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: 1.41.5.sun2
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.
18 Description: resolve race between obd_disconnect and class_disconnect_exports
19 Details : if obd_disconnect will be called to already disconnected export he
20 forget release one reference and osc module can't unloaded.
22 Severity : enhancement
24 Description: Allow tuning service thread via /proc
25 Details : For each service a new
26 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
27 created that can be used to set min/max thread counts, and get the
28 current number of running threads.
32 Descriptoin: don't return error if have particaly created objects for file.
33 Details : lov_update_create_set uses set->set_success as index for created objects,
34 so if some requests will be failed, they will have hole at end of
35 array and we can use qos_shrink_lsm for allocate correct lsm.
37 Severity : enhancement
39 Description: Update OFED support to 1.4.1
43 Description: Reduce small size read RPC
44 Details : Set read-ahead limite for every file and only do read-ahead when
45 available read-ahead pages are bigger than 1M to avoid small size
48 Severity : enhancement
50 Description: move AT tunable parameters for more consistent usage
51 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
53 Severity : enhancement
55 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
56 Details : allow skip disconnected ost for send statfs request and hide error
62 Description: fix racy locking of mballoc block bitmaps causing BUG
63 Details : The locking of the mballoc buddy bitmap and the in-memory
64 block bitmap was using two different spin locks in some
65 cases. This made it possible to incorrectly access the
66 mballoc bitmap while another process was modifying it,
67 causing a sanity assertion to fail. While no on-disk corruption
68 was reported, there was some risk of this happening.
71 Frequency : rare, on llog test 6
73 Descriptoin: don't allow connect to already connected import
74 Details : allowing connect to already connected import is hide connecting problem.
79 Descriptoin: allow kill process which wait statahead result
80 Details : for some reasons 'ls' can stick in waiting result from statahead,
81 in this case need way for kill this process.
83 Severity : enhancement
85 Description: Add state history info file, enhance import info file
86 Details : Track import connection state changes in a new osc/mdc proc file;
87 add overview-type data to the osc/mdc import proc file.
89 Severity : enhancement
91 Description: MDS create should not wait for statfs RPC while holding DLM lock.
93 Severity : enhancement
95 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
100 Descriptoin: OST grant too much space to client even there are not enough space.
101 Details : Client will shrink its grant cache to OST if there are no write
102 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
103 this grant cache if there are already not enough avaible space
104 (left_space < total_clients * 32M).
107 Frequency : start MDS on uncleanly shutdowned MDS device
109 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
110 Details : stay in waiting mds<>ost recovery finished produce random bugs
111 due race between two ll_sync thread for one lov target. send
112 ACTIVATE event only if connect realy finished and import have
116 Frequency : rare, connect and disconnect target at same time
118 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
119 Details : don't call obd_disconnect under lov_lock. this long time
120 operation and can block ptlrpcd which answer to connect request.
125 Descriptoin: don't lose wakeup for imp_recovery_waitq
126 Details : recover_import_no_retry or invalidate_import and import_close can
127 both sleep on imp_recovery_waitq, but we was send only one wakeup
131 Frequency : always with long access acl
133 Descriptoin: mds can't pack reply with long acl.
134 Details : mds don't control size of acl but they limited by reint/getattr
137 Severity : enhancement
139 Description: Update to SLES10 kernel-2.6.16.60-0.33.
141 Severity : enhancement
143 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
146 Frequency : start MDS on uncleanly shutdowned MDS device
148 Descriptoin: aborting recovery hang on MDS
149 Details : don't throttle destroy RPCs for the MDT.
152 Frequency : on remount
154 Description: external journal device not working after the remount
155 Details : clear dev_rdonly flag for external journal devices in
161 Description: shutdown vs evict race
162 Details : client_disconnect_export vs connect request race.
163 if client will evicted at this time - we start invalidate
164 thread without referece to import and import can be freed
168 Frequency : rare, need acl's on inode.
170 Description: client can't handle ost additional correctly
171 Details : if ost was added after client connected to mds client can have
172 hit lnet_try_match_md ... to big messages to wide striped files.
173 in this case need teach client to handle config events about add
174 lov target and update client max ea size at that event.
176 Severity : enhancement
178 Description: Changelogs
179 Details : Changelogs are a lightweight mechanism to track filesystem
180 metadata and namespace changes. The changelog is recorded
181 permanently on the MDTs, and is periodically "consumed" / purged
182 when records are no longer needed.
184 Severity : enhancement
186 Description: compact fld format with extents
187 Details : Store range of seq rather than every seq in FLD. Seq
188 controller update FLD rather than clients. In Case of CMD, mdt0
189 has FLD, all other metadata server act as non persistent proxy
190 for FLD queries and cache fld entries in fld cache.
195 Description: don't skip ost target if they assigned to file
196 Details : Drop slow OSCs if we can, but not for requested start idx.
197 This means "if OSC is slow and it is not the requested
198 start OST, then it can be skipped, otherwise skip it only
199 if it is inactive/recovering/out-of-space.
203 Description: more cleanup in mds_lov
204 Details : not send LOV EA under replay, we can't know about they size at this
205 time. Don't allow client connect to mds before any ost connected,
206 for avoid problems with LOV EA size and returning EIO to client.
208 Severity : enhancement
210 Description: Interoperability at server side (Disk interoperability)
212 Severity : enhancement
214 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
216 Severity : enhancement
218 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
220 Severity : enhancement
222 Description: New client IO stack (CLIO).
224 Severity : enhancement
226 Description: Commit on sharing. Eliminate inter-client dependencies between
227 uncommitted transactions by doing transaction commits.
228 Thereby clients may recovery independently.
231 Frequency : Create a symlink file with a very long name
233 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
234 Details : If there is no extra space in the request for early cancels,
235 ldlm_req_handles_avail() returns 0 instead of a negative value.
237 Severity : enhancement
239 Description: Add /proc entry for import status
240 Details : The mdc, osc, and mgc import directories now have
241 an import directory that contains useful import data for debugging
244 Severity : enhancement
246 Description: Re-disable certain /proc logging
247 Details : Enable and disable client's offset_stats, extents_stats and
248 extents_stats_per_process stats logging on the fly.
251 Frequency : Only on FC kernels 2.6.22+
253 Description: oops in statahead
254 Details : Do not drop reference count for the dentry from VFS when lookup,
255 VFS will do that by itself.
257 Severity : enhancement
259 Description: Generic /proc file permissions
260 Details : Set /Proc file permissions in a more generic way to enable non-
261 root users operate on some /proc files.
265 Description: Hitting mdc_commit_close() ASSERTION
266 Details : Properly handle request reference release in
267 ll_release_openhandle().
271 Description: quota recovery deadlock during mds failover
272 Details : This patch includes att18982, att18236, att18237 in bz14840.
274 1. fix osts hang when mds does failover with quotaon
275 2. prevent watchdog storm when osts threads wait for the
280 Frequency : only patchless client
281 Description: add workaround for race between add/remove dentry from hash
283 Severity : enhancement
285 Description: Allow OST glimpses to return PW locks
289 Description: LBUG when llog conf file is full
290 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
294 Description: Prevent import from entering FULL state when server in recovery
298 Description: service mount cannot take device name with ":"
299 Details : Only when device name contains ":/" will mount treat it as
305 Description: replace ptlrpcd with the statahead thread to interpret the async
306 statahead RPC callback
310 Frequency : on recovery
311 Description: I/O failures after umount during fail back
312 Details : if client reconnected to restarted server we need join to recovery
313 instead of find server handler is changed and process self
314 eviction with cancel all locks.
316 Severity : enhancement
318 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
320 Severity : enhancement
322 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
324 Severity : enhancement
326 Description: Upcall on Lustre log has been dumped
327 Details : Allow for a user mode script to be called once a Lustre log has
328 been dumped. It passes the filename of the dumped log to the
329 script, the location of the script can be specified via
330 /proc/sys/lnet/debug_log_upcall.
335 Description: avoid idr_remove called for id which is not allocated.
336 Details : Move assigment s_dev for clustered nfs to end of initialization,
337 to avoid problem with error handling.
342 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
343 Details : When connection is reused this not moved from CONN_UNUSED_HASH
344 into CONN_USED_HASH and this prodice warning when put connection
345 again in unused hash.
348 Severity : enhancement
350 Description: File striping can now be set to use an arbitrary pool of OSTs.
352 Severity : enhancement
354 Description: Export bytes_read/bytes_write count on OSC/OST.
358 Description: Early reply size mismatch, MGC loses connection
359 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
360 the connect flags are properly negotiated.
365 Description: quotas are not honored with O_DIRECT
366 Details : all writes with the flag O_DIRECT will use grants which leads to
367 this problem. Now using OBD_BRW_SYNC to guard this.
371 Description: add quota statistics
372 Details : 1. sort out quota proc entries and proc code.
373 2. add quota statistics
375 Severity : enhancement
377 Description: enable quota support for HEAD.
381 Description: Properly propagate oinfo flags from lov to osc for statfs
382 Details : restore missing copy oi_flags to lov requests.
384 Severity : enhancement
386 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
387 : l_getgroups(8), lst(8), routerstat(8)
388 Details : included man pages for llobdstat(8), llstat(8),
389 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
391 Severity : enhancement
393 Description: configure's --enable-quota should check the
394 : kernel .config for CONFIG_QUOTA
395 Details : configure is terminated if --enable-quota is passed but
396 : no quota support is in kernel
400 Description: Remove portals compatibility
401 Details : Remove portals compatibility, not interoperable with releases
406 Description: Resolve device initialization race
407 Details : Prevent proc handler from accessing devices added to the
408 obd_devs array but yet be intialized.
410 Severity : enhancement
412 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
414 Severity : enhancement
416 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
421 Description: Using wrong pointer in osc_brw_prep_request
422 Details : Access to array[-1] can produce panic if kernel compiled with
423 CONFIG_PAGE_ALLOC enabled
427 Description: Client runs out of low memory
428 Details : Consider only lowmem when counting initial number of llap pages
432 Description: *optional* service tags registration
433 Details : if the "service tags" package is installed on a Lustre node
434 When the filesystem is mounted, a local-node service tag will
435 be created. See http://inventory.sun.com/ for more information
436 about the Service Tags asset management system.
440 Description: Kernel BUG tries to release flock
441 Details : Lustre does not destroy flock lock before last reference goes
442 away. So always drop flock locks when client is evicted and
443 perform unlock regardless of successfulness of speaking to MDS.
447 Description: add refcount protection for osc callbacks, avoid panic on shutdown
451 Description: sanity test 65a fails if stripecount of -1 is set
452 Details : handle -1 striping on filesystem in ll_dirstripe_verify
457 Description: ASSERTION(CheckWriteback(page,cmd)) failed
458 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
461 Severity : enhancement
463 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
467 Description: do not process already freed flock
468 Details : flock can possibly be freed by another thread before it reaches
469 to ldlm_flock_completion_ast.
473 Description: LBUG during stress test
474 Details : Need properly lock accesses the flock deadlock detection list.
478 Description: oops in page fault handler
479 Details : kernel page fault handler can return two special 'pages' in error
480 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
484 Description: timeout with invalidate import.
485 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
486 should be handled by ptlrpcd. This produce long age waiting and
487 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
489 Severity : enhancement
491 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
495 Description: enable MGS and MDT services start separately
496 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
497 a co-located MGS without starting the MGS, which is a complement
498 to 'nosvc' mount option.
502 Frequency : after recovery
503 Description: precreate to many object's after del orphan.
504 Details : del orphan st in oscc last_id == next_id and this triger growing
505 count of precreated objects. Set flag LOW to skip increase count
506 of precreated objects.
510 Frequency : rare, on clear nid stats
511 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
512 Details : when clean nid stats sometimes try destroy live entry,
513 and this produce panic in free.
517 Description: Stack overflow during MDS log replay
518 ease stack pressure by using a thread dealing llog_process.
522 Description: wait until IO finished before start new when do lock cancel.
523 Details : VM protocol want old IO finished before start new, in this case
524 need wait until PG_writeback is cleared until check dirty flag and
525 call writepages in lock cancel callback.
527 Severity : enhancement
529 Description: using special macro for print time and cleanup in includes.
533 Description: mds_mfd_close() ASSERTION(rc == 0)
534 Details : In mds_mfd_close(), we need protect inode's writecount change
535 within its orphan write semaphore to prevent possible races.
539 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
543 Frequency : rare, on shutdown ost
544 Description: don't hit live lock with umount ost.
545 Details : shrink_dcache_parent can be in long loop with destroy dentries,
546 use shrink_dcache_sb instead.
550 Description: don't panic with use echo client
551 Details : echo client pass NULL as client nid pointer and this produce null
556 Description: fix build on ppc32
557 Details : compile code with -m64 flag produce wrong object file for ppc32.
561 Description: add message levels for liblustreapi
565 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
566 Details : Move the 'good_osts' check before the 'total_bavail' check. This
567 will result in an -EAGAIN and in the exit call path we call
568 alloc_rr() which will with increasing aggressiveness attempt to
569 aquire precreated objects on the minimum number of required OSCs.
573 Description: Use old size assignment to avoid deadlock
574 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
575 the scheduling while holding a spinlock. We do not need locking
576 for size in ll_update_inode() because size is only updated from
577 the MDS for directories or files without objects, so there is no
578 other place to do the update, and concurrent access to such inodes
579 are protected by the inode lock.
583 Description: resolve "_IOWR redefined" build error on SLES10
587 Description: dump the memory debugging after all modules are unloaded to
588 suppress false negative in conf_sanity test 39
590 Severity : enhancement
592 Description: build kernel-ib packages for OFED 1.3 in our release cycle
597 Description: fix SLES kernel versioning
598 Details : the kernel version for our SLES 10 kernel did not include a "-"
599 before the "smp" at the end. while this was not a problem in
600 general, it did mean that software trying to use the kernel
601 version to try to detect a vendor specific kernel would fail.
602 this was most evident by the OFED build scripts.
606 Description: Don't update lov_desc members until making sure they are valid
607 Details : When updating lov_desc members via proc fs, need fix their
608 validities before doing the real update.
612 Description: don't put request into delay list while invalidate in flight.
613 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
614 import in flight. this produce timeout for invalidate and sometimes
615 can cause stale data.
619 Frequency : on ppc only
620 Description: not convert ost objects for directory because it's not exist.
621 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
625 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
627 Severity : enhancement
629 Description: Update to support 2.6.22.14 vanilla kernel.
633 Frequency : rare, on recovery
634 Description: read procfs can produce deadlock in some situation
635 Details : Holding lprocfs lock which send rpc can produce block for destroy
636 obd objects and this also block reconnect with -EALREADY. This
637 isn't fix all lprocfs bugs - but make it rare.
639 Severity : enhancement
641 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
644 Frequency : frequent on X2 node
646 Description: mdc_set_open_replay_data LBUG
647 Details : Set replay data for requests that are eligible for replay.
651 Description: lustre_mgs: operation 101 on unconnected MGS
652 Details : When MGC is disconnected from MGS long enough, MGS will evict the
653 MGC, and late on MGC cannot successfully connect to MGS and a lot
654 of the error messages complaining that MGS is not connected.
658 Frequency : on network error
659 Description: panic with double free request if network error
660 Details : mdc_finish_enqueue is finish request if any network error ocuring,
661 but it's true only for synchronus enqueue, for async enqueue
662 (via ptlrpcd) this incorrect and ptlrpcd want finish request
665 Severity : enhancement
667 Description: client-side metadata stat-ahead during readdir(directory readahead)
668 Details : perform client-side metadata stat-ahead when the client detects
669 readdir and sequential stat of dir entries therein
672 Frequency : on start mds
674 Description: Implement get_info(last_id) in obdfilter.
677 Frequency : occasional
679 Description: Correctly check stale fid, not start epoch if ost not support SOM
680 Details : open with flag O_CREATE need set old fid in op_fid3 because
681 op_fid2 was overwritten with new generated fid, but mds can answer
682 with one of these two fids and both is not stale. Setattr
683 incorrectly started an epoch and assume will be called
684 done_writing, but without SOM done_writing ever being called.
687 Frequency : rare, depends on device drivers and load
689 Description: MDS or OSS nodes crash due to stack overflow
690 Details : Code changes in 1.8.0 increased the stack usage of some functions.
691 In some cases, in conjunction with device drivers that use a lot
692 of stack the MDS (or possibly OSS) service threads could overflow
693 the stack. One change which was identified to consume additional
694 stack has been reworked to avoid the extra stack usage.
697 Frequency : occasional
699 Description: Do not fail import if osc_interpret_create gets -EAGAIN
700 Details : If osc_interpret_create got -EAGAIN it immediately exits and
701 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
702 oscc_has_objects and see OSC has no objests and call
703 oscc_internal_create to resend create request.
705 Severity : enhancement
707 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
709 Severity : enhancement
711 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
714 Frequency : very rare
716 Description: Fix replay if there is an un-replied request and open
717 Details : In some cases, older replay request will revert the
718 mcd->mcd_last_xid on MDS which is used to record the client's
721 Severity : enhancement
723 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
725 Severity : enhancement
727 Description: Add rhel5 support to HEAD.
729 Serverity : enhancement
731 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
736 Description: Don't allow skipping OSTs if index has been specified.
737 Details : Don't allow skipping OSTs if index has been specified, make
738 locking in internal create lots better.
742 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
743 Details : access to bitfield in structure is always rounded to long
744 and this produce problem with not atomic change any bit.
748 Description: Lustre make rpms failed.
749 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
750 builds Lustre rpms from tarball.
753 Frequency : rare on shutdown ost
755 Description: If llog cancel was not send before clean_exports phase, this can
756 produce deadlock in llog code.
757 Details : If llog thread has last reference to obd and call class_import_put
758 this produce deadlock because llog_cleanup_commit_master wait when
759 last llog_commit_thread exited, but this never success because was
760 called from llog_commit_thread.
764 Description: allow userland application know is lost one of stripes.
765 Details : fill lvb_blocks with error code on ost and return it to
766 application if error flag found.
770 Description: NULL lov_tgts causing MDS oops
771 Details : more safe checks for NULL lov_tgts for avoid oops.
773 Severity : enhancement
775 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
779 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
780 Details : If an OST is down the MDS will hang indefinitely in
781 obd_statfs_rqset() waiting for the statfs data. While for
782 MDS QOS usage of statfs, it should not stuck in waiting.
784 Severity : enhancement
786 Description: remote_acl support
787 Details : Support ACL-based permission check for remote user.
788 Support setfacl/getfacl for remote user with the utils
789 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
790 the system "{s,g}etfacl" utils.
792 Severity : enhancement
794 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
796 Severity : enhancement
798 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
802 Description: make mgs_setparam() handle fsname containing dash
803 Details : fsname containing a dash does not work with lctl conf_param
805 Severity : enhancement
807 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
809 Severity : enhancement
811 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
815 Frequency : rare, at shutdown
816 Description: access already free / zero obd_namespace.
817 Details : if client_disconnect_export was called without force flag set,
818 and exist connect request in flight, this can produce access to
819 NULL pointer (or already free pointer) when connect_interpret
820 store ocd flags in obd_namespace.
824 Frequency : only at startup
825 Description: not alloc memory with spinlock held.
826 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
827 if any spinlock held.
829 Severity : enhancement
831 Description: make lustre randomly fail allocating memory
832 Details : Make lustre randomly failed allocating memory for testing purpose.
834 Severity : enhancement
836 Description: lost problems with lov objid file
837 Details : Fixes some scability and access to not inited memory problems
838 in work with lov objdid file.
843 Description: lfs find does not continue on file error
844 Details : Continue other files processing when a file/dir is absent.
848 Description: Inconsistent usage of lustre_pack_reply()
849 Details : Standardize the usage of lustre_pack_reply() such that it
850 always generate a CERROR on failure.
853 Frequency : occasional
855 Description: MDS hang or stay in waiting lock
856 Details : If client receive lock with CBPENDING flag ldlm need send lock
857 cancel as separate rpc, to avoid situation when cancel request
858 can't processed due all i/o threads stay in wait lock.
862 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
866 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
870 Description: Update to SLES9 kernel-2.6.5-7.287.3.
874 Description: Update to RHEL4 latest kernel.
876 Severity : enhancement
878 Description: Build SLES10 patchless client fails
879 Details : The configure was broken by run ./configure with
880 --with-linux-obj=.... argument for patchless client. When the
881 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
882 can't search header adequately. Use absolute path such as
883 -I($LINUX)/include instead.
887 Description: interrupt oig_wait produce painc on resend.
888 Details : brw_redo_request can be used for resend requests from ptlrpcd and
889 private set, and this produce situation when rq_ptlrpcd_data not
890 copyed to new allocated request and triggered LBUG on assert
891 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
892 ptlrpcd set if request is changed and can be safety changed to use
897 Description: LASSERT_{REQ,REP}SWAB macros are buggy
898 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
899 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
900 these macros and replace them with their difinitions instead.
904 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
905 Details : Update which_patch & target file for SLES10 latest kernel.
908 Description: Remove client patches from SLES 10 kernel.
909 Details : This causes SLES 10 clients to behave as patchless clients
910 even on a Lustre-patched (server) kernel.
912 Severity : enhancement
914 Description: self-adjustable client's lru lists
915 Details : use adaptive algorithm for managing client cached locks lru
916 lists according to current server load, other client's work
917 pattern, memory activities, etc. Both, server and client
918 side namespaces provide number of proc tunables for controlling
921 Severity : enhancement
923 Description: light-weight GSS support
924 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
925 and reduce performance overhead.
927 Severity : enhancement
929 Description: Linux keyring support
930 Details : Support using service of Linux keyring for Lustre GSS internal
931 context refresh/cache mechanism.
935 Description: Fix errors in lfs documentation
936 Details : Fixes man pages
940 Description: don't use GFP_* in generic Lustre code.
941 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
945 Description: obdclass is limited by single OBD_ALLOC(idarray)
946 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
950 Description: warnings with build patchless client with vanila 2.6.19 and up
951 Details : change old ctl_table style and replace ctl_table/ctl_table_header
952 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
956 Frequency : Only for SLES
957 Description: sanity_quota fail test_1
958 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
959 file, which makes TSTID[2] inunique.
963 Frequency : Only for SLES9
964 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
965 Details : Update target/ChangeLog/which_patch .
969 Frequency : Only for test_10 in sanity-quota.sh
970 Description: a bug in quota test code
971 Details : checking a test flag in a wrong place
975 Frequency : Only for test_13 in sanity-quota.sh
976 Description: a bug in quota test script
977 Details : shouldn't check overfull allocated quota
981 Description: jbd statistics
982 Details : Port older jbd statistics patch for sles10
984 Severity : enhancement
986 Description: Add printing inode info into message about error in writepage.
991 Description: reply_lock_interpret crash due to race with it and lock cancel.
992 Details : Do not replay locks that are being cancelled. Do not reference
993 locks by their address during replay, just by their handle.
997 Frequency : When flocks are used.
998 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
999 Details : Flock locks might destroy just granted lock if it could be merged
1000 with another existing flock, this is done in completion handler,
1001 so teach ldlm_cli_enquque_fini that this is a valid case for
1007 Description: Oops in read and write path when failing to allocate lock.
1008 Details : Check if lock allocation failed and return error back.
1012 Description: lstripe command fails for valid OST index
1013 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1014 instead of lov->desc.ld_active_tgt_count.
1018 Description: sanity.sh failed test 103
1019 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1024 Description: df doesn't work properly if diskfs blocksize != 4K
1025 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1029 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1030 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1032 Severity : enhancement
1034 Description: Adaptive timeouts
1035 Details : RPC timeouts adapt to changing server load and network
1036 conditions to reduce resend attempts and improve recovery time.
1040 Description: llapi_file_create() does not allow some changes
1041 Details : add llapi_file_open() that allows specifying the mode and
1042 open flags, and also returns an open file handle.
1046 Description: merge and cleanup kernel patches.
1047 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1051 Description: Add journal checksum support.(Kernel part)
1052 Details : The journal checksum feature adds two new flags i.e
1053 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1054 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1055 indicates that the commit block contains the checksum for
1056 the blocks described by the descriptor blocks. Now commit
1057 record can be sent to disk without waiting for descriptor
1058 blocks to be written to disk. This behavior is controlled
1059 using JBD2_FEATURE_ASYNC_COMMIT flag.
1063 Description: OSS needs mutliple precreate threads
1064 Details : Add ability to start more than one create thread per OSS.
1068 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1069 Details : execute lfs setstripe on client
1073 Description: mds_obd_create error creating tmp object
1074 Details : When the user sets quota on root, llog will be affected and can't
1075 create files and write files.
1078 Frequency : Always on ia64 patchless client, and possibly others.
1080 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1081 Details : This allows the patchless client to be loaded on architectures
1082 without this export.
1086 Description: RedHat Update kernel for RHEL5
1087 Details : Add the kernel config file for RHEL5.
1091 Description: RedHat Update kernel for RHEL5
1092 Details : Modify the kernel config file more closer RHEL5.
1096 Description: Build failure against Centos5 (RHEL5)
1097 Details : Define PAGE_SIZE when it isn't present.
1101 Description: test_7 on sanity-quota.sh can't be used on separate machines
1102 Details : add facet to handle it
1106 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1107 Details : with NFS, the anon dentry's parent was set to itself in
1108 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1109 corresponding dentry other than use rec->ur_name.
1111 Severity : enhancement
1113 Description: lfs setstripe enhancement
1114 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1118 Description: enable data checksumming by default
1119 Details : enable checksum by default, allow --disable-checksum
1120 configure option and "-o nochecksum" mount option. Checksums
1121 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1125 Description: lustre support for RHEL5
1126 Details : Add support for RHEL5.
1130 Description: Client eviction due to failover config
1131 Details : after a connection loss, the lustre client should attempt to
1132 reconnect to the last active server first before trying the
1133 other potential connections.
1137 Description: when mds and osts use different quota unit(32bit and 64bit),
1138 quota will be released repeatly.
1139 Details : void sending multiple quota reqs to mds, which will keep the
1140 status between the reqs.
1144 Description: osts not allocated evenly to files
1145 Details : change the condition to increase offset_idx
1148 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1150 Description: Data corruption for OSTs that are formatted larger than 2TB
1152 Details : When generating the bio request for lustre file writes the
1153 sector number would overflow a temporary variable before being
1154 used for the IO. The data reads correctly from Lustre (which
1155 will overflow in a similar manner) but other file data or
1156 filesystem metadata may be corrupted in some cases.
1160 Description: Tune the kernel for good SCSI performance.
1161 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1162 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1167 Description: rewrite ext2-derived code in obdclass/uuid.c
1168 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1169 from scratch preserving functionality.
1173 Description: "lfs find -obd UUID" prints directories
1174 Details : "lfs find -obd UUID" will return all directory names instead
1175 of just file names. It is incorrect because the directories
1176 do not reside on the OSTs.
1180 Description: use i_size_read and i_size_write in 2.6 port
1181 Details : replace inode->i_size access with i_size_read/write()
1184 Frequency : when removing large files
1186 Description: scheduling issue during removal of large Lustre files
1187 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1188 It causes scheduling issues when removing large files (17TB in the
1192 Frequency : only with liblustre clients on XT3
1194 Description: evictions taking too long
1195 Details : allow llrd to evict clients directly on OSTs
1198 Frequency : only on ppc
1200 Description: /proc/fs/lustre/devices broken on ppc
1201 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1202 We should make sure the type of 'index' is loff_t and then cast
1203 explicitly as needed below. Do not assign an explicitly cast
1207 Frequency : only for rhel5
1209 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1210 Details : Modify the target file & which_kernel.
1212 Severity : enhancement
1214 Description: omit set fsid for export NFS
1215 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1216 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1220 Frequency : Always, for kernels after 2.6.16
1221 Description: Fix warning idr_remove called for id=.. which is not allocated.
1222 Details : Last kernels save old s_dev before kill super and not allow
1223 to restore from callback - restore it before call kill_anon_super.
1227 Description: buffer overruns could theoretically occur
1228 Details : llapi_semantic_traverse() modifies the "path" argument by
1229 appending values to the end of the origin string, and a
1230 overrun may occur. Adding buffer overrun check in liblustreapi.
1234 Description: Fix error on 'ls .' at the top of the Lustre mount.
1235 Details : Don't revalidate dentry if it is a root dentry.
1239 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1240 Details : Modify vm-tunables-rhel4.patch.
1244 Description: Kernel config for 2.6.18-vanilla.
1245 Details : Modify targets/2.6-vanilla.target.in.
1246 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1247 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1248 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1249 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1253 Description: improve handling recoverable errors
1254 Details : if request processig with error which can be recoverable on server
1255 request should be resend, otherwise page released from cache and
1260 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1261 Details : Modify target file & which_patch.
1262 A flaw was found in the IA32 system call emulation provided
1263 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1264 value could be stored in the %RAX register, which could trigger an
1265 out-of-bounds system call table access. An untrusted local user
1266 could exploit this flaw to run code in the kernel
1267 (ie a root privilege escalation). (CVE-2007-4573).
1271 Description: O_DIRECT bypasses client statistics.
1272 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1273 read_ahead_stats were updated. copied the stats section from
1274 osc_send_oap_rpc() into async_internal().
1278 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1282 Description: change order of libsysio includes
1283 Details : '#include sysio.h' should always come before '#include xtio.h'
1287 Description: handle "serious error: objid * already exists" more gracefully
1288 Details : If LAST_ID value on disk is smaller than the objects existing in
1289 the O/0/d* directories, it indicates disk corruption and causes an
1290 LBUG(). If the object is 0-length, then we should use the existing
1291 object. This will help to avoid a full fsck in most cases.
1293 Severity : enhancement
1295 Description: adapt the lustre_config script to support the upgrade case
1296 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1297 server targets to 1.6 in parallel.
1301 Description: To avoid grant space > avaible space when the disk is almost
1302 full. Without this patch you might see the error "grant XXXX >
1303 available" or some LBUG about grant, when the disk is almost
1305 Details : In filter_check_grant, for non_grant cache write, we should
1306 check the left space by if (*left > ungranted + bytes), instead
1307 of (*left > ungranted), because only we are sure the left space
1308 is enough for another "bytes", then the ungrant space should be
1309 increase. In client, we should update cl_avail_grant only there
1310 is OBD_MD_FLGRANT in the reply.
1314 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1315 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1316 described in CVE-2007-4573. This problem could allow untrusted
1317 local users to gain root access.
1320 Frequency : when using O_DIRECT and quotas
1322 Description: Incorrect file ownership on O_DIRECT output files
1323 Details : block usage reported by 'lfs quota' does not take into account
1324 files that have been written with O_DIRECT.
1329 Description: touch file failed when fs is not full
1330 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1331 otherwise we can get ENOSP while fs is not full.
1335 Description: parallel lock callbacks
1336 Details : Instead of sending blocking and completion callbacks as separated
1337 requests, adding them to a set and sending in parallel.
1340 Frequency : only for Cray XT3
1341 Bugzilla : 12829/13455
1342 Description: Changing primary group doesn't change the group lustre assigns to
1344 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1345 overridden with the primary group provided by the group upcall,
1346 whereas the supplied fsgid can be trusted if it is in the list of
1347 supplementary groups returned by the group upcall.
1349 Severity : enhancement
1351 Description: Allow masking D_WARNING, D_ERROR messages from console
1352 Details : Console messages can now be disabled via lnet.printk.
1356 Description: User code with malformed file open parameter crashes client node
1357 Details : Before packing join_file req, all the related reference should be
1358 checked carefully in case some malformed flags cause fake
1359 join_file req on client.
1363 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1364 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1365 mode is equal to the granted mode. However, it should also check
1366 that LDLM_FL_CP_REQD is not set.
1370 Description: Heavy nfs access might result in deadlocks
1371 Details : After ELC code landed, it is now improper to enqueue any mds
1372 locks under och_sem, because enqueue might want to decide to
1373 cancel open locks for same inode we are holding och_sem for.
1377 Description: Client eviction while running blogbench
1378 Details : A lot of unlink operations with concurrent I/O can lead to a
1379 deadlock causing evictions. To address the problem, the number of
1380 oustanding OST_DESTROY requests is now throttled to
1381 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1382 callbacks are processed in priority.
1386 Description: enable ACLs on MDS by default
1387 Details : ACLs must be enabled on MDS by default.
1390 Frequency : PPC/PPC64 only
1392 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1393 Details : fix duplicated definitions between asm-ppc64/types.h and
1394 lustre_types.h on PPC.
1397 Frequency : PPC/PPC64 only
1399 Description: asm-ppc/segment.h does not exist
1400 Details : fix compile issue on PPC.
1404 Description: better handle error messages in extents code
1407 Frequency : RHEL4 only
1409 Description: mkfs is very slow on IA64/RHEL4
1410 Details : A performance regression has been discovered in the MPT Fusion
1411 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1412 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1413 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1415 Severity : enhancement
1417 Description: SNMP support enhancement
1418 Details : Adding total number of sampled request for an MDS node in snmp
1421 Severity : enhancement
1423 Description: Optimize ldlm waiting list processing for PR extent locks
1424 Details : When processing waiting list for read extent lock and meeting read
1425 lock that is same or wider to it that is not contended, skip
1426 processing rest of the list and immediatelly return current
1427 status of conflictness, since we are guaranteed there are no
1428 conflicting locks in the rest of the list.
1432 Description: Time out and refuse to reconnect
1433 Details : When the failover node is the primary node, it is possible
1434 to have two identical connections in imp_conn_list. We must
1435 compare not conn's pointers but NIDs, otherwise we can defeat
1436 connection throttling.
1440 Description: port llog fixes from b1_6 into HEAD
1441 Details : Port llog reference couting and some llog cleanups from b1_6
1442 (bug 10800) into HEAD, for protect from panic and access to already
1443 free llog structures.
1447 Description: Detect stride IO mode in read-ahead
1448 Details : When a client does stride read, read-ahead should detect that and
1449 read-ahead pages according to the detected stride pattern.
1453 Description: data checksumming impacts single node performance
1454 Details : add support for several checksum algorithm. Currently, only CRC32
1455 and Adler-32 are supported. The checksum type can be changed on
1456 the fly via /proc/fs/lustre/osc/*/checksum_type.
1460 Description: use adler32 for page checksums
1461 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1466 Description: build for x2 fails
1467 Details : fix compile issue on Cray systems.
1471 Description: Properly match for duplicate locks
1472 Details : Due to different lock order from skiplists code, we need to
1473 traverse entire list for now
1476 Frequency : only on PPC/SLES10
1478 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1479 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1480 be defined. Add a hack in mkfs_lustre.c to work around this compile
1485 Description: LASSERT on MDS when client holding flock lock dies
1486 Details : ldlm pool logic depends on number of granted locks equal to
1487 number of released locks which is not true for flock locks, so
1488 just exclude such locks from consideration.
1492 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1493 Details : Use fsfilt_sync() for both the whole filesystem sync and
1494 individual file sync to eliminate dangerous inode locking
1495 with I_LOCK that can lead to a deadlock.
1499 Description: performance in 1.6.3
1500 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1501 sure that 1MB requests are not fragmented by the block layer.
1503 Severity : enhancement
1505 Description: organize the server-side client stats on per-nid basis
1506 Details : Change the structure of stats under obdfilter and mds to
1514 The "uuid"s file would list the uuids of _active_ exports.
1515 And the clear entry is to clear all stats and stale nids.
1517 Severity : enhancement
1519 Description: eliminate client locks in face of contention
1520 Details : file contention detection and lockless i/o implementation
1521 for contended files.
1525 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1528 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1530 Description: MDS may lose file striping (and hence file data) in some cases
1531 Details : If there are additional extended attributes stored on the MDS,
1532 in particular ACLs, SELinux, or user attributes (if user_xattr
1533 is specified for the client mount options) then there is a risk
1534 of attribute loss. Additionally, the Lustre file striping
1535 needs to be larger than default (e.g. striped over all OSTs),
1536 and an additional attribute must be stored initially in the
1537 inode and then increase in size enough to be moved to the
1538 external attribute block (e.g. ACL growing in size) for file
1543 Description: skiplist implementation simplification
1544 Details : skiplists are used to group compatible locks on granted list
1545 that was implemented as tracking first and last lock of each
1546 lock group the patch changes that to using doubly linked lists
1550 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1551 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1552 because LASSERT on both the data supplied by a client, and the
1553 data on disk is dangerous and incorrect.
1555 Severity : enhancement
1557 Description: Slow truncate/writes to huge files at high offsets.
1558 Details : Directly associate cached pages to lock that protect those pages,
1559 this allows us to quickly find what pages to write and remove
1560 once lock callback is received.
1564 Description: more ldlm soft lockups
1565 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1566 starve other threads from the resource lock for a long time in
1567 case of long waiting queue, so change the debug level from
1568 D_OTHER to the less frequently used D_INFO.
1570 Severity : enhancement
1572 Description: add -gid, -group, -uid, -user options to lfs find
1576 Description: Hung threads in invalidate_inode_pages2_range
1577 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1578 once one is completed. As a result, some RPCs are stuck in the
1579 queue and are never sent.
1583 Description: filter threads hungs on waiting journal commit
1584 Details : Cleanup filter group llog code, then only filter group llog will
1585 be only created in the MDS/OST syncing process.
1589 Description: Procfs and llog threads access destoryed import sometimes.
1590 Details : Sync the import destoryed process with procfs and llog threads by
1591 the import refcount and semaphore.
1593 Severity : enhancement
1595 Description: openlock cache of b1_6 port to HEAD
1600 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1601 Details : If insertion of an extent fails, then discard the inode
1602 preallocation and free data blocks else it can lead to duplicate
1607 Description: don't always update ctime in ext3_xattr_set_handle()
1608 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1609 In some cases the ctime should not be updated, for example for
1610 2.0->1.8 compatibility it is necessary to delete an xattr and it
1611 should not update the ctime.
1615 Bugzilla : 15713/16362
1616 Description: Assertion in iopen_connect_dentry in 1.6.3
1617 Details : looking up an inode via iopen with the wrong generation number can
1618 populate the dcache with a disconneced dentry while the inode
1619 number is in the process of being reallocated. This causes an
1620 assertion failure in iopen since the inode's dentry list contains
1621 both a connected and disconnected dentry.
1625 Description: assertion failure in ldlm_handle2lock()
1626 Details : fix a race between class_handle_unhash() and class_handle2object()
1627 introduced in lustre 1.6.5 by bug 13622.
1632 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1633 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1638 Description: oops in sd_iostats_seq_show()
1639 Details : unloading/reloading the scsi low level driver triggers a kernel
1640 bug when trying to access the sd iostat file.
1645 Description: Kernel panics during QLogic driver reload
1646 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1647 patch, causing memory corruption.
1652 Description: journal_dev option does not work in b1_6
1653 Details : pass mount option during pre-mount.
1655 Severity : enhancement
1657 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1658 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1659 extent information of a file. It can be used to map logical blocks
1660 in a file to physical blocks in the block device.
1664 Description: LDLM soft lockups - improvement
1665 Details : It is be possible to send the lock handle along with each read
1666 or write request because the client is already doing a lock match
1667 itself so there isn't any reason the OST should have to re-do that
1671 Frequency : only on Cray X2
1673 Description: X2 build failures
1674 Details : fix build failures on Cray X2.
1678 Description: xid & resent requests
1679 Details : Initialize RPC XID from clock at startup (randomly if clock is
1682 Severity : enhancement
1684 Description: Add lustre_start utility to start or stop multiple Lustre servers
1689 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1690 Details : In case of memory pressure, list_del() can be called twice on
1691 req->rq_history_list, causing a kernel oops.
1695 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1696 Details : fix stack overflow in the distributed lock manager by defering
1697 export eviction after a failed AST to the elt thread instead of
1698 handling it in the dlm interpret routine.
1702 Description: Convert some comments to new format.
1703 Details : Update documenting comments to match doxygen conventions.
1707 Description: Grammar fixes.
1708 Details : A couple of trivial spelling fixes.
1712 Description: OSD_COUNTERS-mandatory
1713 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1714 to simplify the code.
1718 Description: simplify cmm_device freeing logic.
1719 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1723 Description: Add lockdep support to dt_object_operations locking interface.
1724 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1725 parameter indicating lock ordering. Update mdd code to use new
1730 Description: Introduce failloc constants for lockless IO tests.
1731 Details : Add two new failloc constants to test lockless IO. Only one of
1732 them in implemented---another is checked in yet to be landed
1737 Description: Add lockdep support for inode mutex.
1738 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1739 in the situations where Lustre has to lock more than one inode
1744 Description: Add optional invariants checking support.
1745 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1746 configure switch. This macro is to be used for consistency and
1747 sanity checks that are too expensive to be left in `production'
1752 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1753 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1758 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1759 Details : Use spin_lock_nested() in (the only) situation where more than
1760 one ldlm_lock is locked simultaneously. Also, fix possible
1761 dead-lock in ldlm_lock_change_resource() by enforcing particular
1766 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1767 Details : Instead of specifying each ldlm_lock call-back through separate
1768 parameter, wrap them into struct ldlm_callback_suite.
1772 Description: Kill join_lru obd method and its callers.
1773 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1774 in memory---a requirement implemented through ->o_join_lru() obd
1775 method in HEAD. Remove this method and its users.
1779 Description: Add asynchronous ldlm ENQUEUE completion handler.
1780 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1781 case that stock ldlm_completion_ast() cannot handle as it waits
1782 until lock is granted. Introduce new ldlm_completion_ast_async()
1783 for this. Also comment ldlm_completion_ast().
1787 Description: ldlm_error <-> errno conversion.
1788 Details : Add functions to map (rather arbitrary) between LDLM error codes
1789 and standard errno values. CLIO needs this to prevent LDLM specific
1790 constants from escaping ldlm and osc.
1794 Description: Kill unused ldlm_handle2lock_ns() function.
1795 Details : Kill unused ldlm_handle2lock_ns() function.
1799 Description: Add lu_ref support to ldlm_lock
1800 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1801 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1802 and ldlm_resource. LDLM interface has to be changed, because code that
1803 releases a reference on a lock, has to "know" what reference this is.
1804 In the most frequent case
1806 lock = ldlm_handle2lock(handle);
1808 LDLM_LOCK_PUT(lock);
1810 no changes are required. When any other reference (received _not_ from
1811 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1812 instead of LDLM_LOCK_PUT().
1814 Arguably, changes are pervasive, and interface requires some discipline
1815 for proper use. On the other hand, it was very instrumental in finding
1816 a few leaked lock references.
1820 Description: Add ldlm_lock_addref_try().
1821 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1822 attempts to addref a lock that might be being canceled
1827 Description: Add ldlm_weigh_callback().
1828 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1829 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1830 hard-coded `number of pages' logic.
1834 Description: Add lockdep annotations to llog code.
1835 Details : Use appropriately tagged _nested() locking calls in the places
1836 where llog takes more than one ->lgh_lock lock.
1840 Description: Add loi_kms_set().
1841 Details : Wrap kms updates into a helper function.
1845 Description: Constify instances of struct lsm_operations.
1846 Details : Constify instances of struct lsm_operations.
1850 Description: lu_conf support.
1851 Details : On a server, a file system object is uniquely identified
1852 by a fid, which is sufficient to locate and load all object
1853 state (inode). On a client, on the other hand, more data are
1854 necessary instantiate an object. Change lu_object_find() and
1855 friends to take additional `lu_conf' argument describing object.
1856 Typically this includes layout information.
1860 Description: lu_context fixes.
1861 Details : Introduce new lu_context functions that are needed on the client
1862 side, where some system threads (ptlrpcd) are shared by multiple
1863 modules, and so cannot be stopped during module shutdown.
1867 Description: Add start and stop methods to lu_device_type_operations.
1868 Details : Introduce two new methods in lu_device_type_operations, that are
1869 invoked when first instance of a given type is created and last one
1870 is destroyed respectively. This is need by CLIO.
1874 Description: Add lu_ref support to struct lu_device.
1875 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1876 track leaked references.
1880 Description: Introduce lu_kmem_descr.
1881 Details : lu_kmem_descr and its companion interface allow to create
1882 and destroy a number of kmem caches at once.
1886 Description: Fix lu_object finalization race.
1887 Details : Fix a race between lu_object_find() finding an object and its
1888 concurrent finalization. This race is (most likely) not possible
1889 on the server, but might happen on the client.
1893 Description: Introduce lu_ref interface.
1894 Details : lu_ref is a debugging module allowing to track references to
1895 a given object. It is quite cpu expensive, and has to be
1896 explicitly enabled with --enable-lu_ref. See usage description
1901 Description: Factor lu_site procfs stats into a separate function.
1902 Details : Separate lu_site stats printing code into a separate function
1903 that can be reused on a client.
1907 Description: Constify instances of struct {lu,dt,md}_device_operations.
1908 Details : Constify instances of struct {lu,dt,md}_device_operations.
1912 Description: Introduce struct md_site and move meta-data specific parts of
1913 struct lu_site here.
1914 Details : Move md-specific fields out of struct lu_site into special struct
1915 md_site, so that lu_site can be used on a client.
1919 Description: Kill mdd_lov_destroy().
1920 Details : Remove unused mdd code.
1924 Description: Add st_block checking to multistat.c.
1925 Details : Add st_block checking to multistat.c.
1929 Description: Add lu_ref support to struct obd_device.
1930 Details : Add lu_ref tracking to obd_device.
1934 Description: Kill obd_set_fail_loc().
1935 Details : Remove unused code.
1939 Description: Add special type for ptlrpc_request interpret functions.
1940 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1941 there. Actual usage will be in CLIO.
1945 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1946 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1947 doesn't work with lockdep.
1951 Description: Add rwv.c test program.
1952 Details : New testing program exercising readv(2) and writev(2) (Qian).
1956 Description: Add sendfile.c test program.
1957 Details : New testing program exercising sendfile(2) (Jay).
1961 Description: Ratelimit a message that can be very frequent.
1962 Details : Ratelimit a memory allocation failure message that can
1967 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1968 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1972 Description: Liblustre build fixes.
1973 Details : Liblustre build fixes.
1977 Description: libcfs: add cfs_{need,cond}_resched() interface.
1978 Details : libcfs: add cfs_{need,cond}_resched() definition and
1979 implementations for Linux, NT, and liblustre.
1981 Severity : enhancement
1983 Description: More exported tunables for mballoc
1984 Details : Add support for tunable preallocation window and new tunables for
1985 large/small requests
1989 Description: Detect corruption of block bitmap and checking for preallocations
1990 Details : Checks validity of on-disk block bitmap. Also it does better
1991 checking of number of applied preallocations. When corruption is
1992 found, it turns filesystem readonly to prevent further corruptions.
1996 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1997 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1998 window to 0. If it does hit cache miss, and read-pattern is still
1999 stride-io mode, does not reset the stride window, but also does not
2000 increase the stride window length in this case.
2004 Frequency : only for big-endian servers
2005 Description: Check if system is big-endian while mounting fs with extents feature
2006 Details : Mounting a filesystem with extents feature will fail on big-endian
2007 systems since ext3-based ldiskfs is not supported on big-endian
2008 systems. This can be over-riden with "bigendian_extents" mount option.
2010 Severity : enhancement
2012 Description: The root squash functionality
2013 Details : A security feature, which is to prevent users from being able
2014 to mount lustre on their desktop, run as root, and delete
2015 all of the files in the filesystem. The goal is accomplished by
2016 remapping user id (UID) and group id (GID) of the root user to
2017 a UID and GID specified by the system administartor via Lustre
2018 configuration management server (MGS). The functionality also
2019 allows to specify sets of clients for which the remapping does
2024 Description: Excessive recovery window
2025 Details : With AT enabled, the recovery window can be excessively long (6000+
2026 seconds). To address this problem, we no longer use
2027 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2028 timeout no longer depends on the service time, it is set to
2029 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2030 time via pb_service_time.
2034 Description: Watchdog triggered on MDS failover
2035 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2036 the OSTs know that the MDS "UUID" can be reused for the same export
2037 from a different NID, so we do not need to wait for the export to be
2041 Frequency : rare, only if using MMP with Linux RAID
2043 Description: MMP doesn't work with Linux RAID
2044 Details : While using HA for Lustre servers with Linux RAID, it is possible
2045 that MMP will not detect multiple mounts. To make this work we
2046 need to unplug the device queue in RAID when the MMP block is being
2047 written. Also while reading the MMP block, we should read it from
2048 disk and not the cached one.
2050 Severity : enhancement
2052 Description: open file using fid
2053 Details : A file can be opened using just its fid, like
2054 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2057 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2060 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2061 Details : add proper locking for creation of jbd_4k slab cache
2065 Description: MMP check in ext3_remount() fails without displaying any error
2066 Details : When multiple mount protection fails during remount, proper error
2069 --------------------------------------------------------------------------------
2071 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2073 * Support for kernels:
2074 2.6.5-7.283 (SLES 9),
2075 2.6.9-55.EL (RHEL 4),
2076 2.6.16.46-0.14 (SLES 10),
2077 2.6.18.8 vanilla (kernel.org)
2078 * Client support for unpatched kernels:
2079 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2080 2.6.16 - 2.6.22 vanilla (kernel.org)
2081 * Due to recently discovered recovery problems, we do not recommend
2082 using patchless RHEL 4 clients with this or any earlier release.
2083 * Recommended e2fsprogs version: 1.39.cfs8
2084 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2085 * Starting with this release, the ldiskfs backing filesystem required
2086 by Lustre is now in its own package, lustre-ldiskfs. This package
2087 should be installed. It is versioned separately from Lustre and
2088 may be released separately in future.
2092 Description: block reactivating mgc import until all deactivates complete
2093 Details : Fix race when failing back MDT/MGS to itself (testing)
2096 Frequency : at statup only
2098 Description: mds_lov_synchronize race leads to various problems
2099 Details : simultaneous MDT->OST connections at startup can cause the
2100 sync to abort, leaving the OSC in a bad state.
2102 Severity : enhancement
2104 Description: add optional extra BUILD_VERSION info
2105 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2106 to override the lustre version.
2109 Frequency : 2.6.18 servers only
2111 Description: ll_kern_mount() doesn't release the module reference
2112 Details : The ldiskfs module reference count never drops down to 0
2113 because ll_kern_mount() doesn't release the module reference.
2118 Description: server LBUG when using old ost_num_threads parameter
2119 Details : Accept the old ost_num_threads parameter but warn that it
2120 is deprecated, and fix an off-by-one error that caused an LBUG.
2125 Description: Transient SCSI error results in persistent IO issue
2126 Details : iobuf->dr_error is not reinitialized to 0 between two
2130 Frequency : sometimes when underlying device returns I/O errors
2132 Description: OSTs not going read-only during write failures
2133 Details : OSTs are not remounted read-only when the journal commit threads
2134 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2135 instead of the ext3 wrappers.
2139 Description: poor connect scaling with increasing client count
2140 Details : Don't run filter_grant_sanity_check for more than 100 exports
2141 to improve scaling for large numbers of clients.
2144 Frequency : SLES10 only
2146 Description: sanity-quota.sh quotacheck failed: rc = -22
2147 Details : Quotas cannot be enabled on SLES10.
2150 Frequency : liblustre clients only
2152 Description: getdirentries does not give error when run on compute nodes
2153 Details : getdirentries does not fail when the size specified as an argument
2154 is too small to contain at least one entry
2156 Severity : enhancement
2158 Description: Add LNET router traceability for debug purposes
2159 Details : If a checksum failure occurs with a router as part of the
2160 IO path, the NID of the last router that forwarded the bulk data
2161 is printed so it can be identified.
2166 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2167 Details : Due to a race condition, liblustre clients were occasionally
2168 evicted incorrectly.
2170 Severity : enhancement
2172 Description: lfs setstripe use optional parameters instead of postional
2175 Severity : enhancement
2177 Description: Nanosecond timestamp support for ldiskfs
2178 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2179 resolution timestamps. There is not yet support for this at
2180 the Lustre filesystem level.
2183 Frequency : during server recovery
2185 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2186 Details : request with rq_no_resend flag not awake l_wait_event if they get
2190 Frequency : nfs export on patchless client
2192 Description: connectathon hang when test nfs export over patchless client
2193 Details : Disconnected dentry cannot be found with lookup, so we do not need
2194 to unhash it or make it invalid
2197 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2198 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2199 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2204 Description: open req refcounting wrong on reconnect
2205 Details : If reconnect happened between getting open reply from server and
2206 call to mdc_set_replay_data in ll_file_open, we will schedule
2207 replay for unreferenced request that we are about to free.
2208 Subsequent close will crash in variety of ways.
2209 Check that request is still eligible for replay in
2210 mdc_set_replay_data().
2215 Description: disable writes to filesystem when reading health_check file
2216 Details : the default for reading the health_check proc file has changed
2217 to NOT do a journal transaction and write to disk, because this
2218 can cause reads of the /proc file to hang and block HA state
2219 checking on a healthy but otherwise heavily loaded system. It
2220 is possible to return to the previous behaviour during configure
2221 with --enable-health-write.
2223 Severity : enhancement
2225 Description: 64-bit inode version
2226 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2227 to the inode. This will be required for version-based recovery.
2232 Description: MDS fails to start if a duplicate client export is detected
2233 Details : in some rare cases it was possible for a client to connect to
2234 an MDS multiple times. Upon recovery the MDS would detect this
2235 and fail during startup. Handle this more gracefully.
2237 Severity : enhancement
2239 Description: Add -o localflock option to simulate old noflock
2241 Details : This will achieve local-only flock/fcntl locks
2247 Description: log_commit_thread vs filter_destroy race leads to crash
2248 Details : Take import reference before releasing llog record semaphore
2253 Description: Wrong request locking in request set processing
2254 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2255 imp_delayed_list, in this place should be used imp_lock.
2258 Frequency : when reconnection
2260 Description: Grant Leak when osc reconnect to OST
2261 Details : When osc reconnect ost, OST(filter) should check whether it
2262 should grant more space to client by comparing fed_grant and
2263 cl_avail_grant, and return the granted space to client instead
2264 of "new granted" space, because client will call osc_init_grant
2265 to update the client grant space info.
2268 Frequency : when client reconnect to OST
2270 Description: Grant Leak when osc do resend and replay bulk write
2271 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2272 bulk write request, because the grant info while be sync between
2273 OSC and OST when reconnect, and we should ignore the grant info
2274 these of resend/replay write req.
2279 Description: Grant space more than avaiable left space sometimes.
2280 Details : When then OST is about to be full, if two bulk writing from
2281 different clients came to OST. Accord the avaliable space of the
2282 OST, the first req should be permitted, and the second one
2283 should be denied by ENOSPC. But if the seconde arrived before
2284 the first one is commited. The OST might wrongly permit second
2285 writing, which will cause grant space > avaiable space.
2288 Frequency : when client is evicted
2290 Description: Grant might be wrongly erased when osc is evicted by OST
2291 Details : when the import is evicted by server, it will fork another
2292 thread ptlrpc_invalidate_import_thread to invalidate the
2293 import, where the grant will be set to 0. While the original
2294 thread will update the grant it got when connecting. So if
2295 the former happened latter, the grant will be wrongly errased
2296 because of this race.
2301 Description: Checking Stale with correct fid
2302 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2303 to check whether it is stale, because sometimes, we want the
2304 enqueue happened anyway, and op_data.fid2 will not be initialized.
2306 Severity : enhancement
2308 Description: update patchless client
2309 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2312 Frequency : only with 2.4 kernel
2314 Description: random memory corruption
2315 Details : size of struct ll_inode_info is to big for union inode.u and this
2316 can be cause of random memory corruption.
2321 Description: Memory leak in recovery
2322 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2323 It should also check obd_fail before fsfilt_start, otherwise if
2324 fsfilt_start return -EROFS,(failover mds during mds recovery).
2325 then the req will return with repmsg->transno = 0 and rc = EROFS.
2326 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2327 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2328 be freed no matter whether fsfilt_commit success or not.
2331 Frequency : only with huge count clients
2333 Description: Prevents from taking the superblock lock in llap_from_page for
2335 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2336 for llap_from_page call in ll_removepage prevents from taking the
2337 superblock lock for a soon died page.
2342 Description: Not check open intent error before release open handle
2343 Details : in some rare cases, the open intent error is not checked before
2344 release open handle, which may cause
2345 ASSERTION(open_req->rq_transno != 0), because it tries to release
2346 the failed open handle.
2351 Description: Set cat log bitmap only after create log success.
2352 Details : in some rare cases, the cat log bitmap is set too early. and it
2353 should be set only after create log success.
2357 Description: Accessing a block bevice can re-enable I/O when Lustre is
2358 tearing down a device.
2359 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2364 Description: service threads may hog cpus when there are a lot of requests
2365 Details : Insert cond_resched to give other threads a chance to use some CPU
2370 Description: the cat log was not initialized in recovery
2371 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2372 unlink log on mds will not be initialized until mds post
2373 recovery. And also in mds post recovery, the unlink log will
2374 initialization will be done asynchronausly, so there will be race
2375 between add unlink log and unlink log initialization.
2379 Description: brw_stats were being printed incorrectly
2380 Details : brw_stats were being printed as log2 but all of them were not
2381 recorded as log2. Also remove some code duplication arising from
2382 filter_tally_{read,write}.
2386 Frequency : rare, only in recovery.
2387 Description: ASSERTION(req->rq_type != LI_POISON) failed
2388 Details : imp_lock should be held while iterating over imp_sending_list for
2389 prevent destroy request after get timeout in ptlrpc_queue_wait.
2393 Description: replay-single.sh test 52 fails
2394 Details : A lock's skiplist need to be cleanup when it being unlinked
2395 from its resource list.
2399 Description: Short directio read returns full requested size rather than
2401 Details : Direct I/O operations should return actual amount of bytes
2402 transferred rather than requested size.
2404 Severity : enhancement
2406 Description: metadata RPC reduction (e.g. for rm performance)
2407 Details : decrease the amount of synchronous RPC between clients and servers
2408 by canceling conflicing lock before the operation on the client
2409 and packing thier handles into the main operation RPC to server.
2411 Severity : enhancement
2413 Description: Async OSC create to avoid the blocking unnecessarily.
2414 Details : If a OST has no remain object, system will block on the creating
2415 when need to create a new object on this OST. Now, ways use
2416 pre-created objects when available, instead of blocking on an
2417 empty osc while others are not empty. If we must block, we block
2418 for the shortest possible period of time.
2422 Description: improve handling recoverable errors
2423 Details : if request processig with error which can be recoverable on server
2424 request should be resend, otherwise page released from cache and
2427 Severity : enhancement
2429 Description: refine locking for avoid write wrong info into lov_objid file
2430 Details : fix possible races with add new target and write/update data in
2433 --------------------------------------------------------------------------------
2435 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2440 Frequency : on some architectures
2442 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2443 Details : Uninitialized flags sometimes cause configuration commands to
2447 Frequency : patchless clients only
2449 Description: missing __iget() symbol export
2450 Details : The __iget() symbol export is missing. To avoid the need for
2451 this on patchless clients the deathrow inode reaper is turned
2452 off, and we depend on the VM to clean up old inodes. This
2453 dependency was during via the fix for bug 12181.
2455 --------------------------------------------------------------------------------
2457 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2459 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2460 INTEROPERATE with older versions automatically. In many cases a
2461 special upgrade step is needed. Please read the
2462 user documentation before upgrading any part of a live system.
2463 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2464 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2465 * WARNING: Lustre configuration and startup changes are required with
2466 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2470 Severity : enhancement
2472 Description: MountConf
2473 Details : Lustre configuration is now managed via mkfs and mount
2474 commands instead of lmc and lconf. New obd types (MGS, MGC)
2475 are added for dynamic configuration management. See
2476 https://mail.clusterfs.com/wikis/lustre/MountConf for
2479 Severity : enhancement
2481 Description: dynamic OST addition
2482 Details : OSTs can now be added to a live filesystem
2484 Severity : enhancement
2486 Description: startup order invariance
2487 Details : MDTs and OSTs can be started in any order. Clients only
2488 require the MDT to complete startup.
2490 Severity : enhancement
2492 Description: parallel, asynchronous orphan cleanup
2493 Details : orphan cleanup is now performed in separate threads for each
2494 OST, allowing parallel non-blocking operation.
2496 Severity : enhancement
2498 Description: optimized stripe assignment
2499 Details : stripe assignments are now made based on ost space available,
2500 ost previous usage, and OSS previous usage, in order to try
2501 to optimize storage space and networking resources.
2503 Severity : enhancement
2505 Description: Permanently set tunables
2506 Details : All writable /proc/fs/lustre tunables can now be permanently
2507 set on a per-server basis, at mkfs time or on a live system.
2509 Severity : enhancement
2511 Description: Lustre message v2
2512 Details : Add lustre message format v2.
2514 Severity : enhancement
2516 Description: client OST exclusion list
2517 Details : Clients can be started with a list of OSTs that should be
2518 declared "inactive" for known non-responsive OSTs.
2522 Description: SPEC SFS validation failure on NFS v2 over lustre.
2523 Details : Changes the blocksize for regular files to be 2x RPC size,
2524 and not depend on stripe size.
2526 Severity : enhancement
2528 Description: Multiple MD RPCs in flight.
2529 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2530 To avoid overly-overloading MDS, introduce a limit on number of
2531 MDS RPCs in flight for a single client and add /proc controls
2532 to adjust this limit.
2534 Severity : enhancement
2536 Description: client read/write statistics
2537 Details : Add client read/write call usage stats for performance
2538 analysis of user processes.
2539 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2540 file access. extents_stats shows chunk size distribution.
2541 extents_stats_per_process show chunk size distribution per
2544 Severity : enhancement
2546 Description: mds statistics
2547 Details : Add detailed mds operations statistics in
2548 /proc/fs/lustre/mds/*/stats.
2552 Description: Failure of copying files with lustre special EAs.
2553 Details : Client side always return success for setxattr call for lustre
2554 special xattr (currently only "trusted.lov").
2559 Description: Refcount LNET uuids
2560 Details : The global LNET uuid list grew linearly with every startup;
2561 refcount repeated list entries instead of always adding to
2564 Severity : enhancement
2566 Description: Dynamic service threads
2567 Details : Within a small range, start extra service threads
2568 automatically when the request queue builds up.
2571 Frequency : mixed-endian client/server environments
2573 Description: mixed-endian crashes
2574 Details : The new msg_v2 system had some failures in mixed-endian
2577 Severity : enhancement
2579 Description: Easy OST removal
2580 Details : OSTs can be permanently deactivated with e.g. 'lctl
2581 conf_param lustre-OST0001.osc.active=0'
2583 Severity : enhancement
2585 Description: MGS proc entries
2586 Details : Added basic proc entries for the MGS showing what filesystems
2589 Severity : enhancement
2591 Description: provide MGS failover
2592 Details : Added config lock reacquisition after MGS server failover.
2594 Severity : enhancement
2596 Description: add Linux 2.4 support
2597 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2601 Description: a large application tries to do I/O to the same resource and dies
2602 in the middle of it.
2603 Details : Check the req->rq_arrival time after the call to
2604 ost_brw_lock_get(), but before we do anything about
2605 processing it & sending the BULK transfer request. This
2606 should help move old stale pending locks off the queue as
2607 quickly as obd_timeout.
2610 Frequency : when an incorrect nid is specified during startup
2612 Description: ptlrpc connect to non-existant node causes kernel crash
2613 Details : LNET can't be re-entered from an event callback, which
2614 happened when we expire a message after the export has been
2615 cleaned up. Instead, hand the zombie cleanup off to another
2618 Severity : enhancement
2620 Description: plain/inodebits lock performance improvement
2621 Details : Grouping plain/inodebits in granted list by their request modes
2622 and bits policy, thus improving the performance of search through
2626 Frequency : only if OST filesystem is corrupted
2628 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2629 Details : for a short time RPCs with bulk IO are in the replay list,
2630 but replay of bulk IOs is unimplemented. If the OST filesystem
2631 is corrupted due to disk cache incoherency and then replay is
2632 started it is possible to trip an assertion. Avoid putting
2633 committed RPCs into the replay list at all to avoid this issue.
2636 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2638 Description: System hang on startup
2639 Details : This bug allowed the liblustre (e.g. catamount) client to
2640 return to the app before handling all startup RPCs. This
2641 could leave the node unresponsive to lustre network traffic
2642 and manifested as a server ptllnd timeout.
2644 Severity : enhancement
2646 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2647 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2648 boolean to control whether to print peer debug info when a
2649 client's RPC times out.
2652 Frequency : only for kernels with patches from Lustre below 1.4.3
2654 Description: Remove old rdonly API
2655 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2658 Frequency : only for devices with external journals
2660 Description: Set external device read-only also
2661 Details : During a commanded failover stop, we set the disk device
2662 read-only while the server shuts down. We now also set any
2663 external journal device read-only at the same time.
2666 Frequency : when upgrading from 1.4 while trying to change parameters
2668 Description: The wrong (new) MDC name was used when setting parameters for
2669 upgraded MDT's. Also allows changing of OSC (and MDC)
2670 parameters if --writeconf is specified at tunefs upgrade time.
2673 Frequency : when setting specific ost indicies
2675 Description: QOS code breaks on skipped indicies
2676 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2677 created with --index need not be sequential.
2679 Severity : enhancement
2681 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2682 Details : The uninit_groups feature works in conjunction with the kernel
2683 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2684 up the pass1 processing of e2fsck. This is a read-only feature
2685 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2686 mount filesystems that have had this feature enabled.
2688 Severity : enhancement
2690 Description: Improve multi-block allocation algorithm to avoid fragmentation
2691 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2692 allocation locality and avoid filesystem fragmentation.
2694 ------------------------------------------------------------------------------
2696 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2698 * Support for kernels:
2699 2.4.21-47.0.1.EL (RHEL 3)
2700 2.6.5-7.283 (SLES 9)
2701 2.6.9-42.0.10.EL (RHEL 4)
2702 2.6.12.6 vanilla (kernel.org)
2703 2.6.16.27-0.9 (SLES 10)
2704 * Recommended e2fsprogs version: 1.39.cfs5
2706 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2710 Frequency : occasional, depends on client load and configuration
2711 Bugzilla : 12181, 12203
2712 Description: data loss for recently-modified files
2714 Details : In some cases it is possible that recently written or created
2715 files may not be written to disk in a timely manner (this should
2716 normally be within 30s unless client IO load is very high).
2717 The problem appears as zero-length files or files that are a
2718 multiple of 1MB in size after a client crash or client eviction
2719 that are missing data at the end of the file.
2721 This problem is more likely to be hit on clients where files are
2722 repeatedly created and unlinked in the same directory, clients
2723 have a large amount of RAM, have many CPUs, the filesystem has
2724 many OSTs, the clients are rebooted frequently, and/or the files
2725 are not accessed by other nodes after being written.
2727 The presence of the problem can be detected by looking at
2728 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2729 smaller than the second (nr_unused) then dirty files will not
2730 be flushed automatically to disk. "sync; sleep 10" should be
2731 run several times on the node before unmounting it to update
2732 Lustre (this is also safe to run on nodes without this problem).
2734 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2735 that can cause this same problem, so customers using that kernel
2736 also need to update the kernel in addition to Lustre. In order
2737 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2739 It is normal that files written just before a client crash (less
2740 than 30s) may not yet have been flushed to disk, even for local
2744 Frequency : frequent on thin XT3 nodes
2746 Description: UUID collision on thin XT3 Linux nodes
2747 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2748 randomly, since we relied on an insufficiently-seeded PRNG.
2753 Description: OSS hangs after "All ost request buffers busy"
2754 Details : A deadlock between quota and journal operations caused OSS
2755 hangs after printing "All ost request buffers busy."
2758 Frequency : always on liblustre builds
2760 Description: Cleanup compiler warnings on liblustre
2763 Frequency : always on liblustre builds on XT3
2765 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2769 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2770 > 32000 subdirectories
2771 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2772 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2773 subdirectory count crosses 32000. This will aid e2fsck to
2774 correctly handle more than 32000 subdirectories.
2777 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2779 Description: System hang on startup
2780 Details : This bug allowed the liblustre (e.g. catamount) client to
2781 return to the app before handling all startup RPCs. This
2782 could leave the node unresponsive to lustre network traffic
2783 and manifested as a server ptllnd timeout.
2785 Severity : enhancement
2787 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2788 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2789 boolean to control whether to print peer debug info when a
2790 client's RPC times out.
2795 Description: make O_SYNC working on 2.6 kernels
2796 Details : 2.6 kernels use different method for mark pages for write,
2797 so need add a code to lustre for O_SYNC work.
2802 Description: Failure to close file and release space on NFS
2803 Details : Put inode details into lock acquired in ll_intent_file_open.
2804 Use mdc_intent_lock in ll_intent_open to properly
2805 detect all kind of errors unhandled by mdc_enqueue.
2810 Description: proc file read during shutdown sometimes raced obd removal,
2812 Details : Add lock to prevent obd access after proc file removal.
2815 Frequency : Only for files larger than 4GB on 32-bit clients.
2817 Description: improperly doing page alignment of locks
2818 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2819 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2824 Description: RPCs being resent when they shouldn't be
2825 Details : Some RPCs that should not be resent are being resent. This
2826 can cause inconsistencies in the RPC state machine. Do not
2827 resend such requests.
2830 Frequency : rare, only with NFS export
2832 Description: Crash on NFS re-export node
2833 Details : under very unusual load conditions an assertion is hit in
2834 ll_intent_file_open()
2837 Frequency : only if OST filesystem is corrupted
2839 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2840 Details : for a short time RPCs with bulk IO are in the replay list,
2841 but replay of bulk IOs is unimplemented. If the OST filesystem
2842 is corrupted due to disk cache incoherency and then replay is
2843 started it is possible to trip an assertion. Avoid putting
2844 committed RPCs into the replay list at all to avoid this issue.
2849 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2850 Details : Large single O_DIRECT read and write calls can fail to allocate
2851 a sufficiently large buffer to process the request. In case of
2852 allocation failure the allocation is retried with a smaller
2853 buffer and broken into smaller requests.
2855 Severity : enhancement
2857 Description: Add -o localflock option to simulate old noflock behaviour.
2858 Details : This will achieve local-only flock/fcntl locks coherentness.
2863 Description: versioning check is incomplete
2864 Details : Checking the version difference of client vs. server, report
2865 error if the gap is too big.
2870 Description: add support PG_writeback bit
2871 Details : add support for PG_writeback bit for Lustre, for more carefull
2872 work with page cache in 2.6 kernel. This also fix some deadlocks
2873 and remove hack for work O_SYNC with 2.6 kernel.
2875 Severity : enhancement
2877 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2878 Details : The uninit_groups feature works in conjunction with the kernel
2879 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2880 up the pass1 processing of e2fsck. This is a read-only feature
2881 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2882 mount filesystems that have had this feature enabled.
2884 Severity : enhancement
2886 Description: Improve multi-block allocation algorithm to avoid fragmentation
2887 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2888 allocation locality and avoid filesystem fragmentation.
2890 ------------------------------------------------------------------------------
2892 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2894 * Support for kernels:
2895 2.6.9-42.0.3.EL (RHEL 4)
2896 2.6.5-7.276 (SLES 9)
2897 2.4.21-47.0.1.EL (RHEL 3)
2898 2.6.12.6 vanilla (kernel.org)
2899 2.6.16.21-0.8 (SLES10)
2900 * Recommended e2fsprogs version: 1.39.cfs2-0
2902 * The backwards-compatible /proc/sys/portals symlink has been removed
2903 in this release. Before upgrading, please ensure that you change
2904 any configuration scripts or /etc/sysctl.conf files that access
2905 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2906 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2907 in advance of the upgrade on any system running Lustre 1.4.6 or
2908 newer, since /proc/sys/lnet was added in that version.
2909 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2913 Frequency : only when quota is used
2915 Description: avoid scanning export list for quota master
2916 Details : Change the algorithms to avoid scanning export list in order
2917 to improve the efficiency.
2920 Frequency : MDS failover only, very rarely
2922 Description: "went back in time" messages on mds failover
2923 Details : The greatest transno may be lost when the current operation
2924 finishes with an error (transno==0) and the client's last_rcvd
2925 record is over-written. Save the greatest transno in the
2926 mds_last_transno for this case.
2929 Frequency : always for specific kernels and striping counts
2931 Description: client may get "Matching packet too big" without ACL support
2932 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2933 when trying to access files in certain configurations. The
2934 clients should in fact be denied when mounting because they do
2935 not understand ACLs.
2938 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2940 Description: many clients connecting with IO in progress causes connect timeouts
2941 Details : Avoid synchronous journal commits to avoid delays caused by many
2942 clients connecting/disconnecting when bulk IO is in progress.
2943 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2944 OST_IO_PORTAL to avoid delays behind potentially many pending
2948 Frequency : occasionally with multiple writers to a single file
2950 Description: shared writes to file may result in wrong size reported by stat()
2951 Details : Allow growing of kms when extent lock is cancelled
2954 Frequency : always with random mmap IO to multi-striped file
2956 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2957 Details : Only if the hole is at the end of OST object so that kms is too
2958 small. Fix is to increase kms accordingly in ll_nopage.
2961 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2963 Description: writes to a missing object would leak memory on the OST
2964 Details : If there is an inconsistency between the MDS and OST filesystems,
2965 such that the MDS references an object that doesn't exist, writes
2966 to that object will leak memory due to incorrect cleanup in the
2967 error handling path, eventually running out of memory on the OST.
2972 Description: Creating too long symlink causes lustre errors
2973 Details : Check symlink and name lengths before sending requests to MDS.
2976 Frequency : only if flock is enabled (not on by default)
2978 Description: posix locks not released on fd closure on 2.6.9+
2979 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2980 kernels, this caused such locks not to be released on fd close and
2981 then assertions on fs unmount about still used locks.
2984 Frequency : MDS failover only, very rarely
2986 Description: clients may get ASSERTION(granted_lock != NULL)
2987 Details : When request was taking a long time, and a client was resending
2988 a getattr by name lock request. The were multiple lock requests
2989 with the same client lock handle and
2990 mds_getattr_name->fixup_handle_for_resent_request found one of the
2991 lock handles but later failed with ASSERTION(granted_lock != NULL).
2996 Description: handle->h_buffer_credits > 0, assertion failure
2997 Details : h_buffer_credits is zero after truncate, causing assertion
2998 failure. This patch extends the transaction or creates a new
3002 Frequency : NFS re-export or patchless client
3003 Bugzilla : 11179, 10796
3004 Description: Crash on NFS re-export node (__d_move)
3005 Details : We do not want to hash the dentry if we don't have a lock.
3006 But if this dentry is later used in d_move, we'd hit uninitialised
3007 list head d_hash, so we just do this to init d_hash field but
3008 leave dentry unhashed.
3011 Frequency : NFS re-export or patchless client
3013 Description: NFS exports has problem with symbolic link
3014 Details : lustre client didn't properly install dentry when re-exported
3015 to NFS or running patchless client.
3018 Frequency : NFS re-export or patchless client
3020 Description: Various nfs/patchless fixes.
3021 Details : fixes reuse disconected alias for lookup process - this fixes
3022 warning "find_exported_dentry: npd != pd",
3023 fix permission error with open files at nfs.
3024 fix apply umask when do revalidate.
3027 Frequency : occasional
3029 Description: Crash on NFS re-export node
3030 Details : calling clear_page() on the wrong pointer triggered oops in
3031 generic_mapping_read().
3034 Frequency : rarely, using O_DIRECT IO
3036 Description: unaligned directio crashes client with LASSERT
3037 Details : check for unaligned buffers before trying any requests.
3040 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3042 Description: stale data returned from RAID cache
3043 Details : If only a small amount of IO is done to the RAID device before
3044 reading it again it is possible to get stale data from the RAID
3045 cache instead of reading it from disk.
3048 Frequency : always for sles10 kernel
3050 Description: sles10 support
3051 Details : ll_follow_link: compile fixes and using of nd_set_link
3052 under newer kernels.
3055 Frequency : depends on arch, kernel and compiler version, always on sles10
3058 Description: recursive or deep enough symlinks cause stack overflow
3059 Details : getting rid of large stack-allocated variable in
3063 Frequency : depends on hardware
3065 Description: lustre write performance loss in the SLES10 kernel
3066 Details : the performance loss is caused by using of write barriers in the
3067 ext3 code. The SLES10 kernel turns barrier support on by
3068 default. The fix is to undo that change for ldiskfs.
3070 ------------------------------------------------------------------------------
3072 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3074 * Support for kernels:
3075 2.6.9-42.0.3EL (RHEL 4)
3076 2.6.5-7.276 (SLES 9)
3077 2.4.21-47.0.1.EL (RHEL 3)
3078 2.6.12.6 vanilla (kernel.org)
3082 Frequency : quota enabled and large files being deleted
3084 Description: releasing more than 4GB of quota at once hangs OST
3085 Details : If a user deletes more than 4GB of files on a single OST it
3086 will cause the OST to spin in an infinite loop. Release
3087 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3092 Description: statfs data retrieved from /proc may be stale or zero
3093 Details : When reading per-device statfs data from /proc, in the
3094 {kbytes,files}_{total,free,avail} files, it may appear
3095 as zero or be out of date.
3098 Frequency : systems with MD RAID1 external journal devices
3100 Description: lconf's call to blkid is confused by RAID1 journal devices
3101 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3102 of returning all block devices that match the journal UUID.
3105 Frequency : always, for aggregate stripe size over 4GB
3107 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3108 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3109 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3112 Frequency : always if "lfs find" used on a local file/directory
3114 Description: "lfs find" segfaults if used on a local file/directory
3115 Details : The case where a directory component was not specified wasn't
3116 handled correctly. Handle this properly.
3119 Frequency : always on ppc64
3121 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3122 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3123 perform endianness conversion.
3126 Frequency : rarely (truncate to non-zero file size after write under load)
3127 Bugzilla : 10730, 10687
3128 Description: Files padded with zeros to next 4K multiple
3129 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3130 it is possible that files that are truncated to a non-zero size
3131 immediately after being written are filled with zero bytes beyond
3132 the truncated size. No file data is lost.
3134 Severity : enhancement
3136 Description: Allow recovery/failover for liblustre clients.
3137 Details : liblustre clients were unaware of failover configurations until
3140 Severity : enhancement
3142 Description: user file locks should fail when not mounting with flock option
3143 Details : Set up an error-returning stub in ll_file_operations.lock field
3144 to prevent incorrect behaviour when client is mounted without
3145 flock option. Also, set up properly f_op->flock field for
3149 Frequency : always on ia64
3151 Description: "lfs df" loops on printing out MDS statfs information
3152 Details : The obd_ioctl_data was not initialized and in some systems
3153 this caused a failure during the ioctl that did not return
3154 an error. Initialize the struct and return an error on failure.
3157 Frequency : SLES 9 only
3159 Description: Error of copying files with lustre special EAs as root
3160 Details : Client side always return success for setxattr call for lustre
3161 special xattr (currently only "trusted.lov").
3164 Frequency : rarely on clusters with both ia64+i386 clients
3166 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3167 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3168 clients are doing concurrent writes to the same file it is
3169 possible that the ia64 clients may LASSERT because the OST
3170 extent locks are not PAGE_SIZE aligned. Ensure that grown
3171 locks are always aligned on the request boundary.
3174 Frequency : specific use, occasional
3176 Description: Overwriting in use executable truncates on-disk binary image
3177 Details : If one node attempts to overwrite an executable in use by
3178 another node, we now correctly return ETXTBSY instead of
3179 truncating the file.
3181 Severity : enhancement
3183 Description: Async OSC create to avoid the blocking unnecessarily.
3184 Details : If a OST has no remain object, system will block on the creating
3185 when need to create a new object on this OST. Now, ways use
3186 pre-created objects when available, instead of blocking on an
3187 empty osc while others are not empty. If we must block, we block
3188 for the shortest possible period of time.
3193 Description: chmod on Lustre root is propagated to other clients
3194 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3195 invalid by the follow_mount time.
3198 Frequency : liblustre clients only
3200 Description: Race in 'instant cancel' lock handling could lead to such locks
3201 never to be granted in case of SMP MDS
3202 Details : Do not destroy not yet granted but cbpending locks in
3206 Frequency : replay/resend of open
3208 Description: non null lock assetion failure in mds_intent_policy
3209 Details : Trying to replay/resend lockless open requests resulted in
3210 mds_open() returning 0 with no lock. Now it sets a flag if
3211 a lock is going to be returned.
3213 Severity : enhancement
3215 Description: Checksum enhancements
3216 Details : New checksum enhancements allow for resending RPCs that failed
3219 Severity : enhancement
3221 Description: Tunables on number of dirty pages in cacche
3222 Details : Allow to set limit on number of dirty pages cached.
3227 Description: client crash on unmount - lock still has references
3228 Details : In some error handling cases it was possible to leak a lock
3229 reference on a client while accessing a file. This was not
3230 harmful to the client during operation, but would cause the
3231 client to crash when the filesystem is unmounted.
3234 Frequency : specific case, rare
3236 Description: ETXTBSY on mds though file not in use by client
3237 Details : ETXTBSY is no longer incorrectly returned when attempting to
3238 chmod or chown a directory that the user previously tried to
3239 execute or a currently-executing binary.
3242 Frequency : extremely rare except on liblustre-based clients
3244 Description: Lustre space not freed when files are deleted
3245 Details : Clean up open-unlinked files after client eviction. Previously
3246 the unlink was skipped and the files remained as orphans.
3251 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3252 Details : In some cases it was possible to send a blocking callback to
3253 a client doing a glimpse, even though that client didn't get
3254 a lock granted. When the glimpse lock is cancelled on the OST
3255 the freed lock is left on the waiting list and corrupted the list.
3258 Frequency : all core dumps
3260 Description: Broke core dumps to lustre
3261 Details : Negative dentry may be unhashed if parent does not have UPDATE
3262 lock, but some callers, e.g. do_coredump, expect dentry to be
3263 hashed after successful create, hash it in ll_create_it.
3265 ------------------------------------------------------------------------------
3267 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3269 * Support for kernels:
3270 2.6.9-42.0.2.EL (RHEL 4)
3271 2.6.5-7.276 (SLES 9)
3272 2.4.21-40.EL (RHEL 3)
3273 2.6.12.6 vanilla (kernel.org)
3277 Frequency : always on RHEL 3
3279 Description: Number of open files grows over time
3280 Details : The number of open files grows over time, whether or not
3281 Lustre is started. This was due to a filp leak introduced
3282 by one of our kernel patches.
3284 ------------------------------------------------------------------------------
3286 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3288 * Support for kernels:
3289 2.6.9-42.EL (RHEL 4)
3290 2.6.5-7.276 (SLES 9)
3291 2.4.21-40.EL (RHEL 3)
3292 2.6.12.6 vanilla (kernel.org)
3297 Bugzilla : 5719, 9635, 9792, 9684
3298 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3299 Details : If a server is under heavy load and cannot reply to new
3300 connection requests before the client resends the (re)connect,
3301 the connection handling code can behave badly if two service
3302 threads are concurrently handing separate (re)connections from
3303 the same client. Add better locking to the connection handling
3304 code, and ensure that only a single connection will be processed
3305 for a given client UUID, even if the lock is dropped.
3307 Severity : enhancement
3309 Description: add TCP zero-copy support to kernel
3310 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3311 sends if the hardware supports scatter-gather and checksumming.
3312 This allows socklnd to do client-write and server-read more
3313 efficiently and reduce CPU utilization from skbuf copying.
3316 Frequency : only if NFS exporting from client
3318 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3319 Details : If setting attributes on a file created under NFS that had
3320 never been opened it would be possible to oops the client
3321 if the file had no objects.
3324 Frequency : always for liblustre
3326 Description: liblustre client does MDS+OSTs setattr RPC for each write
3327 Details : When doing a write from a liblustre client, the client
3328 incorrectly issued an RPC to the MDS and each OST the file was
3329 striped over in order to update the timestamps. When writing
3330 with small chunks and many clients this could overwhelm the MDS
3331 with RPCs. In all cases it would slow down the write because
3332 these RPCs are unnecessary.
3334 Severity : enhancement
3336 Description: allow number of MDS service threads to be changed at module load
3337 Details : It is now possible to change the number of MDS service threads
3338 running. Adding "options mds mds_num_threads={N}" to the MDS's
3339 /etc/modprobe.conf will set the number of threads for the next
3340 time Lustre is restarted (assuming the "mds" module is also
3341 reloaded at that time). The default number of threads will
3342 stay the same, 32 for most systems.
3347 Description: OST crash if filesystem is unformatted or corrupt
3348 Details : If an OST is started on a device that has never been formatted
3349 or if the filesystem is corrupt and cannot even mount then the
3350 error handling cleanup routines would dereference a NULL pointer.
3355 Description: NULL pointer deref in llap_from_page.
3356 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3357 incorrect) mapping if the page was truncated/reclaimed while it was
3358 searched for. Check for this condition and skip such pages when
3359 doing readahead. Introduce extra check to llap_from_page() to
3360 verify page->mapping->host is non-NULL (so page is not anonymous).
3363 Frequency : Sometimes when using sys_sendfile
3365 Description: "page not covered by a lock" warnings from ll_readpage
3366 Details : sendfile called ll_readpage without right page locks present.
3367 Now we introduced ll_file_sendfile that does necessary locking
3368 around call to generic_file_sendfile() much like we do in
3372 Frequency : with certain MDS communication failures at client mount time
3374 Description: NULL pointer deref after failed client mount
3375 Details : a client connection request may delayed by the network layer
3376 and not be sent until after the PTLRPC layer has timed out the
3377 request. If the client fails the mount immediately it will try
3378 to clean up before the network times out the request. Add a
3379 reference from the request import to the obd device and delay
3380 the cleanup until the network drops the request.
3383 Frequency : occasionally during client (re)connect
3385 Description: assertion failure during client (re)connect
3386 Details : processing a client connection request may be delayed by the
3387 client or server longer than the client connect timeout. This
3388 causes the client to resend the connection request. If the
3389 original connection request is replied in this interval, the
3390 client may trip an assertion failure in ptlrpc_connect_interpret()
3391 which thought it would be the only running connect process.
3394 Frequency : only with obd_echo servers and clients that are rebooted
3396 Description: kernel BUG accessing uninitialized data structure
3397 Details : When running an obd_echo server it did not start the ping_evictor
3398 thread, and when a client was evicted an uninitialized data
3399 structure was accessed. Start the ping_evictor in the RPC
3400 service startup instead of the OBD startup.
3402 Severity : enhancement
3403 Bugzilla : 10193 (patchless)
3404 Description: Remove dependency on various unexported kernel interfaces.
3405 Details : No longer need reparent_to_init, exit_mm, exit_files,
3406 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3409 Frequency : rare (only users of deprecated and unsupported LDAP config)
3411 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3412 Details : LDAP apparently contains 'lustreName' attributes instead of
3413 'name'. A simple remapping of the name is sufficient.
3416 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3418 Description: waiting_locks_callback trips kernel BUG if client is evicted
3419 Details : Running with the dump_on_timeout debug flag turned on makes
3420 it possible that the waiting_locks_callback() can try to dump
3421 the Lustre kernel debug logs from an interrupt handler. Defer
3422 this log dumping to the expired_lock_main() thread.
3424 Severity : enhancement
3426 Description: Support NFS exporting on 2.6 kernels.
3427 Details : Implement non-rawops metadata methods for NFS server to use without
3428 changing NFS server code.
3431 Frequency : very rare (synthetic metadata workload only)
3433 Description: two racing renames might cause an MDS thread to deadlock
3434 Details : Running the "racer" program may cause one MDS thread to rename
3435 a file from being the source of a rename to being the target of
3436 a rename at exactly the same time that another thread is doing
3437 so, and the second thread has already enqueued these locks after
3438 doing a lookup of the target and is trying to relock them in
3439 order. Ensure that we don't try to re-lock the same resource.
3442 Frequency : only very large systems with liblustre clients
3444 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3445 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3446 all OSTs in parallel. This allows the eviction of stale liblustre
3447 clients to proceed much faster than if they were done in series,
3448 and also offers similar improvements for other set_info RPCs.
3453 Description: excessive CPU usage during initial read phase on client
3454 Details : During the initial read phase on a client, it would agressively
3455 retry readahead on the file, consuming too much CPU and impacting
3456 performance (since 1.4.5.8). Improve the readahead algorithm
3457 to avoid this, and also improve some other common cases (read
3458 of small files in particular, where "small" is files smaller than
3459 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3464 Description: MDS crash when receiving packet with unknown intent.
3465 Details : Do not LBUG in unknown intent case, just return -EFAULT
3467 Severity : enhancement
3468 Bugzilla : 9293, 9385
3469 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3470 Details : Do not serialize getattr (non-intent version) and statfs.
3473 Frequency : occasional, when OST network is overloaded/intermittent
3475 Description: client evicted by OST after bulk IO timeout
3476 Details : If a client sends a bulk IO request (read or write) the OST
3477 may evict the client if it is unresposive to its data GET/PUT
3478 request. This is incorrect if the network is overloaded (takes
3479 too long to transfer the RPC data) or dropped the OST GET/PUT
3480 request. There is no need to evict the client at all, since
3481 the pinger and/or lock callbacks will handle this, and the
3482 client can restart the bulk request.
3485 Frequency : Always when mmapping file with no objects
3487 Description: client crashes when mmapping file with no objects
3488 Details : Check that we actually have objects in a file before doing any
3489 operations on objects in ll_vm_open, ll_vm_close and
3495 Description: Request leak when working with deleted CWD
3496 Details : Introduce advanced request refcount tracking for requests
3497 referenced from lustre intent.
3499 Severity : Enhancement
3501 Description: Cache open file handles on client.
3502 Details : MDS now will return special lock along with openhandle, if
3503 requested and client is allowed to hold openhandle, even if unused,
3504 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3505 closing files for every read/write openration.
3507 Severity : Enhancement
3509 Description: Cache open negative dentries on client when possible.
3510 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3511 negative dentries on lock revocation.
3516 Description: Remounting a client read-only wasn't possible with a zconf mount
3517 Details : It wasn't possible to remount a client read-only with llmount.
3519 Severity : enhancement
3520 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3521 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3522 MPICH is included to add Lustre-specific ADIO interfaces.
3523 This is based closely on the UFS ADIO layer and only differs
3524 in file creation, in order to allow the OST striping to be set.
3525 This is user-contributed code and not supported by CFS.
3530 Description: extended inode attributes (immutable, append-only) work improperly
3531 when 2.4 and 2.6 kernels are used on client/server or vice versa
3532 Details : Introduce kernel-independent values for these flags.
3534 Severity : enhancement
3537 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3538 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3539 granularity. Now, Fractional MB granularity is be supported,
3540 this is very useful for low memory system.
3542 Severity : enhancement
3544 Description: Getattr by fid
3545 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3546 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3547 have locks and avoids getattr rpc requests for it on every stat.
3550 Frequency : Always, for filesystems larger than 2TB
3552 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3553 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3554 filesystems larger than 2TB to be mounted with mballoc enabled.
3557 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3559 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3560 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3561 filesystems larger than 2TB is mounted then the kernel will
3562 silently corrupt the start of the filesystem. CONFIG_LBD is
3563 enabled for all CFS-supported kernels, but the possibility of
3564 this happening with a modified kernel config exists.
3566 Severity : enhancement
3568 Description: add client O_DIRECT support for 2.6 kernels
3569 Details : It is now possible to do O_DIRECT reads and writes to files
3570 in the Lustre client mountpoint on 2.6 kernel clients.
3572 Severity : enhancement
3574 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3575 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3576 parallel, not waiting for response from every OST before sending
3577 a rpc to the next OST.
3582 Description: setattr vs write race when updating file timestamps
3583 Details : Client processes that update a file timestamp into the past
3584 right after writing to the file (e.g. tar) it is possible that
3585 the updated file modification time can be reset to the current
3586 time due to a race between processing the setattr and write RPC.
3588 Severity : enhancement
3590 Description: Bring 'lfs find' closer in line with regular Linux find.
3591 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3592 -print0 options and obtains all the needed info through the lustre
3595 Severity : enhancement
3597 Description: support up to 1024 configured devices on one node
3598 Details : change obd_dev array from statically allocated to dynamically
3599 allocated structs as they are first used to reduce memory usage
3604 Description: Flush dirty partially truncated pages during truncate
3605 Details : Immediatelly flush partially truncated pages in filter_setattr,
3606 this way we completely avoid having any pages in page cache on OST
3607 and can retire ugly workarounds during writes to flush such pages.
3612 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3613 Details : protect inode from truncation within vfs_unlink() context
3614 just take a reference before calling vfs_unlink() and release it
3615 when parent's i_sem is free.
3620 Description: last_id value checked outside lock on OST caused LASSERT failure
3621 Details : If there were multiple MDS->OST object precreate requests in
3622 flight, it was possible that the OST's last object id was checked
3623 outside a lock and incorrectly tripped an assertion. Move checks
3624 inside locks, and discard old precreate requests.
3627 Frequency : always, if extents are used on OSTs
3629 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3630 Details : OSTs using the extents format would not zero the high 16 bits of
3631 the index physical block number. This is not a problem for any
3632 OST filesystems smaller than 16TB, and no kernels support ext3
3633 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3634 new/modified files) and can be fixed for existing filesystems
3635 with e2fsprogs-1.39-cfs1.
3640 Description: import connection selection may be incorrect if timer wraps
3641 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3642 import connections to be ignored if the 32-bit jiffies counter
3643 wraps. Use a 64-bit jiffies counter.
3646 Frequency : very large clusters immediately after boot
3648 Description: LNET request buffers exhausted under heavy short-term load
3649 Details : If a large number of client requests are generated on a service
3650 that has previously never seen so many requests it is possible
3651 that the request buffer growth cannot keep up with the spike in
3652 demand. Instead of dropping incoming requests, they are held in
3653 the LND until the RPC service can accept more requests.
3656 Frequency : Sometimes during replay
3658 Description: Assertion failure in ll_local_open after replay.
3659 Details : If replay happened on an open request reply before we were able
3660 to set replay handler, reply will become not swabbed tripping the
3661 assertion in ll_local_open. Now we set the handler right after
3662 recognising of open request
3665 Frequency : very rare
3667 Description: kernel reports "badness in vsnprintf"
3668 Details : Reading from the "recovery_status" /proc file in small chunks
3669 may cause a negative length in lprocfs_obd_rd_recovery_status()
3670 call to vsnprintf() (which is otherwise harmless). Exit early
3671 if there is no more space in the output buffer.
3673 Severity : enhancement
3675 Description: clear OBD RPC statistics by writing to them
3676 Details : It is now possible to clear the OBD RPC statistics by writing
3677 to the "stats" file.
3682 Description: Client mtime is not the same on different clients after utimes
3683 Details : In some cases, the client was using the utimes() syscall on
3684 a file cached on another node. The clients now validate the
3685 ctime from the MDS + OSTs to determine which one is right.
3690 Description: Inability to activate failout mode
3691 Details : lconf script incorrectly assumed that in pythong string's numeric
3692 value is used in comparisons.
3695 Frequency : always with multiple stripes per file
3697 Description: Inefficient object allocation for mutli-stripe files
3698 Details : When selecting which OSTs to stripe files over, for files with
3699 a stripe count that divides evenly into the number of OSTs,
3700 the MDS is always picking the same starting OST for each file.
3701 Return the OST selection heuristic to the original design.
3706 Description: mount failures may take full timeout to return an error
3707 Details : Under some heavy load conditions it is possible that a
3708 failed mount can wait for the full obd_timeout interval,
3709 possibly several minutes, before reporting an error.
3710 Instead return an error as soon as the status is known.
3712 Frequency : quota enabled and large files being deleted
3714 Description: releasing more than 4GB of quota at once hangs OST
3715 Details : If a user deletes more than 4GB of files on a single OST it
3716 will cause the OST to spin in an infinite loop. Release
3717 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3722 Description: statfs data retrieved from /proc may be stale or zero
3723 Details : When reading per-device statfs data from /proc, in the
3724 {kbytes,files}_{total,free,avail} files, it may appear
3725 as zero or be out of date.
3728 Frequency : systems with MD RAID1 external journal devices
3730 Description: lconf's call to blkid is confused by RAID1 journal devices
3731 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3732 of returning all block devices that match the journal UUID.
3735 Frequency : always, for aggregate stripe size over 4GB
3737 Description: assertion fails when trying to use 4GB stripe size
3738 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3739 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3742 Frequency : always on ppc64
3744 Description: the first write on an ext3 filesystem with mballoc got stuck
3745 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3746 perform endianness conversion.
3748 ------------------------------------------------------------------------------
3750 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3752 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3753 INTEROPERATE with older versions automatically. Please read the
3754 user documentation before upgrading any part of a live system.
3755 * WARNING: Lustre networking configuration changes are required with
3756 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3759 * Support for kernels:
3760 2.6.9-22.0.2.EL (RHEL 4)
3761 2.6.5-7.244 (SLES 9)
3762 2.6.12.6 vanilla (kernel.org)
3765 Severity : enhancement
3766 Bugzilla : 7981/8208
3767 Description: Introduced Lustre Networking (LNET)
3768 Details : LNET is new networking infrastructure for Lustre, it includes
3769 a reorganized network configuration mode (see the user
3770 documentation for full details) as well as support for routing
3771 between different network fabrics. Lustre Networking Devices
3772 (LNDs) for the supported network fabrics have also been
3773 created for this new infrastructure.
3775 Severity : enhancement
3776 Description: Introduced Access control lists
3777 Details : clients can set ACLs on files and directories in order to have
3778 more fine-grained permissions than the standard Unix UGO+RWX.
3779 The MDS must be started with the "-o acl" mount option.
3781 Severity : enhancement
3782 Description: Introduced filesystem quotas
3783 Details : Administrators may now establish per-user quotas on the
3786 Severity : enhancement
3788 Description: Configuration change for the XT3
3789 The PTLLND is now used to run Lustre over Portals on the XT3
3790 The configure option(s) --with-cray-portals are no longer used.
3791 Rather --with-portals=<path-to-portals-includes> is used to
3792 enable building on the XT3. In addition to enable XT3 specific
3793 features the option --enable-cray-xt3 must be used.
3798 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3799 Details : A race between allocating a new llcd and re-getting the llcd_lock
3800 allowed another thread to grab newly-allocated llcd.
3802 Severity : enhancement
3804 Description: 2.6 OST async journal commit and locking fix to improve performance
3805 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3806 2.6 kernels are now async as they already were in 2.4 kernels so
3807 that they can commit concurrently with the network bulk transfer.
3808 For block-allocated files the filter allocation semaphore is held
3809 to avoid filesystem fragmentation during allocation. BKL lock
3810 removed for 2.6 xattr operations where it is no longer needed.
3815 Description: lconf incorrectly determined whether two IP networks could talk
3816 Details : In some more complicated routing and multiple-network
3817 configurations, lconf will avoid trying to make a network
3818 connection to a disjoint part of the IP space. It was doing the
3819 math incorrectly for one set of cases.
3824 Description: Fix for potential infinite loop processing records in an llog.
3825 Details : If an llog record is corrupted/zeroed, it is possible to loop
3826 forever in llog_process(). Validate the llog record length
3827 and skip the remainder of the block on error.
3830 Frequency : occasional (liblustre only)
3832 Description: liblustre could not open files whose last component is a symlink
3833 Details : sysio_path_walk() would incorrectly pass the open intent to
3834 intermediate path components.
3837 Frequency : rare (liblustre only with non-standard tuning)
3838 Bugzilla : 7201 (7350)
3839 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3840 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3841 after it has been released, only do this for referenced locks
3843 Severity : enhancement
3845 Description: specifying an (invalid) directory default stripe_size of -1
3846 would reset the directory default striping
3847 Details : stripe_size -1 was used internally to signal directory stripe
3848 removal, now use "all default" to signal dir stripe removal
3849 as a directory striping of "all default" is not useful
3852 Frequency : common for large clusters running liblustre clients
3854 Description: doing an ls when liblustre clients are running is slow
3855 Details : sending a glimpse AST to a liblustre client waits for every AST
3856 to time out, as liblustre clients will not respond. Since they
3857 cannot cache data we refresh the OST lock LVB from disk instead.
3859 Severity : enhancement
3861 Description: doing an ls at the same time as file IO can be slow
3862 Details : enqueue and other "small" requests can be blocked behind many
3863 large IO requests. Create a new OST IO portal for non-IO
3864 requests so they can be processed faster.
3867 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3869 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3870 created file with mode 000
3871 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3872 allow an ftruncate() as a non-root user to a file with mode 000.
3873 Lustre now respects this flag to disable mode checks when
3874 truncating a file owned by the user
3877 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3880 Description: Revoking locks from clients that went dead or catatonic might take
3882 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3883 cancellation of such locks instant on servers without waiting for
3884 any reply from clients. Clients drops these locks when cancel
3885 notification from server is received without replying.
3888 Frequency : liblustre-only, when liblustre client dies or becomes busy
3890 Description: Doing ls on Linux clients can take a long time with active
3892 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3893 granting such locks to it in the first place if possible. Locks
3894 are taken by proxy on the OST during the read or write and
3895 dropped immediately afterward. Add connect flags handling, do
3896 not grant locks to liblustre clients for glimpse ASTs.
3898 Severity : enhancement
3900 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3901 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3902 sequential reads and behaves badly with applications doing random
3903 reads. Improve it by reading ahead at least read region, and
3904 avoiding excessive large RPC for small reads.
3906 Severity : enhancement
3908 Description: Creating more than 1000 files for a single job may cause a load
3909 imbalance on the OSTs if there are also a large number of OSTs.
3910 Details : qos_prep_create() uses an OST index reseed value that is an
3911 even multiple of the number of available OSTs so that if the
3912 reseed happens in the middle of the object allocation it will
3913 still utilize the OSTs as uniformly as possible.
3918 Description: OST or MDS may oops in ping_evictor_main()
3919 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3920 but doesn't restart at beginning of obd_exports_timed list
3923 Severity : enhancement
3925 Description: improve by-nid export eviction on the MDS and OST
3926 Details : allow multiple exports with the same NID to be evicted at one
3927 time without re-searching the exports list.
3930 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3932 Description: MDS may oops in groups_free()
3933 Details : in rare race conditions a newly allocated group_info struct is
3934 freed again, and this can be NULL. The 2.4 compatibility code
3935 for groups_free() checked for a NULL pointer, but 2.6 did not.
3938 Frequency : common for liblustre clients doing little filesystem IO
3939 Bugzilla : 9352, 7313
3940 Description: server may evict liblustre clients accessing contended locks
3941 Details : if a client is granted a lock or receives a completion AST
3942 with a blocking AST already set it would not reply to the AST
3943 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3946 Frequency : lfs setstripe, only systems with more than 160 OSTs
3948 Description: unable to set striping with a starting offset beyond OST 160
3949 Details : llapi_create_file() incorrectly limited the starting stripe
3950 index to the maximum single-file stripe count.
3953 Frequency : LDAP users only
3955 Description: lconf did not handle in-kernel recovery with LDAP properly
3956 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3958 Severity : enhancement
3960 Description: bind OST threads to NUMA nodes to improve performance
3961 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3962 node and do their allocations there to localize memory access
3964 Severity : enhancement
3966 Description: llmount can determine client NID directly from Myrinet (GM)
3967 Details : the client NID code from gmnalnid was moved directly into
3968 llmount, removing the need to use this or specifying the
3969 client NID explicitly when mounting GM clients with zeroconf
3972 Frequency : if client is started with down MDS
3974 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3975 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3976 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3977 ptlrpc_import_delayed_req() because we didn't check if the
3978 request was interrupted, and we also didn't break out of the
3979 event loop if there was no timeout
3984 Description: data loss during non-page-aligned writes to a single file from
3985 both multiple nodes and multiple threads on one node at same time
3986 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3987 inconsistency led to false short-reads, that were cached and later
3988 used by ->prepare_write() to fill in partially written page,
3989 leading to data loss.
3992 Frequency : always, if lconf --abort_recovery used
3994 Description: lconf --abort_recovery fails with 'Operation not supported'
3995 Details : lconf was attempting to abort recovery on the MDT device and not
3998 Severity : enhancement
4000 Description: remove cleanup logs
4001 Details : replace lconf-generated cleanup logs with lustre internal
4002 cleanup routines. Eliminates the need for client-cleanup and
4005 Severity : enhancement
4007 Description: add support for EAs (user and system) on lustre filesystems
4008 Details : it is now possible to store extended attributes in the Lustre
4009 client filesystem, and with the user_xattr mount option it
4010 is possible to allow users to store EAs on their files also
4012 Severity : enhancement
4014 Description: Add possibility (config option) to show minimal available OST free
4016 Details : When compiled with --enable-mindf configure option, statfs(2)
4017 (and so, df) will return least minimal free space available from
4018 all OSTs as amount of free space on FS, instead of summary of
4019 free spaces of all OSTs.
4021 Severity : enhancement
4023 Description: do not expand extent locks acquired on OST-side
4024 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4025 by server: they are not cached anyway.
4028 Frequency : when mmap is used/binaries executed from Lustre
4030 Description: Unmmap pages before throwing them away from read cache.
4031 Details : llap_shrink cache now attempts to unmap pages before discarding
4032 them (if unmapping failed - do not discard). SLES9 kernel has
4033 extra checks that trigger if this unmapping is not done first.
4038 Description: lconf didn't resolve symlinks before checking to see whether a
4039 given mountpoint was already in use
4042 Frequency : when migrating failover services
4043 Bugzilla : 6395, 9514
4044 Description: When migrating a subset of services from a node (e.g. failback
4045 from a failover service node) the remaining services would
4046 time out and evict clients.
4047 Details : lconf --force (implied by --failover) sets the global obd_timeout
4048 to 5 seconds in order to quickly disconnect, but this caused
4049 other RPCs to time out too quickly. Do not change the global
4050 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4052 Severity : enhancement
4053 Frequency : if MDS is started with down OST
4054 Bugzilla : 9439,5706
4055 Description: Allow startup/shutdown of an MDS without depending on the
4056 availability of the OSTs.
4057 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4058 Add appropriate locking and lov-osc refcounts for safe
4059 cleaning. Add osc abort_inflight calls in case the
4060 synchronize never started.
4063 Frequency : occasional (Cray XT3 only)
4065 Description: root not authorized to access files in CRAY_PORTALS environment
4066 Details : The client process capabilities were not honoured on the MDS in
4067 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4068 been accessed by an authorized user then root was able to access
4069 the file on the local client also. The root user capabilities
4070 are now allowed on the MDS, as this environment has secure UID.
4073 Frequency : occasional
4075 Description: ldiskfs "too long searching" message happens too often
4076 Details : A debugging message (otherwise harmless) prints too often on
4077 the OST console. This has been reduced to only happen when
4078 there are fragmentation problems on the filesystem.
4083 Description: Division by zero in statfs when all OSCs are inactive
4084 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4085 lov_statfs divides by value returned by lov_get_stripecnt().
4089 Bugzilla : 9489, 3273
4090 Description: First write from each client to each OST was only 4kB in size,
4091 to initialize client writeback cache, which caused sub-optimal
4092 RPCs and poor layout on disk for the first writen file.
4093 Details : Clients now request an initial cache grant at (re)connect time
4094 and so that they can start streaming writes to the cache right
4095 away and always do full-sized RPCs if there is enough data.
4096 If the OST is rebooted the client also re-establishes its grant
4097 so that client cached writes will be honoured under the grant.
4102 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4103 Details : Now I/O RPCs go to different portal number and (presumably) fast
4104 lock requests (and glimses) and other RPCs get their own service
4105 threads pool that should be able to service those RPCs
4108 Severity : enhancement
4110 Description: Ability to exchange lustre version between client and servers and
4111 issue warnings at client side if client is too old. Also for
4112 liblustre clients there is ability to refuse connection of too old
4114 Details : New 'version' field is added to connect data structure that is
4115 filled with version info. That info is later checked by server and
4119 Frequency : rare, liblustre only.
4120 Bugzilla : 9296, 9581
4121 Description: Two simultaneous writes from liblustre at offset within same page
4122 might proceed at the same time overwriting eachother with stale
4124 Details : I/O lock withing llu_file_prwv was released too early, before data
4125 actually was hitting the wire. Extended lock-holding time until
4126 server acknowledges receiving data.
4129 Frequency : extremely rare. Never observed in practice.
4131 Description: avoid generating lustre_handle cookie of 0.
4132 Details : class_handle_hash() generates handle cookies by incrementing
4133 global counter, and can hit 0 occasionaly (this is unlikely, but
4134 not impossible, because initial value of cookie counter is
4135 selected randonly). Value of 0 is used as a sentinel meaning
4136 "unassigned handle" --- avoid it. Also coalesce two critical
4137 sections in this function into one.
4139 Severity : enhancement
4141 Description: allow liblustre clients to delegate truncate locking to OST
4142 Details : To avoid overhead of locking, liblustre client instructs OST to
4143 take extent lock in ost_punch() on client's behalf. New connection
4144 flag is added to handle backward compatibility.
4146 Severity : enhancement
4147 Bugzilla : 4928, 7341, 9758
4148 Description: allow number of OST service threads to be specified
4149 Details : a module parameter allows the number of OST service threads
4150 to be specified via "options ost ost_num_threads={N}" in the
4151 OSS's /etc/modules.conf or /etc/modprobe.conf.
4155 Bugzilla : 6146, 9635, 9895
4156 Description: servers crash with bad pointer in target_handle_connect()
4157 Details : In rare cases when a client is reconnecting it was possible that
4158 the connection request was the last reference for that export.
4159 We would temporarily drop the export reference and get a new
4160 one, but this may have been the last reference and the export
4161 was just destroyed. Get new reference before dropping old one.
4163 Severity : enhancement
4164 Frequency : if client is started with failover MDS
4166 Description: Allow multiple MDS hostnames in the mount command
4167 Details : Try to read the configuration from all specified MDS
4168 hostnames during a client mount in case the "primary"
4171 Severity : enhancement
4173 Description: Stop sending data to evicted clients as soon as possible.
4174 Details : Check if the client we are about to send or are sending data to
4175 was evicted already. (Check is done every second of waiting,
4176 for which l_wait_event interface was extended to allow checking
4177 of exit condition at specified intervals).
4180 Frequency : rare, normally only when NFS exporting is done from client
4182 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4184 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4187 Frequency : always, if srand() is called before liblustre initialization
4189 Description: Liblustre uses system PRNG disturbing its usage by user application
4190 Details : Introduce internal to lustre fast and high-quality PRNG for
4191 lustre usage and make liblustre and some other places in generic
4192 lustre code to use it.
4194 Severity : enhancement
4195 Bugzilla : 9477, 9557, 9870
4196 Description: Verify that the MDS configuration logs are updated when xml is
4197 Details : Check if the .xml configuration logs are newer than the config
4198 logs stored on the MDS and report an error if this is the case.
4199 Request --write-conf, or allow starting with --old_conf.
4201 Severity : enhancement
4203 Description: Handle symlinks in the path when checking if Lustre is mounted.
4204 Details : Resolve intermediate symlinks when checking if a client has
4205 mounted a filesystem to avoid duplicate client mounts.
4210 Description: lconf can hit an error exception but still return success.
4211 Details : The lconf command catches the Command error exception at the top
4212 level script context and will exit with the associated exit
4213 status, but doesn't ensure that this exit status is non-zero.
4218 Description: failure of ptlrpc thread startup can cause oops
4219 Details : Starting a ptlrpc service thread can fail if there are a large
4220 number of threads or the server memory is very fragmented.
4221 Handle this without oopsing.
4224 Frequency : always, only if liblustre and non-default acceptor port was used
4226 Description: liblustre cannot connect to servers with non-default acceptor port
4227 Details : tcpnal_set_default_params() was not called and was therefore
4228 ignoring the environment varaible TCPNAL_PORT, as well as other
4229 TCPNAL_ environment variables
4234 Description: two objects could be created on the same OST for a single file
4235 Details : If an OST is down, in some cases it was possible to create two
4236 objects on a single OST for a single file. No problems other
4237 than potential performance impact and spurious error messages.
4241 Bugzilla : 5681, 9562
4242 Description: Client may oops in ll_unhash_aliases
4243 Details : Client dcache may become inconsistent in race condition.
4244 In some cases "getcwd" can fail if the current directory is
4250 Description: Inode refcounting problems in NFS export code
4251 Details : link_raw functions used to call d_instantiate without obtaining
4252 extra inode reference first.
4256 Bugzilla : 9942, 9903
4257 Description: Referencing freed requests leading to crash, memleaks with NFS.
4258 Details : We used to require that call to ll_revalidate_it was always
4259 followed by ll_lookup_it. Also with revalidate_special() it is
4260 possible to call ll_revalidate_it() twice for the same dentry
4261 even if first occurence returned success. This fix changes semantic
4262 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4263 reference on a request referred from the intent.
4264 ll_intent_release() then releases such a request.
4267 Frequency : rare, normally benchmark loads only
4269 Description: unlinked inodes were kept in memory on the client
4270 Details : If a client is repeatedly creating and unlinking files it
4271 can accumulate a lot of stale inodes in the inode slab cache.
4272 If there is no other client load running this can cause the
4273 client node to run out of memory. Instead flush old inodes
4274 from client cache that have the same inode number as a new inode.
4277 Frequency : SLES9 2.6.5 kernel and long filenames only
4278 Bugzilla : 9969, 10379
4279 Description: utime reports stale NFS file handle
4280 Details : SLES9 uses out-of-dentry names in some cases, which confused
4281 the lustre dentry revalidation. Change it to always use the
4285 Frequency : rare, unless heavy write-truncate concurrency is continuous
4286 Bugzilla : 4180, 6984, 7171, 9963, 9331
4287 Description: OST becomes very slow and/or deadlocked during object unlink
4288 Details : filter_destroy() was holding onto the parent directory lock
4289 while truncating+unlinking objects. For very large objects this
4290 may block other threads for a long time and slow overall OST
4291 responsiveness. It may also be possible to get a lock ordering
4292 deadlock in this case, or run out of journal credits because of
4293 the combined truncate+unlink. Solution is to do object truncate
4294 first in one transaction without parent lock, and then do the
4295 final unlink in a new transaction with the parent lock. This
4296 reduces the lock hold time dramatically.
4299 Frequency : rare, 2.4 kernels only
4301 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4302 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4303 before unmounting their filesystem in order to avoid deadlock.
4304 The kernel_locked() function in 2.4 kernels only checks whether
4305 the kernel lock is held, not whether it is this process that is
4306 holding it as 2.6 kernels do.
4311 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4312 Details : The client ptlrpc code may be trying to reconnect to a down
4313 server before a previous connection attempt has timed out.
4314 Increase the reconnect interval to be longer than the connection
4315 timeout interval to avoid sending duplicate connections to
4319 Frequency : echo_client brw_test command
4321 Description: fix echo_client to work with OST preallocated code
4322 Details : OST preallocation code (5137) didn't take echo_client IO path
4323 into account: echo_client calls filter methods outside of any
4324 OST thread and, hence, there is no per-thread preallocated
4325 pages and buffers to use. Solution: hijack pga pages for IO. As
4326 a byproduct, this avoids unnecessary data copying.
4330 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4331 Description: Client can oops in mdc_commit_close() after open replay
4332 Details : It was possible for the MDS to return an open request with no
4333 transaction number in mds_finish_transno() if the client was
4334 evicted, but without actually returning an error. Clients
4335 would later try to replay that open and may trip an assertion
4336 Simplify the client close codepath, and always return an error
4337 from the MDS in case the open is not successful.
4340 Frequency : rare, 2.6 OSTs only
4342 Description: OST may deadlock under high load on fragmented files
4343 Details : If there was a heavy load and highly-fragmented OST filesystems
4344 it was possible to have all the OST threads deadlock waiting on
4345 allocation of biovecs, because the biovecs were not released
4346 until the entire RPC IO was completed. Instead, release biovecs
4347 as soon as they are complete to ensure forward IO progress.
4349 Severity : enhancement
4351 Description: Support for specifying external journal device at mount
4352 Details : If an OST or MDS device is formatted with an external journal
4353 device, this device major/minor is stored in the ext3 superblock
4354 and may not be valid for failover. Allow detecting and
4355 specifying the external journal at mount time.
4360 Description: Mounting an MDS with pending unlinked files may cause oops
4361 Details : target_finish_recovery() calls mds_postrecov() which returned
4362 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4363 considers this an error and immediately begins cleaning up the
4364 lov, just after starting the mds_lov process
4366 Severity : enhancement
4368 Description: Implement 'lfs df' to report actual free space on per-OST basis
4369 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4370 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4371 usage of objects; '-h' to report in human readable format.
4373 ------------------------------------------------------------------------------
4375 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4382 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4383 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4384 at startup during orphan unlinking. Instead, skip buddy update
4385 before setup, as it will be regenerated after recovery is complete.
4390 Description: If an OST is inactive, its locks might reference stale inodes.
4391 Details : lov_change_cbdata() must iterate over all namespaces, even if
4392 they are inactive to clear inode references from the lock.
4394 Severity : enhancement
4395 Frequency : occasional, if non-standard max_dirty_mb used
4397 Description: Client will block write RPCs if not enough grant
4398 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4399 then the client will block writes while waiting for another RPC
4400 to complete instead of consuming its dirty limit. With change
4401 we get improved performance when max_dirty_mb is small.
4403 Severity : enhancement
4404 Bugzilla : 3389, 6253
4405 Description: Add support for supplementary groups on the MDS.
4406 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4407 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4408 lookups for user supplementary groups into a cache.
4412 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4413 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4414 for liblustre O_EXCL is enforced by MDS.
4418 Description: readdir never returns NULL in liblustre.
4419 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4420 dentry in struct dirent.
4424 Frequency : liblustre only, depends on application IO pattern
4425 Description: liblustre clients evicted if not contacting servers
4426 Details : Don't put liblustre clients into the ping_evictor list, so
4427 they will not be evicted by the pinger ever.
4429 Severity : enhancement
4431 Description: Add ability to evict clients by NID from MDS.
4432 Details : By echoing "nid:$NID" string into
4433 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4434 $NID would be instantly evicted from this MDS and from all active
4435 OSTs connected to it.
4439 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4440 Details : lookup_it_finish() used to query file size from OSTs that was not
4445 Description: service threads change working directory to that of init
4446 Details : Starting lustre service threads may pin the working directory
4447 of the parent thread, making that filesystem busy. Threads
4448 now change to the working directory of init to avoid this.
4452 Frequency : during shutdown only
4453 Description: shutdown with a failed MDS or OST can cause unmount to hang
4454 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4459 Frequency : 2.6 only
4460 Description: chmod/chown may include an extra supplementary group
4461 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4462 supplementary group and if none is specified this is used.
4465 Bugzilla : 5479 (6816)
4467 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4468 Details : If lookup is in progress on a file that is unlinked we might try
4469 to revalidate the inode and fail in revalidate after lookup is
4470 complete and ll_file_open() enqueues the open again but
4471 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4474 Frequency : always, if lconf --abort_recovery used
4476 Description: lconf --abort_recovery fails with 'Operation not supported'
4477 Details : lconf was attempting to abort recovery on the MDT device and not
4480 ------------------------------------------------------------------------------
4482 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4487 Frequency : rare (only unsupported configurations with a node running as an
4489 Bugzilla : 6514, 5137
4490 Description: Mounting a Lustre file system on a node running as an OST could
4492 Details : OSTs now preallocates memory needed to write out data at
4493 startup, instead of when needed, to avoid having to
4494 allocate memory in possibly low memory situations.
4495 Specifically, if the file system is mounted on on OST,
4496 memory pressure could force it to try to write out data,
4497 which it needed to allocate memory to do. Due to the low
4498 memory, it would be unable to do so and the node would
4499 become unresponsive.
4501 Severity : enhancement
4503 Description: Addition of lconf --service command line option
4504 Details : lconf now accepts a '--service <arg>' option, which is
4505 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4507 Severity : enhancement
4509 Description: Failover mode is now the default for OSTs.
4510 Details : By default, OSTs will now run in failover mode. To return to
4511 the old behaviour, add '--failout' to the lmc line for OSTs.
4513 Severity : enhancement
4515 Description: Health checks are now provided for MDS and OSTs
4516 Details : Additional detailed health check information on MSD and OSTs
4517 is now provided through the procfs health_check value.
4520 Frequency : occasional, depends on IO load
4522 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4523 after numerous create/delete cycles
4524 Details : The ext3 inode allocation policy would not allocate new inodes
4525 very well on the OSTs because there are no new directories
4526 being created. Instead we look for groups with free space if
4527 the parent directories are nearly full.
4532 Description: Network or server problems during mount may cause partially
4533 mounted clients instead of returning an error.
4534 Details : The config llog parsing code may overwrite the error return
4535 code during mount error handling, returning success instead
4541 Description: MDS can fail to allocate large reply buffers
4542 Details : After long uptimes the MDS can fail to allocate large reply
4543 buffers (e.g. zconf client mount config records) due to memory
4544 fragmentation or consumption by the buffer cache. Preallocate
4545 some large reply buffers so that these replies can be sent even
4546 under memory pressure.
4550 Frequency : rare (liblustre)
4551 Description: fsx running with liblustre complained that using truncate() to
4552 extend the file doesn't work. This patch corrects that issue.
4553 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4554 ATTR_SIZE and lsm use in llu_setattr_raw.
4558 Frequency : rare, only 2.6 kernels
4559 Description: Unusual file access patterns on the MDS may result in inode
4560 data being lost in very rare circumstances.
4561 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4562 code in the 2.6 kernel caused the inode and/or EA data not to
4563 be read from disk, causing single-file corruption.
4567 Frequency : rare, only 2.6 filesystems using extents
4568 Description: Heavy concurrent write and delete load may cause data corruption.
4569 Details : It was possible under high-load situations to have an extent
4570 metadata block in the block device cache from a just-unlinked
4571 file overwrite a newly-allocated data block. We now unmap any
4572 metadata buffers that alias just-allocated data blocks.
4576 Frequency : filesystems with default stripe_count larger than 77
4577 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4578 Details : lconf specifies an inode size of 4096 bytes when the default
4579 stripe_count is larger than 77. This conflicts with the default
4580 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4581 this case to avoid pinning too much memory for large EAs.
4583 ------------------------------------------------------------------------------
4585 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4590 Frequency : rare (extremely heavy IO load with hundreds of clients)
4592 Description: Client is evicted, gets IO error writing to file
4593 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4594 caused clients to be evicted for AST timeouts. The fixes in
4595 bug 5192 mean we no longer need to have such short AST timeouts
4596 so ldlm_timeout has been increased.
4599 Frequency : occasional during --force or --failover shutdown under load
4600 Bugzilla : 5949, 4834
4601 Description: Server oops/LBUG if stopped with --force or --failover under load
4602 Details : a collection of import/export refcount and cleanup ordering
4603 issues fixed for safer force cleanup
4606 Frequency : only filesystems larger than 120 OSTs
4607 Bugzilla : 5990, 6223
4608 Description: lfs getstripe would oops on a very large filesystem
4609 Details : lov_getconfig used kfree on vmalloc'd memory
4612 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4613 Bugzilla : 6242, 6243
4614 Description: reading from files that had been truncated to a non-zero size
4615 but never opened returned no data
4616 Details : ll_file_read() reads zeros from no-object files to EOF
4621 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4622 Details : There's little harm in aborting MDS/OSS recovery and letting it
4623 try again, so I removed the LASSERT and return an error instead.
4625 Severity : enhancement
4627 Description: New debugging infrastructure for tracking down data corruption
4628 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4629 (b) cover more of the client-side code path, and (c) try to narrow
4630 down where problems occurred
4634 Bugzilla : 3819, 4364, 4397, 6313
4635 Description: Racing close and eviction MDS could cause assertion in mds_close
4636 Details : It was possible to get multiple mfd references during close and
4637 client eviction, leading to one thread referencing a freed mfd.
4639 Severity: : enhancement
4640 Bugzilla : 3262, 6359
4641 Description: Attempts to reconnect to servers are now more aggressive.
4642 Details : This builds on the enhanced upcall-less recovery that was added
4643 in 1.4.2. When trying to reconnect to servers, clients will
4644 now try each server in the failover group every 10 seconds. By
4645 default, clients would previously try one server every 25 seconds.
4650 Description: After recovery, certain operations trigger a failed
4651 assertion on a client.
4652 Details : Failing over an mds, using lconf -d --failover, while a
4653 client was doing a readdir() call would cause the client to
4654 LBUG after recovery completed and the readdir() was resent.
4656 Severity : enhancement
4658 Description: Default groups are now added by lconf
4659 Details : You can now run lconf --group <servicename> without having to
4660 manually add groups with lmc.
4663 Frequency : occasional
4665 Description: Nodes with an elan id of 0 trigger a failed assertion
4668 Frequency : always when accessing e.g. tty/console device nodes
4670 Description: tty and some other devices nodes cannot be used on lustre
4671 Details : file's private_data field is used by device data and lustre
4672 values in there got lost. New field was added to struct file to
4673 store fs-specific private data.
4676 Frequency : when exporting Lustre via NFS
4678 Description: NFSD failed occasionally when looking up a path component
4679 Details : NFSD is looking up ".." which was broken in ext3 directories
4680 that had grown large enough to become hashed.
4683 Frequency : Clusters with multiple interfaces not on the same subnet
4685 Description: Nodes will repeatedly try to reconnect to an interface which it
4686 cannot reach and report an error to the log.
4687 Details : Extra peer list entries will be created by lconf with some peers
4688 unreachable. lconf now validates the peer before adding it.
4691 Frequency : Only if a default stripe is set on the filesystem root.
4693 Description: Setting a default stripe on the filesystem root prevented the
4694 filesystem from being remounted.
4695 Details : The client was sending extra request flags in the root getattr
4696 request and did not allocate a reply buffer for the dir EA.
4699 Frequency : occasional, higher if lots of files are accessed by one client
4700 Bugzilla : 6159, 6097
4701 Description: Client trips assertion regarding lsm mismatch/magic
4702 Details : While revalidating inodes the VFS looks up inodes with ifind()
4703 and in rare cases can find an inode that is being freed.
4704 The ll_test_inode() code will free the lsm during ifind()
4705 when it finds an existing inode and then the VFS later attaches
4706 this free lsm to a new inode.
4710 Bugzilla : 6422, 7030
4711 Description: MDS deadlock between mkdir and client eviction
4712 Details : Creating a new file via mkdir or mknod (starting a transaction
4713 and getting the ns lock) can deadlock with client eviction
4714 (gets ns lock and trying to finish a synchronous transaction).
4717 Frequency : occasional
4718 Description: While starting a server, the fsfilt_ext3 module could not be
4720 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4721 kernels. Previously, lconf would still use the ext3 name
4722 when trying to load modules. Now, it will correctly use
4723 ext3 on 2.4 and ldiskfs on 2.6.
4725 Severity : enhancement
4726 Description: The default stripe count has been changed to 1
4727 Details : The interpretation of the default stripe count (0, to lfs
4728 or lmc) has been changed to mean striping across a single
4729 OST, rather than all available. For general usage we have
4730 found a stripe count of 1 or 2 works best.
4732 Severity : enhancement
4733 Description: Add support for compiling against Cray portals.
4734 Details : Conditional compiling for some areas that are different
4738 Frequency : occasional
4739 Bugzilla : 6409, 6834
4740 Description: Creating files with an explicit stripe count may lead to
4741 a failed assertion on the MDS
4742 Details : If some OSTs are full or unavailable, creating files may
4743 trigger a failed assertion on the MDS. Now, Lustre will
4744 try to use other servers or return an error to the
4748 Frequency : occasional
4750 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4751 caused lock timeout during readahead (since 1.4.2).
4752 Details : Processes doing readahead might match a lock that hasn't been
4753 granted yet if there are overlapping and conflicting lock
4754 requests. The readahead process waits on ungranted lock
4755 (original lock is CBPENDING), while OST waits for that process
4756 to cancel CBPENDING read lock and eventually evicts client.
4758 Severity : enhancement
4760 Description: Initial enabling of flock support for clients
4761 Details : Implements fcntl advisory locking and file status functions.
4762 This feature is provided as an optional mount flag (default
4763 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4764 locking are implemented yet, and those that are are not guaranteed
4765 to be completely correct in production environments.
4766 mount -t lustre -o [flock|noflock] ...
4769 Frequency : occasional
4771 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4772 assertion in the ext3 JBD (journaling) layer.
4773 Details : The b_committed_data struct is protected by the big kernel lock
4774 in 2.4 kernels, serializing journal_commit_transaction() and
4775 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4776 access to this struct. In 2.6 kernels there is finer grained
4777 locking to improve SMP performance of the JBD layer.
4781 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4786 Description: merge and cleanup kernel patches.
4789 -----------------------------------------------------------------------------
4791 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4793 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4794 versions of Lustre. Please update all servers and clients to
4795 version 1.4.2 or later at the same time. You must also run
4796 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4797 to update the configuration logs.
4799 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4800 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4801 - swab llog records of type '0' so we get proper header size/idx (5861)
4802 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4803 - fix rename of one directory over another leaking an inode (5953)
4804 - avoid SetPageDirty on 2.6 (5981)
4805 - don't re-add just-being-destroyed locks to the waiting list (5653)
4806 - when creating new directories, inherit the parent's custom
4807 striping settings if present parent (3048)
4808 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4809 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4810 - don't hold client locks on temporary worklist from l_lru (5666)
4811 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4812 - automatically evict dead clients (5921)
4813 - Update file size properly in create+truncate+fstat case (6196)
4814 - Do not unhash mountpoint dentries, do not allow removal of
4816 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4817 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4818 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4819 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4820 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4821 - Correctly handle reads of files with no objects (6243)
4822 - lctl recover will also mark a device active if deactivate used (5933)
4824 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4825 - allow --write-conf on an MDS with different nettype than client (5619)
4826 - don't write config llogs to MDS for mounts not from that MDS (5617)
4827 - lconf should create multiple TCP connections from a client (5201)
4828 - init scripts are now turned off by default; run chkconfig --on
4829 lustre and chkconfig --on lustrefs to use them
4830 - upcalls are no longer needed for clients to recover to failover
4832 - add --abort-recovery option to lconf to abort recovery on device
4834 - add support for an arbitrary number of OSTs (3026)
4835 - Quota support protocol changes.
4836 - forward compatibility changes to wire structs (6007)
4837 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4838 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4839 - improved "lustre status" script
4840 - initialize blocksize for non-regular files (6062)
4841 - added --disable-server and --disable-client configure options (5782)
4842 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4843 - Vanilla 2.4.29 support
4844 - increase maximum number of obd devices to 520 (6242)
4845 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4846 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4847 SLES 9 2.6.5 kernels
4848 - limit stripes per file to 160 (the maximum EA size) (6093)
4850 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4853 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4854 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4855 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4856 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4857 - small ext3 extents cleanups and fixes (5733)
4858 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4859 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4860 - avoid needless client->OST connect, fix handle mismatch (5317)
4861 - fix DLM error path that led to out-of-sync client, long delays (5779)
4862 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4863 - fix several locking issues related to i_size (5492,5624,5654,5672)
4864 - don't move pending lock onto export if it is already evicted (5683)
4865 - fix kernel oops when creating .foo in unlinked directory (5548)
4866 - fix deadlock in obdfilter statistics vs. object create (5811)
4867 - use time_{before,after} to avoid timer jiffies wrap (5882)
4868 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4869 - Do not leak request if server was not able to process it (5154)
4870 - If mds_open unable to find parent dir, make that negative lookup(5154)
4871 - don't create new directories with extent-mapping (5909, 5936)
4873 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4874 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4875 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4876 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4877 - support for RHEL4 kernel on i686 (5773)
4878 - provide error messages when incompatible logs are encountered (5898)
4880 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4881 * version 1.4.0.10 (1.4.1 release candidate 1)
4883 - don't keep a lock reference when lock is not granted (4238)
4884 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4885 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4886 - drop import inflight refcount on signal_completed_replay error (5255)
4887 - unlock page after async write error during send (3677)
4888 - handle missing objects in filter_preprw_read properly (5265)
4889 - no transno return for symlink open, don't save no-trasno open (3440)
4890 - don't try to complete elan receive that already failed (4012)
4891 - free RPC server reply state on error (5406)
4892 - clean up thread from ptlrpc_start_thread() on error (5160)
4893 - readahead could read extra page into cache that wasn't ejected (5388)
4894 - prevent races in class_attach/setup/cleanup/detach (5260)
4895 - don't dereference de->d_inode after l_dput of de (5458)
4896 - use "int" for stripe value returned from lock_to_stripe (5544)
4897 - mballoc allocation and error-checking fixes in 2.6 (5504)
4898 - block device patches to fix I/O request sizes in 2.6 (5482)
4899 - look up hostnames for IB nals (5602)
4900 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4901 - don't start multiple acceptors for the same port (5277)
4902 - fix incorrect LASSERT in mds_getattr_name (5635)
4903 - export a proc file for general "ping" checking (5628)
4904 - fix "lfs check" to not block when the MDS is down (5628)
4906 - service request history (4965)
4907 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4908 - create an "evict_client" /proc entry on OSTs, like the MDS has
4909 - fix mount usage message, return errors per mount(8) (5168)
4910 - change grep [] to grep "[]" in tests so they work in more UMLs
4911 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4912 - remove ancient LOV_MAGIC_V0 EA support (5047)
4913 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4914 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4915 - allow readahead allocations to fail when low on memory (5383)
4916 - mmap locking landed again, after considerable improvement (2828)
4917 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4919 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4922 - send OST transaction number in read/write reply to free req (4966)
4923 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4924 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4925 - the watchdog thread now runs as interruptible (5246)
4926 - flock/lockf fixes (but it's still disabled, pending 5135)
4927 - don't use EXT3 constants in llite code (5094)
4928 - memory shortage at startup could cause assertion (5176)
4930 - reorganization of lov code
4931 - single portals codebase
4933 - add extents/mballoc support (5025)
4934 - direct I/O reads in the obdfilter (4048)
4935 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4937 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4940 - send OST transaction number in read/write reply to free req (4966)
4941 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4942 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4943 - the watchdog thread now runs as interruptible (5246)
4944 - handle missing objects in filter_preprw_read properly (5265)
4945 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4946 - drop import inflight refcount on signal_completed_replay error (5255)
4947 - unlock page after async write error during send (3677)
4948 - return original error code on reconstructed replies (3761)
4949 - no transno return for symlink open, don't save no-trasno open (3440)
4951 - add pid to ldlm debugging output (4922)
4952 - bump the watchdog timeouts -- we can't handle 30sec yet
4953 - extra debugging for orphan dentry/inode bug (5259)
4955 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4958 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4959 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4960 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4961 - fix NULL dereference in /proc/sys/portals/routes (4827)
4962 - allow failed mdc_close() operations to be interrupted (4561)
4963 - stop precreate on OST before MDS would time out on it (4778)
4964 - don't send partial-page writes before EOF from client (4410)
4965 - discard client grant for sub-page writes on large-page clients (4520)
4966 - don't free dentries not owned by NFS code, check generation (4806)
4967 - fix lsm leak if mds_create_objects() fails (4801)
4968 - limit debug_daemon file size, always print CERROR messages (4789)
4969 - use transno after validating reply (3892)
4970 - process timed out requests if import state changes (3754)
4971 - update mtime on OST during writes, return in glimpse (4829)
4972 - add mkfsoptions to LDAP (4679)
4973 - use ->max_readahead method instead of zapping global ra (5039)
4974 - don't interrupt __l_wait_event() during strace
4976 - add software watchdogs to catch hung threads quickly (4941)
4977 - make lustrefs init script start after nfs is mounted
4978 - fix CWARN/ERROR duplication (4930)
4979 - return async write errors to application if possible (2248)
4980 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4981 - print NAL number in %x format (4645)
4982 - update barely-supported suse-2.4.21-171 series (4842)
4983 - support for sles 9 %post scripts
4984 - support for building 2.6 kernel-source packages
4985 - support for sles km_* packages
4987 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4990 - ignore -ENOENT errors in osc_destroy (3639)
4991 - notify osc create thread that OSC is being cleaned up (4600)
4992 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4993 - reconstruct ost_handle() like mds_handle() (4657)
4994 - create a new thread to do import eviction to avoid deadlock (3969)
4995 - let lconf resolve symlinked-to devices (4629)
4996 - don't unlink "objects" from directory with default EA (4554)
4997 - hold socknal file ref over connect in case target is down (4394)
4998 - allow more than 32000 subdirectories in a single directory (3244)
4999 - fix blocks count for O_DIRECT writes (3751)
5000 - OST returns ENOSPC from object create when no space left (4539)
5001 - don't send truncate RPC if file size isn't changing (4410)
5002 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5003 - bind to privileged port in socknal and tcpnal (3689)
5005 - rate limit CERROR/CWARN console message to avoid overload (4519)
5006 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5007 - basic mmap support (3918)
5008 - kernel patch series update from b1_4 (4711)
5010 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5013 - avoid crash during MDS cleanup with OST shut down (2775)
5014 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5015 - don't use bad inodes on the MDS (3744)
5016 - dynamic object preallocation to improve recovery speed (4236)
5017 - don't hold spinlock over lock dumping or change debug flags (4401)
5018 - don't zero obd_dev when it is force cleaned (3651)
5019 - print grants to console if they go negative (4431)
5020 - "lctl deactivate" will stop automatic recovery attempts (3406)
5021 - look for existing locks in ldlm_handle_enqueue() (3764)
5022 - don't resolve lock handle twice in recovery avoiding race (4401)
5023 - revalidate should check working dir is a directory (4134)
5025 - don't always mark "slow" obdfilter messages as errors (4418)
5027 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5030 - don't close LustreDB during write_conf until it is done (3860)
5031 - fix typo in lconf for_each_profile (3821)
5032 - allow dumping logs from multiple threads at one time (3820)
5033 - don't allow multiple threads in OSC recovery (3812)
5034 - fix debug_size parameters (3864)
5035 - fix mds_postrecov to initialize import for llog ctxt (3121)
5036 - replace config semaphore with spinlock (3306)
5037 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5038 - don't allow enqueue to complete on a destroyed export (3822)
5039 - down write_lock before checking llog header bitmap (3825)
5040 - recover from lock replay timeout (3764)
5041 - up llog sem before sending rpc (3652)
5042 - reduce ns lock hold times when setting kms (3267)
5043 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5044 - fix NULL deref and obd_dev leak on setup error (3312)
5045 - replace some LBUG about llog ops with error handling (3841)
5046 - don't match INVALID dentries from d_lookup and spin (3784)
5047 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5048 - fix invalid assertion in ptlrpc_set_wait (3880)
5050 - add libwrap support for the TCP acceptor (3996)
5051 - add /proc/sys/portals/routes for non-root route listing (3994)
5052 - allow setting MDS UUID in .xml (2580)
5053 - print the stack of a process that LBUGs (4228)
5055 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5058 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5059 - make sure to unset replay flag from failed open requests (3440)
5060 - if default stripe count is 0, use OST count for inode size (3636)
5061 - update parent mtime/ctime on client for create/unlink (2611)
5062 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5063 - free recovery state on server during a forced cleanup (3571)
5064 - unregister_reply for resent reqs (3063)
5065 - loop back devices mounting and status check on 2.6 (3563)
5066 - fix resource-creation race that can provoke i_size == 0 (3513)
5067 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5068 - more debugging for page-accounting assertion (3746)
5069 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5070 - avoid deadlock after precreation failure (3758)
5071 - fix race and lock order deadlock in orphan handling (3450, 3750)
5072 - add validity checks when grabbing inodes from l_ast_data (3599)
5074 - add /proc/.../recovery_status to obdfilter (3428)
5075 - lightweight CDEBUG infrastructure, debug daemon (3668)
5076 - change default OSC RPC parameters to be better on small clusters
5077 - turn off OST read cache for files smaller than 32MB
5078 - install man pages and include them in rpms (3100)
5079 - add new init script for (un)mounting lustre filesystems (2593)
5080 - run chkconfig in %post for init scripts (3701)
5081 - drop scimac NAL (unmaintained)
5083 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5086 - clean kiobufs before and after use (3485)
5087 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5088 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5089 - change init script to fail more clearly if not run as root (1528)
5090 - allow clients to reconnect during replay (1742)
5091 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5092 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5093 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5094 - fix cond_resched (3554)
5095 - only evict unfinished clients after recovery (3515)
5096 - allow bulk resend, prevent data loss (3570)
5097 - dynamic ptlrpc request buffer allocation (2102)
5098 - don't allow unlinking open directory if it isn't empty (2904)
5099 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5100 - remove extraneous obd dereference causing LASSERT failure (3334)
5101 - don't use get_cycles() when creating temp. files on the mds (3156)
5102 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5103 - handle EEXIST for set-stripe, set proper directory name (3336)
5105 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5106 - fix ksocknal_fmb_callback() error messages (2918)
5108 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5111 - don't copy lvb into (possibly NULL) reply on error (2983)
5112 - don't deref dentry after dput, don't free lvb on error (2922)
5113 - use the kms to determine writeback rpc length (2947)
5114 - increment oti_logcookies when osc is inactive (2948)
5115 - update client's i_blocks count via lvb messages (2543)
5116 - handle intent open/close of special files properly (1557)
5117 - mount MDS with errors=remount-ro, like obdfilter (2009)
5118 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5119 - don't use cancelling-locks' kms values (2947)
5120 - use highest lock extent for kms, not last one (2925)
5121 - don't dereference ERR_PTR() dentry in error handling path (3107)
5122 - fix thread race in portals_debug_dumplog() (3122)
5123 - create lprocfs device entries at setup instead of at attach (1519)
5124 - common AST error handler, don't evict client on completion race (3145)
5125 - zero nameidata in detach_mnt in 2.6 (3118)
5126 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5127 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5128 - initialize RPC timeout timer earlier for 2.6 (3219)
5129 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5130 - print nal/nid for unknown nid (3258)
5131 - additional checks for oscc recovery before doing precreate (3284)
5132 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5133 - don't crash in mdc_close for bad permissions on open (3285)
5134 - zero i_rdev for non-device files (3147)
5135 - clear page->private before handing to FS, better assertion (3119)
5136 - tune the read pipeline (3236)
5137 - fix incorrect decref of invalidated dentry (2350)
5138 - provide read-ahead stats and refine rpc in flight stats (3328)
5139 - don't hold journal transaction open across create RPC (3313)
5140 - update atime on MDS at close time (3265)
5141 - close LDAP connection when recovering to avoid server load (3315)
5142 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5143 - don't leak open file on MDS after open resend (3325)
5144 - serialize filter_precreate and filter_destroy_precreated (3329)
5145 - loop device shouldn't call sync_dev() for nul device (3092)
5146 - clear page cache after eviction (2766)
5147 - resynchronize MDS->OST in background (2824)
5148 - refuse to mount the same filesystem twice on same mountpoint (3394)
5149 - allow llmount to create routes for mounting behind routers (3320)
5150 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5151 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5152 - fix rare problem with rename on htree directories (3417)
5154 - allow default OST striping configuration per directory (1414)
5155 - fix compilation for qswnal for 2.6 kernels (3125)
5156 - increase maximum number of MDS request buffers for large systems
5157 - change liblustreapi to be useful for external progs like lfsck (3098)
5158 - increase local configuration timeout for slow disks (3353)
5159 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5161 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5164 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5165 - don't overwrite extent policy data in reply if lock was blocked (2901)
5166 - drop filter export grants atomically with removal from device (2663)
5167 - del obd_self_export from work_list in class_disconnect_exports (2908)
5168 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5169 - swab reply message in mdc_close, other PPC fixes (2464)
5170 - fix destroying of named logs (2325)
5171 - overwrite old logs when running lconf --write_conf (2264)
5172 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5173 - fix race in target_handle_connect (2898)
5174 - mds_reint_create() should take same inode create lock (2926)
5175 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5176 - don't close files for self_export to avoid uninitialized obd (2936)
5177 - allow MDS with the same name as client node (2939)
5178 - hold dentry reference for closed log files for unlink (2325)
5179 - reserve space for all logs during transactions (2059)
5180 - don't evict page beyond end of stripe extent (2925)
5181 - don't oops on a deleted current working directory (2399)
5182 - handle hard links to targets without a parent properly (2517)
5183 - don't dereference NULL lock when racing during eviction (2867)
5184 - don't grow lock extents when lots of conflicting locks (2919)
5186 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5189 - account for cache space usage on clients to avoid data loss (974)
5190 - lfsck support in lustre kernel code (2349)
5191 - reduce journal credits needed for BRW writes (2370)
5192 - orphan handling to avoid losing space on client/server crashes
5193 - ptlrpcd can be blocked, stopping ALL progress (2477)
5194 - use lock value blocks to assist in proper KMS, faster stat (1021)
5195 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5196 - recovery for initial connections (2355)
5197 - fixes for mds_cleanup_orphans (1934)
5198 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5199 - block all file creations until orphan recovery completes (1901)
5200 - client remove rq_connection from request struct (2423)
5201 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5202 - recovery timer race (2670)
5203 - mdc_close recovey bug (2532)
5204 - ptlrpc cleanup bug (2710)
5205 - mds timeout on local locks (2588)
5206 - namespace lock held during RPCs (2431)
5207 - handle interrupted sync write properly (2503)
5208 - don't try to handle a message that hasn't been replied to (2699)
5209 - client assert failure during cleanup after abort recovery (2701)
5210 - leak mdc device after failed mount (2712)
5211 - ptlrpc_check_set allows timedout requests to complete (2714)
5212 - wait for inflight reqs when ptlrpcd finishes (2710)
5213 - make sure unregistered services are removed from the srv_list
5214 - reset bulk XID's when resending them (caught by 1138 test)
5215 - unregister_bulk after timeout
5216 - fix lconf error (2694)
5217 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5218 - readahead locks pages, leaves pending causing memory pressure (2673)
5219 - increase OST request buffers to 4096 on large machines (2729)
5220 - fix up permission of existing directories in simple_mkdir (2661)
5221 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5222 - don't assign transno to errored transactions (2742)
5223 - don't delete objects on OST if given a bogus objid from MDS (2751)
5224 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5225 - if rq_no_resend, then timeout request after recovery (2432)
5226 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5227 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5228 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5229 - MDS recovery shouldn't delete objects that it has given out (2730)
5230 - if enqueue arrives after completion, don't clobber LVB (2819)
5231 - don't unlock pages twice when trigger_group_io returns error (2814)
5232 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5233 - don't write pages to disk if there was an error (1450)
5234 - don't ping imports that have recovery disabled (2676)
5235 - take buffered bytes into account when balancing socknal conn (2817)
5236 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5237 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5238 - remove little-endian swabbing of llog records (1987)
5239 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5240 - retry reposting request buffers if they fail (1191)
5241 - grow extent at grant time to avoid granting a revoked lock (2809)
5242 - lock revoke doesn't evict page if covered by a second lock (2765)
5243 - disable VM readahead to avoid reading outside lock extents (2805)
5245 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5246 - updated kernel patches for hp-2.4.20 kernel (2681)
5248 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5251 - fix truncated write corruption (2366)
5252 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5254 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5255 - ensure len(uuid) < 37 in lmc (1171)
5256 - fix ia64 OOPS in llog_test (2255)
5257 - zero end of page at obdfilter for partial page writes (2648)
5258 - don't leave stale dentries around after renames (bug 2428)
5259 - fix timeouts when evicting a client with a single lock held (2642)
5260 - set deadline for the initial HELLO message to drain (2634)
5261 - print out dotted-quad IP addresses in the socknal (2302)
5263 - additional debugging for MDS client eviction problem (2443)
5264 - fix mkfsoptions support for osts (2603, 2604)
5266 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5269 - add series for the vanilla 2.6.0 kernel
5270 - add series for the vanilla 2.4.24 kernel
5271 - add series for a cray x86/64 UL kernel drop
5272 - fix xattr patches for the vanilla 2.4.19 series
5274 - generate true UUIDs in lmc (1171)
5275 - have portals stack dumping break in UML (2466)
5276 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5277 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5278 - wake callback waiting threads on client eviction (2460)
5279 - Add --ptldebug and --subsystem to lmc (1719)
5280 - update assertion to allow safe interrupt allocation
5281 - set rq_no_resend for cancel requests (2432)
5282 - recalculate ptlrpcd timeout after resend (2494)
5283 - call vfs_rmdir when removing pending directories (2368)
5284 - fix renaming a file to itself (2429)
5285 - lmc creates a default one-stripe lov (2454)
5286 - expand procfs space to handle large clusters (2326)
5287 - increase UML stack to avoid overflow
5288 - update lconf's list of debug and subsystem masks
5289 - fix lfs find --obd (2510)
5290 - /proc tunable for disabling filter read caching (2591)
5291 - stop rpm packages from altering slapd.conf (2301)
5292 - disable nagle in the socknal under 0conf (2578)
5293 - choose mds inode size based on stripe count (2572)
5294 - fix kernel-source rpm problems (2516)
5296 - add --disable-doc to avoid pdf generation (2421)
5297 - update documentation, tests, type-os, comments
5298 - avoid format warnings on ia64
5299 - remove the TOE NAL
5300 - tiny code cleanups by removing unused fields
5302 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5305 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5306 - remove the most blatant lies from BUILDING file (2371)
5307 - change default debug level to reasonable production setting
5308 - reduce client side cache size to reduce cache flush time
5309 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5310 - make TCP zerocopy and pinger support enabled by default (2476)
5311 - sync writes completed after process exits caused crashes (2319)
5312 - maintain correct mount count on the MDS (2356)
5313 - backout 1557, because 2316 wasn't really fixed
5314 - better file I/O statistics gathering in /proc
5315 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5316 - another kernel patch to fix zero-copy TCP function export
5317 - don't take duplicate lock when processing re-sent getattr (2420)
5318 - lctl uses obd_self_export instead of creating new conn (2353)
5319 - MDS/OST recovery case which requires object creation asserted (2425)
5320 - move lfs from /usr/sbin to /usr/bin in packages
5321 - fix race between mds_client_add and mds_client_free (2417)
5322 - use kmalloc instead of slabs in portals (2430)
5323 - don't create duplicate records when a failover MDS is present (2442)
5324 - remove unnecessary mount age check (2332)
5325 - don't remove directory inodes from locks prematurely (2451)
5326 - don't break if MDS service name is the same as hostname (2103)
5327 - fix races in client write RPC generation when cache full (2482)
5329 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5332 - remove now-unused request->rq_obd (278)
5333 - if an allocation fails, print out how much memory we've used (1933)
5334 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5335 - add the "configurable stack size" patch to most series files (1256)
5336 - ability to write large log records, for 100+ OST configs (2306)
5337 - fix NULL deref when filter_prep fails (2314)
5338 - fix operator precedence error in filter_sync
5339 - dynamic allocation of socknal TX descriptors (2315)
5340 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5341 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5342 - ensure that CWARN messages in Portals always get to the syslog
5343 - __init/__exit are not for prototype decls (ldlm_init/exit)
5344 - x86-64 compile warning fixes
5345 - fix gateway LMC keyword conflict (2318)
5346 - fix MDS lock inversions in getattr/reint paths (1844)
5347 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5348 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5349 - DLM race condition prevented some lock evictions (2328)
5350 - ENOMEM detection and retry on socknal sends (2230)
5351 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5352 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5353 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5354 - better error messages when a client is rejected during recovery (1505)
5355 - avoid cancelling locks which were never granted, after failure (2330)
5356 - fix i_sem/journal inversion in mds_client_add (2333)
5357 - fix truncate/getattr lock cycle deadlock (2334)
5358 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5359 - fix two rare exit paths which could leak an l_lock() ref (2321)
5360 - fencepost error in MDS/OST orphan recovery (2226)
5361 - make log record alignment 8 bytes (1988)
5362 - lstripe now fails when requested offset > ost_count (2237)
5363 - ensure that all kernel series have a complete list.h (1607)
5364 - fix crashes in special-file operations (2316)
5365 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5367 - allow configurable automake binary, for testing new versions
5368 - small update to the lfs documentation
5370 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5372 * fix negative export reference count in fsfilt_sync (2312)
5374 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5375 * release candidate 0.9.1
5377 - orphans are moved into the PENDING directory for possible recovery
5378 - replayed opens now open by fid for orphan/rename safety (1042)
5379 - last close of an orphan inode generates a transno (683)
5380 - chdir() and mount() now pin the directory entry (1020)
5381 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5382 - discard very old requests without processing them (1502)
5383 - remove some common, well-understood CERRORs (1505)
5384 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5385 - clear "grant" flags in OST replies until OST grant code lands (1644)
5386 - fix read performance by not clobbering i_blksize on client (1598)
5387 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5388 - make LRU size a /proc tunable, clears locks when reduced (707)
5389 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5390 - server locks take references on exports now (1558)
5391 - build fixes for 2.4.20-rh trees (1663)
5392 - return an error from lov_create if all OSCs are inactive (1751)
5393 - fix import levels when a reconnect happens without a timeout (1597)
5394 - exit early from mds_open if we get a lookup error (1749)
5395 - partial page read at EOF wouldn't wait for disk before sending (1642)
5396 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5397 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5398 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5399 - write/truncate lock inversion (1639)
5400 - Don't auto-load obdclass, portals modules during cleanup (1495)
5401 - fix timestamps from jumping to "now" (1763)
5402 - extra journal assertions (1648)
5403 - add an extra multiunlink test (1771)
5404 - fix read_record/write_record API (1776)
5405 - fix leak of offset_extent, possible incorrect i_size later (1772)
5406 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5407 - Debugging for the kqswnal_get_idle_tx problems (1820)
5408 - Allow recovery to be attempted multiple times (1536)
5409 - Write out MDS last_rcvd file after it is first created (1600)
5410 - Fix tx_descriptor leak in failed transmit situations (1827)
5411 - ext3 journaling fixes for assertion failure after IO error (1871)
5412 - class_export_put() on freed export after completion AST error (1896)
5413 - Fix revalidate looping in VFS (1322)
5414 - Don't access a freed export during MDS_REINT timeout (1521)
5415 - Add open-unlink recovery support on the MDS (1673,1764)
5416 - Return an error if no MDS data was read from last_rcvd (1946)
5417 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5418 - Don't setup a disk device that doesn't match exported UUID (317)
5419 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5420 - avoid committing NULL handle in force close
5421 - local.sh is now a one-stripe LOV configuration
5422 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5423 - fix ext3 htree duplicate directory entry corruption (1516)
5424 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5425 - update to new LOV EA format (2097)
5426 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5428 2003-06-15 Phil Schwan <phil@clusterfs.com>
5431 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5432 - per-import/export recovery handling (958, 931, 959)
5433 - multiple last-rcvd slots, for serving multiple FSes (949)
5434 - connections are again shared between multiple imp/exports (963, 964)
5435 - "umount -f" would hang if any requests needed to be sent (393, 978)
5436 - avoid pinning large req buffer by copying for queued messages (989)
5437 - add "uuid" to "lctl device" command to help upcalls (991)
5438 - "open" RPCs with transnos would confuse recovery counters (1037)
5439 - do proper endian conversion of all wire messages (288, 340, 891)
5440 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5441 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5442 - fix ptlrpc_connection leak in target_handle_connect (1174)
5443 - fix import refcounting bug in OST and MDS cleanup (1134)
5444 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5445 - fix very unlikely obd_types race condition (501)
5446 - remove osc_open hack for echo_client (1187)
5447 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5448 - a failure in read_inode2 leads to deadlock (1139)
5449 - cancel ack-locks as soon as transaction is committed (1072)
5450 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5451 - make sure to commitrw after any preprw to avoid deadlock (1162)
5452 - failing to execute a file in a lustre FS would lock inode (1203)
5453 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5454 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5455 - obd_uuid2tgt would walk off the end of the list (1255)
5456 - on IA64 the osc would give portals incorrect bulk size (1258)
5457 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5458 - fix lock inversion caused by new llite matching code (1282)
5459 - limit the number of dirty pages on a client to 10MB (1286)
5460 - timed out locks were not being corrected cancelled (1289)
5461 - fix O_DIRECT above 4GB on IA-32 (1292)
5462 * major user-visible changes
5463 - fail out/fail over policy now controlled by the upcall (993)
5465 - add OBD_PING to check server availability and failure (954)
5466 - lustre messages are now sent in sending host order (288, 340, 891)
5467 - add eadatalen to MDS getattr reply (340)
5468 - OST read replies may contain second buffer, with per-page status (593)
5470 2003-03-11 Phil Schwan <phil@clusterfs.com>
5473 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5474 - failed open()s could cause deadlock; fixed (867, 869)
5475 - stop cancelling OST locks when files are closed (481)
5476 - overlapping XID spaces caused network corruption (851, 853)
5477 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5478 - setattr_raw added, to do single-RPC, server-side setattrs
5479 - lmc/lconf syntax change for OST UUIDs
5480 - fix crashy race condition between ptlrpc_free_req and osc_close
5481 - don't use request in mdc_enqueue if we hit a timeout (889)
5482 - don't set the inode i_size for regular files from the MDS (896)
5483 - handle out of order completion AST (842)
5484 - don't LBUG if a lock request times out after receiving AST (913)
5485 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5486 - if a bad lock AST arrives, send an error instead of dropping entirely
5487 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5488 - fix leak in bulk IO when only partially completed (899, 900, 926)
5489 - fix O_DIRECT for ia64 (55)
5490 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5491 - C-z after timeout could hang a process forever; fixed (977)
5493 - client-side I/O cache (678, 924, 929, 941, 970)
5495 - READPAGE and SETATTRs which don't take server-side locks get
5498 2003-02-11 Phil Schwan <phil@clusterfs.com>
5501 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5502 - Fix fsfilt_extN_readpage() to read a full page of directory
5503 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5504 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5505 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5506 - Most intent operations were removed, in favour of a new RPC mode
5507 that does a single RPC to the server and bypasses most of the VFS
5508 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5509 - Aggressively cancel local locks on DLM servers
5510 - mds_reint_unlink sends EA to the client if it's the last nlink.
5511 client uses that EA to unlink OST objects.
5512 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5513 - recursive symlinks were fixed (439)
5514 - fixed NULL deref in DEBUG_REQ
5515 - filter_update_lastobjid no longer calls sync, which annoyed extN
5516 - fixed multi-client small-writes to a single file problem (445)
5517 - fixed mtime updates during file writes (607)
5518 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5519 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5520 - fixed Linux OST size reporting problem (444, 656)
5521 - OST now updates object mtime with writes or setattr (607, 619)
5522 - client verifies file size before zeroing page past EOF (445)
5523 - OST now writes last allocated objid to disk with allocation (108)
5524 - LOV on echo now works (409)
5526 - mds_reint_unlink sends a new buffer, with the EA included. this
5527 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5528 set if a regular file was being unlinked, and it was the last link
5529 - use PtlGet from the target for bulk writes (315)
5530 - OST now updates object mtime with writes or setattr (607, 619)
5531 - LDLM now has a grant-time callback to revalidate locked items, if
5533 - Many MDS operations were reorganized to combat race conditions
5535 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5536 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5538 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5541 - Fully reactivate OST imports after reconnection (512, others)
5542 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5543 - More graceful error handling for truncating on dead OST (515)
5544 - Don't error out unless we're actually accessing dead stripes (474)
5545 - Fix garbage sizes when stripes are missing (410)
5546 - LRU counters were broken, causing constant lock purge (433, 432)
5547 - garbage on read from stripes with failed OSTs (441)
5548 - mark OSCs as active before reconnecting during recovery (438)
5549 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5550 - lfind did not preserve OST order in output (443)
5551 - symlinks cause hung clients, incorrect data (439)
5552 - stop dereferencing request after dropping refcount (457)
5553 - don't LASSERT(spin_is_locked) on non-SMP (455)
5554 - fixes for many rename() bugs
5555 - fstat didn't correctly synchronize attributes (399)
5556 - server must handle lock cancellation during blocking AST prep (487)
5557 - bulk descriptors were free()d too soon (511)
5558 - fix paths in lconf, which would load incorrect modules (451, 507)
5559 - fix confusing lconf 'host not found' error message (386)
5560 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5561 - fix race condition in mdc_blocking_ast() for inode access (526)
5562 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5563 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5564 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5565 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5566 - fix lstripe in several ways (kernel side) (527)
5567 - fix request leak in ldlm_cli_enqueue (262)
5568 - incorrect OSC was marked inactive after OST failure
5569 - call mds_fs_cleanup before unmounting filesystem (524)
5570 - fix races between taking ns_lock and ldlm_lock_change_resource
5571 - fix races updating LOV export open file list
5572 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5573 - fix recovery NULL deref in ldlm_cli_cancel_unused
5574 - fix some DLM races by using new hash table for lock handles (419)
5575 - permit the client to specify desired inodes, at replay
5576 - duplicate requests when we queue them for replay reintegration
5577 - fix last_rcvd offset calculation
5578 - sync after each recovered transaction, so we always make progress
5579 - never, not always, ERESTART requests without transnos
5580 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5582 - skip replay if the MDS says that the client is already connected
5583 - don't check for a recovery-enabled export to match lctl's UUID
5584 - don't INC_USE_COUNT for phantom exports
5585 - don't crash when cleaning up phantom exports (567)
5586 - don't double-finish or set replay data for errored mdc_open requests
5587 - abort requests when they time out, so we don't get old replies
5588 - send/receive replies for AST messages again
5589 - if the client says that it doesn't have the lock, cancel it on the
5591 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5592 mark it as destroyed, it will all work out when decref is called
5593 - fix module use counts (22, 581)
5595 - ASTs now expect a reply (server cancels lock on error reply)
5597 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5600 - fix many simultaneous client startup (392)
5601 - fix dentry->d_it clobbering
5602 - credentials weren't being shipped for readdir/getattr operations
5603 - remove invalid assertions triggered during some concurrent MD
5605 - proper Lustre versions added (336, 389)
5606 - fix memory leak for create error case (398)
5607 - fix LOV locking bug that would get cli/srv out of sync
5608 - fix echo client over LOV (409)
5609 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5610 - fix double-O_EXCL intent crash (424)
5611 - avoid sending multiple lock CANCELs (352)
5613 - MDS can do multi-client recovery (modulo bugs in new code)
5615 - many updates, edits, cleanups
5617 2002-11-18 Phil Schwan <phil@clusterfs.com>
5620 - fix null d_it dereference (346)
5621 - fix full OST/dbench hang (333)
5622 - fix permission problem with file removal (286)
5623 - fix removal of OSCs from LOV when they fail
5624 - fix NULL deref during bulk timeout (214)
5625 - fix problems related to multiple filesystems on one MDS (241)
5626 - fixed serious subtle metadata locking bugs
5627 - free locks on clients when inodes are removed due to memory
5629 - fix inode pointer in lock data (285)
5630 - partial support for multiple MDS on a single host (241)
5631 - data locks weren't cancelled at clear_inode time (290, 311)
5632 - intent locks could lead to unbounded lock growth (205)
5633 - added a maximum lock count, an LRU list, and a flusher
5634 - fix multiple rename (365)
5635 - properly abstracted the echo client
5636 - OSC locked 1 byte too many; fixed
5637 - rewrote brw callback code:
5638 - fixed recovery bugs related to LOVs (306)
5639 - fixed too-many-pages-in-one-write crash (191)
5640 - fixed (again) crash in sync_io_timeout (214)
5641 - probably fixed callback-related race (385)
5643 - Add capability to MDS protocol
5644 - LDLM cancellations and callbacks on different portals
5646 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5649 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5650 - defer open object destruction to close time (601981, 138)
5651 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5652 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5653 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5654 - don't call obd_setattr in ll_file_release for destroyed objects
5655 * protocol change to lustre_msg: move |version| and add |flags|
5656 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5657 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5658 * added replay of create, unlink, link and rename operations during
5659 MDS failover; recovery should be much more robust now
5660 * remove failed OSCs from LOVs (only lov_create uses this so far)
5661 * the lustre-HOWTO was brought (more) up to date (582544)
5663 2002-10-23 Phil Schwan <phil@clusterfs.com>
5666 - in-use dentries weren't being reused properly (617851)
5667 - prevent multiple LDLM setup (599178)
5668 - fix LOV size calculations for truncate (617853)
5669 - fix client handling of MDS intent errors (POSIX)
5670 - fix permission bug in lovstripe.c test (624321)
5671 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5672 - truncate past end of file could corrupt data
5673 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5674 - a race in recovery could return ETIMEDOUT to apps (623947)
5675 - building outside the source directory was fixed
5676 * the lustre-HOWTO was brought (more) up to date (582544)
5677 * major progress was made on recovery functionality
5679 2002-10-10 Phil Schwan <phil@clusterfs.com>
5682 - recovery deadlock fix
5683 - rm -rf causes LBUG fix (617817)
5684 - file open by multiple tasks fix (618962)
5685 - directory permissions bugs (602707 and 620007)
5686 - journal_stop fixed with locking (611313)
5687 - O_APPEND failures resolved (618273, perhaps 614459)
5688 - lconf PATH fix (619770)
5689 - IA64 build fix (621450)
5690 - RPC buffer sizes scale with amount of memory
5692 2002-10-01 Phil Schwan <phil@clusterfs.com>
5695 - locks would be cancelled without throwing away data pages,
5696 resulting in inconsistent data (605627)
5697 - inode attributes were not always being refreshed (605627, 612449)
5698 - lconf now continues to cleanup after lctl reports an error
5699 - MDS now enforces user permissions (602707)
5700 - lprocfs cleanup fixed, but not yet enabled (614157)
5701 - fixed infinite server hang, should a client not respond to an AST
5702 - avoid going into recovery if user calls readlink() with a buffer
5703 that's too small (613941)
5704 - AST RPCs no longer require replies (614867) -- this may be changed
5705 - don't crash server if client sends an IOV that's too big (611336)
5706 - fixed lock conversion deadlock (611892)
5707 - fixed the following of symlinks (614622)
5708 * recovery: the server can remove locks from a client that dies, other
5709 clients can make progress
5710 * more extN patch fixes
5711 * compile-time configurable ptlrpc buffer allocations
5713 - collaborative read cache document
5714 - Lustre Lite Performance CDR document-in-progress
5716 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5719 - fix typo in patch-2.4.18
5721 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5724 - clear ptlrpc request each time in handle_incoming_request()
5725 - unlink of files now destroys the object on the OST
5727 2002-09-19 Peter Braam <braam@clusterfs.com>
5729 * add hard link support
5730 * change obdfile creation method
5731 * kernel patch changed
5733 2002-09-19 Peter Braam <braam@clusterfs.com>
5736 - stack overflow bug in extN fixed
5738 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5740 * documentation updates
5741 - add man pages for config tools
5742 - update tests/README to describe testing with new config tools
5743 - finish metadata API descriptions
5744 * bug fixes and cleanups
5745 - statfs workaround for 16TB limit
5746 - LOV stripe allocation improved, can stripe on subset of OSTs
5747 - LOV file size/IO offset was wrong for files > 4GB in size
5748 - object EA data was being dropped, caused files to be unreadable
5749 - memory overflow with non-LOV OST caused memory corruption
5750 - fixed regression tests to work with new config tools, obdfilter
5751 - fixed bug when directory size became larger than 1 block
5752 - fixed bug (for single client case) when PWD was deleted
5753 - invalidate local directory pages when doing intent-based ops
5754 - avoid LDLM oops when lock callback contained bad data
5756 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5758 * documentation updates
5759 * bug fixes and cleanups
5760 - configuration tools
5763 - 64-bit compile warnings
5764 - 64-bit internal statfs data
5766 * test_brw on persistent OST devices
5768 * lprocfs (disabled)
5770 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5772 * documentation updates
5773 * bug fixes and cleanups
5774 * configuration tools
5776 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5779 * many small fixes to 0_5_4
5780 * io/network handling
5781 * thinkos in MDS operations
5783 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5786 * crucial basic fixes to 0.5.3
5787 * IOR, Iozone work over Elan
5790 2002-08-07 Phil Schwan <phil@clusterfs.com>
5791 * version 0_5_3, our first alpha
5792 * we use the new Portals iovs
5793 * documentation updates
5794 * bug fixes and cleanups
5795 * small changes in the DLM wire protocol
5797 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5798 * version 0_5_1 with some initial stability,
5799 * locking on MD and file I/O.
5800 * documentation updates
5801 * several bug fixes since 0.5.0
5802 * small changes in wire protocol
5804 2002-07-18 Phil Schwan <phil@clusterfs.com>
5806 * delivered as Lustre Light Alpha
5807 * fixed a crash after handling invalid MDS requests
5808 * fixed directory pages for architectures with non-4k pages sizes
5810 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5811 * release version v0_4_4
5812 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5814 * Fixes a number of interruption problems with OST operations.
5815 * Update documentation for portals header changes
5816 * Move all wire protocol structs/defines to lustre_idl.h
5817 * Fixes symlink length bug.
5818 * Add tcpdump to repository.
5820 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5821 * release version v0_4_3
5822 * Fixes statfs for inodes on extN.
5823 * Fixes bug in runtests which would delete /etc/hosts.
5824 * Use 64-bit object IDs wherever possible (not into VFS though)
5825 Remove ost_get_info, which is unused by lustre, and out of date.
5827 2002-07-03 Peter Braam <braam@clusterfs.com>
5828 * release version v0_4_2 Fixes a lookup error (type not passed)
5829 * move forward to head of Portals
5830 * move forward to latest Lustre kernel
5832 2002-06-25 Peter Braam <braam@clusterfs.com>
5833 * release version v0_4_1. Hopefully stable on single node use.