1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.39.3 (SLES 10),
5 2.6.18-128.7.1.el5 (RHEL 5),
6 2.6.18-128.7.1.el5 (OEL 5),
8 2.6.27 vanilla (kernel.org).
9 * Client support for unpatched kernels:
10 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
11 2.6.16 - 2.6.27 vanilla (kernel.org)
12 * Recommended e2fsprogs version: 1.41.6.sun1
13 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
14 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
15 removed cwd "./" (refer to Bugzilla 14399).
16 * File join has been disabled in this release, refer to Bugzilla 16929.
18 Severity : enhancement
20 Description: Add 2.6.27 and SLES11 (patchless client) support
22 Severity : enhancement
24 Description: Adjust locks' extents on their first enqueue, so that at the time
25 they get granted, there is no need for another pass through the
26 queues since they are already shaped into the proper forms.
30 Description: Fix in ptlrpc_expire_one_request() to print the signed time
33 Severity : enhancement
35 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
38 Severity : enhancement
40 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
42 Severity : enhancement
44 Description: correctly handle big reply message.
45 Details : send LNet event if reply is bigger then buffer and adjust this buffer
50 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
52 Severity : enhancement
54 Description: Add support for OEL5.
58 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
62 Description: client could not reconnect to OST because of an active request.
63 Details : The client now retries to reconnect to the same server, if a
64 connect request failed with EBUSY or -EAGAIN. Test case 35b
65 conf-sanity.sh has been created.
69 Description: Don't allow make backward step on assignin osc next id.
70 Details : race between allocation next id and ll_sync thread can be cause
71 of set wrong osc next id and can be kill valid ost objects.
75 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
79 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
83 Description: Error handling in osc_statfs_interpret() has been improved.
84 Details : Check in osc_statfs_interpret() for EBADR.
88 Description: Do not send statfs() requests to OSTs disabled by administrator.
89 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
91 Severity : enhancement
93 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
95 Severity : enhancement
97 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
99 Severity : enhancement
101 Description: Enhancement for lfs(1) command to use numeric uid/gid.
105 Description: Improved cleanup procedure for conf-sanity test_46a.
109 Description: Do not start bulk transfer, if client tried to reconnect.
110 Details : Do not start bulk transfer, if client tried to reconnect,
111 but failed due to active rpcs. Check for exp_abort_active_req
112 has been implemented.
116 Description: Prevent inconsistences between linux and lustre mount structures.
117 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
118 Make the sleep interruptible.
120 Severity : enhancement
122 Description: provide server to client comms path
123 Details : server to client communications path via new LDLM_SET_INFO rpc
127 Description: truncate starts GFP_FS allocation under transaction causing deadlock
128 Details : ldiskfs_truncate calls grab_cache_page which may start page
129 allocation under an open transaction. This may lead to
130 calling prune_icache with consequent lustre reentrance.
134 Description: send correctly lsm on open replay
135 Details : MDS is trust to LSM size on replay open, but client can set wrong size
140 Description: enable client interface failover
141 Details : When a child reconnects from another NID, properly update export
142 nid hash position and ldlm reverse import.
144 Severity : enhancement
146 Description: Communicate OST degraded/readonly state via statfs to MDS
147 Details : Flags in the statfs returned from OSTs indicate whether the
148 OST is in a degraded RAID state, or if the filesystem has
149 turned read-only after a filesystem error is detected.
154 Description: don't panic if EPROTO was hit when reading symlink
155 Details : correctly handling request reference in error cases.
157 Severity : enhancement
159 Description: Add LustreNetLink, a kernel-userspace communcation path.
161 Severity : enhancement
163 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
168 Description: access to llog context before init.
169 Details : move handling CATALOGS file at osc layer and forbid access to llog
173 Frequency : rare, only when client is writing to OST on local node
175 Description: Avoid deadlock for local client writes
176 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
177 memory freeing context. This allows OST threads to set the
178 PF_MEMALLOC flag on task structures in order to allocate memory
179 from reserved pools and complete IO.
180 Use GFP_HIGHUSER for OST allocations for non-local client writes,
181 so that the OST threads generate memory pressure and allow
182 inactive pages to be reclaimed.
184 Severity : enhancement
186 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
191 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
192 Details : move ldlm namespace creation in setup phase to avoid grab
193 _lprocfs_lock with cli_sem held.
197 Description: Temporarily disable grant shrink.
198 Details : Disable the feature for debugging.
201 Frequency : only when formatting test filesystems
203 Description: Unable to run several mkfs.lustre on loop devices at the same time
204 Details : mkfs.lustre returns error 256 on the concurrent loop devices
205 formatting. The solution is to proper handle the error.
207 Severity : enhancement
209 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
211 Severity : enhancement
213 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
218 Description: resolve race between obd_disconnect and class_disconnect_exports
219 Details : if obd_disconnect will be called to already disconnected export he
220 forget release one reference and osc module can't unloaded.
222 Severity : enhancement
224 Description: Allow tuning service thread via /proc
225 Details : For each service a new
226 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
227 created that can be used to set min/max thread counts, and get the
228 current number of running threads.
233 Descriptoin: don't return error if have partially created objects for file.
234 Details : lov_update_create_set uses set->set_success as index for created
235 objects, so if some requests will be failed, they will have hole
236 at end of array and qos_shrink_lsm for allocate correct lsm.
238 Severity : enhancement
240 Description: Update OFED support to 1.4.1
245 Description: Reduce small size read RPC
246 Details : Set read-ahead limit for every file and only do read-ahead when
247 available read-ahead pages are bigger than 1M to avoid small size
250 Severity : enhancement
252 Description: move AT tunable parameters for more consistent usage
253 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
255 Severity : enhancement
257 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
258 Details : allow skip disconnected ost for send statfs request and hide error
264 Description: fix racy locking of mballoc block bitmaps causing BUG
265 Details : The locking of the mballoc buddy bitmap and the in-memory
266 block bitmap was using two different spin locks in some
267 cases. This made it possible to incorrectly access the
268 mballoc bitmap while another process was modifying it,
269 causing a sanity assertion to fail. While no on-disk corruption
270 was reported, there was some risk of this happening.
273 Frequency : rare, on llog test 6
275 Descriptoin: don't allow connect to already connected import
276 Details : allowing connect to already connected import is hide connecting problem.
281 Descriptoin: allow kill process which wait statahead result
282 Details : for some reasons 'ls' can stick in waiting result from statahead,
283 in this case need way for kill this process.
285 Severity : enhancement
287 Description: Add state history info file, enhance import info file
288 Details : Track import connection state changes in a new osc/mdc proc file;
289 add overview-type data to the osc/mdc import proc file.
291 Severity : enhancement
293 Description: MDS create should not wait for statfs RPC while holding DLM lock.
295 Severity : enhancement
297 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
302 Descriptoin: OST grant too much space to client even there are not enough space.
303 Details : Client will shrink its grant cache to OST if there are no write
304 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
305 this grant cache if there are already not enough avaible space
306 (left_space < total_clients * 32M).
309 Frequency : start MDS on uncleanly shutdowned MDS device
311 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
312 Details : stay in waiting mds<>ost recovery finished produce random bugs
313 due race between two ll_sync thread for one lov target. send
314 ACTIVATE event only if connect realy finished and import have
318 Frequency : rare, connect and disconnect target at same time
320 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
321 Details : don't call obd_disconnect under lov_lock. this long time
322 operation and can block ptlrpcd which answer to connect request.
327 Descriptoin: don't lose wakeup for imp_recovery_waitq
328 Details : recover_import_no_retry or invalidate_import and import_close can
329 both sleep on imp_recovery_waitq, but we was send only one wakeup
333 Frequency : always with long access acl
335 Descriptoin: mds can't pack reply with long acl.
336 Details : mds don't control size of acl but they limited by reint/getattr
339 Severity : enhancement
341 Description: Update to SLES10 kernel-2.6.16.60-0.33.
343 Severity : enhancement
345 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
348 Frequency : start MDS on uncleanly shutdowned MDS device
350 Descriptoin: aborting recovery hang on MDS
351 Details : don't throttle destroy RPCs for the MDT.
354 Frequency : on remount
356 Description: external journal device not working after the remount
357 Details : clear dev_rdonly flag for external journal devices in
363 Description: shutdown vs evict race
364 Details : client_disconnect_export vs connect request race.
365 if client will evicted at this time - we start invalidate
366 thread without referece to import and import can be freed
370 Frequency : rare, need acl's on inode.
372 Description: client can't handle ost additional correctly
373 Details : if ost was added after client connected to mds client can have
374 hit lnet_try_match_md ... to big messages to wide striped files.
375 in this case need teach client to handle config events about add
376 lov target and update client max ea size at that event.
378 Severity : enhancement
380 Description: Changelogs
381 Details : Changelogs are a lightweight mechanism to track filesystem
382 metadata and namespace changes. The changelog is recorded
383 permanently on the MDTs, and is periodically "consumed" / purged
384 when records are no longer needed.
386 Severity : enhancement
388 Description: compact fld format with extents
389 Details : Store range of seq rather than every seq in FLD. Seq
390 controller update FLD rather than clients. In Case of CMD, mdt0
391 has FLD, all other metadata server act as non persistent proxy
392 for FLD queries and cache fld entries in fld cache.
397 Description: don't skip ost target if they assigned to file
398 Details : Drop slow OSCs if we can, but not for requested start idx.
399 This means "if OSC is slow and it is not the requested
400 start OST, then it can be skipped, otherwise skip it only
401 if it is inactive/recovering/out-of-space.
405 Description: more cleanup in mds_lov
406 Details : not send LOV EA under replay, we can't know about they size at this
407 time. Don't allow client connect to mds before any ost connected,
408 for avoid problems with LOV EA size and returning EIO to client.
410 Severity : enhancement
412 Description: Interoperability at server side (Disk interoperability)
414 Severity : enhancement
416 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
418 Severity : enhancement
420 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
422 Severity : enhancement
424 Description: New client IO stack (CLIO).
426 Severity : enhancement
428 Description: Commit on sharing. Eliminate inter-client dependencies between
429 uncommitted transactions by doing transaction commits.
430 Thereby clients may recovery independently.
433 Frequency : Create a symlink file with a very long name
435 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
436 Details : If there is no extra space in the request for early cancels,
437 ldlm_req_handles_avail() returns 0 instead of a negative value.
439 Severity : enhancement
441 Description: Add /proc entry for import status
442 Details : The mdc, osc, and mgc import directories now have
443 an import directory that contains useful import data for debugging
446 Severity : enhancement
448 Description: Re-disable certain /proc logging
449 Details : Enable and disable client's offset_stats, extents_stats and
450 extents_stats_per_process stats logging on the fly.
453 Frequency : Only on FC kernels 2.6.22+
455 Description: oops in statahead
456 Details : Do not drop reference count for the dentry from VFS when lookup,
457 VFS will do that by itself.
459 Severity : enhancement
461 Description: Generic /proc file permissions
462 Details : Set /Proc file permissions in a more generic way to enable non-
463 root users operate on some /proc files.
467 Description: Hitting mdc_commit_close() ASSERTION
468 Details : Properly handle request reference release in
469 ll_release_openhandle().
473 Description: quota recovery deadlock during mds failover
474 Details : This patch includes att18982, att18236, att18237 in bz14840.
476 1. fix osts hang when mds does failover with quotaon
477 2. prevent watchdog storm when osts threads wait for the
482 Frequency : only patchless client
483 Description: add workaround for race between add/remove dentry from hash
485 Severity : enhancement
487 Description: Allow OST glimpses to return PW locks
491 Description: LBUG when llog conf file is full
492 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
496 Description: Prevent import from entering FULL state when server in recovery
500 Description: service mount cannot take device name with ":"
501 Details : Only when device name contains ":/" will mount treat it as
507 Description: replace ptlrpcd with the statahead thread to interpret the async
508 statahead RPC callback
512 Frequency : on recovery
513 Description: I/O failures after umount during fail back
514 Details : if client reconnected to restarted server we need join to recovery
515 instead of find server handler is changed and process self
516 eviction with cancel all locks.
518 Severity : enhancement
520 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
522 Severity : enhancement
524 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
526 Severity : enhancement
528 Description: Upcall on Lustre log has been dumped
529 Details : Allow for a user mode script to be called once a Lustre log has
530 been dumped. It passes the filename of the dumped log to the
531 script, the location of the script can be specified via
532 /proc/sys/lnet/debug_log_upcall.
537 Description: avoid idr_remove called for id which is not allocated.
538 Details : Move assigment s_dev for clustered nfs to end of initialization,
539 to avoid problem with error handling.
544 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
545 Details : When connection is reused this not moved from CONN_UNUSED_HASH
546 into CONN_USED_HASH and this prodice warning when put connection
547 again in unused hash.
550 Severity : enhancement
552 Description: File striping can now be set to use an arbitrary pool of OSTs.
554 Severity : enhancement
556 Description: Export bytes_read/bytes_write count on OSC/OST.
560 Description: Early reply size mismatch, MGC loses connection
561 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
562 the connect flags are properly negotiated.
567 Description: quotas are not honored with O_DIRECT
568 Details : all writes with the flag O_DIRECT will use grants which leads to
569 this problem. Now using OBD_BRW_SYNC to guard this.
573 Description: add quota statistics
574 Details : 1. sort out quota proc entries and proc code.
575 2. add quota statistics
577 Severity : enhancement
579 Description: enable quota support for HEAD.
583 Description: Properly propagate oinfo flags from lov to osc for statfs
584 Details : restore missing copy oi_flags to lov requests.
586 Severity : enhancement
588 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
589 : l_getgroups(8), lst(8), routerstat(8)
590 Details : included man pages for llobdstat(8), llstat(8),
591 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
593 Severity : enhancement
595 Description: configure's --enable-quota should check the
596 : kernel .config for CONFIG_QUOTA
597 Details : configure is terminated if --enable-quota is passed but
598 : no quota support is in kernel
602 Description: Remove portals compatibility
603 Details : Remove portals compatibility, not interoperable with releases
608 Description: Resolve device initialization race
609 Details : Prevent proc handler from accessing devices added to the
610 obd_devs array but yet be intialized.
612 Severity : enhancement
614 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
616 Severity : enhancement
618 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
623 Description: Using wrong pointer in osc_brw_prep_request
624 Details : Access to array[-1] can produce panic if kernel compiled with
625 CONFIG_PAGE_ALLOC enabled
629 Description: Client runs out of low memory
630 Details : Consider only lowmem when counting initial number of llap pages
634 Description: *optional* service tags registration
635 Details : if the "service tags" package is installed on a Lustre node
636 When the filesystem is mounted, a local-node service tag will
637 be created. See http://inventory.sun.com/ for more information
638 about the Service Tags asset management system.
642 Description: Kernel BUG tries to release flock
643 Details : Lustre does not destroy flock lock before last reference goes
644 away. So always drop flock locks when client is evicted and
645 perform unlock regardless of successfulness of speaking to MDS.
649 Description: add refcount protection for osc callbacks, avoid panic on shutdown
653 Description: sanity test 65a fails if stripecount of -1 is set
654 Details : handle -1 striping on filesystem in ll_dirstripe_verify
659 Description: ASSERTION(CheckWriteback(page,cmd)) failed
660 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
663 Severity : enhancement
665 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
669 Description: do not process already freed flock
670 Details : flock can possibly be freed by another thread before it reaches
671 to ldlm_flock_completion_ast.
675 Description: LBUG during stress test
676 Details : Need properly lock accesses the flock deadlock detection list.
680 Description: oops in page fault handler
681 Details : kernel page fault handler can return two special 'pages' in error
682 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
686 Description: timeout with invalidate import.
687 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
688 should be handled by ptlrpcd. This produce long age waiting and
689 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
691 Severity : enhancement
693 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
697 Description: enable MGS and MDT services start separately
698 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
699 a co-located MGS without starting the MGS, which is a complement
700 to 'nosvc' mount option.
704 Frequency : after recovery
705 Description: precreate to many object's after del orphan.
706 Details : del orphan st in oscc last_id == next_id and this triger growing
707 count of precreated objects. Set flag LOW to skip increase count
708 of precreated objects.
712 Frequency : rare, on clear nid stats
713 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
714 Details : when clean nid stats sometimes try destroy live entry,
715 and this produce panic in free.
719 Description: Stack overflow during MDS log replay
720 ease stack pressure by using a thread dealing llog_process.
724 Description: wait until IO finished before start new when do lock cancel.
725 Details : VM protocol want old IO finished before start new, in this case
726 need wait until PG_writeback is cleared until check dirty flag and
727 call writepages in lock cancel callback.
729 Severity : enhancement
731 Description: using special macro for print time and cleanup in includes.
735 Description: mds_mfd_close() ASSERTION(rc == 0)
736 Details : In mds_mfd_close(), we need protect inode's writecount change
737 within its orphan write semaphore to prevent possible races.
741 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
745 Frequency : rare, on shutdown ost
746 Description: don't hit live lock with umount ost.
747 Details : shrink_dcache_parent can be in long loop with destroy dentries,
748 use shrink_dcache_sb instead.
752 Description: don't panic with use echo client
753 Details : echo client pass NULL as client nid pointer and this produce null
758 Description: fix build on ppc32
759 Details : compile code with -m64 flag produce wrong object file for ppc32.
763 Description: add message levels for liblustreapi
767 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
768 Details : Move the 'good_osts' check before the 'total_bavail' check. This
769 will result in an -EAGAIN and in the exit call path we call
770 alloc_rr() which will with increasing aggressiveness attempt to
771 aquire precreated objects on the minimum number of required OSCs.
775 Description: Use old size assignment to avoid deadlock
776 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
777 the scheduling while holding a spinlock. We do not need locking
778 for size in ll_update_inode() because size is only updated from
779 the MDS for directories or files without objects, so there is no
780 other place to do the update, and concurrent access to such inodes
781 are protected by the inode lock.
785 Description: resolve "_IOWR redefined" build error on SLES10
789 Description: dump the memory debugging after all modules are unloaded to
790 suppress false negative in conf_sanity test 39
792 Severity : enhancement
794 Description: build kernel-ib packages for OFED 1.3 in our release cycle
799 Description: fix SLES kernel versioning
800 Details : the kernel version for our SLES 10 kernel did not include a "-"
801 before the "smp" at the end. while this was not a problem in
802 general, it did mean that software trying to use the kernel
803 version to try to detect a vendor specific kernel would fail.
804 this was most evident by the OFED build scripts.
808 Description: Don't update lov_desc members until making sure they are valid
809 Details : When updating lov_desc members via proc fs, need fix their
810 validities before doing the real update.
814 Description: don't put request into delay list while invalidate in flight.
815 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
816 import in flight. this produce timeout for invalidate and sometimes
817 can cause stale data.
821 Frequency : on ppc only
822 Description: not convert ost objects for directory because it's not exist.
823 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
827 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
829 Severity : enhancement
831 Description: Update to support 2.6.22.14 vanilla kernel.
835 Frequency : rare, on recovery
836 Description: read procfs can produce deadlock in some situation
837 Details : Holding lprocfs lock which send rpc can produce block for destroy
838 obd objects and this also block reconnect with -EALREADY. This
839 isn't fix all lprocfs bugs - but make it rare.
841 Severity : enhancement
843 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
846 Frequency : frequent on X2 node
848 Description: mdc_set_open_replay_data LBUG
849 Details : Set replay data for requests that are eligible for replay.
853 Description: lustre_mgs: operation 101 on unconnected MGS
854 Details : When MGC is disconnected from MGS long enough, MGS will evict the
855 MGC, and late on MGC cannot successfully connect to MGS and a lot
856 of the error messages complaining that MGS is not connected.
860 Frequency : on network error
861 Description: panic with double free request if network error
862 Details : mdc_finish_enqueue is finish request if any network error ocuring,
863 but it's true only for synchronus enqueue, for async enqueue
864 (via ptlrpcd) this incorrect and ptlrpcd want finish request
867 Severity : enhancement
869 Description: client-side metadata stat-ahead during readdir(directory readahead)
870 Details : perform client-side metadata stat-ahead when the client detects
871 readdir and sequential stat of dir entries therein
874 Frequency : on start mds
876 Description: Implement get_info(last_id) in obdfilter.
879 Frequency : occasional
881 Description: Correctly check stale fid, not start epoch if ost not support SOM
882 Details : open with flag O_CREATE need set old fid in op_fid3 because
883 op_fid2 was overwritten with new generated fid, but mds can answer
884 with one of these two fids and both is not stale. Setattr
885 incorrectly started an epoch and assume will be called
886 done_writing, but without SOM done_writing ever being called.
889 Frequency : rare, depends on device drivers and load
891 Description: MDS or OSS nodes crash due to stack overflow
892 Details : Code changes in 1.8.0 increased the stack usage of some functions.
893 In some cases, in conjunction with device drivers that use a lot
894 of stack the MDS (or possibly OSS) service threads could overflow
895 the stack. One change which was identified to consume additional
896 stack has been reworked to avoid the extra stack usage.
899 Frequency : occasional
901 Description: Do not fail import if osc_interpret_create gets -EAGAIN
902 Details : If osc_interpret_create got -EAGAIN it immediately exits and
903 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
904 oscc_has_objects and see OSC has no objests and call
905 oscc_internal_create to resend create request.
907 Severity : enhancement
909 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
911 Severity : enhancement
913 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
916 Frequency : very rare
918 Description: Fix replay if there is an un-replied request and open
919 Details : In some cases, older replay request will revert the
920 mcd->mcd_last_xid on MDS which is used to record the client's
923 Severity : enhancement
925 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
927 Severity : enhancement
929 Description: Add rhel5 support to HEAD.
931 Serverity : enhancement
933 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
938 Description: Don't allow skipping OSTs if index has been specified.
939 Details : Don't allow skipping OSTs if index has been specified, make
940 locking in internal create lots better.
944 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
945 Details : access to bitfield in structure is always rounded to long
946 and this produce problem with not atomic change any bit.
950 Description: Lustre make rpms failed.
951 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
952 builds Lustre rpms from tarball.
955 Frequency : rare on shutdown ost
957 Description: If llog cancel was not send before clean_exports phase, this can
958 produce deadlock in llog code.
959 Details : If llog thread has last reference to obd and call class_import_put
960 this produce deadlock because llog_cleanup_commit_master wait when
961 last llog_commit_thread exited, but this never success because was
962 called from llog_commit_thread.
966 Description: allow userland application know is lost one of stripes.
967 Details : fill lvb_blocks with error code on ost and return it to
968 application if error flag found.
972 Description: NULL lov_tgts causing MDS oops
973 Details : more safe checks for NULL lov_tgts for avoid oops.
975 Severity : enhancement
977 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
981 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
982 Details : If an OST is down the MDS will hang indefinitely in
983 obd_statfs_rqset() waiting for the statfs data. While for
984 MDS QOS usage of statfs, it should not stuck in waiting.
986 Severity : enhancement
988 Description: remote_acl support
989 Details : Support ACL-based permission check for remote user.
990 Support setfacl/getfacl for remote user with the utils
991 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
992 the system "{s,g}etfacl" utils.
994 Severity : enhancement
996 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
998 Severity : enhancement
1000 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1004 Description: make mgs_setparam() handle fsname containing dash
1005 Details : fsname containing a dash does not work with lctl conf_param
1007 Severity : enhancement
1009 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1011 Severity : enhancement
1013 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1017 Frequency : rare, at shutdown
1018 Description: access already free / zero obd_namespace.
1019 Details : if client_disconnect_export was called without force flag set,
1020 and exist connect request in flight, this can produce access to
1021 NULL pointer (or already free pointer) when connect_interpret
1022 store ocd flags in obd_namespace.
1026 Frequency : only at startup
1027 Description: not alloc memory with spinlock held.
1028 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1029 if any spinlock held.
1031 Severity : enhancement
1033 Description: make lustre randomly fail allocating memory
1034 Details : Make lustre randomly failed allocating memory for testing purpose.
1036 Severity : enhancement
1038 Description: lost problems with lov objid file
1039 Details : Fixes some scability and access to not inited memory problems
1040 in work with lov objdid file.
1045 Description: lfs find does not continue on file error
1046 Details : Continue other files processing when a file/dir is absent.
1050 Description: Inconsistent usage of lustre_pack_reply()
1051 Details : Standardize the usage of lustre_pack_reply() such that it
1052 always generate a CERROR on failure.
1055 Frequency : occasional
1057 Description: MDS hang or stay in waiting lock
1058 Details : If client receive lock with CBPENDING flag ldlm need send lock
1059 cancel as separate rpc, to avoid situation when cancel request
1060 can't processed due all i/o threads stay in wait lock.
1064 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1068 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1072 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1076 Description: Update to RHEL4 latest kernel.
1078 Severity : enhancement
1080 Description: Build SLES10 patchless client fails
1081 Details : The configure was broken by run ./configure with
1082 --with-linux-obj=.... argument for patchless client. When the
1083 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1084 can't search header adequately. Use absolute path such as
1085 -I($LINUX)/include instead.
1089 Description: interrupt oig_wait produce painc on resend.
1090 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1091 private set, and this produce situation when rq_ptlrpcd_data not
1092 copyed to new allocated request and triggered LBUG on assert
1093 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1094 ptlrpcd set if request is changed and can be safety changed to use
1099 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1100 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1101 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1102 these macros and replace them with their difinitions instead.
1106 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1107 Details : Update which_patch & target file for SLES10 latest kernel.
1110 Description: Remove client patches from SLES 10 kernel.
1111 Details : This causes SLES 10 clients to behave as patchless clients
1112 even on a Lustre-patched (server) kernel.
1114 Severity : enhancement
1116 Description: self-adjustable client's lru lists
1117 Details : use adaptive algorithm for managing client cached locks lru
1118 lists according to current server load, other client's work
1119 pattern, memory activities, etc. Both, server and client
1120 side namespaces provide number of proc tunables for controlling
1123 Severity : enhancement
1125 Description: light-weight GSS support
1126 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1127 and reduce performance overhead.
1129 Severity : enhancement
1131 Description: Linux keyring support
1132 Details : Support using service of Linux keyring for Lustre GSS internal
1133 context refresh/cache mechanism.
1137 Description: Fix errors in lfs documentation
1138 Details : Fixes man pages
1142 Description: don't use GFP_* in generic Lustre code.
1143 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1147 Description: obdclass is limited by single OBD_ALLOC(idarray)
1148 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1152 Description: warnings with build patchless client with vanila 2.6.19 and up
1153 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1154 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1158 Frequency : Only for SLES
1159 Description: sanity_quota fail test_1
1160 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1161 file, which makes TSTID[2] inunique.
1165 Frequency : Only for SLES9
1166 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1167 Details : Update target/ChangeLog/which_patch .
1171 Frequency : Only for test_10 in sanity-quota.sh
1172 Description: a bug in quota test code
1173 Details : checking a test flag in a wrong place
1177 Frequency : Only for test_13 in sanity-quota.sh
1178 Description: a bug in quota test script
1179 Details : shouldn't check overfull allocated quota
1183 Description: jbd statistics
1184 Details : Port older jbd statistics patch for sles10
1186 Severity : enhancement
1188 Description: Add printing inode info into message about error in writepage.
1193 Description: reply_lock_interpret crash due to race with it and lock cancel.
1194 Details : Do not replay locks that are being cancelled. Do not reference
1195 locks by their address during replay, just by their handle.
1199 Frequency : When flocks are used.
1200 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1201 Details : Flock locks might destroy just granted lock if it could be merged
1202 with another existing flock, this is done in completion handler,
1203 so teach ldlm_cli_enquque_fini that this is a valid case for
1209 Description: Oops in read and write path when failing to allocate lock.
1210 Details : Check if lock allocation failed and return error back.
1214 Description: lstripe command fails for valid OST index
1215 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1216 instead of lov->desc.ld_active_tgt_count.
1220 Description: sanity.sh failed test 103
1221 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1226 Description: df doesn't work properly if diskfs blocksize != 4K
1227 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1231 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1232 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1234 Severity : enhancement
1236 Description: Adaptive timeouts
1237 Details : RPC timeouts adapt to changing server load and network
1238 conditions to reduce resend attempts and improve recovery time.
1242 Description: llapi_file_create() does not allow some changes
1243 Details : add llapi_file_open() that allows specifying the mode and
1244 open flags, and also returns an open file handle.
1248 Description: merge and cleanup kernel patches.
1249 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1253 Description: Add journal checksum support.(Kernel part)
1254 Details : The journal checksum feature adds two new flags i.e
1255 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1256 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1257 indicates that the commit block contains the checksum for
1258 the blocks described by the descriptor blocks. Now commit
1259 record can be sent to disk without waiting for descriptor
1260 blocks to be written to disk. This behavior is controlled
1261 using JBD2_FEATURE_ASYNC_COMMIT flag.
1265 Description: OSS needs mutliple precreate threads
1266 Details : Add ability to start more than one create thread per OSS.
1270 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1271 Details : execute lfs setstripe on client
1275 Description: mds_obd_create error creating tmp object
1276 Details : When the user sets quota on root, llog will be affected and can't
1277 create files and write files.
1280 Frequency : Always on ia64 patchless client, and possibly others.
1282 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1283 Details : This allows the patchless client to be loaded on architectures
1284 without this export.
1288 Description: RedHat Update kernel for RHEL5
1289 Details : Add the kernel config file for RHEL5.
1293 Description: RedHat Update kernel for RHEL5
1294 Details : Modify the kernel config file more closer RHEL5.
1298 Description: Build failure against Centos5 (RHEL5)
1299 Details : Define PAGE_SIZE when it isn't present.
1303 Description: test_7 on sanity-quota.sh can't be used on separate machines
1304 Details : add facet to handle it
1308 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1309 Details : with NFS, the anon dentry's parent was set to itself in
1310 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1311 corresponding dentry other than use rec->ur_name.
1313 Severity : enhancement
1315 Description: lfs setstripe enhancement
1316 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1320 Description: enable data checksumming by default
1321 Details : enable checksum by default, allow --disable-checksum
1322 configure option and "-o nochecksum" mount option. Checksums
1323 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1327 Description: lustre support for RHEL5
1328 Details : Add support for RHEL5.
1332 Description: Client eviction due to failover config
1333 Details : after a connection loss, the lustre client should attempt to
1334 reconnect to the last active server first before trying the
1335 other potential connections.
1339 Description: when mds and osts use different quota unit(32bit and 64bit),
1340 quota will be released repeatly.
1341 Details : void sending multiple quota reqs to mds, which will keep the
1342 status between the reqs.
1346 Description: osts not allocated evenly to files
1347 Details : change the condition to increase offset_idx
1350 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1352 Description: Data corruption for OSTs that are formatted larger than 2TB
1354 Details : When generating the bio request for lustre file writes the
1355 sector number would overflow a temporary variable before being
1356 used for the IO. The data reads correctly from Lustre (which
1357 will overflow in a similar manner) but other file data or
1358 filesystem metadata may be corrupted in some cases.
1362 Description: Tune the kernel for good SCSI performance.
1363 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1364 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1369 Description: rewrite ext2-derived code in obdclass/uuid.c
1370 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1371 from scratch preserving functionality.
1375 Description: "lfs find -obd UUID" prints directories
1376 Details : "lfs find -obd UUID" will return all directory names instead
1377 of just file names. It is incorrect because the directories
1378 do not reside on the OSTs.
1382 Description: use i_size_read and i_size_write in 2.6 port
1383 Details : replace inode->i_size access with i_size_read/write()
1386 Frequency : when removing large files
1388 Description: scheduling issue during removal of large Lustre files
1389 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1390 It causes scheduling issues when removing large files (17TB in the
1394 Frequency : only with liblustre clients on XT3
1396 Description: evictions taking too long
1397 Details : allow llrd to evict clients directly on OSTs
1400 Frequency : only on ppc
1402 Description: /proc/fs/lustre/devices broken on ppc
1403 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1404 We should make sure the type of 'index' is loff_t and then cast
1405 explicitly as needed below. Do not assign an explicitly cast
1409 Frequency : only for rhel5
1411 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1412 Details : Modify the target file & which_kernel.
1414 Severity : enhancement
1416 Description: omit set fsid for export NFS
1417 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1418 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1422 Frequency : Always, for kernels after 2.6.16
1423 Description: Fix warning idr_remove called for id=.. which is not allocated.
1424 Details : Last kernels save old s_dev before kill super and not allow
1425 to restore from callback - restore it before call kill_anon_super.
1429 Description: buffer overruns could theoretically occur
1430 Details : llapi_semantic_traverse() modifies the "path" argument by
1431 appending values to the end of the origin string, and a
1432 overrun may occur. Adding buffer overrun check in liblustreapi.
1436 Description: Fix error on 'ls .' at the top of the Lustre mount.
1437 Details : Don't revalidate dentry if it is a root dentry.
1441 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1442 Details : Modify vm-tunables-rhel4.patch.
1446 Description: Kernel config for 2.6.18-vanilla.
1447 Details : Modify targets/2.6-vanilla.target.in.
1448 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1449 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1450 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1451 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1455 Description: improve handling recoverable errors
1456 Details : if request processig with error which can be recoverable on server
1457 request should be resend, otherwise page released from cache and
1462 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1463 Details : Modify target file & which_patch.
1464 A flaw was found in the IA32 system call emulation provided
1465 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1466 value could be stored in the %RAX register, which could trigger an
1467 out-of-bounds system call table access. An untrusted local user
1468 could exploit this flaw to run code in the kernel
1469 (ie a root privilege escalation). (CVE-2007-4573).
1473 Description: O_DIRECT bypasses client statistics.
1474 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1475 read_ahead_stats were updated. copied the stats section from
1476 osc_send_oap_rpc() into async_internal().
1480 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1484 Description: change order of libsysio includes
1485 Details : '#include sysio.h' should always come before '#include xtio.h'
1489 Description: handle "serious error: objid * already exists" more gracefully
1490 Details : If LAST_ID value on disk is smaller than the objects existing in
1491 the O/0/d* directories, it indicates disk corruption and causes an
1492 LBUG(). If the object is 0-length, then we should use the existing
1493 object. This will help to avoid a full fsck in most cases.
1495 Severity : enhancement
1497 Description: adapt the lustre_config script to support the upgrade case
1498 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1499 server targets to 1.6 in parallel.
1503 Description: To avoid grant space > avaible space when the disk is almost
1504 full. Without this patch you might see the error "grant XXXX >
1505 available" or some LBUG about grant, when the disk is almost
1507 Details : In filter_check_grant, for non_grant cache write, we should
1508 check the left space by if (*left > ungranted + bytes), instead
1509 of (*left > ungranted), because only we are sure the left space
1510 is enough for another "bytes", then the ungrant space should be
1511 increase. In client, we should update cl_avail_grant only there
1512 is OBD_MD_FLGRANT in the reply.
1516 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1517 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1518 described in CVE-2007-4573. This problem could allow untrusted
1519 local users to gain root access.
1522 Frequency : when using O_DIRECT and quotas
1524 Description: Incorrect file ownership on O_DIRECT output files
1525 Details : block usage reported by 'lfs quota' does not take into account
1526 files that have been written with O_DIRECT.
1531 Description: touch file failed when fs is not full
1532 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1533 otherwise we can get ENOSP while fs is not full.
1537 Description: parallel lock callbacks
1538 Details : Instead of sending blocking and completion callbacks as separated
1539 requests, adding them to a set and sending in parallel.
1542 Frequency : only for Cray XT3
1543 Bugzilla : 12829/13455
1544 Description: Changing primary group doesn't change the group lustre assigns to
1546 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1547 overridden with the primary group provided by the group upcall,
1548 whereas the supplied fsgid can be trusted if it is in the list of
1549 supplementary groups returned by the group upcall.
1551 Severity : enhancement
1553 Description: Allow masking D_WARNING, D_ERROR messages from console
1554 Details : Console messages can now be disabled via lnet.printk.
1558 Description: User code with malformed file open parameter crashes client node
1559 Details : Before packing join_file req, all the related reference should be
1560 checked carefully in case some malformed flags cause fake
1561 join_file req on client.
1565 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1566 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1567 mode is equal to the granted mode. However, it should also check
1568 that LDLM_FL_CP_REQD is not set.
1572 Description: Heavy nfs access might result in deadlocks
1573 Details : After ELC code landed, it is now improper to enqueue any mds
1574 locks under och_sem, because enqueue might want to decide to
1575 cancel open locks for same inode we are holding och_sem for.
1579 Description: Client eviction while running blogbench
1580 Details : A lot of unlink operations with concurrent I/O can lead to a
1581 deadlock causing evictions. To address the problem, the number of
1582 oustanding OST_DESTROY requests is now throttled to
1583 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1584 callbacks are processed in priority.
1588 Description: enable ACLs on MDS by default
1589 Details : ACLs must be enabled on MDS by default.
1592 Frequency : PPC/PPC64 only
1594 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1595 Details : fix duplicated definitions between asm-ppc64/types.h and
1596 lustre_types.h on PPC.
1599 Frequency : PPC/PPC64 only
1601 Description: asm-ppc/segment.h does not exist
1602 Details : fix compile issue on PPC.
1606 Description: better handle error messages in extents code
1609 Frequency : RHEL4 only
1611 Description: mkfs is very slow on IA64/RHEL4
1612 Details : A performance regression has been discovered in the MPT Fusion
1613 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1614 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1615 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1617 Severity : enhancement
1619 Description: SNMP support enhancement
1620 Details : Adding total number of sampled request for an MDS node in snmp
1623 Severity : enhancement
1625 Description: Optimize ldlm waiting list processing for PR extent locks
1626 Details : When processing waiting list for read extent lock and meeting read
1627 lock that is same or wider to it that is not contended, skip
1628 processing rest of the list and immediatelly return current
1629 status of conflictness, since we are guaranteed there are no
1630 conflicting locks in the rest of the list.
1634 Description: Time out and refuse to reconnect
1635 Details : When the failover node is the primary node, it is possible
1636 to have two identical connections in imp_conn_list. We must
1637 compare not conn's pointers but NIDs, otherwise we can defeat
1638 connection throttling.
1642 Description: port llog fixes from b1_6 into HEAD
1643 Details : Port llog reference couting and some llog cleanups from b1_6
1644 (bug 10800) into HEAD, for protect from panic and access to already
1645 free llog structures.
1649 Description: Detect stride IO mode in read-ahead
1650 Details : When a client does stride read, read-ahead should detect that and
1651 read-ahead pages according to the detected stride pattern.
1655 Description: data checksumming impacts single node performance
1656 Details : add support for several checksum algorithm. Currently, only CRC32
1657 and Adler-32 are supported. The checksum type can be changed on
1658 the fly via /proc/fs/lustre/osc/*/checksum_type.
1662 Description: use adler32 for page checksums
1663 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1668 Description: build for x2 fails
1669 Details : fix compile issue on Cray systems.
1673 Description: Properly match for duplicate locks
1674 Details : Due to different lock order from skiplists code, we need to
1675 traverse entire list for now
1678 Frequency : only on PPC/SLES10
1680 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1681 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1682 be defined. Add a hack in mkfs_lustre.c to work around this compile
1687 Description: LASSERT on MDS when client holding flock lock dies
1688 Details : ldlm pool logic depends on number of granted locks equal to
1689 number of released locks which is not true for flock locks, so
1690 just exclude such locks from consideration.
1694 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1695 Details : Use fsfilt_sync() for both the whole filesystem sync and
1696 individual file sync to eliminate dangerous inode locking
1697 with I_LOCK that can lead to a deadlock.
1701 Description: performance in 1.6.3
1702 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1703 sure that 1MB requests are not fragmented by the block layer.
1705 Severity : enhancement
1707 Description: organize the server-side client stats on per-nid basis
1708 Details : Change the structure of stats under obdfilter and mds to
1716 The "uuid"s file would list the uuids of _active_ exports.
1717 And the clear entry is to clear all stats and stale nids.
1719 Severity : enhancement
1721 Description: eliminate client locks in face of contention
1722 Details : file contention detection and lockless i/o implementation
1723 for contended files.
1727 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1730 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1732 Description: MDS may lose file striping (and hence file data) in some cases
1733 Details : If there are additional extended attributes stored on the MDS,
1734 in particular ACLs, SELinux, or user attributes (if user_xattr
1735 is specified for the client mount options) then there is a risk
1736 of attribute loss. Additionally, the Lustre file striping
1737 needs to be larger than default (e.g. striped over all OSTs),
1738 and an additional attribute must be stored initially in the
1739 inode and then increase in size enough to be moved to the
1740 external attribute block (e.g. ACL growing in size) for file
1745 Description: skiplist implementation simplification
1746 Details : skiplists are used to group compatible locks on granted list
1747 that was implemented as tracking first and last lock of each
1748 lock group the patch changes that to using doubly linked lists
1752 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1753 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1754 because LASSERT on both the data supplied by a client, and the
1755 data on disk is dangerous and incorrect.
1757 Severity : enhancement
1759 Description: Slow truncate/writes to huge files at high offsets.
1760 Details : Directly associate cached pages to lock that protect those pages,
1761 this allows us to quickly find what pages to write and remove
1762 once lock callback is received.
1766 Description: more ldlm soft lockups
1767 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1768 starve other threads from the resource lock for a long time in
1769 case of long waiting queue, so change the debug level from
1770 D_OTHER to the less frequently used D_INFO.
1772 Severity : enhancement
1774 Description: add -gid, -group, -uid, -user options to lfs find
1778 Description: Hung threads in invalidate_inode_pages2_range
1779 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1780 once one is completed. As a result, some RPCs are stuck in the
1781 queue and are never sent.
1785 Description: filter threads hungs on waiting journal commit
1786 Details : Cleanup filter group llog code, then only filter group llog will
1787 be only created in the MDS/OST syncing process.
1791 Description: Procfs and llog threads access destoryed import sometimes.
1792 Details : Sync the import destoryed process with procfs and llog threads by
1793 the import refcount and semaphore.
1795 Severity : enhancement
1797 Description: openlock cache of b1_6 port to HEAD
1802 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1803 Details : If insertion of an extent fails, then discard the inode
1804 preallocation and free data blocks else it can lead to duplicate
1809 Description: don't always update ctime in ext3_xattr_set_handle()
1810 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1811 In some cases the ctime should not be updated, for example for
1812 2.0->1.8 compatibility it is necessary to delete an xattr and it
1813 should not update the ctime.
1817 Bugzilla : 15713/16362
1818 Description: Assertion in iopen_connect_dentry in 1.6.3
1819 Details : looking up an inode via iopen with the wrong generation number can
1820 populate the dcache with a disconneced dentry while the inode
1821 number is in the process of being reallocated. This causes an
1822 assertion failure in iopen since the inode's dentry list contains
1823 both a connected and disconnected dentry.
1827 Description: assertion failure in ldlm_handle2lock()
1828 Details : fix a race between class_handle_unhash() and class_handle2object()
1829 introduced in lustre 1.6.5 by bug 13622.
1834 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1835 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1840 Description: oops in sd_iostats_seq_show()
1841 Details : unloading/reloading the scsi low level driver triggers a kernel
1842 bug when trying to access the sd iostat file.
1847 Description: Kernel panics during QLogic driver reload
1848 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1849 patch, causing memory corruption.
1854 Description: journal_dev option does not work in b1_6
1855 Details : pass mount option during pre-mount.
1857 Severity : enhancement
1859 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1860 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1861 extent information of a file. It can be used to map logical blocks
1862 in a file to physical blocks in the block device.
1866 Description: LDLM soft lockups - improvement
1867 Details : It is be possible to send the lock handle along with each read
1868 or write request because the client is already doing a lock match
1869 itself so there isn't any reason the OST should have to re-do that
1873 Frequency : only on Cray X2
1875 Description: X2 build failures
1876 Details : fix build failures on Cray X2.
1880 Description: xid & resent requests
1881 Details : Initialize RPC XID from clock at startup (randomly if clock is
1884 Severity : enhancement
1886 Description: Add lustre_start utility to start or stop multiple Lustre servers
1891 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1892 Details : In case of memory pressure, list_del() can be called twice on
1893 req->rq_history_list, causing a kernel oops.
1897 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1898 Details : fix stack overflow in the distributed lock manager by defering
1899 export eviction after a failed AST to the elt thread instead of
1900 handling it in the dlm interpret routine.
1904 Description: Convert some comments to new format.
1905 Details : Update documenting comments to match doxygen conventions.
1909 Description: Grammar fixes.
1910 Details : A couple of trivial spelling fixes.
1914 Description: OSD_COUNTERS-mandatory
1915 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1916 to simplify the code.
1920 Description: simplify cmm_device freeing logic.
1921 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1925 Description: Add lockdep support to dt_object_operations locking interface.
1926 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1927 parameter indicating lock ordering. Update mdd code to use new
1932 Description: Introduce failloc constants for lockless IO tests.
1933 Details : Add two new failloc constants to test lockless IO. Only one of
1934 them in implemented---another is checked in yet to be landed
1939 Description: Add lockdep support for inode mutex.
1940 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1941 in the situations where Lustre has to lock more than one inode
1946 Description: Add optional invariants checking support.
1947 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1948 configure switch. This macro is to be used for consistency and
1949 sanity checks that are too expensive to be left in `production'
1954 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1955 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1960 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1961 Details : Use spin_lock_nested() in (the only) situation where more than
1962 one ldlm_lock is locked simultaneously. Also, fix possible
1963 dead-lock in ldlm_lock_change_resource() by enforcing particular
1968 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1969 Details : Instead of specifying each ldlm_lock call-back through separate
1970 parameter, wrap them into struct ldlm_callback_suite.
1974 Description: Kill join_lru obd method and its callers.
1975 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1976 in memory---a requirement implemented through ->o_join_lru() obd
1977 method in HEAD. Remove this method and its users.
1981 Description: Add asynchronous ldlm ENQUEUE completion handler.
1982 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1983 case that stock ldlm_completion_ast() cannot handle as it waits
1984 until lock is granted. Introduce new ldlm_completion_ast_async()
1985 for this. Also comment ldlm_completion_ast().
1989 Description: ldlm_error <-> errno conversion.
1990 Details : Add functions to map (rather arbitrary) between LDLM error codes
1991 and standard errno values. CLIO needs this to prevent LDLM specific
1992 constants from escaping ldlm and osc.
1996 Description: Kill unused ldlm_handle2lock_ns() function.
1997 Details : Kill unused ldlm_handle2lock_ns() function.
2001 Description: Add lu_ref support to ldlm_lock
2002 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2003 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2004 and ldlm_resource. LDLM interface has to be changed, because code that
2005 releases a reference on a lock, has to "know" what reference this is.
2006 In the most frequent case
2008 lock = ldlm_handle2lock(handle);
2010 LDLM_LOCK_PUT(lock);
2012 no changes are required. When any other reference (received _not_ from
2013 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2014 instead of LDLM_LOCK_PUT().
2016 Arguably, changes are pervasive, and interface requires some discipline
2017 for proper use. On the other hand, it was very instrumental in finding
2018 a few leaked lock references.
2022 Description: Add ldlm_lock_addref_try().
2023 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2024 attempts to addref a lock that might be being canceled
2029 Description: Add ldlm_weigh_callback().
2030 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2031 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2032 hard-coded `number of pages' logic.
2036 Description: Add lockdep annotations to llog code.
2037 Details : Use appropriately tagged _nested() locking calls in the places
2038 where llog takes more than one ->lgh_lock lock.
2042 Description: Add loi_kms_set().
2043 Details : Wrap kms updates into a helper function.
2047 Description: Constify instances of struct lsm_operations.
2048 Details : Constify instances of struct lsm_operations.
2052 Description: lu_conf support.
2053 Details : On a server, a file system object is uniquely identified
2054 by a fid, which is sufficient to locate and load all object
2055 state (inode). On a client, on the other hand, more data are
2056 necessary instantiate an object. Change lu_object_find() and
2057 friends to take additional `lu_conf' argument describing object.
2058 Typically this includes layout information.
2062 Description: lu_context fixes.
2063 Details : Introduce new lu_context functions that are needed on the client
2064 side, where some system threads (ptlrpcd) are shared by multiple
2065 modules, and so cannot be stopped during module shutdown.
2069 Description: Add start and stop methods to lu_device_type_operations.
2070 Details : Introduce two new methods in lu_device_type_operations, that are
2071 invoked when first instance of a given type is created and last one
2072 is destroyed respectively. This is need by CLIO.
2076 Description: Add lu_ref support to struct lu_device.
2077 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2078 track leaked references.
2082 Description: Introduce lu_kmem_descr.
2083 Details : lu_kmem_descr and its companion interface allow to create
2084 and destroy a number of kmem caches at once.
2088 Description: Fix lu_object finalization race.
2089 Details : Fix a race between lu_object_find() finding an object and its
2090 concurrent finalization. This race is (most likely) not possible
2091 on the server, but might happen on the client.
2095 Description: Introduce lu_ref interface.
2096 Details : lu_ref is a debugging module allowing to track references to
2097 a given object. It is quite cpu expensive, and has to be
2098 explicitly enabled with --enable-lu_ref. See usage description
2103 Description: Factor lu_site procfs stats into a separate function.
2104 Details : Separate lu_site stats printing code into a separate function
2105 that can be reused on a client.
2109 Description: Constify instances of struct {lu,dt,md}_device_operations.
2110 Details : Constify instances of struct {lu,dt,md}_device_operations.
2114 Description: Introduce struct md_site and move meta-data specific parts of
2115 struct lu_site here.
2116 Details : Move md-specific fields out of struct lu_site into special struct
2117 md_site, so that lu_site can be used on a client.
2121 Description: Kill mdd_lov_destroy().
2122 Details : Remove unused mdd code.
2126 Description: Add st_block checking to multistat.c.
2127 Details : Add st_block checking to multistat.c.
2131 Description: Add lu_ref support to struct obd_device.
2132 Details : Add lu_ref tracking to obd_device.
2136 Description: Kill obd_set_fail_loc().
2137 Details : Remove unused code.
2141 Description: Add special type for ptlrpc_request interpret functions.
2142 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2143 there. Actual usage will be in CLIO.
2147 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2148 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2149 doesn't work with lockdep.
2153 Description: Add rwv.c test program.
2154 Details : New testing program exercising readv(2) and writev(2) (Qian).
2158 Description: Add sendfile.c test program.
2159 Details : New testing program exercising sendfile(2) (Jay).
2163 Description: Ratelimit a message that can be very frequent.
2164 Details : Ratelimit a memory allocation failure message that can
2169 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2170 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2174 Description: Liblustre build fixes.
2175 Details : Liblustre build fixes.
2179 Description: libcfs: add cfs_{need,cond}_resched() interface.
2180 Details : libcfs: add cfs_{need,cond}_resched() definition and
2181 implementations for Linux, NT, and liblustre.
2183 Severity : enhancement
2185 Description: More exported tunables for mballoc
2186 Details : Add support for tunable preallocation window and new tunables for
2187 large/small requests
2191 Description: Detect corruption of block bitmap and checking for preallocations
2192 Details : Checks validity of on-disk block bitmap. Also it does better
2193 checking of number of applied preallocations. When corruption is
2194 found, it turns filesystem readonly to prevent further corruptions.
2198 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2199 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2200 window to 0. If it does hit cache miss, and read-pattern is still
2201 stride-io mode, does not reset the stride window, but also does not
2202 increase the stride window length in this case.
2206 Frequency : only for big-endian servers
2207 Description: Check if system is big-endian while mounting fs with extents feature
2208 Details : Mounting a filesystem with extents feature will fail on big-endian
2209 systems since ext3-based ldiskfs is not supported on big-endian
2210 systems. This can be over-riden with "bigendian_extents" mount option.
2212 Severity : enhancement
2214 Description: The root squash functionality
2215 Details : A security feature, which is to prevent users from being able
2216 to mount lustre on their desktop, run as root, and delete
2217 all of the files in the filesystem. The goal is accomplished by
2218 remapping user id (UID) and group id (GID) of the root user to
2219 a UID and GID specified by the system administartor via Lustre
2220 configuration management server (MGS). The functionality also
2221 allows to specify sets of clients for which the remapping does
2226 Description: Excessive recovery window
2227 Details : With AT enabled, the recovery window can be excessively long (6000+
2228 seconds). To address this problem, we no longer use
2229 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2230 timeout no longer depends on the service time, it is set to
2231 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2232 time via pb_service_time.
2236 Description: Watchdog triggered on MDS failover
2237 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2238 the OSTs know that the MDS "UUID" can be reused for the same export
2239 from a different NID, so we do not need to wait for the export to be
2243 Frequency : rare, only if using MMP with Linux RAID
2245 Description: MMP doesn't work with Linux RAID
2246 Details : While using HA for Lustre servers with Linux RAID, it is possible
2247 that MMP will not detect multiple mounts. To make this work we
2248 need to unplug the device queue in RAID when the MMP block is being
2249 written. Also while reading the MMP block, we should read it from
2250 disk and not the cached one.
2252 Severity : enhancement
2254 Description: open file using fid
2255 Details : A file can be opened using just its fid, like
2256 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2259 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2262 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2263 Details : add proper locking for creation of jbd_4k slab cache
2267 Description: MMP check in ext3_remount() fails without displaying any error
2268 Details : When multiple mount protection fails during remount, proper error
2271 Severity : enhancement
2273 Description: Allow stripe size to be up to 4G-64k
2274 Details : Fix math logic to allow large stripe sizes.
2278 Description: add check for >8TB ldiskfs filesystems
2279 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2280 Don't allow >8TB ldiskfs filesystems to be mounted without
2281 force_over_8tb mount option
2286 Descriptoin: OSS complains "too many blocks" and crashes
2287 Details : In filter_commitrw_write():
2288 1. lquota_chkquota();
2289 2. obdo_from_inode();
2290 3. lquota_pending_commit();
2291 "2." has changed uid and gid in this case, then caused
2292 lquota_chkquota() and lquota_pending_commit() to use
2293 different uid and gid. That is the root of the bug.
2297 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2298 Details : Detect and print out the 'stripe count = -1' case properly.
2302 Description: Out or order replies might be lost on replay
2303 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2304 request with tid smaller then one currently being added, add it
2305 to the start, not end of the list.
2307 --------------------------------------------------------------------------------
2309 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2311 * Support for kernels:
2312 2.6.5-7.283 (SLES 9),
2313 2.6.9-55.EL (RHEL 4),
2314 2.6.16.46-0.14 (SLES 10),
2315 2.6.18.8 vanilla (kernel.org)
2316 * Client support for unpatched kernels:
2317 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2318 2.6.16 - 2.6.22 vanilla (kernel.org)
2319 * Due to recently discovered recovery problems, we do not recommend
2320 using patchless RHEL 4 clients with this or any earlier release.
2321 * Recommended e2fsprogs version: 1.39.cfs8
2322 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2323 * Starting with this release, the ldiskfs backing filesystem required
2324 by Lustre is now in its own package, lustre-ldiskfs. This package
2325 should be installed. It is versioned separately from Lustre and
2326 may be released separately in future.
2330 Description: block reactivating mgc import until all deactivates complete
2331 Details : Fix race when failing back MDT/MGS to itself (testing)
2334 Frequency : at statup only
2336 Description: mds_lov_synchronize race leads to various problems
2337 Details : simultaneous MDT->OST connections at startup can cause the
2338 sync to abort, leaving the OSC in a bad state.
2340 Severity : enhancement
2342 Description: add optional extra BUILD_VERSION info
2343 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2344 to override the lustre version.
2347 Frequency : 2.6.18 servers only
2349 Description: ll_kern_mount() doesn't release the module reference
2350 Details : The ldiskfs module reference count never drops down to 0
2351 because ll_kern_mount() doesn't release the module reference.
2356 Description: server LBUG when using old ost_num_threads parameter
2357 Details : Accept the old ost_num_threads parameter but warn that it
2358 is deprecated, and fix an off-by-one error that caused an LBUG.
2363 Description: Transient SCSI error results in persistent IO issue
2364 Details : iobuf->dr_error is not reinitialized to 0 between two
2368 Frequency : sometimes when underlying device returns I/O errors
2370 Description: OSTs not going read-only during write failures
2371 Details : OSTs are not remounted read-only when the journal commit threads
2372 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2373 instead of the ext3 wrappers.
2377 Description: poor connect scaling with increasing client count
2378 Details : Don't run filter_grant_sanity_check for more than 100 exports
2379 to improve scaling for large numbers of clients.
2382 Frequency : SLES10 only
2384 Description: sanity-quota.sh quotacheck failed: rc = -22
2385 Details : Quotas cannot be enabled on SLES10.
2388 Frequency : liblustre clients only
2390 Description: getdirentries does not give error when run on compute nodes
2391 Details : getdirentries does not fail when the size specified as an argument
2392 is too small to contain at least one entry
2394 Severity : enhancement
2396 Description: Add LNET router traceability for debug purposes
2397 Details : If a checksum failure occurs with a router as part of the
2398 IO path, the NID of the last router that forwarded the bulk data
2399 is printed so it can be identified.
2404 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2405 Details : Due to a race condition, liblustre clients were occasionally
2406 evicted incorrectly.
2408 Severity : enhancement
2410 Description: lfs setstripe use optional parameters instead of postional
2413 Severity : enhancement
2415 Description: Nanosecond timestamp support for ldiskfs
2416 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2417 resolution timestamps. There is not yet support for this at
2418 the Lustre filesystem level.
2421 Frequency : during server recovery
2423 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2424 Details : request with rq_no_resend flag not awake l_wait_event if they get
2428 Frequency : nfs export on patchless client
2430 Description: connectathon hang when test nfs export over patchless client
2431 Details : Disconnected dentry cannot be found with lookup, so we do not need
2432 to unhash it or make it invalid
2435 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2436 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2437 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2442 Description: open req refcounting wrong on reconnect
2443 Details : If reconnect happened between getting open reply from server and
2444 call to mdc_set_replay_data in ll_file_open, we will schedule
2445 replay for unreferenced request that we are about to free.
2446 Subsequent close will crash in variety of ways.
2447 Check that request is still eligible for replay in
2448 mdc_set_replay_data().
2453 Description: disable writes to filesystem when reading health_check file
2454 Details : the default for reading the health_check proc file has changed
2455 to NOT do a journal transaction and write to disk, because this
2456 can cause reads of the /proc file to hang and block HA state
2457 checking on a healthy but otherwise heavily loaded system. It
2458 is possible to return to the previous behaviour during configure
2459 with --enable-health-write.
2461 Severity : enhancement
2463 Description: 64-bit inode version
2464 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2465 to the inode. This will be required for version-based recovery.
2470 Description: MDS fails to start if a duplicate client export is detected
2471 Details : in some rare cases it was possible for a client to connect to
2472 an MDS multiple times. Upon recovery the MDS would detect this
2473 and fail during startup. Handle this more gracefully.
2475 Severity : enhancement
2477 Description: Add -o localflock option to simulate old noflock
2479 Details : This will achieve local-only flock/fcntl locks
2485 Description: log_commit_thread vs filter_destroy race leads to crash
2486 Details : Take import reference before releasing llog record semaphore
2491 Description: Wrong request locking in request set processing
2492 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2493 imp_delayed_list, in this place should be used imp_lock.
2496 Frequency : when reconnection
2498 Description: Grant Leak when osc reconnect to OST
2499 Details : When osc reconnect ost, OST(filter) should check whether it
2500 should grant more space to client by comparing fed_grant and
2501 cl_avail_grant, and return the granted space to client instead
2502 of "new granted" space, because client will call osc_init_grant
2503 to update the client grant space info.
2506 Frequency : when client reconnect to OST
2508 Description: Grant Leak when osc do resend and replay bulk write
2509 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2510 bulk write request, because the grant info while be sync between
2511 OSC and OST when reconnect, and we should ignore the grant info
2512 these of resend/replay write req.
2517 Description: Grant space more than avaiable left space sometimes.
2518 Details : When then OST is about to be full, if two bulk writing from
2519 different clients came to OST. Accord the avaliable space of the
2520 OST, the first req should be permitted, and the second one
2521 should be denied by ENOSPC. But if the seconde arrived before
2522 the first one is commited. The OST might wrongly permit second
2523 writing, which will cause grant space > avaiable space.
2526 Frequency : when client is evicted
2528 Description: Grant might be wrongly erased when osc is evicted by OST
2529 Details : when the import is evicted by server, it will fork another
2530 thread ptlrpc_invalidate_import_thread to invalidate the
2531 import, where the grant will be set to 0. While the original
2532 thread will update the grant it got when connecting. So if
2533 the former happened latter, the grant will be wrongly errased
2534 because of this race.
2539 Description: Checking Stale with correct fid
2540 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2541 to check whether it is stale, because sometimes, we want the
2542 enqueue happened anyway, and op_data.fid2 will not be initialized.
2544 Severity : enhancement
2546 Description: update patchless client
2547 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2550 Frequency : only with 2.4 kernel
2552 Description: random memory corruption
2553 Details : size of struct ll_inode_info is to big for union inode.u and this
2554 can be cause of random memory corruption.
2559 Description: Memory leak in recovery
2560 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2561 It should also check obd_fail before fsfilt_start, otherwise if
2562 fsfilt_start return -EROFS,(failover mds during mds recovery).
2563 then the req will return with repmsg->transno = 0 and rc = EROFS.
2564 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2565 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2566 be freed no matter whether fsfilt_commit success or not.
2569 Frequency : only with huge count clients
2571 Description: Prevents from taking the superblock lock in llap_from_page for
2573 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2574 for llap_from_page call in ll_removepage prevents from taking the
2575 superblock lock for a soon died page.
2580 Description: Not check open intent error before release open handle
2581 Details : in some rare cases, the open intent error is not checked before
2582 release open handle, which may cause
2583 ASSERTION(open_req->rq_transno != 0), because it tries to release
2584 the failed open handle.
2589 Description: Set cat log bitmap only after create log success.
2590 Details : in some rare cases, the cat log bitmap is set too early. and it
2591 should be set only after create log success.
2595 Description: Accessing a block bevice can re-enable I/O when Lustre is
2596 tearing down a device.
2597 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2602 Description: service threads may hog cpus when there are a lot of requests
2603 Details : Insert cond_resched to give other threads a chance to use some CPU
2608 Description: the cat log was not initialized in recovery
2609 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2610 unlink log on mds will not be initialized until mds post
2611 recovery. And also in mds post recovery, the unlink log will
2612 initialization will be done asynchronausly, so there will be race
2613 between add unlink log and unlink log initialization.
2617 Description: brw_stats were being printed incorrectly
2618 Details : brw_stats were being printed as log2 but all of them were not
2619 recorded as log2. Also remove some code duplication arising from
2620 filter_tally_{read,write}.
2624 Frequency : rare, only in recovery.
2625 Description: ASSERTION(req->rq_type != LI_POISON) failed
2626 Details : imp_lock should be held while iterating over imp_sending_list for
2627 prevent destroy request after get timeout in ptlrpc_queue_wait.
2631 Description: replay-single.sh test 52 fails
2632 Details : A lock's skiplist need to be cleanup when it being unlinked
2633 from its resource list.
2637 Description: Short directio read returns full requested size rather than
2639 Details : Direct I/O operations should return actual amount of bytes
2640 transferred rather than requested size.
2642 Severity : enhancement
2644 Description: metadata RPC reduction (e.g. for rm performance)
2645 Details : decrease the amount of synchronous RPC between clients and servers
2646 by canceling conflicing lock before the operation on the client
2647 and packing thier handles into the main operation RPC to server.
2649 Severity : enhancement
2651 Description: Async OSC create to avoid the blocking unnecessarily.
2652 Details : If a OST has no remain object, system will block on the creating
2653 when need to create a new object on this OST. Now, ways use
2654 pre-created objects when available, instead of blocking on an
2655 empty osc while others are not empty. If we must block, we block
2656 for the shortest possible period of time.
2660 Description: improve handling recoverable errors
2661 Details : if request processig with error which can be recoverable on server
2662 request should be resend, otherwise page released from cache and
2665 Severity : enhancement
2667 Description: refine locking for avoid write wrong info into lov_objid file
2668 Details : fix possible races with add new target and write/update data in
2671 --------------------------------------------------------------------------------
2673 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2678 Frequency : on some architectures
2680 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2681 Details : Uninitialized flags sometimes cause configuration commands to
2685 Frequency : patchless clients only
2687 Description: missing __iget() symbol export
2688 Details : The __iget() symbol export is missing. To avoid the need for
2689 this on patchless clients the deathrow inode reaper is turned
2690 off, and we depend on the VM to clean up old inodes. This
2691 dependency was during via the fix for bug 12181.
2693 --------------------------------------------------------------------------------
2695 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2697 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2698 INTEROPERATE with older versions automatically. In many cases a
2699 special upgrade step is needed. Please read the
2700 user documentation before upgrading any part of a live system.
2701 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2702 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2703 * WARNING: Lustre configuration and startup changes are required with
2704 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2708 Severity : enhancement
2710 Description: MountConf
2711 Details : Lustre configuration is now managed via mkfs and mount
2712 commands instead of lmc and lconf. New obd types (MGS, MGC)
2713 are added for dynamic configuration management. See
2714 https://mail.clusterfs.com/wikis/lustre/MountConf for
2717 Severity : enhancement
2719 Description: dynamic OST addition
2720 Details : OSTs can now be added to a live filesystem
2722 Severity : enhancement
2724 Description: startup order invariance
2725 Details : MDTs and OSTs can be started in any order. Clients only
2726 require the MDT to complete startup.
2728 Severity : enhancement
2730 Description: parallel, asynchronous orphan cleanup
2731 Details : orphan cleanup is now performed in separate threads for each
2732 OST, allowing parallel non-blocking operation.
2734 Severity : enhancement
2736 Description: optimized stripe assignment
2737 Details : stripe assignments are now made based on ost space available,
2738 ost previous usage, and OSS previous usage, in order to try
2739 to optimize storage space and networking resources.
2741 Severity : enhancement
2743 Description: Permanently set tunables
2744 Details : All writable /proc/fs/lustre tunables can now be permanently
2745 set on a per-server basis, at mkfs time or on a live system.
2747 Severity : enhancement
2749 Description: Lustre message v2
2750 Details : Add lustre message format v2.
2752 Severity : enhancement
2754 Description: client OST exclusion list
2755 Details : Clients can be started with a list of OSTs that should be
2756 declared "inactive" for known non-responsive OSTs.
2760 Description: SPEC SFS validation failure on NFS v2 over lustre.
2761 Details : Changes the blocksize for regular files to be 2x RPC size,
2762 and not depend on stripe size.
2764 Severity : enhancement
2766 Description: Multiple MD RPCs in flight.
2767 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2768 To avoid overly-overloading MDS, introduce a limit on number of
2769 MDS RPCs in flight for a single client and add /proc controls
2770 to adjust this limit.
2772 Severity : enhancement
2774 Description: client read/write statistics
2775 Details : Add client read/write call usage stats for performance
2776 analysis of user processes.
2777 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2778 file access. extents_stats shows chunk size distribution.
2779 extents_stats_per_process show chunk size distribution per
2782 Severity : enhancement
2784 Description: mds statistics
2785 Details : Add detailed mds operations statistics in
2786 /proc/fs/lustre/mds/*/stats.
2790 Description: Failure of copying files with lustre special EAs.
2791 Details : Client side always return success for setxattr call for lustre
2792 special xattr (currently only "trusted.lov").
2797 Description: Refcount LNET uuids
2798 Details : The global LNET uuid list grew linearly with every startup;
2799 refcount repeated list entries instead of always adding to
2802 Severity : enhancement
2804 Description: Dynamic service threads
2805 Details : Within a small range, start extra service threads
2806 automatically when the request queue builds up.
2809 Frequency : mixed-endian client/server environments
2811 Description: mixed-endian crashes
2812 Details : The new msg_v2 system had some failures in mixed-endian
2815 Severity : enhancement
2817 Description: Easy OST removal
2818 Details : OSTs can be permanently deactivated with e.g. 'lctl
2819 conf_param lustre-OST0001.osc.active=0'
2821 Severity : enhancement
2823 Description: MGS proc entries
2824 Details : Added basic proc entries for the MGS showing what filesystems
2827 Severity : enhancement
2829 Description: provide MGS failover
2830 Details : Added config lock reacquisition after MGS server failover.
2832 Severity : enhancement
2834 Description: add Linux 2.4 support
2835 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2839 Description: a large application tries to do I/O to the same resource and dies
2840 in the middle of it.
2841 Details : Check the req->rq_arrival time after the call to
2842 ost_brw_lock_get(), but before we do anything about
2843 processing it & sending the BULK transfer request. This
2844 should help move old stale pending locks off the queue as
2845 quickly as obd_timeout.
2848 Frequency : when an incorrect nid is specified during startup
2850 Description: ptlrpc connect to non-existant node causes kernel crash
2851 Details : LNET can't be re-entered from an event callback, which
2852 happened when we expire a message after the export has been
2853 cleaned up. Instead, hand the zombie cleanup off to another
2856 Severity : enhancement
2858 Description: plain/inodebits lock performance improvement
2859 Details : Grouping plain/inodebits in granted list by their request modes
2860 and bits policy, thus improving the performance of search through
2864 Frequency : only if OST filesystem is corrupted
2866 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2867 Details : for a short time RPCs with bulk IO are in the replay list,
2868 but replay of bulk IOs is unimplemented. If the OST filesystem
2869 is corrupted due to disk cache incoherency and then replay is
2870 started it is possible to trip an assertion. Avoid putting
2871 committed RPCs into the replay list at all to avoid this issue.
2874 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2876 Description: System hang on startup
2877 Details : This bug allowed the liblustre (e.g. catamount) client to
2878 return to the app before handling all startup RPCs. This
2879 could leave the node unresponsive to lustre network traffic
2880 and manifested as a server ptllnd timeout.
2882 Severity : enhancement
2884 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2885 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2886 boolean to control whether to print peer debug info when a
2887 client's RPC times out.
2890 Frequency : only for kernels with patches from Lustre below 1.4.3
2892 Description: Remove old rdonly API
2893 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2896 Frequency : only for devices with external journals
2898 Description: Set external device read-only also
2899 Details : During a commanded failover stop, we set the disk device
2900 read-only while the server shuts down. We now also set any
2901 external journal device read-only at the same time.
2904 Frequency : when upgrading from 1.4 while trying to change parameters
2906 Description: The wrong (new) MDC name was used when setting parameters for
2907 upgraded MDT's. Also allows changing of OSC (and MDC)
2908 parameters if --writeconf is specified at tunefs upgrade time.
2911 Frequency : when setting specific ost indicies
2913 Description: QOS code breaks on skipped indicies
2914 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2915 created with --index need not be sequential.
2917 Severity : enhancement
2919 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2920 Details : The uninit_groups feature works in conjunction with the kernel
2921 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2922 up the pass1 processing of e2fsck. This is a read-only feature
2923 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2924 mount filesystems that have had this feature enabled.
2926 Severity : enhancement
2928 Description: Improve multi-block allocation algorithm to avoid fragmentation
2929 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2930 allocation locality and avoid filesystem fragmentation.
2932 ------------------------------------------------------------------------------
2934 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2936 * Support for kernels:
2937 2.4.21-47.0.1.EL (RHEL 3)
2938 2.6.5-7.283 (SLES 9)
2939 2.6.9-42.0.10.EL (RHEL 4)
2940 2.6.12.6 vanilla (kernel.org)
2941 2.6.16.27-0.9 (SLES 10)
2942 * Recommended e2fsprogs version: 1.39.cfs5
2944 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2948 Frequency : occasional, depends on client load and configuration
2949 Bugzilla : 12181, 12203
2950 Description: data loss for recently-modified files
2952 Details : In some cases it is possible that recently written or created
2953 files may not be written to disk in a timely manner (this should
2954 normally be within 30s unless client IO load is very high).
2955 The problem appears as zero-length files or files that are a
2956 multiple of 1MB in size after a client crash or client eviction
2957 that are missing data at the end of the file.
2959 This problem is more likely to be hit on clients where files are
2960 repeatedly created and unlinked in the same directory, clients
2961 have a large amount of RAM, have many CPUs, the filesystem has
2962 many OSTs, the clients are rebooted frequently, and/or the files
2963 are not accessed by other nodes after being written.
2965 The presence of the problem can be detected by looking at
2966 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2967 smaller than the second (nr_unused) then dirty files will not
2968 be flushed automatically to disk. "sync; sleep 10" should be
2969 run several times on the node before unmounting it to update
2970 Lustre (this is also safe to run on nodes without this problem).
2972 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2973 that can cause this same problem, so customers using that kernel
2974 also need to update the kernel in addition to Lustre. In order
2975 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2977 It is normal that files written just before a client crash (less
2978 than 30s) may not yet have been flushed to disk, even for local
2982 Frequency : frequent on thin XT3 nodes
2984 Description: UUID collision on thin XT3 Linux nodes
2985 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2986 randomly, since we relied on an insufficiently-seeded PRNG.
2991 Description: OSS hangs after "All ost request buffers busy"
2992 Details : A deadlock between quota and journal operations caused OSS
2993 hangs after printing "All ost request buffers busy."
2996 Frequency : always on liblustre builds
2998 Description: Cleanup compiler warnings on liblustre
3001 Frequency : always on liblustre builds on XT3
3003 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3007 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3008 > 32000 subdirectories
3009 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3010 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3011 subdirectory count crosses 32000. This will aid e2fsck to
3012 correctly handle more than 32000 subdirectories.
3015 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3017 Description: System hang on startup
3018 Details : This bug allowed the liblustre (e.g. catamount) client to
3019 return to the app before handling all startup RPCs. This
3020 could leave the node unresponsive to lustre network traffic
3021 and manifested as a server ptllnd timeout.
3023 Severity : enhancement
3025 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3026 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3027 boolean to control whether to print peer debug info when a
3028 client's RPC times out.
3033 Description: make O_SYNC working on 2.6 kernels
3034 Details : 2.6 kernels use different method for mark pages for write,
3035 so need add a code to lustre for O_SYNC work.
3040 Description: Failure to close file and release space on NFS
3041 Details : Put inode details into lock acquired in ll_intent_file_open.
3042 Use mdc_intent_lock in ll_intent_open to properly
3043 detect all kind of errors unhandled by mdc_enqueue.
3048 Description: proc file read during shutdown sometimes raced obd removal,
3050 Details : Add lock to prevent obd access after proc file removal.
3053 Frequency : Only for files larger than 4GB on 32-bit clients.
3055 Description: improperly doing page alignment of locks
3056 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3057 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3062 Description: RPCs being resent when they shouldn't be
3063 Details : Some RPCs that should not be resent are being resent. This
3064 can cause inconsistencies in the RPC state machine. Do not
3065 resend such requests.
3068 Frequency : rare, only with NFS export
3070 Description: Crash on NFS re-export node
3071 Details : under very unusual load conditions an assertion is hit in
3072 ll_intent_file_open()
3075 Frequency : only if OST filesystem is corrupted
3077 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3078 Details : for a short time RPCs with bulk IO are in the replay list,
3079 but replay of bulk IOs is unimplemented. If the OST filesystem
3080 is corrupted due to disk cache incoherency and then replay is
3081 started it is possible to trip an assertion. Avoid putting
3082 committed RPCs into the replay list at all to avoid this issue.
3087 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3088 Details : Large single O_DIRECT read and write calls can fail to allocate
3089 a sufficiently large buffer to process the request. In case of
3090 allocation failure the allocation is retried with a smaller
3091 buffer and broken into smaller requests.
3093 Severity : enhancement
3095 Description: Add -o localflock option to simulate old noflock behaviour.
3096 Details : This will achieve local-only flock/fcntl locks coherentness.
3101 Description: versioning check is incomplete
3102 Details : Checking the version difference of client vs. server, report
3103 error if the gap is too big.
3108 Description: add support PG_writeback bit
3109 Details : add support for PG_writeback bit for Lustre, for more carefull
3110 work with page cache in 2.6 kernel. This also fix some deadlocks
3111 and remove hack for work O_SYNC with 2.6 kernel.
3113 Severity : enhancement
3115 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3116 Details : The uninit_groups feature works in conjunction with the kernel
3117 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3118 up the pass1 processing of e2fsck. This is a read-only feature
3119 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3120 mount filesystems that have had this feature enabled.
3122 Severity : enhancement
3124 Description: Improve multi-block allocation algorithm to avoid fragmentation
3125 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3126 allocation locality and avoid filesystem fragmentation.
3128 ------------------------------------------------------------------------------
3130 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3132 * Support for kernels:
3133 2.6.9-42.0.3.EL (RHEL 4)
3134 2.6.5-7.276 (SLES 9)
3135 2.4.21-47.0.1.EL (RHEL 3)
3136 2.6.12.6 vanilla (kernel.org)
3137 2.6.16.21-0.8 (SLES10)
3138 * Recommended e2fsprogs version: 1.39.cfs2-0
3140 * The backwards-compatible /proc/sys/portals symlink has been removed
3141 in this release. Before upgrading, please ensure that you change
3142 any configuration scripts or /etc/sysctl.conf files that access
3143 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3144 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3145 in advance of the upgrade on any system running Lustre 1.4.6 or
3146 newer, since /proc/sys/lnet was added in that version.
3147 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3151 Frequency : only when quota is used
3153 Description: avoid scanning export list for quota master
3154 Details : Change the algorithms to avoid scanning export list in order
3155 to improve the efficiency.
3158 Frequency : MDS failover only, very rarely
3160 Description: "went back in time" messages on mds failover
3161 Details : The greatest transno may be lost when the current operation
3162 finishes with an error (transno==0) and the client's last_rcvd
3163 record is over-written. Save the greatest transno in the
3164 mds_last_transno for this case.
3167 Frequency : always for specific kernels and striping counts
3169 Description: client may get "Matching packet too big" without ACL support
3170 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3171 when trying to access files in certain configurations. The
3172 clients should in fact be denied when mounting because they do
3173 not understand ACLs.
3176 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3178 Description: many clients connecting with IO in progress causes connect timeouts
3179 Details : Avoid synchronous journal commits to avoid delays caused by many
3180 clients connecting/disconnecting when bulk IO is in progress.
3181 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3182 OST_IO_PORTAL to avoid delays behind potentially many pending
3186 Frequency : occasionally with multiple writers to a single file
3188 Description: shared writes to file may result in wrong size reported by stat()
3189 Details : Allow growing of kms when extent lock is cancelled
3192 Frequency : always with random mmap IO to multi-striped file
3194 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3195 Details : Only if the hole is at the end of OST object so that kms is too
3196 small. Fix is to increase kms accordingly in ll_nopage.
3199 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3201 Description: writes to a missing object would leak memory on the OST
3202 Details : If there is an inconsistency between the MDS and OST filesystems,
3203 such that the MDS references an object that doesn't exist, writes
3204 to that object will leak memory due to incorrect cleanup in the
3205 error handling path, eventually running out of memory on the OST.
3210 Description: Creating too long symlink causes lustre errors
3211 Details : Check symlink and name lengths before sending requests to MDS.
3214 Frequency : only if flock is enabled (not on by default)
3216 Description: posix locks not released on fd closure on 2.6.9+
3217 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3218 kernels, this caused such locks not to be released on fd close and
3219 then assertions on fs unmount about still used locks.
3222 Frequency : MDS failover only, very rarely
3224 Description: clients may get ASSERTION(granted_lock != NULL)
3225 Details : When request was taking a long time, and a client was resending
3226 a getattr by name lock request. The were multiple lock requests
3227 with the same client lock handle and
3228 mds_getattr_name->fixup_handle_for_resent_request found one of the
3229 lock handles but later failed with ASSERTION(granted_lock != NULL).
3234 Description: handle->h_buffer_credits > 0, assertion failure
3235 Details : h_buffer_credits is zero after truncate, causing assertion
3236 failure. This patch extends the transaction or creates a new
3240 Frequency : NFS re-export or patchless client
3241 Bugzilla : 11179, 10796
3242 Description: Crash on NFS re-export node (__d_move)
3243 Details : We do not want to hash the dentry if we don't have a lock.
3244 But if this dentry is later used in d_move, we'd hit uninitialised
3245 list head d_hash, so we just do this to init d_hash field but
3246 leave dentry unhashed.
3249 Frequency : NFS re-export or patchless client
3251 Description: NFS exports has problem with symbolic link
3252 Details : lustre client didn't properly install dentry when re-exported
3253 to NFS or running patchless client.
3256 Frequency : NFS re-export or patchless client
3258 Description: Various nfs/patchless fixes.
3259 Details : fixes reuse disconected alias for lookup process - this fixes
3260 warning "find_exported_dentry: npd != pd",
3261 fix permission error with open files at nfs.
3262 fix apply umask when do revalidate.
3265 Frequency : occasional
3267 Description: Crash on NFS re-export node
3268 Details : calling clear_page() on the wrong pointer triggered oops in
3269 generic_mapping_read().
3272 Frequency : rarely, using O_DIRECT IO
3274 Description: unaligned directio crashes client with LASSERT
3275 Details : check for unaligned buffers before trying any requests.
3278 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3280 Description: stale data returned from RAID cache
3281 Details : If only a small amount of IO is done to the RAID device before
3282 reading it again it is possible to get stale data from the RAID
3283 cache instead of reading it from disk.
3286 Frequency : always for sles10 kernel
3288 Description: sles10 support
3289 Details : ll_follow_link: compile fixes and using of nd_set_link
3290 under newer kernels.
3293 Frequency : depends on arch, kernel and compiler version, always on sles10
3296 Description: recursive or deep enough symlinks cause stack overflow
3297 Details : getting rid of large stack-allocated variable in
3301 Frequency : depends on hardware
3303 Description: lustre write performance loss in the SLES10 kernel
3304 Details : the performance loss is caused by using of write barriers in the
3305 ext3 code. The SLES10 kernel turns barrier support on by
3306 default. The fix is to undo that change for ldiskfs.
3308 ------------------------------------------------------------------------------
3310 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3312 * Support for kernels:
3313 2.6.9-42.0.3EL (RHEL 4)
3314 2.6.5-7.276 (SLES 9)
3315 2.4.21-47.0.1.EL (RHEL 3)
3316 2.6.12.6 vanilla (kernel.org)
3320 Frequency : quota enabled and large files being deleted
3322 Description: releasing more than 4GB of quota at once hangs OST
3323 Details : If a user deletes more than 4GB of files on a single OST it
3324 will cause the OST to spin in an infinite loop. Release
3325 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3330 Description: statfs data retrieved from /proc may be stale or zero
3331 Details : When reading per-device statfs data from /proc, in the
3332 {kbytes,files}_{total,free,avail} files, it may appear
3333 as zero or be out of date.
3336 Frequency : systems with MD RAID1 external journal devices
3338 Description: lconf's call to blkid is confused by RAID1 journal devices
3339 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3340 of returning all block devices that match the journal UUID.
3343 Frequency : always, for aggregate stripe size over 4GB
3345 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3346 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3347 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3350 Frequency : always if "lfs find" used on a local file/directory
3352 Description: "lfs find" segfaults if used on a local file/directory
3353 Details : The case where a directory component was not specified wasn't
3354 handled correctly. Handle this properly.
3357 Frequency : always on ppc64
3359 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3360 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3361 perform endianness conversion.
3364 Frequency : rarely (truncate to non-zero file size after write under load)
3365 Bugzilla : 10730, 10687
3366 Description: Files padded with zeros to next 4K multiple
3367 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3368 it is possible that files that are truncated to a non-zero size
3369 immediately after being written are filled with zero bytes beyond
3370 the truncated size. No file data is lost.
3372 Severity : enhancement
3374 Description: Allow recovery/failover for liblustre clients.
3375 Details : liblustre clients were unaware of failover configurations until
3378 Severity : enhancement
3380 Description: user file locks should fail when not mounting with flock option
3381 Details : Set up an error-returning stub in ll_file_operations.lock field
3382 to prevent incorrect behaviour when client is mounted without
3383 flock option. Also, set up properly f_op->flock field for
3387 Frequency : always on ia64
3389 Description: "lfs df" loops on printing out MDS statfs information
3390 Details : The obd_ioctl_data was not initialized and in some systems
3391 this caused a failure during the ioctl that did not return
3392 an error. Initialize the struct and return an error on failure.
3395 Frequency : SLES 9 only
3397 Description: Error of copying files with lustre special EAs as root
3398 Details : Client side always return success for setxattr call for lustre
3399 special xattr (currently only "trusted.lov").
3402 Frequency : rarely on clusters with both ia64+i386 clients
3404 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3405 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3406 clients are doing concurrent writes to the same file it is
3407 possible that the ia64 clients may LASSERT because the OST
3408 extent locks are not PAGE_SIZE aligned. Ensure that grown
3409 locks are always aligned on the request boundary.
3412 Frequency : specific use, occasional
3414 Description: Overwriting in use executable truncates on-disk binary image
3415 Details : If one node attempts to overwrite an executable in use by
3416 another node, we now correctly return ETXTBSY instead of
3417 truncating the file.
3419 Severity : enhancement
3421 Description: Async OSC create to avoid the blocking unnecessarily.
3422 Details : If a OST has no remain object, system will block on the creating
3423 when need to create a new object on this OST. Now, ways use
3424 pre-created objects when available, instead of blocking on an
3425 empty osc while others are not empty. If we must block, we block
3426 for the shortest possible period of time.
3431 Description: chmod on Lustre root is propagated to other clients
3432 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3433 invalid by the follow_mount time.
3436 Frequency : liblustre clients only
3438 Description: Race in 'instant cancel' lock handling could lead to such locks
3439 never to be granted in case of SMP MDS
3440 Details : Do not destroy not yet granted but cbpending locks in
3444 Frequency : replay/resend of open
3446 Description: non null lock assetion failure in mds_intent_policy
3447 Details : Trying to replay/resend lockless open requests resulted in
3448 mds_open() returning 0 with no lock. Now it sets a flag if
3449 a lock is going to be returned.
3451 Severity : enhancement
3453 Description: Checksum enhancements
3454 Details : New checksum enhancements allow for resending RPCs that failed
3457 Severity : enhancement
3459 Description: Tunables on number of dirty pages in cacche
3460 Details : Allow to set limit on number of dirty pages cached.
3465 Description: client crash on unmount - lock still has references
3466 Details : In some error handling cases it was possible to leak a lock
3467 reference on a client while accessing a file. This was not
3468 harmful to the client during operation, but would cause the
3469 client to crash when the filesystem is unmounted.
3472 Frequency : specific case, rare
3474 Description: ETXTBSY on mds though file not in use by client
3475 Details : ETXTBSY is no longer incorrectly returned when attempting to
3476 chmod or chown a directory that the user previously tried to
3477 execute or a currently-executing binary.
3480 Frequency : extremely rare except on liblustre-based clients
3482 Description: Lustre space not freed when files are deleted
3483 Details : Clean up open-unlinked files after client eviction. Previously
3484 the unlink was skipped and the files remained as orphans.
3489 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3490 Details : In some cases it was possible to send a blocking callback to
3491 a client doing a glimpse, even though that client didn't get
3492 a lock granted. When the glimpse lock is cancelled on the OST
3493 the freed lock is left on the waiting list and corrupted the list.
3496 Frequency : all core dumps
3498 Description: Broke core dumps to lustre
3499 Details : Negative dentry may be unhashed if parent does not have UPDATE
3500 lock, but some callers, e.g. do_coredump, expect dentry to be
3501 hashed after successful create, hash it in ll_create_it.
3503 ------------------------------------------------------------------------------
3505 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3507 * Support for kernels:
3508 2.6.9-42.0.2.EL (RHEL 4)
3509 2.6.5-7.276 (SLES 9)
3510 2.4.21-40.EL (RHEL 3)
3511 2.6.12.6 vanilla (kernel.org)
3515 Frequency : always on RHEL 3
3517 Description: Number of open files grows over time
3518 Details : The number of open files grows over time, whether or not
3519 Lustre is started. This was due to a filp leak introduced
3520 by one of our kernel patches.
3522 ------------------------------------------------------------------------------
3524 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3526 * Support for kernels:
3527 2.6.9-42.EL (RHEL 4)
3528 2.6.5-7.276 (SLES 9)
3529 2.4.21-40.EL (RHEL 3)
3530 2.6.12.6 vanilla (kernel.org)
3535 Bugzilla : 5719, 9635, 9792, 9684
3536 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3537 Details : If a server is under heavy load and cannot reply to new
3538 connection requests before the client resends the (re)connect,
3539 the connection handling code can behave badly if two service
3540 threads are concurrently handing separate (re)connections from
3541 the same client. Add better locking to the connection handling
3542 code, and ensure that only a single connection will be processed
3543 for a given client UUID, even if the lock is dropped.
3545 Severity : enhancement
3547 Description: add TCP zero-copy support to kernel
3548 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3549 sends if the hardware supports scatter-gather and checksumming.
3550 This allows socklnd to do client-write and server-read more
3551 efficiently and reduce CPU utilization from skbuf copying.
3554 Frequency : only if NFS exporting from client
3556 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3557 Details : If setting attributes on a file created under NFS that had
3558 never been opened it would be possible to oops the client
3559 if the file had no objects.
3562 Frequency : always for liblustre
3564 Description: liblustre client does MDS+OSTs setattr RPC for each write
3565 Details : When doing a write from a liblustre client, the client
3566 incorrectly issued an RPC to the MDS and each OST the file was
3567 striped over in order to update the timestamps. When writing
3568 with small chunks and many clients this could overwhelm the MDS
3569 with RPCs. In all cases it would slow down the write because
3570 these RPCs are unnecessary.
3572 Severity : enhancement
3574 Description: allow number of MDS service threads to be changed at module load
3575 Details : It is now possible to change the number of MDS service threads
3576 running. Adding "options mds mds_num_threads={N}" to the MDS's
3577 /etc/modprobe.conf will set the number of threads for the next
3578 time Lustre is restarted (assuming the "mds" module is also
3579 reloaded at that time). The default number of threads will
3580 stay the same, 32 for most systems.
3585 Description: OST crash if filesystem is unformatted or corrupt
3586 Details : If an OST is started on a device that has never been formatted
3587 or if the filesystem is corrupt and cannot even mount then the
3588 error handling cleanup routines would dereference a NULL pointer.
3593 Description: NULL pointer deref in llap_from_page.
3594 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3595 incorrect) mapping if the page was truncated/reclaimed while it was
3596 searched for. Check for this condition and skip such pages when
3597 doing readahead. Introduce extra check to llap_from_page() to
3598 verify page->mapping->host is non-NULL (so page is not anonymous).
3601 Frequency : Sometimes when using sys_sendfile
3603 Description: "page not covered by a lock" warnings from ll_readpage
3604 Details : sendfile called ll_readpage without right page locks present.
3605 Now we introduced ll_file_sendfile that does necessary locking
3606 around call to generic_file_sendfile() much like we do in
3610 Frequency : with certain MDS communication failures at client mount time
3612 Description: NULL pointer deref after failed client mount
3613 Details : a client connection request may delayed by the network layer
3614 and not be sent until after the PTLRPC layer has timed out the
3615 request. If the client fails the mount immediately it will try
3616 to clean up before the network times out the request. Add a
3617 reference from the request import to the obd device and delay
3618 the cleanup until the network drops the request.
3621 Frequency : occasionally during client (re)connect
3623 Description: assertion failure during client (re)connect
3624 Details : processing a client connection request may be delayed by the
3625 client or server longer than the client connect timeout. This
3626 causes the client to resend the connection request. If the
3627 original connection request is replied in this interval, the
3628 client may trip an assertion failure in ptlrpc_connect_interpret()
3629 which thought it would be the only running connect process.
3632 Frequency : only with obd_echo servers and clients that are rebooted
3634 Description: kernel BUG accessing uninitialized data structure
3635 Details : When running an obd_echo server it did not start the ping_evictor
3636 thread, and when a client was evicted an uninitialized data
3637 structure was accessed. Start the ping_evictor in the RPC
3638 service startup instead of the OBD startup.
3640 Severity : enhancement
3641 Bugzilla : 10193 (patchless)
3642 Description: Remove dependency on various unexported kernel interfaces.
3643 Details : No longer need reparent_to_init, exit_mm, exit_files,
3644 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3647 Frequency : rare (only users of deprecated and unsupported LDAP config)
3649 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3650 Details : LDAP apparently contains 'lustreName' attributes instead of
3651 'name'. A simple remapping of the name is sufficient.
3654 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3656 Description: waiting_locks_callback trips kernel BUG if client is evicted
3657 Details : Running with the dump_on_timeout debug flag turned on makes
3658 it possible that the waiting_locks_callback() can try to dump
3659 the Lustre kernel debug logs from an interrupt handler. Defer
3660 this log dumping to the expired_lock_main() thread.
3662 Severity : enhancement
3664 Description: Support NFS exporting on 2.6 kernels.
3665 Details : Implement non-rawops metadata methods for NFS server to use without
3666 changing NFS server code.
3669 Frequency : very rare (synthetic metadata workload only)
3671 Description: two racing renames might cause an MDS thread to deadlock
3672 Details : Running the "racer" program may cause one MDS thread to rename
3673 a file from being the source of a rename to being the target of
3674 a rename at exactly the same time that another thread is doing
3675 so, and the second thread has already enqueued these locks after
3676 doing a lookup of the target and is trying to relock them in
3677 order. Ensure that we don't try to re-lock the same resource.
3680 Frequency : only very large systems with liblustre clients
3682 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3683 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3684 all OSTs in parallel. This allows the eviction of stale liblustre
3685 clients to proceed much faster than if they were done in series,
3686 and also offers similar improvements for other set_info RPCs.
3691 Description: excessive CPU usage during initial read phase on client
3692 Details : During the initial read phase on a client, it would agressively
3693 retry readahead on the file, consuming too much CPU and impacting
3694 performance (since 1.4.5.8). Improve the readahead algorithm
3695 to avoid this, and also improve some other common cases (read
3696 of small files in particular, where "small" is files smaller than
3697 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3702 Description: MDS crash when receiving packet with unknown intent.
3703 Details : Do not LBUG in unknown intent case, just return -EFAULT
3705 Severity : enhancement
3706 Bugzilla : 9293, 9385
3707 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3708 Details : Do not serialize getattr (non-intent version) and statfs.
3711 Frequency : occasional, when OST network is overloaded/intermittent
3713 Description: client evicted by OST after bulk IO timeout
3714 Details : If a client sends a bulk IO request (read or write) the OST
3715 may evict the client if it is unresposive to its data GET/PUT
3716 request. This is incorrect if the network is overloaded (takes
3717 too long to transfer the RPC data) or dropped the OST GET/PUT
3718 request. There is no need to evict the client at all, since
3719 the pinger and/or lock callbacks will handle this, and the
3720 client can restart the bulk request.
3723 Frequency : Always when mmapping file with no objects
3725 Description: client crashes when mmapping file with no objects
3726 Details : Check that we actually have objects in a file before doing any
3727 operations on objects in ll_vm_open, ll_vm_close and
3733 Description: Request leak when working with deleted CWD
3734 Details : Introduce advanced request refcount tracking for requests
3735 referenced from lustre intent.
3737 Severity : Enhancement
3739 Description: Cache open file handles on client.
3740 Details : MDS now will return special lock along with openhandle, if
3741 requested and client is allowed to hold openhandle, even if unused,
3742 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3743 closing files for every read/write openration.
3745 Severity : Enhancement
3747 Description: Cache open negative dentries on client when possible.
3748 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3749 negative dentries on lock revocation.
3754 Description: Remounting a client read-only wasn't possible with a zconf mount
3755 Details : It wasn't possible to remount a client read-only with llmount.
3757 Severity : enhancement
3758 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3759 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3760 MPICH is included to add Lustre-specific ADIO interfaces.
3761 This is based closely on the UFS ADIO layer and only differs
3762 in file creation, in order to allow the OST striping to be set.
3763 This is user-contributed code and not supported by CFS.
3768 Description: extended inode attributes (immutable, append-only) work improperly
3769 when 2.4 and 2.6 kernels are used on client/server or vice versa
3770 Details : Introduce kernel-independent values for these flags.
3772 Severity : enhancement
3775 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3776 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3777 granularity. Now, Fractional MB granularity is be supported,
3778 this is very useful for low memory system.
3780 Severity : enhancement
3782 Description: Getattr by fid
3783 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3784 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3785 have locks and avoids getattr rpc requests for it on every stat.
3788 Frequency : Always, for filesystems larger than 2TB
3790 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3791 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3792 filesystems larger than 2TB to be mounted with mballoc enabled.
3795 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3797 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3798 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3799 filesystems larger than 2TB is mounted then the kernel will
3800 silently corrupt the start of the filesystem. CONFIG_LBD is
3801 enabled for all CFS-supported kernels, but the possibility of
3802 this happening with a modified kernel config exists.
3804 Severity : enhancement
3806 Description: add client O_DIRECT support for 2.6 kernels
3807 Details : It is now possible to do O_DIRECT reads and writes to files
3808 in the Lustre client mountpoint on 2.6 kernel clients.
3810 Severity : enhancement
3812 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3813 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3814 parallel, not waiting for response from every OST before sending
3815 a rpc to the next OST.
3820 Description: setattr vs write race when updating file timestamps
3821 Details : Client processes that update a file timestamp into the past
3822 right after writing to the file (e.g. tar) it is possible that
3823 the updated file modification time can be reset to the current
3824 time due to a race between processing the setattr and write RPC.
3826 Severity : enhancement
3828 Description: Bring 'lfs find' closer in line with regular Linux find.
3829 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3830 -print0 options and obtains all the needed info through the lustre
3833 Severity : enhancement
3835 Description: support up to 1024 configured devices on one node
3836 Details : change obd_dev array from statically allocated to dynamically
3837 allocated structs as they are first used to reduce memory usage
3842 Description: Flush dirty partially truncated pages during truncate
3843 Details : Immediatelly flush partially truncated pages in filter_setattr,
3844 this way we completely avoid having any pages in page cache on OST
3845 and can retire ugly workarounds during writes to flush such pages.
3850 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3851 Details : protect inode from truncation within vfs_unlink() context
3852 just take a reference before calling vfs_unlink() and release it
3853 when parent's i_sem is free.
3858 Description: last_id value checked outside lock on OST caused LASSERT failure
3859 Details : If there were multiple MDS->OST object precreate requests in
3860 flight, it was possible that the OST's last object id was checked
3861 outside a lock and incorrectly tripped an assertion. Move checks
3862 inside locks, and discard old precreate requests.
3865 Frequency : always, if extents are used on OSTs
3867 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3868 Details : OSTs using the extents format would not zero the high 16 bits of
3869 the index physical block number. This is not a problem for any
3870 OST filesystems smaller than 16TB, and no kernels support ext3
3871 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3872 new/modified files) and can be fixed for existing filesystems
3873 with e2fsprogs-1.39-cfs1.
3878 Description: import connection selection may be incorrect if timer wraps
3879 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3880 import connections to be ignored if the 32-bit jiffies counter
3881 wraps. Use a 64-bit jiffies counter.
3884 Frequency : very large clusters immediately after boot
3886 Description: LNET request buffers exhausted under heavy short-term load
3887 Details : If a large number of client requests are generated on a service
3888 that has previously never seen so many requests it is possible
3889 that the request buffer growth cannot keep up with the spike in
3890 demand. Instead of dropping incoming requests, they are held in
3891 the LND until the RPC service can accept more requests.
3894 Frequency : Sometimes during replay
3896 Description: Assertion failure in ll_local_open after replay.
3897 Details : If replay happened on an open request reply before we were able
3898 to set replay handler, reply will become not swabbed tripping the
3899 assertion in ll_local_open. Now we set the handler right after
3900 recognising of open request
3903 Frequency : very rare
3905 Description: kernel reports "badness in vsnprintf"
3906 Details : Reading from the "recovery_status" /proc file in small chunks
3907 may cause a negative length in lprocfs_obd_rd_recovery_status()
3908 call to vsnprintf() (which is otherwise harmless). Exit early
3909 if there is no more space in the output buffer.
3911 Severity : enhancement
3913 Description: clear OBD RPC statistics by writing to them
3914 Details : It is now possible to clear the OBD RPC statistics by writing
3915 to the "stats" file.
3920 Description: Client mtime is not the same on different clients after utimes
3921 Details : In some cases, the client was using the utimes() syscall on
3922 a file cached on another node. The clients now validate the
3923 ctime from the MDS + OSTs to determine which one is right.
3928 Description: Inability to activate failout mode
3929 Details : lconf script incorrectly assumed that in pythong string's numeric
3930 value is used in comparisons.
3933 Frequency : always with multiple stripes per file
3935 Description: Inefficient object allocation for mutli-stripe files
3936 Details : When selecting which OSTs to stripe files over, for files with
3937 a stripe count that divides evenly into the number of OSTs,
3938 the MDS is always picking the same starting OST for each file.
3939 Return the OST selection heuristic to the original design.
3944 Description: mount failures may take full timeout to return an error
3945 Details : Under some heavy load conditions it is possible that a
3946 failed mount can wait for the full obd_timeout interval,
3947 possibly several minutes, before reporting an error.
3948 Instead return an error as soon as the status is known.
3950 Frequency : quota enabled and large files being deleted
3952 Description: releasing more than 4GB of quota at once hangs OST
3953 Details : If a user deletes more than 4GB of files on a single OST it
3954 will cause the OST to spin in an infinite loop. Release
3955 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3960 Description: statfs data retrieved from /proc may be stale or zero
3961 Details : When reading per-device statfs data from /proc, in the
3962 {kbytes,files}_{total,free,avail} files, it may appear
3963 as zero or be out of date.
3966 Frequency : systems with MD RAID1 external journal devices
3968 Description: lconf's call to blkid is confused by RAID1 journal devices
3969 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3970 of returning all block devices that match the journal UUID.
3973 Frequency : always, for aggregate stripe size over 4GB
3975 Description: assertion fails when trying to use 4GB stripe size
3976 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3977 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3980 Frequency : always on ppc64
3982 Description: the first write on an ext3 filesystem with mballoc got stuck
3983 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3984 perform endianness conversion.
3986 ------------------------------------------------------------------------------
3988 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3990 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3991 INTEROPERATE with older versions automatically. Please read the
3992 user documentation before upgrading any part of a live system.
3993 * WARNING: Lustre networking configuration changes are required with
3994 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3997 * Support for kernels:
3998 2.6.9-22.0.2.EL (RHEL 4)
3999 2.6.5-7.244 (SLES 9)
4000 2.6.12.6 vanilla (kernel.org)
4003 Severity : enhancement
4004 Bugzilla : 7981/8208
4005 Description: Introduced Lustre Networking (LNET)
4006 Details : LNET is new networking infrastructure for Lustre, it includes
4007 a reorganized network configuration mode (see the user
4008 documentation for full details) as well as support for routing
4009 between different network fabrics. Lustre Networking Devices
4010 (LNDs) for the supported network fabrics have also been
4011 created for this new infrastructure.
4013 Severity : enhancement
4014 Description: Introduced Access control lists
4015 Details : clients can set ACLs on files and directories in order to have
4016 more fine-grained permissions than the standard Unix UGO+RWX.
4017 The MDS must be started with the "-o acl" mount option.
4019 Severity : enhancement
4020 Description: Introduced filesystem quotas
4021 Details : Administrators may now establish per-user quotas on the
4024 Severity : enhancement
4026 Description: Configuration change for the XT3
4027 The PTLLND is now used to run Lustre over Portals on the XT3
4028 The configure option(s) --with-cray-portals are no longer used.
4029 Rather --with-portals=<path-to-portals-includes> is used to
4030 enable building on the XT3. In addition to enable XT3 specific
4031 features the option --enable-cray-xt3 must be used.
4036 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4037 Details : A race between allocating a new llcd and re-getting the llcd_lock
4038 allowed another thread to grab newly-allocated llcd.
4040 Severity : enhancement
4042 Description: 2.6 OST async journal commit and locking fix to improve performance
4043 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4044 2.6 kernels are now async as they already were in 2.4 kernels so
4045 that they can commit concurrently with the network bulk transfer.
4046 For block-allocated files the filter allocation semaphore is held
4047 to avoid filesystem fragmentation during allocation. BKL lock
4048 removed for 2.6 xattr operations where it is no longer needed.
4053 Description: lconf incorrectly determined whether two IP networks could talk
4054 Details : In some more complicated routing and multiple-network
4055 configurations, lconf will avoid trying to make a network
4056 connection to a disjoint part of the IP space. It was doing the
4057 math incorrectly for one set of cases.
4062 Description: Fix for potential infinite loop processing records in an llog.
4063 Details : If an llog record is corrupted/zeroed, it is possible to loop
4064 forever in llog_process(). Validate the llog record length
4065 and skip the remainder of the block on error.
4068 Frequency : occasional (liblustre only)
4070 Description: liblustre could not open files whose last component is a symlink
4071 Details : sysio_path_walk() would incorrectly pass the open intent to
4072 intermediate path components.
4075 Frequency : rare (liblustre only with non-standard tuning)
4076 Bugzilla : 7201 (7350)
4077 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4078 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4079 after it has been released, only do this for referenced locks
4081 Severity : enhancement
4083 Description: specifying an (invalid) directory default stripe_size of -1
4084 would reset the directory default striping
4085 Details : stripe_size -1 was used internally to signal directory stripe
4086 removal, now use "all default" to signal dir stripe removal
4087 as a directory striping of "all default" is not useful
4090 Frequency : common for large clusters running liblustre clients
4092 Description: doing an ls when liblustre clients are running is slow
4093 Details : sending a glimpse AST to a liblustre client waits for every AST
4094 to time out, as liblustre clients will not respond. Since they
4095 cannot cache data we refresh the OST lock LVB from disk instead.
4097 Severity : enhancement
4099 Description: doing an ls at the same time as file IO can be slow
4100 Details : enqueue and other "small" requests can be blocked behind many
4101 large IO requests. Create a new OST IO portal for non-IO
4102 requests so they can be processed faster.
4105 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4107 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4108 created file with mode 000
4109 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4110 allow an ftruncate() as a non-root user to a file with mode 000.
4111 Lustre now respects this flag to disable mode checks when
4112 truncating a file owned by the user
4115 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4118 Description: Revoking locks from clients that went dead or catatonic might take
4120 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4121 cancellation of such locks instant on servers without waiting for
4122 any reply from clients. Clients drops these locks when cancel
4123 notification from server is received without replying.
4126 Frequency : liblustre-only, when liblustre client dies or becomes busy
4128 Description: Doing ls on Linux clients can take a long time with active
4130 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4131 granting such locks to it in the first place if possible. Locks
4132 are taken by proxy on the OST during the read or write and
4133 dropped immediately afterward. Add connect flags handling, do
4134 not grant locks to liblustre clients for glimpse ASTs.
4136 Severity : enhancement
4138 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4139 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4140 sequential reads and behaves badly with applications doing random
4141 reads. Improve it by reading ahead at least read region, and
4142 avoiding excessive large RPC for small reads.
4144 Severity : enhancement
4146 Description: Creating more than 1000 files for a single job may cause a load
4147 imbalance on the OSTs if there are also a large number of OSTs.
4148 Details : qos_prep_create() uses an OST index reseed value that is an
4149 even multiple of the number of available OSTs so that if the
4150 reseed happens in the middle of the object allocation it will
4151 still utilize the OSTs as uniformly as possible.
4156 Description: OST or MDS may oops in ping_evictor_main()
4157 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4158 but doesn't restart at beginning of obd_exports_timed list
4161 Severity : enhancement
4163 Description: improve by-nid export eviction on the MDS and OST
4164 Details : allow multiple exports with the same NID to be evicted at one
4165 time without re-searching the exports list.
4168 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4170 Description: MDS may oops in groups_free()
4171 Details : in rare race conditions a newly allocated group_info struct is
4172 freed again, and this can be NULL. The 2.4 compatibility code
4173 for groups_free() checked for a NULL pointer, but 2.6 did not.
4176 Frequency : common for liblustre clients doing little filesystem IO
4177 Bugzilla : 9352, 7313
4178 Description: server may evict liblustre clients accessing contended locks
4179 Details : if a client is granted a lock or receives a completion AST
4180 with a blocking AST already set it would not reply to the AST
4181 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4184 Frequency : lfs setstripe, only systems with more than 160 OSTs
4186 Description: unable to set striping with a starting offset beyond OST 160
4187 Details : llapi_create_file() incorrectly limited the starting stripe
4188 index to the maximum single-file stripe count.
4191 Frequency : LDAP users only
4193 Description: lconf did not handle in-kernel recovery with LDAP properly
4194 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4196 Severity : enhancement
4198 Description: bind OST threads to NUMA nodes to improve performance
4199 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4200 node and do their allocations there to localize memory access
4202 Severity : enhancement
4204 Description: llmount can determine client NID directly from Myrinet (GM)
4205 Details : the client NID code from gmnalnid was moved directly into
4206 llmount, removing the need to use this or specifying the
4207 client NID explicitly when mounting GM clients with zeroconf
4210 Frequency : if client is started with down MDS
4212 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4213 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4214 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4215 ptlrpc_import_delayed_req() because we didn't check if the
4216 request was interrupted, and we also didn't break out of the
4217 event loop if there was no timeout
4222 Description: data loss during non-page-aligned writes to a single file from
4223 both multiple nodes and multiple threads on one node at same time
4224 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4225 inconsistency led to false short-reads, that were cached and later
4226 used by ->prepare_write() to fill in partially written page,
4227 leading to data loss.
4230 Frequency : always, if lconf --abort_recovery used
4232 Description: lconf --abort_recovery fails with 'Operation not supported'
4233 Details : lconf was attempting to abort recovery on the MDT device and not
4236 Severity : enhancement
4238 Description: remove cleanup logs
4239 Details : replace lconf-generated cleanup logs with lustre internal
4240 cleanup routines. Eliminates the need for client-cleanup and
4243 Severity : enhancement
4245 Description: add support for EAs (user and system) on lustre filesystems
4246 Details : it is now possible to store extended attributes in the Lustre
4247 client filesystem, and with the user_xattr mount option it
4248 is possible to allow users to store EAs on their files also
4250 Severity : enhancement
4252 Description: Add possibility (config option) to show minimal available OST free
4254 Details : When compiled with --enable-mindf configure option, statfs(2)
4255 (and so, df) will return least minimal free space available from
4256 all OSTs as amount of free space on FS, instead of summary of
4257 free spaces of all OSTs.
4259 Severity : enhancement
4261 Description: do not expand extent locks acquired on OST-side
4262 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4263 by server: they are not cached anyway.
4266 Frequency : when mmap is used/binaries executed from Lustre
4268 Description: Unmmap pages before throwing them away from read cache.
4269 Details : llap_shrink cache now attempts to unmap pages before discarding
4270 them (if unmapping failed - do not discard). SLES9 kernel has
4271 extra checks that trigger if this unmapping is not done first.
4276 Description: lconf didn't resolve symlinks before checking to see whether a
4277 given mountpoint was already in use
4280 Frequency : when migrating failover services
4281 Bugzilla : 6395, 9514
4282 Description: When migrating a subset of services from a node (e.g. failback
4283 from a failover service node) the remaining services would
4284 time out and evict clients.
4285 Details : lconf --force (implied by --failover) sets the global obd_timeout
4286 to 5 seconds in order to quickly disconnect, but this caused
4287 other RPCs to time out too quickly. Do not change the global
4288 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4290 Severity : enhancement
4291 Frequency : if MDS is started with down OST
4292 Bugzilla : 9439,5706
4293 Description: Allow startup/shutdown of an MDS without depending on the
4294 availability of the OSTs.
4295 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4296 Add appropriate locking and lov-osc refcounts for safe
4297 cleaning. Add osc abort_inflight calls in case the
4298 synchronize never started.
4301 Frequency : occasional (Cray XT3 only)
4303 Description: root not authorized to access files in CRAY_PORTALS environment
4304 Details : The client process capabilities were not honoured on the MDS in
4305 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4306 been accessed by an authorized user then root was able to access
4307 the file on the local client also. The root user capabilities
4308 are now allowed on the MDS, as this environment has secure UID.
4311 Frequency : occasional
4313 Description: ldiskfs "too long searching" message happens too often
4314 Details : A debugging message (otherwise harmless) prints too often on
4315 the OST console. This has been reduced to only happen when
4316 there are fragmentation problems on the filesystem.
4321 Description: Division by zero in statfs when all OSCs are inactive
4322 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4323 lov_statfs divides by value returned by lov_get_stripecnt().
4327 Bugzilla : 9489, 3273
4328 Description: First write from each client to each OST was only 4kB in size,
4329 to initialize client writeback cache, which caused sub-optimal
4330 RPCs and poor layout on disk for the first writen file.
4331 Details : Clients now request an initial cache grant at (re)connect time
4332 and so that they can start streaming writes to the cache right
4333 away and always do full-sized RPCs if there is enough data.
4334 If the OST is rebooted the client also re-establishes its grant
4335 so that client cached writes will be honoured under the grant.
4340 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4341 Details : Now I/O RPCs go to different portal number and (presumably) fast
4342 lock requests (and glimses) and other RPCs get their own service
4343 threads pool that should be able to service those RPCs
4346 Severity : enhancement
4348 Description: Ability to exchange lustre version between client and servers and
4349 issue warnings at client side if client is too old. Also for
4350 liblustre clients there is ability to refuse connection of too old
4352 Details : New 'version' field is added to connect data structure that is
4353 filled with version info. That info is later checked by server and
4357 Frequency : rare, liblustre only.
4358 Bugzilla : 9296, 9581
4359 Description: Two simultaneous writes from liblustre at offset within same page
4360 might proceed at the same time overwriting eachother with stale
4362 Details : I/O lock withing llu_file_prwv was released too early, before data
4363 actually was hitting the wire. Extended lock-holding time until
4364 server acknowledges receiving data.
4367 Frequency : extremely rare. Never observed in practice.
4369 Description: avoid generating lustre_handle cookie of 0.
4370 Details : class_handle_hash() generates handle cookies by incrementing
4371 global counter, and can hit 0 occasionaly (this is unlikely, but
4372 not impossible, because initial value of cookie counter is
4373 selected randonly). Value of 0 is used as a sentinel meaning
4374 "unassigned handle" --- avoid it. Also coalesce two critical
4375 sections in this function into one.
4377 Severity : enhancement
4379 Description: allow liblustre clients to delegate truncate locking to OST
4380 Details : To avoid overhead of locking, liblustre client instructs OST to
4381 take extent lock in ost_punch() on client's behalf. New connection
4382 flag is added to handle backward compatibility.
4384 Severity : enhancement
4385 Bugzilla : 4928, 7341, 9758
4386 Description: allow number of OST service threads to be specified
4387 Details : a module parameter allows the number of OST service threads
4388 to be specified via "options ost ost_num_threads={N}" in the
4389 OSS's /etc/modules.conf or /etc/modprobe.conf.
4393 Bugzilla : 6146, 9635, 9895
4394 Description: servers crash with bad pointer in target_handle_connect()
4395 Details : In rare cases when a client is reconnecting it was possible that
4396 the connection request was the last reference for that export.
4397 We would temporarily drop the export reference and get a new
4398 one, but this may have been the last reference and the export
4399 was just destroyed. Get new reference before dropping old one.
4401 Severity : enhancement
4402 Frequency : if client is started with failover MDS
4404 Description: Allow multiple MDS hostnames in the mount command
4405 Details : Try to read the configuration from all specified MDS
4406 hostnames during a client mount in case the "primary"
4409 Severity : enhancement
4411 Description: Stop sending data to evicted clients as soon as possible.
4412 Details : Check if the client we are about to send or are sending data to
4413 was evicted already. (Check is done every second of waiting,
4414 for which l_wait_event interface was extended to allow checking
4415 of exit condition at specified intervals).
4418 Frequency : rare, normally only when NFS exporting is done from client
4420 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4422 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4425 Frequency : always, if srand() is called before liblustre initialization
4427 Description: Liblustre uses system PRNG disturbing its usage by user application
4428 Details : Introduce internal to lustre fast and high-quality PRNG for
4429 lustre usage and make liblustre and some other places in generic
4430 lustre code to use it.
4432 Severity : enhancement
4433 Bugzilla : 9477, 9557, 9870
4434 Description: Verify that the MDS configuration logs are updated when xml is
4435 Details : Check if the .xml configuration logs are newer than the config
4436 logs stored on the MDS and report an error if this is the case.
4437 Request --write-conf, or allow starting with --old_conf.
4439 Severity : enhancement
4441 Description: Handle symlinks in the path when checking if Lustre is mounted.
4442 Details : Resolve intermediate symlinks when checking if a client has
4443 mounted a filesystem to avoid duplicate client mounts.
4448 Description: lconf can hit an error exception but still return success.
4449 Details : The lconf command catches the Command error exception at the top
4450 level script context and will exit with the associated exit
4451 status, but doesn't ensure that this exit status is non-zero.
4456 Description: failure of ptlrpc thread startup can cause oops
4457 Details : Starting a ptlrpc service thread can fail if there are a large
4458 number of threads or the server memory is very fragmented.
4459 Handle this without oopsing.
4462 Frequency : always, only if liblustre and non-default acceptor port was used
4464 Description: liblustre cannot connect to servers with non-default acceptor port
4465 Details : tcpnal_set_default_params() was not called and was therefore
4466 ignoring the environment varaible TCPNAL_PORT, as well as other
4467 TCPNAL_ environment variables
4472 Description: two objects could be created on the same OST for a single file
4473 Details : If an OST is down, in some cases it was possible to create two
4474 objects on a single OST for a single file. No problems other
4475 than potential performance impact and spurious error messages.
4479 Bugzilla : 5681, 9562
4480 Description: Client may oops in ll_unhash_aliases
4481 Details : Client dcache may become inconsistent in race condition.
4482 In some cases "getcwd" can fail if the current directory is
4488 Description: Inode refcounting problems in NFS export code
4489 Details : link_raw functions used to call d_instantiate without obtaining
4490 extra inode reference first.
4494 Bugzilla : 9942, 9903
4495 Description: Referencing freed requests leading to crash, memleaks with NFS.
4496 Details : We used to require that call to ll_revalidate_it was always
4497 followed by ll_lookup_it. Also with revalidate_special() it is
4498 possible to call ll_revalidate_it() twice for the same dentry
4499 even if first occurence returned success. This fix changes semantic
4500 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4501 reference on a request referred from the intent.
4502 ll_intent_release() then releases such a request.
4505 Frequency : rare, normally benchmark loads only
4507 Description: unlinked inodes were kept in memory on the client
4508 Details : If a client is repeatedly creating and unlinking files it
4509 can accumulate a lot of stale inodes in the inode slab cache.
4510 If there is no other client load running this can cause the
4511 client node to run out of memory. Instead flush old inodes
4512 from client cache that have the same inode number as a new inode.
4515 Frequency : SLES9 2.6.5 kernel and long filenames only
4516 Bugzilla : 9969, 10379
4517 Description: utime reports stale NFS file handle
4518 Details : SLES9 uses out-of-dentry names in some cases, which confused
4519 the lustre dentry revalidation. Change it to always use the
4523 Frequency : rare, unless heavy write-truncate concurrency is continuous
4524 Bugzilla : 4180, 6984, 7171, 9963, 9331
4525 Description: OST becomes very slow and/or deadlocked during object unlink
4526 Details : filter_destroy() was holding onto the parent directory lock
4527 while truncating+unlinking objects. For very large objects this
4528 may block other threads for a long time and slow overall OST
4529 responsiveness. It may also be possible to get a lock ordering
4530 deadlock in this case, or run out of journal credits because of
4531 the combined truncate+unlink. Solution is to do object truncate
4532 first in one transaction without parent lock, and then do the
4533 final unlink in a new transaction with the parent lock. This
4534 reduces the lock hold time dramatically.
4537 Frequency : rare, 2.4 kernels only
4539 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4540 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4541 before unmounting their filesystem in order to avoid deadlock.
4542 The kernel_locked() function in 2.4 kernels only checks whether
4543 the kernel lock is held, not whether it is this process that is
4544 holding it as 2.6 kernels do.
4549 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4550 Details : The client ptlrpc code may be trying to reconnect to a down
4551 server before a previous connection attempt has timed out.
4552 Increase the reconnect interval to be longer than the connection
4553 timeout interval to avoid sending duplicate connections to
4557 Frequency : echo_client brw_test command
4559 Description: fix echo_client to work with OST preallocated code
4560 Details : OST preallocation code (5137) didn't take echo_client IO path
4561 into account: echo_client calls filter methods outside of any
4562 OST thread and, hence, there is no per-thread preallocated
4563 pages and buffers to use. Solution: hijack pga pages for IO. As
4564 a byproduct, this avoids unnecessary data copying.
4568 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4569 Description: Client can oops in mdc_commit_close() after open replay
4570 Details : It was possible for the MDS to return an open request with no
4571 transaction number in mds_finish_transno() if the client was
4572 evicted, but without actually returning an error. Clients
4573 would later try to replay that open and may trip an assertion
4574 Simplify the client close codepath, and always return an error
4575 from the MDS in case the open is not successful.
4578 Frequency : rare, 2.6 OSTs only
4580 Description: OST may deadlock under high load on fragmented files
4581 Details : If there was a heavy load and highly-fragmented OST filesystems
4582 it was possible to have all the OST threads deadlock waiting on
4583 allocation of biovecs, because the biovecs were not released
4584 until the entire RPC IO was completed. Instead, release biovecs
4585 as soon as they are complete to ensure forward IO progress.
4587 Severity : enhancement
4589 Description: Support for specifying external journal device at mount
4590 Details : If an OST or MDS device is formatted with an external journal
4591 device, this device major/minor is stored in the ext3 superblock
4592 and may not be valid for failover. Allow detecting and
4593 specifying the external journal at mount time.
4598 Description: Mounting an MDS with pending unlinked files may cause oops
4599 Details : target_finish_recovery() calls mds_postrecov() which returned
4600 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4601 considers this an error and immediately begins cleaning up the
4602 lov, just after starting the mds_lov process
4604 Severity : enhancement
4606 Description: Implement 'lfs df' to report actual free space on per-OST basis
4607 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4608 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4609 usage of objects; '-h' to report in human readable format.
4611 ------------------------------------------------------------------------------
4613 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4620 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4621 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4622 at startup during orphan unlinking. Instead, skip buddy update
4623 before setup, as it will be regenerated after recovery is complete.
4628 Description: If an OST is inactive, its locks might reference stale inodes.
4629 Details : lov_change_cbdata() must iterate over all namespaces, even if
4630 they are inactive to clear inode references from the lock.
4632 Severity : enhancement
4633 Frequency : occasional, if non-standard max_dirty_mb used
4635 Description: Client will block write RPCs if not enough grant
4636 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4637 then the client will block writes while waiting for another RPC
4638 to complete instead of consuming its dirty limit. With change
4639 we get improved performance when max_dirty_mb is small.
4641 Severity : enhancement
4642 Bugzilla : 3389, 6253
4643 Description: Add support for supplementary groups on the MDS.
4644 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4645 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4646 lookups for user supplementary groups into a cache.
4650 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4651 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4652 for liblustre O_EXCL is enforced by MDS.
4656 Description: readdir never returns NULL in liblustre.
4657 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4658 dentry in struct dirent.
4662 Frequency : liblustre only, depends on application IO pattern
4663 Description: liblustre clients evicted if not contacting servers
4664 Details : Don't put liblustre clients into the ping_evictor list, so
4665 they will not be evicted by the pinger ever.
4667 Severity : enhancement
4669 Description: Add ability to evict clients by NID from MDS.
4670 Details : By echoing "nid:$NID" string into
4671 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4672 $NID would be instantly evicted from this MDS and from all active
4673 OSTs connected to it.
4677 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4678 Details : lookup_it_finish() used to query file size from OSTs that was not
4683 Description: service threads change working directory to that of init
4684 Details : Starting lustre service threads may pin the working directory
4685 of the parent thread, making that filesystem busy. Threads
4686 now change to the working directory of init to avoid this.
4690 Frequency : during shutdown only
4691 Description: shutdown with a failed MDS or OST can cause unmount to hang
4692 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4697 Frequency : 2.6 only
4698 Description: chmod/chown may include an extra supplementary group
4699 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4700 supplementary group and if none is specified this is used.
4703 Bugzilla : 5479 (6816)
4705 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4706 Details : If lookup is in progress on a file that is unlinked we might try
4707 to revalidate the inode and fail in revalidate after lookup is
4708 complete and ll_file_open() enqueues the open again but
4709 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4712 Frequency : always, if lconf --abort_recovery used
4714 Description: lconf --abort_recovery fails with 'Operation not supported'
4715 Details : lconf was attempting to abort recovery on the MDT device and not
4718 ------------------------------------------------------------------------------
4720 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4725 Frequency : rare (only unsupported configurations with a node running as an
4727 Bugzilla : 6514, 5137
4728 Description: Mounting a Lustre file system on a node running as an OST could
4730 Details : OSTs now preallocates memory needed to write out data at
4731 startup, instead of when needed, to avoid having to
4732 allocate memory in possibly low memory situations.
4733 Specifically, if the file system is mounted on on OST,
4734 memory pressure could force it to try to write out data,
4735 which it needed to allocate memory to do. Due to the low
4736 memory, it would be unable to do so and the node would
4737 become unresponsive.
4739 Severity : enhancement
4741 Description: Addition of lconf --service command line option
4742 Details : lconf now accepts a '--service <arg>' option, which is
4743 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4745 Severity : enhancement
4747 Description: Failover mode is now the default for OSTs.
4748 Details : By default, OSTs will now run in failover mode. To return to
4749 the old behaviour, add '--failout' to the lmc line for OSTs.
4751 Severity : enhancement
4753 Description: Health checks are now provided for MDS and OSTs
4754 Details : Additional detailed health check information on MSD and OSTs
4755 is now provided through the procfs health_check value.
4758 Frequency : occasional, depends on IO load
4760 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4761 after numerous create/delete cycles
4762 Details : The ext3 inode allocation policy would not allocate new inodes
4763 very well on the OSTs because there are no new directories
4764 being created. Instead we look for groups with free space if
4765 the parent directories are nearly full.
4770 Description: Network or server problems during mount may cause partially
4771 mounted clients instead of returning an error.
4772 Details : The config llog parsing code may overwrite the error return
4773 code during mount error handling, returning success instead
4779 Description: MDS can fail to allocate large reply buffers
4780 Details : After long uptimes the MDS can fail to allocate large reply
4781 buffers (e.g. zconf client mount config records) due to memory
4782 fragmentation or consumption by the buffer cache. Preallocate
4783 some large reply buffers so that these replies can be sent even
4784 under memory pressure.
4788 Frequency : rare (liblustre)
4789 Description: fsx running with liblustre complained that using truncate() to
4790 extend the file doesn't work. This patch corrects that issue.
4791 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4792 ATTR_SIZE and lsm use in llu_setattr_raw.
4796 Frequency : rare, only 2.6 kernels
4797 Description: Unusual file access patterns on the MDS may result in inode
4798 data being lost in very rare circumstances.
4799 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4800 code in the 2.6 kernel caused the inode and/or EA data not to
4801 be read from disk, causing single-file corruption.
4805 Frequency : rare, only 2.6 filesystems using extents
4806 Description: Heavy concurrent write and delete load may cause data corruption.
4807 Details : It was possible under high-load situations to have an extent
4808 metadata block in the block device cache from a just-unlinked
4809 file overwrite a newly-allocated data block. We now unmap any
4810 metadata buffers that alias just-allocated data blocks.
4814 Frequency : filesystems with default stripe_count larger than 77
4815 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4816 Details : lconf specifies an inode size of 4096 bytes when the default
4817 stripe_count is larger than 77. This conflicts with the default
4818 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4819 this case to avoid pinning too much memory for large EAs.
4821 ------------------------------------------------------------------------------
4823 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4828 Frequency : rare (extremely heavy IO load with hundreds of clients)
4830 Description: Client is evicted, gets IO error writing to file
4831 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4832 caused clients to be evicted for AST timeouts. The fixes in
4833 bug 5192 mean we no longer need to have such short AST timeouts
4834 so ldlm_timeout has been increased.
4837 Frequency : occasional during --force or --failover shutdown under load
4838 Bugzilla : 5949, 4834
4839 Description: Server oops/LBUG if stopped with --force or --failover under load
4840 Details : a collection of import/export refcount and cleanup ordering
4841 issues fixed for safer force cleanup
4844 Frequency : only filesystems larger than 120 OSTs
4845 Bugzilla : 5990, 6223
4846 Description: lfs getstripe would oops on a very large filesystem
4847 Details : lov_getconfig used kfree on vmalloc'd memory
4850 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4851 Bugzilla : 6242, 6243
4852 Description: reading from files that had been truncated to a non-zero size
4853 but never opened returned no data
4854 Details : ll_file_read() reads zeros from no-object files to EOF
4859 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4860 Details : There's little harm in aborting MDS/OSS recovery and letting it
4861 try again, so I removed the LASSERT and return an error instead.
4863 Severity : enhancement
4865 Description: New debugging infrastructure for tracking down data corruption
4866 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4867 (b) cover more of the client-side code path, and (c) try to narrow
4868 down where problems occurred
4872 Bugzilla : 3819, 4364, 4397, 6313
4873 Description: Racing close and eviction MDS could cause assertion in mds_close
4874 Details : It was possible to get multiple mfd references during close and
4875 client eviction, leading to one thread referencing a freed mfd.
4877 Severity: : enhancement
4878 Bugzilla : 3262, 6359
4879 Description: Attempts to reconnect to servers are now more aggressive.
4880 Details : This builds on the enhanced upcall-less recovery that was added
4881 in 1.4.2. When trying to reconnect to servers, clients will
4882 now try each server in the failover group every 10 seconds. By
4883 default, clients would previously try one server every 25 seconds.
4888 Description: After recovery, certain operations trigger a failed
4889 assertion on a client.
4890 Details : Failing over an mds, using lconf -d --failover, while a
4891 client was doing a readdir() call would cause the client to
4892 LBUG after recovery completed and the readdir() was resent.
4894 Severity : enhancement
4896 Description: Default groups are now added by lconf
4897 Details : You can now run lconf --group <servicename> without having to
4898 manually add groups with lmc.
4901 Frequency : occasional
4903 Description: Nodes with an elan id of 0 trigger a failed assertion
4906 Frequency : always when accessing e.g. tty/console device nodes
4908 Description: tty and some other devices nodes cannot be used on lustre
4909 Details : file's private_data field is used by device data and lustre
4910 values in there got lost. New field was added to struct file to
4911 store fs-specific private data.
4914 Frequency : when exporting Lustre via NFS
4916 Description: NFSD failed occasionally when looking up a path component
4917 Details : NFSD is looking up ".." which was broken in ext3 directories
4918 that had grown large enough to become hashed.
4921 Frequency : Clusters with multiple interfaces not on the same subnet
4923 Description: Nodes will repeatedly try to reconnect to an interface which it
4924 cannot reach and report an error to the log.
4925 Details : Extra peer list entries will be created by lconf with some peers
4926 unreachable. lconf now validates the peer before adding it.
4929 Frequency : Only if a default stripe is set on the filesystem root.
4931 Description: Setting a default stripe on the filesystem root prevented the
4932 filesystem from being remounted.
4933 Details : The client was sending extra request flags in the root getattr
4934 request and did not allocate a reply buffer for the dir EA.
4937 Frequency : occasional, higher if lots of files are accessed by one client
4938 Bugzilla : 6159, 6097
4939 Description: Client trips assertion regarding lsm mismatch/magic
4940 Details : While revalidating inodes the VFS looks up inodes with ifind()
4941 and in rare cases can find an inode that is being freed.
4942 The ll_test_inode() code will free the lsm during ifind()
4943 when it finds an existing inode and then the VFS later attaches
4944 this free lsm to a new inode.
4948 Bugzilla : 6422, 7030
4949 Description: MDS deadlock between mkdir and client eviction
4950 Details : Creating a new file via mkdir or mknod (starting a transaction
4951 and getting the ns lock) can deadlock with client eviction
4952 (gets ns lock and trying to finish a synchronous transaction).
4955 Frequency : occasional
4956 Description: While starting a server, the fsfilt_ext3 module could not be
4958 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4959 kernels. Previously, lconf would still use the ext3 name
4960 when trying to load modules. Now, it will correctly use
4961 ext3 on 2.4 and ldiskfs on 2.6.
4963 Severity : enhancement
4964 Description: The default stripe count has been changed to 1
4965 Details : The interpretation of the default stripe count (0, to lfs
4966 or lmc) has been changed to mean striping across a single
4967 OST, rather than all available. For general usage we have
4968 found a stripe count of 1 or 2 works best.
4970 Severity : enhancement
4971 Description: Add support for compiling against Cray portals.
4972 Details : Conditional compiling for some areas that are different
4976 Frequency : occasional
4977 Bugzilla : 6409, 6834
4978 Description: Creating files with an explicit stripe count may lead to
4979 a failed assertion on the MDS
4980 Details : If some OSTs are full or unavailable, creating files may
4981 trigger a failed assertion on the MDS. Now, Lustre will
4982 try to use other servers or return an error to the
4986 Frequency : occasional
4988 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4989 caused lock timeout during readahead (since 1.4.2).
4990 Details : Processes doing readahead might match a lock that hasn't been
4991 granted yet if there are overlapping and conflicting lock
4992 requests. The readahead process waits on ungranted lock
4993 (original lock is CBPENDING), while OST waits for that process
4994 to cancel CBPENDING read lock and eventually evicts client.
4996 Severity : enhancement
4998 Description: Initial enabling of flock support for clients
4999 Details : Implements fcntl advisory locking and file status functions.
5000 This feature is provided as an optional mount flag (default
5001 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5002 locking are implemented yet, and those that are are not guaranteed
5003 to be completely correct in production environments.
5004 mount -t lustre -o [flock|noflock] ...
5007 Frequency : occasional
5009 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5010 assertion in the ext3 JBD (journaling) layer.
5011 Details : The b_committed_data struct is protected by the big kernel lock
5012 in 2.4 kernels, serializing journal_commit_transaction() and
5013 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5014 access to this struct. In 2.6 kernels there is finer grained
5015 locking to improve SMP performance of the JBD layer.
5019 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5024 Description: merge and cleanup kernel patches.
5027 -----------------------------------------------------------------------------
5029 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5031 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5032 versions of Lustre. Please update all servers and clients to
5033 version 1.4.2 or later at the same time. You must also run
5034 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5035 to update the configuration logs.
5037 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5038 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5039 - swab llog records of type '0' so we get proper header size/idx (5861)
5040 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5041 - fix rename of one directory over another leaking an inode (5953)
5042 - avoid SetPageDirty on 2.6 (5981)
5043 - don't re-add just-being-destroyed locks to the waiting list (5653)
5044 - when creating new directories, inherit the parent's custom
5045 striping settings if present parent (3048)
5046 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5047 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5048 - don't hold client locks on temporary worklist from l_lru (5666)
5049 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5050 - automatically evict dead clients (5921)
5051 - Update file size properly in create+truncate+fstat case (6196)
5052 - Do not unhash mountpoint dentries, do not allow removal of
5054 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5055 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5056 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5057 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5058 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5059 - Correctly handle reads of files with no objects (6243)
5060 - lctl recover will also mark a device active if deactivate used (5933)
5062 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5063 - allow --write-conf on an MDS with different nettype than client (5619)
5064 - don't write config llogs to MDS for mounts not from that MDS (5617)
5065 - lconf should create multiple TCP connections from a client (5201)
5066 - init scripts are now turned off by default; run chkconfig --on
5067 lustre and chkconfig --on lustrefs to use them
5068 - upcalls are no longer needed for clients to recover to failover
5070 - add --abort-recovery option to lconf to abort recovery on device
5072 - add support for an arbitrary number of OSTs (3026)
5073 - Quota support protocol changes.
5074 - forward compatibility changes to wire structs (6007)
5075 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5076 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5077 - improved "lustre status" script
5078 - initialize blocksize for non-regular files (6062)
5079 - added --disable-server and --disable-client configure options (5782)
5080 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5081 - Vanilla 2.4.29 support
5082 - increase maximum number of obd devices to 520 (6242)
5083 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5084 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5085 SLES 9 2.6.5 kernels
5086 - limit stripes per file to 160 (the maximum EA size) (6093)
5088 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5091 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5092 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5093 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5094 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5095 - small ext3 extents cleanups and fixes (5733)
5096 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5097 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5098 - avoid needless client->OST connect, fix handle mismatch (5317)
5099 - fix DLM error path that led to out-of-sync client, long delays (5779)
5100 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5101 - fix several locking issues related to i_size (5492,5624,5654,5672)
5102 - don't move pending lock onto export if it is already evicted (5683)
5103 - fix kernel oops when creating .foo in unlinked directory (5548)
5104 - fix deadlock in obdfilter statistics vs. object create (5811)
5105 - use time_{before,after} to avoid timer jiffies wrap (5882)
5106 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5107 - Do not leak request if server was not able to process it (5154)
5108 - If mds_open unable to find parent dir, make that negative lookup(5154)
5109 - don't create new directories with extent-mapping (5909, 5936)
5111 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5112 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5113 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5114 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5115 - support for RHEL4 kernel on i686 (5773)
5116 - provide error messages when incompatible logs are encountered (5898)
5118 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5119 * version 1.4.0.10 (1.4.1 release candidate 1)
5121 - don't keep a lock reference when lock is not granted (4238)
5122 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5123 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5124 - drop import inflight refcount on signal_completed_replay error (5255)
5125 - unlock page after async write error during send (3677)
5126 - handle missing objects in filter_preprw_read properly (5265)
5127 - no transno return for symlink open, don't save no-trasno open (3440)
5128 - don't try to complete elan receive that already failed (4012)
5129 - free RPC server reply state on error (5406)
5130 - clean up thread from ptlrpc_start_thread() on error (5160)
5131 - readahead could read extra page into cache that wasn't ejected (5388)
5132 - prevent races in class_attach/setup/cleanup/detach (5260)
5133 - don't dereference de->d_inode after l_dput of de (5458)
5134 - use "int" for stripe value returned from lock_to_stripe (5544)
5135 - mballoc allocation and error-checking fixes in 2.6 (5504)
5136 - block device patches to fix I/O request sizes in 2.6 (5482)
5137 - look up hostnames for IB nals (5602)
5138 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5139 - don't start multiple acceptors for the same port (5277)
5140 - fix incorrect LASSERT in mds_getattr_name (5635)
5141 - export a proc file for general "ping" checking (5628)
5142 - fix "lfs check" to not block when the MDS is down (5628)
5144 - service request history (4965)
5145 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5146 - create an "evict_client" /proc entry on OSTs, like the MDS has
5147 - fix mount usage message, return errors per mount(8) (5168)
5148 - change grep [] to grep "[]" in tests so they work in more UMLs
5149 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5150 - remove ancient LOV_MAGIC_V0 EA support (5047)
5151 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5152 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5153 - allow readahead allocations to fail when low on memory (5383)
5154 - mmap locking landed again, after considerable improvement (2828)
5155 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5157 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5160 - send OST transaction number in read/write reply to free req (4966)
5161 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5162 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5163 - the watchdog thread now runs as interruptible (5246)
5164 - flock/lockf fixes (but it's still disabled, pending 5135)
5165 - don't use EXT3 constants in llite code (5094)
5166 - memory shortage at startup could cause assertion (5176)
5168 - reorganization of lov code
5169 - single portals codebase
5171 - add extents/mballoc support (5025)
5172 - direct I/O reads in the obdfilter (4048)
5173 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5175 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5178 - send OST transaction number in read/write reply to free req (4966)
5179 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5180 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5181 - the watchdog thread now runs as interruptible (5246)
5182 - handle missing objects in filter_preprw_read properly (5265)
5183 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5184 - drop import inflight refcount on signal_completed_replay error (5255)
5185 - unlock page after async write error during send (3677)
5186 - return original error code on reconstructed replies (3761)
5187 - no transno return for symlink open, don't save no-trasno open (3440)
5189 - add pid to ldlm debugging output (4922)
5190 - bump the watchdog timeouts -- we can't handle 30sec yet
5191 - extra debugging for orphan dentry/inode bug (5259)
5193 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5196 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5197 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5198 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5199 - fix NULL dereference in /proc/sys/portals/routes (4827)
5200 - allow failed mdc_close() operations to be interrupted (4561)
5201 - stop precreate on OST before MDS would time out on it (4778)
5202 - don't send partial-page writes before EOF from client (4410)
5203 - discard client grant for sub-page writes on large-page clients (4520)
5204 - don't free dentries not owned by NFS code, check generation (4806)
5205 - fix lsm leak if mds_create_objects() fails (4801)
5206 - limit debug_daemon file size, always print CERROR messages (4789)
5207 - use transno after validating reply (3892)
5208 - process timed out requests if import state changes (3754)
5209 - update mtime on OST during writes, return in glimpse (4829)
5210 - add mkfsoptions to LDAP (4679)
5211 - use ->max_readahead method instead of zapping global ra (5039)
5212 - don't interrupt __l_wait_event() during strace
5214 - add software watchdogs to catch hung threads quickly (4941)
5215 - make lustrefs init script start after nfs is mounted
5216 - fix CWARN/ERROR duplication (4930)
5217 - return async write errors to application if possible (2248)
5218 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5219 - print NAL number in %x format (4645)
5220 - update barely-supported suse-2.4.21-171 series (4842)
5221 - support for sles 9 %post scripts
5222 - support for building 2.6 kernel-source packages
5223 - support for sles km_* packages
5225 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5228 - ignore -ENOENT errors in osc_destroy (3639)
5229 - notify osc create thread that OSC is being cleaned up (4600)
5230 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5231 - reconstruct ost_handle() like mds_handle() (4657)
5232 - create a new thread to do import eviction to avoid deadlock (3969)
5233 - let lconf resolve symlinked-to devices (4629)
5234 - don't unlink "objects" from directory with default EA (4554)
5235 - hold socknal file ref over connect in case target is down (4394)
5236 - allow more than 32000 subdirectories in a single directory (3244)
5237 - fix blocks count for O_DIRECT writes (3751)
5238 - OST returns ENOSPC from object create when no space left (4539)
5239 - don't send truncate RPC if file size isn't changing (4410)
5240 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5241 - bind to privileged port in socknal and tcpnal (3689)
5243 - rate limit CERROR/CWARN console message to avoid overload (4519)
5244 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5245 - basic mmap support (3918)
5246 - kernel patch series update from b1_4 (4711)
5248 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5251 - avoid crash during MDS cleanup with OST shut down (2775)
5252 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5253 - don't use bad inodes on the MDS (3744)
5254 - dynamic object preallocation to improve recovery speed (4236)
5255 - don't hold spinlock over lock dumping or change debug flags (4401)
5256 - don't zero obd_dev when it is force cleaned (3651)
5257 - print grants to console if they go negative (4431)
5258 - "lctl deactivate" will stop automatic recovery attempts (3406)
5259 - look for existing locks in ldlm_handle_enqueue() (3764)
5260 - don't resolve lock handle twice in recovery avoiding race (4401)
5261 - revalidate should check working dir is a directory (4134)
5263 - don't always mark "slow" obdfilter messages as errors (4418)
5265 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5268 - don't close LustreDB during write_conf until it is done (3860)
5269 - fix typo in lconf for_each_profile (3821)
5270 - allow dumping logs from multiple threads at one time (3820)
5271 - don't allow multiple threads in OSC recovery (3812)
5272 - fix debug_size parameters (3864)
5273 - fix mds_postrecov to initialize import for llog ctxt (3121)
5274 - replace config semaphore with spinlock (3306)
5275 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5276 - don't allow enqueue to complete on a destroyed export (3822)
5277 - down write_lock before checking llog header bitmap (3825)
5278 - recover from lock replay timeout (3764)
5279 - up llog sem before sending rpc (3652)
5280 - reduce ns lock hold times when setting kms (3267)
5281 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5282 - fix NULL deref and obd_dev leak on setup error (3312)
5283 - replace some LBUG about llog ops with error handling (3841)
5284 - don't match INVALID dentries from d_lookup and spin (3784)
5285 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5286 - fix invalid assertion in ptlrpc_set_wait (3880)
5288 - add libwrap support for the TCP acceptor (3996)
5289 - add /proc/sys/portals/routes for non-root route listing (3994)
5290 - allow setting MDS UUID in .xml (2580)
5291 - print the stack of a process that LBUGs (4228)
5293 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5296 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5297 - make sure to unset replay flag from failed open requests (3440)
5298 - if default stripe count is 0, use OST count for inode size (3636)
5299 - update parent mtime/ctime on client for create/unlink (2611)
5300 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5301 - free recovery state on server during a forced cleanup (3571)
5302 - unregister_reply for resent reqs (3063)
5303 - loop back devices mounting and status check on 2.6 (3563)
5304 - fix resource-creation race that can provoke i_size == 0 (3513)
5305 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5306 - more debugging for page-accounting assertion (3746)
5307 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5308 - avoid deadlock after precreation failure (3758)
5309 - fix race and lock order deadlock in orphan handling (3450, 3750)
5310 - add validity checks when grabbing inodes from l_ast_data (3599)
5312 - add /proc/.../recovery_status to obdfilter (3428)
5313 - lightweight CDEBUG infrastructure, debug daemon (3668)
5314 - change default OSC RPC parameters to be better on small clusters
5315 - turn off OST read cache for files smaller than 32MB
5316 - install man pages and include them in rpms (3100)
5317 - add new init script for (un)mounting lustre filesystems (2593)
5318 - run chkconfig in %post for init scripts (3701)
5319 - drop scimac NAL (unmaintained)
5321 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5324 - clean kiobufs before and after use (3485)
5325 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5326 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5327 - change init script to fail more clearly if not run as root (1528)
5328 - allow clients to reconnect during replay (1742)
5329 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5330 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5331 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5332 - fix cond_resched (3554)
5333 - only evict unfinished clients after recovery (3515)
5334 - allow bulk resend, prevent data loss (3570)
5335 - dynamic ptlrpc request buffer allocation (2102)
5336 - don't allow unlinking open directory if it isn't empty (2904)
5337 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5338 - remove extraneous obd dereference causing LASSERT failure (3334)
5339 - don't use get_cycles() when creating temp. files on the mds (3156)
5340 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5341 - handle EEXIST for set-stripe, set proper directory name (3336)
5343 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5344 - fix ksocknal_fmb_callback() error messages (2918)
5346 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5349 - don't copy lvb into (possibly NULL) reply on error (2983)
5350 - don't deref dentry after dput, don't free lvb on error (2922)
5351 - use the kms to determine writeback rpc length (2947)
5352 - increment oti_logcookies when osc is inactive (2948)
5353 - update client's i_blocks count via lvb messages (2543)
5354 - handle intent open/close of special files properly (1557)
5355 - mount MDS with errors=remount-ro, like obdfilter (2009)
5356 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5357 - don't use cancelling-locks' kms values (2947)
5358 - use highest lock extent for kms, not last one (2925)
5359 - don't dereference ERR_PTR() dentry in error handling path (3107)
5360 - fix thread race in portals_debug_dumplog() (3122)
5361 - create lprocfs device entries at setup instead of at attach (1519)
5362 - common AST error handler, don't evict client on completion race (3145)
5363 - zero nameidata in detach_mnt in 2.6 (3118)
5364 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5365 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5366 - initialize RPC timeout timer earlier for 2.6 (3219)
5367 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5368 - print nal/nid for unknown nid (3258)
5369 - additional checks for oscc recovery before doing precreate (3284)
5370 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5371 - don't crash in mdc_close for bad permissions on open (3285)
5372 - zero i_rdev for non-device files (3147)
5373 - clear page->private before handing to FS, better assertion (3119)
5374 - tune the read pipeline (3236)
5375 - fix incorrect decref of invalidated dentry (2350)
5376 - provide read-ahead stats and refine rpc in flight stats (3328)
5377 - don't hold journal transaction open across create RPC (3313)
5378 - update atime on MDS at close time (3265)
5379 - close LDAP connection when recovering to avoid server load (3315)
5380 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5381 - don't leak open file on MDS after open resend (3325)
5382 - serialize filter_precreate and filter_destroy_precreated (3329)
5383 - loop device shouldn't call sync_dev() for nul device (3092)
5384 - clear page cache after eviction (2766)
5385 - resynchronize MDS->OST in background (2824)
5386 - refuse to mount the same filesystem twice on same mountpoint (3394)
5387 - allow llmount to create routes for mounting behind routers (3320)
5388 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5389 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5390 - fix rare problem with rename on htree directories (3417)
5392 - allow default OST striping configuration per directory (1414)
5393 - fix compilation for qswnal for 2.6 kernels (3125)
5394 - increase maximum number of MDS request buffers for large systems
5395 - change liblustreapi to be useful for external progs like lfsck (3098)
5396 - increase local configuration timeout for slow disks (3353)
5397 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5399 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5402 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5403 - don't overwrite extent policy data in reply if lock was blocked (2901)
5404 - drop filter export grants atomically with removal from device (2663)
5405 - del obd_self_export from work_list in class_disconnect_exports (2908)
5406 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5407 - swab reply message in mdc_close, other PPC fixes (2464)
5408 - fix destroying of named logs (2325)
5409 - overwrite old logs when running lconf --write_conf (2264)
5410 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5411 - fix race in target_handle_connect (2898)
5412 - mds_reint_create() should take same inode create lock (2926)
5413 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5414 - don't close files for self_export to avoid uninitialized obd (2936)
5415 - allow MDS with the same name as client node (2939)
5416 - hold dentry reference for closed log files for unlink (2325)
5417 - reserve space for all logs during transactions (2059)
5418 - don't evict page beyond end of stripe extent (2925)
5419 - don't oops on a deleted current working directory (2399)
5420 - handle hard links to targets without a parent properly (2517)
5421 - don't dereference NULL lock when racing during eviction (2867)
5422 - don't grow lock extents when lots of conflicting locks (2919)
5424 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5427 - account for cache space usage on clients to avoid data loss (974)
5428 - lfsck support in lustre kernel code (2349)
5429 - reduce journal credits needed for BRW writes (2370)
5430 - orphan handling to avoid losing space on client/server crashes
5431 - ptlrpcd can be blocked, stopping ALL progress (2477)
5432 - use lock value blocks to assist in proper KMS, faster stat (1021)
5433 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5434 - recovery for initial connections (2355)
5435 - fixes for mds_cleanup_orphans (1934)
5436 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5437 - block all file creations until orphan recovery completes (1901)
5438 - client remove rq_connection from request struct (2423)
5439 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5440 - recovery timer race (2670)
5441 - mdc_close recovey bug (2532)
5442 - ptlrpc cleanup bug (2710)
5443 - mds timeout on local locks (2588)
5444 - namespace lock held during RPCs (2431)
5445 - handle interrupted sync write properly (2503)
5446 - don't try to handle a message that hasn't been replied to (2699)
5447 - client assert failure during cleanup after abort recovery (2701)
5448 - leak mdc device after failed mount (2712)
5449 - ptlrpc_check_set allows timedout requests to complete (2714)
5450 - wait for inflight reqs when ptlrpcd finishes (2710)
5451 - make sure unregistered services are removed from the srv_list
5452 - reset bulk XID's when resending them (caught by 1138 test)
5453 - unregister_bulk after timeout
5454 - fix lconf error (2694)
5455 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5456 - readahead locks pages, leaves pending causing memory pressure (2673)
5457 - increase OST request buffers to 4096 on large machines (2729)
5458 - fix up permission of existing directories in simple_mkdir (2661)
5459 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5460 - don't assign transno to errored transactions (2742)
5461 - don't delete objects on OST if given a bogus objid from MDS (2751)
5462 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5463 - if rq_no_resend, then timeout request after recovery (2432)
5464 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5465 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5466 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5467 - MDS recovery shouldn't delete objects that it has given out (2730)
5468 - if enqueue arrives after completion, don't clobber LVB (2819)
5469 - don't unlock pages twice when trigger_group_io returns error (2814)
5470 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5471 - don't write pages to disk if there was an error (1450)
5472 - don't ping imports that have recovery disabled (2676)
5473 - take buffered bytes into account when balancing socknal conn (2817)
5474 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5475 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5476 - remove little-endian swabbing of llog records (1987)
5477 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5478 - retry reposting request buffers if they fail (1191)
5479 - grow extent at grant time to avoid granting a revoked lock (2809)
5480 - lock revoke doesn't evict page if covered by a second lock (2765)
5481 - disable VM readahead to avoid reading outside lock extents (2805)
5483 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5484 - updated kernel patches for hp-2.4.20 kernel (2681)
5486 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5489 - fix truncated write corruption (2366)
5490 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5492 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5493 - ensure len(uuid) < 37 in lmc (1171)
5494 - fix ia64 OOPS in llog_test (2255)
5495 - zero end of page at obdfilter for partial page writes (2648)
5496 - don't leave stale dentries around after renames (bug 2428)
5497 - fix timeouts when evicting a client with a single lock held (2642)
5498 - set deadline for the initial HELLO message to drain (2634)
5499 - print out dotted-quad IP addresses in the socknal (2302)
5501 - additional debugging for MDS client eviction problem (2443)
5502 - fix mkfsoptions support for osts (2603, 2604)
5504 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5507 - add series for the vanilla 2.6.0 kernel
5508 - add series for the vanilla 2.4.24 kernel
5509 - add series for a cray x86/64 UL kernel drop
5510 - fix xattr patches for the vanilla 2.4.19 series
5512 - generate true UUIDs in lmc (1171)
5513 - have portals stack dumping break in UML (2466)
5514 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5515 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5516 - wake callback waiting threads on client eviction (2460)
5517 - Add --ptldebug and --subsystem to lmc (1719)
5518 - update assertion to allow safe interrupt allocation
5519 - set rq_no_resend for cancel requests (2432)
5520 - recalculate ptlrpcd timeout after resend (2494)
5521 - call vfs_rmdir when removing pending directories (2368)
5522 - fix renaming a file to itself (2429)
5523 - lmc creates a default one-stripe lov (2454)
5524 - expand procfs space to handle large clusters (2326)
5525 - increase UML stack to avoid overflow
5526 - update lconf's list of debug and subsystem masks
5527 - fix lfs find --obd (2510)
5528 - /proc tunable for disabling filter read caching (2591)
5529 - stop rpm packages from altering slapd.conf (2301)
5530 - disable nagle in the socknal under 0conf (2578)
5531 - choose mds inode size based on stripe count (2572)
5532 - fix kernel-source rpm problems (2516)
5534 - add --disable-doc to avoid pdf generation (2421)
5535 - update documentation, tests, type-os, comments
5536 - avoid format warnings on ia64
5537 - remove the TOE NAL
5538 - tiny code cleanups by removing unused fields
5540 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5543 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5544 - remove the most blatant lies from BUILDING file (2371)
5545 - change default debug level to reasonable production setting
5546 - reduce client side cache size to reduce cache flush time
5547 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5548 - make TCP zerocopy and pinger support enabled by default (2476)
5549 - sync writes completed after process exits caused crashes (2319)
5550 - maintain correct mount count on the MDS (2356)
5551 - backout 1557, because 2316 wasn't really fixed
5552 - better file I/O statistics gathering in /proc
5553 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5554 - another kernel patch to fix zero-copy TCP function export
5555 - don't take duplicate lock when processing re-sent getattr (2420)
5556 - lctl uses obd_self_export instead of creating new conn (2353)
5557 - MDS/OST recovery case which requires object creation asserted (2425)
5558 - move lfs from /usr/sbin to /usr/bin in packages
5559 - fix race between mds_client_add and mds_client_free (2417)
5560 - use kmalloc instead of slabs in portals (2430)
5561 - don't create duplicate records when a failover MDS is present (2442)
5562 - remove unnecessary mount age check (2332)
5563 - don't remove directory inodes from locks prematurely (2451)
5564 - don't break if MDS service name is the same as hostname (2103)
5565 - fix races in client write RPC generation when cache full (2482)
5567 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5570 - remove now-unused request->rq_obd (278)
5571 - if an allocation fails, print out how much memory we've used (1933)
5572 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5573 - add the "configurable stack size" patch to most series files (1256)
5574 - ability to write large log records, for 100+ OST configs (2306)
5575 - fix NULL deref when filter_prep fails (2314)
5576 - fix operator precedence error in filter_sync
5577 - dynamic allocation of socknal TX descriptors (2315)
5578 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5579 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5580 - ensure that CWARN messages in Portals always get to the syslog
5581 - __init/__exit are not for prototype decls (ldlm_init/exit)
5582 - x86-64 compile warning fixes
5583 - fix gateway LMC keyword conflict (2318)
5584 - fix MDS lock inversions in getattr/reint paths (1844)
5585 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5586 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5587 - DLM race condition prevented some lock evictions (2328)
5588 - ENOMEM detection and retry on socknal sends (2230)
5589 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5590 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5591 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5592 - better error messages when a client is rejected during recovery (1505)
5593 - avoid cancelling locks which were never granted, after failure (2330)
5594 - fix i_sem/journal inversion in mds_client_add (2333)
5595 - fix truncate/getattr lock cycle deadlock (2334)
5596 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5597 - fix two rare exit paths which could leak an l_lock() ref (2321)
5598 - fencepost error in MDS/OST orphan recovery (2226)
5599 - make log record alignment 8 bytes (1988)
5600 - lstripe now fails when requested offset > ost_count (2237)
5601 - ensure that all kernel series have a complete list.h (1607)
5602 - fix crashes in special-file operations (2316)
5603 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5605 - allow configurable automake binary, for testing new versions
5606 - small update to the lfs documentation
5608 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5610 * fix negative export reference count in fsfilt_sync (2312)
5612 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5613 * release candidate 0.9.1
5615 - orphans are moved into the PENDING directory for possible recovery
5616 - replayed opens now open by fid for orphan/rename safety (1042)
5617 - last close of an orphan inode generates a transno (683)
5618 - chdir() and mount() now pin the directory entry (1020)
5619 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5620 - discard very old requests without processing them (1502)
5621 - remove some common, well-understood CERRORs (1505)
5622 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5623 - clear "grant" flags in OST replies until OST grant code lands (1644)
5624 - fix read performance by not clobbering i_blksize on client (1598)
5625 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5626 - make LRU size a /proc tunable, clears locks when reduced (707)
5627 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5628 - server locks take references on exports now (1558)
5629 - build fixes for 2.4.20-rh trees (1663)
5630 - return an error from lov_create if all OSCs are inactive (1751)
5631 - fix import levels when a reconnect happens without a timeout (1597)
5632 - exit early from mds_open if we get a lookup error (1749)
5633 - partial page read at EOF wouldn't wait for disk before sending (1642)
5634 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5635 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5636 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5637 - write/truncate lock inversion (1639)
5638 - Don't auto-load obdclass, portals modules during cleanup (1495)
5639 - fix timestamps from jumping to "now" (1763)
5640 - extra journal assertions (1648)
5641 - add an extra multiunlink test (1771)
5642 - fix read_record/write_record API (1776)
5643 - fix leak of offset_extent, possible incorrect i_size later (1772)
5644 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5645 - Debugging for the kqswnal_get_idle_tx problems (1820)
5646 - Allow recovery to be attempted multiple times (1536)
5647 - Write out MDS last_rcvd file after it is first created (1600)
5648 - Fix tx_descriptor leak in failed transmit situations (1827)
5649 - ext3 journaling fixes for assertion failure after IO error (1871)
5650 - class_export_put() on freed export after completion AST error (1896)
5651 - Fix revalidate looping in VFS (1322)
5652 - Don't access a freed export during MDS_REINT timeout (1521)
5653 - Add open-unlink recovery support on the MDS (1673,1764)
5654 - Return an error if no MDS data was read from last_rcvd (1946)
5655 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5656 - Don't setup a disk device that doesn't match exported UUID (317)
5657 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5658 - avoid committing NULL handle in force close
5659 - local.sh is now a one-stripe LOV configuration
5660 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5661 - fix ext3 htree duplicate directory entry corruption (1516)
5662 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5663 - update to new LOV EA format (2097)
5664 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5666 2003-06-15 Phil Schwan <phil@clusterfs.com>
5669 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5670 - per-import/export recovery handling (958, 931, 959)
5671 - multiple last-rcvd slots, for serving multiple FSes (949)
5672 - connections are again shared between multiple imp/exports (963, 964)
5673 - "umount -f" would hang if any requests needed to be sent (393, 978)
5674 - avoid pinning large req buffer by copying for queued messages (989)
5675 - add "uuid" to "lctl device" command to help upcalls (991)
5676 - "open" RPCs with transnos would confuse recovery counters (1037)
5677 - do proper endian conversion of all wire messages (288, 340, 891)
5678 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5679 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5680 - fix ptlrpc_connection leak in target_handle_connect (1174)
5681 - fix import refcounting bug in OST and MDS cleanup (1134)
5682 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5683 - fix very unlikely obd_types race condition (501)
5684 - remove osc_open hack for echo_client (1187)
5685 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5686 - a failure in read_inode2 leads to deadlock (1139)
5687 - cancel ack-locks as soon as transaction is committed (1072)
5688 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5689 - make sure to commitrw after any preprw to avoid deadlock (1162)
5690 - failing to execute a file in a lustre FS would lock inode (1203)
5691 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5692 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5693 - obd_uuid2tgt would walk off the end of the list (1255)
5694 - on IA64 the osc would give portals incorrect bulk size (1258)
5695 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5696 - fix lock inversion caused by new llite matching code (1282)
5697 - limit the number of dirty pages on a client to 10MB (1286)
5698 - timed out locks were not being corrected cancelled (1289)
5699 - fix O_DIRECT above 4GB on IA-32 (1292)
5700 * major user-visible changes
5701 - fail out/fail over policy now controlled by the upcall (993)
5703 - add OBD_PING to check server availability and failure (954)
5704 - lustre messages are now sent in sending host order (288, 340, 891)
5705 - add eadatalen to MDS getattr reply (340)
5706 - OST read replies may contain second buffer, with per-page status (593)
5708 2003-03-11 Phil Schwan <phil@clusterfs.com>
5711 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5712 - failed open()s could cause deadlock; fixed (867, 869)
5713 - stop cancelling OST locks when files are closed (481)
5714 - overlapping XID spaces caused network corruption (851, 853)
5715 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5716 - setattr_raw added, to do single-RPC, server-side setattrs
5717 - lmc/lconf syntax change for OST UUIDs
5718 - fix crashy race condition between ptlrpc_free_req and osc_close
5719 - don't use request in mdc_enqueue if we hit a timeout (889)
5720 - don't set the inode i_size for regular files from the MDS (896)
5721 - handle out of order completion AST (842)
5722 - don't LBUG if a lock request times out after receiving AST (913)
5723 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5724 - if a bad lock AST arrives, send an error instead of dropping entirely
5725 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5726 - fix leak in bulk IO when only partially completed (899, 900, 926)
5727 - fix O_DIRECT for ia64 (55)
5728 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5729 - C-z after timeout could hang a process forever; fixed (977)
5731 - client-side I/O cache (678, 924, 929, 941, 970)
5733 - READPAGE and SETATTRs which don't take server-side locks get
5736 2003-02-11 Phil Schwan <phil@clusterfs.com>
5739 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5740 - Fix fsfilt_extN_readpage() to read a full page of directory
5741 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5742 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5743 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5744 - Most intent operations were removed, in favour of a new RPC mode
5745 that does a single RPC to the server and bypasses most of the VFS
5746 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5747 - Aggressively cancel local locks on DLM servers
5748 - mds_reint_unlink sends EA to the client if it's the last nlink.
5749 client uses that EA to unlink OST objects.
5750 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5751 - recursive symlinks were fixed (439)
5752 - fixed NULL deref in DEBUG_REQ
5753 - filter_update_lastobjid no longer calls sync, which annoyed extN
5754 - fixed multi-client small-writes to a single file problem (445)
5755 - fixed mtime updates during file writes (607)
5756 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5757 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5758 - fixed Linux OST size reporting problem (444, 656)
5759 - OST now updates object mtime with writes or setattr (607, 619)
5760 - client verifies file size before zeroing page past EOF (445)
5761 - OST now writes last allocated objid to disk with allocation (108)
5762 - LOV on echo now works (409)
5764 - mds_reint_unlink sends a new buffer, with the EA included. this
5765 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5766 set if a regular file was being unlinked, and it was the last link
5767 - use PtlGet from the target for bulk writes (315)
5768 - OST now updates object mtime with writes or setattr (607, 619)
5769 - LDLM now has a grant-time callback to revalidate locked items, if
5771 - Many MDS operations were reorganized to combat race conditions
5773 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5774 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5776 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5779 - Fully reactivate OST imports after reconnection (512, others)
5780 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5781 - More graceful error handling for truncating on dead OST (515)
5782 - Don't error out unless we're actually accessing dead stripes (474)
5783 - Fix garbage sizes when stripes are missing (410)
5784 - LRU counters were broken, causing constant lock purge (433, 432)
5785 - garbage on read from stripes with failed OSTs (441)
5786 - mark OSCs as active before reconnecting during recovery (438)
5787 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5788 - lfind did not preserve OST order in output (443)
5789 - symlinks cause hung clients, incorrect data (439)
5790 - stop dereferencing request after dropping refcount (457)
5791 - don't LASSERT(spin_is_locked) on non-SMP (455)
5792 - fixes for many rename() bugs
5793 - fstat didn't correctly synchronize attributes (399)
5794 - server must handle lock cancellation during blocking AST prep (487)
5795 - bulk descriptors were free()d too soon (511)
5796 - fix paths in lconf, which would load incorrect modules (451, 507)
5797 - fix confusing lconf 'host not found' error message (386)
5798 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5799 - fix race condition in mdc_blocking_ast() for inode access (526)
5800 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5801 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5802 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5803 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5804 - fix lstripe in several ways (kernel side) (527)
5805 - fix request leak in ldlm_cli_enqueue (262)
5806 - incorrect OSC was marked inactive after OST failure
5807 - call mds_fs_cleanup before unmounting filesystem (524)
5808 - fix races between taking ns_lock and ldlm_lock_change_resource
5809 - fix races updating LOV export open file list
5810 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5811 - fix recovery NULL deref in ldlm_cli_cancel_unused
5812 - fix some DLM races by using new hash table for lock handles (419)
5813 - permit the client to specify desired inodes, at replay
5814 - duplicate requests when we queue them for replay reintegration
5815 - fix last_rcvd offset calculation
5816 - sync after each recovered transaction, so we always make progress
5817 - never, not always, ERESTART requests without transnos
5818 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5820 - skip replay if the MDS says that the client is already connected
5821 - don't check for a recovery-enabled export to match lctl's UUID
5822 - don't INC_USE_COUNT for phantom exports
5823 - don't crash when cleaning up phantom exports (567)
5824 - don't double-finish or set replay data for errored mdc_open requests
5825 - abort requests when they time out, so we don't get old replies
5826 - send/receive replies for AST messages again
5827 - if the client says that it doesn't have the lock, cancel it on the
5829 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5830 mark it as destroyed, it will all work out when decref is called
5831 - fix module use counts (22, 581)
5833 - ASTs now expect a reply (server cancels lock on error reply)
5835 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5838 - fix many simultaneous client startup (392)
5839 - fix dentry->d_it clobbering
5840 - credentials weren't being shipped for readdir/getattr operations
5841 - remove invalid assertions triggered during some concurrent MD
5843 - proper Lustre versions added (336, 389)
5844 - fix memory leak for create error case (398)
5845 - fix LOV locking bug that would get cli/srv out of sync
5846 - fix echo client over LOV (409)
5847 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5848 - fix double-O_EXCL intent crash (424)
5849 - avoid sending multiple lock CANCELs (352)
5851 - MDS can do multi-client recovery (modulo bugs in new code)
5853 - many updates, edits, cleanups
5855 2002-11-18 Phil Schwan <phil@clusterfs.com>
5858 - fix null d_it dereference (346)
5859 - fix full OST/dbench hang (333)
5860 - fix permission problem with file removal (286)
5861 - fix removal of OSCs from LOV when they fail
5862 - fix NULL deref during bulk timeout (214)
5863 - fix problems related to multiple filesystems on one MDS (241)
5864 - fixed serious subtle metadata locking bugs
5865 - free locks on clients when inodes are removed due to memory
5867 - fix inode pointer in lock data (285)
5868 - partial support for multiple MDS on a single host (241)
5869 - data locks weren't cancelled at clear_inode time (290, 311)
5870 - intent locks could lead to unbounded lock growth (205)
5871 - added a maximum lock count, an LRU list, and a flusher
5872 - fix multiple rename (365)
5873 - properly abstracted the echo client
5874 - OSC locked 1 byte too many; fixed
5875 - rewrote brw callback code:
5876 - fixed recovery bugs related to LOVs (306)
5877 - fixed too-many-pages-in-one-write crash (191)
5878 - fixed (again) crash in sync_io_timeout (214)
5879 - probably fixed callback-related race (385)
5881 - Add capability to MDS protocol
5882 - LDLM cancellations and callbacks on different portals
5884 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5887 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5888 - defer open object destruction to close time (601981, 138)
5889 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5890 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5891 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5892 - don't call obd_setattr in ll_file_release for destroyed objects
5893 * protocol change to lustre_msg: move |version| and add |flags|
5894 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5895 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5896 * added replay of create, unlink, link and rename operations during
5897 MDS failover; recovery should be much more robust now
5898 * remove failed OSCs from LOVs (only lov_create uses this so far)
5899 * the lustre-HOWTO was brought (more) up to date (582544)
5901 2002-10-23 Phil Schwan <phil@clusterfs.com>
5904 - in-use dentries weren't being reused properly (617851)
5905 - prevent multiple LDLM setup (599178)
5906 - fix LOV size calculations for truncate (617853)
5907 - fix client handling of MDS intent errors (POSIX)
5908 - fix permission bug in lovstripe.c test (624321)
5909 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5910 - truncate past end of file could corrupt data
5911 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5912 - a race in recovery could return ETIMEDOUT to apps (623947)
5913 - building outside the source directory was fixed
5914 * the lustre-HOWTO was brought (more) up to date (582544)
5915 * major progress was made on recovery functionality
5917 2002-10-10 Phil Schwan <phil@clusterfs.com>
5920 - recovery deadlock fix
5921 - rm -rf causes LBUG fix (617817)
5922 - file open by multiple tasks fix (618962)
5923 - directory permissions bugs (602707 and 620007)
5924 - journal_stop fixed with locking (611313)
5925 - O_APPEND failures resolved (618273, perhaps 614459)
5926 - lconf PATH fix (619770)
5927 - IA64 build fix (621450)
5928 - RPC buffer sizes scale with amount of memory
5930 2002-10-01 Phil Schwan <phil@clusterfs.com>
5933 - locks would be cancelled without throwing away data pages,
5934 resulting in inconsistent data (605627)
5935 - inode attributes were not always being refreshed (605627, 612449)
5936 - lconf now continues to cleanup after lctl reports an error
5937 - MDS now enforces user permissions (602707)
5938 - lprocfs cleanup fixed, but not yet enabled (614157)
5939 - fixed infinite server hang, should a client not respond to an AST
5940 - avoid going into recovery if user calls readlink() with a buffer
5941 that's too small (613941)
5942 - AST RPCs no longer require replies (614867) -- this may be changed
5943 - don't crash server if client sends an IOV that's too big (611336)
5944 - fixed lock conversion deadlock (611892)
5945 - fixed the following of symlinks (614622)
5946 * recovery: the server can remove locks from a client that dies, other
5947 clients can make progress
5948 * more extN patch fixes
5949 * compile-time configurable ptlrpc buffer allocations
5951 - collaborative read cache document
5952 - Lustre Lite Performance CDR document-in-progress
5954 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5957 - fix typo in patch-2.4.18
5959 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5962 - clear ptlrpc request each time in handle_incoming_request()
5963 - unlink of files now destroys the object on the OST
5965 2002-09-19 Peter Braam <braam@clusterfs.com>
5967 * add hard link support
5968 * change obdfile creation method
5969 * kernel patch changed
5971 2002-09-19 Peter Braam <braam@clusterfs.com>
5974 - stack overflow bug in extN fixed
5976 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5978 * documentation updates
5979 - add man pages for config tools
5980 - update tests/README to describe testing with new config tools
5981 - finish metadata API descriptions
5982 * bug fixes and cleanups
5983 - statfs workaround for 16TB limit
5984 - LOV stripe allocation improved, can stripe on subset of OSTs
5985 - LOV file size/IO offset was wrong for files > 4GB in size
5986 - object EA data was being dropped, caused files to be unreadable
5987 - memory overflow with non-LOV OST caused memory corruption
5988 - fixed regression tests to work with new config tools, obdfilter
5989 - fixed bug when directory size became larger than 1 block
5990 - fixed bug (for single client case) when PWD was deleted
5991 - invalidate local directory pages when doing intent-based ops
5992 - avoid LDLM oops when lock callback contained bad data
5994 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5996 * documentation updates
5997 * bug fixes and cleanups
5998 - configuration tools
6001 - 64-bit compile warnings
6002 - 64-bit internal statfs data
6004 * test_brw on persistent OST devices
6006 * lprocfs (disabled)
6008 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6010 * documentation updates
6011 * bug fixes and cleanups
6012 * configuration tools
6014 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6017 * many small fixes to 0_5_4
6018 * io/network handling
6019 * thinkos in MDS operations
6021 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6024 * crucial basic fixes to 0.5.3
6025 * IOR, Iozone work over Elan
6028 2002-08-07 Phil Schwan <phil@clusterfs.com>
6029 * version 0_5_3, our first alpha
6030 * we use the new Portals iovs
6031 * documentation updates
6032 * bug fixes and cleanups
6033 * small changes in the DLM wire protocol
6035 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6036 * version 0_5_1 with some initial stability,
6037 * locking on MD and file I/O.
6038 * documentation updates
6039 * several bug fixes since 0.5.0
6040 * small changes in wire protocol
6042 2002-07-18 Phil Schwan <phil@clusterfs.com>
6044 * delivered as Lustre Light Alpha
6045 * fixed a crash after handling invalid MDS requests
6046 * fixed directory pages for architectures with non-4k pages sizes
6048 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6049 * release version v0_4_4
6050 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6052 * Fixes a number of interruption problems with OST operations.
6053 * Update documentation for portals header changes
6054 * Move all wire protocol structs/defines to lustre_idl.h
6055 * Fixes symlink length bug.
6056 * Add tcpdump to repository.
6058 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6059 * release version v0_4_3
6060 * Fixes statfs for inodes on extN.
6061 * Fixes bug in runtests which would delete /etc/hosts.
6062 * Use 64-bit object IDs wherever possible (not into VFS though)
6063 Remove ost_get_info, which is unused by lustre, and out of date.
6065 2002-07-03 Peter Braam <braam@clusterfs.com>
6066 * release version v0_4_2 Fixes a lookup error (type not passed)
6067 * move forward to head of Portals
6068 * move forward to latest Lustre kernel
6070 2002-06-25 Peter Braam <braam@clusterfs.com>
6071 * release version v0_4_1. Hopefully stable on single node use.