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 Frequency : rare, on clear nid stats
18 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
19 Details : when clean nid stats sometimes try destroy live entry,
20 and this produce panic in free.
24 Description: Stack overflow during MDS log replay
25 ease stack pressure by using a thread dealing llog_process.
29 Description: wait until IO finished before start new when do lock cancel.
30 Details : VM protocol want old IO finished before start new, in this case
31 need wait until PG_writeback is cleared until check dirty flag and
32 call writepages in lock cancel callback.
34 Severity : enhancement
36 Description: using special macro for print time and cleanup in includes.
40 Description: mds_mfd_close() ASSERTION(rc == 0)
41 Details : In mds_mfd_close(), we need protect inode's writecount change
42 within its orphan write semaphore to prevent possible races.
46 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
51 Description: don't hit live lock with umount ost.
52 Details : shrink_dcache_parent can be in long loop with destroy dentries,
53 use shrink_dcache_sb instead.
57 Description: don't panic with use echo client
58 Details : echo client pass NULL as client nid pointer and this produce null
63 Description: fix build on ppc32
64 Details : compile code with -m64 flag produce wrong object file for ppc32.
68 Description: add message levels for liblustreapi
72 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
73 Details : Move the 'good_osts' check before the 'total_bavail' check. This
74 will result in an -EAGAIN and in the exit call path we call
75 alloc_rr() which will with increasing aggressiveness attempt to
76 aquire precreated objects on the minimum number of required OSCs.
80 Description: Use old size assignment to avoid deadlock
81 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
82 the scheduling while holding a spinlock. We do not need locking
83 for size in ll_update_inode() because size is only updated from
84 the MDS for directories or files without objects, so there is no
85 other place to do the update, and concurrent access to such inodes
86 are protected by the inode lock.
90 Description: resolve "_IOWR redefined" build error on SLES10
94 Description: dump the memory debugging after all modules are unloaded to
95 suppress false negative in conf_sanity test 39
97 Severity : enhancement
99 Description: build kernel-ib packages for OFED 1.3 in our release cycle
104 Description: fix SLES kernel versioning
105 Details : the kernel version for our SLES 10 kernel did not include a "-"
106 before the "smp" at the end. while this was not a problem in
107 general, it did mean that software trying to use the kernel
108 version to try to detect a vendor specific kernel would fail.
109 this was most evident by the OFED build scripts.
113 Description: Don't update lov_desc members until making sure they are valid
114 Details : When updating lov_desc members via proc fs, need fix their
115 validities before doing the real update.
119 Description: don't put request into delay list while invalidate in flight.
120 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
121 import in flight. this produce timeout for invalidate and sometimes
122 can cause stale data.
126 Frequency : on ppc only
127 Description: not convert ost objects for directory because it's not exist.
128 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
132 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
134 Severity : enhancement
136 Description: Update to support 2.6.22.14 vanilla kernel.
140 Frequency : rare, on recovery
141 Description: read procfs can produce deadlock in some situation
142 Details : Holding lprocfs lock which send rpc can produce block for destroy
143 obd objects and this also block reconnect with -EALREADY. This isn't
144 fix all lprocfs bugs - but make it rare.
146 Severity : enhancement
148 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
151 Frequency : frequent on X2 node
153 Description: mdc_set_open_replay_data LBUG
154 Details : Set replay data for requests that are eligible for replay.
158 Description: lustre_mgs: operation 101 on unconnected MGS
159 Details : When MGC is disconnected from MGS long enough, MGS will evict the
160 MGC, and late on MGC cannot successfully connect to MGS and a lot
161 of the error messages complaining that MGS is not connected.
164 Frequency : on start mds
166 Description: Implement get_info(last_id) in obdfilter.
169 Frequency : occasional
171 Description: Correctly check stale fid, not start epoch if ost not support SOM
172 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
173 overwrited with new generated fid, but mds can anwer with one of these
174 two fids and both is not stale. setattr incorectly start epoch and
175 assume will be called done_writeting, but without SOM done_writing
179 Frequency : rare, depends on device drivers and load
181 Description: MDS or OSS nodes crash due to stack overflow
182 Details : Code changes in 1.8.0 increased the stack usage of some functions.
183 In some cases, in conjunction with device drivers that use a lot
184 of stack the MDS (or possibly OSS) service threads could overflow
185 the stack. One change which was identified to consume additional
186 stack has been reworked to avoid the extra stack usage.
189 Frequency : occasional
191 Description: Do not fail import if osc_interpret_create gets -EAGAIN
192 Details : If osc_interpret_create got -EAGAIN it immediately exits and
193 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
194 oscc_has_objects and see OSC has no objests and call
195 oscc_internal_create to resend create request.
197 Severity : enhancement
199 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
201 Severity : enhancement
203 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
206 Frequency : very rare
208 Description: Fix replay if there is an un-replied request and open
209 Details : In some cases, older replay request will revert the
210 mcd->mcd_last_xid on MDS which is used to record the client's
213 Severity : enhancement
215 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
217 Severity : enhancement
219 Description: Add rhel5 support to HEAD.
221 Serverity : enhancement
223 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
228 Description: Don't allow skipping OSTs if index has been specified.
229 Details : Don't allow skipping OSTs if index has been specified, make locking
230 in internal create lots better.
234 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
235 Details : access to bitfield in structure is always rounded to long
236 and this produce problem with not atomic change any bit.
240 Description: Lustre make rpms failed.
241 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
242 builds Lustre rpms from tarball.
245 Frequency : rare on shutdown ost
247 Description: If llog cancel was not send before clean_exports phase, this can
248 produce deadlock in llog code.
249 Details : If llog thread has last reference to obd and call class_import_put
250 this produce deadlock because llog_cleanup_commit_master wait when
251 last llog_commit_thread exited, but this never success because was
252 called from llog_commit_thread.
256 Description: allow userland application know is lost one of stripes.
257 Details : fill lvb_blocks with error code on ost and return it to
258 application if error flag found.
262 Description: NULL lov_tgts causing MDS oops
263 Details : more safe checks for NULL lov_tgts for avoid oops.
265 Severity : enhancement
267 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
271 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
272 Details : If an OST is down the MDS will hang indefinitely in
273 obd_statfs_rqset() waiting for the statfs data. While for
274 MDS QOS usage of statfs, it should not stuck in waiting.
276 Severity : enhancement
278 Description: remote_acl support
279 Details : Support ACL-based permission check for remote user.
280 Support setfacl/getfacl for remote user with the utils
281 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
282 the system "{s,g}etfacl" utils.
284 Severity : enhancement
286 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
288 Severity : enhancement
290 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
294 Description: make mgs_setparam() handle fsname containing dash
295 Details : fsname containing a dash does not work with lctl conf_param
297 Severity : enhancement
299 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
301 Severity : enhancement
303 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
307 Frequency : rare, at shutdown
308 Description: access already free / zero obd_namespace.
309 Details : if client_disconnect_export was called without force flag set,
310 and exist connect request in flight, this can produce access to
311 NULL pointer (or already free pointer) when connect_interpret
312 store ocd flags in obd_namespace.
316 Frequency : only at startup
317 Description: not alloc memory with spinlock held.
318 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
319 if any spinlock held.
321 Severity : enhancement
323 Description: make lustre randomly fail allocating memory
324 Details : Make lustre randomly failed allocating memory for testing purpose.
326 Severity : enhancement
328 Description: lost problems with lov objid file
329 Details : Fixes some scability and access to not inited memory problems
330 in work with lov objdid file.
335 Description: lfs find does not continue on file error
336 Details : Continue other files processing when a file/dir is absent.
340 Description: Inconsistent usage of lustre_pack_reply()
341 Details : Standardize the usage of lustre_pack_reply() such that it
342 always generate a CERROR on failure.
345 Frequency : occasional
347 Description: MDS hang or stay in waiting lock
348 Details : If client receive lock with CBPENDING flag ldlm need send lock
349 cancel as separate rpc, to avoid situation when cancel request
350 can't processed due all i/o threads stay in wait lock.
354 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
358 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
362 Description: Update to SLES9 kernel-2.6.5-7.287.3.
366 Description: Update to RHEL4 latest kernel.
368 Severity : enhancement
370 Description: Build SLES10 patchless client fails
371 Details : The configure was broken by run ./configure with
372 --with-linux-obj=.... argument for patchless client. When the
373 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
374 can't search header adequately. Use absolute path such as
375 -I($LINUX)/include instead.
379 Description: interrupt oig_wait produce painc on resend.
380 Details : brw_redo_request can be used for resend requests from ptlrpcd and
381 private set, and this produce situation when rq_ptlrpcd_data not
382 copyed to new allocated request and triggered LBUG on assert
383 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
384 ptlrpcd set if request is changed and can be safety changed to use
389 Description: LASSERT_{REQ,REP}SWAB macros are buggy
390 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
391 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
392 these macros and replace them with their difinitions instead.
396 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
397 Details : Update which_patch & target file for SLES10 latest kernel.
400 Description: Remove client patches from SLES 10 kernel.
401 Details : This causes SLES 10 clients to behave as patchless clients
402 even on a Lustre-patched (server) kernel.
404 Severity : enhancement
406 Description: self-adjustable client's lru lists
407 Details : use adaptive algorithm for managing client cached locks lru
408 lists according to current server load, other client's work
409 pattern, memory activities, etc. Both, server and client
410 side namespaces provide number of proc tunables for controlling
413 Severity : enhancement
415 Description: light-weight GSS support
416 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
417 and reduce performance overhead.
419 Severity : enhancement
421 Description: Linux keyring support
422 Details : Support using service of Linux keyring for Lustre GSS internal
423 context refresh/cache mechanism.
427 Description: Fix errors in lfs documentation
428 Details : Fixes man pages
432 Description: don't use GFP_* in generic Lustre code.
433 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
437 Description: obdclass is limited by single OBD_ALLOC(idarray)
438 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
442 Description: warnings with build patchless client with vanila 2.6.19 and up
443 Details : change old ctl_table style and replace ctl_table/ctl_table_header
444 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
448 Frequency : Only for SLES
449 Description: sanity_quota fail test_1
450 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
451 file, which makes TSTID[2] inunique.
455 Frequency : Only for SLES9
456 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
457 Details : Update target/ChangeLog/which_patch .
461 Frequency : Only for test_10 in sanity-quota.sh
462 Description: a bug in quota test code
463 Details : checking a test flag in a wrong place
467 Frequency : Only for test_13 in sanity-quota.sh
468 Description: a bug in quota test script
469 Details : shouldn't check overfull allocated quota
473 Description: jbd statistics
474 Details : Port older jbd statistics patch for sles10
476 Severity : enhancement
478 Description: Add printing inode info into message about error in writepage.
483 Description: reply_lock_interpret crash due to race with it and lock cancel.
484 Details : Do not replay locks that are being cancelled. Do not reference
485 locks by their address during replay, just by their handle.
489 Frequency : When flocks are used.
490 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
491 Details : Flock locks might destroy just granted lock if it could be merged
492 with another existing flock, this is done in completion handler,
493 so teach ldlm_cli_enquque_fini that this is a valid case for
499 Description: Oops in read and write path when failing to allocate lock.
500 Details : Check if lock allocation failed and return error back.
504 Description: lstripe command fails for valid OST index
505 Details : The stripe offset is compared to lov->desc.ld_tgt_count
506 instead of lov->desc.ld_active_tgt_count.
510 Description: sanity.sh failed test 103
511 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
516 Description: df doesn't work properly if diskfs blocksize != 4K
517 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
521 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
522 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
526 Description: obd_health_check_timeout too short
527 Details : set obd_health_check_timeout as 1.5x of obd_timeout
531 Description: llapi_file_create() does not allow some changes
532 Details : add llapi_file_open() that allows specifying the mode and
533 open flags, and also returns an open file handle.
537 Description: merge and cleanup kernel patches.
538 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
542 Description: Add journal checksum support.(Kernel part)
543 Details : The journal checksum feature adds two new flags i.e
544 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
545 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
546 indicates that the commit block contains the checksum for
547 the blocks described by the descriptor blocks. Now commit
548 record can be sent to disk without waiting for descriptor
549 blocks to be written to disk. This behavior is controlled
550 using JBD2_FEATURE_ASYNC_COMMIT flag.
554 Description: OSS needs mutliple precreate threads
555 Details : Add ability to start more than one create thread per OSS.
559 Description: Sanity.sh test_65k failed: setstripe should have succeeded
560 Details : execute lfs setstripe on client
564 Description: mds_obd_create error creating tmp object
565 Details : When the user sets quota on root, llog will be affected and can't
566 create files and write files.
569 Frequency : Always on ia64 patchless client, and possibly others.
571 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
572 Details : This allows the patchless client to be loaded on architectures
577 Description: RedHat Update kernel for RHEL5
578 Details : Add the kernel config file for RHEL5.
582 Description: RedHat Update kernel for RHEL5
583 Details : Modify the kernel config file more closer RHEL5.
587 Description: Build failure against Centos5 (RHEL5)
588 Details : Define PAGE_SIZE when it isn't present.
592 Description: test_7 on sanity-quota.sh can't be used on separate machines
593 Details : add facet to handle it
597 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
598 Details : with NFS, the anon dentry's parent was set to itself in
599 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
600 corresponding dentry other than use rec->ur_name.
602 Severity : enhancement
604 Description: lfs setstripe enhancement
605 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
609 Description: enable data checksumming by default
610 Details : enable checksum by default, allow --disable-checksum
611 configure option and "-o nochecksum" mount option. Checksums
612 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
616 Description: lustre support for RHEL5
617 Details : Add support for RHEL5.
621 Description: Client eviction due to failover config
622 Details : after a connection loss, the lustre client should attempt to
623 reconnect to the last active server first before trying the
624 other potential connections.
628 Description: when mds and osts use different quota unit(32bit and 64bit),
629 quota will be released repeatly.
630 Details : void sending multiple quota reqs to mds, which will keep the
631 status between the reqs.
635 Description: osts not allocated evenly to files
636 Details : change the condition to increase offset_idx
639 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
641 Description: Data corruption for OSTs that are formatted larger than 2TB
643 Details : When generating the bio request for lustre file writes the
644 sector number would overflow a temporary variable before being
645 used for the IO. The data reads correctly from Lustre (which
646 will overflow in a similar manner) but other file data or
647 filesystem metadata may be corrupted in some cases.
651 Description: Tune the kernel for good SCSI performance.
652 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
653 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
658 Description: rewrite ext2-derived code in obdclass/uuid.c
659 Details : rewrite inherited code (uuid parsing code from ext2 utils)
660 from scratch preserving functionality.
664 Description: "lfs find -obd UUID" prints directories
665 Details : "lfs find -obd UUID" will return all directory names instead
666 of just file names. It is incorrect because the directories
667 do not reside on the OSTs.
671 Description: use i_size_read and i_size_write in 2.6 port
672 Details : replace inode->i_size access with i_size_read/write()
675 Frequency : when removing large files
677 Description: scheduling issue during removal of large Lustre files
678 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
679 It causes scheduling issues when removing large files (17TB in the
683 Frequency : only with liblustre clients on XT3
685 Description: evictions taking too long
686 Details : allow llrd to evict clients directly on OSTs
689 Frequency : only on ppc
691 Description: /proc/fs/lustre/devices broken on ppc
692 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
693 We should make sure the type of 'index' is loff_t and then cast
694 explicitly as needed below. Do not assign an explicitly cast
698 Frequency : only for rhel5
700 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
701 Details : Modify the target file & which_kernel.
703 Severity : enhancement
705 Description: omit set fsid for export NFS
706 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
707 as FS_REQUIRES_DEV for avoid problems with generate fsid.
711 Frequency : Always, for kernels after 2.6.16
712 Description: Fix warning idr_remove called for id=.. which is not allocated.
713 Details : Last kernels save old s_dev before kill super and not allow
714 to restore from callback - restore it before call kill_anon_super.
718 Description: buffer overruns could theoretically occur
719 Details : llapi_semantic_traverse() modifies the "path" argument by
720 appending values to the end of the origin string, and a
721 overrun may occur. Adding buffer overrun check in liblustreapi.
725 Description: Fix error on 'ls .' at the top of the Lustre mount.
726 Details : Don't revalidate dentry if it is a root dentry.
730 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
731 Details : Modify vm-tunables-rhel4.patch.
735 Description: Kernel config for 2.6.18-vanilla.
736 Details : Modify targets/2.6-vanilla.target.in.
737 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
738 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
739 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
740 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
744 Description: improve handling recoverable errors
745 Details : if request processig with error which can be recoverable on server
746 request should be resend, otherwise page released from cache and
751 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
752 Details : Modify target file & which_patch.
753 A flaw was found in the IA32 system call emulation provided
754 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
755 value could be stored in the %RAX register, which could trigger an
756 out-of-bounds system call table access. An untrusted local user
757 could exploit this flaw to run code in the kernel
758 (ie a root privilege escalation). (CVE-2007-4573).
762 Description: O_DIRECT bypasses client statistics.
763 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
764 read_ahead_stats were updated. copied the stats section from
765 osc_send_oap_rpc() into async_internal().
769 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
773 Description: change order of libsysio includes
774 Details : '#include sysio.h' should always come before '#include xtio.h'
778 Description: handle "serious error: objid * already exists" more gracefully
779 Details : If LAST_ID value on disk is smaller than the objects existing in
780 the O/0/d* directories, it indicates disk corruption and causes an
781 LBUG(). If the object is 0-length, then we should use the existing
782 object. This will help to avoid a full fsck in most cases.
784 Severity : enhancement
786 Description: adapt the lustre_config script to support the upgrade case
787 Details : Add "-u" option for lustre_config script to support upgrading 1.4
788 server targets to 1.6 in parallel.
792 Description: To avoid grant space > avaible space when the disk is almost
793 full. Without this patch you might see the error "grant XXXX >
794 available" or some LBUG about grant, when the disk is almost
796 Details : In filter_check_grant, for non_grant cache write, we should
797 check the left space by if (*left > ungranted + bytes), instead
798 of (*left > ungranted), because only we are sure the left space
799 is enough for another "bytes", then the ungrant space should be
800 increase. In client, we should update cl_avail_grant only there
801 is OBD_MD_FLGRANT in the reply.
805 Description: Update RHEL 4 kernel to fix local root privilege escalation.
806 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
807 described in CVE-2007-4573. This problem could allow untrusted
808 local users to gain root access.
811 Frequency : when using O_DIRECT and quotas
813 Description: Incorrect file ownership on O_DIRECT output files
814 Details : block usage reported by 'lfs quota' does not take into account
815 files that have been written with O_DIRECT.
820 Description: touch file failed when fs is not full
821 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
822 otherwise we can get ENOSP while fs is not full.
826 Description: parallel lock callbacks
827 Details : Instead of sending blocking and completion callbacks as separated
828 requests, adding them to a set and sending in parallel.
831 Frequency : only for Cray XT3
832 Bugzilla : 12829/13455
833 Description: Changing primary group doesn't change the group lustre assigns to
835 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
836 overridden with the primary group provided by the group upcall,
837 whereas the supplied fsgid can be trusted if it is in the list of
838 supplementary groups returned by the group upcall.
840 Severity : enhancement
842 Description: Allow masking D_WARNING, D_ERROR messages from console
843 Details : Console messages can now be disabled via lnet.printk.
847 Description: User code with malformed file open parameter crashes client node
848 Details : Before packing join_file req, all the related reference should be
849 checked carefully in case some malformed flags cause fake
850 join_file req on client.
854 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
855 Details : ldlm_completion_ast() assumes that a lock is granted when the req
856 mode is equal to the granted mode. However, it should also check
857 that LDLM_FL_CP_REQD is not set.
861 Description: Heavy nfs access might result in deadlocks
862 Details : After ELC code landed, it is now improper to enqueue any mds
863 locks under och_sem, because enqueue might want to decide to
864 cancel open locks for same inode we are holding och_sem for.
868 Description: Client eviction while running blogbench
869 Details : A lot of unlink operations with concurrent I/O can lead to a
870 deadlock causing evictions. To address the problem, the number of
871 oustanding OST_DESTROY requests is now throttled to
872 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
873 callbacks are processed in priority.
877 Description: enable ACLs on MDS by default
878 Details : ACLs must be enabled on MDS by default.
881 Frequency : PPC/PPC64 only
883 Description: conflicts between asm-ppc64/types.h and lustre_types.h
884 Details : fix duplicated definitions between asm-ppc64/types.h and
885 lustre_types.h on PPC.
888 Frequency : PPC/PPC64 only
890 Description: asm-ppc/segment.h does not exist
891 Details : fix compile issue on PPC.
895 Description: better handle error messages in extents code
898 Frequency : RHEL4 only
900 Description: mkfs is very slow on IA64/RHEL4
901 Details : A performance regression has been discovered in the MPT Fusion
902 driver between versions 3.02.73rh and 3.02.99.00rh. As a
903 consequence, we have downgraded the MPT Fusion driver in the RHEL4
904 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
906 Severity : enhancement
908 Description: SNMP support enhancement
909 Details : Adding total number of sampled request for an MDS node in snmp
912 Severity : enhancement
914 Description: Optimize ldlm waiting list processing for PR extent locks
915 Details : When processing waiting list for read extent lock and meeting read
916 lock that is same or wider to it that is not contended, skip
917 processing rest of the list and immediatelly return current
918 status of conflictness, since we are guaranteed there are no
919 conflicting locks in the rest of the list.
923 Description: Time out and refuse to reconnect
924 Details : When the failover node is the primary node, it is possible
925 to have two identical connections in imp_conn_list. We must
926 compare not conn's pointers but NIDs, otherwise we can defeat
927 connection throttling.
931 Description: port llog fixes from b1_6 into HEAD
932 Details : Port llog reference couting and some llog cleanups from b1_6
933 (bug 10800) into HEAD, for protect from panic and access to already
934 free llog structures.
938 Description: Detect stride IO mode in read-ahead
939 Details : When a client does stride read, read-ahead should detect that and
940 read-ahead pages according to the detected stride pattern.
944 Description: data checksumming impacts single node performance
945 Details : add support for several checksum algorithm. Currently, only CRC32
946 and Adler-32 are supported. The checksum type can be changed on
947 the fly via /proc/fs/lustre/osc/*/checksum_type.
951 Description: use adler32 for page checksums
952 Details : when available, use the Adler-32 algorithm instead of CRC32 for
957 Description: build for x2 fails
958 Details : fix compile issue on Cray systems.
962 Description: Properly match for duplicate locks
963 Details : Due to different lock order from skiplists code, we need to
964 traverse entire list for now
967 Frequency : only on PPC/SLES10
969 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
970 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
971 be defined. Add a hack in mkfs_lustre.c to work around this compile
976 Description: LASSERT on MDS when client holding flock lock dies
977 Details : ldlm pool logic depends on number of granted locks equal to
978 number of released locks which is not true for flock locks, so
979 just exclude such locks from consideration.
983 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
984 Details : Use fsfilt_sync() for both the whole filesystem sync and
985 individual file sync to eliminate dangerous inode locking
986 with I_LOCK that can lead to a deadlock.
990 Description: performance in 1.6.3
991 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
992 sure that 1MB requests are not fragmented by the block layer.
994 Severity : enhancement
996 Description: organize the server-side client stats on per-nid basis
997 Details : Change the structure of stats under obdfilter and mds to
1005 The "uuid"s file would list the uuids of _active_ exports.
1006 And the clear entry is to clear all stats and stale nids.
1008 Severity : enhancement
1010 Description: eliminate client locks in face of contention
1011 Details : file contention detection and lockless i/o implementation
1012 for contended files.
1016 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1019 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1021 Description: MDS may lose file striping (and hence file data) in some cases
1022 Details : If there are additional extended attributes stored on the MDS,
1023 in particular ACLs, SELinux, or user attributes (if user_xattr
1024 is specified for the client mount options) then there is a risk
1025 of attribute loss. Additionally, the Lustre file striping
1026 needs to be larger than default (e.g. striped over all OSTs),
1027 and an additional attribute must be stored initially in the
1028 inode and then increase in size enough to be moved to the
1029 external attribute block (e.g. ACL growing in size) for file
1032 --------------------------------------------------------------------------------
1034 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1036 * Support for kernels:
1037 2.6.5-7.283 (SLES 9),
1038 2.6.9-55.EL (RHEL 4),
1039 2.6.16.46-0.14 (SLES 10),
1040 2.6.18.8 vanilla (kernel.org)
1041 * Client support for unpatched kernels:
1042 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1043 2.6.16 - 2.6.22 vanilla (kernel.org)
1044 * Due to recently discovered recovery problems, we do not recommend
1045 using patchless RHEL 4 clients with this or any earlier release.
1046 * Recommended e2fsprogs version: 1.39.cfs8
1047 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1048 * Starting with this release, the ldiskfs backing filesystem required
1049 by Lustre is now in its own package, lustre-ldiskfs. This package
1050 should be installed. It is versioned separately from Lustre and
1051 may be released separately in future.
1055 Description: block reactivating mgc import until all deactivates complete
1056 Details : Fix race when failing back MDT/MGS to itself (testing)
1059 Frequency : at statup only
1061 Description: mds_lov_synchronize race leads to various problems
1062 Details : simultaneous MDT->OST connections at startup can cause the
1063 sync to abort, leaving the OSC in a bad state.
1065 Severity : enhancement
1067 Description: add optional extra BUILD_VERSION info
1068 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1069 to override the lustre version.
1072 Frequency : 2.6.18 servers only
1074 Description: ll_kern_mount() doesn't release the module reference
1075 Details : The ldiskfs module reference count never drops down to 0
1076 because ll_kern_mount() doesn't release the module reference.
1081 Description: server LBUG when using old ost_num_threads parameter
1082 Details : Accept the old ost_num_threads parameter but warn that it
1083 is deprecated, and fix an off-by-one error that caused an LBUG.
1088 Description: Transient SCSI error results in persistent IO issue
1089 Details : iobuf->dr_error is not reinitialized to 0 between two
1093 Frequency : sometimes when underlying device returns I/O errors
1095 Description: OSTs not going read-only during write failures
1096 Details : OSTs are not remounted read-only when the journal commit threads
1097 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1098 instead of the ext3 wrappers.
1102 Description: poor connect scaling with increasing client count
1103 Details : Don't run filter_grant_sanity_check for more than 100 exports
1104 to improve scaling for large numbers of clients.
1107 Frequency : SLES10 only
1109 Description: sanity-quota.sh quotacheck failed: rc = -22
1110 Details : Quotas cannot be enabled on SLES10.
1113 Frequency : liblustre clients only
1115 Description: getdirentries does not give error when run on compute nodes
1116 Details : getdirentries does not fail when the size specified as an argument
1117 is too small to contain at least one entry
1119 Severity : enhancement
1121 Description: Add LNET router traceability for debug purposes
1122 Details : If a checksum failure occurs with a router as part of the
1123 IO path, the NID of the last router that forwarded the bulk data
1124 is printed so it can be identified.
1129 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1130 Details : Due to a race condition, liblustre clients were occasionally
1131 evicted incorrectly.
1133 Severity : enhancement
1135 Description: lfs setstripe use optional parameters instead of postional
1138 Severity : enhancement
1140 Description: Nanosecond timestamp support for ldiskfs
1141 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1142 resolution timestamps. There is not yet support for this at
1143 the Lustre filesystem level.
1146 Frequency : during server recovery
1148 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1149 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1153 Frequency : nfs export on patchless client
1155 Description: connectathon hang when test nfs export over patchless client
1156 Details : Disconnected dentry cannot be found with lookup, so we do not need
1157 to unhash it or make it invalid
1160 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1161 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1162 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1167 Description: open req refcounting wrong on reconnect
1168 Details : If reconnect happened between getting open reply from server and
1169 call to mdc_set_replay_data in ll_file_open, we will schedule
1170 replay for unreferenced request that we are about to free.
1171 Subsequent close will crash in variety of ways.
1172 Check that request is still eligible for replay in
1173 mdc_set_replay_data().
1178 Description: disable writes to filesystem when reading health_check file
1179 Details : the default for reading the health_check proc file has changed
1180 to NOT do a journal transaction and write to disk, because this
1181 can cause reads of the /proc file to hang and block HA state
1182 checking on a healthy but otherwise heavily loaded system. It
1183 is possible to return to the previous behaviour during configure
1184 with --enable-health-write.
1186 Severity : enhancement
1188 Description: 64-bit inode version
1189 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1190 to the inode. This will be required for version-based recovery.
1195 Description: MDS fails to start if a duplicate client export is detected
1196 Details : in some rare cases it was possible for a client to connect to
1197 an MDS multiple times. Upon recovery the MDS would detect this
1198 and fail during startup. Handle this more gracefully.
1200 Severity : enhancement
1202 Description: Add -o localflock option to simulate old noflock
1204 Details : This will achieve local-only flock/fcntl locks
1210 Description: log_commit_thread vs filter_destroy race leads to crash
1211 Details : Take import reference before releasing llog record semaphore
1216 Description: Wrong request locking in request set processing
1217 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1218 imp_delayed_list, in this place should be used imp_lock.
1221 Frequency : when reconnection
1223 Description: Grant Leak when osc reconnect to OST
1224 Details : When osc reconnect ost, OST(filter) should check whether it
1225 should grant more space to client by comparing fed_grant and
1226 cl_avail_grant, and return the granted space to client instead
1227 of "new granted" space, because client will call osc_init_grant
1228 to update the client grant space info.
1231 Frequency : when client reconnect to OST
1233 Description: Grant Leak when osc do resend and replay bulk write
1234 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1235 bulk write request, because the grant info while be sync between
1236 OSC and OST when reconnect, and we should ignore the grant info
1237 these of resend/replay write req.
1242 Description: Grant space more than avaiable left space sometimes.
1243 Details : When then OST is about to be full, if two bulk writing from
1244 different clients came to OST. Accord the avaliable space of the
1245 OST, the first req should be permitted, and the second one
1246 should be denied by ENOSPC. But if the seconde arrived before
1247 the first one is commited. The OST might wrongly permit second
1248 writing, which will cause grant space > avaiable space.
1251 Frequency : when client is evicted
1253 Description: Grant might be wrongly erased when osc is evicted by OST
1254 Details : when the import is evicted by server, it will fork another
1255 thread ptlrpc_invalidate_import_thread to invalidate the
1256 import, where the grant will be set to 0. While the original
1257 thread will update the grant it got when connecting. So if
1258 the former happened latter, the grant will be wrongly errased
1259 because of this race.
1264 Description: Checking Stale with correct fid
1265 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1266 to check whether it is stale, because sometimes, we want the
1267 enqueue happened anyway, and op_data.fid2 will not be initialized.
1269 Severity : enhancement
1271 Description: update patchless client
1272 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1275 Frequency : only with 2.4 kernel
1277 Description: random memory corruption
1278 Details : size of struct ll_inode_info is to big for union inode.u and this
1279 can be cause of random memory corruption.
1284 Description: Memory leak in recovery
1285 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1286 It should also check obd_fail before fsfilt_start, otherwise if
1287 fsfilt_start return -EROFS,(failover mds during mds recovery).
1288 then the req will return with repmsg->transno = 0 and rc = EROFS.
1289 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1290 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1291 be freed no matter whether fsfilt_commit success or not.
1294 Frequency : only with huge count clients
1296 Description: Prevents from taking the superblock lock in llap_from_page for
1298 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1299 for llap_from_page call in ll_removepage prevents from taking the
1300 superblock lock for a soon died page.
1305 Description: Not check open intent error before release open handle
1306 Details : in some rare cases, the open intent error is not checked before
1307 release open handle, which may cause
1308 ASSERTION(open_req->rq_transno != 0), because it tries to release
1309 the failed open handle.
1314 Description: Set cat log bitmap only after create log success.
1315 Details : in some rare cases, the cat log bitmap is set too early. and it
1316 should be set only after create log success.
1320 Description: Accessing a block bevice can re-enable I/O when Lustre is
1321 tearing down a device.
1322 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1327 Description: service threads may hog cpus when there are a lot of requests
1328 Details : Insert cond_resched to give other threads a chance to use some CPU
1333 Description: the cat log was not initialized in recovery
1334 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1335 unlink log on mds will not be initialized until mds post
1336 recovery. And also in mds post recovery, the unlink log will
1337 initialization will be done asynchronausly, so there will be race
1338 between add unlink log and unlink log initialization.
1342 Description: brw_stats were being printed incorrectly
1343 Details : brw_stats were being printed as log2 but all of them were not
1344 recorded as log2. Also remove some code duplication arising from
1345 filter_tally_{read,write}.
1349 Frequency : rare, only in recovery.
1350 Description: ASSERTION(req->rq_type != LI_POISON) failed
1351 Details : imp_lock should be held while iterating over imp_sending_list for
1352 prevent destroy request after get timeout in ptlrpc_queue_wait.
1356 Description: replay-single.sh test 52 fails
1357 Details : A lock's skiplist need to be cleanup when it being unlinked
1358 from its resource list.
1362 Description: Short directio read returns full requested size rather than
1364 Details : Direct I/O operations should return actual amount of bytes
1365 transferred rather than requested size.
1367 Severity : enhancement
1369 Description: metadata RPC reduction (e.g. for rm performance)
1370 Details : decrease the amount of synchronous RPC between clients and servers
1371 by canceling conflicing lock before the operation on the client
1372 and packing thier handles into the main operation RPC to server.
1374 Severity : enhancement
1376 Description: Async OSC create to avoid the blocking unnecessarily.
1377 Details : If a OST has no remain object, system will block on the creating
1378 when need to create a new object on this OST. Now, ways use
1379 pre-created objects when available, instead of blocking on an
1380 empty osc while others are not empty. If we must block, we block
1381 for the shortest possible period of time.
1383 Severity : enhancement
1385 Description: refine locking for avoid write wrong info into lov_objid file
1386 Details : fix possible races with add new target and write/update data in
1389 --------------------------------------------------------------------------------
1391 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1396 Frequency : on some architectures
1398 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1399 Details : Uninitialized flags sometimes cause configuration commands to
1403 Frequency : patchless clients only
1405 Description: missing __iget() symbol export
1406 Details : The __iget() symbol export is missing. To avoid the need for
1407 this on patchless clients the deathrow inode reaper is turned
1408 off, and we depend on the VM to clean up old inodes. This
1409 dependency was during via the fix for bug 12181.
1411 --------------------------------------------------------------------------------
1413 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1415 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1416 INTEROPERATE with older versions automatically. In many cases a
1417 special upgrade step is needed. Please read the
1418 user documentation before upgrading any part of a live system.
1419 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1420 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1421 * WARNING: Lustre configuration and startup changes are required with
1422 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1428 Severity : enhancement
1430 Description: MountConf
1431 Details : Lustre configuration is now managed via mkfs and mount
1432 commands instead of lmc and lconf. New obd types (MGS, MGC)
1433 are added for dynamic configuration management. See
1434 https://mail.clusterfs.com/wikis/lustre/MountConf for
1437 Severity : enhancement
1439 Description: dynamic OST addition
1440 Details : OSTs can now be added to a live filesystem
1442 Severity : enhancement
1444 Description: startup order invariance
1445 Details : MDTs and OSTs can be started in any order. Clients only
1446 require the MDT to complete startup.
1448 Severity : enhancement
1450 Description: parallel, asynchronous orphan cleanup
1451 Details : orphan cleanup is now performed in separate threads for each
1452 OST, allowing parallel non-blocking operation.
1454 Severity : enhancement
1456 Description: optimized stripe assignment
1457 Details : stripe assignments are now made based on ost space available,
1458 ost previous usage, and OSS previous usage, in order to try
1459 to optimize storage space and networking resources.
1461 Severity : enhancement
1463 Description: Permanently set tunables
1464 Details : All writable /proc/fs/lustre tunables can now be permanently
1465 set on a per-server basis, at mkfs time or on a live system.
1467 Severity : enhancement
1469 Description: Lustre message v2
1470 Details : Add lustre message format v2.
1472 Severity : enhancement
1474 Description: client OST exclusion list
1475 Details : Clients can be started with a list of OSTs that should be
1476 declared "inactive" for known non-responsive OSTs.
1480 Description: SPEC SFS validation failure on NFS v2 over lustre.
1481 Details : Changes the blocksize for regular files to be 2x RPC size,
1482 and not depend on stripe size.
1484 Severity : enhancement
1486 Description: Multiple MD RPCs in flight.
1487 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1488 To avoid overly-overloading MDS, introduce a limit on number of
1489 MDS RPCs in flight for a single client and add /proc controls
1490 to adjust this limit.
1492 Severity : enhancement
1494 Description: client read/write statistics
1495 Details : Add client read/write call usage stats for performance
1496 analysis of user processes.
1497 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1498 file access. extents_stats shows chunk size distribution.
1499 extents_stats_per_process show chunk size distribution per
1502 Severity : enhancement
1504 Description: mds statistics
1505 Details : Add detailed mds operations statistics in
1506 /proc/fs/lustre/mds/*/stats.
1510 Description: Failure of copying files with lustre special EAs.
1511 Details : Client side always return success for setxattr call for lustre
1512 special xattr (currently only "trusted.lov").
1517 Description: Refcount LNET uuids
1518 Details : The global LNET uuid list grew linearly with every startup;
1519 refcount repeated list entries instead of always adding to
1522 Severity : enhancement
1524 Description: Dynamic service threads
1525 Details : Within a small range, start extra service threads
1526 automatically when the request queue builds up.
1529 Frequency : mixed-endian client/server environments
1531 Description: mixed-endian crashes
1532 Details : The new msg_v2 system had some failures in mixed-endian
1535 Severity : enhancement
1537 Description: Easy OST removal
1538 Details : OSTs can be permanently deactivated with e.g. 'lctl
1539 conf_param lustre-OST0001.osc.active=0'
1541 Severity : enhancement
1543 Description: MGS proc entries
1544 Details : Added basic proc entries for the MGS showing what filesystems
1547 Severity : enhancement
1549 Description: provide MGS failover
1550 Details : Added config lock reacquisition after MGS server failover.
1552 Severity : enhancement
1554 Description: add Linux 2.4 support
1555 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1559 Description: a large application tries to do I/O to the same resource and dies
1560 in the middle of it.
1561 Details : Check the req->rq_arrival time after the call to
1562 ost_brw_lock_get(), but before we do anything about
1563 processing it & sending the BULK transfer request. This
1564 should help move old stale pending locks off the queue as
1565 quickly as obd_timeout.
1568 Frequency : when an incorrect nid is specified during startup
1570 Description: ptlrpc connect to non-existant node causes kernel crash
1571 Details : LNET can't be re-entered from an event callback, which
1572 happened when we expire a message after the export has been
1573 cleaned up. Instead, hand the zombie cleanup off to another
1576 Severity : enhancement
1578 Description: plain/inodebits lock performance improvement
1579 Details : Grouping plain/inodebits in granted list by their request modes
1580 and bits policy, thus improving the performance of search through
1584 Frequency : only if OST filesystem is corrupted
1586 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1587 Details : for a short time RPCs with bulk IO are in the replay list,
1588 but replay of bulk IOs is unimplemented. If the OST filesystem
1589 is corrupted due to disk cache incoherency and then replay is
1590 started it is possible to trip an assertion. Avoid putting
1591 committed RPCs into the replay list at all to avoid this issue.
1594 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1596 Description: System hang on startup
1597 Details : This bug allowed the liblustre (e.g. catamount) client to
1598 return to the app before handling all startup RPCs. This
1599 could leave the node unresponsive to lustre network traffic
1600 and manifested as a server ptllnd timeout.
1602 Severity : enhancement
1604 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1605 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1606 boolean to control whether to print peer debug info when a
1607 client's RPC times out.
1610 Frequency : only for kernels with patches from Lustre below 1.4.3
1612 Description: Remove old rdonly API
1613 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1616 Frequency : only for devices with external journals
1618 Description: Set external device read-only also
1619 Details : During a commanded failover stop, we set the disk device
1620 read-only while the server shuts down. We now also set any
1621 external journal device read-only at the same time.
1624 Frequency : when upgrading from 1.4 while trying to change parameters
1626 Description: The wrong (new) MDC name was used when setting parameters for
1627 upgraded MDT's. Also allows changing of OSC (and MDC)
1628 parameters if --writeconf is specified at tunefs upgrade time.
1631 Frequency : when setting specific ost indicies
1633 Description: QOS code breaks on skipped indicies
1634 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1635 created with --index need not be sequential.
1637 Severity : enhancement
1639 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1640 Details : The uninit_groups feature works in conjunction with the kernel
1641 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1642 up the pass1 processing of e2fsck. This is a read-only feature
1643 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1644 mount filesystems that have had this feature enabled.
1646 Severity : enhancement
1648 Description: Improve multi-block allocation algorithm to avoid fragmentation
1649 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1650 allocation locality and avoid filesystem fragmentation.
1652 ------------------------------------------------------------------------------
1654 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1656 * Support for kernels:
1657 2.4.21-47.0.1.EL (RHEL 3)
1658 2.6.5-7.283 (SLES 9)
1659 2.6.9-42.0.10.EL (RHEL 4)
1660 2.6.12.6 vanilla (kernel.org)
1661 2.6.16.27-0.9 (SLES 10)
1662 * Recommended e2fsprogs version: 1.39.cfs5
1664 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1668 Frequency : occasional, depends on client load and configuration
1669 Bugzilla : 12181, 12203
1670 Description: data loss for recently-modified files
1672 Details : In some cases it is possible that recently written or created
1673 files may not be written to disk in a timely manner (this should
1674 normally be within 30s unless client IO load is very high).
1675 The problem appears as zero-length files or files that are a
1676 multiple of 1MB in size after a client crash or client eviction
1677 that are missing data at the end of the file.
1679 This problem is more likely to be hit on clients where files are
1680 repeatedly created and unlinked in the same directory, clients
1681 have a large amount of RAM, have many CPUs, the filesystem has
1682 many OSTs, the clients are rebooted frequently, and/or the files
1683 are not accessed by other nodes after being written.
1685 The presence of the problem can be detected by looking at
1686 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1687 smaller than the second (nr_unused) then dirty files will not
1688 be flushed automatically to disk. "sync; sleep 10" should be
1689 run several times on the node before unmounting it to update
1690 Lustre (this is also safe to run on nodes without this problem).
1692 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1693 that can cause this same problem, so customers using that kernel
1694 also need to update the kernel in addition to Lustre. In order
1695 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1697 It is normal that files written just before a client crash (less
1698 than 30s) may not yet have been flushed to disk, even for local
1702 Frequency : frequent on thin XT3 nodes
1704 Description: UUID collision on thin XT3 Linux nodes
1705 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1706 randomly, since we relied on an insufficiently-seeded PRNG.
1711 Description: OSS hangs after "All ost request buffers busy"
1712 Details : A deadlock between quota and journal operations caused OSS
1713 hangs after printing "All ost request buffers busy."
1716 Frequency : always on liblustre builds
1718 Description: Cleanup compiler warnings on liblustre
1721 Frequency : always on liblustre builds on XT3
1723 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1727 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1728 > 32000 subdirectories
1729 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1730 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1731 subdirectory count crosses 32000. This will aid e2fsck to
1732 correctly handle more than 32000 subdirectories.
1735 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1737 Description: System hang on startup
1738 Details : This bug allowed the liblustre (e.g. catamount) client to
1739 return to the app before handling all startup RPCs. This
1740 could leave the node unresponsive to lustre network traffic
1741 and manifested as a server ptllnd timeout.
1743 Severity : enhancement
1745 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1746 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1747 boolean to control whether to print peer debug info when a
1748 client's RPC times out.
1753 Description: make O_SYNC working on 2.6 kernels
1754 Details : 2.6 kernels use different method for mark pages for write,
1755 so need add a code to lustre for O_SYNC work.
1760 Description: Failure to close file and release space on NFS
1761 Details : Put inode details into lock acquired in ll_intent_file_open.
1762 Use mdc_intent_lock in ll_intent_open to properly
1763 detect all kind of errors unhandled by mdc_enqueue.
1768 Description: proc file read during shutdown sometimes raced obd removal,
1770 Details : Add lock to prevent obd access after proc file removal.
1773 Frequency : Only for files larger than 4GB on 32-bit clients.
1775 Description: improperly doing page alignment of locks
1776 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1777 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1782 Description: RPCs being resent when they shouldn't be
1783 Details : Some RPCs that should not be resent are being resent. This
1784 can cause inconsistencies in the RPC state machine. Do not
1785 resend such requests.
1788 Frequency : rare, only with NFS export
1790 Description: Crash on NFS re-export node
1791 Details : under very unusual load conditions an assertion is hit in
1792 ll_intent_file_open()
1795 Frequency : only if OST filesystem is corrupted
1797 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1798 Details : for a short time RPCs with bulk IO are in the replay list,
1799 but replay of bulk IOs is unimplemented. If the OST filesystem
1800 is corrupted due to disk cache incoherency and then replay is
1801 started it is possible to trip an assertion. Avoid putting
1802 committed RPCs into the replay list at all to avoid this issue.
1807 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1808 Details : Large single O_DIRECT read and write calls can fail to allocate
1809 a sufficiently large buffer to process the request. In case of
1810 allocation failure the allocation is retried with a smaller
1811 buffer and broken into smaller requests.
1813 Severity : enhancement
1815 Description: Add -o localflock option to simulate old noflock behaviour.
1816 Details : This will achieve local-only flock/fcntl locks coherentness.
1821 Description: versioning check is incomplete
1822 Details : Checking the version difference of client vs. server, report
1823 error if the gap is too big.
1828 Description: add support PG_writeback bit
1829 Details : add support for PG_writeback bit for Lustre, for more carefull
1830 work with page cache in 2.6 kernel. This also fix some deadlocks
1831 and remove hack for work O_SYNC with 2.6 kernel.
1833 Severity : enhancement
1835 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1836 Details : The uninit_groups feature works in conjunction with the kernel
1837 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1838 up the pass1 processing of e2fsck. This is a read-only feature
1839 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1840 mount filesystems that have had this feature enabled.
1842 Severity : enhancement
1844 Description: Improve multi-block allocation algorithm to avoid fragmentation
1845 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1846 allocation locality and avoid filesystem fragmentation.
1848 ------------------------------------------------------------------------------
1850 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1852 * Support for kernels:
1853 2.6.9-42.0.3.EL (RHEL 4)
1854 2.6.5-7.276 (SLES 9)
1855 2.4.21-47.0.1.EL (RHEL 3)
1856 2.6.12.6 vanilla (kernel.org)
1857 2.6.16.21-0.8 (SLES10)
1858 * Recommended e2fsprogs version: 1.39.cfs2-0
1860 * The backwards-compatible /proc/sys/portals symlink has been removed
1861 in this release. Before upgrading, please ensure that you change
1862 any configuration scripts or /etc/sysctl.conf files that access
1863 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1864 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1865 in advance of the upgrade on any system running Lustre 1.4.6 or
1866 newer, since /proc/sys/lnet was added in that version.
1867 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1871 Frequency : only when quota is used
1873 Description: avoid scanning export list for quota master
1874 Details : Change the algorithms to avoid scanning export list in order
1875 to improve the efficiency.
1878 Frequency : MDS failover only, very rarely
1880 Description: "went back in time" messages on mds failover
1881 Details : The greatest transno may be lost when the current operation
1882 finishes with an error (transno==0) and the client's last_rcvd
1883 record is over-written. Save the greatest transno in the
1884 mds_last_transno for this case.
1887 Frequency : always for specific kernels and striping counts
1889 Description: client may get "Matching packet too big" without ACL support
1890 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1891 when trying to access files in certain configurations. The
1892 clients should in fact be denied when mounting because they do
1893 not understand ACLs.
1896 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1898 Description: many clients connecting with IO in progress causes connect timeouts
1899 Details : Avoid synchronous journal commits to avoid delays caused by many
1900 clients connecting/disconnecting when bulk IO is in progress.
1901 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1902 OST_IO_PORTAL to avoid delays behind potentially many pending
1906 Frequency : occasionally with multiple writers to a single file
1908 Description: shared writes to file may result in wrong size reported by stat()
1909 Details : Allow growing of kms when extent lock is cancelled
1912 Frequency : always with random mmap IO to multi-striped file
1914 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1915 Details : Only if the hole is at the end of OST object so that kms is too
1916 small. Fix is to increase kms accordingly in ll_nopage.
1919 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1921 Description: writes to a missing object would leak memory on the OST
1922 Details : If there is an inconsistency between the MDS and OST filesystems,
1923 such that the MDS references an object that doesn't exist, writes
1924 to that object will leak memory due to incorrect cleanup in the
1925 error handling path, eventually running out of memory on the OST.
1930 Description: Creating too long symlink causes lustre errors
1931 Details : Check symlink and name lengths before sending requests to MDS.
1934 Frequency : only if flock is enabled (not on by default)
1936 Description: posix locks not released on fd closure on 2.6.9+
1937 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1938 kernels, this caused such locks not to be released on fd close and
1939 then assertions on fs unmount about still used locks.
1942 Frequency : MDS failover only, very rarely
1944 Description: clients may get ASSERTION(granted_lock != NULL)
1945 Details : When request was taking a long time, and a client was resending
1946 a getattr by name lock request. The were multiple lock requests
1947 with the same client lock handle and
1948 mds_getattr_name->fixup_handle_for_resent_request found one of the
1949 lock handles but later failed with ASSERTION(granted_lock != NULL).
1954 Description: handle->h_buffer_credits > 0, assertion failure
1955 Details : h_buffer_credits is zero after truncate, causing assertion
1956 failure. This patch extends the transaction or creates a new
1960 Frequency : NFS re-export or patchless client
1961 Bugzilla : 11179, 10796
1962 Description: Crash on NFS re-export node (__d_move)
1963 Details : We do not want to hash the dentry if we don't have a lock.
1964 But if this dentry is later used in d_move, we'd hit uninitialised
1965 list head d_hash, so we just do this to init d_hash field but
1966 leave dentry unhashed.
1969 Frequency : NFS re-export or patchless client
1971 Description: NFS exports has problem with symbolic link
1972 Details : lustre client didn't properly install dentry when re-exported
1973 to NFS or running patchless client.
1976 Frequency : NFS re-export or patchless client
1978 Description: Various nfs/patchless fixes.
1979 Details : fixes reuse disconected alias for lookup process - this fixes
1980 warning "find_exported_dentry: npd != pd",
1981 fix permission error with open files at nfs.
1982 fix apply umask when do revalidate.
1985 Frequency : occasional
1987 Description: Crash on NFS re-export node
1988 Details : calling clear_page() on the wrong pointer triggered oops in
1989 generic_mapping_read().
1992 Frequency : rarely, using O_DIRECT IO
1994 Description: unaligned directio crashes client with LASSERT
1995 Details : check for unaligned buffers before trying any requests.
1998 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2000 Description: stale data returned from RAID cache
2001 Details : If only a small amount of IO is done to the RAID device before
2002 reading it again it is possible to get stale data from the RAID
2003 cache instead of reading it from disk.
2006 Frequency : always for sles10 kernel
2008 Description: sles10 support
2009 Details : ll_follow_link: compile fixes and using of nd_set_link
2010 under newer kernels.
2013 Frequency : depends on arch, kernel and compiler version, always on sles10
2016 Description: recursive or deep enough symlinks cause stack overflow
2017 Details : getting rid of large stack-allocated variable in
2021 Frequency : depends on hardware
2023 Description: lustre write performance loss in the SLES10 kernel
2024 Details : the performance loss is caused by using of write barriers in the
2025 ext3 code. The SLES10 kernel turns barrier support on by
2026 default. The fix is to undo that change for ldiskfs.
2028 ------------------------------------------------------------------------------
2030 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2032 * Support for kernels:
2033 2.6.9-42.0.3EL (RHEL 4)
2034 2.6.5-7.276 (SLES 9)
2035 2.4.21-47.0.1.EL (RHEL 3)
2036 2.6.12.6 vanilla (kernel.org)
2040 Frequency : quota enabled and large files being deleted
2042 Description: releasing more than 4GB of quota at once hangs OST
2043 Details : If a user deletes more than 4GB of files on a single OST it
2044 will cause the OST to spin in an infinite loop. Release
2045 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2050 Description: statfs data retrieved from /proc may be stale or zero
2051 Details : When reading per-device statfs data from /proc, in the
2052 {kbytes,files}_{total,free,avail} files, it may appear
2053 as zero or be out of date.
2056 Frequency : systems with MD RAID1 external journal devices
2058 Description: lconf's call to blkid is confused by RAID1 journal devices
2059 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2060 of returning all block devices that match the journal UUID.
2063 Frequency : always, for aggregate stripe size over 4GB
2065 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2066 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2067 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2070 Frequency : always if "lfs find" used on a local file/directory
2072 Description: "lfs find" segfaults if used on a local file/directory
2073 Details : The case where a directory component was not specified wasn't
2074 handled correctly. Handle this properly.
2077 Frequency : always on ppc64
2079 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2080 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2081 perform endianness conversion.
2084 Frequency : rarely (truncate to non-zero file size after write under load)
2085 Bugzilla : 10730, 10687
2086 Description: Files padded with zeros to next 4K multiple
2087 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2088 it is possible that files that are truncated to a non-zero size
2089 immediately after being written are filled with zero bytes beyond
2090 the truncated size. No file data is lost.
2092 Severity : enhancement
2094 Description: Allow recovery/failover for liblustre clients.
2095 Details : liblustre clients were unaware of failover configurations until
2098 Severity : enhancement
2100 Description: user file locks should fail when not mounting with flock option
2101 Details : Set up an error-returning stub in ll_file_operations.lock field
2102 to prevent incorrect behaviour when client is mounted without
2103 flock option. Also, set up properly f_op->flock field for
2107 Frequency : always on ia64
2109 Description: "lfs df" loops on printing out MDS statfs information
2110 Details : The obd_ioctl_data was not initialized and in some systems
2111 this caused a failure during the ioctl that did not return
2112 an error. Initialize the struct and return an error on failure.
2115 Frequency : SLES 9 only
2117 Description: Error of copying files with lustre special EAs as root
2118 Details : Client side always return success for setxattr call for lustre
2119 special xattr (currently only "trusted.lov").
2122 Frequency : rarely on clusters with both ia64+i386 clients
2124 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2125 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2126 clients are doing concurrent writes to the same file it is
2127 possible that the ia64 clients may LASSERT because the OST
2128 extent locks are not PAGE_SIZE aligned. Ensure that grown
2129 locks are always aligned on the request boundary.
2132 Frequency : specific use, occasional
2134 Description: Overwriting in use executable truncates on-disk binary image
2135 Details : If one node attempts to overwrite an executable in use by
2136 another node, we now correctly return ETXTBSY instead of
2137 truncating the file.
2139 Severity : enhancement
2141 Description: Async OSC create to avoid the blocking unnecessarily.
2142 Details : If a OST has no remain object, system will block on the creating
2143 when need to create a new object on this OST. Now, ways use
2144 pre-created objects when available, instead of blocking on an
2145 empty osc while others are not empty. If we must block, we block
2146 for the shortest possible period of time.
2151 Description: chmod on Lustre root is propagated to other clients
2152 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2153 invalid by the follow_mount time.
2156 Frequency : liblustre clients only
2158 Description: Race in 'instant cancel' lock handling could lead to such locks
2159 never to be granted in case of SMP MDS
2160 Details : Do not destroy not yet granted but cbpending locks in
2164 Frequency : replay/resend of open
2166 Description: non null lock assetion failure in mds_intent_policy
2167 Details : Trying to replay/resend lockless open requests resulted in
2168 mds_open() returning 0 with no lock. Now it sets a flag if
2169 a lock is going to be returned.
2171 Severity : enhancement
2173 Description: Checksum enhancements
2174 Details : New checksum enhancements allow for resending RPCs that failed
2177 Severity : enhancement
2179 Description: Tunables on number of dirty pages in cacche
2180 Details : Allow to set limit on number of dirty pages cached.
2185 Description: client crash on unmount - lock still has references
2186 Details : In some error handling cases it was possible to leak a lock
2187 reference on a client while accessing a file. This was not
2188 harmful to the client during operation, but would cause the
2189 client to crash when the filesystem is unmounted.
2192 Frequency : specific case, rare
2194 Description: ETXTBSY on mds though file not in use by client
2195 Details : ETXTBSY is no longer incorrectly returned when attempting to
2196 chmod or chown a directory that the user previously tried to
2197 execute or a currently-executing binary.
2200 Frequency : extremely rare except on liblustre-based clients
2202 Description: Lustre space not freed when files are deleted
2203 Details : Clean up open-unlinked files after client eviction. Previously
2204 the unlink was skipped and the files remained as orphans.
2209 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2210 Details : In some cases it was possible to send a blocking callback to
2211 a client doing a glimpse, even though that client didn't get
2212 a lock granted. When the glimpse lock is cancelled on the OST
2213 the freed lock is left on the waiting list and corrupted the list.
2216 Frequency : all core dumps
2218 Description: Broke core dumps to lustre
2219 Details : Negative dentry may be unhashed if parent does not have UPDATE
2220 lock, but some callers, e.g. do_coredump, expect dentry to be
2221 hashed after successful create, hash it in ll_create_it.
2223 ------------------------------------------------------------------------------
2225 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2227 * Support for kernels:
2228 2.6.9-42.0.2.EL (RHEL 4)
2229 2.6.5-7.276 (SLES 9)
2230 2.4.21-40.EL (RHEL 3)
2231 2.6.12.6 vanilla (kernel.org)
2235 Frequency : always on RHEL 3
2237 Description: Number of open files grows over time
2238 Details : The number of open files grows over time, whether or not
2239 Lustre is started. This was due to a filp leak introduced
2240 by one of our kernel patches.
2242 ------------------------------------------------------------------------------
2244 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2246 * Support for kernels:
2247 2.6.9-42.EL (RHEL 4)
2248 2.6.5-7.276 (SLES 9)
2249 2.4.21-40.EL (RHEL 3)
2250 2.6.12.6 vanilla (kernel.org)
2255 Bugzilla : 5719, 9635, 9792, 9684
2256 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2257 Details : If a server is under heavy load and cannot reply to new
2258 connection requests before the client resends the (re)connect,
2259 the connection handling code can behave badly if two service
2260 threads are concurrently handing separate (re)connections from
2261 the same client. Add better locking to the connection handling
2262 code, and ensure that only a single connection will be processed
2263 for a given client UUID, even if the lock is dropped.
2265 Severity : enhancement
2267 Description: add TCP zero-copy support to kernel
2268 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2269 sends if the hardware supports scatter-gather and checksumming.
2270 This allows socklnd to do client-write and server-read more
2271 efficiently and reduce CPU utilization from skbuf copying.
2274 Frequency : only if NFS exporting from client
2276 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2277 Details : If setting attributes on a file created under NFS that had
2278 never been opened it would be possible to oops the client
2279 if the file had no objects.
2282 Frequency : always for liblustre
2284 Description: liblustre client does MDS+OSTs setattr RPC for each write
2285 Details : When doing a write from a liblustre client, the client
2286 incorrectly issued an RPC to the MDS and each OST the file was
2287 striped over in order to update the timestamps. When writing
2288 with small chunks and many clients this could overwhelm the MDS
2289 with RPCs. In all cases it would slow down the write because
2290 these RPCs are unnecessary.
2292 Severity : enhancement
2294 Description: allow number of MDS service threads to be changed at module load
2295 Details : It is now possible to change the number of MDS service threads
2296 running. Adding "options mds mds_num_threads={N}" to the MDS's
2297 /etc/modprobe.conf will set the number of threads for the next
2298 time Lustre is restarted (assuming the "mds" module is also
2299 reloaded at that time). The default number of threads will
2300 stay the same, 32 for most systems.
2305 Description: OST crash if filesystem is unformatted or corrupt
2306 Details : If an OST is started on a device that has never been formatted
2307 or if the filesystem is corrupt and cannot even mount then the
2308 error handling cleanup routines would dereference a NULL pointer.
2313 Description: NULL pointer deref in llap_from_page.
2314 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2315 incorrect) mapping if the page was truncated/reclaimed while it was
2316 searched for. Check for this condition and skip such pages when
2317 doing readahead. Introduce extra check to llap_from_page() to
2318 verify page->mapping->host is non-NULL (so page is not anonymous).
2321 Frequency : Sometimes when using sys_sendfile
2323 Description: "page not covered by a lock" warnings from ll_readpage
2324 Details : sendfile called ll_readpage without right page locks present.
2325 Now we introduced ll_file_sendfile that does necessary locking
2326 around call to generic_file_sendfile() much like we do in
2330 Frequency : with certain MDS communication failures at client mount time
2332 Description: NULL pointer deref after failed client mount
2333 Details : a client connection request may delayed by the network layer
2334 and not be sent until after the PTLRPC layer has timed out the
2335 request. If the client fails the mount immediately it will try
2336 to clean up before the network times out the request. Add a
2337 reference from the request import to the obd device and delay
2338 the cleanup until the network drops the request.
2341 Frequency : occasionally during client (re)connect
2343 Description: assertion failure during client (re)connect
2344 Details : processing a client connection request may be delayed by the
2345 client or server longer than the client connect timeout. This
2346 causes the client to resend the connection request. If the
2347 original connection request is replied in this interval, the
2348 client may trip an assertion failure in ptlrpc_connect_interpret()
2349 which thought it would be the only running connect process.
2352 Frequency : only with obd_echo servers and clients that are rebooted
2354 Description: kernel BUG accessing uninitialized data structure
2355 Details : When running an obd_echo server it did not start the ping_evictor
2356 thread, and when a client was evicted an uninitialized data
2357 structure was accessed. Start the ping_evictor in the RPC
2358 service startup instead of the OBD startup.
2360 Severity : enhancement
2361 Bugzilla : 10193 (patchless)
2362 Description: Remove dependency on various unexported kernel interfaces.
2363 Details : No longer need reparent_to_init, exit_mm, exit_files,
2364 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2367 Frequency : rare (only users of deprecated and unsupported LDAP config)
2369 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2370 Details : LDAP apparently contains 'lustreName' attributes instead of
2371 'name'. A simple remapping of the name is sufficient.
2374 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2376 Description: waiting_locks_callback trips kernel BUG if client is evicted
2377 Details : Running with the dump_on_timeout debug flag turned on makes
2378 it possible that the waiting_locks_callback() can try to dump
2379 the Lustre kernel debug logs from an interrupt handler. Defer
2380 this log dumping to the expired_lock_main() thread.
2382 Severity : enhancement
2384 Description: Support NFS exporting on 2.6 kernels.
2385 Details : Implement non-rawops metadata methods for NFS server to use without
2386 changing NFS server code.
2389 Frequency : very rare (synthetic metadata workload only)
2391 Description: two racing renames might cause an MDS thread to deadlock
2392 Details : Running the "racer" program may cause one MDS thread to rename
2393 a file from being the source of a rename to being the target of
2394 a rename at exactly the same time that another thread is doing
2395 so, and the second thread has already enqueued these locks after
2396 doing a lookup of the target and is trying to relock them in
2397 order. Ensure that we don't try to re-lock the same resource.
2400 Frequency : only very large systems with liblustre clients
2402 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2403 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2404 all OSTs in parallel. This allows the eviction of stale liblustre
2405 clients to proceed much faster than if they were done in series,
2406 and also offers similar improvements for other set_info RPCs.
2411 Description: excessive CPU usage during initial read phase on client
2412 Details : During the initial read phase on a client, it would agressively
2413 retry readahead on the file, consuming too much CPU and impacting
2414 performance (since 1.4.5.8). Improve the readahead algorithm
2415 to avoid this, and also improve some other common cases (read
2416 of small files in particular, where "small" is files smaller than
2417 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2422 Description: MDS crash when receiving packet with unknown intent.
2423 Details : Do not LBUG in unknown intent case, just return -EFAULT
2425 Severity : enhancement
2426 Bugzilla : 9293, 9385
2427 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2428 Details : Do not serialize getattr (non-intent version) and statfs.
2431 Frequency : occasional, when OST network is overloaded/intermittent
2433 Description: client evicted by OST after bulk IO timeout
2434 Details : If a client sends a bulk IO request (read or write) the OST
2435 may evict the client if it is unresposive to its data GET/PUT
2436 request. This is incorrect if the network is overloaded (takes
2437 too long to transfer the RPC data) or dropped the OST GET/PUT
2438 request. There is no need to evict the client at all, since
2439 the pinger and/or lock callbacks will handle this, and the
2440 client can restart the bulk request.
2443 Frequency : Always when mmapping file with no objects
2445 Description: client crashes when mmapping file with no objects
2446 Details : Check that we actually have objects in a file before doing any
2447 operations on objects in ll_vm_open, ll_vm_close and
2453 Description: Request leak when working with deleted CWD
2454 Details : Introduce advanced request refcount tracking for requests
2455 referenced from lustre intent.
2457 Severity : Enhancement
2459 Description: Cache open file handles on client.
2460 Details : MDS now will return special lock along with openhandle, if
2461 requested and client is allowed to hold openhandle, even if unused,
2462 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2463 closing files for every read/write openration.
2465 Severity : Enhancement
2467 Description: Cache open negative dentries on client when possible.
2468 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2469 negative dentries on lock revocation.
2474 Description: Remounting a client read-only wasn't possible with a zconf mount
2475 Details : It wasn't possible to remount a client read-only with llmount.
2477 Severity : enhancement
2478 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2479 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2480 MPICH is included to add Lustre-specific ADIO interfaces.
2481 This is based closely on the UFS ADIO layer and only differs
2482 in file creation, in order to allow the OST striping to be set.
2483 This is user-contributed code and not supported by CFS.
2488 Description: extended inode attributes (immutable, append-only) work improperly
2489 when 2.4 and 2.6 kernels are used on client/server or vice versa
2490 Details : Introduce kernel-independent values for these flags.
2492 Severity : enhancement
2495 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2496 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2497 granularity. Now, Fractional MB granularity is be supported,
2498 this is very useful for low memory system.
2500 Severity : enhancement
2502 Description: Getattr by fid
2503 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2504 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2505 have locks and avoids getattr rpc requests for it on every stat.
2508 Frequency : Always, for filesystems larger than 2TB
2510 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2511 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2512 filesystems larger than 2TB to be mounted with mballoc enabled.
2515 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2517 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2518 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2519 filesystems larger than 2TB is mounted then the kernel will
2520 silently corrupt the start of the filesystem. CONFIG_LBD is
2521 enabled for all CFS-supported kernels, but the possibility of
2522 this happening with a modified kernel config exists.
2524 Severity : enhancement
2526 Description: add client O_DIRECT support for 2.6 kernels
2527 Details : It is now possible to do O_DIRECT reads and writes to files
2528 in the Lustre client mountpoint on 2.6 kernel clients.
2530 Severity : enhancement
2532 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2533 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2534 parallel, not waiting for response from every OST before sending
2535 a rpc to the next OST.
2540 Description: setattr vs write race when updating file timestamps
2541 Details : Client processes that update a file timestamp into the past
2542 right after writing to the file (e.g. tar) it is possible that
2543 the updated file modification time can be reset to the current
2544 time due to a race between processing the setattr and write RPC.
2546 Severity : enhancement
2548 Description: Bring 'lfs find' closer in line with regular Linux find.
2549 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2550 -print0 options and obtains all the needed info through the lustre
2553 Severity : enhancement
2555 Description: support up to 1024 configured devices on one node
2556 Details : change obd_dev array from statically allocated to dynamically
2557 allocated structs as they are first used to reduce memory usage
2562 Description: Flush dirty partially truncated pages during truncate
2563 Details : Immediatelly flush partially truncated pages in filter_setattr,
2564 this way we completely avoid having any pages in page cache on OST
2565 and can retire ugly workarounds during writes to flush such pages.
2570 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2571 Details : protect inode from truncation within vfs_unlink() context
2572 just take a reference before calling vfs_unlink() and release it
2573 when parent's i_sem is free.
2578 Description: last_id value checked outside lock on OST caused LASSERT failure
2579 Details : If there were multiple MDS->OST object precreate requests in
2580 flight, it was possible that the OST's last object id was checked
2581 outside a lock and incorrectly tripped an assertion. Move checks
2582 inside locks, and discard old precreate requests.
2585 Frequency : always, if extents are used on OSTs
2587 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2588 Details : OSTs using the extents format would not zero the high 16 bits of
2589 the index physical block number. This is not a problem for any
2590 OST filesystems smaller than 16TB, and no kernels support ext3
2591 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2592 new/modified files) and can be fixed for existing filesystems
2593 with e2fsprogs-1.39-cfs1.
2598 Description: import connection selection may be incorrect if timer wraps
2599 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2600 import connections to be ignored if the 32-bit jiffies counter
2601 wraps. Use a 64-bit jiffies counter.
2604 Frequency : very large clusters immediately after boot
2606 Description: LNET request buffers exhausted under heavy short-term load
2607 Details : If a large number of client requests are generated on a service
2608 that has previously never seen so many requests it is possible
2609 that the request buffer growth cannot keep up with the spike in
2610 demand. Instead of dropping incoming requests, they are held in
2611 the LND until the RPC service can accept more requests.
2614 Frequency : Sometimes during replay
2616 Description: Assertion failure in ll_local_open after replay.
2617 Details : If replay happened on an open request reply before we were able
2618 to set replay handler, reply will become not swabbed tripping the
2619 assertion in ll_local_open. Now we set the handler right after
2620 recognising of open request
2623 Frequency : very rare
2625 Description: kernel reports "badness in vsnprintf"
2626 Details : Reading from the "recovery_status" /proc file in small chunks
2627 may cause a negative length in lprocfs_obd_rd_recovery_status()
2628 call to vsnprintf() (which is otherwise harmless). Exit early
2629 if there is no more space in the output buffer.
2631 Severity : enhancement
2633 Description: clear OBD RPC statistics by writing to them
2634 Details : It is now possible to clear the OBD RPC statistics by writing
2635 to the "stats" file.
2640 Description: Client mtime is not the same on different clients after utimes
2641 Details : In some cases, the client was using the utimes() syscall on
2642 a file cached on another node. The clients now validate the
2643 ctime from the MDS + OSTs to determine which one is right.
2648 Description: Inability to activate failout mode
2649 Details : lconf script incorrectly assumed that in pythong string's numeric
2650 value is used in comparisons.
2653 Frequency : always with multiple stripes per file
2655 Description: Inefficient object allocation for mutli-stripe files
2656 Details : When selecting which OSTs to stripe files over, for files with
2657 a stripe count that divides evenly into the number of OSTs,
2658 the MDS is always picking the same starting OST for each file.
2659 Return the OST selection heuristic to the original design.
2664 Description: mount failures may take full timeout to return an error
2665 Details : Under some heavy load conditions it is possible that a
2666 failed mount can wait for the full obd_timeout interval,
2667 possibly several minutes, before reporting an error.
2668 Instead return an error as soon as the status is known.
2670 Frequency : quota enabled and large files being deleted
2672 Description: releasing more than 4GB of quota at once hangs OST
2673 Details : If a user deletes more than 4GB of files on a single OST it
2674 will cause the OST to spin in an infinite loop. Release
2675 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2680 Description: statfs data retrieved from /proc may be stale or zero
2681 Details : When reading per-device statfs data from /proc, in the
2682 {kbytes,files}_{total,free,avail} files, it may appear
2683 as zero or be out of date.
2686 Frequency : systems with MD RAID1 external journal devices
2688 Description: lconf's call to blkid is confused by RAID1 journal devices
2689 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2690 of returning all block devices that match the journal UUID.
2693 Frequency : always, for aggregate stripe size over 4GB
2695 Description: assertion fails when trying to use 4GB stripe size
2696 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2697 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2700 Frequency : always on ppc64
2702 Description: the first write on an ext3 filesystem with mballoc got stuck
2703 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2704 perform endianness conversion.
2706 ------------------------------------------------------------------------------
2708 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2710 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2711 INTEROPERATE with older versions automatically. Please read the
2712 user documentation before upgrading any part of a live system.
2713 * WARNING: Lustre networking configuration changes are required with
2714 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2717 * Support for kernels:
2718 2.6.9-22.0.2.EL (RHEL 4)
2719 2.6.5-7.244 (SLES 9)
2720 2.6.12.6 vanilla (kernel.org)
2723 Severity : enhancement
2724 Bugzilla : 7981/8208
2725 Description: Introduced Lustre Networking (LNET)
2726 Details : LNET is new networking infrastructure for Lustre, it includes
2727 a reorganized network configuration mode (see the user
2728 documentation for full details) as well as support for routing
2729 between different network fabrics. Lustre Networking Devices
2730 (LNDs) for the supported network fabrics have also been
2731 created for this new infrastructure.
2733 Severity : enhancement
2734 Description: Introduced Access control lists
2735 Details : clients can set ACLs on files and directories in order to have
2736 more fine-grained permissions than the standard Unix UGO+RWX.
2737 The MDS must be started with the "-o acl" mount option.
2739 Severity : enhancement
2740 Description: Introduced filesystem quotas
2741 Details : Administrators may now establish per-user quotas on the
2744 Severity : enhancement
2746 Description: Configuration change for the XT3
2747 The PTLLND is now used to run Lustre over Portals on the XT3
2748 The configure option(s) --with-cray-portals are no longer used.
2749 Rather --with-portals=<path-to-portals-includes> is used to
2750 enable building on the XT3. In addition to enable XT3 specific
2751 features the option --enable-cray-xt3 must be used.
2756 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2757 Details : A race between allocating a new llcd and re-getting the llcd_lock
2758 allowed another thread to grab newly-allocated llcd.
2760 Severity : enhancement
2762 Description: 2.6 OST async journal commit and locking fix to improve performance
2763 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2764 2.6 kernels are now async as they already were in 2.4 kernels so
2765 that they can commit concurrently with the network bulk transfer.
2766 For block-allocated files the filter allocation semaphore is held
2767 to avoid filesystem fragmentation during allocation. BKL lock
2768 removed for 2.6 xattr operations where it is no longer needed.
2773 Description: lconf incorrectly determined whether two IP networks could talk
2774 Details : In some more complicated routing and multiple-network
2775 configurations, lconf will avoid trying to make a network
2776 connection to a disjoint part of the IP space. It was doing the
2777 math incorrectly for one set of cases.
2782 Description: Fix for potential infinite loop processing records in an llog.
2783 Details : If an llog record is corrupted/zeroed, it is possible to loop
2784 forever in llog_process(). Validate the llog record length
2785 and skip the remainder of the block on error.
2788 Frequency : occasional (liblustre only)
2790 Description: liblustre could not open files whose last component is a symlink
2791 Details : sysio_path_walk() would incorrectly pass the open intent to
2792 intermediate path components.
2795 Frequency : rare (liblustre only with non-standard tuning)
2796 Bugzilla : 7201 (7350)
2797 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2798 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2799 after it has been released, only do this for referenced locks
2801 Severity : enhancement
2803 Description: specifying an (invalid) directory default stripe_size of -1
2804 would reset the directory default striping
2805 Details : stripe_size -1 was used internally to signal directory stripe
2806 removal, now use "all default" to signal dir stripe removal
2807 as a directory striping of "all default" is not useful
2810 Frequency : common for large clusters running liblustre clients
2812 Description: doing an ls when liblustre clients are running is slow
2813 Details : sending a glimpse AST to a liblustre client waits for every AST
2814 to time out, as liblustre clients will not respond. Since they
2815 cannot cache data we refresh the OST lock LVB from disk instead.
2817 Severity : enhancement
2819 Description: doing an ls at the same time as file IO can be slow
2820 Details : enqueue and other "small" requests can be blocked behind many
2821 large IO requests. Create a new OST IO portal for non-IO
2822 requests so they can be processed faster.
2825 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2827 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2828 created file with mode 000
2829 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2830 allow an ftruncate() as a non-root user to a file with mode 000.
2831 Lustre now respects this flag to disable mode checks when
2832 truncating a file owned by the user
2835 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2838 Description: Revoking locks from clients that went dead or catatonic might take
2840 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2841 cancellation of such locks instant on servers without waiting for
2842 any reply from clients. Clients drops these locks when cancel
2843 notification from server is received without replying.
2846 Frequency : liblustre-only, when liblustre client dies or becomes busy
2848 Description: Doing ls on Linux clients can take a long time with active
2850 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2851 granting such locks to it in the first place if possible. Locks
2852 are taken by proxy on the OST during the read or write and
2853 dropped immediately afterward. Add connect flags handling, do
2854 not grant locks to liblustre clients for glimpse ASTs.
2856 Severity : enhancement
2858 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2859 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2860 sequential reads and behaves badly with applications doing random
2861 reads. Improve it by reading ahead at least read region, and
2862 avoiding excessive large RPC for small reads.
2864 Severity : enhancement
2866 Description: Creating more than 1000 files for a single job may cause a load
2867 imbalance on the OSTs if there are also a large number of OSTs.
2868 Details : qos_prep_create() uses an OST index reseed value that is an
2869 even multiple of the number of available OSTs so that if the
2870 reseed happens in the middle of the object allocation it will
2871 still utilize the OSTs as uniformly as possible.
2876 Description: OST or MDS may oops in ping_evictor_main()
2877 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2878 but doesn't restart at beginning of obd_exports_timed list
2881 Severity : enhancement
2883 Description: improve by-nid export eviction on the MDS and OST
2884 Details : allow multiple exports with the same NID to be evicted at one
2885 time without re-searching the exports list.
2888 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2890 Description: MDS may oops in groups_free()
2891 Details : in rare race conditions a newly allocated group_info struct is
2892 freed again, and this can be NULL. The 2.4 compatibility code
2893 for groups_free() checked for a NULL pointer, but 2.6 did not.
2896 Frequency : common for liblustre clients doing little filesystem IO
2897 Bugzilla : 9352, 7313
2898 Description: server may evict liblustre clients accessing contended locks
2899 Details : if a client is granted a lock or receives a completion AST
2900 with a blocking AST already set it would not reply to the AST
2901 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2904 Frequency : lfs setstripe, only systems with more than 160 OSTs
2906 Description: unable to set striping with a starting offset beyond OST 160
2907 Details : llapi_create_file() incorrectly limited the starting stripe
2908 index to the maximum single-file stripe count.
2911 Frequency : LDAP users only
2913 Description: lconf did not handle in-kernel recovery with LDAP properly
2914 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2916 Severity : enhancement
2918 Description: bind OST threads to NUMA nodes to improve performance
2919 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2920 node and do their allocations there to localize memory access
2922 Severity : enhancement
2924 Description: llmount can determine client NID directly from Myrinet (GM)
2925 Details : the client NID code from gmnalnid was moved directly into
2926 llmount, removing the need to use this or specifying the
2927 client NID explicitly when mounting GM clients with zeroconf
2930 Frequency : if client is started with down MDS
2932 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2933 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2934 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2935 ptlrpc_import_delayed_req() because we didn't check if the
2936 request was interrupted, and we also didn't break out of the
2937 event loop if there was no timeout
2942 Description: data loss during non-page-aligned writes to a single file from
2943 both multiple nodes and multiple threads on one node at same time
2944 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2945 inconsistency led to false short-reads, that were cached and later
2946 used by ->prepare_write() to fill in partially written page,
2947 leading to data loss.
2950 Frequency : always, if lconf --abort_recovery used
2952 Description: lconf --abort_recovery fails with 'Operation not supported'
2953 Details : lconf was attempting to abort recovery on the MDT device and not
2956 Severity : enhancement
2958 Description: remove cleanup logs
2959 Details : replace lconf-generated cleanup logs with lustre internal
2960 cleanup routines. Eliminates the need for client-cleanup and
2963 Severity : enhancement
2965 Description: add support for EAs (user and system) on lustre filesystems
2966 Details : it is now possible to store extended attributes in the Lustre
2967 client filesystem, and with the user_xattr mount option it
2968 is possible to allow users to store EAs on their files also
2970 Severity : enhancement
2972 Description: Add possibility (config option) to show minimal available OST free
2974 Details : When compiled with --enable-mindf configure option, statfs(2)
2975 (and so, df) will return least minimal free space available from
2976 all OSTs as amount of free space on FS, instead of summary of
2977 free spaces of all OSTs.
2979 Severity : enhancement
2981 Description: do not expand extent locks acquired on OST-side
2982 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2983 by server: they are not cached anyway.
2986 Frequency : when mmap is used/binaries executed from Lustre
2988 Description: Unmmap pages before throwing them away from read cache.
2989 Details : llap_shrink cache now attempts to unmap pages before discarding
2990 them (if unmapping failed - do not discard). SLES9 kernel has
2991 extra checks that trigger if this unmapping is not done first.
2996 Description: lconf didn't resolve symlinks before checking to see whether a
2997 given mountpoint was already in use
3000 Frequency : when migrating failover services
3001 Bugzilla : 6395, 9514
3002 Description: When migrating a subset of services from a node (e.g. failback
3003 from a failover service node) the remaining services would
3004 time out and evict clients.
3005 Details : lconf --force (implied by --failover) sets the global obd_timeout
3006 to 5 seconds in order to quickly disconnect, but this caused
3007 other RPCs to time out too quickly. Do not change the global
3008 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3010 Severity : enhancement
3011 Frequency : if MDS is started with down OST
3012 Bugzilla : 9439,5706
3013 Description: Allow startup/shutdown of an MDS without depending on the
3014 availability of the OSTs.
3015 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3016 Add appropriate locking and lov-osc refcounts for safe
3017 cleaning. Add osc abort_inflight calls in case the
3018 synchronize never started.
3021 Frequency : occasional (Cray XT3 only)
3023 Description: root not authorized to access files in CRAY_PORTALS environment
3024 Details : The client process capabilities were not honoured on the MDS in
3025 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3026 been accessed by an authorized user then root was able to access
3027 the file on the local client also. The root user capabilities
3028 are now allowed on the MDS, as this environment has secure UID.
3031 Frequency : occasional
3033 Description: ldiskfs "too long searching" message happens too often
3034 Details : A debugging message (otherwise harmless) prints too often on
3035 the OST console. This has been reduced to only happen when
3036 there are fragmentation problems on the filesystem.
3041 Description: Division by zero in statfs when all OSCs are inactive
3042 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3043 lov_statfs divides by value returned by lov_get_stripecnt().
3047 Bugzilla : 9489, 3273
3048 Description: First write from each client to each OST was only 4kB in size,
3049 to initialize client writeback cache, which caused sub-optimal
3050 RPCs and poor layout on disk for the first writen file.
3051 Details : Clients now request an initial cache grant at (re)connect time
3052 and so that they can start streaming writes to the cache right
3053 away and always do full-sized RPCs if there is enough data.
3054 If the OST is rebooted the client also re-establishes its grant
3055 so that client cached writes will be honoured under the grant.
3060 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3061 Details : Now I/O RPCs go to different portal number and (presumably) fast
3062 lock requests (and glimses) and other RPCs get their own service
3063 threads pool that should be able to service those RPCs
3066 Severity : enhancement
3068 Description: Ability to exchange lustre version between client and servers and
3069 issue warnings at client side if client is too old. Also for
3070 liblustre clients there is ability to refuse connection of too old
3072 Details : New 'version' field is added to connect data structure that is
3073 filled with version info. That info is later checked by server and
3077 Frequency : rare, liblustre only.
3078 Bugzilla : 9296, 9581
3079 Description: Two simultaneous writes from liblustre at offset within same page
3080 might proceed at the same time overwriting eachother with stale
3082 Details : I/O lock withing llu_file_prwv was released too early, before data
3083 actually was hitting the wire. Extended lock-holding time until
3084 server acknowledges receiving data.
3087 Frequency : extremely rare. Never observed in practice.
3089 Description: avoid generating lustre_handle cookie of 0.
3090 Details : class_handle_hash() generates handle cookies by incrementing
3091 global counter, and can hit 0 occasionaly (this is unlikely, but
3092 not impossible, because initial value of cookie counter is
3093 selected randonly). Value of 0 is used as a sentinel meaning
3094 "unassigned handle" --- avoid it. Also coalesce two critical
3095 sections in this function into one.
3097 Severity : enhancement
3099 Description: allow liblustre clients to delegate truncate locking to OST
3100 Details : To avoid overhead of locking, liblustre client instructs OST to
3101 take extent lock in ost_punch() on client's behalf. New connection
3102 flag is added to handle backward compatibility.
3104 Severity : enhancement
3105 Bugzilla : 4928, 7341, 9758
3106 Description: allow number of OST service threads to be specified
3107 Details : a module parameter allows the number of OST service threads
3108 to be specified via "options ost ost_num_threads={N}" in the
3109 OSS's /etc/modules.conf or /etc/modprobe.conf.
3113 Bugzilla : 6146, 9635, 9895
3114 Description: servers crash with bad pointer in target_handle_connect()
3115 Details : In rare cases when a client is reconnecting it was possible that
3116 the connection request was the last reference for that export.
3117 We would temporarily drop the export reference and get a new
3118 one, but this may have been the last reference and the export
3119 was just destroyed. Get new reference before dropping old one.
3121 Severity : enhancement
3122 Frequency : if client is started with failover MDS
3124 Description: Allow multiple MDS hostnames in the mount command
3125 Details : Try to read the configuration from all specified MDS
3126 hostnames during a client mount in case the "primary"
3129 Severity : enhancement
3131 Description: Stop sending data to evicted clients as soon as possible.
3132 Details : Check if the client we are about to send or are sending data to
3133 was evicted already. (Check is done every second of waiting,
3134 for which l_wait_event interface was extended to allow checking
3135 of exit condition at specified intervals).
3138 Frequency : rare, normally only when NFS exporting is done from client
3140 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3142 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3145 Frequency : always, if srand() is called before liblustre initialization
3147 Description: Liblustre uses system PRNG disturbing its usage by user application
3148 Details : Introduce internal to lustre fast and high-quality PRNG for
3149 lustre usage and make liblustre and some other places in generic
3150 lustre code to use it.
3152 Severity : enhancement
3153 Bugzilla : 9477, 9557, 9870
3154 Description: Verify that the MDS configuration logs are updated when xml is
3155 Details : Check if the .xml configuration logs are newer than the config
3156 logs stored on the MDS and report an error if this is the case.
3157 Request --write-conf, or allow starting with --old_conf.
3159 Severity : enhancement
3161 Description: Handle symlinks in the path when checking if Lustre is mounted.
3162 Details : Resolve intermediate symlinks when checking if a client has
3163 mounted a filesystem to avoid duplicate client mounts.
3168 Description: lconf can hit an error exception but still return success.
3169 Details : The lconf command catches the Command error exception at the top
3170 level script context and will exit with the associated exit
3171 status, but doesn't ensure that this exit status is non-zero.
3176 Description: failure of ptlrpc thread startup can cause oops
3177 Details : Starting a ptlrpc service thread can fail if there are a large
3178 number of threads or the server memory is very fragmented.
3179 Handle this without oopsing.
3182 Frequency : always, only if liblustre and non-default acceptor port was used
3184 Description: liblustre cannot connect to servers with non-default acceptor port
3185 Details : tcpnal_set_default_params() was not called and was therefore
3186 ignoring the environment varaible TCPNAL_PORT, as well as other
3187 TCPNAL_ environment variables
3192 Description: two objects could be created on the same OST for a single file
3193 Details : If an OST is down, in some cases it was possible to create two
3194 objects on a single OST for a single file. No problems other
3195 than potential performance impact and spurious error messages.
3199 Bugzilla : 5681, 9562
3200 Description: Client may oops in ll_unhash_aliases
3201 Details : Client dcache may become inconsistent in race condition.
3202 In some cases "getcwd" can fail if the current directory is
3208 Description: Inode refcounting problems in NFS export code
3209 Details : link_raw functions used to call d_instantiate without obtaining
3210 extra inode reference first.
3214 Bugzilla : 9942, 9903
3215 Description: Referencing freed requests leading to crash, memleaks with NFS.
3216 Details : We used to require that call to ll_revalidate_it was always
3217 followed by ll_lookup_it. Also with revalidate_special() it is
3218 possible to call ll_revalidate_it() twice for the same dentry
3219 even if first occurence returned success. This fix changes semantic
3220 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3221 reference on a request referred from the intent.
3222 ll_intent_release() then releases such a request.
3225 Frequency : rare, normally benchmark loads only
3227 Description: unlinked inodes were kept in memory on the client
3228 Details : If a client is repeatedly creating and unlinking files it
3229 can accumulate a lot of stale inodes in the inode slab cache.
3230 If there is no other client load running this can cause the
3231 client node to run out of memory. Instead flush old inodes
3232 from client cache that have the same inode number as a new inode.
3235 Frequency : SLES9 2.6.5 kernel and long filenames only
3236 Bugzilla : 9969, 10379
3237 Description: utime reports stale NFS file handle
3238 Details : SLES9 uses out-of-dentry names in some cases, which confused
3239 the lustre dentry revalidation. Change it to always use the
3243 Frequency : rare, unless heavy write-truncate concurrency is continuous
3244 Bugzilla : 4180, 6984, 7171, 9963, 9331
3245 Description: OST becomes very slow and/or deadlocked during object unlink
3246 Details : filter_destroy() was holding onto the parent directory lock
3247 while truncating+unlinking objects. For very large objects this
3248 may block other threads for a long time and slow overall OST
3249 responsiveness. It may also be possible to get a lock ordering
3250 deadlock in this case, or run out of journal credits because of
3251 the combined truncate+unlink. Solution is to do object truncate
3252 first in one transaction without parent lock, and then do the
3253 final unlink in a new transaction with the parent lock. This
3254 reduces the lock hold time dramatically.
3257 Frequency : rare, 2.4 kernels only
3259 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3260 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3261 before unmounting their filesystem in order to avoid deadlock.
3262 The kernel_locked() function in 2.4 kernels only checks whether
3263 the kernel lock is held, not whether it is this process that is
3264 holding it as 2.6 kernels do.
3269 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3270 Details : The client ptlrpc code may be trying to reconnect to a down
3271 server before a previous connection attempt has timed out.
3272 Increase the reconnect interval to be longer than the connection
3273 timeout interval to avoid sending duplicate connections to
3277 Frequency : echo_client brw_test command
3279 Description: fix echo_client to work with OST preallocated code
3280 Details : OST preallocation code (5137) didn't take echo_client IO path
3281 into account: echo_client calls filter methods outside of any
3282 OST thread and, hence, there is no per-thread preallocated
3283 pages and buffers to use. Solution: hijack pga pages for IO. As
3284 a byproduct, this avoids unnecessary data copying.
3288 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3289 Description: Client can oops in mdc_commit_close() after open replay
3290 Details : It was possible for the MDS to return an open request with no
3291 transaction number in mds_finish_transno() if the client was
3292 evicted, but without actually returning an error. Clients
3293 would later try to replay that open and may trip an assertion
3294 Simplify the client close codepath, and always return an error
3295 from the MDS in case the open is not successful.
3298 Frequency : rare, 2.6 OSTs only
3300 Description: OST may deadlock under high load on fragmented files
3301 Details : If there was a heavy load and highly-fragmented OST filesystems
3302 it was possible to have all the OST threads deadlock waiting on
3303 allocation of biovecs, because the biovecs were not released
3304 until the entire RPC IO was completed. Instead, release biovecs
3305 as soon as they are complete to ensure forward IO progress.
3307 Severity : enhancement
3309 Description: Support for specifying external journal device at mount
3310 Details : If an OST or MDS device is formatted with an external journal
3311 device, this device major/minor is stored in the ext3 superblock
3312 and may not be valid for failover. Allow detecting and
3313 specifying the external journal at mount time.
3318 Description: Mounting an MDS with pending unlinked files may cause oops
3319 Details : target_finish_recovery() calls mds_postrecov() which returned
3320 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3321 considers this an error and immediately begins cleaning up the
3322 lov, just after starting the mds_lov process
3324 Severity : enhancement
3326 Description: Implement 'lfs df' to report actual free space on per-OST basis
3327 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3328 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3329 usage of objects; '-h' to report in human readable format.
3331 ------------------------------------------------------------------------------
3333 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3340 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3341 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3342 at startup during orphan unlinking. Instead, skip buddy update
3343 before setup, as it will be regenerated after recovery is complete.
3348 Description: If an OST is inactive, its locks might reference stale inodes.
3349 Details : lov_change_cbdata() must iterate over all namespaces, even if
3350 they are inactive to clear inode references from the lock.
3352 Severity : enhancement
3353 Frequency : occasional, if non-standard max_dirty_mb used
3355 Description: Client will block write RPCs if not enough grant
3356 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3357 then the client will block writes while waiting for another RPC
3358 to complete instead of consuming its dirty limit. With change
3359 we get improved performance when max_dirty_mb is small.
3361 Severity : enhancement
3362 Bugzilla : 3389, 6253
3363 Description: Add support for supplementary groups on the MDS.
3364 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3365 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3366 lookups for user supplementary groups into a cache.
3370 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3371 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3372 for liblustre O_EXCL is enforced by MDS.
3376 Description: readdir never returns NULL in liblustre.
3377 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3378 dentry in struct dirent.
3382 Frequency : liblustre only, depends on application IO pattern
3383 Description: liblustre clients evicted if not contacting servers
3384 Details : Don't put liblustre clients into the ping_evictor list, so
3385 they will not be evicted by the pinger ever.
3387 Severity : enhancement
3389 Description: Add ability to evict clients by NID from MDS.
3390 Details : By echoing "nid:$NID" string into
3391 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3392 $NID would be instantly evicted from this MDS and from all active
3393 OSTs connected to it.
3397 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3398 Details : lookup_it_finish() used to query file size from OSTs that was not
3403 Description: service threads change working directory to that of init
3404 Details : Starting lustre service threads may pin the working directory
3405 of the parent thread, making that filesystem busy. Threads
3406 now change to the working directory of init to avoid this.
3410 Frequency : during shutdown only
3411 Description: shutdown with a failed MDS or OST can cause unmount to hang
3412 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3417 Frequency : 2.6 only
3418 Description: chmod/chown may include an extra supplementary group
3419 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3420 supplementary group and if none is specified this is used.
3423 Bugzilla : 5479 (6816)
3425 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3426 Details : If lookup is in progress on a file that is unlinked we might try
3427 to revalidate the inode and fail in revalidate after lookup is
3428 complete and ll_file_open() enqueues the open again but
3429 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3432 Frequency : always, if lconf --abort_recovery used
3434 Description: lconf --abort_recovery fails with 'Operation not supported'
3435 Details : lconf was attempting to abort recovery on the MDT device and not
3438 ------------------------------------------------------------------------------
3440 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3445 Frequency : rare (only unsupported configurations with a node running as an
3447 Bugzilla : 6514, 5137
3448 Description: Mounting a Lustre file system on a node running as an OST could
3450 Details : OSTs now preallocates memory needed to write out data at
3451 startup, instead of when needed, to avoid having to
3452 allocate memory in possibly low memory situations.
3453 Specifically, if the file system is mounted on on OST,
3454 memory pressure could force it to try to write out data,
3455 which it needed to allocate memory to do. Due to the low
3456 memory, it would be unable to do so and the node would
3457 become unresponsive.
3459 Severity : enhancement
3461 Description: Addition of lconf --service command line option
3462 Details : lconf now accepts a '--service <arg>' option, which is
3463 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3465 Severity : enhancement
3467 Description: Failover mode is now the default for OSTs.
3468 Details : By default, OSTs will now run in failover mode. To return to
3469 the old behaviour, add '--failout' to the lmc line for OSTs.
3471 Severity : enhancement
3473 Description: Health checks are now provided for MDS and OSTs
3474 Details : Additional detailed health check information on MSD and OSTs
3475 is now provided through the procfs health_check value.
3478 Frequency : occasional, depends on IO load
3480 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3481 after numerous create/delete cycles
3482 Details : The ext3 inode allocation policy would not allocate new inodes
3483 very well on the OSTs because there are no new directories
3484 being created. Instead we look for groups with free space if
3485 the parent directories are nearly full.
3490 Description: Network or server problems during mount may cause partially
3491 mounted clients instead of returning an error.
3492 Details : The config llog parsing code may overwrite the error return
3493 code during mount error handling, returning success instead
3499 Description: MDS can fail to allocate large reply buffers
3500 Details : After long uptimes the MDS can fail to allocate large reply
3501 buffers (e.g. zconf client mount config records) due to memory
3502 fragmentation or consumption by the buffer cache. Preallocate
3503 some large reply buffers so that these replies can be sent even
3504 under memory pressure.
3508 Frequency : rare (liblustre)
3509 Description: fsx running with liblustre complained that using truncate() to
3510 extend the file doesn't work. This patch corrects that issue.
3511 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3512 ATTR_SIZE and lsm use in llu_setattr_raw.
3516 Frequency : rare, only 2.6 kernels
3517 Description: Unusual file access patterns on the MDS may result in inode
3518 data being lost in very rare circumstances.
3519 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3520 code in the 2.6 kernel caused the inode and/or EA data not to
3521 be read from disk, causing single-file corruption.
3525 Frequency : rare, only 2.6 filesystems using extents
3526 Description: Heavy concurrent write and delete load may cause data corruption.
3527 Details : It was possible under high-load situations to have an extent
3528 metadata block in the block device cache from a just-unlinked
3529 file overwrite a newly-allocated data block. We now unmap any
3530 metadata buffers that alias just-allocated data blocks.
3534 Frequency : filesystems with default stripe_count larger than 77
3535 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3536 Details : lconf specifies an inode size of 4096 bytes when the default
3537 stripe_count is larger than 77. This conflicts with the default
3538 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3539 this case to avoid pinning too much memory for large EAs.
3541 ------------------------------------------------------------------------------
3543 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3548 Frequency : rare (extremely heavy IO load with hundreds of clients)
3550 Description: Client is evicted, gets IO error writing to file
3551 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3552 caused clients to be evicted for AST timeouts. The fixes in
3553 bug 5192 mean we no longer need to have such short AST timeouts
3554 so ldlm_timeout has been increased.
3557 Frequency : occasional during --force or --failover shutdown under load
3558 Bugzilla : 5949, 4834
3559 Description: Server oops/LBUG if stopped with --force or --failover under load
3560 Details : a collection of import/export refcount and cleanup ordering
3561 issues fixed for safer force cleanup
3564 Frequency : only filesystems larger than 120 OSTs
3565 Bugzilla : 5990, 6223
3566 Description: lfs getstripe would oops on a very large filesystem
3567 Details : lov_getconfig used kfree on vmalloc'd memory
3570 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3571 Bugzilla : 6242, 6243
3572 Description: reading from files that had been truncated to a non-zero size
3573 but never opened returned no data
3574 Details : ll_file_read() reads zeros from no-object files to EOF
3579 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3580 Details : There's little harm in aborting MDS/OSS recovery and letting it
3581 try again, so I removed the LASSERT and return an error instead.
3583 Severity : enhancement
3585 Description: New debugging infrastructure for tracking down data corruption
3586 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3587 (b) cover more of the client-side code path, and (c) try to narrow
3588 down where problems occurred
3592 Bugzilla : 3819, 4364, 4397, 6313
3593 Description: Racing close and eviction MDS could cause assertion in mds_close
3594 Details : It was possible to get multiple mfd references during close and
3595 client eviction, leading to one thread referencing a freed mfd.
3597 Severity: : enhancement
3598 Bugzilla : 3262, 6359
3599 Description: Attempts to reconnect to servers are now more aggressive.
3600 Details : This builds on the enhanced upcall-less recovery that was added
3601 in 1.4.2. When trying to reconnect to servers, clients will
3602 now try each server in the failover group every 10 seconds. By
3603 default, clients would previously try one server every 25 seconds.
3608 Description: After recovery, certain operations trigger a failed
3609 assertion on a client.
3610 Details : Failing over an mds, using lconf -d --failover, while a
3611 client was doing a readdir() call would cause the client to
3612 LBUG after recovery completed and the readdir() was resent.
3614 Severity : enhancement
3616 Description: Default groups are now added by lconf
3617 Details : You can now run lconf --group <servicename> without having to
3618 manually add groups with lmc.
3621 Frequency : occasional
3623 Description: Nodes with an elan id of 0 trigger a failed assertion
3626 Frequency : always when accessing e.g. tty/console device nodes
3628 Description: tty and some other devices nodes cannot be used on lustre
3629 Details : file's private_data field is used by device data and lustre
3630 values in there got lost. New field was added to struct file to
3631 store fs-specific private data.
3634 Frequency : when exporting Lustre via NFS
3636 Description: NFSD failed occasionally when looking up a path component
3637 Details : NFSD is looking up ".." which was broken in ext3 directories
3638 that had grown large enough to become hashed.
3641 Frequency : Clusters with multiple interfaces not on the same subnet
3643 Description: Nodes will repeatedly try to reconnect to an interface which it
3644 cannot reach and report an error to the log.
3645 Details : Extra peer list entries will be created by lconf with some peers
3646 unreachable. lconf now validates the peer before adding it.
3649 Frequency : Only if a default stripe is set on the filesystem root.
3651 Description: Setting a default stripe on the filesystem root prevented the
3652 filesystem from being remounted.
3653 Details : The client was sending extra request flags in the root getattr
3654 request and did not allocate a reply buffer for the dir EA.
3657 Frequency : occasional, higher if lots of files are accessed by one client
3658 Bugzilla : 6159, 6097
3659 Description: Client trips assertion regarding lsm mismatch/magic
3660 Details : While revalidating inodes the VFS looks up inodes with ifind()
3661 and in rare cases can find an inode that is being freed.
3662 The ll_test_inode() code will free the lsm during ifind()
3663 when it finds an existing inode and then the VFS later attaches
3664 this free lsm to a new inode.
3668 Bugzilla : 6422, 7030
3669 Description: MDS deadlock between mkdir and client eviction
3670 Details : Creating a new file via mkdir or mknod (starting a transaction
3671 and getting the ns lock) can deadlock with client eviction
3672 (gets ns lock and trying to finish a synchronous transaction).
3675 Frequency : occasional
3676 Description: While starting a server, the fsfilt_ext3 module could not be
3678 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3679 kernels. Previously, lconf would still use the ext3 name
3680 when trying to load modules. Now, it will correctly use
3681 ext3 on 2.4 and ldiskfs on 2.6.
3683 Severity : enhancement
3684 Description: The default stripe count has been changed to 1
3685 Details : The interpretation of the default stripe count (0, to lfs
3686 or lmc) has been changed to mean striping across a single
3687 OST, rather than all available. For general usage we have
3688 found a stripe count of 1 or 2 works best.
3690 Severity : enhancement
3691 Description: Add support for compiling against Cray portals.
3692 Details : Conditional compiling for some areas that are different
3696 Frequency : occasional
3697 Bugzilla : 6409, 6834
3698 Description: Creating files with an explicit stripe count may lead to
3699 a failed assertion on the MDS
3700 Details : If some OSTs are full or unavailable, creating files may
3701 trigger a failed assertion on the MDS. Now, Lustre will
3702 try to use other servers or return an error to the
3706 Frequency : occasional
3708 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3709 caused lock timeout during readahead (since 1.4.2).
3710 Details : Processes doing readahead might match a lock that hasn't been
3711 granted yet if there are overlapping and conflicting lock
3712 requests. The readahead process waits on ungranted lock
3713 (original lock is CBPENDING), while OST waits for that process
3714 to cancel CBPENDING read lock and eventually evicts client.
3716 Severity : enhancement
3718 Description: Initial enabling of flock support for clients
3719 Details : Implements fcntl advisory locking and file status functions.
3720 This feature is provided as an optional mount flag (default
3721 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3722 locking are implemented yet, and those that are are not guaranteed
3723 to be completely correct in production environments.
3724 mount -t lustre -o [flock|noflock] ...
3727 Frequency : occasional
3729 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3730 assertion in the ext3 JBD (journaling) layer.
3731 Details : The b_committed_data struct is protected by the big kernel lock
3732 in 2.4 kernels, serializing journal_commit_transaction() and
3733 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3734 access to this struct. In 2.6 kernels there is finer grained
3735 locking to improve SMP performance of the JBD layer.
3739 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3744 Description: merge and cleanup kernel patches.
3747 -----------------------------------------------------------------------------
3749 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3751 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3752 versions of Lustre. Please update all servers and clients to
3753 version 1.4.2 or later at the same time. You must also run
3754 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3755 to update the configuration logs.
3757 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3758 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3759 - swab llog records of type '0' so we get proper header size/idx (5861)
3760 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3761 - fix rename of one directory over another leaking an inode (5953)
3762 - avoid SetPageDirty on 2.6 (5981)
3763 - don't re-add just-being-destroyed locks to the waiting list (5653)
3764 - when creating new directories, inherit the parent's custom
3765 striping settings if present parent (3048)
3766 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3767 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3768 - don't hold client locks on temporary worklist from l_lru (5666)
3769 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3770 - automatically evict dead clients (5921)
3771 - Update file size properly in create+truncate+fstat case (6196)
3772 - Do not unhash mountpoint dentries, do not allow removal of
3774 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3775 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3776 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3777 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3778 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3779 - Correctly handle reads of files with no objects (6243)
3780 - lctl recover will also mark a device active if deactivate used (5933)
3782 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3783 - allow --write-conf on an MDS with different nettype than client (5619)
3784 - don't write config llogs to MDS for mounts not from that MDS (5617)
3785 - lconf should create multiple TCP connections from a client (5201)
3786 - init scripts are now turned off by default; run chkconfig --on
3787 lustre and chkconfig --on lustrefs to use them
3788 - upcalls are no longer needed for clients to recover to failover
3790 - add --abort-recovery option to lconf to abort recovery on device
3792 - add support for an arbitrary number of OSTs (3026)
3793 - Quota support protocol changes.
3794 - forward compatibility changes to wire structs (6007)
3795 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3796 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3797 - improved "lustre status" script
3798 - initialize blocksize for non-regular files (6062)
3799 - added --disable-server and --disable-client configure options (5782)
3800 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3801 - Vanilla 2.4.29 support
3802 - increase maximum number of obd devices to 520 (6242)
3803 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3804 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3805 SLES 9 2.6.5 kernels
3806 - limit stripes per file to 160 (the maximum EA size) (6093)
3808 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3811 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3812 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3813 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3814 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3815 - small ext3 extents cleanups and fixes (5733)
3816 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3817 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3818 - avoid needless client->OST connect, fix handle mismatch (5317)
3819 - fix DLM error path that led to out-of-sync client, long delays (5779)
3820 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3821 - fix several locking issues related to i_size (5492,5624,5654,5672)
3822 - don't move pending lock onto export if it is already evicted (5683)
3823 - fix kernel oops when creating .foo in unlinked directory (5548)
3824 - fix deadlock in obdfilter statistics vs. object create (5811)
3825 - use time_{before,after} to avoid timer jiffies wrap (5882)
3826 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3827 - Do not leak request if server was not able to process it (5154)
3828 - If mds_open unable to find parent dir, make that negative lookup(5154)
3829 - don't create new directories with extent-mapping (5909, 5936)
3831 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3832 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3833 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3834 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3835 - support for RHEL4 kernel on i686 (5773)
3836 - provide error messages when incompatible logs are encountered (5898)
3838 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3839 * version 1.4.0.10 (1.4.1 release candidate 1)
3841 - don't keep a lock reference when lock is not granted (4238)
3842 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3843 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3844 - drop import inflight refcount on signal_completed_replay error (5255)
3845 - unlock page after async write error during send (3677)
3846 - handle missing objects in filter_preprw_read properly (5265)
3847 - no transno return for symlink open, don't save no-trasno open (3440)
3848 - don't try to complete elan receive that already failed (4012)
3849 - free RPC server reply state on error (5406)
3850 - clean up thread from ptlrpc_start_thread() on error (5160)
3851 - readahead could read extra page into cache that wasn't ejected (5388)
3852 - prevent races in class_attach/setup/cleanup/detach (5260)
3853 - don't dereference de->d_inode after l_dput of de (5458)
3854 - use "int" for stripe value returned from lock_to_stripe (5544)
3855 - mballoc allocation and error-checking fixes in 2.6 (5504)
3856 - block device patches to fix I/O request sizes in 2.6 (5482)
3857 - look up hostnames for IB nals (5602)
3858 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3859 - don't start multiple acceptors for the same port (5277)
3860 - fix incorrect LASSERT in mds_getattr_name (5635)
3861 - export a proc file for general "ping" checking (5628)
3862 - fix "lfs check" to not block when the MDS is down (5628)
3864 - service request history (4965)
3865 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3866 - create an "evict_client" /proc entry on OSTs, like the MDS has
3867 - fix mount usage message, return errors per mount(8) (5168)
3868 - change grep [] to grep "[]" in tests so they work in more UMLs
3869 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3870 - remove ancient LOV_MAGIC_V0 EA support (5047)
3871 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3872 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3873 - allow readahead allocations to fail when low on memory (5383)
3874 - mmap locking landed again, after considerable improvement (2828)
3875 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3877 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3880 - send OST transaction number in read/write reply to free req (4966)
3881 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3882 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3883 - the watchdog thread now runs as interruptible (5246)
3884 - flock/lockf fixes (but it's still disabled, pending 5135)
3885 - don't use EXT3 constants in llite code (5094)
3886 - memory shortage at startup could cause assertion (5176)
3888 - reorganization of lov code
3889 - single portals codebase
3891 - add extents/mballoc support (5025)
3892 - direct I/O reads in the obdfilter (4048)
3893 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3895 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3898 - send OST transaction number in read/write reply to free req (4966)
3899 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3900 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3901 - the watchdog thread now runs as interruptible (5246)
3902 - handle missing objects in filter_preprw_read properly (5265)
3903 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3904 - drop import inflight refcount on signal_completed_replay error (5255)
3905 - unlock page after async write error during send (3677)
3906 - return original error code on reconstructed replies (3761)
3907 - no transno return for symlink open, don't save no-trasno open (3440)
3909 - add pid to ldlm debugging output (4922)
3910 - bump the watchdog timeouts -- we can't handle 30sec yet
3911 - extra debugging for orphan dentry/inode bug (5259)
3913 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3916 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3917 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3918 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3919 - fix NULL dereference in /proc/sys/portals/routes (4827)
3920 - allow failed mdc_close() operations to be interrupted (4561)
3921 - stop precreate on OST before MDS would time out on it (4778)
3922 - don't send partial-page writes before EOF from client (4410)
3923 - discard client grant for sub-page writes on large-page clients (4520)
3924 - don't free dentries not owned by NFS code, check generation (4806)
3925 - fix lsm leak if mds_create_objects() fails (4801)
3926 - limit debug_daemon file size, always print CERROR messages (4789)
3927 - use transno after validating reply (3892)
3928 - process timed out requests if import state changes (3754)
3929 - update mtime on OST during writes, return in glimpse (4829)
3930 - add mkfsoptions to LDAP (4679)
3931 - use ->max_readahead method instead of zapping global ra (5039)
3932 - don't interrupt __l_wait_event() during strace
3934 - add software watchdogs to catch hung threads quickly (4941)
3935 - make lustrefs init script start after nfs is mounted
3936 - fix CWARN/ERROR duplication (4930)
3937 - return async write errors to application if possible (2248)
3938 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3939 - print NAL number in %x format (4645)
3940 - update barely-supported suse-2.4.21-171 series (4842)
3941 - support for sles 9 %post scripts
3942 - support for building 2.6 kernel-source packages
3943 - support for sles km_* packages
3945 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3948 - ignore -ENOENT errors in osc_destroy (3639)
3949 - notify osc create thread that OSC is being cleaned up (4600)
3950 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3951 - reconstruct ost_handle() like mds_handle() (4657)
3952 - create a new thread to do import eviction to avoid deadlock (3969)
3953 - let lconf resolve symlinked-to devices (4629)
3954 - don't unlink "objects" from directory with default EA (4554)
3955 - hold socknal file ref over connect in case target is down (4394)
3956 - allow more than 32000 subdirectories in a single directory (3244)
3957 - fix blocks count for O_DIRECT writes (3751)
3958 - OST returns ENOSPC from object create when no space left (4539)
3959 - don't send truncate RPC if file size isn't changing (4410)
3960 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3961 - bind to privileged port in socknal and tcpnal (3689)
3963 - rate limit CERROR/CWARN console message to avoid overload (4519)
3964 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3965 - basic mmap support (3918)
3966 - kernel patch series update from b1_4 (4711)
3968 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3971 - avoid crash during MDS cleanup with OST shut down (2775)
3972 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3973 - don't use bad inodes on the MDS (3744)
3974 - dynamic object preallocation to improve recovery speed (4236)
3975 - don't hold spinlock over lock dumping or change debug flags (4401)
3976 - don't zero obd_dev when it is force cleaned (3651)
3977 - print grants to console if they go negative (4431)
3978 - "lctl deactivate" will stop automatic recovery attempts (3406)
3979 - look for existing locks in ldlm_handle_enqueue() (3764)
3980 - don't resolve lock handle twice in recovery avoiding race (4401)
3981 - revalidate should check working dir is a directory (4134)
3983 - don't always mark "slow" obdfilter messages as errors (4418)
3985 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3988 - don't close LustreDB during write_conf until it is done (3860)
3989 - fix typo in lconf for_each_profile (3821)
3990 - allow dumping logs from multiple threads at one time (3820)
3991 - don't allow multiple threads in OSC recovery (3812)
3992 - fix debug_size parameters (3864)
3993 - fix mds_postrecov to initialize import for llog ctxt (3121)
3994 - replace config semaphore with spinlock (3306)
3995 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3996 - don't allow enqueue to complete on a destroyed export (3822)
3997 - down write_lock before checking llog header bitmap (3825)
3998 - recover from lock replay timeout (3764)
3999 - up llog sem before sending rpc (3652)
4000 - reduce ns lock hold times when setting kms (3267)
4001 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4002 - fix NULL deref and obd_dev leak on setup error (3312)
4003 - replace some LBUG about llog ops with error handling (3841)
4004 - don't match INVALID dentries from d_lookup and spin (3784)
4005 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4006 - fix invalid assertion in ptlrpc_set_wait (3880)
4008 - add libwrap support for the TCP acceptor (3996)
4009 - add /proc/sys/portals/routes for non-root route listing (3994)
4010 - allow setting MDS UUID in .xml (2580)
4011 - print the stack of a process that LBUGs (4228)
4013 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4016 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4017 - make sure to unset replay flag from failed open requests (3440)
4018 - if default stripe count is 0, use OST count for inode size (3636)
4019 - update parent mtime/ctime on client for create/unlink (2611)
4020 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4021 - free recovery state on server during a forced cleanup (3571)
4022 - unregister_reply for resent reqs (3063)
4023 - loop back devices mounting and status check on 2.6 (3563)
4024 - fix resource-creation race that can provoke i_size == 0 (3513)
4025 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4026 - more debugging for page-accounting assertion (3746)
4027 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4028 - avoid deadlock after precreation failure (3758)
4029 - fix race and lock order deadlock in orphan handling (3450, 3750)
4030 - add validity checks when grabbing inodes from l_ast_data (3599)
4032 - add /proc/.../recovery_status to obdfilter (3428)
4033 - lightweight CDEBUG infrastructure, debug daemon (3668)
4034 - change default OSC RPC parameters to be better on small clusters
4035 - turn off OST read cache for files smaller than 32MB
4036 - install man pages and include them in rpms (3100)
4037 - add new init script for (un)mounting lustre filesystems (2593)
4038 - run chkconfig in %post for init scripts (3701)
4039 - drop scimac NAL (unmaintained)
4041 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4044 - clean kiobufs before and after use (3485)
4045 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4046 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4047 - change init script to fail more clearly if not run as root (1528)
4048 - allow clients to reconnect during replay (1742)
4049 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4050 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4051 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4052 - fix cond_resched (3554)
4053 - only evict unfinished clients after recovery (3515)
4054 - allow bulk resend, prevent data loss (3570)
4055 - dynamic ptlrpc request buffer allocation (2102)
4056 - don't allow unlinking open directory if it isn't empty (2904)
4057 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4058 - remove extraneous obd dereference causing LASSERT failure (3334)
4059 - don't use get_cycles() when creating temp. files on the mds (3156)
4060 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4061 - handle EEXIST for set-stripe, set proper directory name (3336)
4063 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4064 - fix ksocknal_fmb_callback() error messages (2918)
4066 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4069 - don't copy lvb into (possibly NULL) reply on error (2983)
4070 - don't deref dentry after dput, don't free lvb on error (2922)
4071 - use the kms to determine writeback rpc length (2947)
4072 - increment oti_logcookies when osc is inactive (2948)
4073 - update client's i_blocks count via lvb messages (2543)
4074 - handle intent open/close of special files properly (1557)
4075 - mount MDS with errors=remount-ro, like obdfilter (2009)
4076 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4077 - don't use cancelling-locks' kms values (2947)
4078 - use highest lock extent for kms, not last one (2925)
4079 - don't dereference ERR_PTR() dentry in error handling path (3107)
4080 - fix thread race in portals_debug_dumplog() (3122)
4081 - create lprocfs device entries at setup instead of at attach (1519)
4082 - common AST error handler, don't evict client on completion race (3145)
4083 - zero nameidata in detach_mnt in 2.6 (3118)
4084 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4085 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4086 - initialize RPC timeout timer earlier for 2.6 (3219)
4087 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4088 - print nal/nid for unknown nid (3258)
4089 - additional checks for oscc recovery before doing precreate (3284)
4090 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4091 - don't crash in mdc_close for bad permissions on open (3285)
4092 - zero i_rdev for non-device files (3147)
4093 - clear page->private before handing to FS, better assertion (3119)
4094 - tune the read pipeline (3236)
4095 - fix incorrect decref of invalidated dentry (2350)
4096 - provide read-ahead stats and refine rpc in flight stats (3328)
4097 - don't hold journal transaction open across create RPC (3313)
4098 - update atime on MDS at close time (3265)
4099 - close LDAP connection when recovering to avoid server load (3315)
4100 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4101 - don't leak open file on MDS after open resend (3325)
4102 - serialize filter_precreate and filter_destroy_precreated (3329)
4103 - loop device shouldn't call sync_dev() for nul device (3092)
4104 - clear page cache after eviction (2766)
4105 - resynchronize MDS->OST in background (2824)
4106 - refuse to mount the same filesystem twice on same mountpoint (3394)
4107 - allow llmount to create routes for mounting behind routers (3320)
4108 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4109 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4110 - fix rare problem with rename on htree directories (3417)
4112 - allow default OST striping configuration per directory (1414)
4113 - fix compilation for qswnal for 2.6 kernels (3125)
4114 - increase maximum number of MDS request buffers for large systems
4115 - change liblustreapi to be useful for external progs like lfsck (3098)
4116 - increase local configuration timeout for slow disks (3353)
4117 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4119 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4122 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4123 - don't overwrite extent policy data in reply if lock was blocked (2901)
4124 - drop filter export grants atomically with removal from device (2663)
4125 - del obd_self_export from work_list in class_disconnect_exports (2908)
4126 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4127 - swab reply message in mdc_close, other PPC fixes (2464)
4128 - fix destroying of named logs (2325)
4129 - overwrite old logs when running lconf --write_conf (2264)
4130 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4131 - fix race in target_handle_connect (2898)
4132 - mds_reint_create() should take same inode create lock (2926)
4133 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4134 - don't close files for self_export to avoid uninitialized obd (2936)
4135 - allow MDS with the same name as client node (2939)
4136 - hold dentry reference for closed log files for unlink (2325)
4137 - reserve space for all logs during transactions (2059)
4138 - don't evict page beyond end of stripe extent (2925)
4139 - don't oops on a deleted current working directory (2399)
4140 - handle hard links to targets without a parent properly (2517)
4141 - don't dereference NULL lock when racing during eviction (2867)
4142 - don't grow lock extents when lots of conflicting locks (2919)
4144 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4147 - account for cache space usage on clients to avoid data loss (974)
4148 - lfsck support in lustre kernel code (2349)
4149 - reduce journal credits needed for BRW writes (2370)
4150 - orphan handling to avoid losing space on client/server crashes
4151 - ptlrpcd can be blocked, stopping ALL progress (2477)
4152 - use lock value blocks to assist in proper KMS, faster stat (1021)
4153 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4154 - recovery for initial connections (2355)
4155 - fixes for mds_cleanup_orphans (1934)
4156 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4157 - block all file creations until orphan recovery completes (1901)
4158 - client remove rq_connection from request struct (2423)
4159 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4160 - recovery timer race (2670)
4161 - mdc_close recovey bug (2532)
4162 - ptlrpc cleanup bug (2710)
4163 - mds timeout on local locks (2588)
4164 - namespace lock held during RPCs (2431)
4165 - handle interrupted sync write properly (2503)
4166 - don't try to handle a message that hasn't been replied to (2699)
4167 - client assert failure during cleanup after abort recovery (2701)
4168 - leak mdc device after failed mount (2712)
4169 - ptlrpc_check_set allows timedout requests to complete (2714)
4170 - wait for inflight reqs when ptlrpcd finishes (2710)
4171 - make sure unregistered services are removed from the srv_list
4172 - reset bulk XID's when resending them (caught by 1138 test)
4173 - unregister_bulk after timeout
4174 - fix lconf error (2694)
4175 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4176 - readahead locks pages, leaves pending causing memory pressure (2673)
4177 - increase OST request buffers to 4096 on large machines (2729)
4178 - fix up permission of existing directories in simple_mkdir (2661)
4179 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4180 - don't assign transno to errored transactions (2742)
4181 - don't delete objects on OST if given a bogus objid from MDS (2751)
4182 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4183 - if rq_no_resend, then timeout request after recovery (2432)
4184 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4185 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4186 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4187 - MDS recovery shouldn't delete objects that it has given out (2730)
4188 - if enqueue arrives after completion, don't clobber LVB (2819)
4189 - don't unlock pages twice when trigger_group_io returns error (2814)
4190 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4191 - don't write pages to disk if there was an error (1450)
4192 - don't ping imports that have recovery disabled (2676)
4193 - take buffered bytes into account when balancing socknal conn (2817)
4194 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4195 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4196 - remove little-endian swabbing of llog records (1987)
4197 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4198 - retry reposting request buffers if they fail (1191)
4199 - grow extent at grant time to avoid granting a revoked lock (2809)
4200 - lock revoke doesn't evict page if covered by a second lock (2765)
4201 - disable VM readahead to avoid reading outside lock extents (2805)
4203 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4204 - updated kernel patches for hp-2.4.20 kernel (2681)
4206 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4209 - fix truncated write corruption (2366)
4210 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4212 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4213 - ensure len(uuid) < 37 in lmc (1171)
4214 - fix ia64 OOPS in llog_test (2255)
4215 - zero end of page at obdfilter for partial page writes (2648)
4216 - don't leave stale dentries around after renames (bug 2428)
4217 - fix timeouts when evicting a client with a single lock held (2642)
4218 - set deadline for the initial HELLO message to drain (2634)
4219 - print out dotted-quad IP addresses in the socknal (2302)
4221 - additional debugging for MDS client eviction problem (2443)
4222 - fix mkfsoptions support for osts (2603, 2604)
4224 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4227 - add series for the vanilla 2.6.0 kernel
4228 - add series for the vanilla 2.4.24 kernel
4229 - add series for a cray x86/64 UL kernel drop
4230 - fix xattr patches for the vanilla 2.4.19 series
4232 - generate true UUIDs in lmc (1171)
4233 - have portals stack dumping break in UML (2466)
4234 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4235 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4236 - wake callback waiting threads on client eviction (2460)
4237 - Add --ptldebug and --subsystem to lmc (1719)
4238 - update assertion to allow safe interrupt allocation
4239 - set rq_no_resend for cancel requests (2432)
4240 - recalculate ptlrpcd timeout after resend (2494)
4241 - call vfs_rmdir when removing pending directories (2368)
4242 - fix renaming a file to itself (2429)
4243 - lmc creates a default one-stripe lov (2454)
4244 - expand procfs space to handle large clusters (2326)
4245 - increase UML stack to avoid overflow
4246 - update lconf's list of debug and subsystem masks
4247 - fix lfs find --obd (2510)
4248 - /proc tunable for disabling filter read caching (2591)
4249 - stop rpm packages from altering slapd.conf (2301)
4250 - disable nagle in the socknal under 0conf (2578)
4251 - choose mds inode size based on stripe count (2572)
4252 - fix kernel-source rpm problems (2516)
4254 - add --disable-doc to avoid pdf generation (2421)
4255 - update documentation, tests, type-os, comments
4256 - avoid format warnings on ia64
4257 - remove the TOE NAL
4258 - tiny code cleanups by removing unused fields
4260 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4263 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4264 - remove the most blatant lies from BUILDING file (2371)
4265 - change default debug level to reasonable production setting
4266 - reduce client side cache size to reduce cache flush time
4267 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4268 - make TCP zerocopy and pinger support enabled by default (2476)
4269 - sync writes completed after process exits caused crashes (2319)
4270 - maintain correct mount count on the MDS (2356)
4271 - backout 1557, because 2316 wasn't really fixed
4272 - better file I/O statistics gathering in /proc
4273 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4274 - another kernel patch to fix zero-copy TCP function export
4275 - don't take duplicate lock when processing re-sent getattr (2420)
4276 - lctl uses obd_self_export instead of creating new conn (2353)
4277 - MDS/OST recovery case which requires object creation asserted (2425)
4278 - move lfs from /usr/sbin to /usr/bin in packages
4279 - fix race between mds_client_add and mds_client_free (2417)
4280 - use kmalloc instead of slabs in portals (2430)
4281 - don't create duplicate records when a failover MDS is present (2442)
4282 - remove unnecessary mount age check (2332)
4283 - don't remove directory inodes from locks prematurely (2451)
4284 - don't break if MDS service name is the same as hostname (2103)
4285 - fix races in client write RPC generation when cache full (2482)
4287 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4290 - remove now-unused request->rq_obd (278)
4291 - if an allocation fails, print out how much memory we've used (1933)
4292 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4293 - add the "configurable stack size" patch to most series files (1256)
4294 - ability to write large log records, for 100+ OST configs (2306)
4295 - fix NULL deref when filter_prep fails (2314)
4296 - fix operator precedence error in filter_sync
4297 - dynamic allocation of socknal TX descriptors (2315)
4298 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4299 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4300 - ensure that CWARN messages in Portals always get to the syslog
4301 - __init/__exit are not for prototype decls (ldlm_init/exit)
4302 - x86-64 compile warning fixes
4303 - fix gateway LMC keyword conflict (2318)
4304 - fix MDS lock inversions in getattr/reint paths (1844)
4305 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4306 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4307 - DLM race condition prevented some lock evictions (2328)
4308 - ENOMEM detection and retry on socknal sends (2230)
4309 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4310 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4311 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4312 - better error messages when a client is rejected during recovery (1505)
4313 - avoid cancelling locks which were never granted, after failure (2330)
4314 - fix i_sem/journal inversion in mds_client_add (2333)
4315 - fix truncate/getattr lock cycle deadlock (2334)
4316 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4317 - fix two rare exit paths which could leak an l_lock() ref (2321)
4318 - fencepost error in MDS/OST orphan recovery (2226)
4319 - make log record alignment 8 bytes (1988)
4320 - lstripe now fails when requested offset > ost_count (2237)
4321 - ensure that all kernel series have a complete list.h (1607)
4322 - fix crashes in special-file operations (2316)
4323 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4325 - allow configurable automake binary, for testing new versions
4326 - small update to the lfs documentation
4328 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4330 * fix negative export reference count in fsfilt_sync (2312)
4332 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4333 * release candidate 0.9.1
4335 - orphans are moved into the PENDING directory for possible recovery
4336 - replayed opens now open by fid for orphan/rename safety (1042)
4337 - last close of an orphan inode generates a transno (683)
4338 - chdir() and mount() now pin the directory entry (1020)
4339 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4340 - discard very old requests without processing them (1502)
4341 - remove some common, well-understood CERRORs (1505)
4342 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4343 - clear "grant" flags in OST replies until OST grant code lands (1644)
4344 - fix read performance by not clobbering i_blksize on client (1598)
4345 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4346 - make LRU size a /proc tunable, clears locks when reduced (707)
4347 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4348 - server locks take references on exports now (1558)
4349 - build fixes for 2.4.20-rh trees (1663)
4350 - return an error from lov_create if all OSCs are inactive (1751)
4351 - fix import levels when a reconnect happens without a timeout (1597)
4352 - exit early from mds_open if we get a lookup error (1749)
4353 - partial page read at EOF wouldn't wait for disk before sending (1642)
4354 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4355 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4356 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4357 - write/truncate lock inversion (1639)
4358 - Don't auto-load obdclass, portals modules during cleanup (1495)
4359 - fix timestamps from jumping to "now" (1763)
4360 - extra journal assertions (1648)
4361 - add an extra multiunlink test (1771)
4362 - fix read_record/write_record API (1776)
4363 - fix leak of offset_extent, possible incorrect i_size later (1772)
4364 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4365 - Debugging for the kqswnal_get_idle_tx problems (1820)
4366 - Allow recovery to be attempted multiple times (1536)
4367 - Write out MDS last_rcvd file after it is first created (1600)
4368 - Fix tx_descriptor leak in failed transmit situations (1827)
4369 - ext3 journaling fixes for assertion failure after IO error (1871)
4370 - class_export_put() on freed export after completion AST error (1896)
4371 - Fix revalidate looping in VFS (1322)
4372 - Don't access a freed export during MDS_REINT timeout (1521)
4373 - Add open-unlink recovery support on the MDS (1673,1764)
4374 - Return an error if no MDS data was read from last_rcvd (1946)
4375 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4376 - Don't setup a disk device that doesn't match exported UUID (317)
4377 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4378 - avoid committing NULL handle in force close
4379 - local.sh is now a one-stripe LOV configuration
4380 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4381 - fix ext3 htree duplicate directory entry corruption (1516)
4382 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4383 - update to new LOV EA format (2097)
4384 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4386 2003-06-15 Phil Schwan <phil@clusterfs.com>
4389 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4390 - per-import/export recovery handling (958, 931, 959)
4391 - multiple last-rcvd slots, for serving multiple FSes (949)
4392 - connections are again shared between multiple imp/exports (963, 964)
4393 - "umount -f" would hang if any requests needed to be sent (393, 978)
4394 - avoid pinning large req buffer by copying for queued messages (989)
4395 - add "uuid" to "lctl device" command to help upcalls (991)
4396 - "open" RPCs with transnos would confuse recovery counters (1037)
4397 - do proper endian conversion of all wire messages (288, 340, 891)
4398 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4399 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4400 - fix ptlrpc_connection leak in target_handle_connect (1174)
4401 - fix import refcounting bug in OST and MDS cleanup (1134)
4402 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4403 - fix very unlikely obd_types race condition (501)
4404 - remove osc_open hack for echo_client (1187)
4405 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4406 - a failure in read_inode2 leads to deadlock (1139)
4407 - cancel ack-locks as soon as transaction is committed (1072)
4408 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4409 - make sure to commitrw after any preprw to avoid deadlock (1162)
4410 - failing to execute a file in a lustre FS would lock inode (1203)
4411 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4412 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4413 - obd_uuid2tgt would walk off the end of the list (1255)
4414 - on IA64 the osc would give portals incorrect bulk size (1258)
4415 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4416 - fix lock inversion caused by new llite matching code (1282)
4417 - limit the number of dirty pages on a client to 10MB (1286)
4418 - timed out locks were not being corrected cancelled (1289)
4419 - fix O_DIRECT above 4GB on IA-32 (1292)
4420 * major user-visible changes
4421 - fail out/fail over policy now controlled by the upcall (993)
4423 - add OBD_PING to check server availability and failure (954)
4424 - lustre messages are now sent in sending host order (288, 340, 891)
4425 - add eadatalen to MDS getattr reply (340)
4426 - OST read replies may contain second buffer, with per-page status (593)
4428 2003-03-11 Phil Schwan <phil@clusterfs.com>
4431 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4432 - failed open()s could cause deadlock; fixed (867, 869)
4433 - stop cancelling OST locks when files are closed (481)
4434 - overlapping XID spaces caused network corruption (851, 853)
4435 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4436 - setattr_raw added, to do single-RPC, server-side setattrs
4437 - lmc/lconf syntax change for OST UUIDs
4438 - fix crashy race condition between ptlrpc_free_req and osc_close
4439 - don't use request in mdc_enqueue if we hit a timeout (889)
4440 - don't set the inode i_size for regular files from the MDS (896)
4441 - handle out of order completion AST (842)
4442 - don't LBUG if a lock request times out after receiving AST (913)
4443 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4444 - if a bad lock AST arrives, send an error instead of dropping entirely
4445 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4446 - fix leak in bulk IO when only partially completed (899, 900, 926)
4447 - fix O_DIRECT for ia64 (55)
4448 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4449 - C-z after timeout could hang a process forever; fixed (977)
4451 - client-side I/O cache (678, 924, 929, 941, 970)
4453 - READPAGE and SETATTRs which don't take server-side locks get
4456 2003-02-11 Phil Schwan <phil@clusterfs.com>
4459 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4460 - Fix fsfilt_extN_readpage() to read a full page of directory
4461 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4462 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4463 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4464 - Most intent operations were removed, in favour of a new RPC mode
4465 that does a single RPC to the server and bypasses most of the VFS
4466 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4467 - Aggressively cancel local locks on DLM servers
4468 - mds_reint_unlink sends EA to the client if it's the last nlink.
4469 client uses that EA to unlink OST objects.
4470 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4471 - recursive symlinks were fixed (439)
4472 - fixed NULL deref in DEBUG_REQ
4473 - filter_update_lastobjid no longer calls sync, which annoyed extN
4474 - fixed multi-client small-writes to a single file problem (445)
4475 - fixed mtime updates during file writes (607)
4476 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4477 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4478 - fixed Linux OST size reporting problem (444, 656)
4479 - OST now updates object mtime with writes or setattr (607, 619)
4480 - client verifies file size before zeroing page past EOF (445)
4481 - OST now writes last allocated objid to disk with allocation (108)
4482 - LOV on echo now works (409)
4484 - mds_reint_unlink sends a new buffer, with the EA included. this
4485 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4486 set if a regular file was being unlinked, and it was the last link
4487 - use PtlGet from the target for bulk writes (315)
4488 - OST now updates object mtime with writes or setattr (607, 619)
4489 - LDLM now has a grant-time callback to revalidate locked items, if
4491 - Many MDS operations were reorganized to combat race conditions
4493 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4494 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4496 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4499 - Fully reactivate OST imports after reconnection (512, others)
4500 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4501 - More graceful error handling for truncating on dead OST (515)
4502 - Don't error out unless we're actually accessing dead stripes (474)
4503 - Fix garbage sizes when stripes are missing (410)
4504 - LRU counters were broken, causing constant lock purge (433, 432)
4505 - garbage on read from stripes with failed OSTs (441)
4506 - mark OSCs as active before reconnecting during recovery (438)
4507 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4508 - lfind did not preserve OST order in output (443)
4509 - symlinks cause hung clients, incorrect data (439)
4510 - stop dereferencing request after dropping refcount (457)
4511 - don't LASSERT(spin_is_locked) on non-SMP (455)
4512 - fixes for many rename() bugs
4513 - fstat didn't correctly synchronize attributes (399)
4514 - server must handle lock cancellation during blocking AST prep (487)
4515 - bulk descriptors were free()d too soon (511)
4516 - fix paths in lconf, which would load incorrect modules (451, 507)
4517 - fix confusing lconf 'host not found' error message (386)
4518 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4519 - fix race condition in mdc_blocking_ast() for inode access (526)
4520 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4521 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4522 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4523 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4524 - fix lstripe in several ways (kernel side) (527)
4525 - fix request leak in ldlm_cli_enqueue (262)
4526 - incorrect OSC was marked inactive after OST failure
4527 - call mds_fs_cleanup before unmounting filesystem (524)
4528 - fix races between taking ns_lock and ldlm_lock_change_resource
4529 - fix races updating LOV export open file list
4530 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4531 - fix recovery NULL deref in ldlm_cli_cancel_unused
4532 - fix some DLM races by using new hash table for lock handles (419)
4533 - permit the client to specify desired inodes, at replay
4534 - duplicate requests when we queue them for replay reintegration
4535 - fix last_rcvd offset calculation
4536 - sync after each recovered transaction, so we always make progress
4537 - never, not always, ERESTART requests without transnos
4538 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4540 - skip replay if the MDS says that the client is already connected
4541 - don't check for a recovery-enabled export to match lctl's UUID
4542 - don't INC_USE_COUNT for phantom exports
4543 - don't crash when cleaning up phantom exports (567)
4544 - don't double-finish or set replay data for errored mdc_open requests
4545 - abort requests when they time out, so we don't get old replies
4546 - send/receive replies for AST messages again
4547 - if the client says that it doesn't have the lock, cancel it on the
4549 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4550 mark it as destroyed, it will all work out when decref is called
4551 - fix module use counts (22, 581)
4553 - ASTs now expect a reply (server cancels lock on error reply)
4555 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4558 - fix many simultaneous client startup (392)
4559 - fix dentry->d_it clobbering
4560 - credentials weren't being shipped for readdir/getattr operations
4561 - remove invalid assertions triggered during some concurrent MD
4563 - proper Lustre versions added (336, 389)
4564 - fix memory leak for create error case (398)
4565 - fix LOV locking bug that would get cli/srv out of sync
4566 - fix echo client over LOV (409)
4567 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4568 - fix double-O_EXCL intent crash (424)
4569 - avoid sending multiple lock CANCELs (352)
4571 - MDS can do multi-client recovery (modulo bugs in new code)
4573 - many updates, edits, cleanups
4575 2002-11-18 Phil Schwan <phil@clusterfs.com>
4578 - fix null d_it dereference (346)
4579 - fix full OST/dbench hang (333)
4580 - fix permission problem with file removal (286)
4581 - fix removal of OSCs from LOV when they fail
4582 - fix NULL deref during bulk timeout (214)
4583 - fix problems related to multiple filesystems on one MDS (241)
4584 - fixed serious subtle metadata locking bugs
4585 - free locks on clients when inodes are removed due to memory
4587 - fix inode pointer in lock data (285)
4588 - partial support for multiple MDS on a single host (241)
4589 - data locks weren't cancelled at clear_inode time (290, 311)
4590 - intent locks could lead to unbounded lock growth (205)
4591 - added a maximum lock count, an LRU list, and a flusher
4592 - fix multiple rename (365)
4593 - properly abstracted the echo client
4594 - OSC locked 1 byte too many; fixed
4595 - rewrote brw callback code:
4596 - fixed recovery bugs related to LOVs (306)
4597 - fixed too-many-pages-in-one-write crash (191)
4598 - fixed (again) crash in sync_io_timeout (214)
4599 - probably fixed callback-related race (385)
4601 - Add capability to MDS protocol
4602 - LDLM cancellations and callbacks on different portals
4604 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4607 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4608 - defer open object destruction to close time (601981, 138)
4609 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4610 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4611 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4612 - don't call obd_setattr in ll_file_release for destroyed objects
4613 * protocol change to lustre_msg: move |version| and add |flags|
4614 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4615 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4616 * added replay of create, unlink, link and rename operations during
4617 MDS failover; recovery should be much more robust now
4618 * remove failed OSCs from LOVs (only lov_create uses this so far)
4619 * the lustre-HOWTO was brought (more) up to date (582544)
4621 2002-10-23 Phil Schwan <phil@clusterfs.com>
4624 - in-use dentries weren't being reused properly (617851)
4625 - prevent multiple LDLM setup (599178)
4626 - fix LOV size calculations for truncate (617853)
4627 - fix client handling of MDS intent errors (POSIX)
4628 - fix permission bug in lovstripe.c test (624321)
4629 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4630 - truncate past end of file could corrupt data
4631 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4632 - a race in recovery could return ETIMEDOUT to apps (623947)
4633 - building outside the source directory was fixed
4634 * the lustre-HOWTO was brought (more) up to date (582544)
4635 * major progress was made on recovery functionality
4637 2002-10-10 Phil Schwan <phil@clusterfs.com>
4640 - recovery deadlock fix
4641 - rm -rf causes LBUG fix (617817)
4642 - file open by multiple tasks fix (618962)
4643 - directory permissions bugs (602707 and 620007)
4644 - journal_stop fixed with locking (611313)
4645 - O_APPEND failures resolved (618273, perhaps 614459)
4646 - lconf PATH fix (619770)
4647 - IA64 build fix (621450)
4648 - RPC buffer sizes scale with amount of memory
4650 2002-10-01 Phil Schwan <phil@clusterfs.com>
4653 - locks would be cancelled without throwing away data pages,
4654 resulting in inconsistent data (605627)
4655 - inode attributes were not always being refreshed (605627, 612449)
4656 - lconf now continues to cleanup after lctl reports an error
4657 - MDS now enforces user permissions (602707)
4658 - lprocfs cleanup fixed, but not yet enabled (614157)
4659 - fixed infinite server hang, should a client not respond to an AST
4660 - avoid going into recovery if user calls readlink() with a buffer
4661 that's too small (613941)
4662 - AST RPCs no longer require replies (614867) -- this may be changed
4663 - don't crash server if client sends an IOV that's too big (611336)
4664 - fixed lock conversion deadlock (611892)
4665 - fixed the following of symlinks (614622)
4666 * recovery: the server can remove locks from a client that dies, other
4667 clients can make progress
4668 * more extN patch fixes
4669 * compile-time configurable ptlrpc buffer allocations
4671 - collaborative read cache document
4672 - Lustre Lite Performance CDR document-in-progress
4674 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4677 - fix typo in patch-2.4.18
4679 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4682 - clear ptlrpc request each time in handle_incoming_request()
4683 - unlink of files now destroys the object on the OST
4685 2002-09-19 Peter Braam <braam@clusterfs.com>
4687 * add hard link support
4688 * change obdfile creation method
4689 * kernel patch changed
4691 2002-09-19 Peter Braam <braam@clusterfs.com>
4694 - stack overflow bug in extN fixed
4696 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4698 * documentation updates
4699 - add man pages for config tools
4700 - update tests/README to describe testing with new config tools
4701 - finish metadata API descriptions
4702 * bug fixes and cleanups
4703 - statfs workaround for 16TB limit
4704 - LOV stripe allocation improved, can stripe on subset of OSTs
4705 - LOV file size/IO offset was wrong for files > 4GB in size
4706 - object EA data was being dropped, caused files to be unreadable
4707 - memory overflow with non-LOV OST caused memory corruption
4708 - fixed regression tests to work with new config tools, obdfilter
4709 - fixed bug when directory size became larger than 1 block
4710 - fixed bug (for single client case) when PWD was deleted
4711 - invalidate local directory pages when doing intent-based ops
4712 - avoid LDLM oops when lock callback contained bad data
4714 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4716 * documentation updates
4717 * bug fixes and cleanups
4718 - configuration tools
4721 - 64-bit compile warnings
4722 - 64-bit internal statfs data
4724 * test_brw on persistent OST devices
4726 * lprocfs (disabled)
4728 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4730 * documentation updates
4731 * bug fixes and cleanups
4732 * configuration tools
4734 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4737 * many small fixes to 0_5_4
4738 * io/network handling
4739 * thinkos in MDS operations
4741 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4744 * crucial basic fixes to 0.5.3
4745 * IOR, Iozone work over Elan
4748 2002-08-07 Phil Schwan <phil@clusterfs.com>
4749 * version 0_5_3, our first alpha
4750 * we use the new Portals iovs
4751 * documentation updates
4752 * bug fixes and cleanups
4753 * small changes in the DLM wire protocol
4755 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4756 * version 0_5_1 with some initial stability,
4757 * locking on MD and file I/O.
4758 * documentation updates
4759 * several bug fixes since 0.5.0
4760 * small changes in wire protocol
4762 2002-07-18 Phil Schwan <phil@clusterfs.com>
4764 * delivered as Lustre Light Alpha
4765 * fixed a crash after handling invalid MDS requests
4766 * fixed directory pages for architectures with non-4k pages sizes
4768 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4769 * release version v0_4_4
4770 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4772 * Fixes a number of interruption problems with OST operations.
4773 * Update documentation for portals header changes
4774 * Move all wire protocol structs/defines to lustre_idl.h
4775 * Fixes symlink length bug.
4776 * Add tcpdump to repository.
4778 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4779 * release version v0_4_3
4780 * Fixes statfs for inodes on extN.
4781 * Fixes bug in runtests which would delete /etc/hosts.
4782 * Use 64-bit object IDs wherever possible (not into VFS though)
4783 Remove ost_get_info, which is unused by lustre, and out of date.
4785 2002-07-03 Peter Braam <braam@clusterfs.com>
4786 * release version v0_4_2 Fixes a lookup error (type not passed)
4787 * move forward to head of Portals
4788 * move forward to latest Lustre kernel
4790 2002-06-25 Peter Braam <braam@clusterfs.com>
4791 * release version v0_4_1. Hopefully stable on single node use.