
296 these macros and replace them with their difinitions instead.
300 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
301 Details : Update which_patch & target file for SLES10 latest kernel.
304 Description: Remove client patches from SLES 10 kernel.
305 Details : This causes SLES 10 clients to behave as patchless clients
306 even on a Lustre-patched (server) kernel.
308 Severity : enhancement
310 Description: self-adjustable client's lru lists
311 Details : use adaptive algorithm for managing client cached locks lru
312 lists according to current server load, other client's work
313 pattern, memory activities, etc. Both, server and client
314 side namespaces provide number of proc tunables for controlling
317 Severity : enhancement
319 Description: light-weight GSS support
320 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
321 and reduce performance overhead.
323 Severity : enhancement
325 Description: Linux keyring support
326 Details : Support using service of Linux keyring for Lustre GSS internal
327 context refresh/cache mechanism.
331 Description: Fix errors in lfs documentation
332 Details : Fixes man pages
336 Description: don't use GFP_* in generic Lustre code.
337 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
341 Description: obdclass is limited by single OBD_ALLOC(idarray)
342 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
346 Description: warnings with build patchless client with vanila 2.6.19 and up
347 Details : change old ctl_table style and replace ctl_table/ctl_table_header
348 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
352 Frequency : Only for SLES
353 Description: sanity_quota fail test_1
354 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
355 file, which makes TSTID[2] inunique.
359 Frequency : Only for SLES9
360 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
361 Details : Update target/ChangeLog/which_patch .
365 Frequency : Only for test_10 in sanity-quota.sh
366 Description: a bug in quota test code
367 Details : checking a test flag in a wrong place
371 Frequency : Only for test_13 in sanity-quota.sh
372 Description: a bug in quota test script
373 Details : shouldn't check overfull allocated quota
377 Description: jbd statistics
378 Details : Port older jbd statistics patch for sles10
380 Severity : enhancement
382 Description: Add printing inode info into message about error in writepage.
387 Description: reply_lock_interpret crash due to race with it and lock cancel.
388 Details : Do not replay locks that are being cancelled. Do not reference
389 locks by their address during replay, just by their handle.
393 Frequency : When flocks are used.
394 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
395 Details : Flock locks might destroy just granted lock if it could be merged
396 with another existing flock, this is done in completion handler,
397 so teach ldlm_cli_enquque_fini that this is a valid case for
403 Description: Oops in read and write path when failing to allocate lock.
404 Details : Check if lock allocation failed and return error back.
408 Description: lstripe command fails for valid OST index
409 Details : The stripe offset is compared to lov->desc.ld_tgt_count
410 instead of lov->desc.ld_active_tgt_count.
414 Description: sanity.sh failed test 103
415 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
420 Description: df doesn't work properly if diskfs blocksize != 4K
421 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
425 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
426 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
430 Description: obd_health_check_timeout too short
431 Details : set obd_health_check_timeout as 1.5x of obd_timeout
435 Description: llapi_file_create() does not allow some changes
436 Details : add llapi_file_open() that allows specifying the mode and
437 open flags, and also returns an open file handle.
441 Description: merge and cleanup kernel patches.
442 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
446 Description: Add journal checksum support.(Kernel part)
447 Details : The journal checksum feature adds two new flags i.e
448 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
449 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
450 indicates that the commit block contains the checksum for
451 the blocks described by the descriptor blocks. Now commit
452 record can be sent to disk without waiting for descriptor
453 blocks to be written to disk. This behavior is controlled
454 using JBD2_FEATURE_ASYNC_COMMIT flag.
458 Description: OSS needs mutliple precreate threads
459 Details : Add ability to start more than one create thread per OSS.
463 Description: Sanity.sh test_65k failed: setstripe should have succeeded
464 Details : execute lfs setstripe on client
468 Description: mds_obd_create error creating tmp object
469 Details : When the user sets quota on root, llog will be affected and can't
470 create files and write files.
473 Frequency : Always on ia64 patchless client, and possibly others.
475 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
476 Details : This allows the patchless client to be loaded on architectures
481 Description: RedHat Update kernel for RHEL5
482 Details : Add the kernel config file for RHEL5.
486 Description: RedHat Update kernel for RHEL5
487 Details : Modify the kernel config file more closer RHEL5.
491 Description: Build failure against Centos5 (RHEL5)
492 Details : Define PAGE_SIZE when it isn't present.
496 Description: test_7 on sanity-quota.sh can't be used on separate machines
497 Details : add facet to handle it
501 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
502 Details : with NFS, the anon dentry's parent was set to itself in
503 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
504 corresponding dentry other than use rec->ur_name.
506 Severity : enhancement
508 Description: lfs setstripe enhancement
509 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
513 Description: enable data checksumming by default
514 Details : enable checksum by default, allow --disable-checksum
515 configure option and "-o nochecksum" mount option. Checksums
516 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
520 Description: lustre support for RHEL5
521 Details : Add support for RHEL5.
525 Description: Client eviction due to failover config
526 Details : after a connection loss, the lustre client should attempt to
527 reconnect to the last active server first before trying the
528 other potential connections.
532 Description: when mds and osts use different quota unit(32bit and 64bit),
533 quota will be released repeatly.
534 Details : void sending multiple quota reqs to mds, which will keep the
535 status between the reqs.
539 Description: osts not allocated evenly to files
540 Details : change the condition to increase offset_idx
543 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
545 Description: Data corruption for OSTs that are formatted larger than 2TB
547 Details : When generating the bio request for lustre file writes the
548 sector number would overflow a temporary variable before being
549 used for the IO. The data reads correctly from Lustre (which
550 will overflow in a similar manner) but other file data or
551 filesystem metadata may be corrupted in some cases.
555 Description: Tune the kernel for good SCSI performance.
556 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
557 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
562 Description: rewrite ext2-derived code in obdclass/uuid.c
563 Details : rewrite inherited code (uuid parsing code from ext2 utils)
564 from scratch preserving functionality.
568 Description: "lfs find -obd UUID" prints directories
569 Details : "lfs find -obd UUID" will return all directory names instead
570 of just file names. It is incorrect because the directories
571 do not reside on the OSTs.
575 Description: use i_size_read and i_size_write in 2.6 port
576 Details : replace inode->i_size access with i_size_read/write()
579 Frequency : when removing large files
581 Description: scheduling issue during removal of large Lustre files
582 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
583 It causes scheduling issues when removing large files (17TB in the
587 Frequency : only with liblustre clients on XT3
589 Description: evictions taking too long
590 Details : allow llrd to evict clients directly on OSTs
593 Frequency : only on ppc
595 Description: /proc/fs/lustre/devices broken on ppc
596 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
597 We should make sure the type of 'index' is loff_t and then cast
598 explicitly as needed below. Do not assign an explicitly cast
602 Frequency : only for rhel5
604 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
605 Details : Modify the target file & which_kernel.
607 Severity : enhancement
609 Description: omit set fsid for export NFS
610 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
611 as FS_REQUIRES_DEV for avoid problems with generate fsid.
615 Frequency : Always, for kernels after 2.6.16
616 Description: Fix warning idr_remove called for id=.. which is not allocated.
617 Details : Last kernels save old s_dev before kill super and not allow
618 to restore from callback - restore it before call kill_anon_super.
622 Description: buffer overruns could theoretically occur
623 Details : llapi_semantic_traverse() modifies the "path" argument by
624 appending values to the end of the origin string, and a
625 overrun may occur. Adding buffer overrun check in liblustreapi.
629 Description: Fix error on 'ls .' at the top of the Lustre mount.
630 Details : Don't revalidate dentry if it is a root dentry.
634 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
635 Details : Modify vm-tunables-rhel4.patch.
639 Description: Kernel config for 2.6.18-vanilla.
640 Details : Modify targets/2.6-vanilla.target.in.
641 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
642 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
643 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
644 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
648 Description: improve handling recoverable errors
649 Details : if request processig with error which can be recoverable on server
650 request should be resend, otherwise page released from cache and
655 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
656 Details : Modify target file & which_patch.
657 A flaw was found in the IA32 system call emulation provided
658 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
659 value could be stored in the %RAX register, which could trigger an
660 out-of-bounds system call table access. An untrusted local user
661 could exploit this flaw to run code in the kernel
662 (ie a root privilege escalation). (CVE-2007-4573).
666 Description: O_DIRECT bypasses client statistics.
667 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
668 read_ahead_stats were updated. copied the stats section from
669 osc_send_oap_rpc() into async_internal().
673 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
677 Description: change order of libsysio includes
678 Details : '#include sysio.h' should always come before '#include xtio.h'
682 Description: handle "serious error: objid * already exists" more gracefully
683 Details : If LAST_ID value on disk is smaller than the objects existing in
684 the O/0/d* directories, it indicates disk corruption and causes an
685 LBUG(). If the object is 0-length, then we should use the existing
686 object. This will help to avoid a full fsck in most cases.
688 Severity : enhancement
690 Description: adapt the lustre_config script to support the upgrade case
691 Details : Add "-u" option for lustre_config script to support upgrading 1.4
692 server targets to 1.6 in parallel.
696 Description: To avoid grant space > avaible space when the disk is almost
697 full. Without this patch you might see the error "grant XXXX >
698 available" or some LBUG about grant, when the disk is almost
700 Details : In filter_check_grant, for non_grant cache write, we should
701 check the left space by if (*left > ungranted + bytes), instead
702 of (*left > ungranted), because only we are sure the left space
703 is enough for another "bytes", then the ungrant space should be
704 increase. In client, we should update cl_avail_grant only there
705 is OBD_MD_FLGRANT in the reply.
709 Description: Update RHEL 4 kernel to fix local root privilege escalation.
710 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
711 described in CVE-2007-4573. This problem could allow untrusted
712 local users to gain root access.
715 Frequency : when using O_DIRECT and quotas
717 Description: Incorrect file ownership on O_DIRECT output files
718 Details : block usage reported by 'lfs quota' does not take into account
719 files that have been written with O_DIRECT.
724 Description: touch file failed when fs is not full
725 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
726 otherwise we can get ENOSP while fs is not full.
730 Description: parallel lock callbacks
731 Details : Instead of sending blocking and completion callbacks as separated
732 requests, adding them to a set and sending in parallel.
735 Frequency : only for Cray XT3
736 Bugzilla : 12829/13455
737 Description: Changing primary group doesn't change the group lustre assigns to
739 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
740 overridden with the primary group provided by the group upcall,
741 whereas the supplied fsgid can be trusted if it is in the list of
742 supplementary groups returned by the group upcall.
744 Severity : enhancement
746 Description: Allow masking D_WARNING, D_ERROR messages from console
747 Details : Console messages can now be disabled via lnet.printk.
751 Description: User code with malformed file open parameter crashes client node
752 Details : Before packing join_file req, all the related reference should be
753 checked carefully in case some malformed flags cause fake
754 join_file req on client.
758 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
759 Details : ldlm_completion_ast() assumes that a lock is granted when the req
760 mode is equal to the granted mode. However, it should also check
761 that LDLM_FL_CP_REQD is not set.
765 Description: Heavy nfs access might result in deadlocks
766 Details : After ELC code landed, it is now improper to enqueue any mds
767 locks under och_sem, because enqueue might want to decide to
768 cancel open locks for same inode we are holding och_sem for.
772 Description: Client eviction while running blogbench
773 Details : A lot of unlink operations with concurrent I/O can lead to a
774 deadlock causing evictions. To address the problem, the number of
775 oustanding OST_DESTROY requests is now throttled to
776 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
777 callbacks are processed in priority.
781 Description: enable ACLs on MDS by default
782 Details : ACLs must be enabled on MDS by default.
785 Frequency : PPC/PPC64 only
787 Description: conflicts between asm-ppc64/types.h and lustre_types.h
788 Details : fix duplicated definitions between asm-ppc64/types.h and
789 lustre_types.h on PPC.
792 Frequency : PPC/PPC64 only
794 Description: asm-ppc/segment.h does not exist
795 Details : fix compile issue on PPC.
799 Description: better handle error messages in extents code
802 Frequency : RHEL4 only
804 Description: mkfs is very slow on IA64/RHEL4
805 Details : A performance regression has been discovered in the MPT Fusion
806 driver between versions 3.02.73rh and 3.02.99.00rh. As a
807 consequence, we have downgraded the MPT Fusion driver in the RHEL4
808 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
810 Severity : enhancement
812 Description: SNMP support enhancement
813 Details : Adding total number of sampled request for an MDS node in snmp
816 Severity : enhancement
818 Description: Optimize ldlm waiting list processing for PR extent locks
819 Details : When processing waiting list for read extent lock and meeting read
820 lock that is same or wider to it that is not contended, skip
821 processing rest of the list and immediatelly return current
822 status of conflictness, since we are guaranteed there are no
823 conflicting locks in the rest of the list.
827 Description: Time out and refuse to reconnect
828 Details : When the failover node is the primary node, it is possible
829 to have two identical connections in imp_conn_list. We must
830 compare not conn's pointers but NIDs, otherwise we can defeat
831 connection throttling.
835 Description: port llog fixes from b1_6 into HEAD
836 Details : Port llog reference couting and some llog cleanups from b1_6
837 (bug 10800) into HEAD, for protect from panic and access to already
838 free llog structures.
842 Description: Detect stride IO mode in read-ahead
843 Details : When a client does stride read, read-ahead should detect that and
844 read-ahead pages according to the detected stride pattern.
848 Description: data checksumming impacts single node performance
849 Details : add support for several checksum algorithm. Currently, only CRC32
850 and Adler-32 are supported. The checksum type can be changed on
851 the fly via /proc/fs/lustre/osc/*/checksum_type.
855 Description: use adler32 for page checksums
856 Details : when available, use the Adler-32 algorithm instead of CRC32 for
861 Description: build for x2 fails
862 Details : fix compile issue on Cray systems.
866 Description: Properly match for duplicate locks
867 Details : Due to different lock order from skiplists code, we need to
868 traverse entire list for now
871 Frequency : only on PPC/SLES10
873 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
874 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
875 be defined. Add a hack in mkfs_lustre.c to work around this compile
880 Description: LASSERT on MDS when client holding flock lock dies
881 Details : ldlm pool logic depends on number of granted locks equal to
882 number of released locks which is not true for flock locks, so
883 just exclude such locks from consideration.
887 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
888 Details : Use fsfilt_sync() for both the whole filesystem sync and
889 individual file sync to eliminate dangerous inode locking
890 with I_LOCK that can lead to a deadlock.
894 Description: performance in 1.6.3
895 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
896 sure that 1MB requests are not fragmented by the block layer.
898 --------------------------------------------------------------------------------
900 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
902 * Support for kernels:
903 2.6.5-7.283 (SLES 9),
904 2.6.9-55.EL (RHEL 4),
905 2.6.16.46-0.14 (SLES 10),
906 2.6.18.8 vanilla (kernel.org)
907 * Client support for unpatched kernels:
908 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
909 2.6.16 - 2.6.22 vanilla (kernel.org)
910 * Due to recently discovered recovery problems, we do not recommend
911 using patchless RHEL 4 clients with this or any earlier release.
912 * Recommended e2fsprogs version: 1.39.cfs8
913 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
914 * Starting with this release, the ldiskfs backing filesystem required
915 by Lustre is now in its own package, lustre-ldiskfs. This package
916 should be installed. It is versioned separately from Lustre and
917 may be released separately in future.
921 Description: block reactivating mgc import until all deactivates complete
922 Details : Fix race when failing back MDT/MGS to itself (testing)
925 Frequency : at statup only
927 Description: mds_lov_synchronize race leads to various problems
928 Details : simultaneous MDT->OST connections at startup can cause the
929 sync to abort, leaving the OSC in a bad state.
931 Severity : enhancement
933 Description: add optional extra BUILD_VERSION info
934 Details : add a new environment variable (namely LUSTRE_VERS) which allows
935 to override the lustre version.
938 Frequency : 2.6.18 servers only
940 Description: ll_kern_mount() doesn't release the module reference
941 Details : The ldiskfs module reference count never drops down to 0
942 because ll_kern_mount() doesn't release the module reference.
947 Description: server LBUG when using old ost_num_threads parameter
948 Details : Accept the old ost_num_threads parameter but warn that it
949 is deprecated, and fix an off-by-one error that caused an LBUG.
954 Description: Transient SCSI error results in persistent IO issue
955 Details : iobuf->dr_error is not reinitialized to 0 between two
959 Frequency : sometimes when underlying device returns I/O errors
961 Description: OSTs not going read-only during write failures
962 Details : OSTs are not remounted read-only when the journal commit threads
963 get I/O errors because fsfilt_ext3 calls journal_start/stop()
964 instead of the ext3 wrappers.
968 Description: poor connect scaling with increasing client count
969 Details : Don't run filter_grant_sanity_check for more than 100 exports
970 to improve scaling for large numbers of clients.
973 Frequency : SLES10 only
975 Description: sanity-quota.sh quotacheck failed: rc = -22
976 Details : Quotas cannot be enabled on SLES10.
979 Frequency : liblustre clients only
981 Description: getdirentries does not give error when run on compute nodes
982 Details : getdirentries does not fail when the size specified as an argument
983 is too small to contain at least one entry
985 Severity : enhancement
987 Description: Add LNET router traceability for debug purposes
988 Details : If a checksum failure occurs with a router as part of the
989 IO path, the NID of the last router that forwarded the bulk data
990 is printed so it can be identified.
995 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
996 Details : Due to a race condition, liblustre clients were occasionally
999 Severity : enhancement
1001 Description: lfs setstripe use optional parameters instead of postional
1004 Severity : enhancement
1006 Description: Nanosecond timestamp support for ldiskfs
1007 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1008 resolution timestamps. There is not yet support for this at
1009 the Lustre filesystem level.
1012 Frequency : during server recovery
1014 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1015 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1019 Frequency : nfs export on patchless client
1021 Description: connectathon hang when test nfs export over patchless client
1022 Details : Disconnected dentry cannot be found with lookup, so we do not need
1023 to unhash it or make it invalid
1026 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1027 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1028 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1033 Description: open req refcounting wrong on reconnect
1034 Details : If reconnect happened between getting open reply from server and
1035 call to mdc_set_replay_data in ll_file_open, we will schedule
1036 replay for unreferenced request that we are about to free.
1037 Subsequent close will crash in variety of ways.
1038 Check that request is still eligible for replay in
1039 mdc_set_replay_data().
1044 Description: disable writes to filesystem when reading health_check file
1045 Details : the default for reading the health_check proc file has changed
1046 to NOT do a journal transaction and write to disk, because this
1047 can cause reads of the /proc file to hang and block HA state
1048 checking on a healthy but otherwise heavily loaded system. It
1049 is possible to return to the previous behaviour during configure
1050 with --enable-health-write.
1052 Severity : enhancement
1054 Description: 64-bit inode version
1055 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1056 to the inode. This will be required for version-based recovery.
1061 Description: MDS fails to start if a duplicate client export is detected
1062 Details : in some rare cases it was possible for a client to connect to
1063 an MDS multiple times. Upon recovery the MDS would detect this
1064 and fail during startup. Handle this more gracefully.
1066 Severity : enhancement
1068 Description: Add -o localflock option to simulate old noflock
1070 Details : This will achieve local-only flock/fcntl locks
1076 Description: log_commit_thread vs filter_destroy race leads to crash
1077 Details : Take import reference before releasing llog record semaphore
1082 Description: Wrong request locking in request set processing
1083 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1084 imp_delayed_list, in this place should be used imp_lock.
1087 Frequency : when reconnection
1089 Description: Grant Leak when osc reconnect to OST
1090 Details : When osc reconnect ost, OST(filter) should check whether it
1091 should grant more space to client by comparing fed_grant and
1092 cl_avail_grant, and return the granted space to client instead
1093 of "new granted" space, because client will call osc_init_grant
1094 to update the client grant space info.
1097 Frequency : when client reconnect to OST
1099 Description: Grant Leak when osc do resend and replay bulk write
1100 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1101 bulk write request, because the grant info while be sync between
1102 OSC and OST when reconnect, and we should ignore the grant info
1103 these of resend/replay write req.
1108 Description: Grant space more than avaiable left space sometimes.
1109 Details : When then OST is about to be full, if two bulk writing from
1110 different clients came to OST. Accord the avaliable space of the
1111 OST, the first req should be permitted, and the second one
1112 should be denied by ENOSPC. But if the seconde arrived before
1113 the first one is commited. The OST might wrongly permit second
1114 writing, which will cause grant space > avaiable space.
1117 Frequency : when client is evicted
1119 Description: Grant might be wrongly erased when osc is evicted by OST
1120 Details : when the import is evicted by server, it will fork another
1121 thread ptlrpc_invalidate_import_thread to invalidate the
1122 import, where the grant will be set to 0. While the original
1123 thread will update the grant it got when connecting. So if
1124 the former happened latter, the grant will be wrongly errased
1125 because of this race.
1130 Description: Checking Stale with correct fid
1131 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1132 to check whether it is stale, because sometimes, we want the
1133 enqueue happened anyway, and op_data.fid2 will not be initialized.
1135 Severity : enhancement
1137 Description: update patchless client
1138 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1141 Frequency : only with 2.4 kernel
1143 Description: random memory corruption
1144 Details : size of struct ll_inode_info is to big for union inode.u and this
1145 can be cause of random memory corruption.
1150 Description: Memory leak in recovery
1151 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1152 It should also check obd_fail before fsfilt_start, otherwise if
1153 fsfilt_start return -EROFS,(failover mds during mds recovery).
1154 then the req will return with repmsg->transno = 0 and rc = EROFS.
1155 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1156 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1157 be freed no matter whether fsfilt_commit success or not.
1160 Frequency : only with huge count clients
1162 Description: Prevents from taking the superblock lock in llap_from_page for
1164 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1165 for llap_from_page call in ll_removepage prevents from taking the
1166 superblock lock for a soon died page.
1171 Description: Not check open intent error before release open handle
1172 Details : in some rare cases, the open intent error is not checked before
1173 release open handle, which may cause
1174 ASSERTION(open_req->rq_transno != 0), because it tries to release
1175 the failed open handle.
1180 Description: Set cat log bitmap only after create log success.
1181 Details : in some rare cases, the cat log bitmap is set too early. and it
1182 should be set only after create log success.
1186 Description: Accessing a block bevice can re-enable I/O when Lustre is
1187 tearing down a device.
1188 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1193 Description: service threads may hog cpus when there are a lot of requests
1194 Details : Insert cond_resched to give other threads a chance to use some CPU
1199 Description: the cat log was not initialized in recovery
1200 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1201 unlink log on mds will not be initialized until mds post
1202 recovery. And also in mds post recovery, the unlink log will
1203 initialization will be done asynchronausly, so there will be race
1204 between add unlink log and unlink log initialization.
1208 Description: brw_stats were being printed incorrectly
1209 Details : brw_stats were being printed as log2 but all of them were not
1210 recorded as log2. Also remove some code duplication arising from
1211 filter_tally_{read,write}.
1215 Frequency : rare, only in recovery.
1216 Description: ASSERTION(req->rq_type != LI_POISON) failed
1217 Details : imp_lock should be held while iterating over imp_sending_list for
1218 prevent destroy request after get timeout in ptlrpc_queue_wait.
1222 Description: replay-single.sh test 52 fails
1223 Details : A lock's skiplist need to be cleanup when it being unlinked
1224 from its resource list.
1228 Description: Short directio read returns full requested size rather than
1230 Details : Direct I/O operations should return actual amount of bytes
1231 transferred rather than requested size.
1233 Severity : enhancement
1235 Description: metadata RPC reduction (e.g. for rm performance)
1236 Details : decrease the amount of synchronous RPC between clients and servers
1237 by canceling conflicing lock before the operation on the client
1238 and packing thier handles into the main operation RPC to server.
1240 Severity : enhancement
1242 Description: Async OSC create to avoid the blocking unnecessarily.
1243 Details : If a OST has no remain object, system will block on the creating
1244 when need to create a new object on this OST. Now, ways use
1245 pre-created objects when available, instead of blocking on an
1246 empty osc while others are not empty. If we must block, we block
1247 for the shortest possible period of time.
1249 Severity : enhancement
1251 Description: refine locking for avoid write wrong info into lov_objid file
1252 Details : fix possible races with add new target and write/update data in
1255 --------------------------------------------------------------------------------
1257 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1262 Frequency : on some architectures
1264 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1265 Details : Uninitialized flags sometimes cause configuration commands to
1269 Frequency : patchless clients only
1271 Description: missing __iget() symbol export
1272 Details : The __iget() symbol export is missing. To avoid the need for
1273 this on patchless clients the deathrow inode reaper is turned
1274 off, and we depend on the VM to clean up old inodes. This
1275 dependency was during via the fix for bug 12181.
1277 --------------------------------------------------------------------------------
1279 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1281 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1282 INTEROPERATE with older versions automatically. In many cases a
1283 special upgrade step is needed. Please read the
1284 user documentation before upgrading any part of a live system.
1285 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1286 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1287 * WARNING: Lustre configuration and startup changes are required with
1288 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1294 Severity : enhancement
1296 Description: MountConf
1297 Details : Lustre configuration is now managed via mkfs and mount
1298 commands instead of lmc and lconf. New obd types (MGS, MGC)
1299 are added for dynamic configuration management. See
1300 https://mail.clusterfs.com/wikis/lustre/MountConf for
1303 Severity : enhancement
1305 Description: dynamic OST addition
1306 Details : OSTs can now be added to a live filesystem
1308 Severity : enhancement
1310 Description: startup order invariance
1311 Details : MDTs and OSTs can be started in any order. Clients only
1312 require the MDT to complete startup.
1314 Severity : enhancement
1316 Description: parallel, asynchronous orphan cleanup
1317 Details : orphan cleanup is now performed in separate threads for each
1318 OST, allowing parallel non-blocking operation.
1320 Severity : enhancement
1322 Description: optimized stripe assignment
1323 Details : stripe assignments are now made based on ost space available,
1324 ost previous usage, and OSS previous usage, in order to try
1325 to optimize storage space and networking resources.
1327 Severity : enhancement
1329 Description: Permanently set tunables
1330 Details : All writable /proc/fs/lustre tunables can now be permanently
1331 set on a per-server basis, at mkfs time or on a live system.
1333 Severity : enhancement
1335 Description: Lustre message v2
1336 Details : Add lustre message format v2.
1338 Severity : enhancement
1340 Description: client OST exclusion list
1341 Details : Clients can be started with a list of OSTs that should be
1342 declared "inactive" for known non-responsive OSTs.
1346 Description: SPEC SFS validation failure on NFS v2 over lustre.
1347 Details : Changes the blocksize for regular files to be 2x RPC size,
1348 and not depend on stripe size.
1350 Severity : enhancement
1352 Description: Multiple MD RPCs in flight.
1353 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1354 To avoid overly-overloading MDS, introduce a limit on number of
1355 MDS RPCs in flight for a single client and add /proc controls
1356 to adjust this limit.
1358 Severity : enhancement
1360 Description: client read/write statistics
1361 Details : Add client read/write call usage stats for performance
1362 analysis of user processes.
1363 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1364 file access. extents_stats shows chunk size distribution.
1365 extents_stats_per_process show chunk size distribution per
1368 Severity : enhancement
1370 Description: mds statistics
1371 Details : Add detailed mds operations statistics in
1372 /proc/fs/lustre/mds/*/stats.
1376 Description: Failure of copying files with lustre special EAs.
1377 Details : Client side always return success for setxattr call for lustre
1378 special xattr (currently only "trusted.lov").
1383 Description: Refcount LNET uuids
1384 Details : The global LNET uuid list grew linearly with every startup;
1385 refcount repeated list entries instead of always adding to
1388 Severity : enhancement
1390 Description: Dynamic service threads
1391 Details : Within a small range, start extra service threads
1392 automatically when the request queue builds up.
1395 Frequency : mixed-endian client/server environments
1397 Description: mixed-endian crashes
1398 Details : The new msg_v2 system had some failures in mixed-endian
1401 Severity : enhancement
1403 Description: Easy OST removal
1404 Details : OSTs can be permanently deactivated with e.g. 'lctl
1405 conf_param lustre-OST0001.osc.active=0'
1407 Severity : enhancement
1409 Description: MGS proc entries
1410 Details : Added basic proc entries for the MGS showing what filesystems
1413 Severity : enhancement
1415 Description: provide MGS failover
1416 Details : Added config lock reacquisition after MGS server failover.
1418 Severity : enhancement
1420 Description: add Linux 2.4 support
1421 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1425 Description: a large application tries to do I/O to the same resource and dies
1426 in the middle of it.
1427 Details : Check the req->rq_arrival time after the call to
1428 ost_brw_lock_get(), but before we do anything about
1429 processing it & sending the BULK transfer request. This
1430 should help move old stale pending locks off the queue as
1431 quickly as obd_timeout.
1434 Frequency : when an incorrect nid is specified during startup
1436 Description: ptlrpc connect to non-existant node causes kernel crash
1437 Details : LNET can't be re-entered from an event callback, which
1438 happened when we expire a message after the export has been
1439 cleaned up. Instead, hand the zombie cleanup off to another
1442 Severity : enhancement
1444 Description: plain/inodebits lock performance improvement
1445 Details : Grouping plain/inodebits in granted list by their request modes
1446 and bits policy, thus improving the performance of search through
1450 Frequency : only if OST filesystem is corrupted
1452 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1453 Details : for a short time RPCs with bulk IO are in the replay list,
1454 but replay of bulk IOs is unimplemented. If the OST filesystem
1455 is corrupted due to disk cache incoherency and then replay is
1456 started it is possible to trip an assertion. Avoid putting
1457 committed RPCs into the replay list at all to avoid this issue.
1460 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1462 Description: System hang on startup
1463 Details : This bug allowed the liblustre (e.g. catamount) client to
1464 return to the app before handling all startup RPCs. This
1465 could leave the node unresponsive to lustre network traffic
1466 and manifested as a server ptllnd timeout.
1468 Severity : enhancement
1470 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1471 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1472 boolean to control whether to print peer debug info when a
1473 client's RPC times out.
1476 Frequency : only for kernels with patches from Lustre below 1.4.3
1478 Description: Remove old rdonly API
1479 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1482 Frequency : only for devices with external journals
1484 Description: Set external device read-only also
1485 Details : During a commanded failover stop, we set the disk device
1486 read-only while the server shuts down. We now also set any
1487 external journal device read-only at the same time.
1490 Frequency : when upgrading from 1.4 while trying to change parameters
1492 Description: The wrong (new) MDC name was used when setting parameters for
1493 upgraded MDT's. Also allows changing of OSC (and MDC)
1494 parameters if --writeconf is specified at tunefs upgrade time.
1497 Frequency : when setting specific ost indicies
1499 Description: QOS code breaks on skipped indicies
1500 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1501 created with --index need not be sequential.
1503 Severity : enhancement
1505 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1506 Details : The uninit_groups feature works in conjunction with the kernel
1507 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1508 up the pass1 processing of e2fsck. This is a read-only feature
1509 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1510 mount filesystems that have had this feature enabled.
1512 Severity : enhancement
1514 Description: Improve multi-block allocation algorithm to avoid fragmentation
1515 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1516 allocation locality and avoid filesystem fragmentation.
1518 ------------------------------------------------------------------------------
1520 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1522 * Support for kernels:
1523 2.4.21-47.0.1.EL (RHEL 3)
1524 2.6.5-7.283 (SLES 9)
1525 2.6.9-42.0.10.EL (RHEL 4)
1526 2.6.12.6 vanilla (kernel.org)
1527 2.6.16.27-0.9 (SLES 10)
1528 * Recommended e2fsprogs version: 1.39.cfs5
1530 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1534 Frequency : occasional, depends on client load and configuration
1535 Bugzilla : 12181, 12203
1536 Description: data loss for recently-modified files
1538 Details : In some cases it is possible that recently written or created
1539 files may not be written to disk in a timely manner (this should
1540 normally be within 30s unless client IO load is very high).
1541 The problem appears as zero-length files or files that are a
1542 multiple of 1MB in size after a client crash or client eviction
1543 that are missing data at the end of the file.
1545 This problem is more likely to be hit on clients where files are
1546 repeatedly created and unlinked in the same directory, clients
1547 have a large amount of RAM, have many CPUs, the filesystem has
1548 many OSTs, the clients are rebooted frequently, and/or the files
1549 are not accessed by other nodes after being written.
1551 The presence of the problem can be detected by looking at
1552 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1553 smaller than the second (nr_unused) then dirty files will not
1554 be flushed automatically to disk. "sync; sleep 10" should be
1555 run several times on the node before unmounting it to update
1556 Lustre (this is also safe to run on nodes without this problem).
1558 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1559 that can cause this same problem, so customers using that kernel
1560 also need to update the kernel in addition to Lustre. In order
1561 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1563 It is normal that files written just before a client crash (less
1564 than 30s) may not yet have been flushed to disk, even for local
1568 Frequency : frequent on thin XT3 nodes
1570 Description: UUID collision on thin XT3 Linux nodes
1571 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1572 randomly, since we relied on an insufficiently-seeded PRNG.
1577 Description: OSS hangs after "All ost request buffers busy"
1578 Details : A deadlock between quota and journal operations caused OSS
1579 hangs after printing "All ost request buffers busy."
1582 Frequency : always on liblustre builds
1584 Description: Cleanup compiler warnings on liblustre
1587 Frequency : always on liblustre builds on XT3
1589 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1593 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1594 > 32000 subdirectories
1595 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1596 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1597 subdirectory count crosses 32000. This will aid e2fsck to
1598 correctly handle more than 32000 subdirectories.
1601 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1603 Description: System hang on startup
1604 Details : This bug allowed the liblustre (e.g. catamount) client to
1605 return to the app before handling all startup RPCs. This
1606 could leave the node unresponsive to lustre network traffic
1607 and manifested as a server ptllnd timeout.
1609 Severity : enhancement
1611 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1612 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1613 boolean to control whether to print peer debug info when a
1614 client's RPC times out.
1619 Description: make O_SYNC working on 2.6 kernels
1620 Details : 2.6 kernels use different method for mark pages for write,
1621 so need add a code to lustre for O_SYNC work.
1626 Description: Failure to close file and release space on NFS
1627 Details : Put inode details into lock acquired in ll_intent_file_open.
1628 Use mdc_intent_lock in ll_intent_open to properly
1629 detect all kind of errors unhandled by mdc_enqueue.
1634 Description: proc file read during shutdown sometimes raced obd removal,
1636 Details : Add lock to prevent obd access after proc file removal.
1639 Frequency : Only for files larger than 4GB on 32-bit clients.
1641 Description: improperly doing page alignment of locks
1642 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1643 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1648 Description: RPCs being resent when they shouldn't be
1649 Details : Some RPCs that should not be resent are being resent. This
1650 can cause inconsistencies in the RPC state machine. Do not
1651 resend such requests.
1654 Frequency : rare, only with NFS export
1656 Description: Crash on NFS re-export node
1657 Details : under very unusual load conditions an assertion is hit in
1658 ll_intent_file_open()
1661 Frequency : only if OST filesystem is corrupted
1663 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1664 Details : for a short time RPCs with bulk IO are in the replay list,
1665 but replay of bulk IOs is unimplemented. If the OST filesystem
1666 is corrupted due to disk cache incoherency and then replay is
1667 started it is possible to trip an assertion. Avoid putting
1668 committed RPCs into the replay list at all to avoid this issue.
1673 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1674 Details : Large single O_DIRECT read and write calls can fail to allocate
1675 a sufficiently large buffer to process the request. In case of
1676 allocation failure the allocation is retried with a smaller
1677 buffer and broken into smaller requests.
1679 Severity : enhancement
1681 Description: Add -o localflock option to simulate old noflock behaviour.
1682 Details : This will achieve local-only flock/fcntl locks coherentness.
1687 Description: versioning check is incomplete
1688 Details : Checking the version difference of client vs. server, report
1689 error if the gap is too big.
1694 Description: add support PG_writeback bit
1695 Details : add support for PG_writeback bit for Lustre, for more carefull
1696 work with page cache in 2.6 kernel. This also fix some deadlocks
1697 and remove hack for work O_SYNC with 2.6 kernel.
1699 Severity : enhancement
1701 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1702 Details : The uninit_groups feature works in conjunction with the kernel
1703 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1704 up the pass1 processing of e2fsck. This is a read-only feature
1705 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1706 mount filesystems that have had this feature enabled.
1708 Severity : enhancement
1710 Description: Improve multi-block allocation algorithm to avoid fragmentation
1711 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1712 allocation locality and avoid filesystem fragmentation.
1714 ------------------------------------------------------------------------------
1716 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1718 * Support for kernels:
1719 2.6.9-42.0.3.EL (RHEL 4)
1720 2.6.5-7.276 (SLES 9)
1721 2.4.21-47.0.1.EL (RHEL 3)
1722 2.6.12.6 vanilla (kernel.org)
1723 2.6.16.21-0.8 (SLES10)
1724 * Recommended e2fsprogs version: 1.39.cfs2-0
1726 * The backwards-compatible /proc/sys/portals symlink has been removed
1727 in this release. Before upgrading, please ensure that you change
1728 any configuration scripts or /etc/sysctl.conf files that access
1729 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1730 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1731 in advance of the upgrade on any system running Lustre 1.4.6 or
1732 newer, since /proc/sys/lnet was added in that version.
1733 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1737 Frequency : only when quota is used
1739 Description: avoid scanning export list for quota master
1740 Details : Change the algorithms to avoid scanning export list in order
1741 to improve the efficiency.
1744 Frequency : MDS failover only, very rarely
1746 Description: "went back in time" messages on mds failover
1747 Details : The greatest transno may be lost when the current operation
1748 finishes with an error (transno==0) and the client's last_rcvd
1749 record is over-written. Save the greatest transno in the
1750 mds_last_transno for this case.
1753 Frequency : always for specific kernels and striping counts
1755 Description: client may get "Matching packet too big" without ACL support
1756 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1757 when trying to access files in certain configurations. The
1758 clients should in fact be denied when mounting because they do
1759 not understand ACLs.
1762 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1764 Description: many clients connecting with IO in progress causes connect timeouts
1765 Details : Avoid synchronous journal commits to avoid delays caused by many
1766 clients connecting/disconnecting when bulk IO is in progress.
1767 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1768 OST_IO_PORTAL to avoid delays behind potentially many pending
1772 Frequency : occasionally with multiple writers to a single file
1774 Description: shared writes to file may result in wrong size reported by stat()
1775 Details : Allow growing of kms when extent lock is cancelled
1778 Frequency : always with random mmap IO to multi-striped file
1780 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1781 Details : Only if the hole is at the end of OST object so that kms is too
1782 small. Fix is to increase kms accordingly in ll_nopage.
1785 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1787 Description: writes to a missing object would leak memory on the OST
1788 Details : If there is an inconsistency between the MDS and OST filesystems,
1789 such that the MDS references an object that doesn't exist, writes
1790 to that object will leak memory due to incorrect cleanup in the
1791 error handling path, eventually running out of memory on the OST.
1796 Description: Creating too long symlink causes lustre errors
1797 Details : Check symlink and name lengths before sending requests to MDS.
1800 Frequency : only if flock is enabled (not on by default)
1802 Description: posix locks not released on fd closure on 2.6.9+
1803 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1804 kernels, this caused such locks not to be released on fd close and
1805 then assertions on fs unmount about still used locks.
1808 Frequency : MDS failover only, very rarely
1810 Description: clients may get ASSERTION(granted_lock != NULL)
1811 Details : When request was taking a long time, and a client was resending
1812 a getattr by name lock request. The were multiple lock requests
1813 with the same client lock handle and
1814 mds_getattr_name->fixup_handle_for_resent_request found one of the
1815 lock handles but later failed with ASSERTION(granted_lock != NULL).
1820 Description: handle->h_buffer_credits > 0, assertion failure
1821 Details : h_buffer_credits is zero after truncate, causing assertion
1822 failure. This patch extends the transaction or creates a new
1826 Frequency : NFS re-export or patchless client
1827 Bugzilla : 11179, 10796
1828 Description: Crash on NFS re-export node (__d_move)
1829 Details : We do not want to hash the dentry if we don't have a lock.
1830 But if this dentry is later used in d_move, we'd hit uninitialised
1831 list head d_hash, so we just do this to init d_hash field but
1832 leave dentry unhashed.
1835 Frequency : NFS re-export or patchless client
1837 Description: NFS exports has problem with symbolic link
1838 Details : lustre client didn't properly install dentry when re-exported
1839 to NFS or running patchless client.
1842 Frequency : NFS re-export or patchless client
1844 Description: Various nfs/patchless fixes.
1845 Details : fixes reuse disconected alias for lookup process - this fixes
1846 warning "find_exported_dentry: npd != pd",
1847 fix permission error with open files at nfs.
1848 fix apply umask when do revalidate.
1851 Frequency : occasional
1853 Description: Crash on NFS re-export node
1854 Details : calling clear_page() on the wrong pointer triggered oops in
1855 generic_mapping_read().
1858 Frequency : rarely, using O_DIRECT IO
1860 Description: unaligned directio crashes client with LASSERT
1861 Details : check for unaligned buffers before trying any requests.
1864 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1866 Description: stale data returned from RAID cache
1867 Details : If only a small amount of IO is done to the RAID device before
1868 reading it again it is possible to get stale data from the RAID
1869 cache instead of reading it from disk.
1872 Frequency : always for sles10 kernel
1874 Description: sles10 support
1875 Details : ll_follow_link: compile fixes and using of nd_set_link
1876 under newer kernels.
1879 Frequency : depends on arch, kernel and compiler version, always on sles10
1882 Description: recursive or deep enough symlinks cause stack overflow
1883 Details : getting rid of large stack-allocated variable in
1887 Frequency : depends on hardware
1889 Description: lustre write performance loss in the SLES10 kernel
1890 Details : the performance loss is caused by using of write barriers in the
1891 ext3 code. The SLES10 kernel turns barrier support on by
1892 default. The fix is to undo that change for ldiskfs.
1894 ------------------------------------------------------------------------------
1896 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1898 * Support for kernels:
1899 2.6.9-42.0.3EL (RHEL 4)
1900 2.6.5-7.276 (SLES 9)
1901 2.4.21-47.0.1.EL (RHEL 3)
1902 2.6.12.6 vanilla (kernel.org)
1906 Frequency : quota enabled and large files being deleted
1908 Description: releasing more than 4GB of quota at once hangs OST
1909 Details : If a user deletes more than 4GB of files on a single OST it
1910 will cause the OST to spin in an infinite loop. Release
1911 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1916 Description: statfs data retrieved from /proc may be stale or zero
1917 Details : When reading per-device statfs data from /proc, in the
1918 {kbytes,files}_{total,free,avail} files, it may appear
1919 as zero or be out of date.
1922 Frequency : systems with MD RAID1 external journal devices
1924 Description: lconf's call to blkid is confused by RAID1 journal devices
1925 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1926 of returning all block devices that match the journal UUID.
1929 Frequency : always, for aggregate stripe size over 4GB
1931 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1932 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1933 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1936 Frequency : always if "lfs find" used on a local file/directory
1938 Description: "lfs find" segfaults if used on a local file/directory
1939 Details : The case where a directory component was not specified wasn't
1940 handled correctly. Handle this properly.
1943 Frequency : always on ppc64
1945 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1946 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1947 perform endianness conversion.
1950 Frequency : rarely (truncate to non-zero file size after write under load)
1951 Bugzilla : 10730, 10687
1952 Description: Files padded with zeros to next 4K multiple
1953 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1954 it is possible that files that are truncated to a non-zero size
1955 immediately after being written are filled with zero bytes beyond
1956 the truncated size. No file data is lost.
1958 Severity : enhancement
1960 Description: Allow recovery/failover for liblustre clients.
1961 Details : liblustre clients were unaware of failover configurations until
1964 Severity : enhancement
1966 Description: user file locks should fail when not mounting with flock option
1967 Details : Set up an error-returning stub in ll_file_operations.lock field
1968 to prevent incorrect behaviour when client is mounted without
1969 flock option. Also, set up properly f_op->flock field for
1973 Frequency : always on ia64
1975 Description: "lfs df" loops on printing out MDS statfs information
1976 Details : The obd_ioctl_data was not initialized and in some systems
1977 this caused a failure during the ioctl that did not return
1978 an error. Initialize the struct and return an error on failure.
1981 Frequency : SLES 9 only
1983 Description: Error of copying files with lustre special EAs as root
1984 Details : Client side always return success for setxattr call for lustre
1985 special xattr (currently only "trusted.lov").
1988 Frequency : rarely on clusters with both ia64+i386 clients
1990 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1991 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1992 clients are doing concurrent writes to the same file it is
1993 possible that the ia64 clients may LASSERT because the OST
1994 extent locks are not PAGE_SIZE aligned. Ensure that grown
1995 locks are always aligned on the request boundary.
1998 Frequency : specific use, occasional
2000 Description: Overwriting in use executable truncates on-disk binary image
2001 Details : If one node attempts to overwrite an executable in use by
2002 another node, we now correctly return ETXTBSY instead of
2003 truncating the file.
2005 Severity : enhancement
2007 Description: Async OSC create to avoid the blocking unnecessarily.
2008 Details : If a OST has no remain object, system will block on the creating
2009 when need to create a new object on this OST. Now, ways use
2010 pre-created objects when available, instead of blocking on an
2011 empty osc while others are not empty. If we must block, we block
2012 for the shortest possible period of time.
2017 Description: chmod on Lustre root is propagated to other clients
2018 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2019 invalid by the follow_mount time.
2022 Frequency : liblustre clients only
2024 Description: Race in 'instant cancel' lock handling could lead to such locks
2025 never to be granted in case of SMP MDS
2026 Details : Do not destroy not yet granted but cbpending locks in
2030 Frequency : replay/resend of open
2032 Description: non null lock assetion failure in mds_intent_policy
2033 Details : Trying to replay/resend lockless open requests resulted in
2034 mds_open() returning 0 with no lock. Now it sets a flag if
2035 a lock is going to be returned.
2037 Severity : enhancement
2039 Description: Checksum enhancements
2040 Details : New checksum enhancements allow for resending RPCs that failed
2043 Severity : enhancement
2045 Description: Tunables on number of dirty pages in cacche
2046 Details : Allow to set limit on number of dirty pages cached.
2051 Description: client crash on unmount - lock still has references
2052 Details : In some error handling cases it was possible to leak a lock
2053 reference on a client while accessing a file. This was not
2054 harmful to the client during operation, but would cause the
2055 client to crash when the filesystem is unmounted.
2058 Frequency : specific case, rare
2060 Description: ETXTBSY on mds though file not in use by client
2061 Details : ETXTBSY is no longer incorrectly returned when attempting to
2062 chmod or chown a directory that the user previously tried to
2063 execute or a currently-executing binary.
2066 Frequency : extremely rare except on liblustre-based clients
2068 Description: Lustre space not freed when files are deleted
2069 Details : Clean up open-unlinked files after client eviction. Previously
2070 the unlink was skipped and the files remained as orphans.
2075 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2076 Details : In some cases it was possible to send a blocking callback to
2077 a client doing a glimpse, even though that client didn't get
2078 a lock granted. When the glimpse lock is cancelled on the OST
2079 the freed lock is left on the waiting list and corrupted the list.
2082 Frequency : all core dumps
2084 Description: Broke core dumps to lustre
2085 Details : Negative dentry may be unhashed if parent does not have UPDATE
2086 lock, but some callers, e.g. do_coredump, expect dentry to be
2087 hashed after successful create, hash it in ll_create_it.
2089 ------------------------------------------------------------------------------
2091 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2093 * Support for kernels:
2094 2.6.9-42.0.2.EL (RHEL 4)
2095 2.6.5-7.276 (SLES 9)
2096 2.4.21-40.EL (RHEL 3)
2097 2.6.12.6 vanilla (kernel.org)
2101 Frequency : always on RHEL 3
2103 Description: Number of open files grows over time
2104 Details : The number of open files grows over time, whether or not
2105 Lustre is started. This was due to a filp leak introduced
2106 by one of our kernel patches.
2108 ------------------------------------------------------------------------------
2110 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2112 * Support for kernels:
2113 2.6.9-42.EL (RHEL 4)
2114 2.6.5-7.276 (SLES 9)
2115 2.4.21-40.EL (RHEL 3)
2116 2.6.12.6 vanilla (kernel.org)
2121 Bugzilla : 5719, 9635, 9792, 9684
2122 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2123 Details : If a server is under heavy load and cannot reply to new
2124 connection requests before the client resends the (re)connect,
2125 the connection handling code can behave badly if two service
2126 threads are concurrently handing separate (re)connections from
2127 the same client. Add better locking to the connection handling
2128 code, and ensure that only a single connection will be processed
2129 for a given client UUID, even if the lock is dropped.
2131 Severity : enhancement
2133 Description: add TCP zero-copy support to kernel
2134 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2135 sends if the hardware supports scatter-gather and checksumming.
2136 This allows socklnd to do client-write and server-read more
2137 efficiently and reduce CPU utilization from skbuf copying.
2140 Frequency : only if NFS exporting from client
2142 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2143 Details : If setting attributes on a file created under NFS that had
2144 never been opened it would be possible to oops the client
2145 if the file had no objects.
2148 Frequency : always for liblustre
2150 Description: liblustre client does MDS+OSTs setattr RPC for each write
2151 Details : When doing a write from a liblustre client, the client
2152 incorrectly issued an RPC to the MDS and each OST the file was
2153 striped over in order to update the timestamps. When writing
2154 with small chunks and many clients this could overwhelm the MDS
2155 with RPCs. In all cases it would slow down the write because
2156 these RPCs are unnecessary.
2158 Severity : enhancement
2160 Description: allow number of MDS service threads to be changed at module load
2161 Details : It is now possible to change the number of MDS service threads
2162 running. Adding "options mds mds_num_threads={N}" to the MDS's
2163 /etc/modprobe.conf will set the number of threads for the next
2164 time Lustre is restarted (assuming the "mds" module is also
2165 reloaded at that time). The default number of threads will
2166 stay the same, 32 for most systems.
2171 Description: OST crash if filesystem is unformatted or corrupt
2172 Details : If an OST is started on a device that has never been formatted
2173 or if the filesystem is corrupt and cannot even mount then the
2174 error handling cleanup routines would dereference a NULL pointer.
2179 Description: NULL pointer deref in llap_from_page.
2180 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2181 incorrect) mapping if the page was truncated/reclaimed while it was
2182 searched for. Check for this condition and skip such pages when
2183 doing readahead. Introduce extra check to llap_from_page() to
2184 verify page->mapping->host is non-NULL (so page is not anonymous).
2187 Frequency : Sometimes when using sys_sendfile
2189 Description: "page not covered by a lock" warnings from ll_readpage
2190 Details : sendfile called ll_readpage without right page locks present.
2191 Now we introduced ll_file_sendfile that does necessary locking
2192 around call to generic_file_sendfile() much like we do in
2196 Frequency : with certain MDS communication failures at client mount time
2198 Description: NULL pointer deref after failed client mount
2199 Details : a client connection request may delayed by the network layer
2200 and not be sent until after the PTLRPC layer has timed out the
2201 request. If the client fails the mount immediately it will try
2202 to clean up before the network times out the request. Add a
2203 reference from the request import to the obd device and delay
2204 the cleanup until the network drops the request.
2207 Frequency : occasionally during client (re)connect
2209 Description: assertion failure during client (re)connect
2210 Details : processing a client connection request may be delayed by the
2211 client or server longer than the client connect timeout. This
2212 causes the client to resend the connection request. If the
2213 original connection request is replied in this interval, the
2214 client may trip an assertion failure in ptlrpc_connect_interpret()
2215 which thought it would be the only running connect process.
2218 Frequency : only with obd_echo servers and clients that are rebooted
2220 Description: kernel BUG accessing uninitialized data structure
2221 Details : When running an obd_echo server it did not start the ping_evictor
2222 thread, and when a client was evicted an uninitialized data
2223 structure was accessed. Start the ping_evictor in the RPC
2224 service startup instead of the OBD startup.
2226 Severity : enhancement
2227 Bugzilla : 10193 (patchless)
2228 Description: Remove dependency on various unexported kernel interfaces.
2229 Details : No longer need reparent_to_init, exit_mm, exit_files,
2230 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2233 Frequency : rare (only users of deprecated and unsupported LDAP config)
2235 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2236 Details : LDAP apparently contains 'lustreName' attributes instead of
2237 'name'. A simple remapping of the name is sufficient.
2240 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2242 Description: waiting_locks_callback trips kernel BUG if client is evicted
2243 Details : Running with the dump_on_timeout debug flag turned on makes
2244 it possible that the waiting_locks_callback() can try to dump
2245 the Lustre kernel debug logs from an interrupt handler. Defer
2246 this log dumping to the expired_lock_main() thread.
2248 Severity : enhancement
2250 Description: Support NFS exporting on 2.6 kernels.
2251 Details : Implement non-rawops metadata methods for NFS server to use without
2252 changing NFS server code.
2255 Frequency : very rare (synthetic metadata workload only)
2257 Description: two racing renames might cause an MDS thread to deadlock
2258 Details : Running the "racer" program may cause one MDS thread to rename
2259 a file from being the source of a rename to being the target of
2260 a rename at exactly the same time that another thread is doing
2261 so, and the second thread has already enqueued these locks after
2262 doing a lookup of the target and is trying to relock them in
2263 order. Ensure that we don't try to re-lock the same resource.
2266 Frequency : only very large systems with liblustre clients
2268 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2269 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2270 all OSTs in parallel. This allows the eviction of stale liblustre
2271 clients to proceed much faster than if they were done in series,
2272 and also offers similar improvements for other set_info RPCs.
2277 Description: excessive CPU usage during initial read phase on client
2278 Details : During the initial read phase on a client, it would agressively
2279 retry readahead on the file, consuming too much CPU and impacting
2280 performance (since 1.4.5.8). Improve the readahead algorithm
2281 to avoid this, and also improve some other common cases (read
2282 of small files in particular, where "small" is files smaller than
2283 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2288 Description: MDS crash when receiving packet with unknown intent.
2289 Details : Do not LBUG in unknown intent case, just return -EFAULT
2291 Severity : enhancement
2292 Bugzilla : 9293, 9385
2293 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2294 Details : Do not serialize getattr (non-intent version) and statfs.
2297 Frequency : occasional, when OST network is overloaded/intermittent
2299 Description: client evicted by OST after bulk IO timeout
2300 Details : If a client sends a bulk IO request (read or write) the OST
2301 may evict the client if it is unresposive to its data GET/PUT
2302 request. This is incorrect if the network is overloaded (takes
2303 too long to transfer the RPC data) or dropped the OST GET/PUT
2304 request. There is no need to evict the client at all, since
2305 the pinger and/or lock callbacks will handle this, and the
2306 client can restart the bulk request.
2309 Frequency : Always when mmapping file with no objects
2311 Description: client crashes when mmapping file with no objects
2312 Details : Check that we actually have objects in a file before doing any
2313 operations on objects in ll_vm_open, ll_vm_close and
2319 Description: Request leak when working with deleted CWD
2320 Details : Introduce advanced request refcount tracking for requests
2321 referenced from lustre intent.
2323 Severity : Enhancement
2325 Description: Cache open file handles on client.
2326 Details : MDS now will return special lock along with openhandle, if
2327 requested and client is allowed to hold openhandle, even if unused,
2328 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2329 closing files for every read/write openration.
2331 Severity : Enhancement
2333 Description: Cache open negative dentries on client when possible.
2334 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2335 negative dentries on lock revocation.
2340 Description: Remounting a client read-only wasn't possible with a zconf mount
2341 Details : It wasn't possible to remount a client read-only with llmount.
2343 Severity : enhancement
2344 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2345 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2346 MPICH is included to add Lustre-specific ADIO interfaces.
2347 This is based closely on the UFS ADIO layer and only differs
2348 in file creation, in order to allow the OST striping to be set.
2349 This is user-contributed code and not supported by CFS.
2354 Description: extended inode attributes (immutable, append-only) work improperly
2355 when 2.4 and 2.6 kernels are used on client/server or vice versa
2356 Details : Introduce kernel-independent values for these flags.
2358 Severity : enhancement
2361 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2362 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2363 granularity. Now, Fractional MB granularity is be supported,
2364 this is very useful for low memory system.
2366 Severity : enhancement
2368 Description: Getattr by fid
2369 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2370 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2371 have locks and avoids getattr rpc requests for it on every stat.
2374 Frequency : Always, for filesystems larger than 2TB
2376 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2377 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2378 filesystems larger than 2TB to be mounted with mballoc enabled.
2381 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2383 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2384 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2385 filesystems larger than 2TB is mounted then the kernel will
2386 silently corrupt the start of the filesystem. CONFIG_LBD is
2387 enabled for all CFS-supported kernels, but the possibility of
2388 this happening with a modified kernel config exists.
2390 Severity : enhancement
2392 Description: add client O_DIRECT support for 2.6 kernels
2393 Details : It is now possible to do O_DIRECT reads and writes to files
2394 in the Lustre client mountpoint on 2.6 kernel clients.
2396 Severity : enhancement
2398 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2399 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2400 parallel, not waiting for response from every OST before sending
2401 a rpc to the next OST.
2406 Description: setattr vs write race when updating file timestamps
2407 Details : Client processes that update a file timestamp into the past
2408 right after writing to the file (e.g. tar) it is possible that
2409 the updated file modification time can be reset to the current
2410 time due to a race between processing the setattr and write RPC.
2412 Severity : enhancement
2414 Description: Bring 'lfs find' closer in line with regular Linux find.
2415 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2416 -print0 options and obtains all the needed info through the lustre
2419 Severity : enhancement
2421 Description: support up to 1024 configured devices on one node
2422 Details : change obd_dev array from statically allocated to dynamically
2423 allocated structs as they are first used to reduce memory usage
2428 Description: Flush dirty partially truncated pages during truncate
2429 Details : Immediatelly flush partially truncated pages in filter_setattr,
2430 this way we completely avoid having any pages in page cache on OST
2431 and can retire ugly workarounds during writes to flush such pages.
2436 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2437 Details : protect inode from truncation within vfs_unlink() context
2438 just take a reference before calling vfs_unlink() and release it
2439 when parent's i_sem is free.
2444 Description: last_id value checked outside lock on OST caused LASSERT failure
2445 Details : If there were multiple MDS->OST object precreate requests in
2446 flight, it was possible that the OST's last object id was checked
2447 outside a lock and incorrectly tripped an assertion. Move checks
2448 inside locks, and discard old precreate requests.
2451 Frequency : always, if extents are used on OSTs
2453 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2454 Details : OSTs using the extents format would not zero the high 16 bits of
2455 the index physical block number. This is not a problem for any
2456 OST filesystems smaller than 16TB, and no kernels support ext3
2457 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2458 new/modified files) and can be fixed for existing filesystems
2459 with e2fsprogs-1.39-cfs1.
2464 Description: import connection selection may be incorrect if timer wraps
2465 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2466 import connections to be ignored if the 32-bit jiffies counter
2467 wraps. Use a 64-bit jiffies counter.
2470 Frequency : very large clusters immediately after boot
2472 Description: LNET request buffers exhausted under heavy short-term load
2473 Details : If a large number of client requests are generated on a service
2474 that has previously never seen so many requests it is possible
2475 that the request buffer growth cannot keep up with the spike in
2476 demand. Instead of dropping incoming requests, they are held in
2477 the LND until the RPC service can accept more requests.
2480 Frequency : Sometimes during replay
2482 Description: Assertion failure in ll_local_open after replay.
2483 Details : If replay happened on an open request reply before we were able
2484 to set replay handler, reply will become not swabbed tripping the
2485 assertion in ll_local_open. Now we set the handler right after
2486 recognising of open request
2489 Frequency : very rare
2491 Description: kernel reports "badness in vsnprintf"
2492 Details : Reading from the "recovery_status" /proc file in small chunks
2493 may cause a negative length in lprocfs_obd_rd_recovery_status()
2494 call to vsnprintf() (which is otherwise harmless). Exit early
2495 if there is no more space in the output buffer.
2497 Severity : enhancement
2499 Description: clear OBD RPC statistics by writing to them
2500 Details : It is now possible to clear the OBD RPC statistics by writing
2501 to the "stats" file.
2506 Description: Client mtime is not the same on different clients after utimes
2507 Details : In some cases, the client was using the utimes() syscall on
2508 a file cached on another node. The clients now validate the
2509 ctime from the MDS + OSTs to determine which one is right.
2514 Description: Inability to activate failout mode
2515 Details : lconf script incorrectly assumed that in pythong string's numeric
2516 value is used in comparisons.
2519 Frequency : always with multiple stripes per file
2521 Description: Inefficient object allocation for mutli-stripe files
2522 Details : When selecting which OSTs to stripe files over, for files with
2523 a stripe count that divides evenly into the number of OSTs,
2524 the MDS is always picking the same starting OST for each file.
2525 Return the OST selection heuristic to the original design.
2530 Description: mount failures may take full timeout to return an error
2531 Details : Under some heavy load conditions it is possible that a
2532 failed mount can wait for the full obd_timeout interval,
2533 possibly several minutes, before reporting an error.
2534 Instead return an error as soon as the status is known.
2536 Frequency : quota enabled and large files being deleted
2538 Description: releasing more than 4GB of quota at once hangs OST
2539 Details : If a user deletes more than 4GB of files on a single OST it
2540 will cause the OST to spin in an infinite loop. Release
2541 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2546 Description: statfs data retrieved from /proc may be stale or zero
2547 Details : When reading per-device statfs data from /proc, in the
2548 {kbytes,files}_{total,free,avail} files, it may appear
2549 as zero or be out of date.
2552 Frequency : systems with MD RAID1 external journal devices
2554 Description: lconf's call to blkid is confused by RAID1 journal devices
2555 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2556 of returning all block devices that match the journal UUID.
2559 Frequency : always, for aggregate stripe size over 4GB
2561 Description: assertion fails when trying to use 4GB stripe size
2562 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2563 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2566 Frequency : always on ppc64
2568 Description: the first write on an ext3 filesystem with mballoc got stuck
2569 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2570 perform endianness conversion.
2572 ------------------------------------------------------------------------------
2574 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2576 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2577 INTEROPERATE with older versions automatically. Please read the
2578 user documentation before upgrading any part of a live system.
2579 * WARNING: Lustre networking configuration changes are required with
2580 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2583 * Support for kernels:
2584 2.6.9-22.0.2.EL (RHEL 4)
2585 2.6.5-7.244 (SLES 9)
2586 2.6.12.6 vanilla (kernel.org)
2589 Severity : enhancement
2590 Bugzilla : 7981/8208
2591 Description: Introduced Lustre Networking (LNET)
2592 Details : LNET is new networking infrastructure for Lustre, it includes
2593 a reorganized network configuration mode (see the user
2594 documentation for full details) as well as support for routing
2595 between different network fabrics. Lustre Networking Devices
2596 (LNDs) for the supported network fabrics have also been
2597 created for this new infrastructure.
2599 Severity : enhancement
2600 Description: Introduced Access control lists
2601 Details : clients can set ACLs on files and directories in order to have
2602 more fine-grained permissions than the standard Unix UGO+RWX.
2603 The MDS must be started with the "-o acl" mount option.
2605 Severity : enhancement
2606 Description: Introduced filesystem quotas
2607 Details : Administrators may now establish per-user quotas on the
2610 Severity : enhancement
2612 Description: Configuration change for the XT3
2613 The PTLLND is now used to run Lustre over Portals on the XT3
2614 The configure option(s) --with-cray-portals are no longer used.
2615 Rather --with-portals=<path-to-portals-includes> is used to
2616 enable building on the XT3. In addition to enable XT3 specific
2617 features the option --enable-cray-xt3 must be used.
2622 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2623 Details : A race between allocating a new llcd and re-getting the llcd_lock
2624 allowed another thread to grab newly-allocated llcd.
2626 Severity : enhancement
2628 Description: 2.6 OST async journal commit and locking fix to improve performance
2629 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2630 2.6 kernels are now async as they already were in 2.4 kernels so
2631 that they can commit concurrently with the network bulk transfer.
2632 For block-allocated files the filter allocation semaphore is held
2633 to avoid filesystem fragmentation during allocation. BKL lock
2634 removed for 2.6 xattr operations where it is no longer needed.
2639 Description: lconf incorrectly determined whether two IP networks could talk
2640 Details : In some more complicated routing and multiple-network
2641 configurations, lconf will avoid trying to make a network
2642 connection to a disjoint part of the IP space. It was doing the
2643 math incorrectly for one set of cases.
2648 Description: Fix for potential infinite loop processing records in an llog.
2649 Details : If an llog record is corrupted/zeroed, it is possible to loop
2650 forever in llog_process(). Validate the llog record length
2651 and skip the remainder of the block on error.
2654 Frequency : occasional (liblustre only)
2656 Description: liblustre could not open files whose last component is a symlink
2657 Details : sysio_path_walk() would incorrectly pass the open intent to
2658 intermediate path components.
2661 Frequency : rare (liblustre only with non-standard tuning)
2662 Bugzilla : 7201 (7350)
2663 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2664 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2665 after it has been released, only do this for referenced locks
2667 Severity : enhancement
2669 Description: specifying an (invalid) directory default stripe_size of -1
2670 would reset the directory default striping
2671 Details : stripe_size -1 was used internally to signal directory stripe
2672 removal, now use "all default" to signal dir stripe removal
2673 as a directory striping of "all default" is not useful
2676 Frequency : common for large clusters running liblustre clients
2678 Description: doing an ls when liblustre clients are running is slow
2679 Details : sending a glimpse AST to a liblustre client waits for every AST
2680 to time out, as liblustre clients will not respond. Since they
2681 cannot cache data we refresh the OST lock LVB from disk instead.
2683 Severity : enhancement
2685 Description: doing an ls at the same time as file IO can be slow
2686 Details : enqueue and other "small" requests can be blocked behind many
2687 large IO requests. Create a new OST IO portal for non-IO
2688 requests so they can be processed faster.
2691 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2693 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2694 created file with mode 000
2695 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2696 allow an ftruncate() as a non-root user to a file with mode 000.
2697 Lustre now respects this flag to disable mode checks when
2698 truncating a file owned by the user
2701 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2704 Description: Revoking locks from clients that went dead or catatonic might take
2706 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2707 cancellation of such locks instant on servers without waiting for
2708 any reply from clients. Clients drops these locks when cancel
2709 notification from server is received without replying.
2712 Frequency : liblustre-only, when liblustre client dies or becomes busy
2714 Description: Doing ls on Linux clients can take a long time with active
2716 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2717 granting such locks to it in the first place if possible. Locks
2718 are taken by proxy on the OST during the read or write and
2719 dropped immediately afterward. Add connect flags handling, do
2720 not grant locks to liblustre clients for glimpse ASTs.
2722 Severity : enhancement
2724 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2725 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2726 sequential reads and behaves badly with applications doing random
2727 reads. Improve it by reading ahead at least read region, and
2728 avoiding excessive large RPC for small reads.
2730 Severity : enhancement
2732 Description: Creating more than 1000 files for a single job may cause a load
2733 imbalance on the OSTs if there are also a large number of OSTs.
2734 Details : qos_prep_create() uses an OST index reseed value that is an
2735 even multiple of the number of available OSTs so that if the
2736 reseed happens in the middle of the object allocation it will
2737 still utilize the OSTs as uniformly as possible.
2742 Description: OST or MDS may oops in ping_evictor_main()
2743 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2744 but doesn't restart at beginning of obd_exports_timed list
2747 Severity : enhancement
2749 Description: improve by-nid export eviction on the MDS and OST
2750 Details : allow multiple exports with the same NID to be evicted at one
2751 time without re-searching the exports list.
2754 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2756 Description: MDS may oops in groups_free()
2757 Details : in rare race conditions a newly allocated group_info struct is
2758 freed again, and this can be NULL. The 2.4 compatibility code
2759 for groups_free() checked for a NULL pointer, but 2.6 did not.
2762 Frequency : common for liblustre clients doing little filesystem IO
2763 Bugzilla : 9352, 7313
2764 Description: server may evict liblustre clients accessing contended locks
2765 Details : if a client is granted a lock or receives a completion AST
2766 with a blocking AST already set it would not reply to the AST
2767 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2770 Frequency : lfs setstripe, only systems with more than 160 OSTs
2772 Description: unable to set striping with a starting offset beyond OST 160
2773 Details : llapi_create_file() incorrectly limited the starting stripe
2774 index to the maximum single-file stripe count.
2777 Frequency : LDAP users only
2779 Description: lconf did not handle in-kernel recovery with LDAP properly
2780 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2782 Severity : enhancement
2784 Description: bind OST threads to NUMA nodes to improve performance
2785 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2786 node and do their allocations there to localize memory access
2788 Severity : enhancement
2790 Description: llmount can determine client NID directly from Myrinet (GM)
2791 Details : the client NID code from gmnalnid was moved directly into
2792 llmount, removing the need to use this or specifying the
2793 client NID explicitly when mounting GM clients with zeroconf
2796 Frequency : if client is started with down MDS
2798 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2799 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2800 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2801 ptlrpc_import_delayed_req() because we didn't check if the
2802 request was interrupted, and we also didn't break out of the
2803 event loop if there was no timeout
2808 Description: data loss during non-page-aligned writes to a single file from
2809 both multiple nodes and multiple threads on one node at same time
2810 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2811 inconsistency led to false short-reads, that were cached and later
2812 used by ->prepare_write() to fill in partially written page,
2813 leading to data loss.
2816 Frequency : always, if lconf --abort_recovery used
2818 Description: lconf --abort_recovery fails with 'Operation not supported'
2819 Details : lconf was attempting to abort recovery on the MDT device and not
2822 Severity : enhancement
2824 Description: remove cleanup logs
2825 Details : replace lconf-generated cleanup logs with lustre internal
2826 cleanup routines. Eliminates the need for client-cleanup and
2829 Severity : enhancement
2831 Description: add support for EAs (user and system) on lustre filesystems
2832 Details : it is now possible to store extended attributes in the Lustre
2833 client filesystem, and with the user_xattr mount option it
2834 is possible to allow users to store EAs on their files also
2836 Severity : enhancement
2838 Description: Add possibility (config option) to show minimal available OST free
2840 Details : When compiled with --enable-mindf configure option, statfs(2)
2841 (and so, df) will return least minimal free space available from
2842 all OSTs as amount of free space on FS, instead of summary of
2843 free spaces of all OSTs.
2845 Severity : enhancement
2847 Description: do not expand extent locks acquired on OST-side
2848 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2849 by server: they are not cached anyway.
2852 Frequency : when mmap is used/binaries executed from Lustre
2854 Description: Unmmap pages before throwing them away from read cache.
2855 Details : llap_shrink cache now attempts to unmap pages before discarding
2856 them (if unmapping failed - do not discard). SLES9 kernel has
2857 extra checks that trigger if this unmapping is not done first.
2862 Description: lconf didn't resolve symlinks before checking to see whether a
2863 given mountpoint was already in use
2866 Frequency : when migrating failover services
2867 Bugzilla : 6395, 9514
2868 Description: When migrating a subset of services from a node (e.g. failback
2869 from a failover service node) the remaining services would
2870 time out and evict clients.
2871 Details : lconf --force (implied by --failover) sets the global obd_timeout
2872 to 5 seconds in order to quickly disconnect, but this caused
2873 other RPCs to time out too quickly. Do not change the global
2874 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2876 Severity : enhancement
2877 Frequency : if MDS is started with down OST
2878 Bugzilla : 9439,5706
2879 Description: Allow startup/shutdown of an MDS without depending on the
2880 availability of the OSTs.
2881 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2882 Add appropriate locking and lov-osc refcounts for safe
2883 cleaning. Add osc abort_inflight calls in case the
2884 synchronize never started.
2887 Frequency : occasional (Cray XT3 only)
2889 Description: root not authorized to access files in CRAY_PORTALS environment
2890 Details : The client process capabilities were not honoured on the MDS in
2891 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2892 been accessed by an authorized user then root was able to access
2893 the file on the local client also. The root user capabilities
2894 are now allowed on the MDS, as this environment has secure UID.
2897 Frequency : occasional
2899 Description: ldiskfs "too long searching" message happens too often
2900 Details : A debugging message (otherwise harmless) prints too often on
2901 the OST console. This has been reduced to only happen when
2902 there are fragmentation problems on the filesystem.
2907 Description: Division by zero in statfs when all OSCs are inactive
2908 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2909 lov_statfs divides by value returned by lov_get_stripecnt().
2913 Bugzilla : 9489, 3273
2914 Description: First write from each client to each OST was only 4kB in size,
2915 to initialize client writeback cache, which caused sub-optimal
2916 RPCs and poor layout on disk for the first writen file.
2917 Details : Clients now request an initial cache grant at (re)connect time
2918 and so that they can start streaming writes to the cache right
2919 away and always do full-sized RPCs if there is enough data.
2920 If the OST is rebooted the client also re-establishes its grant
2921 so that client cached writes will be honoured under the grant.
2926 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2927 Details : Now I/O RPCs go to different portal number and (presumably) fast
2928 lock requests (and glimses) and other RPCs get their own service
2929 threads pool that should be able to service those RPCs
2932 Severity : enhancement
2934 Description: Ability to exchange lustre version between client and servers and
2935 issue warnings at client side if client is too old. Also for
2936 liblustre clients there is ability to refuse connection of too old
2938 Details : New 'version' field is added to connect data structure that is
2939 filled with version info. That info is later checked by server and
2943 Frequency : rare, liblustre only.
2944 Bugzilla : 9296, 9581
2945 Description: Two simultaneous writes from liblustre at offset within same page
2946 might proceed at the same time overwriting eachother with stale
2948 Details : I/O lock withing llu_file_prwv was released too early, before data
2949 actually was hitting the wire. Extended lock-holding time until
2950 server acknowledges receiving data.
2953 Frequency : extremely rare. Never observed in practice.
2955 Description: avoid generating lustre_handle cookie of 0.
2956 Details : class_handle_hash() generates handle cookies by incrementing
2957 global counter, and can hit 0 occasionaly (this is unlikely, but
2958 not impossible, because initial value of cookie counter is
2959 selected randonly). Value of 0 is used as a sentinel meaning
2960 "unassigned handle" --- avoid it. Also coalesce two critical
2961 sections in this function into one.
2963 Severity : enhancement
2965 Description: allow liblustre clients to delegate truncate locking to OST
2966 Details : To avoid overhead of locking, liblustre client instructs OST to
2967 take extent lock in ost_punch() on client's behalf. New connection
2968 flag is added to handle backward compatibility.
2970 Severity : enhancement
2971 Bugzilla : 4928, 7341, 9758
2972 Description: allow number of OST service threads to be specified
2973 Details : a module parameter allows the number of OST service threads
2974 to be specified via "options ost ost_num_threads={N}" in the
2975 OSS's /etc/modules.conf or /etc/modprobe.conf.
2979 Bugzilla : 6146, 9635, 9895
2980 Description: servers crash with bad pointer in target_handle_connect()
2981 Details : In rare cases when a client is reconnecting it was possible that
2982 the connection request was the last reference for that export.
2983 We would temporarily drop the export reference and get a new
2984 one, but this may have been the last reference and the export
2985 was just destroyed. Get new reference before dropping old one.
2987 Severity : enhancement
2988 Frequency : if client is started with failover MDS
2990 Description: Allow multiple MDS hostnames in the mount command
2991 Details : Try to read the configuration from all specified MDS
2992 hostnames during a client mount in case the "primary"
2995 Severity : enhancement
2997 Description: Stop sending data to evicted clients as soon as possible.
2998 Details : Check if the client we are about to send or are sending data to
2999 was evicted already. (Check is done every second of waiting,
3000 for which l_wait_event interface was extended to allow checking
3001 of exit condition at specified intervals).
3004 Frequency : rare, normally only when NFS exporting is done from client
3006 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3008 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3011 Frequency : always, if srand() is called before liblustre initialization
3013 Description: Liblustre uses system PRNG disturbing its usage by user application
3014 Details : Introduce internal to lustre fast and high-quality PRNG for
3015 lustre usage and make liblustre and some other places in generic
3016 lustre code to use it.
3018 Severity : enhancement
3019 Bugzilla : 9477, 9557, 9870
3020 Description: Verify that the MDS configuration logs are updated when xml is
3021 Details : Check if the .xml configuration logs are newer than the config
3022 logs stored on the MDS and report an error if this is the case.
3023 Request --write-conf, or allow starting with --old_conf.
3025 Severity : enhancement
3027 Description: Handle symlinks in the path when checking if Lustre is mounted.
3028 Details : Resolve intermediate symlinks when checking if a client has
3029 mounted a filesystem to avoid duplicate client mounts.
3034 Description: lconf can hit an error exception but still return success.
3035 Details : The lconf command catches the Command error exception at the top
3036 level script context and will exit with the associated exit
3037 status, but doesn't ensure that this exit status is non-zero.
3042 Description: failure of ptlrpc thread startup can cause oops
3043 Details : Starting a ptlrpc service thread can fail if there are a large
3044 number of threads or the server memory is very fragmented.
3045 Handle this without oopsing.
3048 Frequency : always, only if liblustre and non-default acceptor port was used
3050 Description: liblustre cannot connect to servers with non-default acceptor port
3051 Details : tcpnal_set_default_params() was not called and was therefore
3052 ignoring the environment varaible TCPNAL_PORT, as well as other
3053 TCPNAL_ environment variables
3058 Description: two objects could be created on the same OST for a single file
3059 Details : If an OST is down, in some cases it was possible to create two
3060 objects on a single OST for a single file. No problems other
3061 than potential performance impact and spurious error messages.
3065 Bugzilla : 5681, 9562
3066 Description: Client may oops in ll_unhash_aliases
3067 Details : Client dcache may become inconsistent in race condition.
3068 In some cases "getcwd" can fail if the current directory is
3074 Description: Inode refcounting problems in NFS export code
3075 Details : link_raw functions used to call d_instantiate without obtaining
3076 extra inode reference first.
3080 Bugzilla : 9942, 9903
3081 Description: Referencing freed requests leading to crash, memleaks with NFS.
3082 Details : We used to require that call to ll_revalidate_it was always
3083 followed by ll_lookup_it. Also with revalidate_special() it is
3084 possible to call ll_revalidate_it() twice for the same dentry
3085 even if first occurence returned success. This fix changes semantic
3086 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3087 reference on a request referred from the intent.
3088 ll_intent_release() then releases such a request.
3091 Frequency : rare, normally benchmark loads only
3093 Description: unlinked inodes were kept in memory on the client
3094 Details : If a client is repeatedly creating and unlinking files it
3095 can accumulate a lot of stale inodes in the inode slab cache.
3096 If there is no other client load running this can cause the
3097 client node to run out of memory. Instead flush old inodes
3098 from client cache that have the same inode number as a new inode.
3101 Frequency : SLES9 2.6.5 kernel and long filenames only
3102 Bugzilla : 9969, 10379
3103 Description: utime reports stale NFS file handle
3104 Details : SLES9 uses out-of-dentry names in some cases, which confused
3105 the lustre dentry revalidation. Change it to always use the
3109 Frequency : rare, unless heavy write-truncate concurrency is continuous
3110 Bugzilla : 4180, 6984, 7171, 9963, 9331
3111 Description: OST becomes very slow and/or deadlocked during object unlink
3112 Details : filter_destroy() was holding onto the parent directory lock
3113 while truncating+unlinking objects. For very large objects this
3114 may block other threads for a long time and slow overall OST
3115 responsiveness. It may also be possible to get a lock ordering
3116 deadlock in this case, or run out of journal credits because of
3117 the combined truncate+unlink. Solution is to do object truncate
3118 first in one transaction without parent lock, and then do the
3119 final unlink in a new transaction with the parent lock. This
3120 reduces the lock hold time dramatically.
3123 Frequency : rare, 2.4 kernels only
3125 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3126 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3127 before unmounting their filesystem in order to avoid deadlock.
3128 The kernel_locked() function in 2.4 kernels only checks whether
3129 the kernel lock is held, not whether it is this process that is
3130 holding it as 2.6 kernels do.
3135 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3136 Details : The client ptlrpc code may be trying to reconnect to a down
3137 server before a previous connection attempt has timed out.
3138 Increase the reconnect interval to be longer than the connection
3139 timeout interval to avoid sending duplicate connections to
3143 Frequency : echo_client brw_test command
3145 Description: fix echo_client to work with OST preallocated code
3146 Details : OST preallocation code (5137) didn't take echo_client IO path
3147 into account: echo_client calls filter methods outside of any
3148 OST thread and, hence, there is no per-thread preallocated
3149 pages and buffers to use. Solution: hijack pga pages for IO. As
3150 a byproduct, this avoids unnecessary data copying.
3154 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3155 Description: Client can oops in mdc_commit_close() after open replay
3156 Details : It was possible for the MDS to return an open request with no
3157 transaction number in mds_finish_transno() if the client was
3158 evicted, but without actually returning an error. Clients
3159 would later try to replay that open and may trip an assertion
3160 Simplify the client close codepath, and always return an error
3161 from the MDS in case the open is not successful.
3164 Frequency : rare, 2.6 OSTs only
3166 Description: OST may deadlock under high load on fragmented files
3167 Details : If there was a heavy load and highly-fragmented OST filesystems
3168 it was possible to have all the OST threads deadlock waiting on
3169 allocation of biovecs, because the biovecs were not released
3170 until the entire RPC IO was completed. Instead, release biovecs
3171 as soon as they are complete to ensure forward IO progress.
3173 Severity : enhancement
3175 Description: Support for specifying external journal device at mount
3176 Details : If an OST or MDS device is formatted with an external journal
3177 device, this device major/minor is stored in the ext3 superblock
3178 and may not be valid for failover. Allow detecting and
3179 specifying the external journal at mount time.
3184 Description: Mounting an MDS with pending unlinked files may cause oops
3185 Details : target_finish_recovery() calls mds_postrecov() which returned
3186 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3187 considers this an error and immediately begins cleaning up the
3188 lov, just after starting the mds_lov process
3190 Severity : enhancement
3192 Description: Implement 'lfs df' to report actual free space on per-OST basis
3193 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3194 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3195 usage of objects; '-h' to report in human readable format.
3197 ------------------------------------------------------------------------------
3199 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3206 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3207 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3208 at startup during orphan unlinking. Instead, skip buddy update
3209 before setup, as it will be regenerated after recovery is complete.
3214 Description: If an OST is inactive, its locks might reference stale inodes.
3215 Details : lov_change_cbdata() must iterate over all namespaces, even if
3216 they are inactive to clear inode references from the lock.
3218 Severity : enhancement
3219 Frequency : occasional, if non-standard max_dirty_mb used
3221 Description: Client will block write RPCs if not enough grant
3222 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3223 then the client will block writes while waiting for another RPC
3224 to complete instead of consuming its dirty limit. With change
3225 we get improved performance when max_dirty_mb is small.
3227 Severity : enhancement
3228 Bugzilla : 3389, 6253
3229 Description: Add support for supplementary groups on the MDS.
3230 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3231 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3232 lookups for user supplementary groups into a cache.
3236 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3237 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3238 for liblustre O_EXCL is enforced by MDS.
3242 Description: readdir never returns NULL in liblustre.
3243 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3244 dentry in struct dirent.
3248 Frequency : liblustre only, depends on application IO pattern
3249 Description: liblustre clients evicted if not contacting servers
3250 Details : Don't put liblustre clients into the ping_evictor list, so
3251 they will not be evicted by the pinger ever.
3253 Severity : enhancement
3255 Description: Add ability to evict clients by NID from MDS.
3256 Details : By echoing "nid:$NID" string into
3257 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3258 $NID would be instantly evicted from this MDS and from all active
3259 OSTs connected to it.
3263 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3264 Details : lookup_it_finish() used to query file size from OSTs that was not
3269 Description: service threads change working directory to that of init
3270 Details : Starting lustre service threads may pin the working directory
3271 of the parent thread, making that filesystem busy. Threads
3272 now change to the working directory of init to avoid this.
3276 Frequency : during shutdown only
3277 Description: shutdown with a failed MDS or OST can cause unmount to hang
3278 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3283 Frequency : 2.6 only
3284 Description: chmod/chown may include an extra supplementary group
3285 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3286 supplementary group and if none is specified this is used.
3289 Bugzilla : 5479 (6816)
3291 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3292 Details : If lookup is in progress on a file that is unlinked we might try
3293 to revalidate the inode and fail in revalidate after lookup is
3294 complete and ll_file_open() enqueues the open again but
3295 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3298 Frequency : always, if lconf --abort_recovery used
3300 Description: lconf --abort_recovery fails with 'Operation not supported'
3301 Details : lconf was attempting to abort recovery on the MDT device and not
3304 ------------------------------------------------------------------------------
3306 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3311 Frequency : rare (only unsupported configurations with a node running as an
3313 Bugzilla : 6514, 5137
3314 Description: Mounting a Lustre file system on a node running as an OST could
3316 Details : OSTs now preallocates memory needed to write out data at
3317 startup, instead of when needed, to avoid having to
3318 allocate memory in possibly low memory situations.
3319 Specifically, if the file system is mounted on on OST,
3320 memory pressure could force it to try to write out data,
3321 which it needed to allocate memory to do. Due to the low
3322 memory, it would be unable to do so and the node would
3323 become unresponsive.
3325 Severity : enhancement
3327 Description: Addition of lconf --service command line option
3328 Details : lconf now accepts a '--service <arg>' option, which is
3329 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3331 Severity : enhancement
3333 Description: Failover mode is now the default for OSTs.
3334 Details : By default, OSTs will now run in failover mode. To return to
3335 the old behaviour, add '--failout' to the lmc line for OSTs.
3337 Severity : enhancement
3339 Description: Health checks are now provided for MDS and OSTs
3340 Details : Additional detailed health check information on MSD and OSTs
3341 is now provided through the procfs health_check value.
3344 Frequency : occasional, depends on IO load
3346 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3347 after numerous create/delete cycles
3348 Details : The ext3 inode allocation policy would not allocate new inodes
3349 very well on the OSTs because there are no new directories
3350 being created. Instead we look for groups with free space if
3351 the parent directories are nearly full.
3356 Description: Network or server problems during mount may cause partially
3357 mounted clients instead of returning an error.
3358 Details : The config llog parsing code may overwrite the error return
3359 code during mount error handling, returning success instead
3365 Description: MDS can fail to allocate large reply buffers
3366 Details : After long uptimes the MDS can fail to allocate large reply
3367 buffers (e.g. zconf client mount config records) due to memory
3368 fragmentation or consumption by the buffer cache. Preallocate
3369 some large reply buffers so that these replies can be sent even
3370 under memory pressure.
3374 Frequency : rare (liblustre)
3375 Description: fsx running with liblustre complained that using truncate() to
3376 extend the file doesn't work. This patch corrects that issue.
3377 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3378 ATTR_SIZE and lsm use in llu_setattr_raw.
3382 Frequency : rare, only 2.6 kernels
3383 Description: Unusual file access patterns on the MDS may result in inode
3384 data being lost in very rare circumstances.
3385 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3386 code in the 2.6 kernel caused the inode and/or EA data not to
3387 be read from disk, causing single-file corruption.
3391 Frequency : rare, only 2.6 filesystems using extents
3392 Description: Heavy concurrent write and delete load may cause data corruption.
3393 Details : It was possible under high-load situations to have an extent
3394 metadata block in the block device cache from a just-unlinked
3395 file overwrite a newly-allocated data block. We now unmap any
3396 metadata buffers that alias just-allocated data blocks.
3400 Frequency : filesystems with default stripe_count larger than 77
3401 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3402 Details : lconf specifies an inode size of 4096 bytes when the default
3403 stripe_count is larger than 77. This conflicts with the default
3404 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3405 this case to avoid pinning too much memory for large EAs.
3407 ------------------------------------------------------------------------------
3409 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3414 Frequency : rare (extremely heavy IO load with hundreds of clients)
3416 Description: Client is evicted, gets IO error writing to file
3417 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3418 caused clients to be evicted for AST timeouts. The fixes in
3419 bug 5192 mean we no longer need to have such short AST timeouts
3420 so ldlm_timeout has been increased.
3423 Frequency : occasional during --force or --failover shutdown under load
3424 Bugzilla : 5949, 4834
3425 Description: Server oops/LBUG if stopped with --force or --failover under load
3426 Details : a collection of import/export refcount and cleanup ordering
3427 issues fixed for safer force cleanup
3430 Frequency : only filesystems larger than 120 OSTs
3431 Bugzilla : 5990, 6223
3432 Description: lfs getstripe would oops on a very large filesystem
3433 Details : lov_getconfig used kfree on vmalloc'd memory
3436 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3437 Bugzilla : 6242, 6243
3438 Description: reading from files that had been truncated to a non-zero size
3439 but never opened returned no data
3440 Details : ll_file_read() reads zeros from no-object files to EOF
3445 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3446 Details : There's little harm in aborting MDS/OSS recovery and letting it
3447 try again, so I removed the LASSERT and return an error instead.
3449 Severity : enhancement
3451 Description: New debugging infrastructure for tracking down data corruption
3452 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3453 (b) cover more of the client-side code path, and (c) try to narrow
3454 down where problems occurred
3458 Bugzilla : 3819, 4364, 4397, 6313
3459 Description: Racing close and eviction MDS could cause assertion in mds_close
3460 Details : It was possible to get multiple mfd references during close and
3461 client eviction, leading to one thread referencing a freed mfd.
3463 Severity: : enhancement
3464 Bugzilla : 3262, 6359
3465 Description: Attempts to reconnect to servers are now more aggressive.
3466 Details : This builds on the enhanced upcall-less recovery that was added
3467 in 1.4.2. When trying to reconnect to servers, clients will
3468 now try each server in the failover group every 10 seconds. By
3469 default, clients would previously try one server every 25 seconds.
3474 Description: After recovery, certain operations trigger a failed
3475 assertion on a client.
3476 Details : Failing over an mds, using lconf -d --failover, while a
3477 client was doing a readdir() call would cause the client to
3478 LBUG after recovery completed and the readdir() was resent.
3480 Severity : enhancement
3482 Description: Default groups are now added by lconf
3483 Details : You can now run lconf --group <servicename> without having to
3484 manually add groups with lmc.
3487 Frequency : occasional
3489 Description: Nodes with an elan id of 0 trigger a failed assertion
3492 Frequency : always when accessing e.g. tty/console device nodes
3494 Description: tty and some other devices nodes cannot be used on lustre
3495 Details : file's private_data field is used by device data and lustre
3496 values in there got lost. New field was added to struct file to
3497 store fs-specific private data.
3500 Frequency : when exporting Lustre via NFS
3502 Description: NFSD failed occasionally when looking up a path component
3503 Details : NFSD is looking up ".." which was broken in ext3 directories
3504 that had grown large enough to become hashed.
3507 Frequency : Clusters with multiple interfaces not on the same subnet
3509 Description: Nodes will repeatedly try to reconnect to an interface which it
3510 cannot reach and report an error to the log.
3511 Details : Extra peer list entries will be created by lconf with some peers
3512 unreachable. lconf now validates the peer before adding it.
3515 Frequency : Only if a default stripe is set on the filesystem root.
3517 Description: Setting a default stripe on the filesystem root prevented the
3518 filesystem from being remounted.
3519 Details : The client was sending extra request flags in the root getattr
3520 request and did not allocate a reply buffer for the dir EA.
3523 Frequency : occasional, higher if lots of files are accessed by one client
3524 Bugzilla : 6159, 6097
3525 Description: Client trips assertion regarding lsm mismatch/magic
3526 Details : While revalidating inodes the VFS looks up inodes with ifind()
3527 and in rare cases can find an inode that is being freed.
3528 The ll_test_inode() code will free the lsm during ifind()
3529 when it finds an existing inode and then the VFS later attaches
3530 this free lsm to a new inode.
3534 Bugzilla : 6422, 7030
3535 Description: MDS deadlock between mkdir and client eviction
3536 Details : Creating a new file via mkdir or mknod (starting a transaction
3537 and getting the ns lock) can deadlock with client eviction
3538 (gets ns lock and trying to finish a synchronous transaction).
3541 Frequency : occasional
3542 Description: While starting a server, the fsfilt_ext3 module could not be
3544 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3545 kernels. Previously, lconf would still use the ext3 name
3546 when trying to load modules. Now, it will correctly use
3547 ext3 on 2.4 and ldiskfs on 2.6.
3549 Severity : enhancement
3550 Description: The default stripe count has been changed to 1
3551 Details : The interpretation of the default stripe count (0, to lfs
3552 or lmc) has been changed to mean striping across a single
3553 OST, rather than all available. For general usage we have
3554 found a stripe count of 1 or 2 works best.
3556 Severity : enhancement
3557 Description: Add support for compiling against Cray portals.
3558 Details : Conditional compiling for some areas that are different
3562 Frequency : occasional
3563 Bugzilla : 6409, 6834
3564 Description: Creating files with an explicit stripe count may lead to
3565 a failed assertion on the MDS
3566 Details : If some OSTs are full or unavailable, creating files may
3567 trigger a failed assertion on the MDS. Now, Lustre will
3568 try to use other servers or return an error to the
3572 Frequency : occasional
3574 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3575 caused lock timeout during readahead (since 1.4.2).
3576 Details : Processes doing readahead might match a lock that hasn't been
3577 granted yet if there are overlapping and conflicting lock
3578 requests. The readahead process waits on ungranted lock
3579 (original lock is CBPENDING), while OST waits for that process
3580 to cancel CBPENDING read lock and eventually evicts client.
3582 Severity : enhancement
3584 Description: Initial enabling of flock support for clients
3585 Details : Implements fcntl advisory locking and file status functions.
3586 This feature is provided as an optional mount flag (default
3587 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3588 locking are implemented yet, and those that are are not guaranteed
3589 to be completely correct in production environments.
3590 mount -t lustre -o [flock|noflock] ...
3593 Frequency : occasional
3595 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3596 assertion in the ext3 JBD (journaling) layer.
3597 Details : The b_committed_data struct is protected by the big kernel lock
3598 in 2.4 kernels, serializing journal_commit_transaction() and
3599 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3600 access to this struct. In 2.6 kernels there is finer grained
3601 locking to improve SMP performance of the JBD layer.
3605 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3610 Description: merge and cleanup kernel patches.
3613 -----------------------------------------------------------------------------
3615 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3617 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3618 versions of Lustre. Please update all servers and clients to
3619 version 1.4.2 or later at the same time. You must also run
3620 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3621 to update the configuration logs.
3623 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3624 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3625 - swab llog records of type '0' so we get proper header size/idx (5861)
3626 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3627 - fix rename of one directory over another leaking an inode (5953)
3628 - avoid SetPageDirty on 2.6 (5981)
3629 - don't re-add just-being-destroyed locks to the waiting list (5653)
3630 - when creating new directories, inherit the parent's custom
3631 striping settings if present parent (3048)
3632 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3633 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3634 - don't hold client locks on temporary worklist from l_lru (5666)
3635 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3636 - automatically evict dead clients (5921)
3637 - Update file size properly in create+truncate+fstat case (6196)
3638 - Do not unhash mountpoint dentries, do not allow removal of
3640 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3641 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3642 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3643 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3644 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3645 - Correctly handle reads of files with no objects (6243)
3646 - lctl recover will also mark a device active if deactivate used (5933)
3648 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3649 - allow --write-conf on an MDS with different nettype than client (5619)
3650 - don't write config llogs to MDS for mounts not from that MDS (5617)
3651 - lconf should create multiple TCP connections from a client (5201)
3652 - init scripts are now turned off by default; run chkconfig --on
3653 lustre and chkconfig --on lustrefs to use them
3654 - upcalls are no longer needed for clients to recover to failover
3656 - add --abort-recovery option to lconf to abort recovery on device
3658 - add support for an arbitrary number of OSTs (3026)
3659 - Quota support protocol changes.
3660 - forward compatibility changes to wire structs (6007)
3661 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3662 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3663 - improved "lustre status" script
3664 - initialize blocksize for non-regular files (6062)
3665 - added --disable-server and --disable-client configure options (5782)
3666 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3667 - Vanilla 2.4.29 support
3668 - increase maximum number of obd devices to 520 (6242)
3669 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3670 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3671 SLES 9 2.6.5 kernels
3672 - limit stripes per file to 160 (the maximum EA size) (6093)
3674 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3677 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3678 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3679 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3680 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3681 - small ext3 extents cleanups and fixes (5733)
3682 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3683 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3684 - avoid needless client->OST connect, fix handle mismatch (5317)
3685 - fix DLM error path that led to out-of-sync client, long delays (5779)
3686 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3687 - fix several locking issues related to i_size (5492,5624,5654,5672)
3688 - don't move pending lock onto export if it is already evicted (5683)
3689 - fix kernel oops when creating .foo in unlinked directory (5548)
3690 - fix deadlock in obdfilter statistics vs. object create (5811)
3691 - use time_{before,after} to avoid timer jiffies wrap (5882)
3692 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3693 - Do not leak request if server was not able to process it (5154)
3694 - If mds_open unable to find parent dir, make that negative lookup(5154)
3695 - don't create new directories with extent-mapping (5909, 5936)
3697 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3698 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3699 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3700 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3701 - support for RHEL4 kernel on i686 (5773)
3702 - provide error messages when incompatible logs are encountered (5898)
3704 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3705 * version 1.4.0.10 (1.4.1 release candidate 1)
3707 - don't keep a lock reference when lock is not granted (4238)
3708 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3709 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3710 - drop import inflight refcount on signal_completed_replay error (5255)
3711 - unlock page after async write error during send (3677)
3712 - handle missing objects in filter_preprw_read properly (5265)
3713 - no transno return for symlink open, don't save no-trasno open (3440)
3714 - don't try to complete elan receive that already failed (4012)
3715 - free RPC server reply state on error (5406)
3716 - clean up thread from ptlrpc_start_thread() on error (5160)
3717 - readahead could read extra page into cache that wasn't ejected (5388)
3718 - prevent races in class_attach/setup/cleanup/detach (5260)
3719 - don't dereference de->d_inode after l_dput of de (5458)
3720 - use "int" for stripe value returned from lock_to_stripe (5544)
3721 - mballoc allocation and error-checking fixes in 2.6 (5504)
3722 - block device patches to fix I/O request sizes in 2.6 (5482)
3723 - look up hostnames for IB nals (5602)
3724 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3725 - don't start multiple acceptors for the same port (5277)
3726 - fix incorrect LASSERT in mds_getattr_name (5635)
3727 - export a proc file for general "ping" checking (5628)
3728 - fix "lfs check" to not block when the MDS is down (5628)
3730 - service request history (4965)
3731 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3732 - create an "evict_client" /proc entry on OSTs, like the MDS has
3733 - fix mount usage message, return errors per mount(8) (5168)
3734 - change grep [] to grep "[]" in tests so they work in more UMLs
3735 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3736 - remove ancient LOV_MAGIC_V0 EA support (5047)
3737 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3738 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3739 - allow readahead allocations to fail when low on memory (5383)
3740 - mmap locking landed again, after considerable improvement (2828)
3741 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3743 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3746 - send OST transaction number in read/write reply to free req (4966)
3747 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3748 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3749 - the watchdog thread now runs as interruptible (5246)
3750 - flock/lockf fixes (but it's still disabled, pending 5135)
3751 - don't use EXT3 constants in llite code (5094)
3752 - memory shortage at startup could cause assertion (5176)
3754 - reorganization of lov code
3755 - single portals codebase
3757 - add extents/mballoc support (5025)
3758 - direct I/O reads in the obdfilter (4048)
3759 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3761 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3764 - send OST transaction number in read/write reply to free req (4966)
3765 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3766 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3767 - the watchdog thread now runs as interruptible (5246)
3768 - handle missing objects in filter_preprw_read properly (5265)
3769 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3770 - drop import inflight refcount on signal_completed_replay error (5255)
3771 - unlock page after async write error during send (3677)
3772 - return original error code on reconstructed replies (3761)
3773 - no transno return for symlink open, don't save no-trasno open (3440)
3775 - add pid to ldlm debugging output (4922)
3776 - bump the watchdog timeouts -- we can't handle 30sec yet
3777 - extra debugging for orphan dentry/inode bug (5259)
3779 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3782 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3783 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3784 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3785 - fix NULL dereference in /proc/sys/portals/routes (4827)
3786 - allow failed mdc_close() operations to be interrupted (4561)
3787 - stop precreate on OST before MDS would time out on it (4778)
3788 - don't send partial-page writes before EOF from client (4410)
3789 - discard client grant for sub-page writes on large-page clients (4520)
3790 - don't free dentries not owned by NFS code, check generation (4806)
3791 - fix lsm leak if mds_create_objects() fails (4801)
3792 - limit debug_daemon file size, always print CERROR messages (4789)
3793 - use transno after validating reply (3892)
3794 - process timed out requests if import state changes (3754)
3795 - update mtime on OST during writes, return in glimpse (4829)
3796 - add mkfsoptions to LDAP (4679)
3797 - use ->max_readahead method instead of zapping global ra (5039)
3798 - don't interrupt __l_wait_event() during strace
3800 - add software watchdogs to catch hung threads quickly (4941)
3801 - make lustrefs init script start after nfs is mounted
3802 - fix CWARN/ERROR duplication (4930)
3803 - return async write errors to application if possible (2248)
3804 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3805 - print NAL number in %x format (4645)
3806 - update barely-supported suse-2.4.21-171 series (4842)
3807 - support for sles 9 %post scripts
3808 - support for building 2.6 kernel-source packages
3809 - support for sles km_* packages
3811 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3814 - ignore -ENOENT errors in osc_destroy (3639)
3815 - notify osc create thread that OSC is being cleaned up (4600)
3816 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3817 - reconstruct ost_handle() like mds_handle() (4657)
3818 - create a new thread to do import eviction to avoid deadlock (3969)
3819 - let lconf resolve symlinked-to devices (4629)
3820 - don't unlink "objects" from directory with default EA (4554)
3821 - hold socknal file ref over connect in case target is down (4394)
3822 - allow more than 32000 subdirectories in a single directory (3244)
3823 - fix blocks count for O_DIRECT writes (3751)
3824 - OST returns ENOSPC from object create when no space left (4539)
3825 - don't send truncate RPC if file size isn't changing (4410)
3826 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3827 - bind to privileged port in socknal and tcpnal (3689)
3829 - rate limit CERROR/CWARN console message to avoid overload (4519)
3830 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3831 - basic mmap support (3918)
3832 - kernel patch series update from b1_4 (4711)
3834 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3837 - avoid crash during MDS cleanup with OST shut down (2775)
3838 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3839 - don't use bad inodes on the MDS (3744)
3840 - dynamic object preallocation to improve recovery speed (4236)
3841 - don't hold spinlock over lock dumping or change debug flags (4401)
3842 - don't zero obd_dev when it is force cleaned (3651)
3843 - print grants to console if they go negative (4431)
3844 - "lctl deactivate" will stop automatic recovery attempts (3406)
3845 - look for existing locks in ldlm_handle_enqueue() (3764)
3846 - don't resolve lock handle twice in recovery avoiding race (4401)
3847 - revalidate should check working dir is a directory (4134)
3849 - don't always mark "slow" obdfilter messages as errors (4418)
3851 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3854 - don't close LustreDB during write_conf until it is done (3860)
3855 - fix typo in lconf for_each_profile (3821)
3856 - allow dumping logs from multiple threads at one time (3820)
3857 - don't allow multiple threads in OSC recovery (3812)
3858 - fix debug_size parameters (3864)
3859 - fix mds_postrecov to initialize import for llog ctxt (3121)
3860 - replace config semaphore with spinlock (3306)
3861 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3862 - don't allow enqueue to complete on a destroyed export (3822)
3863 - down write_lock before checking llog header bitmap (3825)
3864 - recover from lock replay timeout (3764)
3865 - up llog sem before sending rpc (3652)
3866 - reduce ns lock hold times when setting kms (3267)
3867 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3868 - fix NULL deref and obd_dev leak on setup error (3312)
3869 - replace some LBUG about llog ops with error handling (3841)
3870 - don't match INVALID dentries from d_lookup and spin (3784)
3871 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3872 - fix invalid assertion in ptlrpc_set_wait (3880)
3874 - add libwrap support for the TCP acceptor (3996)
3875 - add /proc/sys/portals/routes for non-root route listing (3994)
3876 - allow setting MDS UUID in .xml (2580)
3877 - print the stack of a process that LBUGs (4228)
3879 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3882 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3883 - make sure to unset replay flag from failed open requests (3440)
3884 - if default stripe count is 0, use OST count for inode size (3636)
3885 - update parent mtime/ctime on client for create/unlink (2611)
3886 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3887 - free recovery state on server during a forced cleanup (3571)
3888 - unregister_reply for resent reqs (3063)
3889 - loop back devices mounting and status check on 2.6 (3563)
3890 - fix resource-creation race that can provoke i_size == 0 (3513)
3891 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3892 - more debugging for page-accounting assertion (3746)
3893 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3894 - avoid deadlock after precreation failure (3758)
3895 - fix race and lock order deadlock in orphan handling (3450, 3750)
3896 - add validity checks when grabbing inodes from l_ast_data (3599)
3898 - add /proc/.../recovery_status to obdfilter (3428)
3899 - lightweight CDEBUG infrastructure, debug daemon (3668)
3900 - change default OSC RPC parameters to be better on small clusters
3901 - turn off OST read cache for files smaller than 32MB
3902 - install man pages and include them in rpms (3100)
3903 - add new init script for (un)mounting lustre filesystems (2593)
3904 - run chkconfig in %post for init scripts (3701)
3905 - drop scimac NAL (unmaintained)
3907 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3910 - clean kiobufs before and after use (3485)
3911 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3912 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3913 - change init script to fail more clearly if not run as root (1528)
3914 - allow clients to reconnect during replay (1742)
3915 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3916 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3917 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3918 - fix cond_resched (3554)
3919 - only evict unfinished clients after recovery (3515)
3920 - allow bulk resend, prevent data loss (3570)
3921 - dynamic ptlrpc request buffer allocation (2102)
3922 - don't allow unlinking open directory if it isn't empty (2904)
3923 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3924 - remove extraneous obd dereference causing LASSERT failure (3334)
3925 - don't use get_cycles() when creating temp. files on the mds (3156)
3926 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3927 - handle EEXIST for set-stripe, set proper directory name (3336)
3929 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3930 - fix ksocknal_fmb_callback() error messages (2918)
3932 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3935 - don't copy lvb into (possibly NULL) reply on error (2983)
3936 - don't deref dentry after dput, don't free lvb on error (2922)
3937 - use the kms to determine writeback rpc length (2947)
3938 - increment oti_logcookies when osc is inactive (2948)
3939 - update client's i_blocks count via lvb messages (2543)
3940 - handle intent open/close of special files properly (1557)
3941 - mount MDS with errors=remount-ro, like obdfilter (2009)
3942 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3943 - don't use cancelling-locks' kms values (2947)
3944 - use highest lock extent for kms, not last one (2925)
3945 - don't dereference ERR_PTR() dentry in error handling path (3107)
3946 - fix thread race in portals_debug_dumplog() (3122)
3947 - create lprocfs device entries at setup instead of at attach (1519)
3948 - common AST error handler, don't evict client on completion race (3145)
3949 - zero nameidata in detach_mnt in 2.6 (3118)
3950 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3951 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3952 - initialize RPC timeout timer earlier for 2.6 (3219)
3953 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3954 - print nal/nid for unknown nid (3258)
3955 - additional checks for oscc recovery before doing precreate (3284)
3956 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3957 - don't crash in mdc_close for bad permissions on open (3285)
3958 - zero i_rdev for non-device files (3147)
3959 - clear page->private before handing to FS, better assertion (3119)
3960 - tune the read pipeline (3236)
3961 - fix incorrect decref of invalidated dentry (2350)
3962 - provide read-ahead stats and refine rpc in flight stats (3328)
3963 - don't hold journal transaction open across create RPC (3313)
3964 - update atime on MDS at close time (3265)
3965 - close LDAP connection when recovering to avoid server load (3315)
3966 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3967 - don't leak open file on MDS after open resend (3325)
3968 - serialize filter_precreate and filter_destroy_precreated (3329)
3969 - loop device shouldn't call sync_dev() for nul device (3092)
3970 - clear page cache after eviction (2766)
3971 - resynchronize MDS->OST in background (2824)
3972 - refuse to mount the same filesystem twice on same mountpoint (3394)
3973 - allow llmount to create routes for mounting behind routers (3320)
3974 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3975 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3976 - fix rare problem with rename on htree directories (3417)
3978 - allow default OST striping configuration per directory (1414)
3979 - fix compilation for qswnal for 2.6 kernels (3125)
3980 - increase maximum number of MDS request buffers for large systems
3981 - change liblustreapi to be useful for external progs like lfsck (3098)
3982 - increase local configuration timeout for slow disks (3353)
3983 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3985 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3988 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3989 - don't overwrite extent policy data in reply if lock was blocked (2901)
3990 - drop filter export grants atomically with removal from device (2663)
3991 - del obd_self_export from work_list in class_disconnect_exports (2908)
3992 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3993 - swab reply message in mdc_close, other PPC fixes (2464)
3994 - fix destroying of named logs (2325)
3995 - overwrite old logs when running lconf --write_conf (2264)
3996 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3997 - fix race in target_handle_connect (2898)
3998 - mds_reint_create() should take same inode create lock (2926)
3999 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4000 - don't close files for self_export to avoid uninitialized obd (2936)
4001 - allow MDS with the same name as client node (2939)
4002 - hold dentry reference for closed log files for unlink (2325)
4003 - reserve space for all logs during transactions (2059)
4004 - don't evict page beyond end of stripe extent (2925)
4005 - don't oops on a deleted current working directory (2399)
4006 - handle hard links to targets without a parent properly (2517)
4007 - don't dereference NULL lock when racing during eviction (2867)
4008 - don't grow lock extents when lots of conflicting locks (2919)
4010 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4013 - account for cache space usage on clients to avoid data loss (974)
4014 - lfsck support in lustre kernel code (2349)
4015 - reduce journal credits needed for BRW writes (2370)
4016 - orphan handling to avoid losing space on client/server crashes
4017 - ptlrpcd can be blocked, stopping ALL progress (2477)
4018 - use lock value blocks to assist in proper KMS, faster stat (1021)
4019 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4020 - recovery for initial connections (2355)
4021 - fixes for mds_cleanup_orphans (1934)
4022 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4023 - block all file creations until orphan recovery completes (1901)
4024 - client remove rq_connection from request struct (2423)
4025 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4026 - recovery timer race (2670)
4027 - mdc_close recovey bug (2532)
4028 - ptlrpc cleanup bug (2710)
4029 - mds timeout on local locks (2588)
4030 - namespace lock held during RPCs (2431)
4031 - handle interrupted sync write properly (2503)
4032 - don't try to handle a message that hasn't been replied to (2699)
4033 - client assert failure during cleanup after abort recovery (2701)
4034 - leak mdc device after failed mount (2712)
4035 - ptlrpc_check_set allows timedout requests to complete (2714)
4036 - wait for inflight reqs when ptlrpcd finishes (2710)
4037 - make sure unregistered services are removed from the srv_list
4038 - reset bulk XID's when resending them (caught by 1138 test)
4039 - unregister_bulk after timeout
4040 - fix lconf error (2694)
4041 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4042 - readahead locks pages, leaves pending causing memory pressure (2673)
4043 - increase OST request buffers to 4096 on large machines (2729)
4044 - fix up permission of existing directories in simple_mkdir (2661)
4045 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4046 - don't assign transno to errored transactions (2742)
4047 - don't delete objects on OST if given a bogus objid from MDS (2751)
4048 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4049 - if rq_no_resend, then timeout request after recovery (2432)
4050 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4051 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4052 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4053 - MDS recovery shouldn't delete objects that it has given out (2730)
4054 - if enqueue arrives after completion, don't clobber LVB (2819)
4055 - don't unlock pages twice when trigger_group_io returns error (2814)
4056 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4057 - don't write pages to disk if there was an error (1450)
4058 - don't ping imports that have recovery disabled (2676)
4059 - take buffered bytes into account when balancing socknal conn (2817)
4060 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4061 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4062 - remove little-endian swabbing of llog records (1987)
4063 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4064 - retry reposting request buffers if they fail (1191)
4065 - grow extent at grant time to avoid granting a revoked lock (2809)
4066 - lock revoke doesn't evict page if covered by a second lock (2765)
4067 - disable VM readahead to avoid reading outside lock extents (2805)
4069 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4070 - updated kernel patches for hp-2.4.20 kernel (2681)
4072 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4075 - fix truncated write corruption (2366)
4076 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4078 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4079 - ensure len(uuid) < 37 in lmc (1171)
4080 - fix ia64 OOPS in llog_test (2255)
4081 - zero end of page at obdfilter for partial page writes (2648)
4082 - don't leave stale dentries around after renames (bug 2428)
4083 - fix timeouts when evicting a client with a single lock held (2642)
4084 - set deadline for the initial HELLO message to drain (2634)
4085 - print out dotted-quad IP addresses in the socknal (2302)
4087 - additional debugging for MDS client eviction problem (2443)
4088 - fix mkfsoptions support for osts (2603, 2604)
4090 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4093 - add series for the vanilla 2.6.0 kernel
4094 - add series for the vanilla 2.4.24 kernel
4095 - add series for a cray x86/64 UL kernel drop
4096 - fix xattr patches for the vanilla 2.4.19 series
4098 - generate true UUIDs in lmc (1171)
4099 - have portals stack dumping break in UML (2466)
4100 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4101 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4102 - wake callback waiting threads on client eviction (2460)
4103 - Add --ptldebug and --subsystem to lmc (1719)
4104 - update assertion to allow safe interrupt allocation
4105 - set rq_no_resend for cancel requests (2432)
4106 - recalculate ptlrpcd timeout after resend (2494)
4107 - call vfs_rmdir when removing pending directories (2368)
4108 - fix renaming a file to itself (2429)
4109 - lmc creates a default one-stripe lov (2454)
4110 - expand procfs space to handle large clusters (2326)
4111 - increase UML stack to avoid overflow
4112 - update lconf's list of debug and subsystem masks
4113 - fix lfs find --obd (2510)
4114 - /proc tunable for disabling filter read caching (2591)
4115 - stop rpm packages from altering slapd.conf (2301)
4116 - disable nagle in the socknal under 0conf (2578)
4117 - choose mds inode size based on stripe count (2572)
4118 - fix kernel-source rpm problems (2516)
4120 - add --disable-doc to avoid pdf generation (2421)
4121 - update documentation, tests, type-os, comments
4122 - avoid format warnings on ia64
4123 - remove the TOE NAL
4124 - tiny code cleanups by removing unused fields
4126 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4129 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4130 - remove the most blatant lies from BUILDING file (2371)
4131 - change default debug level to reasonable production setting
4132 - reduce client side cache size to reduce cache flush time
4133 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4134 - make TCP zerocopy and pinger support enabled by default (2476)
4135 - sync writes completed after process exits caused crashes (2319)
4136 - maintain correct mount count on the MDS (2356)
4137 - backout 1557, because 2316 wasn't really fixed
4138 - better file I/O statistics gathering in /proc
4139 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4140 - another kernel patch to fix zero-copy TCP function export
4141 - don't take duplicate lock when processing re-sent getattr (2420)
4142 - lctl uses obd_self_export instead of creating new conn (2353)
4143 - MDS/OST recovery case which requires object creation asserted (2425)
4144 - move lfs from /usr/sbin to /usr/bin in packages
4145 - fix race between mds_client_add and mds_client_free (2417)
4146 - use kmalloc instead of slabs in portals (2430)
4147 - don't create duplicate records when a failover MDS is present (2442)
4148 - remove unnecessary mount age check (2332)
4149 - don't remove directory inodes from locks prematurely (2451)
4150 - don't break if MDS service name is the same as hostname (2103)
4151 - fix races in client write RPC generation when cache full (2482)
4153 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4156 - remove now-unused request->rq_obd (278)
4157 - if an allocation fails, print out how much memory we've used (1933)
4158 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4159 - add the "configurable stack size" patch to most series files (1256)
4160 - ability to write large log records, for 100+ OST configs (2306)
4161 - fix NULL deref when filter_prep fails (2314)
4162 - fix operator precedence error in filter_sync
4163 - dynamic allocation of socknal TX descriptors (2315)
4164 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4165 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4166 - ensure that CWARN messages in Portals always get to the syslog
4167 - __init/__exit are not for prototype decls (ldlm_init/exit)
4168 - x86-64 compile warning fixes
4169 - fix gateway LMC keyword conflict (2318)
4170 - fix MDS lock inversions in getattr/reint paths (1844)
4171 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4172 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4173 - DLM race condition prevented some lock evictions (2328)
4174 - ENOMEM detection and retry on socknal sends (2230)
4175 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4176 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4177 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4178 - better error messages when a client is rejected during recovery (1505)
4179 - avoid cancelling locks which were never granted, after failure (2330)
4180 - fix i_sem/journal inversion in mds_client_add (2333)
4181 - fix truncate/getattr lock cycle deadlock (2334)
4182 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4183 - fix two rare exit paths which could leak an l_lock() ref (2321)
4184 - fencepost error in MDS/OST orphan recovery (2226)
4185 - make log record alignment 8 bytes (1988)
4186 - lstripe now fails when requested offset > ost_count (2237)
4187 - ensure that all kernel series have a complete list.h (1607)
4188 - fix crashes in special-file operations (2316)
4189 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4191 - allow configurable automake binary, for testing new versions
4192 - small update to the lfs documentation
4194 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4196 * fix negative export reference count in fsfilt_sync (2312)
4198 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4199 * release candidate 0.9.1
4201 - orphans are moved into the PENDING directory for possible recovery
4202 - replayed opens now open by fid for orphan/rename safety (1042)
4203 - last close of an orphan inode generates a transno (683)
4204 - chdir() and mount() now pin the directory entry (1020)
4205 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4206 - discard very old requests without processing them (1502)
4207 - remove some common, well-understood CERRORs (1505)
4208 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4209 - clear "grant" flags in OST replies until OST grant code lands (1644)
4210 - fix read performance by not clobbering i_blksize on client (1598)
4211 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4212 - make LRU size a /proc tunable, clears locks when reduced (707)
4213 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4214 - server locks take references on exports now (1558)
4215 - build fixes for 2.4.20-rh trees (1663)
4216 - return an error from lov_create if all OSCs are inactive (1751)
4217 - fix import levels when a reconnect happens without a timeout (1597)
4218 - exit early from mds_open if we get a lookup error (1749)
4219 - partial page read at EOF wouldn't wait for disk before sending (1642)
4220 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4221 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4222 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4223 - write/truncate lock inversion (1639)
4224 - Don't auto-load obdclass, portals modules during cleanup (1495)
4225 - fix timestamps from jumping to "now" (1763)
4226 - extra journal assertions (1648)
4227 - add an extra multiunlink test (1771)
4228 - fix read_record/write_record API (1776)
4229 - fix leak of offset_extent, possible incorrect i_size later (1772)
4230 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4231 - Debugging for the kqswnal_get_idle_tx problems (1820)
4232 - Allow recovery to be attempted multiple times (1536)
4233 - Write out MDS last_rcvd file after it is first created (1600)
4234 - Fix tx_descriptor leak in failed transmit situations (1827)
4235 - ext3 journaling fixes for assertion failure after IO error (1871)
4236 - class_export_put() on freed export after completion AST error (1896)
4237 - Fix revalidate looping in VFS (1322)
4238 - Don't access a freed export during MDS_REINT timeout (1521)
4239 - Add open-unlink recovery support on the MDS (1673,1764)
4240 - Return an error if no MDS data was read from last_rcvd (1946)
4241 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4242 - Don't setup a disk device that doesn't match exported UUID (317)
4243 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4244 - avoid committing NULL handle in force close
4245 - local.sh is now a one-stripe LOV configuration
4246 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4247 - fix ext3 htree duplicate directory entry corruption (1516)
4248 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4249 - update to new LOV EA format (2097)
4250 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4252 2003-06-15 Phil Schwan <phil@clusterfs.com>
4255 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4256 - per-import/export recovery handling (958, 931, 959)
4257 - multiple last-rcvd slots, for serving multiple FSes (949)
4258 - connections are again shared between multiple imp/exports (963, 964)
4259 - "umount -f" would hang if any requests needed to be sent (393, 978)
4260 - avoid pinning large req buffer by copying for queued messages (989)
4261 - add "uuid" to "lctl device" command to help upcalls (991)
4262 - "open" RPCs with transnos would confuse recovery counters (1037)
4263 - do proper endian conversion of all wire messages (288, 340, 891)
4264 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4265 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4266 - fix ptlrpc_connection leak in target_handle_connect (1174)
4267 - fix import refcounting bug in OST and MDS cleanup (1134)
4268 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4269 - fix very unlikely obd_types race condition (501)
4270 - remove osc_open hack for echo_client (1187)
4271 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4272 - a failure in read_inode2 leads to deadlock (1139)
4273 - cancel ack-locks as soon as transaction is committed (1072)
4274 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4275 - make sure to commitrw after any preprw to avoid deadlock (1162)
4276 - failing to execute a file in a lustre FS would lock inode (1203)
4277 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4278 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4279 - obd_uuid2tgt would walk off the end of the list (1255)
4280 - on IA64 the osc would give portals incorrect bulk size (1258)
4281 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4282 - fix lock inversion caused by new llite matching code (1282)
4283 - limit the number of dirty pages on a client to 10MB (1286)
4284 - timed out locks were not being corrected cancelled (1289)
4285 - fix O_DIRECT above 4GB on IA-32 (1292)
4286 * major user-visible changes
4287 - fail out/fail over policy now controlled by the upcall (993)
4289 - add OBD_PING to check server availability and failure (954)
4290 - lustre messages are now sent in sending host order (288, 340, 891)
4291 - add eadatalen to MDS getattr reply (340)
4292 - OST read replies may contain second buffer, with per-page status (593)
4294 2003-03-11 Phil Schwan <phil@clusterfs.com>
4297 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4298 - failed open()s could cause deadlock; fixed (867, 869)
4299 - stop cancelling OST locks when files are closed (481)
4300 - overlapping XID spaces caused network corruption (851, 853)
4301 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4302 - setattr_raw added, to do single-RPC, server-side setattrs
4303 - lmc/lconf syntax change for OST UUIDs
4304 - fix crashy race condition between ptlrpc_free_req and osc_close
4305 - don't use request in mdc_enqueue if we hit a timeout (889)
4306 - don't set the inode i_size for regular files from the MDS (896)
4307 - handle out of order completion AST (842)
4308 - don't LBUG if a lock request times out after receiving AST (913)
4309 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4310 - if a bad lock AST arrives, send an error instead of dropping entirely
4311 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4312 - fix leak in bulk IO when only partially completed (899, 900, 926)
4313 - fix O_DIRECT for ia64 (55)
4314 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4315 - C-z after timeout could hang a process forever; fixed (977)
4317 - client-side I/O cache (678, 924, 929, 941, 970)
4319 - READPAGE and SETATTRs which don't take server-side locks get
4322 2003-02-11 Phil Schwan <phil@clusterfs.com>
4325 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4326 - Fix fsfilt_extN_readpage() to read a full page of directory
4327 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4328 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4329 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4330 - Most intent operations were removed, in favour of a new RPC mode
4331 that does a single RPC to the server and bypasses most of the VFS
4332 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4333 - Aggressively cancel local locks on DLM servers
4334 - mds_reint_unlink sends EA to the client if it's the last nlink.
4335 client uses that EA to unlink OST objects.
4336 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4337 - recursive symlinks were fixed (439)
4338 - fixed NULL deref in DEBUG_REQ
4339 - filter_update_lastobjid no longer calls sync, which annoyed extN
4340 - fixed multi-client small-writes to a single file problem (445)
4341 - fixed mtime updates during file writes (607)
4342 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4343 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4344 - fixed Linux OST size reporting problem (444, 656)
4345 - OST now updates object mtime with writes or setattr (607, 619)
4346 - client verifies file size before zeroing page past EOF (445)
4347 - OST now writes last allocated objid to disk with allocation (108)
4348 - LOV on echo now works (409)
4350 - mds_reint_unlink sends a new buffer, with the EA included. this
4351 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4352 set if a regular file was being unlinked, and it was the last link
4353 - use PtlGet from the target for bulk writes (315)
4354 - OST now updates object mtime with writes or setattr (607, 619)
4355 - LDLM now has a grant-time callback to revalidate locked items, if
4357 - Many MDS operations were reorganized to combat race conditions
4359 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4360 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4362 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4365 - Fully reactivate OST imports after reconnection (512, others)
4366 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4367 - More graceful error handling for truncating on dead OST (515)
4368 - Don't error out unless we're actually accessing dead stripes (474)
4369 - Fix garbage sizes when stripes are missing (410)
4370 - LRU counters were broken, causing constant lock purge (433, 432)
4371 - garbage on read from stripes with failed OSTs (441)
4372 - mark OSCs as active before reconnecting during recovery (438)
4373 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4374 - lfind did not preserve OST order in output (443)
4375 - symlinks cause hung clients, incorrect data (439)
4376 - stop dereferencing request after dropping refcount (457)
4377 - don't LASSERT(spin_is_locked) on non-SMP (455)
4378 - fixes for many rename() bugs
4379 - fstat didn't correctly synchronize attributes (399)
4380 - server must handle lock cancellation during blocking AST prep (487)
4381 - bulk descriptors were free()d too soon (511)
4382 - fix paths in lconf, which would load incorrect modules (451, 507)
4383 - fix confusing lconf 'host not found' error message (386)
4384 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4385 - fix race condition in mdc_blocking_ast() for inode access (526)
4386 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4387 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4388 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4389 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4390 - fix lstripe in several ways (kernel side) (527)
4391 - fix request leak in ldlm_cli_enqueue (262)
4392 - incorrect OSC was marked inactive after OST failure
4393 - call mds_fs_cleanup before unmounting filesystem (524)
4394 - fix races between taking ns_lock and ldlm_lock_change_resource
4395 - fix races updating LOV export open file list
4396 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4397 - fix recovery NULL deref in ldlm_cli_cancel_unused
4398 - fix some DLM races by using new hash table for lock handles (419)
4399 - permit the client to specify desired inodes, at replay
4400 - duplicate requests when we queue them for replay reintegration
4401 - fix last_rcvd offset calculation
4402 - sync after each recovered transaction, so we always make progress
4403 - never, not always, ERESTART requests without transnos
4404 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4406 - skip replay if the MDS says that the client is already connected
4407 - don't check for a recovery-enabled export to match lctl's UUID
4408 - don't INC_USE_COUNT for phantom exports
4409 - don't crash when cleaning up phantom exports (567)
4410 - don't double-finish or set replay data for errored mdc_open requests
4411 - abort requests when they time out, so we don't get old replies
4412 - send/receive replies for AST messages again
4413 - if the client says that it doesn't have the lock, cancel it on the
4415 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4416 mark it as destroyed, it will all work out when decref is called
4417 - fix module use counts (22, 581)
4419 - ASTs now expect a reply (server cancels lock on error reply)
4421 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4424 - fix many simultaneous client startup (392)
4425 - fix dentry->d_it clobbering
4426 - credentials weren't being shipped for readdir/getattr operations
4427 - remove invalid assertions triggered during some concurrent MD
4429 - proper Lustre versions added (336, 389)
4430 - fix memory leak for create error case (398)
4431 - fix LOV locking bug that would get cli/srv out of sync
4432 - fix echo client over LOV (409)
4433 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4434 - fix double-O_EXCL intent crash (424)
4435 - avoid sending multiple lock CANCELs (352)
4437 - MDS can do multi-client recovery (modulo bugs in new code)
4439 - many updates, edits, cleanups
4441 2002-11-18 Phil Schwan <phil@clusterfs.com>
4444 - fix null d_it dereference (346)
4445 - fix full OST/dbench hang (333)
4446 - fix permission problem with file removal (286)
4447 - fix removal of OSCs from LOV when they fail
4448 - fix NULL deref during bulk timeout (214)
4449 - fix problems related to multiple filesystems on one MDS (241)
4450 - fixed serious subtle metadata locking bugs
4451 - free locks on clients when inodes are removed due to memory
4453 - fix inode pointer in lock data (285)
4454 - partial support for multiple MDS on a single host (241)
4455 - data locks weren't cancelled at clear_inode time (290, 311)
4456 - intent locks could lead to unbounded lock growth (205)
4457 - added a maximum lock count, an LRU list, and a flusher
4458 - fix multiple rename (365)
4459 - properly abstracted the echo client
4460 - OSC locked 1 byte too many; fixed
4461 - rewrote brw callback code:
4462 - fixed recovery bugs related to LOVs (306)
4463 - fixed too-many-pages-in-one-write crash (191)
4464 - fixed (again) crash in sync_io_timeout (214)
4465 - probably fixed callback-related race (385)
4467 - Add capability to MDS protocol
4468 - LDLM cancellations and callbacks on different portals
4470 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4473 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4474 - defer open object destruction to close time (601981, 138)
4475 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4476 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4477 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4478 - don't call obd_setattr in ll_file_release for destroyed objects
4479 * protocol change to lustre_msg: move |version| and add |flags|
4480 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4481 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4482 * added replay of create, unlink, link and rename operations during
4483 MDS failover; recovery should be much more robust now
4484 * remove failed OSCs from LOVs (only lov_create uses this so far)
4485 * the lustre-HOWTO was brought (more) up to date (582544)
4487 2002-10-23 Phil Schwan <phil@clusterfs.com>
4490 - in-use dentries weren't being reused properly (617851)
4491 - prevent multiple LDLM setup (599178)
4492 - fix LOV size calculations for truncate (617853)
4493 - fix client handling of MDS intent errors (POSIX)
4494 - fix permission bug in lovstripe.c test (624321)
4495 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4496 - truncate past end of file could corrupt data
4497 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4498 - a race in recovery could return ETIMEDOUT to apps (623947)
4499 - building outside the source directory was fixed
4500 * the lustre-HOWTO was brought (more) up to date (582544)
4501 * major progress was made on recovery functionality
4503 2002-10-10 Phil Schwan <phil@clusterfs.com>
4506 - recovery deadlock fix
4507 - rm -rf causes LBUG fix (617817)
4508 - file open by multiple tasks fix (618962)
4509 - directory permissions bugs (602707 and 620007)
4510 - journal_stop fixed with locking (611313)
4511 - O_APPEND failures resolved (618273, perhaps 614459)
4512 - lconf PATH fix (619770)
4513 - IA64 build fix (621450)
4514 - RPC buffer sizes scale with amount of memory
4516 2002-10-01 Phil Schwan <phil@clusterfs.com>
4519 - locks would be cancelled without throwing away data pages,
4520 resulting in inconsistent data (605627)
4521 - inode attributes were not always being refreshed (605627, 612449)
4522 - lconf now continues to cleanup after lctl reports an error
4523 - MDS now enforces user permissions (602707)
4524 - lprocfs cleanup fixed, but not yet enabled (614157)
4525 - fixed infinite server hang, should a client not respond to an AST
4526 - avoid going into recovery if user calls readlink() with a buffer
4527 that's too small (613941)
4528 - AST RPCs no longer require replies (614867) -- this may be changed
4529 - don't crash server if client sends an IOV that's too big (611336)
4530 - fixed lock conversion deadlock (611892)
4531 - fixed the following of symlinks (614622)
4532 * recovery: the server can remove locks from a client that dies, other
4533 clients can make progress
4534 * more extN patch fixes
4535 * compile-time configurable ptlrpc buffer allocations
4537 - collaborative read cache document
4538 - Lustre Lite Performance CDR document-in-progress
4540 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4543 - fix typo in patch-2.4.18
4545 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4548 - clear ptlrpc request each time in handle_incoming_request()
4549 - unlink of files now destroys the object on the OST
4551 2002-09-19 Peter Braam <braam@clusterfs.com>
4553 * add hard link support
4554 * change obdfile creation method
4555 * kernel patch changed
4557 2002-09-19 Peter Braam <braam@clusterfs.com>
4560 - stack overflow bug in extN fixed
4562 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4564 * documentation updates
4565 - add man pages for config tools
4566 - update tests/README to describe testing with new config tools
4567 - finish metadata API descriptions
4568 * bug fixes and cleanups
4569 - statfs workaround for 16TB limit
4570 - LOV stripe allocation improved, can stripe on subset of OSTs
4571 - LOV file size/IO offset was wrong for files > 4GB in size
4572 - object EA data was being dropped, caused files to be unreadable
4573 - memory overflow with non-LOV OST caused memory corruption
4574 - fixed regression tests to work with new config tools, obdfilter
4575 - fixed bug when directory size became larger than 1 block
4576 - fixed bug (for single client case) when PWD was deleted
4577 - invalidate local directory pages when doing intent-based ops
4578 - avoid LDLM oops when lock callback contained bad data
4580 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4582 * documentation updates
4583 * bug fixes and cleanups
4584 - configuration tools
4587 - 64-bit compile warnings
4588 - 64-bit internal statfs data
4590 * test_brw on persistent OST devices
4592 * lprocfs (disabled)
4594 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4596 * documentation updates
4597 * bug fixes and cleanups
4598 * configuration tools
4600 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4603 * many small fixes to 0_5_4
4604 * io/network handling
4605 * thinkos in MDS operations
4607 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4610 * crucial basic fixes to 0.5.3
4611 * IOR, Iozone work over Elan
4614 2002-08-07 Phil Schwan <phil@clusterfs.com>
4615 * version 0_5_3, our first alpha
4616 * we use the new Portals iovs
4617 * documentation updates
4618 * bug fixes and cleanups
4619 * small changes in the DLM wire protocol
4621 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4622 * version 0_5_1 with some initial stability,
4623 * locking on MD and file I/O.
4624 * documentation updates
4625 * several bug fixes since 0.5.0
4626 * small changes in wire protocol
4628 2002-07-18 Phil Schwan <phil@clusterfs.com>
4630 * delivered as Lustre Light Alpha
4631 * fixed a crash after handling invalid MDS requests
4632 * fixed directory pages for architectures with non-4k pages sizes
4634 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4635 * release version v0_4_4
4636 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4638 * Fixes a number of interruption problems with OST operations.
4639 * Update documentation for portals header changes
4640 * Move all wire protocol structs/defines to lustre_idl.h
4641 * Fixes symlink length bug.
4642 * Add tcpdump to repository.
4644 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4645 * release version v0_4_3
4646 * Fixes statfs for inodes on extN.
4647 * Fixes bug in runtests which would delete /etc/hosts.
4648 * Use 64-bit object IDs wherever possible (not into VFS though)
4649 Remove ost_get_info, which is unused by lustre, and out of date.
4651 2002-07-03 Peter Braam <braam@clusterfs.com>
4652 * release version v0_4_2 Fixes a lookup error (type not passed)
4653 * move forward to head of Portals
4654 * move forward to latest Lustre kernel
4656 2002-06-25 Peter Braam <braam@clusterfs.com>
4657 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob