1 tbd Sun Microsystems, Inc.
4 2.6.16.54-0.2.5 (SLES 10),
5 2.6.18-53.1.14.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.40.7-sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
17 Description: add message levels for liblustreapi
21 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
22 Details : Move the 'good_osts' check before the 'total_bavail' check. This
23 will result in an -EAGAIN and in the exit call path we call
24 alloc_rr() which will with increasing aggressiveness attempt to
25 aquire precreated objects on the minimum number of required OSCs.
29 Description: Use old size assignment to avoid deadlock
30 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
31 the scheduling while holding a spinlock. We do not need locking
32 for size in ll_update_inode() because size is only updated from
33 the MDS for directories or files without objects, so there is no
34 other place to do the update, and concurrent access to such inodes
35 are protected by the inode lock.
39 Description: resolve "_IOWR redefined" build error on SLES10
43 Description: dump the memory debugging after all modules are unloaded to
44 suppress false negative in conf_sanity test 39
46 Severity : enhancement
48 Description: build kernel-ib packages for OFED 1.3 in our release cycle
53 Description: fix SLES kernel versioning
54 Details : the kernel version for our SLES 10 kernel did not include a "-"
55 before the "smp" at the end. while this was not a problem in
56 general, it did mean that software trying to use the kernel
57 version to try to detect a vendor specific kernel would fail.
58 this was most evident by the OFED build scripts.
62 Description: Don't update lov_desc members until making sure they are valid
63 Details : When updating lov_desc members via proc fs, need fix their
64 validities before doing the real update.
68 Description: don't put request into delay list while invalidate in flight.
69 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
70 import in flight. this produce timeout for invalidate and sometimes
75 Frequency : on ppc only
76 Description: not convert ost objects for directory because it's not exist.
77 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
81 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
83 Severity : enhancement
85 Description: Update to support 2.6.22.14 vanilla kernel.
89 Frequency : rare, on recovery
90 Description: read procfs can produce deadlock in some situation
91 Details : Holding lprocfs lock which send rpc can produce block for destroy
92 obd objects and this also block reconnect with -EALREADY. This isn't
93 fix all lprocfs bugs - but make it rare.
95 Severity : enhancement
97 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
100 Frequency : frequent on X2 node
102 Description: mdc_set_open_replay_data LBUG
103 Details : Set replay data for requests that are eligible for replay.
107 Description: lustre_mgs: operation 101 on unconnected MGS
108 Details : When MGC is disconnected from MGS long enough, MGS will evict the
109 MGC, and late on MGC cannot successfully connect to MGS and a lot
110 of the error messages complaining that MGS is not connected.
113 Frequency : on start mds
115 Description: Implement get_info(last_id) in obdfilter.
118 Frequency : occasional
120 Description: Correctly check stale fid, not start epoch if ost not support SOM
121 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
122 overwrited with new generated fid, but mds can anwer with one of these
123 two fids and both is not stale. setattr incorectly start epoch and
124 assume will be called done_writeting, but without SOM done_writing
128 Frequency : rare, depends on device drivers and load
130 Description: MDS or OSS nodes crash due to stack overflow
131 Details : Code changes in 1.8.0 increased the stack usage of some functions.
132 In some cases, in conjunction with device drivers that use a lot
133 of stack the MDS (or possibly OSS) service threads could overflow
134 the stack. One change which was identified to consume additional
135 stack has been reworked to avoid the extra stack usage.
138 Frequency : occasional
140 Description: Do not fail import if osc_interpret_create gets -EAGAIN
141 Details : If osc_interpret_create got -EAGAIN it immediately exits and
142 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
143 oscc_has_objects and see OSC has no objests and call
144 oscc_internal_create to resend create request.
146 Severity : enhancement
148 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
150 Severity : enhancement
152 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
155 Frequency : very rare
157 Description: Fix replay if there is an un-replied request and open
158 Details : In some cases, older replay request will revert the
159 mcd->mcd_last_xid on MDS which is used to record the client's
162 Severity : enhancement
164 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
166 Severity : enhancement
168 Description: Add rhel5 support to HEAD.
170 Serverity : enhancement
172 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
177 Description: Don't allow skipping OSTs if index has been specified.
178 Details : Don't allow skipping OSTs if index has been specified, make locking
179 in internal create lots better.
183 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
184 Details : access to bitfield in structure is always rounded to long
185 and this produce problem with not atomic change any bit.
189 Description: Lustre make rpms failed.
190 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
191 builds Lustre rpms from tarball.
194 Frequency : rare on shutdown ost
196 Description: If llog cancel was not send before clean_exports phase, this can
197 produce deadlock in llog code.
198 Details : If llog thread has last reference to obd and call class_import_put
199 this produce deadlock because llog_cleanup_commit_master wait when
200 last llog_commit_thread exited, but this never success because was
201 called from llog_commit_thread.
205 Description: allow userland application know is lost one of stripes.
206 Details : fill lvb_blocks with error code on ost and return it to
207 application if error flag found.
211 Description: NULL lov_tgts causing MDS oops
212 Details : more safe checks for NULL lov_tgts for avoid oops.
214 Severity : enhancement
216 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
220 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
221 Details : If an OST is down the MDS will hang indefinitely in
222 obd_statfs_rqset() waiting for the statfs data. While for
223 MDS QOS usage of statfs, it should not stuck in waiting.
225 Severity : enhancement
227 Description: remote_acl support
228 Details : Support ACL-based permission check for remote user.
229 Support setfacl/getfacl for remote user with the utils
230 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
231 the system "{s,g}etfacl" utils.
233 Severity : enhancement
235 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
237 Severity : enhancement
239 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
243 Description: make mgs_setparam() handle fsname containing dash
244 Details : fsname containing a dash does not work with lctl conf_param
246 Severity : enhancement
248 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
250 Severity : enhancement
252 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
256 Frequency : rare, at shutdown
257 Description: access already free / zero obd_namespace.
258 Details : if client_disconnect_export was called without force flag set,
259 and exist connect request in flight, this can produce access to
260 NULL pointer (or already free pointer) when connect_interpret
261 store ocd flags in obd_namespace.
265 Frequency : only at startup
266 Description: not alloc memory with spinlock held.
267 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
268 if any spinlock held.
270 Severity : enhancement
272 Description: make lustre randomly fail allocating memory
273 Details : Make lustre randomly failed allocating memory for testing purpose.
275 Severity : enhancement
277 Description: lost problems with lov objid file
278 Details : Fixes some scability and access to not inited memory problems
279 in work with lov objdid file.
284 Description: lfs find does not continue on file error
285 Details : Continue other files processing when a file/dir is absent.
289 Description: Inconsistent usage of lustre_pack_reply()
290 Details : Standardize the usage of lustre_pack_reply() such that it
291 always generate a CERROR on failure.
294 Frequency : occasional
296 Description: MDS hang or stay in waiting lock
297 Details : If client receive lock with CBPENDING flag ldlm need send lock
298 cancel as separate rpc, to avoid situation when cancel request
299 can't processed due all i/o threads stay in wait lock.
303 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
307 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
311 Description: Update to SLES9 kernel-2.6.5-7.287.3.
315 Description: Update to RHEL4 latest kernel.
317 Severity : enhancement
319 Description: Build SLES10 patchless client fails
320 Details : The configure was broken by run ./configure with
321 --with-linux-obj=.... argument for patchless client. When the
322 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
323 can't search header adequately. Use absolute path such as
324 -I($LINUX)/include instead.
328 Description: interrupt oig_wait produce painc on resend.
329 Details : brw_redo_request can be used for resend requests from ptlrpcd and
330 private set, and this produce situation when rq_ptlrpcd_data not
331 copyed to new allocated request and triggered LBUG on assert
332 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
333 ptlrpcd set if request is changed and can be safety changed to use
338 Description: LASSERT_{REQ,REP}SWAB macros are buggy
339 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
340 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
341 these macros and replace them with their difinitions instead.
345 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
346 Details : Update which_patch & target file for SLES10 latest kernel.
349 Description: Remove client patches from SLES 10 kernel.
350 Details : This causes SLES 10 clients to behave as patchless clients
351 even on a Lustre-patched (server) kernel.
353 Severity : enhancement
355 Description: self-adjustable client's lru lists
356 Details : use adaptive algorithm for managing client cached locks lru
357 lists according to current server load, other client's work
358 pattern, memory activities, etc. Both, server and client
359 side namespaces provide number of proc tunables for controlling
362 Severity : enhancement
364 Description: light-weight GSS support
365 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
366 and reduce performance overhead.
368 Severity : enhancement
370 Description: Linux keyring support
371 Details : Support using service of Linux keyring for Lustre GSS internal
372 context refresh/cache mechanism.
376 Description: Fix errors in lfs documentation
377 Details : Fixes man pages
381 Description: don't use GFP_* in generic Lustre code.
382 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
386 Description: obdclass is limited by single OBD_ALLOC(idarray)
387 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
391 Description: warnings with build patchless client with vanila 2.6.19 and up
392 Details : change old ctl_table style and replace ctl_table/ctl_table_header
393 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
397 Frequency : Only for SLES
398 Description: sanity_quota fail test_1
399 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
400 file, which makes TSTID[2] inunique.
404 Frequency : Only for SLES9
405 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
406 Details : Update target/ChangeLog/which_patch .
410 Frequency : Only for test_10 in sanity-quota.sh
411 Description: a bug in quota test code
412 Details : checking a test flag in a wrong place
416 Frequency : Only for test_13 in sanity-quota.sh
417 Description: a bug in quota test script
418 Details : shouldn't check overfull allocated quota
422 Description: jbd statistics
423 Details : Port older jbd statistics patch for sles10
425 Severity : enhancement
427 Description: Add printing inode info into message about error in writepage.
432 Description: reply_lock_interpret crash due to race with it and lock cancel.
433 Details : Do not replay locks that are being cancelled. Do not reference
434 locks by their address during replay, just by their handle.
438 Frequency : When flocks are used.
439 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
440 Details : Flock locks might destroy just granted lock if it could be merged
441 with another existing flock, this is done in completion handler,
442 so teach ldlm_cli_enquque_fini that this is a valid case for
448 Description: Oops in read and write path when failing to allocate lock.
449 Details : Check if lock allocation failed and return error back.
453 Description: lstripe command fails for valid OST index
454 Details : The stripe offset is compared to lov->desc.ld_tgt_count
455 instead of lov->desc.ld_active_tgt_count.
459 Description: sanity.sh failed test 103
460 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
465 Description: df doesn't work properly if diskfs blocksize != 4K
466 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
470 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
471 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
475 Description: obd_health_check_timeout too short
476 Details : set obd_health_check_timeout as 1.5x of obd_timeout
480 Description: llapi_file_create() does not allow some changes
481 Details : add llapi_file_open() that allows specifying the mode and
482 open flags, and also returns an open file handle.
486 Description: merge and cleanup kernel patches.
487 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
491 Description: Add journal checksum support.(Kernel part)
492 Details : The journal checksum feature adds two new flags i.e
493 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
494 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
495 indicates that the commit block contains the checksum for
496 the blocks described by the descriptor blocks. Now commit
497 record can be sent to disk without waiting for descriptor
498 blocks to be written to disk. This behavior is controlled
499 using JBD2_FEATURE_ASYNC_COMMIT flag.
503 Description: OSS needs mutliple precreate threads
504 Details : Add ability to start more than one create thread per OSS.
508 Description: Sanity.sh test_65k failed: setstripe should have succeeded
509 Details : execute lfs setstripe on client
513 Description: mds_obd_create error creating tmp object
514 Details : When the user sets quota on root, llog will be affected and can't
515 create files and write files.
518 Frequency : Always on ia64 patchless client, and possibly others.
520 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
521 Details : This allows the patchless client to be loaded on architectures
526 Description: RedHat Update kernel for RHEL5
527 Details : Add the kernel config file for RHEL5.
531 Description: RedHat Update kernel for RHEL5
532 Details : Modify the kernel config file more closer RHEL5.
536 Description: Build failure against Centos5 (RHEL5)
537 Details : Define PAGE_SIZE when it isn't present.
541 Description: test_7 on sanity-quota.sh can't be used on separate machines
542 Details : add facet to handle it
546 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
547 Details : with NFS, the anon dentry's parent was set to itself in
548 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
549 corresponding dentry other than use rec->ur_name.
551 Severity : enhancement
553 Description: lfs setstripe enhancement
554 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
558 Description: enable data checksumming by default
559 Details : enable checksum by default, allow --disable-checksum
560 configure option and "-o nochecksum" mount option. Checksums
561 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
565 Description: lustre support for RHEL5
566 Details : Add support for RHEL5.
570 Description: Client eviction due to failover config
571 Details : after a connection loss, the lustre client should attempt to
572 reconnect to the last active server first before trying the
573 other potential connections.
577 Description: when mds and osts use different quota unit(32bit and 64bit),
578 quota will be released repeatly.
579 Details : void sending multiple quota reqs to mds, which will keep the
580 status between the reqs.
584 Description: osts not allocated evenly to files
585 Details : change the condition to increase offset_idx
588 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
590 Description: Data corruption for OSTs that are formatted larger than 2TB
592 Details : When generating the bio request for lustre file writes the
593 sector number would overflow a temporary variable before being
594 used for the IO. The data reads correctly from Lustre (which
595 will overflow in a similar manner) but other file data or
596 filesystem metadata may be corrupted in some cases.
600 Description: Tune the kernel for good SCSI performance.
601 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
602 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
607 Description: rewrite ext2-derived code in obdclass/uuid.c
608 Details : rewrite inherited code (uuid parsing code from ext2 utils)
609 from scratch preserving functionality.
613 Description: "lfs find -obd UUID" prints directories
614 Details : "lfs find -obd UUID" will return all directory names instead
615 of just file names. It is incorrect because the directories
616 do not reside on the OSTs.
620 Description: use i_size_read and i_size_write in 2.6 port
621 Details : replace inode->i_size access with i_size_read/write()
624 Frequency : when removing large files
626 Description: scheduling issue during removal of large Lustre files
627 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
628 It causes scheduling issues when removing large files (17TB in the
632 Frequency : only with liblustre clients on XT3
634 Description: evictions taking too long
635 Details : allow llrd to evict clients directly on OSTs
638 Frequency : only on ppc
640 Description: /proc/fs/lustre/devices broken on ppc
641 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
642 We should make sure the type of 'index' is loff_t and then cast
643 explicitly as needed below. Do not assign an explicitly cast
647 Frequency : only for rhel5
649 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
650 Details : Modify the target file & which_kernel.
652 Severity : enhancement
654 Description: omit set fsid for export NFS
655 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
656 as FS_REQUIRES_DEV for avoid problems with generate fsid.
660 Frequency : Always, for kernels after 2.6.16
661 Description: Fix warning idr_remove called for id=.. which is not allocated.
662 Details : Last kernels save old s_dev before kill super and not allow
663 to restore from callback - restore it before call kill_anon_super.
667 Description: buffer overruns could theoretically occur
668 Details : llapi_semantic_traverse() modifies the "path" argument by
669 appending values to the end of the origin string, and a
670 overrun may occur. Adding buffer overrun check in liblustreapi.
674 Description: Fix error on 'ls .' at the top of the Lustre mount.
675 Details : Don't revalidate dentry if it is a root dentry.
679 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
680 Details : Modify vm-tunables-rhel4.patch.
684 Description: Kernel config for 2.6.18-vanilla.
685 Details : Modify targets/2.6-vanilla.target.in.
686 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
687 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
688 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
689 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
693 Description: improve handling recoverable errors
694 Details : if request processig with error which can be recoverable on server
695 request should be resend, otherwise page released from cache and
700 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
701 Details : Modify target file & which_patch.
702 A flaw was found in the IA32 system call emulation provided
703 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
704 value could be stored in the %RAX register, which could trigger an
705 out-of-bounds system call table access. An untrusted local user
706 could exploit this flaw to run code in the kernel
707 (ie a root privilege escalation). (CVE-2007-4573).
711 Description: O_DIRECT bypasses client statistics.
712 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
713 read_ahead_stats were updated. copied the stats section from
714 osc_send_oap_rpc() into async_internal().
718 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
722 Description: change order of libsysio includes
723 Details : '#include sysio.h' should always come before '#include xtio.h'
727 Description: handle "serious error: objid * already exists" more gracefully
728 Details : If LAST_ID value on disk is smaller than the objects existing in
729 the O/0/d* directories, it indicates disk corruption and causes an
730 LBUG(). If the object is 0-length, then we should use the existing
731 object. This will help to avoid a full fsck in most cases.
733 Severity : enhancement
735 Description: adapt the lustre_config script to support the upgrade case
736 Details : Add "-u" option for lustre_config script to support upgrading 1.4
737 server targets to 1.6 in parallel.
741 Description: To avoid grant space > avaible space when the disk is almost
742 full. Without this patch you might see the error "grant XXXX >
743 available" or some LBUG about grant, when the disk is almost
745 Details : In filter_check_grant, for non_grant cache write, we should
746 check the left space by if (*left > ungranted + bytes), instead
747 of (*left > ungranted), because only we are sure the left space
748 is enough for another "bytes", then the ungrant space should be
749 increase. In client, we should update cl_avail_grant only there
750 is OBD_MD_FLGRANT in the reply.
754 Description: Update RHEL 4 kernel to fix local root privilege escalation.
755 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
756 described in CVE-2007-4573. This problem could allow untrusted
757 local users to gain root access.
760 Frequency : when using O_DIRECT and quotas
762 Description: Incorrect file ownership on O_DIRECT output files
763 Details : block usage reported by 'lfs quota' does not take into account
764 files that have been written with O_DIRECT.
769 Description: touch file failed when fs is not full
770 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
771 otherwise we can get ENOSP while fs is not full.
775 Description: parallel lock callbacks
776 Details : Instead of sending blocking and completion callbacks as separated
777 requests, adding them to a set and sending in parallel.
780 Frequency : only for Cray XT3
781 Bugzilla : 12829/13455
782 Description: Changing primary group doesn't change the group lustre assigns to
784 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
785 overridden with the primary group provided by the group upcall,
786 whereas the supplied fsgid can be trusted if it is in the list of
787 supplementary groups returned by the group upcall.
789 Severity : enhancement
791 Description: Allow masking D_WARNING, D_ERROR messages from console
792 Details : Console messages can now be disabled via lnet.printk.
796 Description: User code with malformed file open parameter crashes client node
797 Details : Before packing join_file req, all the related reference should be
798 checked carefully in case some malformed flags cause fake
799 join_file req on client.
803 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
804 Details : ldlm_completion_ast() assumes that a lock is granted when the req
805 mode is equal to the granted mode. However, it should also check
806 that LDLM_FL_CP_REQD is not set.
810 Description: Heavy nfs access might result in deadlocks
811 Details : After ELC code landed, it is now improper to enqueue any mds
812 locks under och_sem, because enqueue might want to decide to
813 cancel open locks for same inode we are holding och_sem for.
817 Description: Client eviction while running blogbench
818 Details : A lot of unlink operations with concurrent I/O can lead to a
819 deadlock causing evictions. To address the problem, the number of
820 oustanding OST_DESTROY requests is now throttled to
821 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
822 callbacks are processed in priority.
826 Description: enable ACLs on MDS by default
827 Details : ACLs must be enabled on MDS by default.
830 Frequency : PPC/PPC64 only
832 Description: conflicts between asm-ppc64/types.h and lustre_types.h
833 Details : fix duplicated definitions between asm-ppc64/types.h and
834 lustre_types.h on PPC.
837 Frequency : PPC/PPC64 only
839 Description: asm-ppc/segment.h does not exist
840 Details : fix compile issue on PPC.
844 Description: better handle error messages in extents code
847 Frequency : RHEL4 only
849 Description: mkfs is very slow on IA64/RHEL4
850 Details : A performance regression has been discovered in the MPT Fusion
851 driver between versions 3.02.73rh and 3.02.99.00rh. As a
852 consequence, we have downgraded the MPT Fusion driver in the RHEL4
853 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
855 Severity : enhancement
857 Description: SNMP support enhancement
858 Details : Adding total number of sampled request for an MDS node in snmp
861 Severity : enhancement
863 Description: Optimize ldlm waiting list processing for PR extent locks
864 Details : When processing waiting list for read extent lock and meeting read
865 lock that is same or wider to it that is not contended, skip
866 processing rest of the list and immediatelly return current
867 status of conflictness, since we are guaranteed there are no
868 conflicting locks in the rest of the list.
872 Description: Time out and refuse to reconnect
873 Details : When the failover node is the primary node, it is possible
874 to have two identical connections in imp_conn_list. We must
875 compare not conn's pointers but NIDs, otherwise we can defeat
876 connection throttling.
880 Description: port llog fixes from b1_6 into HEAD
881 Details : Port llog reference couting and some llog cleanups from b1_6
882 (bug 10800) into HEAD, for protect from panic and access to already
883 free llog structures.
887 Description: Detect stride IO mode in read-ahead
888 Details : When a client does stride read, read-ahead should detect that and
889 read-ahead pages according to the detected stride pattern.
893 Description: data checksumming impacts single node performance
894 Details : add support for several checksum algorithm. Currently, only CRC32
895 and Adler-32 are supported. The checksum type can be changed on
896 the fly via /proc/fs/lustre/osc/*/checksum_type.
900 Description: use adler32 for page checksums
901 Details : when available, use the Adler-32 algorithm instead of CRC32 for
906 Description: build for x2 fails
907 Details : fix compile issue on Cray systems.
911 Description: Properly match for duplicate locks
912 Details : Due to different lock order from skiplists code, we need to
913 traverse entire list for now
916 Frequency : only on PPC/SLES10
918 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
919 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
920 be defined. Add a hack in mkfs_lustre.c to work around this compile
925 Description: LASSERT on MDS when client holding flock lock dies
926 Details : ldlm pool logic depends on number of granted locks equal to
927 number of released locks which is not true for flock locks, so
928 just exclude such locks from consideration.
932 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
933 Details : Use fsfilt_sync() for both the whole filesystem sync and
934 individual file sync to eliminate dangerous inode locking
935 with I_LOCK that can lead to a deadlock.
939 Description: performance in 1.6.3
940 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
941 sure that 1MB requests are not fragmented by the block layer.
943 Severity : enhancement
945 Description: organize the server-side client stats on per-nid basis
946 Details : Change the structure of stats under obdfilter and mds to
954 The "uuid"s file would list the uuids of _active_ exports.
955 And the clear entry is to clear all stats and stale nids.
957 --------------------------------------------------------------------------------
959 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
961 * Support for kernels:
962 2.6.5-7.283 (SLES 9),
963 2.6.9-55.EL (RHEL 4),
964 2.6.16.46-0.14 (SLES 10),
965 2.6.18.8 vanilla (kernel.org)
966 * Client support for unpatched kernels:
967 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
968 2.6.16 - 2.6.22 vanilla (kernel.org)
969 * Due to recently discovered recovery problems, we do not recommend
970 using patchless RHEL 4 clients with this or any earlier release.
971 * Recommended e2fsprogs version: 1.39.cfs8
972 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
973 * Starting with this release, the ldiskfs backing filesystem required
974 by Lustre is now in its own package, lustre-ldiskfs. This package
975 should be installed. It is versioned separately from Lustre and
976 may be released separately in future.
980 Description: block reactivating mgc import until all deactivates complete
981 Details : Fix race when failing back MDT/MGS to itself (testing)
984 Frequency : at statup only
986 Description: mds_lov_synchronize race leads to various problems
987 Details : simultaneous MDT->OST connections at startup can cause the
988 sync to abort, leaving the OSC in a bad state.
990 Severity : enhancement
992 Description: add optional extra BUILD_VERSION info
993 Details : add a new environment variable (namely LUSTRE_VERS) which allows
994 to override the lustre version.
997 Frequency : 2.6.18 servers only
999 Description: ll_kern_mount() doesn't release the module reference
1000 Details : The ldiskfs module reference count never drops down to 0
1001 because ll_kern_mount() doesn't release the module reference.
1006 Description: server LBUG when using old ost_num_threads parameter
1007 Details : Accept the old ost_num_threads parameter but warn that it
1008 is deprecated, and fix an off-by-one error that caused an LBUG.
1013 Description: Transient SCSI error results in persistent IO issue
1014 Details : iobuf->dr_error is not reinitialized to 0 between two
1018 Frequency : sometimes when underlying device returns I/O errors
1020 Description: OSTs not going read-only during write failures
1021 Details : OSTs are not remounted read-only when the journal commit threads
1022 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1023 instead of the ext3 wrappers.
1027 Description: poor connect scaling with increasing client count
1028 Details : Don't run filter_grant_sanity_check for more than 100 exports
1029 to improve scaling for large numbers of clients.
1032 Frequency : SLES10 only
1034 Description: sanity-quota.sh quotacheck failed: rc = -22
1035 Details : Quotas cannot be enabled on SLES10.
1038 Frequency : liblustre clients only
1040 Description: getdirentries does not give error when run on compute nodes
1041 Details : getdirentries does not fail when the size specified as an argument
1042 is too small to contain at least one entry
1044 Severity : enhancement
1046 Description: Add LNET router traceability for debug purposes
1047 Details : If a checksum failure occurs with a router as part of the
1048 IO path, the NID of the last router that forwarded the bulk data
1049 is printed so it can be identified.
1054 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1055 Details : Due to a race condition, liblustre clients were occasionally
1056 evicted incorrectly.
1058 Severity : enhancement
1060 Description: lfs setstripe use optional parameters instead of postional
1063 Severity : enhancement
1065 Description: Nanosecond timestamp support for ldiskfs
1066 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1067 resolution timestamps. There is not yet support for this at
1068 the Lustre filesystem level.
1071 Frequency : during server recovery
1073 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1074 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1078 Frequency : nfs export on patchless client
1080 Description: connectathon hang when test nfs export over patchless client
1081 Details : Disconnected dentry cannot be found with lookup, so we do not need
1082 to unhash it or make it invalid
1085 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1086 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1087 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1092 Description: open req refcounting wrong on reconnect
1093 Details : If reconnect happened between getting open reply from server and
1094 call to mdc_set_replay_data in ll_file_open, we will schedule
1095 replay for unreferenced request that we are about to free.
1096 Subsequent close will crash in variety of ways.
1097 Check that request is still eligible for replay in
1098 mdc_set_replay_data().
1103 Description: disable writes to filesystem when reading health_check file
1104 Details : the default for reading the health_check proc file has changed
1105 to NOT do a journal transaction and write to disk, because this
1106 can cause reads of the /proc file to hang and block HA state
1107 checking on a healthy but otherwise heavily loaded system. It
1108 is possible to return to the previous behaviour during configure
1109 with --enable-health-write.
1111 Severity : enhancement
1113 Description: 64-bit inode version
1114 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1115 to the inode. This will be required for version-based recovery.
1120 Description: MDS fails to start if a duplicate client export is detected
1121 Details : in some rare cases it was possible for a client to connect to
1122 an MDS multiple times. Upon recovery the MDS would detect this
1123 and fail during startup. Handle this more gracefully.
1125 Severity : enhancement
1127 Description: Add -o localflock option to simulate old noflock
1129 Details : This will achieve local-only flock/fcntl locks
1135 Description: log_commit_thread vs filter_destroy race leads to crash
1136 Details : Take import reference before releasing llog record semaphore
1141 Description: Wrong request locking in request set processing
1142 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1143 imp_delayed_list, in this place should be used imp_lock.
1146 Frequency : when reconnection
1148 Description: Grant Leak when osc reconnect to OST
1149 Details : When osc reconnect ost, OST(filter) should check whether it
1150 should grant more space to client by comparing fed_grant and
1151 cl_avail_grant, and return the granted space to client instead
1152 of "new granted" space, because client will call osc_init_grant
1153 to update the client grant space info.
1156 Frequency : when client reconnect to OST
1158 Description: Grant Leak when osc do resend and replay bulk write
1159 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1160 bulk write request, because the grant info while be sync between
1161 OSC and OST when reconnect, and we should ignore the grant info
1162 these of resend/replay write req.
1167 Description: Grant space more than avaiable left space sometimes.
1168 Details : When then OST is about to be full, if two bulk writing from
1169 different clients came to OST. Accord the avaliable space of the
1170 OST, the first req should be permitted, and the second one
1171 should be denied by ENOSPC. But if the seconde arrived before
1172 the first one is commited. The OST might wrongly permit second
1173 writing, which will cause grant space > avaiable space.
1176 Frequency : when client is evicted
1178 Description: Grant might be wrongly erased when osc is evicted by OST
1179 Details : when the import is evicted by server, it will fork another
1180 thread ptlrpc_invalidate_import_thread to invalidate the
1181 import, where the grant will be set to 0. While the original
1182 thread will update the grant it got when connecting. So if
1183 the former happened latter, the grant will be wrongly errased
1184 because of this race.
1189 Description: Checking Stale with correct fid
1190 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1191 to check whether it is stale, because sometimes, we want the
1192 enqueue happened anyway, and op_data.fid2 will not be initialized.
1194 Severity : enhancement
1196 Description: update patchless client
1197 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1200 Frequency : only with 2.4 kernel
1202 Description: random memory corruption
1203 Details : size of struct ll_inode_info is to big for union inode.u and this
1204 can be cause of random memory corruption.
1209 Description: Memory leak in recovery
1210 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1211 It should also check obd_fail before fsfilt_start, otherwise if
1212 fsfilt_start return -EROFS,(failover mds during mds recovery).
1213 then the req will return with repmsg->transno = 0 and rc = EROFS.
1214 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1215 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1216 be freed no matter whether fsfilt_commit success or not.
1219 Frequency : only with huge count clients
1221 Description: Prevents from taking the superblock lock in llap_from_page for
1223 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1224 for llap_from_page call in ll_removepage prevents from taking the
1225 superblock lock for a soon died page.
1230 Description: Not check open intent error before release open handle
1231 Details : in some rare cases, the open intent error is not checked before
1232 release open handle, which may cause
1233 ASSERTION(open_req->rq_transno != 0), because it tries to release
1234 the failed open handle.
1239 Description: Set cat log bitmap only after create log success.
1240 Details : in some rare cases, the cat log bitmap is set too early. and it
1241 should be set only after create log success.
1245 Description: Accessing a block bevice can re-enable I/O when Lustre is
1246 tearing down a device.
1247 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1252 Description: service threads may hog cpus when there are a lot of requests
1253 Details : Insert cond_resched to give other threads a chance to use some CPU
1258 Description: the cat log was not initialized in recovery
1259 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1260 unlink log on mds will not be initialized until mds post
1261 recovery. And also in mds post recovery, the unlink log will
1262 initialization will be done asynchronausly, so there will be race
1263 between add unlink log and unlink log initialization.
1267 Description: brw_stats were being printed incorrectly
1268 Details : brw_stats were being printed as log2 but all of them were not
1269 recorded as log2. Also remove some code duplication arising from
1270 filter_tally_{read,write}.
1274 Frequency : rare, only in recovery.
1275 Description: ASSERTION(req->rq_type != LI_POISON) failed
1276 Details : imp_lock should be held while iterating over imp_sending_list for
1277 prevent destroy request after get timeout in ptlrpc_queue_wait.
1281 Description: replay-single.sh test 52 fails
1282 Details : A lock's skiplist need to be cleanup when it being unlinked
1283 from its resource list.
1287 Description: Short directio read returns full requested size rather than
1289 Details : Direct I/O operations should return actual amount of bytes
1290 transferred rather than requested size.
1292 Severity : enhancement
1294 Description: metadata RPC reduction (e.g. for rm performance)
1295 Details : decrease the amount of synchronous RPC between clients and servers
1296 by canceling conflicing lock before the operation on the client
1297 and packing thier handles into the main operation RPC to server.
1299 Severity : enhancement
1301 Description: Async OSC create to avoid the blocking unnecessarily.
1302 Details : If a OST has no remain object, system will block on the creating
1303 when need to create a new object on this OST. Now, ways use
1304 pre-created objects when available, instead of blocking on an
1305 empty osc while others are not empty. If we must block, we block
1306 for the shortest possible period of time.
1308 Severity : enhancement
1310 Description: refine locking for avoid write wrong info into lov_objid file
1311 Details : fix possible races with add new target and write/update data in
1314 --------------------------------------------------------------------------------
1316 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1321 Frequency : on some architectures
1323 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1324 Details : Uninitialized flags sometimes cause configuration commands to
1328 Frequency : patchless clients only
1330 Description: missing __iget() symbol export
1331 Details : The __iget() symbol export is missing. To avoid the need for
1332 this on patchless clients the deathrow inode reaper is turned
1333 off, and we depend on the VM to clean up old inodes. This
1334 dependency was during via the fix for bug 12181.
1336 --------------------------------------------------------------------------------
1338 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1340 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1341 INTEROPERATE with older versions automatically. In many cases a
1342 special upgrade step is needed. Please read the
1343 user documentation before upgrading any part of a live system.
1344 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1345 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1346 * WARNING: Lustre configuration and startup changes are required with
1347 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1353 Severity : enhancement
1355 Description: MountConf
1356 Details : Lustre configuration is now managed via mkfs and mount
1357 commands instead of lmc and lconf. New obd types (MGS, MGC)
1358 are added for dynamic configuration management. See
1359 https://mail.clusterfs.com/wikis/lustre/MountConf for
1362 Severity : enhancement
1364 Description: dynamic OST addition
1365 Details : OSTs can now be added to a live filesystem
1367 Severity : enhancement
1369 Description: startup order invariance
1370 Details : MDTs and OSTs can be started in any order. Clients only
1371 require the MDT to complete startup.
1373 Severity : enhancement
1375 Description: parallel, asynchronous orphan cleanup
1376 Details : orphan cleanup is now performed in separate threads for each
1377 OST, allowing parallel non-blocking operation.
1379 Severity : enhancement
1381 Description: optimized stripe assignment
1382 Details : stripe assignments are now made based on ost space available,
1383 ost previous usage, and OSS previous usage, in order to try
1384 to optimize storage space and networking resources.
1386 Severity : enhancement
1388 Description: Permanently set tunables
1389 Details : All writable /proc/fs/lustre tunables can now be permanently
1390 set on a per-server basis, at mkfs time or on a live system.
1392 Severity : enhancement
1394 Description: Lustre message v2
1395 Details : Add lustre message format v2.
1397 Severity : enhancement
1399 Description: client OST exclusion list
1400 Details : Clients can be started with a list of OSTs that should be
1401 declared "inactive" for known non-responsive OSTs.
1405 Description: SPEC SFS validation failure on NFS v2 over lustre.
1406 Details : Changes the blocksize for regular files to be 2x RPC size,
1407 and not depend on stripe size.
1409 Severity : enhancement
1411 Description: Multiple MD RPCs in flight.
1412 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1413 To avoid overly-overloading MDS, introduce a limit on number of
1414 MDS RPCs in flight for a single client and add /proc controls
1415 to adjust this limit.
1417 Severity : enhancement
1419 Description: client read/write statistics
1420 Details : Add client read/write call usage stats for performance
1421 analysis of user processes.
1422 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1423 file access. extents_stats shows chunk size distribution.
1424 extents_stats_per_process show chunk size distribution per
1427 Severity : enhancement
1429 Description: mds statistics
1430 Details : Add detailed mds operations statistics in
1431 /proc/fs/lustre/mds/*/stats.
1435 Description: Failure of copying files with lustre special EAs.
1436 Details : Client side always return success for setxattr call for lustre
1437 special xattr (currently only "trusted.lov").
1442 Description: Refcount LNET uuids
1443 Details : The global LNET uuid list grew linearly with every startup;
1444 refcount repeated list entries instead of always adding to
1447 Severity : enhancement
1449 Description: Dynamic service threads
1450 Details : Within a small range, start extra service threads
1451 automatically when the request queue builds up.
1454 Frequency : mixed-endian client/server environments
1456 Description: mixed-endian crashes
1457 Details : The new msg_v2 system had some failures in mixed-endian
1460 Severity : enhancement
1462 Description: Easy OST removal
1463 Details : OSTs can be permanently deactivated with e.g. 'lctl
1464 conf_param lustre-OST0001.osc.active=0'
1466 Severity : enhancement
1468 Description: MGS proc entries
1469 Details : Added basic proc entries for the MGS showing what filesystems
1472 Severity : enhancement
1474 Description: provide MGS failover
1475 Details : Added config lock reacquisition after MGS server failover.
1477 Severity : enhancement
1479 Description: add Linux 2.4 support
1480 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1484 Description: a large application tries to do I/O to the same resource and dies
1485 in the middle of it.
1486 Details : Check the req->rq_arrival time after the call to
1487 ost_brw_lock_get(), but before we do anything about
1488 processing it & sending the BULK transfer request. This
1489 should help move old stale pending locks off the queue as
1490 quickly as obd_timeout.
1493 Frequency : when an incorrect nid is specified during startup
1495 Description: ptlrpc connect to non-existant node causes kernel crash
1496 Details : LNET can't be re-entered from an event callback, which
1497 happened when we expire a message after the export has been
1498 cleaned up. Instead, hand the zombie cleanup off to another
1501 Severity : enhancement
1503 Description: plain/inodebits lock performance improvement
1504 Details : Grouping plain/inodebits in granted list by their request modes
1505 and bits policy, thus improving the performance of search through
1509 Frequency : only if OST filesystem is corrupted
1511 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1512 Details : for a short time RPCs with bulk IO are in the replay list,
1513 but replay of bulk IOs is unimplemented. If the OST filesystem
1514 is corrupted due to disk cache incoherency and then replay is
1515 started it is possible to trip an assertion. Avoid putting
1516 committed RPCs into the replay list at all to avoid this issue.
1519 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1521 Description: System hang on startup
1522 Details : This bug allowed the liblustre (e.g. catamount) client to
1523 return to the app before handling all startup RPCs. This
1524 could leave the node unresponsive to lustre network traffic
1525 and manifested as a server ptllnd timeout.
1527 Severity : enhancement
1529 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1530 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1531 boolean to control whether to print peer debug info when a
1532 client's RPC times out.
1535 Frequency : only for kernels with patches from Lustre below 1.4.3
1537 Description: Remove old rdonly API
1538 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1541 Frequency : only for devices with external journals
1543 Description: Set external device read-only also
1544 Details : During a commanded failover stop, we set the disk device
1545 read-only while the server shuts down. We now also set any
1546 external journal device read-only at the same time.
1549 Frequency : when upgrading from 1.4 while trying to change parameters
1551 Description: The wrong (new) MDC name was used when setting parameters for
1552 upgraded MDT's. Also allows changing of OSC (and MDC)
1553 parameters if --writeconf is specified at tunefs upgrade time.
1556 Frequency : when setting specific ost indicies
1558 Description: QOS code breaks on skipped indicies
1559 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1560 created with --index need not be sequential.
1562 Severity : enhancement
1564 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1565 Details : The uninit_groups feature works in conjunction with the kernel
1566 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1567 up the pass1 processing of e2fsck. This is a read-only feature
1568 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1569 mount filesystems that have had this feature enabled.
1571 Severity : enhancement
1573 Description: Improve multi-block allocation algorithm to avoid fragmentation
1574 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1575 allocation locality and avoid filesystem fragmentation.
1577 ------------------------------------------------------------------------------
1579 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1581 * Support for kernels:
1582 2.4.21-47.0.1.EL (RHEL 3)
1583 2.6.5-7.283 (SLES 9)
1584 2.6.9-42.0.10.EL (RHEL 4)
1585 2.6.12.6 vanilla (kernel.org)
1586 2.6.16.27-0.9 (SLES 10)
1587 * Recommended e2fsprogs version: 1.39.cfs5
1589 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1593 Frequency : occasional, depends on client load and configuration
1594 Bugzilla : 12181, 12203
1595 Description: data loss for recently-modified files
1597 Details : In some cases it is possible that recently written or created
1598 files may not be written to disk in a timely manner (this should
1599 normally be within 30s unless client IO load is very high).
1600 The problem appears as zero-length files or files that are a
1601 multiple of 1MB in size after a client crash or client eviction
1602 that are missing data at the end of the file.
1604 This problem is more likely to be hit on clients where files are
1605 repeatedly created and unlinked in the same directory, clients
1606 have a large amount of RAM, have many CPUs, the filesystem has
1607 many OSTs, the clients are rebooted frequently, and/or the files
1608 are not accessed by other nodes after being written.
1610 The presence of the problem can be detected by looking at
1611 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1612 smaller than the second (nr_unused) then dirty files will not
1613 be flushed automatically to disk. "sync; sleep 10" should be
1614 run several times on the node before unmounting it to update
1615 Lustre (this is also safe to run on nodes without this problem).
1617 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1618 that can cause this same problem, so customers using that kernel
1619 also need to update the kernel in addition to Lustre. In order
1620 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1622 It is normal that files written just before a client crash (less
1623 than 30s) may not yet have been flushed to disk, even for local
1627 Frequency : frequent on thin XT3 nodes
1629 Description: UUID collision on thin XT3 Linux nodes
1630 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1631 randomly, since we relied on an insufficiently-seeded PRNG.
1636 Description: OSS hangs after "All ost request buffers busy"
1637 Details : A deadlock between quota and journal operations caused OSS
1638 hangs after printing "All ost request buffers busy."
1641 Frequency : always on liblustre builds
1643 Description: Cleanup compiler warnings on liblustre
1646 Frequency : always on liblustre builds on XT3
1648 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1652 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1653 > 32000 subdirectories
1654 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1655 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1656 subdirectory count crosses 32000. This will aid e2fsck to
1657 correctly handle more than 32000 subdirectories.
1660 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1662 Description: System hang on startup
1663 Details : This bug allowed the liblustre (e.g. catamount) client to
1664 return to the app before handling all startup RPCs. This
1665 could leave the node unresponsive to lustre network traffic
1666 and manifested as a server ptllnd timeout.
1668 Severity : enhancement
1670 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1671 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1672 boolean to control whether to print peer debug info when a
1673 client's RPC times out.
1678 Description: make O_SYNC working on 2.6 kernels
1679 Details : 2.6 kernels use different method for mark pages for write,
1680 so need add a code to lustre for O_SYNC work.
1685 Description: Failure to close file and release space on NFS
1686 Details : Put inode details into lock acquired in ll_intent_file_open.
1687 Use mdc_intent_lock in ll_intent_open to properly
1688 detect all kind of errors unhandled by mdc_enqueue.
1693 Description: proc file read during shutdown sometimes raced obd removal,
1695 Details : Add lock to prevent obd access after proc file removal.
1698 Frequency : Only for files larger than 4GB on 32-bit clients.
1700 Description: improperly doing page alignment of locks
1701 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1702 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1707 Description: RPCs being resent when they shouldn't be
1708 Details : Some RPCs that should not be resent are being resent. This
1709 can cause inconsistencies in the RPC state machine. Do not
1710 resend such requests.
1713 Frequency : rare, only with NFS export
1715 Description: Crash on NFS re-export node
1716 Details : under very unusual load conditions an assertion is hit in
1717 ll_intent_file_open()
1720 Frequency : only if OST filesystem is corrupted
1722 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1723 Details : for a short time RPCs with bulk IO are in the replay list,
1724 but replay of bulk IOs is unimplemented. If the OST filesystem
1725 is corrupted due to disk cache incoherency and then replay is
1726 started it is possible to trip an assertion. Avoid putting
1727 committed RPCs into the replay list at all to avoid this issue.
1732 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1733 Details : Large single O_DIRECT read and write calls can fail to allocate
1734 a sufficiently large buffer to process the request. In case of
1735 allocation failure the allocation is retried with a smaller
1736 buffer and broken into smaller requests.
1738 Severity : enhancement
1740 Description: Add -o localflock option to simulate old noflock behaviour.
1741 Details : This will achieve local-only flock/fcntl locks coherentness.
1746 Description: versioning check is incomplete
1747 Details : Checking the version difference of client vs. server, report
1748 error if the gap is too big.
1753 Description: add support PG_writeback bit
1754 Details : add support for PG_writeback bit for Lustre, for more carefull
1755 work with page cache in 2.6 kernel. This also fix some deadlocks
1756 and remove hack for work O_SYNC with 2.6 kernel.
1758 Severity : enhancement
1760 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1761 Details : The uninit_groups feature works in conjunction with the kernel
1762 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1763 up the pass1 processing of e2fsck. This is a read-only feature
1764 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1765 mount filesystems that have had this feature enabled.
1767 Severity : enhancement
1769 Description: Improve multi-block allocation algorithm to avoid fragmentation
1770 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1771 allocation locality and avoid filesystem fragmentation.
1773 ------------------------------------------------------------------------------
1775 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1777 * Support for kernels:
1778 2.6.9-42.0.3.EL (RHEL 4)
1779 2.6.5-7.276 (SLES 9)
1780 2.4.21-47.0.1.EL (RHEL 3)
1781 2.6.12.6 vanilla (kernel.org)
1782 2.6.16.21-0.8 (SLES10)
1783 * Recommended e2fsprogs version: 1.39.cfs2-0
1785 * The backwards-compatible /proc/sys/portals symlink has been removed
1786 in this release. Before upgrading, please ensure that you change
1787 any configuration scripts or /etc/sysctl.conf files that access
1788 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1789 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1790 in advance of the upgrade on any system running Lustre 1.4.6 or
1791 newer, since /proc/sys/lnet was added in that version.
1792 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1796 Frequency : only when quota is used
1798 Description: avoid scanning export list for quota master
1799 Details : Change the algorithms to avoid scanning export list in order
1800 to improve the efficiency.
1803 Frequency : MDS failover only, very rarely
1805 Description: "went back in time" messages on mds failover
1806 Details : The greatest transno may be lost when the current operation
1807 finishes with an error (transno==0) and the client's last_rcvd
1808 record is over-written. Save the greatest transno in the
1809 mds_last_transno for this case.
1812 Frequency : always for specific kernels and striping counts
1814 Description: client may get "Matching packet too big" without ACL support
1815 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1816 when trying to access files in certain configurations. The
1817 clients should in fact be denied when mounting because they do
1818 not understand ACLs.
1821 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1823 Description: many clients connecting with IO in progress causes connect timeouts
1824 Details : Avoid synchronous journal commits to avoid delays caused by many
1825 clients connecting/disconnecting when bulk IO is in progress.
1826 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1827 OST_IO_PORTAL to avoid delays behind potentially many pending
1831 Frequency : occasionally with multiple writers to a single file
1833 Description: shared writes to file may result in wrong size reported by stat()
1834 Details : Allow growing of kms when extent lock is cancelled
1837 Frequency : always with random mmap IO to multi-striped file
1839 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1840 Details : Only if the hole is at the end of OST object so that kms is too
1841 small. Fix is to increase kms accordingly in ll_nopage.
1844 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1846 Description: writes to a missing object would leak memory on the OST
1847 Details : If there is an inconsistency between the MDS and OST filesystems,
1848 such that the MDS references an object that doesn't exist, writes
1849 to that object will leak memory due to incorrect cleanup in the
1850 error handling path, eventually running out of memory on the OST.
1855 Description: Creating too long symlink causes lustre errors
1856 Details : Check symlink and name lengths before sending requests to MDS.
1859 Frequency : only if flock is enabled (not on by default)
1861 Description: posix locks not released on fd closure on 2.6.9+
1862 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1863 kernels, this caused such locks not to be released on fd close and
1864 then assertions on fs unmount about still used locks.
1867 Frequency : MDS failover only, very rarely
1869 Description: clients may get ASSERTION(granted_lock != NULL)
1870 Details : When request was taking a long time, and a client was resending
1871 a getattr by name lock request. The were multiple lock requests
1872 with the same client lock handle and
1873 mds_getattr_name->fixup_handle_for_resent_request found one of the
1874 lock handles but later failed with ASSERTION(granted_lock != NULL).
1879 Description: handle->h_buffer_credits > 0, assertion failure
1880 Details : h_buffer_credits is zero after truncate, causing assertion
1881 failure. This patch extends the transaction or creates a new
1885 Frequency : NFS re-export or patchless client
1886 Bugzilla : 11179, 10796
1887 Description: Crash on NFS re-export node (__d_move)
1888 Details : We do not want to hash the dentry if we don't have a lock.
1889 But if this dentry is later used in d_move, we'd hit uninitialised
1890 list head d_hash, so we just do this to init d_hash field but
1891 leave dentry unhashed.
1894 Frequency : NFS re-export or patchless client
1896 Description: NFS exports has problem with symbolic link
1897 Details : lustre client didn't properly install dentry when re-exported
1898 to NFS or running patchless client.
1901 Frequency : NFS re-export or patchless client
1903 Description: Various nfs/patchless fixes.
1904 Details : fixes reuse disconected alias for lookup process - this fixes
1905 warning "find_exported_dentry: npd != pd",
1906 fix permission error with open files at nfs.
1907 fix apply umask when do revalidate.
1910 Frequency : occasional
1912 Description: Crash on NFS re-export node
1913 Details : calling clear_page() on the wrong pointer triggered oops in
1914 generic_mapping_read().
1917 Frequency : rarely, using O_DIRECT IO
1919 Description: unaligned directio crashes client with LASSERT
1920 Details : check for unaligned buffers before trying any requests.
1923 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1925 Description: stale data returned from RAID cache
1926 Details : If only a small amount of IO is done to the RAID device before
1927 reading it again it is possible to get stale data from the RAID
1928 cache instead of reading it from disk.
1931 Frequency : always for sles10 kernel
1933 Description: sles10 support
1934 Details : ll_follow_link: compile fixes and using of nd_set_link
1935 under newer kernels.
1938 Frequency : depends on arch, kernel and compiler version, always on sles10
1941 Description: recursive or deep enough symlinks cause stack overflow
1942 Details : getting rid of large stack-allocated variable in
1946 Frequency : depends on hardware
1948 Description: lustre write performance loss in the SLES10 kernel
1949 Details : the performance loss is caused by using of write barriers in the
1950 ext3 code. The SLES10 kernel turns barrier support on by
1951 default. The fix is to undo that change for ldiskfs.
1953 ------------------------------------------------------------------------------
1955 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1957 * Support for kernels:
1958 2.6.9-42.0.3EL (RHEL 4)
1959 2.6.5-7.276 (SLES 9)
1960 2.4.21-47.0.1.EL (RHEL 3)
1961 2.6.12.6 vanilla (kernel.org)
1965 Frequency : quota enabled and large files being deleted
1967 Description: releasing more than 4GB of quota at once hangs OST
1968 Details : If a user deletes more than 4GB of files on a single OST it
1969 will cause the OST to spin in an infinite loop. Release
1970 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1975 Description: statfs data retrieved from /proc may be stale or zero
1976 Details : When reading per-device statfs data from /proc, in the
1977 {kbytes,files}_{total,free,avail} files, it may appear
1978 as zero or be out of date.
1981 Frequency : systems with MD RAID1 external journal devices
1983 Description: lconf's call to blkid is confused by RAID1 journal devices
1984 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1985 of returning all block devices that match the journal UUID.
1988 Frequency : always, for aggregate stripe size over 4GB
1990 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1991 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1992 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1995 Frequency : always if "lfs find" used on a local file/directory
1997 Description: "lfs find" segfaults if used on a local file/directory
1998 Details : The case where a directory component was not specified wasn't
1999 handled correctly. Handle this properly.
2002 Frequency : always on ppc64
2004 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2005 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2006 perform endianness conversion.
2009 Frequency : rarely (truncate to non-zero file size after write under load)
2010 Bugzilla : 10730, 10687
2011 Description: Files padded with zeros to next 4K multiple
2012 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2013 it is possible that files that are truncated to a non-zero size
2014 immediately after being written are filled with zero bytes beyond
2015 the truncated size. No file data is lost.
2017 Severity : enhancement
2019 Description: Allow recovery/failover for liblustre clients.
2020 Details : liblustre clients were unaware of failover configurations until
2023 Severity : enhancement
2025 Description: user file locks should fail when not mounting with flock option
2026 Details : Set up an error-returning stub in ll_file_operations.lock field
2027 to prevent incorrect behaviour when client is mounted without
2028 flock option. Also, set up properly f_op->flock field for
2032 Frequency : always on ia64
2034 Description: "lfs df" loops on printing out MDS statfs information
2035 Details : The obd_ioctl_data was not initialized and in some systems
2036 this caused a failure during the ioctl that did not return
2037 an error. Initialize the struct and return an error on failure.
2040 Frequency : SLES 9 only
2042 Description: Error of copying files with lustre special EAs as root
2043 Details : Client side always return success for setxattr call for lustre
2044 special xattr (currently only "trusted.lov").
2047 Frequency : rarely on clusters with both ia64+i386 clients
2049 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2050 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2051 clients are doing concurrent writes to the same file it is
2052 possible that the ia64 clients may LASSERT because the OST
2053 extent locks are not PAGE_SIZE aligned. Ensure that grown
2054 locks are always aligned on the request boundary.
2057 Frequency : specific use, occasional
2059 Description: Overwriting in use executable truncates on-disk binary image
2060 Details : If one node attempts to overwrite an executable in use by
2061 another node, we now correctly return ETXTBSY instead of
2062 truncating the file.
2064 Severity : enhancement
2066 Description: Async OSC create to avoid the blocking unnecessarily.
2067 Details : If a OST has no remain object, system will block on the creating
2068 when need to create a new object on this OST. Now, ways use
2069 pre-created objects when available, instead of blocking on an
2070 empty osc while others are not empty. If we must block, we block
2071 for the shortest possible period of time.
2076 Description: chmod on Lustre root is propagated to other clients
2077 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2078 invalid by the follow_mount time.
2081 Frequency : liblustre clients only
2083 Description: Race in 'instant cancel' lock handling could lead to such locks
2084 never to be granted in case of SMP MDS
2085 Details : Do not destroy not yet granted but cbpending locks in
2089 Frequency : replay/resend of open
2091 Description: non null lock assetion failure in mds_intent_policy
2092 Details : Trying to replay/resend lockless open requests resulted in
2093 mds_open() returning 0 with no lock. Now it sets a flag if
2094 a lock is going to be returned.
2096 Severity : enhancement
2098 Description: Checksum enhancements
2099 Details : New checksum enhancements allow for resending RPCs that failed
2102 Severity : enhancement
2104 Description: Tunables on number of dirty pages in cacche
2105 Details : Allow to set limit on number of dirty pages cached.
2110 Description: client crash on unmount - lock still has references
2111 Details : In some error handling cases it was possible to leak a lock
2112 reference on a client while accessing a file. This was not
2113 harmful to the client during operation, but would cause the
2114 client to crash when the filesystem is unmounted.
2117 Frequency : specific case, rare
2119 Description: ETXTBSY on mds though file not in use by client
2120 Details : ETXTBSY is no longer incorrectly returned when attempting to
2121 chmod or chown a directory that the user previously tried to
2122 execute or a currently-executing binary.
2125 Frequency : extremely rare except on liblustre-based clients
2127 Description: Lustre space not freed when files are deleted
2128 Details : Clean up open-unlinked files after client eviction. Previously
2129 the unlink was skipped and the files remained as orphans.
2134 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2135 Details : In some cases it was possible to send a blocking callback to
2136 a client doing a glimpse, even though that client didn't get
2137 a lock granted. When the glimpse lock is cancelled on the OST
2138 the freed lock is left on the waiting list and corrupted the list.
2141 Frequency : all core dumps
2143 Description: Broke core dumps to lustre
2144 Details : Negative dentry may be unhashed if parent does not have UPDATE
2145 lock, but some callers, e.g. do_coredump, expect dentry to be
2146 hashed after successful create, hash it in ll_create_it.
2148 ------------------------------------------------------------------------------
2150 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2152 * Support for kernels:
2153 2.6.9-42.0.2.EL (RHEL 4)
2154 2.6.5-7.276 (SLES 9)
2155 2.4.21-40.EL (RHEL 3)
2156 2.6.12.6 vanilla (kernel.org)
2160 Frequency : always on RHEL 3
2162 Description: Number of open files grows over time
2163 Details : The number of open files grows over time, whether or not
2164 Lustre is started. This was due to a filp leak introduced
2165 by one of our kernel patches.
2167 ------------------------------------------------------------------------------
2169 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2171 * Support for kernels:
2172 2.6.9-42.EL (RHEL 4)
2173 2.6.5-7.276 (SLES 9)
2174 2.4.21-40.EL (RHEL 3)
2175 2.6.12.6 vanilla (kernel.org)
2180 Bugzilla : 5719, 9635, 9792, 9684
2181 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2182 Details : If a server is under heavy load and cannot reply to new
2183 connection requests before the client resends the (re)connect,
2184 the connection handling code can behave badly if two service
2185 threads are concurrently handing separate (re)connections from
2186 the same client. Add better locking to the connection handling
2187 code, and ensure that only a single connection will be processed
2188 for a given client UUID, even if the lock is dropped.
2190 Severity : enhancement
2192 Description: add TCP zero-copy support to kernel
2193 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2194 sends if the hardware supports scatter-gather and checksumming.
2195 This allows socklnd to do client-write and server-read more
2196 efficiently and reduce CPU utilization from skbuf copying.
2199 Frequency : only if NFS exporting from client
2201 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2202 Details : If setting attributes on a file created under NFS that had
2203 never been opened it would be possible to oops the client
2204 if the file had no objects.
2207 Frequency : always for liblustre
2209 Description: liblustre client does MDS+OSTs setattr RPC for each write
2210 Details : When doing a write from a liblustre client, the client
2211 incorrectly issued an RPC to the MDS and each OST the file was
2212 striped over in order to update the timestamps. When writing
2213 with small chunks and many clients this could overwhelm the MDS
2214 with RPCs. In all cases it would slow down the write because
2215 these RPCs are unnecessary.
2217 Severity : enhancement
2219 Description: allow number of MDS service threads to be changed at module load
2220 Details : It is now possible to change the number of MDS service threads
2221 running. Adding "options mds mds_num_threads={N}" to the MDS's
2222 /etc/modprobe.conf will set the number of threads for the next
2223 time Lustre is restarted (assuming the "mds" module is also
2224 reloaded at that time). The default number of threads will
2225 stay the same, 32 for most systems.
2230 Description: OST crash if filesystem is unformatted or corrupt
2231 Details : If an OST is started on a device that has never been formatted
2232 or if the filesystem is corrupt and cannot even mount then the
2233 error handling cleanup routines would dereference a NULL pointer.
2238 Description: NULL pointer deref in llap_from_page.
2239 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2240 incorrect) mapping if the page was truncated/reclaimed while it was
2241 searched for. Check for this condition and skip such pages when
2242 doing readahead. Introduce extra check to llap_from_page() to
2243 verify page->mapping->host is non-NULL (so page is not anonymous).
2246 Frequency : Sometimes when using sys_sendfile
2248 Description: "page not covered by a lock" warnings from ll_readpage
2249 Details : sendfile called ll_readpage without right page locks present.
2250 Now we introduced ll_file_sendfile that does necessary locking
2251 around call to generic_file_sendfile() much like we do in
2255 Frequency : with certain MDS communication failures at client mount time
2257 Description: NULL pointer deref after failed client mount
2258 Details : a client connection request may delayed by the network layer
2259 and not be sent until after the PTLRPC layer has timed out the
2260 request. If the client fails the mount immediately it will try
2261 to clean up before the network times out the request. Add a
2262 reference from the request import to the obd device and delay
2263 the cleanup until the network drops the request.
2266 Frequency : occasionally during client (re)connect
2268 Description: assertion failure during client (re)connect
2269 Details : processing a client connection request may be delayed by the
2270 client or server longer than the client connect timeout. This
2271 causes the client to resend the connection request. If the
2272 original connection request is replied in this interval, the
2273 client may trip an assertion failure in ptlrpc_connect_interpret()
2274 which thought it would be the only running connect process.
2277 Frequency : only with obd_echo servers and clients that are rebooted
2279 Description: kernel BUG accessing uninitialized data structure
2280 Details : When running an obd_echo server it did not start the ping_evictor
2281 thread, and when a client was evicted an uninitialized data
2282 structure was accessed. Start the ping_evictor in the RPC
2283 service startup instead of the OBD startup.
2285 Severity : enhancement
2286 Bugzilla : 10193 (patchless)
2287 Description: Remove dependency on various unexported kernel interfaces.
2288 Details : No longer need reparent_to_init, exit_mm, exit_files,
2289 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2292 Frequency : rare (only users of deprecated and unsupported LDAP config)
2294 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2295 Details : LDAP apparently contains 'lustreName' attributes instead of
2296 'name'. A simple remapping of the name is sufficient.
2299 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2301 Description: waiting_locks_callback trips kernel BUG if client is evicted
2302 Details : Running with the dump_on_timeout debug flag turned on makes
2303 it possible that the waiting_locks_callback() can try to dump
2304 the Lustre kernel debug logs from an interrupt handler. Defer
2305 this log dumping to the expired_lock_main() thread.
2307 Severity : enhancement
2309 Description: Support NFS exporting on 2.6 kernels.
2310 Details : Implement non-rawops metadata methods for NFS server to use without
2311 changing NFS server code.
2314 Frequency : very rare (synthetic metadata workload only)
2316 Description: two racing renames might cause an MDS thread to deadlock
2317 Details : Running the "racer" program may cause one MDS thread to rename
2318 a file from being the source of a rename to being the target of
2319 a rename at exactly the same time that another thread is doing
2320 so, and the second thread has already enqueued these locks after
2321 doing a lookup of the target and is trying to relock them in
2322 order. Ensure that we don't try to re-lock the same resource.
2325 Frequency : only very large systems with liblustre clients
2327 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2328 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2329 all OSTs in parallel. This allows the eviction of stale liblustre
2330 clients to proceed much faster than if they were done in series,
2331 and also offers similar improvements for other set_info RPCs.
2336 Description: excessive CPU usage during initial read phase on client
2337 Details : During the initial read phase on a client, it would agressively
2338 retry readahead on the file, consuming too much CPU and impacting
2339 performance (since 1.4.5.8). Improve the readahead algorithm
2340 to avoid this, and also improve some other common cases (read
2341 of small files in particular, where "small" is files smaller than
2342 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2347 Description: MDS crash when receiving packet with unknown intent.
2348 Details : Do not LBUG in unknown intent case, just return -EFAULT
2350 Severity : enhancement
2351 Bugzilla : 9293, 9385
2352 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2353 Details : Do not serialize getattr (non-intent version) and statfs.
2356 Frequency : occasional, when OST network is overloaded/intermittent
2358 Description: client evicted by OST after bulk IO timeout
2359 Details : If a client sends a bulk IO request (read or write) the OST
2360 may evict the client if it is unresposive to its data GET/PUT
2361 request. This is incorrect if the network is overloaded (takes
2362 too long to transfer the RPC data) or dropped the OST GET/PUT
2363 request. There is no need to evict the client at all, since
2364 the pinger and/or lock callbacks will handle this, and the
2365 client can restart the bulk request.
2368 Frequency : Always when mmapping file with no objects
2370 Description: client crashes when mmapping file with no objects
2371 Details : Check that we actually have objects in a file before doing any
2372 operations on objects in ll_vm_open, ll_vm_close and
2378 Description: Request leak when working with deleted CWD
2379 Details : Introduce advanced request refcount tracking for requests
2380 referenced from lustre intent.
2382 Severity : Enhancement
2384 Description: Cache open file handles on client.
2385 Details : MDS now will return special lock along with openhandle, if
2386 requested and client is allowed to hold openhandle, even if unused,
2387 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2388 closing files for every read/write openration.
2390 Severity : Enhancement
2392 Description: Cache open negative dentries on client when possible.
2393 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2394 negative dentries on lock revocation.
2399 Description: Remounting a client read-only wasn't possible with a zconf mount
2400 Details : It wasn't possible to remount a client read-only with llmount.
2402 Severity : enhancement
2403 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2404 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2405 MPICH is included to add Lustre-specific ADIO interfaces.
2406 This is based closely on the UFS ADIO layer and only differs
2407 in file creation, in order to allow the OST striping to be set.
2408 This is user-contributed code and not supported by CFS.
2413 Description: extended inode attributes (immutable, append-only) work improperly
2414 when 2.4 and 2.6 kernels are used on client/server or vice versa
2415 Details : Introduce kernel-independent values for these flags.
2417 Severity : enhancement
2420 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2421 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2422 granularity. Now, Fractional MB granularity is be supported,
2423 this is very useful for low memory system.
2425 Severity : enhancement
2427 Description: Getattr by fid
2428 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2429 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2430 have locks and avoids getattr rpc requests for it on every stat.
2433 Frequency : Always, for filesystems larger than 2TB
2435 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2436 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2437 filesystems larger than 2TB to be mounted with mballoc enabled.
2440 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2442 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2443 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2444 filesystems larger than 2TB is mounted then the kernel will
2445 silently corrupt the start of the filesystem. CONFIG_LBD is
2446 enabled for all CFS-supported kernels, but the possibility of
2447 this happening with a modified kernel config exists.
2449 Severity : enhancement
2451 Description: add client O_DIRECT support for 2.6 kernels
2452 Details : It is now possible to do O_DIRECT reads and writes to files
2453 in the Lustre client mountpoint on 2.6 kernel clients.
2455 Severity : enhancement
2457 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2458 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2459 parallel, not waiting for response from every OST before sending
2460 a rpc to the next OST.
2465 Description: setattr vs write race when updating file timestamps
2466 Details : Client processes that update a file timestamp into the past
2467 right after writing to the file (e.g. tar) it is possible that
2468 the updated file modification time can be reset to the current
2469 time due to a race between processing the setattr and write RPC.
2471 Severity : enhancement
2473 Description: Bring 'lfs find' closer in line with regular Linux find.
2474 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2475 -print0 options and obtains all the needed info through the lustre
2478 Severity : enhancement
2480 Description: support up to 1024 configured devices on one node
2481 Details : change obd_dev array from statically allocated to dynamically
2482 allocated structs as they are first used to reduce memory usage
2487 Description: Flush dirty partially truncated pages during truncate
2488 Details : Immediatelly flush partially truncated pages in filter_setattr,
2489 this way we completely avoid having any pages in page cache on OST
2490 and can retire ugly workarounds during writes to flush such pages.
2495 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2496 Details : protect inode from truncation within vfs_unlink() context
2497 just take a reference before calling vfs_unlink() and release it
2498 when parent's i_sem is free.
2503 Description: last_id value checked outside lock on OST caused LASSERT failure
2504 Details : If there were multiple MDS->OST object precreate requests in
2505 flight, it was possible that the OST's last object id was checked
2506 outside a lock and incorrectly tripped an assertion. Move checks
2507 inside locks, and discard old precreate requests.
2510 Frequency : always, if extents are used on OSTs
2512 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2513 Details : OSTs using the extents format would not zero the high 16 bits of
2514 the index physical block number. This is not a problem for any
2515 OST filesystems smaller than 16TB, and no kernels support ext3
2516 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2517 new/modified files) and can be fixed for existing filesystems
2518 with e2fsprogs-1.39-cfs1.
2523 Description: import connection selection may be incorrect if timer wraps
2524 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2525 import connections to be ignored if the 32-bit jiffies counter
2526 wraps. Use a 64-bit jiffies counter.
2529 Frequency : very large clusters immediately after boot
2531 Description: LNET request buffers exhausted under heavy short-term load
2532 Details : If a large number of client requests are generated on a service
2533 that has previously never seen so many requests it is possible
2534 that the request buffer growth cannot keep up with the spike in
2535 demand. Instead of dropping incoming requests, they are held in
2536 the LND until the RPC service can accept more requests.
2539 Frequency : Sometimes during replay
2541 Description: Assertion failure in ll_local_open after replay.
2542 Details : If replay happened on an open request reply before we were able
2543 to set replay handler, reply will become not swabbed tripping the
2544 assertion in ll_local_open. Now we set the handler right after
2545 recognising of open request
2548 Frequency : very rare
2550 Description: kernel reports "badness in vsnprintf"
2551 Details : Reading from the "recovery_status" /proc file in small chunks
2552 may cause a negative length in lprocfs_obd_rd_recovery_status()
2553 call to vsnprintf() (which is otherwise harmless). Exit early
2554 if there is no more space in the output buffer.
2556 Severity : enhancement
2558 Description: clear OBD RPC statistics by writing to them
2559 Details : It is now possible to clear the OBD RPC statistics by writing
2560 to the "stats" file.
2565 Description: Client mtime is not the same on different clients after utimes
2566 Details : In some cases, the client was using the utimes() syscall on
2567 a file cached on another node. The clients now validate the
2568 ctime from the MDS + OSTs to determine which one is right.
2573 Description: Inability to activate failout mode
2574 Details : lconf script incorrectly assumed that in pythong string's numeric
2575 value is used in comparisons.
2578 Frequency : always with multiple stripes per file
2580 Description: Inefficient object allocation for mutli-stripe files
2581 Details : When selecting which OSTs to stripe files over, for files with
2582 a stripe count that divides evenly into the number of OSTs,
2583 the MDS is always picking the same starting OST for each file.
2584 Return the OST selection heuristic to the original design.
2589 Description: mount failures may take full timeout to return an error
2590 Details : Under some heavy load conditions it is possible that a
2591 failed mount can wait for the full obd_timeout interval,
2592 possibly several minutes, before reporting an error.
2593 Instead return an error as soon as the status is known.
2595 Frequency : quota enabled and large files being deleted
2597 Description: releasing more than 4GB of quota at once hangs OST
2598 Details : If a user deletes more than 4GB of files on a single OST it
2599 will cause the OST to spin in an infinite loop. Release
2600 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2605 Description: statfs data retrieved from /proc may be stale or zero
2606 Details : When reading per-device statfs data from /proc, in the
2607 {kbytes,files}_{total,free,avail} files, it may appear
2608 as zero or be out of date.
2611 Frequency : systems with MD RAID1 external journal devices
2613 Description: lconf's call to blkid is confused by RAID1 journal devices
2614 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2615 of returning all block devices that match the journal UUID.
2618 Frequency : always, for aggregate stripe size over 4GB
2620 Description: assertion fails when trying to use 4GB stripe size
2621 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2622 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2625 Frequency : always on ppc64
2627 Description: the first write on an ext3 filesystem with mballoc got stuck
2628 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2629 perform endianness conversion.
2631 ------------------------------------------------------------------------------
2633 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2635 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2636 INTEROPERATE with older versions automatically. Please read the
2637 user documentation before upgrading any part of a live system.
2638 * WARNING: Lustre networking configuration changes are required with
2639 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2642 * Support for kernels:
2643 2.6.9-22.0.2.EL (RHEL 4)
2644 2.6.5-7.244 (SLES 9)
2645 2.6.12.6 vanilla (kernel.org)
2648 Severity : enhancement
2649 Bugzilla : 7981/8208
2650 Description: Introduced Lustre Networking (LNET)
2651 Details : LNET is new networking infrastructure for Lustre, it includes
2652 a reorganized network configuration mode (see the user
2653 documentation for full details) as well as support for routing
2654 between different network fabrics. Lustre Networking Devices
2655 (LNDs) for the supported network fabrics have also been
2656 created for this new infrastructure.
2658 Severity : enhancement
2659 Description: Introduced Access control lists
2660 Details : clients can set ACLs on files and directories in order to have
2661 more fine-grained permissions than the standard Unix UGO+RWX.
2662 The MDS must be started with the "-o acl" mount option.
2664 Severity : enhancement
2665 Description: Introduced filesystem quotas
2666 Details : Administrators may now establish per-user quotas on the
2669 Severity : enhancement
2671 Description: Configuration change for the XT3
2672 The PTLLND is now used to run Lustre over Portals on the XT3
2673 The configure option(s) --with-cray-portals are no longer used.
2674 Rather --with-portals=<path-to-portals-includes> is used to
2675 enable building on the XT3. In addition to enable XT3 specific
2676 features the option --enable-cray-xt3 must be used.
2681 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2682 Details : A race between allocating a new llcd and re-getting the llcd_lock
2683 allowed another thread to grab newly-allocated llcd.
2685 Severity : enhancement
2687 Description: 2.6 OST async journal commit and locking fix to improve performance
2688 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2689 2.6 kernels are now async as they already were in 2.4 kernels so
2690 that they can commit concurrently with the network bulk transfer.
2691 For block-allocated files the filter allocation semaphore is held
2692 to avoid filesystem fragmentation during allocation. BKL lock
2693 removed for 2.6 xattr operations where it is no longer needed.
2698 Description: lconf incorrectly determined whether two IP networks could talk
2699 Details : In some more complicated routing and multiple-network
2700 configurations, lconf will avoid trying to make a network
2701 connection to a disjoint part of the IP space. It was doing the
2702 math incorrectly for one set of cases.
2707 Description: Fix for potential infinite loop processing records in an llog.
2708 Details : If an llog record is corrupted/zeroed, it is possible to loop
2709 forever in llog_process(). Validate the llog record length
2710 and skip the remainder of the block on error.
2713 Frequency : occasional (liblustre only)
2715 Description: liblustre could not open files whose last component is a symlink
2716 Details : sysio_path_walk() would incorrectly pass the open intent to
2717 intermediate path components.
2720 Frequency : rare (liblustre only with non-standard tuning)
2721 Bugzilla : 7201 (7350)
2722 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2723 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2724 after it has been released, only do this for referenced locks
2726 Severity : enhancement
2728 Description: specifying an (invalid) directory default stripe_size of -1
2729 would reset the directory default striping
2730 Details : stripe_size -1 was used internally to signal directory stripe
2731 removal, now use "all default" to signal dir stripe removal
2732 as a directory striping of "all default" is not useful
2735 Frequency : common for large clusters running liblustre clients
2737 Description: doing an ls when liblustre clients are running is slow
2738 Details : sending a glimpse AST to a liblustre client waits for every AST
2739 to time out, as liblustre clients will not respond. Since they
2740 cannot cache data we refresh the OST lock LVB from disk instead.
2742 Severity : enhancement
2744 Description: doing an ls at the same time as file IO can be slow
2745 Details : enqueue and other "small" requests can be blocked behind many
2746 large IO requests. Create a new OST IO portal for non-IO
2747 requests so they can be processed faster.
2750 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2752 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2753 created file with mode 000
2754 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2755 allow an ftruncate() as a non-root user to a file with mode 000.
2756 Lustre now respects this flag to disable mode checks when
2757 truncating a file owned by the user
2760 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2763 Description: Revoking locks from clients that went dead or catatonic might take
2765 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2766 cancellation of such locks instant on servers without waiting for
2767 any reply from clients. Clients drops these locks when cancel
2768 notification from server is received without replying.
2771 Frequency : liblustre-only, when liblustre client dies or becomes busy
2773 Description: Doing ls on Linux clients can take a long time with active
2775 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2776 granting such locks to it in the first place if possible. Locks
2777 are taken by proxy on the OST during the read or write and
2778 dropped immediately afterward. Add connect flags handling, do
2779 not grant locks to liblustre clients for glimpse ASTs.
2781 Severity : enhancement
2783 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2784 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2785 sequential reads and behaves badly with applications doing random
2786 reads. Improve it by reading ahead at least read region, and
2787 avoiding excessive large RPC for small reads.
2789 Severity : enhancement
2791 Description: Creating more than 1000 files for a single job may cause a load
2792 imbalance on the OSTs if there are also a large number of OSTs.
2793 Details : qos_prep_create() uses an OST index reseed value that is an
2794 even multiple of the number of available OSTs so that if the
2795 reseed happens in the middle of the object allocation it will
2796 still utilize the OSTs as uniformly as possible.
2801 Description: OST or MDS may oops in ping_evictor_main()
2802 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2803 but doesn't restart at beginning of obd_exports_timed list
2806 Severity : enhancement
2808 Description: improve by-nid export eviction on the MDS and OST
2809 Details : allow multiple exports with the same NID to be evicted at one
2810 time without re-searching the exports list.
2813 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2815 Description: MDS may oops in groups_free()
2816 Details : in rare race conditions a newly allocated group_info struct is
2817 freed again, and this can be NULL. The 2.4 compatibility code
2818 for groups_free() checked for a NULL pointer, but 2.6 did not.
2821 Frequency : common for liblustre clients doing little filesystem IO
2822 Bugzilla : 9352, 7313
2823 Description: server may evict liblustre clients accessing contended locks
2824 Details : if a client is granted a lock or receives a completion AST
2825 with a blocking AST already set it would not reply to the AST
2826 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2829 Frequency : lfs setstripe, only systems with more than 160 OSTs
2831 Description: unable to set striping with a starting offset beyond OST 160
2832 Details : llapi_create_file() incorrectly limited the starting stripe
2833 index to the maximum single-file stripe count.
2836 Frequency : LDAP users only
2838 Description: lconf did not handle in-kernel recovery with LDAP properly
2839 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2841 Severity : enhancement
2843 Description: bind OST threads to NUMA nodes to improve performance
2844 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2845 node and do their allocations there to localize memory access
2847 Severity : enhancement
2849 Description: llmount can determine client NID directly from Myrinet (GM)
2850 Details : the client NID code from gmnalnid was moved directly into
2851 llmount, removing the need to use this or specifying the
2852 client NID explicitly when mounting GM clients with zeroconf
2855 Frequency : if client is started with down MDS
2857 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2858 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2859 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2860 ptlrpc_import_delayed_req() because we didn't check if the
2861 request was interrupted, and we also didn't break out of the
2862 event loop if there was no timeout
2867 Description: data loss during non-page-aligned writes to a single file from
2868 both multiple nodes and multiple threads on one node at same time
2869 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2870 inconsistency led to false short-reads, that were cached and later
2871 used by ->prepare_write() to fill in partially written page,
2872 leading to data loss.
2875 Frequency : always, if lconf --abort_recovery used
2877 Description: lconf --abort_recovery fails with 'Operation not supported'
2878 Details : lconf was attempting to abort recovery on the MDT device and not
2881 Severity : enhancement
2883 Description: remove cleanup logs
2884 Details : replace lconf-generated cleanup logs with lustre internal
2885 cleanup routines. Eliminates the need for client-cleanup and
2888 Severity : enhancement
2890 Description: add support for EAs (user and system) on lustre filesystems
2891 Details : it is now possible to store extended attributes in the Lustre
2892 client filesystem, and with the user_xattr mount option it
2893 is possible to allow users to store EAs on their files also
2895 Severity : enhancement
2897 Description: Add possibility (config option) to show minimal available OST free
2899 Details : When compiled with --enable-mindf configure option, statfs(2)
2900 (and so, df) will return least minimal free space available from
2901 all OSTs as amount of free space on FS, instead of summary of
2902 free spaces of all OSTs.
2904 Severity : enhancement
2906 Description: do not expand extent locks acquired on OST-side
2907 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2908 by server: they are not cached anyway.
2911 Frequency : when mmap is used/binaries executed from Lustre
2913 Description: Unmmap pages before throwing them away from read cache.
2914 Details : llap_shrink cache now attempts to unmap pages before discarding
2915 them (if unmapping failed - do not discard). SLES9 kernel has
2916 extra checks that trigger if this unmapping is not done first.
2921 Description: lconf didn't resolve symlinks before checking to see whether a
2922 given mountpoint was already in use
2925 Frequency : when migrating failover services
2926 Bugzilla : 6395, 9514
2927 Description: When migrating a subset of services from a node (e.g. failback
2928 from a failover service node) the remaining services would
2929 time out and evict clients.
2930 Details : lconf --force (implied by --failover) sets the global obd_timeout
2931 to 5 seconds in order to quickly disconnect, but this caused
2932 other RPCs to time out too quickly. Do not change the global
2933 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2935 Severity : enhancement
2936 Frequency : if MDS is started with down OST
2937 Bugzilla : 9439,5706
2938 Description: Allow startup/shutdown of an MDS without depending on the
2939 availability of the OSTs.
2940 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2941 Add appropriate locking and lov-osc refcounts for safe
2942 cleaning. Add osc abort_inflight calls in case the
2943 synchronize never started.
2946 Frequency : occasional (Cray XT3 only)
2948 Description: root not authorized to access files in CRAY_PORTALS environment
2949 Details : The client process capabilities were not honoured on the MDS in
2950 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2951 been accessed by an authorized user then root was able to access
2952 the file on the local client also. The root user capabilities
2953 are now allowed on the MDS, as this environment has secure UID.
2956 Frequency : occasional
2958 Description: ldiskfs "too long searching" message happens too often
2959 Details : A debugging message (otherwise harmless) prints too often on
2960 the OST console. This has been reduced to only happen when
2961 there are fragmentation problems on the filesystem.
2966 Description: Division by zero in statfs when all OSCs are inactive
2967 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2968 lov_statfs divides by value returned by lov_get_stripecnt().
2972 Bugzilla : 9489, 3273
2973 Description: First write from each client to each OST was only 4kB in size,
2974 to initialize client writeback cache, which caused sub-optimal
2975 RPCs and poor layout on disk for the first writen file.
2976 Details : Clients now request an initial cache grant at (re)connect time
2977 and so that they can start streaming writes to the cache right
2978 away and always do full-sized RPCs if there is enough data.
2979 If the OST is rebooted the client also re-establishes its grant
2980 so that client cached writes will be honoured under the grant.
2985 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2986 Details : Now I/O RPCs go to different portal number and (presumably) fast
2987 lock requests (and glimses) and other RPCs get their own service
2988 threads pool that should be able to service those RPCs
2991 Severity : enhancement
2993 Description: Ability to exchange lustre version between client and servers and
2994 issue warnings at client side if client is too old. Also for
2995 liblustre clients there is ability to refuse connection of too old
2997 Details : New 'version' field is added to connect data structure that is
2998 filled with version info. That info is later checked by server and
3002 Frequency : rare, liblustre only.
3003 Bugzilla : 9296, 9581
3004 Description: Two simultaneous writes from liblustre at offset within same page
3005 might proceed at the same time overwriting eachother with stale
3007 Details : I/O lock withing llu_file_prwv was released too early, before data
3008 actually was hitting the wire. Extended lock-holding time until
3009 server acknowledges receiving data.
3012 Frequency : extremely rare. Never observed in practice.
3014 Description: avoid generating lustre_handle cookie of 0.
3015 Details : class_handle_hash() generates handle cookies by incrementing
3016 global counter, and can hit 0 occasionaly (this is unlikely, but
3017 not impossible, because initial value of cookie counter is
3018 selected randonly). Value of 0 is used as a sentinel meaning
3019 "unassigned handle" --- avoid it. Also coalesce two critical
3020 sections in this function into one.
3022 Severity : enhancement
3024 Description: allow liblustre clients to delegate truncate locking to OST
3025 Details : To avoid overhead of locking, liblustre client instructs OST to
3026 take extent lock in ost_punch() on client's behalf. New connection
3027 flag is added to handle backward compatibility.
3029 Severity : enhancement
3030 Bugzilla : 4928, 7341, 9758
3031 Description: allow number of OST service threads to be specified
3032 Details : a module parameter allows the number of OST service threads
3033 to be specified via "options ost ost_num_threads={N}" in the
3034 OSS's /etc/modules.conf or /etc/modprobe.conf.
3038 Bugzilla : 6146, 9635, 9895
3039 Description: servers crash with bad pointer in target_handle_connect()
3040 Details : In rare cases when a client is reconnecting it was possible that
3041 the connection request was the last reference for that export.
3042 We would temporarily drop the export reference and get a new
3043 one, but this may have been the last reference and the export
3044 was just destroyed. Get new reference before dropping old one.
3046 Severity : enhancement
3047 Frequency : if client is started with failover MDS
3049 Description: Allow multiple MDS hostnames in the mount command
3050 Details : Try to read the configuration from all specified MDS
3051 hostnames during a client mount in case the "primary"
3054 Severity : enhancement
3056 Description: Stop sending data to evicted clients as soon as possible.
3057 Details : Check if the client we are about to send or are sending data to
3058 was evicted already. (Check is done every second of waiting,
3059 for which l_wait_event interface was extended to allow checking
3060 of exit condition at specified intervals).
3063 Frequency : rare, normally only when NFS exporting is done from client
3065 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3067 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3070 Frequency : always, if srand() is called before liblustre initialization
3072 Description: Liblustre uses system PRNG disturbing its usage by user application
3073 Details : Introduce internal to lustre fast and high-quality PRNG for
3074 lustre usage and make liblustre and some other places in generic
3075 lustre code to use it.
3077 Severity : enhancement
3078 Bugzilla : 9477, 9557, 9870
3079 Description: Verify that the MDS configuration logs are updated when xml is
3080 Details : Check if the .xml configuration logs are newer than the config
3081 logs stored on the MDS and report an error if this is the case.
3082 Request --write-conf, or allow starting with --old_conf.
3084 Severity : enhancement
3086 Description: Handle symlinks in the path when checking if Lustre is mounted.
3087 Details : Resolve intermediate symlinks when checking if a client has
3088 mounted a filesystem to avoid duplicate client mounts.
3093 Description: lconf can hit an error exception but still return success.
3094 Details : The lconf command catches the Command error exception at the top
3095 level script context and will exit with the associated exit
3096 status, but doesn't ensure that this exit status is non-zero.
3101 Description: failure of ptlrpc thread startup can cause oops
3102 Details : Starting a ptlrpc service thread can fail if there are a large
3103 number of threads or the server memory is very fragmented.
3104 Handle this without oopsing.
3107 Frequency : always, only if liblustre and non-default acceptor port was used
3109 Description: liblustre cannot connect to servers with non-default acceptor port
3110 Details : tcpnal_set_default_params() was not called and was therefore
3111 ignoring the environment varaible TCPNAL_PORT, as well as other
3112 TCPNAL_ environment variables
3117 Description: two objects could be created on the same OST for a single file
3118 Details : If an OST is down, in some cases it was possible to create two
3119 objects on a single OST for a single file. No problems other
3120 than potential performance impact and spurious error messages.
3124 Bugzilla : 5681, 9562
3125 Description: Client may oops in ll_unhash_aliases
3126 Details : Client dcache may become inconsistent in race condition.
3127 In some cases "getcwd" can fail if the current directory is
3133 Description: Inode refcounting problems in NFS export code
3134 Details : link_raw functions used to call d_instantiate without obtaining
3135 extra inode reference first.
3139 Bugzilla : 9942, 9903
3140 Description: Referencing freed requests leading to crash, memleaks with NFS.
3141 Details : We used to require that call to ll_revalidate_it was always
3142 followed by ll_lookup_it. Also with revalidate_special() it is
3143 possible to call ll_revalidate_it() twice for the same dentry
3144 even if first occurence returned success. This fix changes semantic
3145 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3146 reference on a request referred from the intent.
3147 ll_intent_release() then releases such a request.
3150 Frequency : rare, normally benchmark loads only
3152 Description: unlinked inodes were kept in memory on the client
3153 Details : If a client is repeatedly creating and unlinking files it
3154 can accumulate a lot of stale inodes in the inode slab cache.
3155 If there is no other client load running this can cause the
3156 client node to run out of memory. Instead flush old inodes
3157 from client cache that have the same inode number as a new inode.
3160 Frequency : SLES9 2.6.5 kernel and long filenames only
3161 Bugzilla : 9969, 10379
3162 Description: utime reports stale NFS file handle
3163 Details : SLES9 uses out-of-dentry names in some cases, which confused
3164 the lustre dentry revalidation. Change it to always use the
3168 Frequency : rare, unless heavy write-truncate concurrency is continuous
3169 Bugzilla : 4180, 6984, 7171, 9963, 9331
3170 Description: OST becomes very slow and/or deadlocked during object unlink
3171 Details : filter_destroy() was holding onto the parent directory lock
3172 while truncating+unlinking objects. For very large objects this
3173 may block other threads for a long time and slow overall OST
3174 responsiveness. It may also be possible to get a lock ordering
3175 deadlock in this case, or run out of journal credits because of
3176 the combined truncate+unlink. Solution is to do object truncate
3177 first in one transaction without parent lock, and then do the
3178 final unlink in a new transaction with the parent lock. This
3179 reduces the lock hold time dramatically.
3182 Frequency : rare, 2.4 kernels only
3184 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3185 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3186 before unmounting their filesystem in order to avoid deadlock.
3187 The kernel_locked() function in 2.4 kernels only checks whether
3188 the kernel lock is held, not whether it is this process that is
3189 holding it as 2.6 kernels do.
3194 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3195 Details : The client ptlrpc code may be trying to reconnect to a down
3196 server before a previous connection attempt has timed out.
3197 Increase the reconnect interval to be longer than the connection
3198 timeout interval to avoid sending duplicate connections to
3202 Frequency : echo_client brw_test command
3204 Description: fix echo_client to work with OST preallocated code
3205 Details : OST preallocation code (5137) didn't take echo_client IO path
3206 into account: echo_client calls filter methods outside of any
3207 OST thread and, hence, there is no per-thread preallocated
3208 pages and buffers to use. Solution: hijack pga pages for IO. As
3209 a byproduct, this avoids unnecessary data copying.
3213 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3214 Description: Client can oops in mdc_commit_close() after open replay
3215 Details : It was possible for the MDS to return an open request with no
3216 transaction number in mds_finish_transno() if the client was
3217 evicted, but without actually returning an error. Clients
3218 would later try to replay that open and may trip an assertion
3219 Simplify the client close codepath, and always return an error
3220 from the MDS in case the open is not successful.
3223 Frequency : rare, 2.6 OSTs only
3225 Description: OST may deadlock under high load on fragmented files
3226 Details : If there was a heavy load and highly-fragmented OST filesystems
3227 it was possible to have all the OST threads deadlock waiting on
3228 allocation of biovecs, because the biovecs were not released
3229 until the entire RPC IO was completed. Instead, release biovecs
3230 as soon as they are complete to ensure forward IO progress.
3232 Severity : enhancement
3234 Description: Support for specifying external journal device at mount
3235 Details : If an OST or MDS device is formatted with an external journal
3236 device, this device major/minor is stored in the ext3 superblock
3237 and may not be valid for failover. Allow detecting and
3238 specifying the external journal at mount time.
3243 Description: Mounting an MDS with pending unlinked files may cause oops
3244 Details : target_finish_recovery() calls mds_postrecov() which returned
3245 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3246 considers this an error and immediately begins cleaning up the
3247 lov, just after starting the mds_lov process
3249 Severity : enhancement
3251 Description: Implement 'lfs df' to report actual free space on per-OST basis
3252 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3253 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3254 usage of objects; '-h' to report in human readable format.
3256 ------------------------------------------------------------------------------
3258 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3265 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3266 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3267 at startup during orphan unlinking. Instead, skip buddy update
3268 before setup, as it will be regenerated after recovery is complete.
3273 Description: If an OST is inactive, its locks might reference stale inodes.
3274 Details : lov_change_cbdata() must iterate over all namespaces, even if
3275 they are inactive to clear inode references from the lock.
3277 Severity : enhancement
3278 Frequency : occasional, if non-standard max_dirty_mb used
3280 Description: Client will block write RPCs if not enough grant
3281 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3282 then the client will block writes while waiting for another RPC
3283 to complete instead of consuming its dirty limit. With change
3284 we get improved performance when max_dirty_mb is small.
3286 Severity : enhancement
3287 Bugzilla : 3389, 6253
3288 Description: Add support for supplementary groups on the MDS.
3289 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3290 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3291 lookups for user supplementary groups into a cache.
3295 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3296 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3297 for liblustre O_EXCL is enforced by MDS.
3301 Description: readdir never returns NULL in liblustre.
3302 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3303 dentry in struct dirent.
3307 Frequency : liblustre only, depends on application IO pattern
3308 Description: liblustre clients evicted if not contacting servers
3309 Details : Don't put liblustre clients into the ping_evictor list, so
3310 they will not be evicted by the pinger ever.
3312 Severity : enhancement
3314 Description: Add ability to evict clients by NID from MDS.
3315 Details : By echoing "nid:$NID" string into
3316 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3317 $NID would be instantly evicted from this MDS and from all active
3318 OSTs connected to it.
3322 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3323 Details : lookup_it_finish() used to query file size from OSTs that was not
3328 Description: service threads change working directory to that of init
3329 Details : Starting lustre service threads may pin the working directory
3330 of the parent thread, making that filesystem busy. Threads
3331 now change to the working directory of init to avoid this.
3335 Frequency : during shutdown only
3336 Description: shutdown with a failed MDS or OST can cause unmount to hang
3337 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3342 Frequency : 2.6 only
3343 Description: chmod/chown may include an extra supplementary group
3344 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3345 supplementary group and if none is specified this is used.
3348 Bugzilla : 5479 (6816)
3350 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3351 Details : If lookup is in progress on a file that is unlinked we might try
3352 to revalidate the inode and fail in revalidate after lookup is
3353 complete and ll_file_open() enqueues the open again but
3354 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3357 Frequency : always, if lconf --abort_recovery used
3359 Description: lconf --abort_recovery fails with 'Operation not supported'
3360 Details : lconf was attempting to abort recovery on the MDT device and not
3363 ------------------------------------------------------------------------------
3365 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3370 Frequency : rare (only unsupported configurations with a node running as an
3372 Bugzilla : 6514, 5137
3373 Description: Mounting a Lustre file system on a node running as an OST could
3375 Details : OSTs now preallocates memory needed to write out data at
3376 startup, instead of when needed, to avoid having to
3377 allocate memory in possibly low memory situations.
3378 Specifically, if the file system is mounted on on OST,
3379 memory pressure could force it to try to write out data,
3380 which it needed to allocate memory to do. Due to the low
3381 memory, it would be unable to do so and the node would
3382 become unresponsive.
3384 Severity : enhancement
3386 Description: Addition of lconf --service command line option
3387 Details : lconf now accepts a '--service <arg>' option, which is
3388 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3390 Severity : enhancement
3392 Description: Failover mode is now the default for OSTs.
3393 Details : By default, OSTs will now run in failover mode. To return to
3394 the old behaviour, add '--failout' to the lmc line for OSTs.
3396 Severity : enhancement
3398 Description: Health checks are now provided for MDS and OSTs
3399 Details : Additional detailed health check information on MSD and OSTs
3400 is now provided through the procfs health_check value.
3403 Frequency : occasional, depends on IO load
3405 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3406 after numerous create/delete cycles
3407 Details : The ext3 inode allocation policy would not allocate new inodes
3408 very well on the OSTs because there are no new directories
3409 being created. Instead we look for groups with free space if
3410 the parent directories are nearly full.
3415 Description: Network or server problems during mount may cause partially
3416 mounted clients instead of returning an error.
3417 Details : The config llog parsing code may overwrite the error return
3418 code during mount error handling, returning success instead
3424 Description: MDS can fail to allocate large reply buffers
3425 Details : After long uptimes the MDS can fail to allocate large reply
3426 buffers (e.g. zconf client mount config records) due to memory
3427 fragmentation or consumption by the buffer cache. Preallocate
3428 some large reply buffers so that these replies can be sent even
3429 under memory pressure.
3433 Frequency : rare (liblustre)
3434 Description: fsx running with liblustre complained that using truncate() to
3435 extend the file doesn't work. This patch corrects that issue.
3436 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3437 ATTR_SIZE and lsm use in llu_setattr_raw.
3441 Frequency : rare, only 2.6 kernels
3442 Description: Unusual file access patterns on the MDS may result in inode
3443 data being lost in very rare circumstances.
3444 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3445 code in the 2.6 kernel caused the inode and/or EA data not to
3446 be read from disk, causing single-file corruption.
3450 Frequency : rare, only 2.6 filesystems using extents
3451 Description: Heavy concurrent write and delete load may cause data corruption.
3452 Details : It was possible under high-load situations to have an extent
3453 metadata block in the block device cache from a just-unlinked
3454 file overwrite a newly-allocated data block. We now unmap any
3455 metadata buffers that alias just-allocated data blocks.
3459 Frequency : filesystems with default stripe_count larger than 77
3460 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3461 Details : lconf specifies an inode size of 4096 bytes when the default
3462 stripe_count is larger than 77. This conflicts with the default
3463 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3464 this case to avoid pinning too much memory for large EAs.
3466 ------------------------------------------------------------------------------
3468 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3473 Frequency : rare (extremely heavy IO load with hundreds of clients)
3475 Description: Client is evicted, gets IO error writing to file
3476 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3477 caused clients to be evicted for AST timeouts. The fixes in
3478 bug 5192 mean we no longer need to have such short AST timeouts
3479 so ldlm_timeout has been increased.
3482 Frequency : occasional during --force or --failover shutdown under load
3483 Bugzilla : 5949, 4834
3484 Description: Server oops/LBUG if stopped with --force or --failover under load
3485 Details : a collection of import/export refcount and cleanup ordering
3486 issues fixed for safer force cleanup
3489 Frequency : only filesystems larger than 120 OSTs
3490 Bugzilla : 5990, 6223
3491 Description: lfs getstripe would oops on a very large filesystem
3492 Details : lov_getconfig used kfree on vmalloc'd memory
3495 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3496 Bugzilla : 6242, 6243
3497 Description: reading from files that had been truncated to a non-zero size
3498 but never opened returned no data
3499 Details : ll_file_read() reads zeros from no-object files to EOF
3504 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3505 Details : There's little harm in aborting MDS/OSS recovery and letting it
3506 try again, so I removed the LASSERT and return an error instead.
3508 Severity : enhancement
3510 Description: New debugging infrastructure for tracking down data corruption
3511 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3512 (b) cover more of the client-side code path, and (c) try to narrow
3513 down where problems occurred
3517 Bugzilla : 3819, 4364, 4397, 6313
3518 Description: Racing close and eviction MDS could cause assertion in mds_close
3519 Details : It was possible to get multiple mfd references during close and
3520 client eviction, leading to one thread referencing a freed mfd.
3522 Severity: : enhancement
3523 Bugzilla : 3262, 6359
3524 Description: Attempts to reconnect to servers are now more aggressive.
3525 Details : This builds on the enhanced upcall-less recovery that was added
3526 in 1.4.2. When trying to reconnect to servers, clients will
3527 now try each server in the failover group every 10 seconds. By
3528 default, clients would previously try one server every 25 seconds.
3533 Description: After recovery, certain operations trigger a failed
3534 assertion on a client.
3535 Details : Failing over an mds, using lconf -d --failover, while a
3536 client was doing a readdir() call would cause the client to
3537 LBUG after recovery completed and the readdir() was resent.
3539 Severity : enhancement
3541 Description: Default groups are now added by lconf
3542 Details : You can now run lconf --group <servicename> without having to
3543 manually add groups with lmc.
3546 Frequency : occasional
3548 Description: Nodes with an elan id of 0 trigger a failed assertion
3551 Frequency : always when accessing e.g. tty/console device nodes
3553 Description: tty and some other devices nodes cannot be used on lustre
3554 Details : file's private_data field is used by device data and lustre
3555 values in there got lost. New field was added to struct file to
3556 store fs-specific private data.
3559 Frequency : when exporting Lustre via NFS
3561 Description: NFSD failed occasionally when looking up a path component
3562 Details : NFSD is looking up ".." which was broken in ext3 directories
3563 that had grown large enough to become hashed.
3566 Frequency : Clusters with multiple interfaces not on the same subnet
3568 Description: Nodes will repeatedly try to reconnect to an interface which it
3569 cannot reach and report an error to the log.
3570 Details : Extra peer list entries will be created by lconf with some peers
3571 unreachable. lconf now validates the peer before adding it.
3574 Frequency : Only if a default stripe is set on the filesystem root.
3576 Description: Setting a default stripe on the filesystem root prevented the
3577 filesystem from being remounted.
3578 Details : The client was sending extra request flags in the root getattr
3579 request and did not allocate a reply buffer for the dir EA.
3582 Frequency : occasional, higher if lots of files are accessed by one client
3583 Bugzilla : 6159, 6097
3584 Description: Client trips assertion regarding lsm mismatch/magic
3585 Details : While revalidating inodes the VFS looks up inodes with ifind()
3586 and in rare cases can find an inode that is being freed.
3587 The ll_test_inode() code will free the lsm during ifind()
3588 when it finds an existing inode and then the VFS later attaches
3589 this free lsm to a new inode.
3593 Bugzilla : 6422, 7030
3594 Description: MDS deadlock between mkdir and client eviction
3595 Details : Creating a new file via mkdir or mknod (starting a transaction
3596 and getting the ns lock) can deadlock with client eviction
3597 (gets ns lock and trying to finish a synchronous transaction).
3600 Frequency : occasional
3601 Description: While starting a server, the fsfilt_ext3 module could not be
3603 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3604 kernels. Previously, lconf would still use the ext3 name
3605 when trying to load modules. Now, it will correctly use
3606 ext3 on 2.4 and ldiskfs on 2.6.
3608 Severity : enhancement
3609 Description: The default stripe count has been changed to 1
3610 Details : The interpretation of the default stripe count (0, to lfs
3611 or lmc) has been changed to mean striping across a single
3612 OST, rather than all available. For general usage we have
3613 found a stripe count of 1 or 2 works best.
3615 Severity : enhancement
3616 Description: Add support for compiling against Cray portals.
3617 Details : Conditional compiling for some areas that are different
3621 Frequency : occasional
3622 Bugzilla : 6409, 6834
3623 Description: Creating files with an explicit stripe count may lead to
3624 a failed assertion on the MDS
3625 Details : If some OSTs are full or unavailable, creating files may
3626 trigger a failed assertion on the MDS. Now, Lustre will
3627 try to use other servers or return an error to the
3631 Frequency : occasional
3633 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3634 caused lock timeout during readahead (since 1.4.2).
3635 Details : Processes doing readahead might match a lock that hasn't been
3636 granted yet if there are overlapping and conflicting lock
3637 requests. The readahead process waits on ungranted lock
3638 (original lock is CBPENDING), while OST waits for that process
3639 to cancel CBPENDING read lock and eventually evicts client.
3641 Severity : enhancement
3643 Description: Initial enabling of flock support for clients
3644 Details : Implements fcntl advisory locking and file status functions.
3645 This feature is provided as an optional mount flag (default
3646 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3647 locking are implemented yet, and those that are are not guaranteed
3648 to be completely correct in production environments.
3649 mount -t lustre -o [flock|noflock] ...
3652 Frequency : occasional
3654 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3655 assertion in the ext3 JBD (journaling) layer.
3656 Details : The b_committed_data struct is protected by the big kernel lock
3657 in 2.4 kernels, serializing journal_commit_transaction() and
3658 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3659 access to this struct. In 2.6 kernels there is finer grained
3660 locking to improve SMP performance of the JBD layer.
3664 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3669 Description: merge and cleanup kernel patches.
3672 -----------------------------------------------------------------------------
3674 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3676 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3677 versions of Lustre. Please update all servers and clients to
3678 version 1.4.2 or later at the same time. You must also run
3679 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3680 to update the configuration logs.
3682 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3683 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3684 - swab llog records of type '0' so we get proper header size/idx (5861)
3685 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3686 - fix rename of one directory over another leaking an inode (5953)
3687 - avoid SetPageDirty on 2.6 (5981)
3688 - don't re-add just-being-destroyed locks to the waiting list (5653)
3689 - when creating new directories, inherit the parent's custom
3690 striping settings if present parent (3048)
3691 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3692 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3693 - don't hold client locks on temporary worklist from l_lru (5666)
3694 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3695 - automatically evict dead clients (5921)
3696 - Update file size properly in create+truncate+fstat case (6196)
3697 - Do not unhash mountpoint dentries, do not allow removal of
3699 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3700 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3701 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3702 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3703 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3704 - Correctly handle reads of files with no objects (6243)
3705 - lctl recover will also mark a device active if deactivate used (5933)
3707 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3708 - allow --write-conf on an MDS with different nettype than client (5619)
3709 - don't write config llogs to MDS for mounts not from that MDS (5617)
3710 - lconf should create multiple TCP connections from a client (5201)
3711 - init scripts are now turned off by default; run chkconfig --on
3712 lustre and chkconfig --on lustrefs to use them
3713 - upcalls are no longer needed for clients to recover to failover
3715 - add --abort-recovery option to lconf to abort recovery on device
3717 - add support for an arbitrary number of OSTs (3026)
3718 - Quota support protocol changes.
3719 - forward compatibility changes to wire structs (6007)
3720 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3721 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3722 - improved "lustre status" script
3723 - initialize blocksize for non-regular files (6062)
3724 - added --disable-server and --disable-client configure options (5782)
3725 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3726 - Vanilla 2.4.29 support
3727 - increase maximum number of obd devices to 520 (6242)
3728 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3729 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3730 SLES 9 2.6.5 kernels
3731 - limit stripes per file to 160 (the maximum EA size) (6093)
3733 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3736 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3737 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3738 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3739 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3740 - small ext3 extents cleanups and fixes (5733)
3741 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3742 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3743 - avoid needless client->OST connect, fix handle mismatch (5317)
3744 - fix DLM error path that led to out-of-sync client, long delays (5779)
3745 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3746 - fix several locking issues related to i_size (5492,5624,5654,5672)
3747 - don't move pending lock onto export if it is already evicted (5683)
3748 - fix kernel oops when creating .foo in unlinked directory (5548)
3749 - fix deadlock in obdfilter statistics vs. object create (5811)
3750 - use time_{before,after} to avoid timer jiffies wrap (5882)
3751 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3752 - Do not leak request if server was not able to process it (5154)
3753 - If mds_open unable to find parent dir, make that negative lookup(5154)
3754 - don't create new directories with extent-mapping (5909, 5936)
3756 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3757 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3758 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3759 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3760 - support for RHEL4 kernel on i686 (5773)
3761 - provide error messages when incompatible logs are encountered (5898)
3763 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3764 * version 1.4.0.10 (1.4.1 release candidate 1)
3766 - don't keep a lock reference when lock is not granted (4238)
3767 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3768 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3769 - drop import inflight refcount on signal_completed_replay error (5255)
3770 - unlock page after async write error during send (3677)
3771 - handle missing objects in filter_preprw_read properly (5265)
3772 - no transno return for symlink open, don't save no-trasno open (3440)
3773 - don't try to complete elan receive that already failed (4012)
3774 - free RPC server reply state on error (5406)
3775 - clean up thread from ptlrpc_start_thread() on error (5160)
3776 - readahead could read extra page into cache that wasn't ejected (5388)
3777 - prevent races in class_attach/setup/cleanup/detach (5260)
3778 - don't dereference de->d_inode after l_dput of de (5458)
3779 - use "int" for stripe value returned from lock_to_stripe (5544)
3780 - mballoc allocation and error-checking fixes in 2.6 (5504)
3781 - block device patches to fix I/O request sizes in 2.6 (5482)
3782 - look up hostnames for IB nals (5602)
3783 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3784 - don't start multiple acceptors for the same port (5277)
3785 - fix incorrect LASSERT in mds_getattr_name (5635)
3786 - export a proc file for general "ping" checking (5628)
3787 - fix "lfs check" to not block when the MDS is down (5628)
3789 - service request history (4965)
3790 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3791 - create an "evict_client" /proc entry on OSTs, like the MDS has
3792 - fix mount usage message, return errors per mount(8) (5168)
3793 - change grep [] to grep "[]" in tests so they work in more UMLs
3794 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3795 - remove ancient LOV_MAGIC_V0 EA support (5047)
3796 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3797 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3798 - allow readahead allocations to fail when low on memory (5383)
3799 - mmap locking landed again, after considerable improvement (2828)
3800 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3802 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3805 - send OST transaction number in read/write reply to free req (4966)
3806 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3807 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3808 - the watchdog thread now runs as interruptible (5246)
3809 - flock/lockf fixes (but it's still disabled, pending 5135)
3810 - don't use EXT3 constants in llite code (5094)
3811 - memory shortage at startup could cause assertion (5176)
3813 - reorganization of lov code
3814 - single portals codebase
3816 - add extents/mballoc support (5025)
3817 - direct I/O reads in the obdfilter (4048)
3818 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3820 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3823 - send OST transaction number in read/write reply to free req (4966)
3824 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3825 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3826 - the watchdog thread now runs as interruptible (5246)
3827 - handle missing objects in filter_preprw_read properly (5265)
3828 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3829 - drop import inflight refcount on signal_completed_replay error (5255)
3830 - unlock page after async write error during send (3677)
3831 - return original error code on reconstructed replies (3761)
3832 - no transno return for symlink open, don't save no-trasno open (3440)
3834 - add pid to ldlm debugging output (4922)
3835 - bump the watchdog timeouts -- we can't handle 30sec yet
3836 - extra debugging for orphan dentry/inode bug (5259)
3838 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3841 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3842 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3843 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3844 - fix NULL dereference in /proc/sys/portals/routes (4827)
3845 - allow failed mdc_close() operations to be interrupted (4561)
3846 - stop precreate on OST before MDS would time out on it (4778)
3847 - don't send partial-page writes before EOF from client (4410)
3848 - discard client grant for sub-page writes on large-page clients (4520)
3849 - don't free dentries not owned by NFS code, check generation (4806)
3850 - fix lsm leak if mds_create_objects() fails (4801)
3851 - limit debug_daemon file size, always print CERROR messages (4789)
3852 - use transno after validating reply (3892)
3853 - process timed out requests if import state changes (3754)
3854 - update mtime on OST during writes, return in glimpse (4829)
3855 - add mkfsoptions to LDAP (4679)
3856 - use ->max_readahead method instead of zapping global ra (5039)
3857 - don't interrupt __l_wait_event() during strace
3859 - add software watchdogs to catch hung threads quickly (4941)
3860 - make lustrefs init script start after nfs is mounted
3861 - fix CWARN/ERROR duplication (4930)
3862 - return async write errors to application if possible (2248)
3863 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3864 - print NAL number in %x format (4645)
3865 - update barely-supported suse-2.4.21-171 series (4842)
3866 - support for sles 9 %post scripts
3867 - support for building 2.6 kernel-source packages
3868 - support for sles km_* packages
3870 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3873 - ignore -ENOENT errors in osc_destroy (3639)
3874 - notify osc create thread that OSC is being cleaned up (4600)
3875 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3876 - reconstruct ost_handle() like mds_handle() (4657)
3877 - create a new thread to do import eviction to avoid deadlock (3969)
3878 - let lconf resolve symlinked-to devices (4629)
3879 - don't unlink "objects" from directory with default EA (4554)
3880 - hold socknal file ref over connect in case target is down (4394)
3881 - allow more than 32000 subdirectories in a single directory (3244)
3882 - fix blocks count for O_DIRECT writes (3751)
3883 - OST returns ENOSPC from object create when no space left (4539)
3884 - don't send truncate RPC if file size isn't changing (4410)
3885 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3886 - bind to privileged port in socknal and tcpnal (3689)
3888 - rate limit CERROR/CWARN console message to avoid overload (4519)
3889 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3890 - basic mmap support (3918)
3891 - kernel patch series update from b1_4 (4711)
3893 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3896 - avoid crash during MDS cleanup with OST shut down (2775)
3897 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3898 - don't use bad inodes on the MDS (3744)
3899 - dynamic object preallocation to improve recovery speed (4236)
3900 - don't hold spinlock over lock dumping or change debug flags (4401)
3901 - don't zero obd_dev when it is force cleaned (3651)
3902 - print grants to console if they go negative (4431)
3903 - "lctl deactivate" will stop automatic recovery attempts (3406)
3904 - look for existing locks in ldlm_handle_enqueue() (3764)
3905 - don't resolve lock handle twice in recovery avoiding race (4401)
3906 - revalidate should check working dir is a directory (4134)
3908 - don't always mark "slow" obdfilter messages as errors (4418)
3910 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3913 - don't close LustreDB during write_conf until it is done (3860)
3914 - fix typo in lconf for_each_profile (3821)
3915 - allow dumping logs from multiple threads at one time (3820)
3916 - don't allow multiple threads in OSC recovery (3812)
3917 - fix debug_size parameters (3864)
3918 - fix mds_postrecov to initialize import for llog ctxt (3121)
3919 - replace config semaphore with spinlock (3306)
3920 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3921 - don't allow enqueue to complete on a destroyed export (3822)
3922 - down write_lock before checking llog header bitmap (3825)
3923 - recover from lock replay timeout (3764)
3924 - up llog sem before sending rpc (3652)
3925 - reduce ns lock hold times when setting kms (3267)
3926 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3927 - fix NULL deref and obd_dev leak on setup error (3312)
3928 - replace some LBUG about llog ops with error handling (3841)
3929 - don't match INVALID dentries from d_lookup and spin (3784)
3930 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3931 - fix invalid assertion in ptlrpc_set_wait (3880)
3933 - add libwrap support for the TCP acceptor (3996)
3934 - add /proc/sys/portals/routes for non-root route listing (3994)
3935 - allow setting MDS UUID in .xml (2580)
3936 - print the stack of a process that LBUGs (4228)
3938 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3941 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3942 - make sure to unset replay flag from failed open requests (3440)
3943 - if default stripe count is 0, use OST count for inode size (3636)
3944 - update parent mtime/ctime on client for create/unlink (2611)
3945 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3946 - free recovery state on server during a forced cleanup (3571)
3947 - unregister_reply for resent reqs (3063)
3948 - loop back devices mounting and status check on 2.6 (3563)
3949 - fix resource-creation race that can provoke i_size == 0 (3513)
3950 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3951 - more debugging for page-accounting assertion (3746)
3952 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3953 - avoid deadlock after precreation failure (3758)
3954 - fix race and lock order deadlock in orphan handling (3450, 3750)
3955 - add validity checks when grabbing inodes from l_ast_data (3599)
3957 - add /proc/.../recovery_status to obdfilter (3428)
3958 - lightweight CDEBUG infrastructure, debug daemon (3668)
3959 - change default OSC RPC parameters to be better on small clusters
3960 - turn off OST read cache for files smaller than 32MB
3961 - install man pages and include them in rpms (3100)
3962 - add new init script for (un)mounting lustre filesystems (2593)
3963 - run chkconfig in %post for init scripts (3701)
3964 - drop scimac NAL (unmaintained)
3966 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3969 - clean kiobufs before and after use (3485)
3970 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3971 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3972 - change init script to fail more clearly if not run as root (1528)
3973 - allow clients to reconnect during replay (1742)
3974 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3975 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3976 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3977 - fix cond_resched (3554)
3978 - only evict unfinished clients after recovery (3515)
3979 - allow bulk resend, prevent data loss (3570)
3980 - dynamic ptlrpc request buffer allocation (2102)
3981 - don't allow unlinking open directory if it isn't empty (2904)
3982 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3983 - remove extraneous obd dereference causing LASSERT failure (3334)
3984 - don't use get_cycles() when creating temp. files on the mds (3156)
3985 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3986 - handle EEXIST for set-stripe, set proper directory name (3336)
3988 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3989 - fix ksocknal_fmb_callback() error messages (2918)
3991 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3994 - don't copy lvb into (possibly NULL) reply on error (2983)
3995 - don't deref dentry after dput, don't free lvb on error (2922)
3996 - use the kms to determine writeback rpc length (2947)
3997 - increment oti_logcookies when osc is inactive (2948)
3998 - update client's i_blocks count via lvb messages (2543)
3999 - handle intent open/close of special files properly (1557)
4000 - mount MDS with errors=remount-ro, like obdfilter (2009)
4001 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4002 - don't use cancelling-locks' kms values (2947)
4003 - use highest lock extent for kms, not last one (2925)
4004 - don't dereference ERR_PTR() dentry in error handling path (3107)
4005 - fix thread race in portals_debug_dumplog() (3122)
4006 - create lprocfs device entries at setup instead of at attach (1519)
4007 - common AST error handler, don't evict client on completion race (3145)
4008 - zero nameidata in detach_mnt in 2.6 (3118)
4009 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4010 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4011 - initialize RPC timeout timer earlier for 2.6 (3219)
4012 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4013 - print nal/nid for unknown nid (3258)
4014 - additional checks for oscc recovery before doing precreate (3284)
4015 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4016 - don't crash in mdc_close for bad permissions on open (3285)
4017 - zero i_rdev for non-device files (3147)
4018 - clear page->private before handing to FS, better assertion (3119)
4019 - tune the read pipeline (3236)
4020 - fix incorrect decref of invalidated dentry (2350)
4021 - provide read-ahead stats and refine rpc in flight stats (3328)
4022 - don't hold journal transaction open across create RPC (3313)
4023 - update atime on MDS at close time (3265)
4024 - close LDAP connection when recovering to avoid server load (3315)
4025 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4026 - don't leak open file on MDS after open resend (3325)
4027 - serialize filter_precreate and filter_destroy_precreated (3329)
4028 - loop device shouldn't call sync_dev() for nul device (3092)
4029 - clear page cache after eviction (2766)
4030 - resynchronize MDS->OST in background (2824)
4031 - refuse to mount the same filesystem twice on same mountpoint (3394)
4032 - allow llmount to create routes for mounting behind routers (3320)
4033 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4034 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4035 - fix rare problem with rename on htree directories (3417)
4037 - allow default OST striping configuration per directory (1414)
4038 - fix compilation for qswnal for 2.6 kernels (3125)
4039 - increase maximum number of MDS request buffers for large systems
4040 - change liblustreapi to be useful for external progs like lfsck (3098)
4041 - increase local configuration timeout for slow disks (3353)
4042 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4044 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4047 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4048 - don't overwrite extent policy data in reply if lock was blocked (2901)
4049 - drop filter export grants atomically with removal from device (2663)
4050 - del obd_self_export from work_list in class_disconnect_exports (2908)
4051 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4052 - swab reply message in mdc_close, other PPC fixes (2464)
4053 - fix destroying of named logs (2325)
4054 - overwrite old logs when running lconf --write_conf (2264)
4055 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4056 - fix race in target_handle_connect (2898)
4057 - mds_reint_create() should take same inode create lock (2926)
4058 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4059 - don't close files for self_export to avoid uninitialized obd (2936)
4060 - allow MDS with the same name as client node (2939)
4061 - hold dentry reference for closed log files for unlink (2325)
4062 - reserve space for all logs during transactions (2059)
4063 - don't evict page beyond end of stripe extent (2925)
4064 - don't oops on a deleted current working directory (2399)
4065 - handle hard links to targets without a parent properly (2517)
4066 - don't dereference NULL lock when racing during eviction (2867)
4067 - don't grow lock extents when lots of conflicting locks (2919)
4069 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4072 - account for cache space usage on clients to avoid data loss (974)
4073 - lfsck support in lustre kernel code (2349)
4074 - reduce journal credits needed for BRW writes (2370)
4075 - orphan handling to avoid losing space on client/server crashes
4076 - ptlrpcd can be blocked, stopping ALL progress (2477)
4077 - use lock value blocks to assist in proper KMS, faster stat (1021)
4078 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4079 - recovery for initial connections (2355)
4080 - fixes for mds_cleanup_orphans (1934)
4081 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4082 - block all file creations until orphan recovery completes (1901)
4083 - client remove rq_connection from request struct (2423)
4084 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4085 - recovery timer race (2670)
4086 - mdc_close recovey bug (2532)
4087 - ptlrpc cleanup bug (2710)
4088 - mds timeout on local locks (2588)
4089 - namespace lock held during RPCs (2431)
4090 - handle interrupted sync write properly (2503)
4091 - don't try to handle a message that hasn't been replied to (2699)
4092 - client assert failure during cleanup after abort recovery (2701)
4093 - leak mdc device after failed mount (2712)
4094 - ptlrpc_check_set allows timedout requests to complete (2714)
4095 - wait for inflight reqs when ptlrpcd finishes (2710)
4096 - make sure unregistered services are removed from the srv_list
4097 - reset bulk XID's when resending them (caught by 1138 test)
4098 - unregister_bulk after timeout
4099 - fix lconf error (2694)
4100 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4101 - readahead locks pages, leaves pending causing memory pressure (2673)
4102 - increase OST request buffers to 4096 on large machines (2729)
4103 - fix up permission of existing directories in simple_mkdir (2661)
4104 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4105 - don't assign transno to errored transactions (2742)
4106 - don't delete objects on OST if given a bogus objid from MDS (2751)
4107 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4108 - if rq_no_resend, then timeout request after recovery (2432)
4109 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4110 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4111 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4112 - MDS recovery shouldn't delete objects that it has given out (2730)
4113 - if enqueue arrives after completion, don't clobber LVB (2819)
4114 - don't unlock pages twice when trigger_group_io returns error (2814)
4115 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4116 - don't write pages to disk if there was an error (1450)
4117 - don't ping imports that have recovery disabled (2676)
4118 - take buffered bytes into account when balancing socknal conn (2817)
4119 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4120 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4121 - remove little-endian swabbing of llog records (1987)
4122 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4123 - retry reposting request buffers if they fail (1191)
4124 - grow extent at grant time to avoid granting a revoked lock (2809)
4125 - lock revoke doesn't evict page if covered by a second lock (2765)
4126 - disable VM readahead to avoid reading outside lock extents (2805)
4128 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4129 - updated kernel patches for hp-2.4.20 kernel (2681)
4131 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4134 - fix truncated write corruption (2366)
4135 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4137 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4138 - ensure len(uuid) < 37 in lmc (1171)
4139 - fix ia64 OOPS in llog_test (2255)
4140 - zero end of page at obdfilter for partial page writes (2648)
4141 - don't leave stale dentries around after renames (bug 2428)
4142 - fix timeouts when evicting a client with a single lock held (2642)
4143 - set deadline for the initial HELLO message to drain (2634)
4144 - print out dotted-quad IP addresses in the socknal (2302)
4146 - additional debugging for MDS client eviction problem (2443)
4147 - fix mkfsoptions support for osts (2603, 2604)
4149 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4152 - add series for the vanilla 2.6.0 kernel
4153 - add series for the vanilla 2.4.24 kernel
4154 - add series for a cray x86/64 UL kernel drop
4155 - fix xattr patches for the vanilla 2.4.19 series
4157 - generate true UUIDs in lmc (1171)
4158 - have portals stack dumping break in UML (2466)
4159 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4160 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4161 - wake callback waiting threads on client eviction (2460)
4162 - Add --ptldebug and --subsystem to lmc (1719)
4163 - update assertion to allow safe interrupt allocation
4164 - set rq_no_resend for cancel requests (2432)
4165 - recalculate ptlrpcd timeout after resend (2494)
4166 - call vfs_rmdir when removing pending directories (2368)
4167 - fix renaming a file to itself (2429)
4168 - lmc creates a default one-stripe lov (2454)
4169 - expand procfs space to handle large clusters (2326)
4170 - increase UML stack to avoid overflow
4171 - update lconf's list of debug and subsystem masks
4172 - fix lfs find --obd (2510)
4173 - /proc tunable for disabling filter read caching (2591)
4174 - stop rpm packages from altering slapd.conf (2301)
4175 - disable nagle in the socknal under 0conf (2578)
4176 - choose mds inode size based on stripe count (2572)
4177 - fix kernel-source rpm problems (2516)
4179 - add --disable-doc to avoid pdf generation (2421)
4180 - update documentation, tests, type-os, comments
4181 - avoid format warnings on ia64
4182 - remove the TOE NAL
4183 - tiny code cleanups by removing unused fields
4185 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4188 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4189 - remove the most blatant lies from BUILDING file (2371)
4190 - change default debug level to reasonable production setting
4191 - reduce client side cache size to reduce cache flush time
4192 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4193 - make TCP zerocopy and pinger support enabled by default (2476)
4194 - sync writes completed after process exits caused crashes (2319)
4195 - maintain correct mount count on the MDS (2356)
4196 - backout 1557, because 2316 wasn't really fixed
4197 - better file I/O statistics gathering in /proc
4198 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4199 - another kernel patch to fix zero-copy TCP function export
4200 - don't take duplicate lock when processing re-sent getattr (2420)
4201 - lctl uses obd_self_export instead of creating new conn (2353)
4202 - MDS/OST recovery case which requires object creation asserted (2425)
4203 - move lfs from /usr/sbin to /usr/bin in packages
4204 - fix race between mds_client_add and mds_client_free (2417)
4205 - use kmalloc instead of slabs in portals (2430)
4206 - don't create duplicate records when a failover MDS is present (2442)
4207 - remove unnecessary mount age check (2332)
4208 - don't remove directory inodes from locks prematurely (2451)
4209 - don't break if MDS service name is the same as hostname (2103)
4210 - fix races in client write RPC generation when cache full (2482)
4212 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4215 - remove now-unused request->rq_obd (278)
4216 - if an allocation fails, print out how much memory we've used (1933)
4217 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4218 - add the "configurable stack size" patch to most series files (1256)
4219 - ability to write large log records, for 100+ OST configs (2306)
4220 - fix NULL deref when filter_prep fails (2314)
4221 - fix operator precedence error in filter_sync
4222 - dynamic allocation of socknal TX descriptors (2315)
4223 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4224 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4225 - ensure that CWARN messages in Portals always get to the syslog
4226 - __init/__exit are not for prototype decls (ldlm_init/exit)
4227 - x86-64 compile warning fixes
4228 - fix gateway LMC keyword conflict (2318)
4229 - fix MDS lock inversions in getattr/reint paths (1844)
4230 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4231 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4232 - DLM race condition prevented some lock evictions (2328)
4233 - ENOMEM detection and retry on socknal sends (2230)
4234 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4235 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4236 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4237 - better error messages when a client is rejected during recovery (1505)
4238 - avoid cancelling locks which were never granted, after failure (2330)
4239 - fix i_sem/journal inversion in mds_client_add (2333)
4240 - fix truncate/getattr lock cycle deadlock (2334)
4241 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4242 - fix two rare exit paths which could leak an l_lock() ref (2321)
4243 - fencepost error in MDS/OST orphan recovery (2226)
4244 - make log record alignment 8 bytes (1988)
4245 - lstripe now fails when requested offset > ost_count (2237)
4246 - ensure that all kernel series have a complete list.h (1607)
4247 - fix crashes in special-file operations (2316)
4248 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4250 - allow configurable automake binary, for testing new versions
4251 - small update to the lfs documentation
4253 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4255 * fix negative export reference count in fsfilt_sync (2312)
4257 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4258 * release candidate 0.9.1
4260 - orphans are moved into the PENDING directory for possible recovery
4261 - replayed opens now open by fid for orphan/rename safety (1042)
4262 - last close of an orphan inode generates a transno (683)
4263 - chdir() and mount() now pin the directory entry (1020)
4264 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4265 - discard very old requests without processing them (1502)
4266 - remove some common, well-understood CERRORs (1505)
4267 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4268 - clear "grant" flags in OST replies until OST grant code lands (1644)
4269 - fix read performance by not clobbering i_blksize on client (1598)
4270 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4271 - make LRU size a /proc tunable, clears locks when reduced (707)
4272 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4273 - server locks take references on exports now (1558)
4274 - build fixes for 2.4.20-rh trees (1663)
4275 - return an error from lov_create if all OSCs are inactive (1751)
4276 - fix import levels when a reconnect happens without a timeout (1597)
4277 - exit early from mds_open if we get a lookup error (1749)
4278 - partial page read at EOF wouldn't wait for disk before sending (1642)
4279 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4280 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4281 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4282 - write/truncate lock inversion (1639)
4283 - Don't auto-load obdclass, portals modules during cleanup (1495)
4284 - fix timestamps from jumping to "now" (1763)
4285 - extra journal assertions (1648)
4286 - add an extra multiunlink test (1771)
4287 - fix read_record/write_record API (1776)
4288 - fix leak of offset_extent, possible incorrect i_size later (1772)
4289 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4290 - Debugging for the kqswnal_get_idle_tx problems (1820)
4291 - Allow recovery to be attempted multiple times (1536)
4292 - Write out MDS last_rcvd file after it is first created (1600)
4293 - Fix tx_descriptor leak in failed transmit situations (1827)
4294 - ext3 journaling fixes for assertion failure after IO error (1871)
4295 - class_export_put() on freed export after completion AST error (1896)
4296 - Fix revalidate looping in VFS (1322)
4297 - Don't access a freed export during MDS_REINT timeout (1521)
4298 - Add open-unlink recovery support on the MDS (1673,1764)
4299 - Return an error if no MDS data was read from last_rcvd (1946)
4300 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4301 - Don't setup a disk device that doesn't match exported UUID (317)
4302 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4303 - avoid committing NULL handle in force close
4304 - local.sh is now a one-stripe LOV configuration
4305 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4306 - fix ext3 htree duplicate directory entry corruption (1516)
4307 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4308 - update to new LOV EA format (2097)
4309 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4311 2003-06-15 Phil Schwan <phil@clusterfs.com>
4314 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4315 - per-import/export recovery handling (958, 931, 959)
4316 - multiple last-rcvd slots, for serving multiple FSes (949)
4317 - connections are again shared between multiple imp/exports (963, 964)
4318 - "umount -f" would hang if any requests needed to be sent (393, 978)
4319 - avoid pinning large req buffer by copying for queued messages (989)
4320 - add "uuid" to "lctl device" command to help upcalls (991)
4321 - "open" RPCs with transnos would confuse recovery counters (1037)
4322 - do proper endian conversion of all wire messages (288, 340, 891)
4323 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4324 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4325 - fix ptlrpc_connection leak in target_handle_connect (1174)
4326 - fix import refcounting bug in OST and MDS cleanup (1134)
4327 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4328 - fix very unlikely obd_types race condition (501)
4329 - remove osc_open hack for echo_client (1187)
4330 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4331 - a failure in read_inode2 leads to deadlock (1139)
4332 - cancel ack-locks as soon as transaction is committed (1072)
4333 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4334 - make sure to commitrw after any preprw to avoid deadlock (1162)
4335 - failing to execute a file in a lustre FS would lock inode (1203)
4336 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4337 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4338 - obd_uuid2tgt would walk off the end of the list (1255)
4339 - on IA64 the osc would give portals incorrect bulk size (1258)
4340 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4341 - fix lock inversion caused by new llite matching code (1282)
4342 - limit the number of dirty pages on a client to 10MB (1286)
4343 - timed out locks were not being corrected cancelled (1289)
4344 - fix O_DIRECT above 4GB on IA-32 (1292)
4345 * major user-visible changes
4346 - fail out/fail over policy now controlled by the upcall (993)
4348 - add OBD_PING to check server availability and failure (954)
4349 - lustre messages are now sent in sending host order (288, 340, 891)
4350 - add eadatalen to MDS getattr reply (340)
4351 - OST read replies may contain second buffer, with per-page status (593)
4353 2003-03-11 Phil Schwan <phil@clusterfs.com>
4356 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4357 - failed open()s could cause deadlock; fixed (867, 869)
4358 - stop cancelling OST locks when files are closed (481)
4359 - overlapping XID spaces caused network corruption (851, 853)
4360 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4361 - setattr_raw added, to do single-RPC, server-side setattrs
4362 - lmc/lconf syntax change for OST UUIDs
4363 - fix crashy race condition between ptlrpc_free_req and osc_close
4364 - don't use request in mdc_enqueue if we hit a timeout (889)
4365 - don't set the inode i_size for regular files from the MDS (896)
4366 - handle out of order completion AST (842)
4367 - don't LBUG if a lock request times out after receiving AST (913)
4368 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4369 - if a bad lock AST arrives, send an error instead of dropping entirely
4370 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4371 - fix leak in bulk IO when only partially completed (899, 900, 926)
4372 - fix O_DIRECT for ia64 (55)
4373 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4374 - C-z after timeout could hang a process forever; fixed (977)
4376 - client-side I/O cache (678, 924, 929, 941, 970)
4378 - READPAGE and SETATTRs which don't take server-side locks get
4381 2003-02-11 Phil Schwan <phil@clusterfs.com>
4384 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4385 - Fix fsfilt_extN_readpage() to read a full page of directory
4386 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4387 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4388 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4389 - Most intent operations were removed, in favour of a new RPC mode
4390 that does a single RPC to the server and bypasses most of the VFS
4391 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4392 - Aggressively cancel local locks on DLM servers
4393 - mds_reint_unlink sends EA to the client if it's the last nlink.
4394 client uses that EA to unlink OST objects.
4395 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4396 - recursive symlinks were fixed (439)
4397 - fixed NULL deref in DEBUG_REQ
4398 - filter_update_lastobjid no longer calls sync, which annoyed extN
4399 - fixed multi-client small-writes to a single file problem (445)
4400 - fixed mtime updates during file writes (607)
4401 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4402 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4403 - fixed Linux OST size reporting problem (444, 656)
4404 - OST now updates object mtime with writes or setattr (607, 619)
4405 - client verifies file size before zeroing page past EOF (445)
4406 - OST now writes last allocated objid to disk with allocation (108)
4407 - LOV on echo now works (409)
4409 - mds_reint_unlink sends a new buffer, with the EA included. this
4410 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4411 set if a regular file was being unlinked, and it was the last link
4412 - use PtlGet from the target for bulk writes (315)
4413 - OST now updates object mtime with writes or setattr (607, 619)
4414 - LDLM now has a grant-time callback to revalidate locked items, if
4416 - Many MDS operations were reorganized to combat race conditions
4418 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4419 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4421 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4424 - Fully reactivate OST imports after reconnection (512, others)
4425 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4426 - More graceful error handling for truncating on dead OST (515)
4427 - Don't error out unless we're actually accessing dead stripes (474)
4428 - Fix garbage sizes when stripes are missing (410)
4429 - LRU counters were broken, causing constant lock purge (433, 432)
4430 - garbage on read from stripes with failed OSTs (441)
4431 - mark OSCs as active before reconnecting during recovery (438)
4432 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4433 - lfind did not preserve OST order in output (443)
4434 - symlinks cause hung clients, incorrect data (439)
4435 - stop dereferencing request after dropping refcount (457)
4436 - don't LASSERT(spin_is_locked) on non-SMP (455)
4437 - fixes for many rename() bugs
4438 - fstat didn't correctly synchronize attributes (399)
4439 - server must handle lock cancellation during blocking AST prep (487)
4440 - bulk descriptors were free()d too soon (511)
4441 - fix paths in lconf, which would load incorrect modules (451, 507)
4442 - fix confusing lconf 'host not found' error message (386)
4443 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4444 - fix race condition in mdc_blocking_ast() for inode access (526)
4445 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4446 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4447 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4448 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4449 - fix lstripe in several ways (kernel side) (527)
4450 - fix request leak in ldlm_cli_enqueue (262)
4451 - incorrect OSC was marked inactive after OST failure
4452 - call mds_fs_cleanup before unmounting filesystem (524)
4453 - fix races between taking ns_lock and ldlm_lock_change_resource
4454 - fix races updating LOV export open file list
4455 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4456 - fix recovery NULL deref in ldlm_cli_cancel_unused
4457 - fix some DLM races by using new hash table for lock handles (419)
4458 - permit the client to specify desired inodes, at replay
4459 - duplicate requests when we queue them for replay reintegration
4460 - fix last_rcvd offset calculation
4461 - sync after each recovered transaction, so we always make progress
4462 - never, not always, ERESTART requests without transnos
4463 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4465 - skip replay if the MDS says that the client is already connected
4466 - don't check for a recovery-enabled export to match lctl's UUID
4467 - don't INC_USE_COUNT for phantom exports
4468 - don't crash when cleaning up phantom exports (567)
4469 - don't double-finish or set replay data for errored mdc_open requests
4470 - abort requests when they time out, so we don't get old replies
4471 - send/receive replies for AST messages again
4472 - if the client says that it doesn't have the lock, cancel it on the
4474 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4475 mark it as destroyed, it will all work out when decref is called
4476 - fix module use counts (22, 581)
4478 - ASTs now expect a reply (server cancels lock on error reply)
4480 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4483 - fix many simultaneous client startup (392)
4484 - fix dentry->d_it clobbering
4485 - credentials weren't being shipped for readdir/getattr operations
4486 - remove invalid assertions triggered during some concurrent MD
4488 - proper Lustre versions added (336, 389)
4489 - fix memory leak for create error case (398)
4490 - fix LOV locking bug that would get cli/srv out of sync
4491 - fix echo client over LOV (409)
4492 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4493 - fix double-O_EXCL intent crash (424)
4494 - avoid sending multiple lock CANCELs (352)
4496 - MDS can do multi-client recovery (modulo bugs in new code)
4498 - many updates, edits, cleanups
4500 2002-11-18 Phil Schwan <phil@clusterfs.com>
4503 - fix null d_it dereference (346)
4504 - fix full OST/dbench hang (333)
4505 - fix permission problem with file removal (286)
4506 - fix removal of OSCs from LOV when they fail
4507 - fix NULL deref during bulk timeout (214)
4508 - fix problems related to multiple filesystems on one MDS (241)
4509 - fixed serious subtle metadata locking bugs
4510 - free locks on clients when inodes are removed due to memory
4512 - fix inode pointer in lock data (285)
4513 - partial support for multiple MDS on a single host (241)
4514 - data locks weren't cancelled at clear_inode time (290, 311)
4515 - intent locks could lead to unbounded lock growth (205)
4516 - added a maximum lock count, an LRU list, and a flusher
4517 - fix multiple rename (365)
4518 - properly abstracted the echo client
4519 - OSC locked 1 byte too many; fixed
4520 - rewrote brw callback code:
4521 - fixed recovery bugs related to LOVs (306)
4522 - fixed too-many-pages-in-one-write crash (191)
4523 - fixed (again) crash in sync_io_timeout (214)
4524 - probably fixed callback-related race (385)
4526 - Add capability to MDS protocol
4527 - LDLM cancellations and callbacks on different portals
4529 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4532 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4533 - defer open object destruction to close time (601981, 138)
4534 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4535 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4536 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4537 - don't call obd_setattr in ll_file_release for destroyed objects
4538 * protocol change to lustre_msg: move |version| and add |flags|
4539 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4540 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4541 * added replay of create, unlink, link and rename operations during
4542 MDS failover; recovery should be much more robust now
4543 * remove failed OSCs from LOVs (only lov_create uses this so far)
4544 * the lustre-HOWTO was brought (more) up to date (582544)
4546 2002-10-23 Phil Schwan <phil@clusterfs.com>
4549 - in-use dentries weren't being reused properly (617851)
4550 - prevent multiple LDLM setup (599178)
4551 - fix LOV size calculations for truncate (617853)
4552 - fix client handling of MDS intent errors (POSIX)
4553 - fix permission bug in lovstripe.c test (624321)
4554 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4555 - truncate past end of file could corrupt data
4556 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4557 - a race in recovery could return ETIMEDOUT to apps (623947)
4558 - building outside the source directory was fixed
4559 * the lustre-HOWTO was brought (more) up to date (582544)
4560 * major progress was made on recovery functionality
4562 2002-10-10 Phil Schwan <phil@clusterfs.com>
4565 - recovery deadlock fix
4566 - rm -rf causes LBUG fix (617817)
4567 - file open by multiple tasks fix (618962)
4568 - directory permissions bugs (602707 and 620007)
4569 - journal_stop fixed with locking (611313)
4570 - O_APPEND failures resolved (618273, perhaps 614459)
4571 - lconf PATH fix (619770)
4572 - IA64 build fix (621450)
4573 - RPC buffer sizes scale with amount of memory
4575 2002-10-01 Phil Schwan <phil@clusterfs.com>
4578 - locks would be cancelled without throwing away data pages,
4579 resulting in inconsistent data (605627)
4580 - inode attributes were not always being refreshed (605627, 612449)
4581 - lconf now continues to cleanup after lctl reports an error
4582 - MDS now enforces user permissions (602707)
4583 - lprocfs cleanup fixed, but not yet enabled (614157)
4584 - fixed infinite server hang, should a client not respond to an AST
4585 - avoid going into recovery if user calls readlink() with a buffer
4586 that's too small (613941)
4587 - AST RPCs no longer require replies (614867) -- this may be changed
4588 - don't crash server if client sends an IOV that's too big (611336)
4589 - fixed lock conversion deadlock (611892)
4590 - fixed the following of symlinks (614622)
4591 * recovery: the server can remove locks from a client that dies, other
4592 clients can make progress
4593 * more extN patch fixes
4594 * compile-time configurable ptlrpc buffer allocations
4596 - collaborative read cache document
4597 - Lustre Lite Performance CDR document-in-progress
4599 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4602 - fix typo in patch-2.4.18
4604 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4607 - clear ptlrpc request each time in handle_incoming_request()
4608 - unlink of files now destroys the object on the OST
4610 2002-09-19 Peter Braam <braam@clusterfs.com>
4612 * add hard link support
4613 * change obdfile creation method
4614 * kernel patch changed
4616 2002-09-19 Peter Braam <braam@clusterfs.com>
4619 - stack overflow bug in extN fixed
4621 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4623 * documentation updates
4624 - add man pages for config tools
4625 - update tests/README to describe testing with new config tools
4626 - finish metadata API descriptions
4627 * bug fixes and cleanups
4628 - statfs workaround for 16TB limit
4629 - LOV stripe allocation improved, can stripe on subset of OSTs
4630 - LOV file size/IO offset was wrong for files > 4GB in size
4631 - object EA data was being dropped, caused files to be unreadable
4632 - memory overflow with non-LOV OST caused memory corruption
4633 - fixed regression tests to work with new config tools, obdfilter
4634 - fixed bug when directory size became larger than 1 block
4635 - fixed bug (for single client case) when PWD was deleted
4636 - invalidate local directory pages when doing intent-based ops
4637 - avoid LDLM oops when lock callback contained bad data
4639 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4641 * documentation updates
4642 * bug fixes and cleanups
4643 - configuration tools
4646 - 64-bit compile warnings
4647 - 64-bit internal statfs data
4649 * test_brw on persistent OST devices
4651 * lprocfs (disabled)
4653 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4655 * documentation updates
4656 * bug fixes and cleanups
4657 * configuration tools
4659 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4662 * many small fixes to 0_5_4
4663 * io/network handling
4664 * thinkos in MDS operations
4666 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4669 * crucial basic fixes to 0.5.3
4670 * IOR, Iozone work over Elan
4673 2002-08-07 Phil Schwan <phil@clusterfs.com>
4674 * version 0_5_3, our first alpha
4675 * we use the new Portals iovs
4676 * documentation updates
4677 * bug fixes and cleanups
4678 * small changes in the DLM wire protocol
4680 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4681 * version 0_5_1 with some initial stability,
4682 * locking on MD and file I/O.
4683 * documentation updates
4684 * several bug fixes since 0.5.0
4685 * small changes in wire protocol
4687 2002-07-18 Phil Schwan <phil@clusterfs.com>
4689 * delivered as Lustre Light Alpha
4690 * fixed a crash after handling invalid MDS requests
4691 * fixed directory pages for architectures with non-4k pages sizes
4693 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4694 * release version v0_4_4
4695 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4697 * Fixes a number of interruption problems with OST operations.
4698 * Update documentation for portals header changes
4699 * Move all wire protocol structs/defines to lustre_idl.h
4700 * Fixes symlink length bug.
4701 * Add tcpdump to repository.
4703 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4704 * release version v0_4_3
4705 * Fixes statfs for inodes on extN.
4706 * Fixes bug in runtests which would delete /etc/hosts.
4707 * Use 64-bit object IDs wherever possible (not into VFS though)
4708 Remove ost_get_info, which is unused by lustre, and out of date.
4710 2002-07-03 Peter Braam <braam@clusterfs.com>
4711 * release version v0_4_2 Fixes a lookup error (type not passed)
4712 * move forward to head of Portals
4713 * move forward to latest Lustre kernel
4715 2002-06-25 Peter Braam <braam@clusterfs.com>
4716 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob