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 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.41.6.sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
14 * File join has been disabled in this release, refer to Bugzilla 16929.
18 Description: Don't allow make backward step on assignin osc next id.
19 Details : race between allocation next id and ll_sync thread can be cause
20 of set wrong osc next id and can be kill valid ost objects.
24 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
28 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
32 Description: Error handling in osc_statfs_interpret() has been improved.
33 Details : Check in osc_statfs_interpret() for EBADR.
37 Description: Do not send statfs() requests to OSTs disabled by administrator.
38 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
40 Severity : enhancement
42 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
44 Severity : enhancement
46 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
48 Severity : enhancement
50 Description: Enhancement for lfs(1) command to use numeric uid/gid.
54 Description: Improved cleanup procedure for conf-sanity test_46a.
58 Description: Do not start bulk transfer, if client tried to reconnect.
59 Details : Do not start bulk transfer, if client tried to reconnect,
60 but failed due to active rpcs. Check for exp_abort_active_req
65 Description: Prevent inconsistences between linux and lustre mount structures.
66 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
67 Make the sleep interruptible.
69 Severity : enhancement
71 Description: provide server to client comms path
72 Details : server to client communications path via new LDLM_SET_INFO rpc
76 Description: truncate starts GFP_FS allocation under transaction causing deadlock
77 Details : ldiskfs_truncate calls grab_cache_page which may start page
78 allocation under an open transaction. This may lead to
79 calling prune_icache with consequent lustre reentrance.
83 Description: send correctly lsm on open replay
84 Details : MDS is trust to LSM size on replay open, but client can set wrong size
89 Description: enable client interface failover
90 Details : When a child reconnects from another NID, properly update export
91 nid hash position and ldlm reverse import.
93 Severity : enhancement
95 Description: Communicate OST degraded/readonly state via statfs to MDS
96 Details : Flags in the statfs returned from OSTs indicate whether the
97 OST is in a degraded RAID state, or if the filesystem has
98 turned read-only after a filesystem error is detected.
103 Description: don't panic if EPROTO was hit when reading symlink
104 Details : correctly handling request reference in error cases.
106 Severity : enhancement
108 Description: Add LustreNetLink, a kernel-userspace communcation path.
110 Severity : enhancement
112 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
117 Description: access to llog context before init.
118 Details : move handling CATALOGS file at osc layer and forbid access to llog
122 Frequency : rare, only when client is writing to OST on local node
124 Description: Avoid deadlock for local client writes
125 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
126 memory freeing context. This allows OST threads to set the
127 PF_MEMALLOC flag on task structures in order to allocate memory
128 from reserved pools and complete IO.
129 Use GFP_HIGHUSER for OST allocations for non-local client writes,
130 so that the OST threads generate memory pressure and allow
131 inactive pages to be reclaimed.
133 Severity : enhancement
135 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
140 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
141 Details : move ldlm namespace creation in setup phase to avoid grab
142 _lprocfs_lock with cli_sem held.
146 Description: Temporarily disable grant shrink.
147 Details : Disable the feature for debugging.
150 Frequency : only when formatting test filesystems
152 Description: Unable to run several mkfs.lustre on loop devices at the same time
153 Details : mkfs.lustre returns error 256 on the concurrent loop devices
154 formatting. The solution is to proper handle the error.
156 Severity : enhancement
158 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
160 Severity : enhancement
162 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
167 Description: resolve race between obd_disconnect and class_disconnect_exports
168 Details : if obd_disconnect will be called to already disconnected export he
169 forget release one reference and osc module can't unloaded.
171 Severity : enhancement
173 Description: Allow tuning service thread via /proc
174 Details : For each service a new
175 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
176 created that can be used to set min/max thread counts, and get the
177 current number of running threads.
182 Descriptoin: don't return error if have partially created objects for file.
183 Details : lov_update_create_set uses set->set_success as index for created
184 objects, so if some requests will be failed, they will have hole
185 at end of array and qos_shrink_lsm for allocate correct lsm.
187 Severity : enhancement
189 Description: Update OFED support to 1.4.1
194 Description: Reduce small size read RPC
195 Details : Set read-ahead limit for every file and only do read-ahead when
196 available read-ahead pages are bigger than 1M to avoid small size
199 Severity : enhancement
201 Description: move AT tunable parameters for more consistent usage
202 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
204 Severity : enhancement
206 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
207 Details : allow skip disconnected ost for send statfs request and hide error
213 Description: fix racy locking of mballoc block bitmaps causing BUG
214 Details : The locking of the mballoc buddy bitmap and the in-memory
215 block bitmap was using two different spin locks in some
216 cases. This made it possible to incorrectly access the
217 mballoc bitmap while another process was modifying it,
218 causing a sanity assertion to fail. While no on-disk corruption
219 was reported, there was some risk of this happening.
222 Frequency : rare, on llog test 6
224 Descriptoin: don't allow connect to already connected import
225 Details : allowing connect to already connected import is hide connecting problem.
230 Descriptoin: allow kill process which wait statahead result
231 Details : for some reasons 'ls' can stick in waiting result from statahead,
232 in this case need way for kill this process.
234 Severity : enhancement
236 Description: Add state history info file, enhance import info file
237 Details : Track import connection state changes in a new osc/mdc proc file;
238 add overview-type data to the osc/mdc import proc file.
240 Severity : enhancement
242 Description: MDS create should not wait for statfs RPC while holding DLM lock.
244 Severity : enhancement
246 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
251 Descriptoin: OST grant too much space to client even there are not enough space.
252 Details : Client will shrink its grant cache to OST if there are no write
253 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
254 this grant cache if there are already not enough avaible space
255 (left_space < total_clients * 32M).
258 Frequency : start MDS on uncleanly shutdowned MDS device
260 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
261 Details : stay in waiting mds<>ost recovery finished produce random bugs
262 due race between two ll_sync thread for one lov target. send
263 ACTIVATE event only if connect realy finished and import have
267 Frequency : rare, connect and disconnect target at same time
269 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
270 Details : don't call obd_disconnect under lov_lock. this long time
271 operation and can block ptlrpcd which answer to connect request.
276 Descriptoin: don't lose wakeup for imp_recovery_waitq
277 Details : recover_import_no_retry or invalidate_import and import_close can
278 both sleep on imp_recovery_waitq, but we was send only one wakeup
282 Frequency : always with long access acl
284 Descriptoin: mds can't pack reply with long acl.
285 Details : mds don't control size of acl but they limited by reint/getattr
288 Severity : enhancement
290 Description: Update to SLES10 kernel-2.6.16.60-0.33.
292 Severity : enhancement
294 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
297 Frequency : start MDS on uncleanly shutdowned MDS device
299 Descriptoin: aborting recovery hang on MDS
300 Details : don't throttle destroy RPCs for the MDT.
303 Frequency : on remount
305 Description: external journal device not working after the remount
306 Details : clear dev_rdonly flag for external journal devices in
312 Description: shutdown vs evict race
313 Details : client_disconnect_export vs connect request race.
314 if client will evicted at this time - we start invalidate
315 thread without referece to import and import can be freed
319 Frequency : rare, need acl's on inode.
321 Description: client can't handle ost additional correctly
322 Details : if ost was added after client connected to mds client can have
323 hit lnet_try_match_md ... to big messages to wide striped files.
324 in this case need teach client to handle config events about add
325 lov target and update client max ea size at that event.
327 Severity : enhancement
329 Description: Changelogs
330 Details : Changelogs are a lightweight mechanism to track filesystem
331 metadata and namespace changes. The changelog is recorded
332 permanently on the MDTs, and is periodically "consumed" / purged
333 when records are no longer needed.
335 Severity : enhancement
337 Description: compact fld format with extents
338 Details : Store range of seq rather than every seq in FLD. Seq
339 controller update FLD rather than clients. In Case of CMD, mdt0
340 has FLD, all other metadata server act as non persistent proxy
341 for FLD queries and cache fld entries in fld cache.
346 Description: don't skip ost target if they assigned to file
347 Details : Drop slow OSCs if we can, but not for requested start idx.
348 This means "if OSC is slow and it is not the requested
349 start OST, then it can be skipped, otherwise skip it only
350 if it is inactive/recovering/out-of-space.
354 Description: more cleanup in mds_lov
355 Details : not send LOV EA under replay, we can't know about they size at this
356 time. Don't allow client connect to mds before any ost connected,
357 for avoid problems with LOV EA size and returning EIO to client.
359 Severity : enhancement
361 Description: Interoperability at server side (Disk interoperability)
363 Severity : enhancement
365 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
367 Severity : enhancement
369 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
371 Severity : enhancement
373 Description: New client IO stack (CLIO).
375 Severity : enhancement
377 Description: Commit on sharing. Eliminate inter-client dependencies between
378 uncommitted transactions by doing transaction commits.
379 Thereby clients may recovery independently.
382 Frequency : Create a symlink file with a very long name
384 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
385 Details : If there is no extra space in the request for early cancels,
386 ldlm_req_handles_avail() returns 0 instead of a negative value.
388 Severity : enhancement
390 Description: Add /proc entry for import status
391 Details : The mdc, osc, and mgc import directories now have
392 an import directory that contains useful import data for debugging
395 Severity : enhancement
397 Description: Re-disable certain /proc logging
398 Details : Enable and disable client's offset_stats, extents_stats and
399 extents_stats_per_process stats logging on the fly.
402 Frequency : Only on FC kernels 2.6.22+
404 Description: oops in statahead
405 Details : Do not drop reference count for the dentry from VFS when lookup,
406 VFS will do that by itself.
408 Severity : enhancement
410 Description: Generic /proc file permissions
411 Details : Set /Proc file permissions in a more generic way to enable non-
412 root users operate on some /proc files.
416 Description: Hitting mdc_commit_close() ASSERTION
417 Details : Properly handle request reference release in
418 ll_release_openhandle().
422 Description: quota recovery deadlock during mds failover
423 Details : This patch includes att18982, att18236, att18237 in bz14840.
425 1. fix osts hang when mds does failover with quotaon
426 2. prevent watchdog storm when osts threads wait for the
431 Frequency : only patchless client
432 Description: add workaround for race between add/remove dentry from hash
434 Severity : enhancement
436 Description: Allow OST glimpses to return PW locks
440 Description: LBUG when llog conf file is full
441 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
445 Description: Prevent import from entering FULL state when server in recovery
449 Description: service mount cannot take device name with ":"
450 Details : Only when device name contains ":/" will mount treat it as
456 Description: replace ptlrpcd with the statahead thread to interpret the async
457 statahead RPC callback
461 Frequency : on recovery
462 Description: I/O failures after umount during fail back
463 Details : if client reconnected to restarted server we need join to recovery
464 instead of find server handler is changed and process self
465 eviction with cancel all locks.
467 Severity : enhancement
469 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
471 Severity : enhancement
473 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
475 Severity : enhancement
477 Description: Upcall on Lustre log has been dumped
478 Details : Allow for a user mode script to be called once a Lustre log has
479 been dumped. It passes the filename of the dumped log to the
480 script, the location of the script can be specified via
481 /proc/sys/lnet/debug_log_upcall.
486 Description: avoid idr_remove called for id which is not allocated.
487 Details : Move assigment s_dev for clustered nfs to end of initialization,
488 to avoid problem with error handling.
493 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
494 Details : When connection is reused this not moved from CONN_UNUSED_HASH
495 into CONN_USED_HASH and this prodice warning when put connection
496 again in unused hash.
499 Severity : enhancement
501 Description: File striping can now be set to use an arbitrary pool of OSTs.
503 Severity : enhancement
505 Description: Export bytes_read/bytes_write count on OSC/OST.
509 Description: Early reply size mismatch, MGC loses connection
510 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
511 the connect flags are properly negotiated.
516 Description: quotas are not honored with O_DIRECT
517 Details : all writes with the flag O_DIRECT will use grants which leads to
518 this problem. Now using OBD_BRW_SYNC to guard this.
522 Description: add quota statistics
523 Details : 1. sort out quota proc entries and proc code.
524 2. add quota statistics
526 Severity : enhancement
528 Description: enable quota support for HEAD.
532 Description: Properly propagate oinfo flags from lov to osc for statfs
533 Details : restore missing copy oi_flags to lov requests.
535 Severity : enhancement
537 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
538 : l_getgroups(8), lst(8), routerstat(8)
539 Details : included man pages for llobdstat(8), llstat(8),
540 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
542 Severity : enhancement
544 Description: configure's --enable-quota should check the
545 : kernel .config for CONFIG_QUOTA
546 Details : configure is terminated if --enable-quota is passed but
547 : no quota support is in kernel
551 Description: Remove portals compatibility
552 Details : Remove portals compatibility, not interoperable with releases
557 Description: Resolve device initialization race
558 Details : Prevent proc handler from accessing devices added to the
559 obd_devs array but yet be intialized.
561 Severity : enhancement
563 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
565 Severity : enhancement
567 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
572 Description: Using wrong pointer in osc_brw_prep_request
573 Details : Access to array[-1] can produce panic if kernel compiled with
574 CONFIG_PAGE_ALLOC enabled
578 Description: Client runs out of low memory
579 Details : Consider only lowmem when counting initial number of llap pages
583 Description: *optional* service tags registration
584 Details : if the "service tags" package is installed on a Lustre node
585 When the filesystem is mounted, a local-node service tag will
586 be created. See http://inventory.sun.com/ for more information
587 about the Service Tags asset management system.
591 Description: Kernel BUG tries to release flock
592 Details : Lustre does not destroy flock lock before last reference goes
593 away. So always drop flock locks when client is evicted and
594 perform unlock regardless of successfulness of speaking to MDS.
598 Description: add refcount protection for osc callbacks, avoid panic on shutdown
602 Description: sanity test 65a fails if stripecount of -1 is set
603 Details : handle -1 striping on filesystem in ll_dirstripe_verify
608 Description: ASSERTION(CheckWriteback(page,cmd)) failed
609 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
612 Severity : enhancement
614 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
618 Description: do not process already freed flock
619 Details : flock can possibly be freed by another thread before it reaches
620 to ldlm_flock_completion_ast.
624 Description: LBUG during stress test
625 Details : Need properly lock accesses the flock deadlock detection list.
629 Description: oops in page fault handler
630 Details : kernel page fault handler can return two special 'pages' in error
631 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
635 Description: timeout with invalidate import.
636 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
637 should be handled by ptlrpcd. This produce long age waiting and
638 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
640 Severity : enhancement
642 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
646 Description: enable MGS and MDT services start separately
647 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
648 a co-located MGS without starting the MGS, which is a complement
649 to 'nosvc' mount option.
653 Frequency : after recovery
654 Description: precreate to many object's after del orphan.
655 Details : del orphan st in oscc last_id == next_id and this triger growing
656 count of precreated objects. Set flag LOW to skip increase count
657 of precreated objects.
661 Frequency : rare, on clear nid stats
662 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
663 Details : when clean nid stats sometimes try destroy live entry,
664 and this produce panic in free.
668 Description: Stack overflow during MDS log replay
669 ease stack pressure by using a thread dealing llog_process.
673 Description: wait until IO finished before start new when do lock cancel.
674 Details : VM protocol want old IO finished before start new, in this case
675 need wait until PG_writeback is cleared until check dirty flag and
676 call writepages in lock cancel callback.
678 Severity : enhancement
680 Description: using special macro for print time and cleanup in includes.
684 Description: mds_mfd_close() ASSERTION(rc == 0)
685 Details : In mds_mfd_close(), we need protect inode's writecount change
686 within its orphan write semaphore to prevent possible races.
690 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
694 Frequency : rare, on shutdown ost
695 Description: don't hit live lock with umount ost.
696 Details : shrink_dcache_parent can be in long loop with destroy dentries,
697 use shrink_dcache_sb instead.
701 Description: don't panic with use echo client
702 Details : echo client pass NULL as client nid pointer and this produce null
707 Description: fix build on ppc32
708 Details : compile code with -m64 flag produce wrong object file for ppc32.
712 Description: add message levels for liblustreapi
716 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
717 Details : Move the 'good_osts' check before the 'total_bavail' check. This
718 will result in an -EAGAIN and in the exit call path we call
719 alloc_rr() which will with increasing aggressiveness attempt to
720 aquire precreated objects on the minimum number of required OSCs.
724 Description: Use old size assignment to avoid deadlock
725 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
726 the scheduling while holding a spinlock. We do not need locking
727 for size in ll_update_inode() because size is only updated from
728 the MDS for directories or files without objects, so there is no
729 other place to do the update, and concurrent access to such inodes
730 are protected by the inode lock.
734 Description: resolve "_IOWR redefined" build error on SLES10
738 Description: dump the memory debugging after all modules are unloaded to
739 suppress false negative in conf_sanity test 39
741 Severity : enhancement
743 Description: build kernel-ib packages for OFED 1.3 in our release cycle
748 Description: fix SLES kernel versioning
749 Details : the kernel version for our SLES 10 kernel did not include a "-"
750 before the "smp" at the end. while this was not a problem in
751 general, it did mean that software trying to use the kernel
752 version to try to detect a vendor specific kernel would fail.
753 this was most evident by the OFED build scripts.
757 Description: Don't update lov_desc members until making sure they are valid
758 Details : When updating lov_desc members via proc fs, need fix their
759 validities before doing the real update.
763 Description: don't put request into delay list while invalidate in flight.
764 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
765 import in flight. this produce timeout for invalidate and sometimes
766 can cause stale data.
770 Frequency : on ppc only
771 Description: not convert ost objects for directory because it's not exist.
772 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
776 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
778 Severity : enhancement
780 Description: Update to support 2.6.22.14 vanilla kernel.
784 Frequency : rare, on recovery
785 Description: read procfs can produce deadlock in some situation
786 Details : Holding lprocfs lock which send rpc can produce block for destroy
787 obd objects and this also block reconnect with -EALREADY. This
788 isn't fix all lprocfs bugs - but make it rare.
790 Severity : enhancement
792 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
795 Frequency : frequent on X2 node
797 Description: mdc_set_open_replay_data LBUG
798 Details : Set replay data for requests that are eligible for replay.
802 Description: lustre_mgs: operation 101 on unconnected MGS
803 Details : When MGC is disconnected from MGS long enough, MGS will evict the
804 MGC, and late on MGC cannot successfully connect to MGS and a lot
805 of the error messages complaining that MGS is not connected.
809 Frequency : on network error
810 Description: panic with double free request if network error
811 Details : mdc_finish_enqueue is finish request if any network error ocuring,
812 but it's true only for synchronus enqueue, for async enqueue
813 (via ptlrpcd) this incorrect and ptlrpcd want finish request
816 Severity : enhancement
818 Description: client-side metadata stat-ahead during readdir(directory readahead)
819 Details : perform client-side metadata stat-ahead when the client detects
820 readdir and sequential stat of dir entries therein
823 Frequency : on start mds
825 Description: Implement get_info(last_id) in obdfilter.
828 Frequency : occasional
830 Description: Correctly check stale fid, not start epoch if ost not support SOM
831 Details : open with flag O_CREATE need set old fid in op_fid3 because
832 op_fid2 was overwritten with new generated fid, but mds can answer
833 with one of these two fids and both is not stale. Setattr
834 incorrectly started an epoch and assume will be called
835 done_writing, but without SOM done_writing ever being called.
838 Frequency : rare, depends on device drivers and load
840 Description: MDS or OSS nodes crash due to stack overflow
841 Details : Code changes in 1.8.0 increased the stack usage of some functions.
842 In some cases, in conjunction with device drivers that use a lot
843 of stack the MDS (or possibly OSS) service threads could overflow
844 the stack. One change which was identified to consume additional
845 stack has been reworked to avoid the extra stack usage.
848 Frequency : occasional
850 Description: Do not fail import if osc_interpret_create gets -EAGAIN
851 Details : If osc_interpret_create got -EAGAIN it immediately exits and
852 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
853 oscc_has_objects and see OSC has no objests and call
854 oscc_internal_create to resend create request.
856 Severity : enhancement
858 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
860 Severity : enhancement
862 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
865 Frequency : very rare
867 Description: Fix replay if there is an un-replied request and open
868 Details : In some cases, older replay request will revert the
869 mcd->mcd_last_xid on MDS which is used to record the client's
872 Severity : enhancement
874 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
876 Severity : enhancement
878 Description: Add rhel5 support to HEAD.
880 Serverity : enhancement
882 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
887 Description: Don't allow skipping OSTs if index has been specified.
888 Details : Don't allow skipping OSTs if index has been specified, make
889 locking in internal create lots better.
893 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
894 Details : access to bitfield in structure is always rounded to long
895 and this produce problem with not atomic change any bit.
899 Description: Lustre make rpms failed.
900 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
901 builds Lustre rpms from tarball.
904 Frequency : rare on shutdown ost
906 Description: If llog cancel was not send before clean_exports phase, this can
907 produce deadlock in llog code.
908 Details : If llog thread has last reference to obd and call class_import_put
909 this produce deadlock because llog_cleanup_commit_master wait when
910 last llog_commit_thread exited, but this never success because was
911 called from llog_commit_thread.
915 Description: allow userland application know is lost one of stripes.
916 Details : fill lvb_blocks with error code on ost and return it to
917 application if error flag found.
921 Description: NULL lov_tgts causing MDS oops
922 Details : more safe checks for NULL lov_tgts for avoid oops.
924 Severity : enhancement
926 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
930 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
931 Details : If an OST is down the MDS will hang indefinitely in
932 obd_statfs_rqset() waiting for the statfs data. While for
933 MDS QOS usage of statfs, it should not stuck in waiting.
935 Severity : enhancement
937 Description: remote_acl support
938 Details : Support ACL-based permission check for remote user.
939 Support setfacl/getfacl for remote user with the utils
940 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
941 the system "{s,g}etfacl" utils.
943 Severity : enhancement
945 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
947 Severity : enhancement
949 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
953 Description: make mgs_setparam() handle fsname containing dash
954 Details : fsname containing a dash does not work with lctl conf_param
956 Severity : enhancement
958 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
960 Severity : enhancement
962 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
966 Frequency : rare, at shutdown
967 Description: access already free / zero obd_namespace.
968 Details : if client_disconnect_export was called without force flag set,
969 and exist connect request in flight, this can produce access to
970 NULL pointer (or already free pointer) when connect_interpret
971 store ocd flags in obd_namespace.
975 Frequency : only at startup
976 Description: not alloc memory with spinlock held.
977 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
978 if any spinlock held.
980 Severity : enhancement
982 Description: make lustre randomly fail allocating memory
983 Details : Make lustre randomly failed allocating memory for testing purpose.
985 Severity : enhancement
987 Description: lost problems with lov objid file
988 Details : Fixes some scability and access to not inited memory problems
989 in work with lov objdid file.
994 Description: lfs find does not continue on file error
995 Details : Continue other files processing when a file/dir is absent.
999 Description: Inconsistent usage of lustre_pack_reply()
1000 Details : Standardize the usage of lustre_pack_reply() such that it
1001 always generate a CERROR on failure.
1004 Frequency : occasional
1006 Description: MDS hang or stay in waiting lock
1007 Details : If client receive lock with CBPENDING flag ldlm need send lock
1008 cancel as separate rpc, to avoid situation when cancel request
1009 can't processed due all i/o threads stay in wait lock.
1013 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1017 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1021 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1025 Description: Update to RHEL4 latest kernel.
1027 Severity : enhancement
1029 Description: Build SLES10 patchless client fails
1030 Details : The configure was broken by run ./configure with
1031 --with-linux-obj=.... argument for patchless client. When the
1032 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1033 can't search header adequately. Use absolute path such as
1034 -I($LINUX)/include instead.
1038 Description: interrupt oig_wait produce painc on resend.
1039 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1040 private set, and this produce situation when rq_ptlrpcd_data not
1041 copyed to new allocated request and triggered LBUG on assert
1042 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1043 ptlrpcd set if request is changed and can be safety changed to use
1048 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1049 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1050 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1051 these macros and replace them with their difinitions instead.
1055 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1056 Details : Update which_patch & target file for SLES10 latest kernel.
1059 Description: Remove client patches from SLES 10 kernel.
1060 Details : This causes SLES 10 clients to behave as patchless clients
1061 even on a Lustre-patched (server) kernel.
1063 Severity : enhancement
1065 Description: self-adjustable client's lru lists
1066 Details : use adaptive algorithm for managing client cached locks lru
1067 lists according to current server load, other client's work
1068 pattern, memory activities, etc. Both, server and client
1069 side namespaces provide number of proc tunables for controlling
1072 Severity : enhancement
1074 Description: light-weight GSS support
1075 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1076 and reduce performance overhead.
1078 Severity : enhancement
1080 Description: Linux keyring support
1081 Details : Support using service of Linux keyring for Lustre GSS internal
1082 context refresh/cache mechanism.
1086 Description: Fix errors in lfs documentation
1087 Details : Fixes man pages
1091 Description: don't use GFP_* in generic Lustre code.
1092 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1096 Description: obdclass is limited by single OBD_ALLOC(idarray)
1097 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1101 Description: warnings with build patchless client with vanila 2.6.19 and up
1102 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1103 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1107 Frequency : Only for SLES
1108 Description: sanity_quota fail test_1
1109 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1110 file, which makes TSTID[2] inunique.
1114 Frequency : Only for SLES9
1115 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1116 Details : Update target/ChangeLog/which_patch .
1120 Frequency : Only for test_10 in sanity-quota.sh
1121 Description: a bug in quota test code
1122 Details : checking a test flag in a wrong place
1126 Frequency : Only for test_13 in sanity-quota.sh
1127 Description: a bug in quota test script
1128 Details : shouldn't check overfull allocated quota
1132 Description: jbd statistics
1133 Details : Port older jbd statistics patch for sles10
1135 Severity : enhancement
1137 Description: Add printing inode info into message about error in writepage.
1142 Description: reply_lock_interpret crash due to race with it and lock cancel.
1143 Details : Do not replay locks that are being cancelled. Do not reference
1144 locks by their address during replay, just by their handle.
1148 Frequency : When flocks are used.
1149 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1150 Details : Flock locks might destroy just granted lock if it could be merged
1151 with another existing flock, this is done in completion handler,
1152 so teach ldlm_cli_enquque_fini that this is a valid case for
1158 Description: Oops in read and write path when failing to allocate lock.
1159 Details : Check if lock allocation failed and return error back.
1163 Description: lstripe command fails for valid OST index
1164 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1165 instead of lov->desc.ld_active_tgt_count.
1169 Description: sanity.sh failed test 103
1170 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1175 Description: df doesn't work properly if diskfs blocksize != 4K
1176 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1180 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1181 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1183 Severity : enhancement
1185 Description: Adaptive timeouts
1186 Details : RPC timeouts adapt to changing server load and network
1187 conditions to reduce resend attempts and improve recovery time.
1191 Description: llapi_file_create() does not allow some changes
1192 Details : add llapi_file_open() that allows specifying the mode and
1193 open flags, and also returns an open file handle.
1197 Description: merge and cleanup kernel patches.
1198 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1202 Description: Add journal checksum support.(Kernel part)
1203 Details : The journal checksum feature adds two new flags i.e
1204 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1205 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1206 indicates that the commit block contains the checksum for
1207 the blocks described by the descriptor blocks. Now commit
1208 record can be sent to disk without waiting for descriptor
1209 blocks to be written to disk. This behavior is controlled
1210 using JBD2_FEATURE_ASYNC_COMMIT flag.
1214 Description: OSS needs mutliple precreate threads
1215 Details : Add ability to start more than one create thread per OSS.
1219 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1220 Details : execute lfs setstripe on client
1224 Description: mds_obd_create error creating tmp object
1225 Details : When the user sets quota on root, llog will be affected and can't
1226 create files and write files.
1229 Frequency : Always on ia64 patchless client, and possibly others.
1231 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1232 Details : This allows the patchless client to be loaded on architectures
1233 without this export.
1237 Description: RedHat Update kernel for RHEL5
1238 Details : Add the kernel config file for RHEL5.
1242 Description: RedHat Update kernel for RHEL5
1243 Details : Modify the kernel config file more closer RHEL5.
1247 Description: Build failure against Centos5 (RHEL5)
1248 Details : Define PAGE_SIZE when it isn't present.
1252 Description: test_7 on sanity-quota.sh can't be used on separate machines
1253 Details : add facet to handle it
1257 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1258 Details : with NFS, the anon dentry's parent was set to itself in
1259 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1260 corresponding dentry other than use rec->ur_name.
1262 Severity : enhancement
1264 Description: lfs setstripe enhancement
1265 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1269 Description: enable data checksumming by default
1270 Details : enable checksum by default, allow --disable-checksum
1271 configure option and "-o nochecksum" mount option. Checksums
1272 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1276 Description: lustre support for RHEL5
1277 Details : Add support for RHEL5.
1281 Description: Client eviction due to failover config
1282 Details : after a connection loss, the lustre client should attempt to
1283 reconnect to the last active server first before trying the
1284 other potential connections.
1288 Description: when mds and osts use different quota unit(32bit and 64bit),
1289 quota will be released repeatly.
1290 Details : void sending multiple quota reqs to mds, which will keep the
1291 status between the reqs.
1295 Description: osts not allocated evenly to files
1296 Details : change the condition to increase offset_idx
1299 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1301 Description: Data corruption for OSTs that are formatted larger than 2TB
1303 Details : When generating the bio request for lustre file writes the
1304 sector number would overflow a temporary variable before being
1305 used for the IO. The data reads correctly from Lustre (which
1306 will overflow in a similar manner) but other file data or
1307 filesystem metadata may be corrupted in some cases.
1311 Description: Tune the kernel for good SCSI performance.
1312 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1313 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1318 Description: rewrite ext2-derived code in obdclass/uuid.c
1319 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1320 from scratch preserving functionality.
1324 Description: "lfs find -obd UUID" prints directories
1325 Details : "lfs find -obd UUID" will return all directory names instead
1326 of just file names. It is incorrect because the directories
1327 do not reside on the OSTs.
1331 Description: use i_size_read and i_size_write in 2.6 port
1332 Details : replace inode->i_size access with i_size_read/write()
1335 Frequency : when removing large files
1337 Description: scheduling issue during removal of large Lustre files
1338 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1339 It causes scheduling issues when removing large files (17TB in the
1343 Frequency : only with liblustre clients on XT3
1345 Description: evictions taking too long
1346 Details : allow llrd to evict clients directly on OSTs
1349 Frequency : only on ppc
1351 Description: /proc/fs/lustre/devices broken on ppc
1352 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1353 We should make sure the type of 'index' is loff_t and then cast
1354 explicitly as needed below. Do not assign an explicitly cast
1358 Frequency : only for rhel5
1360 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1361 Details : Modify the target file & which_kernel.
1363 Severity : enhancement
1365 Description: omit set fsid for export NFS
1366 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1367 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1371 Frequency : Always, for kernels after 2.6.16
1372 Description: Fix warning idr_remove called for id=.. which is not allocated.
1373 Details : Last kernels save old s_dev before kill super and not allow
1374 to restore from callback - restore it before call kill_anon_super.
1378 Description: buffer overruns could theoretically occur
1379 Details : llapi_semantic_traverse() modifies the "path" argument by
1380 appending values to the end of the origin string, and a
1381 overrun may occur. Adding buffer overrun check in liblustreapi.
1385 Description: Fix error on 'ls .' at the top of the Lustre mount.
1386 Details : Don't revalidate dentry if it is a root dentry.
1390 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1391 Details : Modify vm-tunables-rhel4.patch.
1395 Description: Kernel config for 2.6.18-vanilla.
1396 Details : Modify targets/2.6-vanilla.target.in.
1397 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1398 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1399 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1400 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1404 Description: improve handling recoverable errors
1405 Details : if request processig with error which can be recoverable on server
1406 request should be resend, otherwise page released from cache and
1411 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1412 Details : Modify target file & which_patch.
1413 A flaw was found in the IA32 system call emulation provided
1414 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1415 value could be stored in the %RAX register, which could trigger an
1416 out-of-bounds system call table access. An untrusted local user
1417 could exploit this flaw to run code in the kernel
1418 (ie a root privilege escalation). (CVE-2007-4573).
1422 Description: O_DIRECT bypasses client statistics.
1423 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1424 read_ahead_stats were updated. copied the stats section from
1425 osc_send_oap_rpc() into async_internal().
1429 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1433 Description: change order of libsysio includes
1434 Details : '#include sysio.h' should always come before '#include xtio.h'
1438 Description: handle "serious error: objid * already exists" more gracefully
1439 Details : If LAST_ID value on disk is smaller than the objects existing in
1440 the O/0/d* directories, it indicates disk corruption and causes an
1441 LBUG(). If the object is 0-length, then we should use the existing
1442 object. This will help to avoid a full fsck in most cases.
1444 Severity : enhancement
1446 Description: adapt the lustre_config script to support the upgrade case
1447 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1448 server targets to 1.6 in parallel.
1452 Description: To avoid grant space > avaible space when the disk is almost
1453 full. Without this patch you might see the error "grant XXXX >
1454 available" or some LBUG about grant, when the disk is almost
1456 Details : In filter_check_grant, for non_grant cache write, we should
1457 check the left space by if (*left > ungranted + bytes), instead
1458 of (*left > ungranted), because only we are sure the left space
1459 is enough for another "bytes", then the ungrant space should be
1460 increase. In client, we should update cl_avail_grant only there
1461 is OBD_MD_FLGRANT in the reply.
1465 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1466 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1467 described in CVE-2007-4573. This problem could allow untrusted
1468 local users to gain root access.
1471 Frequency : when using O_DIRECT and quotas
1473 Description: Incorrect file ownership on O_DIRECT output files
1474 Details : block usage reported by 'lfs quota' does not take into account
1475 files that have been written with O_DIRECT.
1480 Description: touch file failed when fs is not full
1481 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1482 otherwise we can get ENOSP while fs is not full.
1486 Description: parallel lock callbacks
1487 Details : Instead of sending blocking and completion callbacks as separated
1488 requests, adding them to a set and sending in parallel.
1491 Frequency : only for Cray XT3
1492 Bugzilla : 12829/13455
1493 Description: Changing primary group doesn't change the group lustre assigns to
1495 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1496 overridden with the primary group provided by the group upcall,
1497 whereas the supplied fsgid can be trusted if it is in the list of
1498 supplementary groups returned by the group upcall.
1500 Severity : enhancement
1502 Description: Allow masking D_WARNING, D_ERROR messages from console
1503 Details : Console messages can now be disabled via lnet.printk.
1507 Description: User code with malformed file open parameter crashes client node
1508 Details : Before packing join_file req, all the related reference should be
1509 checked carefully in case some malformed flags cause fake
1510 join_file req on client.
1514 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1515 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1516 mode is equal to the granted mode. However, it should also check
1517 that LDLM_FL_CP_REQD is not set.
1521 Description: Heavy nfs access might result in deadlocks
1522 Details : After ELC code landed, it is now improper to enqueue any mds
1523 locks under och_sem, because enqueue might want to decide to
1524 cancel open locks for same inode we are holding och_sem for.
1528 Description: Client eviction while running blogbench
1529 Details : A lot of unlink operations with concurrent I/O can lead to a
1530 deadlock causing evictions. To address the problem, the number of
1531 oustanding OST_DESTROY requests is now throttled to
1532 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1533 callbacks are processed in priority.
1537 Description: enable ACLs on MDS by default
1538 Details : ACLs must be enabled on MDS by default.
1541 Frequency : PPC/PPC64 only
1543 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1544 Details : fix duplicated definitions between asm-ppc64/types.h and
1545 lustre_types.h on PPC.
1548 Frequency : PPC/PPC64 only
1550 Description: asm-ppc/segment.h does not exist
1551 Details : fix compile issue on PPC.
1555 Description: better handle error messages in extents code
1558 Frequency : RHEL4 only
1560 Description: mkfs is very slow on IA64/RHEL4
1561 Details : A performance regression has been discovered in the MPT Fusion
1562 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1563 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1564 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1566 Severity : enhancement
1568 Description: SNMP support enhancement
1569 Details : Adding total number of sampled request for an MDS node in snmp
1572 Severity : enhancement
1574 Description: Optimize ldlm waiting list processing for PR extent locks
1575 Details : When processing waiting list for read extent lock and meeting read
1576 lock that is same or wider to it that is not contended, skip
1577 processing rest of the list and immediatelly return current
1578 status of conflictness, since we are guaranteed there are no
1579 conflicting locks in the rest of the list.
1583 Description: Time out and refuse to reconnect
1584 Details : When the failover node is the primary node, it is possible
1585 to have two identical connections in imp_conn_list. We must
1586 compare not conn's pointers but NIDs, otherwise we can defeat
1587 connection throttling.
1591 Description: port llog fixes from b1_6 into HEAD
1592 Details : Port llog reference couting and some llog cleanups from b1_6
1593 (bug 10800) into HEAD, for protect from panic and access to already
1594 free llog structures.
1598 Description: Detect stride IO mode in read-ahead
1599 Details : When a client does stride read, read-ahead should detect that and
1600 read-ahead pages according to the detected stride pattern.
1604 Description: data checksumming impacts single node performance
1605 Details : add support for several checksum algorithm. Currently, only CRC32
1606 and Adler-32 are supported. The checksum type can be changed on
1607 the fly via /proc/fs/lustre/osc/*/checksum_type.
1611 Description: use adler32 for page checksums
1612 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1617 Description: build for x2 fails
1618 Details : fix compile issue on Cray systems.
1622 Description: Properly match for duplicate locks
1623 Details : Due to different lock order from skiplists code, we need to
1624 traverse entire list for now
1627 Frequency : only on PPC/SLES10
1629 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1630 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1631 be defined. Add a hack in mkfs_lustre.c to work around this compile
1636 Description: LASSERT on MDS when client holding flock lock dies
1637 Details : ldlm pool logic depends on number of granted locks equal to
1638 number of released locks which is not true for flock locks, so
1639 just exclude such locks from consideration.
1643 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1644 Details : Use fsfilt_sync() for both the whole filesystem sync and
1645 individual file sync to eliminate dangerous inode locking
1646 with I_LOCK that can lead to a deadlock.
1650 Description: performance in 1.6.3
1651 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1652 sure that 1MB requests are not fragmented by the block layer.
1654 Severity : enhancement
1656 Description: organize the server-side client stats on per-nid basis
1657 Details : Change the structure of stats under obdfilter and mds to
1665 The "uuid"s file would list the uuids of _active_ exports.
1666 And the clear entry is to clear all stats and stale nids.
1668 Severity : enhancement
1670 Description: eliminate client locks in face of contention
1671 Details : file contention detection and lockless i/o implementation
1672 for contended files.
1676 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1679 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1681 Description: MDS may lose file striping (and hence file data) in some cases
1682 Details : If there are additional extended attributes stored on the MDS,
1683 in particular ACLs, SELinux, or user attributes (if user_xattr
1684 is specified for the client mount options) then there is a risk
1685 of attribute loss. Additionally, the Lustre file striping
1686 needs to be larger than default (e.g. striped over all OSTs),
1687 and an additional attribute must be stored initially in the
1688 inode and then increase in size enough to be moved to the
1689 external attribute block (e.g. ACL growing in size) for file
1694 Description: skiplist implementation simplification
1695 Details : skiplists are used to group compatible locks on granted list
1696 that was implemented as tracking first and last lock of each
1697 lock group the patch changes that to using doubly linked lists
1701 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1702 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1703 because LASSERT on both the data supplied by a client, and the
1704 data on disk is dangerous and incorrect.
1706 Severity : enhancement
1708 Description: Slow truncate/writes to huge files at high offsets.
1709 Details : Directly associate cached pages to lock that protect those pages,
1710 this allows us to quickly find what pages to write and remove
1711 once lock callback is received.
1715 Description: more ldlm soft lockups
1716 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1717 starve other threads from the resource lock for a long time in
1718 case of long waiting queue, so change the debug level from
1719 D_OTHER to the less frequently used D_INFO.
1721 Severity : enhancement
1723 Description: add -gid, -group, -uid, -user options to lfs find
1727 Description: Hung threads in invalidate_inode_pages2_range
1728 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1729 once one is completed. As a result, some RPCs are stuck in the
1730 queue and are never sent.
1734 Description: filter threads hungs on waiting journal commit
1735 Details : Cleanup filter group llog code, then only filter group llog will
1736 be only created in the MDS/OST syncing process.
1740 Description: Procfs and llog threads access destoryed import sometimes.
1741 Details : Sync the import destoryed process with procfs and llog threads by
1742 the import refcount and semaphore.
1744 Severity : enhancement
1746 Description: openlock cache of b1_6 port to HEAD
1751 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1752 Details : If insertion of an extent fails, then discard the inode
1753 preallocation and free data blocks else it can lead to duplicate
1758 Description: don't always update ctime in ext3_xattr_set_handle()
1759 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1760 In some cases the ctime should not be updated, for example for
1761 2.0->1.8 compatibility it is necessary to delete an xattr and it
1762 should not update the ctime.
1766 Bugzilla : 15713/16362
1767 Description: Assertion in iopen_connect_dentry in 1.6.3
1768 Details : looking up an inode via iopen with the wrong generation number can
1769 populate the dcache with a disconneced dentry while the inode
1770 number is in the process of being reallocated. This causes an
1771 assertion failure in iopen since the inode's dentry list contains
1772 both a connected and disconnected dentry.
1776 Description: assertion failure in ldlm_handle2lock()
1777 Details : fix a race between class_handle_unhash() and class_handle2object()
1778 introduced in lustre 1.6.5 by bug 13622.
1783 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1784 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1789 Description: oops in sd_iostats_seq_show()
1790 Details : unloading/reloading the scsi low level driver triggers a kernel
1791 bug when trying to access the sd iostat file.
1796 Description: Kernel panics during QLogic driver reload
1797 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1798 patch, causing memory corruption.
1803 Description: journal_dev option does not work in b1_6
1804 Details : pass mount option during pre-mount.
1806 Severity : enhancement
1808 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1809 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1810 extent information of a file. It can be used to map logical blocks
1811 in a file to physical blocks in the block device.
1815 Description: LDLM soft lockups - improvement
1816 Details : It is be possible to send the lock handle along with each read
1817 or write request because the client is already doing a lock match
1818 itself so there isn't any reason the OST should have to re-do that
1822 Frequency : only on Cray X2
1824 Description: X2 build failures
1825 Details : fix build failures on Cray X2.
1829 Description: xid & resent requests
1830 Details : Initialize RPC XID from clock at startup (randomly if clock is
1833 Severity : enhancement
1835 Description: Add lustre_start utility to start or stop multiple Lustre servers
1840 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1841 Details : In case of memory pressure, list_del() can be called twice on
1842 req->rq_history_list, causing a kernel oops.
1846 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1847 Details : fix stack overflow in the distributed lock manager by defering
1848 export eviction after a failed AST to the elt thread instead of
1849 handling it in the dlm interpret routine.
1853 Description: Convert some comments to new format.
1854 Details : Update documenting comments to match doxygen conventions.
1858 Description: Grammar fixes.
1859 Details : A couple of trivial spelling fixes.
1863 Description: OSD_COUNTERS-mandatory
1864 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1865 to simplify the code.
1869 Description: simplify cmm_device freeing logic.
1870 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1874 Description: Add lockdep support to dt_object_operations locking interface.
1875 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1876 parameter indicating lock ordering. Update mdd code to use new
1881 Description: Introduce failloc constants for lockless IO tests.
1882 Details : Add two new failloc constants to test lockless IO. Only one of
1883 them in implemented---another is checked in yet to be landed
1888 Description: Add lockdep support for inode mutex.
1889 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1890 in the situations where Lustre has to lock more than one inode
1895 Description: Add optional invariants checking support.
1896 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1897 configure switch. This macro is to be used for consistency and
1898 sanity checks that are too expensive to be left in `production'
1903 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1904 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1909 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1910 Details : Use spin_lock_nested() in (the only) situation where more than
1911 one ldlm_lock is locked simultaneously. Also, fix possible
1912 dead-lock in ldlm_lock_change_resource() by enforcing particular
1917 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1918 Details : Instead of specifying each ldlm_lock call-back through separate
1919 parameter, wrap them into struct ldlm_callback_suite.
1923 Description: Kill join_lru obd method and its callers.
1924 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1925 in memory---a requirement implemented through ->o_join_lru() obd
1926 method in HEAD. Remove this method and its users.
1930 Description: Add asynchronous ldlm ENQUEUE completion handler.
1931 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1932 case that stock ldlm_completion_ast() cannot handle as it waits
1933 until lock is granted. Introduce new ldlm_completion_ast_async()
1934 for this. Also comment ldlm_completion_ast().
1938 Description: ldlm_error <-> errno conversion.
1939 Details : Add functions to map (rather arbitrary) between LDLM error codes
1940 and standard errno values. CLIO needs this to prevent LDLM specific
1941 constants from escaping ldlm and osc.
1945 Description: Kill unused ldlm_handle2lock_ns() function.
1946 Details : Kill unused ldlm_handle2lock_ns() function.
1950 Description: Add lu_ref support to ldlm_lock
1951 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1952 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1953 and ldlm_resource. LDLM interface has to be changed, because code that
1954 releases a reference on a lock, has to "know" what reference this is.
1955 In the most frequent case
1957 lock = ldlm_handle2lock(handle);
1959 LDLM_LOCK_PUT(lock);
1961 no changes are required. When any other reference (received _not_ from
1962 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1963 instead of LDLM_LOCK_PUT().
1965 Arguably, changes are pervasive, and interface requires some discipline
1966 for proper use. On the other hand, it was very instrumental in finding
1967 a few leaked lock references.
1971 Description: Add ldlm_lock_addref_try().
1972 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1973 attempts to addref a lock that might be being canceled
1978 Description: Add ldlm_weigh_callback().
1979 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1980 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1981 hard-coded `number of pages' logic.
1985 Description: Add lockdep annotations to llog code.
1986 Details : Use appropriately tagged _nested() locking calls in the places
1987 where llog takes more than one ->lgh_lock lock.
1991 Description: Add loi_kms_set().
1992 Details : Wrap kms updates into a helper function.
1996 Description: Constify instances of struct lsm_operations.
1997 Details : Constify instances of struct lsm_operations.
2001 Description: lu_conf support.
2002 Details : On a server, a file system object is uniquely identified
2003 by a fid, which is sufficient to locate and load all object
2004 state (inode). On a client, on the other hand, more data are
2005 necessary instantiate an object. Change lu_object_find() and
2006 friends to take additional `lu_conf' argument describing object.
2007 Typically this includes layout information.
2011 Description: lu_context fixes.
2012 Details : Introduce new lu_context functions that are needed on the client
2013 side, where some system threads (ptlrpcd) are shared by multiple
2014 modules, and so cannot be stopped during module shutdown.
2018 Description: Add start and stop methods to lu_device_type_operations.
2019 Details : Introduce two new methods in lu_device_type_operations, that are
2020 invoked when first instance of a given type is created and last one
2021 is destroyed respectively. This is need by CLIO.
2025 Description: Add lu_ref support to struct lu_device.
2026 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2027 track leaked references.
2031 Description: Introduce lu_kmem_descr.
2032 Details : lu_kmem_descr and its companion interface allow to create
2033 and destroy a number of kmem caches at once.
2037 Description: Fix lu_object finalization race.
2038 Details : Fix a race between lu_object_find() finding an object and its
2039 concurrent finalization. This race is (most likely) not possible
2040 on the server, but might happen on the client.
2044 Description: Introduce lu_ref interface.
2045 Details : lu_ref is a debugging module allowing to track references to
2046 a given object. It is quite cpu expensive, and has to be
2047 explicitly enabled with --enable-lu_ref. See usage description
2052 Description: Factor lu_site procfs stats into a separate function.
2053 Details : Separate lu_site stats printing code into a separate function
2054 that can be reused on a client.
2058 Description: Constify instances of struct {lu,dt,md}_device_operations.
2059 Details : Constify instances of struct {lu,dt,md}_device_operations.
2063 Description: Introduce struct md_site and move meta-data specific parts of
2064 struct lu_site here.
2065 Details : Move md-specific fields out of struct lu_site into special struct
2066 md_site, so that lu_site can be used on a client.
2070 Description: Kill mdd_lov_destroy().
2071 Details : Remove unused mdd code.
2075 Description: Add st_block checking to multistat.c.
2076 Details : Add st_block checking to multistat.c.
2080 Description: Add lu_ref support to struct obd_device.
2081 Details : Add lu_ref tracking to obd_device.
2085 Description: Kill obd_set_fail_loc().
2086 Details : Remove unused code.
2090 Description: Add special type for ptlrpc_request interpret functions.
2091 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2092 there. Actual usage will be in CLIO.
2096 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2097 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2098 doesn't work with lockdep.
2102 Description: Add rwv.c test program.
2103 Details : New testing program exercising readv(2) and writev(2) (Qian).
2107 Description: Add sendfile.c test program.
2108 Details : New testing program exercising sendfile(2) (Jay).
2112 Description: Ratelimit a message that can be very frequent.
2113 Details : Ratelimit a memory allocation failure message that can
2118 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2119 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2123 Description: Liblustre build fixes.
2124 Details : Liblustre build fixes.
2128 Description: libcfs: add cfs_{need,cond}_resched() interface.
2129 Details : libcfs: add cfs_{need,cond}_resched() definition and
2130 implementations for Linux, NT, and liblustre.
2132 Severity : enhancement
2134 Description: More exported tunables for mballoc
2135 Details : Add support for tunable preallocation window and new tunables for
2136 large/small requests
2140 Description: Detect corruption of block bitmap and checking for preallocations
2141 Details : Checks validity of on-disk block bitmap. Also it does better
2142 checking of number of applied preallocations. When corruption is
2143 found, it turns filesystem readonly to prevent further corruptions.
2147 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2148 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2149 window to 0. If it does hit cache miss, and read-pattern is still
2150 stride-io mode, does not reset the stride window, but also does not
2151 increase the stride window length in this case.
2155 Frequency : only for big-endian servers
2156 Description: Check if system is big-endian while mounting fs with extents feature
2157 Details : Mounting a filesystem with extents feature will fail on big-endian
2158 systems since ext3-based ldiskfs is not supported on big-endian
2159 systems. This can be over-riden with "bigendian_extents" mount option.
2161 Severity : enhancement
2163 Description: The root squash functionality
2164 Details : A security feature, which is to prevent users from being able
2165 to mount lustre on their desktop, run as root, and delete
2166 all of the files in the filesystem. The goal is accomplished by
2167 remapping user id (UID) and group id (GID) of the root user to
2168 a UID and GID specified by the system administartor via Lustre
2169 configuration management server (MGS). The functionality also
2170 allows to specify sets of clients for which the remapping does
2175 Description: Excessive recovery window
2176 Details : With AT enabled, the recovery window can be excessively long (6000+
2177 seconds). To address this problem, we no longer use
2178 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2179 timeout no longer depends on the service time, it is set to
2180 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2181 time via pb_service_time.
2185 Description: Watchdog triggered on MDS failover
2186 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2187 the OSTs know that the MDS "UUID" can be reused for the same export
2188 from a different NID, so we do not need to wait for the export to be
2192 Frequency : rare, only if using MMP with Linux RAID
2194 Description: MMP doesn't work with Linux RAID
2195 Details : While using HA for Lustre servers with Linux RAID, it is possible
2196 that MMP will not detect multiple mounts. To make this work we
2197 need to unplug the device queue in RAID when the MMP block is being
2198 written. Also while reading the MMP block, we should read it from
2199 disk and not the cached one.
2201 Severity : enhancement
2203 Description: open file using fid
2204 Details : A file can be opened using just its fid, like
2205 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2208 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2211 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2212 Details : add proper locking for creation of jbd_4k slab cache
2216 Description: MMP check in ext3_remount() fails without displaying any error
2217 Details : When multiple mount protection fails during remount, proper error
2220 Severity : enhancement
2222 Description: Allow stripe size to be up to 4G-64k
2223 Details : Fix math logic to allow large stripe sizes.
2227 Description: add check for >8TB ldiskfs filesystems
2228 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2229 Don't allow >8TB ldiskfs filesystems to be mounted without
2230 force_over_8tb mount option
2235 Descriptoin: OSS complains "too many blocks" and crashes
2236 Details : In filter_commitrw_write():
2237 1. lquota_chkquota();
2238 2. obdo_from_inode();
2239 3. lquota_pending_commit();
2240 "2." has changed uid and gid in this case, then caused
2241 lquota_chkquota() and lquota_pending_commit() to use
2242 different uid and gid. That is the root of the bug.
2246 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2247 Details : Detect and print out the 'stripe count = -1' case properly.
2249 --------------------------------------------------------------------------------
2251 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2253 * Support for kernels:
2254 2.6.5-7.283 (SLES 9),
2255 2.6.9-55.EL (RHEL 4),
2256 2.6.16.46-0.14 (SLES 10),
2257 2.6.18.8 vanilla (kernel.org)
2258 * Client support for unpatched kernels:
2259 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2260 2.6.16 - 2.6.22 vanilla (kernel.org)
2261 * Due to recently discovered recovery problems, we do not recommend
2262 using patchless RHEL 4 clients with this or any earlier release.
2263 * Recommended e2fsprogs version: 1.39.cfs8
2264 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2265 * Starting with this release, the ldiskfs backing filesystem required
2266 by Lustre is now in its own package, lustre-ldiskfs. This package
2267 should be installed. It is versioned separately from Lustre and
2268 may be released separately in future.
2272 Description: block reactivating mgc import until all deactivates complete
2273 Details : Fix race when failing back MDT/MGS to itself (testing)
2276 Frequency : at statup only
2278 Description: mds_lov_synchronize race leads to various problems
2279 Details : simultaneous MDT->OST connections at startup can cause the
2280 sync to abort, leaving the OSC in a bad state.
2282 Severity : enhancement
2284 Description: add optional extra BUILD_VERSION info
2285 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2286 to override the lustre version.
2289 Frequency : 2.6.18 servers only
2291 Description: ll_kern_mount() doesn't release the module reference
2292 Details : The ldiskfs module reference count never drops down to 0
2293 because ll_kern_mount() doesn't release the module reference.
2298 Description: server LBUG when using old ost_num_threads parameter
2299 Details : Accept the old ost_num_threads parameter but warn that it
2300 is deprecated, and fix an off-by-one error that caused an LBUG.
2305 Description: Transient SCSI error results in persistent IO issue
2306 Details : iobuf->dr_error is not reinitialized to 0 between two
2310 Frequency : sometimes when underlying device returns I/O errors
2312 Description: OSTs not going read-only during write failures
2313 Details : OSTs are not remounted read-only when the journal commit threads
2314 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2315 instead of the ext3 wrappers.
2319 Description: poor connect scaling with increasing client count
2320 Details : Don't run filter_grant_sanity_check for more than 100 exports
2321 to improve scaling for large numbers of clients.
2324 Frequency : SLES10 only
2326 Description: sanity-quota.sh quotacheck failed: rc = -22
2327 Details : Quotas cannot be enabled on SLES10.
2330 Frequency : liblustre clients only
2332 Description: getdirentries does not give error when run on compute nodes
2333 Details : getdirentries does not fail when the size specified as an argument
2334 is too small to contain at least one entry
2336 Severity : enhancement
2338 Description: Add LNET router traceability for debug purposes
2339 Details : If a checksum failure occurs with a router as part of the
2340 IO path, the NID of the last router that forwarded the bulk data
2341 is printed so it can be identified.
2346 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2347 Details : Due to a race condition, liblustre clients were occasionally
2348 evicted incorrectly.
2350 Severity : enhancement
2352 Description: lfs setstripe use optional parameters instead of postional
2355 Severity : enhancement
2357 Description: Nanosecond timestamp support for ldiskfs
2358 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2359 resolution timestamps. There is not yet support for this at
2360 the Lustre filesystem level.
2363 Frequency : during server recovery
2365 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2366 Details : request with rq_no_resend flag not awake l_wait_event if they get
2370 Frequency : nfs export on patchless client
2372 Description: connectathon hang when test nfs export over patchless client
2373 Details : Disconnected dentry cannot be found with lookup, so we do not need
2374 to unhash it or make it invalid
2377 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2378 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2379 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2384 Description: open req refcounting wrong on reconnect
2385 Details : If reconnect happened between getting open reply from server and
2386 call to mdc_set_replay_data in ll_file_open, we will schedule
2387 replay for unreferenced request that we are about to free.
2388 Subsequent close will crash in variety of ways.
2389 Check that request is still eligible for replay in
2390 mdc_set_replay_data().
2395 Description: disable writes to filesystem when reading health_check file
2396 Details : the default for reading the health_check proc file has changed
2397 to NOT do a journal transaction and write to disk, because this
2398 can cause reads of the /proc file to hang and block HA state
2399 checking on a healthy but otherwise heavily loaded system. It
2400 is possible to return to the previous behaviour during configure
2401 with --enable-health-write.
2403 Severity : enhancement
2405 Description: 64-bit inode version
2406 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2407 to the inode. This will be required for version-based recovery.
2412 Description: MDS fails to start if a duplicate client export is detected
2413 Details : in some rare cases it was possible for a client to connect to
2414 an MDS multiple times. Upon recovery the MDS would detect this
2415 and fail during startup. Handle this more gracefully.
2417 Severity : enhancement
2419 Description: Add -o localflock option to simulate old noflock
2421 Details : This will achieve local-only flock/fcntl locks
2427 Description: log_commit_thread vs filter_destroy race leads to crash
2428 Details : Take import reference before releasing llog record semaphore
2433 Description: Wrong request locking in request set processing
2434 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2435 imp_delayed_list, in this place should be used imp_lock.
2438 Frequency : when reconnection
2440 Description: Grant Leak when osc reconnect to OST
2441 Details : When osc reconnect ost, OST(filter) should check whether it
2442 should grant more space to client by comparing fed_grant and
2443 cl_avail_grant, and return the granted space to client instead
2444 of "new granted" space, because client will call osc_init_grant
2445 to update the client grant space info.
2448 Frequency : when client reconnect to OST
2450 Description: Grant Leak when osc do resend and replay bulk write
2451 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2452 bulk write request, because the grant info while be sync between
2453 OSC and OST when reconnect, and we should ignore the grant info
2454 these of resend/replay write req.
2459 Description: Grant space more than avaiable left space sometimes.
2460 Details : When then OST is about to be full, if two bulk writing from
2461 different clients came to OST. Accord the avaliable space of the
2462 OST, the first req should be permitted, and the second one
2463 should be denied by ENOSPC. But if the seconde arrived before
2464 the first one is commited. The OST might wrongly permit second
2465 writing, which will cause grant space > avaiable space.
2468 Frequency : when client is evicted
2470 Description: Grant might be wrongly erased when osc is evicted by OST
2471 Details : when the import is evicted by server, it will fork another
2472 thread ptlrpc_invalidate_import_thread to invalidate the
2473 import, where the grant will be set to 0. While the original
2474 thread will update the grant it got when connecting. So if
2475 the former happened latter, the grant will be wrongly errased
2476 because of this race.
2481 Description: Checking Stale with correct fid
2482 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2483 to check whether it is stale, because sometimes, we want the
2484 enqueue happened anyway, and op_data.fid2 will not be initialized.
2486 Severity : enhancement
2488 Description: update patchless client
2489 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2492 Frequency : only with 2.4 kernel
2494 Description: random memory corruption
2495 Details : size of struct ll_inode_info is to big for union inode.u and this
2496 can be cause of random memory corruption.
2501 Description: Memory leak in recovery
2502 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2503 It should also check obd_fail before fsfilt_start, otherwise if
2504 fsfilt_start return -EROFS,(failover mds during mds recovery).
2505 then the req will return with repmsg->transno = 0 and rc = EROFS.
2506 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2507 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2508 be freed no matter whether fsfilt_commit success or not.
2511 Frequency : only with huge count clients
2513 Description: Prevents from taking the superblock lock in llap_from_page for
2515 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2516 for llap_from_page call in ll_removepage prevents from taking the
2517 superblock lock for a soon died page.
2522 Description: Not check open intent error before release open handle
2523 Details : in some rare cases, the open intent error is not checked before
2524 release open handle, which may cause
2525 ASSERTION(open_req->rq_transno != 0), because it tries to release
2526 the failed open handle.
2531 Description: Set cat log bitmap only after create log success.
2532 Details : in some rare cases, the cat log bitmap is set too early. and it
2533 should be set only after create log success.
2537 Description: Accessing a block bevice can re-enable I/O when Lustre is
2538 tearing down a device.
2539 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2544 Description: service threads may hog cpus when there are a lot of requests
2545 Details : Insert cond_resched to give other threads a chance to use some CPU
2550 Description: the cat log was not initialized in recovery
2551 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2552 unlink log on mds will not be initialized until mds post
2553 recovery. And also in mds post recovery, the unlink log will
2554 initialization will be done asynchronausly, so there will be race
2555 between add unlink log and unlink log initialization.
2559 Description: brw_stats were being printed incorrectly
2560 Details : brw_stats were being printed as log2 but all of them were not
2561 recorded as log2. Also remove some code duplication arising from
2562 filter_tally_{read,write}.
2566 Frequency : rare, only in recovery.
2567 Description: ASSERTION(req->rq_type != LI_POISON) failed
2568 Details : imp_lock should be held while iterating over imp_sending_list for
2569 prevent destroy request after get timeout in ptlrpc_queue_wait.
2573 Description: replay-single.sh test 52 fails
2574 Details : A lock's skiplist need to be cleanup when it being unlinked
2575 from its resource list.
2579 Description: Short directio read returns full requested size rather than
2581 Details : Direct I/O operations should return actual amount of bytes
2582 transferred rather than requested size.
2584 Severity : enhancement
2586 Description: metadata RPC reduction (e.g. for rm performance)
2587 Details : decrease the amount of synchronous RPC between clients and servers
2588 by canceling conflicing lock before the operation on the client
2589 and packing thier handles into the main operation RPC to server.
2591 Severity : enhancement
2593 Description: Async OSC create to avoid the blocking unnecessarily.
2594 Details : If a OST has no remain object, system will block on the creating
2595 when need to create a new object on this OST. Now, ways use
2596 pre-created objects when available, instead of blocking on an
2597 empty osc while others are not empty. If we must block, we block
2598 for the shortest possible period of time.
2602 Description: improve handling recoverable errors
2603 Details : if request processig with error which can be recoverable on server
2604 request should be resend, otherwise page released from cache and
2607 Severity : enhancement
2609 Description: refine locking for avoid write wrong info into lov_objid file
2610 Details : fix possible races with add new target and write/update data in
2613 --------------------------------------------------------------------------------
2615 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2620 Frequency : on some architectures
2622 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2623 Details : Uninitialized flags sometimes cause configuration commands to
2627 Frequency : patchless clients only
2629 Description: missing __iget() symbol export
2630 Details : The __iget() symbol export is missing. To avoid the need for
2631 this on patchless clients the deathrow inode reaper is turned
2632 off, and we depend on the VM to clean up old inodes. This
2633 dependency was during via the fix for bug 12181.
2635 --------------------------------------------------------------------------------
2637 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2639 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2640 INTEROPERATE with older versions automatically. In many cases a
2641 special upgrade step is needed. Please read the
2642 user documentation before upgrading any part of a live system.
2643 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2644 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2645 * WARNING: Lustre configuration and startup changes are required with
2646 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2650 Severity : enhancement
2652 Description: MountConf
2653 Details : Lustre configuration is now managed via mkfs and mount
2654 commands instead of lmc and lconf. New obd types (MGS, MGC)
2655 are added for dynamic configuration management. See
2656 https://mail.clusterfs.com/wikis/lustre/MountConf for
2659 Severity : enhancement
2661 Description: dynamic OST addition
2662 Details : OSTs can now be added to a live filesystem
2664 Severity : enhancement
2666 Description: startup order invariance
2667 Details : MDTs and OSTs can be started in any order. Clients only
2668 require the MDT to complete startup.
2670 Severity : enhancement
2672 Description: parallel, asynchronous orphan cleanup
2673 Details : orphan cleanup is now performed in separate threads for each
2674 OST, allowing parallel non-blocking operation.
2676 Severity : enhancement
2678 Description: optimized stripe assignment
2679 Details : stripe assignments are now made based on ost space available,
2680 ost previous usage, and OSS previous usage, in order to try
2681 to optimize storage space and networking resources.
2683 Severity : enhancement
2685 Description: Permanently set tunables
2686 Details : All writable /proc/fs/lustre tunables can now be permanently
2687 set on a per-server basis, at mkfs time or on a live system.
2689 Severity : enhancement
2691 Description: Lustre message v2
2692 Details : Add lustre message format v2.
2694 Severity : enhancement
2696 Description: client OST exclusion list
2697 Details : Clients can be started with a list of OSTs that should be
2698 declared "inactive" for known non-responsive OSTs.
2702 Description: SPEC SFS validation failure on NFS v2 over lustre.
2703 Details : Changes the blocksize for regular files to be 2x RPC size,
2704 and not depend on stripe size.
2706 Severity : enhancement
2708 Description: Multiple MD RPCs in flight.
2709 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2710 To avoid overly-overloading MDS, introduce a limit on number of
2711 MDS RPCs in flight for a single client and add /proc controls
2712 to adjust this limit.
2714 Severity : enhancement
2716 Description: client read/write statistics
2717 Details : Add client read/write call usage stats for performance
2718 analysis of user processes.
2719 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2720 file access. extents_stats shows chunk size distribution.
2721 extents_stats_per_process show chunk size distribution per
2724 Severity : enhancement
2726 Description: mds statistics
2727 Details : Add detailed mds operations statistics in
2728 /proc/fs/lustre/mds/*/stats.
2732 Description: Failure of copying files with lustre special EAs.
2733 Details : Client side always return success for setxattr call for lustre
2734 special xattr (currently only "trusted.lov").
2739 Description: Refcount LNET uuids
2740 Details : The global LNET uuid list grew linearly with every startup;
2741 refcount repeated list entries instead of always adding to
2744 Severity : enhancement
2746 Description: Dynamic service threads
2747 Details : Within a small range, start extra service threads
2748 automatically when the request queue builds up.
2751 Frequency : mixed-endian client/server environments
2753 Description: mixed-endian crashes
2754 Details : The new msg_v2 system had some failures in mixed-endian
2757 Severity : enhancement
2759 Description: Easy OST removal
2760 Details : OSTs can be permanently deactivated with e.g. 'lctl
2761 conf_param lustre-OST0001.osc.active=0'
2763 Severity : enhancement
2765 Description: MGS proc entries
2766 Details : Added basic proc entries for the MGS showing what filesystems
2769 Severity : enhancement
2771 Description: provide MGS failover
2772 Details : Added config lock reacquisition after MGS server failover.
2774 Severity : enhancement
2776 Description: add Linux 2.4 support
2777 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2781 Description: a large application tries to do I/O to the same resource and dies
2782 in the middle of it.
2783 Details : Check the req->rq_arrival time after the call to
2784 ost_brw_lock_get(), but before we do anything about
2785 processing it & sending the BULK transfer request. This
2786 should help move old stale pending locks off the queue as
2787 quickly as obd_timeout.
2790 Frequency : when an incorrect nid is specified during startup
2792 Description: ptlrpc connect to non-existant node causes kernel crash
2793 Details : LNET can't be re-entered from an event callback, which
2794 happened when we expire a message after the export has been
2795 cleaned up. Instead, hand the zombie cleanup off to another
2798 Severity : enhancement
2800 Description: plain/inodebits lock performance improvement
2801 Details : Grouping plain/inodebits in granted list by their request modes
2802 and bits policy, thus improving the performance of search through
2806 Frequency : only if OST filesystem is corrupted
2808 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2809 Details : for a short time RPCs with bulk IO are in the replay list,
2810 but replay of bulk IOs is unimplemented. If the OST filesystem
2811 is corrupted due to disk cache incoherency and then replay is
2812 started it is possible to trip an assertion. Avoid putting
2813 committed RPCs into the replay list at all to avoid this issue.
2816 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2818 Description: System hang on startup
2819 Details : This bug allowed the liblustre (e.g. catamount) client to
2820 return to the app before handling all startup RPCs. This
2821 could leave the node unresponsive to lustre network traffic
2822 and manifested as a server ptllnd timeout.
2824 Severity : enhancement
2826 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2827 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2828 boolean to control whether to print peer debug info when a
2829 client's RPC times out.
2832 Frequency : only for kernels with patches from Lustre below 1.4.3
2834 Description: Remove old rdonly API
2835 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2838 Frequency : only for devices with external journals
2840 Description: Set external device read-only also
2841 Details : During a commanded failover stop, we set the disk device
2842 read-only while the server shuts down. We now also set any
2843 external journal device read-only at the same time.
2846 Frequency : when upgrading from 1.4 while trying to change parameters
2848 Description: The wrong (new) MDC name was used when setting parameters for
2849 upgraded MDT's. Also allows changing of OSC (and MDC)
2850 parameters if --writeconf is specified at tunefs upgrade time.
2853 Frequency : when setting specific ost indicies
2855 Description: QOS code breaks on skipped indicies
2856 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2857 created with --index need not be sequential.
2859 Severity : enhancement
2861 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2862 Details : The uninit_groups feature works in conjunction with the kernel
2863 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2864 up the pass1 processing of e2fsck. This is a read-only feature
2865 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2866 mount filesystems that have had this feature enabled.
2868 Severity : enhancement
2870 Description: Improve multi-block allocation algorithm to avoid fragmentation
2871 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2872 allocation locality and avoid filesystem fragmentation.
2874 ------------------------------------------------------------------------------
2876 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2878 * Support for kernels:
2879 2.4.21-47.0.1.EL (RHEL 3)
2880 2.6.5-7.283 (SLES 9)
2881 2.6.9-42.0.10.EL (RHEL 4)
2882 2.6.12.6 vanilla (kernel.org)
2883 2.6.16.27-0.9 (SLES 10)
2884 * Recommended e2fsprogs version: 1.39.cfs5
2886 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2890 Frequency : occasional, depends on client load and configuration
2891 Bugzilla : 12181, 12203
2892 Description: data loss for recently-modified files
2894 Details : In some cases it is possible that recently written or created
2895 files may not be written to disk in a timely manner (this should
2896 normally be within 30s unless client IO load is very high).
2897 The problem appears as zero-length files or files that are a
2898 multiple of 1MB in size after a client crash or client eviction
2899 that are missing data at the end of the file.
2901 This problem is more likely to be hit on clients where files are
2902 repeatedly created and unlinked in the same directory, clients
2903 have a large amount of RAM, have many CPUs, the filesystem has
2904 many OSTs, the clients are rebooted frequently, and/or the files
2905 are not accessed by other nodes after being written.
2907 The presence of the problem can be detected by looking at
2908 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2909 smaller than the second (nr_unused) then dirty files will not
2910 be flushed automatically to disk. "sync; sleep 10" should be
2911 run several times on the node before unmounting it to update
2912 Lustre (this is also safe to run on nodes without this problem).
2914 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2915 that can cause this same problem, so customers using that kernel
2916 also need to update the kernel in addition to Lustre. In order
2917 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2919 It is normal that files written just before a client crash (less
2920 than 30s) may not yet have been flushed to disk, even for local
2924 Frequency : frequent on thin XT3 nodes
2926 Description: UUID collision on thin XT3 Linux nodes
2927 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2928 randomly, since we relied on an insufficiently-seeded PRNG.
2933 Description: OSS hangs after "All ost request buffers busy"
2934 Details : A deadlock between quota and journal operations caused OSS
2935 hangs after printing "All ost request buffers busy."
2938 Frequency : always on liblustre builds
2940 Description: Cleanup compiler warnings on liblustre
2943 Frequency : always on liblustre builds on XT3
2945 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2949 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2950 > 32000 subdirectories
2951 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2952 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2953 subdirectory count crosses 32000. This will aid e2fsck to
2954 correctly handle more than 32000 subdirectories.
2957 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2959 Description: System hang on startup
2960 Details : This bug allowed the liblustre (e.g. catamount) client to
2961 return to the app before handling all startup RPCs. This
2962 could leave the node unresponsive to lustre network traffic
2963 and manifested as a server ptllnd timeout.
2965 Severity : enhancement
2967 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2968 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2969 boolean to control whether to print peer debug info when a
2970 client's RPC times out.
2975 Description: make O_SYNC working on 2.6 kernels
2976 Details : 2.6 kernels use different method for mark pages for write,
2977 so need add a code to lustre for O_SYNC work.
2982 Description: Failure to close file and release space on NFS
2983 Details : Put inode details into lock acquired in ll_intent_file_open.
2984 Use mdc_intent_lock in ll_intent_open to properly
2985 detect all kind of errors unhandled by mdc_enqueue.
2990 Description: proc file read during shutdown sometimes raced obd removal,
2992 Details : Add lock to prevent obd access after proc file removal.
2995 Frequency : Only for files larger than 4GB on 32-bit clients.
2997 Description: improperly doing page alignment of locks
2998 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2999 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3004 Description: RPCs being resent when they shouldn't be
3005 Details : Some RPCs that should not be resent are being resent. This
3006 can cause inconsistencies in the RPC state machine. Do not
3007 resend such requests.
3010 Frequency : rare, only with NFS export
3012 Description: Crash on NFS re-export node
3013 Details : under very unusual load conditions an assertion is hit in
3014 ll_intent_file_open()
3017 Frequency : only if OST filesystem is corrupted
3019 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3020 Details : for a short time RPCs with bulk IO are in the replay list,
3021 but replay of bulk IOs is unimplemented. If the OST filesystem
3022 is corrupted due to disk cache incoherency and then replay is
3023 started it is possible to trip an assertion. Avoid putting
3024 committed RPCs into the replay list at all to avoid this issue.
3029 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3030 Details : Large single O_DIRECT read and write calls can fail to allocate
3031 a sufficiently large buffer to process the request. In case of
3032 allocation failure the allocation is retried with a smaller
3033 buffer and broken into smaller requests.
3035 Severity : enhancement
3037 Description: Add -o localflock option to simulate old noflock behaviour.
3038 Details : This will achieve local-only flock/fcntl locks coherentness.
3043 Description: versioning check is incomplete
3044 Details : Checking the version difference of client vs. server, report
3045 error if the gap is too big.
3050 Description: add support PG_writeback bit
3051 Details : add support for PG_writeback bit for Lustre, for more carefull
3052 work with page cache in 2.6 kernel. This also fix some deadlocks
3053 and remove hack for work O_SYNC with 2.6 kernel.
3055 Severity : enhancement
3057 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3058 Details : The uninit_groups feature works in conjunction with the kernel
3059 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3060 up the pass1 processing of e2fsck. This is a read-only feature
3061 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3062 mount filesystems that have had this feature enabled.
3064 Severity : enhancement
3066 Description: Improve multi-block allocation algorithm to avoid fragmentation
3067 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3068 allocation locality and avoid filesystem fragmentation.
3070 ------------------------------------------------------------------------------
3072 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3074 * Support for kernels:
3075 2.6.9-42.0.3.EL (RHEL 4)
3076 2.6.5-7.276 (SLES 9)
3077 2.4.21-47.0.1.EL (RHEL 3)
3078 2.6.12.6 vanilla (kernel.org)
3079 2.6.16.21-0.8 (SLES10)
3080 * Recommended e2fsprogs version: 1.39.cfs2-0
3082 * The backwards-compatible /proc/sys/portals symlink has been removed
3083 in this release. Before upgrading, please ensure that you change
3084 any configuration scripts or /etc/sysctl.conf files that access
3085 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3086 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3087 in advance of the upgrade on any system running Lustre 1.4.6 or
3088 newer, since /proc/sys/lnet was added in that version.
3089 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3093 Frequency : only when quota is used
3095 Description: avoid scanning export list for quota master
3096 Details : Change the algorithms to avoid scanning export list in order
3097 to improve the efficiency.
3100 Frequency : MDS failover only, very rarely
3102 Description: "went back in time" messages on mds failover
3103 Details : The greatest transno may be lost when the current operation
3104 finishes with an error (transno==0) and the client's last_rcvd
3105 record is over-written. Save the greatest transno in the
3106 mds_last_transno for this case.
3109 Frequency : always for specific kernels and striping counts
3111 Description: client may get "Matching packet too big" without ACL support
3112 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3113 when trying to access files in certain configurations. The
3114 clients should in fact be denied when mounting because they do
3115 not understand ACLs.
3118 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3120 Description: many clients connecting with IO in progress causes connect timeouts
3121 Details : Avoid synchronous journal commits to avoid delays caused by many
3122 clients connecting/disconnecting when bulk IO is in progress.
3123 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3124 OST_IO_PORTAL to avoid delays behind potentially many pending
3128 Frequency : occasionally with multiple writers to a single file
3130 Description: shared writes to file may result in wrong size reported by stat()
3131 Details : Allow growing of kms when extent lock is cancelled
3134 Frequency : always with random mmap IO to multi-striped file
3136 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3137 Details : Only if the hole is at the end of OST object so that kms is too
3138 small. Fix is to increase kms accordingly in ll_nopage.
3141 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3143 Description: writes to a missing object would leak memory on the OST
3144 Details : If there is an inconsistency between the MDS and OST filesystems,
3145 such that the MDS references an object that doesn't exist, writes
3146 to that object will leak memory due to incorrect cleanup in the
3147 error handling path, eventually running out of memory on the OST.
3152 Description: Creating too long symlink causes lustre errors
3153 Details : Check symlink and name lengths before sending requests to MDS.
3156 Frequency : only if flock is enabled (not on by default)
3158 Description: posix locks not released on fd closure on 2.6.9+
3159 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3160 kernels, this caused such locks not to be released on fd close and
3161 then assertions on fs unmount about still used locks.
3164 Frequency : MDS failover only, very rarely
3166 Description: clients may get ASSERTION(granted_lock != NULL)
3167 Details : When request was taking a long time, and a client was resending
3168 a getattr by name lock request. The were multiple lock requests
3169 with the same client lock handle and
3170 mds_getattr_name->fixup_handle_for_resent_request found one of the
3171 lock handles but later failed with ASSERTION(granted_lock != NULL).
3176 Description: handle->h_buffer_credits > 0, assertion failure
3177 Details : h_buffer_credits is zero after truncate, causing assertion
3178 failure. This patch extends the transaction or creates a new
3182 Frequency : NFS re-export or patchless client
3183 Bugzilla : 11179, 10796
3184 Description: Crash on NFS re-export node (__d_move)
3185 Details : We do not want to hash the dentry if we don't have a lock.
3186 But if this dentry is later used in d_move, we'd hit uninitialised
3187 list head d_hash, so we just do this to init d_hash field but
3188 leave dentry unhashed.
3191 Frequency : NFS re-export or patchless client
3193 Description: NFS exports has problem with symbolic link
3194 Details : lustre client didn't properly install dentry when re-exported
3195 to NFS or running patchless client.
3198 Frequency : NFS re-export or patchless client
3200 Description: Various nfs/patchless fixes.
3201 Details : fixes reuse disconected alias for lookup process - this fixes
3202 warning "find_exported_dentry: npd != pd",
3203 fix permission error with open files at nfs.
3204 fix apply umask when do revalidate.
3207 Frequency : occasional
3209 Description: Crash on NFS re-export node
3210 Details : calling clear_page() on the wrong pointer triggered oops in
3211 generic_mapping_read().
3214 Frequency : rarely, using O_DIRECT IO
3216 Description: unaligned directio crashes client with LASSERT
3217 Details : check for unaligned buffers before trying any requests.
3220 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3222 Description: stale data returned from RAID cache
3223 Details : If only a small amount of IO is done to the RAID device before
3224 reading it again it is possible to get stale data from the RAID
3225 cache instead of reading it from disk.
3228 Frequency : always for sles10 kernel
3230 Description: sles10 support
3231 Details : ll_follow_link: compile fixes and using of nd_set_link
3232 under newer kernels.
3235 Frequency : depends on arch, kernel and compiler version, always on sles10
3238 Description: recursive or deep enough symlinks cause stack overflow
3239 Details : getting rid of large stack-allocated variable in
3243 Frequency : depends on hardware
3245 Description: lustre write performance loss in the SLES10 kernel
3246 Details : the performance loss is caused by using of write barriers in the
3247 ext3 code. The SLES10 kernel turns barrier support on by
3248 default. The fix is to undo that change for ldiskfs.
3250 ------------------------------------------------------------------------------
3252 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3254 * Support for kernels:
3255 2.6.9-42.0.3EL (RHEL 4)
3256 2.6.5-7.276 (SLES 9)
3257 2.4.21-47.0.1.EL (RHEL 3)
3258 2.6.12.6 vanilla (kernel.org)
3262 Frequency : quota enabled and large files being deleted
3264 Description: releasing more than 4GB of quota at once hangs OST
3265 Details : If a user deletes more than 4GB of files on a single OST it
3266 will cause the OST to spin in an infinite loop. Release
3267 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3272 Description: statfs data retrieved from /proc may be stale or zero
3273 Details : When reading per-device statfs data from /proc, in the
3274 {kbytes,files}_{total,free,avail} files, it may appear
3275 as zero or be out of date.
3278 Frequency : systems with MD RAID1 external journal devices
3280 Description: lconf's call to blkid is confused by RAID1 journal devices
3281 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3282 of returning all block devices that match the journal UUID.
3285 Frequency : always, for aggregate stripe size over 4GB
3287 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3288 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3289 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3292 Frequency : always if "lfs find" used on a local file/directory
3294 Description: "lfs find" segfaults if used on a local file/directory
3295 Details : The case where a directory component was not specified wasn't
3296 handled correctly. Handle this properly.
3299 Frequency : always on ppc64
3301 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3302 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3303 perform endianness conversion.
3306 Frequency : rarely (truncate to non-zero file size after write under load)
3307 Bugzilla : 10730, 10687
3308 Description: Files padded with zeros to next 4K multiple
3309 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3310 it is possible that files that are truncated to a non-zero size
3311 immediately after being written are filled with zero bytes beyond
3312 the truncated size. No file data is lost.
3314 Severity : enhancement
3316 Description: Allow recovery/failover for liblustre clients.
3317 Details : liblustre clients were unaware of failover configurations until
3320 Severity : enhancement
3322 Description: user file locks should fail when not mounting with flock option
3323 Details : Set up an error-returning stub in ll_file_operations.lock field
3324 to prevent incorrect behaviour when client is mounted without
3325 flock option. Also, set up properly f_op->flock field for
3329 Frequency : always on ia64
3331 Description: "lfs df" loops on printing out MDS statfs information
3332 Details : The obd_ioctl_data was not initialized and in some systems
3333 this caused a failure during the ioctl that did not return
3334 an error. Initialize the struct and return an error on failure.
3337 Frequency : SLES 9 only
3339 Description: Error of copying files with lustre special EAs as root
3340 Details : Client side always return success for setxattr call for lustre
3341 special xattr (currently only "trusted.lov").
3344 Frequency : rarely on clusters with both ia64+i386 clients
3346 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3347 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3348 clients are doing concurrent writes to the same file it is
3349 possible that the ia64 clients may LASSERT because the OST
3350 extent locks are not PAGE_SIZE aligned. Ensure that grown
3351 locks are always aligned on the request boundary.
3354 Frequency : specific use, occasional
3356 Description: Overwriting in use executable truncates on-disk binary image
3357 Details : If one node attempts to overwrite an executable in use by
3358 another node, we now correctly return ETXTBSY instead of
3359 truncating the file.
3361 Severity : enhancement
3363 Description: Async OSC create to avoid the blocking unnecessarily.
3364 Details : If a OST has no remain object, system will block on the creating
3365 when need to create a new object on this OST. Now, ways use
3366 pre-created objects when available, instead of blocking on an
3367 empty osc while others are not empty. If we must block, we block
3368 for the shortest possible period of time.
3373 Description: chmod on Lustre root is propagated to other clients
3374 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3375 invalid by the follow_mount time.
3378 Frequency : liblustre clients only
3380 Description: Race in 'instant cancel' lock handling could lead to such locks
3381 never to be granted in case of SMP MDS
3382 Details : Do not destroy not yet granted but cbpending locks in
3386 Frequency : replay/resend of open
3388 Description: non null lock assetion failure in mds_intent_policy
3389 Details : Trying to replay/resend lockless open requests resulted in
3390 mds_open() returning 0 with no lock. Now it sets a flag if
3391 a lock is going to be returned.
3393 Severity : enhancement
3395 Description: Checksum enhancements
3396 Details : New checksum enhancements allow for resending RPCs that failed
3399 Severity : enhancement
3401 Description: Tunables on number of dirty pages in cacche
3402 Details : Allow to set limit on number of dirty pages cached.
3407 Description: client crash on unmount - lock still has references
3408 Details : In some error handling cases it was possible to leak a lock
3409 reference on a client while accessing a file. This was not
3410 harmful to the client during operation, but would cause the
3411 client to crash when the filesystem is unmounted.
3414 Frequency : specific case, rare
3416 Description: ETXTBSY on mds though file not in use by client
3417 Details : ETXTBSY is no longer incorrectly returned when attempting to
3418 chmod or chown a directory that the user previously tried to
3419 execute or a currently-executing binary.
3422 Frequency : extremely rare except on liblustre-based clients
3424 Description: Lustre space not freed when files are deleted
3425 Details : Clean up open-unlinked files after client eviction. Previously
3426 the unlink was skipped and the files remained as orphans.
3431 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3432 Details : In some cases it was possible to send a blocking callback to
3433 a client doing a glimpse, even though that client didn't get
3434 a lock granted. When the glimpse lock is cancelled on the OST
3435 the freed lock is left on the waiting list and corrupted the list.
3438 Frequency : all core dumps
3440 Description: Broke core dumps to lustre
3441 Details : Negative dentry may be unhashed if parent does not have UPDATE
3442 lock, but some callers, e.g. do_coredump, expect dentry to be
3443 hashed after successful create, hash it in ll_create_it.
3445 ------------------------------------------------------------------------------
3447 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3449 * Support for kernels:
3450 2.6.9-42.0.2.EL (RHEL 4)
3451 2.6.5-7.276 (SLES 9)
3452 2.4.21-40.EL (RHEL 3)
3453 2.6.12.6 vanilla (kernel.org)
3457 Frequency : always on RHEL 3
3459 Description: Number of open files grows over time
3460 Details : The number of open files grows over time, whether or not
3461 Lustre is started. This was due to a filp leak introduced
3462 by one of our kernel patches.
3464 ------------------------------------------------------------------------------
3466 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3468 * Support for kernels:
3469 2.6.9-42.EL (RHEL 4)
3470 2.6.5-7.276 (SLES 9)
3471 2.4.21-40.EL (RHEL 3)
3472 2.6.12.6 vanilla (kernel.org)
3477 Bugzilla : 5719, 9635, 9792, 9684
3478 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3479 Details : If a server is under heavy load and cannot reply to new
3480 connection requests before the client resends the (re)connect,
3481 the connection handling code can behave badly if two service
3482 threads are concurrently handing separate (re)connections from
3483 the same client. Add better locking to the connection handling
3484 code, and ensure that only a single connection will be processed
3485 for a given client UUID, even if the lock is dropped.
3487 Severity : enhancement
3489 Description: add TCP zero-copy support to kernel
3490 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3491 sends if the hardware supports scatter-gather and checksumming.
3492 This allows socklnd to do client-write and server-read more
3493 efficiently and reduce CPU utilization from skbuf copying.
3496 Frequency : only if NFS exporting from client
3498 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3499 Details : If setting attributes on a file created under NFS that had
3500 never been opened it would be possible to oops the client
3501 if the file had no objects.
3504 Frequency : always for liblustre
3506 Description: liblustre client does MDS+OSTs setattr RPC for each write
3507 Details : When doing a write from a liblustre client, the client
3508 incorrectly issued an RPC to the MDS and each OST the file was
3509 striped over in order to update the timestamps. When writing
3510 with small chunks and many clients this could overwhelm the MDS
3511 with RPCs. In all cases it would slow down the write because
3512 these RPCs are unnecessary.
3514 Severity : enhancement
3516 Description: allow number of MDS service threads to be changed at module load
3517 Details : It is now possible to change the number of MDS service threads
3518 running. Adding "options mds mds_num_threads={N}" to the MDS's
3519 /etc/modprobe.conf will set the number of threads for the next
3520 time Lustre is restarted (assuming the "mds" module is also
3521 reloaded at that time). The default number of threads will
3522 stay the same, 32 for most systems.
3527 Description: OST crash if filesystem is unformatted or corrupt
3528 Details : If an OST is started on a device that has never been formatted
3529 or if the filesystem is corrupt and cannot even mount then the
3530 error handling cleanup routines would dereference a NULL pointer.
3535 Description: NULL pointer deref in llap_from_page.
3536 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3537 incorrect) mapping if the page was truncated/reclaimed while it was
3538 searched for. Check for this condition and skip such pages when
3539 doing readahead. Introduce extra check to llap_from_page() to
3540 verify page->mapping->host is non-NULL (so page is not anonymous).
3543 Frequency : Sometimes when using sys_sendfile
3545 Description: "page not covered by a lock" warnings from ll_readpage
3546 Details : sendfile called ll_readpage without right page locks present.
3547 Now we introduced ll_file_sendfile that does necessary locking
3548 around call to generic_file_sendfile() much like we do in
3552 Frequency : with certain MDS communication failures at client mount time
3554 Description: NULL pointer deref after failed client mount
3555 Details : a client connection request may delayed by the network layer
3556 and not be sent until after the PTLRPC layer has timed out the
3557 request. If the client fails the mount immediately it will try
3558 to clean up before the network times out the request. Add a
3559 reference from the request import to the obd device and delay
3560 the cleanup until the network drops the request.
3563 Frequency : occasionally during client (re)connect
3565 Description: assertion failure during client (re)connect
3566 Details : processing a client connection request may be delayed by the
3567 client or server longer than the client connect timeout. This
3568 causes the client to resend the connection request. If the
3569 original connection request is replied in this interval, the
3570 client may trip an assertion failure in ptlrpc_connect_interpret()
3571 which thought it would be the only running connect process.
3574 Frequency : only with obd_echo servers and clients that are rebooted
3576 Description: kernel BUG accessing uninitialized data structure
3577 Details : When running an obd_echo server it did not start the ping_evictor
3578 thread, and when a client was evicted an uninitialized data
3579 structure was accessed. Start the ping_evictor in the RPC
3580 service startup instead of the OBD startup.
3582 Severity : enhancement
3583 Bugzilla : 10193 (patchless)
3584 Description: Remove dependency on various unexported kernel interfaces.
3585 Details : No longer need reparent_to_init, exit_mm, exit_files,
3586 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3589 Frequency : rare (only users of deprecated and unsupported LDAP config)
3591 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3592 Details : LDAP apparently contains 'lustreName' attributes instead of
3593 'name'. A simple remapping of the name is sufficient.
3596 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3598 Description: waiting_locks_callback trips kernel BUG if client is evicted
3599 Details : Running with the dump_on_timeout debug flag turned on makes
3600 it possible that the waiting_locks_callback() can try to dump
3601 the Lustre kernel debug logs from an interrupt handler. Defer
3602 this log dumping to the expired_lock_main() thread.
3604 Severity : enhancement
3606 Description: Support NFS exporting on 2.6 kernels.
3607 Details : Implement non-rawops metadata methods for NFS server to use without
3608 changing NFS server code.
3611 Frequency : very rare (synthetic metadata workload only)
3613 Description: two racing renames might cause an MDS thread to deadlock
3614 Details : Running the "racer" program may cause one MDS thread to rename
3615 a file from being the source of a rename to being the target of
3616 a rename at exactly the same time that another thread is doing
3617 so, and the second thread has already enqueued these locks after
3618 doing a lookup of the target and is trying to relock them in
3619 order. Ensure that we don't try to re-lock the same resource.
3622 Frequency : only very large systems with liblustre clients
3624 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3625 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3626 all OSTs in parallel. This allows the eviction of stale liblustre
3627 clients to proceed much faster than if they were done in series,
3628 and also offers similar improvements for other set_info RPCs.
3633 Description: excessive CPU usage during initial read phase on client
3634 Details : During the initial read phase on a client, it would agressively
3635 retry readahead on the file, consuming too much CPU and impacting
3636 performance (since 1.4.5.8). Improve the readahead algorithm
3637 to avoid this, and also improve some other common cases (read
3638 of small files in particular, where "small" is files smaller than
3639 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3644 Description: MDS crash when receiving packet with unknown intent.
3645 Details : Do not LBUG in unknown intent case, just return -EFAULT
3647 Severity : enhancement
3648 Bugzilla : 9293, 9385
3649 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3650 Details : Do not serialize getattr (non-intent version) and statfs.
3653 Frequency : occasional, when OST network is overloaded/intermittent
3655 Description: client evicted by OST after bulk IO timeout
3656 Details : If a client sends a bulk IO request (read or write) the OST
3657 may evict the client if it is unresposive to its data GET/PUT
3658 request. This is incorrect if the network is overloaded (takes
3659 too long to transfer the RPC data) or dropped the OST GET/PUT
3660 request. There is no need to evict the client at all, since
3661 the pinger and/or lock callbacks will handle this, and the
3662 client can restart the bulk request.
3665 Frequency : Always when mmapping file with no objects
3667 Description: client crashes when mmapping file with no objects
3668 Details : Check that we actually have objects in a file before doing any
3669 operations on objects in ll_vm_open, ll_vm_close and
3675 Description: Request leak when working with deleted CWD
3676 Details : Introduce advanced request refcount tracking for requests
3677 referenced from lustre intent.
3679 Severity : Enhancement
3681 Description: Cache open file handles on client.
3682 Details : MDS now will return special lock along with openhandle, if
3683 requested and client is allowed to hold openhandle, even if unused,
3684 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3685 closing files for every read/write openration.
3687 Severity : Enhancement
3689 Description: Cache open negative dentries on client when possible.
3690 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3691 negative dentries on lock revocation.
3696 Description: Remounting a client read-only wasn't possible with a zconf mount
3697 Details : It wasn't possible to remount a client read-only with llmount.
3699 Severity : enhancement
3700 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3701 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3702 MPICH is included to add Lustre-specific ADIO interfaces.
3703 This is based closely on the UFS ADIO layer and only differs
3704 in file creation, in order to allow the OST striping to be set.
3705 This is user-contributed code and not supported by CFS.
3710 Description: extended inode attributes (immutable, append-only) work improperly
3711 when 2.4 and 2.6 kernels are used on client/server or vice versa
3712 Details : Introduce kernel-independent values for these flags.
3714 Severity : enhancement
3717 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3718 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3719 granularity. Now, Fractional MB granularity is be supported,
3720 this is very useful for low memory system.
3722 Severity : enhancement
3724 Description: Getattr by fid
3725 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3726 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3727 have locks and avoids getattr rpc requests for it on every stat.
3730 Frequency : Always, for filesystems larger than 2TB
3732 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3733 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3734 filesystems larger than 2TB to be mounted with mballoc enabled.
3737 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3739 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3740 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3741 filesystems larger than 2TB is mounted then the kernel will
3742 silently corrupt the start of the filesystem. CONFIG_LBD is
3743 enabled for all CFS-supported kernels, but the possibility of
3744 this happening with a modified kernel config exists.
3746 Severity : enhancement
3748 Description: add client O_DIRECT support for 2.6 kernels
3749 Details : It is now possible to do O_DIRECT reads and writes to files
3750 in the Lustre client mountpoint on 2.6 kernel clients.
3752 Severity : enhancement
3754 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3755 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3756 parallel, not waiting for response from every OST before sending
3757 a rpc to the next OST.
3762 Description: setattr vs write race when updating file timestamps
3763 Details : Client processes that update a file timestamp into the past
3764 right after writing to the file (e.g. tar) it is possible that
3765 the updated file modification time can be reset to the current
3766 time due to a race between processing the setattr and write RPC.
3768 Severity : enhancement
3770 Description: Bring 'lfs find' closer in line with regular Linux find.
3771 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3772 -print0 options and obtains all the needed info through the lustre
3775 Severity : enhancement
3777 Description: support up to 1024 configured devices on one node
3778 Details : change obd_dev array from statically allocated to dynamically
3779 allocated structs as they are first used to reduce memory usage
3784 Description: Flush dirty partially truncated pages during truncate
3785 Details : Immediatelly flush partially truncated pages in filter_setattr,
3786 this way we completely avoid having any pages in page cache on OST
3787 and can retire ugly workarounds during writes to flush such pages.
3792 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3793 Details : protect inode from truncation within vfs_unlink() context
3794 just take a reference before calling vfs_unlink() and release it
3795 when parent's i_sem is free.
3800 Description: last_id value checked outside lock on OST caused LASSERT failure
3801 Details : If there were multiple MDS->OST object precreate requests in
3802 flight, it was possible that the OST's last object id was checked
3803 outside a lock and incorrectly tripped an assertion. Move checks
3804 inside locks, and discard old precreate requests.
3807 Frequency : always, if extents are used on OSTs
3809 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3810 Details : OSTs using the extents format would not zero the high 16 bits of
3811 the index physical block number. This is not a problem for any
3812 OST filesystems smaller than 16TB, and no kernels support ext3
3813 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3814 new/modified files) and can be fixed for existing filesystems
3815 with e2fsprogs-1.39-cfs1.
3820 Description: import connection selection may be incorrect if timer wraps
3821 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3822 import connections to be ignored if the 32-bit jiffies counter
3823 wraps. Use a 64-bit jiffies counter.
3826 Frequency : very large clusters immediately after boot
3828 Description: LNET request buffers exhausted under heavy short-term load
3829 Details : If a large number of client requests are generated on a service
3830 that has previously never seen so many requests it is possible
3831 that the request buffer growth cannot keep up with the spike in
3832 demand. Instead of dropping incoming requests, they are held in
3833 the LND until the RPC service can accept more requests.
3836 Frequency : Sometimes during replay
3838 Description: Assertion failure in ll_local_open after replay.
3839 Details : If replay happened on an open request reply before we were able
3840 to set replay handler, reply will become not swabbed tripping the
3841 assertion in ll_local_open. Now we set the handler right after
3842 recognising of open request
3845 Frequency : very rare
3847 Description: kernel reports "badness in vsnprintf"
3848 Details : Reading from the "recovery_status" /proc file in small chunks
3849 may cause a negative length in lprocfs_obd_rd_recovery_status()
3850 call to vsnprintf() (which is otherwise harmless). Exit early
3851 if there is no more space in the output buffer.
3853 Severity : enhancement
3855 Description: clear OBD RPC statistics by writing to them
3856 Details : It is now possible to clear the OBD RPC statistics by writing
3857 to the "stats" file.
3862 Description: Client mtime is not the same on different clients after utimes
3863 Details : In some cases, the client was using the utimes() syscall on
3864 a file cached on another node. The clients now validate the
3865 ctime from the MDS + OSTs to determine which one is right.
3870 Description: Inability to activate failout mode
3871 Details : lconf script incorrectly assumed that in pythong string's numeric
3872 value is used in comparisons.
3875 Frequency : always with multiple stripes per file
3877 Description: Inefficient object allocation for mutli-stripe files
3878 Details : When selecting which OSTs to stripe files over, for files with
3879 a stripe count that divides evenly into the number of OSTs,
3880 the MDS is always picking the same starting OST for each file.
3881 Return the OST selection heuristic to the original design.
3886 Description: mount failures may take full timeout to return an error
3887 Details : Under some heavy load conditions it is possible that a
3888 failed mount can wait for the full obd_timeout interval,
3889 possibly several minutes, before reporting an error.
3890 Instead return an error as soon as the status is known.
3892 Frequency : quota enabled and large files being deleted
3894 Description: releasing more than 4GB of quota at once hangs OST
3895 Details : If a user deletes more than 4GB of files on a single OST it
3896 will cause the OST to spin in an infinite loop. Release
3897 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3902 Description: statfs data retrieved from /proc may be stale or zero
3903 Details : When reading per-device statfs data from /proc, in the
3904 {kbytes,files}_{total,free,avail} files, it may appear
3905 as zero or be out of date.
3908 Frequency : systems with MD RAID1 external journal devices
3910 Description: lconf's call to blkid is confused by RAID1 journal devices
3911 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3912 of returning all block devices that match the journal UUID.
3915 Frequency : always, for aggregate stripe size over 4GB
3917 Description: assertion fails when trying to use 4GB stripe size
3918 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3919 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3922 Frequency : always on ppc64
3924 Description: the first write on an ext3 filesystem with mballoc got stuck
3925 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3926 perform endianness conversion.
3928 ------------------------------------------------------------------------------
3930 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3932 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3933 INTEROPERATE with older versions automatically. Please read the
3934 user documentation before upgrading any part of a live system.
3935 * WARNING: Lustre networking configuration changes are required with
3936 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3939 * Support for kernels:
3940 2.6.9-22.0.2.EL (RHEL 4)
3941 2.6.5-7.244 (SLES 9)
3942 2.6.12.6 vanilla (kernel.org)
3945 Severity : enhancement
3946 Bugzilla : 7981/8208
3947 Description: Introduced Lustre Networking (LNET)
3948 Details : LNET is new networking infrastructure for Lustre, it includes
3949 a reorganized network configuration mode (see the user
3950 documentation for full details) as well as support for routing
3951 between different network fabrics. Lustre Networking Devices
3952 (LNDs) for the supported network fabrics have also been
3953 created for this new infrastructure.
3955 Severity : enhancement
3956 Description: Introduced Access control lists
3957 Details : clients can set ACLs on files and directories in order to have
3958 more fine-grained permissions than the standard Unix UGO+RWX.
3959 The MDS must be started with the "-o acl" mount option.
3961 Severity : enhancement
3962 Description: Introduced filesystem quotas
3963 Details : Administrators may now establish per-user quotas on the
3966 Severity : enhancement
3968 Description: Configuration change for the XT3
3969 The PTLLND is now used to run Lustre over Portals on the XT3
3970 The configure option(s) --with-cray-portals are no longer used.
3971 Rather --with-portals=<path-to-portals-includes> is used to
3972 enable building on the XT3. In addition to enable XT3 specific
3973 features the option --enable-cray-xt3 must be used.
3978 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3979 Details : A race between allocating a new llcd and re-getting the llcd_lock
3980 allowed another thread to grab newly-allocated llcd.
3982 Severity : enhancement
3984 Description: 2.6 OST async journal commit and locking fix to improve performance
3985 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3986 2.6 kernels are now async as they already were in 2.4 kernels so
3987 that they can commit concurrently with the network bulk transfer.
3988 For block-allocated files the filter allocation semaphore is held
3989 to avoid filesystem fragmentation during allocation. BKL lock
3990 removed for 2.6 xattr operations where it is no longer needed.
3995 Description: lconf incorrectly determined whether two IP networks could talk
3996 Details : In some more complicated routing and multiple-network
3997 configurations, lconf will avoid trying to make a network
3998 connection to a disjoint part of the IP space. It was doing the
3999 math incorrectly for one set of cases.
4004 Description: Fix for potential infinite loop processing records in an llog.
4005 Details : If an llog record is corrupted/zeroed, it is possible to loop
4006 forever in llog_process(). Validate the llog record length
4007 and skip the remainder of the block on error.
4010 Frequency : occasional (liblustre only)
4012 Description: liblustre could not open files whose last component is a symlink
4013 Details : sysio_path_walk() would incorrectly pass the open intent to
4014 intermediate path components.
4017 Frequency : rare (liblustre only with non-standard tuning)
4018 Bugzilla : 7201 (7350)
4019 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4020 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4021 after it has been released, only do this for referenced locks
4023 Severity : enhancement
4025 Description: specifying an (invalid) directory default stripe_size of -1
4026 would reset the directory default striping
4027 Details : stripe_size -1 was used internally to signal directory stripe
4028 removal, now use "all default" to signal dir stripe removal
4029 as a directory striping of "all default" is not useful
4032 Frequency : common for large clusters running liblustre clients
4034 Description: doing an ls when liblustre clients are running is slow
4035 Details : sending a glimpse AST to a liblustre client waits for every AST
4036 to time out, as liblustre clients will not respond. Since they
4037 cannot cache data we refresh the OST lock LVB from disk instead.
4039 Severity : enhancement
4041 Description: doing an ls at the same time as file IO can be slow
4042 Details : enqueue and other "small" requests can be blocked behind many
4043 large IO requests. Create a new OST IO portal for non-IO
4044 requests so they can be processed faster.
4047 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4049 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4050 created file with mode 000
4051 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4052 allow an ftruncate() as a non-root user to a file with mode 000.
4053 Lustre now respects this flag to disable mode checks when
4054 truncating a file owned by the user
4057 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4060 Description: Revoking locks from clients that went dead or catatonic might take
4062 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4063 cancellation of such locks instant on servers without waiting for
4064 any reply from clients. Clients drops these locks when cancel
4065 notification from server is received without replying.
4068 Frequency : liblustre-only, when liblustre client dies or becomes busy
4070 Description: Doing ls on Linux clients can take a long time with active
4072 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4073 granting such locks to it in the first place if possible. Locks
4074 are taken by proxy on the OST during the read or write and
4075 dropped immediately afterward. Add connect flags handling, do
4076 not grant locks to liblustre clients for glimpse ASTs.
4078 Severity : enhancement
4080 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4081 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4082 sequential reads and behaves badly with applications doing random
4083 reads. Improve it by reading ahead at least read region, and
4084 avoiding excessive large RPC for small reads.
4086 Severity : enhancement
4088 Description: Creating more than 1000 files for a single job may cause a load
4089 imbalance on the OSTs if there are also a large number of OSTs.
4090 Details : qos_prep_create() uses an OST index reseed value that is an
4091 even multiple of the number of available OSTs so that if the
4092 reseed happens in the middle of the object allocation it will
4093 still utilize the OSTs as uniformly as possible.
4098 Description: OST or MDS may oops in ping_evictor_main()
4099 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4100 but doesn't restart at beginning of obd_exports_timed list
4103 Severity : enhancement
4105 Description: improve by-nid export eviction on the MDS and OST
4106 Details : allow multiple exports with the same NID to be evicted at one
4107 time without re-searching the exports list.
4110 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4112 Description: MDS may oops in groups_free()
4113 Details : in rare race conditions a newly allocated group_info struct is
4114 freed again, and this can be NULL. The 2.4 compatibility code
4115 for groups_free() checked for a NULL pointer, but 2.6 did not.
4118 Frequency : common for liblustre clients doing little filesystem IO
4119 Bugzilla : 9352, 7313
4120 Description: server may evict liblustre clients accessing contended locks
4121 Details : if a client is granted a lock or receives a completion AST
4122 with a blocking AST already set it would not reply to the AST
4123 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4126 Frequency : lfs setstripe, only systems with more than 160 OSTs
4128 Description: unable to set striping with a starting offset beyond OST 160
4129 Details : llapi_create_file() incorrectly limited the starting stripe
4130 index to the maximum single-file stripe count.
4133 Frequency : LDAP users only
4135 Description: lconf did not handle in-kernel recovery with LDAP properly
4136 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4138 Severity : enhancement
4140 Description: bind OST threads to NUMA nodes to improve performance
4141 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4142 node and do their allocations there to localize memory access
4144 Severity : enhancement
4146 Description: llmount can determine client NID directly from Myrinet (GM)
4147 Details : the client NID code from gmnalnid was moved directly into
4148 llmount, removing the need to use this or specifying the
4149 client NID explicitly when mounting GM clients with zeroconf
4152 Frequency : if client is started with down MDS
4154 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4155 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4156 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4157 ptlrpc_import_delayed_req() because we didn't check if the
4158 request was interrupted, and we also didn't break out of the
4159 event loop if there was no timeout
4164 Description: data loss during non-page-aligned writes to a single file from
4165 both multiple nodes and multiple threads on one node at same time
4166 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4167 inconsistency led to false short-reads, that were cached and later
4168 used by ->prepare_write() to fill in partially written page,
4169 leading to data loss.
4172 Frequency : always, if lconf --abort_recovery used
4174 Description: lconf --abort_recovery fails with 'Operation not supported'
4175 Details : lconf was attempting to abort recovery on the MDT device and not
4178 Severity : enhancement
4180 Description: remove cleanup logs
4181 Details : replace lconf-generated cleanup logs with lustre internal
4182 cleanup routines. Eliminates the need for client-cleanup and
4185 Severity : enhancement
4187 Description: add support for EAs (user and system) on lustre filesystems
4188 Details : it is now possible to store extended attributes in the Lustre
4189 client filesystem, and with the user_xattr mount option it
4190 is possible to allow users to store EAs on their files also
4192 Severity : enhancement
4194 Description: Add possibility (config option) to show minimal available OST free
4196 Details : When compiled with --enable-mindf configure option, statfs(2)
4197 (and so, df) will return least minimal free space available from
4198 all OSTs as amount of free space on FS, instead of summary of
4199 free spaces of all OSTs.
4201 Severity : enhancement
4203 Description: do not expand extent locks acquired on OST-side
4204 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4205 by server: they are not cached anyway.
4208 Frequency : when mmap is used/binaries executed from Lustre
4210 Description: Unmmap pages before throwing them away from read cache.
4211 Details : llap_shrink cache now attempts to unmap pages before discarding
4212 them (if unmapping failed - do not discard). SLES9 kernel has
4213 extra checks that trigger if this unmapping is not done first.
4218 Description: lconf didn't resolve symlinks before checking to see whether a
4219 given mountpoint was already in use
4222 Frequency : when migrating failover services
4223 Bugzilla : 6395, 9514
4224 Description: When migrating a subset of services from a node (e.g. failback
4225 from a failover service node) the remaining services would
4226 time out and evict clients.
4227 Details : lconf --force (implied by --failover) sets the global obd_timeout
4228 to 5 seconds in order to quickly disconnect, but this caused
4229 other RPCs to time out too quickly. Do not change the global
4230 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4232 Severity : enhancement
4233 Frequency : if MDS is started with down OST
4234 Bugzilla : 9439,5706
4235 Description: Allow startup/shutdown of an MDS without depending on the
4236 availability of the OSTs.
4237 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4238 Add appropriate locking and lov-osc refcounts for safe
4239 cleaning. Add osc abort_inflight calls in case the
4240 synchronize never started.
4243 Frequency : occasional (Cray XT3 only)
4245 Description: root not authorized to access files in CRAY_PORTALS environment
4246 Details : The client process capabilities were not honoured on the MDS in
4247 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4248 been accessed by an authorized user then root was able to access
4249 the file on the local client also. The root user capabilities
4250 are now allowed on the MDS, as this environment has secure UID.
4253 Frequency : occasional
4255 Description: ldiskfs "too long searching" message happens too often
4256 Details : A debugging message (otherwise harmless) prints too often on
4257 the OST console. This has been reduced to only happen when
4258 there are fragmentation problems on the filesystem.
4263 Description: Division by zero in statfs when all OSCs are inactive
4264 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4265 lov_statfs divides by value returned by lov_get_stripecnt().
4269 Bugzilla : 9489, 3273
4270 Description: First write from each client to each OST was only 4kB in size,
4271 to initialize client writeback cache, which caused sub-optimal
4272 RPCs and poor layout on disk for the first writen file.
4273 Details : Clients now request an initial cache grant at (re)connect time
4274 and so that they can start streaming writes to the cache right
4275 away and always do full-sized RPCs if there is enough data.
4276 If the OST is rebooted the client also re-establishes its grant
4277 so that client cached writes will be honoured under the grant.
4282 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4283 Details : Now I/O RPCs go to different portal number and (presumably) fast
4284 lock requests (and glimses) and other RPCs get their own service
4285 threads pool that should be able to service those RPCs
4288 Severity : enhancement
4290 Description: Ability to exchange lustre version between client and servers and
4291 issue warnings at client side if client is too old. Also for
4292 liblustre clients there is ability to refuse connection of too old
4294 Details : New 'version' field is added to connect data structure that is
4295 filled with version info. That info is later checked by server and
4299 Frequency : rare, liblustre only.
4300 Bugzilla : 9296, 9581
4301 Description: Two simultaneous writes from liblustre at offset within same page
4302 might proceed at the same time overwriting eachother with stale
4304 Details : I/O lock withing llu_file_prwv was released too early, before data
4305 actually was hitting the wire. Extended lock-holding time until
4306 server acknowledges receiving data.
4309 Frequency : extremely rare. Never observed in practice.
4311 Description: avoid generating lustre_handle cookie of 0.
4312 Details : class_handle_hash() generates handle cookies by incrementing
4313 global counter, and can hit 0 occasionaly (this is unlikely, but
4314 not impossible, because initial value of cookie counter is
4315 selected randonly). Value of 0 is used as a sentinel meaning
4316 "unassigned handle" --- avoid it. Also coalesce two critical
4317 sections in this function into one.
4319 Severity : enhancement
4321 Description: allow liblustre clients to delegate truncate locking to OST
4322 Details : To avoid overhead of locking, liblustre client instructs OST to
4323 take extent lock in ost_punch() on client's behalf. New connection
4324 flag is added to handle backward compatibility.
4326 Severity : enhancement
4327 Bugzilla : 4928, 7341, 9758
4328 Description: allow number of OST service threads to be specified
4329 Details : a module parameter allows the number of OST service threads
4330 to be specified via "options ost ost_num_threads={N}" in the
4331 OSS's /etc/modules.conf or /etc/modprobe.conf.
4335 Bugzilla : 6146, 9635, 9895
4336 Description: servers crash with bad pointer in target_handle_connect()
4337 Details : In rare cases when a client is reconnecting it was possible that
4338 the connection request was the last reference for that export.
4339 We would temporarily drop the export reference and get a new
4340 one, but this may have been the last reference and the export
4341 was just destroyed. Get new reference before dropping old one.
4343 Severity : enhancement
4344 Frequency : if client is started with failover MDS
4346 Description: Allow multiple MDS hostnames in the mount command
4347 Details : Try to read the configuration from all specified MDS
4348 hostnames during a client mount in case the "primary"
4351 Severity : enhancement
4353 Description: Stop sending data to evicted clients as soon as possible.
4354 Details : Check if the client we are about to send or are sending data to
4355 was evicted already. (Check is done every second of waiting,
4356 for which l_wait_event interface was extended to allow checking
4357 of exit condition at specified intervals).
4360 Frequency : rare, normally only when NFS exporting is done from client
4362 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4364 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4367 Frequency : always, if srand() is called before liblustre initialization
4369 Description: Liblustre uses system PRNG disturbing its usage by user application
4370 Details : Introduce internal to lustre fast and high-quality PRNG for
4371 lustre usage and make liblustre and some other places in generic
4372 lustre code to use it.
4374 Severity : enhancement
4375 Bugzilla : 9477, 9557, 9870
4376 Description: Verify that the MDS configuration logs are updated when xml is
4377 Details : Check if the .xml configuration logs are newer than the config
4378 logs stored on the MDS and report an error if this is the case.
4379 Request --write-conf, or allow starting with --old_conf.
4381 Severity : enhancement
4383 Description: Handle symlinks in the path when checking if Lustre is mounted.
4384 Details : Resolve intermediate symlinks when checking if a client has
4385 mounted a filesystem to avoid duplicate client mounts.
4390 Description: lconf can hit an error exception but still return success.
4391 Details : The lconf command catches the Command error exception at the top
4392 level script context and will exit with the associated exit
4393 status, but doesn't ensure that this exit status is non-zero.
4398 Description: failure of ptlrpc thread startup can cause oops
4399 Details : Starting a ptlrpc service thread can fail if there are a large
4400 number of threads or the server memory is very fragmented.
4401 Handle this without oopsing.
4404 Frequency : always, only if liblustre and non-default acceptor port was used
4406 Description: liblustre cannot connect to servers with non-default acceptor port
4407 Details : tcpnal_set_default_params() was not called and was therefore
4408 ignoring the environment varaible TCPNAL_PORT, as well as other
4409 TCPNAL_ environment variables
4414 Description: two objects could be created on the same OST for a single file
4415 Details : If an OST is down, in some cases it was possible to create two
4416 objects on a single OST for a single file. No problems other
4417 than potential performance impact and spurious error messages.
4421 Bugzilla : 5681, 9562
4422 Description: Client may oops in ll_unhash_aliases
4423 Details : Client dcache may become inconsistent in race condition.
4424 In some cases "getcwd" can fail if the current directory is
4430 Description: Inode refcounting problems in NFS export code
4431 Details : link_raw functions used to call d_instantiate without obtaining
4432 extra inode reference first.
4436 Bugzilla : 9942, 9903
4437 Description: Referencing freed requests leading to crash, memleaks with NFS.
4438 Details : We used to require that call to ll_revalidate_it was always
4439 followed by ll_lookup_it. Also with revalidate_special() it is
4440 possible to call ll_revalidate_it() twice for the same dentry
4441 even if first occurence returned success. This fix changes semantic
4442 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4443 reference on a request referred from the intent.
4444 ll_intent_release() then releases such a request.
4447 Frequency : rare, normally benchmark loads only
4449 Description: unlinked inodes were kept in memory on the client
4450 Details : If a client is repeatedly creating and unlinking files it
4451 can accumulate a lot of stale inodes in the inode slab cache.
4452 If there is no other client load running this can cause the
4453 client node to run out of memory. Instead flush old inodes
4454 from client cache that have the same inode number as a new inode.
4457 Frequency : SLES9 2.6.5 kernel and long filenames only
4458 Bugzilla : 9969, 10379
4459 Description: utime reports stale NFS file handle
4460 Details : SLES9 uses out-of-dentry names in some cases, which confused
4461 the lustre dentry revalidation. Change it to always use the
4465 Frequency : rare, unless heavy write-truncate concurrency is continuous
4466 Bugzilla : 4180, 6984, 7171, 9963, 9331
4467 Description: OST becomes very slow and/or deadlocked during object unlink
4468 Details : filter_destroy() was holding onto the parent directory lock
4469 while truncating+unlinking objects. For very large objects this
4470 may block other threads for a long time and slow overall OST
4471 responsiveness. It may also be possible to get a lock ordering
4472 deadlock in this case, or run out of journal credits because of
4473 the combined truncate+unlink. Solution is to do object truncate
4474 first in one transaction without parent lock, and then do the
4475 final unlink in a new transaction with the parent lock. This
4476 reduces the lock hold time dramatically.
4479 Frequency : rare, 2.4 kernels only
4481 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4482 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4483 before unmounting their filesystem in order to avoid deadlock.
4484 The kernel_locked() function in 2.4 kernels only checks whether
4485 the kernel lock is held, not whether it is this process that is
4486 holding it as 2.6 kernels do.
4491 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4492 Details : The client ptlrpc code may be trying to reconnect to a down
4493 server before a previous connection attempt has timed out.
4494 Increase the reconnect interval to be longer than the connection
4495 timeout interval to avoid sending duplicate connections to
4499 Frequency : echo_client brw_test command
4501 Description: fix echo_client to work with OST preallocated code
4502 Details : OST preallocation code (5137) didn't take echo_client IO path
4503 into account: echo_client calls filter methods outside of any
4504 OST thread and, hence, there is no per-thread preallocated
4505 pages and buffers to use. Solution: hijack pga pages for IO. As
4506 a byproduct, this avoids unnecessary data copying.
4510 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4511 Description: Client can oops in mdc_commit_close() after open replay
4512 Details : It was possible for the MDS to return an open request with no
4513 transaction number in mds_finish_transno() if the client was
4514 evicted, but without actually returning an error. Clients
4515 would later try to replay that open and may trip an assertion
4516 Simplify the client close codepath, and always return an error
4517 from the MDS in case the open is not successful.
4520 Frequency : rare, 2.6 OSTs only
4522 Description: OST may deadlock under high load on fragmented files
4523 Details : If there was a heavy load and highly-fragmented OST filesystems
4524 it was possible to have all the OST threads deadlock waiting on
4525 allocation of biovecs, because the biovecs were not released
4526 until the entire RPC IO was completed. Instead, release biovecs
4527 as soon as they are complete to ensure forward IO progress.
4529 Severity : enhancement
4531 Description: Support for specifying external journal device at mount
4532 Details : If an OST or MDS device is formatted with an external journal
4533 device, this device major/minor is stored in the ext3 superblock
4534 and may not be valid for failover. Allow detecting and
4535 specifying the external journal at mount time.
4540 Description: Mounting an MDS with pending unlinked files may cause oops
4541 Details : target_finish_recovery() calls mds_postrecov() which returned
4542 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4543 considers this an error and immediately begins cleaning up the
4544 lov, just after starting the mds_lov process
4546 Severity : enhancement
4548 Description: Implement 'lfs df' to report actual free space on per-OST basis
4549 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4550 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4551 usage of objects; '-h' to report in human readable format.
4553 ------------------------------------------------------------------------------
4555 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4562 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4563 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4564 at startup during orphan unlinking. Instead, skip buddy update
4565 before setup, as it will be regenerated after recovery is complete.
4570 Description: If an OST is inactive, its locks might reference stale inodes.
4571 Details : lov_change_cbdata() must iterate over all namespaces, even if
4572 they are inactive to clear inode references from the lock.
4574 Severity : enhancement
4575 Frequency : occasional, if non-standard max_dirty_mb used
4577 Description: Client will block write RPCs if not enough grant
4578 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4579 then the client will block writes while waiting for another RPC
4580 to complete instead of consuming its dirty limit. With change
4581 we get improved performance when max_dirty_mb is small.
4583 Severity : enhancement
4584 Bugzilla : 3389, 6253
4585 Description: Add support for supplementary groups on the MDS.
4586 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4587 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4588 lookups for user supplementary groups into a cache.
4592 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4593 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4594 for liblustre O_EXCL is enforced by MDS.
4598 Description: readdir never returns NULL in liblustre.
4599 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4600 dentry in struct dirent.
4604 Frequency : liblustre only, depends on application IO pattern
4605 Description: liblustre clients evicted if not contacting servers
4606 Details : Don't put liblustre clients into the ping_evictor list, so
4607 they will not be evicted by the pinger ever.
4609 Severity : enhancement
4611 Description: Add ability to evict clients by NID from MDS.
4612 Details : By echoing "nid:$NID" string into
4613 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4614 $NID would be instantly evicted from this MDS and from all active
4615 OSTs connected to it.
4619 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4620 Details : lookup_it_finish() used to query file size from OSTs that was not
4625 Description: service threads change working directory to that of init
4626 Details : Starting lustre service threads may pin the working directory
4627 of the parent thread, making that filesystem busy. Threads
4628 now change to the working directory of init to avoid this.
4632 Frequency : during shutdown only
4633 Description: shutdown with a failed MDS or OST can cause unmount to hang
4634 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4639 Frequency : 2.6 only
4640 Description: chmod/chown may include an extra supplementary group
4641 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4642 supplementary group and if none is specified this is used.
4645 Bugzilla : 5479 (6816)
4647 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4648 Details : If lookup is in progress on a file that is unlinked we might try
4649 to revalidate the inode and fail in revalidate after lookup is
4650 complete and ll_file_open() enqueues the open again but
4651 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4654 Frequency : always, if lconf --abort_recovery used
4656 Description: lconf --abort_recovery fails with 'Operation not supported'
4657 Details : lconf was attempting to abort recovery on the MDT device and not
4660 ------------------------------------------------------------------------------
4662 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4667 Frequency : rare (only unsupported configurations with a node running as an
4669 Bugzilla : 6514, 5137
4670 Description: Mounting a Lustre file system on a node running as an OST could
4672 Details : OSTs now preallocates memory needed to write out data at
4673 startup, instead of when needed, to avoid having to
4674 allocate memory in possibly low memory situations.
4675 Specifically, if the file system is mounted on on OST,
4676 memory pressure could force it to try to write out data,
4677 which it needed to allocate memory to do. Due to the low
4678 memory, it would be unable to do so and the node would
4679 become unresponsive.
4681 Severity : enhancement
4683 Description: Addition of lconf --service command line option
4684 Details : lconf now accepts a '--service <arg>' option, which is
4685 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4687 Severity : enhancement
4689 Description: Failover mode is now the default for OSTs.
4690 Details : By default, OSTs will now run in failover mode. To return to
4691 the old behaviour, add '--failout' to the lmc line for OSTs.
4693 Severity : enhancement
4695 Description: Health checks are now provided for MDS and OSTs
4696 Details : Additional detailed health check information on MSD and OSTs
4697 is now provided through the procfs health_check value.
4700 Frequency : occasional, depends on IO load
4702 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4703 after numerous create/delete cycles
4704 Details : The ext3 inode allocation policy would not allocate new inodes
4705 very well on the OSTs because there are no new directories
4706 being created. Instead we look for groups with free space if
4707 the parent directories are nearly full.
4712 Description: Network or server problems during mount may cause partially
4713 mounted clients instead of returning an error.
4714 Details : The config llog parsing code may overwrite the error return
4715 code during mount error handling, returning success instead
4721 Description: MDS can fail to allocate large reply buffers
4722 Details : After long uptimes the MDS can fail to allocate large reply
4723 buffers (e.g. zconf client mount config records) due to memory
4724 fragmentation or consumption by the buffer cache. Preallocate
4725 some large reply buffers so that these replies can be sent even
4726 under memory pressure.
4730 Frequency : rare (liblustre)
4731 Description: fsx running with liblustre complained that using truncate() to
4732 extend the file doesn't work. This patch corrects that issue.
4733 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4734 ATTR_SIZE and lsm use in llu_setattr_raw.
4738 Frequency : rare, only 2.6 kernels
4739 Description: Unusual file access patterns on the MDS may result in inode
4740 data being lost in very rare circumstances.
4741 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4742 code in the 2.6 kernel caused the inode and/or EA data not to
4743 be read from disk, causing single-file corruption.
4747 Frequency : rare, only 2.6 filesystems using extents
4748 Description: Heavy concurrent write and delete load may cause data corruption.
4749 Details : It was possible under high-load situations to have an extent
4750 metadata block in the block device cache from a just-unlinked
4751 file overwrite a newly-allocated data block. We now unmap any
4752 metadata buffers that alias just-allocated data blocks.
4756 Frequency : filesystems with default stripe_count larger than 77
4757 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4758 Details : lconf specifies an inode size of 4096 bytes when the default
4759 stripe_count is larger than 77. This conflicts with the default
4760 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4761 this case to avoid pinning too much memory for large EAs.
4763 ------------------------------------------------------------------------------
4765 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4770 Frequency : rare (extremely heavy IO load with hundreds of clients)
4772 Description: Client is evicted, gets IO error writing to file
4773 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4774 caused clients to be evicted for AST timeouts. The fixes in
4775 bug 5192 mean we no longer need to have such short AST timeouts
4776 so ldlm_timeout has been increased.
4779 Frequency : occasional during --force or --failover shutdown under load
4780 Bugzilla : 5949, 4834
4781 Description: Server oops/LBUG if stopped with --force or --failover under load
4782 Details : a collection of import/export refcount and cleanup ordering
4783 issues fixed for safer force cleanup
4786 Frequency : only filesystems larger than 120 OSTs
4787 Bugzilla : 5990, 6223
4788 Description: lfs getstripe would oops on a very large filesystem
4789 Details : lov_getconfig used kfree on vmalloc'd memory
4792 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4793 Bugzilla : 6242, 6243
4794 Description: reading from files that had been truncated to a non-zero size
4795 but never opened returned no data
4796 Details : ll_file_read() reads zeros from no-object files to EOF
4801 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4802 Details : There's little harm in aborting MDS/OSS recovery and letting it
4803 try again, so I removed the LASSERT and return an error instead.
4805 Severity : enhancement
4807 Description: New debugging infrastructure for tracking down data corruption
4808 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4809 (b) cover more of the client-side code path, and (c) try to narrow
4810 down where problems occurred
4814 Bugzilla : 3819, 4364, 4397, 6313
4815 Description: Racing close and eviction MDS could cause assertion in mds_close
4816 Details : It was possible to get multiple mfd references during close and
4817 client eviction, leading to one thread referencing a freed mfd.
4819 Severity: : enhancement
4820 Bugzilla : 3262, 6359
4821 Description: Attempts to reconnect to servers are now more aggressive.
4822 Details : This builds on the enhanced upcall-less recovery that was added
4823 in 1.4.2. When trying to reconnect to servers, clients will
4824 now try each server in the failover group every 10 seconds. By
4825 default, clients would previously try one server every 25 seconds.
4830 Description: After recovery, certain operations trigger a failed
4831 assertion on a client.
4832 Details : Failing over an mds, using lconf -d --failover, while a
4833 client was doing a readdir() call would cause the client to
4834 LBUG after recovery completed and the readdir() was resent.
4836 Severity : enhancement
4838 Description: Default groups are now added by lconf
4839 Details : You can now run lconf --group <servicename> without having to
4840 manually add groups with lmc.
4843 Frequency : occasional
4845 Description: Nodes with an elan id of 0 trigger a failed assertion
4848 Frequency : always when accessing e.g. tty/console device nodes
4850 Description: tty and some other devices nodes cannot be used on lustre
4851 Details : file's private_data field is used by device data and lustre
4852 values in there got lost. New field was added to struct file to
4853 store fs-specific private data.
4856 Frequency : when exporting Lustre via NFS
4858 Description: NFSD failed occasionally when looking up a path component
4859 Details : NFSD is looking up ".." which was broken in ext3 directories
4860 that had grown large enough to become hashed.
4863 Frequency : Clusters with multiple interfaces not on the same subnet
4865 Description: Nodes will repeatedly try to reconnect to an interface which it
4866 cannot reach and report an error to the log.
4867 Details : Extra peer list entries will be created by lconf with some peers
4868 unreachable. lconf now validates the peer before adding it.
4871 Frequency : Only if a default stripe is set on the filesystem root.
4873 Description: Setting a default stripe on the filesystem root prevented the
4874 filesystem from being remounted.
4875 Details : The client was sending extra request flags in the root getattr
4876 request and did not allocate a reply buffer for the dir EA.
4879 Frequency : occasional, higher if lots of files are accessed by one client
4880 Bugzilla : 6159, 6097
4881 Description: Client trips assertion regarding lsm mismatch/magic
4882 Details : While revalidating inodes the VFS looks up inodes with ifind()
4883 and in rare cases can find an inode that is being freed.
4884 The ll_test_inode() code will free the lsm during ifind()
4885 when it finds an existing inode and then the VFS later attaches
4886 this free lsm to a new inode.
4890 Bugzilla : 6422, 7030
4891 Description: MDS deadlock between mkdir and client eviction
4892 Details : Creating a new file via mkdir or mknod (starting a transaction
4893 and getting the ns lock) can deadlock with client eviction
4894 (gets ns lock and trying to finish a synchronous transaction).
4897 Frequency : occasional
4898 Description: While starting a server, the fsfilt_ext3 module could not be
4900 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4901 kernels. Previously, lconf would still use the ext3 name
4902 when trying to load modules. Now, it will correctly use
4903 ext3 on 2.4 and ldiskfs on 2.6.
4905 Severity : enhancement
4906 Description: The default stripe count has been changed to 1
4907 Details : The interpretation of the default stripe count (0, to lfs
4908 or lmc) has been changed to mean striping across a single
4909 OST, rather than all available. For general usage we have
4910 found a stripe count of 1 or 2 works best.
4912 Severity : enhancement
4913 Description: Add support for compiling against Cray portals.
4914 Details : Conditional compiling for some areas that are different
4918 Frequency : occasional
4919 Bugzilla : 6409, 6834
4920 Description: Creating files with an explicit stripe count may lead to
4921 a failed assertion on the MDS
4922 Details : If some OSTs are full or unavailable, creating files may
4923 trigger a failed assertion on the MDS. Now, Lustre will
4924 try to use other servers or return an error to the
4928 Frequency : occasional
4930 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4931 caused lock timeout during readahead (since 1.4.2).
4932 Details : Processes doing readahead might match a lock that hasn't been
4933 granted yet if there are overlapping and conflicting lock
4934 requests. The readahead process waits on ungranted lock
4935 (original lock is CBPENDING), while OST waits for that process
4936 to cancel CBPENDING read lock and eventually evicts client.
4938 Severity : enhancement
4940 Description: Initial enabling of flock support for clients
4941 Details : Implements fcntl advisory locking and file status functions.
4942 This feature is provided as an optional mount flag (default
4943 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4944 locking are implemented yet, and those that are are not guaranteed
4945 to be completely correct in production environments.
4946 mount -t lustre -o [flock|noflock] ...
4949 Frequency : occasional
4951 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4952 assertion in the ext3 JBD (journaling) layer.
4953 Details : The b_committed_data struct is protected by the big kernel lock
4954 in 2.4 kernels, serializing journal_commit_transaction() and
4955 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4956 access to this struct. In 2.6 kernels there is finer grained
4957 locking to improve SMP performance of the JBD layer.
4961 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4966 Description: merge and cleanup kernel patches.
4969 -----------------------------------------------------------------------------
4971 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4973 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4974 versions of Lustre. Please update all servers and clients to
4975 version 1.4.2 or later at the same time. You must also run
4976 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4977 to update the configuration logs.
4979 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4980 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4981 - swab llog records of type '0' so we get proper header size/idx (5861)
4982 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4983 - fix rename of one directory over another leaking an inode (5953)
4984 - avoid SetPageDirty on 2.6 (5981)
4985 - don't re-add just-being-destroyed locks to the waiting list (5653)
4986 - when creating new directories, inherit the parent's custom
4987 striping settings if present parent (3048)
4988 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4989 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4990 - don't hold client locks on temporary worklist from l_lru (5666)
4991 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4992 - automatically evict dead clients (5921)
4993 - Update file size properly in create+truncate+fstat case (6196)
4994 - Do not unhash mountpoint dentries, do not allow removal of
4996 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4997 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4998 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4999 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5000 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5001 - Correctly handle reads of files with no objects (6243)
5002 - lctl recover will also mark a device active if deactivate used (5933)
5004 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5005 - allow --write-conf on an MDS with different nettype than client (5619)
5006 - don't write config llogs to MDS for mounts not from that MDS (5617)
5007 - lconf should create multiple TCP connections from a client (5201)
5008 - init scripts are now turned off by default; run chkconfig --on
5009 lustre and chkconfig --on lustrefs to use them
5010 - upcalls are no longer needed for clients to recover to failover
5012 - add --abort-recovery option to lconf to abort recovery on device
5014 - add support for an arbitrary number of OSTs (3026)
5015 - Quota support protocol changes.
5016 - forward compatibility changes to wire structs (6007)
5017 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5018 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5019 - improved "lustre status" script
5020 - initialize blocksize for non-regular files (6062)
5021 - added --disable-server and --disable-client configure options (5782)
5022 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5023 - Vanilla 2.4.29 support
5024 - increase maximum number of obd devices to 520 (6242)
5025 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5026 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5027 SLES 9 2.6.5 kernels
5028 - limit stripes per file to 160 (the maximum EA size) (6093)
5030 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5033 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5034 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5035 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5036 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5037 - small ext3 extents cleanups and fixes (5733)
5038 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5039 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5040 - avoid needless client->OST connect, fix handle mismatch (5317)
5041 - fix DLM error path that led to out-of-sync client, long delays (5779)
5042 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5043 - fix several locking issues related to i_size (5492,5624,5654,5672)
5044 - don't move pending lock onto export if it is already evicted (5683)
5045 - fix kernel oops when creating .foo in unlinked directory (5548)
5046 - fix deadlock in obdfilter statistics vs. object create (5811)
5047 - use time_{before,after} to avoid timer jiffies wrap (5882)
5048 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5049 - Do not leak request if server was not able to process it (5154)
5050 - If mds_open unable to find parent dir, make that negative lookup(5154)
5051 - don't create new directories with extent-mapping (5909, 5936)
5053 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5054 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5055 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5056 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5057 - support for RHEL4 kernel on i686 (5773)
5058 - provide error messages when incompatible logs are encountered (5898)
5060 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5061 * version 1.4.0.10 (1.4.1 release candidate 1)
5063 - don't keep a lock reference when lock is not granted (4238)
5064 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5065 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5066 - drop import inflight refcount on signal_completed_replay error (5255)
5067 - unlock page after async write error during send (3677)
5068 - handle missing objects in filter_preprw_read properly (5265)
5069 - no transno return for symlink open, don't save no-trasno open (3440)
5070 - don't try to complete elan receive that already failed (4012)
5071 - free RPC server reply state on error (5406)
5072 - clean up thread from ptlrpc_start_thread() on error (5160)
5073 - readahead could read extra page into cache that wasn't ejected (5388)
5074 - prevent races in class_attach/setup/cleanup/detach (5260)
5075 - don't dereference de->d_inode after l_dput of de (5458)
5076 - use "int" for stripe value returned from lock_to_stripe (5544)
5077 - mballoc allocation and error-checking fixes in 2.6 (5504)
5078 - block device patches to fix I/O request sizes in 2.6 (5482)
5079 - look up hostnames for IB nals (5602)
5080 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5081 - don't start multiple acceptors for the same port (5277)
5082 - fix incorrect LASSERT in mds_getattr_name (5635)
5083 - export a proc file for general "ping" checking (5628)
5084 - fix "lfs check" to not block when the MDS is down (5628)
5086 - service request history (4965)
5087 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5088 - create an "evict_client" /proc entry on OSTs, like the MDS has
5089 - fix mount usage message, return errors per mount(8) (5168)
5090 - change grep [] to grep "[]" in tests so they work in more UMLs
5091 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5092 - remove ancient LOV_MAGIC_V0 EA support (5047)
5093 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5094 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5095 - allow readahead allocations to fail when low on memory (5383)
5096 - mmap locking landed again, after considerable improvement (2828)
5097 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5099 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5102 - send OST transaction number in read/write reply to free req (4966)
5103 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5104 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5105 - the watchdog thread now runs as interruptible (5246)
5106 - flock/lockf fixes (but it's still disabled, pending 5135)
5107 - don't use EXT3 constants in llite code (5094)
5108 - memory shortage at startup could cause assertion (5176)
5110 - reorganization of lov code
5111 - single portals codebase
5113 - add extents/mballoc support (5025)
5114 - direct I/O reads in the obdfilter (4048)
5115 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5117 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5120 - send OST transaction number in read/write reply to free req (4966)
5121 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5122 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5123 - the watchdog thread now runs as interruptible (5246)
5124 - handle missing objects in filter_preprw_read properly (5265)
5125 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5126 - drop import inflight refcount on signal_completed_replay error (5255)
5127 - unlock page after async write error during send (3677)
5128 - return original error code on reconstructed replies (3761)
5129 - no transno return for symlink open, don't save no-trasno open (3440)
5131 - add pid to ldlm debugging output (4922)
5132 - bump the watchdog timeouts -- we can't handle 30sec yet
5133 - extra debugging for orphan dentry/inode bug (5259)
5135 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5138 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5139 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5140 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5141 - fix NULL dereference in /proc/sys/portals/routes (4827)
5142 - allow failed mdc_close() operations to be interrupted (4561)
5143 - stop precreate on OST before MDS would time out on it (4778)
5144 - don't send partial-page writes before EOF from client (4410)
5145 - discard client grant for sub-page writes on large-page clients (4520)
5146 - don't free dentries not owned by NFS code, check generation (4806)
5147 - fix lsm leak if mds_create_objects() fails (4801)
5148 - limit debug_daemon file size, always print CERROR messages (4789)
5149 - use transno after validating reply (3892)
5150 - process timed out requests if import state changes (3754)
5151 - update mtime on OST during writes, return in glimpse (4829)
5152 - add mkfsoptions to LDAP (4679)
5153 - use ->max_readahead method instead of zapping global ra (5039)
5154 - don't interrupt __l_wait_event() during strace
5156 - add software watchdogs to catch hung threads quickly (4941)
5157 - make lustrefs init script start after nfs is mounted
5158 - fix CWARN/ERROR duplication (4930)
5159 - return async write errors to application if possible (2248)
5160 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5161 - print NAL number in %x format (4645)
5162 - update barely-supported suse-2.4.21-171 series (4842)
5163 - support for sles 9 %post scripts
5164 - support for building 2.6 kernel-source packages
5165 - support for sles km_* packages
5167 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5170 - ignore -ENOENT errors in osc_destroy (3639)
5171 - notify osc create thread that OSC is being cleaned up (4600)
5172 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5173 - reconstruct ost_handle() like mds_handle() (4657)
5174 - create a new thread to do import eviction to avoid deadlock (3969)
5175 - let lconf resolve symlinked-to devices (4629)
5176 - don't unlink "objects" from directory with default EA (4554)
5177 - hold socknal file ref over connect in case target is down (4394)
5178 - allow more than 32000 subdirectories in a single directory (3244)
5179 - fix blocks count for O_DIRECT writes (3751)
5180 - OST returns ENOSPC from object create when no space left (4539)
5181 - don't send truncate RPC if file size isn't changing (4410)
5182 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5183 - bind to privileged port in socknal and tcpnal (3689)
5185 - rate limit CERROR/CWARN console message to avoid overload (4519)
5186 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5187 - basic mmap support (3918)
5188 - kernel patch series update from b1_4 (4711)
5190 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5193 - avoid crash during MDS cleanup with OST shut down (2775)
5194 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5195 - don't use bad inodes on the MDS (3744)
5196 - dynamic object preallocation to improve recovery speed (4236)
5197 - don't hold spinlock over lock dumping or change debug flags (4401)
5198 - don't zero obd_dev when it is force cleaned (3651)
5199 - print grants to console if they go negative (4431)
5200 - "lctl deactivate" will stop automatic recovery attempts (3406)
5201 - look for existing locks in ldlm_handle_enqueue() (3764)
5202 - don't resolve lock handle twice in recovery avoiding race (4401)
5203 - revalidate should check working dir is a directory (4134)
5205 - don't always mark "slow" obdfilter messages as errors (4418)
5207 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5210 - don't close LustreDB during write_conf until it is done (3860)
5211 - fix typo in lconf for_each_profile (3821)
5212 - allow dumping logs from multiple threads at one time (3820)
5213 - don't allow multiple threads in OSC recovery (3812)
5214 - fix debug_size parameters (3864)
5215 - fix mds_postrecov to initialize import for llog ctxt (3121)
5216 - replace config semaphore with spinlock (3306)
5217 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5218 - don't allow enqueue to complete on a destroyed export (3822)
5219 - down write_lock before checking llog header bitmap (3825)
5220 - recover from lock replay timeout (3764)
5221 - up llog sem before sending rpc (3652)
5222 - reduce ns lock hold times when setting kms (3267)
5223 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5224 - fix NULL deref and obd_dev leak on setup error (3312)
5225 - replace some LBUG about llog ops with error handling (3841)
5226 - don't match INVALID dentries from d_lookup and spin (3784)
5227 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5228 - fix invalid assertion in ptlrpc_set_wait (3880)
5230 - add libwrap support for the TCP acceptor (3996)
5231 - add /proc/sys/portals/routes for non-root route listing (3994)
5232 - allow setting MDS UUID in .xml (2580)
5233 - print the stack of a process that LBUGs (4228)
5235 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5238 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5239 - make sure to unset replay flag from failed open requests (3440)
5240 - if default stripe count is 0, use OST count for inode size (3636)
5241 - update parent mtime/ctime on client for create/unlink (2611)
5242 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5243 - free recovery state on server during a forced cleanup (3571)
5244 - unregister_reply for resent reqs (3063)
5245 - loop back devices mounting and status check on 2.6 (3563)
5246 - fix resource-creation race that can provoke i_size == 0 (3513)
5247 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5248 - more debugging for page-accounting assertion (3746)
5249 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5250 - avoid deadlock after precreation failure (3758)
5251 - fix race and lock order deadlock in orphan handling (3450, 3750)
5252 - add validity checks when grabbing inodes from l_ast_data (3599)
5254 - add /proc/.../recovery_status to obdfilter (3428)
5255 - lightweight CDEBUG infrastructure, debug daemon (3668)
5256 - change default OSC RPC parameters to be better on small clusters
5257 - turn off OST read cache for files smaller than 32MB
5258 - install man pages and include them in rpms (3100)
5259 - add new init script for (un)mounting lustre filesystems (2593)
5260 - run chkconfig in %post for init scripts (3701)
5261 - drop scimac NAL (unmaintained)
5263 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5266 - clean kiobufs before and after use (3485)
5267 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5268 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5269 - change init script to fail more clearly if not run as root (1528)
5270 - allow clients to reconnect during replay (1742)
5271 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5272 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5273 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5274 - fix cond_resched (3554)
5275 - only evict unfinished clients after recovery (3515)
5276 - allow bulk resend, prevent data loss (3570)
5277 - dynamic ptlrpc request buffer allocation (2102)
5278 - don't allow unlinking open directory if it isn't empty (2904)
5279 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5280 - remove extraneous obd dereference causing LASSERT failure (3334)
5281 - don't use get_cycles() when creating temp. files on the mds (3156)
5282 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5283 - handle EEXIST for set-stripe, set proper directory name (3336)
5285 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5286 - fix ksocknal_fmb_callback() error messages (2918)
5288 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5291 - don't copy lvb into (possibly NULL) reply on error (2983)
5292 - don't deref dentry after dput, don't free lvb on error (2922)
5293 - use the kms to determine writeback rpc length (2947)
5294 - increment oti_logcookies when osc is inactive (2948)
5295 - update client's i_blocks count via lvb messages (2543)
5296 - handle intent open/close of special files properly (1557)
5297 - mount MDS with errors=remount-ro, like obdfilter (2009)
5298 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5299 - don't use cancelling-locks' kms values (2947)
5300 - use highest lock extent for kms, not last one (2925)
5301 - don't dereference ERR_PTR() dentry in error handling path (3107)
5302 - fix thread race in portals_debug_dumplog() (3122)
5303 - create lprocfs device entries at setup instead of at attach (1519)
5304 - common AST error handler, don't evict client on completion race (3145)
5305 - zero nameidata in detach_mnt in 2.6 (3118)
5306 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5307 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5308 - initialize RPC timeout timer earlier for 2.6 (3219)
5309 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5310 - print nal/nid for unknown nid (3258)
5311 - additional checks for oscc recovery before doing precreate (3284)
5312 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5313 - don't crash in mdc_close for bad permissions on open (3285)
5314 - zero i_rdev for non-device files (3147)
5315 - clear page->private before handing to FS, better assertion (3119)
5316 - tune the read pipeline (3236)
5317 - fix incorrect decref of invalidated dentry (2350)
5318 - provide read-ahead stats and refine rpc in flight stats (3328)
5319 - don't hold journal transaction open across create RPC (3313)
5320 - update atime on MDS at close time (3265)
5321 - close LDAP connection when recovering to avoid server load (3315)
5322 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5323 - don't leak open file on MDS after open resend (3325)
5324 - serialize filter_precreate and filter_destroy_precreated (3329)
5325 - loop device shouldn't call sync_dev() for nul device (3092)
5326 - clear page cache after eviction (2766)
5327 - resynchronize MDS->OST in background (2824)
5328 - refuse to mount the same filesystem twice on same mountpoint (3394)
5329 - allow llmount to create routes for mounting behind routers (3320)
5330 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5331 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5332 - fix rare problem with rename on htree directories (3417)
5334 - allow default OST striping configuration per directory (1414)
5335 - fix compilation for qswnal for 2.6 kernels (3125)
5336 - increase maximum number of MDS request buffers for large systems
5337 - change liblustreapi to be useful for external progs like lfsck (3098)
5338 - increase local configuration timeout for slow disks (3353)
5339 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5341 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5344 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5345 - don't overwrite extent policy data in reply if lock was blocked (2901)
5346 - drop filter export grants atomically with removal from device (2663)
5347 - del obd_self_export from work_list in class_disconnect_exports (2908)
5348 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5349 - swab reply message in mdc_close, other PPC fixes (2464)
5350 - fix destroying of named logs (2325)
5351 - overwrite old logs when running lconf --write_conf (2264)
5352 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5353 - fix race in target_handle_connect (2898)
5354 - mds_reint_create() should take same inode create lock (2926)
5355 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5356 - don't close files for self_export to avoid uninitialized obd (2936)
5357 - allow MDS with the same name as client node (2939)
5358 - hold dentry reference for closed log files for unlink (2325)
5359 - reserve space for all logs during transactions (2059)
5360 - don't evict page beyond end of stripe extent (2925)
5361 - don't oops on a deleted current working directory (2399)
5362 - handle hard links to targets without a parent properly (2517)
5363 - don't dereference NULL lock when racing during eviction (2867)
5364 - don't grow lock extents when lots of conflicting locks (2919)
5366 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5369 - account for cache space usage on clients to avoid data loss (974)
5370 - lfsck support in lustre kernel code (2349)
5371 - reduce journal credits needed for BRW writes (2370)
5372 - orphan handling to avoid losing space on client/server crashes
5373 - ptlrpcd can be blocked, stopping ALL progress (2477)
5374 - use lock value blocks to assist in proper KMS, faster stat (1021)
5375 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5376 - recovery for initial connections (2355)
5377 - fixes for mds_cleanup_orphans (1934)
5378 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5379 - block all file creations until orphan recovery completes (1901)
5380 - client remove rq_connection from request struct (2423)
5381 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5382 - recovery timer race (2670)
5383 - mdc_close recovey bug (2532)
5384 - ptlrpc cleanup bug (2710)
5385 - mds timeout on local locks (2588)
5386 - namespace lock held during RPCs (2431)
5387 - handle interrupted sync write properly (2503)
5388 - don't try to handle a message that hasn't been replied to (2699)
5389 - client assert failure during cleanup after abort recovery (2701)
5390 - leak mdc device after failed mount (2712)
5391 - ptlrpc_check_set allows timedout requests to complete (2714)
5392 - wait for inflight reqs when ptlrpcd finishes (2710)
5393 - make sure unregistered services are removed from the srv_list
5394 - reset bulk XID's when resending them (caught by 1138 test)
5395 - unregister_bulk after timeout
5396 - fix lconf error (2694)
5397 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5398 - readahead locks pages, leaves pending causing memory pressure (2673)
5399 - increase OST request buffers to 4096 on large machines (2729)
5400 - fix up permission of existing directories in simple_mkdir (2661)
5401 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5402 - don't assign transno to errored transactions (2742)
5403 - don't delete objects on OST if given a bogus objid from MDS (2751)
5404 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5405 - if rq_no_resend, then timeout request after recovery (2432)
5406 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5407 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5408 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5409 - MDS recovery shouldn't delete objects that it has given out (2730)
5410 - if enqueue arrives after completion, don't clobber LVB (2819)
5411 - don't unlock pages twice when trigger_group_io returns error (2814)
5412 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5413 - don't write pages to disk if there was an error (1450)
5414 - don't ping imports that have recovery disabled (2676)
5415 - take buffered bytes into account when balancing socknal conn (2817)
5416 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5417 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5418 - remove little-endian swabbing of llog records (1987)
5419 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5420 - retry reposting request buffers if they fail (1191)
5421 - grow extent at grant time to avoid granting a revoked lock (2809)
5422 - lock revoke doesn't evict page if covered by a second lock (2765)
5423 - disable VM readahead to avoid reading outside lock extents (2805)
5425 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5426 - updated kernel patches for hp-2.4.20 kernel (2681)
5428 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5431 - fix truncated write corruption (2366)
5432 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5434 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5435 - ensure len(uuid) < 37 in lmc (1171)
5436 - fix ia64 OOPS in llog_test (2255)
5437 - zero end of page at obdfilter for partial page writes (2648)
5438 - don't leave stale dentries around after renames (bug 2428)
5439 - fix timeouts when evicting a client with a single lock held (2642)
5440 - set deadline for the initial HELLO message to drain (2634)
5441 - print out dotted-quad IP addresses in the socknal (2302)
5443 - additional debugging for MDS client eviction problem (2443)
5444 - fix mkfsoptions support for osts (2603, 2604)
5446 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5449 - add series for the vanilla 2.6.0 kernel
5450 - add series for the vanilla 2.4.24 kernel
5451 - add series for a cray x86/64 UL kernel drop
5452 - fix xattr patches for the vanilla 2.4.19 series
5454 - generate true UUIDs in lmc (1171)
5455 - have portals stack dumping break in UML (2466)
5456 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5457 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5458 - wake callback waiting threads on client eviction (2460)
5459 - Add --ptldebug and --subsystem to lmc (1719)
5460 - update assertion to allow safe interrupt allocation
5461 - set rq_no_resend for cancel requests (2432)
5462 - recalculate ptlrpcd timeout after resend (2494)
5463 - call vfs_rmdir when removing pending directories (2368)
5464 - fix renaming a file to itself (2429)
5465 - lmc creates a default one-stripe lov (2454)
5466 - expand procfs space to handle large clusters (2326)
5467 - increase UML stack to avoid overflow
5468 - update lconf's list of debug and subsystem masks
5469 - fix lfs find --obd (2510)
5470 - /proc tunable for disabling filter read caching (2591)
5471 - stop rpm packages from altering slapd.conf (2301)
5472 - disable nagle in the socknal under 0conf (2578)
5473 - choose mds inode size based on stripe count (2572)
5474 - fix kernel-source rpm problems (2516)
5476 - add --disable-doc to avoid pdf generation (2421)
5477 - update documentation, tests, type-os, comments
5478 - avoid format warnings on ia64
5479 - remove the TOE NAL
5480 - tiny code cleanups by removing unused fields
5482 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5485 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5486 - remove the most blatant lies from BUILDING file (2371)
5487 - change default debug level to reasonable production setting
5488 - reduce client side cache size to reduce cache flush time
5489 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5490 - make TCP zerocopy and pinger support enabled by default (2476)
5491 - sync writes completed after process exits caused crashes (2319)
5492 - maintain correct mount count on the MDS (2356)
5493 - backout 1557, because 2316 wasn't really fixed
5494 - better file I/O statistics gathering in /proc
5495 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5496 - another kernel patch to fix zero-copy TCP function export
5497 - don't take duplicate lock when processing re-sent getattr (2420)
5498 - lctl uses obd_self_export instead of creating new conn (2353)
5499 - MDS/OST recovery case which requires object creation asserted (2425)
5500 - move lfs from /usr/sbin to /usr/bin in packages
5501 - fix race between mds_client_add and mds_client_free (2417)
5502 - use kmalloc instead of slabs in portals (2430)
5503 - don't create duplicate records when a failover MDS is present (2442)
5504 - remove unnecessary mount age check (2332)
5505 - don't remove directory inodes from locks prematurely (2451)
5506 - don't break if MDS service name is the same as hostname (2103)
5507 - fix races in client write RPC generation when cache full (2482)
5509 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5512 - remove now-unused request->rq_obd (278)
5513 - if an allocation fails, print out how much memory we've used (1933)
5514 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5515 - add the "configurable stack size" patch to most series files (1256)
5516 - ability to write large log records, for 100+ OST configs (2306)
5517 - fix NULL deref when filter_prep fails (2314)
5518 - fix operator precedence error in filter_sync
5519 - dynamic allocation of socknal TX descriptors (2315)
5520 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5521 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5522 - ensure that CWARN messages in Portals always get to the syslog
5523 - __init/__exit are not for prototype decls (ldlm_init/exit)
5524 - x86-64 compile warning fixes
5525 - fix gateway LMC keyword conflict (2318)
5526 - fix MDS lock inversions in getattr/reint paths (1844)
5527 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5528 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5529 - DLM race condition prevented some lock evictions (2328)
5530 - ENOMEM detection and retry on socknal sends (2230)
5531 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5532 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5533 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5534 - better error messages when a client is rejected during recovery (1505)
5535 - avoid cancelling locks which were never granted, after failure (2330)
5536 - fix i_sem/journal inversion in mds_client_add (2333)
5537 - fix truncate/getattr lock cycle deadlock (2334)
5538 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5539 - fix two rare exit paths which could leak an l_lock() ref (2321)
5540 - fencepost error in MDS/OST orphan recovery (2226)
5541 - make log record alignment 8 bytes (1988)
5542 - lstripe now fails when requested offset > ost_count (2237)
5543 - ensure that all kernel series have a complete list.h (1607)
5544 - fix crashes in special-file operations (2316)
5545 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5547 - allow configurable automake binary, for testing new versions
5548 - small update to the lfs documentation
5550 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5552 * fix negative export reference count in fsfilt_sync (2312)
5554 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5555 * release candidate 0.9.1
5557 - orphans are moved into the PENDING directory for possible recovery
5558 - replayed opens now open by fid for orphan/rename safety (1042)
5559 - last close of an orphan inode generates a transno (683)
5560 - chdir() and mount() now pin the directory entry (1020)
5561 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5562 - discard very old requests without processing them (1502)
5563 - remove some common, well-understood CERRORs (1505)
5564 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5565 - clear "grant" flags in OST replies until OST grant code lands (1644)
5566 - fix read performance by not clobbering i_blksize on client (1598)
5567 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5568 - make LRU size a /proc tunable, clears locks when reduced (707)
5569 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5570 - server locks take references on exports now (1558)
5571 - build fixes for 2.4.20-rh trees (1663)
5572 - return an error from lov_create if all OSCs are inactive (1751)
5573 - fix import levels when a reconnect happens without a timeout (1597)
5574 - exit early from mds_open if we get a lookup error (1749)
5575 - partial page read at EOF wouldn't wait for disk before sending (1642)
5576 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5577 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5578 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5579 - write/truncate lock inversion (1639)
5580 - Don't auto-load obdclass, portals modules during cleanup (1495)
5581 - fix timestamps from jumping to "now" (1763)
5582 - extra journal assertions (1648)
5583 - add an extra multiunlink test (1771)
5584 - fix read_record/write_record API (1776)
5585 - fix leak of offset_extent, possible incorrect i_size later (1772)
5586 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5587 - Debugging for the kqswnal_get_idle_tx problems (1820)
5588 - Allow recovery to be attempted multiple times (1536)
5589 - Write out MDS last_rcvd file after it is first created (1600)
5590 - Fix tx_descriptor leak in failed transmit situations (1827)
5591 - ext3 journaling fixes for assertion failure after IO error (1871)
5592 - class_export_put() on freed export after completion AST error (1896)
5593 - Fix revalidate looping in VFS (1322)
5594 - Don't access a freed export during MDS_REINT timeout (1521)
5595 - Add open-unlink recovery support on the MDS (1673,1764)
5596 - Return an error if no MDS data was read from last_rcvd (1946)
5597 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5598 - Don't setup a disk device that doesn't match exported UUID (317)
5599 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5600 - avoid committing NULL handle in force close
5601 - local.sh is now a one-stripe LOV configuration
5602 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5603 - fix ext3 htree duplicate directory entry corruption (1516)
5604 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5605 - update to new LOV EA format (2097)
5606 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5608 2003-06-15 Phil Schwan <phil@clusterfs.com>
5611 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5612 - per-import/export recovery handling (958, 931, 959)
5613 - multiple last-rcvd slots, for serving multiple FSes (949)
5614 - connections are again shared between multiple imp/exports (963, 964)
5615 - "umount -f" would hang if any requests needed to be sent (393, 978)
5616 - avoid pinning large req buffer by copying for queued messages (989)
5617 - add "uuid" to "lctl device" command to help upcalls (991)
5618 - "open" RPCs with transnos would confuse recovery counters (1037)
5619 - do proper endian conversion of all wire messages (288, 340, 891)
5620 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5621 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5622 - fix ptlrpc_connection leak in target_handle_connect (1174)
5623 - fix import refcounting bug in OST and MDS cleanup (1134)
5624 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5625 - fix very unlikely obd_types race condition (501)
5626 - remove osc_open hack for echo_client (1187)
5627 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5628 - a failure in read_inode2 leads to deadlock (1139)
5629 - cancel ack-locks as soon as transaction is committed (1072)
5630 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5631 - make sure to commitrw after any preprw to avoid deadlock (1162)
5632 - failing to execute a file in a lustre FS would lock inode (1203)
5633 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5634 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5635 - obd_uuid2tgt would walk off the end of the list (1255)
5636 - on IA64 the osc would give portals incorrect bulk size (1258)
5637 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5638 - fix lock inversion caused by new llite matching code (1282)
5639 - limit the number of dirty pages on a client to 10MB (1286)
5640 - timed out locks were not being corrected cancelled (1289)
5641 - fix O_DIRECT above 4GB on IA-32 (1292)
5642 * major user-visible changes
5643 - fail out/fail over policy now controlled by the upcall (993)
5645 - add OBD_PING to check server availability and failure (954)
5646 - lustre messages are now sent in sending host order (288, 340, 891)
5647 - add eadatalen to MDS getattr reply (340)
5648 - OST read replies may contain second buffer, with per-page status (593)
5650 2003-03-11 Phil Schwan <phil@clusterfs.com>
5653 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5654 - failed open()s could cause deadlock; fixed (867, 869)
5655 - stop cancelling OST locks when files are closed (481)
5656 - overlapping XID spaces caused network corruption (851, 853)
5657 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5658 - setattr_raw added, to do single-RPC, server-side setattrs
5659 - lmc/lconf syntax change for OST UUIDs
5660 - fix crashy race condition between ptlrpc_free_req and osc_close
5661 - don't use request in mdc_enqueue if we hit a timeout (889)
5662 - don't set the inode i_size for regular files from the MDS (896)
5663 - handle out of order completion AST (842)
5664 - don't LBUG if a lock request times out after receiving AST (913)
5665 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5666 - if a bad lock AST arrives, send an error instead of dropping entirely
5667 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5668 - fix leak in bulk IO when only partially completed (899, 900, 926)
5669 - fix O_DIRECT for ia64 (55)
5670 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5671 - C-z after timeout could hang a process forever; fixed (977)
5673 - client-side I/O cache (678, 924, 929, 941, 970)
5675 - READPAGE and SETATTRs which don't take server-side locks get
5678 2003-02-11 Phil Schwan <phil@clusterfs.com>
5681 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5682 - Fix fsfilt_extN_readpage() to read a full page of directory
5683 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5684 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5685 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5686 - Most intent operations were removed, in favour of a new RPC mode
5687 that does a single RPC to the server and bypasses most of the VFS
5688 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5689 - Aggressively cancel local locks on DLM servers
5690 - mds_reint_unlink sends EA to the client if it's the last nlink.
5691 client uses that EA to unlink OST objects.
5692 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5693 - recursive symlinks were fixed (439)
5694 - fixed NULL deref in DEBUG_REQ
5695 - filter_update_lastobjid no longer calls sync, which annoyed extN
5696 - fixed multi-client small-writes to a single file problem (445)
5697 - fixed mtime updates during file writes (607)
5698 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5699 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5700 - fixed Linux OST size reporting problem (444, 656)
5701 - OST now updates object mtime with writes or setattr (607, 619)
5702 - client verifies file size before zeroing page past EOF (445)
5703 - OST now writes last allocated objid to disk with allocation (108)
5704 - LOV on echo now works (409)
5706 - mds_reint_unlink sends a new buffer, with the EA included. this
5707 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5708 set if a regular file was being unlinked, and it was the last link
5709 - use PtlGet from the target for bulk writes (315)
5710 - OST now updates object mtime with writes or setattr (607, 619)
5711 - LDLM now has a grant-time callback to revalidate locked items, if
5713 - Many MDS operations were reorganized to combat race conditions
5715 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5716 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5718 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5721 - Fully reactivate OST imports after reconnection (512, others)
5722 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5723 - More graceful error handling for truncating on dead OST (515)
5724 - Don't error out unless we're actually accessing dead stripes (474)
5725 - Fix garbage sizes when stripes are missing (410)
5726 - LRU counters were broken, causing constant lock purge (433, 432)
5727 - garbage on read from stripes with failed OSTs (441)
5728 - mark OSCs as active before reconnecting during recovery (438)
5729 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5730 - lfind did not preserve OST order in output (443)
5731 - symlinks cause hung clients, incorrect data (439)
5732 - stop dereferencing request after dropping refcount (457)
5733 - don't LASSERT(spin_is_locked) on non-SMP (455)
5734 - fixes for many rename() bugs
5735 - fstat didn't correctly synchronize attributes (399)
5736 - server must handle lock cancellation during blocking AST prep (487)
5737 - bulk descriptors were free()d too soon (511)
5738 - fix paths in lconf, which would load incorrect modules (451, 507)
5739 - fix confusing lconf 'host not found' error message (386)
5740 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5741 - fix race condition in mdc_blocking_ast() for inode access (526)
5742 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5743 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5744 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5745 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5746 - fix lstripe in several ways (kernel side) (527)
5747 - fix request leak in ldlm_cli_enqueue (262)
5748 - incorrect OSC was marked inactive after OST failure
5749 - call mds_fs_cleanup before unmounting filesystem (524)
5750 - fix races between taking ns_lock and ldlm_lock_change_resource
5751 - fix races updating LOV export open file list
5752 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5753 - fix recovery NULL deref in ldlm_cli_cancel_unused
5754 - fix some DLM races by using new hash table for lock handles (419)
5755 - permit the client to specify desired inodes, at replay
5756 - duplicate requests when we queue them for replay reintegration
5757 - fix last_rcvd offset calculation
5758 - sync after each recovered transaction, so we always make progress
5759 - never, not always, ERESTART requests without transnos
5760 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5762 - skip replay if the MDS says that the client is already connected
5763 - don't check for a recovery-enabled export to match lctl's UUID
5764 - don't INC_USE_COUNT for phantom exports
5765 - don't crash when cleaning up phantom exports (567)
5766 - don't double-finish or set replay data for errored mdc_open requests
5767 - abort requests when they time out, so we don't get old replies
5768 - send/receive replies for AST messages again
5769 - if the client says that it doesn't have the lock, cancel it on the
5771 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5772 mark it as destroyed, it will all work out when decref is called
5773 - fix module use counts (22, 581)
5775 - ASTs now expect a reply (server cancels lock on error reply)
5777 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5780 - fix many simultaneous client startup (392)
5781 - fix dentry->d_it clobbering
5782 - credentials weren't being shipped for readdir/getattr operations
5783 - remove invalid assertions triggered during some concurrent MD
5785 - proper Lustre versions added (336, 389)
5786 - fix memory leak for create error case (398)
5787 - fix LOV locking bug that would get cli/srv out of sync
5788 - fix echo client over LOV (409)
5789 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5790 - fix double-O_EXCL intent crash (424)
5791 - avoid sending multiple lock CANCELs (352)
5793 - MDS can do multi-client recovery (modulo bugs in new code)
5795 - many updates, edits, cleanups
5797 2002-11-18 Phil Schwan <phil@clusterfs.com>
5800 - fix null d_it dereference (346)
5801 - fix full OST/dbench hang (333)
5802 - fix permission problem with file removal (286)
5803 - fix removal of OSCs from LOV when they fail
5804 - fix NULL deref during bulk timeout (214)
5805 - fix problems related to multiple filesystems on one MDS (241)
5806 - fixed serious subtle metadata locking bugs
5807 - free locks on clients when inodes are removed due to memory
5809 - fix inode pointer in lock data (285)
5810 - partial support for multiple MDS on a single host (241)
5811 - data locks weren't cancelled at clear_inode time (290, 311)
5812 - intent locks could lead to unbounded lock growth (205)
5813 - added a maximum lock count, an LRU list, and a flusher
5814 - fix multiple rename (365)
5815 - properly abstracted the echo client
5816 - OSC locked 1 byte too many; fixed
5817 - rewrote brw callback code:
5818 - fixed recovery bugs related to LOVs (306)
5819 - fixed too-many-pages-in-one-write crash (191)
5820 - fixed (again) crash in sync_io_timeout (214)
5821 - probably fixed callback-related race (385)
5823 - Add capability to MDS protocol
5824 - LDLM cancellations and callbacks on different portals
5826 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5829 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5830 - defer open object destruction to close time (601981, 138)
5831 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5832 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5833 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5834 - don't call obd_setattr in ll_file_release for destroyed objects
5835 * protocol change to lustre_msg: move |version| and add |flags|
5836 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5837 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5838 * added replay of create, unlink, link and rename operations during
5839 MDS failover; recovery should be much more robust now
5840 * remove failed OSCs from LOVs (only lov_create uses this so far)
5841 * the lustre-HOWTO was brought (more) up to date (582544)
5843 2002-10-23 Phil Schwan <phil@clusterfs.com>
5846 - in-use dentries weren't being reused properly (617851)
5847 - prevent multiple LDLM setup (599178)
5848 - fix LOV size calculations for truncate (617853)
5849 - fix client handling of MDS intent errors (POSIX)
5850 - fix permission bug in lovstripe.c test (624321)
5851 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5852 - truncate past end of file could corrupt data
5853 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5854 - a race in recovery could return ETIMEDOUT to apps (623947)
5855 - building outside the source directory was fixed
5856 * the lustre-HOWTO was brought (more) up to date (582544)
5857 * major progress was made on recovery functionality
5859 2002-10-10 Phil Schwan <phil@clusterfs.com>
5862 - recovery deadlock fix
5863 - rm -rf causes LBUG fix (617817)
5864 - file open by multiple tasks fix (618962)
5865 - directory permissions bugs (602707 and 620007)
5866 - journal_stop fixed with locking (611313)
5867 - O_APPEND failures resolved (618273, perhaps 614459)
5868 - lconf PATH fix (619770)
5869 - IA64 build fix (621450)
5870 - RPC buffer sizes scale with amount of memory
5872 2002-10-01 Phil Schwan <phil@clusterfs.com>
5875 - locks would be cancelled without throwing away data pages,
5876 resulting in inconsistent data (605627)
5877 - inode attributes were not always being refreshed (605627, 612449)
5878 - lconf now continues to cleanup after lctl reports an error
5879 - MDS now enforces user permissions (602707)
5880 - lprocfs cleanup fixed, but not yet enabled (614157)
5881 - fixed infinite server hang, should a client not respond to an AST
5882 - avoid going into recovery if user calls readlink() with a buffer
5883 that's too small (613941)
5884 - AST RPCs no longer require replies (614867) -- this may be changed
5885 - don't crash server if client sends an IOV that's too big (611336)
5886 - fixed lock conversion deadlock (611892)
5887 - fixed the following of symlinks (614622)
5888 * recovery: the server can remove locks from a client that dies, other
5889 clients can make progress
5890 * more extN patch fixes
5891 * compile-time configurable ptlrpc buffer allocations
5893 - collaborative read cache document
5894 - Lustre Lite Performance CDR document-in-progress
5896 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5899 - fix typo in patch-2.4.18
5901 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5904 - clear ptlrpc request each time in handle_incoming_request()
5905 - unlink of files now destroys the object on the OST
5907 2002-09-19 Peter Braam <braam@clusterfs.com>
5909 * add hard link support
5910 * change obdfile creation method
5911 * kernel patch changed
5913 2002-09-19 Peter Braam <braam@clusterfs.com>
5916 - stack overflow bug in extN fixed
5918 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5920 * documentation updates
5921 - add man pages for config tools
5922 - update tests/README to describe testing with new config tools
5923 - finish metadata API descriptions
5924 * bug fixes and cleanups
5925 - statfs workaround for 16TB limit
5926 - LOV stripe allocation improved, can stripe on subset of OSTs
5927 - LOV file size/IO offset was wrong for files > 4GB in size
5928 - object EA data was being dropped, caused files to be unreadable
5929 - memory overflow with non-LOV OST caused memory corruption
5930 - fixed regression tests to work with new config tools, obdfilter
5931 - fixed bug when directory size became larger than 1 block
5932 - fixed bug (for single client case) when PWD was deleted
5933 - invalidate local directory pages when doing intent-based ops
5934 - avoid LDLM oops when lock callback contained bad data
5936 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5938 * documentation updates
5939 * bug fixes and cleanups
5940 - configuration tools
5943 - 64-bit compile warnings
5944 - 64-bit internal statfs data
5946 * test_brw on persistent OST devices
5948 * lprocfs (disabled)
5950 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5952 * documentation updates
5953 * bug fixes and cleanups
5954 * configuration tools
5956 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5959 * many small fixes to 0_5_4
5960 * io/network handling
5961 * thinkos in MDS operations
5963 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5966 * crucial basic fixes to 0.5.3
5967 * IOR, Iozone work over Elan
5970 2002-08-07 Phil Schwan <phil@clusterfs.com>
5971 * version 0_5_3, our first alpha
5972 * we use the new Portals iovs
5973 * documentation updates
5974 * bug fixes and cleanups
5975 * small changes in the DLM wire protocol
5977 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5978 * version 0_5_1 with some initial stability,
5979 * locking on MD and file I/O.
5980 * documentation updates
5981 * several bug fixes since 0.5.0
5982 * small changes in wire protocol
5984 2002-07-18 Phil Schwan <phil@clusterfs.com>
5986 * delivered as Lustre Light Alpha
5987 * fixed a crash after handling invalid MDS requests
5988 * fixed directory pages for architectures with non-4k pages sizes
5990 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5991 * release version v0_4_4
5992 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5994 * Fixes a number of interruption problems with OST operations.
5995 * Update documentation for portals header changes
5996 * Move all wire protocol structs/defines to lustre_idl.h
5997 * Fixes symlink length bug.
5998 * Add tcpdump to repository.
6000 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6001 * release version v0_4_3
6002 * Fixes statfs for inodes on extN.
6003 * Fixes bug in runtests which would delete /etc/hosts.
6004 * Use 64-bit object IDs wherever possible (not into VFS though)
6005 Remove ost_get_info, which is unused by lustre, and out of date.
6007 2002-07-03 Peter Braam <braam@clusterfs.com>
6008 * release version v0_4_2 Fixes a lookup error (type not passed)
6009 * move forward to head of Portals
6010 * move forward to latest Lustre kernel
6012 2002-06-25 Peter Braam <braam@clusterfs.com>
6013 * release version v0_4_1. Hopefully stable on single node use.