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 put request into delay list while invalidate in flight.
18 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
19 import in flight. this produce timeout for invalidate and sometimes
24 Frequency : on ppc only
25 Description: not convert ost objects for directory because it's not exist.
26 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
30 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
32 Severity : enhancement
34 Description: Update to support 2.6.22.14 vanilla kernel.
38 Frequency : rare, on recovery
39 Description: read procfs can produce deadlock in some situation
40 Details : Holding lprocfs lock which send rpc can produce block for destroy
41 obd objects and this also block reconnect with -EALREADY. This isn't
42 fix all lprocfs bugs - but make it rare.
44 Severity : enhancement
46 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
49 Frequency : frequent on X2 node
51 Description: mdc_set_open_replay_data LBUG
52 Details : Set replay data for requests that are eligible for replay.
56 Description: lustre_mgs: operation 101 on unconnected MGS
57 Details : When MGC is disconnected from MGS long enough, MGS will evict the
58 MGC, and late on MGC cannot successfully connect to MGS and a lot
59 of the error messages complaining that MGS is not connected.
62 Frequency : on start mds
64 Description: Implement get_info(last_id) in obdfilter.
67 Frequency : occasional
69 Description: Correctly check stale fid, not start epoch if ost not support SOM
70 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
71 overwrited with new generated fid, but mds can anwer with one of these
72 two fids and both is not stale. setattr incorectly start epoch and
73 assume will be called done_writeting, but without SOM done_writing
77 Frequency : rare, depends on device drivers and load
79 Description: MDS or OSS nodes crash due to stack overflow
80 Details : Code changes in 1.8.0 increased the stack usage of some functions.
81 In some cases, in conjunction with device drivers that use a lot
82 of stack the MDS (or possibly OSS) service threads could overflow
83 the stack. One change which was identified to consume additional
84 stack has been reworked to avoid the extra stack usage.
87 Frequency : occasional
89 Description: Do not fail import if osc_interpret_create gets -EAGAIN
90 Details : If osc_interpret_create got -EAGAIN it immediately exits and
91 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
92 oscc_has_objects and see OSC has no objests and call
93 oscc_internal_create to resend create request.
95 Severity : enhancement
97 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
99 Severity : enhancement
101 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
104 Frequency : very rare
106 Description: Fix replay if there is an un-replied request and open
107 Details : In some cases, older replay request will revert the
108 mcd->mcd_last_xid on MDS which is used to record the client's
111 Severity : enhancement
113 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
115 Severity : enhancement
117 Description: Add rhel5 support to HEAD.
119 Serverity : enhancement
121 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
126 Description: Don't allow skipping OSTs if index has been specified.
127 Details : Don't allow skipping OSTs if index has been specified, make locking
128 in internal create lots better.
132 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
133 Details : access to bitfield in structure is always rounded to long
134 and this produce problem with not atomic change any bit.
138 Description: Lustre make rpms failed.
139 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
140 builds Lustre rpms from tarball.
143 Frequency : rare on shutdown ost
145 Description: If llog cancel was not send before clean_exports phase, this can
146 produce deadlock in llog code.
147 Details : If llog thread has last reference to obd and call class_import_put
148 this produce deadlock because llog_cleanup_commit_master wait when
149 last llog_commit_thread exited, but this never success because was
150 called from llog_commit_thread.
154 Description: allow userland application know is lost one of stripes.
155 Details : fill lvb_blocks with error code on ost and return it to
156 application if error flag found.
160 Description: NULL lov_tgts causing MDS oops
161 Details : more safe checks for NULL lov_tgts for avoid oops.
163 Severity : enhancement
165 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
169 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
170 Details : If an OST is down the MDS will hang indefinitely in
171 obd_statfs_rqset() waiting for the statfs data. While for
172 MDS QOS usage of statfs, it should not stuck in waiting.
174 Severity : enhancement
176 Description: remote_acl support
177 Details : Support ACL-based permission check for remote user.
178 Support setfacl/getfacl for remote user with the utils
179 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
180 the system "{s,g}etfacl" utils.
182 Severity : enhancement
184 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
186 Severity : enhancement
188 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
192 Description: make mgs_setparam() handle fsname containing dash
193 Details : fsname containing a dash does not work with lctl conf_param
195 Severity : enhancement
197 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
199 Severity : enhancement
201 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
205 Frequency : rare, at shutdown
206 Description: access already free / zero obd_namespace.
207 Details : if client_disconnect_export was called without force flag set,
208 and exist connect request in flight, this can produce access to
209 NULL pointer (or already free pointer) when connect_interpret
210 store ocd flags in obd_namespace.
214 Frequency : only at startup
215 Description: not alloc memory with spinlock held.
216 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
217 if any spinlock held.
219 Severity : enhancement
221 Description: make lustre randomly fail allocating memory
222 Details : Make lustre randomly failed allocating memory for testing purpose.
224 Severity : enhancement
226 Description: lost problems with lov objid file
227 Details : Fixes some scability and access to not inited memory problems
228 in work with lov objdid file.
233 Description: lfs find does not continue on file error
234 Details : Continue other files processing when a file/dir is absent.
238 Description: Inconsistent usage of lustre_pack_reply()
239 Details : Standardize the usage of lustre_pack_reply() such that it
240 always generate a CERROR on failure.
243 Frequency : occasional
245 Description: MDS hang or stay in waiting lock
246 Details : If client receive lock with CBPENDING flag ldlm need send lock
247 cancel as separate rpc, to avoid situation when cancel request
248 can't processed due all i/o threads stay in wait lock.
252 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
256 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
260 Description: Update to SLES9 kernel-2.6.5-7.287.3.
264 Description: Update to RHEL4 latest kernel.
266 Severity : enhancement
268 Description: Build SLES10 patchless client fails
269 Details : The configure was broken by run ./configure with
270 --with-linux-obj=.... argument for patchless client. When the
271 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
272 can't search header adequately. Use absolute path such as
273 -I($LINUX)/include instead.
277 Description: interrupt oig_wait produce painc on resend.
278 Details : brw_redo_request can be used for resend requests from ptlrpcd and
279 private set, and this produce situation when rq_ptlrpcd_data not
280 copyed to new allocated request and triggered LBUG on assert
281 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
282 ptlrpcd set if request is changed and can be safety changed to use
287 Description: LASSERT_{REQ,REP}SWAB macros are buggy
288 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
289 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
290 these macros and replace them with their difinitions instead.
294 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
295 Details : Update which_patch & target file for SLES10 latest kernel.
298 Description: Remove client patches from SLES 10 kernel.
299 Details : This causes SLES 10 clients to behave as patchless clients
300 even on a Lustre-patched (server) kernel.
302 Severity : enhancement
304 Description: self-adjustable client's lru lists
305 Details : use adaptive algorithm for managing client cached locks lru
306 lists according to current server load, other client's work
307 pattern, memory activities, etc. Both, server and client
308 side namespaces provide number of proc tunables for controlling
311 Severity : enhancement
313 Description: light-weight GSS support
314 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
315 and reduce performance overhead.
317 Severity : enhancement
319 Description: Linux keyring support
320 Details : Support using service of Linux keyring for Lustre GSS internal
321 context refresh/cache mechanism.
325 Description: Fix errors in lfs documentation
326 Details : Fixes man pages
330 Description: don't use GFP_* in generic Lustre code.
331 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
335 Description: obdclass is limited by single OBD_ALLOC(idarray)
336 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
340 Description: warnings with build patchless client with vanila 2.6.19 and up
341 Details : change old ctl_table style and replace ctl_table/ctl_table_header
342 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
346 Frequency : Only for SLES
347 Description: sanity_quota fail test_1
348 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
349 file, which makes TSTID[2] inunique.
353 Frequency : Only for SLES9
354 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
355 Details : Update target/ChangeLog/which_patch .
359 Frequency : Only for test_10 in sanity-quota.sh
360 Description: a bug in quota test code
361 Details : checking a test flag in a wrong place
365 Frequency : Only for test_13 in sanity-quota.sh
366 Description: a bug in quota test script
367 Details : shouldn't check overfull allocated quota
371 Description: jbd statistics
372 Details : Port older jbd statistics patch for sles10
374 Severity : enhancement
376 Description: Add printing inode info into message about error in writepage.
381 Description: reply_lock_interpret crash due to race with it and lock cancel.
382 Details : Do not replay locks that are being cancelled. Do not reference
383 locks by their address during replay, just by their handle.
387 Frequency : When flocks are used.
388 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
389 Details : Flock locks might destroy just granted lock if it could be merged
390 with another existing flock, this is done in completion handler,
391 so teach ldlm_cli_enquque_fini that this is a valid case for
397 Description: Oops in read and write path when failing to allocate lock.
398 Details : Check if lock allocation failed and return error back.
402 Description: lstripe command fails for valid OST index
403 Details : The stripe offset is compared to lov->desc.ld_tgt_count
404 instead of lov->desc.ld_active_tgt_count.
408 Description: sanity.sh failed test 103
409 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
414 Description: df doesn't work properly if diskfs blocksize != 4K
415 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
419 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
420 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
424 Description: obd_health_check_timeout too short
425 Details : set obd_health_check_timeout as 1.5x of obd_timeout
429 Description: llapi_file_create() does not allow some changes
430 Details : add llapi_file_open() that allows specifying the mode and
431 open flags, and also returns an open file handle.
435 Description: merge and cleanup kernel patches.
436 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
440 Description: Add journal checksum support.(Kernel part)
441 Details : The journal checksum feature adds two new flags i.e
442 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
443 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
444 indicates that the commit block contains the checksum for
445 the blocks described by the descriptor blocks. Now commit
446 record can be sent to disk without waiting for descriptor
447 blocks to be written to disk. This behavior is controlled
448 using JBD2_FEATURE_ASYNC_COMMIT flag.
452 Description: OSS needs mutliple precreate threads
453 Details : Add ability to start more than one create thread per OSS.
457 Description: Sanity.sh test_65k failed: setstripe should have succeeded
458 Details : execute lfs setstripe on client
462 Description: mds_obd_create error creating tmp object
463 Details : When the user sets quota on root, llog will be affected and can't
464 create files and write files.
467 Frequency : Always on ia64 patchless client, and possibly others.
469 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
470 Details : This allows the patchless client to be loaded on architectures
475 Description: RedHat Update kernel for RHEL5
476 Details : Add the kernel config file for RHEL5.
480 Description: RedHat Update kernel for RHEL5
481 Details : Modify the kernel config file more closer RHEL5.
485 Description: Build failure against Centos5 (RHEL5)
486 Details : Define PAGE_SIZE when it isn't present.
490 Description: test_7 on sanity-quota.sh can't be used on separate machines
491 Details : add facet to handle it
495 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
496 Details : with NFS, the anon dentry's parent was set to itself in
497 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
498 corresponding dentry other than use rec->ur_name.
500 Severity : enhancement
502 Description: lfs setstripe enhancement
503 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
507 Description: enable data checksumming by default
508 Details : enable checksum by default, allow --disable-checksum
509 configure option and "-o nochecksum" mount option. Checksums
510 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
514 Description: lustre support for RHEL5
515 Details : Add support for RHEL5.
519 Description: Client eviction due to failover config
520 Details : after a connection loss, the lustre client should attempt to
521 reconnect to the last active server first before trying the
522 other potential connections.
526 Description: when mds and osts use different quota unit(32bit and 64bit),
527 quota will be released repeatly.
528 Details : void sending multiple quota reqs to mds, which will keep the
529 status between the reqs.
533 Description: osts not allocated evenly to files
534 Details : change the condition to increase offset_idx
537 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
539 Description: Data corruption for OSTs that are formatted larger than 2TB
541 Details : When generating the bio request for lustre file writes the
542 sector number would overflow a temporary variable before being
543 used for the IO. The data reads correctly from Lustre (which
544 will overflow in a similar manner) but other file data or
545 filesystem metadata may be corrupted in some cases.
549 Description: Tune the kernel for good SCSI performance.
550 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
551 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
556 Description: rewrite ext2-derived code in obdclass/uuid.c
557 Details : rewrite inherited code (uuid parsing code from ext2 utils)
558 from scratch preserving functionality.
562 Description: "lfs find -obd UUID" prints directories
563 Details : "lfs find -obd UUID" will return all directory names instead
564 of just file names. It is incorrect because the directories
565 do not reside on the OSTs.
569 Description: use i_size_read and i_size_write in 2.6 port
570 Details : replace inode->i_size access with i_size_read/write()
573 Frequency : when removing large files
575 Description: scheduling issue during removal of large Lustre files
576 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
577 It causes scheduling issues when removing large files (17TB in the
581 Frequency : only with liblustre clients on XT3
583 Description: evictions taking too long
584 Details : allow llrd to evict clients directly on OSTs
587 Frequency : only on ppc
589 Description: /proc/fs/lustre/devices broken on ppc
590 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
591 We should make sure the type of 'index' is loff_t and then cast
592 explicitly as needed below. Do not assign an explicitly cast
596 Frequency : only for rhel5
598 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
599 Details : Modify the target file & which_kernel.
601 Severity : enhancement
603 Description: omit set fsid for export NFS
604 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
605 as FS_REQUIRES_DEV for avoid problems with generate fsid.
609 Frequency : Always, for kernels after 2.6.16
610 Description: Fix warning idr_remove called for id=.. which is not allocated.
611 Details : Last kernels save old s_dev before kill super and not allow
612 to restore from callback - restore it before call kill_anon_super.
616 Description: buffer overruns could theoretically occur
617 Details : llapi_semantic_traverse() modifies the "path" argument by
618 appending values to the end of the origin string, and a
619 overrun may occur. Adding buffer overrun check in liblustreapi.
623 Description: Fix error on 'ls .' at the top of the Lustre mount.
624 Details : Don't revalidate dentry if it is a root dentry.
628 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
629 Details : Modify vm-tunables-rhel4.patch.
633 Description: Kernel config for 2.6.18-vanilla.
634 Details : Modify targets/2.6-vanilla.target.in.
635 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
636 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
637 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
638 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
642 Description: improve handling recoverable errors
643 Details : if request processig with error which can be recoverable on server
644 request should be resend, otherwise page released from cache and
649 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
650 Details : Modify target file & which_patch.
651 A flaw was found in the IA32 system call emulation provided
652 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
653 value could be stored in the %RAX register, which could trigger an
654 out-of-bounds system call table access. An untrusted local user
655 could exploit this flaw to run code in the kernel
656 (ie a root privilege escalation). (CVE-2007-4573).
660 Description: O_DIRECT bypasses client statistics.
661 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
662 read_ahead_stats were updated. copied the stats section from
663 osc_send_oap_rpc() into async_internal().
667 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
671 Description: change order of libsysio includes
672 Details : '#include sysio.h' should always come before '#include xtio.h'
676 Description: handle "serious error: objid * already exists" more gracefully
677 Details : If LAST_ID value on disk is smaller than the objects existing in
678 the O/0/d* directories, it indicates disk corruption and causes an
679 LBUG(). If the object is 0-length, then we should use the existing
680 object. This will help to avoid a full fsck in most cases.
682 Severity : enhancement
684 Description: adapt the lustre_config script to support the upgrade case
685 Details : Add "-u" option for lustre_config script to support upgrading 1.4
686 server targets to 1.6 in parallel.
690 Description: To avoid grant space > avaible space when the disk is almost
691 full. Without this patch you might see the error "grant XXXX >
692 available" or some LBUG about grant, when the disk is almost
694 Details : In filter_check_grant, for non_grant cache write, we should
695 check the left space by if (*left > ungranted + bytes), instead
696 of (*left > ungranted), because only we are sure the left space
697 is enough for another "bytes", then the ungrant space should be
698 increase. In client, we should update cl_avail_grant only there
699 is OBD_MD_FLGRANT in the reply.
703 Description: Update RHEL 4 kernel to fix local root privilege escalation.
704 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
705 described in CVE-2007-4573. This problem could allow untrusted
706 local users to gain root access.
709 Frequency : when using O_DIRECT and quotas
711 Description: Incorrect file ownership on O_DIRECT output files
712 Details : block usage reported by 'lfs quota' does not take into account
713 files that have been written with O_DIRECT.
718 Description: touch file failed when fs is not full
719 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
720 otherwise we can get ENOSP while fs is not full.
724 Description: parallel lock callbacks
725 Details : Instead of sending blocking and completion callbacks as separated
726 requests, adding them to a set and sending in parallel.
729 Frequency : only for Cray XT3
730 Bugzilla : 12829/13455
731 Description: Changing primary group doesn't change the group lustre assigns to
733 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
734 overridden with the primary group provided by the group upcall,
735 whereas the supplied fsgid can be trusted if it is in the list of
736 supplementary groups returned by the group upcall.
738 Severity : enhancement
740 Description: Allow masking D_WARNING, D_ERROR messages from console
741 Details : Console messages can now be disabled via lnet.printk.
745 Description: User code with malformed file open parameter crashes client node
746 Details : Before packing join_file req, all the related reference should be
747 checked carefully in case some malformed flags cause fake
748 join_file req on client.
752 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
753 Details : ldlm_completion_ast() assumes that a lock is granted when the req
754 mode is equal to the granted mode. However, it should also check
755 that LDLM_FL_CP_REQD is not set.
759 Description: Heavy nfs access might result in deadlocks
760 Details : After ELC code landed, it is now improper to enqueue any mds
761 locks under och_sem, because enqueue might want to decide to
762 cancel open locks for same inode we are holding och_sem for.
766 Description: Client eviction while running blogbench
767 Details : A lot of unlink operations with concurrent I/O can lead to a
768 deadlock causing evictions. To address the problem, the number of
769 oustanding OST_DESTROY requests is now throttled to
770 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
771 callbacks are processed in priority.
775 Description: enable ACLs on MDS by default
776 Details : ACLs must be enabled on MDS by default.
779 Frequency : PPC/PPC64 only
781 Description: conflicts between asm-ppc64/types.h and lustre_types.h
782 Details : fix duplicated definitions between asm-ppc64/types.h and
783 lustre_types.h on PPC.
786 Frequency : PPC/PPC64 only
788 Description: asm-ppc/segment.h does not exist
789 Details : fix compile issue on PPC.
793 Description: better handle error messages in extents code
796 Frequency : RHEL4 only
798 Description: mkfs is very slow on IA64/RHEL4
799 Details : A performance regression has been discovered in the MPT Fusion
800 driver between versions 3.02.73rh and 3.02.99.00rh. As a
801 consequence, we have downgraded the MPT Fusion driver in the RHEL4
802 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
804 Severity : enhancement
806 Description: SNMP support enhancement
807 Details : Adding total number of sampled request for an MDS node in snmp
810 Severity : enhancement
812 Description: Optimize ldlm waiting list processing for PR extent locks
813 Details : When processing waiting list for read extent lock and meeting read
814 lock that is same or wider to it that is not contended, skip
815 processing rest of the list and immediatelly return current
816 status of conflictness, since we are guaranteed there are no
817 conflicting locks in the rest of the list.
821 Description: Time out and refuse to reconnect
822 Details : When the failover node is the primary node, it is possible
823 to have two identical connections in imp_conn_list. We must
824 compare not conn's pointers but NIDs, otherwise we can defeat
825 connection throttling.
829 Description: port llog fixes from b1_6 into HEAD
830 Details : Port llog reference couting and some llog cleanups from b1_6
831 (bug 10800) into HEAD, for protect from panic and access to already
832 free llog structures.
836 Description: Detect stride IO mode in read-ahead
837 Details : When a client does stride read, read-ahead should detect that and
838 read-ahead pages according to the detected stride pattern.
842 Description: data checksumming impacts single node performance
843 Details : add support for several checksum algorithm. Currently, only CRC32
844 and Adler-32 are supported. The checksum type can be changed on
845 the fly via /proc/fs/lustre/osc/*/checksum_type.
849 Description: use adler32 for page checksums
850 Details : when available, use the Adler-32 algorithm instead of CRC32 for
855 Description: build for x2 fails
856 Details : fix compile issue on Cray systems.
860 Description: Properly match for duplicate locks
861 Details : Due to different lock order from skiplists code, we need to
862 traverse entire list for now
865 Frequency : only on PPC/SLES10
867 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
868 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
869 be defined. Add a hack in mkfs_lustre.c to work around this compile
874 Description: LASSERT on MDS when client holding flock lock dies
875 Details : ldlm pool logic depends on number of granted locks equal to
876 number of released locks which is not true for flock locks, so
877 just exclude such locks from consideration.
881 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
882 Details : Use fsfilt_sync() for both the whole filesystem sync and
883 individual file sync to eliminate dangerous inode locking
884 with I_LOCK that can lead to a deadlock.
886 --------------------------------------------------------------------------------
888 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
890 * Support for kernels:
891 2.6.5-7.283 (SLES 9),
892 2.6.9-55.EL (RHEL 4),
893 2.6.16.46-0.14 (SLES 10),
894 2.6.18.8 vanilla (kernel.org)
895 * Client support for unpatched kernels:
896 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
897 2.6.16 - 2.6.22 vanilla (kernel.org)
898 * Due to recently discovered recovery problems, we do not recommend
899 using patchless RHEL 4 clients with this or any earlier release.
900 * Recommended e2fsprogs version: 1.39.cfs8
901 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
902 * Starting with this release, the ldiskfs backing filesystem required
903 by Lustre is now in its own package, lustre-ldiskfs. This package
904 should be installed. It is versioned separately from Lustre and
905 may be released separately in future.
909 Description: block reactivating mgc import until all deactivates complete
910 Details : Fix race when failing back MDT/MGS to itself (testing)
913 Frequency : at statup only
915 Description: mds_lov_synchronize race leads to various problems
916 Details : simultaneous MDT->OST connections at startup can cause the
917 sync to abort, leaving the OSC in a bad state.
919 Severity : enhancement
921 Description: add optional extra BUILD_VERSION info
922 Details : add a new environment variable (namely LUSTRE_VERS) which allows
923 to override the lustre version.
926 Frequency : 2.6.18 servers only
928 Description: ll_kern_mount() doesn't release the module reference
929 Details : The ldiskfs module reference count never drops down to 0
930 because ll_kern_mount() doesn't release the module reference.
935 Description: server LBUG when using old ost_num_threads parameter
936 Details : Accept the old ost_num_threads parameter but warn that it
937 is deprecated, and fix an off-by-one error that caused an LBUG.
942 Description: Transient SCSI error results in persistent IO issue
943 Details : iobuf->dr_error is not reinitialized to 0 between two
947 Frequency : sometimes when underlying device returns I/O errors
949 Description: OSTs not going read-only during write failures
950 Details : OSTs are not remounted read-only when the journal commit threads
951 get I/O errors because fsfilt_ext3 calls journal_start/stop()
952 instead of the ext3 wrappers.
956 Description: poor connect scaling with increasing client count
957 Details : Don't run filter_grant_sanity_check for more than 100 exports
958 to improve scaling for large numbers of clients.
961 Frequency : SLES10 only
963 Description: sanity-quota.sh quotacheck failed: rc = -22
964 Details : Quotas cannot be enabled on SLES10.
967 Frequency : liblustre clients only
969 Description: getdirentries does not give error when run on compute nodes
970 Details : getdirentries does not fail when the size specified as an argument
971 is too small to contain at least one entry
973 Severity : enhancement
975 Description: Add LNET router traceability for debug purposes
976 Details : If a checksum failure occurs with a router as part of the
977 IO path, the NID of the last router that forwarded the bulk data
978 is printed so it can be identified.
983 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
984 Details : Due to a race condition, liblustre clients were occasionally
987 Severity : enhancement
989 Description: lfs setstripe use optional parameters instead of postional
992 Severity : enhancement
994 Description: Nanosecond timestamp support for ldiskfs
995 Details : The on-disk ldiskfs filesystem has added support for nanosecond
996 resolution timestamps. There is not yet support for this at
997 the Lustre filesystem level.
1000 Frequency : during server recovery
1002 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1003 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1007 Frequency : nfs export on patchless client
1009 Description: connectathon hang when test nfs export over patchless client
1010 Details : Disconnected dentry cannot be found with lookup, so we do not need
1011 to unhash it or make it invalid
1014 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1015 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1016 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1021 Description: open req refcounting wrong on reconnect
1022 Details : If reconnect happened between getting open reply from server and
1023 call to mdc_set_replay_data in ll_file_open, we will schedule
1024 replay for unreferenced request that we are about to free.
1025 Subsequent close will crash in variety of ways.
1026 Check that request is still eligible for replay in
1027 mdc_set_replay_data().
1032 Description: disable writes to filesystem when reading health_check file
1033 Details : the default for reading the health_check proc file has changed
1034 to NOT do a journal transaction and write to disk, because this
1035 can cause reads of the /proc file to hang and block HA state
1036 checking on a healthy but otherwise heavily loaded system. It
1037 is possible to return to the previous behaviour during configure
1038 with --enable-health-write.
1040 Severity : enhancement
1042 Description: 64-bit inode version
1043 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1044 to the inode. This will be required for version-based recovery.
1049 Description: MDS fails to start if a duplicate client export is detected
1050 Details : in some rare cases it was possible for a client to connect to
1051 an MDS multiple times. Upon recovery the MDS would detect this
1052 and fail during startup. Handle this more gracefully.
1054 Severity : enhancement
1056 Description: Add -o localflock option to simulate old noflock
1058 Details : This will achieve local-only flock/fcntl locks
1064 Description: log_commit_thread vs filter_destroy race leads to crash
1065 Details : Take import reference before releasing llog record semaphore
1070 Description: Wrong request locking in request set processing
1071 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1072 imp_delayed_list, in this place should be used imp_lock.
1075 Frequency : when reconnection
1077 Description: Grant Leak when osc reconnect to OST
1078 Details : When osc reconnect ost, OST(filter) should check whether it
1079 should grant more space to client by comparing fed_grant and
1080 cl_avail_grant, and return the granted space to client instead
1081 of "new granted" space, because client will call osc_init_grant
1082 to update the client grant space info.
1085 Frequency : when client reconnect to OST
1087 Description: Grant Leak when osc do resend and replay bulk write
1088 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1089 bulk write request, because the grant info while be sync between
1090 OSC and OST when reconnect, and we should ignore the grant info
1091 these of resend/replay write req.
1096 Description: Grant space more than avaiable left space sometimes.
1097 Details : When then OST is about to be full, if two bulk writing from
1098 different clients came to OST. Accord the avaliable space of the
1099 OST, the first req should be permitted, and the second one
1100 should be denied by ENOSPC. But if the seconde arrived before
1101 the first one is commited. The OST might wrongly permit second
1102 writing, which will cause grant space > avaiable space.
1105 Frequency : when client is evicted
1107 Description: Grant might be wrongly erased when osc is evicted by OST
1108 Details : when the import is evicted by server, it will fork another
1109 thread ptlrpc_invalidate_import_thread to invalidate the
1110 import, where the grant will be set to 0. While the original
1111 thread will update the grant it got when connecting. So if
1112 the former happened latter, the grant will be wrongly errased
1113 because of this race.
1118 Description: Checking Stale with correct fid
1119 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1120 to check whether it is stale, because sometimes, we want the
1121 enqueue happened anyway, and op_data.fid2 will not be initialized.
1123 Severity : enhancement
1125 Description: update patchless client
1126 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1129 Frequency : only with 2.4 kernel
1131 Description: random memory corruption
1132 Details : size of struct ll_inode_info is to big for union inode.u and this
1133 can be cause of random memory corruption.
1138 Description: Memory leak in recovery
1139 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1140 It should also check obd_fail before fsfilt_start, otherwise if
1141 fsfilt_start return -EROFS,(failover mds during mds recovery).
1142 then the req will return with repmsg->transno = 0 and rc = EROFS.
1143 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1144 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1145 be freed no matter whether fsfilt_commit success or not.
1148 Frequency : only with huge count clients
1150 Description: Prevents from taking the superblock lock in llap_from_page for
1152 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1153 for llap_from_page call in ll_removepage prevents from taking the
1154 superblock lock for a soon died page.
1159 Description: Not check open intent error before release open handle
1160 Details : in some rare cases, the open intent error is not checked before
1161 release open handle, which may cause
1162 ASSERTION(open_req->rq_transno != 0), because it tries to release
1163 the failed open handle.
1168 Description: Set cat log bitmap only after create log success.
1169 Details : in some rare cases, the cat log bitmap is set too early. and it
1170 should be set only after create log success.
1174 Description: Accessing a block bevice can re-enable I/O when Lustre is
1175 tearing down a device.
1176 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1181 Description: service threads may hog cpus when there are a lot of requests
1182 Details : Insert cond_resched to give other threads a chance to use some CPU
1187 Description: the cat log was not initialized in recovery
1188 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1189 unlink log on mds will not be initialized until mds post
1190 recovery. And also in mds post recovery, the unlink log will
1191 initialization will be done asynchronausly, so there will be race
1192 between add unlink log and unlink log initialization.
1196 Description: brw_stats were being printed incorrectly
1197 Details : brw_stats were being printed as log2 but all of them were not
1198 recorded as log2. Also remove some code duplication arising from
1199 filter_tally_{read,write}.
1203 Frequency : rare, only in recovery.
1204 Description: ASSERTION(req->rq_type != LI_POISON) failed
1205 Details : imp_lock should be held while iterating over imp_sending_list for
1206 prevent destroy request after get timeout in ptlrpc_queue_wait.
1210 Description: replay-single.sh test 52 fails
1211 Details : A lock's skiplist need to be cleanup when it being unlinked
1212 from its resource list.
1216 Description: Short directio read returns full requested size rather than
1218 Details : Direct I/O operations should return actual amount of bytes
1219 transferred rather than requested size.
1221 Severity : enhancement
1223 Description: metadata RPC reduction (e.g. for rm performance)
1224 Details : decrease the amount of synchronous RPC between clients and servers
1225 by canceling conflicing lock before the operation on the client
1226 and packing thier handles into the main operation RPC to server.
1228 Severity : enhancement
1230 Description: Async OSC create to avoid the blocking unnecessarily.
1231 Details : If a OST has no remain object, system will block on the creating
1232 when need to create a new object on this OST. Now, ways use
1233 pre-created objects when available, instead of blocking on an
1234 empty osc while others are not empty. If we must block, we block
1235 for the shortest possible period of time.
1237 Severity : enhancement
1239 Description: refine locking for avoid write wrong info into lov_objid file
1240 Details : fix possible races with add new target and write/update data in
1243 --------------------------------------------------------------------------------
1245 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1250 Frequency : on some architectures
1252 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1253 Details : Uninitialized flags sometimes cause configuration commands to
1257 Frequency : patchless clients only
1259 Description: missing __iget() symbol export
1260 Details : The __iget() symbol export is missing. To avoid the need for
1261 this on patchless clients the deathrow inode reaper is turned
1262 off, and we depend on the VM to clean up old inodes. This
1263 dependency was during via the fix for bug 12181.
1265 --------------------------------------------------------------------------------
1267 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1269 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1270 INTEROPERATE with older versions automatically. In many cases a
1271 special upgrade step is needed. Please read the
1272 user documentation before upgrading any part of a live system.
1273 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1274 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1275 * WARNING: Lustre configuration and startup changes are required with
1276 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1282 Severity : enhancement
1284 Description: MountConf
1285 Details : Lustre configuration is now managed via mkfs and mount
1286 commands instead of lmc and lconf. New obd types (MGS, MGC)
1287 are added for dynamic configuration management. See
1288 https://mail.clusterfs.com/wikis/lustre/MountConf for
1291 Severity : enhancement
1293 Description: dynamic OST addition
1294 Details : OSTs can now be added to a live filesystem
1296 Severity : enhancement
1298 Description: startup order invariance
1299 Details : MDTs and OSTs can be started in any order. Clients only
1300 require the MDT to complete startup.
1302 Severity : enhancement
1304 Description: parallel, asynchronous orphan cleanup
1305 Details : orphan cleanup is now performed in separate threads for each
1306 OST, allowing parallel non-blocking operation.
1308 Severity : enhancement
1310 Description: optimized stripe assignment
1311 Details : stripe assignments are now made based on ost space available,
1312 ost previous usage, and OSS previous usage, in order to try
1313 to optimize storage space and networking resources.
1315 Severity : enhancement
1317 Description: Permanently set tunables
1318 Details : All writable /proc/fs/lustre tunables can now be permanently
1319 set on a per-server basis, at mkfs time or on a live system.
1321 Severity : enhancement
1323 Description: Lustre message v2
1324 Details : Add lustre message format v2.
1326 Severity : enhancement
1328 Description: client OST exclusion list
1329 Details : Clients can be started with a list of OSTs that should be
1330 declared "inactive" for known non-responsive OSTs.
1334 Description: SPEC SFS validation failure on NFS v2 over lustre.
1335 Details : Changes the blocksize for regular files to be 2x RPC size,
1336 and not depend on stripe size.
1338 Severity : enhancement
1340 Description: Multiple MD RPCs in flight.
1341 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1342 To avoid overly-overloading MDS, introduce a limit on number of
1343 MDS RPCs in flight for a single client and add /proc controls
1344 to adjust this limit.
1346 Severity : enhancement
1348 Description: client read/write statistics
1349 Details : Add client read/write call usage stats for performance
1350 analysis of user processes.
1351 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1352 file access. extents_stats shows chunk size distribution.
1353 extents_stats_per_process show chunk size distribution per
1356 Severity : enhancement
1358 Description: mds statistics
1359 Details : Add detailed mds operations statistics in
1360 /proc/fs/lustre/mds/*/stats.
1364 Description: Failure of copying files with lustre special EAs.
1365 Details : Client side always return success for setxattr call for lustre
1366 special xattr (currently only "trusted.lov").
1371 Description: Refcount LNET uuids
1372 Details : The global LNET uuid list grew linearly with every startup;
1373 refcount repeated list entries instead of always adding to
1376 Severity : enhancement
1378 Description: Dynamic service threads
1379 Details : Within a small range, start extra service threads
1380 automatically when the request queue builds up.
1383 Frequency : mixed-endian client/server environments
1385 Description: mixed-endian crashes
1386 Details : The new msg_v2 system had some failures in mixed-endian
1389 Severity : enhancement
1391 Description: Easy OST removal
1392 Details : OSTs can be permanently deactivated with e.g. 'lctl
1393 conf_param lustre-OST0001.osc.active=0'
1395 Severity : enhancement
1397 Description: MGS proc entries
1398 Details : Added basic proc entries for the MGS showing what filesystems
1401 Severity : enhancement
1403 Description: provide MGS failover
1404 Details : Added config lock reacquisition after MGS server failover.
1406 Severity : enhancement
1408 Description: add Linux 2.4 support
1409 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1413 Description: a large application tries to do I/O to the same resource and dies
1414 in the middle of it.
1415 Details : Check the req->rq_arrival time after the call to
1416 ost_brw_lock_get(), but before we do anything about
1417 processing it & sending the BULK transfer request. This
1418 should help move old stale pending locks off the queue as
1419 quickly as obd_timeout.
1422 Frequency : when an incorrect nid is specified during startup
1424 Description: ptlrpc connect to non-existant node causes kernel crash
1425 Details : LNET can't be re-entered from an event callback, which
1426 happened when we expire a message after the export has been
1427 cleaned up. Instead, hand the zombie cleanup off to another
1430 Severity : enhancement
1432 Description: plain/inodebits lock performance improvement
1433 Details : Grouping plain/inodebits in granted list by their request modes
1434 and bits policy, thus improving the performance of search through
1438 Frequency : only if OST filesystem is corrupted
1440 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1441 Details : for a short time RPCs with bulk IO are in the replay list,
1442 but replay of bulk IOs is unimplemented. If the OST filesystem
1443 is corrupted due to disk cache incoherency and then replay is
1444 started it is possible to trip an assertion. Avoid putting
1445 committed RPCs into the replay list at all to avoid this issue.
1448 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1450 Description: System hang on startup
1451 Details : This bug allowed the liblustre (e.g. catamount) client to
1452 return to the app before handling all startup RPCs. This
1453 could leave the node unresponsive to lustre network traffic
1454 and manifested as a server ptllnd timeout.
1456 Severity : enhancement
1458 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1459 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1460 boolean to control whether to print peer debug info when a
1461 client's RPC times out.
1464 Frequency : only for kernels with patches from Lustre below 1.4.3
1466 Description: Remove old rdonly API
1467 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1470 Frequency : only for devices with external journals
1472 Description: Set external device read-only also
1473 Details : During a commanded failover stop, we set the disk device
1474 read-only while the server shuts down. We now also set any
1475 external journal device read-only at the same time.
1478 Frequency : when upgrading from 1.4 while trying to change parameters
1480 Description: The wrong (new) MDC name was used when setting parameters for
1481 upgraded MDT's. Also allows changing of OSC (and MDC)
1482 parameters if --writeconf is specified at tunefs upgrade time.
1485 Frequency : when setting specific ost indicies
1487 Description: QOS code breaks on skipped indicies
1488 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1489 created with --index need not be sequential.
1491 Severity : enhancement
1493 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1494 Details : The uninit_groups feature works in conjunction with the kernel
1495 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1496 up the pass1 processing of e2fsck. This is a read-only feature
1497 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1498 mount filesystems that have had this feature enabled.
1500 Severity : enhancement
1502 Description: Improve multi-block allocation algorithm to avoid fragmentation
1503 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1504 allocation locality and avoid filesystem fragmentation.
1506 ------------------------------------------------------------------------------
1508 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1510 * Support for kernels:
1511 2.4.21-47.0.1.EL (RHEL 3)
1512 2.6.5-7.283 (SLES 9)
1513 2.6.9-42.0.10.EL (RHEL 4)
1514 2.6.12.6 vanilla (kernel.org)
1515 2.6.16.27-0.9 (SLES 10)
1516 * Recommended e2fsprogs version: 1.39.cfs5
1518 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1522 Frequency : occasional, depends on client load and configuration
1523 Bugzilla : 12181, 12203
1524 Description: data loss for recently-modified files
1526 Details : In some cases it is possible that recently written or created
1527 files may not be written to disk in a timely manner (this should
1528 normally be within 30s unless client IO load is very high).
1529 The problem appears as zero-length files or files that are a
1530 multiple of 1MB in size after a client crash or client eviction
1531 that are missing data at the end of the file.
1533 This problem is more likely to be hit on clients where files are
1534 repeatedly created and unlinked in the same directory, clients
1535 have a large amount of RAM, have many CPUs, the filesystem has
1536 many OSTs, the clients are rebooted frequently, and/or the files
1537 are not accessed by other nodes after being written.
1539 The presence of the problem can be detected by looking at
1540 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1541 smaller than the second (nr_unused) then dirty files will not
1542 be flushed automatically to disk. "sync; sleep 10" should be
1543 run several times on the node before unmounting it to update
1544 Lustre (this is also safe to run on nodes without this problem).
1546 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1547 that can cause this same problem, so customers using that kernel
1548 also need to update the kernel in addition to Lustre. In order
1549 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1551 It is normal that files written just before a client crash (less
1552 than 30s) may not yet have been flushed to disk, even for local
1556 Frequency : frequent on thin XT3 nodes
1558 Description: UUID collision on thin XT3 Linux nodes
1559 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1560 randomly, since we relied on an insufficiently-seeded PRNG.
1565 Description: OSS hangs after "All ost request buffers busy"
1566 Details : A deadlock between quota and journal operations caused OSS
1567 hangs after printing "All ost request buffers busy."
1570 Frequency : always on liblustre builds
1572 Description: Cleanup compiler warnings on liblustre
1575 Frequency : always on liblustre builds on XT3
1577 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1581 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1582 > 32000 subdirectories
1583 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1584 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1585 subdirectory count crosses 32000. This will aid e2fsck to
1586 correctly handle more than 32000 subdirectories.
1589 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1591 Description: System hang on startup
1592 Details : This bug allowed the liblustre (e.g. catamount) client to
1593 return to the app before handling all startup RPCs. This
1594 could leave the node unresponsive to lustre network traffic
1595 and manifested as a server ptllnd timeout.
1597 Severity : enhancement
1599 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1600 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1601 boolean to control whether to print peer debug info when a
1602 client's RPC times out.
1607 Description: make O_SYNC working on 2.6 kernels
1608 Details : 2.6 kernels use different method for mark pages for write,
1609 so need add a code to lustre for O_SYNC work.
1614 Description: Failure to close file and release space on NFS
1615 Details : Put inode details into lock acquired in ll_intent_file_open.
1616 Use mdc_intent_lock in ll_intent_open to properly
1617 detect all kind of errors unhandled by mdc_enqueue.
1622 Description: proc file read during shutdown sometimes raced obd removal,
1624 Details : Add lock to prevent obd access after proc file removal.
1627 Frequency : Only for files larger than 4GB on 32-bit clients.
1629 Description: improperly doing page alignment of locks
1630 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1631 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1636 Description: RPCs being resent when they shouldn't be
1637 Details : Some RPCs that should not be resent are being resent. This
1638 can cause inconsistencies in the RPC state machine. Do not
1639 resend such requests.
1642 Frequency : rare, only with NFS export
1644 Description: Crash on NFS re-export node
1645 Details : under very unusual load conditions an assertion is hit in
1646 ll_intent_file_open()
1649 Frequency : only if OST filesystem is corrupted
1651 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1652 Details : for a short time RPCs with bulk IO are in the replay list,
1653 but replay of bulk IOs is unimplemented. If the OST filesystem
1654 is corrupted due to disk cache incoherency and then replay is
1655 started it is possible to trip an assertion. Avoid putting
1656 committed RPCs into the replay list at all to avoid this issue.
1661 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1662 Details : Large single O_DIRECT read and write calls can fail to allocate
1663 a sufficiently large buffer to process the request. In case of
1664 allocation failure the allocation is retried with a smaller
1665 buffer and broken into smaller requests.
1667 Severity : enhancement
1669 Description: Add -o localflock option to simulate old noflock behaviour.
1670 Details : This will achieve local-only flock/fcntl locks coherentness.
1675 Description: versioning check is incomplete
1676 Details : Checking the version difference of client vs. server, report
1677 error if the gap is too big.
1682 Description: add support PG_writeback bit
1683 Details : add support for PG_writeback bit for Lustre, for more carefull
1684 work with page cache in 2.6 kernel. This also fix some deadlocks
1685 and remove hack for work O_SYNC with 2.6 kernel.
1687 Severity : enhancement
1689 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1690 Details : The uninit_groups feature works in conjunction with the kernel
1691 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1692 up the pass1 processing of e2fsck. This is a read-only feature
1693 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1694 mount filesystems that have had this feature enabled.
1696 Severity : enhancement
1698 Description: Improve multi-block allocation algorithm to avoid fragmentation
1699 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1700 allocation locality and avoid filesystem fragmentation.
1702 ------------------------------------------------------------------------------
1704 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1706 * Support for kernels:
1707 2.6.9-42.0.3.EL (RHEL 4)
1708 2.6.5-7.276 (SLES 9)
1709 2.4.21-47.0.1.EL (RHEL 3)
1710 2.6.12.6 vanilla (kernel.org)
1711 2.6.16.21-0.8 (SLES10)
1712 * Recommended e2fsprogs version: 1.39.cfs2-0
1714 * The backwards-compatible /proc/sys/portals symlink has been removed
1715 in this release. Before upgrading, please ensure that you change
1716 any configuration scripts or /etc/sysctl.conf files that access
1717 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1718 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1719 in advance of the upgrade on any system running Lustre 1.4.6 or
1720 newer, since /proc/sys/lnet was added in that version.
1721 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1725 Frequency : only when quota is used
1727 Description: avoid scanning export list for quota master
1728 Details : Change the algorithms to avoid scanning export list in order
1729 to improve the efficiency.
1732 Frequency : MDS failover only, very rarely
1734 Description: "went back in time" messages on mds failover
1735 Details : The greatest transno may be lost when the current operation
1736 finishes with an error (transno==0) and the client's last_rcvd
1737 record is over-written. Save the greatest transno in the
1738 mds_last_transno for this case.
1741 Frequency : always for specific kernels and striping counts
1743 Description: client may get "Matching packet too big" without ACL support
1744 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1745 when trying to access files in certain configurations. The
1746 clients should in fact be denied when mounting because they do
1747 not understand ACLs.
1750 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1752 Description: many clients connecting with IO in progress causes connect timeouts
1753 Details : Avoid synchronous journal commits to avoid delays caused by many
1754 clients connecting/disconnecting when bulk IO is in progress.
1755 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1756 OST_IO_PORTAL to avoid delays behind potentially many pending
1760 Frequency : occasionally with multiple writers to a single file
1762 Description: shared writes to file may result in wrong size reported by stat()
1763 Details : Allow growing of kms when extent lock is cancelled
1766 Frequency : always with random mmap IO to multi-striped file
1768 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1769 Details : Only if the hole is at the end of OST object so that kms is too
1770 small. Fix is to increase kms accordingly in ll_nopage.
1773 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1775 Description: writes to a missing object would leak memory on the OST
1776 Details : If there is an inconsistency between the MDS and OST filesystems,
1777 such that the MDS references an object that doesn't exist, writes
1778 to that object will leak memory due to incorrect cleanup in the
1779 error handling path, eventually running out of memory on the OST.
1784 Description: Creating too long symlink causes lustre errors
1785 Details : Check symlink and name lengths before sending requests to MDS.
1788 Frequency : only if flock is enabled (not on by default)
1790 Description: posix locks not released on fd closure on 2.6.9+
1791 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1792 kernels, this caused such locks not to be released on fd close and
1793 then assertions on fs unmount about still used locks.
1796 Frequency : MDS failover only, very rarely
1798 Description: clients may get ASSERTION(granted_lock != NULL)
1799 Details : When request was taking a long time, and a client was resending
1800 a getattr by name lock request. The were multiple lock requests
1801 with the same client lock handle and
1802 mds_getattr_name->fixup_handle_for_resent_request found one of the
1803 lock handles but later failed with ASSERTION(granted_lock != NULL).
1808 Description: handle->h_buffer_credits > 0, assertion failure
1809 Details : h_buffer_credits is zero after truncate, causing assertion
1810 failure. This patch extends the transaction or creates a new
1814 Frequency : NFS re-export or patchless client
1815 Bugzilla : 11179, 10796
1816 Description: Crash on NFS re-export node (__d_move)
1817 Details : We do not want to hash the dentry if we don't have a lock.
1818 But if this dentry is later used in d_move, we'd hit uninitialised
1819 list head d_hash, so we just do this to init d_hash field but
1820 leave dentry unhashed.
1823 Frequency : NFS re-export or patchless client
1825 Description: NFS exports has problem with symbolic link
1826 Details : lustre client didn't properly install dentry when re-exported
1827 to NFS or running patchless client.
1830 Frequency : NFS re-export or patchless client
1832 Description: Various nfs/patchless fixes.
1833 Details : fixes reuse disconected alias for lookup process - this fixes
1834 warning "find_exported_dentry: npd != pd",
1835 fix permission error with open files at nfs.
1836 fix apply umask when do revalidate.
1839 Frequency : occasional
1841 Description: Crash on NFS re-export node
1842 Details : calling clear_page() on the wrong pointer triggered oops in
1843 generic_mapping_read().
1846 Frequency : rarely, using O_DIRECT IO
1848 Description: unaligned directio crashes client with LASSERT
1849 Details : check for unaligned buffers before trying any requests.
1852 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1854 Description: stale data returned from RAID cache
1855 Details : If only a small amount of IO is done to the RAID device before
1856 reading it again it is possible to get stale data from the RAID
1857 cache instead of reading it from disk.
1860 Frequency : always for sles10 kernel
1862 Description: sles10 support
1863 Details : ll_follow_link: compile fixes and using of nd_set_link
1864 under newer kernels.
1867 Frequency : depends on arch, kernel and compiler version, always on sles10
1870 Description: recursive or deep enough symlinks cause stack overflow
1871 Details : getting rid of large stack-allocated variable in
1875 Frequency : depends on hardware
1877 Description: lustre write performance loss in the SLES10 kernel
1878 Details : the performance loss is caused by using of write barriers in the
1879 ext3 code. The SLES10 kernel turns barrier support on by
1880 default. The fix is to undo that change for ldiskfs.
1882 ------------------------------------------------------------------------------
1884 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1886 * Support for kernels:
1887 2.6.9-42.0.3EL (RHEL 4)
1888 2.6.5-7.276 (SLES 9)
1889 2.4.21-47.0.1.EL (RHEL 3)
1890 2.6.12.6 vanilla (kernel.org)
1894 Frequency : quota enabled and large files being deleted
1896 Description: releasing more than 4GB of quota at once hangs OST
1897 Details : If a user deletes more than 4GB of files on a single OST it
1898 will cause the OST to spin in an infinite loop. Release
1899 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1904 Description: statfs data retrieved from /proc may be stale or zero
1905 Details : When reading per-device statfs data from /proc, in the
1906 {kbytes,files}_{total,free,avail} files, it may appear
1907 as zero or be out of date.
1910 Frequency : systems with MD RAID1 external journal devices
1912 Description: lconf's call to blkid is confused by RAID1 journal devices
1913 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1914 of returning all block devices that match the journal UUID.
1917 Frequency : always, for aggregate stripe size over 4GB
1919 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1920 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1921 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1924 Frequency : always if "lfs find" used on a local file/directory
1926 Description: "lfs find" segfaults if used on a local file/directory
1927 Details : The case where a directory component was not specified wasn't
1928 handled correctly. Handle this properly.
1931 Frequency : always on ppc64
1933 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1934 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1935 perform endianness conversion.
1938 Frequency : rarely (truncate to non-zero file size after write under load)
1939 Bugzilla : 10730, 10687
1940 Description: Files padded with zeros to next 4K multiple
1941 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1942 it is possible that files that are truncated to a non-zero size
1943 immediately after being written are filled with zero bytes beyond
1944 the truncated size. No file data is lost.
1946 Severity : enhancement
1948 Description: Allow recovery/failover for liblustre clients.
1949 Details : liblustre clients were unaware of failover configurations until
1952 Severity : enhancement
1954 Description: user file locks should fail when not mounting with flock option
1955 Details : Set up an error-returning stub in ll_file_operations.lock field
1956 to prevent incorrect behaviour when client is mounted without
1957 flock option. Also, set up properly f_op->flock field for
1961 Frequency : always on ia64
1963 Description: "lfs df" loops on printing out MDS statfs information
1964 Details : The obd_ioctl_data was not initialized and in some systems
1965 this caused a failure during the ioctl that did not return
1966 an error. Initialize the struct and return an error on failure.
1969 Frequency : SLES 9 only
1971 Description: Error of copying files with lustre special EAs as root
1972 Details : Client side always return success for setxattr call for lustre
1973 special xattr (currently only "trusted.lov").
1976 Frequency : rarely on clusters with both ia64+i386 clients
1978 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1979 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1980 clients are doing concurrent writes to the same file it is
1981 possible that the ia64 clients may LASSERT because the OST
1982 extent locks are not PAGE_SIZE aligned. Ensure that grown
1983 locks are always aligned on the request boundary.
1986 Frequency : specific use, occasional
1988 Description: Overwriting in use executable truncates on-disk binary image
1989 Details : If one node attempts to overwrite an executable in use by
1990 another node, we now correctly return ETXTBSY instead of
1991 truncating the file.
1993 Severity : enhancement
1995 Description: Async OSC create to avoid the blocking unnecessarily.
1996 Details : If a OST has no remain object, system will block on the creating
1997 when need to create a new object on this OST. Now, ways use
1998 pre-created objects when available, instead of blocking on an
1999 empty osc while others are not empty. If we must block, we block
2000 for the shortest possible period of time.
2005 Description: chmod on Lustre root is propagated to other clients
2006 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2007 invalid by the follow_mount time.
2010 Frequency : liblustre clients only
2012 Description: Race in 'instant cancel' lock handling could lead to such locks
2013 never to be granted in case of SMP MDS
2014 Details : Do not destroy not yet granted but cbpending locks in
2018 Frequency : replay/resend of open
2020 Description: non null lock assetion failure in mds_intent_policy
2021 Details : Trying to replay/resend lockless open requests resulted in
2022 mds_open() returning 0 with no lock. Now it sets a flag if
2023 a lock is going to be returned.
2025 Severity : enhancement
2027 Description: Checksum enhancements
2028 Details : New checksum enhancements allow for resending RPCs that failed
2031 Severity : enhancement
2033 Description: Tunables on number of dirty pages in cacche
2034 Details : Allow to set limit on number of dirty pages cached.
2039 Description: client crash on unmount - lock still has references
2040 Details : In some error handling cases it was possible to leak a lock
2041 reference on a client while accessing a file. This was not
2042 harmful to the client during operation, but would cause the
2043 client to crash when the filesystem is unmounted.
2046 Frequency : specific case, rare
2048 Description: ETXTBSY on mds though file not in use by client
2049 Details : ETXTBSY is no longer incorrectly returned when attempting to
2050 chmod or chown a directory that the user previously tried to
2051 execute or a currently-executing binary.
2054 Frequency : extremely rare except on liblustre-based clients
2056 Description: Lustre space not freed when files are deleted
2057 Details : Clean up open-unlinked files after client eviction. Previously
2058 the unlink was skipped and the files remained as orphans.
2063 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2064 Details : In some cases it was possible to send a blocking callback to
2065 a client doing a glimpse, even though that client didn't get
2066 a lock granted. When the glimpse lock is cancelled on the OST
2067 the freed lock is left on the waiting list and corrupted the list.
2070 Frequency : all core dumps
2072 Description: Broke core dumps to lustre
2073 Details : Negative dentry may be unhashed if parent does not have UPDATE
2074 lock, but some callers, e.g. do_coredump, expect dentry to be
2075 hashed after successful create, hash it in ll_create_it.
2077 ------------------------------------------------------------------------------
2079 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2081 * Support for kernels:
2082 2.6.9-42.0.2.EL (RHEL 4)
2083 2.6.5-7.276 (SLES 9)
2084 2.4.21-40.EL (RHEL 3)
2085 2.6.12.6 vanilla (kernel.org)
2089 Frequency : always on RHEL 3
2091 Description: Number of open files grows over time
2092 Details : The number of open files grows over time, whether or not
2093 Lustre is started. This was due to a filp leak introduced
2094 by one of our kernel patches.
2096 ------------------------------------------------------------------------------
2098 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2100 * Support for kernels:
2101 2.6.9-42.EL (RHEL 4)
2102 2.6.5-7.276 (SLES 9)
2103 2.4.21-40.EL (RHEL 3)
2104 2.6.12.6 vanilla (kernel.org)
2109 Bugzilla : 5719, 9635, 9792, 9684
2110 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2111 Details : If a server is under heavy load and cannot reply to new
2112 connection requests before the client resends the (re)connect,
2113 the connection handling code can behave badly if two service
2114 threads are concurrently handing separate (re)connections from
2115 the same client. Add better locking to the connection handling
2116 code, and ensure that only a single connection will be processed
2117 for a given client UUID, even if the lock is dropped.
2119 Severity : enhancement
2121 Description: add TCP zero-copy support to kernel
2122 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2123 sends if the hardware supports scatter-gather and checksumming.
2124 This allows socklnd to do client-write and server-read more
2125 efficiently and reduce CPU utilization from skbuf copying.
2128 Frequency : only if NFS exporting from client
2130 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2131 Details : If setting attributes on a file created under NFS that had
2132 never been opened it would be possible to oops the client
2133 if the file had no objects.
2136 Frequency : always for liblustre
2138 Description: liblustre client does MDS+OSTs setattr RPC for each write
2139 Details : When doing a write from a liblustre client, the client
2140 incorrectly issued an RPC to the MDS and each OST the file was
2141 striped over in order to update the timestamps. When writing
2142 with small chunks and many clients this could overwhelm the MDS
2143 with RPCs. In all cases it would slow down the write because
2144 these RPCs are unnecessary.
2146 Severity : enhancement
2148 Description: allow number of MDS service threads to be changed at module load
2149 Details : It is now possible to change the number of MDS service threads
2150 running. Adding "options mds mds_num_threads={N}" to the MDS's
2151 /etc/modprobe.conf will set the number of threads for the next
2152 time Lustre is restarted (assuming the "mds" module is also
2153 reloaded at that time). The default number of threads will
2154 stay the same, 32 for most systems.
2159 Description: OST crash if filesystem is unformatted or corrupt
2160 Details : If an OST is started on a device that has never been formatted
2161 or if the filesystem is corrupt and cannot even mount then the
2162 error handling cleanup routines would dereference a NULL pointer.
2167 Description: NULL pointer deref in llap_from_page.
2168 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2169 incorrect) mapping if the page was truncated/reclaimed while it was
2170 searched for. Check for this condition and skip such pages when
2171 doing readahead. Introduce extra check to llap_from_page() to
2172 verify page->mapping->host is non-NULL (so page is not anonymous).
2175 Frequency : Sometimes when using sys_sendfile
2177 Description: "page not covered by a lock" warnings from ll_readpage
2178 Details : sendfile called ll_readpage without right page locks present.
2179 Now we introduced ll_file_sendfile that does necessary locking
2180 around call to generic_file_sendfile() much like we do in
2184 Frequency : with certain MDS communication failures at client mount time
2186 Description: NULL pointer deref after failed client mount
2187 Details : a client connection request may delayed by the network layer
2188 and not be sent until after the PTLRPC layer has timed out the
2189 request. If the client fails the mount immediately it will try
2190 to clean up before the network times out the request. Add a
2191 reference from the request import to the obd device and delay
2192 the cleanup until the network drops the request.
2195 Frequency : occasionally during client (re)connect
2197 Description: assertion failure during client (re)connect
2198 Details : processing a client connection request may be delayed by the
2199 client or server longer than the client connect timeout. This
2200 causes the client to resend the connection request. If the
2201 original connection request is replied in this interval, the
2202 client may trip an assertion failure in ptlrpc_connect_interpret()
2203 which thought it would be the only running connect process.
2206 Frequency : only with obd_echo servers and clients that are rebooted
2208 Description: kernel BUG accessing uninitialized data structure
2209 Details : When running an obd_echo server it did not start the ping_evictor
2210 thread, and when a client was evicted an uninitialized data
2211 structure was accessed. Start the ping_evictor in the RPC
2212 service startup instead of the OBD startup.
2214 Severity : enhancement
2215 Bugzilla : 10193 (patchless)
2216 Description: Remove dependency on various unexported kernel interfaces.
2217 Details : No longer need reparent_to_init, exit_mm, exit_files,
2218 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2221 Frequency : rare (only users of deprecated and unsupported LDAP config)
2223 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2224 Details : LDAP apparently contains 'lustreName' attributes instead of
2225 'name'. A simple remapping of the name is sufficient.
2228 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2230 Description: waiting_locks_callback trips kernel BUG if client is evicted
2231 Details : Running with the dump_on_timeout debug flag turned on makes
2232 it possible that the waiting_locks_callback() can try to dump
2233 the Lustre kernel debug logs from an interrupt handler. Defer
2234 this log dumping to the expired_lock_main() thread.
2236 Severity : enhancement
2238 Description: Support NFS exporting on 2.6 kernels.
2239 Details : Implement non-rawops metadata methods for NFS server to use without
2240 changing NFS server code.
2243 Frequency : very rare (synthetic metadata workload only)
2245 Description: two racing renames might cause an MDS thread to deadlock
2246 Details : Running the "racer" program may cause one MDS thread to rename
2247 a file from being the source of a rename to being the target of
2248 a rename at exactly the same time that another thread is doing
2249 so, and the second thread has already enqueued these locks after
2250 doing a lookup of the target and is trying to relock them in
2251 order. Ensure that we don't try to re-lock the same resource.
2254 Frequency : only very large systems with liblustre clients
2256 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2257 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2258 all OSTs in parallel. This allows the eviction of stale liblustre
2259 clients to proceed much faster than if they were done in series,
2260 and also offers similar improvements for other set_info RPCs.
2265 Description: excessive CPU usage during initial read phase on client
2266 Details : During the initial read phase on a client, it would agressively
2267 retry readahead on the file, consuming too much CPU and impacting
2268 performance (since 1.4.5.8). Improve the readahead algorithm
2269 to avoid this, and also improve some other common cases (read
2270 of small files in particular, where "small" is files smaller than
2271 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2276 Description: MDS crash when receiving packet with unknown intent.
2277 Details : Do not LBUG in unknown intent case, just return -EFAULT
2279 Severity : enhancement
2280 Bugzilla : 9293, 9385
2281 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2282 Details : Do not serialize getattr (non-intent version) and statfs.
2285 Frequency : occasional, when OST network is overloaded/intermittent
2287 Description: client evicted by OST after bulk IO timeout
2288 Details : If a client sends a bulk IO request (read or write) the OST
2289 may evict the client if it is unresposive to its data GET/PUT
2290 request. This is incorrect if the network is overloaded (takes
2291 too long to transfer the RPC data) or dropped the OST GET/PUT
2292 request. There is no need to evict the client at all, since
2293 the pinger and/or lock callbacks will handle this, and the
2294 client can restart the bulk request.
2297 Frequency : Always when mmapping file with no objects
2299 Description: client crashes when mmapping file with no objects
2300 Details : Check that we actually have objects in a file before doing any
2301 operations on objects in ll_vm_open, ll_vm_close and
2307 Description: Request leak when working with deleted CWD
2308 Details : Introduce advanced request refcount tracking for requests
2309 referenced from lustre intent.
2311 Severity : Enhancement
2313 Description: Cache open file handles on client.
2314 Details : MDS now will return special lock along with openhandle, if
2315 requested and client is allowed to hold openhandle, even if unused,
2316 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2317 closing files for every read/write openration.
2319 Severity : Enhancement
2321 Description: Cache open negative dentries on client when possible.
2322 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2323 negative dentries on lock revocation.
2328 Description: Remounting a client read-only wasn't possible with a zconf mount
2329 Details : It wasn't possible to remount a client read-only with llmount.
2331 Severity : enhancement
2332 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2333 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2334 MPICH is included to add Lustre-specific ADIO interfaces.
2335 This is based closely on the UFS ADIO layer and only differs
2336 in file creation, in order to allow the OST striping to be set.
2337 This is user-contributed code and not supported by CFS.
2342 Description: extended inode attributes (immutable, append-only) work improperly
2343 when 2.4 and 2.6 kernels are used on client/server or vice versa
2344 Details : Introduce kernel-independent values for these flags.
2346 Severity : enhancement
2349 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2350 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2351 granularity. Now, Fractional MB granularity is be supported,
2352 this is very useful for low memory system.
2354 Severity : enhancement
2356 Description: Getattr by fid
2357 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2358 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2359 have locks and avoids getattr rpc requests for it on every stat.
2362 Frequency : Always, for filesystems larger than 2TB
2364 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2365 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2366 filesystems larger than 2TB to be mounted with mballoc enabled.
2369 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2371 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2372 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2373 filesystems larger than 2TB is mounted then the kernel will
2374 silently corrupt the start of the filesystem. CONFIG_LBD is
2375 enabled for all CFS-supported kernels, but the possibility of
2376 this happening with a modified kernel config exists.
2378 Severity : enhancement
2380 Description: add client O_DIRECT support for 2.6 kernels
2381 Details : It is now possible to do O_DIRECT reads and writes to files
2382 in the Lustre client mountpoint on 2.6 kernel clients.
2384 Severity : enhancement
2386 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2387 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2388 parallel, not waiting for response from every OST before sending
2389 a rpc to the next OST.
2394 Description: setattr vs write race when updating file timestamps
2395 Details : Client processes that update a file timestamp into the past
2396 right after writing to the file (e.g. tar) it is possible that
2397 the updated file modification time can be reset to the current
2398 time due to a race between processing the setattr and write RPC.
2400 Severity : enhancement
2402 Description: Bring 'lfs find' closer in line with regular Linux find.
2403 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2404 -print0 options and obtains all the needed info through the lustre
2407 Severity : enhancement
2409 Description: support up to 1024 configured devices on one node
2410 Details : change obd_dev array from statically allocated to dynamically
2411 allocated structs as they are first used to reduce memory usage
2416 Description: Flush dirty partially truncated pages during truncate
2417 Details : Immediatelly flush partially truncated pages in filter_setattr,
2418 this way we completely avoid having any pages in page cache on OST
2419 and can retire ugly workarounds during writes to flush such pages.
2424 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2425 Details : protect inode from truncation within vfs_unlink() context
2426 just take a reference before calling vfs_unlink() and release it
2427 when parent's i_sem is free.
2432 Description: last_id value checked outside lock on OST caused LASSERT failure
2433 Details : If there were multiple MDS->OST object precreate requests in
2434 flight, it was possible that the OST's last object id was checked
2435 outside a lock and incorrectly tripped an assertion. Move checks
2436 inside locks, and discard old precreate requests.
2439 Frequency : always, if extents are used on OSTs
2441 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2442 Details : OSTs using the extents format would not zero the high 16 bits of
2443 the index physical block number. This is not a problem for any
2444 OST filesystems smaller than 16TB, and no kernels support ext3
2445 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2446 new/modified files) and can be fixed for existing filesystems
2447 with e2fsprogs-1.39-cfs1.
2452 Description: import connection selection may be incorrect if timer wraps
2453 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2454 import connections to be ignored if the 32-bit jiffies counter
2455 wraps. Use a 64-bit jiffies counter.
2458 Frequency : very large clusters immediately after boot
2460 Description: LNET request buffers exhausted under heavy short-term load
2461 Details : If a large number of client requests are generated on a service
2462 that has previously never seen so many requests it is possible
2463 that the request buffer growth cannot keep up with the spike in
2464 demand. Instead of dropping incoming requests, they are held in
2465 the LND until the RPC service can accept more requests.
2468 Frequency : Sometimes during replay
2470 Description: Assertion failure in ll_local_open after replay.
2471 Details : If replay happened on an open request reply before we were able
2472 to set replay handler, reply will become not swabbed tripping the
2473 assertion in ll_local_open. Now we set the handler right after
2474 recognising of open request
2477 Frequency : very rare
2479 Description: kernel reports "badness in vsnprintf"
2480 Details : Reading from the "recovery_status" /proc file in small chunks
2481 may cause a negative length in lprocfs_obd_rd_recovery_status()
2482 call to vsnprintf() (which is otherwise harmless). Exit early
2483 if there is no more space in the output buffer.
2485 Severity : enhancement
2487 Description: clear OBD RPC statistics by writing to them
2488 Details : It is now possible to clear the OBD RPC statistics by writing
2489 to the "stats" file.
2494 Description: Client mtime is not the same on different clients after utimes
2495 Details : In some cases, the client was using the utimes() syscall on
2496 a file cached on another node. The clients now validate the
2497 ctime from the MDS + OSTs to determine which one is right.
2502 Description: Inability to activate failout mode
2503 Details : lconf script incorrectly assumed that in pythong string's numeric
2504 value is used in comparisons.
2507 Frequency : always with multiple stripes per file
2509 Description: Inefficient object allocation for mutli-stripe files
2510 Details : When selecting which OSTs to stripe files over, for files with
2511 a stripe count that divides evenly into the number of OSTs,
2512 the MDS is always picking the same starting OST for each file.
2513 Return the OST selection heuristic to the original design.
2518 Description: mount failures may take full timeout to return an error
2519 Details : Under some heavy load conditions it is possible that a
2520 failed mount can wait for the full obd_timeout interval,
2521 possibly several minutes, before reporting an error.
2522 Instead return an error as soon as the status is known.
2524 Frequency : quota enabled and large files being deleted
2526 Description: releasing more than 4GB of quota at once hangs OST
2527 Details : If a user deletes more than 4GB of files on a single OST it
2528 will cause the OST to spin in an infinite loop. Release
2529 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2534 Description: statfs data retrieved from /proc may be stale or zero
2535 Details : When reading per-device statfs data from /proc, in the
2536 {kbytes,files}_{total,free,avail} files, it may appear
2537 as zero or be out of date.
2540 Frequency : systems with MD RAID1 external journal devices
2542 Description: lconf's call to blkid is confused by RAID1 journal devices
2543 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2544 of returning all block devices that match the journal UUID.
2547 Frequency : always, for aggregate stripe size over 4GB
2549 Description: assertion fails when trying to use 4GB stripe size
2550 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2551 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2554 Frequency : always on ppc64
2556 Description: the first write on an ext3 filesystem with mballoc got stuck
2557 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2558 perform endianness conversion.
2560 ------------------------------------------------------------------------------
2562 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2564 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2565 INTEROPERATE with older versions automatically. Please read the
2566 user documentation before upgrading any part of a live system.
2567 * WARNING: Lustre networking configuration changes are required with
2568 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2571 * Support for kernels:
2572 2.6.9-22.0.2.EL (RHEL 4)
2573 2.6.5-7.244 (SLES 9)
2574 2.6.12.6 vanilla (kernel.org)
2577 Severity : enhancement
2578 Bugzilla : 7981/8208
2579 Description: Introduced Lustre Networking (LNET)
2580 Details : LNET is new networking infrastructure for Lustre, it includes
2581 a reorganized network configuration mode (see the user
2582 documentation for full details) as well as support for routing
2583 between different network fabrics. Lustre Networking Devices
2584 (LNDs) for the supported network fabrics have also been
2585 created for this new infrastructure.
2587 Severity : enhancement
2588 Description: Introduced Access control lists
2589 Details : clients can set ACLs on files and directories in order to have
2590 more fine-grained permissions than the standard Unix UGO+RWX.
2591 The MDS must be started with the "-o acl" mount option.
2593 Severity : enhancement
2594 Description: Introduced filesystem quotas
2595 Details : Administrators may now establish per-user quotas on the
2598 Severity : enhancement
2600 Description: Configuration change for the XT3
2601 The PTLLND is now used to run Lustre over Portals on the XT3
2602 The configure option(s) --with-cray-portals are no longer used.
2603 Rather --with-portals=<path-to-portals-includes> is used to
2604 enable building on the XT3. In addition to enable XT3 specific
2605 features the option --enable-cray-xt3 must be used.
2610 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2611 Details : A race between allocating a new llcd and re-getting the llcd_lock
2612 allowed another thread to grab newly-allocated llcd.
2614 Severity : enhancement
2616 Description: 2.6 OST async journal commit and locking fix to improve performance
2617 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2618 2.6 kernels are now async as they already were in 2.4 kernels so
2619 that they can commit concurrently with the network bulk transfer.
2620 For block-allocated files the filter allocation semaphore is held
2621 to avoid filesystem fragmentation during allocation. BKL lock
2622 removed for 2.6 xattr operations where it is no longer needed.
2627 Description: lconf incorrectly determined whether two IP networks could talk
2628 Details : In some more complicated routing and multiple-network
2629 configurations, lconf will avoid trying to make a network
2630 connection to a disjoint part of the IP space. It was doing the
2631 math incorrectly for one set of cases.
2636 Description: Fix for potential infinite loop processing records in an llog.
2637 Details : If an llog record is corrupted/zeroed, it is possible to loop
2638 forever in llog_process(). Validate the llog record length
2639 and skip the remainder of the block on error.
2642 Frequency : occasional (liblustre only)
2644 Description: liblustre could not open files whose last component is a symlink
2645 Details : sysio_path_walk() would incorrectly pass the open intent to
2646 intermediate path components.
2649 Frequency : rare (liblustre only with non-standard tuning)
2650 Bugzilla : 7201 (7350)
2651 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2652 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2653 after it has been released, only do this for referenced locks
2655 Severity : enhancement
2657 Description: specifying an (invalid) directory default stripe_size of -1
2658 would reset the directory default striping
2659 Details : stripe_size -1 was used internally to signal directory stripe
2660 removal, now use "all default" to signal dir stripe removal
2661 as a directory striping of "all default" is not useful
2664 Frequency : common for large clusters running liblustre clients
2666 Description: doing an ls when liblustre clients are running is slow
2667 Details : sending a glimpse AST to a liblustre client waits for every AST
2668 to time out, as liblustre clients will not respond. Since they
2669 cannot cache data we refresh the OST lock LVB from disk instead.
2671 Severity : enhancement
2673 Description: doing an ls at the same time as file IO can be slow
2674 Details : enqueue and other "small" requests can be blocked behind many
2675 large IO requests. Create a new OST IO portal for non-IO
2676 requests so they can be processed faster.
2679 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2681 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2682 created file with mode 000
2683 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2684 allow an ftruncate() as a non-root user to a file with mode 000.
2685 Lustre now respects this flag to disable mode checks when
2686 truncating a file owned by the user
2689 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2692 Description: Revoking locks from clients that went dead or catatonic might take
2694 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2695 cancellation of such locks instant on servers without waiting for
2696 any reply from clients. Clients drops these locks when cancel
2697 notification from server is received without replying.
2700 Frequency : liblustre-only, when liblustre client dies or becomes busy
2702 Description: Doing ls on Linux clients can take a long time with active
2704 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2705 granting such locks to it in the first place if possible. Locks
2706 are taken by proxy on the OST during the read or write and
2707 dropped immediately afterward. Add connect flags handling, do
2708 not grant locks to liblustre clients for glimpse ASTs.
2710 Severity : enhancement
2712 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2713 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2714 sequential reads and behaves badly with applications doing random
2715 reads. Improve it by reading ahead at least read region, and
2716 avoiding excessive large RPC for small reads.
2718 Severity : enhancement
2720 Description: Creating more than 1000 files for a single job may cause a load
2721 imbalance on the OSTs if there are also a large number of OSTs.
2722 Details : qos_prep_create() uses an OST index reseed value that is an
2723 even multiple of the number of available OSTs so that if the
2724 reseed happens in the middle of the object allocation it will
2725 still utilize the OSTs as uniformly as possible.
2730 Description: OST or MDS may oops in ping_evictor_main()
2731 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2732 but doesn't restart at beginning of obd_exports_timed list
2735 Severity : enhancement
2737 Description: improve by-nid export eviction on the MDS and OST
2738 Details : allow multiple exports with the same NID to be evicted at one
2739 time without re-searching the exports list.
2742 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2744 Description: MDS may oops in groups_free()
2745 Details : in rare race conditions a newly allocated group_info struct is
2746 freed again, and this can be NULL. The 2.4 compatibility code
2747 for groups_free() checked for a NULL pointer, but 2.6 did not.
2750 Frequency : common for liblustre clients doing little filesystem IO
2751 Bugzilla : 9352, 7313
2752 Description: server may evict liblustre clients accessing contended locks
2753 Details : if a client is granted a lock or receives a completion AST
2754 with a blocking AST already set it would not reply to the AST
2755 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2758 Frequency : lfs setstripe, only systems with more than 160 OSTs
2760 Description: unable to set striping with a starting offset beyond OST 160
2761 Details : llapi_create_file() incorrectly limited the starting stripe
2762 index to the maximum single-file stripe count.
2765 Frequency : LDAP users only
2767 Description: lconf did not handle in-kernel recovery with LDAP properly
2768 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2770 Severity : enhancement
2772 Description: bind OST threads to NUMA nodes to improve performance
2773 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2774 node and do their allocations there to localize memory access
2776 Severity : enhancement
2778 Description: llmount can determine client NID directly from Myrinet (GM)
2779 Details : the client NID code from gmnalnid was moved directly into
2780 llmount, removing the need to use this or specifying the
2781 client NID explicitly when mounting GM clients with zeroconf
2784 Frequency : if client is started with down MDS
2786 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2787 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2788 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2789 ptlrpc_import_delayed_req() because we didn't check if the
2790 request was interrupted, and we also didn't break out of the
2791 event loop if there was no timeout
2796 Description: data loss during non-page-aligned writes to a single file from
2797 both multiple nodes and multiple threads on one node at same time
2798 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2799 inconsistency led to false short-reads, that were cached and later
2800 used by ->prepare_write() to fill in partially written page,
2801 leading to data loss.
2804 Frequency : always, if lconf --abort_recovery used
2806 Description: lconf --abort_recovery fails with 'Operation not supported'
2807 Details : lconf was attempting to abort recovery on the MDT device and not
2810 Severity : enhancement
2812 Description: remove cleanup logs
2813 Details : replace lconf-generated cleanup logs with lustre internal
2814 cleanup routines. Eliminates the need for client-cleanup and
2817 Severity : enhancement
2819 Description: add support for EAs (user and system) on lustre filesystems
2820 Details : it is now possible to store extended attributes in the Lustre
2821 client filesystem, and with the user_xattr mount option it
2822 is possible to allow users to store EAs on their files also
2824 Severity : enhancement
2826 Description: Add possibility (config option) to show minimal available OST free
2828 Details : When compiled with --enable-mindf configure option, statfs(2)
2829 (and so, df) will return least minimal free space available from
2830 all OSTs as amount of free space on FS, instead of summary of
2831 free spaces of all OSTs.
2833 Severity : enhancement
2835 Description: do not expand extent locks acquired on OST-side
2836 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2837 by server: they are not cached anyway.
2840 Frequency : when mmap is used/binaries executed from Lustre
2842 Description: Unmmap pages before throwing them away from read cache.
2843 Details : llap_shrink cache now attempts to unmap pages before discarding
2844 them (if unmapping failed - do not discard). SLES9 kernel has
2845 extra checks that trigger if this unmapping is not done first.
2850 Description: lconf didn't resolve symlinks before checking to see whether a
2851 given mountpoint was already in use
2854 Frequency : when migrating failover services
2855 Bugzilla : 6395, 9514
2856 Description: When migrating a subset of services from a node (e.g. failback
2857 from a failover service node) the remaining services would
2858 time out and evict clients.
2859 Details : lconf --force (implied by --failover) sets the global obd_timeout
2860 to 5 seconds in order to quickly disconnect, but this caused
2861 other RPCs to time out too quickly. Do not change the global
2862 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2864 Severity : enhancement
2865 Frequency : if MDS is started with down OST
2866 Bugzilla : 9439,5706
2867 Description: Allow startup/shutdown of an MDS without depending on the
2868 availability of the OSTs.
2869 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2870 Add appropriate locking and lov-osc refcounts for safe
2871 cleaning. Add osc abort_inflight calls in case the
2872 synchronize never started.
2875 Frequency : occasional (Cray XT3 only)
2877 Description: root not authorized to access files in CRAY_PORTALS environment
2878 Details : The client process capabilities were not honoured on the MDS in
2879 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2880 been accessed by an authorized user then root was able to access
2881 the file on the local client also. The root user capabilities
2882 are now allowed on the MDS, as this environment has secure UID.
2885 Frequency : occasional
2887 Description: ldiskfs "too long searching" message happens too often
2888 Details : A debugging message (otherwise harmless) prints too often on
2889 the OST console. This has been reduced to only happen when
2890 there are fragmentation problems on the filesystem.
2895 Description: Division by zero in statfs when all OSCs are inactive
2896 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2897 lov_statfs divides by value returned by lov_get_stripecnt().
2901 Bugzilla : 9489, 3273
2902 Description: First write from each client to each OST was only 4kB in size,
2903 to initialize client writeback cache, which caused sub-optimal
2904 RPCs and poor layout on disk for the first writen file.
2905 Details : Clients now request an initial cache grant at (re)connect time
2906 and so that they can start streaming writes to the cache right
2907 away and always do full-sized RPCs if there is enough data.
2908 If the OST is rebooted the client also re-establishes its grant
2909 so that client cached writes will be honoured under the grant.
2914 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2915 Details : Now I/O RPCs go to different portal number and (presumably) fast
2916 lock requests (and glimses) and other RPCs get their own service
2917 threads pool that should be able to service those RPCs
2920 Severity : enhancement
2922 Description: Ability to exchange lustre version between client and servers and
2923 issue warnings at client side if client is too old. Also for
2924 liblustre clients there is ability to refuse connection of too old
2926 Details : New 'version' field is added to connect data structure that is
2927 filled with version info. That info is later checked by server and
2931 Frequency : rare, liblustre only.
2932 Bugzilla : 9296, 9581
2933 Description: Two simultaneous writes from liblustre at offset within same page
2934 might proceed at the same time overwriting eachother with stale
2936 Details : I/O lock withing llu_file_prwv was released too early, before data
2937 actually was hitting the wire. Extended lock-holding time until
2938 server acknowledges receiving data.
2941 Frequency : extremely rare. Never observed in practice.
2943 Description: avoid generating lustre_handle cookie of 0.
2944 Details : class_handle_hash() generates handle cookies by incrementing
2945 global counter, and can hit 0 occasionaly (this is unlikely, but
2946 not impossible, because initial value of cookie counter is
2947 selected randonly). Value of 0 is used as a sentinel meaning
2948 "unassigned handle" --- avoid it. Also coalesce two critical
2949 sections in this function into one.
2951 Severity : enhancement
2953 Description: allow liblustre clients to delegate truncate locking to OST
2954 Details : To avoid overhead of locking, liblustre client instructs OST to
2955 take extent lock in ost_punch() on client's behalf. New connection
2956 flag is added to handle backward compatibility.
2958 Severity : enhancement
2959 Bugzilla : 4928, 7341, 9758
2960 Description: allow number of OST service threads to be specified
2961 Details : a module parameter allows the number of OST service threads
2962 to be specified via "options ost ost_num_threads={N}" in the
2963 OSS's /etc/modules.conf or /etc/modprobe.conf.
2967 Bugzilla : 6146, 9635, 9895
2968 Description: servers crash with bad pointer in target_handle_connect()
2969 Details : In rare cases when a client is reconnecting it was possible that
2970 the connection request was the last reference for that export.
2971 We would temporarily drop the export reference and get a new
2972 one, but this may have been the last reference and the export
2973 was just destroyed. Get new reference before dropping old one.
2975 Severity : enhancement
2976 Frequency : if client is started with failover MDS
2978 Description: Allow multiple MDS hostnames in the mount command
2979 Details : Try to read the configuration from all specified MDS
2980 hostnames during a client mount in case the "primary"
2983 Severity : enhancement
2985 Description: Stop sending data to evicted clients as soon as possible.
2986 Details : Check if the client we are about to send or are sending data to
2987 was evicted already. (Check is done every second of waiting,
2988 for which l_wait_event interface was extended to allow checking
2989 of exit condition at specified intervals).
2992 Frequency : rare, normally only when NFS exporting is done from client
2994 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2996 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2999 Frequency : always, if srand() is called before liblustre initialization
3001 Description: Liblustre uses system PRNG disturbing its usage by user application
3002 Details : Introduce internal to lustre fast and high-quality PRNG for
3003 lustre usage and make liblustre and some other places in generic
3004 lustre code to use it.
3006 Severity : enhancement
3007 Bugzilla : 9477, 9557, 9870
3008 Description: Verify that the MDS configuration logs are updated when xml is
3009 Details : Check if the .xml configuration logs are newer than the config
3010 logs stored on the MDS and report an error if this is the case.
3011 Request --write-conf, or allow starting with --old_conf.
3013 Severity : enhancement
3015 Description: Handle symlinks in the path when checking if Lustre is mounted.
3016 Details : Resolve intermediate symlinks when checking if a client has
3017 mounted a filesystem to avoid duplicate client mounts.
3022 Description: lconf can hit an error exception but still return success.
3023 Details : The lconf command catches the Command error exception at the top
3024 level script context and will exit with the associated exit
3025 status, but doesn't ensure that this exit status is non-zero.
3030 Description: failure of ptlrpc thread startup can cause oops
3031 Details : Starting a ptlrpc service thread can fail if there are a large
3032 number of threads or the server memory is very fragmented.
3033 Handle this without oopsing.
3036 Frequency : always, only if liblustre and non-default acceptor port was used
3038 Description: liblustre cannot connect to servers with non-default acceptor port
3039 Details : tcpnal_set_default_params() was not called and was therefore
3040 ignoring the environment varaible TCPNAL_PORT, as well as other
3041 TCPNAL_ environment variables
3046 Description: two objects could be created on the same OST for a single file
3047 Details : If an OST is down, in some cases it was possible to create two
3048 objects on a single OST for a single file. No problems other
3049 than potential performance impact and spurious error messages.
3053 Bugzilla : 5681, 9562
3054 Description: Client may oops in ll_unhash_aliases
3055 Details : Client dcache may become inconsistent in race condition.
3056 In some cases "getcwd" can fail if the current directory is
3062 Description: Inode refcounting problems in NFS export code
3063 Details : link_raw functions used to call d_instantiate without obtaining
3064 extra inode reference first.
3068 Bugzilla : 9942, 9903
3069 Description: Referencing freed requests leading to crash, memleaks with NFS.
3070 Details : We used to require that call to ll_revalidate_it was always
3071 followed by ll_lookup_it. Also with revalidate_special() it is
3072 possible to call ll_revalidate_it() twice for the same dentry
3073 even if first occurence returned success. This fix changes semantic
3074 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3075 reference on a request referred from the intent.
3076 ll_intent_release() then releases such a request.
3079 Frequency : rare, normally benchmark loads only
3081 Description: unlinked inodes were kept in memory on the client
3082 Details : If a client is repeatedly creating and unlinking files it
3083 can accumulate a lot of stale inodes in the inode slab cache.
3084 If there is no other client load running this can cause the
3085 client node to run out of memory. Instead flush old inodes
3086 from client cache that have the same inode number as a new inode.
3089 Frequency : SLES9 2.6.5 kernel and long filenames only
3090 Bugzilla : 9969, 10379
3091 Description: utime reports stale NFS file handle
3092 Details : SLES9 uses out-of-dentry names in some cases, which confused
3093 the lustre dentry revalidation. Change it to always use the
3097 Frequency : rare, unless heavy write-truncate concurrency is continuous
3098 Bugzilla : 4180, 6984, 7171, 9963, 9331
3099 Description: OST becomes very slow and/or deadlocked during object unlink
3100 Details : filter_destroy() was holding onto the parent directory lock
3101 while truncating+unlinking objects. For very large objects this
3102 may block other threads for a long time and slow overall OST
3103 responsiveness. It may also be possible to get a lock ordering
3104 deadlock in this case, or run out of journal credits because of
3105 the combined truncate+unlink. Solution is to do object truncate
3106 first in one transaction without parent lock, and then do the
3107 final unlink in a new transaction with the parent lock. This
3108 reduces the lock hold time dramatically.
3111 Frequency : rare, 2.4 kernels only
3113 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3114 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3115 before unmounting their filesystem in order to avoid deadlock.
3116 The kernel_locked() function in 2.4 kernels only checks whether
3117 the kernel lock is held, not whether it is this process that is
3118 holding it as 2.6 kernels do.
3123 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3124 Details : The client ptlrpc code may be trying to reconnect to a down
3125 server before a previous connection attempt has timed out.
3126 Increase the reconnect interval to be longer than the connection
3127 timeout interval to avoid sending duplicate connections to
3131 Frequency : echo_client brw_test command
3133 Description: fix echo_client to work with OST preallocated code
3134 Details : OST preallocation code (5137) didn't take echo_client IO path
3135 into account: echo_client calls filter methods outside of any
3136 OST thread and, hence, there is no per-thread preallocated
3137 pages and buffers to use. Solution: hijack pga pages for IO. As
3138 a byproduct, this avoids unnecessary data copying.
3142 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3143 Description: Client can oops in mdc_commit_close() after open replay
3144 Details : It was possible for the MDS to return an open request with no
3145 transaction number in mds_finish_transno() if the client was
3146 evicted, but without actually returning an error. Clients
3147 would later try to replay that open and may trip an assertion
3148 Simplify the client close codepath, and always return an error
3149 from the MDS in case the open is not successful.
3152 Frequency : rare, 2.6 OSTs only
3154 Description: OST may deadlock under high load on fragmented files
3155 Details : If there was a heavy load and highly-fragmented OST filesystems
3156 it was possible to have all the OST threads deadlock waiting on
3157 allocation of biovecs, because the biovecs were not released
3158 until the entire RPC IO was completed. Instead, release biovecs
3159 as soon as they are complete to ensure forward IO progress.
3161 Severity : enhancement
3163 Description: Support for specifying external journal device at mount
3164 Details : If an OST or MDS device is formatted with an external journal
3165 device, this device major/minor is stored in the ext3 superblock
3166 and may not be valid for failover. Allow detecting and
3167 specifying the external journal at mount time.
3172 Description: Mounting an MDS with pending unlinked files may cause oops
3173 Details : target_finish_recovery() calls mds_postrecov() which returned
3174 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3175 considers this an error and immediately begins cleaning up the
3176 lov, just after starting the mds_lov process
3178 Severity : enhancement
3180 Description: Implement 'lfs df' to report actual free space on per-OST basis
3181 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3182 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3183 usage of objects; '-h' to report in human readable format.
3185 ------------------------------------------------------------------------------
3187 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3194 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3195 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3196 at startup during orphan unlinking. Instead, skip buddy update
3197 before setup, as it will be regenerated after recovery is complete.
3202 Description: If an OST is inactive, its locks might reference stale inodes.
3203 Details : lov_change_cbdata() must iterate over all namespaces, even if
3204 they are inactive to clear inode references from the lock.
3206 Severity : enhancement
3207 Frequency : occasional, if non-standard max_dirty_mb used
3209 Description: Client will block write RPCs if not enough grant
3210 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3211 then the client will block writes while waiting for another RPC
3212 to complete instead of consuming its dirty limit. With change
3213 we get improved performance when max_dirty_mb is small.
3215 Severity : enhancement
3216 Bugzilla : 3389, 6253
3217 Description: Add support for supplementary groups on the MDS.
3218 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3219 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3220 lookups for user supplementary groups into a cache.
3224 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3225 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3226 for liblustre O_EXCL is enforced by MDS.
3230 Description: readdir never returns NULL in liblustre.
3231 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3232 dentry in struct dirent.
3236 Frequency : liblustre only, depends on application IO pattern
3237 Description: liblustre clients evicted if not contacting servers
3238 Details : Don't put liblustre clients into the ping_evictor list, so
3239 they will not be evicted by the pinger ever.
3241 Severity : enhancement
3243 Description: Add ability to evict clients by NID from MDS.
3244 Details : By echoing "nid:$NID" string into
3245 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3246 $NID would be instantly evicted from this MDS and from all active
3247 OSTs connected to it.
3251 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3252 Details : lookup_it_finish() used to query file size from OSTs that was not
3257 Description: service threads change working directory to that of init
3258 Details : Starting lustre service threads may pin the working directory
3259 of the parent thread, making that filesystem busy. Threads
3260 now change to the working directory of init to avoid this.
3264 Frequency : during shutdown only
3265 Description: shutdown with a failed MDS or OST can cause unmount to hang
3266 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3271 Frequency : 2.6 only
3272 Description: chmod/chown may include an extra supplementary group
3273 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3274 supplementary group and if none is specified this is used.
3277 Bugzilla : 5479 (6816)
3279 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3280 Details : If lookup is in progress on a file that is unlinked we might try
3281 to revalidate the inode and fail in revalidate after lookup is
3282 complete and ll_file_open() enqueues the open again but
3283 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3286 Frequency : always, if lconf --abort_recovery used
3288 Description: lconf --abort_recovery fails with 'Operation not supported'
3289 Details : lconf was attempting to abort recovery on the MDT device and not
3292 ------------------------------------------------------------------------------
3294 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3299 Frequency : rare (only unsupported configurations with a node running as an
3301 Bugzilla : 6514, 5137
3302 Description: Mounting a Lustre file system on a node running as an OST could
3304 Details : OSTs now preallocates memory needed to write out data at
3305 startup, instead of when needed, to avoid having to
3306 allocate memory in possibly low memory situations.
3307 Specifically, if the file system is mounted on on OST,
3308 memory pressure could force it to try to write out data,
3309 which it needed to allocate memory to do. Due to the low
3310 memory, it would be unable to do so and the node would
3311 become unresponsive.
3313 Severity : enhancement
3315 Description: Addition of lconf --service command line option
3316 Details : lconf now accepts a '--service <arg>' option, which is
3317 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3319 Severity : enhancement
3321 Description: Failover mode is now the default for OSTs.
3322 Details : By default, OSTs will now run in failover mode. To return to
3323 the old behaviour, add '--failout' to the lmc line for OSTs.
3325 Severity : enhancement
3327 Description: Health checks are now provided for MDS and OSTs
3328 Details : Additional detailed health check information on MSD and OSTs
3329 is now provided through the procfs health_check value.
3332 Frequency : occasional, depends on IO load
3334 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3335 after numerous create/delete cycles
3336 Details : The ext3 inode allocation policy would not allocate new inodes
3337 very well on the OSTs because there are no new directories
3338 being created. Instead we look for groups with free space if
3339 the parent directories are nearly full.
3344 Description: Network or server problems during mount may cause partially
3345 mounted clients instead of returning an error.
3346 Details : The config llog parsing code may overwrite the error return
3347 code during mount error handling, returning success instead
3353 Description: MDS can fail to allocate large reply buffers
3354 Details : After long uptimes the MDS can fail to allocate large reply
3355 buffers (e.g. zconf client mount config records) due to memory
3356 fragmentation or consumption by the buffer cache. Preallocate
3357 some large reply buffers so that these replies can be sent even
3358 under memory pressure.
3362 Frequency : rare (liblustre)
3363 Description: fsx running with liblustre complained that using truncate() to
3364 extend the file doesn't work. This patch corrects that issue.
3365 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3366 ATTR_SIZE and lsm use in llu_setattr_raw.
3370 Frequency : rare, only 2.6 kernels
3371 Description: Unusual file access patterns on the MDS may result in inode
3372 data being lost in very rare circumstances.
3373 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3374 code in the 2.6 kernel caused the inode and/or EA data not to
3375 be read from disk, causing single-file corruption.
3379 Frequency : rare, only 2.6 filesystems using extents
3380 Description: Heavy concurrent write and delete load may cause data corruption.
3381 Details : It was possible under high-load situations to have an extent
3382 metadata block in the block device cache from a just-unlinked
3383 file overwrite a newly-allocated data block. We now unmap any
3384 metadata buffers that alias just-allocated data blocks.
3388 Frequency : filesystems with default stripe_count larger than 77
3389 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3390 Details : lconf specifies an inode size of 4096 bytes when the default
3391 stripe_count is larger than 77. This conflicts with the default
3392 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3393 this case to avoid pinning too much memory for large EAs.
3395 ------------------------------------------------------------------------------
3397 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3402 Frequency : rare (extremely heavy IO load with hundreds of clients)
3404 Description: Client is evicted, gets IO error writing to file
3405 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3406 caused clients to be evicted for AST timeouts. The fixes in
3407 bug 5192 mean we no longer need to have such short AST timeouts
3408 so ldlm_timeout has been increased.
3411 Frequency : occasional during --force or --failover shutdown under load
3412 Bugzilla : 5949, 4834
3413 Description: Server oops/LBUG if stopped with --force or --failover under load
3414 Details : a collection of import/export refcount and cleanup ordering
3415 issues fixed for safer force cleanup
3418 Frequency : only filesystems larger than 120 OSTs
3419 Bugzilla : 5990, 6223
3420 Description: lfs getstripe would oops on a very large filesystem
3421 Details : lov_getconfig used kfree on vmalloc'd memory
3424 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3425 Bugzilla : 6242, 6243
3426 Description: reading from files that had been truncated to a non-zero size
3427 but never opened returned no data
3428 Details : ll_file_read() reads zeros from no-object files to EOF
3433 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3434 Details : There's little harm in aborting MDS/OSS recovery and letting it
3435 try again, so I removed the LASSERT and return an error instead.
3437 Severity : enhancement
3439 Description: New debugging infrastructure for tracking down data corruption
3440 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3441 (b) cover more of the client-side code path, and (c) try to narrow
3442 down where problems occurred
3446 Bugzilla : 3819, 4364, 4397, 6313
3447 Description: Racing close and eviction MDS could cause assertion in mds_close
3448 Details : It was possible to get multiple mfd references during close and
3449 client eviction, leading to one thread referencing a freed mfd.
3451 Severity: : enhancement
3452 Bugzilla : 3262, 6359
3453 Description: Attempts to reconnect to servers are now more aggressive.
3454 Details : This builds on the enhanced upcall-less recovery that was added
3455 in 1.4.2. When trying to reconnect to servers, clients will
3456 now try each server in the failover group every 10 seconds. By
3457 default, clients would previously try one server every 25 seconds.
3462 Description: After recovery, certain operations trigger a failed
3463 assertion on a client.
3464 Details : Failing over an mds, using lconf -d --failover, while a
3465 client was doing a readdir() call would cause the client to
3466 LBUG after recovery completed and the readdir() was resent.
3468 Severity : enhancement
3470 Description: Default groups are now added by lconf
3471 Details : You can now run lconf --group <servicename> without having to
3472 manually add groups with lmc.
3475 Frequency : occasional
3477 Description: Nodes with an elan id of 0 trigger a failed assertion
3480 Frequency : always when accessing e.g. tty/console device nodes
3482 Description: tty and some other devices nodes cannot be used on lustre
3483 Details : file's private_data field is used by device data and lustre
3484 values in there got lost. New field was added to struct file to
3485 store fs-specific private data.
3488 Frequency : when exporting Lustre via NFS
3490 Description: NFSD failed occasionally when looking up a path component
3491 Details : NFSD is looking up ".." which was broken in ext3 directories
3492 that had grown large enough to become hashed.
3495 Frequency : Clusters with multiple interfaces not on the same subnet
3497 Description: Nodes will repeatedly try to reconnect to an interface which it
3498 cannot reach and report an error to the log.
3499 Details : Extra peer list entries will be created by lconf with some peers
3500 unreachable. lconf now validates the peer before adding it.
3503 Frequency : Only if a default stripe is set on the filesystem root.
3505 Description: Setting a default stripe on the filesystem root prevented the
3506 filesystem from being remounted.
3507 Details : The client was sending extra request flags in the root getattr
3508 request and did not allocate a reply buffer for the dir EA.
3511 Frequency : occasional, higher if lots of files are accessed by one client
3512 Bugzilla : 6159, 6097
3513 Description: Client trips assertion regarding lsm mismatch/magic
3514 Details : While revalidating inodes the VFS looks up inodes with ifind()
3515 and in rare cases can find an inode that is being freed.
3516 The ll_test_inode() code will free the lsm during ifind()
3517 when it finds an existing inode and then the VFS later attaches
3518 this free lsm to a new inode.
3522 Bugzilla : 6422, 7030
3523 Description: MDS deadlock between mkdir and client eviction
3524 Details : Creating a new file via mkdir or mknod (starting a transaction
3525 and getting the ns lock) can deadlock with client eviction
3526 (gets ns lock and trying to finish a synchronous transaction).
3529 Frequency : occasional
3530 Description: While starting a server, the fsfilt_ext3 module could not be
3532 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3533 kernels. Previously, lconf would still use the ext3 name
3534 when trying to load modules. Now, it will correctly use
3535 ext3 on 2.4 and ldiskfs on 2.6.
3537 Severity : enhancement
3538 Description: The default stripe count has been changed to 1
3539 Details : The interpretation of the default stripe count (0, to lfs
3540 or lmc) has been changed to mean striping across a single
3541 OST, rather than all available. For general usage we have
3542 found a stripe count of 1 or 2 works best.
3544 Severity : enhancement
3545 Description: Add support for compiling against Cray portals.
3546 Details : Conditional compiling for some areas that are different
3550 Frequency : occasional
3551 Bugzilla : 6409, 6834
3552 Description: Creating files with an explicit stripe count may lead to
3553 a failed assertion on the MDS
3554 Details : If some OSTs are full or unavailable, creating files may
3555 trigger a failed assertion on the MDS. Now, Lustre will
3556 try to use other servers or return an error to the
3560 Frequency : occasional
3562 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3563 caused lock timeout during readahead (since 1.4.2).
3564 Details : Processes doing readahead might match a lock that hasn't been
3565 granted yet if there are overlapping and conflicting lock
3566 requests. The readahead process waits on ungranted lock
3567 (original lock is CBPENDING), while OST waits for that process
3568 to cancel CBPENDING read lock and eventually evicts client.
3570 Severity : enhancement
3572 Description: Initial enabling of flock support for clients
3573 Details : Implements fcntl advisory locking and file status functions.
3574 This feature is provided as an optional mount flag (default
3575 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3576 locking are implemented yet, and those that are are not guaranteed
3577 to be completely correct in production environments.
3578 mount -t lustre -o [flock|noflock] ...
3581 Frequency : occasional
3583 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3584 assertion in the ext3 JBD (journaling) layer.
3585 Details : The b_committed_data struct is protected by the big kernel lock
3586 in 2.4 kernels, serializing journal_commit_transaction() and
3587 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3588 access to this struct. In 2.6 kernels there is finer grained
3589 locking to improve SMP performance of the JBD layer.
3593 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3598 Description: merge and cleanup kernel patches.
3601 -----------------------------------------------------------------------------
3603 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3605 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3606 versions of Lustre. Please update all servers and clients to
3607 version 1.4.2 or later at the same time. You must also run
3608 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3609 to update the configuration logs.
3611 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3612 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3613 - swab llog records of type '0' so we get proper header size/idx (5861)
3614 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3615 - fix rename of one directory over another leaking an inode (5953)
3616 - avoid SetPageDirty on 2.6 (5981)
3617 - don't re-add just-being-destroyed locks to the waiting list (5653)
3618 - when creating new directories, inherit the parent's custom
3619 striping settings if present parent (3048)
3620 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3621 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3622 - don't hold client locks on temporary worklist from l_lru (5666)
3623 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3624 - automatically evict dead clients (5921)
3625 - Update file size properly in create+truncate+fstat case (6196)
3626 - Do not unhash mountpoint dentries, do not allow removal of
3628 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3629 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3630 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3631 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3632 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3633 - Correctly handle reads of files with no objects (6243)
3634 - lctl recover will also mark a device active if deactivate used (5933)
3636 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3637 - allow --write-conf on an MDS with different nettype than client (5619)
3638 - don't write config llogs to MDS for mounts not from that MDS (5617)
3639 - lconf should create multiple TCP connections from a client (5201)
3640 - init scripts are now turned off by default; run chkconfig --on
3641 lustre and chkconfig --on lustrefs to use them
3642 - upcalls are no longer needed for clients to recover to failover
3644 - add --abort-recovery option to lconf to abort recovery on device
3646 - add support for an arbitrary number of OSTs (3026)
3647 - Quota support protocol changes.
3648 - forward compatibility changes to wire structs (6007)
3649 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3650 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3651 - improved "lustre status" script
3652 - initialize blocksize for non-regular files (6062)
3653 - added --disable-server and --disable-client configure options (5782)
3654 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3655 - Vanilla 2.4.29 support
3656 - increase maximum number of obd devices to 520 (6242)
3657 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3658 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3659 SLES 9 2.6.5 kernels
3660 - limit stripes per file to 160 (the maximum EA size) (6093)
3662 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3665 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3666 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3667 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3668 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3669 - small ext3 extents cleanups and fixes (5733)
3670 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3671 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3672 - avoid needless client->OST connect, fix handle mismatch (5317)
3673 - fix DLM error path that led to out-of-sync client, long delays (5779)
3674 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3675 - fix several locking issues related to i_size (5492,5624,5654,5672)
3676 - don't move pending lock onto export if it is already evicted (5683)
3677 - fix kernel oops when creating .foo in unlinked directory (5548)
3678 - fix deadlock in obdfilter statistics vs. object create (5811)
3679 - use time_{before,after} to avoid timer jiffies wrap (5882)
3680 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3681 - Do not leak request if server was not able to process it (5154)
3682 - If mds_open unable to find parent dir, make that negative lookup(5154)
3683 - don't create new directories with extent-mapping (5909, 5936)
3685 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3686 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3687 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3688 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3689 - support for RHEL4 kernel on i686 (5773)
3690 - provide error messages when incompatible logs are encountered (5898)
3692 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3693 * version 1.4.0.10 (1.4.1 release candidate 1)
3695 - don't keep a lock reference when lock is not granted (4238)
3696 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3697 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3698 - drop import inflight refcount on signal_completed_replay error (5255)
3699 - unlock page after async write error during send (3677)
3700 - handle missing objects in filter_preprw_read properly (5265)
3701 - no transno return for symlink open, don't save no-trasno open (3440)
3702 - don't try to complete elan receive that already failed (4012)
3703 - free RPC server reply state on error (5406)
3704 - clean up thread from ptlrpc_start_thread() on error (5160)
3705 - readahead could read extra page into cache that wasn't ejected (5388)
3706 - prevent races in class_attach/setup/cleanup/detach (5260)
3707 - don't dereference de->d_inode after l_dput of de (5458)
3708 - use "int" for stripe value returned from lock_to_stripe (5544)
3709 - mballoc allocation and error-checking fixes in 2.6 (5504)
3710 - block device patches to fix I/O request sizes in 2.6 (5482)
3711 - look up hostnames for IB nals (5602)
3712 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3713 - don't start multiple acceptors for the same port (5277)
3714 - fix incorrect LASSERT in mds_getattr_name (5635)
3715 - export a proc file for general "ping" checking (5628)
3716 - fix "lfs check" to not block when the MDS is down (5628)
3718 - service request history (4965)
3719 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3720 - create an "evict_client" /proc entry on OSTs, like the MDS has
3721 - fix mount usage message, return errors per mount(8) (5168)
3722 - change grep [] to grep "[]" in tests so they work in more UMLs
3723 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3724 - remove ancient LOV_MAGIC_V0 EA support (5047)
3725 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3726 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3727 - allow readahead allocations to fail when low on memory (5383)
3728 - mmap locking landed again, after considerable improvement (2828)
3729 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3731 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3734 - send OST transaction number in read/write reply to free req (4966)
3735 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3736 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3737 - the watchdog thread now runs as interruptible (5246)
3738 - flock/lockf fixes (but it's still disabled, pending 5135)
3739 - don't use EXT3 constants in llite code (5094)
3740 - memory shortage at startup could cause assertion (5176)
3742 - reorganization of lov code
3743 - single portals codebase
3745 - add extents/mballoc support (5025)
3746 - direct I/O reads in the obdfilter (4048)
3747 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3749 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3752 - send OST transaction number in read/write reply to free req (4966)
3753 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3754 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3755 - the watchdog thread now runs as interruptible (5246)
3756 - handle missing objects in filter_preprw_read properly (5265)
3757 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3758 - drop import inflight refcount on signal_completed_replay error (5255)
3759 - unlock page after async write error during send (3677)
3760 - return original error code on reconstructed replies (3761)
3761 - no transno return for symlink open, don't save no-trasno open (3440)
3763 - add pid to ldlm debugging output (4922)
3764 - bump the watchdog timeouts -- we can't handle 30sec yet
3765 - extra debugging for orphan dentry/inode bug (5259)
3767 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3770 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3771 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3772 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3773 - fix NULL dereference in /proc/sys/portals/routes (4827)
3774 - allow failed mdc_close() operations to be interrupted (4561)
3775 - stop precreate on OST before MDS would time out on it (4778)
3776 - don't send partial-page writes before EOF from client (4410)
3777 - discard client grant for sub-page writes on large-page clients (4520)
3778 - don't free dentries not owned by NFS code, check generation (4806)
3779 - fix lsm leak if mds_create_objects() fails (4801)
3780 - limit debug_daemon file size, always print CERROR messages (4789)
3781 - use transno after validating reply (3892)
3782 - process timed out requests if import state changes (3754)
3783 - update mtime on OST during writes, return in glimpse (4829)
3784 - add mkfsoptions to LDAP (4679)
3785 - use ->max_readahead method instead of zapping global ra (5039)
3786 - don't interrupt __l_wait_event() during strace
3788 - add software watchdogs to catch hung threads quickly (4941)
3789 - make lustrefs init script start after nfs is mounted
3790 - fix CWARN/ERROR duplication (4930)
3791 - return async write errors to application if possible (2248)
3792 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3793 - print NAL number in %x format (4645)
3794 - update barely-supported suse-2.4.21-171 series (4842)
3795 - support for sles 9 %post scripts
3796 - support for building 2.6 kernel-source packages
3797 - support for sles km_* packages
3799 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3802 - ignore -ENOENT errors in osc_destroy (3639)
3803 - notify osc create thread that OSC is being cleaned up (4600)
3804 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3805 - reconstruct ost_handle() like mds_handle() (4657)
3806 - create a new thread to do import eviction to avoid deadlock (3969)
3807 - let lconf resolve symlinked-to devices (4629)
3808 - don't unlink "objects" from directory with default EA (4554)
3809 - hold socknal file ref over connect in case target is down (4394)
3810 - allow more than 32000 subdirectories in a single directory (3244)
3811 - fix blocks count for O_DIRECT writes (3751)
3812 - OST returns ENOSPC from object create when no space left (4539)
3813 - don't send truncate RPC if file size isn't changing (4410)
3814 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3815 - bind to privileged port in socknal and tcpnal (3689)
3817 - rate limit CERROR/CWARN console message to avoid overload (4519)
3818 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3819 - basic mmap support (3918)
3820 - kernel patch series update from b1_4 (4711)
3822 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3825 - avoid crash during MDS cleanup with OST shut down (2775)
3826 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3827 - don't use bad inodes on the MDS (3744)
3828 - dynamic object preallocation to improve recovery speed (4236)
3829 - don't hold spinlock over lock dumping or change debug flags (4401)
3830 - don't zero obd_dev when it is force cleaned (3651)
3831 - print grants to console if they go negative (4431)
3832 - "lctl deactivate" will stop automatic recovery attempts (3406)
3833 - look for existing locks in ldlm_handle_enqueue() (3764)
3834 - don't resolve lock handle twice in recovery avoiding race (4401)
3835 - revalidate should check working dir is a directory (4134)
3837 - don't always mark "slow" obdfilter messages as errors (4418)
3839 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3842 - don't close LustreDB during write_conf until it is done (3860)
3843 - fix typo in lconf for_each_profile (3821)
3844 - allow dumping logs from multiple threads at one time (3820)
3845 - don't allow multiple threads in OSC recovery (3812)
3846 - fix debug_size parameters (3864)
3847 - fix mds_postrecov to initialize import for llog ctxt (3121)
3848 - replace config semaphore with spinlock (3306)
3849 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3850 - don't allow enqueue to complete on a destroyed export (3822)
3851 - down write_lock before checking llog header bitmap (3825)
3852 - recover from lock replay timeout (3764)
3853 - up llog sem before sending rpc (3652)
3854 - reduce ns lock hold times when setting kms (3267)
3855 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3856 - fix NULL deref and obd_dev leak on setup error (3312)
3857 - replace some LBUG about llog ops with error handling (3841)
3858 - don't match INVALID dentries from d_lookup and spin (3784)
3859 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3860 - fix invalid assertion in ptlrpc_set_wait (3880)
3862 - add libwrap support for the TCP acceptor (3996)
3863 - add /proc/sys/portals/routes for non-root route listing (3994)
3864 - allow setting MDS UUID in .xml (2580)
3865 - print the stack of a process that LBUGs (4228)
3867 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3870 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3871 - make sure to unset replay flag from failed open requests (3440)
3872 - if default stripe count is 0, use OST count for inode size (3636)
3873 - update parent mtime/ctime on client for create/unlink (2611)
3874 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3875 - free recovery state on server during a forced cleanup (3571)
3876 - unregister_reply for resent reqs (3063)
3877 - loop back devices mounting and status check on 2.6 (3563)
3878 - fix resource-creation race that can provoke i_size == 0 (3513)
3879 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3880 - more debugging for page-accounting assertion (3746)
3881 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3882 - avoid deadlock after precreation failure (3758)
3883 - fix race and lock order deadlock in orphan handling (3450, 3750)
3884 - add validity checks when grabbing inodes from l_ast_data (3599)
3886 - add /proc/.../recovery_status to obdfilter (3428)
3887 - lightweight CDEBUG infrastructure, debug daemon (3668)
3888 - change default OSC RPC parameters to be better on small clusters
3889 - turn off OST read cache for files smaller than 32MB
3890 - install man pages and include them in rpms (3100)
3891 - add new init script for (un)mounting lustre filesystems (2593)
3892 - run chkconfig in %post for init scripts (3701)
3893 - drop scimac NAL (unmaintained)
3895 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3898 - clean kiobufs before and after use (3485)
3899 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3900 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3901 - change init script to fail more clearly if not run as root (1528)
3902 - allow clients to reconnect during replay (1742)
3903 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3904 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3905 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3906 - fix cond_resched (3554)
3907 - only evict unfinished clients after recovery (3515)
3908 - allow bulk resend, prevent data loss (3570)
3909 - dynamic ptlrpc request buffer allocation (2102)
3910 - don't allow unlinking open directory if it isn't empty (2904)
3911 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3912 - remove extraneous obd dereference causing LASSERT failure (3334)
3913 - don't use get_cycles() when creating temp. files on the mds (3156)
3914 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3915 - handle EEXIST for set-stripe, set proper directory name (3336)
3917 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3918 - fix ksocknal_fmb_callback() error messages (2918)
3920 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3923 - don't copy lvb into (possibly NULL) reply on error (2983)
3924 - don't deref dentry after dput, don't free lvb on error (2922)
3925 - use the kms to determine writeback rpc length (2947)
3926 - increment oti_logcookies when osc is inactive (2948)
3927 - update client's i_blocks count via lvb messages (2543)
3928 - handle intent open/close of special files properly (1557)
3929 - mount MDS with errors=remount-ro, like obdfilter (2009)
3930 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3931 - don't use cancelling-locks' kms values (2947)
3932 - use highest lock extent for kms, not last one (2925)
3933 - don't dereference ERR_PTR() dentry in error handling path (3107)
3934 - fix thread race in portals_debug_dumplog() (3122)
3935 - create lprocfs device entries at setup instead of at attach (1519)
3936 - common AST error handler, don't evict client on completion race (3145)
3937 - zero nameidata in detach_mnt in 2.6 (3118)
3938 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3939 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3940 - initialize RPC timeout timer earlier for 2.6 (3219)
3941 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3942 - print nal/nid for unknown nid (3258)
3943 - additional checks for oscc recovery before doing precreate (3284)
3944 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3945 - don't crash in mdc_close for bad permissions on open (3285)
3946 - zero i_rdev for non-device files (3147)
3947 - clear page->private before handing to FS, better assertion (3119)
3948 - tune the read pipeline (3236)
3949 - fix incorrect decref of invalidated dentry (2350)
3950 - provide read-ahead stats and refine rpc in flight stats (3328)
3951 - don't hold journal transaction open across create RPC (3313)
3952 - update atime on MDS at close time (3265)
3953 - close LDAP connection when recovering to avoid server load (3315)
3954 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3955 - don't leak open file on MDS after open resend (3325)
3956 - serialize filter_precreate and filter_destroy_precreated (3329)
3957 - loop device shouldn't call sync_dev() for nul device (3092)
3958 - clear page cache after eviction (2766)
3959 - resynchronize MDS->OST in background (2824)
3960 - refuse to mount the same filesystem twice on same mountpoint (3394)
3961 - allow llmount to create routes for mounting behind routers (3320)
3962 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3963 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3964 - fix rare problem with rename on htree directories (3417)
3966 - allow default OST striping configuration per directory (1414)
3967 - fix compilation for qswnal for 2.6 kernels (3125)
3968 - increase maximum number of MDS request buffers for large systems
3969 - change liblustreapi to be useful for external progs like lfsck (3098)
3970 - increase local configuration timeout for slow disks (3353)
3971 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3973 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3976 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3977 - don't overwrite extent policy data in reply if lock was blocked (2901)
3978 - drop filter export grants atomically with removal from device (2663)
3979 - del obd_self_export from work_list in class_disconnect_exports (2908)
3980 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3981 - swab reply message in mdc_close, other PPC fixes (2464)
3982 - fix destroying of named logs (2325)
3983 - overwrite old logs when running lconf --write_conf (2264)
3984 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3985 - fix race in target_handle_connect (2898)
3986 - mds_reint_create() should take same inode create lock (2926)
3987 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3988 - don't close files for self_export to avoid uninitialized obd (2936)
3989 - allow MDS with the same name as client node (2939)
3990 - hold dentry reference for closed log files for unlink (2325)
3991 - reserve space for all logs during transactions (2059)
3992 - don't evict page beyond end of stripe extent (2925)
3993 - don't oops on a deleted current working directory (2399)
3994 - handle hard links to targets without a parent properly (2517)
3995 - don't dereference NULL lock when racing during eviction (2867)
3996 - don't grow lock extents when lots of conflicting locks (2919)
3998 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4001 - account for cache space usage on clients to avoid data loss (974)
4002 - lfsck support in lustre kernel code (2349)
4003 - reduce journal credits needed for BRW writes (2370)
4004 - orphan handling to avoid losing space on client/server crashes
4005 - ptlrpcd can be blocked, stopping ALL progress (2477)
4006 - use lock value blocks to assist in proper KMS, faster stat (1021)
4007 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4008 - recovery for initial connections (2355)
4009 - fixes for mds_cleanup_orphans (1934)
4010 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4011 - block all file creations until orphan recovery completes (1901)
4012 - client remove rq_connection from request struct (2423)
4013 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4014 - recovery timer race (2670)
4015 - mdc_close recovey bug (2532)
4016 - ptlrpc cleanup bug (2710)
4017 - mds timeout on local locks (2588)
4018 - namespace lock held during RPCs (2431)
4019 - handle interrupted sync write properly (2503)
4020 - don't try to handle a message that hasn't been replied to (2699)
4021 - client assert failure during cleanup after abort recovery (2701)
4022 - leak mdc device after failed mount (2712)
4023 - ptlrpc_check_set allows timedout requests to complete (2714)
4024 - wait for inflight reqs when ptlrpcd finishes (2710)
4025 - make sure unregistered services are removed from the srv_list
4026 - reset bulk XID's when resending them (caught by 1138 test)
4027 - unregister_bulk after timeout
4028 - fix lconf error (2694)
4029 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4030 - readahead locks pages, leaves pending causing memory pressure (2673)
4031 - increase OST request buffers to 4096 on large machines (2729)
4032 - fix up permission of existing directories in simple_mkdir (2661)
4033 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4034 - don't assign transno to errored transactions (2742)
4035 - don't delete objects on OST if given a bogus objid from MDS (2751)
4036 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4037 - if rq_no_resend, then timeout request after recovery (2432)
4038 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4039 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4040 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4041 - MDS recovery shouldn't delete objects that it has given out (2730)
4042 - if enqueue arrives after completion, don't clobber LVB (2819)
4043 - don't unlock pages twice when trigger_group_io returns error (2814)
4044 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4045 - don't write pages to disk if there was an error (1450)
4046 - don't ping imports that have recovery disabled (2676)
4047 - take buffered bytes into account when balancing socknal conn (2817)
4048 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4049 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4050 - remove little-endian swabbing of llog records (1987)
4051 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4052 - retry reposting request buffers if they fail (1191)
4053 - grow extent at grant time to avoid granting a revoked lock (2809)
4054 - lock revoke doesn't evict page if covered by a second lock (2765)
4055 - disable VM readahead to avoid reading outside lock extents (2805)
4057 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4058 - updated kernel patches for hp-2.4.20 kernel (2681)
4060 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4063 - fix truncated write corruption (2366)
4064 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4066 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4067 - ensure len(uuid) < 37 in lmc (1171)
4068 - fix ia64 OOPS in llog_test (2255)
4069 - zero end of page at obdfilter for partial page writes (2648)
4070 - don't leave stale dentries around after renames (bug 2428)
4071 - fix timeouts when evicting a client with a single lock held (2642)
4072 - set deadline for the initial HELLO message to drain (2634)
4073 - print out dotted-quad IP addresses in the socknal (2302)
4075 - additional debugging for MDS client eviction problem (2443)
4076 - fix mkfsoptions support for osts (2603, 2604)
4078 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4081 - add series for the vanilla 2.6.0 kernel
4082 - add series for the vanilla 2.4.24 kernel
4083 - add series for a cray x86/64 UL kernel drop
4084 - fix xattr patches for the vanilla 2.4.19 series
4086 - generate true UUIDs in lmc (1171)
4087 - have portals stack dumping break in UML (2466)
4088 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4089 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4090 - wake callback waiting threads on client eviction (2460)
4091 - Add --ptldebug and --subsystem to lmc (1719)
4092 - update assertion to allow safe interrupt allocation
4093 - set rq_no_resend for cancel requests (2432)
4094 - recalculate ptlrpcd timeout after resend (2494)
4095 - call vfs_rmdir when removing pending directories (2368)
4096 - fix renaming a file to itself (2429)
4097 - lmc creates a default one-stripe lov (2454)
4098 - expand procfs space to handle large clusters (2326)
4099 - increase UML stack to avoid overflow
4100 - update lconf's list of debug and subsystem masks
4101 - fix lfs find --obd (2510)
4102 - /proc tunable for disabling filter read caching (2591)
4103 - stop rpm packages from altering slapd.conf (2301)
4104 - disable nagle in the socknal under 0conf (2578)
4105 - choose mds inode size based on stripe count (2572)
4106 - fix kernel-source rpm problems (2516)
4108 - add --disable-doc to avoid pdf generation (2421)
4109 - update documentation, tests, type-os, comments
4110 - avoid format warnings on ia64
4111 - remove the TOE NAL
4112 - tiny code cleanups by removing unused fields
4114 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4117 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4118 - remove the most blatant lies from BUILDING file (2371)
4119 - change default debug level to reasonable production setting
4120 - reduce client side cache size to reduce cache flush time
4121 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4122 - make TCP zerocopy and pinger support enabled by default (2476)
4123 - sync writes completed after process exits caused crashes (2319)
4124 - maintain correct mount count on the MDS (2356)
4125 - backout 1557, because 2316 wasn't really fixed
4126 - better file I/O statistics gathering in /proc
4127 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4128 - another kernel patch to fix zero-copy TCP function export
4129 - don't take duplicate lock when processing re-sent getattr (2420)
4130 - lctl uses obd_self_export instead of creating new conn (2353)
4131 - MDS/OST recovery case which requires object creation asserted (2425)
4132 - move lfs from /usr/sbin to /usr/bin in packages
4133 - fix race between mds_client_add and mds_client_free (2417)
4134 - use kmalloc instead of slabs in portals (2430)
4135 - don't create duplicate records when a failover MDS is present (2442)
4136 - remove unnecessary mount age check (2332)
4137 - don't remove directory inodes from locks prematurely (2451)
4138 - don't break if MDS service name is the same as hostname (2103)
4139 - fix races in client write RPC generation when cache full (2482)
4141 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4144 - remove now-unused request->rq_obd (278)
4145 - if an allocation fails, print out how much memory we've used (1933)
4146 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4147 - add the "configurable stack size" patch to most series files (1256)
4148 - ability to write large log records, for 100+ OST configs (2306)
4149 - fix NULL deref when filter_prep fails (2314)
4150 - fix operator precedence error in filter_sync
4151 - dynamic allocation of socknal TX descriptors (2315)
4152 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4153 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4154 - ensure that CWARN messages in Portals always get to the syslog
4155 - __init/__exit are not for prototype decls (ldlm_init/exit)
4156 - x86-64 compile warning fixes
4157 - fix gateway LMC keyword conflict (2318)
4158 - fix MDS lock inversions in getattr/reint paths (1844)
4159 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4160 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4161 - DLM race condition prevented some lock evictions (2328)
4162 - ENOMEM detection and retry on socknal sends (2230)
4163 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4164 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4165 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4166 - better error messages when a client is rejected during recovery (1505)
4167 - avoid cancelling locks which were never granted, after failure (2330)
4168 - fix i_sem/journal inversion in mds_client_add (2333)
4169 - fix truncate/getattr lock cycle deadlock (2334)
4170 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4171 - fix two rare exit paths which could leak an l_lock() ref (2321)
4172 - fencepost error in MDS/OST orphan recovery (2226)
4173 - make log record alignment 8 bytes (1988)
4174 - lstripe now fails when requested offset > ost_count (2237)
4175 - ensure that all kernel series have a complete list.h (1607)
4176 - fix crashes in special-file operations (2316)
4177 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4179 - allow configurable automake binary, for testing new versions
4180 - small update to the lfs documentation
4182 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4184 * fix negative export reference count in fsfilt_sync (2312)
4186 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4187 * release candidate 0.9.1
4189 - orphans are moved into the PENDING directory for possible recovery
4190 - replayed opens now open by fid for orphan/rename safety (1042)
4191 - last close of an orphan inode generates a transno (683)
4192 - chdir() and mount() now pin the directory entry (1020)
4193 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4194 - discard very old requests without processing them (1502)
4195 - remove some common, well-understood CERRORs (1505)
4196 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4197 - clear "grant" flags in OST replies until OST grant code lands (1644)
4198 - fix read performance by not clobbering i_blksize on client (1598)
4199 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4200 - make LRU size a /proc tunable, clears locks when reduced (707)
4201 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4202 - server locks take references on exports now (1558)
4203 - build fixes for 2.4.20-rh trees (1663)
4204 - return an error from lov_create if all OSCs are inactive (1751)
4205 - fix import levels when a reconnect happens without a timeout (1597)
4206 - exit early from mds_open if we get a lookup error (1749)
4207 - partial page read at EOF wouldn't wait for disk before sending (1642)
4208 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4209 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4210 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4211 - write/truncate lock inversion (1639)
4212 - Don't auto-load obdclass, portals modules during cleanup (1495)
4213 - fix timestamps from jumping to "now" (1763)
4214 - extra journal assertions (1648)
4215 - add an extra multiunlink test (1771)
4216 - fix read_record/write_record API (1776)
4217 - fix leak of offset_extent, possible incorrect i_size later (1772)
4218 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4219 - Debugging for the kqswnal_get_idle_tx problems (1820)
4220 - Allow recovery to be attempted multiple times (1536)
4221 - Write out MDS last_rcvd file after it is first created (1600)
4222 - Fix tx_descriptor leak in failed transmit situations (1827)
4223 - ext3 journaling fixes for assertion failure after IO error (1871)
4224 - class_export_put() on freed export after completion AST error (1896)
4225 - Fix revalidate looping in VFS (1322)
4226 - Don't access a freed export during MDS_REINT timeout (1521)
4227 - Add open-unlink recovery support on the MDS (1673,1764)
4228 - Return an error if no MDS data was read from last_rcvd (1946)
4229 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4230 - Don't setup a disk device that doesn't match exported UUID (317)
4231 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4232 - avoid committing NULL handle in force close
4233 - local.sh is now a one-stripe LOV configuration
4234 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4235 - fix ext3 htree duplicate directory entry corruption (1516)
4236 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4237 - update to new LOV EA format (2097)
4238 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4240 2003-06-15 Phil Schwan <phil@clusterfs.com>
4243 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4244 - per-import/export recovery handling (958, 931, 959)
4245 - multiple last-rcvd slots, for serving multiple FSes (949)
4246 - connections are again shared between multiple imp/exports (963, 964)
4247 - "umount -f" would hang if any requests needed to be sent (393, 978)
4248 - avoid pinning large req buffer by copying for queued messages (989)
4249 - add "uuid" to "lctl device" command to help upcalls (991)
4250 - "open" RPCs with transnos would confuse recovery counters (1037)
4251 - do proper endian conversion of all wire messages (288, 340, 891)
4252 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4253 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4254 - fix ptlrpc_connection leak in target_handle_connect (1174)
4255 - fix import refcounting bug in OST and MDS cleanup (1134)
4256 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4257 - fix very unlikely obd_types race condition (501)
4258 - remove osc_open hack for echo_client (1187)
4259 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4260 - a failure in read_inode2 leads to deadlock (1139)
4261 - cancel ack-locks as soon as transaction is committed (1072)
4262 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4263 - make sure to commitrw after any preprw to avoid deadlock (1162)
4264 - failing to execute a file in a lustre FS would lock inode (1203)
4265 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4266 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4267 - obd_uuid2tgt would walk off the end of the list (1255)
4268 - on IA64 the osc would give portals incorrect bulk size (1258)
4269 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4270 - fix lock inversion caused by new llite matching code (1282)
4271 - limit the number of dirty pages on a client to 10MB (1286)
4272 - timed out locks were not being corrected cancelled (1289)
4273 - fix O_DIRECT above 4GB on IA-32 (1292)
4274 * major user-visible changes
4275 - fail out/fail over policy now controlled by the upcall (993)
4277 - add OBD_PING to check server availability and failure (954)
4278 - lustre messages are now sent in sending host order (288, 340, 891)
4279 - add eadatalen to MDS getattr reply (340)
4280 - OST read replies may contain second buffer, with per-page status (593)
4282 2003-03-11 Phil Schwan <phil@clusterfs.com>
4285 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4286 - failed open()s could cause deadlock; fixed (867, 869)
4287 - stop cancelling OST locks when files are closed (481)
4288 - overlapping XID spaces caused network corruption (851, 853)
4289 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4290 - setattr_raw added, to do single-RPC, server-side setattrs
4291 - lmc/lconf syntax change for OST UUIDs
4292 - fix crashy race condition between ptlrpc_free_req and osc_close
4293 - don't use request in mdc_enqueue if we hit a timeout (889)
4294 - don't set the inode i_size for regular files from the MDS (896)
4295 - handle out of order completion AST (842)
4296 - don't LBUG if a lock request times out after receiving AST (913)
4297 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4298 - if a bad lock AST arrives, send an error instead of dropping entirely
4299 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4300 - fix leak in bulk IO when only partially completed (899, 900, 926)
4301 - fix O_DIRECT for ia64 (55)
4302 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4303 - C-z after timeout could hang a process forever; fixed (977)
4305 - client-side I/O cache (678, 924, 929, 941, 970)
4307 - READPAGE and SETATTRs which don't take server-side locks get
4310 2003-02-11 Phil Schwan <phil@clusterfs.com>
4313 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4314 - Fix fsfilt_extN_readpage() to read a full page of directory
4315 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4316 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4317 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4318 - Most intent operations were removed, in favour of a new RPC mode
4319 that does a single RPC to the server and bypasses most of the VFS
4320 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4321 - Aggressively cancel local locks on DLM servers
4322 - mds_reint_unlink sends EA to the client if it's the last nlink.
4323 client uses that EA to unlink OST objects.
4324 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4325 - recursive symlinks were fixed (439)
4326 - fixed NULL deref in DEBUG_REQ
4327 - filter_update_lastobjid no longer calls sync, which annoyed extN
4328 - fixed multi-client small-writes to a single file problem (445)
4329 - fixed mtime updates during file writes (607)
4330 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4331 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4332 - fixed Linux OST size reporting problem (444, 656)
4333 - OST now updates object mtime with writes or setattr (607, 619)
4334 - client verifies file size before zeroing page past EOF (445)
4335 - OST now writes last allocated objid to disk with allocation (108)
4336 - LOV on echo now works (409)
4338 - mds_reint_unlink sends a new buffer, with the EA included. this
4339 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4340 set if a regular file was being unlinked, and it was the last link
4341 - use PtlGet from the target for bulk writes (315)
4342 - OST now updates object mtime with writes or setattr (607, 619)
4343 - LDLM now has a grant-time callback to revalidate locked items, if
4345 - Many MDS operations were reorganized to combat race conditions
4347 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4348 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4350 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4353 - Fully reactivate OST imports after reconnection (512, others)
4354 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4355 - More graceful error handling for truncating on dead OST (515)
4356 - Don't error out unless we're actually accessing dead stripes (474)
4357 - Fix garbage sizes when stripes are missing (410)
4358 - LRU counters were broken, causing constant lock purge (433, 432)
4359 - garbage on read from stripes with failed OSTs (441)
4360 - mark OSCs as active before reconnecting during recovery (438)
4361 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4362 - lfind did not preserve OST order in output (443)
4363 - symlinks cause hung clients, incorrect data (439)
4364 - stop dereferencing request after dropping refcount (457)
4365 - don't LASSERT(spin_is_locked) on non-SMP (455)
4366 - fixes for many rename() bugs
4367 - fstat didn't correctly synchronize attributes (399)
4368 - server must handle lock cancellation during blocking AST prep (487)
4369 - bulk descriptors were free()d too soon (511)
4370 - fix paths in lconf, which would load incorrect modules (451, 507)
4371 - fix confusing lconf 'host not found' error message (386)
4372 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4373 - fix race condition in mdc_blocking_ast() for inode access (526)
4374 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4375 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4376 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4377 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4378 - fix lstripe in several ways (kernel side) (527)
4379 - fix request leak in ldlm_cli_enqueue (262)
4380 - incorrect OSC was marked inactive after OST failure
4381 - call mds_fs_cleanup before unmounting filesystem (524)
4382 - fix races between taking ns_lock and ldlm_lock_change_resource
4383 - fix races updating LOV export open file list
4384 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4385 - fix recovery NULL deref in ldlm_cli_cancel_unused
4386 - fix some DLM races by using new hash table for lock handles (419)
4387 - permit the client to specify desired inodes, at replay
4388 - duplicate requests when we queue them for replay reintegration
4389 - fix last_rcvd offset calculation
4390 - sync after each recovered transaction, so we always make progress
4391 - never, not always, ERESTART requests without transnos
4392 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4394 - skip replay if the MDS says that the client is already connected
4395 - don't check for a recovery-enabled export to match lctl's UUID
4396 - don't INC_USE_COUNT for phantom exports
4397 - don't crash when cleaning up phantom exports (567)
4398 - don't double-finish or set replay data for errored mdc_open requests
4399 - abort requests when they time out, so we don't get old replies
4400 - send/receive replies for AST messages again
4401 - if the client says that it doesn't have the lock, cancel it on the
4403 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4404 mark it as destroyed, it will all work out when decref is called
4405 - fix module use counts (22, 581)
4407 - ASTs now expect a reply (server cancels lock on error reply)
4409 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4412 - fix many simultaneous client startup (392)
4413 - fix dentry->d_it clobbering
4414 - credentials weren't being shipped for readdir/getattr operations
4415 - remove invalid assertions triggered during some concurrent MD
4417 - proper Lustre versions added (336, 389)
4418 - fix memory leak for create error case (398)
4419 - fix LOV locking bug that would get cli/srv out of sync
4420 - fix echo client over LOV (409)
4421 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4422 - fix double-O_EXCL intent crash (424)
4423 - avoid sending multiple lock CANCELs (352)
4425 - MDS can do multi-client recovery (modulo bugs in new code)
4427 - many updates, edits, cleanups
4429 2002-11-18 Phil Schwan <phil@clusterfs.com>
4432 - fix null d_it dereference (346)
4433 - fix full OST/dbench hang (333)
4434 - fix permission problem with file removal (286)
4435 - fix removal of OSCs from LOV when they fail
4436 - fix NULL deref during bulk timeout (214)
4437 - fix problems related to multiple filesystems on one MDS (241)
4438 - fixed serious subtle metadata locking bugs
4439 - free locks on clients when inodes are removed due to memory
4441 - fix inode pointer in lock data (285)
4442 - partial support for multiple MDS on a single host (241)
4443 - data locks weren't cancelled at clear_inode time (290, 311)
4444 - intent locks could lead to unbounded lock growth (205)
4445 - added a maximum lock count, an LRU list, and a flusher
4446 - fix multiple rename (365)
4447 - properly abstracted the echo client
4448 - OSC locked 1 byte too many; fixed
4449 - rewrote brw callback code:
4450 - fixed recovery bugs related to LOVs (306)
4451 - fixed too-many-pages-in-one-write crash (191)
4452 - fixed (again) crash in sync_io_timeout (214)
4453 - probably fixed callback-related race (385)
4455 - Add capability to MDS protocol
4456 - LDLM cancellations and callbacks on different portals
4458 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4461 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4462 - defer open object destruction to close time (601981, 138)
4463 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4464 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4465 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4466 - don't call obd_setattr in ll_file_release for destroyed objects
4467 * protocol change to lustre_msg: move |version| and add |flags|
4468 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4469 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4470 * added replay of create, unlink, link and rename operations during
4471 MDS failover; recovery should be much more robust now
4472 * remove failed OSCs from LOVs (only lov_create uses this so far)
4473 * the lustre-HOWTO was brought (more) up to date (582544)
4475 2002-10-23 Phil Schwan <phil@clusterfs.com>
4478 - in-use dentries weren't being reused properly (617851)
4479 - prevent multiple LDLM setup (599178)
4480 - fix LOV size calculations for truncate (617853)
4481 - fix client handling of MDS intent errors (POSIX)
4482 - fix permission bug in lovstripe.c test (624321)
4483 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4484 - truncate past end of file could corrupt data
4485 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4486 - a race in recovery could return ETIMEDOUT to apps (623947)
4487 - building outside the source directory was fixed
4488 * the lustre-HOWTO was brought (more) up to date (582544)
4489 * major progress was made on recovery functionality
4491 2002-10-10 Phil Schwan <phil@clusterfs.com>
4494 - recovery deadlock fix
4495 - rm -rf causes LBUG fix (617817)
4496 - file open by multiple tasks fix (618962)
4497 - directory permissions bugs (602707 and 620007)
4498 - journal_stop fixed with locking (611313)
4499 - O_APPEND failures resolved (618273, perhaps 614459)
4500 - lconf PATH fix (619770)
4501 - IA64 build fix (621450)
4502 - RPC buffer sizes scale with amount of memory
4504 2002-10-01 Phil Schwan <phil@clusterfs.com>
4507 - locks would be cancelled without throwing away data pages,
4508 resulting in inconsistent data (605627)
4509 - inode attributes were not always being refreshed (605627, 612449)
4510 - lconf now continues to cleanup after lctl reports an error
4511 - MDS now enforces user permissions (602707)
4512 - lprocfs cleanup fixed, but not yet enabled (614157)
4513 - fixed infinite server hang, should a client not respond to an AST
4514 - avoid going into recovery if user calls readlink() with a buffer
4515 that's too small (613941)
4516 - AST RPCs no longer require replies (614867) -- this may be changed
4517 - don't crash server if client sends an IOV that's too big (611336)
4518 - fixed lock conversion deadlock (611892)
4519 - fixed the following of symlinks (614622)
4520 * recovery: the server can remove locks from a client that dies, other
4521 clients can make progress
4522 * more extN patch fixes
4523 * compile-time configurable ptlrpc buffer allocations
4525 - collaborative read cache document
4526 - Lustre Lite Performance CDR document-in-progress
4528 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4531 - fix typo in patch-2.4.18
4533 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4536 - clear ptlrpc request each time in handle_incoming_request()
4537 - unlink of files now destroys the object on the OST
4539 2002-09-19 Peter Braam <braam@clusterfs.com>
4541 * add hard link support
4542 * change obdfile creation method
4543 * kernel patch changed
4545 2002-09-19 Peter Braam <braam@clusterfs.com>
4548 - stack overflow bug in extN fixed
4550 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4552 * documentation updates
4553 - add man pages for config tools
4554 - update tests/README to describe testing with new config tools
4555 - finish metadata API descriptions
4556 * bug fixes and cleanups
4557 - statfs workaround for 16TB limit
4558 - LOV stripe allocation improved, can stripe on subset of OSTs
4559 - LOV file size/IO offset was wrong for files > 4GB in size
4560 - object EA data was being dropped, caused files to be unreadable
4561 - memory overflow with non-LOV OST caused memory corruption
4562 - fixed regression tests to work with new config tools, obdfilter
4563 - fixed bug when directory size became larger than 1 block
4564 - fixed bug (for single client case) when PWD was deleted
4565 - invalidate local directory pages when doing intent-based ops
4566 - avoid LDLM oops when lock callback contained bad data
4568 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4570 * documentation updates
4571 * bug fixes and cleanups
4572 - configuration tools
4575 - 64-bit compile warnings
4576 - 64-bit internal statfs data
4578 * test_brw on persistent OST devices
4580 * lprocfs (disabled)
4582 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4584 * documentation updates
4585 * bug fixes and cleanups
4586 * configuration tools
4588 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4591 * many small fixes to 0_5_4
4592 * io/network handling
4593 * thinkos in MDS operations
4595 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4598 * crucial basic fixes to 0.5.3
4599 * IOR, Iozone work over Elan
4602 2002-08-07 Phil Schwan <phil@clusterfs.com>
4603 * version 0_5_3, our first alpha
4604 * we use the new Portals iovs
4605 * documentation updates
4606 * bug fixes and cleanups
4607 * small changes in the DLM wire protocol
4609 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4610 * version 0_5_1 with some initial stability,
4611 * locking on MD and file I/O.
4612 * documentation updates
4613 * several bug fixes since 0.5.0
4614 * small changes in wire protocol
4616 2002-07-18 Phil Schwan <phil@clusterfs.com>
4618 * delivered as Lustre Light Alpha
4619 * fixed a crash after handling invalid MDS requests
4620 * fixed directory pages for architectures with non-4k pages sizes
4622 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4623 * release version v0_4_4
4624 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4626 * Fixes a number of interruption problems with OST operations.
4627 * Update documentation for portals header changes
4628 * Move all wire protocol structs/defines to lustre_idl.h
4629 * Fixes symlink length bug.
4630 * Add tcpdump to repository.
4632 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4633 * release version v0_4_3
4634 * Fixes statfs for inodes on extN.
4635 * Fixes bug in runtests which would delete /etc/hosts.
4636 * Use 64-bit object IDs wherever possible (not into VFS though)
4637 Remove ost_get_info, which is unused by lustre, and out of date.
4639 2002-07-03 Peter Braam <braam@clusterfs.com>
4640 * release version v0_4_2 Fixes a lookup error (type not passed)
4641 * move forward to head of Portals
4642 * move forward to latest Lustre kernel
4644 2002-06-25 Peter Braam <braam@clusterfs.com>
4645 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob