1 tbd Sun Microsystems, Inc.
4 2.6.16.54-0.2.5 (SLES 10),
5 2.6.18-53.1.14.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.40.7-sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
15 Severity : enhancement
17 Description: build kernel-ib packages for OFED 1.3 in our release cycle
22 Description: fix SLES kernel versioning
23 Details : the kernel version for our SLES 10 kernel did not include a "-"
24 before the "smp" at the end. while this was not a problem in
25 general, it did mean that software trying to use the kernel
26 version to try to detect a vendor specific kernel would fail.
27 this was most evident by the OFED build scripts.
31 Description: Don't update lov_desc members until making sure they are valid
32 Details : When updating lov_desc members via proc fs, need fix their
33 validities before doing the real update.
37 Description: don't put request into delay list while invalidate in flight.
38 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
39 import in flight. this produce timeout for invalidate and sometimes
44 Frequency : on ppc only
45 Description: not convert ost objects for directory because it's not exist.
46 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
50 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
52 Severity : enhancement
54 Description: Update to support 2.6.22.14 vanilla kernel.
58 Frequency : rare, on recovery
59 Description: read procfs can produce deadlock in some situation
60 Details : Holding lprocfs lock which send rpc can produce block for destroy
61 obd objects and this also block reconnect with -EALREADY. This isn't
62 fix all lprocfs bugs - but make it rare.
64 Severity : enhancement
66 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
69 Frequency : frequent on X2 node
71 Description: mdc_set_open_replay_data LBUG
72 Details : Set replay data for requests that are eligible for replay.
76 Description: lustre_mgs: operation 101 on unconnected MGS
77 Details : When MGC is disconnected from MGS long enough, MGS will evict the
78 MGC, and late on MGC cannot successfully connect to MGS and a lot
79 of the error messages complaining that MGS is not connected.
82 Frequency : on start mds
84 Description: Implement get_info(last_id) in obdfilter.
87 Frequency : occasional
89 Description: Correctly check stale fid, not start epoch if ost not support SOM
90 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
91 overwrited with new generated fid, but mds can anwer with one of these
92 two fids and both is not stale. setattr incorectly start epoch and
93 assume will be called done_writeting, but without SOM done_writing
97 Frequency : rare, depends on device drivers and load
99 Description: MDS or OSS nodes crash due to stack overflow
100 Details : Code changes in 1.8.0 increased the stack usage of some functions.
101 In some cases, in conjunction with device drivers that use a lot
102 of stack the MDS (or possibly OSS) service threads could overflow
103 the stack. One change which was identified to consume additional
104 stack has been reworked to avoid the extra stack usage.
107 Frequency : occasional
109 Description: Do not fail import if osc_interpret_create gets -EAGAIN
110 Details : If osc_interpret_create got -EAGAIN it immediately exits and
111 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
112 oscc_has_objects and see OSC has no objests and call
113 oscc_internal_create to resend create request.
115 Severity : enhancement
117 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
119 Severity : enhancement
121 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
124 Frequency : very rare
126 Description: Fix replay if there is an un-replied request and open
127 Details : In some cases, older replay request will revert the
128 mcd->mcd_last_xid on MDS which is used to record the client's
131 Severity : enhancement
133 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
135 Severity : enhancement
137 Description: Add rhel5 support to HEAD.
139 Serverity : enhancement
141 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
146 Description: Don't allow skipping OSTs if index has been specified.
147 Details : Don't allow skipping OSTs if index has been specified, make locking
148 in internal create lots better.
152 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
153 Details : access to bitfield in structure is always rounded to long
154 and this produce problem with not atomic change any bit.
158 Description: Lustre make rpms failed.
159 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
160 builds Lustre rpms from tarball.
163 Frequency : rare on shutdown ost
165 Description: If llog cancel was not send before clean_exports phase, this can
166 produce deadlock in llog code.
167 Details : If llog thread has last reference to obd and call class_import_put
168 this produce deadlock because llog_cleanup_commit_master wait when
169 last llog_commit_thread exited, but this never success because was
170 called from llog_commit_thread.
174 Description: allow userland application know is lost one of stripes.
175 Details : fill lvb_blocks with error code on ost and return it to
176 application if error flag found.
180 Description: NULL lov_tgts causing MDS oops
181 Details : more safe checks for NULL lov_tgts for avoid oops.
183 Severity : enhancement
185 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
189 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
190 Details : If an OST is down the MDS will hang indefinitely in
191 obd_statfs_rqset() waiting for the statfs data. While for
192 MDS QOS usage of statfs, it should not stuck in waiting.
194 Severity : enhancement
196 Description: remote_acl support
197 Details : Support ACL-based permission check for remote user.
198 Support setfacl/getfacl for remote user with the utils
199 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
200 the system "{s,g}etfacl" utils.
202 Severity : enhancement
204 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
206 Severity : enhancement
208 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
212 Description: make mgs_setparam() handle fsname containing dash
213 Details : fsname containing a dash does not work with lctl conf_param
215 Severity : enhancement
217 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
219 Severity : enhancement
221 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
225 Frequency : rare, at shutdown
226 Description: access already free / zero obd_namespace.
227 Details : if client_disconnect_export was called without force flag set,
228 and exist connect request in flight, this can produce access to
229 NULL pointer (or already free pointer) when connect_interpret
230 store ocd flags in obd_namespace.
234 Frequency : only at startup
235 Description: not alloc memory with spinlock held.
236 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
237 if any spinlock held.
239 Severity : enhancement
241 Description: make lustre randomly fail allocating memory
242 Details : Make lustre randomly failed allocating memory for testing purpose.
244 Severity : enhancement
246 Description: lost problems with lov objid file
247 Details : Fixes some scability and access to not inited memory problems
248 in work with lov objdid file.
253 Description: lfs find does not continue on file error
254 Details : Continue other files processing when a file/dir is absent.
258 Description: Inconsistent usage of lustre_pack_reply()
259 Details : Standardize the usage of lustre_pack_reply() such that it
260 always generate a CERROR on failure.
263 Frequency : occasional
265 Description: MDS hang or stay in waiting lock
266 Details : If client receive lock with CBPENDING flag ldlm need send lock
267 cancel as separate rpc, to avoid situation when cancel request
268 can't processed due all i/o threads stay in wait lock.
272 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
276 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
280 Description: Update to SLES9 kernel-2.6.5-7.287.3.
284 Description: Update to RHEL4 latest kernel.
286 Severity : enhancement
288 Description: Build SLES10 patchless client fails
289 Details : The configure was broken by run ./configure with
290 --with-linux-obj=.... argument for patchless client. When the
291 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
292 can't search header adequately. Use absolute path such as
293 -I($LINUX)/include instead.
297 Description: interrupt oig_wait produce painc on resend.
298 Details : brw_redo_request can be used for resend requests from ptlrpcd and
299 private set, and this produce situation when rq_ptlrpcd_data not
300 copyed to new allocated request and triggered LBUG on assert
301 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
302 ptlrpcd set if request is changed and can be safety changed to use
307 Description: LASSERT_{REQ,REP}SWAB macros are buggy
308 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
309 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
310 these macros and replace them with their difinitions instead.
314 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
315 Details : Update which_patch & target file for SLES10 latest kernel.
318 Description: Remove client patches from SLES 10 kernel.
319 Details : This causes SLES 10 clients to behave as patchless clients
320 even on a Lustre-patched (server) kernel.
322 Severity : enhancement
324 Description: self-adjustable client's lru lists
325 Details : use adaptive algorithm for managing client cached locks lru
326 lists according to current server load, other client's work
327 pattern, memory activities, etc. Both, server and client
328 side namespaces provide number of proc tunables for controlling
331 Severity : enhancement
333 Description: light-weight GSS support
334 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
335 and reduce performance overhead.
337 Severity : enhancement
339 Description: Linux keyring support
340 Details : Support using service of Linux keyring for Lustre GSS internal
341 context refresh/cache mechanism.
345 Description: Fix errors in lfs documentation
346 Details : Fixes man pages
350 Description: don't use GFP_* in generic Lustre code.
351 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
355 Description: obdclass is limited by single OBD_ALLOC(idarray)
356 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
360 Description: warnings with build patchless client with vanila 2.6.19 and up
361 Details : change old ctl_table style and replace ctl_table/ctl_table_header
362 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
366 Frequency : Only for SLES
367 Description: sanity_quota fail test_1
368 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
369 file, which makes TSTID[2] inunique.
373 Frequency : Only for SLES9
374 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
375 Details : Update target/ChangeLog/which_patch .
379 Frequency : Only for test_10 in sanity-quota.sh
380 Description: a bug in quota test code
381 Details : checking a test flag in a wrong place
385 Frequency : Only for test_13 in sanity-quota.sh
386 Description: a bug in quota test script
387 Details : shouldn't check overfull allocated quota
391 Description: jbd statistics
392 Details : Port older jbd statistics patch for sles10
394 Severity : enhancement
396 Description: Add printing inode info into message about error in writepage.
401 Description: reply_lock_interpret crash due to race with it and lock cancel.
402 Details : Do not replay locks that are being cancelled. Do not reference
403 locks by their address during replay, just by their handle.
407 Frequency : When flocks are used.
408 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
409 Details : Flock locks might destroy just granted lock if it could be merged
410 with another existing flock, this is done in completion handler,
411 so teach ldlm_cli_enquque_fini that this is a valid case for
417 Description: Oops in read and write path when failing to allocate lock.
418 Details : Check if lock allocation failed and return error back.
422 Description: lstripe command fails for valid OST index
423 Details : The stripe offset is compared to lov->desc.ld_tgt_count
424 instead of lov->desc.ld_active_tgt_count.
428 Description: sanity.sh failed test 103
429 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
434 Description: df doesn't work properly if diskfs blocksize != 4K
435 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
439 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
440 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
444 Description: obd_health_check_timeout too short
445 Details : set obd_health_check_timeout as 1.5x of obd_timeout
449 Description: llapi_file_create() does not allow some changes
450 Details : add llapi_file_open() that allows specifying the mode and
451 open flags, and also returns an open file handle.
455 Description: merge and cleanup kernel patches.
456 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
460 Description: Add journal checksum support.(Kernel part)
461 Details : The journal checksum feature adds two new flags i.e
462 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
463 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
464 indicates that the commit block contains the checksum for
465 the blocks described by the descriptor blocks. Now commit
466 record can be sent to disk without waiting for descriptor
467 blocks to be written to disk. This behavior is controlled
468 using JBD2_FEATURE_ASYNC_COMMIT flag.
472 Description: OSS needs mutliple precreate threads
473 Details : Add ability to start more than one create thread per OSS.
477 Description: Sanity.sh test_65k failed: setstripe should have succeeded
478 Details : execute lfs setstripe on client
482 Description: mds_obd_create error creating tmp object
483 Details : When the user sets quota on root, llog will be affected and can't
484 create files and write files.
487 Frequency : Always on ia64 patchless client, and possibly others.
489 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
490 Details : This allows the patchless client to be loaded on architectures
495 Description: RedHat Update kernel for RHEL5
496 Details : Add the kernel config file for RHEL5.
500 Description: RedHat Update kernel for RHEL5
501 Details : Modify the kernel config file more closer RHEL5.
505 Description: Build failure against Centos5 (RHEL5)
506 Details : Define PAGE_SIZE when it isn't present.
510 Description: test_7 on sanity-quota.sh can't be used on separate machines
511 Details : add facet to handle it
515 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
516 Details : with NFS, the anon dentry's parent was set to itself in
517 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
518 corresponding dentry other than use rec->ur_name.
520 Severity : enhancement
522 Description: lfs setstripe enhancement
523 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
527 Description: enable data checksumming by default
528 Details : enable checksum by default, allow --disable-checksum
529 configure option and "-o nochecksum" mount option. Checksums
530 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
534 Description: lustre support for RHEL5
535 Details : Add support for RHEL5.
539 Description: Client eviction due to failover config
540 Details : after a connection loss, the lustre client should attempt to
541 reconnect to the last active server first before trying the
542 other potential connections.
546 Description: when mds and osts use different quota unit(32bit and 64bit),
547 quota will be released repeatly.
548 Details : void sending multiple quota reqs to mds, which will keep the
549 status between the reqs.
553 Description: osts not allocated evenly to files
554 Details : change the condition to increase offset_idx
557 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
559 Description: Data corruption for OSTs that are formatted larger than 2TB
561 Details : When generating the bio request for lustre file writes the
562 sector number would overflow a temporary variable before being
563 used for the IO. The data reads correctly from Lustre (which
564 will overflow in a similar manner) but other file data or
565 filesystem metadata may be corrupted in some cases.
569 Description: Tune the kernel for good SCSI performance.
570 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
571 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
576 Description: rewrite ext2-derived code in obdclass/uuid.c
577 Details : rewrite inherited code (uuid parsing code from ext2 utils)
578 from scratch preserving functionality.
582 Description: "lfs find -obd UUID" prints directories
583 Details : "lfs find -obd UUID" will return all directory names instead
584 of just file names. It is incorrect because the directories
585 do not reside on the OSTs.
589 Description: use i_size_read and i_size_write in 2.6 port
590 Details : replace inode->i_size access with i_size_read/write()
593 Frequency : when removing large files
595 Description: scheduling issue during removal of large Lustre files
596 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
597 It causes scheduling issues when removing large files (17TB in the
601 Frequency : only with liblustre clients on XT3
603 Description: evictions taking too long
604 Details : allow llrd to evict clients directly on OSTs
607 Frequency : only on ppc
609 Description: /proc/fs/lustre/devices broken on ppc
610 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
611 We should make sure the type of 'index' is loff_t and then cast
612 explicitly as needed below. Do not assign an explicitly cast
616 Frequency : only for rhel5
618 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
619 Details : Modify the target file & which_kernel.
621 Severity : enhancement
623 Description: omit set fsid for export NFS
624 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
625 as FS_REQUIRES_DEV for avoid problems with generate fsid.
629 Frequency : Always, for kernels after 2.6.16
630 Description: Fix warning idr_remove called for id=.. which is not allocated.
631 Details : Last kernels save old s_dev before kill super and not allow
632 to restore from callback - restore it before call kill_anon_super.
636 Description: buffer overruns could theoretically occur
637 Details : llapi_semantic_traverse() modifies the "path" argument by
638 appending values to the end of the origin string, and a
639 overrun may occur. Adding buffer overrun check in liblustreapi.
643 Description: Fix error on 'ls .' at the top of the Lustre mount.
644 Details : Don't revalidate dentry if it is a root dentry.
648 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
649 Details : Modify vm-tunables-rhel4.patch.
653 Description: Kernel config for 2.6.18-vanilla.
654 Details : Modify targets/2.6-vanilla.target.in.
655 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
656 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
657 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
658 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
662 Description: improve handling recoverable errors
663 Details : if request processig with error which can be recoverable on server
664 request should be resend, otherwise page released from cache and
669 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
670 Details : Modify target file & which_patch.
671 A flaw was found in the IA32 system call emulation provided
672 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
673 value could be stored in the %RAX register, which could trigger an
674 out-of-bounds system call table access. An untrusted local user
675 could exploit this flaw to run code in the kernel
676 (ie a root privilege escalation). (CVE-2007-4573).
680 Description: O_DIRECT bypasses client statistics.
681 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
682 read_ahead_stats were updated. copied the stats section from
683 osc_send_oap_rpc() into async_internal().
687 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
691 Description: change order of libsysio includes
692 Details : '#include sysio.h' should always come before '#include xtio.h'
696 Description: handle "serious error: objid * already exists" more gracefully
697 Details : If LAST_ID value on disk is smaller than the objects existing in
698 the O/0/d* directories, it indicates disk corruption and causes an
699 LBUG(). If the object is 0-length, then we should use the existing
700 object. This will help to avoid a full fsck in most cases.
702 Severity : enhancement
704 Description: adapt the lustre_config script to support the upgrade case
705 Details : Add "-u" option for lustre_config script to support upgrading 1.4
706 server targets to 1.6 in parallel.
710 Description: To avoid grant space > avaible space when the disk is almost
711 full. Without this patch you might see the error "grant XXXX >
712 available" or some LBUG about grant, when the disk is almost
714 Details : In filter_check_grant, for non_grant cache write, we should
715 check the left space by if (*left > ungranted + bytes), instead
716 of (*left > ungranted), because only we are sure the left space
717 is enough for another "bytes", then the ungrant space should be
718 increase. In client, we should update cl_avail_grant only there
719 is OBD_MD_FLGRANT in the reply.
723 Description: Update RHEL 4 kernel to fix local root privilege escalation.
724 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
725 described in CVE-2007-4573. This problem could allow untrusted
726 local users to gain root access.
729 Frequency : when using O_DIRECT and quotas
731 Description: Incorrect file ownership on O_DIRECT output files
732 Details : block usage reported by 'lfs quota' does not take into account
733 files that have been written with O_DIRECT.
738 Description: touch file failed when fs is not full
739 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
740 otherwise we can get ENOSP while fs is not full.
744 Description: parallel lock callbacks
745 Details : Instead of sending blocking and completion callbacks as separated
746 requests, adding them to a set and sending in parallel.
749 Frequency : only for Cray XT3
750 Bugzilla : 12829/13455
751 Description: Changing primary group doesn't change the group lustre assigns to
753 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
754 overridden with the primary group provided by the group upcall,
755 whereas the supplied fsgid can be trusted if it is in the list of
756 supplementary groups returned by the group upcall.
758 Severity : enhancement
760 Description: Allow masking D_WARNING, D_ERROR messages from console
761 Details : Console messages can now be disabled via lnet.printk.
765 Description: User code with malformed file open parameter crashes client node
766 Details : Before packing join_file req, all the related reference should be
767 checked carefully in case some malformed flags cause fake
768 join_file req on client.
772 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
773 Details : ldlm_completion_ast() assumes that a lock is granted when the req
774 mode is equal to the granted mode. However, it should also check
775 that LDLM_FL_CP_REQD is not set.
779 Description: Heavy nfs access might result in deadlocks
780 Details : After ELC code landed, it is now improper to enqueue any mds
781 locks under och_sem, because enqueue might want to decide to
782 cancel open locks for same inode we are holding och_sem for.
786 Description: Client eviction while running blogbench
787 Details : A lot of unlink operations with concurrent I/O can lead to a
788 deadlock causing evictions. To address the problem, the number of
789 oustanding OST_DESTROY requests is now throttled to
790 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
791 callbacks are processed in priority.
795 Description: enable ACLs on MDS by default
796 Details : ACLs must be enabled on MDS by default.
799 Frequency : PPC/PPC64 only
801 Description: conflicts between asm-ppc64/types.h and lustre_types.h
802 Details : fix duplicated definitions between asm-ppc64/types.h and
803 lustre_types.h on PPC.
806 Frequency : PPC/PPC64 only
808 Description: asm-ppc/segment.h does not exist
809 Details : fix compile issue on PPC.
813 Description: better handle error messages in extents code
816 Frequency : RHEL4 only
818 Description: mkfs is very slow on IA64/RHEL4
819 Details : A performance regression has been discovered in the MPT Fusion
820 driver between versions 3.02.73rh and 3.02.99.00rh. As a
821 consequence, we have downgraded the MPT Fusion driver in the RHEL4
822 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
824 Severity : enhancement
826 Description: SNMP support enhancement
827 Details : Adding total number of sampled request for an MDS node in snmp
830 Severity : enhancement
832 Description: Optimize ldlm waiting list processing for PR extent locks
833 Details : When processing waiting list for read extent lock and meeting read
834 lock that is same or wider to it that is not contended, skip
835 processing rest of the list and immediatelly return current
836 status of conflictness, since we are guaranteed there are no
837 conflicting locks in the rest of the list.
841 Description: Time out and refuse to reconnect
842 Details : When the failover node is the primary node, it is possible
843 to have two identical connections in imp_conn_list. We must
844 compare not conn's pointers but NIDs, otherwise we can defeat
845 connection throttling.
849 Description: port llog fixes from b1_6 into HEAD
850 Details : Port llog reference couting and some llog cleanups from b1_6
851 (bug 10800) into HEAD, for protect from panic and access to already
852 free llog structures.
856 Description: Detect stride IO mode in read-ahead
857 Details : When a client does stride read, read-ahead should detect that and
858 read-ahead pages according to the detected stride pattern.
862 Description: data checksumming impacts single node performance
863 Details : add support for several checksum algorithm. Currently, only CRC32
864 and Adler-32 are supported. The checksum type can be changed on
865 the fly via /proc/fs/lustre/osc/*/checksum_type.
869 Description: use adler32 for page checksums
870 Details : when available, use the Adler-32 algorithm instead of CRC32 for
875 Description: build for x2 fails
876 Details : fix compile issue on Cray systems.
880 Description: Properly match for duplicate locks
881 Details : Due to different lock order from skiplists code, we need to
882 traverse entire list for now
885 Frequency : only on PPC/SLES10
887 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
888 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
889 be defined. Add a hack in mkfs_lustre.c to work around this compile
894 Description: LASSERT on MDS when client holding flock lock dies
895 Details : ldlm pool logic depends on number of granted locks equal to
896 number of released locks which is not true for flock locks, so
897 just exclude such locks from consideration.
901 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
902 Details : Use fsfilt_sync() for both the whole filesystem sync and
903 individual file sync to eliminate dangerous inode locking
904 with I_LOCK that can lead to a deadlock.
908 Description: performance in 1.6.3
909 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
910 sure that 1MB requests are not fragmented by the block layer.
912 --------------------------------------------------------------------------------
914 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
916 * Support for kernels:
917 2.6.5-7.283 (SLES 9),
918 2.6.9-55.EL (RHEL 4),
919 2.6.16.46-0.14 (SLES 10),
920 2.6.18.8 vanilla (kernel.org)
921 * Client support for unpatched kernels:
922 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
923 2.6.16 - 2.6.22 vanilla (kernel.org)
924 * Due to recently discovered recovery problems, we do not recommend
925 using patchless RHEL 4 clients with this or any earlier release.
926 * Recommended e2fsprogs version: 1.39.cfs8
927 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
928 * Starting with this release, the ldiskfs backing filesystem required
929 by Lustre is now in its own package, lustre-ldiskfs. This package
930 should be installed. It is versioned separately from Lustre and
931 may be released separately in future.
935 Description: block reactivating mgc import until all deactivates complete
936 Details : Fix race when failing back MDT/MGS to itself (testing)
939 Frequency : at statup only
941 Description: mds_lov_synchronize race leads to various problems
942 Details : simultaneous MDT->OST connections at startup can cause the
943 sync to abort, leaving the OSC in a bad state.
945 Severity : enhancement
947 Description: add optional extra BUILD_VERSION info
948 Details : add a new environment variable (namely LUSTRE_VERS) which allows
949 to override the lustre version.
952 Frequency : 2.6.18 servers only
954 Description: ll_kern_mount() doesn't release the module reference
955 Details : The ldiskfs module reference count never drops down to 0
956 because ll_kern_mount() doesn't release the module reference.
961 Description: server LBUG when using old ost_num_threads parameter
962 Details : Accept the old ost_num_threads parameter but warn that it
963 is deprecated, and fix an off-by-one error that caused an LBUG.
968 Description: Transient SCSI error results in persistent IO issue
969 Details : iobuf->dr_error is not reinitialized to 0 between two
973 Frequency : sometimes when underlying device returns I/O errors
975 Description: OSTs not going read-only during write failures
976 Details : OSTs are not remounted read-only when the journal commit threads
977 get I/O errors because fsfilt_ext3 calls journal_start/stop()
978 instead of the ext3 wrappers.
982 Description: poor connect scaling with increasing client count
983 Details : Don't run filter_grant_sanity_check for more than 100 exports
984 to improve scaling for large numbers of clients.
987 Frequency : SLES10 only
989 Description: sanity-quota.sh quotacheck failed: rc = -22
990 Details : Quotas cannot be enabled on SLES10.
993 Frequency : liblustre clients only
995 Description: getdirentries does not give error when run on compute nodes
996 Details : getdirentries does not fail when the size specified as an argument
997 is too small to contain at least one entry
999 Severity : enhancement
1001 Description: Add LNET router traceability for debug purposes
1002 Details : If a checksum failure occurs with a router as part of the
1003 IO path, the NID of the last router that forwarded the bulk data
1004 is printed so it can be identified.
1009 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1010 Details : Due to a race condition, liblustre clients were occasionally
1011 evicted incorrectly.
1013 Severity : enhancement
1015 Description: lfs setstripe use optional parameters instead of postional
1018 Severity : enhancement
1020 Description: Nanosecond timestamp support for ldiskfs
1021 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1022 resolution timestamps. There is not yet support for this at
1023 the Lustre filesystem level.
1026 Frequency : during server recovery
1028 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1029 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1033 Frequency : nfs export on patchless client
1035 Description: connectathon hang when test nfs export over patchless client
1036 Details : Disconnected dentry cannot be found with lookup, so we do not need
1037 to unhash it or make it invalid
1040 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1041 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1042 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1047 Description: open req refcounting wrong on reconnect
1048 Details : If reconnect happened between getting open reply from server and
1049 call to mdc_set_replay_data in ll_file_open, we will schedule
1050 replay for unreferenced request that we are about to free.
1051 Subsequent close will crash in variety of ways.
1052 Check that request is still eligible for replay in
1053 mdc_set_replay_data().
1058 Description: disable writes to filesystem when reading health_check file
1059 Details : the default for reading the health_check proc file has changed
1060 to NOT do a journal transaction and write to disk, because this
1061 can cause reads of the /proc file to hang and block HA state
1062 checking on a healthy but otherwise heavily loaded system. It
1063 is possible to return to the previous behaviour during configure
1064 with --enable-health-write.
1066 Severity : enhancement
1068 Description: 64-bit inode version
1069 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1070 to the inode. This will be required for version-based recovery.
1075 Description: MDS fails to start if a duplicate client export is detected
1076 Details : in some rare cases it was possible for a client to connect to
1077 an MDS multiple times. Upon recovery the MDS would detect this
1078 and fail during startup. Handle this more gracefully.
1080 Severity : enhancement
1082 Description: Add -o localflock option to simulate old noflock
1084 Details : This will achieve local-only flock/fcntl locks
1090 Description: log_commit_thread vs filter_destroy race leads to crash
1091 Details : Take import reference before releasing llog record semaphore
1096 Description: Wrong request locking in request set processing
1097 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1098 imp_delayed_list, in this place should be used imp_lock.
1101 Frequency : when reconnection
1103 Description: Grant Leak when osc reconnect to OST
1104 Details : When osc reconnect ost, OST(filter) should check whether it
1105 should grant more space to client by comparing fed_grant and
1106 cl_avail_grant, and return the granted space to client instead
1107 of "new granted" space, because client will call osc_init_grant
1108 to update the client grant space info.
1111 Frequency : when client reconnect to OST
1113 Description: Grant Leak when osc do resend and replay bulk write
1114 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1115 bulk write request, because the grant info while be sync between
1116 OSC and OST when reconnect, and we should ignore the grant info
1117 these of resend/replay write req.
1122 Description: Grant space more than avaiable left space sometimes.
1123 Details : When then OST is about to be full, if two bulk writing from
1124 different clients came to OST. Accord the avaliable space of the
1125 OST, the first req should be permitted, and the second one
1126 should be denied by ENOSPC. But if the seconde arrived before
1127 the first one is commited. The OST might wrongly permit second
1128 writing, which will cause grant space > avaiable space.
1131 Frequency : when client is evicted
1133 Description: Grant might be wrongly erased when osc is evicted by OST
1134 Details : when the import is evicted by server, it will fork another
1135 thread ptlrpc_invalidate_import_thread to invalidate the
1136 import, where the grant will be set to 0. While the original
1137 thread will update the grant it got when connecting. So if
1138 the former happened latter, the grant will be wrongly errased
1139 because of this race.
1144 Description: Checking Stale with correct fid
1145 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1146 to check whether it is stale, because sometimes, we want the
1147 enqueue happened anyway, and op_data.fid2 will not be initialized.
1149 Severity : enhancement
1151 Description: update patchless client
1152 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1155 Frequency : only with 2.4 kernel
1157 Description: random memory corruption
1158 Details : size of struct ll_inode_info is to big for union inode.u and this
1159 can be cause of random memory corruption.
1164 Description: Memory leak in recovery
1165 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1166 It should also check obd_fail before fsfilt_start, otherwise if
1167 fsfilt_start return -EROFS,(failover mds during mds recovery).
1168 then the req will return with repmsg->transno = 0 and rc = EROFS.
1169 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1170 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1171 be freed no matter whether fsfilt_commit success or not.
1174 Frequency : only with huge count clients
1176 Description: Prevents from taking the superblock lock in llap_from_page for
1178 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1179 for llap_from_page call in ll_removepage prevents from taking the
1180 superblock lock for a soon died page.
1185 Description: Not check open intent error before release open handle
1186 Details : in some rare cases, the open intent error is not checked before
1187 release open handle, which may cause
1188 ASSERTION(open_req->rq_transno != 0), because it tries to release
1189 the failed open handle.
1194 Description: Set cat log bitmap only after create log success.
1195 Details : in some rare cases, the cat log bitmap is set too early. and it
1196 should be set only after create log success.
1200 Description: Accessing a block bevice can re-enable I/O when Lustre is
1201 tearing down a device.
1202 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1207 Description: service threads may hog cpus when there are a lot of requests
1208 Details : Insert cond_resched to give other threads a chance to use some CPU
1213 Description: the cat log was not initialized in recovery
1214 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1215 unlink log on mds will not be initialized until mds post
1216 recovery. And also in mds post recovery, the unlink log will
1217 initialization will be done asynchronausly, so there will be race
1218 between add unlink log and unlink log initialization.
1222 Description: brw_stats were being printed incorrectly
1223 Details : brw_stats were being printed as log2 but all of them were not
1224 recorded as log2. Also remove some code duplication arising from
1225 filter_tally_{read,write}.
1229 Frequency : rare, only in recovery.
1230 Description: ASSERTION(req->rq_type != LI_POISON) failed
1231 Details : imp_lock should be held while iterating over imp_sending_list for
1232 prevent destroy request after get timeout in ptlrpc_queue_wait.
1236 Description: replay-single.sh test 52 fails
1237 Details : A lock's skiplist need to be cleanup when it being unlinked
1238 from its resource list.
1242 Description: Short directio read returns full requested size rather than
1244 Details : Direct I/O operations should return actual amount of bytes
1245 transferred rather than requested size.
1247 Severity : enhancement
1249 Description: metadata RPC reduction (e.g. for rm performance)
1250 Details : decrease the amount of synchronous RPC between clients and servers
1251 by canceling conflicing lock before the operation on the client
1252 and packing thier handles into the main operation RPC to server.
1254 Severity : enhancement
1256 Description: Async OSC create to avoid the blocking unnecessarily.
1257 Details : If a OST has no remain object, system will block on the creating
1258 when need to create a new object on this OST. Now, ways use
1259 pre-created objects when available, instead of blocking on an
1260 empty osc while others are not empty. If we must block, we block
1261 for the shortest possible period of time.
1263 Severity : enhancement
1265 Description: refine locking for avoid write wrong info into lov_objid file
1266 Details : fix possible races with add new target and write/update data in
1269 --------------------------------------------------------------------------------
1271 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1276 Frequency : on some architectures
1278 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1279 Details : Uninitialized flags sometimes cause configuration commands to
1283 Frequency : patchless clients only
1285 Description: missing __iget() symbol export
1286 Details : The __iget() symbol export is missing. To avoid the need for
1287 this on patchless clients the deathrow inode reaper is turned
1288 off, and we depend on the VM to clean up old inodes. This
1289 dependency was during via the fix for bug 12181.
1291 --------------------------------------------------------------------------------
1293 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1295 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1296 INTEROPERATE with older versions automatically. In many cases a
1297 special upgrade step is needed. Please read the
1298 user documentation before upgrading any part of a live system.
1299 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1300 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1301 * WARNING: Lustre configuration and startup changes are required with
1302 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1308 Severity : enhancement
1310 Description: MountConf
1311 Details : Lustre configuration is now managed via mkfs and mount
1312 commands instead of lmc and lconf. New obd types (MGS, MGC)
1313 are added for dynamic configuration management. See
1314 https://mail.clusterfs.com/wikis/lustre/MountConf for
1317 Severity : enhancement
1319 Description: dynamic OST addition
1320 Details : OSTs can now be added to a live filesystem
1322 Severity : enhancement
1324 Description: startup order invariance
1325 Details : MDTs and OSTs can be started in any order. Clients only
1326 require the MDT to complete startup.
1328 Severity : enhancement
1330 Description: parallel, asynchronous orphan cleanup
1331 Details : orphan cleanup is now performed in separate threads for each
1332 OST, allowing parallel non-blocking operation.
1334 Severity : enhancement
1336 Description: optimized stripe assignment
1337 Details : stripe assignments are now made based on ost space available,
1338 ost previous usage, and OSS previous usage, in order to try
1339 to optimize storage space and networking resources.
1341 Severity : enhancement
1343 Description: Permanently set tunables
1344 Details : All writable /proc/fs/lustre tunables can now be permanently
1345 set on a per-server basis, at mkfs time or on a live system.
1347 Severity : enhancement
1349 Description: Lustre message v2
1350 Details : Add lustre message format v2.
1352 Severity : enhancement
1354 Description: client OST exclusion list
1355 Details : Clients can be started with a list of OSTs that should be
1356 declared "inactive" for known non-responsive OSTs.
1360 Description: SPEC SFS validation failure on NFS v2 over lustre.
1361 Details : Changes the blocksize for regular files to be 2x RPC size,
1362 and not depend on stripe size.
1364 Severity : enhancement
1366 Description: Multiple MD RPCs in flight.
1367 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1368 To avoid overly-overloading MDS, introduce a limit on number of
1369 MDS RPCs in flight for a single client and add /proc controls
1370 to adjust this limit.
1372 Severity : enhancement
1374 Description: client read/write statistics
1375 Details : Add client read/write call usage stats for performance
1376 analysis of user processes.
1377 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1378 file access. extents_stats shows chunk size distribution.
1379 extents_stats_per_process show chunk size distribution per
1382 Severity : enhancement
1384 Description: mds statistics
1385 Details : Add detailed mds operations statistics in
1386 /proc/fs/lustre/mds/*/stats.
1390 Description: Failure of copying files with lustre special EAs.
1391 Details : Client side always return success for setxattr call for lustre
1392 special xattr (currently only "trusted.lov").
1397 Description: Refcount LNET uuids
1398 Details : The global LNET uuid list grew linearly with every startup;
1399 refcount repeated list entries instead of always adding to
1402 Severity : enhancement
1404 Description: Dynamic service threads
1405 Details : Within a small range, start extra service threads
1406 automatically when the request queue builds up.
1409 Frequency : mixed-endian client/server environments
1411 Description: mixed-endian crashes
1412 Details : The new msg_v2 system had some failures in mixed-endian
1415 Severity : enhancement
1417 Description: Easy OST removal
1418 Details : OSTs can be permanently deactivated with e.g. 'lctl
1419 conf_param lustre-OST0001.osc.active=0'
1421 Severity : enhancement
1423 Description: MGS proc entries
1424 Details : Added basic proc entries for the MGS showing what filesystems
1427 Severity : enhancement
1429 Description: provide MGS failover
1430 Details : Added config lock reacquisition after MGS server failover.
1432 Severity : enhancement
1434 Description: add Linux 2.4 support
1435 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1439 Description: a large application tries to do I/O to the same resource and dies
1440 in the middle of it.
1441 Details : Check the req->rq_arrival time after the call to
1442 ost_brw_lock_get(), but before we do anything about
1443 processing it & sending the BULK transfer request. This
1444 should help move old stale pending locks off the queue as
1445 quickly as obd_timeout.
1448 Frequency : when an incorrect nid is specified during startup
1450 Description: ptlrpc connect to non-existant node causes kernel crash
1451 Details : LNET can't be re-entered from an event callback, which
1452 happened when we expire a message after the export has been
1453 cleaned up. Instead, hand the zombie cleanup off to another
1456 Severity : enhancement
1458 Description: plain/inodebits lock performance improvement
1459 Details : Grouping plain/inodebits in granted list by their request modes
1460 and bits policy, thus improving the performance of search through
1464 Frequency : only if OST filesystem is corrupted
1466 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1467 Details : for a short time RPCs with bulk IO are in the replay list,
1468 but replay of bulk IOs is unimplemented. If the OST filesystem
1469 is corrupted due to disk cache incoherency and then replay is
1470 started it is possible to trip an assertion. Avoid putting
1471 committed RPCs into the replay list at all to avoid this issue.
1474 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1476 Description: System hang on startup
1477 Details : This bug allowed the liblustre (e.g. catamount) client to
1478 return to the app before handling all startup RPCs. This
1479 could leave the node unresponsive to lustre network traffic
1480 and manifested as a server ptllnd timeout.
1482 Severity : enhancement
1484 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1485 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1486 boolean to control whether to print peer debug info when a
1487 client's RPC times out.
1490 Frequency : only for kernels with patches from Lustre below 1.4.3
1492 Description: Remove old rdonly API
1493 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1496 Frequency : only for devices with external journals
1498 Description: Set external device read-only also
1499 Details : During a commanded failover stop, we set the disk device
1500 read-only while the server shuts down. We now also set any
1501 external journal device read-only at the same time.
1504 Frequency : when upgrading from 1.4 while trying to change parameters
1506 Description: The wrong (new) MDC name was used when setting parameters for
1507 upgraded MDT's. Also allows changing of OSC (and MDC)
1508 parameters if --writeconf is specified at tunefs upgrade time.
1511 Frequency : when setting specific ost indicies
1513 Description: QOS code breaks on skipped indicies
1514 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1515 created with --index need not be sequential.
1517 Severity : enhancement
1519 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1520 Details : The uninit_groups feature works in conjunction with the kernel
1521 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1522 up the pass1 processing of e2fsck. This is a read-only feature
1523 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1524 mount filesystems that have had this feature enabled.
1526 Severity : enhancement
1528 Description: Improve multi-block allocation algorithm to avoid fragmentation
1529 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1530 allocation locality and avoid filesystem fragmentation.
1532 ------------------------------------------------------------------------------
1534 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1536 * Support for kernels:
1537 2.4.21-47.0.1.EL (RHEL 3)
1538 2.6.5-7.283 (SLES 9)
1539 2.6.9-42.0.10.EL (RHEL 4)
1540 2.6.12.6 vanilla (kernel.org)
1541 2.6.16.27-0.9 (SLES 10)
1542 * Recommended e2fsprogs version: 1.39.cfs5
1544 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1548 Frequency : occasional, depends on client load and configuration
1549 Bugzilla : 12181, 12203
1550 Description: data loss for recently-modified files
1552 Details : In some cases it is possible that recently written or created
1553 files may not be written to disk in a timely manner (this should
1554 normally be within 30s unless client IO load is very high).
1555 The problem appears as zero-length files or files that are a
1556 multiple of 1MB in size after a client crash or client eviction
1557 that are missing data at the end of the file.
1559 This problem is more likely to be hit on clients where files are
1560 repeatedly created and unlinked in the same directory, clients
1561 have a large amount of RAM, have many CPUs, the filesystem has
1562 many OSTs, the clients are rebooted frequently, and/or the files
1563 are not accessed by other nodes after being written.
1565 The presence of the problem can be detected by looking at
1566 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1567 smaller than the second (nr_unused) then dirty files will not
1568 be flushed automatically to disk. "sync; sleep 10" should be
1569 run several times on the node before unmounting it to update
1570 Lustre (this is also safe to run on nodes without this problem).
1572 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1573 that can cause this same problem, so customers using that kernel
1574 also need to update the kernel in addition to Lustre. In order
1575 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1577 It is normal that files written just before a client crash (less
1578 than 30s) may not yet have been flushed to disk, even for local
1582 Frequency : frequent on thin XT3 nodes
1584 Description: UUID collision on thin XT3 Linux nodes
1585 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1586 randomly, since we relied on an insufficiently-seeded PRNG.
1591 Description: OSS hangs after "All ost request buffers busy"
1592 Details : A deadlock between quota and journal operations caused OSS
1593 hangs after printing "All ost request buffers busy."
1596 Frequency : always on liblustre builds
1598 Description: Cleanup compiler warnings on liblustre
1601 Frequency : always on liblustre builds on XT3
1603 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1607 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1608 > 32000 subdirectories
1609 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1610 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1611 subdirectory count crosses 32000. This will aid e2fsck to
1612 correctly handle more than 32000 subdirectories.
1615 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1617 Description: System hang on startup
1618 Details : This bug allowed the liblustre (e.g. catamount) client to
1619 return to the app before handling all startup RPCs. This
1620 could leave the node unresponsive to lustre network traffic
1621 and manifested as a server ptllnd timeout.
1623 Severity : enhancement
1625 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1626 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1627 boolean to control whether to print peer debug info when a
1628 client's RPC times out.
1633 Description: make O_SYNC working on 2.6 kernels
1634 Details : 2.6 kernels use different method for mark pages for write,
1635 so need add a code to lustre for O_SYNC work.
1640 Description: Failure to close file and release space on NFS
1641 Details : Put inode details into lock acquired in ll_intent_file_open.
1642 Use mdc_intent_lock in ll_intent_open to properly
1643 detect all kind of errors unhandled by mdc_enqueue.
1648 Description: proc file read during shutdown sometimes raced obd removal,
1650 Details : Add lock to prevent obd access after proc file removal.
1653 Frequency : Only for files larger than 4GB on 32-bit clients.
1655 Description: improperly doing page alignment of locks
1656 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1657 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1662 Description: RPCs being resent when they shouldn't be
1663 Details : Some RPCs that should not be resent are being resent. This
1664 can cause inconsistencies in the RPC state machine. Do not
1665 resend such requests.
1668 Frequency : rare, only with NFS export
1670 Description: Crash on NFS re-export node
1671 Details : under very unusual load conditions an assertion is hit in
1672 ll_intent_file_open()
1675 Frequency : only if OST filesystem is corrupted
1677 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1678 Details : for a short time RPCs with bulk IO are in the replay list,
1679 but replay of bulk IOs is unimplemented. If the OST filesystem
1680 is corrupted due to disk cache incoherency and then replay is
1681 started it is possible to trip an assertion. Avoid putting
1682 committed RPCs into the replay list at all to avoid this issue.
1687 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1688 Details : Large single O_DIRECT read and write calls can fail to allocate
1689 a sufficiently large buffer to process the request. In case of
1690 allocation failure the allocation is retried with a smaller
1691 buffer and broken into smaller requests.
1693 Severity : enhancement
1695 Description: Add -o localflock option to simulate old noflock behaviour.
1696 Details : This will achieve local-only flock/fcntl locks coherentness.
1701 Description: versioning check is incomplete
1702 Details : Checking the version difference of client vs. server, report
1703 error if the gap is too big.
1708 Description: add support PG_writeback bit
1709 Details : add support for PG_writeback bit for Lustre, for more carefull
1710 work with page cache in 2.6 kernel. This also fix some deadlocks
1711 and remove hack for work O_SYNC with 2.6 kernel.
1713 Severity : enhancement
1715 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1716 Details : The uninit_groups feature works in conjunction with the kernel
1717 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1718 up the pass1 processing of e2fsck. This is a read-only feature
1719 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1720 mount filesystems that have had this feature enabled.
1722 Severity : enhancement
1724 Description: Improve multi-block allocation algorithm to avoid fragmentation
1725 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1726 allocation locality and avoid filesystem fragmentation.
1728 ------------------------------------------------------------------------------
1730 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1732 * Support for kernels:
1733 2.6.9-42.0.3.EL (RHEL 4)
1734 2.6.5-7.276 (SLES 9)
1735 2.4.21-47.0.1.EL (RHEL 3)
1736 2.6.12.6 vanilla (kernel.org)
1737 2.6.16.21-0.8 (SLES10)
1738 * Recommended e2fsprogs version: 1.39.cfs2-0
1740 * The backwards-compatible /proc/sys/portals symlink has been removed
1741 in this release. Before upgrading, please ensure that you change
1742 any configuration scripts or /etc/sysctl.conf files that access
1743 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1744 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1745 in advance of the upgrade on any system running Lustre 1.4.6 or
1746 newer, since /proc/sys/lnet was added in that version.
1747 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1751 Frequency : only when quota is used
1753 Description: avoid scanning export list for quota master
1754 Details : Change the algorithms to avoid scanning export list in order
1755 to improve the efficiency.
1758 Frequency : MDS failover only, very rarely
1760 Description: "went back in time" messages on mds failover
1761 Details : The greatest transno may be lost when the current operation
1762 finishes with an error (transno==0) and the client's last_rcvd
1763 record is over-written. Save the greatest transno in the
1764 mds_last_transno for this case.
1767 Frequency : always for specific kernels and striping counts
1769 Description: client may get "Matching packet too big" without ACL support
1770 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1771 when trying to access files in certain configurations. The
1772 clients should in fact be denied when mounting because they do
1773 not understand ACLs.
1776 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1778 Description: many clients connecting with IO in progress causes connect timeouts
1779 Details : Avoid synchronous journal commits to avoid delays caused by many
1780 clients connecting/disconnecting when bulk IO is in progress.
1781 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1782 OST_IO_PORTAL to avoid delays behind potentially many pending
1786 Frequency : occasionally with multiple writers to a single file
1788 Description: shared writes to file may result in wrong size reported by stat()
1789 Details : Allow growing of kms when extent lock is cancelled
1792 Frequency : always with random mmap IO to multi-striped file
1794 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1795 Details : Only if the hole is at the end of OST object so that kms is too
1796 small. Fix is to increase kms accordingly in ll_nopage.
1799 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1801 Description: writes to a missing object would leak memory on the OST
1802 Details : If there is an inconsistency between the MDS and OST filesystems,
1803 such that the MDS references an object that doesn't exist, writes
1804 to that object will leak memory due to incorrect cleanup in the
1805 error handling path, eventually running out of memory on the OST.
1810 Description: Creating too long symlink causes lustre errors
1811 Details : Check symlink and name lengths before sending requests to MDS.
1814 Frequency : only if flock is enabled (not on by default)
1816 Description: posix locks not released on fd closure on 2.6.9+
1817 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1818 kernels, this caused such locks not to be released on fd close and
1819 then assertions on fs unmount about still used locks.
1822 Frequency : MDS failover only, very rarely
1824 Description: clients may get ASSERTION(granted_lock != NULL)
1825 Details : When request was taking a long time, and a client was resending
1826 a getattr by name lock request. The were multiple lock requests
1827 with the same client lock handle and
1828 mds_getattr_name->fixup_handle_for_resent_request found one of the
1829 lock handles but later failed with ASSERTION(granted_lock != NULL).
1834 Description: handle->h_buffer_credits > 0, assertion failure
1835 Details : h_buffer_credits is zero after truncate, causing assertion
1836 failure. This patch extends the transaction or creates a new
1840 Frequency : NFS re-export or patchless client
1841 Bugzilla : 11179, 10796
1842 Description: Crash on NFS re-export node (__d_move)
1843 Details : We do not want to hash the dentry if we don't have a lock.
1844 But if this dentry is later used in d_move, we'd hit uninitialised
1845 list head d_hash, so we just do this to init d_hash field but
1846 leave dentry unhashed.
1849 Frequency : NFS re-export or patchless client
1851 Description: NFS exports has problem with symbolic link
1852 Details : lustre client didn't properly install dentry when re-exported
1853 to NFS or running patchless client.
1856 Frequency : NFS re-export or patchless client
1858 Description: Various nfs/patchless fixes.
1859 Details : fixes reuse disconected alias for lookup process - this fixes
1860 warning "find_exported_dentry: npd != pd",
1861 fix permission error with open files at nfs.
1862 fix apply umask when do revalidate.
1865 Frequency : occasional
1867 Description: Crash on NFS re-export node
1868 Details : calling clear_page() on the wrong pointer triggered oops in
1869 generic_mapping_read().
1872 Frequency : rarely, using O_DIRECT IO
1874 Description: unaligned directio crashes client with LASSERT
1875 Details : check for unaligned buffers before trying any requests.
1878 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1880 Description: stale data returned from RAID cache
1881 Details : If only a small amount of IO is done to the RAID device before
1882 reading it again it is possible to get stale data from the RAID
1883 cache instead of reading it from disk.
1886 Frequency : always for sles10 kernel
1888 Description: sles10 support
1889 Details : ll_follow_link: compile fixes and using of nd_set_link
1890 under newer kernels.
1893 Frequency : depends on arch, kernel and compiler version, always on sles10
1896 Description: recursive or deep enough symlinks cause stack overflow
1897 Details : getting rid of large stack-allocated variable in
1901 Frequency : depends on hardware
1903 Description: lustre write performance loss in the SLES10 kernel
1904 Details : the performance loss is caused by using of write barriers in the
1905 ext3 code. The SLES10 kernel turns barrier support on by
1906 default. The fix is to undo that change for ldiskfs.
1908 ------------------------------------------------------------------------------
1910 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1912 * Support for kernels:
1913 2.6.9-42.0.3EL (RHEL 4)
1914 2.6.5-7.276 (SLES 9)
1915 2.4.21-47.0.1.EL (RHEL 3)
1916 2.6.12.6 vanilla (kernel.org)
1920 Frequency : quota enabled and large files being deleted
1922 Description: releasing more than 4GB of quota at once hangs OST
1923 Details : If a user deletes more than 4GB of files on a single OST it
1924 will cause the OST to spin in an infinite loop. Release
1925 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1930 Description: statfs data retrieved from /proc may be stale or zero
1931 Details : When reading per-device statfs data from /proc, in the
1932 {kbytes,files}_{total,free,avail} files, it may appear
1933 as zero or be out of date.
1936 Frequency : systems with MD RAID1 external journal devices
1938 Description: lconf's call to blkid is confused by RAID1 journal devices
1939 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1940 of returning all block devices that match the journal UUID.
1943 Frequency : always, for aggregate stripe size over 4GB
1945 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1946 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1947 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1950 Frequency : always if "lfs find" used on a local file/directory
1952 Description: "lfs find" segfaults if used on a local file/directory
1953 Details : The case where a directory component was not specified wasn't
1954 handled correctly. Handle this properly.
1957 Frequency : always on ppc64
1959 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1960 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1961 perform endianness conversion.
1964 Frequency : rarely (truncate to non-zero file size after write under load)
1965 Bugzilla : 10730, 10687
1966 Description: Files padded with zeros to next 4K multiple
1967 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1968 it is possible that files that are truncated to a non-zero size
1969 immediately after being written are filled with zero bytes beyond
1970 the truncated size. No file data is lost.
1972 Severity : enhancement
1974 Description: Allow recovery/failover for liblustre clients.
1975 Details : liblustre clients were unaware of failover configurations until
1978 Severity : enhancement
1980 Description: user file locks should fail when not mounting with flock option
1981 Details : Set up an error-returning stub in ll_file_operations.lock field
1982 to prevent incorrect behaviour when client is mounted without
1983 flock option. Also, set up properly f_op->flock field for
1987 Frequency : always on ia64
1989 Description: "lfs df" loops on printing out MDS statfs information
1990 Details : The obd_ioctl_data was not initialized and in some systems
1991 this caused a failure during the ioctl that did not return
1992 an error. Initialize the struct and return an error on failure.
1995 Frequency : SLES 9 only
1997 Description: Error of copying files with lustre special EAs as root
1998 Details : Client side always return success for setxattr call for lustre
1999 special xattr (currently only "trusted.lov").
2002 Frequency : rarely on clusters with both ia64+i386 clients
2004 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2005 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2006 clients are doing concurrent writes to the same file it is
2007 possible that the ia64 clients may LASSERT because the OST
2008 extent locks are not PAGE_SIZE aligned. Ensure that grown
2009 locks are always aligned on the request boundary.
2012 Frequency : specific use, occasional
2014 Description: Overwriting in use executable truncates on-disk binary image
2015 Details : If one node attempts to overwrite an executable in use by
2016 another node, we now correctly return ETXTBSY instead of
2017 truncating the file.
2019 Severity : enhancement
2021 Description: Async OSC create to avoid the blocking unnecessarily.
2022 Details : If a OST has no remain object, system will block on the creating
2023 when need to create a new object on this OST. Now, ways use
2024 pre-created objects when available, instead of blocking on an
2025 empty osc while others are not empty. If we must block, we block
2026 for the shortest possible period of time.
2031 Description: chmod on Lustre root is propagated to other clients
2032 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2033 invalid by the follow_mount time.
2036 Frequency : liblustre clients only
2038 Description: Race in 'instant cancel' lock handling could lead to such locks
2039 never to be granted in case of SMP MDS
2040 Details : Do not destroy not yet granted but cbpending locks in
2044 Frequency : replay/resend of open
2046 Description: non null lock assetion failure in mds_intent_policy
2047 Details : Trying to replay/resend lockless open requests resulted in
2048 mds_open() returning 0 with no lock. Now it sets a flag if
2049 a lock is going to be returned.
2051 Severity : enhancement
2053 Description: Checksum enhancements
2054 Details : New checksum enhancements allow for resending RPCs that failed
2057 Severity : enhancement
2059 Description: Tunables on number of dirty pages in cacche
2060 Details : Allow to set limit on number of dirty pages cached.
2065 Description: client crash on unmount - lock still has references
2066 Details : In some error handling cases it was possible to leak a lock
2067 reference on a client while accessing a file. This was not
2068 harmful to the client during operation, but would cause the
2069 client to crash when the filesystem is unmounted.
2072 Frequency : specific case, rare
2074 Description: ETXTBSY on mds though file not in use by client
2075 Details : ETXTBSY is no longer incorrectly returned when attempting to
2076 chmod or chown a directory that the user previously tried to
2077 execute or a currently-executing binary.
2080 Frequency : extremely rare except on liblustre-based clients
2082 Description: Lustre space not freed when files are deleted
2083 Details : Clean up open-unlinked files after client eviction. Previously
2084 the unlink was skipped and the files remained as orphans.
2089 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2090 Details : In some cases it was possible to send a blocking callback to
2091 a client doing a glimpse, even though that client didn't get
2092 a lock granted. When the glimpse lock is cancelled on the OST
2093 the freed lock is left on the waiting list and corrupted the list.
2096 Frequency : all core dumps
2098 Description: Broke core dumps to lustre
2099 Details : Negative dentry may be unhashed if parent does not have UPDATE
2100 lock, but some callers, e.g. do_coredump, expect dentry to be
2101 hashed after successful create, hash it in ll_create_it.
2103 ------------------------------------------------------------------------------
2105 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2107 * Support for kernels:
2108 2.6.9-42.0.2.EL (RHEL 4)
2109 2.6.5-7.276 (SLES 9)
2110 2.4.21-40.EL (RHEL 3)
2111 2.6.12.6 vanilla (kernel.org)
2115 Frequency : always on RHEL 3
2117 Description: Number of open files grows over time
2118 Details : The number of open files grows over time, whether or not
2119 Lustre is started. This was due to a filp leak introduced
2120 by one of our kernel patches.
2122 ------------------------------------------------------------------------------
2124 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2126 * Support for kernels:
2127 2.6.9-42.EL (RHEL 4)
2128 2.6.5-7.276 (SLES 9)
2129 2.4.21-40.EL (RHEL 3)
2130 2.6.12.6 vanilla (kernel.org)
2135 Bugzilla : 5719, 9635, 9792, 9684
2136 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2137 Details : If a server is under heavy load and cannot reply to new
2138 connection requests before the client resends the (re)connect,
2139 the connection handling code can behave badly if two service
2140 threads are concurrently handing separate (re)connections from
2141 the same client. Add better locking to the connection handling
2142 code, and ensure that only a single connection will be processed
2143 for a given client UUID, even if the lock is dropped.
2145 Severity : enhancement
2147 Description: add TCP zero-copy support to kernel
2148 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2149 sends if the hardware supports scatter-gather and checksumming.
2150 This allows socklnd to do client-write and server-read more
2151 efficiently and reduce CPU utilization from skbuf copying.
2154 Frequency : only if NFS exporting from client
2156 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2157 Details : If setting attributes on a file created under NFS that had
2158 never been opened it would be possible to oops the client
2159 if the file had no objects.
2162 Frequency : always for liblustre
2164 Description: liblustre client does MDS+OSTs setattr RPC for each write
2165 Details : When doing a write from a liblustre client, the client
2166 incorrectly issued an RPC to the MDS and each OST the file was
2167 striped over in order to update the timestamps. When writing
2168 with small chunks and many clients this could overwhelm the MDS
2169 with RPCs. In all cases it would slow down the write because
2170 these RPCs are unnecessary.
2172 Severity : enhancement
2174 Description: allow number of MDS service threads to be changed at module load
2175 Details : It is now possible to change the number of MDS service threads
2176 running. Adding "options mds mds_num_threads={N}" to the MDS's
2177 /etc/modprobe.conf will set the number of threads for the next
2178 time Lustre is restarted (assuming the "mds" module is also
2179 reloaded at that time). The default number of threads will
2180 stay the same, 32 for most systems.
2185 Description: OST crash if filesystem is unformatted or corrupt
2186 Details : If an OST is started on a device that has never been formatted
2187 or if the filesystem is corrupt and cannot even mount then the
2188 error handling cleanup routines would dereference a NULL pointer.
2193 Description: NULL pointer deref in llap_from_page.
2194 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2195 incorrect) mapping if the page was truncated/reclaimed while it was
2196 searched for. Check for this condition and skip such pages when
2197 doing readahead. Introduce extra check to llap_from_page() to
2198 verify page->mapping->host is non-NULL (so page is not anonymous).
2201 Frequency : Sometimes when using sys_sendfile
2203 Description: "page not covered by a lock" warnings from ll_readpage
2204 Details : sendfile called ll_readpage without right page locks present.
2205 Now we introduced ll_file_sendfile that does necessary locking
2206 around call to generic_file_sendfile() much like we do in
2210 Frequency : with certain MDS communication failures at client mount time
2212 Description: NULL pointer deref after failed client mount
2213 Details : a client connection request may delayed by the network layer
2214 and not be sent until after the PTLRPC layer has timed out the
2215 request. If the client fails the mount immediately it will try
2216 to clean up before the network times out the request. Add a
2217 reference from the request import to the obd device and delay
2218 the cleanup until the network drops the request.
2221 Frequency : occasionally during client (re)connect
2223 Description: assertion failure during client (re)connect
2224 Details : processing a client connection request may be delayed by the
2225 client or server longer than the client connect timeout. This
2226 causes the client to resend the connection request. If the
2227 original connection request is replied in this interval, the
2228 client may trip an assertion failure in ptlrpc_connect_interpret()
2229 which thought it would be the only running connect process.
2232 Frequency : only with obd_echo servers and clients that are rebooted
2234 Description: kernel BUG accessing uninitialized data structure
2235 Details : When running an obd_echo server it did not start the ping_evictor
2236 thread, and when a client was evicted an uninitialized data
2237 structure was accessed. Start the ping_evictor in the RPC
2238 service startup instead of the OBD startup.
2240 Severity : enhancement
2241 Bugzilla : 10193 (patchless)
2242 Description: Remove dependency on various unexported kernel interfaces.
2243 Details : No longer need reparent_to_init, exit_mm, exit_files,
2244 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2247 Frequency : rare (only users of deprecated and unsupported LDAP config)
2249 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2250 Details : LDAP apparently contains 'lustreName' attributes instead of
2251 'name'. A simple remapping of the name is sufficient.
2254 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2256 Description: waiting_locks_callback trips kernel BUG if client is evicted
2257 Details : Running with the dump_on_timeout debug flag turned on makes
2258 it possible that the waiting_locks_callback() can try to dump
2259 the Lustre kernel debug logs from an interrupt handler. Defer
2260 this log dumping to the expired_lock_main() thread.
2262 Severity : enhancement
2264 Description: Support NFS exporting on 2.6 kernels.
2265 Details : Implement non-rawops metadata methods for NFS server to use without
2266 changing NFS server code.
2269 Frequency : very rare (synthetic metadata workload only)
2271 Description: two racing renames might cause an MDS thread to deadlock
2272 Details : Running the "racer" program may cause one MDS thread to rename
2273 a file from being the source of a rename to being the target of
2274 a rename at exactly the same time that another thread is doing
2275 so, and the second thread has already enqueued these locks after
2276 doing a lookup of the target and is trying to relock them in
2277 order. Ensure that we don't try to re-lock the same resource.
2280 Frequency : only very large systems with liblustre clients
2282 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2283 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2284 all OSTs in parallel. This allows the eviction of stale liblustre
2285 clients to proceed much faster than if they were done in series,
2286 and also offers similar improvements for other set_info RPCs.
2291 Description: excessive CPU usage during initial read phase on client
2292 Details : During the initial read phase on a client, it would agressively
2293 retry readahead on the file, consuming too much CPU and impacting
2294 performance (since 1.4.5.8). Improve the readahead algorithm
2295 to avoid this, and also improve some other common cases (read
2296 of small files in particular, where "small" is files smaller than
2297 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2302 Description: MDS crash when receiving packet with unknown intent.
2303 Details : Do not LBUG in unknown intent case, just return -EFAULT
2305 Severity : enhancement
2306 Bugzilla : 9293, 9385
2307 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2308 Details : Do not serialize getattr (non-intent version) and statfs.
2311 Frequency : occasional, when OST network is overloaded/intermittent
2313 Description: client evicted by OST after bulk IO timeout
2314 Details : If a client sends a bulk IO request (read or write) the OST
2315 may evict the client if it is unresposive to its data GET/PUT
2316 request. This is incorrect if the network is overloaded (takes
2317 too long to transfer the RPC data) or dropped the OST GET/PUT
2318 request. There is no need to evict the client at all, since
2319 the pinger and/or lock callbacks will handle this, and the
2320 client can restart the bulk request.
2323 Frequency : Always when mmapping file with no objects
2325 Description: client crashes when mmapping file with no objects
2326 Details : Check that we actually have objects in a file before doing any
2327 operations on objects in ll_vm_open, ll_vm_close and
2333 Description: Request leak when working with deleted CWD
2334 Details : Introduce advanced request refcount tracking for requests
2335 referenced from lustre intent.
2337 Severity : Enhancement
2339 Description: Cache open file handles on client.
2340 Details : MDS now will return special lock along with openhandle, if
2341 requested and client is allowed to hold openhandle, even if unused,
2342 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2343 closing files for every read/write openration.
2345 Severity : Enhancement
2347 Description: Cache open negative dentries on client when possible.
2348 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2349 negative dentries on lock revocation.
2354 Description: Remounting a client read-only wasn't possible with a zconf mount
2355 Details : It wasn't possible to remount a client read-only with llmount.
2357 Severity : enhancement
2358 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2359 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2360 MPICH is included to add Lustre-specific ADIO interfaces.
2361 This is based closely on the UFS ADIO layer and only differs
2362 in file creation, in order to allow the OST striping to be set.
2363 This is user-contributed code and not supported by CFS.
2368 Description: extended inode attributes (immutable, append-only) work improperly
2369 when 2.4 and 2.6 kernels are used on client/server or vice versa
2370 Details : Introduce kernel-independent values for these flags.
2372 Severity : enhancement
2375 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2376 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2377 granularity. Now, Fractional MB granularity is be supported,
2378 this is very useful for low memory system.
2380 Severity : enhancement
2382 Description: Getattr by fid
2383 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2384 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2385 have locks and avoids getattr rpc requests for it on every stat.
2388 Frequency : Always, for filesystems larger than 2TB
2390 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2391 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2392 filesystems larger than 2TB to be mounted with mballoc enabled.
2395 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2397 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2398 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2399 filesystems larger than 2TB is mounted then the kernel will
2400 silently corrupt the start of the filesystem. CONFIG_LBD is
2401 enabled for all CFS-supported kernels, but the possibility of
2402 this happening with a modified kernel config exists.
2404 Severity : enhancement
2406 Description: add client O_DIRECT support for 2.6 kernels
2407 Details : It is now possible to do O_DIRECT reads and writes to files
2408 in the Lustre client mountpoint on 2.6 kernel clients.
2410 Severity : enhancement
2412 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2413 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2414 parallel, not waiting for response from every OST before sending
2415 a rpc to the next OST.
2420 Description: setattr vs write race when updating file timestamps
2421 Details : Client processes that update a file timestamp into the past
2422 right after writing to the file (e.g. tar) it is possible that
2423 the updated file modification time can be reset to the current
2424 time due to a race between processing the setattr and write RPC.
2426 Severity : enhancement
2428 Description: Bring 'lfs find' closer in line with regular Linux find.
2429 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2430 -print0 options and obtains all the needed info through the lustre
2433 Severity : enhancement
2435 Description: support up to 1024 configured devices on one node
2436 Details : change obd_dev array from statically allocated to dynamically
2437 allocated structs as they are first used to reduce memory usage
2442 Description: Flush dirty partially truncated pages during truncate
2443 Details : Immediatelly flush partially truncated pages in filter_setattr,
2444 this way we completely avoid having any pages in page cache on OST
2445 and can retire ugly workarounds during writes to flush such pages.
2450 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2451 Details : protect inode from truncation within vfs_unlink() context
2452 just take a reference before calling vfs_unlink() and release it
2453 when parent's i_sem is free.
2458 Description: last_id value checked outside lock on OST caused LASSERT failure
2459 Details : If there were multiple MDS->OST object precreate requests in
2460 flight, it was possible that the OST's last object id was checked
2461 outside a lock and incorrectly tripped an assertion. Move checks
2462 inside locks, and discard old precreate requests.
2465 Frequency : always, if extents are used on OSTs
2467 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2468 Details : OSTs using the extents format would not zero the high 16 bits of
2469 the index physical block number. This is not a problem for any
2470 OST filesystems smaller than 16TB, and no kernels support ext3
2471 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2472 new/modified files) and can be fixed for existing filesystems
2473 with e2fsprogs-1.39-cfs1.
2478 Description: import connection selection may be incorrect if timer wraps
2479 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2480 import connections to be ignored if the 32-bit jiffies counter
2481 wraps. Use a 64-bit jiffies counter.
2484 Frequency : very large clusters immediately after boot
2486 Description: LNET request buffers exhausted under heavy short-term load
2487 Details : If a large number of client requests are generated on a service
2488 that has previously never seen so many requests it is possible
2489 that the request buffer growth cannot keep up with the spike in
2490 demand. Instead of dropping incoming requests, they are held in
2491 the LND until the RPC service can accept more requests.
2494 Frequency : Sometimes during replay
2496 Description: Assertion failure in ll_local_open after replay.
2497 Details : If replay happened on an open request reply before we were able
2498 to set replay handler, reply will become not swabbed tripping the
2499 assertion in ll_local_open. Now we set the handler right after
2500 recognising of open request
2503 Frequency : very rare
2505 Description: kernel reports "badness in vsnprintf"
2506 Details : Reading from the "recovery_status" /proc file in small chunks
2507 may cause a negative length in lprocfs_obd_rd_recovery_status()
2508 call to vsnprintf() (which is otherwise harmless). Exit early
2509 if there is no more space in the output buffer.
2511 Severity : enhancement
2513 Description: clear OBD RPC statistics by writing to them
2514 Details : It is now possible to clear the OBD RPC statistics by writing
2515 to the "stats" file.
2520 Description: Client mtime is not the same on different clients after utimes
2521 Details : In some cases, the client was using the utimes() syscall on
2522 a file cached on another node. The clients now validate the
2523 ctime from the MDS + OSTs to determine which one is right.
2528 Description: Inability to activate failout mode
2529 Details : lconf script incorrectly assumed that in pythong string's numeric
2530 value is used in comparisons.
2533 Frequency : always with multiple stripes per file
2535 Description: Inefficient object allocation for mutli-stripe files
2536 Details : When selecting which OSTs to stripe files over, for files with
2537 a stripe count that divides evenly into the number of OSTs,
2538 the MDS is always picking the same starting OST for each file.
2539 Return the OST selection heuristic to the original design.
2544 Description: mount failures may take full timeout to return an error
2545 Details : Under some heavy load conditions it is possible that a
2546 failed mount can wait for the full obd_timeout interval,
2547 possibly several minutes, before reporting an error.
2548 Instead return an error as soon as the status is known.
2550 Frequency : quota enabled and large files being deleted
2552 Description: releasing more than 4GB of quota at once hangs OST
2553 Details : If a user deletes more than 4GB of files on a single OST it
2554 will cause the OST to spin in an infinite loop. Release
2555 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2560 Description: statfs data retrieved from /proc may be stale or zero
2561 Details : When reading per-device statfs data from /proc, in the
2562 {kbytes,files}_{total,free,avail} files, it may appear
2563 as zero or be out of date.
2566 Frequency : systems with MD RAID1 external journal devices
2568 Description: lconf's call to blkid is confused by RAID1 journal devices
2569 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2570 of returning all block devices that match the journal UUID.
2573 Frequency : always, for aggregate stripe size over 4GB
2575 Description: assertion fails when trying to use 4GB stripe size
2576 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2577 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2580 Frequency : always on ppc64
2582 Description: the first write on an ext3 filesystem with mballoc got stuck
2583 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2584 perform endianness conversion.
2586 ------------------------------------------------------------------------------
2588 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2590 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2591 INTEROPERATE with older versions automatically. Please read the
2592 user documentation before upgrading any part of a live system.
2593 * WARNING: Lustre networking configuration changes are required with
2594 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2597 * Support for kernels:
2598 2.6.9-22.0.2.EL (RHEL 4)
2599 2.6.5-7.244 (SLES 9)
2600 2.6.12.6 vanilla (kernel.org)
2603 Severity : enhancement
2604 Bugzilla : 7981/8208
2605 Description: Introduced Lustre Networking (LNET)
2606 Details : LNET is new networking infrastructure for Lustre, it includes
2607 a reorganized network configuration mode (see the user
2608 documentation for full details) as well as support for routing
2609 between different network fabrics. Lustre Networking Devices
2610 (LNDs) for the supported network fabrics have also been
2611 created for this new infrastructure.
2613 Severity : enhancement
2614 Description: Introduced Access control lists
2615 Details : clients can set ACLs on files and directories in order to have
2616 more fine-grained permissions than the standard Unix UGO+RWX.
2617 The MDS must be started with the "-o acl" mount option.
2619 Severity : enhancement
2620 Description: Introduced filesystem quotas
2621 Details : Administrators may now establish per-user quotas on the
2624 Severity : enhancement
2626 Description: Configuration change for the XT3
2627 The PTLLND is now used to run Lustre over Portals on the XT3
2628 The configure option(s) --with-cray-portals are no longer used.
2629 Rather --with-portals=<path-to-portals-includes> is used to
2630 enable building on the XT3. In addition to enable XT3 specific
2631 features the option --enable-cray-xt3 must be used.
2636 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2637 Details : A race between allocating a new llcd and re-getting the llcd_lock
2638 allowed another thread to grab newly-allocated llcd.
2640 Severity : enhancement
2642 Description: 2.6 OST async journal commit and locking fix to improve performance
2643 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2644 2.6 kernels are now async as they already were in 2.4 kernels so
2645 that they can commit concurrently with the network bulk transfer.
2646 For block-allocated files the filter allocation semaphore is held
2647 to avoid filesystem fragmentation during allocation. BKL lock
2648 removed for 2.6 xattr operations where it is no longer needed.
2653 Description: lconf incorrectly determined whether two IP networks could talk
2654 Details : In some more complicated routing and multiple-network
2655 configurations, lconf will avoid trying to make a network
2656 connection to a disjoint part of the IP space. It was doing the
2657 math incorrectly for one set of cases.
2662 Description: Fix for potential infinite loop processing records in an llog.
2663 Details : If an llog record is corrupted/zeroed, it is possible to loop
2664 forever in llog_process(). Validate the llog record length
2665 and skip the remainder of the block on error.
2668 Frequency : occasional (liblustre only)
2670 Description: liblustre could not open files whose last component is a symlink
2671 Details : sysio_path_walk() would incorrectly pass the open intent to
2672 intermediate path components.
2675 Frequency : rare (liblustre only with non-standard tuning)
2676 Bugzilla : 7201 (7350)
2677 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2678 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2679 after it has been released, only do this for referenced locks
2681 Severity : enhancement
2683 Description: specifying an (invalid) directory default stripe_size of -1
2684 would reset the directory default striping
2685 Details : stripe_size -1 was used internally to signal directory stripe
2686 removal, now use "all default" to signal dir stripe removal
2687 as a directory striping of "all default" is not useful
2690 Frequency : common for large clusters running liblustre clients
2692 Description: doing an ls when liblustre clients are running is slow
2693 Details : sending a glimpse AST to a liblustre client waits for every AST
2694 to time out, as liblustre clients will not respond. Since they
2695 cannot cache data we refresh the OST lock LVB from disk instead.
2697 Severity : enhancement
2699 Description: doing an ls at the same time as file IO can be slow
2700 Details : enqueue and other "small" requests can be blocked behind many
2701 large IO requests. Create a new OST IO portal for non-IO
2702 requests so they can be processed faster.
2705 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2707 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2708 created file with mode 000
2709 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2710 allow an ftruncate() as a non-root user to a file with mode 000.
2711 Lustre now respects this flag to disable mode checks when
2712 truncating a file owned by the user
2715 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2718 Description: Revoking locks from clients that went dead or catatonic might take
2720 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2721 cancellation of such locks instant on servers without waiting for
2722 any reply from clients. Clients drops these locks when cancel
2723 notification from server is received without replying.
2726 Frequency : liblustre-only, when liblustre client dies or becomes busy
2728 Description: Doing ls on Linux clients can take a long time with active
2730 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2731 granting such locks to it in the first place if possible. Locks
2732 are taken by proxy on the OST during the read or write and
2733 dropped immediately afterward. Add connect flags handling, do
2734 not grant locks to liblustre clients for glimpse ASTs.
2736 Severity : enhancement
2738 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2739 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2740 sequential reads and behaves badly with applications doing random
2741 reads. Improve it by reading ahead at least read region, and
2742 avoiding excessive large RPC for small reads.
2744 Severity : enhancement
2746 Description: Creating more than 1000 files for a single job may cause a load
2747 imbalance on the OSTs if there are also a large number of OSTs.
2748 Details : qos_prep_create() uses an OST index reseed value that is an
2749 even multiple of the number of available OSTs so that if the
2750 reseed happens in the middle of the object allocation it will
2751 still utilize the OSTs as uniformly as possible.
2756 Description: OST or MDS may oops in ping_evictor_main()
2757 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2758 but doesn't restart at beginning of obd_exports_timed list
2761 Severity : enhancement
2763 Description: improve by-nid export eviction on the MDS and OST
2764 Details : allow multiple exports with the same NID to be evicted at one
2765 time without re-searching the exports list.
2768 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2770 Description: MDS may oops in groups_free()
2771 Details : in rare race conditions a newly allocated group_info struct is
2772 freed again, and this can be NULL. The 2.4 compatibility code
2773 for groups_free() checked for a NULL pointer, but 2.6 did not.
2776 Frequency : common for liblustre clients doing little filesystem IO
2777 Bugzilla : 9352, 7313
2778 Description: server may evict liblustre clients accessing contended locks
2779 Details : if a client is granted a lock or receives a completion AST
2780 with a blocking AST already set it would not reply to the AST
2781 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2784 Frequency : lfs setstripe, only systems with more than 160 OSTs
2786 Description: unable to set striping with a starting offset beyond OST 160
2787 Details : llapi_create_file() incorrectly limited the starting stripe
2788 index to the maximum single-file stripe count.
2791 Frequency : LDAP users only
2793 Description: lconf did not handle in-kernel recovery with LDAP properly
2794 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2796 Severity : enhancement
2798 Description: bind OST threads to NUMA nodes to improve performance
2799 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2800 node and do their allocations there to localize memory access
2802 Severity : enhancement
2804 Description: llmount can determine client NID directly from Myrinet (GM)
2805 Details : the client NID code from gmnalnid was moved directly into
2806 llmount, removing the need to use this or specifying the
2807 client NID explicitly when mounting GM clients with zeroconf
2810 Frequency : if client is started with down MDS
2812 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2813 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2814 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2815 ptlrpc_import_delayed_req() because we didn't check if the
2816 request was interrupted, and we also didn't break out of the
2817 event loop if there was no timeout
2822 Description: data loss during non-page-aligned writes to a single file from
2823 both multiple nodes and multiple threads on one node at same time
2824 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2825 inconsistency led to false short-reads, that were cached and later
2826 used by ->prepare_write() to fill in partially written page,
2827 leading to data loss.
2830 Frequency : always, if lconf --abort_recovery used
2832 Description: lconf --abort_recovery fails with 'Operation not supported'
2833 Details : lconf was attempting to abort recovery on the MDT device and not
2836 Severity : enhancement
2838 Description: remove cleanup logs
2839 Details : replace lconf-generated cleanup logs with lustre internal
2840 cleanup routines. Eliminates the need for client-cleanup and
2843 Severity : enhancement
2845 Description: add support for EAs (user and system) on lustre filesystems
2846 Details : it is now possible to store extended attributes in the Lustre
2847 client filesystem, and with the user_xattr mount option it
2848 is possible to allow users to store EAs on their files also
2850 Severity : enhancement
2852 Description: Add possibility (config option) to show minimal available OST free
2854 Details : When compiled with --enable-mindf configure option, statfs(2)
2855 (and so, df) will return least minimal free space available from
2856 all OSTs as amount of free space on FS, instead of summary of
2857 free spaces of all OSTs.
2859 Severity : enhancement
2861 Description: do not expand extent locks acquired on OST-side
2862 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2863 by server: they are not cached anyway.
2866 Frequency : when mmap is used/binaries executed from Lustre
2868 Description: Unmmap pages before throwing them away from read cache.
2869 Details : llap_shrink cache now attempts to unmap pages before discarding
2870 them (if unmapping failed - do not discard). SLES9 kernel has
2871 extra checks that trigger if this unmapping is not done first.
2876 Description: lconf didn't resolve symlinks before checking to see whether a
2877 given mountpoint was already in use
2880 Frequency : when migrating failover services
2881 Bugzilla : 6395, 9514
2882 Description: When migrating a subset of services from a node (e.g. failback
2883 from a failover service node) the remaining services would
2884 time out and evict clients.
2885 Details : lconf --force (implied by --failover) sets the global obd_timeout
2886 to 5 seconds in order to quickly disconnect, but this caused
2887 other RPCs to time out too quickly. Do not change the global
2888 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2890 Severity : enhancement
2891 Frequency : if MDS is started with down OST
2892 Bugzilla : 9439,5706
2893 Description: Allow startup/shutdown of an MDS without depending on the
2894 availability of the OSTs.
2895 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2896 Add appropriate locking and lov-osc refcounts for safe
2897 cleaning. Add osc abort_inflight calls in case the
2898 synchronize never started.
2901 Frequency : occasional (Cray XT3 only)
2903 Description: root not authorized to access files in CRAY_PORTALS environment
2904 Details : The client process capabilities were not honoured on the MDS in
2905 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2906 been accessed by an authorized user then root was able to access
2907 the file on the local client also. The root user capabilities
2908 are now allowed on the MDS, as this environment has secure UID.
2911 Frequency : occasional
2913 Description: ldiskfs "too long searching" message happens too often
2914 Details : A debugging message (otherwise harmless) prints too often on
2915 the OST console. This has been reduced to only happen when
2916 there are fragmentation problems on the filesystem.
2921 Description: Division by zero in statfs when all OSCs are inactive
2922 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2923 lov_statfs divides by value returned by lov_get_stripecnt().
2927 Bugzilla : 9489, 3273
2928 Description: First write from each client to each OST was only 4kB in size,
2929 to initialize client writeback cache, which caused sub-optimal
2930 RPCs and poor layout on disk for the first writen file.
2931 Details : Clients now request an initial cache grant at (re)connect time
2932 and so that they can start streaming writes to the cache right
2933 away and always do full-sized RPCs if there is enough data.
2934 If the OST is rebooted the client also re-establishes its grant
2935 so that client cached writes will be honoured under the grant.
2940 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2941 Details : Now I/O RPCs go to different portal number and (presumably) fast
2942 lock requests (and glimses) and other RPCs get their own service
2943 threads pool that should be able to service those RPCs
2946 Severity : enhancement
2948 Description: Ability to exchange lustre version between client and servers and
2949 issue warnings at client side if client is too old. Also for
2950 liblustre clients there is ability to refuse connection of too old
2952 Details : New 'version' field is added to connect data structure that is
2953 filled with version info. That info is later checked by server and
2957 Frequency : rare, liblustre only.
2958 Bugzilla : 9296, 9581
2959 Description: Two simultaneous writes from liblustre at offset within same page
2960 might proceed at the same time overwriting eachother with stale
2962 Details : I/O lock withing llu_file_prwv was released too early, before data
2963 actually was hitting the wire. Extended lock-holding time until
2964 server acknowledges receiving data.
2967 Frequency : extremely rare. Never observed in practice.
2969 Description: avoid generating lustre_handle cookie of 0.
2970 Details : class_handle_hash() generates handle cookies by incrementing
2971 global counter, and can hit 0 occasionaly (this is unlikely, but
2972 not impossible, because initial value of cookie counter is
2973 selected randonly). Value of 0 is used as a sentinel meaning
2974 "unassigned handle" --- avoid it. Also coalesce two critical
2975 sections in this function into one.
2977 Severity : enhancement
2979 Description: allow liblustre clients to delegate truncate locking to OST
2980 Details : To avoid overhead of locking, liblustre client instructs OST to
2981 take extent lock in ost_punch() on client's behalf. New connection
2982 flag is added to handle backward compatibility.
2984 Severity : enhancement
2985 Bugzilla : 4928, 7341, 9758
2986 Description: allow number of OST service threads to be specified
2987 Details : a module parameter allows the number of OST service threads
2988 to be specified via "options ost ost_num_threads={N}" in the
2989 OSS's /etc/modules.conf or /etc/modprobe.conf.
2993 Bugzilla : 6146, 9635, 9895
2994 Description: servers crash with bad pointer in target_handle_connect()
2995 Details : In rare cases when a client is reconnecting it was possible that
2996 the connection request was the last reference for that export.
2997 We would temporarily drop the export reference and get a new
2998 one, but this may have been the last reference and the export
2999 was just destroyed. Get new reference before dropping old one.
3001 Severity : enhancement
3002 Frequency : if client is started with failover MDS
3004 Description: Allow multiple MDS hostnames in the mount command
3005 Details : Try to read the configuration from all specified MDS
3006 hostnames during a client mount in case the "primary"
3009 Severity : enhancement
3011 Description: Stop sending data to evicted clients as soon as possible.
3012 Details : Check if the client we are about to send or are sending data to
3013 was evicted already. (Check is done every second of waiting,
3014 for which l_wait_event interface was extended to allow checking
3015 of exit condition at specified intervals).
3018 Frequency : rare, normally only when NFS exporting is done from client
3020 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3022 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3025 Frequency : always, if srand() is called before liblustre initialization
3027 Description: Liblustre uses system PRNG disturbing its usage by user application
3028 Details : Introduce internal to lustre fast and high-quality PRNG for
3029 lustre usage and make liblustre and some other places in generic
3030 lustre code to use it.
3032 Severity : enhancement
3033 Bugzilla : 9477, 9557, 9870
3034 Description: Verify that the MDS configuration logs are updated when xml is
3035 Details : Check if the .xml configuration logs are newer than the config
3036 logs stored on the MDS and report an error if this is the case.
3037 Request --write-conf, or allow starting with --old_conf.
3039 Severity : enhancement
3041 Description: Handle symlinks in the path when checking if Lustre is mounted.
3042 Details : Resolve intermediate symlinks when checking if a client has
3043 mounted a filesystem to avoid duplicate client mounts.
3048 Description: lconf can hit an error exception but still return success.
3049 Details : The lconf command catches the Command error exception at the top
3050 level script context and will exit with the associated exit
3051 status, but doesn't ensure that this exit status is non-zero.
3056 Description: failure of ptlrpc thread startup can cause oops
3057 Details : Starting a ptlrpc service thread can fail if there are a large
3058 number of threads or the server memory is very fragmented.
3059 Handle this without oopsing.
3062 Frequency : always, only if liblustre and non-default acceptor port was used
3064 Description: liblustre cannot connect to servers with non-default acceptor port
3065 Details : tcpnal_set_default_params() was not called and was therefore
3066 ignoring the environment varaible TCPNAL_PORT, as well as other
3067 TCPNAL_ environment variables
3072 Description: two objects could be created on the same OST for a single file
3073 Details : If an OST is down, in some cases it was possible to create two
3074 objects on a single OST for a single file. No problems other
3075 than potential performance impact and spurious error messages.
3079 Bugzilla : 5681, 9562
3080 Description: Client may oops in ll_unhash_aliases
3081 Details : Client dcache may become inconsistent in race condition.
3082 In some cases "getcwd" can fail if the current directory is
3088 Description: Inode refcounting problems in NFS export code
3089 Details : link_raw functions used to call d_instantiate without obtaining
3090 extra inode reference first.
3094 Bugzilla : 9942, 9903
3095 Description: Referencing freed requests leading to crash, memleaks with NFS.
3096 Details : We used to require that call to ll_revalidate_it was always
3097 followed by ll_lookup_it. Also with revalidate_special() it is
3098 possible to call ll_revalidate_it() twice for the same dentry
3099 even if first occurence returned success. This fix changes semantic
3100 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3101 reference on a request referred from the intent.
3102 ll_intent_release() then releases such a request.
3105 Frequency : rare, normally benchmark loads only
3107 Description: unlinked inodes were kept in memory on the client
3108 Details : If a client is repeatedly creating and unlinking files it
3109 can accumulate a lot of stale inodes in the inode slab cache.
3110 If there is no other client load running this can cause the
3111 client node to run out of memory. Instead flush old inodes
3112 from client cache that have the same inode number as a new inode.
3115 Frequency : SLES9 2.6.5 kernel and long filenames only
3116 Bugzilla : 9969, 10379
3117 Description: utime reports stale NFS file handle
3118 Details : SLES9 uses out-of-dentry names in some cases, which confused
3119 the lustre dentry revalidation. Change it to always use the
3123 Frequency : rare, unless heavy write-truncate concurrency is continuous
3124 Bugzilla : 4180, 6984, 7171, 9963, 9331
3125 Description: OST becomes very slow and/or deadlocked during object unlink
3126 Details : filter_destroy() was holding onto the parent directory lock
3127 while truncating+unlinking objects. For very large objects this
3128 may block other threads for a long time and slow overall OST
3129 responsiveness. It may also be possible to get a lock ordering
3130 deadlock in this case, or run out of journal credits because of
3131 the combined truncate+unlink. Solution is to do object truncate
3132 first in one transaction without parent lock, and then do the
3133 final unlink in a new transaction with the parent lock. This
3134 reduces the lock hold time dramatically.
3137 Frequency : rare, 2.4 kernels only
3139 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3140 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3141 before unmounting their filesystem in order to avoid deadlock.
3142 The kernel_locked() function in 2.4 kernels only checks whether
3143 the kernel lock is held, not whether it is this process that is
3144 holding it as 2.6 kernels do.
3149 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3150 Details : The client ptlrpc code may be trying to reconnect to a down
3151 server before a previous connection attempt has timed out.
3152 Increase the reconnect interval to be longer than the connection
3153 timeout interval to avoid sending duplicate connections to
3157 Frequency : echo_client brw_test command
3159 Description: fix echo_client to work with OST preallocated code
3160 Details : OST preallocation code (5137) didn't take echo_client IO path
3161 into account: echo_client calls filter methods outside of any
3162 OST thread and, hence, there is no per-thread preallocated
3163 pages and buffers to use. Solution: hijack pga pages for IO. As
3164 a byproduct, this avoids unnecessary data copying.
3168 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3169 Description: Client can oops in mdc_commit_close() after open replay
3170 Details : It was possible for the MDS to return an open request with no
3171 transaction number in mds_finish_transno() if the client was
3172 evicted, but without actually returning an error. Clients
3173 would later try to replay that open and may trip an assertion
3174 Simplify the client close codepath, and always return an error
3175 from the MDS in case the open is not successful.
3178 Frequency : rare, 2.6 OSTs only
3180 Description: OST may deadlock under high load on fragmented files
3181 Details : If there was a heavy load and highly-fragmented OST filesystems
3182 it was possible to have all the OST threads deadlock waiting on
3183 allocation of biovecs, because the biovecs were not released
3184 until the entire RPC IO was completed. Instead, release biovecs
3185 as soon as they are complete to ensure forward IO progress.
3187 Severity : enhancement
3189 Description: Support for specifying external journal device at mount
3190 Details : If an OST or MDS device is formatted with an external journal
3191 device, this device major/minor is stored in the ext3 superblock
3192 and may not be valid for failover. Allow detecting and
3193 specifying the external journal at mount time.
3198 Description: Mounting an MDS with pending unlinked files may cause oops
3199 Details : target_finish_recovery() calls mds_postrecov() which returned
3200 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3201 considers this an error and immediately begins cleaning up the
3202 lov, just after starting the mds_lov process
3204 Severity : enhancement
3206 Description: Implement 'lfs df' to report actual free space on per-OST basis
3207 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3208 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3209 usage of objects; '-h' to report in human readable format.
3211 ------------------------------------------------------------------------------
3213 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3220 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3221 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3222 at startup during orphan unlinking. Instead, skip buddy update
3223 before setup, as it will be regenerated after recovery is complete.
3228 Description: If an OST is inactive, its locks might reference stale inodes.
3229 Details : lov_change_cbdata() must iterate over all namespaces, even if
3230 they are inactive to clear inode references from the lock.
3232 Severity : enhancement
3233 Frequency : occasional, if non-standard max_dirty_mb used
3235 Description: Client will block write RPCs if not enough grant
3236 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3237 then the client will block writes while waiting for another RPC
3238 to complete instead of consuming its dirty limit. With change
3239 we get improved performance when max_dirty_mb is small.
3241 Severity : enhancement
3242 Bugzilla : 3389, 6253
3243 Description: Add support for supplementary groups on the MDS.
3244 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3245 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3246 lookups for user supplementary groups into a cache.
3250 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3251 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3252 for liblustre O_EXCL is enforced by MDS.
3256 Description: readdir never returns NULL in liblustre.
3257 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3258 dentry in struct dirent.
3262 Frequency : liblustre only, depends on application IO pattern
3263 Description: liblustre clients evicted if not contacting servers
3264 Details : Don't put liblustre clients into the ping_evictor list, so
3265 they will not be evicted by the pinger ever.
3267 Severity : enhancement
3269 Description: Add ability to evict clients by NID from MDS.
3270 Details : By echoing "nid:$NID" string into
3271 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3272 $NID would be instantly evicted from this MDS and from all active
3273 OSTs connected to it.
3277 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3278 Details : lookup_it_finish() used to query file size from OSTs that was not
3283 Description: service threads change working directory to that of init
3284 Details : Starting lustre service threads may pin the working directory
3285 of the parent thread, making that filesystem busy. Threads
3286 now change to the working directory of init to avoid this.
3290 Frequency : during shutdown only
3291 Description: shutdown with a failed MDS or OST can cause unmount to hang
3292 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3297 Frequency : 2.6 only
3298 Description: chmod/chown may include an extra supplementary group
3299 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3300 supplementary group and if none is specified this is used.
3303 Bugzilla : 5479 (6816)
3305 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3306 Details : If lookup is in progress on a file that is unlinked we might try
3307 to revalidate the inode and fail in revalidate after lookup is
3308 complete and ll_file_open() enqueues the open again but
3309 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3312 Frequency : always, if lconf --abort_recovery used
3314 Description: lconf --abort_recovery fails with 'Operation not supported'
3315 Details : lconf was attempting to abort recovery on the MDT device and not
3318 ------------------------------------------------------------------------------
3320 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3325 Frequency : rare (only unsupported configurations with a node running as an
3327 Bugzilla : 6514, 5137
3328 Description: Mounting a Lustre file system on a node running as an OST could
3330 Details : OSTs now preallocates memory needed to write out data at
3331 startup, instead of when needed, to avoid having to
3332 allocate memory in possibly low memory situations.
3333 Specifically, if the file system is mounted on on OST,
3334 memory pressure could force it to try to write out data,
3335 which it needed to allocate memory to do. Due to the low
3336 memory, it would be unable to do so and the node would
3337 become unresponsive.
3339 Severity : enhancement
3341 Description: Addition of lconf --service command line option
3342 Details : lconf now accepts a '--service <arg>' option, which is
3343 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3345 Severity : enhancement
3347 Description: Failover mode is now the default for OSTs.
3348 Details : By default, OSTs will now run in failover mode. To return to
3349 the old behaviour, add '--failout' to the lmc line for OSTs.
3351 Severity : enhancement
3353 Description: Health checks are now provided for MDS and OSTs
3354 Details : Additional detailed health check information on MSD and OSTs
3355 is now provided through the procfs health_check value.
3358 Frequency : occasional, depends on IO load
3360 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3361 after numerous create/delete cycles
3362 Details : The ext3 inode allocation policy would not allocate new inodes
3363 very well on the OSTs because there are no new directories
3364 being created. Instead we look for groups with free space if
3365 the parent directories are nearly full.
3370 Description: Network or server problems during mount may cause partially
3371 mounted clients instead of returning an error.
3372 Details : The config llog parsing code may overwrite the error return
3373 code during mount error handling, returning success instead
3379 Description: MDS can fail to allocate large reply buffers
3380 Details : After long uptimes the MDS can fail to allocate large reply
3381 buffers (e.g. zconf client mount config records) due to memory
3382 fragmentation or consumption by the buffer cache. Preallocate
3383 some large reply buffers so that these replies can be sent even
3384 under memory pressure.
3388 Frequency : rare (liblustre)
3389 Description: fsx running with liblustre complained that using truncate() to
3390 extend the file doesn't work. This patch corrects that issue.
3391 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3392 ATTR_SIZE and lsm use in llu_setattr_raw.
3396 Frequency : rare, only 2.6 kernels
3397 Description: Unusual file access patterns on the MDS may result in inode
3398 data being lost in very rare circumstances.
3399 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3400 code in the 2.6 kernel caused the inode and/or EA data not to
3401 be read from disk, causing single-file corruption.
3405 Frequency : rare, only 2.6 filesystems using extents
3406 Description: Heavy concurrent write and delete load may cause data corruption.
3407 Details : It was possible under high-load situations to have an extent
3408 metadata block in the block device cache from a just-unlinked
3409 file overwrite a newly-allocated data block. We now unmap any
3410 metadata buffers that alias just-allocated data blocks.
3414 Frequency : filesystems with default stripe_count larger than 77
3415 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3416 Details : lconf specifies an inode size of 4096 bytes when the default
3417 stripe_count is larger than 77. This conflicts with the default
3418 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3419 this case to avoid pinning too much memory for large EAs.
3421 ------------------------------------------------------------------------------
3423 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3428 Frequency : rare (extremely heavy IO load with hundreds of clients)
3430 Description: Client is evicted, gets IO error writing to file
3431 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3432 caused clients to be evicted for AST timeouts. The fixes in
3433 bug 5192 mean we no longer need to have such short AST timeouts
3434 so ldlm_timeout has been increased.
3437 Frequency : occasional during --force or --failover shutdown under load
3438 Bugzilla : 5949, 4834
3439 Description: Server oops/LBUG if stopped with --force or --failover under load
3440 Details : a collection of import/export refcount and cleanup ordering
3441 issues fixed for safer force cleanup
3444 Frequency : only filesystems larger than 120 OSTs
3445 Bugzilla : 5990, 6223
3446 Description: lfs getstripe would oops on a very large filesystem
3447 Details : lov_getconfig used kfree on vmalloc'd memory
3450 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3451 Bugzilla : 6242, 6243
3452 Description: reading from files that had been truncated to a non-zero size
3453 but never opened returned no data
3454 Details : ll_file_read() reads zeros from no-object files to EOF
3459 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3460 Details : There's little harm in aborting MDS/OSS recovery and letting it
3461 try again, so I removed the LASSERT and return an error instead.
3463 Severity : enhancement
3465 Description: New debugging infrastructure for tracking down data corruption
3466 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3467 (b) cover more of the client-side code path, and (c) try to narrow
3468 down where problems occurred
3472 Bugzilla : 3819, 4364, 4397, 6313
3473 Description: Racing close and eviction MDS could cause assertion in mds_close
3474 Details : It was possible to get multiple mfd references during close and
3475 client eviction, leading to one thread referencing a freed mfd.
3477 Severity: : enhancement
3478 Bugzilla : 3262, 6359
3479 Description: Attempts to reconnect to servers are now more aggressive.
3480 Details : This builds on the enhanced upcall-less recovery that was added
3481 in 1.4.2. When trying to reconnect to servers, clients will
3482 now try each server in the failover group every 10 seconds. By
3483 default, clients would previously try one server every 25 seconds.
3488 Description: After recovery, certain operations trigger a failed
3489 assertion on a client.
3490 Details : Failing over an mds, using lconf -d --failover, while a
3491 client was doing a readdir() call would cause the client to
3492 LBUG after recovery completed and the readdir() was resent.
3494 Severity : enhancement
3496 Description: Default groups are now added by lconf
3497 Details : You can now run lconf --group <servicename> without having to
3498 manually add groups with lmc.
3501 Frequency : occasional
3503 Description: Nodes with an elan id of 0 trigger a failed assertion
3506 Frequency : always when accessing e.g. tty/console device nodes
3508 Description: tty and some other devices nodes cannot be used on lustre
3509 Details : file's private_data field is used by device data and lustre
3510 values in there got lost. New field was added to struct file to
3511 store fs-specific private data.
3514 Frequency : when exporting Lustre via NFS
3516 Description: NFSD failed occasionally when looking up a path component
3517 Details : NFSD is looking up ".." which was broken in ext3 directories
3518 that had grown large enough to become hashed.
3521 Frequency : Clusters with multiple interfaces not on the same subnet
3523 Description: Nodes will repeatedly try to reconnect to an interface which it
3524 cannot reach and report an error to the log.
3525 Details : Extra peer list entries will be created by lconf with some peers
3526 unreachable. lconf now validates the peer before adding it.
3529 Frequency : Only if a default stripe is set on the filesystem root.
3531 Description: Setting a default stripe on the filesystem root prevented the
3532 filesystem from being remounted.
3533 Details : The client was sending extra request flags in the root getattr
3534 request and did not allocate a reply buffer for the dir EA.
3537 Frequency : occasional, higher if lots of files are accessed by one client
3538 Bugzilla : 6159, 6097
3539 Description: Client trips assertion regarding lsm mismatch/magic
3540 Details : While revalidating inodes the VFS looks up inodes with ifind()
3541 and in rare cases can find an inode that is being freed.
3542 The ll_test_inode() code will free the lsm during ifind()
3543 when it finds an existing inode and then the VFS later attaches
3544 this free lsm to a new inode.
3548 Bugzilla : 6422, 7030
3549 Description: MDS deadlock between mkdir and client eviction
3550 Details : Creating a new file via mkdir or mknod (starting a transaction
3551 and getting the ns lock) can deadlock with client eviction
3552 (gets ns lock and trying to finish a synchronous transaction).
3555 Frequency : occasional
3556 Description: While starting a server, the fsfilt_ext3 module could not be
3558 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3559 kernels. Previously, lconf would still use the ext3 name
3560 when trying to load modules. Now, it will correctly use
3561 ext3 on 2.4 and ldiskfs on 2.6.
3563 Severity : enhancement
3564 Description: The default stripe count has been changed to 1
3565 Details : The interpretation of the default stripe count (0, to lfs
3566 or lmc) has been changed to mean striping across a single
3567 OST, rather than all available. For general usage we have
3568 found a stripe count of 1 or 2 works best.
3570 Severity : enhancement
3571 Description: Add support for compiling against Cray portals.
3572 Details : Conditional compiling for some areas that are different
3576 Frequency : occasional
3577 Bugzilla : 6409, 6834
3578 Description: Creating files with an explicit stripe count may lead to
3579 a failed assertion on the MDS
3580 Details : If some OSTs are full or unavailable, creating files may
3581 trigger a failed assertion on the MDS. Now, Lustre will
3582 try to use other servers or return an error to the
3586 Frequency : occasional
3588 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3589 caused lock timeout during readahead (since 1.4.2).
3590 Details : Processes doing readahead might match a lock that hasn't been
3591 granted yet if there are overlapping and conflicting lock
3592 requests. The readahead process waits on ungranted lock
3593 (original lock is CBPENDING), while OST waits for that process
3594 to cancel CBPENDING read lock and eventually evicts client.
3596 Severity : enhancement
3598 Description: Initial enabling of flock support for clients
3599 Details : Implements fcntl advisory locking and file status functions.
3600 This feature is provided as an optional mount flag (default
3601 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3602 locking are implemented yet, and those that are are not guaranteed
3603 to be completely correct in production environments.
3604 mount -t lustre -o [flock|noflock] ...
3607 Frequency : occasional
3609 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3610 assertion in the ext3 JBD (journaling) layer.
3611 Details : The b_committed_data struct is protected by the big kernel lock
3612 in 2.4 kernels, serializing journal_commit_transaction() and
3613 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3614 access to this struct. In 2.6 kernels there is finer grained
3615 locking to improve SMP performance of the JBD layer.
3619 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3624 Description: merge and cleanup kernel patches.
3627 -----------------------------------------------------------------------------
3629 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3631 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3632 versions of Lustre. Please update all servers and clients to
3633 version 1.4.2 or later at the same time. You must also run
3634 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3635 to update the configuration logs.
3637 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3638 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3639 - swab llog records of type '0' so we get proper header size/idx (5861)
3640 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3641 - fix rename of one directory over another leaking an inode (5953)
3642 - avoid SetPageDirty on 2.6 (5981)
3643 - don't re-add just-being-destroyed locks to the waiting list (5653)
3644 - when creating new directories, inherit the parent's custom
3645 striping settings if present parent (3048)
3646 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3647 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3648 - don't hold client locks on temporary worklist from l_lru (5666)
3649 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3650 - automatically evict dead clients (5921)
3651 - Update file size properly in create+truncate+fstat case (6196)
3652 - Do not unhash mountpoint dentries, do not allow removal of
3654 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3655 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3656 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3657 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3658 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3659 - Correctly handle reads of files with no objects (6243)
3660 - lctl recover will also mark a device active if deactivate used (5933)
3662 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3663 - allow --write-conf on an MDS with different nettype than client (5619)
3664 - don't write config llogs to MDS for mounts not from that MDS (5617)
3665 - lconf should create multiple TCP connections from a client (5201)
3666 - init scripts are now turned off by default; run chkconfig --on
3667 lustre and chkconfig --on lustrefs to use them
3668 - upcalls are no longer needed for clients to recover to failover
3670 - add --abort-recovery option to lconf to abort recovery on device
3672 - add support for an arbitrary number of OSTs (3026)
3673 - Quota support protocol changes.
3674 - forward compatibility changes to wire structs (6007)
3675 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3676 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3677 - improved "lustre status" script
3678 - initialize blocksize for non-regular files (6062)
3679 - added --disable-server and --disable-client configure options (5782)
3680 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3681 - Vanilla 2.4.29 support
3682 - increase maximum number of obd devices to 520 (6242)
3683 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3684 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3685 SLES 9 2.6.5 kernels
3686 - limit stripes per file to 160 (the maximum EA size) (6093)
3688 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3691 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3692 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3693 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3694 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3695 - small ext3 extents cleanups and fixes (5733)
3696 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3697 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3698 - avoid needless client->OST connect, fix handle mismatch (5317)
3699 - fix DLM error path that led to out-of-sync client, long delays (5779)
3700 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3701 - fix several locking issues related to i_size (5492,5624,5654,5672)
3702 - don't move pending lock onto export if it is already evicted (5683)
3703 - fix kernel oops when creating .foo in unlinked directory (5548)
3704 - fix deadlock in obdfilter statistics vs. object create (5811)
3705 - use time_{before,after} to avoid timer jiffies wrap (5882)
3706 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3707 - Do not leak request if server was not able to process it (5154)
3708 - If mds_open unable to find parent dir, make that negative lookup(5154)
3709 - don't create new directories with extent-mapping (5909, 5936)
3711 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3712 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3713 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3714 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3715 - support for RHEL4 kernel on i686 (5773)
3716 - provide error messages when incompatible logs are encountered (5898)
3718 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3719 * version 1.4.0.10 (1.4.1 release candidate 1)
3721 - don't keep a lock reference when lock is not granted (4238)
3722 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3723 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3724 - drop import inflight refcount on signal_completed_replay error (5255)
3725 - unlock page after async write error during send (3677)
3726 - handle missing objects in filter_preprw_read properly (5265)
3727 - no transno return for symlink open, don't save no-trasno open (3440)
3728 - don't try to complete elan receive that already failed (4012)
3729 - free RPC server reply state on error (5406)
3730 - clean up thread from ptlrpc_start_thread() on error (5160)
3731 - readahead could read extra page into cache that wasn't ejected (5388)
3732 - prevent races in class_attach/setup/cleanup/detach (5260)
3733 - don't dereference de->d_inode after l_dput of de (5458)
3734 - use "int" for stripe value returned from lock_to_stripe (5544)
3735 - mballoc allocation and error-checking fixes in 2.6 (5504)
3736 - block device patches to fix I/O request sizes in 2.6 (5482)
3737 - look up hostnames for IB nals (5602)
3738 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3739 - don't start multiple acceptors for the same port (5277)
3740 - fix incorrect LASSERT in mds_getattr_name (5635)
3741 - export a proc file for general "ping" checking (5628)
3742 - fix "lfs check" to not block when the MDS is down (5628)
3744 - service request history (4965)
3745 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3746 - create an "evict_client" /proc entry on OSTs, like the MDS has
3747 - fix mount usage message, return errors per mount(8) (5168)
3748 - change grep [] to grep "[]" in tests so they work in more UMLs
3749 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3750 - remove ancient LOV_MAGIC_V0 EA support (5047)
3751 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3752 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3753 - allow readahead allocations to fail when low on memory (5383)
3754 - mmap locking landed again, after considerable improvement (2828)
3755 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3757 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3760 - send OST transaction number in read/write reply to free req (4966)
3761 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3762 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3763 - the watchdog thread now runs as interruptible (5246)
3764 - flock/lockf fixes (but it's still disabled, pending 5135)
3765 - don't use EXT3 constants in llite code (5094)
3766 - memory shortage at startup could cause assertion (5176)
3768 - reorganization of lov code
3769 - single portals codebase
3771 - add extents/mballoc support (5025)
3772 - direct I/O reads in the obdfilter (4048)
3773 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3775 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3778 - send OST transaction number in read/write reply to free req (4966)
3779 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3780 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3781 - the watchdog thread now runs as interruptible (5246)
3782 - handle missing objects in filter_preprw_read properly (5265)
3783 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3784 - drop import inflight refcount on signal_completed_replay error (5255)
3785 - unlock page after async write error during send (3677)
3786 - return original error code on reconstructed replies (3761)
3787 - no transno return for symlink open, don't save no-trasno open (3440)
3789 - add pid to ldlm debugging output (4922)
3790 - bump the watchdog timeouts -- we can't handle 30sec yet
3791 - extra debugging for orphan dentry/inode bug (5259)
3793 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3796 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3797 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3798 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3799 - fix NULL dereference in /proc/sys/portals/routes (4827)
3800 - allow failed mdc_close() operations to be interrupted (4561)
3801 - stop precreate on OST before MDS would time out on it (4778)
3802 - don't send partial-page writes before EOF from client (4410)
3803 - discard client grant for sub-page writes on large-page clients (4520)
3804 - don't free dentries not owned by NFS code, check generation (4806)
3805 - fix lsm leak if mds_create_objects() fails (4801)
3806 - limit debug_daemon file size, always print CERROR messages (4789)
3807 - use transno after validating reply (3892)
3808 - process timed out requests if import state changes (3754)
3809 - update mtime on OST during writes, return in glimpse (4829)
3810 - add mkfsoptions to LDAP (4679)
3811 - use ->max_readahead method instead of zapping global ra (5039)
3812 - don't interrupt __l_wait_event() during strace
3814 - add software watchdogs to catch hung threads quickly (4941)
3815 - make lustrefs init script start after nfs is mounted
3816 - fix CWARN/ERROR duplication (4930)
3817 - return async write errors to application if possible (2248)
3818 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3819 - print NAL number in %x format (4645)
3820 - update barely-supported suse-2.4.21-171 series (4842)
3821 - support for sles 9 %post scripts
3822 - support for building 2.6 kernel-source packages
3823 - support for sles km_* packages
3825 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3828 - ignore -ENOENT errors in osc_destroy (3639)
3829 - notify osc create thread that OSC is being cleaned up (4600)
3830 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3831 - reconstruct ost_handle() like mds_handle() (4657)
3832 - create a new thread to do import eviction to avoid deadlock (3969)
3833 - let lconf resolve symlinked-to devices (4629)
3834 - don't unlink "objects" from directory with default EA (4554)
3835 - hold socknal file ref over connect in case target is down (4394)
3836 - allow more than 32000 subdirectories in a single directory (3244)
3837 - fix blocks count for O_DIRECT writes (3751)
3838 - OST returns ENOSPC from object create when no space left (4539)
3839 - don't send truncate RPC if file size isn't changing (4410)
3840 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3841 - bind to privileged port in socknal and tcpnal (3689)
3843 - rate limit CERROR/CWARN console message to avoid overload (4519)
3844 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3845 - basic mmap support (3918)
3846 - kernel patch series update from b1_4 (4711)
3848 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3851 - avoid crash during MDS cleanup with OST shut down (2775)
3852 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3853 - don't use bad inodes on the MDS (3744)
3854 - dynamic object preallocation to improve recovery speed (4236)
3855 - don't hold spinlock over lock dumping or change debug flags (4401)
3856 - don't zero obd_dev when it is force cleaned (3651)
3857 - print grants to console if they go negative (4431)
3858 - "lctl deactivate" will stop automatic recovery attempts (3406)
3859 - look for existing locks in ldlm_handle_enqueue() (3764)
3860 - don't resolve lock handle twice in recovery avoiding race (4401)
3861 - revalidate should check working dir is a directory (4134)
3863 - don't always mark "slow" obdfilter messages as errors (4418)
3865 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3868 - don't close LustreDB during write_conf until it is done (3860)
3869 - fix typo in lconf for_each_profile (3821)
3870 - allow dumping logs from multiple threads at one time (3820)
3871 - don't allow multiple threads in OSC recovery (3812)
3872 - fix debug_size parameters (3864)
3873 - fix mds_postrecov to initialize import for llog ctxt (3121)
3874 - replace config semaphore with spinlock (3306)
3875 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3876 - don't allow enqueue to complete on a destroyed export (3822)
3877 - down write_lock before checking llog header bitmap (3825)
3878 - recover from lock replay timeout (3764)
3879 - up llog sem before sending rpc (3652)
3880 - reduce ns lock hold times when setting kms (3267)
3881 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3882 - fix NULL deref and obd_dev leak on setup error (3312)
3883 - replace some LBUG about llog ops with error handling (3841)
3884 - don't match INVALID dentries from d_lookup and spin (3784)
3885 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3886 - fix invalid assertion in ptlrpc_set_wait (3880)
3888 - add libwrap support for the TCP acceptor (3996)
3889 - add /proc/sys/portals/routes for non-root route listing (3994)
3890 - allow setting MDS UUID in .xml (2580)
3891 - print the stack of a process that LBUGs (4228)
3893 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3896 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3897 - make sure to unset replay flag from failed open requests (3440)
3898 - if default stripe count is 0, use OST count for inode size (3636)
3899 - update parent mtime/ctime on client for create/unlink (2611)
3900 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3901 - free recovery state on server during a forced cleanup (3571)
3902 - unregister_reply for resent reqs (3063)
3903 - loop back devices mounting and status check on 2.6 (3563)
3904 - fix resource-creation race that can provoke i_size == 0 (3513)
3905 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3906 - more debugging for page-accounting assertion (3746)
3907 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3908 - avoid deadlock after precreation failure (3758)
3909 - fix race and lock order deadlock in orphan handling (3450, 3750)
3910 - add validity checks when grabbing inodes from l_ast_data (3599)
3912 - add /proc/.../recovery_status to obdfilter (3428)
3913 - lightweight CDEBUG infrastructure, debug daemon (3668)
3914 - change default OSC RPC parameters to be better on small clusters
3915 - turn off OST read cache for files smaller than 32MB
3916 - install man pages and include them in rpms (3100)
3917 - add new init script for (un)mounting lustre filesystems (2593)
3918 - run chkconfig in %post for init scripts (3701)
3919 - drop scimac NAL (unmaintained)
3921 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3924 - clean kiobufs before and after use (3485)
3925 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3926 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3927 - change init script to fail more clearly if not run as root (1528)
3928 - allow clients to reconnect during replay (1742)
3929 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3930 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3931 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3932 - fix cond_resched (3554)
3933 - only evict unfinished clients after recovery (3515)
3934 - allow bulk resend, prevent data loss (3570)
3935 - dynamic ptlrpc request buffer allocation (2102)
3936 - don't allow unlinking open directory if it isn't empty (2904)
3937 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3938 - remove extraneous obd dereference causing LASSERT failure (3334)
3939 - don't use get_cycles() when creating temp. files on the mds (3156)
3940 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3941 - handle EEXIST for set-stripe, set proper directory name (3336)
3943 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3944 - fix ksocknal_fmb_callback() error messages (2918)
3946 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3949 - don't copy lvb into (possibly NULL) reply on error (2983)
3950 - don't deref dentry after dput, don't free lvb on error (2922)
3951 - use the kms to determine writeback rpc length (2947)
3952 - increment oti_logcookies when osc is inactive (2948)
3953 - update client's i_blocks count via lvb messages (2543)
3954 - handle intent open/close of special files properly (1557)
3955 - mount MDS with errors=remount-ro, like obdfilter (2009)
3956 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3957 - don't use cancelling-locks' kms values (2947)
3958 - use highest lock extent for kms, not last one (2925)
3959 - don't dereference ERR_PTR() dentry in error handling path (3107)
3960 - fix thread race in portals_debug_dumplog() (3122)
3961 - create lprocfs device entries at setup instead of at attach (1519)
3962 - common AST error handler, don't evict client on completion race (3145)
3963 - zero nameidata in detach_mnt in 2.6 (3118)
3964 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3965 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3966 - initialize RPC timeout timer earlier for 2.6 (3219)
3967 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3968 - print nal/nid for unknown nid (3258)
3969 - additional checks for oscc recovery before doing precreate (3284)
3970 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3971 - don't crash in mdc_close for bad permissions on open (3285)
3972 - zero i_rdev for non-device files (3147)
3973 - clear page->private before handing to FS, better assertion (3119)
3974 - tune the read pipeline (3236)
3975 - fix incorrect decref of invalidated dentry (2350)
3976 - provide read-ahead stats and refine rpc in flight stats (3328)
3977 - don't hold journal transaction open across create RPC (3313)
3978 - update atime on MDS at close time (3265)
3979 - close LDAP connection when recovering to avoid server load (3315)
3980 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3981 - don't leak open file on MDS after open resend (3325)
3982 - serialize filter_precreate and filter_destroy_precreated (3329)
3983 - loop device shouldn't call sync_dev() for nul device (3092)
3984 - clear page cache after eviction (2766)
3985 - resynchronize MDS->OST in background (2824)
3986 - refuse to mount the same filesystem twice on same mountpoint (3394)
3987 - allow llmount to create routes for mounting behind routers (3320)
3988 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3989 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3990 - fix rare problem with rename on htree directories (3417)
3992 - allow default OST striping configuration per directory (1414)
3993 - fix compilation for qswnal for 2.6 kernels (3125)
3994 - increase maximum number of MDS request buffers for large systems
3995 - change liblustreapi to be useful for external progs like lfsck (3098)
3996 - increase local configuration timeout for slow disks (3353)
3997 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3999 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4002 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4003 - don't overwrite extent policy data in reply if lock was blocked (2901)
4004 - drop filter export grants atomically with removal from device (2663)
4005 - del obd_self_export from work_list in class_disconnect_exports (2908)
4006 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4007 - swab reply message in mdc_close, other PPC fixes (2464)
4008 - fix destroying of named logs (2325)
4009 - overwrite old logs when running lconf --write_conf (2264)
4010 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4011 - fix race in target_handle_connect (2898)
4012 - mds_reint_create() should take same inode create lock (2926)
4013 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4014 - don't close files for self_export to avoid uninitialized obd (2936)
4015 - allow MDS with the same name as client node (2939)
4016 - hold dentry reference for closed log files for unlink (2325)
4017 - reserve space for all logs during transactions (2059)
4018 - don't evict page beyond end of stripe extent (2925)
4019 - don't oops on a deleted current working directory (2399)
4020 - handle hard links to targets without a parent properly (2517)
4021 - don't dereference NULL lock when racing during eviction (2867)
4022 - don't grow lock extents when lots of conflicting locks (2919)
4024 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4027 - account for cache space usage on clients to avoid data loss (974)
4028 - lfsck support in lustre kernel code (2349)
4029 - reduce journal credits needed for BRW writes (2370)
4030 - orphan handling to avoid losing space on client/server crashes
4031 - ptlrpcd can be blocked, stopping ALL progress (2477)
4032 - use lock value blocks to assist in proper KMS, faster stat (1021)
4033 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4034 - recovery for initial connections (2355)
4035 - fixes for mds_cleanup_orphans (1934)
4036 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4037 - block all file creations until orphan recovery completes (1901)
4038 - client remove rq_connection from request struct (2423)
4039 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4040 - recovery timer race (2670)
4041 - mdc_close recovey bug (2532)
4042 - ptlrpc cleanup bug (2710)
4043 - mds timeout on local locks (2588)
4044 - namespace lock held during RPCs (2431)
4045 - handle interrupted sync write properly (2503)
4046 - don't try to handle a message that hasn't been replied to (2699)
4047 - client assert failure during cleanup after abort recovery (2701)
4048 - leak mdc device after failed mount (2712)
4049 - ptlrpc_check_set allows timedout requests to complete (2714)
4050 - wait for inflight reqs when ptlrpcd finishes (2710)
4051 - make sure unregistered services are removed from the srv_list
4052 - reset bulk XID's when resending them (caught by 1138 test)
4053 - unregister_bulk after timeout
4054 - fix lconf error (2694)
4055 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4056 - readahead locks pages, leaves pending causing memory pressure (2673)
4057 - increase OST request buffers to 4096 on large machines (2729)
4058 - fix up permission of existing directories in simple_mkdir (2661)
4059 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4060 - don't assign transno to errored transactions (2742)
4061 - don't delete objects on OST if given a bogus objid from MDS (2751)
4062 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4063 - if rq_no_resend, then timeout request after recovery (2432)
4064 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4065 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4066 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4067 - MDS recovery shouldn't delete objects that it has given out (2730)
4068 - if enqueue arrives after completion, don't clobber LVB (2819)
4069 - don't unlock pages twice when trigger_group_io returns error (2814)
4070 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4071 - don't write pages to disk if there was an error (1450)
4072 - don't ping imports that have recovery disabled (2676)
4073 - take buffered bytes into account when balancing socknal conn (2817)
4074 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4075 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4076 - remove little-endian swabbing of llog records (1987)
4077 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4078 - retry reposting request buffers if they fail (1191)
4079 - grow extent at grant time to avoid granting a revoked lock (2809)
4080 - lock revoke doesn't evict page if covered by a second lock (2765)
4081 - disable VM readahead to avoid reading outside lock extents (2805)
4083 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4084 - updated kernel patches for hp-2.4.20 kernel (2681)
4086 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4089 - fix truncated write corruption (2366)
4090 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4092 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4093 - ensure len(uuid) < 37 in lmc (1171)
4094 - fix ia64 OOPS in llog_test (2255)
4095 - zero end of page at obdfilter for partial page writes (2648)
4096 - don't leave stale dentries around after renames (bug 2428)
4097 - fix timeouts when evicting a client with a single lock held (2642)
4098 - set deadline for the initial HELLO message to drain (2634)
4099 - print out dotted-quad IP addresses in the socknal (2302)
4101 - additional debugging for MDS client eviction problem (2443)
4102 - fix mkfsoptions support for osts (2603, 2604)
4104 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4107 - add series for the vanilla 2.6.0 kernel
4108 - add series for the vanilla 2.4.24 kernel
4109 - add series for a cray x86/64 UL kernel drop
4110 - fix xattr patches for the vanilla 2.4.19 series
4112 - generate true UUIDs in lmc (1171)
4113 - have portals stack dumping break in UML (2466)
4114 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4115 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4116 - wake callback waiting threads on client eviction (2460)
4117 - Add --ptldebug and --subsystem to lmc (1719)
4118 - update assertion to allow safe interrupt allocation
4119 - set rq_no_resend for cancel requests (2432)
4120 - recalculate ptlrpcd timeout after resend (2494)
4121 - call vfs_rmdir when removing pending directories (2368)
4122 - fix renaming a file to itself (2429)
4123 - lmc creates a default one-stripe lov (2454)
4124 - expand procfs space to handle large clusters (2326)
4125 - increase UML stack to avoid overflow
4126 - update lconf's list of debug and subsystem masks
4127 - fix lfs find --obd (2510)
4128 - /proc tunable for disabling filter read caching (2591)
4129 - stop rpm packages from altering slapd.conf (2301)
4130 - disable nagle in the socknal under 0conf (2578)
4131 - choose mds inode size based on stripe count (2572)
4132 - fix kernel-source rpm problems (2516)
4134 - add --disable-doc to avoid pdf generation (2421)
4135 - update documentation, tests, type-os, comments
4136 - avoid format warnings on ia64
4137 - remove the TOE NAL
4138 - tiny code cleanups by removing unused fields
4140 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4143 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4144 - remove the most blatant lies from BUILDING file (2371)
4145 - change default debug level to reasonable production setting
4146 - reduce client side cache size to reduce cache flush time
4147 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4148 - make TCP zerocopy and pinger support enabled by default (2476)
4149 - sync writes completed after process exits caused crashes (2319)
4150 - maintain correct mount count on the MDS (2356)
4151 - backout 1557, because 2316 wasn't really fixed
4152 - better file I/O statistics gathering in /proc
4153 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4154 - another kernel patch to fix zero-copy TCP function export
4155 - don't take duplicate lock when processing re-sent getattr (2420)
4156 - lctl uses obd_self_export instead of creating new conn (2353)
4157 - MDS/OST recovery case which requires object creation asserted (2425)
4158 - move lfs from /usr/sbin to /usr/bin in packages
4159 - fix race between mds_client_add and mds_client_free (2417)
4160 - use kmalloc instead of slabs in portals (2430)
4161 - don't create duplicate records when a failover MDS is present (2442)
4162 - remove unnecessary mount age check (2332)
4163 - don't remove directory inodes from locks prematurely (2451)
4164 - don't break if MDS service name is the same as hostname (2103)
4165 - fix races in client write RPC generation when cache full (2482)
4167 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4170 - remove now-unused request->rq_obd (278)
4171 - if an allocation fails, print out how much memory we've used (1933)
4172 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4173 - add the "configurable stack size" patch to most series files (1256)
4174 - ability to write large log records, for 100+ OST configs (2306)
4175 - fix NULL deref when filter_prep fails (2314)
4176 - fix operator precedence error in filter_sync
4177 - dynamic allocation of socknal TX descriptors (2315)
4178 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4179 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4180 - ensure that CWARN messages in Portals always get to the syslog
4181 - __init/__exit are not for prototype decls (ldlm_init/exit)
4182 - x86-64 compile warning fixes
4183 - fix gateway LMC keyword conflict (2318)
4184 - fix MDS lock inversions in getattr/reint paths (1844)
4185 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4186 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4187 - DLM race condition prevented some lock evictions (2328)
4188 - ENOMEM detection and retry on socknal sends (2230)
4189 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4190 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4191 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4192 - better error messages when a client is rejected during recovery (1505)
4193 - avoid cancelling locks which were never granted, after failure (2330)
4194 - fix i_sem/journal inversion in mds_client_add (2333)
4195 - fix truncate/getattr lock cycle deadlock (2334)
4196 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4197 - fix two rare exit paths which could leak an l_lock() ref (2321)
4198 - fencepost error in MDS/OST orphan recovery (2226)
4199 - make log record alignment 8 bytes (1988)
4200 - lstripe now fails when requested offset > ost_count (2237)
4201 - ensure that all kernel series have a complete list.h (1607)
4202 - fix crashes in special-file operations (2316)
4203 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4205 - allow configurable automake binary, for testing new versions
4206 - small update to the lfs documentation
4208 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4210 * fix negative export reference count in fsfilt_sync (2312)
4212 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4213 * release candidate 0.9.1
4215 - orphans are moved into the PENDING directory for possible recovery
4216 - replayed opens now open by fid for orphan/rename safety (1042)
4217 - last close of an orphan inode generates a transno (683)
4218 - chdir() and mount() now pin the directory entry (1020)
4219 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4220 - discard very old requests without processing them (1502)
4221 - remove some common, well-understood CERRORs (1505)
4222 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4223 - clear "grant" flags in OST replies until OST grant code lands (1644)
4224 - fix read performance by not clobbering i_blksize on client (1598)
4225 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4226 - make LRU size a /proc tunable, clears locks when reduced (707)
4227 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4228 - server locks take references on exports now (1558)
4229 - build fixes for 2.4.20-rh trees (1663)
4230 - return an error from lov_create if all OSCs are inactive (1751)
4231 - fix import levels when a reconnect happens without a timeout (1597)
4232 - exit early from mds_open if we get a lookup error (1749)
4233 - partial page read at EOF wouldn't wait for disk before sending (1642)
4234 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4235 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4236 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4237 - write/truncate lock inversion (1639)
4238 - Don't auto-load obdclass, portals modules during cleanup (1495)
4239 - fix timestamps from jumping to "now" (1763)
4240 - extra journal assertions (1648)
4241 - add an extra multiunlink test (1771)
4242 - fix read_record/write_record API (1776)
4243 - fix leak of offset_extent, possible incorrect i_size later (1772)
4244 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4245 - Debugging for the kqswnal_get_idle_tx problems (1820)
4246 - Allow recovery to be attempted multiple times (1536)
4247 - Write out MDS last_rcvd file after it is first created (1600)
4248 - Fix tx_descriptor leak in failed transmit situations (1827)
4249 - ext3 journaling fixes for assertion failure after IO error (1871)
4250 - class_export_put() on freed export after completion AST error (1896)
4251 - Fix revalidate looping in VFS (1322)
4252 - Don't access a freed export during MDS_REINT timeout (1521)
4253 - Add open-unlink recovery support on the MDS (1673,1764)
4254 - Return an error if no MDS data was read from last_rcvd (1946)
4255 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4256 - Don't setup a disk device that doesn't match exported UUID (317)
4257 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4258 - avoid committing NULL handle in force close
4259 - local.sh is now a one-stripe LOV configuration
4260 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4261 - fix ext3 htree duplicate directory entry corruption (1516)
4262 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4263 - update to new LOV EA format (2097)
4264 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4266 2003-06-15 Phil Schwan <phil@clusterfs.com>
4269 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4270 - per-import/export recovery handling (958, 931, 959)
4271 - multiple last-rcvd slots, for serving multiple FSes (949)
4272 - connections are again shared between multiple imp/exports (963, 964)
4273 - "umount -f" would hang if any requests needed to be sent (393, 978)
4274 - avoid pinning large req buffer by copying for queued messages (989)
4275 - add "uuid" to "lctl device" command to help upcalls (991)
4276 - "open" RPCs with transnos would confuse recovery counters (1037)
4277 - do proper endian conversion of all wire messages (288, 340, 891)
4278 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4279 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4280 - fix ptlrpc_connection leak in target_handle_connect (1174)
4281 - fix import refcounting bug in OST and MDS cleanup (1134)
4282 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4283 - fix very unlikely obd_types race condition (501)
4284 - remove osc_open hack for echo_client (1187)
4285 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4286 - a failure in read_inode2 leads to deadlock (1139)
4287 - cancel ack-locks as soon as transaction is committed (1072)
4288 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4289 - make sure to commitrw after any preprw to avoid deadlock (1162)
4290 - failing to execute a file in a lustre FS would lock inode (1203)
4291 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4292 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4293 - obd_uuid2tgt would walk off the end of the list (1255)
4294 - on IA64 the osc would give portals incorrect bulk size (1258)
4295 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4296 - fix lock inversion caused by new llite matching code (1282)
4297 - limit the number of dirty pages on a client to 10MB (1286)
4298 - timed out locks were not being corrected cancelled (1289)
4299 - fix O_DIRECT above 4GB on IA-32 (1292)
4300 * major user-visible changes
4301 - fail out/fail over policy now controlled by the upcall (993)
4303 - add OBD_PING to check server availability and failure (954)
4304 - lustre messages are now sent in sending host order (288, 340, 891)
4305 - add eadatalen to MDS getattr reply (340)
4306 - OST read replies may contain second buffer, with per-page status (593)
4308 2003-03-11 Phil Schwan <phil@clusterfs.com>
4311 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4312 - failed open()s could cause deadlock; fixed (867, 869)
4313 - stop cancelling OST locks when files are closed (481)
4314 - overlapping XID spaces caused network corruption (851, 853)
4315 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4316 - setattr_raw added, to do single-RPC, server-side setattrs
4317 - lmc/lconf syntax change for OST UUIDs
4318 - fix crashy race condition between ptlrpc_free_req and osc_close
4319 - don't use request in mdc_enqueue if we hit a timeout (889)
4320 - don't set the inode i_size for regular files from the MDS (896)
4321 - handle out of order completion AST (842)
4322 - don't LBUG if a lock request times out after receiving AST (913)
4323 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4324 - if a bad lock AST arrives, send an error instead of dropping entirely
4325 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4326 - fix leak in bulk IO when only partially completed (899, 900, 926)
4327 - fix O_DIRECT for ia64 (55)
4328 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4329 - C-z after timeout could hang a process forever; fixed (977)
4331 - client-side I/O cache (678, 924, 929, 941, 970)
4333 - READPAGE and SETATTRs which don't take server-side locks get
4336 2003-02-11 Phil Schwan <phil@clusterfs.com>
4339 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4340 - Fix fsfilt_extN_readpage() to read a full page of directory
4341 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4342 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4343 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4344 - Most intent operations were removed, in favour of a new RPC mode
4345 that does a single RPC to the server and bypasses most of the VFS
4346 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4347 - Aggressively cancel local locks on DLM servers
4348 - mds_reint_unlink sends EA to the client if it's the last nlink.
4349 client uses that EA to unlink OST objects.
4350 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4351 - recursive symlinks were fixed (439)
4352 - fixed NULL deref in DEBUG_REQ
4353 - filter_update_lastobjid no longer calls sync, which annoyed extN
4354 - fixed multi-client small-writes to a single file problem (445)
4355 - fixed mtime updates during file writes (607)
4356 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4357 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4358 - fixed Linux OST size reporting problem (444, 656)
4359 - OST now updates object mtime with writes or setattr (607, 619)
4360 - client verifies file size before zeroing page past EOF (445)
4361 - OST now writes last allocated objid to disk with allocation (108)
4362 - LOV on echo now works (409)
4364 - mds_reint_unlink sends a new buffer, with the EA included. this
4365 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4366 set if a regular file was being unlinked, and it was the last link
4367 - use PtlGet from the target for bulk writes (315)
4368 - OST now updates object mtime with writes or setattr (607, 619)
4369 - LDLM now has a grant-time callback to revalidate locked items, if
4371 - Many MDS operations were reorganized to combat race conditions
4373 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4374 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4376 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4379 - Fully reactivate OST imports after reconnection (512, others)
4380 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4381 - More graceful error handling for truncating on dead OST (515)
4382 - Don't error out unless we're actually accessing dead stripes (474)
4383 - Fix garbage sizes when stripes are missing (410)
4384 - LRU counters were broken, causing constant lock purge (433, 432)
4385 - garbage on read from stripes with failed OSTs (441)
4386 - mark OSCs as active before reconnecting during recovery (438)
4387 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4388 - lfind did not preserve OST order in output (443)
4389 - symlinks cause hung clients, incorrect data (439)
4390 - stop dereferencing request after dropping refcount (457)
4391 - don't LASSERT(spin_is_locked) on non-SMP (455)
4392 - fixes for many rename() bugs
4393 - fstat didn't correctly synchronize attributes (399)
4394 - server must handle lock cancellation during blocking AST prep (487)
4395 - bulk descriptors were free()d too soon (511)
4396 - fix paths in lconf, which would load incorrect modules (451, 507)
4397 - fix confusing lconf 'host not found' error message (386)
4398 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4399 - fix race condition in mdc_blocking_ast() for inode access (526)
4400 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4401 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4402 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4403 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4404 - fix lstripe in several ways (kernel side) (527)
4405 - fix request leak in ldlm_cli_enqueue (262)
4406 - incorrect OSC was marked inactive after OST failure
4407 - call mds_fs_cleanup before unmounting filesystem (524)
4408 - fix races between taking ns_lock and ldlm_lock_change_resource
4409 - fix races updating LOV export open file list
4410 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4411 - fix recovery NULL deref in ldlm_cli_cancel_unused
4412 - fix some DLM races by using new hash table for lock handles (419)
4413 - permit the client to specify desired inodes, at replay
4414 - duplicate requests when we queue them for replay reintegration
4415 - fix last_rcvd offset calculation
4416 - sync after each recovered transaction, so we always make progress
4417 - never, not always, ERESTART requests without transnos
4418 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4420 - skip replay if the MDS says that the client is already connected
4421 - don't check for a recovery-enabled export to match lctl's UUID
4422 - don't INC_USE_COUNT for phantom exports
4423 - don't crash when cleaning up phantom exports (567)
4424 - don't double-finish or set replay data for errored mdc_open requests
4425 - abort requests when they time out, so we don't get old replies
4426 - send/receive replies for AST messages again
4427 - if the client says that it doesn't have the lock, cancel it on the
4429 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4430 mark it as destroyed, it will all work out when decref is called
4431 - fix module use counts (22, 581)
4433 - ASTs now expect a reply (server cancels lock on error reply)
4435 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4438 - fix many simultaneous client startup (392)
4439 - fix dentry->d_it clobbering
4440 - credentials weren't being shipped for readdir/getattr operations
4441 - remove invalid assertions triggered during some concurrent MD
4443 - proper Lustre versions added (336, 389)
4444 - fix memory leak for create error case (398)
4445 - fix LOV locking bug that would get cli/srv out of sync
4446 - fix echo client over LOV (409)
4447 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4448 - fix double-O_EXCL intent crash (424)
4449 - avoid sending multiple lock CANCELs (352)
4451 - MDS can do multi-client recovery (modulo bugs in new code)
4453 - many updates, edits, cleanups
4455 2002-11-18 Phil Schwan <phil@clusterfs.com>
4458 - fix null d_it dereference (346)
4459 - fix full OST/dbench hang (333)
4460 - fix permission problem with file removal (286)
4461 - fix removal of OSCs from LOV when they fail
4462 - fix NULL deref during bulk timeout (214)
4463 - fix problems related to multiple filesystems on one MDS (241)
4464 - fixed serious subtle metadata locking bugs
4465 - free locks on clients when inodes are removed due to memory
4467 - fix inode pointer in lock data (285)
4468 - partial support for multiple MDS on a single host (241)
4469 - data locks weren't cancelled at clear_inode time (290, 311)
4470 - intent locks could lead to unbounded lock growth (205)
4471 - added a maximum lock count, an LRU list, and a flusher
4472 - fix multiple rename (365)
4473 - properly abstracted the echo client
4474 - OSC locked 1 byte too many; fixed
4475 - rewrote brw callback code:
4476 - fixed recovery bugs related to LOVs (306)
4477 - fixed too-many-pages-in-one-write crash (191)
4478 - fixed (again) crash in sync_io_timeout (214)
4479 - probably fixed callback-related race (385)
4481 - Add capability to MDS protocol
4482 - LDLM cancellations and callbacks on different portals
4484 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4487 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4488 - defer open object destruction to close time (601981, 138)
4489 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4490 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4491 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4492 - don't call obd_setattr in ll_file_release for destroyed objects
4493 * protocol change to lustre_msg: move |version| and add |flags|
4494 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4495 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4496 * added replay of create, unlink, link and rename operations during
4497 MDS failover; recovery should be much more robust now
4498 * remove failed OSCs from LOVs (only lov_create uses this so far)
4499 * the lustre-HOWTO was brought (more) up to date (582544)
4501 2002-10-23 Phil Schwan <phil@clusterfs.com>
4504 - in-use dentries weren't being reused properly (617851)
4505 - prevent multiple LDLM setup (599178)
4506 - fix LOV size calculations for truncate (617853)
4507 - fix client handling of MDS intent errors (POSIX)
4508 - fix permission bug in lovstripe.c test (624321)
4509 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4510 - truncate past end of file could corrupt data
4511 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4512 - a race in recovery could return ETIMEDOUT to apps (623947)
4513 - building outside the source directory was fixed
4514 * the lustre-HOWTO was brought (more) up to date (582544)
4515 * major progress was made on recovery functionality
4517 2002-10-10 Phil Schwan <phil@clusterfs.com>
4520 - recovery deadlock fix
4521 - rm -rf causes LBUG fix (617817)
4522 - file open by multiple tasks fix (618962)
4523 - directory permissions bugs (602707 and 620007)
4524 - journal_stop fixed with locking (611313)
4525 - O_APPEND failures resolved (618273, perhaps 614459)
4526 - lconf PATH fix (619770)
4527 - IA64 build fix (621450)
4528 - RPC buffer sizes scale with amount of memory
4530 2002-10-01 Phil Schwan <phil@clusterfs.com>
4533 - locks would be cancelled without throwing away data pages,
4534 resulting in inconsistent data (605627)
4535 - inode attributes were not always being refreshed (605627, 612449)
4536 - lconf now continues to cleanup after lctl reports an error
4537 - MDS now enforces user permissions (602707)
4538 - lprocfs cleanup fixed, but not yet enabled (614157)
4539 - fixed infinite server hang, should a client not respond to an AST
4540 - avoid going into recovery if user calls readlink() with a buffer
4541 that's too small (613941)
4542 - AST RPCs no longer require replies (614867) -- this may be changed
4543 - don't crash server if client sends an IOV that's too big (611336)
4544 - fixed lock conversion deadlock (611892)
4545 - fixed the following of symlinks (614622)
4546 * recovery: the server can remove locks from a client that dies, other
4547 clients can make progress
4548 * more extN patch fixes
4549 * compile-time configurable ptlrpc buffer allocations
4551 - collaborative read cache document
4552 - Lustre Lite Performance CDR document-in-progress
4554 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4557 - fix typo in patch-2.4.18
4559 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4562 - clear ptlrpc request each time in handle_incoming_request()
4563 - unlink of files now destroys the object on the OST
4565 2002-09-19 Peter Braam <braam@clusterfs.com>
4567 * add hard link support
4568 * change obdfile creation method
4569 * kernel patch changed
4571 2002-09-19 Peter Braam <braam@clusterfs.com>
4574 - stack overflow bug in extN fixed
4576 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4578 * documentation updates
4579 - add man pages for config tools
4580 - update tests/README to describe testing with new config tools
4581 - finish metadata API descriptions
4582 * bug fixes and cleanups
4583 - statfs workaround for 16TB limit
4584 - LOV stripe allocation improved, can stripe on subset of OSTs
4585 - LOV file size/IO offset was wrong for files > 4GB in size
4586 - object EA data was being dropped, caused files to be unreadable
4587 - memory overflow with non-LOV OST caused memory corruption
4588 - fixed regression tests to work with new config tools, obdfilter
4589 - fixed bug when directory size became larger than 1 block
4590 - fixed bug (for single client case) when PWD was deleted
4591 - invalidate local directory pages when doing intent-based ops
4592 - avoid LDLM oops when lock callback contained bad data
4594 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4596 * documentation updates
4597 * bug fixes and cleanups
4598 - configuration tools
4601 - 64-bit compile warnings
4602 - 64-bit internal statfs data
4604 * test_brw on persistent OST devices
4606 * lprocfs (disabled)
4608 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4610 * documentation updates
4611 * bug fixes and cleanups
4612 * configuration tools
4614 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4617 * many small fixes to 0_5_4
4618 * io/network handling
4619 * thinkos in MDS operations
4621 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4624 * crucial basic fixes to 0.5.3
4625 * IOR, Iozone work over Elan
4628 2002-08-07 Phil Schwan <phil@clusterfs.com>
4629 * version 0_5_3, our first alpha
4630 * we use the new Portals iovs
4631 * documentation updates
4632 * bug fixes and cleanups
4633 * small changes in the DLM wire protocol
4635 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4636 * version 0_5_1 with some initial stability,
4637 * locking on MD and file I/O.
4638 * documentation updates
4639 * several bug fixes since 0.5.0
4640 * small changes in wire protocol
4642 2002-07-18 Phil Schwan <phil@clusterfs.com>
4644 * delivered as Lustre Light Alpha
4645 * fixed a crash after handling invalid MDS requests
4646 * fixed directory pages for architectures with non-4k pages sizes
4648 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4649 * release version v0_4_4
4650 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4652 * Fixes a number of interruption problems with OST operations.
4653 * Update documentation for portals header changes
4654 * Move all wire protocol structs/defines to lustre_idl.h
4655 * Fixes symlink length bug.
4656 * Add tcpdump to repository.
4658 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4659 * release version v0_4_3
4660 * Fixes statfs for inodes on extN.
4661 * Fixes bug in runtests which would delete /etc/hosts.
4662 * Use 64-bit object IDs wherever possible (not into VFS though)
4663 Remove ost_get_info, which is unused by lustre, and out of date.
4665 2002-07-03 Peter Braam <braam@clusterfs.com>
4666 * release version v0_4_2 Fixes a lookup error (type not passed)
4667 * move forward to head of Portals
4668 * move forward to latest Lustre kernel
4670 2002-06-25 Peter Braam <braam@clusterfs.com>
4671 * release version v0_4_1. Hopefully stable on single node use.