1 tbd Sun Microsystems, Inc.
4 2.6.9-67.0.4.EL (RHEL 4),
5 2.6.16.54-0.2.5 (SLES 10),
6 2.6.18-53.1.14.el5 (RHEL 5),
7 2.6.22.14 vanilla (kernel.org).
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.16 - 2.6.21 vanilla (kernel.org)
11 * Recommended e2fsprogs version: 1.40.7-sun1
12 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
13 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
14 removed cwd "./" (refer to Bugzilla 14399).
16 Severity : enhancement
18 Description: Update to support 2.6.22.14 vanilla kernel.
22 Frequency : rare, on recovery
23 Description: read procfs can produce deadlock in some situation
24 Details : Holding lprocfs lock which send rpc can produce block for destroy
25 obd objects and this also block reconnect with -EALREADY. This isn't
26 fix all lprocfs bugs - but make it rare.
28 Severity : enhancement
30 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
33 Frequency : frequent on X2 node
35 Description: mdc_set_open_replay_data LBUG
36 Details : Set replay data for requests that are eligible for replay.
40 Description: lustre_mgs: operation 101 on unconnected MGS
41 Details : When MGC is disconnected from MGS long enough, MGS will evict the
42 MGC, and late on MGC cannot successfully connect to MGS and a lot
43 of the error messages complaining that MGS is not connected.
46 Frequency : on start mds
48 Description: Implement get_info(last_id) in obdfilter.
51 Frequency : occasional
53 Description: Correctly check stale fid, not start epoch if ost not support SOM
54 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
55 overwrited with new generated fid, but mds can anwer with one of these
56 two fids and both is not stale. setattr incorectly start epoch and
57 assume will be called done_writeting, but without SOM done_writing
61 Frequency : rare, depends on device drivers and load
63 Description: MDS or OSS nodes crash due to stack overflow
64 Details : Code changes in 1.8.0 increased the stack usage of some functions.
65 In some cases, in conjunction with device drivers that use a lot
66 of stack the MDS (or possibly OSS) service threads could overflow
67 the stack. One change which was identified to consume additional
68 stack has been reworked to avoid the extra stack usage.
71 Frequency : occasional
73 Description: Do not fail import if osc_interpret_create gets -EAGAIN
74 Details : If osc_interpret_create got -EAGAIN it immediately exits and
75 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
76 oscc_has_objects and see OSC has no objests and call
77 oscc_internal_create to resend create request.
79 Severity : enhancement
81 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
83 Severity : enhancement
85 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
90 Description: Fix replay if there is an un-replied request and open
91 Details : In some cases, older replay request will revert the
92 mcd->mcd_last_xid on MDS which is used to record the client's
95 Severity : enhancement
97 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
99 Severity : enhancement
101 Description: Add rhel5 support to HEAD.
103 Serverity : enhancement
105 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
110 Description: Don't allow skipping OSTs if index has been specified.
111 Details : Don't allow skipping OSTs if index has been specified, make locking
112 in internal create lots better.
116 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
117 Details : access to bitfield in structure is always rounded to long
118 and this produce problem with not atomic change any bit.
122 Description: Lustre make rpms failed.
123 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
124 builds Lustre rpms from tarball.
127 Frequency : rare on shutdown ost
129 Description: If llog cancel was not send before clean_exports phase, this can
130 produce deadlock in llog code.
131 Details : If llog thread has last reference to obd and call class_import_put
132 this produce deadlock because llog_cleanup_commit_master wait when
133 last llog_commit_thread exited, but this never success because was
134 called from llog_commit_thread.
138 Description: allow userland application know is lost one of stripes.
139 Details : fill lvb_blocks with error code on ost and return it to
140 application if error flag found.
144 Description: NULL lov_tgts causing MDS oops
145 Details : more safe checks for NULL lov_tgts for avoid oops.
147 Severity : enhancement
149 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
153 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
154 Details : If an OST is down the MDS will hang indefinitely in
155 obd_statfs_rqset() waiting for the statfs data. While for
156 MDS QOS usage of statfs, it should not stuck in waiting.
158 Severity : enhancement
160 Description: remote_acl support
161 Details : Support ACL-based permission check for remote user.
162 Support setfacl/getfacl for remote user with the utils
163 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
164 the system "{s,g}etfacl" utils.
166 Severity : enhancement
168 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
170 Severity : enhancement
172 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
176 Description: make mgs_setparam() handle fsname containing dash
177 Details : fsname containing a dash does not work with lctl conf_param
179 Severity : enhancement
181 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
183 Severity : enhancement
185 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
189 Frequency : rare, at shutdown
190 Description: access already free / zero obd_namespace.
191 Details : if client_disconnect_export was called without force flag set,
192 and exist connect request in flight, this can produce access to
193 NULL pointer (or already free pointer) when connect_interpret
194 store ocd flags in obd_namespace.
198 Frequency : only at startup
199 Description: not alloc memory with spinlock held.
200 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
201 if any spinlock held.
203 Severity : enhancement
205 Description: make lustre randomly fail allocating memory
206 Details : Make lustre randomly failed allocating memory for testing purpose.
208 Severity : enhancement
210 Description: lost problems with lov objid file
211 Details : Fixes some scability and access to not inited memory problems
212 in work with lov objdid file.
217 Description: lfs find does not continue on file error
218 Details : Continue other files processing when a file/dir is absent.
222 Description: Inconsistent usage of lustre_pack_reply()
223 Details : Standardize the usage of lustre_pack_reply() such that it
224 always generate a CERROR on failure.
227 Frequency : occasional
229 Description: MDS hang or stay in waiting lock
230 Details : If client receive lock with CBPENDING flag ldlm need send lock
231 cancel as separate rpc, to avoid situation when cancel request
232 can't processed due all i/o threads stay in wait lock.
236 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
240 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
244 Description: Update to SLES9 kernel-2.6.5-7.287.3.
248 Description: Update to RHEL4 latest kernel.
250 Severity : enhancement
252 Description: Build SLES10 patchless client fails
253 Details : The configure was broken by run ./configure with
254 --with-linux-obj=.... argument for patchless client. When the
255 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
256 can't search header adequately. Use absolute path such as
257 -I($LINUX)/include instead.
261 Description: interrupt oig_wait produce painc on resend.
262 Details : brw_redo_request can be used for resend requests from ptlrpcd and
263 private set, and this produce situation when rq_ptlrpcd_data not
264 copyed to new allocated request and triggered LBUG on assert
265 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
266 ptlrpcd set if request is changed and can be safety changed to use
271 Description: LASSERT_{REQ,REP}SWAB macros are buggy
272 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
273 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
274 these macros and replace them with their difinitions instead.
278 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
279 Details : Update which_patch & target file for SLES10 latest kernel.
282 Description: Remove client patches from SLES 10 kernel.
283 Details : This causes SLES 10 clients to behave as patchless clients
284 even on a Lustre-patched (server) kernel.
286 Severity : enhancement
288 Description: self-adjustable client's lru lists
289 Details : use adaptive algorithm for managing client cached locks lru
290 lists according to current server load, other client's work
291 pattern, memory activities, etc. Both, server and client
292 side namespaces provide number of proc tunables for controlling
295 Severity : enhancement
297 Description: light-weight GSS support
298 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
299 and reduce performance overhead.
301 Severity : enhancement
303 Description: Linux keyring support
304 Details : Support using service of Linux keyring for Lustre GSS internal
305 context refresh/cache mechanism.
309 Description: Fix errors in lfs documentation
310 Details : Fixes man pages
314 Description: don't use GFP_* in generic Lustre code.
315 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
319 Description: obdclass is limited by single OBD_ALLOC(idarray)
320 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
324 Description: warnings with build patchless client with vanila 2.6.19 and up
325 Details : change old ctl_table style and replace ctl_table/ctl_table_header
326 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
330 Frequency : Only for SLES
331 Description: sanity_quota fail test_1
332 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
333 file, which makes TSTID[2] inunique.
337 Frequency : Only for SLES9
338 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
339 Details : Update target/ChangeLog/which_patch .
343 Frequency : Only for test_10 in sanity-quota.sh
344 Description: a bug in quota test code
345 Details : checking a test flag in a wrong place
349 Frequency : Only for test_13 in sanity-quota.sh
350 Description: a bug in quota test script
351 Details : shouldn't check overfull allocated quota
355 Description: jbd statistics
356 Details : Port older jbd statistics patch for sles10
358 Severity : enhancement
360 Description: Add printing inode info into message about error in writepage.
365 Description: reply_lock_interpret crash due to race with it and lock cancel.
366 Details : Do not replay locks that are being cancelled. Do not reference
367 locks by their address during replay, just by their handle.
371 Frequency : When flocks are used.
372 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
373 Details : Flock locks might destroy just granted lock if it could be merged
374 with another existing flock, this is done in completion handler,
375 so teach ldlm_cli_enquque_fini that this is a valid case for
381 Description: Oops in read and write path when failing to allocate lock.
382 Details : Check if lock allocation failed and return error back.
386 Description: lstripe command fails for valid OST index
387 Details : The stripe offset is compared to lov->desc.ld_tgt_count
388 instead of lov->desc.ld_active_tgt_count.
392 Description: sanity.sh failed test 103
393 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
398 Description: df doesn't work properly if diskfs blocksize != 4K
399 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
403 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
404 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
408 Description: obd_health_check_timeout too short
409 Details : set obd_health_check_timeout as 1.5x of obd_timeout
413 Description: llapi_file_create() does not allow some changes
414 Details : add llapi_file_open() that allows specifying the mode and
415 open flags, and also returns an open file handle.
419 Description: merge and cleanup kernel patches.
420 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
424 Description: Add journal checksum support.(Kernel part)
425 Details : The journal checksum feature adds two new flags i.e
426 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
427 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
428 indicates that the commit block contains the checksum for
429 the blocks described by the descriptor blocks. Now commit
430 record can be sent to disk without waiting for descriptor
431 blocks to be written to disk. This behavior is controlled
432 using JBD2_FEATURE_ASYNC_COMMIT flag.
436 Description: OSS needs mutliple precreate threads
437 Details : Add ability to start more than one create thread per OSS.
441 Description: Sanity.sh test_65k failed: setstripe should have succeeded
442 Details : execute lfs setstripe on client
446 Description: mds_obd_create error creating tmp object
447 Details : When the user sets quota on root, llog will be affected and can't
448 create files and write files.
451 Frequency : Always on ia64 patchless client, and possibly others.
453 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
454 Details : This allows the patchless client to be loaded on architectures
459 Description: RedHat Update kernel for RHEL5
460 Details : Add the kernel config file for RHEL5.
464 Description: RedHat Update kernel for RHEL5
465 Details : Modify the kernel config file more closer RHEL5.
469 Description: Build failure against Centos5 (RHEL5)
470 Details : Define PAGE_SIZE when it isn't present.
474 Description: test_7 on sanity-quota.sh can't be used on separate machines
475 Details : add facet to handle it
479 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
480 Details : with NFS, the anon dentry's parent was set to itself in
481 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
482 corresponding dentry other than use rec->ur_name.
484 Severity : enhancement
486 Description: lfs setstripe enhancement
487 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
491 Description: enable data checksumming by default
492 Details : enable checksum by default, allow --disable-checksum
493 configure option and "-o nochecksum" mount option. Checksums
494 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
498 Description: lustre support for RHEL5
499 Details : Add support for RHEL5.
503 Description: Client eviction due to failover config
504 Details : after a connection loss, the lustre client should attempt to
505 reconnect to the last active server first before trying the
506 other potential connections.
510 Description: when mds and osts use different quota unit(32bit and 64bit),
511 quota will be released repeatly.
512 Details : void sending multiple quota reqs to mds, which will keep the
513 status between the reqs.
517 Description: osts not allocated evenly to files
518 Details : change the condition to increase offset_idx
521 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
523 Description: Data corruption for OSTs that are formatted larger than 2TB
525 Details : When generating the bio request for lustre file writes the
526 sector number would overflow a temporary variable before being
527 used for the IO. The data reads correctly from Lustre (which
528 will overflow in a similar manner) but other file data or
529 filesystem metadata may be corrupted in some cases.
533 Description: Tune the kernel for good SCSI performance.
534 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
535 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
540 Description: rewrite ext2-derived code in obdclass/uuid.c
541 Details : rewrite inherited code (uuid parsing code from ext2 utils)
542 from scratch preserving functionality.
546 Description: "lfs find -obd UUID" prints directories
547 Details : "lfs find -obd UUID" will return all directory names instead
548 of just file names. It is incorrect because the directories
549 do not reside on the OSTs.
553 Description: use i_size_read and i_size_write in 2.6 port
554 Details : replace inode->i_size access with i_size_read/write()
557 Frequency : when removing large files
559 Description: scheduling issue during removal of large Lustre files
560 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
561 It causes scheduling issues when removing large files (17TB in the
565 Frequency : only with liblustre clients on XT3
567 Description: evictions taking too long
568 Details : allow llrd to evict clients directly on OSTs
571 Frequency : only on ppc
573 Description: /proc/fs/lustre/devices broken on ppc
574 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
575 We should make sure the type of 'index' is loff_t and then cast
576 explicitly as needed below. Do not assign an explicitly cast
580 Frequency : only for rhel5
582 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
583 Details : Modify the target file & which_kernel.
585 Severity : enhancement
587 Description: omit set fsid for export NFS
588 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
589 as FS_REQUIRES_DEV for avoid problems with generate fsid.
593 Frequency : Always, for kernels after 2.6.16
594 Description: Fix warning idr_remove called for id=.. which is not allocated.
595 Details : Last kernels save old s_dev before kill super and not allow
596 to restore from callback - restore it before call kill_anon_super.
600 Description: buffer overruns could theoretically occur
601 Details : llapi_semantic_traverse() modifies the "path" argument by
602 appending values to the end of the origin string, and a
603 overrun may occur. Adding buffer overrun check in liblustreapi.
607 Description: Fix error on 'ls .' at the top of the Lustre mount.
608 Details : Don't revalidate dentry if it is a root dentry.
612 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
613 Details : Modify vm-tunables-rhel4.patch.
617 Description: Kernel config for 2.6.18-vanilla.
618 Details : Modify targets/2.6-vanilla.target.in.
619 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
620 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
621 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
622 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
626 Description: improve handling recoverable errors
627 Details : if request processig with error which can be recoverable on server
628 request should be resend, otherwise page released from cache and
633 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
634 Details : Modify target file & which_patch.
635 A flaw was found in the IA32 system call emulation provided
636 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
637 value could be stored in the %RAX register, which could trigger an
638 out-of-bounds system call table access. An untrusted local user
639 could exploit this flaw to run code in the kernel
640 (ie a root privilege escalation). (CVE-2007-4573).
644 Description: O_DIRECT bypasses client statistics.
645 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
646 read_ahead_stats were updated. copied the stats section from
647 osc_send_oap_rpc() into async_internal().
651 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
655 Description: change order of libsysio includes
656 Details : '#include sysio.h' should always come before '#include xtio.h'
660 Description: handle "serious error: objid * already exists" more gracefully
661 Details : If LAST_ID value on disk is smaller than the objects existing in
662 the O/0/d* directories, it indicates disk corruption and causes an
663 LBUG(). If the object is 0-length, then we should use the existing
664 object. This will help to avoid a full fsck in most cases.
666 Severity : enhancement
668 Description: adapt the lustre_config script to support the upgrade case
669 Details : Add "-u" option for lustre_config script to support upgrading 1.4
670 server targets to 1.6 in parallel.
674 Description: To avoid grant space > avaible space when the disk is almost
675 full. Without this patch you might see the error "grant XXXX >
676 available" or some LBUG about grant, when the disk is almost
678 Details : In filter_check_grant, for non_grant cache write, we should
679 check the left space by if (*left > ungranted + bytes), instead
680 of (*left > ungranted), because only we are sure the left space
681 is enough for another "bytes", then the ungrant space should be
682 increase. In client, we should update cl_avail_grant only there
683 is OBD_MD_FLGRANT in the reply.
687 Description: Update RHEL 4 kernel to fix local root privilege escalation.
688 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
689 described in CVE-2007-4573. This problem could allow untrusted
690 local users to gain root access.
693 Frequency : when using O_DIRECT and quotas
695 Description: Incorrect file ownership on O_DIRECT output files
696 Details : block usage reported by 'lfs quota' does not take into account
697 files that have been written with O_DIRECT.
702 Description: touch file failed when fs is not full
703 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
704 otherwise we can get ENOSP while fs is not full.
708 Description: parallel lock callbacks
709 Details : Instead of sending blocking and completion callbacks as separated
710 requests, adding them to a set and sending in parallel.
713 Frequency : only for Cray XT3
714 Bugzilla : 12829/13455
715 Description: Changing primary group doesn't change the group lustre assigns to
717 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
718 overridden with the primary group provided by the group upcall,
719 whereas the supplied fsgid can be trusted if it is in the list of
720 supplementary groups returned by the group upcall.
722 Severity : enhancement
724 Description: Allow masking D_WARNING, D_ERROR messages from console
725 Details : Console messages can now be disabled via lnet.printk.
729 Description: User code with malformed file open parameter crashes client node
730 Details : Before packing join_file req, all the related reference should be
731 checked carefully in case some malformed flags cause fake
732 join_file req on client.
736 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
737 Details : ldlm_completion_ast() assumes that a lock is granted when the req
738 mode is equal to the granted mode. However, it should also check
739 that LDLM_FL_CP_REQD is not set.
743 Description: Heavy nfs access might result in deadlocks
744 Details : After ELC code landed, it is now improper to enqueue any mds
745 locks under och_sem, because enqueue might want to decide to
746 cancel open locks for same inode we are holding och_sem for.
750 Description: Client eviction while running blogbench
751 Details : A lot of unlink operations with concurrent I/O can lead to a
752 deadlock causing evictions. To address the problem, the number of
753 oustanding OST_DESTROY requests is now throttled to
754 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
755 callbacks are processed in priority.
759 Description: enable ACLs on MDS by default
760 Details : ACLs must be enabled on MDS by default.
763 Frequency : PPC/PPC64 only
765 Description: conflicts between asm-ppc64/types.h and lustre_types.h
766 Details : fix duplicated definitions between asm-ppc64/types.h and
767 lustre_types.h on PPC.
770 Frequency : PPC/PPC64 only
772 Description: asm-ppc/segment.h does not exist
773 Details : fix compile issue on PPC.
777 Description: better handle error messages in extents code
780 Frequency : RHEL4 only
782 Description: mkfs is very slow on IA64/RHEL4
783 Details : A performance regression has been discovered in the MPT Fusion
784 driver between versions 3.02.73rh and 3.02.99.00rh. As a
785 consequence, we have downgraded the MPT Fusion driver in the RHEL4
786 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
788 Severity : enhancement
790 Description: SNMP support enhancement
791 Details : Adding total number of sampled request for an MDS node in snmp
794 Severity : enhancement
796 Description: Optimize ldlm waiting list processing for PR extent locks
797 Details : When processing waiting list for read extent lock and meeting read
798 lock that is same or wider to it that is not contended, skip
799 processing rest of the list and immediatelly return current
800 status of conflictness, since we are guaranteed there are no
801 conflicting locks in the rest of the list.
805 Description: Time out and refuse to reconnect
806 Details : When the failover node is the primary node, it is possible
807 to have two identical connections in imp_conn_list. We must
808 compare not conn's pointers but NIDs, otherwise we can defeat
809 connection throttling.
813 Description: port llog fixes from b1_6 into HEAD
814 Details : Port llog reference couting and some llog cleanups from b1_6
815 (bug 10800) into HEAD, for protect from panic and access to already
816 free llog structures.
820 Description: Detect stride IO mode in read-ahead
821 Details : When a client does stride read, read-ahead should detect that and
822 read-ahead pages according to the detected stride pattern.
826 Description: data checksumming impacts single node performance
827 Details : add support for several checksum algorithm. Currently, only CRC32
828 and Adler-32 are supported. The checksum type can be changed on
829 the fly via /proc/fs/lustre/osc/*/checksum_type.
833 Description: use adler32 for page checksums
834 Details : when available, use the Adler-32 algorithm instead of CRC32 for
839 Description: build for x2 fails
840 Details : fix compile issue on Cray systems.
844 Description: Properly match for duplicate locks
845 Details : Due to different lock order from skiplists code, we need to
846 traverse entire list for now
849 Frequency : only on PPC/SLES10
851 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
852 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
853 be defined. Add a hack in mkfs_lustre.c to work around this compile
858 Description: LASSERT on MDS when client holding flock lock dies
859 Details : ldlm pool logic depends on number of granted locks equal to
860 number of released locks which is not true for flock locks, so
861 just exclude such locks from consideration.
863 --------------------------------------------------------------------------------
865 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
867 * Support for kernels:
868 2.6.5-7.283 (SLES 9),
869 2.6.9-55.EL (RHEL 4),
870 2.6.16.46-0.14 (SLES 10),
871 2.6.18.8 vanilla (kernel.org)
872 * Client support for unpatched kernels:
873 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
874 2.6.16 - 2.6.22 vanilla (kernel.org)
875 * Due to recently discovered recovery problems, we do not recommend
876 using patchless RHEL 4 clients with this or any earlier release.
877 * Recommended e2fsprogs version: 1.39.cfs8
878 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
879 * Starting with this release, the ldiskfs backing filesystem required
880 by Lustre is now in its own package, lustre-ldiskfs. This package
881 should be installed. It is versioned separately from Lustre and
882 may be released separately in future.
886 Description: block reactivating mgc import until all deactivates complete
887 Details : Fix race when failing back MDT/MGS to itself (testing)
890 Frequency : at statup only
892 Description: mds_lov_synchronize race leads to various problems
893 Details : simultaneous MDT->OST connections at startup can cause the
894 sync to abort, leaving the OSC in a bad state.
896 Severity : enhancement
898 Description: add optional extra BUILD_VERSION info
899 Details : add a new environment variable (namely LUSTRE_VERS) which allows
900 to override the lustre version.
903 Frequency : 2.6.18 servers only
905 Description: ll_kern_mount() doesn't release the module reference
906 Details : The ldiskfs module reference count never drops down to 0
907 because ll_kern_mount() doesn't release the module reference.
912 Description: server LBUG when using old ost_num_threads parameter
913 Details : Accept the old ost_num_threads parameter but warn that it
914 is deprecated, and fix an off-by-one error that caused an LBUG.
919 Description: Transient SCSI error results in persistent IO issue
920 Details : iobuf->dr_error is not reinitialized to 0 between two
924 Frequency : sometimes when underlying device returns I/O errors
926 Description: OSTs not going read-only during write failures
927 Details : OSTs are not remounted read-only when the journal commit threads
928 get I/O errors because fsfilt_ext3 calls journal_start/stop()
929 instead of the ext3 wrappers.
933 Description: poor connect scaling with increasing client count
934 Details : Don't run filter_grant_sanity_check for more than 100 exports
935 to improve scaling for large numbers of clients.
938 Frequency : SLES10 only
940 Description: sanity-quota.sh quotacheck failed: rc = -22
941 Details : Quotas cannot be enabled on SLES10.
944 Frequency : liblustre clients only
946 Description: getdirentries does not give error when run on compute nodes
947 Details : getdirentries does not fail when the size specified as an argument
948 is too small to contain at least one entry
950 Severity : enhancement
952 Description: Add LNET router traceability for debug purposes
953 Details : If a checksum failure occurs with a router as part of the
954 IO path, the NID of the last router that forwarded the bulk data
955 is printed so it can be identified.
960 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
961 Details : Due to a race condition, liblustre clients were occasionally
964 Severity : enhancement
966 Description: lfs setstripe use optional parameters instead of postional
969 Severity : enhancement
971 Description: Nanosecond timestamp support for ldiskfs
972 Details : The on-disk ldiskfs filesystem has added support for nanosecond
973 resolution timestamps. There is not yet support for this at
974 the Lustre filesystem level.
977 Frequency : during server recovery
979 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
980 Details : request with rq_no_resend flag not awake l_wait_event if they get a
984 Frequency : nfs export on patchless client
986 Description: connectathon hang when test nfs export over patchless client
987 Details : Disconnected dentry cannot be found with lookup, so we do not need
988 to unhash it or make it invalid
991 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
992 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
993 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
998 Description: open req refcounting wrong on reconnect
999 Details : If reconnect happened between getting open reply from server and
1000 call to mdc_set_replay_data in ll_file_open, we will schedule
1001 replay for unreferenced request that we are about to free.
1002 Subsequent close will crash in variety of ways.
1003 Check that request is still eligible for replay in
1004 mdc_set_replay_data().
1009 Description: disable writes to filesystem when reading health_check file
1010 Details : the default for reading the health_check proc file has changed
1011 to NOT do a journal transaction and write to disk, because this
1012 can cause reads of the /proc file to hang and block HA state
1013 checking on a healthy but otherwise heavily loaded system. It
1014 is possible to return to the previous behaviour during configure
1015 with --enable-health-write.
1017 Severity : enhancement
1019 Description: 64-bit inode version
1020 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1021 to the inode. This will be required for version-based recovery.
1026 Description: MDS fails to start if a duplicate client export is detected
1027 Details : in some rare cases it was possible for a client to connect to
1028 an MDS multiple times. Upon recovery the MDS would detect this
1029 and fail during startup. Handle this more gracefully.
1031 Severity : enhancement
1033 Description: Add -o localflock option to simulate old noflock
1035 Details : This will achieve local-only flock/fcntl locks
1041 Description: log_commit_thread vs filter_destroy race leads to crash
1042 Details : Take import reference before releasing llog record semaphore
1047 Description: Wrong request locking in request set processing
1048 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1049 imp_delayed_list, in this place should be used imp_lock.
1052 Frequency : when reconnection
1054 Description: Grant Leak when osc reconnect to OST
1055 Details : When osc reconnect ost, OST(filter) should check whether it
1056 should grant more space to client by comparing fed_grant and
1057 cl_avail_grant, and return the granted space to client instead
1058 of "new granted" space, because client will call osc_init_grant
1059 to update the client grant space info.
1062 Frequency : when client reconnect to OST
1064 Description: Grant Leak when osc do resend and replay bulk write
1065 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1066 bulk write request, because the grant info while be sync between
1067 OSC and OST when reconnect, and we should ignore the grant info
1068 these of resend/replay write req.
1073 Description: Grant space more than avaiable left space sometimes.
1074 Details : When then OST is about to be full, if two bulk writing from
1075 different clients came to OST. Accord the avaliable space of the
1076 OST, the first req should be permitted, and the second one
1077 should be denied by ENOSPC. But if the seconde arrived before
1078 the first one is commited. The OST might wrongly permit second
1079 writing, which will cause grant space > avaiable space.
1082 Frequency : when client is evicted
1084 Description: Grant might be wrongly erased when osc is evicted by OST
1085 Details : when the import is evicted by server, it will fork another
1086 thread ptlrpc_invalidate_import_thread to invalidate the
1087 import, where the grant will be set to 0. While the original
1088 thread will update the grant it got when connecting. So if
1089 the former happened latter, the grant will be wrongly errased
1090 because of this race.
1095 Description: Checking Stale with correct fid
1096 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1097 to check whether it is stale, because sometimes, we want the
1098 enqueue happened anyway, and op_data.fid2 will not be initialized.
1100 Severity : enhancement
1102 Description: update patchless client
1103 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1106 Frequency : only with 2.4 kernel
1108 Description: random memory corruption
1109 Details : size of struct ll_inode_info is to big for union inode.u and this
1110 can be cause of random memory corruption.
1115 Description: Memory leak in recovery
1116 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1117 It should also check obd_fail before fsfilt_start, otherwise if
1118 fsfilt_start return -EROFS,(failover mds during mds recovery).
1119 then the req will return with repmsg->transno = 0 and rc = EROFS.
1120 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1121 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1122 be freed no matter whether fsfilt_commit success or not.
1125 Frequency : only with huge count clients
1127 Description: Prevents from taking the superblock lock in llap_from_page for
1129 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1130 for llap_from_page call in ll_removepage prevents from taking the
1131 superblock lock for a soon died page.
1136 Description: Not check open intent error before release open handle
1137 Details : in some rare cases, the open intent error is not checked before
1138 release open handle, which may cause
1139 ASSERTION(open_req->rq_transno != 0), because it tries to release
1140 the failed open handle.
1145 Description: Set cat log bitmap only after create log success.
1146 Details : in some rare cases, the cat log bitmap is set too early. and it
1147 should be set only after create log success.
1151 Description: Accessing a block bevice can re-enable I/O when Lustre is
1152 tearing down a device.
1153 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1158 Description: service threads may hog cpus when there are a lot of requests
1159 Details : Insert cond_resched to give other threads a chance to use some CPU
1164 Description: the cat log was not initialized in recovery
1165 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1166 unlink log on mds will not be initialized until mds post
1167 recovery. And also in mds post recovery, the unlink log will
1168 initialization will be done asynchronausly, so there will be race
1169 between add unlink log and unlink log initialization.
1173 Description: brw_stats were being printed incorrectly
1174 Details : brw_stats were being printed as log2 but all of them were not
1175 recorded as log2. Also remove some code duplication arising from
1176 filter_tally_{read,write}.
1180 Frequency : rare, only in recovery.
1181 Description: ASSERTION(req->rq_type != LI_POISON) failed
1182 Details : imp_lock should be held while iterating over imp_sending_list for
1183 prevent destroy request after get timeout in ptlrpc_queue_wait.
1187 Description: replay-single.sh test 52 fails
1188 Details : A lock's skiplist need to be cleanup when it being unlinked
1189 from its resource list.
1193 Description: Short directio read returns full requested size rather than
1195 Details : Direct I/O operations should return actual amount of bytes
1196 transferred rather than requested size.
1198 Severity : enhancement
1200 Description: metadata RPC reduction (e.g. for rm performance)
1201 Details : decrease the amount of synchronous RPC between clients and servers
1202 by canceling conflicing lock before the operation on the client
1203 and packing thier handles into the main operation RPC to server.
1205 Severity : enhancement
1207 Description: Async OSC create to avoid the blocking unnecessarily.
1208 Details : If a OST has no remain object, system will block on the creating
1209 when need to create a new object on this OST. Now, ways use
1210 pre-created objects when available, instead of blocking on an
1211 empty osc while others are not empty. If we must block, we block
1212 for the shortest possible period of time.
1214 Severity : enhancement
1216 Description: refine locking for avoid write wrong info into lov_objid file
1217 Details : fix possible races with add new target and write/update data in
1220 --------------------------------------------------------------------------------
1222 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1227 Frequency : on some architectures
1229 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1230 Details : Uninitialized flags sometimes cause configuration commands to
1234 Frequency : patchless clients only
1236 Description: missing __iget() symbol export
1237 Details : The __iget() symbol export is missing. To avoid the need for
1238 this on patchless clients the deathrow inode reaper is turned
1239 off, and we depend on the VM to clean up old inodes. This
1240 dependency was during via the fix for bug 12181.
1242 --------------------------------------------------------------------------------
1244 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1246 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1247 INTEROPERATE with older versions automatically. In many cases a
1248 special upgrade step is needed. Please read the
1249 user documentation before upgrading any part of a live system.
1250 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1251 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1252 * WARNING: Lustre configuration and startup changes are required with
1253 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1259 Severity : enhancement
1261 Description: MountConf
1262 Details : Lustre configuration is now managed via mkfs and mount
1263 commands instead of lmc and lconf. New obd types (MGS, MGC)
1264 are added for dynamic configuration management. See
1265 https://mail.clusterfs.com/wikis/lustre/MountConf for
1268 Severity : enhancement
1270 Description: dynamic OST addition
1271 Details : OSTs can now be added to a live filesystem
1273 Severity : enhancement
1275 Description: startup order invariance
1276 Details : MDTs and OSTs can be started in any order. Clients only
1277 require the MDT to complete startup.
1279 Severity : enhancement
1281 Description: parallel, asynchronous orphan cleanup
1282 Details : orphan cleanup is now performed in separate threads for each
1283 OST, allowing parallel non-blocking operation.
1285 Severity : enhancement
1287 Description: optimized stripe assignment
1288 Details : stripe assignments are now made based on ost space available,
1289 ost previous usage, and OSS previous usage, in order to try
1290 to optimize storage space and networking resources.
1292 Severity : enhancement
1294 Description: Permanently set tunables
1295 Details : All writable /proc/fs/lustre tunables can now be permanently
1296 set on a per-server basis, at mkfs time or on a live system.
1298 Severity : enhancement
1300 Description: Lustre message v2
1301 Details : Add lustre message format v2.
1303 Severity : enhancement
1305 Description: client OST exclusion list
1306 Details : Clients can be started with a list of OSTs that should be
1307 declared "inactive" for known non-responsive OSTs.
1311 Description: SPEC SFS validation failure on NFS v2 over lustre.
1312 Details : Changes the blocksize for regular files to be 2x RPC size,
1313 and not depend on stripe size.
1315 Severity : enhancement
1317 Description: Multiple MD RPCs in flight.
1318 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1319 To avoid overly-overloading MDS, introduce a limit on number of
1320 MDS RPCs in flight for a single client and add /proc controls
1321 to adjust this limit.
1323 Severity : enhancement
1325 Description: client read/write statistics
1326 Details : Add client read/write call usage stats for performance
1327 analysis of user processes.
1328 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1329 file access. extents_stats shows chunk size distribution.
1330 extents_stats_per_process show chunk size distribution per
1333 Severity : enhancement
1335 Description: mds statistics
1336 Details : Add detailed mds operations statistics in
1337 /proc/fs/lustre/mds/*/stats.
1341 Description: Failure of copying files with lustre special EAs.
1342 Details : Client side always return success for setxattr call for lustre
1343 special xattr (currently only "trusted.lov").
1348 Description: Refcount LNET uuids
1349 Details : The global LNET uuid list grew linearly with every startup;
1350 refcount repeated list entries instead of always adding to
1353 Severity : enhancement
1355 Description: Dynamic service threads
1356 Details : Within a small range, start extra service threads
1357 automatically when the request queue builds up.
1360 Frequency : mixed-endian client/server environments
1362 Description: mixed-endian crashes
1363 Details : The new msg_v2 system had some failures in mixed-endian
1366 Severity : enhancement
1368 Description: Easy OST removal
1369 Details : OSTs can be permanently deactivated with e.g. 'lctl
1370 conf_param lustre-OST0001.osc.active=0'
1372 Severity : enhancement
1374 Description: MGS proc entries
1375 Details : Added basic proc entries for the MGS showing what filesystems
1378 Severity : enhancement
1380 Description: provide MGS failover
1381 Details : Added config lock reacquisition after MGS server failover.
1383 Severity : enhancement
1385 Description: add Linux 2.4 support
1386 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1390 Description: a large application tries to do I/O to the same resource and dies
1391 in the middle of it.
1392 Details : Check the req->rq_arrival time after the call to
1393 ost_brw_lock_get(), but before we do anything about
1394 processing it & sending the BULK transfer request. This
1395 should help move old stale pending locks off the queue as
1396 quickly as obd_timeout.
1399 Frequency : when an incorrect nid is specified during startup
1401 Description: ptlrpc connect to non-existant node causes kernel crash
1402 Details : LNET can't be re-entered from an event callback, which
1403 happened when we expire a message after the export has been
1404 cleaned up. Instead, hand the zombie cleanup off to another
1407 Severity : enhancement
1409 Description: plain/inodebits lock performance improvement
1410 Details : Grouping plain/inodebits in granted list by their request modes
1411 and bits policy, thus improving the performance of search through
1415 Frequency : only if OST filesystem is corrupted
1417 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1418 Details : for a short time RPCs with bulk IO are in the replay list,
1419 but replay of bulk IOs is unimplemented. If the OST filesystem
1420 is corrupted due to disk cache incoherency and then replay is
1421 started it is possible to trip an assertion. Avoid putting
1422 committed RPCs into the replay list at all to avoid this issue.
1425 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1427 Description: System hang on startup
1428 Details : This bug allowed the liblustre (e.g. catamount) client to
1429 return to the app before handling all startup RPCs. This
1430 could leave the node unresponsive to lustre network traffic
1431 and manifested as a server ptllnd timeout.
1433 Severity : enhancement
1435 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1436 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1437 boolean to control whether to print peer debug info when a
1438 client's RPC times out.
1441 Frequency : only for kernels with patches from Lustre below 1.4.3
1443 Description: Remove old rdonly API
1444 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1447 Frequency : only for devices with external journals
1449 Description: Set external device read-only also
1450 Details : During a commanded failover stop, we set the disk device
1451 read-only while the server shuts down. We now also set any
1452 external journal device read-only at the same time.
1455 Frequency : when upgrading from 1.4 while trying to change parameters
1457 Description: The wrong (new) MDC name was used when setting parameters for
1458 upgraded MDT's. Also allows changing of OSC (and MDC)
1459 parameters if --writeconf is specified at tunefs upgrade time.
1462 Frequency : when setting specific ost indicies
1464 Description: QOS code breaks on skipped indicies
1465 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1466 created with --index need not be sequential.
1468 Severity : enhancement
1470 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1471 Details : The uninit_groups feature works in conjunction with the kernel
1472 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1473 up the pass1 processing of e2fsck. This is a read-only feature
1474 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1475 mount filesystems that have had this feature enabled.
1477 Severity : enhancement
1479 Description: Improve multi-block allocation algorithm to avoid fragmentation
1480 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1481 allocation locality and avoid filesystem fragmentation.
1483 ------------------------------------------------------------------------------
1485 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1487 * Support for kernels:
1488 2.4.21-47.0.1.EL (RHEL 3)
1489 2.6.5-7.283 (SLES 9)
1490 2.6.9-42.0.10.EL (RHEL 4)
1491 2.6.12.6 vanilla (kernel.org)
1492 2.6.16.27-0.9 (SLES 10)
1493 * Recommended e2fsprogs version: 1.39.cfs5
1495 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1499 Frequency : occasional, depends on client load and configuration
1500 Bugzilla : 12181, 12203
1501 Description: data loss for recently-modified files
1503 Details : In some cases it is possible that recently written or created
1504 files may not be written to disk in a timely manner (this should
1505 normally be within 30s unless client IO load is very high).
1506 The problem appears as zero-length files or files that are a
1507 multiple of 1MB in size after a client crash or client eviction
1508 that are missing data at the end of the file.
1510 This problem is more likely to be hit on clients where files are
1511 repeatedly created and unlinked in the same directory, clients
1512 have a large amount of RAM, have many CPUs, the filesystem has
1513 many OSTs, the clients are rebooted frequently, and/or the files
1514 are not accessed by other nodes after being written.
1516 The presence of the problem can be detected by looking at
1517 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1518 smaller than the second (nr_unused) then dirty files will not
1519 be flushed automatically to disk. "sync; sleep 10" should be
1520 run several times on the node before unmounting it to update
1521 Lustre (this is also safe to run on nodes without this problem).
1523 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1524 that can cause this same problem, so customers using that kernel
1525 also need to update the kernel in addition to Lustre. In order
1526 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1528 It is normal that files written just before a client crash (less
1529 than 30s) may not yet have been flushed to disk, even for local
1533 Frequency : frequent on thin XT3 nodes
1535 Description: UUID collision on thin XT3 Linux nodes
1536 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1537 randomly, since we relied on an insufficiently-seeded PRNG.
1542 Description: OSS hangs after "All ost request buffers busy"
1543 Details : A deadlock between quota and journal operations caused OSS
1544 hangs after printing "All ost request buffers busy."
1547 Frequency : always on liblustre builds
1549 Description: Cleanup compiler warnings on liblustre
1552 Frequency : always on liblustre builds on XT3
1554 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1558 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1559 > 32000 subdirectories
1560 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1561 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1562 subdirectory count crosses 32000. This will aid e2fsck to
1563 correctly handle more than 32000 subdirectories.
1566 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1568 Description: System hang on startup
1569 Details : This bug allowed the liblustre (e.g. catamount) client to
1570 return to the app before handling all startup RPCs. This
1571 could leave the node unresponsive to lustre network traffic
1572 and manifested as a server ptllnd timeout.
1574 Severity : enhancement
1576 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1577 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1578 boolean to control whether to print peer debug info when a
1579 client's RPC times out.
1584 Description: make O_SYNC working on 2.6 kernels
1585 Details : 2.6 kernels use different method for mark pages for write,
1586 so need add a code to lustre for O_SYNC work.
1591 Description: Failure to close file and release space on NFS
1592 Details : Put inode details into lock acquired in ll_intent_file_open.
1593 Use mdc_intent_lock in ll_intent_open to properly
1594 detect all kind of errors unhandled by mdc_enqueue.
1599 Description: proc file read during shutdown sometimes raced obd removal,
1601 Details : Add lock to prevent obd access after proc file removal.
1604 Frequency : Only for files larger than 4GB on 32-bit clients.
1606 Description: improperly doing page alignment of locks
1607 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1608 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1613 Description: RPCs being resent when they shouldn't be
1614 Details : Some RPCs that should not be resent are being resent. This
1615 can cause inconsistencies in the RPC state machine. Do not
1616 resend such requests.
1619 Frequency : rare, only with NFS export
1621 Description: Crash on NFS re-export node
1622 Details : under very unusual load conditions an assertion is hit in
1623 ll_intent_file_open()
1626 Frequency : only if OST filesystem is corrupted
1628 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1629 Details : for a short time RPCs with bulk IO are in the replay list,
1630 but replay of bulk IOs is unimplemented. If the OST filesystem
1631 is corrupted due to disk cache incoherency and then replay is
1632 started it is possible to trip an assertion. Avoid putting
1633 committed RPCs into the replay list at all to avoid this issue.
1638 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1639 Details : Large single O_DIRECT read and write calls can fail to allocate
1640 a sufficiently large buffer to process the request. In case of
1641 allocation failure the allocation is retried with a smaller
1642 buffer and broken into smaller requests.
1644 Severity : enhancement
1646 Description: Add -o localflock option to simulate old noflock behaviour.
1647 Details : This will achieve local-only flock/fcntl locks coherentness.
1652 Description: versioning check is incomplete
1653 Details : Checking the version difference of client vs. server, report
1654 error if the gap is too big.
1659 Description: add support PG_writeback bit
1660 Details : add support for PG_writeback bit for Lustre, for more carefull
1661 work with page cache in 2.6 kernel. This also fix some deadlocks
1662 and remove hack for work O_SYNC with 2.6 kernel.
1664 Severity : enhancement
1666 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1667 Details : The uninit_groups feature works in conjunction with the kernel
1668 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1669 up the pass1 processing of e2fsck. This is a read-only feature
1670 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1671 mount filesystems that have had this feature enabled.
1673 Severity : enhancement
1675 Description: Improve multi-block allocation algorithm to avoid fragmentation
1676 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1677 allocation locality and avoid filesystem fragmentation.
1679 ------------------------------------------------------------------------------
1681 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1683 * Support for kernels:
1684 2.6.9-42.0.3.EL (RHEL 4)
1685 2.6.5-7.276 (SLES 9)
1686 2.4.21-47.0.1.EL (RHEL 3)
1687 2.6.12.6 vanilla (kernel.org)
1688 2.6.16.21-0.8 (SLES10)
1689 * Recommended e2fsprogs version: 1.39.cfs2-0
1691 * The backwards-compatible /proc/sys/portals symlink has been removed
1692 in this release. Before upgrading, please ensure that you change
1693 any configuration scripts or /etc/sysctl.conf files that access
1694 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1695 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1696 in advance of the upgrade on any system running Lustre 1.4.6 or
1697 newer, since /proc/sys/lnet was added in that version.
1698 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1702 Frequency : only when quota is used
1704 Description: avoid scanning export list for quota master
1705 Details : Change the algorithms to avoid scanning export list in order
1706 to improve the efficiency.
1709 Frequency : MDS failover only, very rarely
1711 Description: "went back in time" messages on mds failover
1712 Details : The greatest transno may be lost when the current operation
1713 finishes with an error (transno==0) and the client's last_rcvd
1714 record is over-written. Save the greatest transno in the
1715 mds_last_transno for this case.
1718 Frequency : always for specific kernels and striping counts
1720 Description: client may get "Matching packet too big" without ACL support
1721 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1722 when trying to access files in certain configurations. The
1723 clients should in fact be denied when mounting because they do
1724 not understand ACLs.
1727 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1729 Description: many clients connecting with IO in progress causes connect timeouts
1730 Details : Avoid synchronous journal commits to avoid delays caused by many
1731 clients connecting/disconnecting when bulk IO is in progress.
1732 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1733 OST_IO_PORTAL to avoid delays behind potentially many pending
1737 Frequency : occasionally with multiple writers to a single file
1739 Description: shared writes to file may result in wrong size reported by stat()
1740 Details : Allow growing of kms when extent lock is cancelled
1743 Frequency : always with random mmap IO to multi-striped file
1745 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1746 Details : Only if the hole is at the end of OST object so that kms is too
1747 small. Fix is to increase kms accordingly in ll_nopage.
1750 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1752 Description: writes to a missing object would leak memory on the OST
1753 Details : If there is an inconsistency between the MDS and OST filesystems,
1754 such that the MDS references an object that doesn't exist, writes
1755 to that object will leak memory due to incorrect cleanup in the
1756 error handling path, eventually running out of memory on the OST.
1761 Description: Creating too long symlink causes lustre errors
1762 Details : Check symlink and name lengths before sending requests to MDS.
1765 Frequency : only if flock is enabled (not on by default)
1767 Description: posix locks not released on fd closure on 2.6.9+
1768 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1769 kernels, this caused such locks not to be released on fd close and
1770 then assertions on fs unmount about still used locks.
1773 Frequency : MDS failover only, very rarely
1775 Description: clients may get ASSERTION(granted_lock != NULL)
1776 Details : When request was taking a long time, and a client was resending
1777 a getattr by name lock request. The were multiple lock requests
1778 with the same client lock handle and
1779 mds_getattr_name->fixup_handle_for_resent_request found one of the
1780 lock handles but later failed with ASSERTION(granted_lock != NULL).
1785 Description: handle->h_buffer_credits > 0, assertion failure
1786 Details : h_buffer_credits is zero after truncate, causing assertion
1787 failure. This patch extends the transaction or creates a new
1791 Frequency : NFS re-export or patchless client
1792 Bugzilla : 11179, 10796
1793 Description: Crash on NFS re-export node (__d_move)
1794 Details : We do not want to hash the dentry if we don't have a lock.
1795 But if this dentry is later used in d_move, we'd hit uninitialised
1796 list head d_hash, so we just do this to init d_hash field but
1797 leave dentry unhashed.
1800 Frequency : NFS re-export or patchless client
1802 Description: NFS exports has problem with symbolic link
1803 Details : lustre client didn't properly install dentry when re-exported
1804 to NFS or running patchless client.
1807 Frequency : NFS re-export or patchless client
1809 Description: Various nfs/patchless fixes.
1810 Details : fixes reuse disconected alias for lookup process - this fixes
1811 warning "find_exported_dentry: npd != pd",
1812 fix permission error with open files at nfs.
1813 fix apply umask when do revalidate.
1816 Frequency : occasional
1818 Description: Crash on NFS re-export node
1819 Details : calling clear_page() on the wrong pointer triggered oops in
1820 generic_mapping_read().
1823 Frequency : rarely, using O_DIRECT IO
1825 Description: unaligned directio crashes client with LASSERT
1826 Details : check for unaligned buffers before trying any requests.
1829 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1831 Description: stale data returned from RAID cache
1832 Details : If only a small amount of IO is done to the RAID device before
1833 reading it again it is possible to get stale data from the RAID
1834 cache instead of reading it from disk.
1837 Frequency : always for sles10 kernel
1839 Description: sles10 support
1840 Details : ll_follow_link: compile fixes and using of nd_set_link
1841 under newer kernels.
1844 Frequency : depends on arch, kernel and compiler version, always on sles10
1847 Description: recursive or deep enough symlinks cause stack overflow
1848 Details : getting rid of large stack-allocated variable in
1852 Frequency : depends on hardware
1854 Description: lustre write performance loss in the SLES10 kernel
1855 Details : the performance loss is caused by using of write barriers in the
1856 ext3 code. The SLES10 kernel turns barrier support on by
1857 default. The fix is to undo that change for ldiskfs.
1859 ------------------------------------------------------------------------------
1861 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1863 * Support for kernels:
1864 2.6.9-42.0.3EL (RHEL 4)
1865 2.6.5-7.276 (SLES 9)
1866 2.4.21-47.0.1.EL (RHEL 3)
1867 2.6.12.6 vanilla (kernel.org)
1871 Frequency : quota enabled and large files being deleted
1873 Description: releasing more than 4GB of quota at once hangs OST
1874 Details : If a user deletes more than 4GB of files on a single OST it
1875 will cause the OST to spin in an infinite loop. Release
1876 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1881 Description: statfs data retrieved from /proc may be stale or zero
1882 Details : When reading per-device statfs data from /proc, in the
1883 {kbytes,files}_{total,free,avail} files, it may appear
1884 as zero or be out of date.
1887 Frequency : systems with MD RAID1 external journal devices
1889 Description: lconf's call to blkid is confused by RAID1 journal devices
1890 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1891 of returning all block devices that match the journal UUID.
1894 Frequency : always, for aggregate stripe size over 4GB
1896 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1897 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1898 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1901 Frequency : always if "lfs find" used on a local file/directory
1903 Description: "lfs find" segfaults if used on a local file/directory
1904 Details : The case where a directory component was not specified wasn't
1905 handled correctly. Handle this properly.
1908 Frequency : always on ppc64
1910 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1911 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1912 perform endianness conversion.
1915 Frequency : rarely (truncate to non-zero file size after write under load)
1916 Bugzilla : 10730, 10687
1917 Description: Files padded with zeros to next 4K multiple
1918 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1919 it is possible that files that are truncated to a non-zero size
1920 immediately after being written are filled with zero bytes beyond
1921 the truncated size. No file data is lost.
1923 Severity : enhancement
1925 Description: Allow recovery/failover for liblustre clients.
1926 Details : liblustre clients were unaware of failover configurations until
1929 Severity : enhancement
1931 Description: user file locks should fail when not mounting with flock option
1932 Details : Set up an error-returning stub in ll_file_operations.lock field
1933 to prevent incorrect behaviour when client is mounted without
1934 flock option. Also, set up properly f_op->flock field for
1938 Frequency : always on ia64
1940 Description: "lfs df" loops on printing out MDS statfs information
1941 Details : The obd_ioctl_data was not initialized and in some systems
1942 this caused a failure during the ioctl that did not return
1943 an error. Initialize the struct and return an error on failure.
1946 Frequency : SLES 9 only
1948 Description: Error of copying files with lustre special EAs as root
1949 Details : Client side always return success for setxattr call for lustre
1950 special xattr (currently only "trusted.lov").
1953 Frequency : rarely on clusters with both ia64+i386 clients
1955 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1956 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1957 clients are doing concurrent writes to the same file it is
1958 possible that the ia64 clients may LASSERT because the OST
1959 extent locks are not PAGE_SIZE aligned. Ensure that grown
1960 locks are always aligned on the request boundary.
1963 Frequency : specific use, occasional
1965 Description: Overwriting in use executable truncates on-disk binary image
1966 Details : If one node attempts to overwrite an executable in use by
1967 another node, we now correctly return ETXTBSY instead of
1968 truncating the file.
1970 Severity : enhancement
1972 Description: Async OSC create to avoid the blocking unnecessarily.
1973 Details : If a OST has no remain object, system will block on the creating
1974 when need to create a new object on this OST. Now, ways use
1975 pre-created objects when available, instead of blocking on an
1976 empty osc while others are not empty. If we must block, we block
1977 for the shortest possible period of time.
1982 Description: chmod on Lustre root is propagated to other clients
1983 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1984 invalid by the follow_mount time.
1987 Frequency : liblustre clients only
1989 Description: Race in 'instant cancel' lock handling could lead to such locks
1990 never to be granted in case of SMP MDS
1991 Details : Do not destroy not yet granted but cbpending locks in
1995 Frequency : replay/resend of open
1997 Description: non null lock assetion failure in mds_intent_policy
1998 Details : Trying to replay/resend lockless open requests resulted in
1999 mds_open() returning 0 with no lock. Now it sets a flag if
2000 a lock is going to be returned.
2002 Severity : enhancement
2004 Description: Checksum enhancements
2005 Details : New checksum enhancements allow for resending RPCs that failed
2008 Severity : enhancement
2010 Description: Tunables on number of dirty pages in cacche
2011 Details : Allow to set limit on number of dirty pages cached.
2016 Description: client crash on unmount - lock still has references
2017 Details : In some error handling cases it was possible to leak a lock
2018 reference on a client while accessing a file. This was not
2019 harmful to the client during operation, but would cause the
2020 client to crash when the filesystem is unmounted.
2023 Frequency : specific case, rare
2025 Description: ETXTBSY on mds though file not in use by client
2026 Details : ETXTBSY is no longer incorrectly returned when attempting to
2027 chmod or chown a directory that the user previously tried to
2028 execute or a currently-executing binary.
2031 Frequency : extremely rare except on liblustre-based clients
2033 Description: Lustre space not freed when files are deleted
2034 Details : Clean up open-unlinked files after client eviction. Previously
2035 the unlink was skipped and the files remained as orphans.
2040 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2041 Details : In some cases it was possible to send a blocking callback to
2042 a client doing a glimpse, even though that client didn't get
2043 a lock granted. When the glimpse lock is cancelled on the OST
2044 the freed lock is left on the waiting list and corrupted the list.
2047 Frequency : all core dumps
2049 Description: Broke core dumps to lustre
2050 Details : Negative dentry may be unhashed if parent does not have UPDATE
2051 lock, but some callers, e.g. do_coredump, expect dentry to be
2052 hashed after successful create, hash it in ll_create_it.
2054 ------------------------------------------------------------------------------
2056 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2058 * Support for kernels:
2059 2.6.9-42.0.2.EL (RHEL 4)
2060 2.6.5-7.276 (SLES 9)
2061 2.4.21-40.EL (RHEL 3)
2062 2.6.12.6 vanilla (kernel.org)
2066 Frequency : always on RHEL 3
2068 Description: Number of open files grows over time
2069 Details : The number of open files grows over time, whether or not
2070 Lustre is started. This was due to a filp leak introduced
2071 by one of our kernel patches.
2073 ------------------------------------------------------------------------------
2075 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2077 * Support for kernels:
2078 2.6.9-42.EL (RHEL 4)
2079 2.6.5-7.276 (SLES 9)
2080 2.4.21-40.EL (RHEL 3)
2081 2.6.12.6 vanilla (kernel.org)
2086 Bugzilla : 5719, 9635, 9792, 9684
2087 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2088 Details : If a server is under heavy load and cannot reply to new
2089 connection requests before the client resends the (re)connect,
2090 the connection handling code can behave badly if two service
2091 threads are concurrently handing separate (re)connections from
2092 the same client. Add better locking to the connection handling
2093 code, and ensure that only a single connection will be processed
2094 for a given client UUID, even if the lock is dropped.
2096 Severity : enhancement
2098 Description: add TCP zero-copy support to kernel
2099 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2100 sends if the hardware supports scatter-gather and checksumming.
2101 This allows socklnd to do client-write and server-read more
2102 efficiently and reduce CPU utilization from skbuf copying.
2105 Frequency : only if NFS exporting from client
2107 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2108 Details : If setting attributes on a file created under NFS that had
2109 never been opened it would be possible to oops the client
2110 if the file had no objects.
2113 Frequency : always for liblustre
2115 Description: liblustre client does MDS+OSTs setattr RPC for each write
2116 Details : When doing a write from a liblustre client, the client
2117 incorrectly issued an RPC to the MDS and each OST the file was
2118 striped over in order to update the timestamps. When writing
2119 with small chunks and many clients this could overwhelm the MDS
2120 with RPCs. In all cases it would slow down the write because
2121 these RPCs are unnecessary.
2123 Severity : enhancement
2125 Description: allow number of MDS service threads to be changed at module load
2126 Details : It is now possible to change the number of MDS service threads
2127 running. Adding "options mds mds_num_threads={N}" to the MDS's
2128 /etc/modprobe.conf will set the number of threads for the next
2129 time Lustre is restarted (assuming the "mds" module is also
2130 reloaded at that time). The default number of threads will
2131 stay the same, 32 for most systems.
2136 Description: OST crash if filesystem is unformatted or corrupt
2137 Details : If an OST is started on a device that has never been formatted
2138 or if the filesystem is corrupt and cannot even mount then the
2139 error handling cleanup routines would dereference a NULL pointer.
2144 Description: NULL pointer deref in llap_from_page.
2145 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2146 incorrect) mapping if the page was truncated/reclaimed while it was
2147 searched for. Check for this condition and skip such pages when
2148 doing readahead. Introduce extra check to llap_from_page() to
2149 verify page->mapping->host is non-NULL (so page is not anonymous).
2152 Frequency : Sometimes when using sys_sendfile
2154 Description: "page not covered by a lock" warnings from ll_readpage
2155 Details : sendfile called ll_readpage without right page locks present.
2156 Now we introduced ll_file_sendfile that does necessary locking
2157 around call to generic_file_sendfile() much like we do in
2161 Frequency : with certain MDS communication failures at client mount time
2163 Description: NULL pointer deref after failed client mount
2164 Details : a client connection request may delayed by the network layer
2165 and not be sent until after the PTLRPC layer has timed out the
2166 request. If the client fails the mount immediately it will try
2167 to clean up before the network times out the request. Add a
2168 reference from the request import to the obd device and delay
2169 the cleanup until the network drops the request.
2172 Frequency : occasionally during client (re)connect
2174 Description: assertion failure during client (re)connect
2175 Details : processing a client connection request may be delayed by the
2176 client or server longer than the client connect timeout. This
2177 causes the client to resend the connection request. If the
2178 original connection request is replied in this interval, the
2179 client may trip an assertion failure in ptlrpc_connect_interpret()
2180 which thought it would be the only running connect process.
2183 Frequency : only with obd_echo servers and clients that are rebooted
2185 Description: kernel BUG accessing uninitialized data structure
2186 Details : When running an obd_echo server it did not start the ping_evictor
2187 thread, and when a client was evicted an uninitialized data
2188 structure was accessed. Start the ping_evictor in the RPC
2189 service startup instead of the OBD startup.
2191 Severity : enhancement
2192 Bugzilla : 10193 (patchless)
2193 Description: Remove dependency on various unexported kernel interfaces.
2194 Details : No longer need reparent_to_init, exit_mm, exit_files,
2195 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2198 Frequency : rare (only users of deprecated and unsupported LDAP config)
2200 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2201 Details : LDAP apparently contains 'lustreName' attributes instead of
2202 'name'. A simple remapping of the name is sufficient.
2205 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2207 Description: waiting_locks_callback trips kernel BUG if client is evicted
2208 Details : Running with the dump_on_timeout debug flag turned on makes
2209 it possible that the waiting_locks_callback() can try to dump
2210 the Lustre kernel debug logs from an interrupt handler. Defer
2211 this log dumping to the expired_lock_main() thread.
2213 Severity : enhancement
2215 Description: Support NFS exporting on 2.6 kernels.
2216 Details : Implement non-rawops metadata methods for NFS server to use without
2217 changing NFS server code.
2220 Frequency : very rare (synthetic metadata workload only)
2222 Description: two racing renames might cause an MDS thread to deadlock
2223 Details : Running the "racer" program may cause one MDS thread to rename
2224 a file from being the source of a rename to being the target of
2225 a rename at exactly the same time that another thread is doing
2226 so, and the second thread has already enqueued these locks after
2227 doing a lookup of the target and is trying to relock them in
2228 order. Ensure that we don't try to re-lock the same resource.
2231 Frequency : only very large systems with liblustre clients
2233 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2234 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2235 all OSTs in parallel. This allows the eviction of stale liblustre
2236 clients to proceed much faster than if they were done in series,
2237 and also offers similar improvements for other set_info RPCs.
2242 Description: excessive CPU usage during initial read phase on client
2243 Details : During the initial read phase on a client, it would agressively
2244 retry readahead on the file, consuming too much CPU and impacting
2245 performance (since 1.4.5.8). Improve the readahead algorithm
2246 to avoid this, and also improve some other common cases (read
2247 of small files in particular, where "small" is files smaller than
2248 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2253 Description: MDS crash when receiving packet with unknown intent.
2254 Details : Do not LBUG in unknown intent case, just return -EFAULT
2256 Severity : enhancement
2257 Bugzilla : 9293, 9385
2258 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2259 Details : Do not serialize getattr (non-intent version) and statfs.
2262 Frequency : occasional, when OST network is overloaded/intermittent
2264 Description: client evicted by OST after bulk IO timeout
2265 Details : If a client sends a bulk IO request (read or write) the OST
2266 may evict the client if it is unresposive to its data GET/PUT
2267 request. This is incorrect if the network is overloaded (takes
2268 too long to transfer the RPC data) or dropped the OST GET/PUT
2269 request. There is no need to evict the client at all, since
2270 the pinger and/or lock callbacks will handle this, and the
2271 client can restart the bulk request.
2274 Frequency : Always when mmapping file with no objects
2276 Description: client crashes when mmapping file with no objects
2277 Details : Check that we actually have objects in a file before doing any
2278 operations on objects in ll_vm_open, ll_vm_close and
2284 Description: Request leak when working with deleted CWD
2285 Details : Introduce advanced request refcount tracking for requests
2286 referenced from lustre intent.
2288 Severity : Enhancement
2290 Description: Cache open file handles on client.
2291 Details : MDS now will return special lock along with openhandle, if
2292 requested and client is allowed to hold openhandle, even if unused,
2293 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2294 closing files for every read/write openration.
2296 Severity : Enhancement
2298 Description: Cache open negative dentries on client when possible.
2299 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2300 negative dentries on lock revocation.
2305 Description: Remounting a client read-only wasn't possible with a zconf mount
2306 Details : It wasn't possible to remount a client read-only with llmount.
2308 Severity : enhancement
2309 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2310 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2311 MPICH is included to add Lustre-specific ADIO interfaces.
2312 This is based closely on the UFS ADIO layer and only differs
2313 in file creation, in order to allow the OST striping to be set.
2314 This is user-contributed code and not supported by CFS.
2319 Description: extended inode attributes (immutable, append-only) work improperly
2320 when 2.4 and 2.6 kernels are used on client/server or vice versa
2321 Details : Introduce kernel-independent values for these flags.
2323 Severity : enhancement
2326 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2327 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2328 granularity. Now, Fractional MB granularity is be supported,
2329 this is very useful for low memory system.
2331 Severity : enhancement
2333 Description: Getattr by fid
2334 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2335 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2336 have locks and avoids getattr rpc requests for it on every stat.
2339 Frequency : Always, for filesystems larger than 2TB
2341 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2342 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2343 filesystems larger than 2TB to be mounted with mballoc enabled.
2346 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2348 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2349 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2350 filesystems larger than 2TB is mounted then the kernel will
2351 silently corrupt the start of the filesystem. CONFIG_LBD is
2352 enabled for all CFS-supported kernels, but the possibility of
2353 this happening with a modified kernel config exists.
2355 Severity : enhancement
2357 Description: add client O_DIRECT support for 2.6 kernels
2358 Details : It is now possible to do O_DIRECT reads and writes to files
2359 in the Lustre client mountpoint on 2.6 kernel clients.
2361 Severity : enhancement
2363 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2364 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2365 parallel, not waiting for response from every OST before sending
2366 a rpc to the next OST.
2371 Description: setattr vs write race when updating file timestamps
2372 Details : Client processes that update a file timestamp into the past
2373 right after writing to the file (e.g. tar) it is possible that
2374 the updated file modification time can be reset to the current
2375 time due to a race between processing the setattr and write RPC.
2377 Severity : enhancement
2379 Description: Bring 'lfs find' closer in line with regular Linux find.
2380 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2381 -print0 options and obtains all the needed info through the lustre
2384 Severity : enhancement
2386 Description: support up to 1024 configured devices on one node
2387 Details : change obd_dev array from statically allocated to dynamically
2388 allocated structs as they are first used to reduce memory usage
2393 Description: Flush dirty partially truncated pages during truncate
2394 Details : Immediatelly flush partially truncated pages in filter_setattr,
2395 this way we completely avoid having any pages in page cache on OST
2396 and can retire ugly workarounds during writes to flush such pages.
2401 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2402 Details : protect inode from truncation within vfs_unlink() context
2403 just take a reference before calling vfs_unlink() and release it
2404 when parent's i_sem is free.
2409 Description: last_id value checked outside lock on OST caused LASSERT failure
2410 Details : If there were multiple MDS->OST object precreate requests in
2411 flight, it was possible that the OST's last object id was checked
2412 outside a lock and incorrectly tripped an assertion. Move checks
2413 inside locks, and discard old precreate requests.
2416 Frequency : always, if extents are used on OSTs
2418 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2419 Details : OSTs using the extents format would not zero the high 16 bits of
2420 the index physical block number. This is not a problem for any
2421 OST filesystems smaller than 16TB, and no kernels support ext3
2422 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2423 new/modified files) and can be fixed for existing filesystems
2424 with e2fsprogs-1.39-cfs1.
2429 Description: import connection selection may be incorrect if timer wraps
2430 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2431 import connections to be ignored if the 32-bit jiffies counter
2432 wraps. Use a 64-bit jiffies counter.
2435 Frequency : very large clusters immediately after boot
2437 Description: LNET request buffers exhausted under heavy short-term load
2438 Details : If a large number of client requests are generated on a service
2439 that has previously never seen so many requests it is possible
2440 that the request buffer growth cannot keep up with the spike in
2441 demand. Instead of dropping incoming requests, they are held in
2442 the LND until the RPC service can accept more requests.
2445 Frequency : Sometimes during replay
2447 Description: Assertion failure in ll_local_open after replay.
2448 Details : If replay happened on an open request reply before we were able
2449 to set replay handler, reply will become not swabbed tripping the
2450 assertion in ll_local_open. Now we set the handler right after
2451 recognising of open request
2454 Frequency : very rare
2456 Description: kernel reports "badness in vsnprintf"
2457 Details : Reading from the "recovery_status" /proc file in small chunks
2458 may cause a negative length in lprocfs_obd_rd_recovery_status()
2459 call to vsnprintf() (which is otherwise harmless). Exit early
2460 if there is no more space in the output buffer.
2462 Severity : enhancement
2464 Description: clear OBD RPC statistics by writing to them
2465 Details : It is now possible to clear the OBD RPC statistics by writing
2466 to the "stats" file.
2471 Description: Client mtime is not the same on different clients after utimes
2472 Details : In some cases, the client was using the utimes() syscall on
2473 a file cached on another node. The clients now validate the
2474 ctime from the MDS + OSTs to determine which one is right.
2479 Description: Inability to activate failout mode
2480 Details : lconf script incorrectly assumed that in pythong string's numeric
2481 value is used in comparisons.
2484 Frequency : always with multiple stripes per file
2486 Description: Inefficient object allocation for mutli-stripe files
2487 Details : When selecting which OSTs to stripe files over, for files with
2488 a stripe count that divides evenly into the number of OSTs,
2489 the MDS is always picking the same starting OST for each file.
2490 Return the OST selection heuristic to the original design.
2495 Description: mount failures may take full timeout to return an error
2496 Details : Under some heavy load conditions it is possible that a
2497 failed mount can wait for the full obd_timeout interval,
2498 possibly several minutes, before reporting an error.
2499 Instead return an error as soon as the status is known.
2501 Frequency : quota enabled and large files being deleted
2503 Description: releasing more than 4GB of quota at once hangs OST
2504 Details : If a user deletes more than 4GB of files on a single OST it
2505 will cause the OST to spin in an infinite loop. Release
2506 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2511 Description: statfs data retrieved from /proc may be stale or zero
2512 Details : When reading per-device statfs data from /proc, in the
2513 {kbytes,files}_{total,free,avail} files, it may appear
2514 as zero or be out of date.
2517 Frequency : systems with MD RAID1 external journal devices
2519 Description: lconf's call to blkid is confused by RAID1 journal devices
2520 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2521 of returning all block devices that match the journal UUID.
2524 Frequency : always, for aggregate stripe size over 4GB
2526 Description: assertion fails when trying to use 4GB stripe size
2527 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2528 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2531 Frequency : always on ppc64
2533 Description: the first write on an ext3 filesystem with mballoc got stuck
2534 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2535 perform endianness conversion.
2537 ------------------------------------------------------------------------------
2539 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2541 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2542 INTEROPERATE with older versions automatically. Please read the
2543 user documentation before upgrading any part of a live system.
2544 * WARNING: Lustre networking configuration changes are required with
2545 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2548 * Support for kernels:
2549 2.6.9-22.0.2.EL (RHEL 4)
2550 2.6.5-7.244 (SLES 9)
2551 2.6.12.6 vanilla (kernel.org)
2554 Severity : enhancement
2555 Bugzilla : 7981/8208
2556 Description: Introduced Lustre Networking (LNET)
2557 Details : LNET is new networking infrastructure for Lustre, it includes
2558 a reorganized network configuration mode (see the user
2559 documentation for full details) as well as support for routing
2560 between different network fabrics. Lustre Networking Devices
2561 (LNDs) for the supported network fabrics have also been
2562 created for this new infrastructure.
2564 Severity : enhancement
2565 Description: Introduced Access control lists
2566 Details : clients can set ACLs on files and directories in order to have
2567 more fine-grained permissions than the standard Unix UGO+RWX.
2568 The MDS must be started with the "-o acl" mount option.
2570 Severity : enhancement
2571 Description: Introduced filesystem quotas
2572 Details : Administrators may now establish per-user quotas on the
2575 Severity : enhancement
2577 Description: Configuration change for the XT3
2578 The PTLLND is now used to run Lustre over Portals on the XT3
2579 The configure option(s) --with-cray-portals are no longer used.
2580 Rather --with-portals=<path-to-portals-includes> is used to
2581 enable building on the XT3. In addition to enable XT3 specific
2582 features the option --enable-cray-xt3 must be used.
2587 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2588 Details : A race between allocating a new llcd and re-getting the llcd_lock
2589 allowed another thread to grab newly-allocated llcd.
2591 Severity : enhancement
2593 Description: 2.6 OST async journal commit and locking fix to improve performance
2594 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2595 2.6 kernels are now async as they already were in 2.4 kernels so
2596 that they can commit concurrently with the network bulk transfer.
2597 For block-allocated files the filter allocation semaphore is held
2598 to avoid filesystem fragmentation during allocation. BKL lock
2599 removed for 2.6 xattr operations where it is no longer needed.
2604 Description: lconf incorrectly determined whether two IP networks could talk
2605 Details : In some more complicated routing and multiple-network
2606 configurations, lconf will avoid trying to make a network
2607 connection to a disjoint part of the IP space. It was doing the
2608 math incorrectly for one set of cases.
2613 Description: Fix for potential infinite loop processing records in an llog.
2614 Details : If an llog record is corrupted/zeroed, it is possible to loop
2615 forever in llog_process(). Validate the llog record length
2616 and skip the remainder of the block on error.
2619 Frequency : occasional (liblustre only)
2621 Description: liblustre could not open files whose last component is a symlink
2622 Details : sysio_path_walk() would incorrectly pass the open intent to
2623 intermediate path components.
2626 Frequency : rare (liblustre only with non-standard tuning)
2627 Bugzilla : 7201 (7350)
2628 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2629 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2630 after it has been released, only do this for referenced locks
2632 Severity : enhancement
2634 Description: specifying an (invalid) directory default stripe_size of -1
2635 would reset the directory default striping
2636 Details : stripe_size -1 was used internally to signal directory stripe
2637 removal, now use "all default" to signal dir stripe removal
2638 as a directory striping of "all default" is not useful
2641 Frequency : common for large clusters running liblustre clients
2643 Description: doing an ls when liblustre clients are running is slow
2644 Details : sending a glimpse AST to a liblustre client waits for every AST
2645 to time out, as liblustre clients will not respond. Since they
2646 cannot cache data we refresh the OST lock LVB from disk instead.
2648 Severity : enhancement
2650 Description: doing an ls at the same time as file IO can be slow
2651 Details : enqueue and other "small" requests can be blocked behind many
2652 large IO requests. Create a new OST IO portal for non-IO
2653 requests so they can be processed faster.
2656 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2658 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2659 created file with mode 000
2660 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2661 allow an ftruncate() as a non-root user to a file with mode 000.
2662 Lustre now respects this flag to disable mode checks when
2663 truncating a file owned by the user
2666 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2669 Description: Revoking locks from clients that went dead or catatonic might take
2671 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2672 cancellation of such locks instant on servers without waiting for
2673 any reply from clients. Clients drops these locks when cancel
2674 notification from server is received without replying.
2677 Frequency : liblustre-only, when liblustre client dies or becomes busy
2679 Description: Doing ls on Linux clients can take a long time with active
2681 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2682 granting such locks to it in the first place if possible. Locks
2683 are taken by proxy on the OST during the read or write and
2684 dropped immediately afterward. Add connect flags handling, do
2685 not grant locks to liblustre clients for glimpse ASTs.
2687 Severity : enhancement
2689 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2690 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2691 sequential reads and behaves badly with applications doing random
2692 reads. Improve it by reading ahead at least read region, and
2693 avoiding excessive large RPC for small reads.
2695 Severity : enhancement
2697 Description: Creating more than 1000 files for a single job may cause a load
2698 imbalance on the OSTs if there are also a large number of OSTs.
2699 Details : qos_prep_create() uses an OST index reseed value that is an
2700 even multiple of the number of available OSTs so that if the
2701 reseed happens in the middle of the object allocation it will
2702 still utilize the OSTs as uniformly as possible.
2707 Description: OST or MDS may oops in ping_evictor_main()
2708 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2709 but doesn't restart at beginning of obd_exports_timed list
2712 Severity : enhancement
2714 Description: improve by-nid export eviction on the MDS and OST
2715 Details : allow multiple exports with the same NID to be evicted at one
2716 time without re-searching the exports list.
2719 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2721 Description: MDS may oops in groups_free()
2722 Details : in rare race conditions a newly allocated group_info struct is
2723 freed again, and this can be NULL. The 2.4 compatibility code
2724 for groups_free() checked for a NULL pointer, but 2.6 did not.
2727 Frequency : common for liblustre clients doing little filesystem IO
2728 Bugzilla : 9352, 7313
2729 Description: server may evict liblustre clients accessing contended locks
2730 Details : if a client is granted a lock or receives a completion AST
2731 with a blocking AST already set it would not reply to the AST
2732 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2735 Frequency : lfs setstripe, only systems with more than 160 OSTs
2737 Description: unable to set striping with a starting offset beyond OST 160
2738 Details : llapi_create_file() incorrectly limited the starting stripe
2739 index to the maximum single-file stripe count.
2742 Frequency : LDAP users only
2744 Description: lconf did not handle in-kernel recovery with LDAP properly
2745 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2747 Severity : enhancement
2749 Description: bind OST threads to NUMA nodes to improve performance
2750 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2751 node and do their allocations there to localize memory access
2753 Severity : enhancement
2755 Description: llmount can determine client NID directly from Myrinet (GM)
2756 Details : the client NID code from gmnalnid was moved directly into
2757 llmount, removing the need to use this or specifying the
2758 client NID explicitly when mounting GM clients with zeroconf
2761 Frequency : if client is started with down MDS
2763 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2764 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2765 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2766 ptlrpc_import_delayed_req() because we didn't check if the
2767 request was interrupted, and we also didn't break out of the
2768 event loop if there was no timeout
2773 Description: data loss during non-page-aligned writes to a single file from
2774 both multiple nodes and multiple threads on one node at same time
2775 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2776 inconsistency led to false short-reads, that were cached and later
2777 used by ->prepare_write() to fill in partially written page,
2778 leading to data loss.
2781 Frequency : always, if lconf --abort_recovery used
2783 Description: lconf --abort_recovery fails with 'Operation not supported'
2784 Details : lconf was attempting to abort recovery on the MDT device and not
2787 Severity : enhancement
2789 Description: remove cleanup logs
2790 Details : replace lconf-generated cleanup logs with lustre internal
2791 cleanup routines. Eliminates the need for client-cleanup and
2794 Severity : enhancement
2796 Description: add support for EAs (user and system) on lustre filesystems
2797 Details : it is now possible to store extended attributes in the Lustre
2798 client filesystem, and with the user_xattr mount option it
2799 is possible to allow users to store EAs on their files also
2801 Severity : enhancement
2803 Description: Add possibility (config option) to show minimal available OST free
2805 Details : When compiled with --enable-mindf configure option, statfs(2)
2806 (and so, df) will return least minimal free space available from
2807 all OSTs as amount of free space on FS, instead of summary of
2808 free spaces of all OSTs.
2810 Severity : enhancement
2812 Description: do not expand extent locks acquired on OST-side
2813 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2814 by server: they are not cached anyway.
2817 Frequency : when mmap is used/binaries executed from Lustre
2819 Description: Unmmap pages before throwing them away from read cache.
2820 Details : llap_shrink cache now attempts to unmap pages before discarding
2821 them (if unmapping failed - do not discard). SLES9 kernel has
2822 extra checks that trigger if this unmapping is not done first.
2827 Description: lconf didn't resolve symlinks before checking to see whether a
2828 given mountpoint was already in use
2831 Frequency : when migrating failover services
2832 Bugzilla : 6395, 9514
2833 Description: When migrating a subset of services from a node (e.g. failback
2834 from a failover service node) the remaining services would
2835 time out and evict clients.
2836 Details : lconf --force (implied by --failover) sets the global obd_timeout
2837 to 5 seconds in order to quickly disconnect, but this caused
2838 other RPCs to time out too quickly. Do not change the global
2839 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2841 Severity : enhancement
2842 Frequency : if MDS is started with down OST
2843 Bugzilla : 9439,5706
2844 Description: Allow startup/shutdown of an MDS without depending on the
2845 availability of the OSTs.
2846 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2847 Add appropriate locking and lov-osc refcounts for safe
2848 cleaning. Add osc abort_inflight calls in case the
2849 synchronize never started.
2852 Frequency : occasional (Cray XT3 only)
2854 Description: root not authorized to access files in CRAY_PORTALS environment
2855 Details : The client process capabilities were not honoured on the MDS in
2856 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2857 been accessed by an authorized user then root was able to access
2858 the file on the local client also. The root user capabilities
2859 are now allowed on the MDS, as this environment has secure UID.
2862 Frequency : occasional
2864 Description: ldiskfs "too long searching" message happens too often
2865 Details : A debugging message (otherwise harmless) prints too often on
2866 the OST console. This has been reduced to only happen when
2867 there are fragmentation problems on the filesystem.
2872 Description: Division by zero in statfs when all OSCs are inactive
2873 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2874 lov_statfs divides by value returned by lov_get_stripecnt().
2878 Bugzilla : 9489, 3273
2879 Description: First write from each client to each OST was only 4kB in size,
2880 to initialize client writeback cache, which caused sub-optimal
2881 RPCs and poor layout on disk for the first writen file.
2882 Details : Clients now request an initial cache grant at (re)connect time
2883 and so that they can start streaming writes to the cache right
2884 away and always do full-sized RPCs if there is enough data.
2885 If the OST is rebooted the client also re-establishes its grant
2886 so that client cached writes will be honoured under the grant.
2891 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2892 Details : Now I/O RPCs go to different portal number and (presumably) fast
2893 lock requests (and glimses) and other RPCs get their own service
2894 threads pool that should be able to service those RPCs
2897 Severity : enhancement
2899 Description: Ability to exchange lustre version between client and servers and
2900 issue warnings at client side if client is too old. Also for
2901 liblustre clients there is ability to refuse connection of too old
2903 Details : New 'version' field is added to connect data structure that is
2904 filled with version info. That info is later checked by server and
2908 Frequency : rare, liblustre only.
2909 Bugzilla : 9296, 9581
2910 Description: Two simultaneous writes from liblustre at offset within same page
2911 might proceed at the same time overwriting eachother with stale
2913 Details : I/O lock withing llu_file_prwv was released too early, before data
2914 actually was hitting the wire. Extended lock-holding time until
2915 server acknowledges receiving data.
2918 Frequency : extremely rare. Never observed in practice.
2920 Description: avoid generating lustre_handle cookie of 0.
2921 Details : class_handle_hash() generates handle cookies by incrementing
2922 global counter, and can hit 0 occasionaly (this is unlikely, but
2923 not impossible, because initial value of cookie counter is
2924 selected randonly). Value of 0 is used as a sentinel meaning
2925 "unassigned handle" --- avoid it. Also coalesce two critical
2926 sections in this function into one.
2928 Severity : enhancement
2930 Description: allow liblustre clients to delegate truncate locking to OST
2931 Details : To avoid overhead of locking, liblustre client instructs OST to
2932 take extent lock in ost_punch() on client's behalf. New connection
2933 flag is added to handle backward compatibility.
2935 Severity : enhancement
2936 Bugzilla : 4928, 7341, 9758
2937 Description: allow number of OST service threads to be specified
2938 Details : a module parameter allows the number of OST service threads
2939 to be specified via "options ost ost_num_threads={N}" in the
2940 OSS's /etc/modules.conf or /etc/modprobe.conf.
2944 Bugzilla : 6146, 9635, 9895
2945 Description: servers crash with bad pointer in target_handle_connect()
2946 Details : In rare cases when a client is reconnecting it was possible that
2947 the connection request was the last reference for that export.
2948 We would temporarily drop the export reference and get a new
2949 one, but this may have been the last reference and the export
2950 was just destroyed. Get new reference before dropping old one.
2952 Severity : enhancement
2953 Frequency : if client is started with failover MDS
2955 Description: Allow multiple MDS hostnames in the mount command
2956 Details : Try to read the configuration from all specified MDS
2957 hostnames during a client mount in case the "primary"
2960 Severity : enhancement
2962 Description: Stop sending data to evicted clients as soon as possible.
2963 Details : Check if the client we are about to send or are sending data to
2964 was evicted already. (Check is done every second of waiting,
2965 for which l_wait_event interface was extended to allow checking
2966 of exit condition at specified intervals).
2969 Frequency : rare, normally only when NFS exporting is done from client
2971 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2973 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2976 Frequency : always, if srand() is called before liblustre initialization
2978 Description: Liblustre uses system PRNG disturbing its usage by user application
2979 Details : Introduce internal to lustre fast and high-quality PRNG for
2980 lustre usage and make liblustre and some other places in generic
2981 lustre code to use it.
2983 Severity : enhancement
2984 Bugzilla : 9477, 9557, 9870
2985 Description: Verify that the MDS configuration logs are updated when xml is
2986 Details : Check if the .xml configuration logs are newer than the config
2987 logs stored on the MDS and report an error if this is the case.
2988 Request --write-conf, or allow starting with --old_conf.
2990 Severity : enhancement
2992 Description: Handle symlinks in the path when checking if Lustre is mounted.
2993 Details : Resolve intermediate symlinks when checking if a client has
2994 mounted a filesystem to avoid duplicate client mounts.
2999 Description: lconf can hit an error exception but still return success.
3000 Details : The lconf command catches the Command error exception at the top
3001 level script context and will exit with the associated exit
3002 status, but doesn't ensure that this exit status is non-zero.
3007 Description: failure of ptlrpc thread startup can cause oops
3008 Details : Starting a ptlrpc service thread can fail if there are a large
3009 number of threads or the server memory is very fragmented.
3010 Handle this without oopsing.
3013 Frequency : always, only if liblustre and non-default acceptor port was used
3015 Description: liblustre cannot connect to servers with non-default acceptor port
3016 Details : tcpnal_set_default_params() was not called and was therefore
3017 ignoring the environment varaible TCPNAL_PORT, as well as other
3018 TCPNAL_ environment variables
3023 Description: two objects could be created on the same OST for a single file
3024 Details : If an OST is down, in some cases it was possible to create two
3025 objects on a single OST for a single file. No problems other
3026 than potential performance impact and spurious error messages.
3030 Bugzilla : 5681, 9562
3031 Description: Client may oops in ll_unhash_aliases
3032 Details : Client dcache may become inconsistent in race condition.
3033 In some cases "getcwd" can fail if the current directory is
3039 Description: Inode refcounting problems in NFS export code
3040 Details : link_raw functions used to call d_instantiate without obtaining
3041 extra inode reference first.
3045 Bugzilla : 9942, 9903
3046 Description: Referencing freed requests leading to crash, memleaks with NFS.
3047 Details : We used to require that call to ll_revalidate_it was always
3048 followed by ll_lookup_it. Also with revalidate_special() it is
3049 possible to call ll_revalidate_it() twice for the same dentry
3050 even if first occurence returned success. This fix changes semantic
3051 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3052 reference on a request referred from the intent.
3053 ll_intent_release() then releases such a request.
3056 Frequency : rare, normally benchmark loads only
3058 Description: unlinked inodes were kept in memory on the client
3059 Details : If a client is repeatedly creating and unlinking files it
3060 can accumulate a lot of stale inodes in the inode slab cache.
3061 If there is no other client load running this can cause the
3062 client node to run out of memory. Instead flush old inodes
3063 from client cache that have the same inode number as a new inode.
3066 Frequency : SLES9 2.6.5 kernel and long filenames only
3067 Bugzilla : 9969, 10379
3068 Description: utime reports stale NFS file handle
3069 Details : SLES9 uses out-of-dentry names in some cases, which confused
3070 the lustre dentry revalidation. Change it to always use the
3074 Frequency : rare, unless heavy write-truncate concurrency is continuous
3075 Bugzilla : 4180, 6984, 7171, 9963, 9331
3076 Description: OST becomes very slow and/or deadlocked during object unlink
3077 Details : filter_destroy() was holding onto the parent directory lock
3078 while truncating+unlinking objects. For very large objects this
3079 may block other threads for a long time and slow overall OST
3080 responsiveness. It may also be possible to get a lock ordering
3081 deadlock in this case, or run out of journal credits because of
3082 the combined truncate+unlink. Solution is to do object truncate
3083 first in one transaction without parent lock, and then do the
3084 final unlink in a new transaction with the parent lock. This
3085 reduces the lock hold time dramatically.
3088 Frequency : rare, 2.4 kernels only
3090 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3091 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3092 before unmounting their filesystem in order to avoid deadlock.
3093 The kernel_locked() function in 2.4 kernels only checks whether
3094 the kernel lock is held, not whether it is this process that is
3095 holding it as 2.6 kernels do.
3100 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3101 Details : The client ptlrpc code may be trying to reconnect to a down
3102 server before a previous connection attempt has timed out.
3103 Increase the reconnect interval to be longer than the connection
3104 timeout interval to avoid sending duplicate connections to
3108 Frequency : echo_client brw_test command
3110 Description: fix echo_client to work with OST preallocated code
3111 Details : OST preallocation code (5137) didn't take echo_client IO path
3112 into account: echo_client calls filter methods outside of any
3113 OST thread and, hence, there is no per-thread preallocated
3114 pages and buffers to use. Solution: hijack pga pages for IO. As
3115 a byproduct, this avoids unnecessary data copying.
3119 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3120 Description: Client can oops in mdc_commit_close() after open replay
3121 Details : It was possible for the MDS to return an open request with no
3122 transaction number in mds_finish_transno() if the client was
3123 evicted, but without actually returning an error. Clients
3124 would later try to replay that open and may trip an assertion
3125 Simplify the client close codepath, and always return an error
3126 from the MDS in case the open is not successful.
3129 Frequency : rare, 2.6 OSTs only
3131 Description: OST may deadlock under high load on fragmented files
3132 Details : If there was a heavy load and highly-fragmented OST filesystems
3133 it was possible to have all the OST threads deadlock waiting on
3134 allocation of biovecs, because the biovecs were not released
3135 until the entire RPC IO was completed. Instead, release biovecs
3136 as soon as they are complete to ensure forward IO progress.
3138 Severity : enhancement
3140 Description: Support for specifying external journal device at mount
3141 Details : If an OST or MDS device is formatted with an external journal
3142 device, this device major/minor is stored in the ext3 superblock
3143 and may not be valid for failover. Allow detecting and
3144 specifying the external journal at mount time.
3149 Description: Mounting an MDS with pending unlinked files may cause oops
3150 Details : target_finish_recovery() calls mds_postrecov() which returned
3151 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3152 considers this an error and immediately begins cleaning up the
3153 lov, just after starting the mds_lov process
3155 Severity : enhancement
3157 Description: Implement 'lfs df' to report actual free space on per-OST basis
3158 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3159 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3160 usage of objects; '-h' to report in human readable format.
3162 ------------------------------------------------------------------------------
3164 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3171 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3172 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3173 at startup during orphan unlinking. Instead, skip buddy update
3174 before setup, as it will be regenerated after recovery is complete.
3179 Description: If an OST is inactive, its locks might reference stale inodes.
3180 Details : lov_change_cbdata() must iterate over all namespaces, even if
3181 they are inactive to clear inode references from the lock.
3183 Severity : enhancement
3184 Frequency : occasional, if non-standard max_dirty_mb used
3186 Description: Client will block write RPCs if not enough grant
3187 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3188 then the client will block writes while waiting for another RPC
3189 to complete instead of consuming its dirty limit. With change
3190 we get improved performance when max_dirty_mb is small.
3192 Severity : enhancement
3193 Bugzilla : 3389, 6253
3194 Description: Add support for supplementary groups on the MDS.
3195 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3196 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3197 lookups for user supplementary groups into a cache.
3201 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3202 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3203 for liblustre O_EXCL is enforced by MDS.
3207 Description: readdir never returns NULL in liblustre.
3208 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3209 dentry in struct dirent.
3213 Frequency : liblustre only, depends on application IO pattern
3214 Description: liblustre clients evicted if not contacting servers
3215 Details : Don't put liblustre clients into the ping_evictor list, so
3216 they will not be evicted by the pinger ever.
3218 Severity : enhancement
3220 Description: Add ability to evict clients by NID from MDS.
3221 Details : By echoing "nid:$NID" string into
3222 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3223 $NID would be instantly evicted from this MDS and from all active
3224 OSTs connected to it.
3228 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3229 Details : lookup_it_finish() used to query file size from OSTs that was not
3234 Description: service threads change working directory to that of init
3235 Details : Starting lustre service threads may pin the working directory
3236 of the parent thread, making that filesystem busy. Threads
3237 now change to the working directory of init to avoid this.
3241 Frequency : during shutdown only
3242 Description: shutdown with a failed MDS or OST can cause unmount to hang
3243 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3248 Frequency : 2.6 only
3249 Description: chmod/chown may include an extra supplementary group
3250 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3251 supplementary group and if none is specified this is used.
3254 Bugzilla : 5479 (6816)
3256 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3257 Details : If lookup is in progress on a file that is unlinked we might try
3258 to revalidate the inode and fail in revalidate after lookup is
3259 complete and ll_file_open() enqueues the open again but
3260 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3263 Frequency : always, if lconf --abort_recovery used
3265 Description: lconf --abort_recovery fails with 'Operation not supported'
3266 Details : lconf was attempting to abort recovery on the MDT device and not
3269 ------------------------------------------------------------------------------
3271 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3276 Frequency : rare (only unsupported configurations with a node running as an
3278 Bugzilla : 6514, 5137
3279 Description: Mounting a Lustre file system on a node running as an OST could
3281 Details : OSTs now preallocates memory needed to write out data at
3282 startup, instead of when needed, to avoid having to
3283 allocate memory in possibly low memory situations.
3284 Specifically, if the file system is mounted on on OST,
3285 memory pressure could force it to try to write out data,
3286 which it needed to allocate memory to do. Due to the low
3287 memory, it would be unable to do so and the node would
3288 become unresponsive.
3290 Severity : enhancement
3292 Description: Addition of lconf --service command line option
3293 Details : lconf now accepts a '--service <arg>' option, which is
3294 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3296 Severity : enhancement
3298 Description: Failover mode is now the default for OSTs.
3299 Details : By default, OSTs will now run in failover mode. To return to
3300 the old behaviour, add '--failout' to the lmc line for OSTs.
3302 Severity : enhancement
3304 Description: Health checks are now provided for MDS and OSTs
3305 Details : Additional detailed health check information on MSD and OSTs
3306 is now provided through the procfs health_check value.
3309 Frequency : occasional, depends on IO load
3311 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3312 after numerous create/delete cycles
3313 Details : The ext3 inode allocation policy would not allocate new inodes
3314 very well on the OSTs because there are no new directories
3315 being created. Instead we look for groups with free space if
3316 the parent directories are nearly full.
3321 Description: Network or server problems during mount may cause partially
3322 mounted clients instead of returning an error.
3323 Details : The config llog parsing code may overwrite the error return
3324 code during mount error handling, returning success instead
3330 Description: MDS can fail to allocate large reply buffers
3331 Details : After long uptimes the MDS can fail to allocate large reply
3332 buffers (e.g. zconf client mount config records) due to memory
3333 fragmentation or consumption by the buffer cache. Preallocate
3334 some large reply buffers so that these replies can be sent even
3335 under memory pressure.
3339 Frequency : rare (liblustre)
3340 Description: fsx running with liblustre complained that using truncate() to
3341 extend the file doesn't work. This patch corrects that issue.
3342 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3343 ATTR_SIZE and lsm use in llu_setattr_raw.
3347 Frequency : rare, only 2.6 kernels
3348 Description: Unusual file access patterns on the MDS may result in inode
3349 data being lost in very rare circumstances.
3350 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3351 code in the 2.6 kernel caused the inode and/or EA data not to
3352 be read from disk, causing single-file corruption.
3356 Frequency : rare, only 2.6 filesystems using extents
3357 Description: Heavy concurrent write and delete load may cause data corruption.
3358 Details : It was possible under high-load situations to have an extent
3359 metadata block in the block device cache from a just-unlinked
3360 file overwrite a newly-allocated data block. We now unmap any
3361 metadata buffers that alias just-allocated data blocks.
3365 Frequency : filesystems with default stripe_count larger than 77
3366 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3367 Details : lconf specifies an inode size of 4096 bytes when the default
3368 stripe_count is larger than 77. This conflicts with the default
3369 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3370 this case to avoid pinning too much memory for large EAs.
3372 ------------------------------------------------------------------------------
3374 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3379 Frequency : rare (extremely heavy IO load with hundreds of clients)
3381 Description: Client is evicted, gets IO error writing to file
3382 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3383 caused clients to be evicted for AST timeouts. The fixes in
3384 bug 5192 mean we no longer need to have such short AST timeouts
3385 so ldlm_timeout has been increased.
3388 Frequency : occasional during --force or --failover shutdown under load
3389 Bugzilla : 5949, 4834
3390 Description: Server oops/LBUG if stopped with --force or --failover under load
3391 Details : a collection of import/export refcount and cleanup ordering
3392 issues fixed for safer force cleanup
3395 Frequency : only filesystems larger than 120 OSTs
3396 Bugzilla : 5990, 6223
3397 Description: lfs getstripe would oops on a very large filesystem
3398 Details : lov_getconfig used kfree on vmalloc'd memory
3401 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3402 Bugzilla : 6242, 6243
3403 Description: reading from files that had been truncated to a non-zero size
3404 but never opened returned no data
3405 Details : ll_file_read() reads zeros from no-object files to EOF
3410 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3411 Details : There's little harm in aborting MDS/OSS recovery and letting it
3412 try again, so I removed the LASSERT and return an error instead.
3414 Severity : enhancement
3416 Description: New debugging infrastructure for tracking down data corruption
3417 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3418 (b) cover more of the client-side code path, and (c) try to narrow
3419 down where problems occurred
3423 Bugzilla : 3819, 4364, 4397, 6313
3424 Description: Racing close and eviction MDS could cause assertion in mds_close
3425 Details : It was possible to get multiple mfd references during close and
3426 client eviction, leading to one thread referencing a freed mfd.
3428 Severity: : enhancement
3429 Bugzilla : 3262, 6359
3430 Description: Attempts to reconnect to servers are now more aggressive.
3431 Details : This builds on the enhanced upcall-less recovery that was added
3432 in 1.4.2. When trying to reconnect to servers, clients will
3433 now try each server in the failover group every 10 seconds. By
3434 default, clients would previously try one server every 25 seconds.
3439 Description: After recovery, certain operations trigger a failed
3440 assertion on a client.
3441 Details : Failing over an mds, using lconf -d --failover, while a
3442 client was doing a readdir() call would cause the client to
3443 LBUG after recovery completed and the readdir() was resent.
3445 Severity : enhancement
3447 Description: Default groups are now added by lconf
3448 Details : You can now run lconf --group <servicename> without having to
3449 manually add groups with lmc.
3452 Frequency : occasional
3454 Description: Nodes with an elan id of 0 trigger a failed assertion
3457 Frequency : always when accessing e.g. tty/console device nodes
3459 Description: tty and some other devices nodes cannot be used on lustre
3460 Details : file's private_data field is used by device data and lustre
3461 values in there got lost. New field was added to struct file to
3462 store fs-specific private data.
3465 Frequency : when exporting Lustre via NFS
3467 Description: NFSD failed occasionally when looking up a path component
3468 Details : NFSD is looking up ".." which was broken in ext3 directories
3469 that had grown large enough to become hashed.
3472 Frequency : Clusters with multiple interfaces not on the same subnet
3474 Description: Nodes will repeatedly try to reconnect to an interface which it
3475 cannot reach and report an error to the log.
3476 Details : Extra peer list entries will be created by lconf with some peers
3477 unreachable. lconf now validates the peer before adding it.
3480 Frequency : Only if a default stripe is set on the filesystem root.
3482 Description: Setting a default stripe on the filesystem root prevented the
3483 filesystem from being remounted.
3484 Details : The client was sending extra request flags in the root getattr
3485 request and did not allocate a reply buffer for the dir EA.
3488 Frequency : occasional, higher if lots of files are accessed by one client
3489 Bugzilla : 6159, 6097
3490 Description: Client trips assertion regarding lsm mismatch/magic
3491 Details : While revalidating inodes the VFS looks up inodes with ifind()
3492 and in rare cases can find an inode that is being freed.
3493 The ll_test_inode() code will free the lsm during ifind()
3494 when it finds an existing inode and then the VFS later attaches
3495 this free lsm to a new inode.
3499 Bugzilla : 6422, 7030
3500 Description: MDS deadlock between mkdir and client eviction
3501 Details : Creating a new file via mkdir or mknod (starting a transaction
3502 and getting the ns lock) can deadlock with client eviction
3503 (gets ns lock and trying to finish a synchronous transaction).
3506 Frequency : occasional
3507 Description: While starting a server, the fsfilt_ext3 module could not be
3509 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3510 kernels. Previously, lconf would still use the ext3 name
3511 when trying to load modules. Now, it will correctly use
3512 ext3 on 2.4 and ldiskfs on 2.6.
3514 Severity : enhancement
3515 Description: The default stripe count has been changed to 1
3516 Details : The interpretation of the default stripe count (0, to lfs
3517 or lmc) has been changed to mean striping across a single
3518 OST, rather than all available. For general usage we have
3519 found a stripe count of 1 or 2 works best.
3521 Severity : enhancement
3522 Description: Add support for compiling against Cray portals.
3523 Details : Conditional compiling for some areas that are different
3527 Frequency : occasional
3528 Bugzilla : 6409, 6834
3529 Description: Creating files with an explicit stripe count may lead to
3530 a failed assertion on the MDS
3531 Details : If some OSTs are full or unavailable, creating files may
3532 trigger a failed assertion on the MDS. Now, Lustre will
3533 try to use other servers or return an error to the
3537 Frequency : occasional
3539 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3540 caused lock timeout during readahead (since 1.4.2).
3541 Details : Processes doing readahead might match a lock that hasn't been
3542 granted yet if there are overlapping and conflicting lock
3543 requests. The readahead process waits on ungranted lock
3544 (original lock is CBPENDING), while OST waits for that process
3545 to cancel CBPENDING read lock and eventually evicts client.
3547 Severity : enhancement
3549 Description: Initial enabling of flock support for clients
3550 Details : Implements fcntl advisory locking and file status functions.
3551 This feature is provided as an optional mount flag (default
3552 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3553 locking are implemented yet, and those that are are not guaranteed
3554 to be completely correct in production environments.
3555 mount -t lustre -o [flock|noflock] ...
3558 Frequency : occasional
3560 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3561 assertion in the ext3 JBD (journaling) layer.
3562 Details : The b_committed_data struct is protected by the big kernel lock
3563 in 2.4 kernels, serializing journal_commit_transaction() and
3564 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3565 access to this struct. In 2.6 kernels there is finer grained
3566 locking to improve SMP performance of the JBD layer.
3570 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3575 Description: merge and cleanup kernel patches.
3578 -----------------------------------------------------------------------------
3580 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3582 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3583 versions of Lustre. Please update all servers and clients to
3584 version 1.4.2 or later at the same time. You must also run
3585 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3586 to update the configuration logs.
3588 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3589 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3590 - swab llog records of type '0' so we get proper header size/idx (5861)
3591 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3592 - fix rename of one directory over another leaking an inode (5953)
3593 - avoid SetPageDirty on 2.6 (5981)
3594 - don't re-add just-being-destroyed locks to the waiting list (5653)
3595 - when creating new directories, inherit the parent's custom
3596 striping settings if present parent (3048)
3597 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3598 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3599 - don't hold client locks on temporary worklist from l_lru (5666)
3600 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3601 - automatically evict dead clients (5921)
3602 - Update file size properly in create+truncate+fstat case (6196)
3603 - Do not unhash mountpoint dentries, do not allow removal of
3605 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3606 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3607 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3608 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3609 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3610 - Correctly handle reads of files with no objects (6243)
3611 - lctl recover will also mark a device active if deactivate used (5933)
3613 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3614 - allow --write-conf on an MDS with different nettype than client (5619)
3615 - don't write config llogs to MDS for mounts not from that MDS (5617)
3616 - lconf should create multiple TCP connections from a client (5201)
3617 - init scripts are now turned off by default; run chkconfig --on
3618 lustre and chkconfig --on lustrefs to use them
3619 - upcalls are no longer needed for clients to recover to failover
3621 - add --abort-recovery option to lconf to abort recovery on device
3623 - add support for an arbitrary number of OSTs (3026)
3624 - Quota support protocol changes.
3625 - forward compatibility changes to wire structs (6007)
3626 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3627 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3628 - improved "lustre status" script
3629 - initialize blocksize for non-regular files (6062)
3630 - added --disable-server and --disable-client configure options (5782)
3631 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3632 - Vanilla 2.4.29 support
3633 - increase maximum number of obd devices to 520 (6242)
3634 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3635 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3636 SLES 9 2.6.5 kernels
3637 - limit stripes per file to 160 (the maximum EA size) (6093)
3639 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3642 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3643 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3644 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3645 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3646 - small ext3 extents cleanups and fixes (5733)
3647 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3648 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3649 - avoid needless client->OST connect, fix handle mismatch (5317)
3650 - fix DLM error path that led to out-of-sync client, long delays (5779)
3651 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3652 - fix several locking issues related to i_size (5492,5624,5654,5672)
3653 - don't move pending lock onto export if it is already evicted (5683)
3654 - fix kernel oops when creating .foo in unlinked directory (5548)
3655 - fix deadlock in obdfilter statistics vs. object create (5811)
3656 - use time_{before,after} to avoid timer jiffies wrap (5882)
3657 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3658 - Do not leak request if server was not able to process it (5154)
3659 - If mds_open unable to find parent dir, make that negative lookup(5154)
3660 - don't create new directories with extent-mapping (5909, 5936)
3662 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3663 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3664 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3665 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3666 - support for RHEL4 kernel on i686 (5773)
3667 - provide error messages when incompatible logs are encountered (5898)
3669 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3670 * version 1.4.0.10 (1.4.1 release candidate 1)
3672 - don't keep a lock reference when lock is not granted (4238)
3673 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3674 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3675 - drop import inflight refcount on signal_completed_replay error (5255)
3676 - unlock page after async write error during send (3677)
3677 - handle missing objects in filter_preprw_read properly (5265)
3678 - no transno return for symlink open, don't save no-trasno open (3440)
3679 - don't try to complete elan receive that already failed (4012)
3680 - free RPC server reply state on error (5406)
3681 - clean up thread from ptlrpc_start_thread() on error (5160)
3682 - readahead could read extra page into cache that wasn't ejected (5388)
3683 - prevent races in class_attach/setup/cleanup/detach (5260)
3684 - don't dereference de->d_inode after l_dput of de (5458)
3685 - use "int" for stripe value returned from lock_to_stripe (5544)
3686 - mballoc allocation and error-checking fixes in 2.6 (5504)
3687 - block device patches to fix I/O request sizes in 2.6 (5482)
3688 - look up hostnames for IB nals (5602)
3689 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3690 - don't start multiple acceptors for the same port (5277)
3691 - fix incorrect LASSERT in mds_getattr_name (5635)
3692 - export a proc file for general "ping" checking (5628)
3693 - fix "lfs check" to not block when the MDS is down (5628)
3695 - service request history (4965)
3696 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3697 - create an "evict_client" /proc entry on OSTs, like the MDS has
3698 - fix mount usage message, return errors per mount(8) (5168)
3699 - change grep [] to grep "[]" in tests so they work in more UMLs
3700 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3701 - remove ancient LOV_MAGIC_V0 EA support (5047)
3702 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3703 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3704 - allow readahead allocations to fail when low on memory (5383)
3705 - mmap locking landed again, after considerable improvement (2828)
3706 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3708 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3711 - send OST transaction number in read/write reply to free req (4966)
3712 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3713 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3714 - the watchdog thread now runs as interruptible (5246)
3715 - flock/lockf fixes (but it's still disabled, pending 5135)
3716 - don't use EXT3 constants in llite code (5094)
3717 - memory shortage at startup could cause assertion (5176)
3719 - reorganization of lov code
3720 - single portals codebase
3722 - add extents/mballoc support (5025)
3723 - direct I/O reads in the obdfilter (4048)
3724 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3726 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3729 - send OST transaction number in read/write reply to free req (4966)
3730 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3731 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3732 - the watchdog thread now runs as interruptible (5246)
3733 - handle missing objects in filter_preprw_read properly (5265)
3734 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3735 - drop import inflight refcount on signal_completed_replay error (5255)
3736 - unlock page after async write error during send (3677)
3737 - return original error code on reconstructed replies (3761)
3738 - no transno return for symlink open, don't save no-trasno open (3440)
3740 - add pid to ldlm debugging output (4922)
3741 - bump the watchdog timeouts -- we can't handle 30sec yet
3742 - extra debugging for orphan dentry/inode bug (5259)
3744 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3747 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3748 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3749 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3750 - fix NULL dereference in /proc/sys/portals/routes (4827)
3751 - allow failed mdc_close() operations to be interrupted (4561)
3752 - stop precreate on OST before MDS would time out on it (4778)
3753 - don't send partial-page writes before EOF from client (4410)
3754 - discard client grant for sub-page writes on large-page clients (4520)
3755 - don't free dentries not owned by NFS code, check generation (4806)
3756 - fix lsm leak if mds_create_objects() fails (4801)
3757 - limit debug_daemon file size, always print CERROR messages (4789)
3758 - use transno after validating reply (3892)
3759 - process timed out requests if import state changes (3754)
3760 - update mtime on OST during writes, return in glimpse (4829)
3761 - add mkfsoptions to LDAP (4679)
3762 - use ->max_readahead method instead of zapping global ra (5039)
3763 - don't interrupt __l_wait_event() during strace
3765 - add software watchdogs to catch hung threads quickly (4941)
3766 - make lustrefs init script start after nfs is mounted
3767 - fix CWARN/ERROR duplication (4930)
3768 - return async write errors to application if possible (2248)
3769 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3770 - print NAL number in %x format (4645)
3771 - update barely-supported suse-2.4.21-171 series (4842)
3772 - support for sles 9 %post scripts
3773 - support for building 2.6 kernel-source packages
3774 - support for sles km_* packages
3776 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3779 - ignore -ENOENT errors in osc_destroy (3639)
3780 - notify osc create thread that OSC is being cleaned up (4600)
3781 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3782 - reconstruct ost_handle() like mds_handle() (4657)
3783 - create a new thread to do import eviction to avoid deadlock (3969)
3784 - let lconf resolve symlinked-to devices (4629)
3785 - don't unlink "objects" from directory with default EA (4554)
3786 - hold socknal file ref over connect in case target is down (4394)
3787 - allow more than 32000 subdirectories in a single directory (3244)
3788 - fix blocks count for O_DIRECT writes (3751)
3789 - OST returns ENOSPC from object create when no space left (4539)
3790 - don't send truncate RPC if file size isn't changing (4410)
3791 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3792 - bind to privileged port in socknal and tcpnal (3689)
3794 - rate limit CERROR/CWARN console message to avoid overload (4519)
3795 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3796 - basic mmap support (3918)
3797 - kernel patch series update from b1_4 (4711)
3799 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3802 - avoid crash during MDS cleanup with OST shut down (2775)
3803 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3804 - don't use bad inodes on the MDS (3744)
3805 - dynamic object preallocation to improve recovery speed (4236)
3806 - don't hold spinlock over lock dumping or change debug flags (4401)
3807 - don't zero obd_dev when it is force cleaned (3651)
3808 - print grants to console if they go negative (4431)
3809 - "lctl deactivate" will stop automatic recovery attempts (3406)
3810 - look for existing locks in ldlm_handle_enqueue() (3764)
3811 - don't resolve lock handle twice in recovery avoiding race (4401)
3812 - revalidate should check working dir is a directory (4134)
3814 - don't always mark "slow" obdfilter messages as errors (4418)
3816 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3819 - don't close LustreDB during write_conf until it is done (3860)
3820 - fix typo in lconf for_each_profile (3821)
3821 - allow dumping logs from multiple threads at one time (3820)
3822 - don't allow multiple threads in OSC recovery (3812)
3823 - fix debug_size parameters (3864)
3824 - fix mds_postrecov to initialize import for llog ctxt (3121)
3825 - replace config semaphore with spinlock (3306)
3826 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3827 - don't allow enqueue to complete on a destroyed export (3822)
3828 - down write_lock before checking llog header bitmap (3825)
3829 - recover from lock replay timeout (3764)
3830 - up llog sem before sending rpc (3652)
3831 - reduce ns lock hold times when setting kms (3267)
3832 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3833 - fix NULL deref and obd_dev leak on setup error (3312)
3834 - replace some LBUG about llog ops with error handling (3841)
3835 - don't match INVALID dentries from d_lookup and spin (3784)
3836 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3837 - fix invalid assertion in ptlrpc_set_wait (3880)
3839 - add libwrap support for the TCP acceptor (3996)
3840 - add /proc/sys/portals/routes for non-root route listing (3994)
3841 - allow setting MDS UUID in .xml (2580)
3842 - print the stack of a process that LBUGs (4228)
3844 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3847 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3848 - make sure to unset replay flag from failed open requests (3440)
3849 - if default stripe count is 0, use OST count for inode size (3636)
3850 - update parent mtime/ctime on client for create/unlink (2611)
3851 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3852 - free recovery state on server during a forced cleanup (3571)
3853 - unregister_reply for resent reqs (3063)
3854 - loop back devices mounting and status check on 2.6 (3563)
3855 - fix resource-creation race that can provoke i_size == 0 (3513)
3856 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3857 - more debugging for page-accounting assertion (3746)
3858 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3859 - avoid deadlock after precreation failure (3758)
3860 - fix race and lock order deadlock in orphan handling (3450, 3750)
3861 - add validity checks when grabbing inodes from l_ast_data (3599)
3863 - add /proc/.../recovery_status to obdfilter (3428)
3864 - lightweight CDEBUG infrastructure, debug daemon (3668)
3865 - change default OSC RPC parameters to be better on small clusters
3866 - turn off OST read cache for files smaller than 32MB
3867 - install man pages and include them in rpms (3100)
3868 - add new init script for (un)mounting lustre filesystems (2593)
3869 - run chkconfig in %post for init scripts (3701)
3870 - drop scimac NAL (unmaintained)
3872 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3875 - clean kiobufs before and after use (3485)
3876 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3877 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3878 - change init script to fail more clearly if not run as root (1528)
3879 - allow clients to reconnect during replay (1742)
3880 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3881 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3882 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3883 - fix cond_resched (3554)
3884 - only evict unfinished clients after recovery (3515)
3885 - allow bulk resend, prevent data loss (3570)
3886 - dynamic ptlrpc request buffer allocation (2102)
3887 - don't allow unlinking open directory if it isn't empty (2904)
3888 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3889 - remove extraneous obd dereference causing LASSERT failure (3334)
3890 - don't use get_cycles() when creating temp. files on the mds (3156)
3891 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3892 - handle EEXIST for set-stripe, set proper directory name (3336)
3894 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3895 - fix ksocknal_fmb_callback() error messages (2918)
3897 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3900 - don't copy lvb into (possibly NULL) reply on error (2983)
3901 - don't deref dentry after dput, don't free lvb on error (2922)
3902 - use the kms to determine writeback rpc length (2947)
3903 - increment oti_logcookies when osc is inactive (2948)
3904 - update client's i_blocks count via lvb messages (2543)
3905 - handle intent open/close of special files properly (1557)
3906 - mount MDS with errors=remount-ro, like obdfilter (2009)
3907 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3908 - don't use cancelling-locks' kms values (2947)
3909 - use highest lock extent for kms, not last one (2925)
3910 - don't dereference ERR_PTR() dentry in error handling path (3107)
3911 - fix thread race in portals_debug_dumplog() (3122)
3912 - create lprocfs device entries at setup instead of at attach (1519)
3913 - common AST error handler, don't evict client on completion race (3145)
3914 - zero nameidata in detach_mnt in 2.6 (3118)
3915 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3916 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3917 - initialize RPC timeout timer earlier for 2.6 (3219)
3918 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3919 - print nal/nid for unknown nid (3258)
3920 - additional checks for oscc recovery before doing precreate (3284)
3921 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3922 - don't crash in mdc_close for bad permissions on open (3285)
3923 - zero i_rdev for non-device files (3147)
3924 - clear page->private before handing to FS, better assertion (3119)
3925 - tune the read pipeline (3236)
3926 - fix incorrect decref of invalidated dentry (2350)
3927 - provide read-ahead stats and refine rpc in flight stats (3328)
3928 - don't hold journal transaction open across create RPC (3313)
3929 - update atime on MDS at close time (3265)
3930 - close LDAP connection when recovering to avoid server load (3315)
3931 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3932 - don't leak open file on MDS after open resend (3325)
3933 - serialize filter_precreate and filter_destroy_precreated (3329)
3934 - loop device shouldn't call sync_dev() for nul device (3092)
3935 - clear page cache after eviction (2766)
3936 - resynchronize MDS->OST in background (2824)
3937 - refuse to mount the same filesystem twice on same mountpoint (3394)
3938 - allow llmount to create routes for mounting behind routers (3320)
3939 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3940 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3941 - fix rare problem with rename on htree directories (3417)
3943 - allow default OST striping configuration per directory (1414)
3944 - fix compilation for qswnal for 2.6 kernels (3125)
3945 - increase maximum number of MDS request buffers for large systems
3946 - change liblustreapi to be useful for external progs like lfsck (3098)
3947 - increase local configuration timeout for slow disks (3353)
3948 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3950 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3953 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3954 - don't overwrite extent policy data in reply if lock was blocked (2901)
3955 - drop filter export grants atomically with removal from device (2663)
3956 - del obd_self_export from work_list in class_disconnect_exports (2908)
3957 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3958 - swab reply message in mdc_close, other PPC fixes (2464)
3959 - fix destroying of named logs (2325)
3960 - overwrite old logs when running lconf --write_conf (2264)
3961 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3962 - fix race in target_handle_connect (2898)
3963 - mds_reint_create() should take same inode create lock (2926)
3964 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3965 - don't close files for self_export to avoid uninitialized obd (2936)
3966 - allow MDS with the same name as client node (2939)
3967 - hold dentry reference for closed log files for unlink (2325)
3968 - reserve space for all logs during transactions (2059)
3969 - don't evict page beyond end of stripe extent (2925)
3970 - don't oops on a deleted current working directory (2399)
3971 - handle hard links to targets without a parent properly (2517)
3972 - don't dereference NULL lock when racing during eviction (2867)
3973 - don't grow lock extents when lots of conflicting locks (2919)
3975 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3978 - account for cache space usage on clients to avoid data loss (974)
3979 - lfsck support in lustre kernel code (2349)
3980 - reduce journal credits needed for BRW writes (2370)
3981 - orphan handling to avoid losing space on client/server crashes
3982 - ptlrpcd can be blocked, stopping ALL progress (2477)
3983 - use lock value blocks to assist in proper KMS, faster stat (1021)
3984 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3985 - recovery for initial connections (2355)
3986 - fixes for mds_cleanup_orphans (1934)
3987 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3988 - block all file creations until orphan recovery completes (1901)
3989 - client remove rq_connection from request struct (2423)
3990 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3991 - recovery timer race (2670)
3992 - mdc_close recovey bug (2532)
3993 - ptlrpc cleanup bug (2710)
3994 - mds timeout on local locks (2588)
3995 - namespace lock held during RPCs (2431)
3996 - handle interrupted sync write properly (2503)
3997 - don't try to handle a message that hasn't been replied to (2699)
3998 - client assert failure during cleanup after abort recovery (2701)
3999 - leak mdc device after failed mount (2712)
4000 - ptlrpc_check_set allows timedout requests to complete (2714)
4001 - wait for inflight reqs when ptlrpcd finishes (2710)
4002 - make sure unregistered services are removed from the srv_list
4003 - reset bulk XID's when resending them (caught by 1138 test)
4004 - unregister_bulk after timeout
4005 - fix lconf error (2694)
4006 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4007 - readahead locks pages, leaves pending causing memory pressure (2673)
4008 - increase OST request buffers to 4096 on large machines (2729)
4009 - fix up permission of existing directories in simple_mkdir (2661)
4010 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4011 - don't assign transno to errored transactions (2742)
4012 - don't delete objects on OST if given a bogus objid from MDS (2751)
4013 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4014 - if rq_no_resend, then timeout request after recovery (2432)
4015 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4016 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4017 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4018 - MDS recovery shouldn't delete objects that it has given out (2730)
4019 - if enqueue arrives after completion, don't clobber LVB (2819)
4020 - don't unlock pages twice when trigger_group_io returns error (2814)
4021 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4022 - don't write pages to disk if there was an error (1450)
4023 - don't ping imports that have recovery disabled (2676)
4024 - take buffered bytes into account when balancing socknal conn (2817)
4025 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4026 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4027 - remove little-endian swabbing of llog records (1987)
4028 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4029 - retry reposting request buffers if they fail (1191)
4030 - grow extent at grant time to avoid granting a revoked lock (2809)
4031 - lock revoke doesn't evict page if covered by a second lock (2765)
4032 - disable VM readahead to avoid reading outside lock extents (2805)
4034 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4035 - updated kernel patches for hp-2.4.20 kernel (2681)
4037 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4040 - fix truncated write corruption (2366)
4041 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4043 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4044 - ensure len(uuid) < 37 in lmc (1171)
4045 - fix ia64 OOPS in llog_test (2255)
4046 - zero end of page at obdfilter for partial page writes (2648)
4047 - don't leave stale dentries around after renames (bug 2428)
4048 - fix timeouts when evicting a client with a single lock held (2642)
4049 - set deadline for the initial HELLO message to drain (2634)
4050 - print out dotted-quad IP addresses in the socknal (2302)
4052 - additional debugging for MDS client eviction problem (2443)
4053 - fix mkfsoptions support for osts (2603, 2604)
4055 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4058 - add series for the vanilla 2.6.0 kernel
4059 - add series for the vanilla 2.4.24 kernel
4060 - add series for a cray x86/64 UL kernel drop
4061 - fix xattr patches for the vanilla 2.4.19 series
4063 - generate true UUIDs in lmc (1171)
4064 - have portals stack dumping break in UML (2466)
4065 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4066 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4067 - wake callback waiting threads on client eviction (2460)
4068 - Add --ptldebug and --subsystem to lmc (1719)
4069 - update assertion to allow safe interrupt allocation
4070 - set rq_no_resend for cancel requests (2432)
4071 - recalculate ptlrpcd timeout after resend (2494)
4072 - call vfs_rmdir when removing pending directories (2368)
4073 - fix renaming a file to itself (2429)
4074 - lmc creates a default one-stripe lov (2454)
4075 - expand procfs space to handle large clusters (2326)
4076 - increase UML stack to avoid overflow
4077 - update lconf's list of debug and subsystem masks
4078 - fix lfs find --obd (2510)
4079 - /proc tunable for disabling filter read caching (2591)
4080 - stop rpm packages from altering slapd.conf (2301)
4081 - disable nagle in the socknal under 0conf (2578)
4082 - choose mds inode size based on stripe count (2572)
4083 - fix kernel-source rpm problems (2516)
4085 - add --disable-doc to avoid pdf generation (2421)
4086 - update documentation, tests, type-os, comments
4087 - avoid format warnings on ia64
4088 - remove the TOE NAL
4089 - tiny code cleanups by removing unused fields
4091 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4094 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4095 - remove the most blatant lies from BUILDING file (2371)
4096 - change default debug level to reasonable production setting
4097 - reduce client side cache size to reduce cache flush time
4098 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4099 - make TCP zerocopy and pinger support enabled by default (2476)
4100 - sync writes completed after process exits caused crashes (2319)
4101 - maintain correct mount count on the MDS (2356)
4102 - backout 1557, because 2316 wasn't really fixed
4103 - better file I/O statistics gathering in /proc
4104 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4105 - another kernel patch to fix zero-copy TCP function export
4106 - don't take duplicate lock when processing re-sent getattr (2420)
4107 - lctl uses obd_self_export instead of creating new conn (2353)
4108 - MDS/OST recovery case which requires object creation asserted (2425)
4109 - move lfs from /usr/sbin to /usr/bin in packages
4110 - fix race between mds_client_add and mds_client_free (2417)
4111 - use kmalloc instead of slabs in portals (2430)
4112 - don't create duplicate records when a failover MDS is present (2442)
4113 - remove unnecessary mount age check (2332)
4114 - don't remove directory inodes from locks prematurely (2451)
4115 - don't break if MDS service name is the same as hostname (2103)
4116 - fix races in client write RPC generation when cache full (2482)
4118 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4121 - remove now-unused request->rq_obd (278)
4122 - if an allocation fails, print out how much memory we've used (1933)
4123 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4124 - add the "configurable stack size" patch to most series files (1256)
4125 - ability to write large log records, for 100+ OST configs (2306)
4126 - fix NULL deref when filter_prep fails (2314)
4127 - fix operator precedence error in filter_sync
4128 - dynamic allocation of socknal TX descriptors (2315)
4129 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4130 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4131 - ensure that CWARN messages in Portals always get to the syslog
4132 - __init/__exit are not for prototype decls (ldlm_init/exit)
4133 - x86-64 compile warning fixes
4134 - fix gateway LMC keyword conflict (2318)
4135 - fix MDS lock inversions in getattr/reint paths (1844)
4136 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4137 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4138 - DLM race condition prevented some lock evictions (2328)
4139 - ENOMEM detection and retry on socknal sends (2230)
4140 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4141 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4142 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4143 - better error messages when a client is rejected during recovery (1505)
4144 - avoid cancelling locks which were never granted, after failure (2330)
4145 - fix i_sem/journal inversion in mds_client_add (2333)
4146 - fix truncate/getattr lock cycle deadlock (2334)
4147 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4148 - fix two rare exit paths which could leak an l_lock() ref (2321)
4149 - fencepost error in MDS/OST orphan recovery (2226)
4150 - make log record alignment 8 bytes (1988)
4151 - lstripe now fails when requested offset > ost_count (2237)
4152 - ensure that all kernel series have a complete list.h (1607)
4153 - fix crashes in special-file operations (2316)
4154 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4156 - allow configurable automake binary, for testing new versions
4157 - small update to the lfs documentation
4159 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4161 * fix negative export reference count in fsfilt_sync (2312)
4163 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4164 * release candidate 0.9.1
4166 - orphans are moved into the PENDING directory for possible recovery
4167 - replayed opens now open by fid for orphan/rename safety (1042)
4168 - last close of an orphan inode generates a transno (683)
4169 - chdir() and mount() now pin the directory entry (1020)
4170 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4171 - discard very old requests without processing them (1502)
4172 - remove some common, well-understood CERRORs (1505)
4173 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4174 - clear "grant" flags in OST replies until OST grant code lands (1644)
4175 - fix read performance by not clobbering i_blksize on client (1598)
4176 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4177 - make LRU size a /proc tunable, clears locks when reduced (707)
4178 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4179 - server locks take references on exports now (1558)
4180 - build fixes for 2.4.20-rh trees (1663)
4181 - return an error from lov_create if all OSCs are inactive (1751)
4182 - fix import levels when a reconnect happens without a timeout (1597)
4183 - exit early from mds_open if we get a lookup error (1749)
4184 - partial page read at EOF wouldn't wait for disk before sending (1642)
4185 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4186 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4187 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4188 - write/truncate lock inversion (1639)
4189 - Don't auto-load obdclass, portals modules during cleanup (1495)
4190 - fix timestamps from jumping to "now" (1763)
4191 - extra journal assertions (1648)
4192 - add an extra multiunlink test (1771)
4193 - fix read_record/write_record API (1776)
4194 - fix leak of offset_extent, possible incorrect i_size later (1772)
4195 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4196 - Debugging for the kqswnal_get_idle_tx problems (1820)
4197 - Allow recovery to be attempted multiple times (1536)
4198 - Write out MDS last_rcvd file after it is first created (1600)
4199 - Fix tx_descriptor leak in failed transmit situations (1827)
4200 - ext3 journaling fixes for assertion failure after IO error (1871)
4201 - class_export_put() on freed export after completion AST error (1896)
4202 - Fix revalidate looping in VFS (1322)
4203 - Don't access a freed export during MDS_REINT timeout (1521)
4204 - Add open-unlink recovery support on the MDS (1673,1764)
4205 - Return an error if no MDS data was read from last_rcvd (1946)
4206 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4207 - Don't setup a disk device that doesn't match exported UUID (317)
4208 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4209 - avoid committing NULL handle in force close
4210 - local.sh is now a one-stripe LOV configuration
4211 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4212 - fix ext3 htree duplicate directory entry corruption (1516)
4213 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4214 - update to new LOV EA format (2097)
4215 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4217 2003-06-15 Phil Schwan <phil@clusterfs.com>
4220 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4221 - per-import/export recovery handling (958, 931, 959)
4222 - multiple last-rcvd slots, for serving multiple FSes (949)
4223 - connections are again shared between multiple imp/exports (963, 964)
4224 - "umount -f" would hang if any requests needed to be sent (393, 978)
4225 - avoid pinning large req buffer by copying for queued messages (989)
4226 - add "uuid" to "lctl device" command to help upcalls (991)
4227 - "open" RPCs with transnos would confuse recovery counters (1037)
4228 - do proper endian conversion of all wire messages (288, 340, 891)
4229 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4230 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4231 - fix ptlrpc_connection leak in target_handle_connect (1174)
4232 - fix import refcounting bug in OST and MDS cleanup (1134)
4233 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4234 - fix very unlikely obd_types race condition (501)
4235 - remove osc_open hack for echo_client (1187)
4236 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4237 - a failure in read_inode2 leads to deadlock (1139)
4238 - cancel ack-locks as soon as transaction is committed (1072)
4239 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4240 - make sure to commitrw after any preprw to avoid deadlock (1162)
4241 - failing to execute a file in a lustre FS would lock inode (1203)
4242 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4243 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4244 - obd_uuid2tgt would walk off the end of the list (1255)
4245 - on IA64 the osc would give portals incorrect bulk size (1258)
4246 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4247 - fix lock inversion caused by new llite matching code (1282)
4248 - limit the number of dirty pages on a client to 10MB (1286)
4249 - timed out locks were not being corrected cancelled (1289)
4250 - fix O_DIRECT above 4GB on IA-32 (1292)
4251 * major user-visible changes
4252 - fail out/fail over policy now controlled by the upcall (993)
4254 - add OBD_PING to check server availability and failure (954)
4255 - lustre messages are now sent in sending host order (288, 340, 891)
4256 - add eadatalen to MDS getattr reply (340)
4257 - OST read replies may contain second buffer, with per-page status (593)
4259 2003-03-11 Phil Schwan <phil@clusterfs.com>
4262 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4263 - failed open()s could cause deadlock; fixed (867, 869)
4264 - stop cancelling OST locks when files are closed (481)
4265 - overlapping XID spaces caused network corruption (851, 853)
4266 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4267 - setattr_raw added, to do single-RPC, server-side setattrs
4268 - lmc/lconf syntax change for OST UUIDs
4269 - fix crashy race condition between ptlrpc_free_req and osc_close
4270 - don't use request in mdc_enqueue if we hit a timeout (889)
4271 - don't set the inode i_size for regular files from the MDS (896)
4272 - handle out of order completion AST (842)
4273 - don't LBUG if a lock request times out after receiving AST (913)
4274 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4275 - if a bad lock AST arrives, send an error instead of dropping entirely
4276 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4277 - fix leak in bulk IO when only partially completed (899, 900, 926)
4278 - fix O_DIRECT for ia64 (55)
4279 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4280 - C-z after timeout could hang a process forever; fixed (977)
4282 - client-side I/O cache (678, 924, 929, 941, 970)
4284 - READPAGE and SETATTRs which don't take server-side locks get
4287 2003-02-11 Phil Schwan <phil@clusterfs.com>
4290 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4291 - Fix fsfilt_extN_readpage() to read a full page of directory
4292 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4293 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4294 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4295 - Most intent operations were removed, in favour of a new RPC mode
4296 that does a single RPC to the server and bypasses most of the VFS
4297 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4298 - Aggressively cancel local locks on DLM servers
4299 - mds_reint_unlink sends EA to the client if it's the last nlink.
4300 client uses that EA to unlink OST objects.
4301 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4302 - recursive symlinks were fixed (439)
4303 - fixed NULL deref in DEBUG_REQ
4304 - filter_update_lastobjid no longer calls sync, which annoyed extN
4305 - fixed multi-client small-writes to a single file problem (445)
4306 - fixed mtime updates during file writes (607)
4307 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4308 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4309 - fixed Linux OST size reporting problem (444, 656)
4310 - OST now updates object mtime with writes or setattr (607, 619)
4311 - client verifies file size before zeroing page past EOF (445)
4312 - OST now writes last allocated objid to disk with allocation (108)
4313 - LOV on echo now works (409)
4315 - mds_reint_unlink sends a new buffer, with the EA included. this
4316 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4317 set if a regular file was being unlinked, and it was the last link
4318 - use PtlGet from the target for bulk writes (315)
4319 - OST now updates object mtime with writes or setattr (607, 619)
4320 - LDLM now has a grant-time callback to revalidate locked items, if
4322 - Many MDS operations were reorganized to combat race conditions
4324 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4325 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4327 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4330 - Fully reactivate OST imports after reconnection (512, others)
4331 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4332 - More graceful error handling for truncating on dead OST (515)
4333 - Don't error out unless we're actually accessing dead stripes (474)
4334 - Fix garbage sizes when stripes are missing (410)
4335 - LRU counters were broken, causing constant lock purge (433, 432)
4336 - garbage on read from stripes with failed OSTs (441)
4337 - mark OSCs as active before reconnecting during recovery (438)
4338 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4339 - lfind did not preserve OST order in output (443)
4340 - symlinks cause hung clients, incorrect data (439)
4341 - stop dereferencing request after dropping refcount (457)
4342 - don't LASSERT(spin_is_locked) on non-SMP (455)
4343 - fixes for many rename() bugs
4344 - fstat didn't correctly synchronize attributes (399)
4345 - server must handle lock cancellation during blocking AST prep (487)
4346 - bulk descriptors were free()d too soon (511)
4347 - fix paths in lconf, which would load incorrect modules (451, 507)
4348 - fix confusing lconf 'host not found' error message (386)
4349 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4350 - fix race condition in mdc_blocking_ast() for inode access (526)
4351 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4352 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4353 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4354 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4355 - fix lstripe in several ways (kernel side) (527)
4356 - fix request leak in ldlm_cli_enqueue (262)
4357 - incorrect OSC was marked inactive after OST failure
4358 - call mds_fs_cleanup before unmounting filesystem (524)
4359 - fix races between taking ns_lock and ldlm_lock_change_resource
4360 - fix races updating LOV export open file list
4361 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4362 - fix recovery NULL deref in ldlm_cli_cancel_unused
4363 - fix some DLM races by using new hash table for lock handles (419)
4364 - permit the client to specify desired inodes, at replay
4365 - duplicate requests when we queue them for replay reintegration
4366 - fix last_rcvd offset calculation
4367 - sync after each recovered transaction, so we always make progress
4368 - never, not always, ERESTART requests without transnos
4369 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4371 - skip replay if the MDS says that the client is already connected
4372 - don't check for a recovery-enabled export to match lctl's UUID
4373 - don't INC_USE_COUNT for phantom exports
4374 - don't crash when cleaning up phantom exports (567)
4375 - don't double-finish or set replay data for errored mdc_open requests
4376 - abort requests when they time out, so we don't get old replies
4377 - send/receive replies for AST messages again
4378 - if the client says that it doesn't have the lock, cancel it on the
4380 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4381 mark it as destroyed, it will all work out when decref is called
4382 - fix module use counts (22, 581)
4384 - ASTs now expect a reply (server cancels lock on error reply)
4386 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4389 - fix many simultaneous client startup (392)
4390 - fix dentry->d_it clobbering
4391 - credentials weren't being shipped for readdir/getattr operations
4392 - remove invalid assertions triggered during some concurrent MD
4394 - proper Lustre versions added (336, 389)
4395 - fix memory leak for create error case (398)
4396 - fix LOV locking bug that would get cli/srv out of sync
4397 - fix echo client over LOV (409)
4398 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4399 - fix double-O_EXCL intent crash (424)
4400 - avoid sending multiple lock CANCELs (352)
4402 - MDS can do multi-client recovery (modulo bugs in new code)
4404 - many updates, edits, cleanups
4406 2002-11-18 Phil Schwan <phil@clusterfs.com>
4409 - fix null d_it dereference (346)
4410 - fix full OST/dbench hang (333)
4411 - fix permission problem with file removal (286)
4412 - fix removal of OSCs from LOV when they fail
4413 - fix NULL deref during bulk timeout (214)
4414 - fix problems related to multiple filesystems on one MDS (241)
4415 - fixed serious subtle metadata locking bugs
4416 - free locks on clients when inodes are removed due to memory
4418 - fix inode pointer in lock data (285)
4419 - partial support for multiple MDS on a single host (241)
4420 - data locks weren't cancelled at clear_inode time (290, 311)
4421 - intent locks could lead to unbounded lock growth (205)
4422 - added a maximum lock count, an LRU list, and a flusher
4423 - fix multiple rename (365)
4424 - properly abstracted the echo client
4425 - OSC locked 1 byte too many; fixed
4426 - rewrote brw callback code:
4427 - fixed recovery bugs related to LOVs (306)
4428 - fixed too-many-pages-in-one-write crash (191)
4429 - fixed (again) crash in sync_io_timeout (214)
4430 - probably fixed callback-related race (385)
4432 - Add capability to MDS protocol
4433 - LDLM cancellations and callbacks on different portals
4435 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4438 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4439 - defer open object destruction to close time (601981, 138)
4440 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4441 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4442 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4443 - don't call obd_setattr in ll_file_release for destroyed objects
4444 * protocol change to lustre_msg: move |version| and add |flags|
4445 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4446 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4447 * added replay of create, unlink, link and rename operations during
4448 MDS failover; recovery should be much more robust now
4449 * remove failed OSCs from LOVs (only lov_create uses this so far)
4450 * the lustre-HOWTO was brought (more) up to date (582544)
4452 2002-10-23 Phil Schwan <phil@clusterfs.com>
4455 - in-use dentries weren't being reused properly (617851)
4456 - prevent multiple LDLM setup (599178)
4457 - fix LOV size calculations for truncate (617853)
4458 - fix client handling of MDS intent errors (POSIX)
4459 - fix permission bug in lovstripe.c test (624321)
4460 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4461 - truncate past end of file could corrupt data
4462 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4463 - a race in recovery could return ETIMEDOUT to apps (623947)
4464 - building outside the source directory was fixed
4465 * the lustre-HOWTO was brought (more) up to date (582544)
4466 * major progress was made on recovery functionality
4468 2002-10-10 Phil Schwan <phil@clusterfs.com>
4471 - recovery deadlock fix
4472 - rm -rf causes LBUG fix (617817)
4473 - file open by multiple tasks fix (618962)
4474 - directory permissions bugs (602707 and 620007)
4475 - journal_stop fixed with locking (611313)
4476 - O_APPEND failures resolved (618273, perhaps 614459)
4477 - lconf PATH fix (619770)
4478 - IA64 build fix (621450)
4479 - RPC buffer sizes scale with amount of memory
4481 2002-10-01 Phil Schwan <phil@clusterfs.com>
4484 - locks would be cancelled without throwing away data pages,
4485 resulting in inconsistent data (605627)
4486 - inode attributes were not always being refreshed (605627, 612449)
4487 - lconf now continues to cleanup after lctl reports an error
4488 - MDS now enforces user permissions (602707)
4489 - lprocfs cleanup fixed, but not yet enabled (614157)
4490 - fixed infinite server hang, should a client not respond to an AST
4491 - avoid going into recovery if user calls readlink() with a buffer
4492 that's too small (613941)
4493 - AST RPCs no longer require replies (614867) -- this may be changed
4494 - don't crash server if client sends an IOV that's too big (611336)
4495 - fixed lock conversion deadlock (611892)
4496 - fixed the following of symlinks (614622)
4497 * recovery: the server can remove locks from a client that dies, other
4498 clients can make progress
4499 * more extN patch fixes
4500 * compile-time configurable ptlrpc buffer allocations
4502 - collaborative read cache document
4503 - Lustre Lite Performance CDR document-in-progress
4505 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4508 - fix typo in patch-2.4.18
4510 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4513 - clear ptlrpc request each time in handle_incoming_request()
4514 - unlink of files now destroys the object on the OST
4516 2002-09-19 Peter Braam <braam@clusterfs.com>
4518 * add hard link support
4519 * change obdfile creation method
4520 * kernel patch changed
4522 2002-09-19 Peter Braam <braam@clusterfs.com>
4525 - stack overflow bug in extN fixed
4527 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4529 * documentation updates
4530 - add man pages for config tools
4531 - update tests/README to describe testing with new config tools
4532 - finish metadata API descriptions
4533 * bug fixes and cleanups
4534 - statfs workaround for 16TB limit
4535 - LOV stripe allocation improved, can stripe on subset of OSTs
4536 - LOV file size/IO offset was wrong for files > 4GB in size
4537 - object EA data was being dropped, caused files to be unreadable
4538 - memory overflow with non-LOV OST caused memory corruption
4539 - fixed regression tests to work with new config tools, obdfilter
4540 - fixed bug when directory size became larger than 1 block
4541 - fixed bug (for single client case) when PWD was deleted
4542 - invalidate local directory pages when doing intent-based ops
4543 - avoid LDLM oops when lock callback contained bad data
4545 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4547 * documentation updates
4548 * bug fixes and cleanups
4549 - configuration tools
4552 - 64-bit compile warnings
4553 - 64-bit internal statfs data
4555 * test_brw on persistent OST devices
4557 * lprocfs (disabled)
4559 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4561 * documentation updates
4562 * bug fixes and cleanups
4563 * configuration tools
4565 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4568 * many small fixes to 0_5_4
4569 * io/network handling
4570 * thinkos in MDS operations
4572 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4575 * crucial basic fixes to 0.5.3
4576 * IOR, Iozone work over Elan
4579 2002-08-07 Phil Schwan <phil@clusterfs.com>
4580 * version 0_5_3, our first alpha
4581 * we use the new Portals iovs
4582 * documentation updates
4583 * bug fixes and cleanups
4584 * small changes in the DLM wire protocol
4586 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4587 * version 0_5_1 with some initial stability,
4588 * locking on MD and file I/O.
4589 * documentation updates
4590 * several bug fixes since 0.5.0
4591 * small changes in wire protocol
4593 2002-07-18 Phil Schwan <phil@clusterfs.com>
4595 * delivered as Lustre Light Alpha
4596 * fixed a crash after handling invalid MDS requests
4597 * fixed directory pages for architectures with non-4k pages sizes
4599 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4600 * release version v0_4_4
4601 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4603 * Fixes a number of interruption problems with OST operations.
4604 * Update documentation for portals header changes
4605 * Move all wire protocol structs/defines to lustre_idl.h
4606 * Fixes symlink length bug.
4607 * Add tcpdump to repository.
4609 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4610 * release version v0_4_3
4611 * Fixes statfs for inodes on extN.
4612 * Fixes bug in runtests which would delete /etc/hosts.
4613 * Use 64-bit object IDs wherever possible (not into VFS though)
4614 Remove ost_get_info, which is unused by lustre, and out of date.
4616 2002-07-03 Peter Braam <braam@clusterfs.com>
4617 * release version v0_4_2 Fixes a lookup error (type not passed)
4618 * move forward to head of Portals
4619 * move forward to latest Lustre kernel
4621 2002-06-25 Peter Braam <braam@clusterfs.com>
4622 * release version v0_4_1. Hopefully stable on single node use.