1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.39.3 (SLES 10),
5 2.6.18-128.1.14.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.41.6.sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
14 * File join has been disabled in this release, refer to Bugzilla 16929.
16 Severity : enhancement
18 Description: Add LustreNetLink, a kernel-userspace communcation path.
20 Severity : enhancement
22 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
27 Description: access to llog context before init.
28 Details : move handling CATALOGS file at osc layer and forbid access to llog
33 Description: Avoid deadlock for local client writes
34 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
35 memory freeing context. This allows OST threads to set the
36 PF_MEMALLOC flag on task structures in order to allocate memory
37 from reserved pools and complete IO.
38 Use GFP_HIGHUSER for OST allocations for non-local client writes,
39 so that the OST threads generate memory pressure and allow
40 inactive pages to be reclaimed.
42 Severity : enhancement
44 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
49 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
50 Details : move ldlm namespace creation in setup phase to avoid grab
51 _lprocfs_lock with cli_sem held.
55 Description: Temporarily disable grant shrink.
56 Details : Disable the feature for debugging.
60 Description: Unable to run several mkfs.lustre on loop devices at the same
62 Details : mkfs.lustre returns error 256 on the concurrent loop devices
63 formatting. The solution is to proper handle the error.
65 Severity : enhancement
67 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
69 Severity : enhancement
71 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
75 Description: resolve race between obd_disconnect and class_disconnect_exports
76 Details : if obd_disconnect will be called to already disconnected export he
77 forget release one reference and osc module can't unloaded.
79 Severity : enhancement
81 Description: Allow tuning service thread via /proc
82 Details : For each service a new
83 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
84 created that can be used to set min/max thread counts, and get the
85 current number of running threads.
89 Descriptoin: don't return error if have particaly created objects for file.
90 Details : lov_update_create_set uses set->set_success as index for created objects,
91 so if some requests will be failed, they will have hole at end of
92 array and we can use qos_shrink_lsm for allocate correct lsm.
94 Severity : enhancement
96 Description: Update OFED support to 1.4.1
100 Description: Reduce small size read RPC
101 Details : Set read-ahead limite for every file and only do read-ahead when
102 available read-ahead pages are bigger than 1M to avoid small size
105 Severity : enhancement
107 Description: move AT tunable parameters for more consistent usage
108 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
110 Severity : enhancement
112 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
113 Details : allow skip disconnected ost for send statfs request and hide error
119 Description: fix racy locking of mballoc block bitmaps causing BUG
120 Details : The locking of the mballoc buddy bitmap and the in-memory
121 block bitmap was using two different spin locks in some
122 cases. This made it possible to incorrectly access the
123 mballoc bitmap while another process was modifying it,
124 causing a sanity assertion to fail. While no on-disk corruption
125 was reported, there was some risk of this happening.
128 Frequency : rare, on llog test 6
130 Descriptoin: don't allow connect to already connected import
131 Details : allowing connect to already connected import is hide connecting problem.
136 Descriptoin: allow kill process which wait statahead result
137 Details : for some reasons 'ls' can stick in waiting result from statahead,
138 in this case need way for kill this process.
140 Severity : enhancement
142 Description: Add state history info file, enhance import info file
143 Details : Track import connection state changes in a new osc/mdc proc file;
144 add overview-type data to the osc/mdc import proc file.
146 Severity : enhancement
148 Description: MDS create should not wait for statfs RPC while holding DLM lock.
150 Severity : enhancement
152 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
157 Descriptoin: OST grant too much space to client even there are not enough space.
158 Details : Client will shrink its grant cache to OST if there are no write
159 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
160 this grant cache if there are already not enough avaible space
161 (left_space < total_clients * 32M).
164 Frequency : start MDS on uncleanly shutdowned MDS device
166 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
167 Details : stay in waiting mds<>ost recovery finished produce random bugs
168 due race between two ll_sync thread for one lov target. send
169 ACTIVATE event only if connect realy finished and import have
173 Frequency : rare, connect and disconnect target at same time
175 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
176 Details : don't call obd_disconnect under lov_lock. this long time
177 operation and can block ptlrpcd which answer to connect request.
182 Descriptoin: don't lose wakeup for imp_recovery_waitq
183 Details : recover_import_no_retry or invalidate_import and import_close can
184 both sleep on imp_recovery_waitq, but we was send only one wakeup
188 Frequency : always with long access acl
190 Descriptoin: mds can't pack reply with long acl.
191 Details : mds don't control size of acl but they limited by reint/getattr
194 Severity : enhancement
196 Description: Update to SLES10 kernel-2.6.16.60-0.33.
198 Severity : enhancement
200 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
203 Frequency : start MDS on uncleanly shutdowned MDS device
205 Descriptoin: aborting recovery hang on MDS
206 Details : don't throttle destroy RPCs for the MDT.
209 Frequency : on remount
211 Description: external journal device not working after the remount
212 Details : clear dev_rdonly flag for external journal devices in
218 Description: shutdown vs evict race
219 Details : client_disconnect_export vs connect request race.
220 if client will evicted at this time - we start invalidate
221 thread without referece to import and import can be freed
225 Frequency : rare, need acl's on inode.
227 Description: client can't handle ost additional correctly
228 Details : if ost was added after client connected to mds client can have
229 hit lnet_try_match_md ... to big messages to wide striped files.
230 in this case need teach client to handle config events about add
231 lov target and update client max ea size at that event.
233 Severity : enhancement
235 Description: Changelogs
236 Details : Changelogs are a lightweight mechanism to track filesystem
237 metadata and namespace changes. The changelog is recorded
238 permanently on the MDTs, and is periodically "consumed" / purged
239 when records are no longer needed.
241 Severity : enhancement
243 Description: compact fld format with extents
244 Details : Store range of seq rather than every seq in FLD. Seq
245 controller update FLD rather than clients. In Case of CMD, mdt0
246 has FLD, all other metadata server act as non persistent proxy
247 for FLD queries and cache fld entries in fld cache.
252 Description: don't skip ost target if they assigned to file
253 Details : Drop slow OSCs if we can, but not for requested start idx.
254 This means "if OSC is slow and it is not the requested
255 start OST, then it can be skipped, otherwise skip it only
256 if it is inactive/recovering/out-of-space.
260 Description: more cleanup in mds_lov
261 Details : not send LOV EA under replay, we can't know about they size at this
262 time. Don't allow client connect to mds before any ost connected,
263 for avoid problems with LOV EA size and returning EIO to client.
265 Severity : enhancement
267 Description: Interoperability at server side (Disk interoperability)
269 Severity : enhancement
271 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
273 Severity : enhancement
275 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
277 Severity : enhancement
279 Description: New client IO stack (CLIO).
281 Severity : enhancement
283 Description: Commit on sharing. Eliminate inter-client dependencies between
284 uncommitted transactions by doing transaction commits.
285 Thereby clients may recovery independently.
288 Frequency : Create a symlink file with a very long name
290 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
291 Details : If there is no extra space in the request for early cancels,
292 ldlm_req_handles_avail() returns 0 instead of a negative value.
294 Severity : enhancement
296 Description: Add /proc entry for import status
297 Details : The mdc, osc, and mgc import directories now have
298 an import directory that contains useful import data for debugging
301 Severity : enhancement
303 Description: Re-disable certain /proc logging
304 Details : Enable and disable client's offset_stats, extents_stats and
305 extents_stats_per_process stats logging on the fly.
308 Frequency : Only on FC kernels 2.6.22+
310 Description: oops in statahead
311 Details : Do not drop reference count for the dentry from VFS when lookup,
312 VFS will do that by itself.
314 Severity : enhancement
316 Description: Generic /proc file permissions
317 Details : Set /Proc file permissions in a more generic way to enable non-
318 root users operate on some /proc files.
322 Description: Hitting mdc_commit_close() ASSERTION
323 Details : Properly handle request reference release in
324 ll_release_openhandle().
328 Description: quota recovery deadlock during mds failover
329 Details : This patch includes att18982, att18236, att18237 in bz14840.
331 1. fix osts hang when mds does failover with quotaon
332 2. prevent watchdog storm when osts threads wait for the
337 Frequency : only patchless client
338 Description: add workaround for race between add/remove dentry from hash
340 Severity : enhancement
342 Description: Allow OST glimpses to return PW locks
346 Description: LBUG when llog conf file is full
347 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
351 Description: Prevent import from entering FULL state when server in recovery
355 Description: service mount cannot take device name with ":"
356 Details : Only when device name contains ":/" will mount treat it as
362 Description: replace ptlrpcd with the statahead thread to interpret the async
363 statahead RPC callback
367 Frequency : on recovery
368 Description: I/O failures after umount during fail back
369 Details : if client reconnected to restarted server we need join to recovery
370 instead of find server handler is changed and process self
371 eviction with cancel all locks.
373 Severity : enhancement
375 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
377 Severity : enhancement
379 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
381 Severity : enhancement
383 Description: Upcall on Lustre log has been dumped
384 Details : Allow for a user mode script to be called once a Lustre log has
385 been dumped. It passes the filename of the dumped log to the
386 script, the location of the script can be specified via
387 /proc/sys/lnet/debug_log_upcall.
392 Description: avoid idr_remove called for id which is not allocated.
393 Details : Move assigment s_dev for clustered nfs to end of initialization,
394 to avoid problem with error handling.
399 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
400 Details : When connection is reused this not moved from CONN_UNUSED_HASH
401 into CONN_USED_HASH and this prodice warning when put connection
402 again in unused hash.
405 Severity : enhancement
407 Description: File striping can now be set to use an arbitrary pool of OSTs.
409 Severity : enhancement
411 Description: Export bytes_read/bytes_write count on OSC/OST.
415 Description: Early reply size mismatch, MGC loses connection
416 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
417 the connect flags are properly negotiated.
422 Description: quotas are not honored with O_DIRECT
423 Details : all writes with the flag O_DIRECT will use grants which leads to
424 this problem. Now using OBD_BRW_SYNC to guard this.
428 Description: add quota statistics
429 Details : 1. sort out quota proc entries and proc code.
430 2. add quota statistics
432 Severity : enhancement
434 Description: enable quota support for HEAD.
438 Description: Properly propagate oinfo flags from lov to osc for statfs
439 Details : restore missing copy oi_flags to lov requests.
441 Severity : enhancement
443 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
444 : l_getgroups(8), lst(8), routerstat(8)
445 Details : included man pages for llobdstat(8), llstat(8),
446 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
448 Severity : enhancement
450 Description: configure's --enable-quota should check the
451 : kernel .config for CONFIG_QUOTA
452 Details : configure is terminated if --enable-quota is passed but
453 : no quota support is in kernel
457 Description: Remove portals compatibility
458 Details : Remove portals compatibility, not interoperable with releases
463 Description: Resolve device initialization race
464 Details : Prevent proc handler from accessing devices added to the
465 obd_devs array but yet be intialized.
467 Severity : enhancement
469 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
471 Severity : enhancement
473 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
478 Description: Using wrong pointer in osc_brw_prep_request
479 Details : Access to array[-1] can produce panic if kernel compiled with
480 CONFIG_PAGE_ALLOC enabled
484 Description: Client runs out of low memory
485 Details : Consider only lowmem when counting initial number of llap pages
489 Description: *optional* service tags registration
490 Details : if the "service tags" package is installed on a Lustre node
491 When the filesystem is mounted, a local-node service tag will
492 be created. See http://inventory.sun.com/ for more information
493 about the Service Tags asset management system.
497 Description: Kernel BUG tries to release flock
498 Details : Lustre does not destroy flock lock before last reference goes
499 away. So always drop flock locks when client is evicted and
500 perform unlock regardless of successfulness of speaking to MDS.
504 Description: add refcount protection for osc callbacks, avoid panic on shutdown
508 Description: sanity test 65a fails if stripecount of -1 is set
509 Details : handle -1 striping on filesystem in ll_dirstripe_verify
514 Description: ASSERTION(CheckWriteback(page,cmd)) failed
515 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
518 Severity : enhancement
520 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
524 Description: do not process already freed flock
525 Details : flock can possibly be freed by another thread before it reaches
526 to ldlm_flock_completion_ast.
530 Description: LBUG during stress test
531 Details : Need properly lock accesses the flock deadlock detection list.
535 Description: oops in page fault handler
536 Details : kernel page fault handler can return two special 'pages' in error
537 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
541 Description: timeout with invalidate import.
542 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
543 should be handled by ptlrpcd. This produce long age waiting and
544 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
546 Severity : enhancement
548 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
552 Description: enable MGS and MDT services start separately
553 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
554 a co-located MGS without starting the MGS, which is a complement
555 to 'nosvc' mount option.
559 Frequency : after recovery
560 Description: precreate to many object's after del orphan.
561 Details : del orphan st in oscc last_id == next_id and this triger growing
562 count of precreated objects. Set flag LOW to skip increase count
563 of precreated objects.
567 Frequency : rare, on clear nid stats
568 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
569 Details : when clean nid stats sometimes try destroy live entry,
570 and this produce panic in free.
574 Description: Stack overflow during MDS log replay
575 ease stack pressure by using a thread dealing llog_process.
579 Description: wait until IO finished before start new when do lock cancel.
580 Details : VM protocol want old IO finished before start new, in this case
581 need wait until PG_writeback is cleared until check dirty flag and
582 call writepages in lock cancel callback.
584 Severity : enhancement
586 Description: using special macro for print time and cleanup in includes.
590 Description: mds_mfd_close() ASSERTION(rc == 0)
591 Details : In mds_mfd_close(), we need protect inode's writecount change
592 within its orphan write semaphore to prevent possible races.
596 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
600 Frequency : rare, on shutdown ost
601 Description: don't hit live lock with umount ost.
602 Details : shrink_dcache_parent can be in long loop with destroy dentries,
603 use shrink_dcache_sb instead.
607 Description: don't panic with use echo client
608 Details : echo client pass NULL as client nid pointer and this produce null
613 Description: fix build on ppc32
614 Details : compile code with -m64 flag produce wrong object file for ppc32.
618 Description: add message levels for liblustreapi
622 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
623 Details : Move the 'good_osts' check before the 'total_bavail' check. This
624 will result in an -EAGAIN and in the exit call path we call
625 alloc_rr() which will with increasing aggressiveness attempt to
626 aquire precreated objects on the minimum number of required OSCs.
630 Description: Use old size assignment to avoid deadlock
631 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
632 the scheduling while holding a spinlock. We do not need locking
633 for size in ll_update_inode() because size is only updated from
634 the MDS for directories or files without objects, so there is no
635 other place to do the update, and concurrent access to such inodes
636 are protected by the inode lock.
640 Description: resolve "_IOWR redefined" build error on SLES10
644 Description: dump the memory debugging after all modules are unloaded to
645 suppress false negative in conf_sanity test 39
647 Severity : enhancement
649 Description: build kernel-ib packages for OFED 1.3 in our release cycle
654 Description: fix SLES kernel versioning
655 Details : the kernel version for our SLES 10 kernel did not include a "-"
656 before the "smp" at the end. while this was not a problem in
657 general, it did mean that software trying to use the kernel
658 version to try to detect a vendor specific kernel would fail.
659 this was most evident by the OFED build scripts.
663 Description: Don't update lov_desc members until making sure they are valid
664 Details : When updating lov_desc members via proc fs, need fix their
665 validities before doing the real update.
669 Description: don't put request into delay list while invalidate in flight.
670 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
671 import in flight. this produce timeout for invalidate and sometimes
672 can cause stale data.
676 Frequency : on ppc only
677 Description: not convert ost objects for directory because it's not exist.
678 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
682 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
684 Severity : enhancement
686 Description: Update to support 2.6.22.14 vanilla kernel.
690 Frequency : rare, on recovery
691 Description: read procfs can produce deadlock in some situation
692 Details : Holding lprocfs lock which send rpc can produce block for destroy
693 obd objects and this also block reconnect with -EALREADY. This
694 isn't fix all lprocfs bugs - but make it rare.
696 Severity : enhancement
698 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
701 Frequency : frequent on X2 node
703 Description: mdc_set_open_replay_data LBUG
704 Details : Set replay data for requests that are eligible for replay.
708 Description: lustre_mgs: operation 101 on unconnected MGS
709 Details : When MGC is disconnected from MGS long enough, MGS will evict the
710 MGC, and late on MGC cannot successfully connect to MGS and a lot
711 of the error messages complaining that MGS is not connected.
715 Frequency : on network error
716 Description: panic with double free request if network error
717 Details : mdc_finish_enqueue is finish request if any network error ocuring,
718 but it's true only for synchronus enqueue, for async enqueue
719 (via ptlrpcd) this incorrect and ptlrpcd want finish request
722 Severity : enhancement
724 Description: client-side metadata stat-ahead during readdir(directory readahead)
725 Details : perform client-side metadata stat-ahead when the client detects
726 readdir and sequential stat of dir entries therein
729 Frequency : on start mds
731 Description: Implement get_info(last_id) in obdfilter.
734 Frequency : occasional
736 Description: Correctly check stale fid, not start epoch if ost not support SOM
737 Details : open with flag O_CREATE need set old fid in op_fid3 because
738 op_fid2 was overwritten with new generated fid, but mds can answer
739 with one of these two fids and both is not stale. Setattr
740 incorrectly started an epoch and assume will be called
741 done_writing, but without SOM done_writing ever being called.
744 Frequency : rare, depends on device drivers and load
746 Description: MDS or OSS nodes crash due to stack overflow
747 Details : Code changes in 1.8.0 increased the stack usage of some functions.
748 In some cases, in conjunction with device drivers that use a lot
749 of stack the MDS (or possibly OSS) service threads could overflow
750 the stack. One change which was identified to consume additional
751 stack has been reworked to avoid the extra stack usage.
754 Frequency : occasional
756 Description: Do not fail import if osc_interpret_create gets -EAGAIN
757 Details : If osc_interpret_create got -EAGAIN it immediately exits and
758 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
759 oscc_has_objects and see OSC has no objests and call
760 oscc_internal_create to resend create request.
762 Severity : enhancement
764 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
766 Severity : enhancement
768 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
771 Frequency : very rare
773 Description: Fix replay if there is an un-replied request and open
774 Details : In some cases, older replay request will revert the
775 mcd->mcd_last_xid on MDS which is used to record the client's
778 Severity : enhancement
780 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
782 Severity : enhancement
784 Description: Add rhel5 support to HEAD.
786 Serverity : enhancement
788 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
793 Description: Don't allow skipping OSTs if index has been specified.
794 Details : Don't allow skipping OSTs if index has been specified, make
795 locking in internal create lots better.
799 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
800 Details : access to bitfield in structure is always rounded to long
801 and this produce problem with not atomic change any bit.
805 Description: Lustre make rpms failed.
806 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
807 builds Lustre rpms from tarball.
810 Frequency : rare on shutdown ost
812 Description: If llog cancel was not send before clean_exports phase, this can
813 produce deadlock in llog code.
814 Details : If llog thread has last reference to obd and call class_import_put
815 this produce deadlock because llog_cleanup_commit_master wait when
816 last llog_commit_thread exited, but this never success because was
817 called from llog_commit_thread.
821 Description: allow userland application know is lost one of stripes.
822 Details : fill lvb_blocks with error code on ost and return it to
823 application if error flag found.
827 Description: NULL lov_tgts causing MDS oops
828 Details : more safe checks for NULL lov_tgts for avoid oops.
830 Severity : enhancement
832 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
836 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
837 Details : If an OST is down the MDS will hang indefinitely in
838 obd_statfs_rqset() waiting for the statfs data. While for
839 MDS QOS usage of statfs, it should not stuck in waiting.
841 Severity : enhancement
843 Description: remote_acl support
844 Details : Support ACL-based permission check for remote user.
845 Support setfacl/getfacl for remote user with the utils
846 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
847 the system "{s,g}etfacl" utils.
849 Severity : enhancement
851 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
853 Severity : enhancement
855 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
859 Description: make mgs_setparam() handle fsname containing dash
860 Details : fsname containing a dash does not work with lctl conf_param
862 Severity : enhancement
864 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
866 Severity : enhancement
868 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
872 Frequency : rare, at shutdown
873 Description: access already free / zero obd_namespace.
874 Details : if client_disconnect_export was called without force flag set,
875 and exist connect request in flight, this can produce access to
876 NULL pointer (or already free pointer) when connect_interpret
877 store ocd flags in obd_namespace.
881 Frequency : only at startup
882 Description: not alloc memory with spinlock held.
883 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
884 if any spinlock held.
886 Severity : enhancement
888 Description: make lustre randomly fail allocating memory
889 Details : Make lustre randomly failed allocating memory for testing purpose.
891 Severity : enhancement
893 Description: lost problems with lov objid file
894 Details : Fixes some scability and access to not inited memory problems
895 in work with lov objdid file.
900 Description: lfs find does not continue on file error
901 Details : Continue other files processing when a file/dir is absent.
905 Description: Inconsistent usage of lustre_pack_reply()
906 Details : Standardize the usage of lustre_pack_reply() such that it
907 always generate a CERROR on failure.
910 Frequency : occasional
912 Description: MDS hang or stay in waiting lock
913 Details : If client receive lock with CBPENDING flag ldlm need send lock
914 cancel as separate rpc, to avoid situation when cancel request
915 can't processed due all i/o threads stay in wait lock.
919 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
923 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
927 Description: Update to SLES9 kernel-2.6.5-7.287.3.
931 Description: Update to RHEL4 latest kernel.
933 Severity : enhancement
935 Description: Build SLES10 patchless client fails
936 Details : The configure was broken by run ./configure with
937 --with-linux-obj=.... argument for patchless client. When the
938 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
939 can't search header adequately. Use absolute path such as
940 -I($LINUX)/include instead.
944 Description: interrupt oig_wait produce painc on resend.
945 Details : brw_redo_request can be used for resend requests from ptlrpcd and
946 private set, and this produce situation when rq_ptlrpcd_data not
947 copyed to new allocated request and triggered LBUG on assert
948 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
949 ptlrpcd set if request is changed and can be safety changed to use
954 Description: LASSERT_{REQ,REP}SWAB macros are buggy
955 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
956 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
957 these macros and replace them with their difinitions instead.
961 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
962 Details : Update which_patch & target file for SLES10 latest kernel.
965 Description: Remove client patches from SLES 10 kernel.
966 Details : This causes SLES 10 clients to behave as patchless clients
967 even on a Lustre-patched (server) kernel.
969 Severity : enhancement
971 Description: self-adjustable client's lru lists
972 Details : use adaptive algorithm for managing client cached locks lru
973 lists according to current server load, other client's work
974 pattern, memory activities, etc. Both, server and client
975 side namespaces provide number of proc tunables for controlling
978 Severity : enhancement
980 Description: light-weight GSS support
981 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
982 and reduce performance overhead.
984 Severity : enhancement
986 Description: Linux keyring support
987 Details : Support using service of Linux keyring for Lustre GSS internal
988 context refresh/cache mechanism.
992 Description: Fix errors in lfs documentation
993 Details : Fixes man pages
997 Description: don't use GFP_* in generic Lustre code.
998 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1002 Description: obdclass is limited by single OBD_ALLOC(idarray)
1003 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1007 Description: warnings with build patchless client with vanila 2.6.19 and up
1008 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1009 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1013 Frequency : Only for SLES
1014 Description: sanity_quota fail test_1
1015 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1016 file, which makes TSTID[2] inunique.
1020 Frequency : Only for SLES9
1021 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1022 Details : Update target/ChangeLog/which_patch .
1026 Frequency : Only for test_10 in sanity-quota.sh
1027 Description: a bug in quota test code
1028 Details : checking a test flag in a wrong place
1032 Frequency : Only for test_13 in sanity-quota.sh
1033 Description: a bug in quota test script
1034 Details : shouldn't check overfull allocated quota
1038 Description: jbd statistics
1039 Details : Port older jbd statistics patch for sles10
1041 Severity : enhancement
1043 Description: Add printing inode info into message about error in writepage.
1048 Description: reply_lock_interpret crash due to race with it and lock cancel.
1049 Details : Do not replay locks that are being cancelled. Do not reference
1050 locks by their address during replay, just by their handle.
1054 Frequency : When flocks are used.
1055 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1056 Details : Flock locks might destroy just granted lock if it could be merged
1057 with another existing flock, this is done in completion handler,
1058 so teach ldlm_cli_enquque_fini that this is a valid case for
1064 Description: Oops in read and write path when failing to allocate lock.
1065 Details : Check if lock allocation failed and return error back.
1069 Description: lstripe command fails for valid OST index
1070 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1071 instead of lov->desc.ld_active_tgt_count.
1075 Description: sanity.sh failed test 103
1076 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1081 Description: df doesn't work properly if diskfs blocksize != 4K
1082 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1086 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1087 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1089 Severity : enhancement
1091 Description: Adaptive timeouts
1092 Details : RPC timeouts adapt to changing server load and network
1093 conditions to reduce resend attempts and improve recovery time.
1097 Description: llapi_file_create() does not allow some changes
1098 Details : add llapi_file_open() that allows specifying the mode and
1099 open flags, and also returns an open file handle.
1103 Description: merge and cleanup kernel patches.
1104 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1108 Description: Add journal checksum support.(Kernel part)
1109 Details : The journal checksum feature adds two new flags i.e
1110 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1111 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1112 indicates that the commit block contains the checksum for
1113 the blocks described by the descriptor blocks. Now commit
1114 record can be sent to disk without waiting for descriptor
1115 blocks to be written to disk. This behavior is controlled
1116 using JBD2_FEATURE_ASYNC_COMMIT flag.
1120 Description: OSS needs mutliple precreate threads
1121 Details : Add ability to start more than one create thread per OSS.
1125 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1126 Details : execute lfs setstripe on client
1130 Description: mds_obd_create error creating tmp object
1131 Details : When the user sets quota on root, llog will be affected and can't
1132 create files and write files.
1135 Frequency : Always on ia64 patchless client, and possibly others.
1137 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1138 Details : This allows the patchless client to be loaded on architectures
1139 without this export.
1143 Description: RedHat Update kernel for RHEL5
1144 Details : Add the kernel config file for RHEL5.
1148 Description: RedHat Update kernel for RHEL5
1149 Details : Modify the kernel config file more closer RHEL5.
1153 Description: Build failure against Centos5 (RHEL5)
1154 Details : Define PAGE_SIZE when it isn't present.
1158 Description: test_7 on sanity-quota.sh can't be used on separate machines
1159 Details : add facet to handle it
1163 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1164 Details : with NFS, the anon dentry's parent was set to itself in
1165 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1166 corresponding dentry other than use rec->ur_name.
1168 Severity : enhancement
1170 Description: lfs setstripe enhancement
1171 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1175 Description: enable data checksumming by default
1176 Details : enable checksum by default, allow --disable-checksum
1177 configure option and "-o nochecksum" mount option. Checksums
1178 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1182 Description: lustre support for RHEL5
1183 Details : Add support for RHEL5.
1187 Description: Client eviction due to failover config
1188 Details : after a connection loss, the lustre client should attempt to
1189 reconnect to the last active server first before trying the
1190 other potential connections.
1194 Description: when mds and osts use different quota unit(32bit and 64bit),
1195 quota will be released repeatly.
1196 Details : void sending multiple quota reqs to mds, which will keep the
1197 status between the reqs.
1201 Description: osts not allocated evenly to files
1202 Details : change the condition to increase offset_idx
1205 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1207 Description: Data corruption for OSTs that are formatted larger than 2TB
1209 Details : When generating the bio request for lustre file writes the
1210 sector number would overflow a temporary variable before being
1211 used for the IO. The data reads correctly from Lustre (which
1212 will overflow in a similar manner) but other file data or
1213 filesystem metadata may be corrupted in some cases.
1217 Description: Tune the kernel for good SCSI performance.
1218 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1219 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1224 Description: rewrite ext2-derived code in obdclass/uuid.c
1225 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1226 from scratch preserving functionality.
1230 Description: "lfs find -obd UUID" prints directories
1231 Details : "lfs find -obd UUID" will return all directory names instead
1232 of just file names. It is incorrect because the directories
1233 do not reside on the OSTs.
1237 Description: use i_size_read and i_size_write in 2.6 port
1238 Details : replace inode->i_size access with i_size_read/write()
1241 Frequency : when removing large files
1243 Description: scheduling issue during removal of large Lustre files
1244 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1245 It causes scheduling issues when removing large files (17TB in the
1249 Frequency : only with liblustre clients on XT3
1251 Description: evictions taking too long
1252 Details : allow llrd to evict clients directly on OSTs
1255 Frequency : only on ppc
1257 Description: /proc/fs/lustre/devices broken on ppc
1258 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1259 We should make sure the type of 'index' is loff_t and then cast
1260 explicitly as needed below. Do not assign an explicitly cast
1264 Frequency : only for rhel5
1266 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1267 Details : Modify the target file & which_kernel.
1269 Severity : enhancement
1271 Description: omit set fsid for export NFS
1272 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1273 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1277 Frequency : Always, for kernels after 2.6.16
1278 Description: Fix warning idr_remove called for id=.. which is not allocated.
1279 Details : Last kernels save old s_dev before kill super and not allow
1280 to restore from callback - restore it before call kill_anon_super.
1284 Description: buffer overruns could theoretically occur
1285 Details : llapi_semantic_traverse() modifies the "path" argument by
1286 appending values to the end of the origin string, and a
1287 overrun may occur. Adding buffer overrun check in liblustreapi.
1291 Description: Fix error on 'ls .' at the top of the Lustre mount.
1292 Details : Don't revalidate dentry if it is a root dentry.
1296 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1297 Details : Modify vm-tunables-rhel4.patch.
1301 Description: Kernel config for 2.6.18-vanilla.
1302 Details : Modify targets/2.6-vanilla.target.in.
1303 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1304 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1305 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1306 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1310 Description: improve handling recoverable errors
1311 Details : if request processig with error which can be recoverable on server
1312 request should be resend, otherwise page released from cache and
1317 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1318 Details : Modify target file & which_patch.
1319 A flaw was found in the IA32 system call emulation provided
1320 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1321 value could be stored in the %RAX register, which could trigger an
1322 out-of-bounds system call table access. An untrusted local user
1323 could exploit this flaw to run code in the kernel
1324 (ie a root privilege escalation). (CVE-2007-4573).
1328 Description: O_DIRECT bypasses client statistics.
1329 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1330 read_ahead_stats were updated. copied the stats section from
1331 osc_send_oap_rpc() into async_internal().
1335 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1339 Description: change order of libsysio includes
1340 Details : '#include sysio.h' should always come before '#include xtio.h'
1344 Description: handle "serious error: objid * already exists" more gracefully
1345 Details : If LAST_ID value on disk is smaller than the objects existing in
1346 the O/0/d* directories, it indicates disk corruption and causes an
1347 LBUG(). If the object is 0-length, then we should use the existing
1348 object. This will help to avoid a full fsck in most cases.
1350 Severity : enhancement
1352 Description: adapt the lustre_config script to support the upgrade case
1353 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1354 server targets to 1.6 in parallel.
1358 Description: To avoid grant space > avaible space when the disk is almost
1359 full. Without this patch you might see the error "grant XXXX >
1360 available" or some LBUG about grant, when the disk is almost
1362 Details : In filter_check_grant, for non_grant cache write, we should
1363 check the left space by if (*left > ungranted + bytes), instead
1364 of (*left > ungranted), because only we are sure the left space
1365 is enough for another "bytes", then the ungrant space should be
1366 increase. In client, we should update cl_avail_grant only there
1367 is OBD_MD_FLGRANT in the reply.
1371 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1372 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1373 described in CVE-2007-4573. This problem could allow untrusted
1374 local users to gain root access.
1377 Frequency : when using O_DIRECT and quotas
1379 Description: Incorrect file ownership on O_DIRECT output files
1380 Details : block usage reported by 'lfs quota' does not take into account
1381 files that have been written with O_DIRECT.
1386 Description: touch file failed when fs is not full
1387 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1388 otherwise we can get ENOSP while fs is not full.
1392 Description: parallel lock callbacks
1393 Details : Instead of sending blocking and completion callbacks as separated
1394 requests, adding them to a set and sending in parallel.
1397 Frequency : only for Cray XT3
1398 Bugzilla : 12829/13455
1399 Description: Changing primary group doesn't change the group lustre assigns to
1401 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1402 overridden with the primary group provided by the group upcall,
1403 whereas the supplied fsgid can be trusted if it is in the list of
1404 supplementary groups returned by the group upcall.
1406 Severity : enhancement
1408 Description: Allow masking D_WARNING, D_ERROR messages from console
1409 Details : Console messages can now be disabled via lnet.printk.
1413 Description: User code with malformed file open parameter crashes client node
1414 Details : Before packing join_file req, all the related reference should be
1415 checked carefully in case some malformed flags cause fake
1416 join_file req on client.
1420 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1421 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1422 mode is equal to the granted mode. However, it should also check
1423 that LDLM_FL_CP_REQD is not set.
1427 Description: Heavy nfs access might result in deadlocks
1428 Details : After ELC code landed, it is now improper to enqueue any mds
1429 locks under och_sem, because enqueue might want to decide to
1430 cancel open locks for same inode we are holding och_sem for.
1434 Description: Client eviction while running blogbench
1435 Details : A lot of unlink operations with concurrent I/O can lead to a
1436 deadlock causing evictions. To address the problem, the number of
1437 oustanding OST_DESTROY requests is now throttled to
1438 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1439 callbacks are processed in priority.
1443 Description: enable ACLs on MDS by default
1444 Details : ACLs must be enabled on MDS by default.
1447 Frequency : PPC/PPC64 only
1449 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1450 Details : fix duplicated definitions between asm-ppc64/types.h and
1451 lustre_types.h on PPC.
1454 Frequency : PPC/PPC64 only
1456 Description: asm-ppc/segment.h does not exist
1457 Details : fix compile issue on PPC.
1461 Description: better handle error messages in extents code
1464 Frequency : RHEL4 only
1466 Description: mkfs is very slow on IA64/RHEL4
1467 Details : A performance regression has been discovered in the MPT Fusion
1468 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1469 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1470 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1472 Severity : enhancement
1474 Description: SNMP support enhancement
1475 Details : Adding total number of sampled request for an MDS node in snmp
1478 Severity : enhancement
1480 Description: Optimize ldlm waiting list processing for PR extent locks
1481 Details : When processing waiting list for read extent lock and meeting read
1482 lock that is same or wider to it that is not contended, skip
1483 processing rest of the list and immediatelly return current
1484 status of conflictness, since we are guaranteed there are no
1485 conflicting locks in the rest of the list.
1489 Description: Time out and refuse to reconnect
1490 Details : When the failover node is the primary node, it is possible
1491 to have two identical connections in imp_conn_list. We must
1492 compare not conn's pointers but NIDs, otherwise we can defeat
1493 connection throttling.
1497 Description: port llog fixes from b1_6 into HEAD
1498 Details : Port llog reference couting and some llog cleanups from b1_6
1499 (bug 10800) into HEAD, for protect from panic and access to already
1500 free llog structures.
1504 Description: Detect stride IO mode in read-ahead
1505 Details : When a client does stride read, read-ahead should detect that and
1506 read-ahead pages according to the detected stride pattern.
1510 Description: data checksumming impacts single node performance
1511 Details : add support for several checksum algorithm. Currently, only CRC32
1512 and Adler-32 are supported. The checksum type can be changed on
1513 the fly via /proc/fs/lustre/osc/*/checksum_type.
1517 Description: use adler32 for page checksums
1518 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1523 Description: build for x2 fails
1524 Details : fix compile issue on Cray systems.
1528 Description: Properly match for duplicate locks
1529 Details : Due to different lock order from skiplists code, we need to
1530 traverse entire list for now
1533 Frequency : only on PPC/SLES10
1535 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1536 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1537 be defined. Add a hack in mkfs_lustre.c to work around this compile
1542 Description: LASSERT on MDS when client holding flock lock dies
1543 Details : ldlm pool logic depends on number of granted locks equal to
1544 number of released locks which is not true for flock locks, so
1545 just exclude such locks from consideration.
1549 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1550 Details : Use fsfilt_sync() for both the whole filesystem sync and
1551 individual file sync to eliminate dangerous inode locking
1552 with I_LOCK that can lead to a deadlock.
1556 Description: performance in 1.6.3
1557 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1558 sure that 1MB requests are not fragmented by the block layer.
1560 Severity : enhancement
1562 Description: organize the server-side client stats on per-nid basis
1563 Details : Change the structure of stats under obdfilter and mds to
1571 The "uuid"s file would list the uuids of _active_ exports.
1572 And the clear entry is to clear all stats and stale nids.
1574 Severity : enhancement
1576 Description: eliminate client locks in face of contention
1577 Details : file contention detection and lockless i/o implementation
1578 for contended files.
1582 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1585 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1587 Description: MDS may lose file striping (and hence file data) in some cases
1588 Details : If there are additional extended attributes stored on the MDS,
1589 in particular ACLs, SELinux, or user attributes (if user_xattr
1590 is specified for the client mount options) then there is a risk
1591 of attribute loss. Additionally, the Lustre file striping
1592 needs to be larger than default (e.g. striped over all OSTs),
1593 and an additional attribute must be stored initially in the
1594 inode and then increase in size enough to be moved to the
1595 external attribute block (e.g. ACL growing in size) for file
1600 Description: skiplist implementation simplification
1601 Details : skiplists are used to group compatible locks on granted list
1602 that was implemented as tracking first and last lock of each
1603 lock group the patch changes that to using doubly linked lists
1607 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1608 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1609 because LASSERT on both the data supplied by a client, and the
1610 data on disk is dangerous and incorrect.
1612 Severity : enhancement
1614 Description: Slow truncate/writes to huge files at high offsets.
1615 Details : Directly associate cached pages to lock that protect those pages,
1616 this allows us to quickly find what pages to write and remove
1617 once lock callback is received.
1621 Description: more ldlm soft lockups
1622 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1623 starve other threads from the resource lock for a long time in
1624 case of long waiting queue, so change the debug level from
1625 D_OTHER to the less frequently used D_INFO.
1627 Severity : enhancement
1629 Description: add -gid, -group, -uid, -user options to lfs find
1633 Description: Hung threads in invalidate_inode_pages2_range
1634 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1635 once one is completed. As a result, some RPCs are stuck in the
1636 queue and are never sent.
1640 Description: filter threads hungs on waiting journal commit
1641 Details : Cleanup filter group llog code, then only filter group llog will
1642 be only created in the MDS/OST syncing process.
1646 Description: Procfs and llog threads access destoryed import sometimes.
1647 Details : Sync the import destoryed process with procfs and llog threads by
1648 the import refcount and semaphore.
1650 Severity : enhancement
1652 Description: openlock cache of b1_6 port to HEAD
1657 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1658 Details : If insertion of an extent fails, then discard the inode
1659 preallocation and free data blocks else it can lead to duplicate
1664 Description: don't always update ctime in ext3_xattr_set_handle()
1665 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1666 In some cases the ctime should not be updated, for example for
1667 2.0->1.8 compatibility it is necessary to delete an xattr and it
1668 should not update the ctime.
1672 Bugzilla : 15713/16362
1673 Description: Assertion in iopen_connect_dentry in 1.6.3
1674 Details : looking up an inode via iopen with the wrong generation number can
1675 populate the dcache with a disconneced dentry while the inode
1676 number is in the process of being reallocated. This causes an
1677 assertion failure in iopen since the inode's dentry list contains
1678 both a connected and disconnected dentry.
1682 Description: assertion failure in ldlm_handle2lock()
1683 Details : fix a race between class_handle_unhash() and class_handle2object()
1684 introduced in lustre 1.6.5 by bug 13622.
1689 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1690 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1695 Description: oops in sd_iostats_seq_show()
1696 Details : unloading/reloading the scsi low level driver triggers a kernel
1697 bug when trying to access the sd iostat file.
1702 Description: Kernel panics during QLogic driver reload
1703 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1704 patch, causing memory corruption.
1709 Description: journal_dev option does not work in b1_6
1710 Details : pass mount option during pre-mount.
1712 Severity : enhancement
1714 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1715 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1716 extent information of a file. It can be used to map logical blocks
1717 in a file to physical blocks in the block device.
1721 Description: LDLM soft lockups - improvement
1722 Details : It is be possible to send the lock handle along with each read
1723 or write request because the client is already doing a lock match
1724 itself so there isn't any reason the OST should have to re-do that
1728 Frequency : only on Cray X2
1730 Description: X2 build failures
1731 Details : fix build failures on Cray X2.
1735 Description: xid & resent requests
1736 Details : Initialize RPC XID from clock at startup (randomly if clock is
1739 Severity : enhancement
1741 Description: Add lustre_start utility to start or stop multiple Lustre servers
1746 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1747 Details : In case of memory pressure, list_del() can be called twice on
1748 req->rq_history_list, causing a kernel oops.
1752 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1753 Details : fix stack overflow in the distributed lock manager by defering
1754 export eviction after a failed AST to the elt thread instead of
1755 handling it in the dlm interpret routine.
1759 Description: Convert some comments to new format.
1760 Details : Update documenting comments to match doxygen conventions.
1764 Description: Grammar fixes.
1765 Details : A couple of trivial spelling fixes.
1769 Description: OSD_COUNTERS-mandatory
1770 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1771 to simplify the code.
1775 Description: simplify cmm_device freeing logic.
1776 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1780 Description: Add lockdep support to dt_object_operations locking interface.
1781 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1782 parameter indicating lock ordering. Update mdd code to use new
1787 Description: Introduce failloc constants for lockless IO tests.
1788 Details : Add two new failloc constants to test lockless IO. Only one of
1789 them in implemented---another is checked in yet to be landed
1794 Description: Add lockdep support for inode mutex.
1795 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1796 in the situations where Lustre has to lock more than one inode
1801 Description: Add optional invariants checking support.
1802 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1803 configure switch. This macro is to be used for consistency and
1804 sanity checks that are too expensive to be left in `production'
1809 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1810 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1815 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1816 Details : Use spin_lock_nested() in (the only) situation where more than
1817 one ldlm_lock is locked simultaneously. Also, fix possible
1818 dead-lock in ldlm_lock_change_resource() by enforcing particular
1823 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1824 Details : Instead of specifying each ldlm_lock call-back through separate
1825 parameter, wrap them into struct ldlm_callback_suite.
1829 Description: Kill join_lru obd method and its callers.
1830 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1831 in memory---a requirement implemented through ->o_join_lru() obd
1832 method in HEAD. Remove this method and its users.
1836 Description: Add asynchronous ldlm ENQUEUE completion handler.
1837 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1838 case that stock ldlm_completion_ast() cannot handle as it waits
1839 until lock is granted. Introduce new ldlm_completion_ast_async()
1840 for this. Also comment ldlm_completion_ast().
1844 Description: ldlm_error <-> errno conversion.
1845 Details : Add functions to map (rather arbitrary) between LDLM error codes
1846 and standard errno values. CLIO needs this to prevent LDLM specific
1847 constants from escaping ldlm and osc.
1851 Description: Kill unused ldlm_handle2lock_ns() function.
1852 Details : Kill unused ldlm_handle2lock_ns() function.
1856 Description: Add lu_ref support to ldlm_lock
1857 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1858 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1859 and ldlm_resource. LDLM interface has to be changed, because code that
1860 releases a reference on a lock, has to "know" what reference this is.
1861 In the most frequent case
1863 lock = ldlm_handle2lock(handle);
1865 LDLM_LOCK_PUT(lock);
1867 no changes are required. When any other reference (received _not_ from
1868 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1869 instead of LDLM_LOCK_PUT().
1871 Arguably, changes are pervasive, and interface requires some discipline
1872 for proper use. On the other hand, it was very instrumental in finding
1873 a few leaked lock references.
1877 Description: Add ldlm_lock_addref_try().
1878 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1879 attempts to addref a lock that might be being canceled
1884 Description: Add ldlm_weigh_callback().
1885 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1886 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1887 hard-coded `number of pages' logic.
1891 Description: Add lockdep annotations to llog code.
1892 Details : Use appropriately tagged _nested() locking calls in the places
1893 where llog takes more than one ->lgh_lock lock.
1897 Description: Add loi_kms_set().
1898 Details : Wrap kms updates into a helper function.
1902 Description: Constify instances of struct lsm_operations.
1903 Details : Constify instances of struct lsm_operations.
1907 Description: lu_conf support.
1908 Details : On a server, a file system object is uniquely identified
1909 by a fid, which is sufficient to locate and load all object
1910 state (inode). On a client, on the other hand, more data are
1911 necessary instantiate an object. Change lu_object_find() and
1912 friends to take additional `lu_conf' argument describing object.
1913 Typically this includes layout information.
1917 Description: lu_context fixes.
1918 Details : Introduce new lu_context functions that are needed on the client
1919 side, where some system threads (ptlrpcd) are shared by multiple
1920 modules, and so cannot be stopped during module shutdown.
1924 Description: Add start and stop methods to lu_device_type_operations.
1925 Details : Introduce two new methods in lu_device_type_operations, that are
1926 invoked when first instance of a given type is created and last one
1927 is destroyed respectively. This is need by CLIO.
1931 Description: Add lu_ref support to struct lu_device.
1932 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1933 track leaked references.
1937 Description: Introduce lu_kmem_descr.
1938 Details : lu_kmem_descr and its companion interface allow to create
1939 and destroy a number of kmem caches at once.
1943 Description: Fix lu_object finalization race.
1944 Details : Fix a race between lu_object_find() finding an object and its
1945 concurrent finalization. This race is (most likely) not possible
1946 on the server, but might happen on the client.
1950 Description: Introduce lu_ref interface.
1951 Details : lu_ref is a debugging module allowing to track references to
1952 a given object. It is quite cpu expensive, and has to be
1953 explicitly enabled with --enable-lu_ref. See usage description
1958 Description: Factor lu_site procfs stats into a separate function.
1959 Details : Separate lu_site stats printing code into a separate function
1960 that can be reused on a client.
1964 Description: Constify instances of struct {lu,dt,md}_device_operations.
1965 Details : Constify instances of struct {lu,dt,md}_device_operations.
1969 Description: Introduce struct md_site and move meta-data specific parts of
1970 struct lu_site here.
1971 Details : Move md-specific fields out of struct lu_site into special struct
1972 md_site, so that lu_site can be used on a client.
1976 Description: Kill mdd_lov_destroy().
1977 Details : Remove unused mdd code.
1981 Description: Add st_block checking to multistat.c.
1982 Details : Add st_block checking to multistat.c.
1986 Description: Add lu_ref support to struct obd_device.
1987 Details : Add lu_ref tracking to obd_device.
1991 Description: Kill obd_set_fail_loc().
1992 Details : Remove unused code.
1996 Description: Add special type for ptlrpc_request interpret functions.
1997 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1998 there. Actual usage will be in CLIO.
2002 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2003 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2004 doesn't work with lockdep.
2008 Description: Add rwv.c test program.
2009 Details : New testing program exercising readv(2) and writev(2) (Qian).
2013 Description: Add sendfile.c test program.
2014 Details : New testing program exercising sendfile(2) (Jay).
2018 Description: Ratelimit a message that can be very frequent.
2019 Details : Ratelimit a memory allocation failure message that can
2024 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2025 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2029 Description: Liblustre build fixes.
2030 Details : Liblustre build fixes.
2034 Description: libcfs: add cfs_{need,cond}_resched() interface.
2035 Details : libcfs: add cfs_{need,cond}_resched() definition and
2036 implementations for Linux, NT, and liblustre.
2038 Severity : enhancement
2040 Description: More exported tunables for mballoc
2041 Details : Add support for tunable preallocation window and new tunables for
2042 large/small requests
2046 Description: Detect corruption of block bitmap and checking for preallocations
2047 Details : Checks validity of on-disk block bitmap. Also it does better
2048 checking of number of applied preallocations. When corruption is
2049 found, it turns filesystem readonly to prevent further corruptions.
2053 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2054 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2055 window to 0. If it does hit cache miss, and read-pattern is still
2056 stride-io mode, does not reset the stride window, but also does not
2057 increase the stride window length in this case.
2061 Frequency : only for big-endian servers
2062 Description: Check if system is big-endian while mounting fs with extents feature
2063 Details : Mounting a filesystem with extents feature will fail on big-endian
2064 systems since ext3-based ldiskfs is not supported on big-endian
2065 systems. This can be over-riden with "bigendian_extents" mount option.
2067 Severity : enhancement
2069 Description: The root squash functionality
2070 Details : A security feature, which is to prevent users from being able
2071 to mount lustre on their desktop, run as root, and delete
2072 all of the files in the filesystem. The goal is accomplished by
2073 remapping user id (UID) and group id (GID) of the root user to
2074 a UID and GID specified by the system administartor via Lustre
2075 configuration management server (MGS). The functionality also
2076 allows to specify sets of clients for which the remapping does
2081 Description: Excessive recovery window
2082 Details : With AT enabled, the recovery window can be excessively long (6000+
2083 seconds). To address this problem, we no longer use
2084 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2085 timeout no longer depends on the service time, it is set to
2086 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2087 time via pb_service_time.
2091 Description: Watchdog triggered on MDS failover
2092 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2093 the OSTs know that the MDS "UUID" can be reused for the same export
2094 from a different NID, so we do not need to wait for the export to be
2098 Frequency : rare, only if using MMP with Linux RAID
2100 Description: MMP doesn't work with Linux RAID
2101 Details : While using HA for Lustre servers with Linux RAID, it is possible
2102 that MMP will not detect multiple mounts. To make this work we
2103 need to unplug the device queue in RAID when the MMP block is being
2104 written. Also while reading the MMP block, we should read it from
2105 disk and not the cached one.
2107 Severity : enhancement
2109 Description: open file using fid
2110 Details : A file can be opened using just its fid, like
2111 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2114 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2117 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2118 Details : add proper locking for creation of jbd_4k slab cache
2122 Description: MMP check in ext3_remount() fails without displaying any error
2123 Details : When multiple mount protection fails during remount, proper error
2126 Severity : enhancement
2128 Description: Allow stripe size to be up to 4G-64k
2129 Details : Fix math logic to allow large stripe sizes.
2133 Description: add check for >8TB ldiskfs filesystems
2134 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2135 Don't allow >8TB ldiskfs filesystems to be mounted without
2136 force_over_8tb mount option
2138 --------------------------------------------------------------------------------
2140 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2142 * Support for kernels:
2143 2.6.5-7.283 (SLES 9),
2144 2.6.9-55.EL (RHEL 4),
2145 2.6.16.46-0.14 (SLES 10),
2146 2.6.18.8 vanilla (kernel.org)
2147 * Client support for unpatched kernels:
2148 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2149 2.6.16 - 2.6.22 vanilla (kernel.org)
2150 * Due to recently discovered recovery problems, we do not recommend
2151 using patchless RHEL 4 clients with this or any earlier release.
2152 * Recommended e2fsprogs version: 1.39.cfs8
2153 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2154 * Starting with this release, the ldiskfs backing filesystem required
2155 by Lustre is now in its own package, lustre-ldiskfs. This package
2156 should be installed. It is versioned separately from Lustre and
2157 may be released separately in future.
2161 Description: block reactivating mgc import until all deactivates complete
2162 Details : Fix race when failing back MDT/MGS to itself (testing)
2165 Frequency : at statup only
2167 Description: mds_lov_synchronize race leads to various problems
2168 Details : simultaneous MDT->OST connections at startup can cause the
2169 sync to abort, leaving the OSC in a bad state.
2171 Severity : enhancement
2173 Description: add optional extra BUILD_VERSION info
2174 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2175 to override the lustre version.
2178 Frequency : 2.6.18 servers only
2180 Description: ll_kern_mount() doesn't release the module reference
2181 Details : The ldiskfs module reference count never drops down to 0
2182 because ll_kern_mount() doesn't release the module reference.
2187 Description: server LBUG when using old ost_num_threads parameter
2188 Details : Accept the old ost_num_threads parameter but warn that it
2189 is deprecated, and fix an off-by-one error that caused an LBUG.
2194 Description: Transient SCSI error results in persistent IO issue
2195 Details : iobuf->dr_error is not reinitialized to 0 between two
2199 Frequency : sometimes when underlying device returns I/O errors
2201 Description: OSTs not going read-only during write failures
2202 Details : OSTs are not remounted read-only when the journal commit threads
2203 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2204 instead of the ext3 wrappers.
2208 Description: poor connect scaling with increasing client count
2209 Details : Don't run filter_grant_sanity_check for more than 100 exports
2210 to improve scaling for large numbers of clients.
2213 Frequency : SLES10 only
2215 Description: sanity-quota.sh quotacheck failed: rc = -22
2216 Details : Quotas cannot be enabled on SLES10.
2219 Frequency : liblustre clients only
2221 Description: getdirentries does not give error when run on compute nodes
2222 Details : getdirentries does not fail when the size specified as an argument
2223 is too small to contain at least one entry
2225 Severity : enhancement
2227 Description: Add LNET router traceability for debug purposes
2228 Details : If a checksum failure occurs with a router as part of the
2229 IO path, the NID of the last router that forwarded the bulk data
2230 is printed so it can be identified.
2235 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2236 Details : Due to a race condition, liblustre clients were occasionally
2237 evicted incorrectly.
2239 Severity : enhancement
2241 Description: lfs setstripe use optional parameters instead of postional
2244 Severity : enhancement
2246 Description: Nanosecond timestamp support for ldiskfs
2247 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2248 resolution timestamps. There is not yet support for this at
2249 the Lustre filesystem level.
2252 Frequency : during server recovery
2254 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2255 Details : request with rq_no_resend flag not awake l_wait_event if they get
2259 Frequency : nfs export on patchless client
2261 Description: connectathon hang when test nfs export over patchless client
2262 Details : Disconnected dentry cannot be found with lookup, so we do not need
2263 to unhash it or make it invalid
2266 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2267 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2268 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2273 Description: open req refcounting wrong on reconnect
2274 Details : If reconnect happened between getting open reply from server and
2275 call to mdc_set_replay_data in ll_file_open, we will schedule
2276 replay for unreferenced request that we are about to free.
2277 Subsequent close will crash in variety of ways.
2278 Check that request is still eligible for replay in
2279 mdc_set_replay_data().
2284 Description: disable writes to filesystem when reading health_check file
2285 Details : the default for reading the health_check proc file has changed
2286 to NOT do a journal transaction and write to disk, because this
2287 can cause reads of the /proc file to hang and block HA state
2288 checking on a healthy but otherwise heavily loaded system. It
2289 is possible to return to the previous behaviour during configure
2290 with --enable-health-write.
2292 Severity : enhancement
2294 Description: 64-bit inode version
2295 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2296 to the inode. This will be required for version-based recovery.
2301 Description: MDS fails to start if a duplicate client export is detected
2302 Details : in some rare cases it was possible for a client to connect to
2303 an MDS multiple times. Upon recovery the MDS would detect this
2304 and fail during startup. Handle this more gracefully.
2306 Severity : enhancement
2308 Description: Add -o localflock option to simulate old noflock
2310 Details : This will achieve local-only flock/fcntl locks
2316 Description: log_commit_thread vs filter_destroy race leads to crash
2317 Details : Take import reference before releasing llog record semaphore
2322 Description: Wrong request locking in request set processing
2323 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2324 imp_delayed_list, in this place should be used imp_lock.
2327 Frequency : when reconnection
2329 Description: Grant Leak when osc reconnect to OST
2330 Details : When osc reconnect ost, OST(filter) should check whether it
2331 should grant more space to client by comparing fed_grant and
2332 cl_avail_grant, and return the granted space to client instead
2333 of "new granted" space, because client will call osc_init_grant
2334 to update the client grant space info.
2337 Frequency : when client reconnect to OST
2339 Description: Grant Leak when osc do resend and replay bulk write
2340 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2341 bulk write request, because the grant info while be sync between
2342 OSC and OST when reconnect, and we should ignore the grant info
2343 these of resend/replay write req.
2348 Description: Grant space more than avaiable left space sometimes.
2349 Details : When then OST is about to be full, if two bulk writing from
2350 different clients came to OST. Accord the avaliable space of the
2351 OST, the first req should be permitted, and the second one
2352 should be denied by ENOSPC. But if the seconde arrived before
2353 the first one is commited. The OST might wrongly permit second
2354 writing, which will cause grant space > avaiable space.
2357 Frequency : when client is evicted
2359 Description: Grant might be wrongly erased when osc is evicted by OST
2360 Details : when the import is evicted by server, it will fork another
2361 thread ptlrpc_invalidate_import_thread to invalidate the
2362 import, where the grant will be set to 0. While the original
2363 thread will update the grant it got when connecting. So if
2364 the former happened latter, the grant will be wrongly errased
2365 because of this race.
2370 Description: Checking Stale with correct fid
2371 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2372 to check whether it is stale, because sometimes, we want the
2373 enqueue happened anyway, and op_data.fid2 will not be initialized.
2375 Severity : enhancement
2377 Description: update patchless client
2378 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2381 Frequency : only with 2.4 kernel
2383 Description: random memory corruption
2384 Details : size of struct ll_inode_info is to big for union inode.u and this
2385 can be cause of random memory corruption.
2390 Description: Memory leak in recovery
2391 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2392 It should also check obd_fail before fsfilt_start, otherwise if
2393 fsfilt_start return -EROFS,(failover mds during mds recovery).
2394 then the req will return with repmsg->transno = 0 and rc = EROFS.
2395 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2396 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2397 be freed no matter whether fsfilt_commit success or not.
2400 Frequency : only with huge count clients
2402 Description: Prevents from taking the superblock lock in llap_from_page for
2404 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2405 for llap_from_page call in ll_removepage prevents from taking the
2406 superblock lock for a soon died page.
2411 Description: Not check open intent error before release open handle
2412 Details : in some rare cases, the open intent error is not checked before
2413 release open handle, which may cause
2414 ASSERTION(open_req->rq_transno != 0), because it tries to release
2415 the failed open handle.
2420 Description: Set cat log bitmap only after create log success.
2421 Details : in some rare cases, the cat log bitmap is set too early. and it
2422 should be set only after create log success.
2426 Description: Accessing a block bevice can re-enable I/O when Lustre is
2427 tearing down a device.
2428 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2433 Description: service threads may hog cpus when there are a lot of requests
2434 Details : Insert cond_resched to give other threads a chance to use some CPU
2439 Description: the cat log was not initialized in recovery
2440 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2441 unlink log on mds will not be initialized until mds post
2442 recovery. And also in mds post recovery, the unlink log will
2443 initialization will be done asynchronausly, so there will be race
2444 between add unlink log and unlink log initialization.
2448 Description: brw_stats were being printed incorrectly
2449 Details : brw_stats were being printed as log2 but all of them were not
2450 recorded as log2. Also remove some code duplication arising from
2451 filter_tally_{read,write}.
2455 Frequency : rare, only in recovery.
2456 Description: ASSERTION(req->rq_type != LI_POISON) failed
2457 Details : imp_lock should be held while iterating over imp_sending_list for
2458 prevent destroy request after get timeout in ptlrpc_queue_wait.
2462 Description: replay-single.sh test 52 fails
2463 Details : A lock's skiplist need to be cleanup when it being unlinked
2464 from its resource list.
2468 Description: Short directio read returns full requested size rather than
2470 Details : Direct I/O operations should return actual amount of bytes
2471 transferred rather than requested size.
2473 Severity : enhancement
2475 Description: metadata RPC reduction (e.g. for rm performance)
2476 Details : decrease the amount of synchronous RPC between clients and servers
2477 by canceling conflicing lock before the operation on the client
2478 and packing thier handles into the main operation RPC to server.
2480 Severity : enhancement
2482 Description: Async OSC create to avoid the blocking unnecessarily.
2483 Details : If a OST has no remain object, system will block on the creating
2484 when need to create a new object on this OST. Now, ways use
2485 pre-created objects when available, instead of blocking on an
2486 empty osc while others are not empty. If we must block, we block
2487 for the shortest possible period of time.
2491 Description: improve handling recoverable errors
2492 Details : if request processig with error which can be recoverable on server
2493 request should be resend, otherwise page released from cache and
2496 Severity : enhancement
2498 Description: refine locking for avoid write wrong info into lov_objid file
2499 Details : fix possible races with add new target and write/update data in
2502 --------------------------------------------------------------------------------
2504 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2509 Frequency : on some architectures
2511 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2512 Details : Uninitialized flags sometimes cause configuration commands to
2516 Frequency : patchless clients only
2518 Description: missing __iget() symbol export
2519 Details : The __iget() symbol export is missing. To avoid the need for
2520 this on patchless clients the deathrow inode reaper is turned
2521 off, and we depend on the VM to clean up old inodes. This
2522 dependency was during via the fix for bug 12181.
2524 --------------------------------------------------------------------------------
2526 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2528 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2529 INTEROPERATE with older versions automatically. In many cases a
2530 special upgrade step is needed. Please read the
2531 user documentation before upgrading any part of a live system.
2532 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2533 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2534 * WARNING: Lustre configuration and startup changes are required with
2535 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2539 Severity : enhancement
2541 Description: MountConf
2542 Details : Lustre configuration is now managed via mkfs and mount
2543 commands instead of lmc and lconf. New obd types (MGS, MGC)
2544 are added for dynamic configuration management. See
2545 https://mail.clusterfs.com/wikis/lustre/MountConf for
2548 Severity : enhancement
2550 Description: dynamic OST addition
2551 Details : OSTs can now be added to a live filesystem
2553 Severity : enhancement
2555 Description: startup order invariance
2556 Details : MDTs and OSTs can be started in any order. Clients only
2557 require the MDT to complete startup.
2559 Severity : enhancement
2561 Description: parallel, asynchronous orphan cleanup
2562 Details : orphan cleanup is now performed in separate threads for each
2563 OST, allowing parallel non-blocking operation.
2565 Severity : enhancement
2567 Description: optimized stripe assignment
2568 Details : stripe assignments are now made based on ost space available,
2569 ost previous usage, and OSS previous usage, in order to try
2570 to optimize storage space and networking resources.
2572 Severity : enhancement
2574 Description: Permanently set tunables
2575 Details : All writable /proc/fs/lustre tunables can now be permanently
2576 set on a per-server basis, at mkfs time or on a live system.
2578 Severity : enhancement
2580 Description: Lustre message v2
2581 Details : Add lustre message format v2.
2583 Severity : enhancement
2585 Description: client OST exclusion list
2586 Details : Clients can be started with a list of OSTs that should be
2587 declared "inactive" for known non-responsive OSTs.
2591 Description: SPEC SFS validation failure on NFS v2 over lustre.
2592 Details : Changes the blocksize for regular files to be 2x RPC size,
2593 and not depend on stripe size.
2595 Severity : enhancement
2597 Description: Multiple MD RPCs in flight.
2598 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2599 To avoid overly-overloading MDS, introduce a limit on number of
2600 MDS RPCs in flight for a single client and add /proc controls
2601 to adjust this limit.
2603 Severity : enhancement
2605 Description: client read/write statistics
2606 Details : Add client read/write call usage stats for performance
2607 analysis of user processes.
2608 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2609 file access. extents_stats shows chunk size distribution.
2610 extents_stats_per_process show chunk size distribution per
2613 Severity : enhancement
2615 Description: mds statistics
2616 Details : Add detailed mds operations statistics in
2617 /proc/fs/lustre/mds/*/stats.
2621 Description: Failure of copying files with lustre special EAs.
2622 Details : Client side always return success for setxattr call for lustre
2623 special xattr (currently only "trusted.lov").
2628 Description: Refcount LNET uuids
2629 Details : The global LNET uuid list grew linearly with every startup;
2630 refcount repeated list entries instead of always adding to
2633 Severity : enhancement
2635 Description: Dynamic service threads
2636 Details : Within a small range, start extra service threads
2637 automatically when the request queue builds up.
2640 Frequency : mixed-endian client/server environments
2642 Description: mixed-endian crashes
2643 Details : The new msg_v2 system had some failures in mixed-endian
2646 Severity : enhancement
2648 Description: Easy OST removal
2649 Details : OSTs can be permanently deactivated with e.g. 'lctl
2650 conf_param lustre-OST0001.osc.active=0'
2652 Severity : enhancement
2654 Description: MGS proc entries
2655 Details : Added basic proc entries for the MGS showing what filesystems
2658 Severity : enhancement
2660 Description: provide MGS failover
2661 Details : Added config lock reacquisition after MGS server failover.
2663 Severity : enhancement
2665 Description: add Linux 2.4 support
2666 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2670 Description: a large application tries to do I/O to the same resource and dies
2671 in the middle of it.
2672 Details : Check the req->rq_arrival time after the call to
2673 ost_brw_lock_get(), but before we do anything about
2674 processing it & sending the BULK transfer request. This
2675 should help move old stale pending locks off the queue as
2676 quickly as obd_timeout.
2679 Frequency : when an incorrect nid is specified during startup
2681 Description: ptlrpc connect to non-existant node causes kernel crash
2682 Details : LNET can't be re-entered from an event callback, which
2683 happened when we expire a message after the export has been
2684 cleaned up. Instead, hand the zombie cleanup off to another
2687 Severity : enhancement
2689 Description: plain/inodebits lock performance improvement
2690 Details : Grouping plain/inodebits in granted list by their request modes
2691 and bits policy, thus improving the performance of search through
2695 Frequency : only if OST filesystem is corrupted
2697 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2698 Details : for a short time RPCs with bulk IO are in the replay list,
2699 but replay of bulk IOs is unimplemented. If the OST filesystem
2700 is corrupted due to disk cache incoherency and then replay is
2701 started it is possible to trip an assertion. Avoid putting
2702 committed RPCs into the replay list at all to avoid this issue.
2705 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2707 Description: System hang on startup
2708 Details : This bug allowed the liblustre (e.g. catamount) client to
2709 return to the app before handling all startup RPCs. This
2710 could leave the node unresponsive to lustre network traffic
2711 and manifested as a server ptllnd timeout.
2713 Severity : enhancement
2715 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2716 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2717 boolean to control whether to print peer debug info when a
2718 client's RPC times out.
2721 Frequency : only for kernels with patches from Lustre below 1.4.3
2723 Description: Remove old rdonly API
2724 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2727 Frequency : only for devices with external journals
2729 Description: Set external device read-only also
2730 Details : During a commanded failover stop, we set the disk device
2731 read-only while the server shuts down. We now also set any
2732 external journal device read-only at the same time.
2735 Frequency : when upgrading from 1.4 while trying to change parameters
2737 Description: The wrong (new) MDC name was used when setting parameters for
2738 upgraded MDT's. Also allows changing of OSC (and MDC)
2739 parameters if --writeconf is specified at tunefs upgrade time.
2742 Frequency : when setting specific ost indicies
2744 Description: QOS code breaks on skipped indicies
2745 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2746 created with --index need not be sequential.
2748 Severity : enhancement
2750 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2751 Details : The uninit_groups feature works in conjunction with the kernel
2752 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2753 up the pass1 processing of e2fsck. This is a read-only feature
2754 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2755 mount filesystems that have had this feature enabled.
2757 Severity : enhancement
2759 Description: Improve multi-block allocation algorithm to avoid fragmentation
2760 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2761 allocation locality and avoid filesystem fragmentation.
2763 ------------------------------------------------------------------------------
2765 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2767 * Support for kernels:
2768 2.4.21-47.0.1.EL (RHEL 3)
2769 2.6.5-7.283 (SLES 9)
2770 2.6.9-42.0.10.EL (RHEL 4)
2771 2.6.12.6 vanilla (kernel.org)
2772 2.6.16.27-0.9 (SLES 10)
2773 * Recommended e2fsprogs version: 1.39.cfs5
2775 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2779 Frequency : occasional, depends on client load and configuration
2780 Bugzilla : 12181, 12203
2781 Description: data loss for recently-modified files
2783 Details : In some cases it is possible that recently written or created
2784 files may not be written to disk in a timely manner (this should
2785 normally be within 30s unless client IO load is very high).
2786 The problem appears as zero-length files or files that are a
2787 multiple of 1MB in size after a client crash or client eviction
2788 that are missing data at the end of the file.
2790 This problem is more likely to be hit on clients where files are
2791 repeatedly created and unlinked in the same directory, clients
2792 have a large amount of RAM, have many CPUs, the filesystem has
2793 many OSTs, the clients are rebooted frequently, and/or the files
2794 are not accessed by other nodes after being written.
2796 The presence of the problem can be detected by looking at
2797 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2798 smaller than the second (nr_unused) then dirty files will not
2799 be flushed automatically to disk. "sync; sleep 10" should be
2800 run several times on the node before unmounting it to update
2801 Lustre (this is also safe to run on nodes without this problem).
2803 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2804 that can cause this same problem, so customers using that kernel
2805 also need to update the kernel in addition to Lustre. In order
2806 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2808 It is normal that files written just before a client crash (less
2809 than 30s) may not yet have been flushed to disk, even for local
2813 Frequency : frequent on thin XT3 nodes
2815 Description: UUID collision on thin XT3 Linux nodes
2816 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2817 randomly, since we relied on an insufficiently-seeded PRNG.
2822 Description: OSS hangs after "All ost request buffers busy"
2823 Details : A deadlock between quota and journal operations caused OSS
2824 hangs after printing "All ost request buffers busy."
2827 Frequency : always on liblustre builds
2829 Description: Cleanup compiler warnings on liblustre
2832 Frequency : always on liblustre builds on XT3
2834 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2838 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2839 > 32000 subdirectories
2840 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2841 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2842 subdirectory count crosses 32000. This will aid e2fsck to
2843 correctly handle more than 32000 subdirectories.
2846 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2848 Description: System hang on startup
2849 Details : This bug allowed the liblustre (e.g. catamount) client to
2850 return to the app before handling all startup RPCs. This
2851 could leave the node unresponsive to lustre network traffic
2852 and manifested as a server ptllnd timeout.
2854 Severity : enhancement
2856 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2857 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2858 boolean to control whether to print peer debug info when a
2859 client's RPC times out.
2864 Description: make O_SYNC working on 2.6 kernels
2865 Details : 2.6 kernels use different method for mark pages for write,
2866 so need add a code to lustre for O_SYNC work.
2871 Description: Failure to close file and release space on NFS
2872 Details : Put inode details into lock acquired in ll_intent_file_open.
2873 Use mdc_intent_lock in ll_intent_open to properly
2874 detect all kind of errors unhandled by mdc_enqueue.
2879 Description: proc file read during shutdown sometimes raced obd removal,
2881 Details : Add lock to prevent obd access after proc file removal.
2884 Frequency : Only for files larger than 4GB on 32-bit clients.
2886 Description: improperly doing page alignment of locks
2887 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2888 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2893 Description: RPCs being resent when they shouldn't be
2894 Details : Some RPCs that should not be resent are being resent. This
2895 can cause inconsistencies in the RPC state machine. Do not
2896 resend such requests.
2899 Frequency : rare, only with NFS export
2901 Description: Crash on NFS re-export node
2902 Details : under very unusual load conditions an assertion is hit in
2903 ll_intent_file_open()
2906 Frequency : only if OST filesystem is corrupted
2908 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2909 Details : for a short time RPCs with bulk IO are in the replay list,
2910 but replay of bulk IOs is unimplemented. If the OST filesystem
2911 is corrupted due to disk cache incoherency and then replay is
2912 started it is possible to trip an assertion. Avoid putting
2913 committed RPCs into the replay list at all to avoid this issue.
2918 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2919 Details : Large single O_DIRECT read and write calls can fail to allocate
2920 a sufficiently large buffer to process the request. In case of
2921 allocation failure the allocation is retried with a smaller
2922 buffer and broken into smaller requests.
2924 Severity : enhancement
2926 Description: Add -o localflock option to simulate old noflock behaviour.
2927 Details : This will achieve local-only flock/fcntl locks coherentness.
2932 Description: versioning check is incomplete
2933 Details : Checking the version difference of client vs. server, report
2934 error if the gap is too big.
2939 Description: add support PG_writeback bit
2940 Details : add support for PG_writeback bit for Lustre, for more carefull
2941 work with page cache in 2.6 kernel. This also fix some deadlocks
2942 and remove hack for work O_SYNC with 2.6 kernel.
2944 Severity : enhancement
2946 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2947 Details : The uninit_groups feature works in conjunction with the kernel
2948 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2949 up the pass1 processing of e2fsck. This is a read-only feature
2950 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2951 mount filesystems that have had this feature enabled.
2953 Severity : enhancement
2955 Description: Improve multi-block allocation algorithm to avoid fragmentation
2956 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2957 allocation locality and avoid filesystem fragmentation.
2959 ------------------------------------------------------------------------------
2961 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2963 * Support for kernels:
2964 2.6.9-42.0.3.EL (RHEL 4)
2965 2.6.5-7.276 (SLES 9)
2966 2.4.21-47.0.1.EL (RHEL 3)
2967 2.6.12.6 vanilla (kernel.org)
2968 2.6.16.21-0.8 (SLES10)
2969 * Recommended e2fsprogs version: 1.39.cfs2-0
2971 * The backwards-compatible /proc/sys/portals symlink has been removed
2972 in this release. Before upgrading, please ensure that you change
2973 any configuration scripts or /etc/sysctl.conf files that access
2974 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2975 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2976 in advance of the upgrade on any system running Lustre 1.4.6 or
2977 newer, since /proc/sys/lnet was added in that version.
2978 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2982 Frequency : only when quota is used
2984 Description: avoid scanning export list for quota master
2985 Details : Change the algorithms to avoid scanning export list in order
2986 to improve the efficiency.
2989 Frequency : MDS failover only, very rarely
2991 Description: "went back in time" messages on mds failover
2992 Details : The greatest transno may be lost when the current operation
2993 finishes with an error (transno==0) and the client's last_rcvd
2994 record is over-written. Save the greatest transno in the
2995 mds_last_transno for this case.
2998 Frequency : always for specific kernels and striping counts
3000 Description: client may get "Matching packet too big" without ACL support
3001 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3002 when trying to access files in certain configurations. The
3003 clients should in fact be denied when mounting because they do
3004 not understand ACLs.
3007 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3009 Description: many clients connecting with IO in progress causes connect timeouts
3010 Details : Avoid synchronous journal commits to avoid delays caused by many
3011 clients connecting/disconnecting when bulk IO is in progress.
3012 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3013 OST_IO_PORTAL to avoid delays behind potentially many pending
3017 Frequency : occasionally with multiple writers to a single file
3019 Description: shared writes to file may result in wrong size reported by stat()
3020 Details : Allow growing of kms when extent lock is cancelled
3023 Frequency : always with random mmap IO to multi-striped file
3025 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3026 Details : Only if the hole is at the end of OST object so that kms is too
3027 small. Fix is to increase kms accordingly in ll_nopage.
3030 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3032 Description: writes to a missing object would leak memory on the OST
3033 Details : If there is an inconsistency between the MDS and OST filesystems,
3034 such that the MDS references an object that doesn't exist, writes
3035 to that object will leak memory due to incorrect cleanup in the
3036 error handling path, eventually running out of memory on the OST.
3041 Description: Creating too long symlink causes lustre errors
3042 Details : Check symlink and name lengths before sending requests to MDS.
3045 Frequency : only if flock is enabled (not on by default)
3047 Description: posix locks not released on fd closure on 2.6.9+
3048 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3049 kernels, this caused such locks not to be released on fd close and
3050 then assertions on fs unmount about still used locks.
3053 Frequency : MDS failover only, very rarely
3055 Description: clients may get ASSERTION(granted_lock != NULL)
3056 Details : When request was taking a long time, and a client was resending
3057 a getattr by name lock request. The were multiple lock requests
3058 with the same client lock handle and
3059 mds_getattr_name->fixup_handle_for_resent_request found one of the
3060 lock handles but later failed with ASSERTION(granted_lock != NULL).
3065 Description: handle->h_buffer_credits > 0, assertion failure
3066 Details : h_buffer_credits is zero after truncate, causing assertion
3067 failure. This patch extends the transaction or creates a new
3071 Frequency : NFS re-export or patchless client
3072 Bugzilla : 11179, 10796
3073 Description: Crash on NFS re-export node (__d_move)
3074 Details : We do not want to hash the dentry if we don't have a lock.
3075 But if this dentry is later used in d_move, we'd hit uninitialised
3076 list head d_hash, so we just do this to init d_hash field but
3077 leave dentry unhashed.
3080 Frequency : NFS re-export or patchless client
3082 Description: NFS exports has problem with symbolic link
3083 Details : lustre client didn't properly install dentry when re-exported
3084 to NFS or running patchless client.
3087 Frequency : NFS re-export or patchless client
3089 Description: Various nfs/patchless fixes.
3090 Details : fixes reuse disconected alias for lookup process - this fixes
3091 warning "find_exported_dentry: npd != pd",
3092 fix permission error with open files at nfs.
3093 fix apply umask when do revalidate.
3096 Frequency : occasional
3098 Description: Crash on NFS re-export node
3099 Details : calling clear_page() on the wrong pointer triggered oops in
3100 generic_mapping_read().
3103 Frequency : rarely, using O_DIRECT IO
3105 Description: unaligned directio crashes client with LASSERT
3106 Details : check for unaligned buffers before trying any requests.
3109 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3111 Description: stale data returned from RAID cache
3112 Details : If only a small amount of IO is done to the RAID device before
3113 reading it again it is possible to get stale data from the RAID
3114 cache instead of reading it from disk.
3117 Frequency : always for sles10 kernel
3119 Description: sles10 support
3120 Details : ll_follow_link: compile fixes and using of nd_set_link
3121 under newer kernels.
3124 Frequency : depends on arch, kernel and compiler version, always on sles10
3127 Description: recursive or deep enough symlinks cause stack overflow
3128 Details : getting rid of large stack-allocated variable in
3132 Frequency : depends on hardware
3134 Description: lustre write performance loss in the SLES10 kernel
3135 Details : the performance loss is caused by using of write barriers in the
3136 ext3 code. The SLES10 kernel turns barrier support on by
3137 default. The fix is to undo that change for ldiskfs.
3139 ------------------------------------------------------------------------------
3141 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3143 * Support for kernels:
3144 2.6.9-42.0.3EL (RHEL 4)
3145 2.6.5-7.276 (SLES 9)
3146 2.4.21-47.0.1.EL (RHEL 3)
3147 2.6.12.6 vanilla (kernel.org)
3151 Frequency : quota enabled and large files being deleted
3153 Description: releasing more than 4GB of quota at once hangs OST
3154 Details : If a user deletes more than 4GB of files on a single OST it
3155 will cause the OST to spin in an infinite loop. Release
3156 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3161 Description: statfs data retrieved from /proc may be stale or zero
3162 Details : When reading per-device statfs data from /proc, in the
3163 {kbytes,files}_{total,free,avail} files, it may appear
3164 as zero or be out of date.
3167 Frequency : systems with MD RAID1 external journal devices
3169 Description: lconf's call to blkid is confused by RAID1 journal devices
3170 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3171 of returning all block devices that match the journal UUID.
3174 Frequency : always, for aggregate stripe size over 4GB
3176 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3177 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3178 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3181 Frequency : always if "lfs find" used on a local file/directory
3183 Description: "lfs find" segfaults if used on a local file/directory
3184 Details : The case where a directory component was not specified wasn't
3185 handled correctly. Handle this properly.
3188 Frequency : always on ppc64
3190 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3191 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3192 perform endianness conversion.
3195 Frequency : rarely (truncate to non-zero file size after write under load)
3196 Bugzilla : 10730, 10687
3197 Description: Files padded with zeros to next 4K multiple
3198 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3199 it is possible that files that are truncated to a non-zero size
3200 immediately after being written are filled with zero bytes beyond
3201 the truncated size. No file data is lost.
3203 Severity : enhancement
3205 Description: Allow recovery/failover for liblustre clients.
3206 Details : liblustre clients were unaware of failover configurations until
3209 Severity : enhancement
3211 Description: user file locks should fail when not mounting with flock option
3212 Details : Set up an error-returning stub in ll_file_operations.lock field
3213 to prevent incorrect behaviour when client is mounted without
3214 flock option. Also, set up properly f_op->flock field for
3218 Frequency : always on ia64
3220 Description: "lfs df" loops on printing out MDS statfs information
3221 Details : The obd_ioctl_data was not initialized and in some systems
3222 this caused a failure during the ioctl that did not return
3223 an error. Initialize the struct and return an error on failure.
3226 Frequency : SLES 9 only
3228 Description: Error of copying files with lustre special EAs as root
3229 Details : Client side always return success for setxattr call for lustre
3230 special xattr (currently only "trusted.lov").
3233 Frequency : rarely on clusters with both ia64+i386 clients
3235 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3236 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3237 clients are doing concurrent writes to the same file it is
3238 possible that the ia64 clients may LASSERT because the OST
3239 extent locks are not PAGE_SIZE aligned. Ensure that grown
3240 locks are always aligned on the request boundary.
3243 Frequency : specific use, occasional
3245 Description: Overwriting in use executable truncates on-disk binary image
3246 Details : If one node attempts to overwrite an executable in use by
3247 another node, we now correctly return ETXTBSY instead of
3248 truncating the file.
3250 Severity : enhancement
3252 Description: Async OSC create to avoid the blocking unnecessarily.
3253 Details : If a OST has no remain object, system will block on the creating
3254 when need to create a new object on this OST. Now, ways use
3255 pre-created objects when available, instead of blocking on an
3256 empty osc while others are not empty. If we must block, we block
3257 for the shortest possible period of time.
3262 Description: chmod on Lustre root is propagated to other clients
3263 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3264 invalid by the follow_mount time.
3267 Frequency : liblustre clients only
3269 Description: Race in 'instant cancel' lock handling could lead to such locks
3270 never to be granted in case of SMP MDS
3271 Details : Do not destroy not yet granted but cbpending locks in
3275 Frequency : replay/resend of open
3277 Description: non null lock assetion failure in mds_intent_policy
3278 Details : Trying to replay/resend lockless open requests resulted in
3279 mds_open() returning 0 with no lock. Now it sets a flag if
3280 a lock is going to be returned.
3282 Severity : enhancement
3284 Description: Checksum enhancements
3285 Details : New checksum enhancements allow for resending RPCs that failed
3288 Severity : enhancement
3290 Description: Tunables on number of dirty pages in cacche
3291 Details : Allow to set limit on number of dirty pages cached.
3296 Description: client crash on unmount - lock still has references
3297 Details : In some error handling cases it was possible to leak a lock
3298 reference on a client while accessing a file. This was not
3299 harmful to the client during operation, but would cause the
3300 client to crash when the filesystem is unmounted.
3303 Frequency : specific case, rare
3305 Description: ETXTBSY on mds though file not in use by client
3306 Details : ETXTBSY is no longer incorrectly returned when attempting to
3307 chmod or chown a directory that the user previously tried to
3308 execute or a currently-executing binary.
3311 Frequency : extremely rare except on liblustre-based clients
3313 Description: Lustre space not freed when files are deleted
3314 Details : Clean up open-unlinked files after client eviction. Previously
3315 the unlink was skipped and the files remained as orphans.
3320 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3321 Details : In some cases it was possible to send a blocking callback to
3322 a client doing a glimpse, even though that client didn't get
3323 a lock granted. When the glimpse lock is cancelled on the OST
3324 the freed lock is left on the waiting list and corrupted the list.
3327 Frequency : all core dumps
3329 Description: Broke core dumps to lustre
3330 Details : Negative dentry may be unhashed if parent does not have UPDATE
3331 lock, but some callers, e.g. do_coredump, expect dentry to be
3332 hashed after successful create, hash it in ll_create_it.
3334 ------------------------------------------------------------------------------
3336 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3338 * Support for kernels:
3339 2.6.9-42.0.2.EL (RHEL 4)
3340 2.6.5-7.276 (SLES 9)
3341 2.4.21-40.EL (RHEL 3)
3342 2.6.12.6 vanilla (kernel.org)
3346 Frequency : always on RHEL 3
3348 Description: Number of open files grows over time
3349 Details : The number of open files grows over time, whether or not
3350 Lustre is started. This was due to a filp leak introduced
3351 by one of our kernel patches.
3353 ------------------------------------------------------------------------------
3355 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3357 * Support for kernels:
3358 2.6.9-42.EL (RHEL 4)
3359 2.6.5-7.276 (SLES 9)
3360 2.4.21-40.EL (RHEL 3)
3361 2.6.12.6 vanilla (kernel.org)
3366 Bugzilla : 5719, 9635, 9792, 9684
3367 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3368 Details : If a server is under heavy load and cannot reply to new
3369 connection requests before the client resends the (re)connect,
3370 the connection handling code can behave badly if two service
3371 threads are concurrently handing separate (re)connections from
3372 the same client. Add better locking to the connection handling
3373 code, and ensure that only a single connection will be processed
3374 for a given client UUID, even if the lock is dropped.
3376 Severity : enhancement
3378 Description: add TCP zero-copy support to kernel
3379 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3380 sends if the hardware supports scatter-gather and checksumming.
3381 This allows socklnd to do client-write and server-read more
3382 efficiently and reduce CPU utilization from skbuf copying.
3385 Frequency : only if NFS exporting from client
3387 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3388 Details : If setting attributes on a file created under NFS that had
3389 never been opened it would be possible to oops the client
3390 if the file had no objects.
3393 Frequency : always for liblustre
3395 Description: liblustre client does MDS+OSTs setattr RPC for each write
3396 Details : When doing a write from a liblustre client, the client
3397 incorrectly issued an RPC to the MDS and each OST the file was
3398 striped over in order to update the timestamps. When writing
3399 with small chunks and many clients this could overwhelm the MDS
3400 with RPCs. In all cases it would slow down the write because
3401 these RPCs are unnecessary.
3403 Severity : enhancement
3405 Description: allow number of MDS service threads to be changed at module load
3406 Details : It is now possible to change the number of MDS service threads
3407 running. Adding "options mds mds_num_threads={N}" to the MDS's
3408 /etc/modprobe.conf will set the number of threads for the next
3409 time Lustre is restarted (assuming the "mds" module is also
3410 reloaded at that time). The default number of threads will
3411 stay the same, 32 for most systems.
3416 Description: OST crash if filesystem is unformatted or corrupt
3417 Details : If an OST is started on a device that has never been formatted
3418 or if the filesystem is corrupt and cannot even mount then the
3419 error handling cleanup routines would dereference a NULL pointer.
3424 Description: NULL pointer deref in llap_from_page.
3425 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3426 incorrect) mapping if the page was truncated/reclaimed while it was
3427 searched for. Check for this condition and skip such pages when
3428 doing readahead. Introduce extra check to llap_from_page() to
3429 verify page->mapping->host is non-NULL (so page is not anonymous).
3432 Frequency : Sometimes when using sys_sendfile
3434 Description: "page not covered by a lock" warnings from ll_readpage
3435 Details : sendfile called ll_readpage without right page locks present.
3436 Now we introduced ll_file_sendfile that does necessary locking
3437 around call to generic_file_sendfile() much like we do in
3441 Frequency : with certain MDS communication failures at client mount time
3443 Description: NULL pointer deref after failed client mount
3444 Details : a client connection request may delayed by the network layer
3445 and not be sent until after the PTLRPC layer has timed out the
3446 request. If the client fails the mount immediately it will try
3447 to clean up before the network times out the request. Add a
3448 reference from the request import to the obd device and delay
3449 the cleanup until the network drops the request.
3452 Frequency : occasionally during client (re)connect
3454 Description: assertion failure during client (re)connect
3455 Details : processing a client connection request may be delayed by the
3456 client or server longer than the client connect timeout. This
3457 causes the client to resend the connection request. If the
3458 original connection request is replied in this interval, the
3459 client may trip an assertion failure in ptlrpc_connect_interpret()
3460 which thought it would be the only running connect process.
3463 Frequency : only with obd_echo servers and clients that are rebooted
3465 Description: kernel BUG accessing uninitialized data structure
3466 Details : When running an obd_echo server it did not start the ping_evictor
3467 thread, and when a client was evicted an uninitialized data
3468 structure was accessed. Start the ping_evictor in the RPC
3469 service startup instead of the OBD startup.
3471 Severity : enhancement
3472 Bugzilla : 10193 (patchless)
3473 Description: Remove dependency on various unexported kernel interfaces.
3474 Details : No longer need reparent_to_init, exit_mm, exit_files,
3475 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3478 Frequency : rare (only users of deprecated and unsupported LDAP config)
3480 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3481 Details : LDAP apparently contains 'lustreName' attributes instead of
3482 'name'. A simple remapping of the name is sufficient.
3485 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3487 Description: waiting_locks_callback trips kernel BUG if client is evicted
3488 Details : Running with the dump_on_timeout debug flag turned on makes
3489 it possible that the waiting_locks_callback() can try to dump
3490 the Lustre kernel debug logs from an interrupt handler. Defer
3491 this log dumping to the expired_lock_main() thread.
3493 Severity : enhancement
3495 Description: Support NFS exporting on 2.6 kernels.
3496 Details : Implement non-rawops metadata methods for NFS server to use without
3497 changing NFS server code.
3500 Frequency : very rare (synthetic metadata workload only)
3502 Description: two racing renames might cause an MDS thread to deadlock
3503 Details : Running the "racer" program may cause one MDS thread to rename
3504 a file from being the source of a rename to being the target of
3505 a rename at exactly the same time that another thread is doing
3506 so, and the second thread has already enqueued these locks after
3507 doing a lookup of the target and is trying to relock them in
3508 order. Ensure that we don't try to re-lock the same resource.
3511 Frequency : only very large systems with liblustre clients
3513 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3514 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3515 all OSTs in parallel. This allows the eviction of stale liblustre
3516 clients to proceed much faster than if they were done in series,
3517 and also offers similar improvements for other set_info RPCs.
3522 Description: excessive CPU usage during initial read phase on client
3523 Details : During the initial read phase on a client, it would agressively
3524 retry readahead on the file, consuming too much CPU and impacting
3525 performance (since 1.4.5.8). Improve the readahead algorithm
3526 to avoid this, and also improve some other common cases (read
3527 of small files in particular, where "small" is files smaller than
3528 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3533 Description: MDS crash when receiving packet with unknown intent.
3534 Details : Do not LBUG in unknown intent case, just return -EFAULT
3536 Severity : enhancement
3537 Bugzilla : 9293, 9385
3538 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3539 Details : Do not serialize getattr (non-intent version) and statfs.
3542 Frequency : occasional, when OST network is overloaded/intermittent
3544 Description: client evicted by OST after bulk IO timeout
3545 Details : If a client sends a bulk IO request (read or write) the OST
3546 may evict the client if it is unresposive to its data GET/PUT
3547 request. This is incorrect if the network is overloaded (takes
3548 too long to transfer the RPC data) or dropped the OST GET/PUT
3549 request. There is no need to evict the client at all, since
3550 the pinger and/or lock callbacks will handle this, and the
3551 client can restart the bulk request.
3554 Frequency : Always when mmapping file with no objects
3556 Description: client crashes when mmapping file with no objects
3557 Details : Check that we actually have objects in a file before doing any
3558 operations on objects in ll_vm_open, ll_vm_close and
3564 Description: Request leak when working with deleted CWD
3565 Details : Introduce advanced request refcount tracking for requests
3566 referenced from lustre intent.
3568 Severity : Enhancement
3570 Description: Cache open file handles on client.
3571 Details : MDS now will return special lock along with openhandle, if
3572 requested and client is allowed to hold openhandle, even if unused,
3573 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3574 closing files for every read/write openration.
3576 Severity : Enhancement
3578 Description: Cache open negative dentries on client when possible.
3579 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3580 negative dentries on lock revocation.
3585 Description: Remounting a client read-only wasn't possible with a zconf mount
3586 Details : It wasn't possible to remount a client read-only with llmount.
3588 Severity : enhancement
3589 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3590 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3591 MPICH is included to add Lustre-specific ADIO interfaces.
3592 This is based closely on the UFS ADIO layer and only differs
3593 in file creation, in order to allow the OST striping to be set.
3594 This is user-contributed code and not supported by CFS.
3599 Description: extended inode attributes (immutable, append-only) work improperly
3600 when 2.4 and 2.6 kernels are used on client/server or vice versa
3601 Details : Introduce kernel-independent values for these flags.
3603 Severity : enhancement
3606 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3607 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3608 granularity. Now, Fractional MB granularity is be supported,
3609 this is very useful for low memory system.
3611 Severity : enhancement
3613 Description: Getattr by fid
3614 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3615 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3616 have locks and avoids getattr rpc requests for it on every stat.
3619 Frequency : Always, for filesystems larger than 2TB
3621 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3622 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3623 filesystems larger than 2TB to be mounted with mballoc enabled.
3626 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3628 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3629 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3630 filesystems larger than 2TB is mounted then the kernel will
3631 silently corrupt the start of the filesystem. CONFIG_LBD is
3632 enabled for all CFS-supported kernels, but the possibility of
3633 this happening with a modified kernel config exists.
3635 Severity : enhancement
3637 Description: add client O_DIRECT support for 2.6 kernels
3638 Details : It is now possible to do O_DIRECT reads and writes to files
3639 in the Lustre client mountpoint on 2.6 kernel clients.
3641 Severity : enhancement
3643 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3644 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3645 parallel, not waiting for response from every OST before sending
3646 a rpc to the next OST.
3651 Description: setattr vs write race when updating file timestamps
3652 Details : Client processes that update a file timestamp into the past
3653 right after writing to the file (e.g. tar) it is possible that
3654 the updated file modification time can be reset to the current
3655 time due to a race between processing the setattr and write RPC.
3657 Severity : enhancement
3659 Description: Bring 'lfs find' closer in line with regular Linux find.
3660 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3661 -print0 options and obtains all the needed info through the lustre
3664 Severity : enhancement
3666 Description: support up to 1024 configured devices on one node
3667 Details : change obd_dev array from statically allocated to dynamically
3668 allocated structs as they are first used to reduce memory usage
3673 Description: Flush dirty partially truncated pages during truncate
3674 Details : Immediatelly flush partially truncated pages in filter_setattr,
3675 this way we completely avoid having any pages in page cache on OST
3676 and can retire ugly workarounds during writes to flush such pages.
3681 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3682 Details : protect inode from truncation within vfs_unlink() context
3683 just take a reference before calling vfs_unlink() and release it
3684 when parent's i_sem is free.
3689 Description: last_id value checked outside lock on OST caused LASSERT failure
3690 Details : If there were multiple MDS->OST object precreate requests in
3691 flight, it was possible that the OST's last object id was checked
3692 outside a lock and incorrectly tripped an assertion. Move checks
3693 inside locks, and discard old precreate requests.
3696 Frequency : always, if extents are used on OSTs
3698 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3699 Details : OSTs using the extents format would not zero the high 16 bits of
3700 the index physical block number. This is not a problem for any
3701 OST filesystems smaller than 16TB, and no kernels support ext3
3702 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3703 new/modified files) and can be fixed for existing filesystems
3704 with e2fsprogs-1.39-cfs1.
3709 Description: import connection selection may be incorrect if timer wraps
3710 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3711 import connections to be ignored if the 32-bit jiffies counter
3712 wraps. Use a 64-bit jiffies counter.
3715 Frequency : very large clusters immediately after boot
3717 Description: LNET request buffers exhausted under heavy short-term load
3718 Details : If a large number of client requests are generated on a service
3719 that has previously never seen so many requests it is possible
3720 that the request buffer growth cannot keep up with the spike in
3721 demand. Instead of dropping incoming requests, they are held in
3722 the LND until the RPC service can accept more requests.
3725 Frequency : Sometimes during replay
3727 Description: Assertion failure in ll_local_open after replay.
3728 Details : If replay happened on an open request reply before we were able
3729 to set replay handler, reply will become not swabbed tripping the
3730 assertion in ll_local_open. Now we set the handler right after
3731 recognising of open request
3734 Frequency : very rare
3736 Description: kernel reports "badness in vsnprintf"
3737 Details : Reading from the "recovery_status" /proc file in small chunks
3738 may cause a negative length in lprocfs_obd_rd_recovery_status()
3739 call to vsnprintf() (which is otherwise harmless). Exit early
3740 if there is no more space in the output buffer.
3742 Severity : enhancement
3744 Description: clear OBD RPC statistics by writing to them
3745 Details : It is now possible to clear the OBD RPC statistics by writing
3746 to the "stats" file.
3751 Description: Client mtime is not the same on different clients after utimes
3752 Details : In some cases, the client was using the utimes() syscall on
3753 a file cached on another node. The clients now validate the
3754 ctime from the MDS + OSTs to determine which one is right.
3759 Description: Inability to activate failout mode
3760 Details : lconf script incorrectly assumed that in pythong string's numeric
3761 value is used in comparisons.
3764 Frequency : always with multiple stripes per file
3766 Description: Inefficient object allocation for mutli-stripe files
3767 Details : When selecting which OSTs to stripe files over, for files with
3768 a stripe count that divides evenly into the number of OSTs,
3769 the MDS is always picking the same starting OST for each file.
3770 Return the OST selection heuristic to the original design.
3775 Description: mount failures may take full timeout to return an error
3776 Details : Under some heavy load conditions it is possible that a
3777 failed mount can wait for the full obd_timeout interval,
3778 possibly several minutes, before reporting an error.
3779 Instead return an error as soon as the status is known.
3781 Frequency : quota enabled and large files being deleted
3783 Description: releasing more than 4GB of quota at once hangs OST
3784 Details : If a user deletes more than 4GB of files on a single OST it
3785 will cause the OST to spin in an infinite loop. Release
3786 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3791 Description: statfs data retrieved from /proc may be stale or zero
3792 Details : When reading per-device statfs data from /proc, in the
3793 {kbytes,files}_{total,free,avail} files, it may appear
3794 as zero or be out of date.
3797 Frequency : systems with MD RAID1 external journal devices
3799 Description: lconf's call to blkid is confused by RAID1 journal devices
3800 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3801 of returning all block devices that match the journal UUID.
3804 Frequency : always, for aggregate stripe size over 4GB
3806 Description: assertion fails when trying to use 4GB stripe size
3807 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3808 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3811 Frequency : always on ppc64
3813 Description: the first write on an ext3 filesystem with mballoc got stuck
3814 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3815 perform endianness conversion.
3817 ------------------------------------------------------------------------------
3819 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3821 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3822 INTEROPERATE with older versions automatically. Please read the
3823 user documentation before upgrading any part of a live system.
3824 * WARNING: Lustre networking configuration changes are required with
3825 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3828 * Support for kernels:
3829 2.6.9-22.0.2.EL (RHEL 4)
3830 2.6.5-7.244 (SLES 9)
3831 2.6.12.6 vanilla (kernel.org)
3834 Severity : enhancement
3835 Bugzilla : 7981/8208
3836 Description: Introduced Lustre Networking (LNET)
3837 Details : LNET is new networking infrastructure for Lustre, it includes
3838 a reorganized network configuration mode (see the user
3839 documentation for full details) as well as support for routing
3840 between different network fabrics. Lustre Networking Devices
3841 (LNDs) for the supported network fabrics have also been
3842 created for this new infrastructure.
3844 Severity : enhancement
3845 Description: Introduced Access control lists
3846 Details : clients can set ACLs on files and directories in order to have
3847 more fine-grained permissions than the standard Unix UGO+RWX.
3848 The MDS must be started with the "-o acl" mount option.
3850 Severity : enhancement
3851 Description: Introduced filesystem quotas
3852 Details : Administrators may now establish per-user quotas on the
3855 Severity : enhancement
3857 Description: Configuration change for the XT3
3858 The PTLLND is now used to run Lustre over Portals on the XT3
3859 The configure option(s) --with-cray-portals are no longer used.
3860 Rather --with-portals=<path-to-portals-includes> is used to
3861 enable building on the XT3. In addition to enable XT3 specific
3862 features the option --enable-cray-xt3 must be used.
3867 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3868 Details : A race between allocating a new llcd and re-getting the llcd_lock
3869 allowed another thread to grab newly-allocated llcd.
3871 Severity : enhancement
3873 Description: 2.6 OST async journal commit and locking fix to improve performance
3874 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3875 2.6 kernels are now async as they already were in 2.4 kernels so
3876 that they can commit concurrently with the network bulk transfer.
3877 For block-allocated files the filter allocation semaphore is held
3878 to avoid filesystem fragmentation during allocation. BKL lock
3879 removed for 2.6 xattr operations where it is no longer needed.
3884 Description: lconf incorrectly determined whether two IP networks could talk
3885 Details : In some more complicated routing and multiple-network
3886 configurations, lconf will avoid trying to make a network
3887 connection to a disjoint part of the IP space. It was doing the
3888 math incorrectly for one set of cases.
3893 Description: Fix for potential infinite loop processing records in an llog.
3894 Details : If an llog record is corrupted/zeroed, it is possible to loop
3895 forever in llog_process(). Validate the llog record length
3896 and skip the remainder of the block on error.
3899 Frequency : occasional (liblustre only)
3901 Description: liblustre could not open files whose last component is a symlink
3902 Details : sysio_path_walk() would incorrectly pass the open intent to
3903 intermediate path components.
3906 Frequency : rare (liblustre only with non-standard tuning)
3907 Bugzilla : 7201 (7350)
3908 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3909 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3910 after it has been released, only do this for referenced locks
3912 Severity : enhancement
3914 Description: specifying an (invalid) directory default stripe_size of -1
3915 would reset the directory default striping
3916 Details : stripe_size -1 was used internally to signal directory stripe
3917 removal, now use "all default" to signal dir stripe removal
3918 as a directory striping of "all default" is not useful
3921 Frequency : common for large clusters running liblustre clients
3923 Description: doing an ls when liblustre clients are running is slow
3924 Details : sending a glimpse AST to a liblustre client waits for every AST
3925 to time out, as liblustre clients will not respond. Since they
3926 cannot cache data we refresh the OST lock LVB from disk instead.
3928 Severity : enhancement
3930 Description: doing an ls at the same time as file IO can be slow
3931 Details : enqueue and other "small" requests can be blocked behind many
3932 large IO requests. Create a new OST IO portal for non-IO
3933 requests so they can be processed faster.
3936 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3938 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3939 created file with mode 000
3940 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3941 allow an ftruncate() as a non-root user to a file with mode 000.
3942 Lustre now respects this flag to disable mode checks when
3943 truncating a file owned by the user
3946 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3949 Description: Revoking locks from clients that went dead or catatonic might take
3951 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3952 cancellation of such locks instant on servers without waiting for
3953 any reply from clients. Clients drops these locks when cancel
3954 notification from server is received without replying.
3957 Frequency : liblustre-only, when liblustre client dies or becomes busy
3959 Description: Doing ls on Linux clients can take a long time with active
3961 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3962 granting such locks to it in the first place if possible. Locks
3963 are taken by proxy on the OST during the read or write and
3964 dropped immediately afterward. Add connect flags handling, do
3965 not grant locks to liblustre clients for glimpse ASTs.
3967 Severity : enhancement
3969 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3970 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3971 sequential reads and behaves badly with applications doing random
3972 reads. Improve it by reading ahead at least read region, and
3973 avoiding excessive large RPC for small reads.
3975 Severity : enhancement
3977 Description: Creating more than 1000 files for a single job may cause a load
3978 imbalance on the OSTs if there are also a large number of OSTs.
3979 Details : qos_prep_create() uses an OST index reseed value that is an
3980 even multiple of the number of available OSTs so that if the
3981 reseed happens in the middle of the object allocation it will
3982 still utilize the OSTs as uniformly as possible.
3987 Description: OST or MDS may oops in ping_evictor_main()
3988 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3989 but doesn't restart at beginning of obd_exports_timed list
3992 Severity : enhancement
3994 Description: improve by-nid export eviction on the MDS and OST
3995 Details : allow multiple exports with the same NID to be evicted at one
3996 time without re-searching the exports list.
3999 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4001 Description: MDS may oops in groups_free()
4002 Details : in rare race conditions a newly allocated group_info struct is
4003 freed again, and this can be NULL. The 2.4 compatibility code
4004 for groups_free() checked for a NULL pointer, but 2.6 did not.
4007 Frequency : common for liblustre clients doing little filesystem IO
4008 Bugzilla : 9352, 7313
4009 Description: server may evict liblustre clients accessing contended locks
4010 Details : if a client is granted a lock or receives a completion AST
4011 with a blocking AST already set it would not reply to the AST
4012 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4015 Frequency : lfs setstripe, only systems with more than 160 OSTs
4017 Description: unable to set striping with a starting offset beyond OST 160
4018 Details : llapi_create_file() incorrectly limited the starting stripe
4019 index to the maximum single-file stripe count.
4022 Frequency : LDAP users only
4024 Description: lconf did not handle in-kernel recovery with LDAP properly
4025 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4027 Severity : enhancement
4029 Description: bind OST threads to NUMA nodes to improve performance
4030 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4031 node and do their allocations there to localize memory access
4033 Severity : enhancement
4035 Description: llmount can determine client NID directly from Myrinet (GM)
4036 Details : the client NID code from gmnalnid was moved directly into
4037 llmount, removing the need to use this or specifying the
4038 client NID explicitly when mounting GM clients with zeroconf
4041 Frequency : if client is started with down MDS
4043 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4044 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4045 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4046 ptlrpc_import_delayed_req() because we didn't check if the
4047 request was interrupted, and we also didn't break out of the
4048 event loop if there was no timeout
4053 Description: data loss during non-page-aligned writes to a single file from
4054 both multiple nodes and multiple threads on one node at same time
4055 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4056 inconsistency led to false short-reads, that were cached and later
4057 used by ->prepare_write() to fill in partially written page,
4058 leading to data loss.
4061 Frequency : always, if lconf --abort_recovery used
4063 Description: lconf --abort_recovery fails with 'Operation not supported'
4064 Details : lconf was attempting to abort recovery on the MDT device and not
4067 Severity : enhancement
4069 Description: remove cleanup logs
4070 Details : replace lconf-generated cleanup logs with lustre internal
4071 cleanup routines. Eliminates the need for client-cleanup and
4074 Severity : enhancement
4076 Description: add support for EAs (user and system) on lustre filesystems
4077 Details : it is now possible to store extended attributes in the Lustre
4078 client filesystem, and with the user_xattr mount option it
4079 is possible to allow users to store EAs on their files also
4081 Severity : enhancement
4083 Description: Add possibility (config option) to show minimal available OST free
4085 Details : When compiled with --enable-mindf configure option, statfs(2)
4086 (and so, df) will return least minimal free space available from
4087 all OSTs as amount of free space on FS, instead of summary of
4088 free spaces of all OSTs.
4090 Severity : enhancement
4092 Description: do not expand extent locks acquired on OST-side
4093 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4094 by server: they are not cached anyway.
4097 Frequency : when mmap is used/binaries executed from Lustre
4099 Description: Unmmap pages before throwing them away from read cache.
4100 Details : llap_shrink cache now attempts to unmap pages before discarding
4101 them (if unmapping failed - do not discard). SLES9 kernel has
4102 extra checks that trigger if this unmapping is not done first.
4107 Description: lconf didn't resolve symlinks before checking to see whether a
4108 given mountpoint was already in use
4111 Frequency : when migrating failover services
4112 Bugzilla : 6395, 9514
4113 Description: When migrating a subset of services from a node (e.g. failback
4114 from a failover service node) the remaining services would
4115 time out and evict clients.
4116 Details : lconf --force (implied by --failover) sets the global obd_timeout
4117 to 5 seconds in order to quickly disconnect, but this caused
4118 other RPCs to time out too quickly. Do not change the global
4119 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4121 Severity : enhancement
4122 Frequency : if MDS is started with down OST
4123 Bugzilla : 9439,5706
4124 Description: Allow startup/shutdown of an MDS without depending on the
4125 availability of the OSTs.
4126 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4127 Add appropriate locking and lov-osc refcounts for safe
4128 cleaning. Add osc abort_inflight calls in case the
4129 synchronize never started.
4132 Frequency : occasional (Cray XT3 only)
4134 Description: root not authorized to access files in CRAY_PORTALS environment
4135 Details : The client process capabilities were not honoured on the MDS in
4136 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4137 been accessed by an authorized user then root was able to access
4138 the file on the local client also. The root user capabilities
4139 are now allowed on the MDS, as this environment has secure UID.
4142 Frequency : occasional
4144 Description: ldiskfs "too long searching" message happens too often
4145 Details : A debugging message (otherwise harmless) prints too often on
4146 the OST console. This has been reduced to only happen when
4147 there are fragmentation problems on the filesystem.
4152 Description: Division by zero in statfs when all OSCs are inactive
4153 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4154 lov_statfs divides by value returned by lov_get_stripecnt().
4158 Bugzilla : 9489, 3273
4159 Description: First write from each client to each OST was only 4kB in size,
4160 to initialize client writeback cache, which caused sub-optimal
4161 RPCs and poor layout on disk for the first writen file.
4162 Details : Clients now request an initial cache grant at (re)connect time
4163 and so that they can start streaming writes to the cache right
4164 away and always do full-sized RPCs if there is enough data.
4165 If the OST is rebooted the client also re-establishes its grant
4166 so that client cached writes will be honoured under the grant.
4171 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4172 Details : Now I/O RPCs go to different portal number and (presumably) fast
4173 lock requests (and glimses) and other RPCs get their own service
4174 threads pool that should be able to service those RPCs
4177 Severity : enhancement
4179 Description: Ability to exchange lustre version between client and servers and
4180 issue warnings at client side if client is too old. Also for
4181 liblustre clients there is ability to refuse connection of too old
4183 Details : New 'version' field is added to connect data structure that is
4184 filled with version info. That info is later checked by server and
4188 Frequency : rare, liblustre only.
4189 Bugzilla : 9296, 9581
4190 Description: Two simultaneous writes from liblustre at offset within same page
4191 might proceed at the same time overwriting eachother with stale
4193 Details : I/O lock withing llu_file_prwv was released too early, before data
4194 actually was hitting the wire. Extended lock-holding time until
4195 server acknowledges receiving data.
4198 Frequency : extremely rare. Never observed in practice.
4200 Description: avoid generating lustre_handle cookie of 0.
4201 Details : class_handle_hash() generates handle cookies by incrementing
4202 global counter, and can hit 0 occasionaly (this is unlikely, but
4203 not impossible, because initial value of cookie counter is
4204 selected randonly). Value of 0 is used as a sentinel meaning
4205 "unassigned handle" --- avoid it. Also coalesce two critical
4206 sections in this function into one.
4208 Severity : enhancement
4210 Description: allow liblustre clients to delegate truncate locking to OST
4211 Details : To avoid overhead of locking, liblustre client instructs OST to
4212 take extent lock in ost_punch() on client's behalf. New connection
4213 flag is added to handle backward compatibility.
4215 Severity : enhancement
4216 Bugzilla : 4928, 7341, 9758
4217 Description: allow number of OST service threads to be specified
4218 Details : a module parameter allows the number of OST service threads
4219 to be specified via "options ost ost_num_threads={N}" in the
4220 OSS's /etc/modules.conf or /etc/modprobe.conf.
4224 Bugzilla : 6146, 9635, 9895
4225 Description: servers crash with bad pointer in target_handle_connect()
4226 Details : In rare cases when a client is reconnecting it was possible that
4227 the connection request was the last reference for that export.
4228 We would temporarily drop the export reference and get a new
4229 one, but this may have been the last reference and the export
4230 was just destroyed. Get new reference before dropping old one.
4232 Severity : enhancement
4233 Frequency : if client is started with failover MDS
4235 Description: Allow multiple MDS hostnames in the mount command
4236 Details : Try to read the configuration from all specified MDS
4237 hostnames during a client mount in case the "primary"
4240 Severity : enhancement
4242 Description: Stop sending data to evicted clients as soon as possible.
4243 Details : Check if the client we are about to send or are sending data to
4244 was evicted already. (Check is done every second of waiting,
4245 for which l_wait_event interface was extended to allow checking
4246 of exit condition at specified intervals).
4249 Frequency : rare, normally only when NFS exporting is done from client
4251 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4253 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4256 Frequency : always, if srand() is called before liblustre initialization
4258 Description: Liblustre uses system PRNG disturbing its usage by user application
4259 Details : Introduce internal to lustre fast and high-quality PRNG for
4260 lustre usage and make liblustre and some other places in generic
4261 lustre code to use it.
4263 Severity : enhancement
4264 Bugzilla : 9477, 9557, 9870
4265 Description: Verify that the MDS configuration logs are updated when xml is
4266 Details : Check if the .xml configuration logs are newer than the config
4267 logs stored on the MDS and report an error if this is the case.
4268 Request --write-conf, or allow starting with --old_conf.
4270 Severity : enhancement
4272 Description: Handle symlinks in the path when checking if Lustre is mounted.
4273 Details : Resolve intermediate symlinks when checking if a client has
4274 mounted a filesystem to avoid duplicate client mounts.
4279 Description: lconf can hit an error exception but still return success.
4280 Details : The lconf command catches the Command error exception at the top
4281 level script context and will exit with the associated exit
4282 status, but doesn't ensure that this exit status is non-zero.
4287 Description: failure of ptlrpc thread startup can cause oops
4288 Details : Starting a ptlrpc service thread can fail if there are a large
4289 number of threads or the server memory is very fragmented.
4290 Handle this without oopsing.
4293 Frequency : always, only if liblustre and non-default acceptor port was used
4295 Description: liblustre cannot connect to servers with non-default acceptor port
4296 Details : tcpnal_set_default_params() was not called and was therefore
4297 ignoring the environment varaible TCPNAL_PORT, as well as other
4298 TCPNAL_ environment variables
4303 Description: two objects could be created on the same OST for a single file
4304 Details : If an OST is down, in some cases it was possible to create two
4305 objects on a single OST for a single file. No problems other
4306 than potential performance impact and spurious error messages.
4310 Bugzilla : 5681, 9562
4311 Description: Client may oops in ll_unhash_aliases
4312 Details : Client dcache may become inconsistent in race condition.
4313 In some cases "getcwd" can fail if the current directory is
4319 Description: Inode refcounting problems in NFS export code
4320 Details : link_raw functions used to call d_instantiate without obtaining
4321 extra inode reference first.
4325 Bugzilla : 9942, 9903
4326 Description: Referencing freed requests leading to crash, memleaks with NFS.
4327 Details : We used to require that call to ll_revalidate_it was always
4328 followed by ll_lookup_it. Also with revalidate_special() it is
4329 possible to call ll_revalidate_it() twice for the same dentry
4330 even if first occurence returned success. This fix changes semantic
4331 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4332 reference on a request referred from the intent.
4333 ll_intent_release() then releases such a request.
4336 Frequency : rare, normally benchmark loads only
4338 Description: unlinked inodes were kept in memory on the client
4339 Details : If a client is repeatedly creating and unlinking files it
4340 can accumulate a lot of stale inodes in the inode slab cache.
4341 If there is no other client load running this can cause the
4342 client node to run out of memory. Instead flush old inodes
4343 from client cache that have the same inode number as a new inode.
4346 Frequency : SLES9 2.6.5 kernel and long filenames only
4347 Bugzilla : 9969, 10379
4348 Description: utime reports stale NFS file handle
4349 Details : SLES9 uses out-of-dentry names in some cases, which confused
4350 the lustre dentry revalidation. Change it to always use the
4354 Frequency : rare, unless heavy write-truncate concurrency is continuous
4355 Bugzilla : 4180, 6984, 7171, 9963, 9331
4356 Description: OST becomes very slow and/or deadlocked during object unlink
4357 Details : filter_destroy() was holding onto the parent directory lock
4358 while truncating+unlinking objects. For very large objects this
4359 may block other threads for a long time and slow overall OST
4360 responsiveness. It may also be possible to get a lock ordering
4361 deadlock in this case, or run out of journal credits because of
4362 the combined truncate+unlink. Solution is to do object truncate
4363 first in one transaction without parent lock, and then do the
4364 final unlink in a new transaction with the parent lock. This
4365 reduces the lock hold time dramatically.
4368 Frequency : rare, 2.4 kernels only
4370 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4371 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4372 before unmounting their filesystem in order to avoid deadlock.
4373 The kernel_locked() function in 2.4 kernels only checks whether
4374 the kernel lock is held, not whether it is this process that is
4375 holding it as 2.6 kernels do.
4380 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4381 Details : The client ptlrpc code may be trying to reconnect to a down
4382 server before a previous connection attempt has timed out.
4383 Increase the reconnect interval to be longer than the connection
4384 timeout interval to avoid sending duplicate connections to
4388 Frequency : echo_client brw_test command
4390 Description: fix echo_client to work with OST preallocated code
4391 Details : OST preallocation code (5137) didn't take echo_client IO path
4392 into account: echo_client calls filter methods outside of any
4393 OST thread and, hence, there is no per-thread preallocated
4394 pages and buffers to use. Solution: hijack pga pages for IO. As
4395 a byproduct, this avoids unnecessary data copying.
4399 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4400 Description: Client can oops in mdc_commit_close() after open replay
4401 Details : It was possible for the MDS to return an open request with no
4402 transaction number in mds_finish_transno() if the client was
4403 evicted, but without actually returning an error. Clients
4404 would later try to replay that open and may trip an assertion
4405 Simplify the client close codepath, and always return an error
4406 from the MDS in case the open is not successful.
4409 Frequency : rare, 2.6 OSTs only
4411 Description: OST may deadlock under high load on fragmented files
4412 Details : If there was a heavy load and highly-fragmented OST filesystems
4413 it was possible to have all the OST threads deadlock waiting on
4414 allocation of biovecs, because the biovecs were not released
4415 until the entire RPC IO was completed. Instead, release biovecs
4416 as soon as they are complete to ensure forward IO progress.
4418 Severity : enhancement
4420 Description: Support for specifying external journal device at mount
4421 Details : If an OST or MDS device is formatted with an external journal
4422 device, this device major/minor is stored in the ext3 superblock
4423 and may not be valid for failover. Allow detecting and
4424 specifying the external journal at mount time.
4429 Description: Mounting an MDS with pending unlinked files may cause oops
4430 Details : target_finish_recovery() calls mds_postrecov() which returned
4431 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4432 considers this an error and immediately begins cleaning up the
4433 lov, just after starting the mds_lov process
4435 Severity : enhancement
4437 Description: Implement 'lfs df' to report actual free space on per-OST basis
4438 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4439 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4440 usage of objects; '-h' to report in human readable format.
4442 ------------------------------------------------------------------------------
4444 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4451 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4452 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4453 at startup during orphan unlinking. Instead, skip buddy update
4454 before setup, as it will be regenerated after recovery is complete.
4459 Description: If an OST is inactive, its locks might reference stale inodes.
4460 Details : lov_change_cbdata() must iterate over all namespaces, even if
4461 they are inactive to clear inode references from the lock.
4463 Severity : enhancement
4464 Frequency : occasional, if non-standard max_dirty_mb used
4466 Description: Client will block write RPCs if not enough grant
4467 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4468 then the client will block writes while waiting for another RPC
4469 to complete instead of consuming its dirty limit. With change
4470 we get improved performance when max_dirty_mb is small.
4472 Severity : enhancement
4473 Bugzilla : 3389, 6253
4474 Description: Add support for supplementary groups on the MDS.
4475 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4476 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4477 lookups for user supplementary groups into a cache.
4481 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4482 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4483 for liblustre O_EXCL is enforced by MDS.
4487 Description: readdir never returns NULL in liblustre.
4488 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4489 dentry in struct dirent.
4493 Frequency : liblustre only, depends on application IO pattern
4494 Description: liblustre clients evicted if not contacting servers
4495 Details : Don't put liblustre clients into the ping_evictor list, so
4496 they will not be evicted by the pinger ever.
4498 Severity : enhancement
4500 Description: Add ability to evict clients by NID from MDS.
4501 Details : By echoing "nid:$NID" string into
4502 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4503 $NID would be instantly evicted from this MDS and from all active
4504 OSTs connected to it.
4508 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4509 Details : lookup_it_finish() used to query file size from OSTs that was not
4514 Description: service threads change working directory to that of init
4515 Details : Starting lustre service threads may pin the working directory
4516 of the parent thread, making that filesystem busy. Threads
4517 now change to the working directory of init to avoid this.
4521 Frequency : during shutdown only
4522 Description: shutdown with a failed MDS or OST can cause unmount to hang
4523 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4528 Frequency : 2.6 only
4529 Description: chmod/chown may include an extra supplementary group
4530 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4531 supplementary group and if none is specified this is used.
4534 Bugzilla : 5479 (6816)
4536 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4537 Details : If lookup is in progress on a file that is unlinked we might try
4538 to revalidate the inode and fail in revalidate after lookup is
4539 complete and ll_file_open() enqueues the open again but
4540 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4543 Frequency : always, if lconf --abort_recovery used
4545 Description: lconf --abort_recovery fails with 'Operation not supported'
4546 Details : lconf was attempting to abort recovery on the MDT device and not
4549 ------------------------------------------------------------------------------
4551 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4556 Frequency : rare (only unsupported configurations with a node running as an
4558 Bugzilla : 6514, 5137
4559 Description: Mounting a Lustre file system on a node running as an OST could
4561 Details : OSTs now preallocates memory needed to write out data at
4562 startup, instead of when needed, to avoid having to
4563 allocate memory in possibly low memory situations.
4564 Specifically, if the file system is mounted on on OST,
4565 memory pressure could force it to try to write out data,
4566 which it needed to allocate memory to do. Due to the low
4567 memory, it would be unable to do so and the node would
4568 become unresponsive.
4570 Severity : enhancement
4572 Description: Addition of lconf --service command line option
4573 Details : lconf now accepts a '--service <arg>' option, which is
4574 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4576 Severity : enhancement
4578 Description: Failover mode is now the default for OSTs.
4579 Details : By default, OSTs will now run in failover mode. To return to
4580 the old behaviour, add '--failout' to the lmc line for OSTs.
4582 Severity : enhancement
4584 Description: Health checks are now provided for MDS and OSTs
4585 Details : Additional detailed health check information on MSD and OSTs
4586 is now provided through the procfs health_check value.
4589 Frequency : occasional, depends on IO load
4591 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4592 after numerous create/delete cycles
4593 Details : The ext3 inode allocation policy would not allocate new inodes
4594 very well on the OSTs because there are no new directories
4595 being created. Instead we look for groups with free space if
4596 the parent directories are nearly full.
4601 Description: Network or server problems during mount may cause partially
4602 mounted clients instead of returning an error.
4603 Details : The config llog parsing code may overwrite the error return
4604 code during mount error handling, returning success instead
4610 Description: MDS can fail to allocate large reply buffers
4611 Details : After long uptimes the MDS can fail to allocate large reply
4612 buffers (e.g. zconf client mount config records) due to memory
4613 fragmentation or consumption by the buffer cache. Preallocate
4614 some large reply buffers so that these replies can be sent even
4615 under memory pressure.
4619 Frequency : rare (liblustre)
4620 Description: fsx running with liblustre complained that using truncate() to
4621 extend the file doesn't work. This patch corrects that issue.
4622 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4623 ATTR_SIZE and lsm use in llu_setattr_raw.
4627 Frequency : rare, only 2.6 kernels
4628 Description: Unusual file access patterns on the MDS may result in inode
4629 data being lost in very rare circumstances.
4630 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4631 code in the 2.6 kernel caused the inode and/or EA data not to
4632 be read from disk, causing single-file corruption.
4636 Frequency : rare, only 2.6 filesystems using extents
4637 Description: Heavy concurrent write and delete load may cause data corruption.
4638 Details : It was possible under high-load situations to have an extent
4639 metadata block in the block device cache from a just-unlinked
4640 file overwrite a newly-allocated data block. We now unmap any
4641 metadata buffers that alias just-allocated data blocks.
4645 Frequency : filesystems with default stripe_count larger than 77
4646 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4647 Details : lconf specifies an inode size of 4096 bytes when the default
4648 stripe_count is larger than 77. This conflicts with the default
4649 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4650 this case to avoid pinning too much memory for large EAs.
4652 ------------------------------------------------------------------------------
4654 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4659 Frequency : rare (extremely heavy IO load with hundreds of clients)
4661 Description: Client is evicted, gets IO error writing to file
4662 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4663 caused clients to be evicted for AST timeouts. The fixes in
4664 bug 5192 mean we no longer need to have such short AST timeouts
4665 so ldlm_timeout has been increased.
4668 Frequency : occasional during --force or --failover shutdown under load
4669 Bugzilla : 5949, 4834
4670 Description: Server oops/LBUG if stopped with --force or --failover under load
4671 Details : a collection of import/export refcount and cleanup ordering
4672 issues fixed for safer force cleanup
4675 Frequency : only filesystems larger than 120 OSTs
4676 Bugzilla : 5990, 6223
4677 Description: lfs getstripe would oops on a very large filesystem
4678 Details : lov_getconfig used kfree on vmalloc'd memory
4681 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4682 Bugzilla : 6242, 6243
4683 Description: reading from files that had been truncated to a non-zero size
4684 but never opened returned no data
4685 Details : ll_file_read() reads zeros from no-object files to EOF
4690 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4691 Details : There's little harm in aborting MDS/OSS recovery and letting it
4692 try again, so I removed the LASSERT and return an error instead.
4694 Severity : enhancement
4696 Description: New debugging infrastructure for tracking down data corruption
4697 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4698 (b) cover more of the client-side code path, and (c) try to narrow
4699 down where problems occurred
4703 Bugzilla : 3819, 4364, 4397, 6313
4704 Description: Racing close and eviction MDS could cause assertion in mds_close
4705 Details : It was possible to get multiple mfd references during close and
4706 client eviction, leading to one thread referencing a freed mfd.
4708 Severity: : enhancement
4709 Bugzilla : 3262, 6359
4710 Description: Attempts to reconnect to servers are now more aggressive.
4711 Details : This builds on the enhanced upcall-less recovery that was added
4712 in 1.4.2. When trying to reconnect to servers, clients will
4713 now try each server in the failover group every 10 seconds. By
4714 default, clients would previously try one server every 25 seconds.
4719 Description: After recovery, certain operations trigger a failed
4720 assertion on a client.
4721 Details : Failing over an mds, using lconf -d --failover, while a
4722 client was doing a readdir() call would cause the client to
4723 LBUG after recovery completed and the readdir() was resent.
4725 Severity : enhancement
4727 Description: Default groups are now added by lconf
4728 Details : You can now run lconf --group <servicename> without having to
4729 manually add groups with lmc.
4732 Frequency : occasional
4734 Description: Nodes with an elan id of 0 trigger a failed assertion
4737 Frequency : always when accessing e.g. tty/console device nodes
4739 Description: tty and some other devices nodes cannot be used on lustre
4740 Details : file's private_data field is used by device data and lustre
4741 values in there got lost. New field was added to struct file to
4742 store fs-specific private data.
4745 Frequency : when exporting Lustre via NFS
4747 Description: NFSD failed occasionally when looking up a path component
4748 Details : NFSD is looking up ".." which was broken in ext3 directories
4749 that had grown large enough to become hashed.
4752 Frequency : Clusters with multiple interfaces not on the same subnet
4754 Description: Nodes will repeatedly try to reconnect to an interface which it
4755 cannot reach and report an error to the log.
4756 Details : Extra peer list entries will be created by lconf with some peers
4757 unreachable. lconf now validates the peer before adding it.
4760 Frequency : Only if a default stripe is set on the filesystem root.
4762 Description: Setting a default stripe on the filesystem root prevented the
4763 filesystem from being remounted.
4764 Details : The client was sending extra request flags in the root getattr
4765 request and did not allocate a reply buffer for the dir EA.
4768 Frequency : occasional, higher if lots of files are accessed by one client
4769 Bugzilla : 6159, 6097
4770 Description: Client trips assertion regarding lsm mismatch/magic
4771 Details : While revalidating inodes the VFS looks up inodes with ifind()
4772 and in rare cases can find an inode that is being freed.
4773 The ll_test_inode() code will free the lsm during ifind()
4774 when it finds an existing inode and then the VFS later attaches
4775 this free lsm to a new inode.
4779 Bugzilla : 6422, 7030
4780 Description: MDS deadlock between mkdir and client eviction
4781 Details : Creating a new file via mkdir or mknod (starting a transaction
4782 and getting the ns lock) can deadlock with client eviction
4783 (gets ns lock and trying to finish a synchronous transaction).
4786 Frequency : occasional
4787 Description: While starting a server, the fsfilt_ext3 module could not be
4789 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4790 kernels. Previously, lconf would still use the ext3 name
4791 when trying to load modules. Now, it will correctly use
4792 ext3 on 2.4 and ldiskfs on 2.6.
4794 Severity : enhancement
4795 Description: The default stripe count has been changed to 1
4796 Details : The interpretation of the default stripe count (0, to lfs
4797 or lmc) has been changed to mean striping across a single
4798 OST, rather than all available. For general usage we have
4799 found a stripe count of 1 or 2 works best.
4801 Severity : enhancement
4802 Description: Add support for compiling against Cray portals.
4803 Details : Conditional compiling for some areas that are different
4807 Frequency : occasional
4808 Bugzilla : 6409, 6834
4809 Description: Creating files with an explicit stripe count may lead to
4810 a failed assertion on the MDS
4811 Details : If some OSTs are full or unavailable, creating files may
4812 trigger a failed assertion on the MDS. Now, Lustre will
4813 try to use other servers or return an error to the
4817 Frequency : occasional
4819 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4820 caused lock timeout during readahead (since 1.4.2).
4821 Details : Processes doing readahead might match a lock that hasn't been
4822 granted yet if there are overlapping and conflicting lock
4823 requests. The readahead process waits on ungranted lock
4824 (original lock is CBPENDING), while OST waits for that process
4825 to cancel CBPENDING read lock and eventually evicts client.
4827 Severity : enhancement
4829 Description: Initial enabling of flock support for clients
4830 Details : Implements fcntl advisory locking and file status functions.
4831 This feature is provided as an optional mount flag (default
4832 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4833 locking are implemented yet, and those that are are not guaranteed
4834 to be completely correct in production environments.
4835 mount -t lustre -o [flock|noflock] ...
4838 Frequency : occasional
4840 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4841 assertion in the ext3 JBD (journaling) layer.
4842 Details : The b_committed_data struct is protected by the big kernel lock
4843 in 2.4 kernels, serializing journal_commit_transaction() and
4844 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4845 access to this struct. In 2.6 kernels there is finer grained
4846 locking to improve SMP performance of the JBD layer.
4850 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4855 Description: merge and cleanup kernel patches.
4858 -----------------------------------------------------------------------------
4860 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4862 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4863 versions of Lustre. Please update all servers and clients to
4864 version 1.4.2 or later at the same time. You must also run
4865 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4866 to update the configuration logs.
4868 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4869 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4870 - swab llog records of type '0' so we get proper header size/idx (5861)
4871 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4872 - fix rename of one directory over another leaking an inode (5953)
4873 - avoid SetPageDirty on 2.6 (5981)
4874 - don't re-add just-being-destroyed locks to the waiting list (5653)
4875 - when creating new directories, inherit the parent's custom
4876 striping settings if present parent (3048)
4877 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4878 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4879 - don't hold client locks on temporary worklist from l_lru (5666)
4880 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4881 - automatically evict dead clients (5921)
4882 - Update file size properly in create+truncate+fstat case (6196)
4883 - Do not unhash mountpoint dentries, do not allow removal of
4885 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4886 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4887 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4888 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4889 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4890 - Correctly handle reads of files with no objects (6243)
4891 - lctl recover will also mark a device active if deactivate used (5933)
4893 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4894 - allow --write-conf on an MDS with different nettype than client (5619)
4895 - don't write config llogs to MDS for mounts not from that MDS (5617)
4896 - lconf should create multiple TCP connections from a client (5201)
4897 - init scripts are now turned off by default; run chkconfig --on
4898 lustre and chkconfig --on lustrefs to use them
4899 - upcalls are no longer needed for clients to recover to failover
4901 - add --abort-recovery option to lconf to abort recovery on device
4903 - add support for an arbitrary number of OSTs (3026)
4904 - Quota support protocol changes.
4905 - forward compatibility changes to wire structs (6007)
4906 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4907 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4908 - improved "lustre status" script
4909 - initialize blocksize for non-regular files (6062)
4910 - added --disable-server and --disable-client configure options (5782)
4911 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4912 - Vanilla 2.4.29 support
4913 - increase maximum number of obd devices to 520 (6242)
4914 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4915 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4916 SLES 9 2.6.5 kernels
4917 - limit stripes per file to 160 (the maximum EA size) (6093)
4919 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4922 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4923 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4924 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4925 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4926 - small ext3 extents cleanups and fixes (5733)
4927 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4928 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4929 - avoid needless client->OST connect, fix handle mismatch (5317)
4930 - fix DLM error path that led to out-of-sync client, long delays (5779)
4931 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4932 - fix several locking issues related to i_size (5492,5624,5654,5672)
4933 - don't move pending lock onto export if it is already evicted (5683)
4934 - fix kernel oops when creating .foo in unlinked directory (5548)
4935 - fix deadlock in obdfilter statistics vs. object create (5811)
4936 - use time_{before,after} to avoid timer jiffies wrap (5882)
4937 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4938 - Do not leak request if server was not able to process it (5154)
4939 - If mds_open unable to find parent dir, make that negative lookup(5154)
4940 - don't create new directories with extent-mapping (5909, 5936)
4942 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4943 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4944 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4945 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4946 - support for RHEL4 kernel on i686 (5773)
4947 - provide error messages when incompatible logs are encountered (5898)
4949 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4950 * version 1.4.0.10 (1.4.1 release candidate 1)
4952 - don't keep a lock reference when lock is not granted (4238)
4953 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4954 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4955 - drop import inflight refcount on signal_completed_replay error (5255)
4956 - unlock page after async write error during send (3677)
4957 - handle missing objects in filter_preprw_read properly (5265)
4958 - no transno return for symlink open, don't save no-trasno open (3440)
4959 - don't try to complete elan receive that already failed (4012)
4960 - free RPC server reply state on error (5406)
4961 - clean up thread from ptlrpc_start_thread() on error (5160)
4962 - readahead could read extra page into cache that wasn't ejected (5388)
4963 - prevent races in class_attach/setup/cleanup/detach (5260)
4964 - don't dereference de->d_inode after l_dput of de (5458)
4965 - use "int" for stripe value returned from lock_to_stripe (5544)
4966 - mballoc allocation and error-checking fixes in 2.6 (5504)
4967 - block device patches to fix I/O request sizes in 2.6 (5482)
4968 - look up hostnames for IB nals (5602)
4969 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4970 - don't start multiple acceptors for the same port (5277)
4971 - fix incorrect LASSERT in mds_getattr_name (5635)
4972 - export a proc file for general "ping" checking (5628)
4973 - fix "lfs check" to not block when the MDS is down (5628)
4975 - service request history (4965)
4976 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4977 - create an "evict_client" /proc entry on OSTs, like the MDS has
4978 - fix mount usage message, return errors per mount(8) (5168)
4979 - change grep [] to grep "[]" in tests so they work in more UMLs
4980 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4981 - remove ancient LOV_MAGIC_V0 EA support (5047)
4982 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4983 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4984 - allow readahead allocations to fail when low on memory (5383)
4985 - mmap locking landed again, after considerable improvement (2828)
4986 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4988 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4991 - send OST transaction number in read/write reply to free req (4966)
4992 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4993 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4994 - the watchdog thread now runs as interruptible (5246)
4995 - flock/lockf fixes (but it's still disabled, pending 5135)
4996 - don't use EXT3 constants in llite code (5094)
4997 - memory shortage at startup could cause assertion (5176)
4999 - reorganization of lov code
5000 - single portals codebase
5002 - add extents/mballoc support (5025)
5003 - direct I/O reads in the obdfilter (4048)
5004 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5006 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5009 - send OST transaction number in read/write reply to free req (4966)
5010 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5011 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5012 - the watchdog thread now runs as interruptible (5246)
5013 - handle missing objects in filter_preprw_read properly (5265)
5014 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5015 - drop import inflight refcount on signal_completed_replay error (5255)
5016 - unlock page after async write error during send (3677)
5017 - return original error code on reconstructed replies (3761)
5018 - no transno return for symlink open, don't save no-trasno open (3440)
5020 - add pid to ldlm debugging output (4922)
5021 - bump the watchdog timeouts -- we can't handle 30sec yet
5022 - extra debugging for orphan dentry/inode bug (5259)
5024 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5027 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5028 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5029 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5030 - fix NULL dereference in /proc/sys/portals/routes (4827)
5031 - allow failed mdc_close() operations to be interrupted (4561)
5032 - stop precreate on OST before MDS would time out on it (4778)
5033 - don't send partial-page writes before EOF from client (4410)
5034 - discard client grant for sub-page writes on large-page clients (4520)
5035 - don't free dentries not owned by NFS code, check generation (4806)
5036 - fix lsm leak if mds_create_objects() fails (4801)
5037 - limit debug_daemon file size, always print CERROR messages (4789)
5038 - use transno after validating reply (3892)
5039 - process timed out requests if import state changes (3754)
5040 - update mtime on OST during writes, return in glimpse (4829)
5041 - add mkfsoptions to LDAP (4679)
5042 - use ->max_readahead method instead of zapping global ra (5039)
5043 - don't interrupt __l_wait_event() during strace
5045 - add software watchdogs to catch hung threads quickly (4941)
5046 - make lustrefs init script start after nfs is mounted
5047 - fix CWARN/ERROR duplication (4930)
5048 - return async write errors to application if possible (2248)
5049 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5050 - print NAL number in %x format (4645)
5051 - update barely-supported suse-2.4.21-171 series (4842)
5052 - support for sles 9 %post scripts
5053 - support for building 2.6 kernel-source packages
5054 - support for sles km_* packages
5056 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5059 - ignore -ENOENT errors in osc_destroy (3639)
5060 - notify osc create thread that OSC is being cleaned up (4600)
5061 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5062 - reconstruct ost_handle() like mds_handle() (4657)
5063 - create a new thread to do import eviction to avoid deadlock (3969)
5064 - let lconf resolve symlinked-to devices (4629)
5065 - don't unlink "objects" from directory with default EA (4554)
5066 - hold socknal file ref over connect in case target is down (4394)
5067 - allow more than 32000 subdirectories in a single directory (3244)
5068 - fix blocks count for O_DIRECT writes (3751)
5069 - OST returns ENOSPC from object create when no space left (4539)
5070 - don't send truncate RPC if file size isn't changing (4410)
5071 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5072 - bind to privileged port in socknal and tcpnal (3689)
5074 - rate limit CERROR/CWARN console message to avoid overload (4519)
5075 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5076 - basic mmap support (3918)
5077 - kernel patch series update from b1_4 (4711)
5079 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5082 - avoid crash during MDS cleanup with OST shut down (2775)
5083 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5084 - don't use bad inodes on the MDS (3744)
5085 - dynamic object preallocation to improve recovery speed (4236)
5086 - don't hold spinlock over lock dumping or change debug flags (4401)
5087 - don't zero obd_dev when it is force cleaned (3651)
5088 - print grants to console if they go negative (4431)
5089 - "lctl deactivate" will stop automatic recovery attempts (3406)
5090 - look for existing locks in ldlm_handle_enqueue() (3764)
5091 - don't resolve lock handle twice in recovery avoiding race (4401)
5092 - revalidate should check working dir is a directory (4134)
5094 - don't always mark "slow" obdfilter messages as errors (4418)
5096 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5099 - don't close LustreDB during write_conf until it is done (3860)
5100 - fix typo in lconf for_each_profile (3821)
5101 - allow dumping logs from multiple threads at one time (3820)
5102 - don't allow multiple threads in OSC recovery (3812)
5103 - fix debug_size parameters (3864)
5104 - fix mds_postrecov to initialize import for llog ctxt (3121)
5105 - replace config semaphore with spinlock (3306)
5106 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5107 - don't allow enqueue to complete on a destroyed export (3822)
5108 - down write_lock before checking llog header bitmap (3825)
5109 - recover from lock replay timeout (3764)
5110 - up llog sem before sending rpc (3652)
5111 - reduce ns lock hold times when setting kms (3267)
5112 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5113 - fix NULL deref and obd_dev leak on setup error (3312)
5114 - replace some LBUG about llog ops with error handling (3841)
5115 - don't match INVALID dentries from d_lookup and spin (3784)
5116 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5117 - fix invalid assertion in ptlrpc_set_wait (3880)
5119 - add libwrap support for the TCP acceptor (3996)
5120 - add /proc/sys/portals/routes for non-root route listing (3994)
5121 - allow setting MDS UUID in .xml (2580)
5122 - print the stack of a process that LBUGs (4228)
5124 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5127 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5128 - make sure to unset replay flag from failed open requests (3440)
5129 - if default stripe count is 0, use OST count for inode size (3636)
5130 - update parent mtime/ctime on client for create/unlink (2611)
5131 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5132 - free recovery state on server during a forced cleanup (3571)
5133 - unregister_reply for resent reqs (3063)
5134 - loop back devices mounting and status check on 2.6 (3563)
5135 - fix resource-creation race that can provoke i_size == 0 (3513)
5136 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5137 - more debugging for page-accounting assertion (3746)
5138 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5139 - avoid deadlock after precreation failure (3758)
5140 - fix race and lock order deadlock in orphan handling (3450, 3750)
5141 - add validity checks when grabbing inodes from l_ast_data (3599)
5143 - add /proc/.../recovery_status to obdfilter (3428)
5144 - lightweight CDEBUG infrastructure, debug daemon (3668)
5145 - change default OSC RPC parameters to be better on small clusters
5146 - turn off OST read cache for files smaller than 32MB
5147 - install man pages and include them in rpms (3100)
5148 - add new init script for (un)mounting lustre filesystems (2593)
5149 - run chkconfig in %post for init scripts (3701)
5150 - drop scimac NAL (unmaintained)
5152 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5155 - clean kiobufs before and after use (3485)
5156 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5157 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5158 - change init script to fail more clearly if not run as root (1528)
5159 - allow clients to reconnect during replay (1742)
5160 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5161 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5162 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5163 - fix cond_resched (3554)
5164 - only evict unfinished clients after recovery (3515)
5165 - allow bulk resend, prevent data loss (3570)
5166 - dynamic ptlrpc request buffer allocation (2102)
5167 - don't allow unlinking open directory if it isn't empty (2904)
5168 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5169 - remove extraneous obd dereference causing LASSERT failure (3334)
5170 - don't use get_cycles() when creating temp. files on the mds (3156)
5171 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5172 - handle EEXIST for set-stripe, set proper directory name (3336)
5174 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5175 - fix ksocknal_fmb_callback() error messages (2918)
5177 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5180 - don't copy lvb into (possibly NULL) reply on error (2983)
5181 - don't deref dentry after dput, don't free lvb on error (2922)
5182 - use the kms to determine writeback rpc length (2947)
5183 - increment oti_logcookies when osc is inactive (2948)
5184 - update client's i_blocks count via lvb messages (2543)
5185 - handle intent open/close of special files properly (1557)
5186 - mount MDS with errors=remount-ro, like obdfilter (2009)
5187 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5188 - don't use cancelling-locks' kms values (2947)
5189 - use highest lock extent for kms, not last one (2925)
5190 - don't dereference ERR_PTR() dentry in error handling path (3107)
5191 - fix thread race in portals_debug_dumplog() (3122)
5192 - create lprocfs device entries at setup instead of at attach (1519)
5193 - common AST error handler, don't evict client on completion race (3145)
5194 - zero nameidata in detach_mnt in 2.6 (3118)
5195 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5196 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5197 - initialize RPC timeout timer earlier for 2.6 (3219)
5198 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5199 - print nal/nid for unknown nid (3258)
5200 - additional checks for oscc recovery before doing precreate (3284)
5201 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5202 - don't crash in mdc_close for bad permissions on open (3285)
5203 - zero i_rdev for non-device files (3147)
5204 - clear page->private before handing to FS, better assertion (3119)
5205 - tune the read pipeline (3236)
5206 - fix incorrect decref of invalidated dentry (2350)
5207 - provide read-ahead stats and refine rpc in flight stats (3328)
5208 - don't hold journal transaction open across create RPC (3313)
5209 - update atime on MDS at close time (3265)
5210 - close LDAP connection when recovering to avoid server load (3315)
5211 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5212 - don't leak open file on MDS after open resend (3325)
5213 - serialize filter_precreate and filter_destroy_precreated (3329)
5214 - loop device shouldn't call sync_dev() for nul device (3092)
5215 - clear page cache after eviction (2766)
5216 - resynchronize MDS->OST in background (2824)
5217 - refuse to mount the same filesystem twice on same mountpoint (3394)
5218 - allow llmount to create routes for mounting behind routers (3320)
5219 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5220 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5221 - fix rare problem with rename on htree directories (3417)
5223 - allow default OST striping configuration per directory (1414)
5224 - fix compilation for qswnal for 2.6 kernels (3125)
5225 - increase maximum number of MDS request buffers for large systems
5226 - change liblustreapi to be useful for external progs like lfsck (3098)
5227 - increase local configuration timeout for slow disks (3353)
5228 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5230 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5233 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5234 - don't overwrite extent policy data in reply if lock was blocked (2901)
5235 - drop filter export grants atomically with removal from device (2663)
5236 - del obd_self_export from work_list in class_disconnect_exports (2908)
5237 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5238 - swab reply message in mdc_close, other PPC fixes (2464)
5239 - fix destroying of named logs (2325)
5240 - overwrite old logs when running lconf --write_conf (2264)
5241 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5242 - fix race in target_handle_connect (2898)
5243 - mds_reint_create() should take same inode create lock (2926)
5244 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5245 - don't close files for self_export to avoid uninitialized obd (2936)
5246 - allow MDS with the same name as client node (2939)
5247 - hold dentry reference for closed log files for unlink (2325)
5248 - reserve space for all logs during transactions (2059)
5249 - don't evict page beyond end of stripe extent (2925)
5250 - don't oops on a deleted current working directory (2399)
5251 - handle hard links to targets without a parent properly (2517)
5252 - don't dereference NULL lock when racing during eviction (2867)
5253 - don't grow lock extents when lots of conflicting locks (2919)
5255 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5258 - account for cache space usage on clients to avoid data loss (974)
5259 - lfsck support in lustre kernel code (2349)
5260 - reduce journal credits needed for BRW writes (2370)
5261 - orphan handling to avoid losing space on client/server crashes
5262 - ptlrpcd can be blocked, stopping ALL progress (2477)
5263 - use lock value blocks to assist in proper KMS, faster stat (1021)
5264 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5265 - recovery for initial connections (2355)
5266 - fixes for mds_cleanup_orphans (1934)
5267 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5268 - block all file creations until orphan recovery completes (1901)
5269 - client remove rq_connection from request struct (2423)
5270 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5271 - recovery timer race (2670)
5272 - mdc_close recovey bug (2532)
5273 - ptlrpc cleanup bug (2710)
5274 - mds timeout on local locks (2588)
5275 - namespace lock held during RPCs (2431)
5276 - handle interrupted sync write properly (2503)
5277 - don't try to handle a message that hasn't been replied to (2699)
5278 - client assert failure during cleanup after abort recovery (2701)
5279 - leak mdc device after failed mount (2712)
5280 - ptlrpc_check_set allows timedout requests to complete (2714)
5281 - wait for inflight reqs when ptlrpcd finishes (2710)
5282 - make sure unregistered services are removed from the srv_list
5283 - reset bulk XID's when resending them (caught by 1138 test)
5284 - unregister_bulk after timeout
5285 - fix lconf error (2694)
5286 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5287 - readahead locks pages, leaves pending causing memory pressure (2673)
5288 - increase OST request buffers to 4096 on large machines (2729)
5289 - fix up permission of existing directories in simple_mkdir (2661)
5290 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5291 - don't assign transno to errored transactions (2742)
5292 - don't delete objects on OST if given a bogus objid from MDS (2751)
5293 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5294 - if rq_no_resend, then timeout request after recovery (2432)
5295 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5296 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5297 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5298 - MDS recovery shouldn't delete objects that it has given out (2730)
5299 - if enqueue arrives after completion, don't clobber LVB (2819)
5300 - don't unlock pages twice when trigger_group_io returns error (2814)
5301 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5302 - don't write pages to disk if there was an error (1450)
5303 - don't ping imports that have recovery disabled (2676)
5304 - take buffered bytes into account when balancing socknal conn (2817)
5305 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5306 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5307 - remove little-endian swabbing of llog records (1987)
5308 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5309 - retry reposting request buffers if they fail (1191)
5310 - grow extent at grant time to avoid granting a revoked lock (2809)
5311 - lock revoke doesn't evict page if covered by a second lock (2765)
5312 - disable VM readahead to avoid reading outside lock extents (2805)
5314 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5315 - updated kernel patches for hp-2.4.20 kernel (2681)
5317 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5320 - fix truncated write corruption (2366)
5321 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5323 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5324 - ensure len(uuid) < 37 in lmc (1171)
5325 - fix ia64 OOPS in llog_test (2255)
5326 - zero end of page at obdfilter for partial page writes (2648)
5327 - don't leave stale dentries around after renames (bug 2428)
5328 - fix timeouts when evicting a client with a single lock held (2642)
5329 - set deadline for the initial HELLO message to drain (2634)
5330 - print out dotted-quad IP addresses in the socknal (2302)
5332 - additional debugging for MDS client eviction problem (2443)
5333 - fix mkfsoptions support for osts (2603, 2604)
5335 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5338 - add series for the vanilla 2.6.0 kernel
5339 - add series for the vanilla 2.4.24 kernel
5340 - add series for a cray x86/64 UL kernel drop
5341 - fix xattr patches for the vanilla 2.4.19 series
5343 - generate true UUIDs in lmc (1171)
5344 - have portals stack dumping break in UML (2466)
5345 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5346 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5347 - wake callback waiting threads on client eviction (2460)
5348 - Add --ptldebug and --subsystem to lmc (1719)
5349 - update assertion to allow safe interrupt allocation
5350 - set rq_no_resend for cancel requests (2432)
5351 - recalculate ptlrpcd timeout after resend (2494)
5352 - call vfs_rmdir when removing pending directories (2368)
5353 - fix renaming a file to itself (2429)
5354 - lmc creates a default one-stripe lov (2454)
5355 - expand procfs space to handle large clusters (2326)
5356 - increase UML stack to avoid overflow
5357 - update lconf's list of debug and subsystem masks
5358 - fix lfs find --obd (2510)
5359 - /proc tunable for disabling filter read caching (2591)
5360 - stop rpm packages from altering slapd.conf (2301)
5361 - disable nagle in the socknal under 0conf (2578)
5362 - choose mds inode size based on stripe count (2572)
5363 - fix kernel-source rpm problems (2516)
5365 - add --disable-doc to avoid pdf generation (2421)
5366 - update documentation, tests, type-os, comments
5367 - avoid format warnings on ia64
5368 - remove the TOE NAL
5369 - tiny code cleanups by removing unused fields
5371 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5374 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5375 - remove the most blatant lies from BUILDING file (2371)
5376 - change default debug level to reasonable production setting
5377 - reduce client side cache size to reduce cache flush time
5378 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5379 - make TCP zerocopy and pinger support enabled by default (2476)
5380 - sync writes completed after process exits caused crashes (2319)
5381 - maintain correct mount count on the MDS (2356)
5382 - backout 1557, because 2316 wasn't really fixed
5383 - better file I/O statistics gathering in /proc
5384 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5385 - another kernel patch to fix zero-copy TCP function export
5386 - don't take duplicate lock when processing re-sent getattr (2420)
5387 - lctl uses obd_self_export instead of creating new conn (2353)
5388 - MDS/OST recovery case which requires object creation asserted (2425)
5389 - move lfs from /usr/sbin to /usr/bin in packages
5390 - fix race between mds_client_add and mds_client_free (2417)
5391 - use kmalloc instead of slabs in portals (2430)
5392 - don't create duplicate records when a failover MDS is present (2442)
5393 - remove unnecessary mount age check (2332)
5394 - don't remove directory inodes from locks prematurely (2451)
5395 - don't break if MDS service name is the same as hostname (2103)
5396 - fix races in client write RPC generation when cache full (2482)
5398 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5401 - remove now-unused request->rq_obd (278)
5402 - if an allocation fails, print out how much memory we've used (1933)
5403 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5404 - add the "configurable stack size" patch to most series files (1256)
5405 - ability to write large log records, for 100+ OST configs (2306)
5406 - fix NULL deref when filter_prep fails (2314)
5407 - fix operator precedence error in filter_sync
5408 - dynamic allocation of socknal TX descriptors (2315)
5409 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5410 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5411 - ensure that CWARN messages in Portals always get to the syslog
5412 - __init/__exit are not for prototype decls (ldlm_init/exit)
5413 - x86-64 compile warning fixes
5414 - fix gateway LMC keyword conflict (2318)
5415 - fix MDS lock inversions in getattr/reint paths (1844)
5416 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5417 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5418 - DLM race condition prevented some lock evictions (2328)
5419 - ENOMEM detection and retry on socknal sends (2230)
5420 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5421 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5422 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5423 - better error messages when a client is rejected during recovery (1505)
5424 - avoid cancelling locks which were never granted, after failure (2330)
5425 - fix i_sem/journal inversion in mds_client_add (2333)
5426 - fix truncate/getattr lock cycle deadlock (2334)
5427 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5428 - fix two rare exit paths which could leak an l_lock() ref (2321)
5429 - fencepost error in MDS/OST orphan recovery (2226)
5430 - make log record alignment 8 bytes (1988)
5431 - lstripe now fails when requested offset > ost_count (2237)
5432 - ensure that all kernel series have a complete list.h (1607)
5433 - fix crashes in special-file operations (2316)
5434 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5436 - allow configurable automake binary, for testing new versions
5437 - small update to the lfs documentation
5439 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5441 * fix negative export reference count in fsfilt_sync (2312)
5443 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5444 * release candidate 0.9.1
5446 - orphans are moved into the PENDING directory for possible recovery
5447 - replayed opens now open by fid for orphan/rename safety (1042)
5448 - last close of an orphan inode generates a transno (683)
5449 - chdir() and mount() now pin the directory entry (1020)
5450 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5451 - discard very old requests without processing them (1502)
5452 - remove some common, well-understood CERRORs (1505)
5453 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5454 - clear "grant" flags in OST replies until OST grant code lands (1644)
5455 - fix read performance by not clobbering i_blksize on client (1598)
5456 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5457 - make LRU size a /proc tunable, clears locks when reduced (707)
5458 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5459 - server locks take references on exports now (1558)
5460 - build fixes for 2.4.20-rh trees (1663)
5461 - return an error from lov_create if all OSCs are inactive (1751)
5462 - fix import levels when a reconnect happens without a timeout (1597)
5463 - exit early from mds_open if we get a lookup error (1749)
5464 - partial page read at EOF wouldn't wait for disk before sending (1642)
5465 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5466 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5467 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5468 - write/truncate lock inversion (1639)
5469 - Don't auto-load obdclass, portals modules during cleanup (1495)
5470 - fix timestamps from jumping to "now" (1763)
5471 - extra journal assertions (1648)
5472 - add an extra multiunlink test (1771)
5473 - fix read_record/write_record API (1776)
5474 - fix leak of offset_extent, possible incorrect i_size later (1772)
5475 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5476 - Debugging for the kqswnal_get_idle_tx problems (1820)
5477 - Allow recovery to be attempted multiple times (1536)
5478 - Write out MDS last_rcvd file after it is first created (1600)
5479 - Fix tx_descriptor leak in failed transmit situations (1827)
5480 - ext3 journaling fixes for assertion failure after IO error (1871)
5481 - class_export_put() on freed export after completion AST error (1896)
5482 - Fix revalidate looping in VFS (1322)
5483 - Don't access a freed export during MDS_REINT timeout (1521)
5484 - Add open-unlink recovery support on the MDS (1673,1764)
5485 - Return an error if no MDS data was read from last_rcvd (1946)
5486 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5487 - Don't setup a disk device that doesn't match exported UUID (317)
5488 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5489 - avoid committing NULL handle in force close
5490 - local.sh is now a one-stripe LOV configuration
5491 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5492 - fix ext3 htree duplicate directory entry corruption (1516)
5493 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5494 - update to new LOV EA format (2097)
5495 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5497 2003-06-15 Phil Schwan <phil@clusterfs.com>
5500 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5501 - per-import/export recovery handling (958, 931, 959)
5502 - multiple last-rcvd slots, for serving multiple FSes (949)
5503 - connections are again shared between multiple imp/exports (963, 964)
5504 - "umount -f" would hang if any requests needed to be sent (393, 978)
5505 - avoid pinning large req buffer by copying for queued messages (989)
5506 - add "uuid" to "lctl device" command to help upcalls (991)
5507 - "open" RPCs with transnos would confuse recovery counters (1037)
5508 - do proper endian conversion of all wire messages (288, 340, 891)
5509 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5510 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5511 - fix ptlrpc_connection leak in target_handle_connect (1174)
5512 - fix import refcounting bug in OST and MDS cleanup (1134)
5513 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5514 - fix very unlikely obd_types race condition (501)
5515 - remove osc_open hack for echo_client (1187)
5516 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5517 - a failure in read_inode2 leads to deadlock (1139)
5518 - cancel ack-locks as soon as transaction is committed (1072)
5519 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5520 - make sure to commitrw after any preprw to avoid deadlock (1162)
5521 - failing to execute a file in a lustre FS would lock inode (1203)
5522 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5523 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5524 - obd_uuid2tgt would walk off the end of the list (1255)
5525 - on IA64 the osc would give portals incorrect bulk size (1258)
5526 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5527 - fix lock inversion caused by new llite matching code (1282)
5528 - limit the number of dirty pages on a client to 10MB (1286)
5529 - timed out locks were not being corrected cancelled (1289)
5530 - fix O_DIRECT above 4GB on IA-32 (1292)
5531 * major user-visible changes
5532 - fail out/fail over policy now controlled by the upcall (993)
5534 - add OBD_PING to check server availability and failure (954)
5535 - lustre messages are now sent in sending host order (288, 340, 891)
5536 - add eadatalen to MDS getattr reply (340)
5537 - OST read replies may contain second buffer, with per-page status (593)
5539 2003-03-11 Phil Schwan <phil@clusterfs.com>
5542 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5543 - failed open()s could cause deadlock; fixed (867, 869)
5544 - stop cancelling OST locks when files are closed (481)
5545 - overlapping XID spaces caused network corruption (851, 853)
5546 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5547 - setattr_raw added, to do single-RPC, server-side setattrs
5548 - lmc/lconf syntax change for OST UUIDs
5549 - fix crashy race condition between ptlrpc_free_req and osc_close
5550 - don't use request in mdc_enqueue if we hit a timeout (889)
5551 - don't set the inode i_size for regular files from the MDS (896)
5552 - handle out of order completion AST (842)
5553 - don't LBUG if a lock request times out after receiving AST (913)
5554 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5555 - if a bad lock AST arrives, send an error instead of dropping entirely
5556 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5557 - fix leak in bulk IO when only partially completed (899, 900, 926)
5558 - fix O_DIRECT for ia64 (55)
5559 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5560 - C-z after timeout could hang a process forever; fixed (977)
5562 - client-side I/O cache (678, 924, 929, 941, 970)
5564 - READPAGE and SETATTRs which don't take server-side locks get
5567 2003-02-11 Phil Schwan <phil@clusterfs.com>
5570 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5571 - Fix fsfilt_extN_readpage() to read a full page of directory
5572 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5573 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5574 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5575 - Most intent operations were removed, in favour of a new RPC mode
5576 that does a single RPC to the server and bypasses most of the VFS
5577 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5578 - Aggressively cancel local locks on DLM servers
5579 - mds_reint_unlink sends EA to the client if it's the last nlink.
5580 client uses that EA to unlink OST objects.
5581 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5582 - recursive symlinks were fixed (439)
5583 - fixed NULL deref in DEBUG_REQ
5584 - filter_update_lastobjid no longer calls sync, which annoyed extN
5585 - fixed multi-client small-writes to a single file problem (445)
5586 - fixed mtime updates during file writes (607)
5587 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5588 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5589 - fixed Linux OST size reporting problem (444, 656)
5590 - OST now updates object mtime with writes or setattr (607, 619)
5591 - client verifies file size before zeroing page past EOF (445)
5592 - OST now writes last allocated objid to disk with allocation (108)
5593 - LOV on echo now works (409)
5595 - mds_reint_unlink sends a new buffer, with the EA included. this
5596 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5597 set if a regular file was being unlinked, and it was the last link
5598 - use PtlGet from the target for bulk writes (315)
5599 - OST now updates object mtime with writes or setattr (607, 619)
5600 - LDLM now has a grant-time callback to revalidate locked items, if
5602 - Many MDS operations were reorganized to combat race conditions
5604 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5605 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5607 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5610 - Fully reactivate OST imports after reconnection (512, others)
5611 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5612 - More graceful error handling for truncating on dead OST (515)
5613 - Don't error out unless we're actually accessing dead stripes (474)
5614 - Fix garbage sizes when stripes are missing (410)
5615 - LRU counters were broken, causing constant lock purge (433, 432)
5616 - garbage on read from stripes with failed OSTs (441)
5617 - mark OSCs as active before reconnecting during recovery (438)
5618 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5619 - lfind did not preserve OST order in output (443)
5620 - symlinks cause hung clients, incorrect data (439)
5621 - stop dereferencing request after dropping refcount (457)
5622 - don't LASSERT(spin_is_locked) on non-SMP (455)
5623 - fixes for many rename() bugs
5624 - fstat didn't correctly synchronize attributes (399)
5625 - server must handle lock cancellation during blocking AST prep (487)
5626 - bulk descriptors were free()d too soon (511)
5627 - fix paths in lconf, which would load incorrect modules (451, 507)
5628 - fix confusing lconf 'host not found' error message (386)
5629 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5630 - fix race condition in mdc_blocking_ast() for inode access (526)
5631 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5632 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5633 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5634 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5635 - fix lstripe in several ways (kernel side) (527)
5636 - fix request leak in ldlm_cli_enqueue (262)
5637 - incorrect OSC was marked inactive after OST failure
5638 - call mds_fs_cleanup before unmounting filesystem (524)
5639 - fix races between taking ns_lock and ldlm_lock_change_resource
5640 - fix races updating LOV export open file list
5641 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5642 - fix recovery NULL deref in ldlm_cli_cancel_unused
5643 - fix some DLM races by using new hash table for lock handles (419)
5644 - permit the client to specify desired inodes, at replay
5645 - duplicate requests when we queue them for replay reintegration
5646 - fix last_rcvd offset calculation
5647 - sync after each recovered transaction, so we always make progress
5648 - never, not always, ERESTART requests without transnos
5649 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5651 - skip replay if the MDS says that the client is already connected
5652 - don't check for a recovery-enabled export to match lctl's UUID
5653 - don't INC_USE_COUNT for phantom exports
5654 - don't crash when cleaning up phantom exports (567)
5655 - don't double-finish or set replay data for errored mdc_open requests
5656 - abort requests when they time out, so we don't get old replies
5657 - send/receive replies for AST messages again
5658 - if the client says that it doesn't have the lock, cancel it on the
5660 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5661 mark it as destroyed, it will all work out when decref is called
5662 - fix module use counts (22, 581)
5664 - ASTs now expect a reply (server cancels lock on error reply)
5666 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5669 - fix many simultaneous client startup (392)
5670 - fix dentry->d_it clobbering
5671 - credentials weren't being shipped for readdir/getattr operations
5672 - remove invalid assertions triggered during some concurrent MD
5674 - proper Lustre versions added (336, 389)
5675 - fix memory leak for create error case (398)
5676 - fix LOV locking bug that would get cli/srv out of sync
5677 - fix echo client over LOV (409)
5678 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5679 - fix double-O_EXCL intent crash (424)
5680 - avoid sending multiple lock CANCELs (352)
5682 - MDS can do multi-client recovery (modulo bugs in new code)
5684 - many updates, edits, cleanups
5686 2002-11-18 Phil Schwan <phil@clusterfs.com>
5689 - fix null d_it dereference (346)
5690 - fix full OST/dbench hang (333)
5691 - fix permission problem with file removal (286)
5692 - fix removal of OSCs from LOV when they fail
5693 - fix NULL deref during bulk timeout (214)
5694 - fix problems related to multiple filesystems on one MDS (241)
5695 - fixed serious subtle metadata locking bugs
5696 - free locks on clients when inodes are removed due to memory
5698 - fix inode pointer in lock data (285)
5699 - partial support for multiple MDS on a single host (241)
5700 - data locks weren't cancelled at clear_inode time (290, 311)
5701 - intent locks could lead to unbounded lock growth (205)
5702 - added a maximum lock count, an LRU list, and a flusher
5703 - fix multiple rename (365)
5704 - properly abstracted the echo client
5705 - OSC locked 1 byte too many; fixed
5706 - rewrote brw callback code:
5707 - fixed recovery bugs related to LOVs (306)
5708 - fixed too-many-pages-in-one-write crash (191)
5709 - fixed (again) crash in sync_io_timeout (214)
5710 - probably fixed callback-related race (385)
5712 - Add capability to MDS protocol
5713 - LDLM cancellations and callbacks on different portals
5715 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5718 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5719 - defer open object destruction to close time (601981, 138)
5720 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5721 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5722 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5723 - don't call obd_setattr in ll_file_release for destroyed objects
5724 * protocol change to lustre_msg: move |version| and add |flags|
5725 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5726 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5727 * added replay of create, unlink, link and rename operations during
5728 MDS failover; recovery should be much more robust now
5729 * remove failed OSCs from LOVs (only lov_create uses this so far)
5730 * the lustre-HOWTO was brought (more) up to date (582544)
5732 2002-10-23 Phil Schwan <phil@clusterfs.com>
5735 - in-use dentries weren't being reused properly (617851)
5736 - prevent multiple LDLM setup (599178)
5737 - fix LOV size calculations for truncate (617853)
5738 - fix client handling of MDS intent errors (POSIX)
5739 - fix permission bug in lovstripe.c test (624321)
5740 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5741 - truncate past end of file could corrupt data
5742 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5743 - a race in recovery could return ETIMEDOUT to apps (623947)
5744 - building outside the source directory was fixed
5745 * the lustre-HOWTO was brought (more) up to date (582544)
5746 * major progress was made on recovery functionality
5748 2002-10-10 Phil Schwan <phil@clusterfs.com>
5751 - recovery deadlock fix
5752 - rm -rf causes LBUG fix (617817)
5753 - file open by multiple tasks fix (618962)
5754 - directory permissions bugs (602707 and 620007)
5755 - journal_stop fixed with locking (611313)
5756 - O_APPEND failures resolved (618273, perhaps 614459)
5757 - lconf PATH fix (619770)
5758 - IA64 build fix (621450)
5759 - RPC buffer sizes scale with amount of memory
5761 2002-10-01 Phil Schwan <phil@clusterfs.com>
5764 - locks would be cancelled without throwing away data pages,
5765 resulting in inconsistent data (605627)
5766 - inode attributes were not always being refreshed (605627, 612449)
5767 - lconf now continues to cleanup after lctl reports an error
5768 - MDS now enforces user permissions (602707)
5769 - lprocfs cleanup fixed, but not yet enabled (614157)
5770 - fixed infinite server hang, should a client not respond to an AST
5771 - avoid going into recovery if user calls readlink() with a buffer
5772 that's too small (613941)
5773 - AST RPCs no longer require replies (614867) -- this may be changed
5774 - don't crash server if client sends an IOV that's too big (611336)
5775 - fixed lock conversion deadlock (611892)
5776 - fixed the following of symlinks (614622)
5777 * recovery: the server can remove locks from a client that dies, other
5778 clients can make progress
5779 * more extN patch fixes
5780 * compile-time configurable ptlrpc buffer allocations
5782 - collaborative read cache document
5783 - Lustre Lite Performance CDR document-in-progress
5785 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5788 - fix typo in patch-2.4.18
5790 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5793 - clear ptlrpc request each time in handle_incoming_request()
5794 - unlink of files now destroys the object on the OST
5796 2002-09-19 Peter Braam <braam@clusterfs.com>
5798 * add hard link support
5799 * change obdfile creation method
5800 * kernel patch changed
5802 2002-09-19 Peter Braam <braam@clusterfs.com>
5805 - stack overflow bug in extN fixed
5807 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5809 * documentation updates
5810 - add man pages for config tools
5811 - update tests/README to describe testing with new config tools
5812 - finish metadata API descriptions
5813 * bug fixes and cleanups
5814 - statfs workaround for 16TB limit
5815 - LOV stripe allocation improved, can stripe on subset of OSTs
5816 - LOV file size/IO offset was wrong for files > 4GB in size
5817 - object EA data was being dropped, caused files to be unreadable
5818 - memory overflow with non-LOV OST caused memory corruption
5819 - fixed regression tests to work with new config tools, obdfilter
5820 - fixed bug when directory size became larger than 1 block
5821 - fixed bug (for single client case) when PWD was deleted
5822 - invalidate local directory pages when doing intent-based ops
5823 - avoid LDLM oops when lock callback contained bad data
5825 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5827 * documentation updates
5828 * bug fixes and cleanups
5829 - configuration tools
5832 - 64-bit compile warnings
5833 - 64-bit internal statfs data
5835 * test_brw on persistent OST devices
5837 * lprocfs (disabled)
5839 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5841 * documentation updates
5842 * bug fixes and cleanups
5843 * configuration tools
5845 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5848 * many small fixes to 0_5_4
5849 * io/network handling
5850 * thinkos in MDS operations
5852 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5855 * crucial basic fixes to 0.5.3
5856 * IOR, Iozone work over Elan
5859 2002-08-07 Phil Schwan <phil@clusterfs.com>
5860 * version 0_5_3, our first alpha
5861 * we use the new Portals iovs
5862 * documentation updates
5863 * bug fixes and cleanups
5864 * small changes in the DLM wire protocol
5866 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5867 * version 0_5_1 with some initial stability,
5868 * locking on MD and file I/O.
5869 * documentation updates
5870 * several bug fixes since 0.5.0
5871 * small changes in wire protocol
5873 2002-07-18 Phil Schwan <phil@clusterfs.com>
5875 * delivered as Lustre Light Alpha
5876 * fixed a crash after handling invalid MDS requests
5877 * fixed directory pages for architectures with non-4k pages sizes
5879 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5880 * release version v0_4_4
5881 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5883 * Fixes a number of interruption problems with OST operations.
5884 * Update documentation for portals header changes
5885 * Move all wire protocol structs/defines to lustre_idl.h
5886 * Fixes symlink length bug.
5887 * Add tcpdump to repository.
5889 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5890 * release version v0_4_3
5891 * Fixes statfs for inodes on extN.
5892 * Fixes bug in runtests which would delete /etc/hosts.
5893 * Use 64-bit object IDs wherever possible (not into VFS though)
5894 Remove ost_get_info, which is unused by lustre, and out of date.
5896 2002-07-03 Peter Braam <braam@clusterfs.com>
5897 * release version v0_4_2 Fixes a lookup error (type not passed)
5898 * move forward to head of Portals
5899 * move forward to latest Lustre kernel
5901 2002-06-25 Peter Braam <braam@clusterfs.com>
5902 * release version v0_4_1. Hopefully stable on single node use.