1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.33 (SLES 10),
5 2.6.18-128.1.1.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: 1.40.11-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).
14 * File join has been disabled in this release, refer to Bugzilla 16929.
16 Severity : enhancement
18 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
19 Details : allow skip disconnected ost for send statfs request and hide error
25 Description: fix racy locking of mballoc block bitmaps causing BUG
26 Details : The locking of the mballoc buddy bitmap and the in-memory
27 block bitmap was using two different spin locks in some
28 cases. This made it possible to incorrectly access the
29 mballoc bitmap while another process was modifying it,
30 causing a sanity assertion to fail. While no on-disk corruption
31 was reported, there was some risk of this happening.
34 Frequency : rare, on llog test 6
36 Descriptoin: don't allow connect to already connected import
37 Details : allowing connect to already connected import is hide connecting problem.
42 Descriptoin: allow kill process which wait statahead result
43 Details : for some reasons 'ls' can stick in waiting result from statahead,
44 in this case need way for kill this process.
46 Severity : enhancement
48 Description: Add state history info file, enhance import info file
49 Details : Track import connection state changes in a new osc/mdc proc file;
50 add overview-type data to the osc/mdc import proc file.
52 Severity : enhancement
54 Description: MDS create should not wait for statfs RPC while holding DLM lock.
56 Severity : enhancement
58 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
63 Descriptoin: OST grant too much space to client even there are not enough space.
64 Details : Client will shrink its grant cache to OST if there are no write
65 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
66 this grant cache if there are already not enough avaible space
67 (left_space < total_clients * 32M).
70 Frequency : start MDS on uncleanly shutdowned MDS device
72 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
73 Details : stay in waiting mds<>ost recovery finished produce random bugs
74 due race between two ll_sync thread for one lov target. send
75 ACTIVATE event only if connect realy finished and import have
79 Frequency : rare, connect and disconnect target at same time
81 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
82 Details : don't call obd_disconnect under lov_lock. this long time
83 operation and can block ptlrpcd which answer to connect request.
88 Descriptoin: don't lose wakeup for imp_recovery_waitq
89 Details : recover_import_no_retry or invalidate_import and import_close can
90 both sleep on imp_recovery_waitq, but we was send only one wakeup
94 Frequency : always with long access acl
96 Descriptoin: mds can't pack reply with long acl.
97 Details : mds don't control size of acl but they limited by reint/getattr
100 Severity : enhancement
102 Description: Update to SLES10 kernel-2.6.16.60-0.33.
104 Severity : enhancement
106 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
109 Frequency : start MDS on uncleanly shutdowned MDS device
111 Descriptoin: aborting recovery hang on MDS
112 Details : don't throttle destroy RPCs for the MDT.
115 Frequency : on remount
117 Description: external journal device not working after the remount
118 Details : clear dev_rdonly flag for external journal devices in
124 Description: shutdown vs evict race
125 Details : client_disconnect_export vs connect request race.
126 if client will evicted at this time - we start invalidate
127 thread without referece to import and import can be freed
131 Frequency : rare, need acl's on inode.
133 Description: client can't handle ost additional correctly
134 Details : if ost was added after client connected to mds client can have
135 hit lnet_try_match_md ... to big messages to wide striped files.
136 in this case need teach client to handle config events about add
137 lov target and update client max ea size at that event.
139 Severity : enhancement
141 Description: Changelogs
142 Details : Changelogs are a lightweight mechanism to track filesystem
143 metadata and namespace changes. The changelog is recorded
144 permanently on the MDTs, and is periodically "consumed" / purged
145 when records are no longer needed.
147 Severity : enhancement
149 Description: compact fld format with extents
150 Details : Store range of seq rather than every seq in FLD. Seq
151 controller update FLD rather than clients. In Case of CMD, mdt0
152 has FLD, all other metadata server act as non persistent proxy
153 for FLD queries and cache fld entries in fld cache.
158 Description: don't skip ost target if they assigned to file
159 Details : Drop slow OSCs if we can, but not for requested start idx.
160 This means "if OSC is slow and it is not the requested
161 start OST, then it can be skipped, otherwise skip it only
162 if it is inactive/recovering/out-of-space.
166 Description: more cleanup in mds_lov
167 Details : not send LOV EA under replay, we can't know about they size at this
168 time. Don't allow client connect to mds before any ost connected,
169 for avoid problems with LOV EA size and returning EIO to client.
171 Severity : enhancement
173 Description: Interoperability at server side (Disk interoperability)
175 Severity : enhancement
177 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
179 Severity : enhancement
181 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
183 Severity : enhancement
185 Description: New client IO stack (CLIO).
187 Severity : enhancement
189 Description: Commit on sharing. Eliminate inter-client dependencies between
190 uncommitted transactions by doing transaction commits.
191 Thereby clients may recovery independently.
194 Frequency : Create a symlink file with a very long name
196 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
197 Details : If there is no extra space in the request for early cancels,
198 ldlm_req_handles_avail() returns 0 instead of a negative value.
200 Severity : enhancement
202 Description: Add /proc entry for import status
203 Details : The mdc, osc, and mgc import directories now have
204 an import directory that contains useful import data for debugging
207 Severity : enhancement
209 Description: Re-disable certain /proc logging
210 Details : Enable and disable client's offset_stats, extents_stats and
211 extents_stats_per_process stats logging on the fly.
214 Frequency : Only on FC kernels 2.6.22+
216 Description: oops in statahead
217 Details : Do not drop reference count for the dentry from VFS when lookup,
218 VFS will do that by itself.
220 Severity : enhancement
222 Description: Generic /proc file permissions
223 Details : Set /Proc file permissions in a more generic way to enable non-
224 root users operate on some /proc files.
228 Description: Hitting mdc_commit_close() ASSERTION
229 Details : Properly handle request reference release in
230 ll_release_openhandle().
234 Description: quota recovery deadlock during mds failover
235 Details : This patch includes att18982, att18236, att18237 in bz14840.
237 1. fix osts hang when mds does failover with quotaon
238 2. prevent watchdog storm when osts threads wait for the
243 Frequency : only patchless client
244 Description: add workaround for race between add/remove dentry from hash
246 Severity : enhancement
248 Description: Allow OST glimpses to return PW locks
252 Description: LBUG when llog conf file is full
253 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
257 Description: Prevent import from entering FULL state when server in recovery
261 Description: service mount cannot take device name with ":"
262 Details : Only when device name contains ":/" will mount treat it as
268 Description: replace ptlrpcd with the statahead thread to interpret the async
269 statahead RPC callback
273 Frequency : on recovery
274 Description: I/O failures after umount during fail back
275 Details : if client reconnected to restarted server we need join to recovery
276 instead of find server handler is changed and process self
277 eviction with cancel all locks.
279 Severity : enhancement
281 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
283 Severity : enhancement
285 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
287 Severity : enhancement
289 Description: Upcall on Lustre log has been dumped
290 Details : Allow for a user mode script to be called once a Lustre log has
291 been dumped. It passes the filename of the dumped log to the
292 script, the location of the script can be specified via
293 /proc/sys/lnet/debug_log_upcall.
298 Description: avoid idr_remove called for id which is not allocated.
299 Details : Move assigment s_dev for clustered nfs to end of initialization,
300 to avoid problem with error handling.
305 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
306 Details : When connection is reused this not moved from CONN_UNUSED_HASH
307 into CONN_USED_HASH and this prodice warning when put connection
308 again in unused hash.
311 Severity : enhancement
313 Description: File striping can now be set to use an arbitrary pool of OSTs.
315 Severity : enhancement
317 Description: Export bytes_read/bytes_write count on OSC/OST.
321 Description: Early reply size mismatch, MGC loses connection
322 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
323 the connect flags are properly negotiated.
328 Description: quotas are not honored with O_DIRECT
329 Details : all writes with the flag O_DIRECT will use grants which leads to
330 this problem. Now using OBD_BRW_SYNC to guard this.
334 Description: add quota statistics
335 Details : 1. sort out quota proc entries and proc code.
336 2. add quota statistics
338 Severity : enhancement
340 Description: enable quota support for HEAD.
344 Description: Properly propagate oinfo flags from lov to osc for statfs
345 Details : restore missing copy oi_flags to lov requests.
347 Severity : enhancement
349 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
350 : l_getgroups(8), lst(8), routerstat(8)
351 Details : included man pages for llobdstat(8), llstat(8),
352 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
354 Severity : enhancement
356 Description: configure's --enable-quota should check the
357 : kernel .config for CONFIG_QUOTA
358 Details : configure is terminated if --enable-quota is passed but
359 : no quota support is in kernel
363 Description: Remove portals compatibility
364 Details : Remove portals compatibility, not interoperable with releases
369 Description: Resolve device initialization race
370 Details : Prevent proc handler from accessing devices added to the
371 obd_devs array but yet be intialized.
373 Severity : enhancement
375 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
377 Severity : enhancement
379 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
384 Description: Using wrong pointer in osc_brw_prep_request
385 Details : Access to array[-1] can produce panic if kernel compiled with
386 CONFIG_PAGE_ALLOC enabled
390 Description: Client runs out of low memory
391 Details : Consider only lowmem when counting initial number of llap pages
395 Description: *optional* service tags registration
396 Details : if the "service tags" package is installed on a Lustre node
397 When the filesystem is mounted, a local-node service tag will
398 be created. See http://inventory.sun.com/ for more information
399 about the Service Tags asset management system.
403 Description: Kernel BUG tries to release flock
404 Details : Lustre does not destroy flock lock before last reference goes
405 away. So always drop flock locks when client is evicted and
406 perform unlock regardless of successfulness of speaking to MDS.
410 Description: add refcount protection for osc callbacks, avoid panic on shutdown
414 Description: sanity test 65a fails if stripecount of -1 is set
415 Details : handle -1 striping on filesystem in ll_dirstripe_verify
420 Description: ASSERTION(CheckWriteback(page,cmd)) failed
421 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
424 Severity : enhancement
426 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
430 Description: do not process already freed flock
431 Details : flock can possibly be freed by another thread before it reaches
432 to ldlm_flock_completion_ast.
436 Description: LBUG during stress test
437 Details : Need properly lock accesses the flock deadlock detection list.
441 Description: oops in page fault handler
442 Details : kernel page fault handler can return two special 'pages' in error
443 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
447 Description: timeout with invalidate import.
448 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
449 should be handled by ptlrpcd. This produce long age waiting and
450 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
452 Severity : enhancement
454 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
458 Description: enable MGS and MDT services start separately
459 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
460 a co-located MGS without starting the MGS, which is a complement
461 to 'nosvc' mount option.
465 Frequency : after recovery
466 Description: precreate to many object's after del orphan.
467 Details : del orphan st in oscc last_id == next_id and this triger growing
468 count of precreated objects. Set flag LOW to skip increase count
469 of precreated objects.
473 Frequency : rare, on clear nid stats
474 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
475 Details : when clean nid stats sometimes try destroy live entry,
476 and this produce panic in free.
480 Description: Stack overflow during MDS log replay
481 ease stack pressure by using a thread dealing llog_process.
485 Description: wait until IO finished before start new when do lock cancel.
486 Details : VM protocol want old IO finished before start new, in this case
487 need wait until PG_writeback is cleared until check dirty flag and
488 call writepages in lock cancel callback.
490 Severity : enhancement
492 Description: using special macro for print time and cleanup in includes.
496 Description: mds_mfd_close() ASSERTION(rc == 0)
497 Details : In mds_mfd_close(), we need protect inode's writecount change
498 within its orphan write semaphore to prevent possible races.
502 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
506 Frequency : rare, on shutdown ost
507 Description: don't hit live lock with umount ost.
508 Details : shrink_dcache_parent can be in long loop with destroy dentries,
509 use shrink_dcache_sb instead.
513 Description: don't panic with use echo client
514 Details : echo client pass NULL as client nid pointer and this produce null
519 Description: fix build on ppc32
520 Details : compile code with -m64 flag produce wrong object file for ppc32.
524 Description: add message levels for liblustreapi
528 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
529 Details : Move the 'good_osts' check before the 'total_bavail' check. This
530 will result in an -EAGAIN and in the exit call path we call
531 alloc_rr() which will with increasing aggressiveness attempt to
532 aquire precreated objects on the minimum number of required OSCs.
536 Description: Use old size assignment to avoid deadlock
537 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
538 the scheduling while holding a spinlock. We do not need locking
539 for size in ll_update_inode() because size is only updated from
540 the MDS for directories or files without objects, so there is no
541 other place to do the update, and concurrent access to such inodes
542 are protected by the inode lock.
546 Description: resolve "_IOWR redefined" build error on SLES10
550 Description: dump the memory debugging after all modules are unloaded to
551 suppress false negative in conf_sanity test 39
553 Severity : enhancement
555 Description: build kernel-ib packages for OFED 1.3 in our release cycle
560 Description: fix SLES kernel versioning
561 Details : the kernel version for our SLES 10 kernel did not include a "-"
562 before the "smp" at the end. while this was not a problem in
563 general, it did mean that software trying to use the kernel
564 version to try to detect a vendor specific kernel would fail.
565 this was most evident by the OFED build scripts.
569 Description: Don't update lov_desc members until making sure they are valid
570 Details : When updating lov_desc members via proc fs, need fix their
571 validities before doing the real update.
575 Description: don't put request into delay list while invalidate in flight.
576 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
577 import in flight. this produce timeout for invalidate and sometimes
578 can cause stale data.
582 Frequency : on ppc only
583 Description: not convert ost objects for directory because it's not exist.
584 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
588 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
590 Severity : enhancement
592 Description: Update to support 2.6.22.14 vanilla kernel.
596 Frequency : rare, on recovery
597 Description: read procfs can produce deadlock in some situation
598 Details : Holding lprocfs lock which send rpc can produce block for destroy
599 obd objects and this also block reconnect with -EALREADY. This
600 isn't fix all lprocfs bugs - but make it rare.
602 Severity : enhancement
604 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
607 Frequency : frequent on X2 node
609 Description: mdc_set_open_replay_data LBUG
610 Details : Set replay data for requests that are eligible for replay.
614 Description: lustre_mgs: operation 101 on unconnected MGS
615 Details : When MGC is disconnected from MGS long enough, MGS will evict the
616 MGC, and late on MGC cannot successfully connect to MGS and a lot
617 of the error messages complaining that MGS is not connected.
621 Frequency : on network error
622 Description: panic with double free request if network error
623 Details : mdc_finish_enqueue is finish request if any network error ocuring,
624 but it's true only for synchronus enqueue, for async enqueue
625 (via ptlrpcd) this incorrect and ptlrpcd want finish request
628 Severity : enhancement
630 Description: client-side metadata stat-ahead during readdir(directory readahead)
631 Details : perform client-side metadata stat-ahead when the client detects
632 readdir and sequential stat of dir entries therein
635 Frequency : on start mds
637 Description: Implement get_info(last_id) in obdfilter.
640 Frequency : occasional
642 Description: Correctly check stale fid, not start epoch if ost not support SOM
643 Details : open with flag O_CREATE need set old fid in op_fid3 because
644 op_fid2 was overwritten with new generated fid, but mds can answer
645 with one of these two fids and both is not stale. Setattr
646 incorrectly started an epoch and assume will be called
647 done_writing, but without SOM done_writing ever being called.
650 Frequency : rare, depends on device drivers and load
652 Description: MDS or OSS nodes crash due to stack overflow
653 Details : Code changes in 1.8.0 increased the stack usage of some functions.
654 In some cases, in conjunction with device drivers that use a lot
655 of stack the MDS (or possibly OSS) service threads could overflow
656 the stack. One change which was identified to consume additional
657 stack has been reworked to avoid the extra stack usage.
660 Frequency : occasional
662 Description: Do not fail import if osc_interpret_create gets -EAGAIN
663 Details : If osc_interpret_create got -EAGAIN it immediately exits and
664 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
665 oscc_has_objects and see OSC has no objests and call
666 oscc_internal_create to resend create request.
668 Severity : enhancement
670 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
672 Severity : enhancement
674 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
677 Frequency : very rare
679 Description: Fix replay if there is an un-replied request and open
680 Details : In some cases, older replay request will revert the
681 mcd->mcd_last_xid on MDS which is used to record the client's
684 Severity : enhancement
686 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
688 Severity : enhancement
690 Description: Add rhel5 support to HEAD.
692 Serverity : enhancement
694 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
699 Description: Don't allow skipping OSTs if index has been specified.
700 Details : Don't allow skipping OSTs if index has been specified, make
701 locking in internal create lots better.
705 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
706 Details : access to bitfield in structure is always rounded to long
707 and this produce problem with not atomic change any bit.
711 Description: Lustre make rpms failed.
712 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
713 builds Lustre rpms from tarball.
716 Frequency : rare on shutdown ost
718 Description: If llog cancel was not send before clean_exports phase, this can
719 produce deadlock in llog code.
720 Details : If llog thread has last reference to obd and call class_import_put
721 this produce deadlock because llog_cleanup_commit_master wait when
722 last llog_commit_thread exited, but this never success because was
723 called from llog_commit_thread.
727 Description: allow userland application know is lost one of stripes.
728 Details : fill lvb_blocks with error code on ost and return it to
729 application if error flag found.
733 Description: NULL lov_tgts causing MDS oops
734 Details : more safe checks for NULL lov_tgts for avoid oops.
736 Severity : enhancement
738 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
742 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
743 Details : If an OST is down the MDS will hang indefinitely in
744 obd_statfs_rqset() waiting for the statfs data. While for
745 MDS QOS usage of statfs, it should not stuck in waiting.
747 Severity : enhancement
749 Description: remote_acl support
750 Details : Support ACL-based permission check for remote user.
751 Support setfacl/getfacl for remote user with the utils
752 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
753 the system "{s,g}etfacl" utils.
755 Severity : enhancement
757 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
759 Severity : enhancement
761 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
765 Description: make mgs_setparam() handle fsname containing dash
766 Details : fsname containing a dash does not work with lctl conf_param
768 Severity : enhancement
770 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
772 Severity : enhancement
774 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
778 Frequency : rare, at shutdown
779 Description: access already free / zero obd_namespace.
780 Details : if client_disconnect_export was called without force flag set,
781 and exist connect request in flight, this can produce access to
782 NULL pointer (or already free pointer) when connect_interpret
783 store ocd flags in obd_namespace.
787 Frequency : only at startup
788 Description: not alloc memory with spinlock held.
789 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
790 if any spinlock held.
792 Severity : enhancement
794 Description: make lustre randomly fail allocating memory
795 Details : Make lustre randomly failed allocating memory for testing purpose.
797 Severity : enhancement
799 Description: lost problems with lov objid file
800 Details : Fixes some scability and access to not inited memory problems
801 in work with lov objdid file.
806 Description: lfs find does not continue on file error
807 Details : Continue other files processing when a file/dir is absent.
811 Description: Inconsistent usage of lustre_pack_reply()
812 Details : Standardize the usage of lustre_pack_reply() such that it
813 always generate a CERROR on failure.
816 Frequency : occasional
818 Description: MDS hang or stay in waiting lock
819 Details : If client receive lock with CBPENDING flag ldlm need send lock
820 cancel as separate rpc, to avoid situation when cancel request
821 can't processed due all i/o threads stay in wait lock.
825 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
829 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
833 Description: Update to SLES9 kernel-2.6.5-7.287.3.
837 Description: Update to RHEL4 latest kernel.
839 Severity : enhancement
841 Description: Build SLES10 patchless client fails
842 Details : The configure was broken by run ./configure with
843 --with-linux-obj=.... argument for patchless client. When the
844 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
845 can't search header adequately. Use absolute path such as
846 -I($LINUX)/include instead.
850 Description: interrupt oig_wait produce painc on resend.
851 Details : brw_redo_request can be used for resend requests from ptlrpcd and
852 private set, and this produce situation when rq_ptlrpcd_data not
853 copyed to new allocated request and triggered LBUG on assert
854 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
855 ptlrpcd set if request is changed and can be safety changed to use
860 Description: LASSERT_{REQ,REP}SWAB macros are buggy
861 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
862 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
863 these macros and replace them with their difinitions instead.
867 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
868 Details : Update which_patch & target file for SLES10 latest kernel.
871 Description: Remove client patches from SLES 10 kernel.
872 Details : This causes SLES 10 clients to behave as patchless clients
873 even on a Lustre-patched (server) kernel.
875 Severity : enhancement
877 Description: self-adjustable client's lru lists
878 Details : use adaptive algorithm for managing client cached locks lru
879 lists according to current server load, other client's work
880 pattern, memory activities, etc. Both, server and client
881 side namespaces provide number of proc tunables for controlling
884 Severity : enhancement
886 Description: light-weight GSS support
887 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
888 and reduce performance overhead.
890 Severity : enhancement
892 Description: Linux keyring support
893 Details : Support using service of Linux keyring for Lustre GSS internal
894 context refresh/cache mechanism.
898 Description: Fix errors in lfs documentation
899 Details : Fixes man pages
903 Description: don't use GFP_* in generic Lustre code.
904 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
908 Description: obdclass is limited by single OBD_ALLOC(idarray)
909 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
913 Description: warnings with build patchless client with vanila 2.6.19 and up
914 Details : change old ctl_table style and replace ctl_table/ctl_table_header
915 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
919 Frequency : Only for SLES
920 Description: sanity_quota fail test_1
921 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
922 file, which makes TSTID[2] inunique.
926 Frequency : Only for SLES9
927 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
928 Details : Update target/ChangeLog/which_patch .
932 Frequency : Only for test_10 in sanity-quota.sh
933 Description: a bug in quota test code
934 Details : checking a test flag in a wrong place
938 Frequency : Only for test_13 in sanity-quota.sh
939 Description: a bug in quota test script
940 Details : shouldn't check overfull allocated quota
944 Description: jbd statistics
945 Details : Port older jbd statistics patch for sles10
947 Severity : enhancement
949 Description: Add printing inode info into message about error in writepage.
954 Description: reply_lock_interpret crash due to race with it and lock cancel.
955 Details : Do not replay locks that are being cancelled. Do not reference
956 locks by their address during replay, just by their handle.
960 Frequency : When flocks are used.
961 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
962 Details : Flock locks might destroy just granted lock if it could be merged
963 with another existing flock, this is done in completion handler,
964 so teach ldlm_cli_enquque_fini that this is a valid case for
970 Description: Oops in read and write path when failing to allocate lock.
971 Details : Check if lock allocation failed and return error back.
975 Description: lstripe command fails for valid OST index
976 Details : The stripe offset is compared to lov->desc.ld_tgt_count
977 instead of lov->desc.ld_active_tgt_count.
981 Description: sanity.sh failed test 103
982 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
987 Description: df doesn't work properly if diskfs blocksize != 4K
988 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
992 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
993 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
995 Severity : enhancement
997 Description: Adaptive timeouts
998 Details : RPC timeouts adapt to changing server load and network
999 conditions to reduce resend attempts and improve recovery time.
1003 Description: llapi_file_create() does not allow some changes
1004 Details : add llapi_file_open() that allows specifying the mode and
1005 open flags, and also returns an open file handle.
1009 Description: merge and cleanup kernel patches.
1010 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1014 Description: Add journal checksum support.(Kernel part)
1015 Details : The journal checksum feature adds two new flags i.e
1016 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1017 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1018 indicates that the commit block contains the checksum for
1019 the blocks described by the descriptor blocks. Now commit
1020 record can be sent to disk without waiting for descriptor
1021 blocks to be written to disk. This behavior is controlled
1022 using JBD2_FEATURE_ASYNC_COMMIT flag.
1026 Description: OSS needs mutliple precreate threads
1027 Details : Add ability to start more than one create thread per OSS.
1031 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1032 Details : execute lfs setstripe on client
1036 Description: mds_obd_create error creating tmp object
1037 Details : When the user sets quota on root, llog will be affected and can't
1038 create files and write files.
1041 Frequency : Always on ia64 patchless client, and possibly others.
1043 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1044 Details : This allows the patchless client to be loaded on architectures
1045 without this export.
1049 Description: RedHat Update kernel for RHEL5
1050 Details : Add the kernel config file for RHEL5.
1054 Description: RedHat Update kernel for RHEL5
1055 Details : Modify the kernel config file more closer RHEL5.
1059 Description: Build failure against Centos5 (RHEL5)
1060 Details : Define PAGE_SIZE when it isn't present.
1064 Description: test_7 on sanity-quota.sh can't be used on separate machines
1065 Details : add facet to handle it
1069 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1070 Details : with NFS, the anon dentry's parent was set to itself in
1071 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1072 corresponding dentry other than use rec->ur_name.
1074 Severity : enhancement
1076 Description: lfs setstripe enhancement
1077 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1081 Description: enable data checksumming by default
1082 Details : enable checksum by default, allow --disable-checksum
1083 configure option and "-o nochecksum" mount option. Checksums
1084 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1088 Description: lustre support for RHEL5
1089 Details : Add support for RHEL5.
1093 Description: Client eviction due to failover config
1094 Details : after a connection loss, the lustre client should attempt to
1095 reconnect to the last active server first before trying the
1096 other potential connections.
1100 Description: when mds and osts use different quota unit(32bit and 64bit),
1101 quota will be released repeatly.
1102 Details : void sending multiple quota reqs to mds, which will keep the
1103 status between the reqs.
1107 Description: osts not allocated evenly to files
1108 Details : change the condition to increase offset_idx
1111 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1113 Description: Data corruption for OSTs that are formatted larger than 2TB
1115 Details : When generating the bio request for lustre file writes the
1116 sector number would overflow a temporary variable before being
1117 used for the IO. The data reads correctly from Lustre (which
1118 will overflow in a similar manner) but other file data or
1119 filesystem metadata may be corrupted in some cases.
1123 Description: Tune the kernel for good SCSI performance.
1124 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1125 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1130 Description: rewrite ext2-derived code in obdclass/uuid.c
1131 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1132 from scratch preserving functionality.
1136 Description: "lfs find -obd UUID" prints directories
1137 Details : "lfs find -obd UUID" will return all directory names instead
1138 of just file names. It is incorrect because the directories
1139 do not reside on the OSTs.
1143 Description: use i_size_read and i_size_write in 2.6 port
1144 Details : replace inode->i_size access with i_size_read/write()
1147 Frequency : when removing large files
1149 Description: scheduling issue during removal of large Lustre files
1150 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1151 It causes scheduling issues when removing large files (17TB in the
1155 Frequency : only with liblustre clients on XT3
1157 Description: evictions taking too long
1158 Details : allow llrd to evict clients directly on OSTs
1161 Frequency : only on ppc
1163 Description: /proc/fs/lustre/devices broken on ppc
1164 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1165 We should make sure the type of 'index' is loff_t and then cast
1166 explicitly as needed below. Do not assign an explicitly cast
1170 Frequency : only for rhel5
1172 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1173 Details : Modify the target file & which_kernel.
1175 Severity : enhancement
1177 Description: omit set fsid for export NFS
1178 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1179 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1183 Frequency : Always, for kernels after 2.6.16
1184 Description: Fix warning idr_remove called for id=.. which is not allocated.
1185 Details : Last kernels save old s_dev before kill super and not allow
1186 to restore from callback - restore it before call kill_anon_super.
1190 Description: buffer overruns could theoretically occur
1191 Details : llapi_semantic_traverse() modifies the "path" argument by
1192 appending values to the end of the origin string, and a
1193 overrun may occur. Adding buffer overrun check in liblustreapi.
1197 Description: Fix error on 'ls .' at the top of the Lustre mount.
1198 Details : Don't revalidate dentry if it is a root dentry.
1202 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1203 Details : Modify vm-tunables-rhel4.patch.
1207 Description: Kernel config for 2.6.18-vanilla.
1208 Details : Modify targets/2.6-vanilla.target.in.
1209 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1210 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1211 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1212 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1216 Description: improve handling recoverable errors
1217 Details : if request processig with error which can be recoverable on server
1218 request should be resend, otherwise page released from cache and
1223 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1224 Details : Modify target file & which_patch.
1225 A flaw was found in the IA32 system call emulation provided
1226 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1227 value could be stored in the %RAX register, which could trigger an
1228 out-of-bounds system call table access. An untrusted local user
1229 could exploit this flaw to run code in the kernel
1230 (ie a root privilege escalation). (CVE-2007-4573).
1234 Description: O_DIRECT bypasses client statistics.
1235 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1236 read_ahead_stats were updated. copied the stats section from
1237 osc_send_oap_rpc() into async_internal().
1241 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1245 Description: change order of libsysio includes
1246 Details : '#include sysio.h' should always come before '#include xtio.h'
1250 Description: handle "serious error: objid * already exists" more gracefully
1251 Details : If LAST_ID value on disk is smaller than the objects existing in
1252 the O/0/d* directories, it indicates disk corruption and causes an
1253 LBUG(). If the object is 0-length, then we should use the existing
1254 object. This will help to avoid a full fsck in most cases.
1256 Severity : enhancement
1258 Description: adapt the lustre_config script to support the upgrade case
1259 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1260 server targets to 1.6 in parallel.
1264 Description: To avoid grant space > avaible space when the disk is almost
1265 full. Without this patch you might see the error "grant XXXX >
1266 available" or some LBUG about grant, when the disk is almost
1268 Details : In filter_check_grant, for non_grant cache write, we should
1269 check the left space by if (*left > ungranted + bytes), instead
1270 of (*left > ungranted), because only we are sure the left space
1271 is enough for another "bytes", then the ungrant space should be
1272 increase. In client, we should update cl_avail_grant only there
1273 is OBD_MD_FLGRANT in the reply.
1277 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1278 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1279 described in CVE-2007-4573. This problem could allow untrusted
1280 local users to gain root access.
1283 Frequency : when using O_DIRECT and quotas
1285 Description: Incorrect file ownership on O_DIRECT output files
1286 Details : block usage reported by 'lfs quota' does not take into account
1287 files that have been written with O_DIRECT.
1292 Description: touch file failed when fs is not full
1293 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1294 otherwise we can get ENOSP while fs is not full.
1298 Description: parallel lock callbacks
1299 Details : Instead of sending blocking and completion callbacks as separated
1300 requests, adding them to a set and sending in parallel.
1303 Frequency : only for Cray XT3
1304 Bugzilla : 12829/13455
1305 Description: Changing primary group doesn't change the group lustre assigns to
1307 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1308 overridden with the primary group provided by the group upcall,
1309 whereas the supplied fsgid can be trusted if it is in the list of
1310 supplementary groups returned by the group upcall.
1312 Severity : enhancement
1314 Description: Allow masking D_WARNING, D_ERROR messages from console
1315 Details : Console messages can now be disabled via lnet.printk.
1319 Description: User code with malformed file open parameter crashes client node
1320 Details : Before packing join_file req, all the related reference should be
1321 checked carefully in case some malformed flags cause fake
1322 join_file req on client.
1326 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1327 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1328 mode is equal to the granted mode. However, it should also check
1329 that LDLM_FL_CP_REQD is not set.
1333 Description: Heavy nfs access might result in deadlocks
1334 Details : After ELC code landed, it is now improper to enqueue any mds
1335 locks under och_sem, because enqueue might want to decide to
1336 cancel open locks for same inode we are holding och_sem for.
1340 Description: Client eviction while running blogbench
1341 Details : A lot of unlink operations with concurrent I/O can lead to a
1342 deadlock causing evictions. To address the problem, the number of
1343 oustanding OST_DESTROY requests is now throttled to
1344 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1345 callbacks are processed in priority.
1349 Description: enable ACLs on MDS by default
1350 Details : ACLs must be enabled on MDS by default.
1353 Frequency : PPC/PPC64 only
1355 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1356 Details : fix duplicated definitions between asm-ppc64/types.h and
1357 lustre_types.h on PPC.
1360 Frequency : PPC/PPC64 only
1362 Description: asm-ppc/segment.h does not exist
1363 Details : fix compile issue on PPC.
1367 Description: better handle error messages in extents code
1370 Frequency : RHEL4 only
1372 Description: mkfs is very slow on IA64/RHEL4
1373 Details : A performance regression has been discovered in the MPT Fusion
1374 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1375 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1376 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1378 Severity : enhancement
1380 Description: SNMP support enhancement
1381 Details : Adding total number of sampled request for an MDS node in snmp
1384 Severity : enhancement
1386 Description: Optimize ldlm waiting list processing for PR extent locks
1387 Details : When processing waiting list for read extent lock and meeting read
1388 lock that is same or wider to it that is not contended, skip
1389 processing rest of the list and immediatelly return current
1390 status of conflictness, since we are guaranteed there are no
1391 conflicting locks in the rest of the list.
1395 Description: Time out and refuse to reconnect
1396 Details : When the failover node is the primary node, it is possible
1397 to have two identical connections in imp_conn_list. We must
1398 compare not conn's pointers but NIDs, otherwise we can defeat
1399 connection throttling.
1403 Description: port llog fixes from b1_6 into HEAD
1404 Details : Port llog reference couting and some llog cleanups from b1_6
1405 (bug 10800) into HEAD, for protect from panic and access to already
1406 free llog structures.
1410 Description: Detect stride IO mode in read-ahead
1411 Details : When a client does stride read, read-ahead should detect that and
1412 read-ahead pages according to the detected stride pattern.
1416 Description: data checksumming impacts single node performance
1417 Details : add support for several checksum algorithm. Currently, only CRC32
1418 and Adler-32 are supported. The checksum type can be changed on
1419 the fly via /proc/fs/lustre/osc/*/checksum_type.
1423 Description: use adler32 for page checksums
1424 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1429 Description: build for x2 fails
1430 Details : fix compile issue on Cray systems.
1434 Description: Properly match for duplicate locks
1435 Details : Due to different lock order from skiplists code, we need to
1436 traverse entire list for now
1439 Frequency : only on PPC/SLES10
1441 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1442 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1443 be defined. Add a hack in mkfs_lustre.c to work around this compile
1448 Description: LASSERT on MDS when client holding flock lock dies
1449 Details : ldlm pool logic depends on number of granted locks equal to
1450 number of released locks which is not true for flock locks, so
1451 just exclude such locks from consideration.
1455 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1456 Details : Use fsfilt_sync() for both the whole filesystem sync and
1457 individual file sync to eliminate dangerous inode locking
1458 with I_LOCK that can lead to a deadlock.
1462 Description: performance in 1.6.3
1463 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1464 sure that 1MB requests are not fragmented by the block layer.
1466 Severity : enhancement
1468 Description: organize the server-side client stats on per-nid basis
1469 Details : Change the structure of stats under obdfilter and mds to
1477 The "uuid"s file would list the uuids of _active_ exports.
1478 And the clear entry is to clear all stats and stale nids.
1480 Severity : enhancement
1482 Description: eliminate client locks in face of contention
1483 Details : file contention detection and lockless i/o implementation
1484 for contended files.
1488 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1491 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1493 Description: MDS may lose file striping (and hence file data) in some cases
1494 Details : If there are additional extended attributes stored on the MDS,
1495 in particular ACLs, SELinux, or user attributes (if user_xattr
1496 is specified for the client mount options) then there is a risk
1497 of attribute loss. Additionally, the Lustre file striping
1498 needs to be larger than default (e.g. striped over all OSTs),
1499 and an additional attribute must be stored initially in the
1500 inode and then increase in size enough to be moved to the
1501 external attribute block (e.g. ACL growing in size) for file
1506 Description: skiplist implementation simplification
1507 Details : skiplists are used to group compatible locks on granted list
1508 that was implemented as tracking first and last lock of each
1509 lock group the patch changes that to using doubly linked lists
1513 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1514 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1515 because LASSERT on both the data supplied by a client, and the
1516 data on disk is dangerous and incorrect.
1518 Severity : enhancement
1520 Description: Slow truncate/writes to huge files at high offsets.
1521 Details : Directly associate cached pages to lock that protect those pages,
1522 this allows us to quickly find what pages to write and remove
1523 once lock callback is received.
1527 Description: more ldlm soft lockups
1528 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1529 starve other threads from the resource lock for a long time in
1530 case of long waiting queue, so change the debug level from
1531 D_OTHER to the less frequently used D_INFO.
1533 Severity : enhancement
1535 Description: add -gid, -group, -uid, -user options to lfs find
1539 Description: Hung threads in invalidate_inode_pages2_range
1540 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1541 once one is completed. As a result, some RPCs are stuck in the
1542 queue and are never sent.
1546 Description: filter threads hungs on waiting journal commit
1547 Details : Cleanup filter group llog code, then only filter group llog will
1548 be only created in the MDS/OST syncing process.
1552 Description: Procfs and llog threads access destoryed import sometimes.
1553 Details : Sync the import destoryed process with procfs and llog threads by
1554 the import refcount and semaphore.
1556 Severity : enhancement
1558 Description: openlock cache of b1_6 port to HEAD
1563 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1564 Details : If insertion of an extent fails, then discard the inode
1565 preallocation and free data blocks else it can lead to duplicate
1570 Description: don't always update ctime in ext3_xattr_set_handle()
1571 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1572 In some cases the ctime should not be updated, for example for
1573 2.0->1.8 compatibility it is necessary to delete an xattr and it
1574 should not update the ctime.
1578 Bugzilla : 15713/16362
1579 Description: Assertion in iopen_connect_dentry in 1.6.3
1580 Details : looking up an inode via iopen with the wrong generation number can
1581 populate the dcache with a disconneced dentry while the inode
1582 number is in the process of being reallocated. This causes an
1583 assertion failure in iopen since the inode's dentry list contains
1584 both a connected and disconnected dentry.
1588 Description: assertion failure in ldlm_handle2lock()
1589 Details : fix a race between class_handle_unhash() and class_handle2object()
1590 introduced in lustre 1.6.5 by bug 13622.
1595 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1596 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1601 Description: oops in sd_iostats_seq_show()
1602 Details : unloading/reloading the scsi low level driver triggers a kernel
1603 bug when trying to access the sd iostat file.
1608 Description: Kernel panics during QLogic driver reload
1609 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1610 patch, causing memory corruption.
1615 Description: journal_dev option does not work in b1_6
1616 Details : pass mount option during pre-mount.
1618 Severity : enhancement
1620 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1621 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1622 extent information of a file. It can be used to map logical blocks
1623 in a file to physical blocks in the block device.
1627 Description: LDLM soft lockups - improvement
1628 Details : It is be possible to send the lock handle along with each read
1629 or write request because the client is already doing a lock match
1630 itself so there isn't any reason the OST should have to re-do that
1634 Frequency : only on Cray X2
1636 Description: X2 build failures
1637 Details : fix build failures on Cray X2.
1641 Description: xid & resent requests
1642 Details : Initialize RPC XID from clock at startup (randomly if clock is
1645 Severity : enhancement
1647 Description: Add lustre_start utility to start or stop multiple Lustre servers
1652 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1653 Details : In case of memory pressure, list_del() can be called twice on
1654 req->rq_history_list, causing a kernel oops.
1658 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1659 Details : fix stack overflow in the distributed lock manager by defering
1660 export eviction after a failed AST to the elt thread instead of
1661 handling it in the dlm interpret routine.
1665 Description: Convert some comments to new format.
1666 Details : Update documenting comments to match doxygen conventions.
1670 Description: Grammar fixes.
1671 Details : A couple of trivial spelling fixes.
1675 Description: OSD_COUNTERS-mandatory
1676 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1677 to simplify the code.
1681 Description: simplify cmm_device freeing logic.
1682 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1686 Description: Add lockdep support to dt_object_operations locking interface.
1687 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1688 parameter indicating lock ordering. Update mdd code to use new
1693 Description: Introduce failloc constants for lockless IO tests.
1694 Details : Add two new failloc constants to test lockless IO. Only one of
1695 them in implemented---another is checked in yet to be landed
1700 Description: Add lockdep support for inode mutex.
1701 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1702 in the situations where Lustre has to lock more than one inode
1707 Description: Add optional invariants checking support.
1708 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1709 configure switch. This macro is to be used for consistency and
1710 sanity checks that are too expensive to be left in `production'
1715 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1716 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1721 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1722 Details : Use spin_lock_nested() in (the only) situation where more than
1723 one ldlm_lock is locked simultaneously. Also, fix possible
1724 dead-lock in ldlm_lock_change_resource() by enforcing particular
1729 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1730 Details : Instead of specifying each ldlm_lock call-back through separate
1731 parameter, wrap them into struct ldlm_callback_suite.
1735 Description: Kill join_lru obd method and its callers.
1736 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1737 in memory---a requirement implemented through ->o_join_lru() obd
1738 method in HEAD. Remove this method and its users.
1742 Description: Add asynchronous ldlm ENQUEUE completion handler.
1743 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1744 case that stock ldlm_completion_ast() cannot handle as it waits
1745 until lock is granted. Introduce new ldlm_completion_ast_async()
1746 for this. Also comment ldlm_completion_ast().
1750 Description: ldlm_error <-> errno conversion.
1751 Details : Add functions to map (rather arbitrary) between LDLM error codes
1752 and standard errno values. CLIO needs this to prevent LDLM specific
1753 constants from escaping ldlm and osc.
1757 Description: Kill unused ldlm_handle2lock_ns() function.
1758 Details : Kill unused ldlm_handle2lock_ns() function.
1762 Description: Add lu_ref support to ldlm_lock
1763 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1764 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1765 and ldlm_resource. LDLM interface has to be changed, because code that
1766 releases a reference on a lock, has to "know" what reference this is.
1767 In the most frequent case
1769 lock = ldlm_handle2lock(handle);
1771 LDLM_LOCK_PUT(lock);
1773 no changes are required. When any other reference (received _not_ from
1774 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1775 instead of LDLM_LOCK_PUT().
1777 Arguably, changes are pervasive, and interface requires some discipline
1778 for proper use. On the other hand, it was very instrumental in finding
1779 a few leaked lock references.
1783 Description: Add ldlm_lock_addref_try().
1784 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1785 attempts to addref a lock that might be being canceled
1790 Description: Add ldlm_weigh_callback().
1791 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1792 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1793 hard-coded `number of pages' logic.
1797 Description: Add lockdep annotations to llog code.
1798 Details : Use appropriately tagged _nested() locking calls in the places
1799 where llog takes more than one ->lgh_lock lock.
1803 Description: Add loi_kms_set().
1804 Details : Wrap kms updates into a helper function.
1808 Description: Constify instances of struct lsm_operations.
1809 Details : Constify instances of struct lsm_operations.
1813 Description: lu_conf support.
1814 Details : On a server, a file system object is uniquely identified
1815 by a fid, which is sufficient to locate and load all object
1816 state (inode). On a client, on the other hand, more data are
1817 necessary instantiate an object. Change lu_object_find() and
1818 friends to take additional `lu_conf' argument describing object.
1819 Typically this includes layout information.
1823 Description: lu_context fixes.
1824 Details : Introduce new lu_context functions that are needed on the client
1825 side, where some system threads (ptlrpcd) are shared by multiple
1826 modules, and so cannot be stopped during module shutdown.
1830 Description: Add start and stop methods to lu_device_type_operations.
1831 Details : Introduce two new methods in lu_device_type_operations, that are
1832 invoked when first instance of a given type is created and last one
1833 is destroyed respectively. This is need by CLIO.
1837 Description: Add lu_ref support to struct lu_device.
1838 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1839 track leaked references.
1843 Description: Introduce lu_kmem_descr.
1844 Details : lu_kmem_descr and its companion interface allow to create
1845 and destroy a number of kmem caches at once.
1849 Description: Fix lu_object finalization race.
1850 Details : Fix a race between lu_object_find() finding an object and its
1851 concurrent finalization. This race is (most likely) not possible
1852 on the server, but might happen on the client.
1856 Description: Introduce lu_ref interface.
1857 Details : lu_ref is a debugging module allowing to track references to
1858 a given object. It is quite cpu expensive, and has to be
1859 explicitly enabled with --enable-lu_ref. See usage description
1864 Description: Factor lu_site procfs stats into a separate function.
1865 Details : Separate lu_site stats printing code into a separate function
1866 that can be reused on a client.
1870 Description: Constify instances of struct {lu,dt,md}_device_operations.
1871 Details : Constify instances of struct {lu,dt,md}_device_operations.
1875 Description: Introduce struct md_site and move meta-data specific parts of
1876 struct lu_site here.
1877 Details : Move md-specific fields out of struct lu_site into special struct
1878 md_site, so that lu_site can be used on a client.
1882 Description: Kill mdd_lov_destroy().
1883 Details : Remove unused mdd code.
1887 Description: Add st_block checking to multistat.c.
1888 Details : Add st_block checking to multistat.c.
1892 Description: Add lu_ref support to struct obd_device.
1893 Details : Add lu_ref tracking to obd_device.
1897 Description: Kill obd_set_fail_loc().
1898 Details : Remove unused code.
1902 Description: Add special type for ptlrpc_request interpret functions.
1903 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1904 there. Actual usage will be in CLIO.
1908 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1909 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1910 doesn't work with lockdep.
1914 Description: Add rwv.c test program.
1915 Details : New testing program exercising readv(2) and writev(2) (Qian).
1919 Description: Add sendfile.c test program.
1920 Details : New testing program exercising sendfile(2) (Jay).
1924 Description: Ratelimit a message that can be very frequent.
1925 Details : Ratelimit a memory allocation failure message that can
1930 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1931 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1935 Description: Liblustre build fixes.
1936 Details : Liblustre build fixes.
1940 Description: libcfs: add cfs_{need,cond}_resched() interface.
1941 Details : libcfs: add cfs_{need,cond}_resched() definition and
1942 implementations for Linux, NT, and liblustre.
1944 Severity : enhancement
1946 Description: More exported tunables for mballoc
1947 Details : Add support for tunable preallocation window and new tunables for
1948 large/small requests
1952 Description: Detect corruption of block bitmap and checking for preallocations
1953 Details : Checks validity of on-disk block bitmap. Also it does better
1954 checking of number of applied preallocations. When corruption is
1955 found, it turns filesystem readonly to prevent further corruptions.
1959 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1960 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1961 window to 0. If it does hit cache miss, and read-pattern is still
1962 stride-io mode, does not reset the stride window, but also does not
1963 increase the stride window length in this case.
1967 Frequency : only for big-endian servers
1968 Description: Check if system is big-endian while mounting fs with extents feature
1969 Details : Mounting a filesystem with extents feature will fail on big-endian
1970 systems since ext3-based ldiskfs is not supported on big-endian
1971 systems. This can be over-riden with "bigendian_extents" mount option.
1973 Severity : enhancement
1975 Description: The root squash functionality
1976 Details : A security feature, which is to prevent users from being able
1977 to mount lustre on their desktop, run as root, and delete
1978 all of the files in the filesystem. The goal is accomplished by
1979 remapping user id (UID) and group id (GID) of the root user to
1980 a UID and GID specified by the system administartor via Lustre
1981 configuration management server (MGS). The functionality also
1982 allows to specify sets of clients for which the remapping does
1987 Description: Excessive recovery window
1988 Details : With AT enabled, the recovery window can be excessively long (6000+
1989 seconds). To address this problem, we no longer use
1990 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1991 timeout no longer depends on the service time, it is set to
1992 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1993 time via pb_service_time.
1997 Description: Watchdog triggered on MDS failover
1998 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
1999 the OSTs know that the MDS "UUID" can be reused for the same export
2000 from a different NID, so we do not need to wait for the export to be
2004 Frequency : rare, only if using MMP with Linux RAID
2006 Description: MMP doesn't work with Linux RAID
2007 Details : While using HA for Lustre servers with Linux RAID, it is possible
2008 that MMP will not detect multiple mounts. To make this work we
2009 need to unplug the device queue in RAID when the MMP block is being
2010 written. Also while reading the MMP block, we should read it from
2011 disk and not the cached one.
2013 Severity : enhancement
2015 Description: open file using fid
2016 Details : A file can be opened using just its fid, like
2017 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2019 --------------------------------------------------------------------------------
2021 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2023 * Support for kernels:
2024 2.6.5-7.283 (SLES 9),
2025 2.6.9-55.EL (RHEL 4),
2026 2.6.16.46-0.14 (SLES 10),
2027 2.6.18.8 vanilla (kernel.org)
2028 * Client support for unpatched kernels:
2029 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2030 2.6.16 - 2.6.22 vanilla (kernel.org)
2031 * Due to recently discovered recovery problems, we do not recommend
2032 using patchless RHEL 4 clients with this or any earlier release.
2033 * Recommended e2fsprogs version: 1.39.cfs8
2034 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2035 * Starting with this release, the ldiskfs backing filesystem required
2036 by Lustre is now in its own package, lustre-ldiskfs. This package
2037 should be installed. It is versioned separately from Lustre and
2038 may be released separately in future.
2042 Description: block reactivating mgc import until all deactivates complete
2043 Details : Fix race when failing back MDT/MGS to itself (testing)
2046 Frequency : at statup only
2048 Description: mds_lov_synchronize race leads to various problems
2049 Details : simultaneous MDT->OST connections at startup can cause the
2050 sync to abort, leaving the OSC in a bad state.
2052 Severity : enhancement
2054 Description: add optional extra BUILD_VERSION info
2055 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2056 to override the lustre version.
2059 Frequency : 2.6.18 servers only
2061 Description: ll_kern_mount() doesn't release the module reference
2062 Details : The ldiskfs module reference count never drops down to 0
2063 because ll_kern_mount() doesn't release the module reference.
2068 Description: server LBUG when using old ost_num_threads parameter
2069 Details : Accept the old ost_num_threads parameter but warn that it
2070 is deprecated, and fix an off-by-one error that caused an LBUG.
2075 Description: Transient SCSI error results in persistent IO issue
2076 Details : iobuf->dr_error is not reinitialized to 0 between two
2080 Frequency : sometimes when underlying device returns I/O errors
2082 Description: OSTs not going read-only during write failures
2083 Details : OSTs are not remounted read-only when the journal commit threads
2084 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2085 instead of the ext3 wrappers.
2089 Description: poor connect scaling with increasing client count
2090 Details : Don't run filter_grant_sanity_check for more than 100 exports
2091 to improve scaling for large numbers of clients.
2094 Frequency : SLES10 only
2096 Description: sanity-quota.sh quotacheck failed: rc = -22
2097 Details : Quotas cannot be enabled on SLES10.
2100 Frequency : liblustre clients only
2102 Description: getdirentries does not give error when run on compute nodes
2103 Details : getdirentries does not fail when the size specified as an argument
2104 is too small to contain at least one entry
2106 Severity : enhancement
2108 Description: Add LNET router traceability for debug purposes
2109 Details : If a checksum failure occurs with a router as part of the
2110 IO path, the NID of the last router that forwarded the bulk data
2111 is printed so it can be identified.
2116 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2117 Details : Due to a race condition, liblustre clients were occasionally
2118 evicted incorrectly.
2120 Severity : enhancement
2122 Description: lfs setstripe use optional parameters instead of postional
2125 Severity : enhancement
2127 Description: Nanosecond timestamp support for ldiskfs
2128 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2129 resolution timestamps. There is not yet support for this at
2130 the Lustre filesystem level.
2133 Frequency : during server recovery
2135 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2136 Details : request with rq_no_resend flag not awake l_wait_event if they get
2140 Frequency : nfs export on patchless client
2142 Description: connectathon hang when test nfs export over patchless client
2143 Details : Disconnected dentry cannot be found with lookup, so we do not need
2144 to unhash it or make it invalid
2147 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2148 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2149 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2154 Description: open req refcounting wrong on reconnect
2155 Details : If reconnect happened between getting open reply from server and
2156 call to mdc_set_replay_data in ll_file_open, we will schedule
2157 replay for unreferenced request that we are about to free.
2158 Subsequent close will crash in variety of ways.
2159 Check that request is still eligible for replay in
2160 mdc_set_replay_data().
2165 Description: disable writes to filesystem when reading health_check file
2166 Details : the default for reading the health_check proc file has changed
2167 to NOT do a journal transaction and write to disk, because this
2168 can cause reads of the /proc file to hang and block HA state
2169 checking on a healthy but otherwise heavily loaded system. It
2170 is possible to return to the previous behaviour during configure
2171 with --enable-health-write.
2173 Severity : enhancement
2175 Description: 64-bit inode version
2176 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2177 to the inode. This will be required for version-based recovery.
2182 Description: MDS fails to start if a duplicate client export is detected
2183 Details : in some rare cases it was possible for a client to connect to
2184 an MDS multiple times. Upon recovery the MDS would detect this
2185 and fail during startup. Handle this more gracefully.
2187 Severity : enhancement
2189 Description: Add -o localflock option to simulate old noflock
2191 Details : This will achieve local-only flock/fcntl locks
2197 Description: log_commit_thread vs filter_destroy race leads to crash
2198 Details : Take import reference before releasing llog record semaphore
2203 Description: Wrong request locking in request set processing
2204 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2205 imp_delayed_list, in this place should be used imp_lock.
2208 Frequency : when reconnection
2210 Description: Grant Leak when osc reconnect to OST
2211 Details : When osc reconnect ost, OST(filter) should check whether it
2212 should grant more space to client by comparing fed_grant and
2213 cl_avail_grant, and return the granted space to client instead
2214 of "new granted" space, because client will call osc_init_grant
2215 to update the client grant space info.
2218 Frequency : when client reconnect to OST
2220 Description: Grant Leak when osc do resend and replay bulk write
2221 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2222 bulk write request, because the grant info while be sync between
2223 OSC and OST when reconnect, and we should ignore the grant info
2224 these of resend/replay write req.
2229 Description: Grant space more than avaiable left space sometimes.
2230 Details : When then OST is about to be full, if two bulk writing from
2231 different clients came to OST. Accord the avaliable space of the
2232 OST, the first req should be permitted, and the second one
2233 should be denied by ENOSPC. But if the seconde arrived before
2234 the first one is commited. The OST might wrongly permit second
2235 writing, which will cause grant space > avaiable space.
2238 Frequency : when client is evicted
2240 Description: Grant might be wrongly erased when osc is evicted by OST
2241 Details : when the import is evicted by server, it will fork another
2242 thread ptlrpc_invalidate_import_thread to invalidate the
2243 import, where the grant will be set to 0. While the original
2244 thread will update the grant it got when connecting. So if
2245 the former happened latter, the grant will be wrongly errased
2246 because of this race.
2251 Description: Checking Stale with correct fid
2252 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2253 to check whether it is stale, because sometimes, we want the
2254 enqueue happened anyway, and op_data.fid2 will not be initialized.
2256 Severity : enhancement
2258 Description: update patchless client
2259 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2262 Frequency : only with 2.4 kernel
2264 Description: random memory corruption
2265 Details : size of struct ll_inode_info is to big for union inode.u and this
2266 can be cause of random memory corruption.
2271 Description: Memory leak in recovery
2272 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2273 It should also check obd_fail before fsfilt_start, otherwise if
2274 fsfilt_start return -EROFS,(failover mds during mds recovery).
2275 then the req will return with repmsg->transno = 0 and rc = EROFS.
2276 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2277 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2278 be freed no matter whether fsfilt_commit success or not.
2281 Frequency : only with huge count clients
2283 Description: Prevents from taking the superblock lock in llap_from_page for
2285 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2286 for llap_from_page call in ll_removepage prevents from taking the
2287 superblock lock for a soon died page.
2292 Description: Not check open intent error before release open handle
2293 Details : in some rare cases, the open intent error is not checked before
2294 release open handle, which may cause
2295 ASSERTION(open_req->rq_transno != 0), because it tries to release
2296 the failed open handle.
2301 Description: Set cat log bitmap only after create log success.
2302 Details : in some rare cases, the cat log bitmap is set too early. and it
2303 should be set only after create log success.
2307 Description: Accessing a block bevice can re-enable I/O when Lustre is
2308 tearing down a device.
2309 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2314 Description: service threads may hog cpus when there are a lot of requests
2315 Details : Insert cond_resched to give other threads a chance to use some CPU
2320 Description: the cat log was not initialized in recovery
2321 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2322 unlink log on mds will not be initialized until mds post
2323 recovery. And also in mds post recovery, the unlink log will
2324 initialization will be done asynchronausly, so there will be race
2325 between add unlink log and unlink log initialization.
2329 Description: brw_stats were being printed incorrectly
2330 Details : brw_stats were being printed as log2 but all of them were not
2331 recorded as log2. Also remove some code duplication arising from
2332 filter_tally_{read,write}.
2336 Frequency : rare, only in recovery.
2337 Description: ASSERTION(req->rq_type != LI_POISON) failed
2338 Details : imp_lock should be held while iterating over imp_sending_list for
2339 prevent destroy request after get timeout in ptlrpc_queue_wait.
2343 Description: replay-single.sh test 52 fails
2344 Details : A lock's skiplist need to be cleanup when it being unlinked
2345 from its resource list.
2349 Description: Short directio read returns full requested size rather than
2351 Details : Direct I/O operations should return actual amount of bytes
2352 transferred rather than requested size.
2354 Severity : enhancement
2356 Description: metadata RPC reduction (e.g. for rm performance)
2357 Details : decrease the amount of synchronous RPC between clients and servers
2358 by canceling conflicing lock before the operation on the client
2359 and packing thier handles into the main operation RPC to server.
2361 Severity : enhancement
2363 Description: Async OSC create to avoid the blocking unnecessarily.
2364 Details : If a OST has no remain object, system will block on the creating
2365 when need to create a new object on this OST. Now, ways use
2366 pre-created objects when available, instead of blocking on an
2367 empty osc while others are not empty. If we must block, we block
2368 for the shortest possible period of time.
2372 Description: improve handling recoverable errors
2373 Details : if request processig with error which can be recoverable on server
2374 request should be resend, otherwise page released from cache and
2377 Severity : enhancement
2379 Description: refine locking for avoid write wrong info into lov_objid file
2380 Details : fix possible races with add new target and write/update data in
2383 --------------------------------------------------------------------------------
2385 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2390 Frequency : on some architectures
2392 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2393 Details : Uninitialized flags sometimes cause configuration commands to
2397 Frequency : patchless clients only
2399 Description: missing __iget() symbol export
2400 Details : The __iget() symbol export is missing. To avoid the need for
2401 this on patchless clients the deathrow inode reaper is turned
2402 off, and we depend on the VM to clean up old inodes. This
2403 dependency was during via the fix for bug 12181.
2405 --------------------------------------------------------------------------------
2407 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2409 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2410 INTEROPERATE with older versions automatically. In many cases a
2411 special upgrade step is needed. Please read the
2412 user documentation before upgrading any part of a live system.
2413 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2414 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2415 * WARNING: Lustre configuration and startup changes are required with
2416 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2420 Severity : enhancement
2422 Description: MountConf
2423 Details : Lustre configuration is now managed via mkfs and mount
2424 commands instead of lmc and lconf. New obd types (MGS, MGC)
2425 are added for dynamic configuration management. See
2426 https://mail.clusterfs.com/wikis/lustre/MountConf for
2429 Severity : enhancement
2431 Description: dynamic OST addition
2432 Details : OSTs can now be added to a live filesystem
2434 Severity : enhancement
2436 Description: startup order invariance
2437 Details : MDTs and OSTs can be started in any order. Clients only
2438 require the MDT to complete startup.
2440 Severity : enhancement
2442 Description: parallel, asynchronous orphan cleanup
2443 Details : orphan cleanup is now performed in separate threads for each
2444 OST, allowing parallel non-blocking operation.
2446 Severity : enhancement
2448 Description: optimized stripe assignment
2449 Details : stripe assignments are now made based on ost space available,
2450 ost previous usage, and OSS previous usage, in order to try
2451 to optimize storage space and networking resources.
2453 Severity : enhancement
2455 Description: Permanently set tunables
2456 Details : All writable /proc/fs/lustre tunables can now be permanently
2457 set on a per-server basis, at mkfs time or on a live system.
2459 Severity : enhancement
2461 Description: Lustre message v2
2462 Details : Add lustre message format v2.
2464 Severity : enhancement
2466 Description: client OST exclusion list
2467 Details : Clients can be started with a list of OSTs that should be
2468 declared "inactive" for known non-responsive OSTs.
2472 Description: SPEC SFS validation failure on NFS v2 over lustre.
2473 Details : Changes the blocksize for regular files to be 2x RPC size,
2474 and not depend on stripe size.
2476 Severity : enhancement
2478 Description: Multiple MD RPCs in flight.
2479 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2480 To avoid overly-overloading MDS, introduce a limit on number of
2481 MDS RPCs in flight for a single client and add /proc controls
2482 to adjust this limit.
2484 Severity : enhancement
2486 Description: client read/write statistics
2487 Details : Add client read/write call usage stats for performance
2488 analysis of user processes.
2489 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2490 file access. extents_stats shows chunk size distribution.
2491 extents_stats_per_process show chunk size distribution per
2494 Severity : enhancement
2496 Description: mds statistics
2497 Details : Add detailed mds operations statistics in
2498 /proc/fs/lustre/mds/*/stats.
2502 Description: Failure of copying files with lustre special EAs.
2503 Details : Client side always return success for setxattr call for lustre
2504 special xattr (currently only "trusted.lov").
2509 Description: Refcount LNET uuids
2510 Details : The global LNET uuid list grew linearly with every startup;
2511 refcount repeated list entries instead of always adding to
2514 Severity : enhancement
2516 Description: Dynamic service threads
2517 Details : Within a small range, start extra service threads
2518 automatically when the request queue builds up.
2521 Frequency : mixed-endian client/server environments
2523 Description: mixed-endian crashes
2524 Details : The new msg_v2 system had some failures in mixed-endian
2527 Severity : enhancement
2529 Description: Easy OST removal
2530 Details : OSTs can be permanently deactivated with e.g. 'lctl
2531 conf_param lustre-OST0001.osc.active=0'
2533 Severity : enhancement
2535 Description: MGS proc entries
2536 Details : Added basic proc entries for the MGS showing what filesystems
2539 Severity : enhancement
2541 Description: provide MGS failover
2542 Details : Added config lock reacquisition after MGS server failover.
2544 Severity : enhancement
2546 Description: add Linux 2.4 support
2547 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2551 Description: a large application tries to do I/O to the same resource and dies
2552 in the middle of it.
2553 Details : Check the req->rq_arrival time after the call to
2554 ost_brw_lock_get(), but before we do anything about
2555 processing it & sending the BULK transfer request. This
2556 should help move old stale pending locks off the queue as
2557 quickly as obd_timeout.
2560 Frequency : when an incorrect nid is specified during startup
2562 Description: ptlrpc connect to non-existant node causes kernel crash
2563 Details : LNET can't be re-entered from an event callback, which
2564 happened when we expire a message after the export has been
2565 cleaned up. Instead, hand the zombie cleanup off to another
2568 Severity : enhancement
2570 Description: plain/inodebits lock performance improvement
2571 Details : Grouping plain/inodebits in granted list by their request modes
2572 and bits policy, thus improving the performance of search through
2576 Frequency : only if OST filesystem is corrupted
2578 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2579 Details : for a short time RPCs with bulk IO are in the replay list,
2580 but replay of bulk IOs is unimplemented. If the OST filesystem
2581 is corrupted due to disk cache incoherency and then replay is
2582 started it is possible to trip an assertion. Avoid putting
2583 committed RPCs into the replay list at all to avoid this issue.
2586 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2588 Description: System hang on startup
2589 Details : This bug allowed the liblustre (e.g. catamount) client to
2590 return to the app before handling all startup RPCs. This
2591 could leave the node unresponsive to lustre network traffic
2592 and manifested as a server ptllnd timeout.
2594 Severity : enhancement
2596 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2597 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2598 boolean to control whether to print peer debug info when a
2599 client's RPC times out.
2602 Frequency : only for kernels with patches from Lustre below 1.4.3
2604 Description: Remove old rdonly API
2605 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2608 Frequency : only for devices with external journals
2610 Description: Set external device read-only also
2611 Details : During a commanded failover stop, we set the disk device
2612 read-only while the server shuts down. We now also set any
2613 external journal device read-only at the same time.
2616 Frequency : when upgrading from 1.4 while trying to change parameters
2618 Description: The wrong (new) MDC name was used when setting parameters for
2619 upgraded MDT's. Also allows changing of OSC (and MDC)
2620 parameters if --writeconf is specified at tunefs upgrade time.
2623 Frequency : when setting specific ost indicies
2625 Description: QOS code breaks on skipped indicies
2626 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2627 created with --index need not be sequential.
2629 Severity : enhancement
2631 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2632 Details : The uninit_groups feature works in conjunction with the kernel
2633 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2634 up the pass1 processing of e2fsck. This is a read-only feature
2635 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2636 mount filesystems that have had this feature enabled.
2638 Severity : enhancement
2640 Description: Improve multi-block allocation algorithm to avoid fragmentation
2641 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2642 allocation locality and avoid filesystem fragmentation.
2644 ------------------------------------------------------------------------------
2646 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2648 * Support for kernels:
2649 2.4.21-47.0.1.EL (RHEL 3)
2650 2.6.5-7.283 (SLES 9)
2651 2.6.9-42.0.10.EL (RHEL 4)
2652 2.6.12.6 vanilla (kernel.org)
2653 2.6.16.27-0.9 (SLES 10)
2654 * Recommended e2fsprogs version: 1.39.cfs5
2656 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2660 Frequency : occasional, depends on client load and configuration
2661 Bugzilla : 12181, 12203
2662 Description: data loss for recently-modified files
2664 Details : In some cases it is possible that recently written or created
2665 files may not be written to disk in a timely manner (this should
2666 normally be within 30s unless client IO load is very high).
2667 The problem appears as zero-length files or files that are a
2668 multiple of 1MB in size after a client crash or client eviction
2669 that are missing data at the end of the file.
2671 This problem is more likely to be hit on clients where files are
2672 repeatedly created and unlinked in the same directory, clients
2673 have a large amount of RAM, have many CPUs, the filesystem has
2674 many OSTs, the clients are rebooted frequently, and/or the files
2675 are not accessed by other nodes after being written.
2677 The presence of the problem can be detected by looking at
2678 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2679 smaller than the second (nr_unused) then dirty files will not
2680 be flushed automatically to disk. "sync; sleep 10" should be
2681 run several times on the node before unmounting it to update
2682 Lustre (this is also safe to run on nodes without this problem).
2684 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2685 that can cause this same problem, so customers using that kernel
2686 also need to update the kernel in addition to Lustre. In order
2687 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2689 It is normal that files written just before a client crash (less
2690 than 30s) may not yet have been flushed to disk, even for local
2694 Frequency : frequent on thin XT3 nodes
2696 Description: UUID collision on thin XT3 Linux nodes
2697 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2698 randomly, since we relied on an insufficiently-seeded PRNG.
2703 Description: OSS hangs after "All ost request buffers busy"
2704 Details : A deadlock between quota and journal operations caused OSS
2705 hangs after printing "All ost request buffers busy."
2708 Frequency : always on liblustre builds
2710 Description: Cleanup compiler warnings on liblustre
2713 Frequency : always on liblustre builds on XT3
2715 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2719 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2720 > 32000 subdirectories
2721 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2722 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2723 subdirectory count crosses 32000. This will aid e2fsck to
2724 correctly handle more than 32000 subdirectories.
2727 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2729 Description: System hang on startup
2730 Details : This bug allowed the liblustre (e.g. catamount) client to
2731 return to the app before handling all startup RPCs. This
2732 could leave the node unresponsive to lustre network traffic
2733 and manifested as a server ptllnd timeout.
2735 Severity : enhancement
2737 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2738 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2739 boolean to control whether to print peer debug info when a
2740 client's RPC times out.
2745 Description: make O_SYNC working on 2.6 kernels
2746 Details : 2.6 kernels use different method for mark pages for write,
2747 so need add a code to lustre for O_SYNC work.
2752 Description: Failure to close file and release space on NFS
2753 Details : Put inode details into lock acquired in ll_intent_file_open.
2754 Use mdc_intent_lock in ll_intent_open to properly
2755 detect all kind of errors unhandled by mdc_enqueue.
2760 Description: proc file read during shutdown sometimes raced obd removal,
2762 Details : Add lock to prevent obd access after proc file removal.
2765 Frequency : Only for files larger than 4GB on 32-bit clients.
2767 Description: improperly doing page alignment of locks
2768 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2769 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2774 Description: RPCs being resent when they shouldn't be
2775 Details : Some RPCs that should not be resent are being resent. This
2776 can cause inconsistencies in the RPC state machine. Do not
2777 resend such requests.
2780 Frequency : rare, only with NFS export
2782 Description: Crash on NFS re-export node
2783 Details : under very unusual load conditions an assertion is hit in
2784 ll_intent_file_open()
2787 Frequency : only if OST filesystem is corrupted
2789 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2790 Details : for a short time RPCs with bulk IO are in the replay list,
2791 but replay of bulk IOs is unimplemented. If the OST filesystem
2792 is corrupted due to disk cache incoherency and then replay is
2793 started it is possible to trip an assertion. Avoid putting
2794 committed RPCs into the replay list at all to avoid this issue.
2799 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2800 Details : Large single O_DIRECT read and write calls can fail to allocate
2801 a sufficiently large buffer to process the request. In case of
2802 allocation failure the allocation is retried with a smaller
2803 buffer and broken into smaller requests.
2805 Severity : enhancement
2807 Description: Add -o localflock option to simulate old noflock behaviour.
2808 Details : This will achieve local-only flock/fcntl locks coherentness.
2813 Description: versioning check is incomplete
2814 Details : Checking the version difference of client vs. server, report
2815 error if the gap is too big.
2820 Description: add support PG_writeback bit
2821 Details : add support for PG_writeback bit for Lustre, for more carefull
2822 work with page cache in 2.6 kernel. This also fix some deadlocks
2823 and remove hack for work O_SYNC with 2.6 kernel.
2825 Severity : enhancement
2827 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2828 Details : The uninit_groups feature works in conjunction with the kernel
2829 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2830 up the pass1 processing of e2fsck. This is a read-only feature
2831 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2832 mount filesystems that have had this feature enabled.
2834 Severity : enhancement
2836 Description: Improve multi-block allocation algorithm to avoid fragmentation
2837 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2838 allocation locality and avoid filesystem fragmentation.
2840 ------------------------------------------------------------------------------
2842 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2844 * Support for kernels:
2845 2.6.9-42.0.3.EL (RHEL 4)
2846 2.6.5-7.276 (SLES 9)
2847 2.4.21-47.0.1.EL (RHEL 3)
2848 2.6.12.6 vanilla (kernel.org)
2849 2.6.16.21-0.8 (SLES10)
2850 * Recommended e2fsprogs version: 1.39.cfs2-0
2852 * The backwards-compatible /proc/sys/portals symlink has been removed
2853 in this release. Before upgrading, please ensure that you change
2854 any configuration scripts or /etc/sysctl.conf files that access
2855 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2856 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2857 in advance of the upgrade on any system running Lustre 1.4.6 or
2858 newer, since /proc/sys/lnet was added in that version.
2859 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2863 Frequency : only when quota is used
2865 Description: avoid scanning export list for quota master
2866 Details : Change the algorithms to avoid scanning export list in order
2867 to improve the efficiency.
2870 Frequency : MDS failover only, very rarely
2872 Description: "went back in time" messages on mds failover
2873 Details : The greatest transno may be lost when the current operation
2874 finishes with an error (transno==0) and the client's last_rcvd
2875 record is over-written. Save the greatest transno in the
2876 mds_last_transno for this case.
2879 Frequency : always for specific kernels and striping counts
2881 Description: client may get "Matching packet too big" without ACL support
2882 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2883 when trying to access files in certain configurations. The
2884 clients should in fact be denied when mounting because they do
2885 not understand ACLs.
2888 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2890 Description: many clients connecting with IO in progress causes connect timeouts
2891 Details : Avoid synchronous journal commits to avoid delays caused by many
2892 clients connecting/disconnecting when bulk IO is in progress.
2893 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2894 OST_IO_PORTAL to avoid delays behind potentially many pending
2898 Frequency : occasionally with multiple writers to a single file
2900 Description: shared writes to file may result in wrong size reported by stat()
2901 Details : Allow growing of kms when extent lock is cancelled
2904 Frequency : always with random mmap IO to multi-striped file
2906 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2907 Details : Only if the hole is at the end of OST object so that kms is too
2908 small. Fix is to increase kms accordingly in ll_nopage.
2911 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2913 Description: writes to a missing object would leak memory on the OST
2914 Details : If there is an inconsistency between the MDS and OST filesystems,
2915 such that the MDS references an object that doesn't exist, writes
2916 to that object will leak memory due to incorrect cleanup in the
2917 error handling path, eventually running out of memory on the OST.
2922 Description: Creating too long symlink causes lustre errors
2923 Details : Check symlink and name lengths before sending requests to MDS.
2926 Frequency : only if flock is enabled (not on by default)
2928 Description: posix locks not released on fd closure on 2.6.9+
2929 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2930 kernels, this caused such locks not to be released on fd close and
2931 then assertions on fs unmount about still used locks.
2934 Frequency : MDS failover only, very rarely
2936 Description: clients may get ASSERTION(granted_lock != NULL)
2937 Details : When request was taking a long time, and a client was resending
2938 a getattr by name lock request. The were multiple lock requests
2939 with the same client lock handle and
2940 mds_getattr_name->fixup_handle_for_resent_request found one of the
2941 lock handles but later failed with ASSERTION(granted_lock != NULL).
2946 Description: handle->h_buffer_credits > 0, assertion failure
2947 Details : h_buffer_credits is zero after truncate, causing assertion
2948 failure. This patch extends the transaction or creates a new
2952 Frequency : NFS re-export or patchless client
2953 Bugzilla : 11179, 10796
2954 Description: Crash on NFS re-export node (__d_move)
2955 Details : We do not want to hash the dentry if we don't have a lock.
2956 But if this dentry is later used in d_move, we'd hit uninitialised
2957 list head d_hash, so we just do this to init d_hash field but
2958 leave dentry unhashed.
2961 Frequency : NFS re-export or patchless client
2963 Description: NFS exports has problem with symbolic link
2964 Details : lustre client didn't properly install dentry when re-exported
2965 to NFS or running patchless client.
2968 Frequency : NFS re-export or patchless client
2970 Description: Various nfs/patchless fixes.
2971 Details : fixes reuse disconected alias for lookup process - this fixes
2972 warning "find_exported_dentry: npd != pd",
2973 fix permission error with open files at nfs.
2974 fix apply umask when do revalidate.
2977 Frequency : occasional
2979 Description: Crash on NFS re-export node
2980 Details : calling clear_page() on the wrong pointer triggered oops in
2981 generic_mapping_read().
2984 Frequency : rarely, using O_DIRECT IO
2986 Description: unaligned directio crashes client with LASSERT
2987 Details : check for unaligned buffers before trying any requests.
2990 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2992 Description: stale data returned from RAID cache
2993 Details : If only a small amount of IO is done to the RAID device before
2994 reading it again it is possible to get stale data from the RAID
2995 cache instead of reading it from disk.
2998 Frequency : always for sles10 kernel
3000 Description: sles10 support
3001 Details : ll_follow_link: compile fixes and using of nd_set_link
3002 under newer kernels.
3005 Frequency : depends on arch, kernel and compiler version, always on sles10
3008 Description: recursive or deep enough symlinks cause stack overflow
3009 Details : getting rid of large stack-allocated variable in
3013 Frequency : depends on hardware
3015 Description: lustre write performance loss in the SLES10 kernel
3016 Details : the performance loss is caused by using of write barriers in the
3017 ext3 code. The SLES10 kernel turns barrier support on by
3018 default. The fix is to undo that change for ldiskfs.
3020 ------------------------------------------------------------------------------
3022 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3024 * Support for kernels:
3025 2.6.9-42.0.3EL (RHEL 4)
3026 2.6.5-7.276 (SLES 9)
3027 2.4.21-47.0.1.EL (RHEL 3)
3028 2.6.12.6 vanilla (kernel.org)
3032 Frequency : quota enabled and large files being deleted
3034 Description: releasing more than 4GB of quota at once hangs OST
3035 Details : If a user deletes more than 4GB of files on a single OST it
3036 will cause the OST to spin in an infinite loop. Release
3037 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3042 Description: statfs data retrieved from /proc may be stale or zero
3043 Details : When reading per-device statfs data from /proc, in the
3044 {kbytes,files}_{total,free,avail} files, it may appear
3045 as zero or be out of date.
3048 Frequency : systems with MD RAID1 external journal devices
3050 Description: lconf's call to blkid is confused by RAID1 journal devices
3051 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3052 of returning all block devices that match the journal UUID.
3055 Frequency : always, for aggregate stripe size over 4GB
3057 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3058 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3059 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3062 Frequency : always if "lfs find" used on a local file/directory
3064 Description: "lfs find" segfaults if used on a local file/directory
3065 Details : The case where a directory component was not specified wasn't
3066 handled correctly. Handle this properly.
3069 Frequency : always on ppc64
3071 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3072 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3073 perform endianness conversion.
3076 Frequency : rarely (truncate to non-zero file size after write under load)
3077 Bugzilla : 10730, 10687
3078 Description: Files padded with zeros to next 4K multiple
3079 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3080 it is possible that files that are truncated to a non-zero size
3081 immediately after being written are filled with zero bytes beyond
3082 the truncated size. No file data is lost.
3084 Severity : enhancement
3086 Description: Allow recovery/failover for liblustre clients.
3087 Details : liblustre clients were unaware of failover configurations until
3090 Severity : enhancement
3092 Description: user file locks should fail when not mounting with flock option
3093 Details : Set up an error-returning stub in ll_file_operations.lock field
3094 to prevent incorrect behaviour when client is mounted without
3095 flock option. Also, set up properly f_op->flock field for
3099 Frequency : always on ia64
3101 Description: "lfs df" loops on printing out MDS statfs information
3102 Details : The obd_ioctl_data was not initialized and in some systems
3103 this caused a failure during the ioctl that did not return
3104 an error. Initialize the struct and return an error on failure.
3107 Frequency : SLES 9 only
3109 Description: Error of copying files with lustre special EAs as root
3110 Details : Client side always return success for setxattr call for lustre
3111 special xattr (currently only "trusted.lov").
3114 Frequency : rarely on clusters with both ia64+i386 clients
3116 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3117 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3118 clients are doing concurrent writes to the same file it is
3119 possible that the ia64 clients may LASSERT because the OST
3120 extent locks are not PAGE_SIZE aligned. Ensure that grown
3121 locks are always aligned on the request boundary.
3124 Frequency : specific use, occasional
3126 Description: Overwriting in use executable truncates on-disk binary image
3127 Details : If one node attempts to overwrite an executable in use by
3128 another node, we now correctly return ETXTBSY instead of
3129 truncating the file.
3131 Severity : enhancement
3133 Description: Async OSC create to avoid the blocking unnecessarily.
3134 Details : If a OST has no remain object, system will block on the creating
3135 when need to create a new object on this OST. Now, ways use
3136 pre-created objects when available, instead of blocking on an
3137 empty osc while others are not empty. If we must block, we block
3138 for the shortest possible period of time.
3143 Description: chmod on Lustre root is propagated to other clients
3144 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3145 invalid by the follow_mount time.
3148 Frequency : liblustre clients only
3150 Description: Race in 'instant cancel' lock handling could lead to such locks
3151 never to be granted in case of SMP MDS
3152 Details : Do not destroy not yet granted but cbpending locks in
3156 Frequency : replay/resend of open
3158 Description: non null lock assetion failure in mds_intent_policy
3159 Details : Trying to replay/resend lockless open requests resulted in
3160 mds_open() returning 0 with no lock. Now it sets a flag if
3161 a lock is going to be returned.
3163 Severity : enhancement
3165 Description: Checksum enhancements
3166 Details : New checksum enhancements allow for resending RPCs that failed
3169 Severity : enhancement
3171 Description: Tunables on number of dirty pages in cacche
3172 Details : Allow to set limit on number of dirty pages cached.
3177 Description: client crash on unmount - lock still has references
3178 Details : In some error handling cases it was possible to leak a lock
3179 reference on a client while accessing a file. This was not
3180 harmful to the client during operation, but would cause the
3181 client to crash when the filesystem is unmounted.
3184 Frequency : specific case, rare
3186 Description: ETXTBSY on mds though file not in use by client
3187 Details : ETXTBSY is no longer incorrectly returned when attempting to
3188 chmod or chown a directory that the user previously tried to
3189 execute or a currently-executing binary.
3192 Frequency : extremely rare except on liblustre-based clients
3194 Description: Lustre space not freed when files are deleted
3195 Details : Clean up open-unlinked files after client eviction. Previously
3196 the unlink was skipped and the files remained as orphans.
3201 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3202 Details : In some cases it was possible to send a blocking callback to
3203 a client doing a glimpse, even though that client didn't get
3204 a lock granted. When the glimpse lock is cancelled on the OST
3205 the freed lock is left on the waiting list and corrupted the list.
3208 Frequency : all core dumps
3210 Description: Broke core dumps to lustre
3211 Details : Negative dentry may be unhashed if parent does not have UPDATE
3212 lock, but some callers, e.g. do_coredump, expect dentry to be
3213 hashed after successful create, hash it in ll_create_it.
3215 ------------------------------------------------------------------------------
3217 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3219 * Support for kernels:
3220 2.6.9-42.0.2.EL (RHEL 4)
3221 2.6.5-7.276 (SLES 9)
3222 2.4.21-40.EL (RHEL 3)
3223 2.6.12.6 vanilla (kernel.org)
3227 Frequency : always on RHEL 3
3229 Description: Number of open files grows over time
3230 Details : The number of open files grows over time, whether or not
3231 Lustre is started. This was due to a filp leak introduced
3232 by one of our kernel patches.
3234 ------------------------------------------------------------------------------
3236 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3238 * Support for kernels:
3239 2.6.9-42.EL (RHEL 4)
3240 2.6.5-7.276 (SLES 9)
3241 2.4.21-40.EL (RHEL 3)
3242 2.6.12.6 vanilla (kernel.org)
3247 Bugzilla : 5719, 9635, 9792, 9684
3248 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3249 Details : If a server is under heavy load and cannot reply to new
3250 connection requests before the client resends the (re)connect,
3251 the connection handling code can behave badly if two service
3252 threads are concurrently handing separate (re)connections from
3253 the same client. Add better locking to the connection handling
3254 code, and ensure that only a single connection will be processed
3255 for a given client UUID, even if the lock is dropped.
3257 Severity : enhancement
3259 Description: add TCP zero-copy support to kernel
3260 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3261 sends if the hardware supports scatter-gather and checksumming.
3262 This allows socklnd to do client-write and server-read more
3263 efficiently and reduce CPU utilization from skbuf copying.
3266 Frequency : only if NFS exporting from client
3268 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3269 Details : If setting attributes on a file created under NFS that had
3270 never been opened it would be possible to oops the client
3271 if the file had no objects.
3274 Frequency : always for liblustre
3276 Description: liblustre client does MDS+OSTs setattr RPC for each write
3277 Details : When doing a write from a liblustre client, the client
3278 incorrectly issued an RPC to the MDS and each OST the file was
3279 striped over in order to update the timestamps. When writing
3280 with small chunks and many clients this could overwhelm the MDS
3281 with RPCs. In all cases it would slow down the write because
3282 these RPCs are unnecessary.
3284 Severity : enhancement
3286 Description: allow number of MDS service threads to be changed at module load
3287 Details : It is now possible to change the number of MDS service threads
3288 running. Adding "options mds mds_num_threads={N}" to the MDS's
3289 /etc/modprobe.conf will set the number of threads for the next
3290 time Lustre is restarted (assuming the "mds" module is also
3291 reloaded at that time). The default number of threads will
3292 stay the same, 32 for most systems.
3297 Description: OST crash if filesystem is unformatted or corrupt
3298 Details : If an OST is started on a device that has never been formatted
3299 or if the filesystem is corrupt and cannot even mount then the
3300 error handling cleanup routines would dereference a NULL pointer.
3305 Description: NULL pointer deref in llap_from_page.
3306 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3307 incorrect) mapping if the page was truncated/reclaimed while it was
3308 searched for. Check for this condition and skip such pages when
3309 doing readahead. Introduce extra check to llap_from_page() to
3310 verify page->mapping->host is non-NULL (so page is not anonymous).
3313 Frequency : Sometimes when using sys_sendfile
3315 Description: "page not covered by a lock" warnings from ll_readpage
3316 Details : sendfile called ll_readpage without right page locks present.
3317 Now we introduced ll_file_sendfile that does necessary locking
3318 around call to generic_file_sendfile() much like we do in
3322 Frequency : with certain MDS communication failures at client mount time
3324 Description: NULL pointer deref after failed client mount
3325 Details : a client connection request may delayed by the network layer
3326 and not be sent until after the PTLRPC layer has timed out the
3327 request. If the client fails the mount immediately it will try
3328 to clean up before the network times out the request. Add a
3329 reference from the request import to the obd device and delay
3330 the cleanup until the network drops the request.
3333 Frequency : occasionally during client (re)connect
3335 Description: assertion failure during client (re)connect
3336 Details : processing a client connection request may be delayed by the
3337 client or server longer than the client connect timeout. This
3338 causes the client to resend the connection request. If the
3339 original connection request is replied in this interval, the
3340 client may trip an assertion failure in ptlrpc_connect_interpret()
3341 which thought it would be the only running connect process.
3344 Frequency : only with obd_echo servers and clients that are rebooted
3346 Description: kernel BUG accessing uninitialized data structure
3347 Details : When running an obd_echo server it did not start the ping_evictor
3348 thread, and when a client was evicted an uninitialized data
3349 structure was accessed. Start the ping_evictor in the RPC
3350 service startup instead of the OBD startup.
3352 Severity : enhancement
3353 Bugzilla : 10193 (patchless)
3354 Description: Remove dependency on various unexported kernel interfaces.
3355 Details : No longer need reparent_to_init, exit_mm, exit_files,
3356 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3359 Frequency : rare (only users of deprecated and unsupported LDAP config)
3361 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3362 Details : LDAP apparently contains 'lustreName' attributes instead of
3363 'name'. A simple remapping of the name is sufficient.
3366 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3368 Description: waiting_locks_callback trips kernel BUG if client is evicted
3369 Details : Running with the dump_on_timeout debug flag turned on makes
3370 it possible that the waiting_locks_callback() can try to dump
3371 the Lustre kernel debug logs from an interrupt handler. Defer
3372 this log dumping to the expired_lock_main() thread.
3374 Severity : enhancement
3376 Description: Support NFS exporting on 2.6 kernels.
3377 Details : Implement non-rawops metadata methods for NFS server to use without
3378 changing NFS server code.
3381 Frequency : very rare (synthetic metadata workload only)
3383 Description: two racing renames might cause an MDS thread to deadlock
3384 Details : Running the "racer" program may cause one MDS thread to rename
3385 a file from being the source of a rename to being the target of
3386 a rename at exactly the same time that another thread is doing
3387 so, and the second thread has already enqueued these locks after
3388 doing a lookup of the target and is trying to relock them in
3389 order. Ensure that we don't try to re-lock the same resource.
3392 Frequency : only very large systems with liblustre clients
3394 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3395 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3396 all OSTs in parallel. This allows the eviction of stale liblustre
3397 clients to proceed much faster than if they were done in series,
3398 and also offers similar improvements for other set_info RPCs.
3403 Description: excessive CPU usage during initial read phase on client
3404 Details : During the initial read phase on a client, it would agressively
3405 retry readahead on the file, consuming too much CPU and impacting
3406 performance (since 1.4.5.8). Improve the readahead algorithm
3407 to avoid this, and also improve some other common cases (read
3408 of small files in particular, where "small" is files smaller than
3409 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3414 Description: MDS crash when receiving packet with unknown intent.
3415 Details : Do not LBUG in unknown intent case, just return -EFAULT
3417 Severity : enhancement
3418 Bugzilla : 9293, 9385
3419 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3420 Details : Do not serialize getattr (non-intent version) and statfs.
3423 Frequency : occasional, when OST network is overloaded/intermittent
3425 Description: client evicted by OST after bulk IO timeout
3426 Details : If a client sends a bulk IO request (read or write) the OST
3427 may evict the client if it is unresposive to its data GET/PUT
3428 request. This is incorrect if the network is overloaded (takes
3429 too long to transfer the RPC data) or dropped the OST GET/PUT
3430 request. There is no need to evict the client at all, since
3431 the pinger and/or lock callbacks will handle this, and the
3432 client can restart the bulk request.
3435 Frequency : Always when mmapping file with no objects
3437 Description: client crashes when mmapping file with no objects
3438 Details : Check that we actually have objects in a file before doing any
3439 operations on objects in ll_vm_open, ll_vm_close and
3445 Description: Request leak when working with deleted CWD
3446 Details : Introduce advanced request refcount tracking for requests
3447 referenced from lustre intent.
3449 Severity : Enhancement
3451 Description: Cache open file handles on client.
3452 Details : MDS now will return special lock along with openhandle, if
3453 requested and client is allowed to hold openhandle, even if unused,
3454 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3455 closing files for every read/write openration.
3457 Severity : Enhancement
3459 Description: Cache open negative dentries on client when possible.
3460 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3461 negative dentries on lock revocation.
3466 Description: Remounting a client read-only wasn't possible with a zconf mount
3467 Details : It wasn't possible to remount a client read-only with llmount.
3469 Severity : enhancement
3470 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3471 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3472 MPICH is included to add Lustre-specific ADIO interfaces.
3473 This is based closely on the UFS ADIO layer and only differs
3474 in file creation, in order to allow the OST striping to be set.
3475 This is user-contributed code and not supported by CFS.
3480 Description: extended inode attributes (immutable, append-only) work improperly
3481 when 2.4 and 2.6 kernels are used on client/server or vice versa
3482 Details : Introduce kernel-independent values for these flags.
3484 Severity : enhancement
3487 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3488 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3489 granularity. Now, Fractional MB granularity is be supported,
3490 this is very useful for low memory system.
3492 Severity : enhancement
3494 Description: Getattr by fid
3495 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3496 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3497 have locks and avoids getattr rpc requests for it on every stat.
3500 Frequency : Always, for filesystems larger than 2TB
3502 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3503 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3504 filesystems larger than 2TB to be mounted with mballoc enabled.
3507 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3509 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3510 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3511 filesystems larger than 2TB is mounted then the kernel will
3512 silently corrupt the start of the filesystem. CONFIG_LBD is
3513 enabled for all CFS-supported kernels, but the possibility of
3514 this happening with a modified kernel config exists.
3516 Severity : enhancement
3518 Description: add client O_DIRECT support for 2.6 kernels
3519 Details : It is now possible to do O_DIRECT reads and writes to files
3520 in the Lustre client mountpoint on 2.6 kernel clients.
3522 Severity : enhancement
3524 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3525 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3526 parallel, not waiting for response from every OST before sending
3527 a rpc to the next OST.
3532 Description: setattr vs write race when updating file timestamps
3533 Details : Client processes that update a file timestamp into the past
3534 right after writing to the file (e.g. tar) it is possible that
3535 the updated file modification time can be reset to the current
3536 time due to a race between processing the setattr and write RPC.
3538 Severity : enhancement
3540 Description: Bring 'lfs find' closer in line with regular Linux find.
3541 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3542 -print0 options and obtains all the needed info through the lustre
3545 Severity : enhancement
3547 Description: support up to 1024 configured devices on one node
3548 Details : change obd_dev array from statically allocated to dynamically
3549 allocated structs as they are first used to reduce memory usage
3554 Description: Flush dirty partially truncated pages during truncate
3555 Details : Immediatelly flush partially truncated pages in filter_setattr,
3556 this way we completely avoid having any pages in page cache on OST
3557 and can retire ugly workarounds during writes to flush such pages.
3562 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3563 Details : protect inode from truncation within vfs_unlink() context
3564 just take a reference before calling vfs_unlink() and release it
3565 when parent's i_sem is free.
3570 Description: last_id value checked outside lock on OST caused LASSERT failure
3571 Details : If there were multiple MDS->OST object precreate requests in
3572 flight, it was possible that the OST's last object id was checked
3573 outside a lock and incorrectly tripped an assertion. Move checks
3574 inside locks, and discard old precreate requests.
3577 Frequency : always, if extents are used on OSTs
3579 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3580 Details : OSTs using the extents format would not zero the high 16 bits of
3581 the index physical block number. This is not a problem for any
3582 OST filesystems smaller than 16TB, and no kernels support ext3
3583 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3584 new/modified files) and can be fixed for existing filesystems
3585 with e2fsprogs-1.39-cfs1.
3590 Description: import connection selection may be incorrect if timer wraps
3591 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3592 import connections to be ignored if the 32-bit jiffies counter
3593 wraps. Use a 64-bit jiffies counter.
3596 Frequency : very large clusters immediately after boot
3598 Description: LNET request buffers exhausted under heavy short-term load
3599 Details : If a large number of client requests are generated on a service
3600 that has previously never seen so many requests it is possible
3601 that the request buffer growth cannot keep up with the spike in
3602 demand. Instead of dropping incoming requests, they are held in
3603 the LND until the RPC service can accept more requests.
3606 Frequency : Sometimes during replay
3608 Description: Assertion failure in ll_local_open after replay.
3609 Details : If replay happened on an open request reply before we were able
3610 to set replay handler, reply will become not swabbed tripping the
3611 assertion in ll_local_open. Now we set the handler right after
3612 recognising of open request
3615 Frequency : very rare
3617 Description: kernel reports "badness in vsnprintf"
3618 Details : Reading from the "recovery_status" /proc file in small chunks
3619 may cause a negative length in lprocfs_obd_rd_recovery_status()
3620 call to vsnprintf() (which is otherwise harmless). Exit early
3621 if there is no more space in the output buffer.
3623 Severity : enhancement
3625 Description: clear OBD RPC statistics by writing to them
3626 Details : It is now possible to clear the OBD RPC statistics by writing
3627 to the "stats" file.
3632 Description: Client mtime is not the same on different clients after utimes
3633 Details : In some cases, the client was using the utimes() syscall on
3634 a file cached on another node. The clients now validate the
3635 ctime from the MDS + OSTs to determine which one is right.
3640 Description: Inability to activate failout mode
3641 Details : lconf script incorrectly assumed that in pythong string's numeric
3642 value is used in comparisons.
3645 Frequency : always with multiple stripes per file
3647 Description: Inefficient object allocation for mutli-stripe files
3648 Details : When selecting which OSTs to stripe files over, for files with
3649 a stripe count that divides evenly into the number of OSTs,
3650 the MDS is always picking the same starting OST for each file.
3651 Return the OST selection heuristic to the original design.
3656 Description: mount failures may take full timeout to return an error
3657 Details : Under some heavy load conditions it is possible that a
3658 failed mount can wait for the full obd_timeout interval,
3659 possibly several minutes, before reporting an error.
3660 Instead return an error as soon as the status is known.
3662 Frequency : quota enabled and large files being deleted
3664 Description: releasing more than 4GB of quota at once hangs OST
3665 Details : If a user deletes more than 4GB of files on a single OST it
3666 will cause the OST to spin in an infinite loop. Release
3667 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3672 Description: statfs data retrieved from /proc may be stale or zero
3673 Details : When reading per-device statfs data from /proc, in the
3674 {kbytes,files}_{total,free,avail} files, it may appear
3675 as zero or be out of date.
3678 Frequency : systems with MD RAID1 external journal devices
3680 Description: lconf's call to blkid is confused by RAID1 journal devices
3681 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3682 of returning all block devices that match the journal UUID.
3685 Frequency : always, for aggregate stripe size over 4GB
3687 Description: assertion fails when trying to use 4GB stripe size
3688 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3689 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3692 Frequency : always on ppc64
3694 Description: the first write on an ext3 filesystem with mballoc got stuck
3695 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3696 perform endianness conversion.
3698 ------------------------------------------------------------------------------
3700 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3702 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3703 INTEROPERATE with older versions automatically. Please read the
3704 user documentation before upgrading any part of a live system.
3705 * WARNING: Lustre networking configuration changes are required with
3706 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3709 * Support for kernels:
3710 2.6.9-22.0.2.EL (RHEL 4)
3711 2.6.5-7.244 (SLES 9)
3712 2.6.12.6 vanilla (kernel.org)
3715 Severity : enhancement
3716 Bugzilla : 7981/8208
3717 Description: Introduced Lustre Networking (LNET)
3718 Details : LNET is new networking infrastructure for Lustre, it includes
3719 a reorganized network configuration mode (see the user
3720 documentation for full details) as well as support for routing
3721 between different network fabrics. Lustre Networking Devices
3722 (LNDs) for the supported network fabrics have also been
3723 created for this new infrastructure.
3725 Severity : enhancement
3726 Description: Introduced Access control lists
3727 Details : clients can set ACLs on files and directories in order to have
3728 more fine-grained permissions than the standard Unix UGO+RWX.
3729 The MDS must be started with the "-o acl" mount option.
3731 Severity : enhancement
3732 Description: Introduced filesystem quotas
3733 Details : Administrators may now establish per-user quotas on the
3736 Severity : enhancement
3738 Description: Configuration change for the XT3
3739 The PTLLND is now used to run Lustre over Portals on the XT3
3740 The configure option(s) --with-cray-portals are no longer used.
3741 Rather --with-portals=<path-to-portals-includes> is used to
3742 enable building on the XT3. In addition to enable XT3 specific
3743 features the option --enable-cray-xt3 must be used.
3748 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3749 Details : A race between allocating a new llcd and re-getting the llcd_lock
3750 allowed another thread to grab newly-allocated llcd.
3752 Severity : enhancement
3754 Description: 2.6 OST async journal commit and locking fix to improve performance
3755 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3756 2.6 kernels are now async as they already were in 2.4 kernels so
3757 that they can commit concurrently with the network bulk transfer.
3758 For block-allocated files the filter allocation semaphore is held
3759 to avoid filesystem fragmentation during allocation. BKL lock
3760 removed for 2.6 xattr operations where it is no longer needed.
3765 Description: lconf incorrectly determined whether two IP networks could talk
3766 Details : In some more complicated routing and multiple-network
3767 configurations, lconf will avoid trying to make a network
3768 connection to a disjoint part of the IP space. It was doing the
3769 math incorrectly for one set of cases.
3774 Description: Fix for potential infinite loop processing records in an llog.
3775 Details : If an llog record is corrupted/zeroed, it is possible to loop
3776 forever in llog_process(). Validate the llog record length
3777 and skip the remainder of the block on error.
3780 Frequency : occasional (liblustre only)
3782 Description: liblustre could not open files whose last component is a symlink
3783 Details : sysio_path_walk() would incorrectly pass the open intent to
3784 intermediate path components.
3787 Frequency : rare (liblustre only with non-standard tuning)
3788 Bugzilla : 7201 (7350)
3789 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3790 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3791 after it has been released, only do this for referenced locks
3793 Severity : enhancement
3795 Description: specifying an (invalid) directory default stripe_size of -1
3796 would reset the directory default striping
3797 Details : stripe_size -1 was used internally to signal directory stripe
3798 removal, now use "all default" to signal dir stripe removal
3799 as a directory striping of "all default" is not useful
3802 Frequency : common for large clusters running liblustre clients
3804 Description: doing an ls when liblustre clients are running is slow
3805 Details : sending a glimpse AST to a liblustre client waits for every AST
3806 to time out, as liblustre clients will not respond. Since they
3807 cannot cache data we refresh the OST lock LVB from disk instead.
3809 Severity : enhancement
3811 Description: doing an ls at the same time as file IO can be slow
3812 Details : enqueue and other "small" requests can be blocked behind many
3813 large IO requests. Create a new OST IO portal for non-IO
3814 requests so they can be processed faster.
3817 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3819 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3820 created file with mode 000
3821 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3822 allow an ftruncate() as a non-root user to a file with mode 000.
3823 Lustre now respects this flag to disable mode checks when
3824 truncating a file owned by the user
3827 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3830 Description: Revoking locks from clients that went dead or catatonic might take
3832 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3833 cancellation of such locks instant on servers without waiting for
3834 any reply from clients. Clients drops these locks when cancel
3835 notification from server is received without replying.
3838 Frequency : liblustre-only, when liblustre client dies or becomes busy
3840 Description: Doing ls on Linux clients can take a long time with active
3842 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3843 granting such locks to it in the first place if possible. Locks
3844 are taken by proxy on the OST during the read or write and
3845 dropped immediately afterward. Add connect flags handling, do
3846 not grant locks to liblustre clients for glimpse ASTs.
3848 Severity : enhancement
3850 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3851 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3852 sequential reads and behaves badly with applications doing random
3853 reads. Improve it by reading ahead at least read region, and
3854 avoiding excessive large RPC for small reads.
3856 Severity : enhancement
3858 Description: Creating more than 1000 files for a single job may cause a load
3859 imbalance on the OSTs if there are also a large number of OSTs.
3860 Details : qos_prep_create() uses an OST index reseed value that is an
3861 even multiple of the number of available OSTs so that if the
3862 reseed happens in the middle of the object allocation it will
3863 still utilize the OSTs as uniformly as possible.
3868 Description: OST or MDS may oops in ping_evictor_main()
3869 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3870 but doesn't restart at beginning of obd_exports_timed list
3873 Severity : enhancement
3875 Description: improve by-nid export eviction on the MDS and OST
3876 Details : allow multiple exports with the same NID to be evicted at one
3877 time without re-searching the exports list.
3880 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3882 Description: MDS may oops in groups_free()
3883 Details : in rare race conditions a newly allocated group_info struct is
3884 freed again, and this can be NULL. The 2.4 compatibility code
3885 for groups_free() checked for a NULL pointer, but 2.6 did not.
3888 Frequency : common for liblustre clients doing little filesystem IO
3889 Bugzilla : 9352, 7313
3890 Description: server may evict liblustre clients accessing contended locks
3891 Details : if a client is granted a lock or receives a completion AST
3892 with a blocking AST already set it would not reply to the AST
3893 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3896 Frequency : lfs setstripe, only systems with more than 160 OSTs
3898 Description: unable to set striping with a starting offset beyond OST 160
3899 Details : llapi_create_file() incorrectly limited the starting stripe
3900 index to the maximum single-file stripe count.
3903 Frequency : LDAP users only
3905 Description: lconf did not handle in-kernel recovery with LDAP properly
3906 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3908 Severity : enhancement
3910 Description: bind OST threads to NUMA nodes to improve performance
3911 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3912 node and do their allocations there to localize memory access
3914 Severity : enhancement
3916 Description: llmount can determine client NID directly from Myrinet (GM)
3917 Details : the client NID code from gmnalnid was moved directly into
3918 llmount, removing the need to use this or specifying the
3919 client NID explicitly when mounting GM clients with zeroconf
3922 Frequency : if client is started with down MDS
3924 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3925 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3926 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3927 ptlrpc_import_delayed_req() because we didn't check if the
3928 request was interrupted, and we also didn't break out of the
3929 event loop if there was no timeout
3934 Description: data loss during non-page-aligned writes to a single file from
3935 both multiple nodes and multiple threads on one node at same time
3936 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3937 inconsistency led to false short-reads, that were cached and later
3938 used by ->prepare_write() to fill in partially written page,
3939 leading to data loss.
3942 Frequency : always, if lconf --abort_recovery used
3944 Description: lconf --abort_recovery fails with 'Operation not supported'
3945 Details : lconf was attempting to abort recovery on the MDT device and not
3948 Severity : enhancement
3950 Description: remove cleanup logs
3951 Details : replace lconf-generated cleanup logs with lustre internal
3952 cleanup routines. Eliminates the need for client-cleanup and
3955 Severity : enhancement
3957 Description: add support for EAs (user and system) on lustre filesystems
3958 Details : it is now possible to store extended attributes in the Lustre
3959 client filesystem, and with the user_xattr mount option it
3960 is possible to allow users to store EAs on their files also
3962 Severity : enhancement
3964 Description: Add possibility (config option) to show minimal available OST free
3966 Details : When compiled with --enable-mindf configure option, statfs(2)
3967 (and so, df) will return least minimal free space available from
3968 all OSTs as amount of free space on FS, instead of summary of
3969 free spaces of all OSTs.
3971 Severity : enhancement
3973 Description: do not expand extent locks acquired on OST-side
3974 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3975 by server: they are not cached anyway.
3978 Frequency : when mmap is used/binaries executed from Lustre
3980 Description: Unmmap pages before throwing them away from read cache.
3981 Details : llap_shrink cache now attempts to unmap pages before discarding
3982 them (if unmapping failed - do not discard). SLES9 kernel has
3983 extra checks that trigger if this unmapping is not done first.
3988 Description: lconf didn't resolve symlinks before checking to see whether a
3989 given mountpoint was already in use
3992 Frequency : when migrating failover services
3993 Bugzilla : 6395, 9514
3994 Description: When migrating a subset of services from a node (e.g. failback
3995 from a failover service node) the remaining services would
3996 time out and evict clients.
3997 Details : lconf --force (implied by --failover) sets the global obd_timeout
3998 to 5 seconds in order to quickly disconnect, but this caused
3999 other RPCs to time out too quickly. Do not change the global
4000 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4002 Severity : enhancement
4003 Frequency : if MDS is started with down OST
4004 Bugzilla : 9439,5706
4005 Description: Allow startup/shutdown of an MDS without depending on the
4006 availability of the OSTs.
4007 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4008 Add appropriate locking and lov-osc refcounts for safe
4009 cleaning. Add osc abort_inflight calls in case the
4010 synchronize never started.
4013 Frequency : occasional (Cray XT3 only)
4015 Description: root not authorized to access files in CRAY_PORTALS environment
4016 Details : The client process capabilities were not honoured on the MDS in
4017 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4018 been accessed by an authorized user then root was able to access
4019 the file on the local client also. The root user capabilities
4020 are now allowed on the MDS, as this environment has secure UID.
4023 Frequency : occasional
4025 Description: ldiskfs "too long searching" message happens too often
4026 Details : A debugging message (otherwise harmless) prints too often on
4027 the OST console. This has been reduced to only happen when
4028 there are fragmentation problems on the filesystem.
4033 Description: Division by zero in statfs when all OSCs are inactive
4034 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4035 lov_statfs divides by value returned by lov_get_stripecnt().
4039 Bugzilla : 9489, 3273
4040 Description: First write from each client to each OST was only 4kB in size,
4041 to initialize client writeback cache, which caused sub-optimal
4042 RPCs and poor layout on disk for the first writen file.
4043 Details : Clients now request an initial cache grant at (re)connect time
4044 and so that they can start streaming writes to the cache right
4045 away and always do full-sized RPCs if there is enough data.
4046 If the OST is rebooted the client also re-establishes its grant
4047 so that client cached writes will be honoured under the grant.
4052 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4053 Details : Now I/O RPCs go to different portal number and (presumably) fast
4054 lock requests (and glimses) and other RPCs get their own service
4055 threads pool that should be able to service those RPCs
4058 Severity : enhancement
4060 Description: Ability to exchange lustre version between client and servers and
4061 issue warnings at client side if client is too old. Also for
4062 liblustre clients there is ability to refuse connection of too old
4064 Details : New 'version' field is added to connect data structure that is
4065 filled with version info. That info is later checked by server and
4069 Frequency : rare, liblustre only.
4070 Bugzilla : 9296, 9581
4071 Description: Two simultaneous writes from liblustre at offset within same page
4072 might proceed at the same time overwriting eachother with stale
4074 Details : I/O lock withing llu_file_prwv was released too early, before data
4075 actually was hitting the wire. Extended lock-holding time until
4076 server acknowledges receiving data.
4079 Frequency : extremely rare. Never observed in practice.
4081 Description: avoid generating lustre_handle cookie of 0.
4082 Details : class_handle_hash() generates handle cookies by incrementing
4083 global counter, and can hit 0 occasionaly (this is unlikely, but
4084 not impossible, because initial value of cookie counter is
4085 selected randonly). Value of 0 is used as a sentinel meaning
4086 "unassigned handle" --- avoid it. Also coalesce two critical
4087 sections in this function into one.
4089 Severity : enhancement
4091 Description: allow liblustre clients to delegate truncate locking to OST
4092 Details : To avoid overhead of locking, liblustre client instructs OST to
4093 take extent lock in ost_punch() on client's behalf. New connection
4094 flag is added to handle backward compatibility.
4096 Severity : enhancement
4097 Bugzilla : 4928, 7341, 9758
4098 Description: allow number of OST service threads to be specified
4099 Details : a module parameter allows the number of OST service threads
4100 to be specified via "options ost ost_num_threads={N}" in the
4101 OSS's /etc/modules.conf or /etc/modprobe.conf.
4105 Bugzilla : 6146, 9635, 9895
4106 Description: servers crash with bad pointer in target_handle_connect()
4107 Details : In rare cases when a client is reconnecting it was possible that
4108 the connection request was the last reference for that export.
4109 We would temporarily drop the export reference and get a new
4110 one, but this may have been the last reference and the export
4111 was just destroyed. Get new reference before dropping old one.
4113 Severity : enhancement
4114 Frequency : if client is started with failover MDS
4116 Description: Allow multiple MDS hostnames in the mount command
4117 Details : Try to read the configuration from all specified MDS
4118 hostnames during a client mount in case the "primary"
4121 Severity : enhancement
4123 Description: Stop sending data to evicted clients as soon as possible.
4124 Details : Check if the client we are about to send or are sending data to
4125 was evicted already. (Check is done every second of waiting,
4126 for which l_wait_event interface was extended to allow checking
4127 of exit condition at specified intervals).
4130 Frequency : rare, normally only when NFS exporting is done from client
4132 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4134 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4137 Frequency : always, if srand() is called before liblustre initialization
4139 Description: Liblustre uses system PRNG disturbing its usage by user application
4140 Details : Introduce internal to lustre fast and high-quality PRNG for
4141 lustre usage and make liblustre and some other places in generic
4142 lustre code to use it.
4144 Severity : enhancement
4145 Bugzilla : 9477, 9557, 9870
4146 Description: Verify that the MDS configuration logs are updated when xml is
4147 Details : Check if the .xml configuration logs are newer than the config
4148 logs stored on the MDS and report an error if this is the case.
4149 Request --write-conf, or allow starting with --old_conf.
4151 Severity : enhancement
4153 Description: Handle symlinks in the path when checking if Lustre is mounted.
4154 Details : Resolve intermediate symlinks when checking if a client has
4155 mounted a filesystem to avoid duplicate client mounts.
4160 Description: lconf can hit an error exception but still return success.
4161 Details : The lconf command catches the Command error exception at the top
4162 level script context and will exit with the associated exit
4163 status, but doesn't ensure that this exit status is non-zero.
4168 Description: failure of ptlrpc thread startup can cause oops
4169 Details : Starting a ptlrpc service thread can fail if there are a large
4170 number of threads or the server memory is very fragmented.
4171 Handle this without oopsing.
4174 Frequency : always, only if liblustre and non-default acceptor port was used
4176 Description: liblustre cannot connect to servers with non-default acceptor port
4177 Details : tcpnal_set_default_params() was not called and was therefore
4178 ignoring the environment varaible TCPNAL_PORT, as well as other
4179 TCPNAL_ environment variables
4184 Description: two objects could be created on the same OST for a single file
4185 Details : If an OST is down, in some cases it was possible to create two
4186 objects on a single OST for a single file. No problems other
4187 than potential performance impact and spurious error messages.
4191 Bugzilla : 5681, 9562
4192 Description: Client may oops in ll_unhash_aliases
4193 Details : Client dcache may become inconsistent in race condition.
4194 In some cases "getcwd" can fail if the current directory is
4200 Description: Inode refcounting problems in NFS export code
4201 Details : link_raw functions used to call d_instantiate without obtaining
4202 extra inode reference first.
4206 Bugzilla : 9942, 9903
4207 Description: Referencing freed requests leading to crash, memleaks with NFS.
4208 Details : We used to require that call to ll_revalidate_it was always
4209 followed by ll_lookup_it. Also with revalidate_special() it is
4210 possible to call ll_revalidate_it() twice for the same dentry
4211 even if first occurence returned success. This fix changes semantic
4212 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4213 reference on a request referred from the intent.
4214 ll_intent_release() then releases such a request.
4217 Frequency : rare, normally benchmark loads only
4219 Description: unlinked inodes were kept in memory on the client
4220 Details : If a client is repeatedly creating and unlinking files it
4221 can accumulate a lot of stale inodes in the inode slab cache.
4222 If there is no other client load running this can cause the
4223 client node to run out of memory. Instead flush old inodes
4224 from client cache that have the same inode number as a new inode.
4227 Frequency : SLES9 2.6.5 kernel and long filenames only
4228 Bugzilla : 9969, 10379
4229 Description: utime reports stale NFS file handle
4230 Details : SLES9 uses out-of-dentry names in some cases, which confused
4231 the lustre dentry revalidation. Change it to always use the
4235 Frequency : rare, unless heavy write-truncate concurrency is continuous
4236 Bugzilla : 4180, 6984, 7171, 9963, 9331
4237 Description: OST becomes very slow and/or deadlocked during object unlink
4238 Details : filter_destroy() was holding onto the parent directory lock
4239 while truncating+unlinking objects. For very large objects this
4240 may block other threads for a long time and slow overall OST
4241 responsiveness. It may also be possible to get a lock ordering
4242 deadlock in this case, or run out of journal credits because of
4243 the combined truncate+unlink. Solution is to do object truncate
4244 first in one transaction without parent lock, and then do the
4245 final unlink in a new transaction with the parent lock. This
4246 reduces the lock hold time dramatically.
4249 Frequency : rare, 2.4 kernels only
4251 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4252 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4253 before unmounting their filesystem in order to avoid deadlock.
4254 The kernel_locked() function in 2.4 kernels only checks whether
4255 the kernel lock is held, not whether it is this process that is
4256 holding it as 2.6 kernels do.
4261 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4262 Details : The client ptlrpc code may be trying to reconnect to a down
4263 server before a previous connection attempt has timed out.
4264 Increase the reconnect interval to be longer than the connection
4265 timeout interval to avoid sending duplicate connections to
4269 Frequency : echo_client brw_test command
4271 Description: fix echo_client to work with OST preallocated code
4272 Details : OST preallocation code (5137) didn't take echo_client IO path
4273 into account: echo_client calls filter methods outside of any
4274 OST thread and, hence, there is no per-thread preallocated
4275 pages and buffers to use. Solution: hijack pga pages for IO. As
4276 a byproduct, this avoids unnecessary data copying.
4280 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4281 Description: Client can oops in mdc_commit_close() after open replay
4282 Details : It was possible for the MDS to return an open request with no
4283 transaction number in mds_finish_transno() if the client was
4284 evicted, but without actually returning an error. Clients
4285 would later try to replay that open and may trip an assertion
4286 Simplify the client close codepath, and always return an error
4287 from the MDS in case the open is not successful.
4290 Frequency : rare, 2.6 OSTs only
4292 Description: OST may deadlock under high load on fragmented files
4293 Details : If there was a heavy load and highly-fragmented OST filesystems
4294 it was possible to have all the OST threads deadlock waiting on
4295 allocation of biovecs, because the biovecs were not released
4296 until the entire RPC IO was completed. Instead, release biovecs
4297 as soon as they are complete to ensure forward IO progress.
4299 Severity : enhancement
4301 Description: Support for specifying external journal device at mount
4302 Details : If an OST or MDS device is formatted with an external journal
4303 device, this device major/minor is stored in the ext3 superblock
4304 and may not be valid for failover. Allow detecting and
4305 specifying the external journal at mount time.
4310 Description: Mounting an MDS with pending unlinked files may cause oops
4311 Details : target_finish_recovery() calls mds_postrecov() which returned
4312 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4313 considers this an error and immediately begins cleaning up the
4314 lov, just after starting the mds_lov process
4316 Severity : enhancement
4318 Description: Implement 'lfs df' to report actual free space on per-OST basis
4319 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4320 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4321 usage of objects; '-h' to report in human readable format.
4323 ------------------------------------------------------------------------------
4325 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4332 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4333 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4334 at startup during orphan unlinking. Instead, skip buddy update
4335 before setup, as it will be regenerated after recovery is complete.
4340 Description: If an OST is inactive, its locks might reference stale inodes.
4341 Details : lov_change_cbdata() must iterate over all namespaces, even if
4342 they are inactive to clear inode references from the lock.
4344 Severity : enhancement
4345 Frequency : occasional, if non-standard max_dirty_mb used
4347 Description: Client will block write RPCs if not enough grant
4348 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4349 then the client will block writes while waiting for another RPC
4350 to complete instead of consuming its dirty limit. With change
4351 we get improved performance when max_dirty_mb is small.
4353 Severity : enhancement
4354 Bugzilla : 3389, 6253
4355 Description: Add support for supplementary groups on the MDS.
4356 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4357 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4358 lookups for user supplementary groups into a cache.
4362 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4363 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4364 for liblustre O_EXCL is enforced by MDS.
4368 Description: readdir never returns NULL in liblustre.
4369 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4370 dentry in struct dirent.
4374 Frequency : liblustre only, depends on application IO pattern
4375 Description: liblustre clients evicted if not contacting servers
4376 Details : Don't put liblustre clients into the ping_evictor list, so
4377 they will not be evicted by the pinger ever.
4379 Severity : enhancement
4381 Description: Add ability to evict clients by NID from MDS.
4382 Details : By echoing "nid:$NID" string into
4383 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4384 $NID would be instantly evicted from this MDS and from all active
4385 OSTs connected to it.
4389 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4390 Details : lookup_it_finish() used to query file size from OSTs that was not
4395 Description: service threads change working directory to that of init
4396 Details : Starting lustre service threads may pin the working directory
4397 of the parent thread, making that filesystem busy. Threads
4398 now change to the working directory of init to avoid this.
4402 Frequency : during shutdown only
4403 Description: shutdown with a failed MDS or OST can cause unmount to hang
4404 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4409 Frequency : 2.6 only
4410 Description: chmod/chown may include an extra supplementary group
4411 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4412 supplementary group and if none is specified this is used.
4415 Bugzilla : 5479 (6816)
4417 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4418 Details : If lookup is in progress on a file that is unlinked we might try
4419 to revalidate the inode and fail in revalidate after lookup is
4420 complete and ll_file_open() enqueues the open again but
4421 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4424 Frequency : always, if lconf --abort_recovery used
4426 Description: lconf --abort_recovery fails with 'Operation not supported'
4427 Details : lconf was attempting to abort recovery on the MDT device and not
4430 ------------------------------------------------------------------------------
4432 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4437 Frequency : rare (only unsupported configurations with a node running as an
4439 Bugzilla : 6514, 5137
4440 Description: Mounting a Lustre file system on a node running as an OST could
4442 Details : OSTs now preallocates memory needed to write out data at
4443 startup, instead of when needed, to avoid having to
4444 allocate memory in possibly low memory situations.
4445 Specifically, if the file system is mounted on on OST,
4446 memory pressure could force it to try to write out data,
4447 which it needed to allocate memory to do. Due to the low
4448 memory, it would be unable to do so and the node would
4449 become unresponsive.
4451 Severity : enhancement
4453 Description: Addition of lconf --service command line option
4454 Details : lconf now accepts a '--service <arg>' option, which is
4455 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4457 Severity : enhancement
4459 Description: Failover mode is now the default for OSTs.
4460 Details : By default, OSTs will now run in failover mode. To return to
4461 the old behaviour, add '--failout' to the lmc line for OSTs.
4463 Severity : enhancement
4465 Description: Health checks are now provided for MDS and OSTs
4466 Details : Additional detailed health check information on MSD and OSTs
4467 is now provided through the procfs health_check value.
4470 Frequency : occasional, depends on IO load
4472 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4473 after numerous create/delete cycles
4474 Details : The ext3 inode allocation policy would not allocate new inodes
4475 very well on the OSTs because there are no new directories
4476 being created. Instead we look for groups with free space if
4477 the parent directories are nearly full.
4482 Description: Network or server problems during mount may cause partially
4483 mounted clients instead of returning an error.
4484 Details : The config llog parsing code may overwrite the error return
4485 code during mount error handling, returning success instead
4491 Description: MDS can fail to allocate large reply buffers
4492 Details : After long uptimes the MDS can fail to allocate large reply
4493 buffers (e.g. zconf client mount config records) due to memory
4494 fragmentation or consumption by the buffer cache. Preallocate
4495 some large reply buffers so that these replies can be sent even
4496 under memory pressure.
4500 Frequency : rare (liblustre)
4501 Description: fsx running with liblustre complained that using truncate() to
4502 extend the file doesn't work. This patch corrects that issue.
4503 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4504 ATTR_SIZE and lsm use in llu_setattr_raw.
4508 Frequency : rare, only 2.6 kernels
4509 Description: Unusual file access patterns on the MDS may result in inode
4510 data being lost in very rare circumstances.
4511 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4512 code in the 2.6 kernel caused the inode and/or EA data not to
4513 be read from disk, causing single-file corruption.
4517 Frequency : rare, only 2.6 filesystems using extents
4518 Description: Heavy concurrent write and delete load may cause data corruption.
4519 Details : It was possible under high-load situations to have an extent
4520 metadata block in the block device cache from a just-unlinked
4521 file overwrite a newly-allocated data block. We now unmap any
4522 metadata buffers that alias just-allocated data blocks.
4526 Frequency : filesystems with default stripe_count larger than 77
4527 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4528 Details : lconf specifies an inode size of 4096 bytes when the default
4529 stripe_count is larger than 77. This conflicts with the default
4530 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4531 this case to avoid pinning too much memory for large EAs.
4533 ------------------------------------------------------------------------------
4535 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4540 Frequency : rare (extremely heavy IO load with hundreds of clients)
4542 Description: Client is evicted, gets IO error writing to file
4543 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4544 caused clients to be evicted for AST timeouts. The fixes in
4545 bug 5192 mean we no longer need to have such short AST timeouts
4546 so ldlm_timeout has been increased.
4549 Frequency : occasional during --force or --failover shutdown under load
4550 Bugzilla : 5949, 4834
4551 Description: Server oops/LBUG if stopped with --force or --failover under load
4552 Details : a collection of import/export refcount and cleanup ordering
4553 issues fixed for safer force cleanup
4556 Frequency : only filesystems larger than 120 OSTs
4557 Bugzilla : 5990, 6223
4558 Description: lfs getstripe would oops on a very large filesystem
4559 Details : lov_getconfig used kfree on vmalloc'd memory
4562 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4563 Bugzilla : 6242, 6243
4564 Description: reading from files that had been truncated to a non-zero size
4565 but never opened returned no data
4566 Details : ll_file_read() reads zeros from no-object files to EOF
4571 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4572 Details : There's little harm in aborting MDS/OSS recovery and letting it
4573 try again, so I removed the LASSERT and return an error instead.
4575 Severity : enhancement
4577 Description: New debugging infrastructure for tracking down data corruption
4578 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4579 (b) cover more of the client-side code path, and (c) try to narrow
4580 down where problems occurred
4584 Bugzilla : 3819, 4364, 4397, 6313
4585 Description: Racing close and eviction MDS could cause assertion in mds_close
4586 Details : It was possible to get multiple mfd references during close and
4587 client eviction, leading to one thread referencing a freed mfd.
4589 Severity: : enhancement
4590 Bugzilla : 3262, 6359
4591 Description: Attempts to reconnect to servers are now more aggressive.
4592 Details : This builds on the enhanced upcall-less recovery that was added
4593 in 1.4.2. When trying to reconnect to servers, clients will
4594 now try each server in the failover group every 10 seconds. By
4595 default, clients would previously try one server every 25 seconds.
4600 Description: After recovery, certain operations trigger a failed
4601 assertion on a client.
4602 Details : Failing over an mds, using lconf -d --failover, while a
4603 client was doing a readdir() call would cause the client to
4604 LBUG after recovery completed and the readdir() was resent.
4606 Severity : enhancement
4608 Description: Default groups are now added by lconf
4609 Details : You can now run lconf --group <servicename> without having to
4610 manually add groups with lmc.
4613 Frequency : occasional
4615 Description: Nodes with an elan id of 0 trigger a failed assertion
4618 Frequency : always when accessing e.g. tty/console device nodes
4620 Description: tty and some other devices nodes cannot be used on lustre
4621 Details : file's private_data field is used by device data and lustre
4622 values in there got lost. New field was added to struct file to
4623 store fs-specific private data.
4626 Frequency : when exporting Lustre via NFS
4628 Description: NFSD failed occasionally when looking up a path component
4629 Details : NFSD is looking up ".." which was broken in ext3 directories
4630 that had grown large enough to become hashed.
4633 Frequency : Clusters with multiple interfaces not on the same subnet
4635 Description: Nodes will repeatedly try to reconnect to an interface which it
4636 cannot reach and report an error to the log.
4637 Details : Extra peer list entries will be created by lconf with some peers
4638 unreachable. lconf now validates the peer before adding it.
4641 Frequency : Only if a default stripe is set on the filesystem root.
4643 Description: Setting a default stripe on the filesystem root prevented the
4644 filesystem from being remounted.
4645 Details : The client was sending extra request flags in the root getattr
4646 request and did not allocate a reply buffer for the dir EA.
4649 Frequency : occasional, higher if lots of files are accessed by one client
4650 Bugzilla : 6159, 6097
4651 Description: Client trips assertion regarding lsm mismatch/magic
4652 Details : While revalidating inodes the VFS looks up inodes with ifind()
4653 and in rare cases can find an inode that is being freed.
4654 The ll_test_inode() code will free the lsm during ifind()
4655 when it finds an existing inode and then the VFS later attaches
4656 this free lsm to a new inode.
4660 Bugzilla : 6422, 7030
4661 Description: MDS deadlock between mkdir and client eviction
4662 Details : Creating a new file via mkdir or mknod (starting a transaction
4663 and getting the ns lock) can deadlock with client eviction
4664 (gets ns lock and trying to finish a synchronous transaction).
4667 Frequency : occasional
4668 Description: While starting a server, the fsfilt_ext3 module could not be
4670 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4671 kernels. Previously, lconf would still use the ext3 name
4672 when trying to load modules. Now, it will correctly use
4673 ext3 on 2.4 and ldiskfs on 2.6.
4675 Severity : enhancement
4676 Description: The default stripe count has been changed to 1
4677 Details : The interpretation of the default stripe count (0, to lfs
4678 or lmc) has been changed to mean striping across a single
4679 OST, rather than all available. For general usage we have
4680 found a stripe count of 1 or 2 works best.
4682 Severity : enhancement
4683 Description: Add support for compiling against Cray portals.
4684 Details : Conditional compiling for some areas that are different
4688 Frequency : occasional
4689 Bugzilla : 6409, 6834
4690 Description: Creating files with an explicit stripe count may lead to
4691 a failed assertion on the MDS
4692 Details : If some OSTs are full or unavailable, creating files may
4693 trigger a failed assertion on the MDS. Now, Lustre will
4694 try to use other servers or return an error to the
4698 Frequency : occasional
4700 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4701 caused lock timeout during readahead (since 1.4.2).
4702 Details : Processes doing readahead might match a lock that hasn't been
4703 granted yet if there are overlapping and conflicting lock
4704 requests. The readahead process waits on ungranted lock
4705 (original lock is CBPENDING), while OST waits for that process
4706 to cancel CBPENDING read lock and eventually evicts client.
4708 Severity : enhancement
4710 Description: Initial enabling of flock support for clients
4711 Details : Implements fcntl advisory locking and file status functions.
4712 This feature is provided as an optional mount flag (default
4713 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4714 locking are implemented yet, and those that are are not guaranteed
4715 to be completely correct in production environments.
4716 mount -t lustre -o [flock|noflock] ...
4719 Frequency : occasional
4721 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4722 assertion in the ext3 JBD (journaling) layer.
4723 Details : The b_committed_data struct is protected by the big kernel lock
4724 in 2.4 kernels, serializing journal_commit_transaction() and
4725 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4726 access to this struct. In 2.6 kernels there is finer grained
4727 locking to improve SMP performance of the JBD layer.
4731 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4736 Description: merge and cleanup kernel patches.
4739 -----------------------------------------------------------------------------
4741 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4743 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4744 versions of Lustre. Please update all servers and clients to
4745 version 1.4.2 or later at the same time. You must also run
4746 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4747 to update the configuration logs.
4749 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4750 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4751 - swab llog records of type '0' so we get proper header size/idx (5861)
4752 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4753 - fix rename of one directory over another leaking an inode (5953)
4754 - avoid SetPageDirty on 2.6 (5981)
4755 - don't re-add just-being-destroyed locks to the waiting list (5653)
4756 - when creating new directories, inherit the parent's custom
4757 striping settings if present parent (3048)
4758 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4759 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4760 - don't hold client locks on temporary worklist from l_lru (5666)
4761 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4762 - automatically evict dead clients (5921)
4763 - Update file size properly in create+truncate+fstat case (6196)
4764 - Do not unhash mountpoint dentries, do not allow removal of
4766 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4767 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4768 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4769 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4770 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4771 - Correctly handle reads of files with no objects (6243)
4772 - lctl recover will also mark a device active if deactivate used (5933)
4774 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4775 - allow --write-conf on an MDS with different nettype than client (5619)
4776 - don't write config llogs to MDS for mounts not from that MDS (5617)
4777 - lconf should create multiple TCP connections from a client (5201)
4778 - init scripts are now turned off by default; run chkconfig --on
4779 lustre and chkconfig --on lustrefs to use them
4780 - upcalls are no longer needed for clients to recover to failover
4782 - add --abort-recovery option to lconf to abort recovery on device
4784 - add support for an arbitrary number of OSTs (3026)
4785 - Quota support protocol changes.
4786 - forward compatibility changes to wire structs (6007)
4787 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4788 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4789 - improved "lustre status" script
4790 - initialize blocksize for non-regular files (6062)
4791 - added --disable-server and --disable-client configure options (5782)
4792 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4793 - Vanilla 2.4.29 support
4794 - increase maximum number of obd devices to 520 (6242)
4795 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4796 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4797 SLES 9 2.6.5 kernels
4798 - limit stripes per file to 160 (the maximum EA size) (6093)
4800 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4803 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4804 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4805 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4806 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4807 - small ext3 extents cleanups and fixes (5733)
4808 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4809 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4810 - avoid needless client->OST connect, fix handle mismatch (5317)
4811 - fix DLM error path that led to out-of-sync client, long delays (5779)
4812 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4813 - fix several locking issues related to i_size (5492,5624,5654,5672)
4814 - don't move pending lock onto export if it is already evicted (5683)
4815 - fix kernel oops when creating .foo in unlinked directory (5548)
4816 - fix deadlock in obdfilter statistics vs. object create (5811)
4817 - use time_{before,after} to avoid timer jiffies wrap (5882)
4818 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4819 - Do not leak request if server was not able to process it (5154)
4820 - If mds_open unable to find parent dir, make that negative lookup(5154)
4821 - don't create new directories with extent-mapping (5909, 5936)
4823 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4824 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4825 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4826 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4827 - support for RHEL4 kernel on i686 (5773)
4828 - provide error messages when incompatible logs are encountered (5898)
4830 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4831 * version 1.4.0.10 (1.4.1 release candidate 1)
4833 - don't keep a lock reference when lock is not granted (4238)
4834 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4835 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4836 - drop import inflight refcount on signal_completed_replay error (5255)
4837 - unlock page after async write error during send (3677)
4838 - handle missing objects in filter_preprw_read properly (5265)
4839 - no transno return for symlink open, don't save no-trasno open (3440)
4840 - don't try to complete elan receive that already failed (4012)
4841 - free RPC server reply state on error (5406)
4842 - clean up thread from ptlrpc_start_thread() on error (5160)
4843 - readahead could read extra page into cache that wasn't ejected (5388)
4844 - prevent races in class_attach/setup/cleanup/detach (5260)
4845 - don't dereference de->d_inode after l_dput of de (5458)
4846 - use "int" for stripe value returned from lock_to_stripe (5544)
4847 - mballoc allocation and error-checking fixes in 2.6 (5504)
4848 - block device patches to fix I/O request sizes in 2.6 (5482)
4849 - look up hostnames for IB nals (5602)
4850 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4851 - don't start multiple acceptors for the same port (5277)
4852 - fix incorrect LASSERT in mds_getattr_name (5635)
4853 - export a proc file for general "ping" checking (5628)
4854 - fix "lfs check" to not block when the MDS is down (5628)
4856 - service request history (4965)
4857 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4858 - create an "evict_client" /proc entry on OSTs, like the MDS has
4859 - fix mount usage message, return errors per mount(8) (5168)
4860 - change grep [] to grep "[]" in tests so they work in more UMLs
4861 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4862 - remove ancient LOV_MAGIC_V0 EA support (5047)
4863 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4864 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4865 - allow readahead allocations to fail when low on memory (5383)
4866 - mmap locking landed again, after considerable improvement (2828)
4867 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4869 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4872 - send OST transaction number in read/write reply to free req (4966)
4873 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4874 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4875 - the watchdog thread now runs as interruptible (5246)
4876 - flock/lockf fixes (but it's still disabled, pending 5135)
4877 - don't use EXT3 constants in llite code (5094)
4878 - memory shortage at startup could cause assertion (5176)
4880 - reorganization of lov code
4881 - single portals codebase
4883 - add extents/mballoc support (5025)
4884 - direct I/O reads in the obdfilter (4048)
4885 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4887 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4890 - send OST transaction number in read/write reply to free req (4966)
4891 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4892 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4893 - the watchdog thread now runs as interruptible (5246)
4894 - handle missing objects in filter_preprw_read properly (5265)
4895 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4896 - drop import inflight refcount on signal_completed_replay error (5255)
4897 - unlock page after async write error during send (3677)
4898 - return original error code on reconstructed replies (3761)
4899 - no transno return for symlink open, don't save no-trasno open (3440)
4901 - add pid to ldlm debugging output (4922)
4902 - bump the watchdog timeouts -- we can't handle 30sec yet
4903 - extra debugging for orphan dentry/inode bug (5259)
4905 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4908 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4909 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4910 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4911 - fix NULL dereference in /proc/sys/portals/routes (4827)
4912 - allow failed mdc_close() operations to be interrupted (4561)
4913 - stop precreate on OST before MDS would time out on it (4778)
4914 - don't send partial-page writes before EOF from client (4410)
4915 - discard client grant for sub-page writes on large-page clients (4520)
4916 - don't free dentries not owned by NFS code, check generation (4806)
4917 - fix lsm leak if mds_create_objects() fails (4801)
4918 - limit debug_daemon file size, always print CERROR messages (4789)
4919 - use transno after validating reply (3892)
4920 - process timed out requests if import state changes (3754)
4921 - update mtime on OST during writes, return in glimpse (4829)
4922 - add mkfsoptions to LDAP (4679)
4923 - use ->max_readahead method instead of zapping global ra (5039)
4924 - don't interrupt __l_wait_event() during strace
4926 - add software watchdogs to catch hung threads quickly (4941)
4927 - make lustrefs init script start after nfs is mounted
4928 - fix CWARN/ERROR duplication (4930)
4929 - return async write errors to application if possible (2248)
4930 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4931 - print NAL number in %x format (4645)
4932 - update barely-supported suse-2.4.21-171 series (4842)
4933 - support for sles 9 %post scripts
4934 - support for building 2.6 kernel-source packages
4935 - support for sles km_* packages
4937 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4940 - ignore -ENOENT errors in osc_destroy (3639)
4941 - notify osc create thread that OSC is being cleaned up (4600)
4942 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4943 - reconstruct ost_handle() like mds_handle() (4657)
4944 - create a new thread to do import eviction to avoid deadlock (3969)
4945 - let lconf resolve symlinked-to devices (4629)
4946 - don't unlink "objects" from directory with default EA (4554)
4947 - hold socknal file ref over connect in case target is down (4394)
4948 - allow more than 32000 subdirectories in a single directory (3244)
4949 - fix blocks count for O_DIRECT writes (3751)
4950 - OST returns ENOSPC from object create when no space left (4539)
4951 - don't send truncate RPC if file size isn't changing (4410)
4952 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4953 - bind to privileged port in socknal and tcpnal (3689)
4955 - rate limit CERROR/CWARN console message to avoid overload (4519)
4956 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4957 - basic mmap support (3918)
4958 - kernel patch series update from b1_4 (4711)
4960 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4963 - avoid crash during MDS cleanup with OST shut down (2775)
4964 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4965 - don't use bad inodes on the MDS (3744)
4966 - dynamic object preallocation to improve recovery speed (4236)
4967 - don't hold spinlock over lock dumping or change debug flags (4401)
4968 - don't zero obd_dev when it is force cleaned (3651)
4969 - print grants to console if they go negative (4431)
4970 - "lctl deactivate" will stop automatic recovery attempts (3406)
4971 - look for existing locks in ldlm_handle_enqueue() (3764)
4972 - don't resolve lock handle twice in recovery avoiding race (4401)
4973 - revalidate should check working dir is a directory (4134)
4975 - don't always mark "slow" obdfilter messages as errors (4418)
4977 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4980 - don't close LustreDB during write_conf until it is done (3860)
4981 - fix typo in lconf for_each_profile (3821)
4982 - allow dumping logs from multiple threads at one time (3820)
4983 - don't allow multiple threads in OSC recovery (3812)
4984 - fix debug_size parameters (3864)
4985 - fix mds_postrecov to initialize import for llog ctxt (3121)
4986 - replace config semaphore with spinlock (3306)
4987 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4988 - don't allow enqueue to complete on a destroyed export (3822)
4989 - down write_lock before checking llog header bitmap (3825)
4990 - recover from lock replay timeout (3764)
4991 - up llog sem before sending rpc (3652)
4992 - reduce ns lock hold times when setting kms (3267)
4993 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4994 - fix NULL deref and obd_dev leak on setup error (3312)
4995 - replace some LBUG about llog ops with error handling (3841)
4996 - don't match INVALID dentries from d_lookup and spin (3784)
4997 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4998 - fix invalid assertion in ptlrpc_set_wait (3880)
5000 - add libwrap support for the TCP acceptor (3996)
5001 - add /proc/sys/portals/routes for non-root route listing (3994)
5002 - allow setting MDS UUID in .xml (2580)
5003 - print the stack of a process that LBUGs (4228)
5005 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5008 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5009 - make sure to unset replay flag from failed open requests (3440)
5010 - if default stripe count is 0, use OST count for inode size (3636)
5011 - update parent mtime/ctime on client for create/unlink (2611)
5012 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5013 - free recovery state on server during a forced cleanup (3571)
5014 - unregister_reply for resent reqs (3063)
5015 - loop back devices mounting and status check on 2.6 (3563)
5016 - fix resource-creation race that can provoke i_size == 0 (3513)
5017 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5018 - more debugging for page-accounting assertion (3746)
5019 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5020 - avoid deadlock after precreation failure (3758)
5021 - fix race and lock order deadlock in orphan handling (3450, 3750)
5022 - add validity checks when grabbing inodes from l_ast_data (3599)
5024 - add /proc/.../recovery_status to obdfilter (3428)
5025 - lightweight CDEBUG infrastructure, debug daemon (3668)
5026 - change default OSC RPC parameters to be better on small clusters
5027 - turn off OST read cache for files smaller than 32MB
5028 - install man pages and include them in rpms (3100)
5029 - add new init script for (un)mounting lustre filesystems (2593)
5030 - run chkconfig in %post for init scripts (3701)
5031 - drop scimac NAL (unmaintained)
5033 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5036 - clean kiobufs before and after use (3485)
5037 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5038 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5039 - change init script to fail more clearly if not run as root (1528)
5040 - allow clients to reconnect during replay (1742)
5041 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5042 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5043 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5044 - fix cond_resched (3554)
5045 - only evict unfinished clients after recovery (3515)
5046 - allow bulk resend, prevent data loss (3570)
5047 - dynamic ptlrpc request buffer allocation (2102)
5048 - don't allow unlinking open directory if it isn't empty (2904)
5049 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5050 - remove extraneous obd dereference causing LASSERT failure (3334)
5051 - don't use get_cycles() when creating temp. files on the mds (3156)
5052 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5053 - handle EEXIST for set-stripe, set proper directory name (3336)
5055 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5056 - fix ksocknal_fmb_callback() error messages (2918)
5058 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5061 - don't copy lvb into (possibly NULL) reply on error (2983)
5062 - don't deref dentry after dput, don't free lvb on error (2922)
5063 - use the kms to determine writeback rpc length (2947)
5064 - increment oti_logcookies when osc is inactive (2948)
5065 - update client's i_blocks count via lvb messages (2543)
5066 - handle intent open/close of special files properly (1557)
5067 - mount MDS with errors=remount-ro, like obdfilter (2009)
5068 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5069 - don't use cancelling-locks' kms values (2947)
5070 - use highest lock extent for kms, not last one (2925)
5071 - don't dereference ERR_PTR() dentry in error handling path (3107)
5072 - fix thread race in portals_debug_dumplog() (3122)
5073 - create lprocfs device entries at setup instead of at attach (1519)
5074 - common AST error handler, don't evict client on completion race (3145)
5075 - zero nameidata in detach_mnt in 2.6 (3118)
5076 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5077 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5078 - initialize RPC timeout timer earlier for 2.6 (3219)
5079 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5080 - print nal/nid for unknown nid (3258)
5081 - additional checks for oscc recovery before doing precreate (3284)
5082 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5083 - don't crash in mdc_close for bad permissions on open (3285)
5084 - zero i_rdev for non-device files (3147)
5085 - clear page->private before handing to FS, better assertion (3119)
5086 - tune the read pipeline (3236)
5087 - fix incorrect decref of invalidated dentry (2350)
5088 - provide read-ahead stats and refine rpc in flight stats (3328)
5089 - don't hold journal transaction open across create RPC (3313)
5090 - update atime on MDS at close time (3265)
5091 - close LDAP connection when recovering to avoid server load (3315)
5092 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5093 - don't leak open file on MDS after open resend (3325)
5094 - serialize filter_precreate and filter_destroy_precreated (3329)
5095 - loop device shouldn't call sync_dev() for nul device (3092)
5096 - clear page cache after eviction (2766)
5097 - resynchronize MDS->OST in background (2824)
5098 - refuse to mount the same filesystem twice on same mountpoint (3394)
5099 - allow llmount to create routes for mounting behind routers (3320)
5100 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5101 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5102 - fix rare problem with rename on htree directories (3417)
5104 - allow default OST striping configuration per directory (1414)
5105 - fix compilation for qswnal for 2.6 kernels (3125)
5106 - increase maximum number of MDS request buffers for large systems
5107 - change liblustreapi to be useful for external progs like lfsck (3098)
5108 - increase local configuration timeout for slow disks (3353)
5109 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5111 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5114 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5115 - don't overwrite extent policy data in reply if lock was blocked (2901)
5116 - drop filter export grants atomically with removal from device (2663)
5117 - del obd_self_export from work_list in class_disconnect_exports (2908)
5118 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5119 - swab reply message in mdc_close, other PPC fixes (2464)
5120 - fix destroying of named logs (2325)
5121 - overwrite old logs when running lconf --write_conf (2264)
5122 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5123 - fix race in target_handle_connect (2898)
5124 - mds_reint_create() should take same inode create lock (2926)
5125 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5126 - don't close files for self_export to avoid uninitialized obd (2936)
5127 - allow MDS with the same name as client node (2939)
5128 - hold dentry reference for closed log files for unlink (2325)
5129 - reserve space for all logs during transactions (2059)
5130 - don't evict page beyond end of stripe extent (2925)
5131 - don't oops on a deleted current working directory (2399)
5132 - handle hard links to targets without a parent properly (2517)
5133 - don't dereference NULL lock when racing during eviction (2867)
5134 - don't grow lock extents when lots of conflicting locks (2919)
5136 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5139 - account for cache space usage on clients to avoid data loss (974)
5140 - lfsck support in lustre kernel code (2349)
5141 - reduce journal credits needed for BRW writes (2370)
5142 - orphan handling to avoid losing space on client/server crashes
5143 - ptlrpcd can be blocked, stopping ALL progress (2477)
5144 - use lock value blocks to assist in proper KMS, faster stat (1021)
5145 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5146 - recovery for initial connections (2355)
5147 - fixes for mds_cleanup_orphans (1934)
5148 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5149 - block all file creations until orphan recovery completes (1901)
5150 - client remove rq_connection from request struct (2423)
5151 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5152 - recovery timer race (2670)
5153 - mdc_close recovey bug (2532)
5154 - ptlrpc cleanup bug (2710)
5155 - mds timeout on local locks (2588)
5156 - namespace lock held during RPCs (2431)
5157 - handle interrupted sync write properly (2503)
5158 - don't try to handle a message that hasn't been replied to (2699)
5159 - client assert failure during cleanup after abort recovery (2701)
5160 - leak mdc device after failed mount (2712)
5161 - ptlrpc_check_set allows timedout requests to complete (2714)
5162 - wait for inflight reqs when ptlrpcd finishes (2710)
5163 - make sure unregistered services are removed from the srv_list
5164 - reset bulk XID's when resending them (caught by 1138 test)
5165 - unregister_bulk after timeout
5166 - fix lconf error (2694)
5167 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5168 - readahead locks pages, leaves pending causing memory pressure (2673)
5169 - increase OST request buffers to 4096 on large machines (2729)
5170 - fix up permission of existing directories in simple_mkdir (2661)
5171 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5172 - don't assign transno to errored transactions (2742)
5173 - don't delete objects on OST if given a bogus objid from MDS (2751)
5174 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5175 - if rq_no_resend, then timeout request after recovery (2432)
5176 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5177 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5178 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5179 - MDS recovery shouldn't delete objects that it has given out (2730)
5180 - if enqueue arrives after completion, don't clobber LVB (2819)
5181 - don't unlock pages twice when trigger_group_io returns error (2814)
5182 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5183 - don't write pages to disk if there was an error (1450)
5184 - don't ping imports that have recovery disabled (2676)
5185 - take buffered bytes into account when balancing socknal conn (2817)
5186 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5187 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5188 - remove little-endian swabbing of llog records (1987)
5189 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5190 - retry reposting request buffers if they fail (1191)
5191 - grow extent at grant time to avoid granting a revoked lock (2809)
5192 - lock revoke doesn't evict page if covered by a second lock (2765)
5193 - disable VM readahead to avoid reading outside lock extents (2805)
5195 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5196 - updated kernel patches for hp-2.4.20 kernel (2681)
5198 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5201 - fix truncated write corruption (2366)
5202 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5204 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5205 - ensure len(uuid) < 37 in lmc (1171)
5206 - fix ia64 OOPS in llog_test (2255)
5207 - zero end of page at obdfilter for partial page writes (2648)
5208 - don't leave stale dentries around after renames (bug 2428)
5209 - fix timeouts when evicting a client with a single lock held (2642)
5210 - set deadline for the initial HELLO message to drain (2634)
5211 - print out dotted-quad IP addresses in the socknal (2302)
5213 - additional debugging for MDS client eviction problem (2443)
5214 - fix mkfsoptions support for osts (2603, 2604)
5216 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5219 - add series for the vanilla 2.6.0 kernel
5220 - add series for the vanilla 2.4.24 kernel
5221 - add series for a cray x86/64 UL kernel drop
5222 - fix xattr patches for the vanilla 2.4.19 series
5224 - generate true UUIDs in lmc (1171)
5225 - have portals stack dumping break in UML (2466)
5226 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5227 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5228 - wake callback waiting threads on client eviction (2460)
5229 - Add --ptldebug and --subsystem to lmc (1719)
5230 - update assertion to allow safe interrupt allocation
5231 - set rq_no_resend for cancel requests (2432)
5232 - recalculate ptlrpcd timeout after resend (2494)
5233 - call vfs_rmdir when removing pending directories (2368)
5234 - fix renaming a file to itself (2429)
5235 - lmc creates a default one-stripe lov (2454)
5236 - expand procfs space to handle large clusters (2326)
5237 - increase UML stack to avoid overflow
5238 - update lconf's list of debug and subsystem masks
5239 - fix lfs find --obd (2510)
5240 - /proc tunable for disabling filter read caching (2591)
5241 - stop rpm packages from altering slapd.conf (2301)
5242 - disable nagle in the socknal under 0conf (2578)
5243 - choose mds inode size based on stripe count (2572)
5244 - fix kernel-source rpm problems (2516)
5246 - add --disable-doc to avoid pdf generation (2421)
5247 - update documentation, tests, type-os, comments
5248 - avoid format warnings on ia64
5249 - remove the TOE NAL
5250 - tiny code cleanups by removing unused fields
5252 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5255 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5256 - remove the most blatant lies from BUILDING file (2371)
5257 - change default debug level to reasonable production setting
5258 - reduce client side cache size to reduce cache flush time
5259 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5260 - make TCP zerocopy and pinger support enabled by default (2476)
5261 - sync writes completed after process exits caused crashes (2319)
5262 - maintain correct mount count on the MDS (2356)
5263 - backout 1557, because 2316 wasn't really fixed
5264 - better file I/O statistics gathering in /proc
5265 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5266 - another kernel patch to fix zero-copy TCP function export
5267 - don't take duplicate lock when processing re-sent getattr (2420)
5268 - lctl uses obd_self_export instead of creating new conn (2353)
5269 - MDS/OST recovery case which requires object creation asserted (2425)
5270 - move lfs from /usr/sbin to /usr/bin in packages
5271 - fix race between mds_client_add and mds_client_free (2417)
5272 - use kmalloc instead of slabs in portals (2430)
5273 - don't create duplicate records when a failover MDS is present (2442)
5274 - remove unnecessary mount age check (2332)
5275 - don't remove directory inodes from locks prematurely (2451)
5276 - don't break if MDS service name is the same as hostname (2103)
5277 - fix races in client write RPC generation when cache full (2482)
5279 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5282 - remove now-unused request->rq_obd (278)
5283 - if an allocation fails, print out how much memory we've used (1933)
5284 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5285 - add the "configurable stack size" patch to most series files (1256)
5286 - ability to write large log records, for 100+ OST configs (2306)
5287 - fix NULL deref when filter_prep fails (2314)
5288 - fix operator precedence error in filter_sync
5289 - dynamic allocation of socknal TX descriptors (2315)
5290 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5291 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5292 - ensure that CWARN messages in Portals always get to the syslog
5293 - __init/__exit are not for prototype decls (ldlm_init/exit)
5294 - x86-64 compile warning fixes
5295 - fix gateway LMC keyword conflict (2318)
5296 - fix MDS lock inversions in getattr/reint paths (1844)
5297 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5298 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5299 - DLM race condition prevented some lock evictions (2328)
5300 - ENOMEM detection and retry on socknal sends (2230)
5301 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5302 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5303 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5304 - better error messages when a client is rejected during recovery (1505)
5305 - avoid cancelling locks which were never granted, after failure (2330)
5306 - fix i_sem/journal inversion in mds_client_add (2333)
5307 - fix truncate/getattr lock cycle deadlock (2334)
5308 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5309 - fix two rare exit paths which could leak an l_lock() ref (2321)
5310 - fencepost error in MDS/OST orphan recovery (2226)
5311 - make log record alignment 8 bytes (1988)
5312 - lstripe now fails when requested offset > ost_count (2237)
5313 - ensure that all kernel series have a complete list.h (1607)
5314 - fix crashes in special-file operations (2316)
5315 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5317 - allow configurable automake binary, for testing new versions
5318 - small update to the lfs documentation
5320 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5322 * fix negative export reference count in fsfilt_sync (2312)
5324 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5325 * release candidate 0.9.1
5327 - orphans are moved into the PENDING directory for possible recovery
5328 - replayed opens now open by fid for orphan/rename safety (1042)
5329 - last close of an orphan inode generates a transno (683)
5330 - chdir() and mount() now pin the directory entry (1020)
5331 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5332 - discard very old requests without processing them (1502)
5333 - remove some common, well-understood CERRORs (1505)
5334 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5335 - clear "grant" flags in OST replies until OST grant code lands (1644)
5336 - fix read performance by not clobbering i_blksize on client (1598)
5337 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5338 - make LRU size a /proc tunable, clears locks when reduced (707)
5339 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5340 - server locks take references on exports now (1558)
5341 - build fixes for 2.4.20-rh trees (1663)
5342 - return an error from lov_create if all OSCs are inactive (1751)
5343 - fix import levels when a reconnect happens without a timeout (1597)
5344 - exit early from mds_open if we get a lookup error (1749)
5345 - partial page read at EOF wouldn't wait for disk before sending (1642)
5346 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5347 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5348 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5349 - write/truncate lock inversion (1639)
5350 - Don't auto-load obdclass, portals modules during cleanup (1495)
5351 - fix timestamps from jumping to "now" (1763)
5352 - extra journal assertions (1648)
5353 - add an extra multiunlink test (1771)
5354 - fix read_record/write_record API (1776)
5355 - fix leak of offset_extent, possible incorrect i_size later (1772)
5356 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5357 - Debugging for the kqswnal_get_idle_tx problems (1820)
5358 - Allow recovery to be attempted multiple times (1536)
5359 - Write out MDS last_rcvd file after it is first created (1600)
5360 - Fix tx_descriptor leak in failed transmit situations (1827)
5361 - ext3 journaling fixes for assertion failure after IO error (1871)
5362 - class_export_put() on freed export after completion AST error (1896)
5363 - Fix revalidate looping in VFS (1322)
5364 - Don't access a freed export during MDS_REINT timeout (1521)
5365 - Add open-unlink recovery support on the MDS (1673,1764)
5366 - Return an error if no MDS data was read from last_rcvd (1946)
5367 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5368 - Don't setup a disk device that doesn't match exported UUID (317)
5369 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5370 - avoid committing NULL handle in force close
5371 - local.sh is now a one-stripe LOV configuration
5372 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5373 - fix ext3 htree duplicate directory entry corruption (1516)
5374 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5375 - update to new LOV EA format (2097)
5376 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5378 2003-06-15 Phil Schwan <phil@clusterfs.com>
5381 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5382 - per-import/export recovery handling (958, 931, 959)
5383 - multiple last-rcvd slots, for serving multiple FSes (949)
5384 - connections are again shared between multiple imp/exports (963, 964)
5385 - "umount -f" would hang if any requests needed to be sent (393, 978)
5386 - avoid pinning large req buffer by copying for queued messages (989)
5387 - add "uuid" to "lctl device" command to help upcalls (991)
5388 - "open" RPCs with transnos would confuse recovery counters (1037)
5389 - do proper endian conversion of all wire messages (288, 340, 891)
5390 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5391 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5392 - fix ptlrpc_connection leak in target_handle_connect (1174)
5393 - fix import refcounting bug in OST and MDS cleanup (1134)
5394 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5395 - fix very unlikely obd_types race condition (501)
5396 - remove osc_open hack for echo_client (1187)
5397 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5398 - a failure in read_inode2 leads to deadlock (1139)
5399 - cancel ack-locks as soon as transaction is committed (1072)
5400 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5401 - make sure to commitrw after any preprw to avoid deadlock (1162)
5402 - failing to execute a file in a lustre FS would lock inode (1203)
5403 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5404 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5405 - obd_uuid2tgt would walk off the end of the list (1255)
5406 - on IA64 the osc would give portals incorrect bulk size (1258)
5407 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5408 - fix lock inversion caused by new llite matching code (1282)
5409 - limit the number of dirty pages on a client to 10MB (1286)
5410 - timed out locks were not being corrected cancelled (1289)
5411 - fix O_DIRECT above 4GB on IA-32 (1292)
5412 * major user-visible changes
5413 - fail out/fail over policy now controlled by the upcall (993)
5415 - add OBD_PING to check server availability and failure (954)
5416 - lustre messages are now sent in sending host order (288, 340, 891)
5417 - add eadatalen to MDS getattr reply (340)
5418 - OST read replies may contain second buffer, with per-page status (593)
5420 2003-03-11 Phil Schwan <phil@clusterfs.com>
5423 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5424 - failed open()s could cause deadlock; fixed (867, 869)
5425 - stop cancelling OST locks when files are closed (481)
5426 - overlapping XID spaces caused network corruption (851, 853)
5427 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5428 - setattr_raw added, to do single-RPC, server-side setattrs
5429 - lmc/lconf syntax change for OST UUIDs
5430 - fix crashy race condition between ptlrpc_free_req and osc_close
5431 - don't use request in mdc_enqueue if we hit a timeout (889)
5432 - don't set the inode i_size for regular files from the MDS (896)
5433 - handle out of order completion AST (842)
5434 - don't LBUG if a lock request times out after receiving AST (913)
5435 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5436 - if a bad lock AST arrives, send an error instead of dropping entirely
5437 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5438 - fix leak in bulk IO when only partially completed (899, 900, 926)
5439 - fix O_DIRECT for ia64 (55)
5440 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5441 - C-z after timeout could hang a process forever; fixed (977)
5443 - client-side I/O cache (678, 924, 929, 941, 970)
5445 - READPAGE and SETATTRs which don't take server-side locks get
5448 2003-02-11 Phil Schwan <phil@clusterfs.com>
5451 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5452 - Fix fsfilt_extN_readpage() to read a full page of directory
5453 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5454 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5455 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5456 - Most intent operations were removed, in favour of a new RPC mode
5457 that does a single RPC to the server and bypasses most of the VFS
5458 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5459 - Aggressively cancel local locks on DLM servers
5460 - mds_reint_unlink sends EA to the client if it's the last nlink.
5461 client uses that EA to unlink OST objects.
5462 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5463 - recursive symlinks were fixed (439)
5464 - fixed NULL deref in DEBUG_REQ
5465 - filter_update_lastobjid no longer calls sync, which annoyed extN
5466 - fixed multi-client small-writes to a single file problem (445)
5467 - fixed mtime updates during file writes (607)
5468 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5469 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5470 - fixed Linux OST size reporting problem (444, 656)
5471 - OST now updates object mtime with writes or setattr (607, 619)
5472 - client verifies file size before zeroing page past EOF (445)
5473 - OST now writes last allocated objid to disk with allocation (108)
5474 - LOV on echo now works (409)
5476 - mds_reint_unlink sends a new buffer, with the EA included. this
5477 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5478 set if a regular file was being unlinked, and it was the last link
5479 - use PtlGet from the target for bulk writes (315)
5480 - OST now updates object mtime with writes or setattr (607, 619)
5481 - LDLM now has a grant-time callback to revalidate locked items, if
5483 - Many MDS operations were reorganized to combat race conditions
5485 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5486 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5488 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5491 - Fully reactivate OST imports after reconnection (512, others)
5492 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5493 - More graceful error handling for truncating on dead OST (515)
5494 - Don't error out unless we're actually accessing dead stripes (474)
5495 - Fix garbage sizes when stripes are missing (410)
5496 - LRU counters were broken, causing constant lock purge (433, 432)
5497 - garbage on read from stripes with failed OSTs (441)
5498 - mark OSCs as active before reconnecting during recovery (438)
5499 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5500 - lfind did not preserve OST order in output (443)
5501 - symlinks cause hung clients, incorrect data (439)
5502 - stop dereferencing request after dropping refcount (457)
5503 - don't LASSERT(spin_is_locked) on non-SMP (455)
5504 - fixes for many rename() bugs
5505 - fstat didn't correctly synchronize attributes (399)
5506 - server must handle lock cancellation during blocking AST prep (487)
5507 - bulk descriptors were free()d too soon (511)
5508 - fix paths in lconf, which would load incorrect modules (451, 507)
5509 - fix confusing lconf 'host not found' error message (386)
5510 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5511 - fix race condition in mdc_blocking_ast() for inode access (526)
5512 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5513 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5514 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5515 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5516 - fix lstripe in several ways (kernel side) (527)
5517 - fix request leak in ldlm_cli_enqueue (262)
5518 - incorrect OSC was marked inactive after OST failure
5519 - call mds_fs_cleanup before unmounting filesystem (524)
5520 - fix races between taking ns_lock and ldlm_lock_change_resource
5521 - fix races updating LOV export open file list
5522 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5523 - fix recovery NULL deref in ldlm_cli_cancel_unused
5524 - fix some DLM races by using new hash table for lock handles (419)
5525 - permit the client to specify desired inodes, at replay
5526 - duplicate requests when we queue them for replay reintegration
5527 - fix last_rcvd offset calculation
5528 - sync after each recovered transaction, so we always make progress
5529 - never, not always, ERESTART requests without transnos
5530 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5532 - skip replay if the MDS says that the client is already connected
5533 - don't check for a recovery-enabled export to match lctl's UUID
5534 - don't INC_USE_COUNT for phantom exports
5535 - don't crash when cleaning up phantom exports (567)
5536 - don't double-finish or set replay data for errored mdc_open requests
5537 - abort requests when they time out, so we don't get old replies
5538 - send/receive replies for AST messages again
5539 - if the client says that it doesn't have the lock, cancel it on the
5541 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5542 mark it as destroyed, it will all work out when decref is called
5543 - fix module use counts (22, 581)
5545 - ASTs now expect a reply (server cancels lock on error reply)
5547 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5550 - fix many simultaneous client startup (392)
5551 - fix dentry->d_it clobbering
5552 - credentials weren't being shipped for readdir/getattr operations
5553 - remove invalid assertions triggered during some concurrent MD
5555 - proper Lustre versions added (336, 389)
5556 - fix memory leak for create error case (398)
5557 - fix LOV locking bug that would get cli/srv out of sync
5558 - fix echo client over LOV (409)
5559 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5560 - fix double-O_EXCL intent crash (424)
5561 - avoid sending multiple lock CANCELs (352)
5563 - MDS can do multi-client recovery (modulo bugs in new code)
5565 - many updates, edits, cleanups
5567 2002-11-18 Phil Schwan <phil@clusterfs.com>
5570 - fix null d_it dereference (346)
5571 - fix full OST/dbench hang (333)
5572 - fix permission problem with file removal (286)
5573 - fix removal of OSCs from LOV when they fail
5574 - fix NULL deref during bulk timeout (214)
5575 - fix problems related to multiple filesystems on one MDS (241)
5576 - fixed serious subtle metadata locking bugs
5577 - free locks on clients when inodes are removed due to memory
5579 - fix inode pointer in lock data (285)
5580 - partial support for multiple MDS on a single host (241)
5581 - data locks weren't cancelled at clear_inode time (290, 311)
5582 - intent locks could lead to unbounded lock growth (205)
5583 - added a maximum lock count, an LRU list, and a flusher
5584 - fix multiple rename (365)
5585 - properly abstracted the echo client
5586 - OSC locked 1 byte too many; fixed
5587 - rewrote brw callback code:
5588 - fixed recovery bugs related to LOVs (306)
5589 - fixed too-many-pages-in-one-write crash (191)
5590 - fixed (again) crash in sync_io_timeout (214)
5591 - probably fixed callback-related race (385)
5593 - Add capability to MDS protocol
5594 - LDLM cancellations and callbacks on different portals
5596 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5599 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5600 - defer open object destruction to close time (601981, 138)
5601 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5602 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5603 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5604 - don't call obd_setattr in ll_file_release for destroyed objects
5605 * protocol change to lustre_msg: move |version| and add |flags|
5606 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5607 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5608 * added replay of create, unlink, link and rename operations during
5609 MDS failover; recovery should be much more robust now
5610 * remove failed OSCs from LOVs (only lov_create uses this so far)
5611 * the lustre-HOWTO was brought (more) up to date (582544)
5613 2002-10-23 Phil Schwan <phil@clusterfs.com>
5616 - in-use dentries weren't being reused properly (617851)
5617 - prevent multiple LDLM setup (599178)
5618 - fix LOV size calculations for truncate (617853)
5619 - fix client handling of MDS intent errors (POSIX)
5620 - fix permission bug in lovstripe.c test (624321)
5621 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5622 - truncate past end of file could corrupt data
5623 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5624 - a race in recovery could return ETIMEDOUT to apps (623947)
5625 - building outside the source directory was fixed
5626 * the lustre-HOWTO was brought (more) up to date (582544)
5627 * major progress was made on recovery functionality
5629 2002-10-10 Phil Schwan <phil@clusterfs.com>
5632 - recovery deadlock fix
5633 - rm -rf causes LBUG fix (617817)
5634 - file open by multiple tasks fix (618962)
5635 - directory permissions bugs (602707 and 620007)
5636 - journal_stop fixed with locking (611313)
5637 - O_APPEND failures resolved (618273, perhaps 614459)
5638 - lconf PATH fix (619770)
5639 - IA64 build fix (621450)
5640 - RPC buffer sizes scale with amount of memory
5642 2002-10-01 Phil Schwan <phil@clusterfs.com>
5645 - locks would be cancelled without throwing away data pages,
5646 resulting in inconsistent data (605627)
5647 - inode attributes were not always being refreshed (605627, 612449)
5648 - lconf now continues to cleanup after lctl reports an error
5649 - MDS now enforces user permissions (602707)
5650 - lprocfs cleanup fixed, but not yet enabled (614157)
5651 - fixed infinite server hang, should a client not respond to an AST
5652 - avoid going into recovery if user calls readlink() with a buffer
5653 that's too small (613941)
5654 - AST RPCs no longer require replies (614867) -- this may be changed
5655 - don't crash server if client sends an IOV that's too big (611336)
5656 - fixed lock conversion deadlock (611892)
5657 - fixed the following of symlinks (614622)
5658 * recovery: the server can remove locks from a client that dies, other
5659 clients can make progress
5660 * more extN patch fixes
5661 * compile-time configurable ptlrpc buffer allocations
5663 - collaborative read cache document
5664 - Lustre Lite Performance CDR document-in-progress
5666 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5669 - fix typo in patch-2.4.18
5671 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5674 - clear ptlrpc request each time in handle_incoming_request()
5675 - unlink of files now destroys the object on the OST
5677 2002-09-19 Peter Braam <braam@clusterfs.com>
5679 * add hard link support
5680 * change obdfile creation method
5681 * kernel patch changed
5683 2002-09-19 Peter Braam <braam@clusterfs.com>
5686 - stack overflow bug in extN fixed
5688 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5690 * documentation updates
5691 - add man pages for config tools
5692 - update tests/README to describe testing with new config tools
5693 - finish metadata API descriptions
5694 * bug fixes and cleanups
5695 - statfs workaround for 16TB limit
5696 - LOV stripe allocation improved, can stripe on subset of OSTs
5697 - LOV file size/IO offset was wrong for files > 4GB in size
5698 - object EA data was being dropped, caused files to be unreadable
5699 - memory overflow with non-LOV OST caused memory corruption
5700 - fixed regression tests to work with new config tools, obdfilter
5701 - fixed bug when directory size became larger than 1 block
5702 - fixed bug (for single client case) when PWD was deleted
5703 - invalidate local directory pages when doing intent-based ops
5704 - avoid LDLM oops when lock callback contained bad data
5706 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5708 * documentation updates
5709 * bug fixes and cleanups
5710 - configuration tools
5713 - 64-bit compile warnings
5714 - 64-bit internal statfs data
5716 * test_brw on persistent OST devices
5718 * lprocfs (disabled)
5720 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5722 * documentation updates
5723 * bug fixes and cleanups
5724 * configuration tools
5726 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5729 * many small fixes to 0_5_4
5730 * io/network handling
5731 * thinkos in MDS operations
5733 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5736 * crucial basic fixes to 0.5.3
5737 * IOR, Iozone work over Elan
5740 2002-08-07 Phil Schwan <phil@clusterfs.com>
5741 * version 0_5_3, our first alpha
5742 * we use the new Portals iovs
5743 * documentation updates
5744 * bug fixes and cleanups
5745 * small changes in the DLM wire protocol
5747 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5748 * version 0_5_1 with some initial stability,
5749 * locking on MD and file I/O.
5750 * documentation updates
5751 * several bug fixes since 0.5.0
5752 * small changes in wire protocol
5754 2002-07-18 Phil Schwan <phil@clusterfs.com>
5756 * delivered as Lustre Light Alpha
5757 * fixed a crash after handling invalid MDS requests
5758 * fixed directory pages for architectures with non-4k pages sizes
5760 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5761 * release version v0_4_4
5762 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5764 * Fixes a number of interruption problems with OST operations.
5765 * Update documentation for portals header changes
5766 * Move all wire protocol structs/defines to lustre_idl.h
5767 * Fixes symlink length bug.
5768 * Add tcpdump to repository.
5770 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5771 * release version v0_4_3
5772 * Fixes statfs for inodes on extN.
5773 * Fixes bug in runtests which would delete /etc/hosts.
5774 * Use 64-bit object IDs wherever possible (not into VFS though)
5775 Remove ost_get_info, which is unused by lustre, and out of date.
5777 2002-07-03 Peter Braam <braam@clusterfs.com>
5778 * release version v0_4_2 Fixes a lookup error (type not passed)
5779 * move forward to head of Portals
5780 * move forward to latest Lustre kernel
5782 2002-06-25 Peter Braam <braam@clusterfs.com>
5783 * release version v0_4_1. Hopefully stable on single node use.