1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.39.3 (SLES 10),
5 2.6.18-128.1.14.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.41.6.sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
14 * File join has been disabled in this release, refer to Bugzilla 16929.
16 Severity : enhancement
18 Description: provide server to client comms path
19 Details : server to client communications path via new LDLM_SET_INFO rpc
23 Description: truncate starts GFP_FS allocation under transaction causing deadlock
24 Details : ldiskfs_truncate calls grab_cache_page which may start page
25 allocation under an open transaction. This may lead to
26 calling prune_icache with consequent lustre reentrance.
30 Description: enable client interface failover
31 Details : When a child reconnects from another NID, properly update export
32 nid hash position and ldlm reverse import.
34 Severity : enhancement
36 Description: Communicate OST degraded/readonly state via statfs to MDS
37 Details : Flags in the statfs returned from OSTs indicate whether the
38 OST is in a degraded RAID state, or if the filesystem has
39 turned read-only after a filesystem error is detected.
44 Description: don't panic if EPROTO was hit when reading symlink
45 Details : correctly handling request reference in error cases.
47 Severity : enhancement
49 Description: Add LustreNetLink, a kernel-userspace communcation path.
51 Severity : enhancement
53 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
58 Description: access to llog context before init.
59 Details : move handling CATALOGS file at osc layer and forbid access to llog
63 Frequency : rare, only when client is writing to OST on local node
65 Description: Avoid deadlock for local client writes
66 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
67 memory freeing context. This allows OST threads to set the
68 PF_MEMALLOC flag on task structures in order to allocate memory
69 from reserved pools and complete IO.
70 Use GFP_HIGHUSER for OST allocations for non-local client writes,
71 so that the OST threads generate memory pressure and allow
72 inactive pages to be reclaimed.
74 Severity : enhancement
76 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
81 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
82 Details : move ldlm namespace creation in setup phase to avoid grab
83 _lprocfs_lock with cli_sem held.
87 Description: Temporarily disable grant shrink.
88 Details : Disable the feature for debugging.
91 Frequency : only when formatting test filesystems
93 Description: Unable to run several mkfs.lustre on loop devices at the same time
94 Details : mkfs.lustre returns error 256 on the concurrent loop devices
95 formatting. The solution is to proper handle the error.
97 Severity : enhancement
99 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
101 Severity : enhancement
103 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
108 Description: resolve race between obd_disconnect and class_disconnect_exports
109 Details : if obd_disconnect will be called to already disconnected export he
110 forget release one reference and osc module can't unloaded.
112 Severity : enhancement
114 Description: Allow tuning service thread via /proc
115 Details : For each service a new
116 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
117 created that can be used to set min/max thread counts, and get the
118 current number of running threads.
123 Descriptoin: don't return error if have partially created objects for file.
124 Details : lov_update_create_set uses set->set_success as index for created
125 objects, so if some requests will be failed, they will have hole
126 at end of array and qos_shrink_lsm for allocate correct lsm.
128 Severity : enhancement
130 Description: Update OFED support to 1.4.1
135 Description: Reduce small size read RPC
136 Details : Set read-ahead limit for every file and only do read-ahead when
137 available read-ahead pages are bigger than 1M to avoid small size
140 Severity : enhancement
142 Description: move AT tunable parameters for more consistent usage
143 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
145 Severity : enhancement
147 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
148 Details : allow skip disconnected ost for send statfs request and hide error
154 Description: fix racy locking of mballoc block bitmaps causing BUG
155 Details : The locking of the mballoc buddy bitmap and the in-memory
156 block bitmap was using two different spin locks in some
157 cases. This made it possible to incorrectly access the
158 mballoc bitmap while another process was modifying it,
159 causing a sanity assertion to fail. While no on-disk corruption
160 was reported, there was some risk of this happening.
163 Frequency : rare, on llog test 6
165 Descriptoin: don't allow connect to already connected import
166 Details : allowing connect to already connected import is hide connecting problem.
171 Descriptoin: allow kill process which wait statahead result
172 Details : for some reasons 'ls' can stick in waiting result from statahead,
173 in this case need way for kill this process.
175 Severity : enhancement
177 Description: Add state history info file, enhance import info file
178 Details : Track import connection state changes in a new osc/mdc proc file;
179 add overview-type data to the osc/mdc import proc file.
181 Severity : enhancement
183 Description: MDS create should not wait for statfs RPC while holding DLM lock.
185 Severity : enhancement
187 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
192 Descriptoin: OST grant too much space to client even there are not enough space.
193 Details : Client will shrink its grant cache to OST if there are no write
194 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
195 this grant cache if there are already not enough avaible space
196 (left_space < total_clients * 32M).
199 Frequency : start MDS on uncleanly shutdowned MDS device
201 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
202 Details : stay in waiting mds<>ost recovery finished produce random bugs
203 due race between two ll_sync thread for one lov target. send
204 ACTIVATE event only if connect realy finished and import have
208 Frequency : rare, connect and disconnect target at same time
210 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
211 Details : don't call obd_disconnect under lov_lock. this long time
212 operation and can block ptlrpcd which answer to connect request.
217 Descriptoin: don't lose wakeup for imp_recovery_waitq
218 Details : recover_import_no_retry or invalidate_import and import_close can
219 both sleep on imp_recovery_waitq, but we was send only one wakeup
223 Frequency : always with long access acl
225 Descriptoin: mds can't pack reply with long acl.
226 Details : mds don't control size of acl but they limited by reint/getattr
229 Severity : enhancement
231 Description: Update to SLES10 kernel-2.6.16.60-0.33.
233 Severity : enhancement
235 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
238 Frequency : start MDS on uncleanly shutdowned MDS device
240 Descriptoin: aborting recovery hang on MDS
241 Details : don't throttle destroy RPCs for the MDT.
244 Frequency : on remount
246 Description: external journal device not working after the remount
247 Details : clear dev_rdonly flag for external journal devices in
253 Description: shutdown vs evict race
254 Details : client_disconnect_export vs connect request race.
255 if client will evicted at this time - we start invalidate
256 thread without referece to import and import can be freed
260 Frequency : rare, need acl's on inode.
262 Description: client can't handle ost additional correctly
263 Details : if ost was added after client connected to mds client can have
264 hit lnet_try_match_md ... to big messages to wide striped files.
265 in this case need teach client to handle config events about add
266 lov target and update client max ea size at that event.
268 Severity : enhancement
270 Description: Changelogs
271 Details : Changelogs are a lightweight mechanism to track filesystem
272 metadata and namespace changes. The changelog is recorded
273 permanently on the MDTs, and is periodically "consumed" / purged
274 when records are no longer needed.
276 Severity : enhancement
278 Description: compact fld format with extents
279 Details : Store range of seq rather than every seq in FLD. Seq
280 controller update FLD rather than clients. In Case of CMD, mdt0
281 has FLD, all other metadata server act as non persistent proxy
282 for FLD queries and cache fld entries in fld cache.
287 Description: don't skip ost target if they assigned to file
288 Details : Drop slow OSCs if we can, but not for requested start idx.
289 This means "if OSC is slow and it is not the requested
290 start OST, then it can be skipped, otherwise skip it only
291 if it is inactive/recovering/out-of-space.
295 Description: more cleanup in mds_lov
296 Details : not send LOV EA under replay, we can't know about they size at this
297 time. Don't allow client connect to mds before any ost connected,
298 for avoid problems with LOV EA size and returning EIO to client.
300 Severity : enhancement
302 Description: Interoperability at server side (Disk interoperability)
304 Severity : enhancement
306 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
308 Severity : enhancement
310 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
312 Severity : enhancement
314 Description: New client IO stack (CLIO).
316 Severity : enhancement
318 Description: Commit on sharing. Eliminate inter-client dependencies between
319 uncommitted transactions by doing transaction commits.
320 Thereby clients may recovery independently.
323 Frequency : Create a symlink file with a very long name
325 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
326 Details : If there is no extra space in the request for early cancels,
327 ldlm_req_handles_avail() returns 0 instead of a negative value.
329 Severity : enhancement
331 Description: Add /proc entry for import status
332 Details : The mdc, osc, and mgc import directories now have
333 an import directory that contains useful import data for debugging
336 Severity : enhancement
338 Description: Re-disable certain /proc logging
339 Details : Enable and disable client's offset_stats, extents_stats and
340 extents_stats_per_process stats logging on the fly.
343 Frequency : Only on FC kernels 2.6.22+
345 Description: oops in statahead
346 Details : Do not drop reference count for the dentry from VFS when lookup,
347 VFS will do that by itself.
349 Severity : enhancement
351 Description: Generic /proc file permissions
352 Details : Set /Proc file permissions in a more generic way to enable non-
353 root users operate on some /proc files.
357 Description: Hitting mdc_commit_close() ASSERTION
358 Details : Properly handle request reference release in
359 ll_release_openhandle().
363 Description: quota recovery deadlock during mds failover
364 Details : This patch includes att18982, att18236, att18237 in bz14840.
366 1. fix osts hang when mds does failover with quotaon
367 2. prevent watchdog storm when osts threads wait for the
372 Frequency : only patchless client
373 Description: add workaround for race between add/remove dentry from hash
375 Severity : enhancement
377 Description: Allow OST glimpses to return PW locks
381 Description: LBUG when llog conf file is full
382 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
386 Description: Prevent import from entering FULL state when server in recovery
390 Description: service mount cannot take device name with ":"
391 Details : Only when device name contains ":/" will mount treat it as
397 Description: replace ptlrpcd with the statahead thread to interpret the async
398 statahead RPC callback
402 Frequency : on recovery
403 Description: I/O failures after umount during fail back
404 Details : if client reconnected to restarted server we need join to recovery
405 instead of find server handler is changed and process self
406 eviction with cancel all locks.
408 Severity : enhancement
410 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
412 Severity : enhancement
414 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
416 Severity : enhancement
418 Description: Upcall on Lustre log has been dumped
419 Details : Allow for a user mode script to be called once a Lustre log has
420 been dumped. It passes the filename of the dumped log to the
421 script, the location of the script can be specified via
422 /proc/sys/lnet/debug_log_upcall.
427 Description: avoid idr_remove called for id which is not allocated.
428 Details : Move assigment s_dev for clustered nfs to end of initialization,
429 to avoid problem with error handling.
434 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
435 Details : When connection is reused this not moved from CONN_UNUSED_HASH
436 into CONN_USED_HASH and this prodice warning when put connection
437 again in unused hash.
440 Severity : enhancement
442 Description: File striping can now be set to use an arbitrary pool of OSTs.
444 Severity : enhancement
446 Description: Export bytes_read/bytes_write count on OSC/OST.
450 Description: Early reply size mismatch, MGC loses connection
451 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
452 the connect flags are properly negotiated.
457 Description: quotas are not honored with O_DIRECT
458 Details : all writes with the flag O_DIRECT will use grants which leads to
459 this problem. Now using OBD_BRW_SYNC to guard this.
463 Description: add quota statistics
464 Details : 1. sort out quota proc entries and proc code.
465 2. add quota statistics
467 Severity : enhancement
469 Description: enable quota support for HEAD.
473 Description: Properly propagate oinfo flags from lov to osc for statfs
474 Details : restore missing copy oi_flags to lov requests.
476 Severity : enhancement
478 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
479 : l_getgroups(8), lst(8), routerstat(8)
480 Details : included man pages for llobdstat(8), llstat(8),
481 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
483 Severity : enhancement
485 Description: configure's --enable-quota should check the
486 : kernel .config for CONFIG_QUOTA
487 Details : configure is terminated if --enable-quota is passed but
488 : no quota support is in kernel
492 Description: Remove portals compatibility
493 Details : Remove portals compatibility, not interoperable with releases
498 Description: Resolve device initialization race
499 Details : Prevent proc handler from accessing devices added to the
500 obd_devs array but yet be intialized.
502 Severity : enhancement
504 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
506 Severity : enhancement
508 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
513 Description: Using wrong pointer in osc_brw_prep_request
514 Details : Access to array[-1] can produce panic if kernel compiled with
515 CONFIG_PAGE_ALLOC enabled
519 Description: Client runs out of low memory
520 Details : Consider only lowmem when counting initial number of llap pages
524 Description: *optional* service tags registration
525 Details : if the "service tags" package is installed on a Lustre node
526 When the filesystem is mounted, a local-node service tag will
527 be created. See http://inventory.sun.com/ for more information
528 about the Service Tags asset management system.
532 Description: Kernel BUG tries to release flock
533 Details : Lustre does not destroy flock lock before last reference goes
534 away. So always drop flock locks when client is evicted and
535 perform unlock regardless of successfulness of speaking to MDS.
539 Description: add refcount protection for osc callbacks, avoid panic on shutdown
543 Description: sanity test 65a fails if stripecount of -1 is set
544 Details : handle -1 striping on filesystem in ll_dirstripe_verify
549 Description: ASSERTION(CheckWriteback(page,cmd)) failed
550 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
553 Severity : enhancement
555 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
559 Description: do not process already freed flock
560 Details : flock can possibly be freed by another thread before it reaches
561 to ldlm_flock_completion_ast.
565 Description: LBUG during stress test
566 Details : Need properly lock accesses the flock deadlock detection list.
570 Description: oops in page fault handler
571 Details : kernel page fault handler can return two special 'pages' in error
572 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
576 Description: timeout with invalidate import.
577 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
578 should be handled by ptlrpcd. This produce long age waiting and
579 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
581 Severity : enhancement
583 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
587 Description: enable MGS and MDT services start separately
588 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
589 a co-located MGS without starting the MGS, which is a complement
590 to 'nosvc' mount option.
594 Frequency : after recovery
595 Description: precreate to many object's after del orphan.
596 Details : del orphan st in oscc last_id == next_id and this triger growing
597 count of precreated objects. Set flag LOW to skip increase count
598 of precreated objects.
602 Frequency : rare, on clear nid stats
603 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
604 Details : when clean nid stats sometimes try destroy live entry,
605 and this produce panic in free.
609 Description: Stack overflow during MDS log replay
610 ease stack pressure by using a thread dealing llog_process.
614 Description: wait until IO finished before start new when do lock cancel.
615 Details : VM protocol want old IO finished before start new, in this case
616 need wait until PG_writeback is cleared until check dirty flag and
617 call writepages in lock cancel callback.
619 Severity : enhancement
621 Description: using special macro for print time and cleanup in includes.
625 Description: mds_mfd_close() ASSERTION(rc == 0)
626 Details : In mds_mfd_close(), we need protect inode's writecount change
627 within its orphan write semaphore to prevent possible races.
631 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
635 Frequency : rare, on shutdown ost
636 Description: don't hit live lock with umount ost.
637 Details : shrink_dcache_parent can be in long loop with destroy dentries,
638 use shrink_dcache_sb instead.
642 Description: don't panic with use echo client
643 Details : echo client pass NULL as client nid pointer and this produce null
648 Description: fix build on ppc32
649 Details : compile code with -m64 flag produce wrong object file for ppc32.
653 Description: add message levels for liblustreapi
657 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
658 Details : Move the 'good_osts' check before the 'total_bavail' check. This
659 will result in an -EAGAIN and in the exit call path we call
660 alloc_rr() which will with increasing aggressiveness attempt to
661 aquire precreated objects on the minimum number of required OSCs.
665 Description: Use old size assignment to avoid deadlock
666 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
667 the scheduling while holding a spinlock. We do not need locking
668 for size in ll_update_inode() because size is only updated from
669 the MDS for directories or files without objects, so there is no
670 other place to do the update, and concurrent access to such inodes
671 are protected by the inode lock.
675 Description: resolve "_IOWR redefined" build error on SLES10
679 Description: dump the memory debugging after all modules are unloaded to
680 suppress false negative in conf_sanity test 39
682 Severity : enhancement
684 Description: build kernel-ib packages for OFED 1.3 in our release cycle
689 Description: fix SLES kernel versioning
690 Details : the kernel version for our SLES 10 kernel did not include a "-"
691 before the "smp" at the end. while this was not a problem in
692 general, it did mean that software trying to use the kernel
693 version to try to detect a vendor specific kernel would fail.
694 this was most evident by the OFED build scripts.
698 Description: Don't update lov_desc members until making sure they are valid
699 Details : When updating lov_desc members via proc fs, need fix their
700 validities before doing the real update.
704 Description: don't put request into delay list while invalidate in flight.
705 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
706 import in flight. this produce timeout for invalidate and sometimes
707 can cause stale data.
711 Frequency : on ppc only
712 Description: not convert ost objects for directory because it's not exist.
713 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
717 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
719 Severity : enhancement
721 Description: Update to support 2.6.22.14 vanilla kernel.
725 Frequency : rare, on recovery
726 Description: read procfs can produce deadlock in some situation
727 Details : Holding lprocfs lock which send rpc can produce block for destroy
728 obd objects and this also block reconnect with -EALREADY. This
729 isn't fix all lprocfs bugs - but make it rare.
731 Severity : enhancement
733 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
736 Frequency : frequent on X2 node
738 Description: mdc_set_open_replay_data LBUG
739 Details : Set replay data for requests that are eligible for replay.
743 Description: lustre_mgs: operation 101 on unconnected MGS
744 Details : When MGC is disconnected from MGS long enough, MGS will evict the
745 MGC, and late on MGC cannot successfully connect to MGS and a lot
746 of the error messages complaining that MGS is not connected.
750 Frequency : on network error
751 Description: panic with double free request if network error
752 Details : mdc_finish_enqueue is finish request if any network error ocuring,
753 but it's true only for synchronus enqueue, for async enqueue
754 (via ptlrpcd) this incorrect and ptlrpcd want finish request
757 Severity : enhancement
759 Description: client-side metadata stat-ahead during readdir(directory readahead)
760 Details : perform client-side metadata stat-ahead when the client detects
761 readdir and sequential stat of dir entries therein
764 Frequency : on start mds
766 Description: Implement get_info(last_id) in obdfilter.
769 Frequency : occasional
771 Description: Correctly check stale fid, not start epoch if ost not support SOM
772 Details : open with flag O_CREATE need set old fid in op_fid3 because
773 op_fid2 was overwritten with new generated fid, but mds can answer
774 with one of these two fids and both is not stale. Setattr
775 incorrectly started an epoch and assume will be called
776 done_writing, but without SOM done_writing ever being called.
779 Frequency : rare, depends on device drivers and load
781 Description: MDS or OSS nodes crash due to stack overflow
782 Details : Code changes in 1.8.0 increased the stack usage of some functions.
783 In some cases, in conjunction with device drivers that use a lot
784 of stack the MDS (or possibly OSS) service threads could overflow
785 the stack. One change which was identified to consume additional
786 stack has been reworked to avoid the extra stack usage.
789 Frequency : occasional
791 Description: Do not fail import if osc_interpret_create gets -EAGAIN
792 Details : If osc_interpret_create got -EAGAIN it immediately exits and
793 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
794 oscc_has_objects and see OSC has no objests and call
795 oscc_internal_create to resend create request.
797 Severity : enhancement
799 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
801 Severity : enhancement
803 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
806 Frequency : very rare
808 Description: Fix replay if there is an un-replied request and open
809 Details : In some cases, older replay request will revert the
810 mcd->mcd_last_xid on MDS which is used to record the client's
813 Severity : enhancement
815 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
817 Severity : enhancement
819 Description: Add rhel5 support to HEAD.
821 Serverity : enhancement
823 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
828 Description: Don't allow skipping OSTs if index has been specified.
829 Details : Don't allow skipping OSTs if index has been specified, make
830 locking in internal create lots better.
834 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
835 Details : access to bitfield in structure is always rounded to long
836 and this produce problem with not atomic change any bit.
840 Description: Lustre make rpms failed.
841 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
842 builds Lustre rpms from tarball.
845 Frequency : rare on shutdown ost
847 Description: If llog cancel was not send before clean_exports phase, this can
848 produce deadlock in llog code.
849 Details : If llog thread has last reference to obd and call class_import_put
850 this produce deadlock because llog_cleanup_commit_master wait when
851 last llog_commit_thread exited, but this never success because was
852 called from llog_commit_thread.
856 Description: allow userland application know is lost one of stripes.
857 Details : fill lvb_blocks with error code on ost and return it to
858 application if error flag found.
862 Description: NULL lov_tgts causing MDS oops
863 Details : more safe checks for NULL lov_tgts for avoid oops.
865 Severity : enhancement
867 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
871 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
872 Details : If an OST is down the MDS will hang indefinitely in
873 obd_statfs_rqset() waiting for the statfs data. While for
874 MDS QOS usage of statfs, it should not stuck in waiting.
876 Severity : enhancement
878 Description: remote_acl support
879 Details : Support ACL-based permission check for remote user.
880 Support setfacl/getfacl for remote user with the utils
881 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
882 the system "{s,g}etfacl" utils.
884 Severity : enhancement
886 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
888 Severity : enhancement
890 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
894 Description: make mgs_setparam() handle fsname containing dash
895 Details : fsname containing a dash does not work with lctl conf_param
897 Severity : enhancement
899 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
901 Severity : enhancement
903 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
907 Frequency : rare, at shutdown
908 Description: access already free / zero obd_namespace.
909 Details : if client_disconnect_export was called without force flag set,
910 and exist connect request in flight, this can produce access to
911 NULL pointer (or already free pointer) when connect_interpret
912 store ocd flags in obd_namespace.
916 Frequency : only at startup
917 Description: not alloc memory with spinlock held.
918 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
919 if any spinlock held.
921 Severity : enhancement
923 Description: make lustre randomly fail allocating memory
924 Details : Make lustre randomly failed allocating memory for testing purpose.
926 Severity : enhancement
928 Description: lost problems with lov objid file
929 Details : Fixes some scability and access to not inited memory problems
930 in work with lov objdid file.
935 Description: lfs find does not continue on file error
936 Details : Continue other files processing when a file/dir is absent.
940 Description: Inconsistent usage of lustre_pack_reply()
941 Details : Standardize the usage of lustre_pack_reply() such that it
942 always generate a CERROR on failure.
945 Frequency : occasional
947 Description: MDS hang or stay in waiting lock
948 Details : If client receive lock with CBPENDING flag ldlm need send lock
949 cancel as separate rpc, to avoid situation when cancel request
950 can't processed due all i/o threads stay in wait lock.
954 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
958 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
962 Description: Update to SLES9 kernel-2.6.5-7.287.3.
966 Description: Update to RHEL4 latest kernel.
968 Severity : enhancement
970 Description: Build SLES10 patchless client fails
971 Details : The configure was broken by run ./configure with
972 --with-linux-obj=.... argument for patchless client. When the
973 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
974 can't search header adequately. Use absolute path such as
975 -I($LINUX)/include instead.
979 Description: interrupt oig_wait produce painc on resend.
980 Details : brw_redo_request can be used for resend requests from ptlrpcd and
981 private set, and this produce situation when rq_ptlrpcd_data not
982 copyed to new allocated request and triggered LBUG on assert
983 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
984 ptlrpcd set if request is changed and can be safety changed to use
989 Description: LASSERT_{REQ,REP}SWAB macros are buggy
990 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
991 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
992 these macros and replace them with their difinitions instead.
996 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
997 Details : Update which_patch & target file for SLES10 latest kernel.
1000 Description: Remove client patches from SLES 10 kernel.
1001 Details : This causes SLES 10 clients to behave as patchless clients
1002 even on a Lustre-patched (server) kernel.
1004 Severity : enhancement
1006 Description: self-adjustable client's lru lists
1007 Details : use adaptive algorithm for managing client cached locks lru
1008 lists according to current server load, other client's work
1009 pattern, memory activities, etc. Both, server and client
1010 side namespaces provide number of proc tunables for controlling
1013 Severity : enhancement
1015 Description: light-weight GSS support
1016 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1017 and reduce performance overhead.
1019 Severity : enhancement
1021 Description: Linux keyring support
1022 Details : Support using service of Linux keyring for Lustre GSS internal
1023 context refresh/cache mechanism.
1027 Description: Fix errors in lfs documentation
1028 Details : Fixes man pages
1032 Description: don't use GFP_* in generic Lustre code.
1033 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1037 Description: obdclass is limited by single OBD_ALLOC(idarray)
1038 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1042 Description: warnings with build patchless client with vanila 2.6.19 and up
1043 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1044 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1048 Frequency : Only for SLES
1049 Description: sanity_quota fail test_1
1050 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1051 file, which makes TSTID[2] inunique.
1055 Frequency : Only for SLES9
1056 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1057 Details : Update target/ChangeLog/which_patch .
1061 Frequency : Only for test_10 in sanity-quota.sh
1062 Description: a bug in quota test code
1063 Details : checking a test flag in a wrong place
1067 Frequency : Only for test_13 in sanity-quota.sh
1068 Description: a bug in quota test script
1069 Details : shouldn't check overfull allocated quota
1073 Description: jbd statistics
1074 Details : Port older jbd statistics patch for sles10
1076 Severity : enhancement
1078 Description: Add printing inode info into message about error in writepage.
1083 Description: reply_lock_interpret crash due to race with it and lock cancel.
1084 Details : Do not replay locks that are being cancelled. Do not reference
1085 locks by their address during replay, just by their handle.
1089 Frequency : When flocks are used.
1090 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1091 Details : Flock locks might destroy just granted lock if it could be merged
1092 with another existing flock, this is done in completion handler,
1093 so teach ldlm_cli_enquque_fini that this is a valid case for
1099 Description: Oops in read and write path when failing to allocate lock.
1100 Details : Check if lock allocation failed and return error back.
1104 Description: lstripe command fails for valid OST index
1105 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1106 instead of lov->desc.ld_active_tgt_count.
1110 Description: sanity.sh failed test 103
1111 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1116 Description: df doesn't work properly if diskfs blocksize != 4K
1117 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1121 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1122 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1124 Severity : enhancement
1126 Description: Adaptive timeouts
1127 Details : RPC timeouts adapt to changing server load and network
1128 conditions to reduce resend attempts and improve recovery time.
1132 Description: llapi_file_create() does not allow some changes
1133 Details : add llapi_file_open() that allows specifying the mode and
1134 open flags, and also returns an open file handle.
1138 Description: merge and cleanup kernel patches.
1139 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1143 Description: Add journal checksum support.(Kernel part)
1144 Details : The journal checksum feature adds two new flags i.e
1145 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1146 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1147 indicates that the commit block contains the checksum for
1148 the blocks described by the descriptor blocks. Now commit
1149 record can be sent to disk without waiting for descriptor
1150 blocks to be written to disk. This behavior is controlled
1151 using JBD2_FEATURE_ASYNC_COMMIT flag.
1155 Description: OSS needs mutliple precreate threads
1156 Details : Add ability to start more than one create thread per OSS.
1160 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1161 Details : execute lfs setstripe on client
1165 Description: mds_obd_create error creating tmp object
1166 Details : When the user sets quota on root, llog will be affected and can't
1167 create files and write files.
1170 Frequency : Always on ia64 patchless client, and possibly others.
1172 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1173 Details : This allows the patchless client to be loaded on architectures
1174 without this export.
1178 Description: RedHat Update kernel for RHEL5
1179 Details : Add the kernel config file for RHEL5.
1183 Description: RedHat Update kernel for RHEL5
1184 Details : Modify the kernel config file more closer RHEL5.
1188 Description: Build failure against Centos5 (RHEL5)
1189 Details : Define PAGE_SIZE when it isn't present.
1193 Description: test_7 on sanity-quota.sh can't be used on separate machines
1194 Details : add facet to handle it
1198 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1199 Details : with NFS, the anon dentry's parent was set to itself in
1200 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1201 corresponding dentry other than use rec->ur_name.
1203 Severity : enhancement
1205 Description: lfs setstripe enhancement
1206 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1210 Description: enable data checksumming by default
1211 Details : enable checksum by default, allow --disable-checksum
1212 configure option and "-o nochecksum" mount option. Checksums
1213 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1217 Description: lustre support for RHEL5
1218 Details : Add support for RHEL5.
1222 Description: Client eviction due to failover config
1223 Details : after a connection loss, the lustre client should attempt to
1224 reconnect to the last active server first before trying the
1225 other potential connections.
1229 Description: when mds and osts use different quota unit(32bit and 64bit),
1230 quota will be released repeatly.
1231 Details : void sending multiple quota reqs to mds, which will keep the
1232 status between the reqs.
1236 Description: osts not allocated evenly to files
1237 Details : change the condition to increase offset_idx
1240 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1242 Description: Data corruption for OSTs that are formatted larger than 2TB
1244 Details : When generating the bio request for lustre file writes the
1245 sector number would overflow a temporary variable before being
1246 used for the IO. The data reads correctly from Lustre (which
1247 will overflow in a similar manner) but other file data or
1248 filesystem metadata may be corrupted in some cases.
1252 Description: Tune the kernel for good SCSI performance.
1253 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1254 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1259 Description: rewrite ext2-derived code in obdclass/uuid.c
1260 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1261 from scratch preserving functionality.
1265 Description: "lfs find -obd UUID" prints directories
1266 Details : "lfs find -obd UUID" will return all directory names instead
1267 of just file names. It is incorrect because the directories
1268 do not reside on the OSTs.
1272 Description: use i_size_read and i_size_write in 2.6 port
1273 Details : replace inode->i_size access with i_size_read/write()
1276 Frequency : when removing large files
1278 Description: scheduling issue during removal of large Lustre files
1279 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1280 It causes scheduling issues when removing large files (17TB in the
1284 Frequency : only with liblustre clients on XT3
1286 Description: evictions taking too long
1287 Details : allow llrd to evict clients directly on OSTs
1290 Frequency : only on ppc
1292 Description: /proc/fs/lustre/devices broken on ppc
1293 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1294 We should make sure the type of 'index' is loff_t and then cast
1295 explicitly as needed below. Do not assign an explicitly cast
1299 Frequency : only for rhel5
1301 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1302 Details : Modify the target file & which_kernel.
1304 Severity : enhancement
1306 Description: omit set fsid for export NFS
1307 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1308 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1312 Frequency : Always, for kernels after 2.6.16
1313 Description: Fix warning idr_remove called for id=.. which is not allocated.
1314 Details : Last kernels save old s_dev before kill super and not allow
1315 to restore from callback - restore it before call kill_anon_super.
1319 Description: buffer overruns could theoretically occur
1320 Details : llapi_semantic_traverse() modifies the "path" argument by
1321 appending values to the end of the origin string, and a
1322 overrun may occur. Adding buffer overrun check in liblustreapi.
1326 Description: Fix error on 'ls .' at the top of the Lustre mount.
1327 Details : Don't revalidate dentry if it is a root dentry.
1331 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1332 Details : Modify vm-tunables-rhel4.patch.
1336 Description: Kernel config for 2.6.18-vanilla.
1337 Details : Modify targets/2.6-vanilla.target.in.
1338 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1339 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1340 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1341 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1345 Description: improve handling recoverable errors
1346 Details : if request processig with error which can be recoverable on server
1347 request should be resend, otherwise page released from cache and
1352 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1353 Details : Modify target file & which_patch.
1354 A flaw was found in the IA32 system call emulation provided
1355 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1356 value could be stored in the %RAX register, which could trigger an
1357 out-of-bounds system call table access. An untrusted local user
1358 could exploit this flaw to run code in the kernel
1359 (ie a root privilege escalation). (CVE-2007-4573).
1363 Description: O_DIRECT bypasses client statistics.
1364 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1365 read_ahead_stats were updated. copied the stats section from
1366 osc_send_oap_rpc() into async_internal().
1370 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1374 Description: change order of libsysio includes
1375 Details : '#include sysio.h' should always come before '#include xtio.h'
1379 Description: handle "serious error: objid * already exists" more gracefully
1380 Details : If LAST_ID value on disk is smaller than the objects existing in
1381 the O/0/d* directories, it indicates disk corruption and causes an
1382 LBUG(). If the object is 0-length, then we should use the existing
1383 object. This will help to avoid a full fsck in most cases.
1385 Severity : enhancement
1387 Description: adapt the lustre_config script to support the upgrade case
1388 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1389 server targets to 1.6 in parallel.
1393 Description: To avoid grant space > avaible space when the disk is almost
1394 full. Without this patch you might see the error "grant XXXX >
1395 available" or some LBUG about grant, when the disk is almost
1397 Details : In filter_check_grant, for non_grant cache write, we should
1398 check the left space by if (*left > ungranted + bytes), instead
1399 of (*left > ungranted), because only we are sure the left space
1400 is enough for another "bytes", then the ungrant space should be
1401 increase. In client, we should update cl_avail_grant only there
1402 is OBD_MD_FLGRANT in the reply.
1406 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1407 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1408 described in CVE-2007-4573. This problem could allow untrusted
1409 local users to gain root access.
1412 Frequency : when using O_DIRECT and quotas
1414 Description: Incorrect file ownership on O_DIRECT output files
1415 Details : block usage reported by 'lfs quota' does not take into account
1416 files that have been written with O_DIRECT.
1421 Description: touch file failed when fs is not full
1422 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1423 otherwise we can get ENOSP while fs is not full.
1427 Description: parallel lock callbacks
1428 Details : Instead of sending blocking and completion callbacks as separated
1429 requests, adding them to a set and sending in parallel.
1432 Frequency : only for Cray XT3
1433 Bugzilla : 12829/13455
1434 Description: Changing primary group doesn't change the group lustre assigns to
1436 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1437 overridden with the primary group provided by the group upcall,
1438 whereas the supplied fsgid can be trusted if it is in the list of
1439 supplementary groups returned by the group upcall.
1441 Severity : enhancement
1443 Description: Allow masking D_WARNING, D_ERROR messages from console
1444 Details : Console messages can now be disabled via lnet.printk.
1448 Description: User code with malformed file open parameter crashes client node
1449 Details : Before packing join_file req, all the related reference should be
1450 checked carefully in case some malformed flags cause fake
1451 join_file req on client.
1455 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1456 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1457 mode is equal to the granted mode. However, it should also check
1458 that LDLM_FL_CP_REQD is not set.
1462 Description: Heavy nfs access might result in deadlocks
1463 Details : After ELC code landed, it is now improper to enqueue any mds
1464 locks under och_sem, because enqueue might want to decide to
1465 cancel open locks for same inode we are holding och_sem for.
1469 Description: Client eviction while running blogbench
1470 Details : A lot of unlink operations with concurrent I/O can lead to a
1471 deadlock causing evictions. To address the problem, the number of
1472 oustanding OST_DESTROY requests is now throttled to
1473 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1474 callbacks are processed in priority.
1478 Description: enable ACLs on MDS by default
1479 Details : ACLs must be enabled on MDS by default.
1482 Frequency : PPC/PPC64 only
1484 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1485 Details : fix duplicated definitions between asm-ppc64/types.h and
1486 lustre_types.h on PPC.
1489 Frequency : PPC/PPC64 only
1491 Description: asm-ppc/segment.h does not exist
1492 Details : fix compile issue on PPC.
1496 Description: better handle error messages in extents code
1499 Frequency : RHEL4 only
1501 Description: mkfs is very slow on IA64/RHEL4
1502 Details : A performance regression has been discovered in the MPT Fusion
1503 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1504 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1505 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1507 Severity : enhancement
1509 Description: SNMP support enhancement
1510 Details : Adding total number of sampled request for an MDS node in snmp
1513 Severity : enhancement
1515 Description: Optimize ldlm waiting list processing for PR extent locks
1516 Details : When processing waiting list for read extent lock and meeting read
1517 lock that is same or wider to it that is not contended, skip
1518 processing rest of the list and immediatelly return current
1519 status of conflictness, since we are guaranteed there are no
1520 conflicting locks in the rest of the list.
1524 Description: Time out and refuse to reconnect
1525 Details : When the failover node is the primary node, it is possible
1526 to have two identical connections in imp_conn_list. We must
1527 compare not conn's pointers but NIDs, otherwise we can defeat
1528 connection throttling.
1532 Description: port llog fixes from b1_6 into HEAD
1533 Details : Port llog reference couting and some llog cleanups from b1_6
1534 (bug 10800) into HEAD, for protect from panic and access to already
1535 free llog structures.
1539 Description: Detect stride IO mode in read-ahead
1540 Details : When a client does stride read, read-ahead should detect that and
1541 read-ahead pages according to the detected stride pattern.
1545 Description: data checksumming impacts single node performance
1546 Details : add support for several checksum algorithm. Currently, only CRC32
1547 and Adler-32 are supported. The checksum type can be changed on
1548 the fly via /proc/fs/lustre/osc/*/checksum_type.
1552 Description: use adler32 for page checksums
1553 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1558 Description: build for x2 fails
1559 Details : fix compile issue on Cray systems.
1563 Description: Properly match for duplicate locks
1564 Details : Due to different lock order from skiplists code, we need to
1565 traverse entire list for now
1568 Frequency : only on PPC/SLES10
1570 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1571 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1572 be defined. Add a hack in mkfs_lustre.c to work around this compile
1577 Description: LASSERT on MDS when client holding flock lock dies
1578 Details : ldlm pool logic depends on number of granted locks equal to
1579 number of released locks which is not true for flock locks, so
1580 just exclude such locks from consideration.
1584 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1585 Details : Use fsfilt_sync() for both the whole filesystem sync and
1586 individual file sync to eliminate dangerous inode locking
1587 with I_LOCK that can lead to a deadlock.
1591 Description: performance in 1.6.3
1592 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1593 sure that 1MB requests are not fragmented by the block layer.
1595 Severity : enhancement
1597 Description: organize the server-side client stats on per-nid basis
1598 Details : Change the structure of stats under obdfilter and mds to
1606 The "uuid"s file would list the uuids of _active_ exports.
1607 And the clear entry is to clear all stats and stale nids.
1609 Severity : enhancement
1611 Description: eliminate client locks in face of contention
1612 Details : file contention detection and lockless i/o implementation
1613 for contended files.
1617 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1620 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1622 Description: MDS may lose file striping (and hence file data) in some cases
1623 Details : If there are additional extended attributes stored on the MDS,
1624 in particular ACLs, SELinux, or user attributes (if user_xattr
1625 is specified for the client mount options) then there is a risk
1626 of attribute loss. Additionally, the Lustre file striping
1627 needs to be larger than default (e.g. striped over all OSTs),
1628 and an additional attribute must be stored initially in the
1629 inode and then increase in size enough to be moved to the
1630 external attribute block (e.g. ACL growing in size) for file
1635 Description: skiplist implementation simplification
1636 Details : skiplists are used to group compatible locks on granted list
1637 that was implemented as tracking first and last lock of each
1638 lock group the patch changes that to using doubly linked lists
1642 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1643 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1644 because LASSERT on both the data supplied by a client, and the
1645 data on disk is dangerous and incorrect.
1647 Severity : enhancement
1649 Description: Slow truncate/writes to huge files at high offsets.
1650 Details : Directly associate cached pages to lock that protect those pages,
1651 this allows us to quickly find what pages to write and remove
1652 once lock callback is received.
1656 Description: more ldlm soft lockups
1657 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1658 starve other threads from the resource lock for a long time in
1659 case of long waiting queue, so change the debug level from
1660 D_OTHER to the less frequently used D_INFO.
1662 Severity : enhancement
1664 Description: add -gid, -group, -uid, -user options to lfs find
1668 Description: Hung threads in invalidate_inode_pages2_range
1669 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1670 once one is completed. As a result, some RPCs are stuck in the
1671 queue and are never sent.
1675 Description: filter threads hungs on waiting journal commit
1676 Details : Cleanup filter group llog code, then only filter group llog will
1677 be only created in the MDS/OST syncing process.
1681 Description: Procfs and llog threads access destoryed import sometimes.
1682 Details : Sync the import destoryed process with procfs and llog threads by
1683 the import refcount and semaphore.
1685 Severity : enhancement
1687 Description: openlock cache of b1_6 port to HEAD
1692 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1693 Details : If insertion of an extent fails, then discard the inode
1694 preallocation and free data blocks else it can lead to duplicate
1699 Description: don't always update ctime in ext3_xattr_set_handle()
1700 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1701 In some cases the ctime should not be updated, for example for
1702 2.0->1.8 compatibility it is necessary to delete an xattr and it
1703 should not update the ctime.
1707 Bugzilla : 15713/16362
1708 Description: Assertion in iopen_connect_dentry in 1.6.3
1709 Details : looking up an inode via iopen with the wrong generation number can
1710 populate the dcache with a disconneced dentry while the inode
1711 number is in the process of being reallocated. This causes an
1712 assertion failure in iopen since the inode's dentry list contains
1713 both a connected and disconnected dentry.
1717 Description: assertion failure in ldlm_handle2lock()
1718 Details : fix a race between class_handle_unhash() and class_handle2object()
1719 introduced in lustre 1.6.5 by bug 13622.
1724 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1725 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1730 Description: oops in sd_iostats_seq_show()
1731 Details : unloading/reloading the scsi low level driver triggers a kernel
1732 bug when trying to access the sd iostat file.
1737 Description: Kernel panics during QLogic driver reload
1738 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1739 patch, causing memory corruption.
1744 Description: journal_dev option does not work in b1_6
1745 Details : pass mount option during pre-mount.
1747 Severity : enhancement
1749 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1750 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1751 extent information of a file. It can be used to map logical blocks
1752 in a file to physical blocks in the block device.
1756 Description: LDLM soft lockups - improvement
1757 Details : It is be possible to send the lock handle along with each read
1758 or write request because the client is already doing a lock match
1759 itself so there isn't any reason the OST should have to re-do that
1763 Frequency : only on Cray X2
1765 Description: X2 build failures
1766 Details : fix build failures on Cray X2.
1770 Description: xid & resent requests
1771 Details : Initialize RPC XID from clock at startup (randomly if clock is
1774 Severity : enhancement
1776 Description: Add lustre_start utility to start or stop multiple Lustre servers
1781 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1782 Details : In case of memory pressure, list_del() can be called twice on
1783 req->rq_history_list, causing a kernel oops.
1787 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1788 Details : fix stack overflow in the distributed lock manager by defering
1789 export eviction after a failed AST to the elt thread instead of
1790 handling it in the dlm interpret routine.
1794 Description: Convert some comments to new format.
1795 Details : Update documenting comments to match doxygen conventions.
1799 Description: Grammar fixes.
1800 Details : A couple of trivial spelling fixes.
1804 Description: OSD_COUNTERS-mandatory
1805 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1806 to simplify the code.
1810 Description: simplify cmm_device freeing logic.
1811 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1815 Description: Add lockdep support to dt_object_operations locking interface.
1816 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1817 parameter indicating lock ordering. Update mdd code to use new
1822 Description: Introduce failloc constants for lockless IO tests.
1823 Details : Add two new failloc constants to test lockless IO. Only one of
1824 them in implemented---another is checked in yet to be landed
1829 Description: Add lockdep support for inode mutex.
1830 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1831 in the situations where Lustre has to lock more than one inode
1836 Description: Add optional invariants checking support.
1837 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1838 configure switch. This macro is to be used for consistency and
1839 sanity checks that are too expensive to be left in `production'
1844 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1845 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1850 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1851 Details : Use spin_lock_nested() in (the only) situation where more than
1852 one ldlm_lock is locked simultaneously. Also, fix possible
1853 dead-lock in ldlm_lock_change_resource() by enforcing particular
1858 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1859 Details : Instead of specifying each ldlm_lock call-back through separate
1860 parameter, wrap them into struct ldlm_callback_suite.
1864 Description: Kill join_lru obd method and its callers.
1865 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1866 in memory---a requirement implemented through ->o_join_lru() obd
1867 method in HEAD. Remove this method and its users.
1871 Description: Add asynchronous ldlm ENQUEUE completion handler.
1872 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1873 case that stock ldlm_completion_ast() cannot handle as it waits
1874 until lock is granted. Introduce new ldlm_completion_ast_async()
1875 for this. Also comment ldlm_completion_ast().
1879 Description: ldlm_error <-> errno conversion.
1880 Details : Add functions to map (rather arbitrary) between LDLM error codes
1881 and standard errno values. CLIO needs this to prevent LDLM specific
1882 constants from escaping ldlm and osc.
1886 Description: Kill unused ldlm_handle2lock_ns() function.
1887 Details : Kill unused ldlm_handle2lock_ns() function.
1891 Description: Add lu_ref support to ldlm_lock
1892 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1893 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1894 and ldlm_resource. LDLM interface has to be changed, because code that
1895 releases a reference on a lock, has to "know" what reference this is.
1896 In the most frequent case
1898 lock = ldlm_handle2lock(handle);
1900 LDLM_LOCK_PUT(lock);
1902 no changes are required. When any other reference (received _not_ from
1903 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1904 instead of LDLM_LOCK_PUT().
1906 Arguably, changes are pervasive, and interface requires some discipline
1907 for proper use. On the other hand, it was very instrumental in finding
1908 a few leaked lock references.
1912 Description: Add ldlm_lock_addref_try().
1913 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1914 attempts to addref a lock that might be being canceled
1919 Description: Add ldlm_weigh_callback().
1920 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1921 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1922 hard-coded `number of pages' logic.
1926 Description: Add lockdep annotations to llog code.
1927 Details : Use appropriately tagged _nested() locking calls in the places
1928 where llog takes more than one ->lgh_lock lock.
1932 Description: Add loi_kms_set().
1933 Details : Wrap kms updates into a helper function.
1937 Description: Constify instances of struct lsm_operations.
1938 Details : Constify instances of struct lsm_operations.
1942 Description: lu_conf support.
1943 Details : On a server, a file system object is uniquely identified
1944 by a fid, which is sufficient to locate and load all object
1945 state (inode). On a client, on the other hand, more data are
1946 necessary instantiate an object. Change lu_object_find() and
1947 friends to take additional `lu_conf' argument describing object.
1948 Typically this includes layout information.
1952 Description: lu_context fixes.
1953 Details : Introduce new lu_context functions that are needed on the client
1954 side, where some system threads (ptlrpcd) are shared by multiple
1955 modules, and so cannot be stopped during module shutdown.
1959 Description: Add start and stop methods to lu_device_type_operations.
1960 Details : Introduce two new methods in lu_device_type_operations, that are
1961 invoked when first instance of a given type is created and last one
1962 is destroyed respectively. This is need by CLIO.
1966 Description: Add lu_ref support to struct lu_device.
1967 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1968 track leaked references.
1972 Description: Introduce lu_kmem_descr.
1973 Details : lu_kmem_descr and its companion interface allow to create
1974 and destroy a number of kmem caches at once.
1978 Description: Fix lu_object finalization race.
1979 Details : Fix a race between lu_object_find() finding an object and its
1980 concurrent finalization. This race is (most likely) not possible
1981 on the server, but might happen on the client.
1985 Description: Introduce lu_ref interface.
1986 Details : lu_ref is a debugging module allowing to track references to
1987 a given object. It is quite cpu expensive, and has to be
1988 explicitly enabled with --enable-lu_ref. See usage description
1993 Description: Factor lu_site procfs stats into a separate function.
1994 Details : Separate lu_site stats printing code into a separate function
1995 that can be reused on a client.
1999 Description: Constify instances of struct {lu,dt,md}_device_operations.
2000 Details : Constify instances of struct {lu,dt,md}_device_operations.
2004 Description: Introduce struct md_site and move meta-data specific parts of
2005 struct lu_site here.
2006 Details : Move md-specific fields out of struct lu_site into special struct
2007 md_site, so that lu_site can be used on a client.
2011 Description: Kill mdd_lov_destroy().
2012 Details : Remove unused mdd code.
2016 Description: Add st_block checking to multistat.c.
2017 Details : Add st_block checking to multistat.c.
2021 Description: Add lu_ref support to struct obd_device.
2022 Details : Add lu_ref tracking to obd_device.
2026 Description: Kill obd_set_fail_loc().
2027 Details : Remove unused code.
2031 Description: Add special type for ptlrpc_request interpret functions.
2032 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2033 there. Actual usage will be in CLIO.
2037 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2038 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2039 doesn't work with lockdep.
2043 Description: Add rwv.c test program.
2044 Details : New testing program exercising readv(2) and writev(2) (Qian).
2048 Description: Add sendfile.c test program.
2049 Details : New testing program exercising sendfile(2) (Jay).
2053 Description: Ratelimit a message that can be very frequent.
2054 Details : Ratelimit a memory allocation failure message that can
2059 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2060 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2064 Description: Liblustre build fixes.
2065 Details : Liblustre build fixes.
2069 Description: libcfs: add cfs_{need,cond}_resched() interface.
2070 Details : libcfs: add cfs_{need,cond}_resched() definition and
2071 implementations for Linux, NT, and liblustre.
2073 Severity : enhancement
2075 Description: More exported tunables for mballoc
2076 Details : Add support for tunable preallocation window and new tunables for
2077 large/small requests
2081 Description: Detect corruption of block bitmap and checking for preallocations
2082 Details : Checks validity of on-disk block bitmap. Also it does better
2083 checking of number of applied preallocations. When corruption is
2084 found, it turns filesystem readonly to prevent further corruptions.
2088 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2089 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2090 window to 0. If it does hit cache miss, and read-pattern is still
2091 stride-io mode, does not reset the stride window, but also does not
2092 increase the stride window length in this case.
2096 Frequency : only for big-endian servers
2097 Description: Check if system is big-endian while mounting fs with extents feature
2098 Details : Mounting a filesystem with extents feature will fail on big-endian
2099 systems since ext3-based ldiskfs is not supported on big-endian
2100 systems. This can be over-riden with "bigendian_extents" mount option.
2102 Severity : enhancement
2104 Description: The root squash functionality
2105 Details : A security feature, which is to prevent users from being able
2106 to mount lustre on their desktop, run as root, and delete
2107 all of the files in the filesystem. The goal is accomplished by
2108 remapping user id (UID) and group id (GID) of the root user to
2109 a UID and GID specified by the system administartor via Lustre
2110 configuration management server (MGS). The functionality also
2111 allows to specify sets of clients for which the remapping does
2116 Description: Excessive recovery window
2117 Details : With AT enabled, the recovery window can be excessively long (6000+
2118 seconds). To address this problem, we no longer use
2119 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2120 timeout no longer depends on the service time, it is set to
2121 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2122 time via pb_service_time.
2126 Description: Watchdog triggered on MDS failover
2127 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2128 the OSTs know that the MDS "UUID" can be reused for the same export
2129 from a different NID, so we do not need to wait for the export to be
2133 Frequency : rare, only if using MMP with Linux RAID
2135 Description: MMP doesn't work with Linux RAID
2136 Details : While using HA for Lustre servers with Linux RAID, it is possible
2137 that MMP will not detect multiple mounts. To make this work we
2138 need to unplug the device queue in RAID when the MMP block is being
2139 written. Also while reading the MMP block, we should read it from
2140 disk and not the cached one.
2142 Severity : enhancement
2144 Description: open file using fid
2145 Details : A file can be opened using just its fid, like
2146 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2149 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2152 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2153 Details : add proper locking for creation of jbd_4k slab cache
2157 Description: MMP check in ext3_remount() fails without displaying any error
2158 Details : When multiple mount protection fails during remount, proper error
2161 Severity : enhancement
2163 Description: Allow stripe size to be up to 4G-64k
2164 Details : Fix math logic to allow large stripe sizes.
2168 Description: add check for >8TB ldiskfs filesystems
2169 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2170 Don't allow >8TB ldiskfs filesystems to be mounted without
2171 force_over_8tb mount option
2176 Descriptoin: OSS complains "too many blocks" and crashes
2177 Details : In filter_commitrw_write():
2178 1. lquota_chkquota();
2179 2. obdo_from_inode();
2180 3. lquota_pending_commit();
2181 "2." has changed uid and gid in this case, then caused
2182 lquota_chkquota() and lquota_pending_commit() to use
2183 different uid and gid. That is the root of the bug.
2185 --------------------------------------------------------------------------------
2187 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2189 * Support for kernels:
2190 2.6.5-7.283 (SLES 9),
2191 2.6.9-55.EL (RHEL 4),
2192 2.6.16.46-0.14 (SLES 10),
2193 2.6.18.8 vanilla (kernel.org)
2194 * Client support for unpatched kernels:
2195 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2196 2.6.16 - 2.6.22 vanilla (kernel.org)
2197 * Due to recently discovered recovery problems, we do not recommend
2198 using patchless RHEL 4 clients with this or any earlier release.
2199 * Recommended e2fsprogs version: 1.39.cfs8
2200 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2201 * Starting with this release, the ldiskfs backing filesystem required
2202 by Lustre is now in its own package, lustre-ldiskfs. This package
2203 should be installed. It is versioned separately from Lustre and
2204 may be released separately in future.
2208 Description: block reactivating mgc import until all deactivates complete
2209 Details : Fix race when failing back MDT/MGS to itself (testing)
2212 Frequency : at statup only
2214 Description: mds_lov_synchronize race leads to various problems
2215 Details : simultaneous MDT->OST connections at startup can cause the
2216 sync to abort, leaving the OSC in a bad state.
2218 Severity : enhancement
2220 Description: add optional extra BUILD_VERSION info
2221 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2222 to override the lustre version.
2225 Frequency : 2.6.18 servers only
2227 Description: ll_kern_mount() doesn't release the module reference
2228 Details : The ldiskfs module reference count never drops down to 0
2229 because ll_kern_mount() doesn't release the module reference.
2234 Description: server LBUG when using old ost_num_threads parameter
2235 Details : Accept the old ost_num_threads parameter but warn that it
2236 is deprecated, and fix an off-by-one error that caused an LBUG.
2241 Description: Transient SCSI error results in persistent IO issue
2242 Details : iobuf->dr_error is not reinitialized to 0 between two
2246 Frequency : sometimes when underlying device returns I/O errors
2248 Description: OSTs not going read-only during write failures
2249 Details : OSTs are not remounted read-only when the journal commit threads
2250 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2251 instead of the ext3 wrappers.
2255 Description: poor connect scaling with increasing client count
2256 Details : Don't run filter_grant_sanity_check for more than 100 exports
2257 to improve scaling for large numbers of clients.
2260 Frequency : SLES10 only
2262 Description: sanity-quota.sh quotacheck failed: rc = -22
2263 Details : Quotas cannot be enabled on SLES10.
2266 Frequency : liblustre clients only
2268 Description: getdirentries does not give error when run on compute nodes
2269 Details : getdirentries does not fail when the size specified as an argument
2270 is too small to contain at least one entry
2272 Severity : enhancement
2274 Description: Add LNET router traceability for debug purposes
2275 Details : If a checksum failure occurs with a router as part of the
2276 IO path, the NID of the last router that forwarded the bulk data
2277 is printed so it can be identified.
2282 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2283 Details : Due to a race condition, liblustre clients were occasionally
2284 evicted incorrectly.
2286 Severity : enhancement
2288 Description: lfs setstripe use optional parameters instead of postional
2291 Severity : enhancement
2293 Description: Nanosecond timestamp support for ldiskfs
2294 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2295 resolution timestamps. There is not yet support for this at
2296 the Lustre filesystem level.
2299 Frequency : during server recovery
2301 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2302 Details : request with rq_no_resend flag not awake l_wait_event if they get
2306 Frequency : nfs export on patchless client
2308 Description: connectathon hang when test nfs export over patchless client
2309 Details : Disconnected dentry cannot be found with lookup, so we do not need
2310 to unhash it or make it invalid
2313 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2314 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2315 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2320 Description: open req refcounting wrong on reconnect
2321 Details : If reconnect happened between getting open reply from server and
2322 call to mdc_set_replay_data in ll_file_open, we will schedule
2323 replay for unreferenced request that we are about to free.
2324 Subsequent close will crash in variety of ways.
2325 Check that request is still eligible for replay in
2326 mdc_set_replay_data().
2331 Description: disable writes to filesystem when reading health_check file
2332 Details : the default for reading the health_check proc file has changed
2333 to NOT do a journal transaction and write to disk, because this
2334 can cause reads of the /proc file to hang and block HA state
2335 checking on a healthy but otherwise heavily loaded system. It
2336 is possible to return to the previous behaviour during configure
2337 with --enable-health-write.
2339 Severity : enhancement
2341 Description: 64-bit inode version
2342 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2343 to the inode. This will be required for version-based recovery.
2348 Description: MDS fails to start if a duplicate client export is detected
2349 Details : in some rare cases it was possible for a client to connect to
2350 an MDS multiple times. Upon recovery the MDS would detect this
2351 and fail during startup. Handle this more gracefully.
2353 Severity : enhancement
2355 Description: Add -o localflock option to simulate old noflock
2357 Details : This will achieve local-only flock/fcntl locks
2363 Description: log_commit_thread vs filter_destroy race leads to crash
2364 Details : Take import reference before releasing llog record semaphore
2369 Description: Wrong request locking in request set processing
2370 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2371 imp_delayed_list, in this place should be used imp_lock.
2374 Frequency : when reconnection
2376 Description: Grant Leak when osc reconnect to OST
2377 Details : When osc reconnect ost, OST(filter) should check whether it
2378 should grant more space to client by comparing fed_grant and
2379 cl_avail_grant, and return the granted space to client instead
2380 of "new granted" space, because client will call osc_init_grant
2381 to update the client grant space info.
2384 Frequency : when client reconnect to OST
2386 Description: Grant Leak when osc do resend and replay bulk write
2387 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2388 bulk write request, because the grant info while be sync between
2389 OSC and OST when reconnect, and we should ignore the grant info
2390 these of resend/replay write req.
2395 Description: Grant space more than avaiable left space sometimes.
2396 Details : When then OST is about to be full, if two bulk writing from
2397 different clients came to OST. Accord the avaliable space of the
2398 OST, the first req should be permitted, and the second one
2399 should be denied by ENOSPC. But if the seconde arrived before
2400 the first one is commited. The OST might wrongly permit second
2401 writing, which will cause grant space > avaiable space.
2404 Frequency : when client is evicted
2406 Description: Grant might be wrongly erased when osc is evicted by OST
2407 Details : when the import is evicted by server, it will fork another
2408 thread ptlrpc_invalidate_import_thread to invalidate the
2409 import, where the grant will be set to 0. While the original
2410 thread will update the grant it got when connecting. So if
2411 the former happened latter, the grant will be wrongly errased
2412 because of this race.
2417 Description: Checking Stale with correct fid
2418 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2419 to check whether it is stale, because sometimes, we want the
2420 enqueue happened anyway, and op_data.fid2 will not be initialized.
2422 Severity : enhancement
2424 Description: update patchless client
2425 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2428 Frequency : only with 2.4 kernel
2430 Description: random memory corruption
2431 Details : size of struct ll_inode_info is to big for union inode.u and this
2432 can be cause of random memory corruption.
2437 Description: Memory leak in recovery
2438 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2439 It should also check obd_fail before fsfilt_start, otherwise if
2440 fsfilt_start return -EROFS,(failover mds during mds recovery).
2441 then the req will return with repmsg->transno = 0 and rc = EROFS.
2442 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2443 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2444 be freed no matter whether fsfilt_commit success or not.
2447 Frequency : only with huge count clients
2449 Description: Prevents from taking the superblock lock in llap_from_page for
2451 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2452 for llap_from_page call in ll_removepage prevents from taking the
2453 superblock lock for a soon died page.
2458 Description: Not check open intent error before release open handle
2459 Details : in some rare cases, the open intent error is not checked before
2460 release open handle, which may cause
2461 ASSERTION(open_req->rq_transno != 0), because it tries to release
2462 the failed open handle.
2467 Description: Set cat log bitmap only after create log success.
2468 Details : in some rare cases, the cat log bitmap is set too early. and it
2469 should be set only after create log success.
2473 Description: Accessing a block bevice can re-enable I/O when Lustre is
2474 tearing down a device.
2475 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2480 Description: service threads may hog cpus when there are a lot of requests
2481 Details : Insert cond_resched to give other threads a chance to use some CPU
2486 Description: the cat log was not initialized in recovery
2487 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2488 unlink log on mds will not be initialized until mds post
2489 recovery. And also in mds post recovery, the unlink log will
2490 initialization will be done asynchronausly, so there will be race
2491 between add unlink log and unlink log initialization.
2495 Description: brw_stats were being printed incorrectly
2496 Details : brw_stats were being printed as log2 but all of them were not
2497 recorded as log2. Also remove some code duplication arising from
2498 filter_tally_{read,write}.
2502 Frequency : rare, only in recovery.
2503 Description: ASSERTION(req->rq_type != LI_POISON) failed
2504 Details : imp_lock should be held while iterating over imp_sending_list for
2505 prevent destroy request after get timeout in ptlrpc_queue_wait.
2509 Description: replay-single.sh test 52 fails
2510 Details : A lock's skiplist need to be cleanup when it being unlinked
2511 from its resource list.
2515 Description: Short directio read returns full requested size rather than
2517 Details : Direct I/O operations should return actual amount of bytes
2518 transferred rather than requested size.
2520 Severity : enhancement
2522 Description: metadata RPC reduction (e.g. for rm performance)
2523 Details : decrease the amount of synchronous RPC between clients and servers
2524 by canceling conflicing lock before the operation on the client
2525 and packing thier handles into the main operation RPC to server.
2527 Severity : enhancement
2529 Description: Async OSC create to avoid the blocking unnecessarily.
2530 Details : If a OST has no remain object, system will block on the creating
2531 when need to create a new object on this OST. Now, ways use
2532 pre-created objects when available, instead of blocking on an
2533 empty osc while others are not empty. If we must block, we block
2534 for the shortest possible period of time.
2538 Description: improve handling recoverable errors
2539 Details : if request processig with error which can be recoverable on server
2540 request should be resend, otherwise page released from cache and
2543 Severity : enhancement
2545 Description: refine locking for avoid write wrong info into lov_objid file
2546 Details : fix possible races with add new target and write/update data in
2549 --------------------------------------------------------------------------------
2551 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2556 Frequency : on some architectures
2558 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2559 Details : Uninitialized flags sometimes cause configuration commands to
2563 Frequency : patchless clients only
2565 Description: missing __iget() symbol export
2566 Details : The __iget() symbol export is missing. To avoid the need for
2567 this on patchless clients the deathrow inode reaper is turned
2568 off, and we depend on the VM to clean up old inodes. This
2569 dependency was during via the fix for bug 12181.
2571 --------------------------------------------------------------------------------
2573 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2575 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2576 INTEROPERATE with older versions automatically. In many cases a
2577 special upgrade step is needed. Please read the
2578 user documentation before upgrading any part of a live system.
2579 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2580 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2581 * WARNING: Lustre configuration and startup changes are required with
2582 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2586 Severity : enhancement
2588 Description: MountConf
2589 Details : Lustre configuration is now managed via mkfs and mount
2590 commands instead of lmc and lconf. New obd types (MGS, MGC)
2591 are added for dynamic configuration management. See
2592 https://mail.clusterfs.com/wikis/lustre/MountConf for
2595 Severity : enhancement
2597 Description: dynamic OST addition
2598 Details : OSTs can now be added to a live filesystem
2600 Severity : enhancement
2602 Description: startup order invariance
2603 Details : MDTs and OSTs can be started in any order. Clients only
2604 require the MDT to complete startup.
2606 Severity : enhancement
2608 Description: parallel, asynchronous orphan cleanup
2609 Details : orphan cleanup is now performed in separate threads for each
2610 OST, allowing parallel non-blocking operation.
2612 Severity : enhancement
2614 Description: optimized stripe assignment
2615 Details : stripe assignments are now made based on ost space available,
2616 ost previous usage, and OSS previous usage, in order to try
2617 to optimize storage space and networking resources.
2619 Severity : enhancement
2621 Description: Permanently set tunables
2622 Details : All writable /proc/fs/lustre tunables can now be permanently
2623 set on a per-server basis, at mkfs time or on a live system.
2625 Severity : enhancement
2627 Description: Lustre message v2
2628 Details : Add lustre message format v2.
2630 Severity : enhancement
2632 Description: client OST exclusion list
2633 Details : Clients can be started with a list of OSTs that should be
2634 declared "inactive" for known non-responsive OSTs.
2638 Description: SPEC SFS validation failure on NFS v2 over lustre.
2639 Details : Changes the blocksize for regular files to be 2x RPC size,
2640 and not depend on stripe size.
2642 Severity : enhancement
2644 Description: Multiple MD RPCs in flight.
2645 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2646 To avoid overly-overloading MDS, introduce a limit on number of
2647 MDS RPCs in flight for a single client and add /proc controls
2648 to adjust this limit.
2650 Severity : enhancement
2652 Description: client read/write statistics
2653 Details : Add client read/write call usage stats for performance
2654 analysis of user processes.
2655 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2656 file access. extents_stats shows chunk size distribution.
2657 extents_stats_per_process show chunk size distribution per
2660 Severity : enhancement
2662 Description: mds statistics
2663 Details : Add detailed mds operations statistics in
2664 /proc/fs/lustre/mds/*/stats.
2668 Description: Failure of copying files with lustre special EAs.
2669 Details : Client side always return success for setxattr call for lustre
2670 special xattr (currently only "trusted.lov").
2675 Description: Refcount LNET uuids
2676 Details : The global LNET uuid list grew linearly with every startup;
2677 refcount repeated list entries instead of always adding to
2680 Severity : enhancement
2682 Description: Dynamic service threads
2683 Details : Within a small range, start extra service threads
2684 automatically when the request queue builds up.
2687 Frequency : mixed-endian client/server environments
2689 Description: mixed-endian crashes
2690 Details : The new msg_v2 system had some failures in mixed-endian
2693 Severity : enhancement
2695 Description: Easy OST removal
2696 Details : OSTs can be permanently deactivated with e.g. 'lctl
2697 conf_param lustre-OST0001.osc.active=0'
2699 Severity : enhancement
2701 Description: MGS proc entries
2702 Details : Added basic proc entries for the MGS showing what filesystems
2705 Severity : enhancement
2707 Description: provide MGS failover
2708 Details : Added config lock reacquisition after MGS server failover.
2710 Severity : enhancement
2712 Description: add Linux 2.4 support
2713 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2717 Description: a large application tries to do I/O to the same resource and dies
2718 in the middle of it.
2719 Details : Check the req->rq_arrival time after the call to
2720 ost_brw_lock_get(), but before we do anything about
2721 processing it & sending the BULK transfer request. This
2722 should help move old stale pending locks off the queue as
2723 quickly as obd_timeout.
2726 Frequency : when an incorrect nid is specified during startup
2728 Description: ptlrpc connect to non-existant node causes kernel crash
2729 Details : LNET can't be re-entered from an event callback, which
2730 happened when we expire a message after the export has been
2731 cleaned up. Instead, hand the zombie cleanup off to another
2734 Severity : enhancement
2736 Description: plain/inodebits lock performance improvement
2737 Details : Grouping plain/inodebits in granted list by their request modes
2738 and bits policy, thus improving the performance of search through
2742 Frequency : only if OST filesystem is corrupted
2744 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2745 Details : for a short time RPCs with bulk IO are in the replay list,
2746 but replay of bulk IOs is unimplemented. If the OST filesystem
2747 is corrupted due to disk cache incoherency and then replay is
2748 started it is possible to trip an assertion. Avoid putting
2749 committed RPCs into the replay list at all to avoid this issue.
2752 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2754 Description: System hang on startup
2755 Details : This bug allowed the liblustre (e.g. catamount) client to
2756 return to the app before handling all startup RPCs. This
2757 could leave the node unresponsive to lustre network traffic
2758 and manifested as a server ptllnd timeout.
2760 Severity : enhancement
2762 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2763 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2764 boolean to control whether to print peer debug info when a
2765 client's RPC times out.
2768 Frequency : only for kernels with patches from Lustre below 1.4.3
2770 Description: Remove old rdonly API
2771 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2774 Frequency : only for devices with external journals
2776 Description: Set external device read-only also
2777 Details : During a commanded failover stop, we set the disk device
2778 read-only while the server shuts down. We now also set any
2779 external journal device read-only at the same time.
2782 Frequency : when upgrading from 1.4 while trying to change parameters
2784 Description: The wrong (new) MDC name was used when setting parameters for
2785 upgraded MDT's. Also allows changing of OSC (and MDC)
2786 parameters if --writeconf is specified at tunefs upgrade time.
2789 Frequency : when setting specific ost indicies
2791 Description: QOS code breaks on skipped indicies
2792 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2793 created with --index need not be sequential.
2795 Severity : enhancement
2797 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2798 Details : The uninit_groups feature works in conjunction with the kernel
2799 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2800 up the pass1 processing of e2fsck. This is a read-only feature
2801 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2802 mount filesystems that have had this feature enabled.
2804 Severity : enhancement
2806 Description: Improve multi-block allocation algorithm to avoid fragmentation
2807 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2808 allocation locality and avoid filesystem fragmentation.
2810 ------------------------------------------------------------------------------
2812 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2814 * Support for kernels:
2815 2.4.21-47.0.1.EL (RHEL 3)
2816 2.6.5-7.283 (SLES 9)
2817 2.6.9-42.0.10.EL (RHEL 4)
2818 2.6.12.6 vanilla (kernel.org)
2819 2.6.16.27-0.9 (SLES 10)
2820 * Recommended e2fsprogs version: 1.39.cfs5
2822 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2826 Frequency : occasional, depends on client load and configuration
2827 Bugzilla : 12181, 12203
2828 Description: data loss for recently-modified files
2830 Details : In some cases it is possible that recently written or created
2831 files may not be written to disk in a timely manner (this should
2832 normally be within 30s unless client IO load is very high).
2833 The problem appears as zero-length files or files that are a
2834 multiple of 1MB in size after a client crash or client eviction
2835 that are missing data at the end of the file.
2837 This problem is more likely to be hit on clients where files are
2838 repeatedly created and unlinked in the same directory, clients
2839 have a large amount of RAM, have many CPUs, the filesystem has
2840 many OSTs, the clients are rebooted frequently, and/or the files
2841 are not accessed by other nodes after being written.
2843 The presence of the problem can be detected by looking at
2844 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2845 smaller than the second (nr_unused) then dirty files will not
2846 be flushed automatically to disk. "sync; sleep 10" should be
2847 run several times on the node before unmounting it to update
2848 Lustre (this is also safe to run on nodes without this problem).
2850 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2851 that can cause this same problem, so customers using that kernel
2852 also need to update the kernel in addition to Lustre. In order
2853 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2855 It is normal that files written just before a client crash (less
2856 than 30s) may not yet have been flushed to disk, even for local
2860 Frequency : frequent on thin XT3 nodes
2862 Description: UUID collision on thin XT3 Linux nodes
2863 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2864 randomly, since we relied on an insufficiently-seeded PRNG.
2869 Description: OSS hangs after "All ost request buffers busy"
2870 Details : A deadlock between quota and journal operations caused OSS
2871 hangs after printing "All ost request buffers busy."
2874 Frequency : always on liblustre builds
2876 Description: Cleanup compiler warnings on liblustre
2879 Frequency : always on liblustre builds on XT3
2881 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2885 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2886 > 32000 subdirectories
2887 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2888 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2889 subdirectory count crosses 32000. This will aid e2fsck to
2890 correctly handle more than 32000 subdirectories.
2893 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2895 Description: System hang on startup
2896 Details : This bug allowed the liblustre (e.g. catamount) client to
2897 return to the app before handling all startup RPCs. This
2898 could leave the node unresponsive to lustre network traffic
2899 and manifested as a server ptllnd timeout.
2901 Severity : enhancement
2903 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2904 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2905 boolean to control whether to print peer debug info when a
2906 client's RPC times out.
2911 Description: make O_SYNC working on 2.6 kernels
2912 Details : 2.6 kernels use different method for mark pages for write,
2913 so need add a code to lustre for O_SYNC work.
2918 Description: Failure to close file and release space on NFS
2919 Details : Put inode details into lock acquired in ll_intent_file_open.
2920 Use mdc_intent_lock in ll_intent_open to properly
2921 detect all kind of errors unhandled by mdc_enqueue.
2926 Description: proc file read during shutdown sometimes raced obd removal,
2928 Details : Add lock to prevent obd access after proc file removal.
2931 Frequency : Only for files larger than 4GB on 32-bit clients.
2933 Description: improperly doing page alignment of locks
2934 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2935 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2940 Description: RPCs being resent when they shouldn't be
2941 Details : Some RPCs that should not be resent are being resent. This
2942 can cause inconsistencies in the RPC state machine. Do not
2943 resend such requests.
2946 Frequency : rare, only with NFS export
2948 Description: Crash on NFS re-export node
2949 Details : under very unusual load conditions an assertion is hit in
2950 ll_intent_file_open()
2953 Frequency : only if OST filesystem is corrupted
2955 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2956 Details : for a short time RPCs with bulk IO are in the replay list,
2957 but replay of bulk IOs is unimplemented. If the OST filesystem
2958 is corrupted due to disk cache incoherency and then replay is
2959 started it is possible to trip an assertion. Avoid putting
2960 committed RPCs into the replay list at all to avoid this issue.
2965 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2966 Details : Large single O_DIRECT read and write calls can fail to allocate
2967 a sufficiently large buffer to process the request. In case of
2968 allocation failure the allocation is retried with a smaller
2969 buffer and broken into smaller requests.
2971 Severity : enhancement
2973 Description: Add -o localflock option to simulate old noflock behaviour.
2974 Details : This will achieve local-only flock/fcntl locks coherentness.
2979 Description: versioning check is incomplete
2980 Details : Checking the version difference of client vs. server, report
2981 error if the gap is too big.
2986 Description: add support PG_writeback bit
2987 Details : add support for PG_writeback bit for Lustre, for more carefull
2988 work with page cache in 2.6 kernel. This also fix some deadlocks
2989 and remove hack for work O_SYNC with 2.6 kernel.
2991 Severity : enhancement
2993 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2994 Details : The uninit_groups feature works in conjunction with the kernel
2995 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2996 up the pass1 processing of e2fsck. This is a read-only feature
2997 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2998 mount filesystems that have had this feature enabled.
3000 Severity : enhancement
3002 Description: Improve multi-block allocation algorithm to avoid fragmentation
3003 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3004 allocation locality and avoid filesystem fragmentation.
3006 ------------------------------------------------------------------------------
3008 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3010 * Support for kernels:
3011 2.6.9-42.0.3.EL (RHEL 4)
3012 2.6.5-7.276 (SLES 9)
3013 2.4.21-47.0.1.EL (RHEL 3)
3014 2.6.12.6 vanilla (kernel.org)
3015 2.6.16.21-0.8 (SLES10)
3016 * Recommended e2fsprogs version: 1.39.cfs2-0
3018 * The backwards-compatible /proc/sys/portals symlink has been removed
3019 in this release. Before upgrading, please ensure that you change
3020 any configuration scripts or /etc/sysctl.conf files that access
3021 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3022 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3023 in advance of the upgrade on any system running Lustre 1.4.6 or
3024 newer, since /proc/sys/lnet was added in that version.
3025 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3029 Frequency : only when quota is used
3031 Description: avoid scanning export list for quota master
3032 Details : Change the algorithms to avoid scanning export list in order
3033 to improve the efficiency.
3036 Frequency : MDS failover only, very rarely
3038 Description: "went back in time" messages on mds failover
3039 Details : The greatest transno may be lost when the current operation
3040 finishes with an error (transno==0) and the client's last_rcvd
3041 record is over-written. Save the greatest transno in the
3042 mds_last_transno for this case.
3045 Frequency : always for specific kernels and striping counts
3047 Description: client may get "Matching packet too big" without ACL support
3048 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3049 when trying to access files in certain configurations. The
3050 clients should in fact be denied when mounting because they do
3051 not understand ACLs.
3054 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3056 Description: many clients connecting with IO in progress causes connect timeouts
3057 Details : Avoid synchronous journal commits to avoid delays caused by many
3058 clients connecting/disconnecting when bulk IO is in progress.
3059 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3060 OST_IO_PORTAL to avoid delays behind potentially many pending
3064 Frequency : occasionally with multiple writers to a single file
3066 Description: shared writes to file may result in wrong size reported by stat()
3067 Details : Allow growing of kms when extent lock is cancelled
3070 Frequency : always with random mmap IO to multi-striped file
3072 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3073 Details : Only if the hole is at the end of OST object so that kms is too
3074 small. Fix is to increase kms accordingly in ll_nopage.
3077 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3079 Description: writes to a missing object would leak memory on the OST
3080 Details : If there is an inconsistency between the MDS and OST filesystems,
3081 such that the MDS references an object that doesn't exist, writes
3082 to that object will leak memory due to incorrect cleanup in the
3083 error handling path, eventually running out of memory on the OST.
3088 Description: Creating too long symlink causes lustre errors
3089 Details : Check symlink and name lengths before sending requests to MDS.
3092 Frequency : only if flock is enabled (not on by default)
3094 Description: posix locks not released on fd closure on 2.6.9+
3095 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3096 kernels, this caused such locks not to be released on fd close and
3097 then assertions on fs unmount about still used locks.
3100 Frequency : MDS failover only, very rarely
3102 Description: clients may get ASSERTION(granted_lock != NULL)
3103 Details : When request was taking a long time, and a client was resending
3104 a getattr by name lock request. The were multiple lock requests
3105 with the same client lock handle and
3106 mds_getattr_name->fixup_handle_for_resent_request found one of the
3107 lock handles but later failed with ASSERTION(granted_lock != NULL).
3112 Description: handle->h_buffer_credits > 0, assertion failure
3113 Details : h_buffer_credits is zero after truncate, causing assertion
3114 failure. This patch extends the transaction or creates a new
3118 Frequency : NFS re-export or patchless client
3119 Bugzilla : 11179, 10796
3120 Description: Crash on NFS re-export node (__d_move)
3121 Details : We do not want to hash the dentry if we don't have a lock.
3122 But if this dentry is later used in d_move, we'd hit uninitialised
3123 list head d_hash, so we just do this to init d_hash field but
3124 leave dentry unhashed.
3127 Frequency : NFS re-export or patchless client
3129 Description: NFS exports has problem with symbolic link
3130 Details : lustre client didn't properly install dentry when re-exported
3131 to NFS or running patchless client.
3134 Frequency : NFS re-export or patchless client
3136 Description: Various nfs/patchless fixes.
3137 Details : fixes reuse disconected alias for lookup process - this fixes
3138 warning "find_exported_dentry: npd != pd",
3139 fix permission error with open files at nfs.
3140 fix apply umask when do revalidate.
3143 Frequency : occasional
3145 Description: Crash on NFS re-export node
3146 Details : calling clear_page() on the wrong pointer triggered oops in
3147 generic_mapping_read().
3150 Frequency : rarely, using O_DIRECT IO
3152 Description: unaligned directio crashes client with LASSERT
3153 Details : check for unaligned buffers before trying any requests.
3156 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3158 Description: stale data returned from RAID cache
3159 Details : If only a small amount of IO is done to the RAID device before
3160 reading it again it is possible to get stale data from the RAID
3161 cache instead of reading it from disk.
3164 Frequency : always for sles10 kernel
3166 Description: sles10 support
3167 Details : ll_follow_link: compile fixes and using of nd_set_link
3168 under newer kernels.
3171 Frequency : depends on arch, kernel and compiler version, always on sles10
3174 Description: recursive or deep enough symlinks cause stack overflow
3175 Details : getting rid of large stack-allocated variable in
3179 Frequency : depends on hardware
3181 Description: lustre write performance loss in the SLES10 kernel
3182 Details : the performance loss is caused by using of write barriers in the
3183 ext3 code. The SLES10 kernel turns barrier support on by
3184 default. The fix is to undo that change for ldiskfs.
3186 ------------------------------------------------------------------------------
3188 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3190 * Support for kernels:
3191 2.6.9-42.0.3EL (RHEL 4)
3192 2.6.5-7.276 (SLES 9)
3193 2.4.21-47.0.1.EL (RHEL 3)
3194 2.6.12.6 vanilla (kernel.org)
3198 Frequency : quota enabled and large files being deleted
3200 Description: releasing more than 4GB of quota at once hangs OST
3201 Details : If a user deletes more than 4GB of files on a single OST it
3202 will cause the OST to spin in an infinite loop. Release
3203 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3208 Description: statfs data retrieved from /proc may be stale or zero
3209 Details : When reading per-device statfs data from /proc, in the
3210 {kbytes,files}_{total,free,avail} files, it may appear
3211 as zero or be out of date.
3214 Frequency : systems with MD RAID1 external journal devices
3216 Description: lconf's call to blkid is confused by RAID1 journal devices
3217 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3218 of returning all block devices that match the journal UUID.
3221 Frequency : always, for aggregate stripe size over 4GB
3223 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3224 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3225 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3228 Frequency : always if "lfs find" used on a local file/directory
3230 Description: "lfs find" segfaults if used on a local file/directory
3231 Details : The case where a directory component was not specified wasn't
3232 handled correctly. Handle this properly.
3235 Frequency : always on ppc64
3237 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3238 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3239 perform endianness conversion.
3242 Frequency : rarely (truncate to non-zero file size after write under load)
3243 Bugzilla : 10730, 10687
3244 Description: Files padded with zeros to next 4K multiple
3245 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3246 it is possible that files that are truncated to a non-zero size
3247 immediately after being written are filled with zero bytes beyond
3248 the truncated size. No file data is lost.
3250 Severity : enhancement
3252 Description: Allow recovery/failover for liblustre clients.
3253 Details : liblustre clients were unaware of failover configurations until
3256 Severity : enhancement
3258 Description: user file locks should fail when not mounting with flock option
3259 Details : Set up an error-returning stub in ll_file_operations.lock field
3260 to prevent incorrect behaviour when client is mounted without
3261 flock option. Also, set up properly f_op->flock field for
3265 Frequency : always on ia64
3267 Description: "lfs df" loops on printing out MDS statfs information
3268 Details : The obd_ioctl_data was not initialized and in some systems
3269 this caused a failure during the ioctl that did not return
3270 an error. Initialize the struct and return an error on failure.
3273 Frequency : SLES 9 only
3275 Description: Error of copying files with lustre special EAs as root
3276 Details : Client side always return success for setxattr call for lustre
3277 special xattr (currently only "trusted.lov").
3280 Frequency : rarely on clusters with both ia64+i386 clients
3282 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3283 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3284 clients are doing concurrent writes to the same file it is
3285 possible that the ia64 clients may LASSERT because the OST
3286 extent locks are not PAGE_SIZE aligned. Ensure that grown
3287 locks are always aligned on the request boundary.
3290 Frequency : specific use, occasional
3292 Description: Overwriting in use executable truncates on-disk binary image
3293 Details : If one node attempts to overwrite an executable in use by
3294 another node, we now correctly return ETXTBSY instead of
3295 truncating the file.
3297 Severity : enhancement
3299 Description: Async OSC create to avoid the blocking unnecessarily.
3300 Details : If a OST has no remain object, system will block on the creating
3301 when need to create a new object on this OST. Now, ways use
3302 pre-created objects when available, instead of blocking on an
3303 empty osc while others are not empty. If we must block, we block
3304 for the shortest possible period of time.
3309 Description: chmod on Lustre root is propagated to other clients
3310 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3311 invalid by the follow_mount time.
3314 Frequency : liblustre clients only
3316 Description: Race in 'instant cancel' lock handling could lead to such locks
3317 never to be granted in case of SMP MDS
3318 Details : Do not destroy not yet granted but cbpending locks in
3322 Frequency : replay/resend of open
3324 Description: non null lock assetion failure in mds_intent_policy
3325 Details : Trying to replay/resend lockless open requests resulted in
3326 mds_open() returning 0 with no lock. Now it sets a flag if
3327 a lock is going to be returned.
3329 Severity : enhancement
3331 Description: Checksum enhancements
3332 Details : New checksum enhancements allow for resending RPCs that failed
3335 Severity : enhancement
3337 Description: Tunables on number of dirty pages in cacche
3338 Details : Allow to set limit on number of dirty pages cached.
3343 Description: client crash on unmount - lock still has references
3344 Details : In some error handling cases it was possible to leak a lock
3345 reference on a client while accessing a file. This was not
3346 harmful to the client during operation, but would cause the
3347 client to crash when the filesystem is unmounted.
3350 Frequency : specific case, rare
3352 Description: ETXTBSY on mds though file not in use by client
3353 Details : ETXTBSY is no longer incorrectly returned when attempting to
3354 chmod or chown a directory that the user previously tried to
3355 execute or a currently-executing binary.
3358 Frequency : extremely rare except on liblustre-based clients
3360 Description: Lustre space not freed when files are deleted
3361 Details : Clean up open-unlinked files after client eviction. Previously
3362 the unlink was skipped and the files remained as orphans.
3367 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3368 Details : In some cases it was possible to send a blocking callback to
3369 a client doing a glimpse, even though that client didn't get
3370 a lock granted. When the glimpse lock is cancelled on the OST
3371 the freed lock is left on the waiting list and corrupted the list.
3374 Frequency : all core dumps
3376 Description: Broke core dumps to lustre
3377 Details : Negative dentry may be unhashed if parent does not have UPDATE
3378 lock, but some callers, e.g. do_coredump, expect dentry to be
3379 hashed after successful create, hash it in ll_create_it.
3381 ------------------------------------------------------------------------------
3383 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3385 * Support for kernels:
3386 2.6.9-42.0.2.EL (RHEL 4)
3387 2.6.5-7.276 (SLES 9)
3388 2.4.21-40.EL (RHEL 3)
3389 2.6.12.6 vanilla (kernel.org)
3393 Frequency : always on RHEL 3
3395 Description: Number of open files grows over time
3396 Details : The number of open files grows over time, whether or not
3397 Lustre is started. This was due to a filp leak introduced
3398 by one of our kernel patches.
3400 ------------------------------------------------------------------------------
3402 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3404 * Support for kernels:
3405 2.6.9-42.EL (RHEL 4)
3406 2.6.5-7.276 (SLES 9)
3407 2.4.21-40.EL (RHEL 3)
3408 2.6.12.6 vanilla (kernel.org)
3413 Bugzilla : 5719, 9635, 9792, 9684
3414 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3415 Details : If a server is under heavy load and cannot reply to new
3416 connection requests before the client resends the (re)connect,
3417 the connection handling code can behave badly if two service
3418 threads are concurrently handing separate (re)connections from
3419 the same client. Add better locking to the connection handling
3420 code, and ensure that only a single connection will be processed
3421 for a given client UUID, even if the lock is dropped.
3423 Severity : enhancement
3425 Description: add TCP zero-copy support to kernel
3426 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3427 sends if the hardware supports scatter-gather and checksumming.
3428 This allows socklnd to do client-write and server-read more
3429 efficiently and reduce CPU utilization from skbuf copying.
3432 Frequency : only if NFS exporting from client
3434 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3435 Details : If setting attributes on a file created under NFS that had
3436 never been opened it would be possible to oops the client
3437 if the file had no objects.
3440 Frequency : always for liblustre
3442 Description: liblustre client does MDS+OSTs setattr RPC for each write
3443 Details : When doing a write from a liblustre client, the client
3444 incorrectly issued an RPC to the MDS and each OST the file was
3445 striped over in order to update the timestamps. When writing
3446 with small chunks and many clients this could overwhelm the MDS
3447 with RPCs. In all cases it would slow down the write because
3448 these RPCs are unnecessary.
3450 Severity : enhancement
3452 Description: allow number of MDS service threads to be changed at module load
3453 Details : It is now possible to change the number of MDS service threads
3454 running. Adding "options mds mds_num_threads={N}" to the MDS's
3455 /etc/modprobe.conf will set the number of threads for the next
3456 time Lustre is restarted (assuming the "mds" module is also
3457 reloaded at that time). The default number of threads will
3458 stay the same, 32 for most systems.
3463 Description: OST crash if filesystem is unformatted or corrupt
3464 Details : If an OST is started on a device that has never been formatted
3465 or if the filesystem is corrupt and cannot even mount then the
3466 error handling cleanup routines would dereference a NULL pointer.
3471 Description: NULL pointer deref in llap_from_page.
3472 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3473 incorrect) mapping if the page was truncated/reclaimed while it was
3474 searched for. Check for this condition and skip such pages when
3475 doing readahead. Introduce extra check to llap_from_page() to
3476 verify page->mapping->host is non-NULL (so page is not anonymous).
3479 Frequency : Sometimes when using sys_sendfile
3481 Description: "page not covered by a lock" warnings from ll_readpage
3482 Details : sendfile called ll_readpage without right page locks present.
3483 Now we introduced ll_file_sendfile that does necessary locking
3484 around call to generic_file_sendfile() much like we do in
3488 Frequency : with certain MDS communication failures at client mount time
3490 Description: NULL pointer deref after failed client mount
3491 Details : a client connection request may delayed by the network layer
3492 and not be sent until after the PTLRPC layer has timed out the
3493 request. If the client fails the mount immediately it will try
3494 to clean up before the network times out the request. Add a
3495 reference from the request import to the obd device and delay
3496 the cleanup until the network drops the request.
3499 Frequency : occasionally during client (re)connect
3501 Description: assertion failure during client (re)connect
3502 Details : processing a client connection request may be delayed by the
3503 client or server longer than the client connect timeout. This
3504 causes the client to resend the connection request. If the
3505 original connection request is replied in this interval, the
3506 client may trip an assertion failure in ptlrpc_connect_interpret()
3507 which thought it would be the only running connect process.
3510 Frequency : only with obd_echo servers and clients that are rebooted
3512 Description: kernel BUG accessing uninitialized data structure
3513 Details : When running an obd_echo server it did not start the ping_evictor
3514 thread, and when a client was evicted an uninitialized data
3515 structure was accessed. Start the ping_evictor in the RPC
3516 service startup instead of the OBD startup.
3518 Severity : enhancement
3519 Bugzilla : 10193 (patchless)
3520 Description: Remove dependency on various unexported kernel interfaces.
3521 Details : No longer need reparent_to_init, exit_mm, exit_files,
3522 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3525 Frequency : rare (only users of deprecated and unsupported LDAP config)
3527 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3528 Details : LDAP apparently contains 'lustreName' attributes instead of
3529 'name'. A simple remapping of the name is sufficient.
3532 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3534 Description: waiting_locks_callback trips kernel BUG if client is evicted
3535 Details : Running with the dump_on_timeout debug flag turned on makes
3536 it possible that the waiting_locks_callback() can try to dump
3537 the Lustre kernel debug logs from an interrupt handler. Defer
3538 this log dumping to the expired_lock_main() thread.
3540 Severity : enhancement
3542 Description: Support NFS exporting on 2.6 kernels.
3543 Details : Implement non-rawops metadata methods for NFS server to use without
3544 changing NFS server code.
3547 Frequency : very rare (synthetic metadata workload only)
3549 Description: two racing renames might cause an MDS thread to deadlock
3550 Details : Running the "racer" program may cause one MDS thread to rename
3551 a file from being the source of a rename to being the target of
3552 a rename at exactly the same time that another thread is doing
3553 so, and the second thread has already enqueued these locks after
3554 doing a lookup of the target and is trying to relock them in
3555 order. Ensure that we don't try to re-lock the same resource.
3558 Frequency : only very large systems with liblustre clients
3560 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3561 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3562 all OSTs in parallel. This allows the eviction of stale liblustre
3563 clients to proceed much faster than if they were done in series,
3564 and also offers similar improvements for other set_info RPCs.
3569 Description: excessive CPU usage during initial read phase on client
3570 Details : During the initial read phase on a client, it would agressively
3571 retry readahead on the file, consuming too much CPU and impacting
3572 performance (since 1.4.5.8). Improve the readahead algorithm
3573 to avoid this, and also improve some other common cases (read
3574 of small files in particular, where "small" is files smaller than
3575 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3580 Description: MDS crash when receiving packet with unknown intent.
3581 Details : Do not LBUG in unknown intent case, just return -EFAULT
3583 Severity : enhancement
3584 Bugzilla : 9293, 9385
3585 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3586 Details : Do not serialize getattr (non-intent version) and statfs.
3589 Frequency : occasional, when OST network is overloaded/intermittent
3591 Description: client evicted by OST after bulk IO timeout
3592 Details : If a client sends a bulk IO request (read or write) the OST
3593 may evict the client if it is unresposive to its data GET/PUT
3594 request. This is incorrect if the network is overloaded (takes
3595 too long to transfer the RPC data) or dropped the OST GET/PUT
3596 request. There is no need to evict the client at all, since
3597 the pinger and/or lock callbacks will handle this, and the
3598 client can restart the bulk request.
3601 Frequency : Always when mmapping file with no objects
3603 Description: client crashes when mmapping file with no objects
3604 Details : Check that we actually have objects in a file before doing any
3605 operations on objects in ll_vm_open, ll_vm_close and
3611 Description: Request leak when working with deleted CWD
3612 Details : Introduce advanced request refcount tracking for requests
3613 referenced from lustre intent.
3615 Severity : Enhancement
3617 Description: Cache open file handles on client.
3618 Details : MDS now will return special lock along with openhandle, if
3619 requested and client is allowed to hold openhandle, even if unused,
3620 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3621 closing files for every read/write openration.
3623 Severity : Enhancement
3625 Description: Cache open negative dentries on client when possible.
3626 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3627 negative dentries on lock revocation.
3632 Description: Remounting a client read-only wasn't possible with a zconf mount
3633 Details : It wasn't possible to remount a client read-only with llmount.
3635 Severity : enhancement
3636 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3637 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3638 MPICH is included to add Lustre-specific ADIO interfaces.
3639 This is based closely on the UFS ADIO layer and only differs
3640 in file creation, in order to allow the OST striping to be set.
3641 This is user-contributed code and not supported by CFS.
3646 Description: extended inode attributes (immutable, append-only) work improperly
3647 when 2.4 and 2.6 kernels are used on client/server or vice versa
3648 Details : Introduce kernel-independent values for these flags.
3650 Severity : enhancement
3653 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3654 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3655 granularity. Now, Fractional MB granularity is be supported,
3656 this is very useful for low memory system.
3658 Severity : enhancement
3660 Description: Getattr by fid
3661 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3662 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3663 have locks and avoids getattr rpc requests for it on every stat.
3666 Frequency : Always, for filesystems larger than 2TB
3668 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3669 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3670 filesystems larger than 2TB to be mounted with mballoc enabled.
3673 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3675 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3676 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3677 filesystems larger than 2TB is mounted then the kernel will
3678 silently corrupt the start of the filesystem. CONFIG_LBD is
3679 enabled for all CFS-supported kernels, but the possibility of
3680 this happening with a modified kernel config exists.
3682 Severity : enhancement
3684 Description: add client O_DIRECT support for 2.6 kernels
3685 Details : It is now possible to do O_DIRECT reads and writes to files
3686 in the Lustre client mountpoint on 2.6 kernel clients.
3688 Severity : enhancement
3690 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3691 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3692 parallel, not waiting for response from every OST before sending
3693 a rpc to the next OST.
3698 Description: setattr vs write race when updating file timestamps
3699 Details : Client processes that update a file timestamp into the past
3700 right after writing to the file (e.g. tar) it is possible that
3701 the updated file modification time can be reset to the current
3702 time due to a race between processing the setattr and write RPC.
3704 Severity : enhancement
3706 Description: Bring 'lfs find' closer in line with regular Linux find.
3707 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3708 -print0 options and obtains all the needed info through the lustre
3711 Severity : enhancement
3713 Description: support up to 1024 configured devices on one node
3714 Details : change obd_dev array from statically allocated to dynamically
3715 allocated structs as they are first used to reduce memory usage
3720 Description: Flush dirty partially truncated pages during truncate
3721 Details : Immediatelly flush partially truncated pages in filter_setattr,
3722 this way we completely avoid having any pages in page cache on OST
3723 and can retire ugly workarounds during writes to flush such pages.
3728 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3729 Details : protect inode from truncation within vfs_unlink() context
3730 just take a reference before calling vfs_unlink() and release it
3731 when parent's i_sem is free.
3736 Description: last_id value checked outside lock on OST caused LASSERT failure
3737 Details : If there were multiple MDS->OST object precreate requests in
3738 flight, it was possible that the OST's last object id was checked
3739 outside a lock and incorrectly tripped an assertion. Move checks
3740 inside locks, and discard old precreate requests.
3743 Frequency : always, if extents are used on OSTs
3745 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3746 Details : OSTs using the extents format would not zero the high 16 bits of
3747 the index physical block number. This is not a problem for any
3748 OST filesystems smaller than 16TB, and no kernels support ext3
3749 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3750 new/modified files) and can be fixed for existing filesystems
3751 with e2fsprogs-1.39-cfs1.
3756 Description: import connection selection may be incorrect if timer wraps
3757 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3758 import connections to be ignored if the 32-bit jiffies counter
3759 wraps. Use a 64-bit jiffies counter.
3762 Frequency : very large clusters immediately after boot
3764 Description: LNET request buffers exhausted under heavy short-term load
3765 Details : If a large number of client requests are generated on a service
3766 that has previously never seen so many requests it is possible
3767 that the request buffer growth cannot keep up with the spike in
3768 demand. Instead of dropping incoming requests, they are held in
3769 the LND until the RPC service can accept more requests.
3772 Frequency : Sometimes during replay
3774 Description: Assertion failure in ll_local_open after replay.
3775 Details : If replay happened on an open request reply before we were able
3776 to set replay handler, reply will become not swabbed tripping the
3777 assertion in ll_local_open. Now we set the handler right after
3778 recognising of open request
3781 Frequency : very rare
3783 Description: kernel reports "badness in vsnprintf"
3784 Details : Reading from the "recovery_status" /proc file in small chunks
3785 may cause a negative length in lprocfs_obd_rd_recovery_status()
3786 call to vsnprintf() (which is otherwise harmless). Exit early
3787 if there is no more space in the output buffer.
3789 Severity : enhancement
3791 Description: clear OBD RPC statistics by writing to them
3792 Details : It is now possible to clear the OBD RPC statistics by writing
3793 to the "stats" file.
3798 Description: Client mtime is not the same on different clients after utimes
3799 Details : In some cases, the client was using the utimes() syscall on
3800 a file cached on another node. The clients now validate the
3801 ctime from the MDS + OSTs to determine which one is right.
3806 Description: Inability to activate failout mode
3807 Details : lconf script incorrectly assumed that in pythong string's numeric
3808 value is used in comparisons.
3811 Frequency : always with multiple stripes per file
3813 Description: Inefficient object allocation for mutli-stripe files
3814 Details : When selecting which OSTs to stripe files over, for files with
3815 a stripe count that divides evenly into the number of OSTs,
3816 the MDS is always picking the same starting OST for each file.
3817 Return the OST selection heuristic to the original design.
3822 Description: mount failures may take full timeout to return an error
3823 Details : Under some heavy load conditions it is possible that a
3824 failed mount can wait for the full obd_timeout interval,
3825 possibly several minutes, before reporting an error.
3826 Instead return an error as soon as the status is known.
3828 Frequency : quota enabled and large files being deleted
3830 Description: releasing more than 4GB of quota at once hangs OST
3831 Details : If a user deletes more than 4GB of files on a single OST it
3832 will cause the OST to spin in an infinite loop. Release
3833 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3838 Description: statfs data retrieved from /proc may be stale or zero
3839 Details : When reading per-device statfs data from /proc, in the
3840 {kbytes,files}_{total,free,avail} files, it may appear
3841 as zero or be out of date.
3844 Frequency : systems with MD RAID1 external journal devices
3846 Description: lconf's call to blkid is confused by RAID1 journal devices
3847 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3848 of returning all block devices that match the journal UUID.
3851 Frequency : always, for aggregate stripe size over 4GB
3853 Description: assertion fails when trying to use 4GB stripe size
3854 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3855 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3858 Frequency : always on ppc64
3860 Description: the first write on an ext3 filesystem with mballoc got stuck
3861 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3862 perform endianness conversion.
3864 ------------------------------------------------------------------------------
3866 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3868 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3869 INTEROPERATE with older versions automatically. Please read the
3870 user documentation before upgrading any part of a live system.
3871 * WARNING: Lustre networking configuration changes are required with
3872 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3875 * Support for kernels:
3876 2.6.9-22.0.2.EL (RHEL 4)
3877 2.6.5-7.244 (SLES 9)
3878 2.6.12.6 vanilla (kernel.org)
3881 Severity : enhancement
3882 Bugzilla : 7981/8208
3883 Description: Introduced Lustre Networking (LNET)
3884 Details : LNET is new networking infrastructure for Lustre, it includes
3885 a reorganized network configuration mode (see the user
3886 documentation for full details) as well as support for routing
3887 between different network fabrics. Lustre Networking Devices
3888 (LNDs) for the supported network fabrics have also been
3889 created for this new infrastructure.
3891 Severity : enhancement
3892 Description: Introduced Access control lists
3893 Details : clients can set ACLs on files and directories in order to have
3894 more fine-grained permissions than the standard Unix UGO+RWX.
3895 The MDS must be started with the "-o acl" mount option.
3897 Severity : enhancement
3898 Description: Introduced filesystem quotas
3899 Details : Administrators may now establish per-user quotas on the
3902 Severity : enhancement
3904 Description: Configuration change for the XT3
3905 The PTLLND is now used to run Lustre over Portals on the XT3
3906 The configure option(s) --with-cray-portals are no longer used.
3907 Rather --with-portals=<path-to-portals-includes> is used to
3908 enable building on the XT3. In addition to enable XT3 specific
3909 features the option --enable-cray-xt3 must be used.
3914 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3915 Details : A race between allocating a new llcd and re-getting the llcd_lock
3916 allowed another thread to grab newly-allocated llcd.
3918 Severity : enhancement
3920 Description: 2.6 OST async journal commit and locking fix to improve performance
3921 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3922 2.6 kernels are now async as they already were in 2.4 kernels so
3923 that they can commit concurrently with the network bulk transfer.
3924 For block-allocated files the filter allocation semaphore is held
3925 to avoid filesystem fragmentation during allocation. BKL lock
3926 removed for 2.6 xattr operations where it is no longer needed.
3931 Description: lconf incorrectly determined whether two IP networks could talk
3932 Details : In some more complicated routing and multiple-network
3933 configurations, lconf will avoid trying to make a network
3934 connection to a disjoint part of the IP space. It was doing the
3935 math incorrectly for one set of cases.
3940 Description: Fix for potential infinite loop processing records in an llog.
3941 Details : If an llog record is corrupted/zeroed, it is possible to loop
3942 forever in llog_process(). Validate the llog record length
3943 and skip the remainder of the block on error.
3946 Frequency : occasional (liblustre only)
3948 Description: liblustre could not open files whose last component is a symlink
3949 Details : sysio_path_walk() would incorrectly pass the open intent to
3950 intermediate path components.
3953 Frequency : rare (liblustre only with non-standard tuning)
3954 Bugzilla : 7201 (7350)
3955 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3956 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3957 after it has been released, only do this for referenced locks
3959 Severity : enhancement
3961 Description: specifying an (invalid) directory default stripe_size of -1
3962 would reset the directory default striping
3963 Details : stripe_size -1 was used internally to signal directory stripe
3964 removal, now use "all default" to signal dir stripe removal
3965 as a directory striping of "all default" is not useful
3968 Frequency : common for large clusters running liblustre clients
3970 Description: doing an ls when liblustre clients are running is slow
3971 Details : sending a glimpse AST to a liblustre client waits for every AST
3972 to time out, as liblustre clients will not respond. Since they
3973 cannot cache data we refresh the OST lock LVB from disk instead.
3975 Severity : enhancement
3977 Description: doing an ls at the same time as file IO can be slow
3978 Details : enqueue and other "small" requests can be blocked behind many
3979 large IO requests. Create a new OST IO portal for non-IO
3980 requests so they can be processed faster.
3983 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3985 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3986 created file with mode 000
3987 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3988 allow an ftruncate() as a non-root user to a file with mode 000.
3989 Lustre now respects this flag to disable mode checks when
3990 truncating a file owned by the user
3993 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3996 Description: Revoking locks from clients that went dead or catatonic might take
3998 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3999 cancellation of such locks instant on servers without waiting for
4000 any reply from clients. Clients drops these locks when cancel
4001 notification from server is received without replying.
4004 Frequency : liblustre-only, when liblustre client dies or becomes busy
4006 Description: Doing ls on Linux clients can take a long time with active
4008 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4009 granting such locks to it in the first place if possible. Locks
4010 are taken by proxy on the OST during the read or write and
4011 dropped immediately afterward. Add connect flags handling, do
4012 not grant locks to liblustre clients for glimpse ASTs.
4014 Severity : enhancement
4016 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4017 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4018 sequential reads and behaves badly with applications doing random
4019 reads. Improve it by reading ahead at least read region, and
4020 avoiding excessive large RPC for small reads.
4022 Severity : enhancement
4024 Description: Creating more than 1000 files for a single job may cause a load
4025 imbalance on the OSTs if there are also a large number of OSTs.
4026 Details : qos_prep_create() uses an OST index reseed value that is an
4027 even multiple of the number of available OSTs so that if the
4028 reseed happens in the middle of the object allocation it will
4029 still utilize the OSTs as uniformly as possible.
4034 Description: OST or MDS may oops in ping_evictor_main()
4035 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4036 but doesn't restart at beginning of obd_exports_timed list
4039 Severity : enhancement
4041 Description: improve by-nid export eviction on the MDS and OST
4042 Details : allow multiple exports with the same NID to be evicted at one
4043 time without re-searching the exports list.
4046 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4048 Description: MDS may oops in groups_free()
4049 Details : in rare race conditions a newly allocated group_info struct is
4050 freed again, and this can be NULL. The 2.4 compatibility code
4051 for groups_free() checked for a NULL pointer, but 2.6 did not.
4054 Frequency : common for liblustre clients doing little filesystem IO
4055 Bugzilla : 9352, 7313
4056 Description: server may evict liblustre clients accessing contended locks
4057 Details : if a client is granted a lock or receives a completion AST
4058 with a blocking AST already set it would not reply to the AST
4059 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4062 Frequency : lfs setstripe, only systems with more than 160 OSTs
4064 Description: unable to set striping with a starting offset beyond OST 160
4065 Details : llapi_create_file() incorrectly limited the starting stripe
4066 index to the maximum single-file stripe count.
4069 Frequency : LDAP users only
4071 Description: lconf did not handle in-kernel recovery with LDAP properly
4072 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4074 Severity : enhancement
4076 Description: bind OST threads to NUMA nodes to improve performance
4077 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4078 node and do their allocations there to localize memory access
4080 Severity : enhancement
4082 Description: llmount can determine client NID directly from Myrinet (GM)
4083 Details : the client NID code from gmnalnid was moved directly into
4084 llmount, removing the need to use this or specifying the
4085 client NID explicitly when mounting GM clients with zeroconf
4088 Frequency : if client is started with down MDS
4090 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4091 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4092 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4093 ptlrpc_import_delayed_req() because we didn't check if the
4094 request was interrupted, and we also didn't break out of the
4095 event loop if there was no timeout
4100 Description: data loss during non-page-aligned writes to a single file from
4101 both multiple nodes and multiple threads on one node at same time
4102 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4103 inconsistency led to false short-reads, that were cached and later
4104 used by ->prepare_write() to fill in partially written page,
4105 leading to data loss.
4108 Frequency : always, if lconf --abort_recovery used
4110 Description: lconf --abort_recovery fails with 'Operation not supported'
4111 Details : lconf was attempting to abort recovery on the MDT device and not
4114 Severity : enhancement
4116 Description: remove cleanup logs
4117 Details : replace lconf-generated cleanup logs with lustre internal
4118 cleanup routines. Eliminates the need for client-cleanup and
4121 Severity : enhancement
4123 Description: add support for EAs (user and system) on lustre filesystems
4124 Details : it is now possible to store extended attributes in the Lustre
4125 client filesystem, and with the user_xattr mount option it
4126 is possible to allow users to store EAs on their files also
4128 Severity : enhancement
4130 Description: Add possibility (config option) to show minimal available OST free
4132 Details : When compiled with --enable-mindf configure option, statfs(2)
4133 (and so, df) will return least minimal free space available from
4134 all OSTs as amount of free space on FS, instead of summary of
4135 free spaces of all OSTs.
4137 Severity : enhancement
4139 Description: do not expand extent locks acquired on OST-side
4140 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4141 by server: they are not cached anyway.
4144 Frequency : when mmap is used/binaries executed from Lustre
4146 Description: Unmmap pages before throwing them away from read cache.
4147 Details : llap_shrink cache now attempts to unmap pages before discarding
4148 them (if unmapping failed - do not discard). SLES9 kernel has
4149 extra checks that trigger if this unmapping is not done first.
4154 Description: lconf didn't resolve symlinks before checking to see whether a
4155 given mountpoint was already in use
4158 Frequency : when migrating failover services
4159 Bugzilla : 6395, 9514
4160 Description: When migrating a subset of services from a node (e.g. failback
4161 from a failover service node) the remaining services would
4162 time out and evict clients.
4163 Details : lconf --force (implied by --failover) sets the global obd_timeout
4164 to 5 seconds in order to quickly disconnect, but this caused
4165 other RPCs to time out too quickly. Do not change the global
4166 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4168 Severity : enhancement
4169 Frequency : if MDS is started with down OST
4170 Bugzilla : 9439,5706
4171 Description: Allow startup/shutdown of an MDS without depending on the
4172 availability of the OSTs.
4173 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4174 Add appropriate locking and lov-osc refcounts for safe
4175 cleaning. Add osc abort_inflight calls in case the
4176 synchronize never started.
4179 Frequency : occasional (Cray XT3 only)
4181 Description: root not authorized to access files in CRAY_PORTALS environment
4182 Details : The client process capabilities were not honoured on the MDS in
4183 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4184 been accessed by an authorized user then root was able to access
4185 the file on the local client also. The root user capabilities
4186 are now allowed on the MDS, as this environment has secure UID.
4189 Frequency : occasional
4191 Description: ldiskfs "too long searching" message happens too often
4192 Details : A debugging message (otherwise harmless) prints too often on
4193 the OST console. This has been reduced to only happen when
4194 there are fragmentation problems on the filesystem.
4199 Description: Division by zero in statfs when all OSCs are inactive
4200 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4201 lov_statfs divides by value returned by lov_get_stripecnt().
4205 Bugzilla : 9489, 3273
4206 Description: First write from each client to each OST was only 4kB in size,
4207 to initialize client writeback cache, which caused sub-optimal
4208 RPCs and poor layout on disk for the first writen file.
4209 Details : Clients now request an initial cache grant at (re)connect time
4210 and so that they can start streaming writes to the cache right
4211 away and always do full-sized RPCs if there is enough data.
4212 If the OST is rebooted the client also re-establishes its grant
4213 so that client cached writes will be honoured under the grant.
4218 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4219 Details : Now I/O RPCs go to different portal number and (presumably) fast
4220 lock requests (and glimses) and other RPCs get their own service
4221 threads pool that should be able to service those RPCs
4224 Severity : enhancement
4226 Description: Ability to exchange lustre version between client and servers and
4227 issue warnings at client side if client is too old. Also for
4228 liblustre clients there is ability to refuse connection of too old
4230 Details : New 'version' field is added to connect data structure that is
4231 filled with version info. That info is later checked by server and
4235 Frequency : rare, liblustre only.
4236 Bugzilla : 9296, 9581
4237 Description: Two simultaneous writes from liblustre at offset within same page
4238 might proceed at the same time overwriting eachother with stale
4240 Details : I/O lock withing llu_file_prwv was released too early, before data
4241 actually was hitting the wire. Extended lock-holding time until
4242 server acknowledges receiving data.
4245 Frequency : extremely rare. Never observed in practice.
4247 Description: avoid generating lustre_handle cookie of 0.
4248 Details : class_handle_hash() generates handle cookies by incrementing
4249 global counter, and can hit 0 occasionaly (this is unlikely, but
4250 not impossible, because initial value of cookie counter is
4251 selected randonly). Value of 0 is used as a sentinel meaning
4252 "unassigned handle" --- avoid it. Also coalesce two critical
4253 sections in this function into one.
4255 Severity : enhancement
4257 Description: allow liblustre clients to delegate truncate locking to OST
4258 Details : To avoid overhead of locking, liblustre client instructs OST to
4259 take extent lock in ost_punch() on client's behalf. New connection
4260 flag is added to handle backward compatibility.
4262 Severity : enhancement
4263 Bugzilla : 4928, 7341, 9758
4264 Description: allow number of OST service threads to be specified
4265 Details : a module parameter allows the number of OST service threads
4266 to be specified via "options ost ost_num_threads={N}" in the
4267 OSS's /etc/modules.conf or /etc/modprobe.conf.
4271 Bugzilla : 6146, 9635, 9895
4272 Description: servers crash with bad pointer in target_handle_connect()
4273 Details : In rare cases when a client is reconnecting it was possible that
4274 the connection request was the last reference for that export.
4275 We would temporarily drop the export reference and get a new
4276 one, but this may have been the last reference and the export
4277 was just destroyed. Get new reference before dropping old one.
4279 Severity : enhancement
4280 Frequency : if client is started with failover MDS
4282 Description: Allow multiple MDS hostnames in the mount command
4283 Details : Try to read the configuration from all specified MDS
4284 hostnames during a client mount in case the "primary"
4287 Severity : enhancement
4289 Description: Stop sending data to evicted clients as soon as possible.
4290 Details : Check if the client we are about to send or are sending data to
4291 was evicted already. (Check is done every second of waiting,
4292 for which l_wait_event interface was extended to allow checking
4293 of exit condition at specified intervals).
4296 Frequency : rare, normally only when NFS exporting is done from client
4298 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4300 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4303 Frequency : always, if srand() is called before liblustre initialization
4305 Description: Liblustre uses system PRNG disturbing its usage by user application
4306 Details : Introduce internal to lustre fast and high-quality PRNG for
4307 lustre usage and make liblustre and some other places in generic
4308 lustre code to use it.
4310 Severity : enhancement
4311 Bugzilla : 9477, 9557, 9870
4312 Description: Verify that the MDS configuration logs are updated when xml is
4313 Details : Check if the .xml configuration logs are newer than the config
4314 logs stored on the MDS and report an error if this is the case.
4315 Request --write-conf, or allow starting with --old_conf.
4317 Severity : enhancement
4319 Description: Handle symlinks in the path when checking if Lustre is mounted.
4320 Details : Resolve intermediate symlinks when checking if a client has
4321 mounted a filesystem to avoid duplicate client mounts.
4326 Description: lconf can hit an error exception but still return success.
4327 Details : The lconf command catches the Command error exception at the top
4328 level script context and will exit with the associated exit
4329 status, but doesn't ensure that this exit status is non-zero.
4334 Description: failure of ptlrpc thread startup can cause oops
4335 Details : Starting a ptlrpc service thread can fail if there are a large
4336 number of threads or the server memory is very fragmented.
4337 Handle this without oopsing.
4340 Frequency : always, only if liblustre and non-default acceptor port was used
4342 Description: liblustre cannot connect to servers with non-default acceptor port
4343 Details : tcpnal_set_default_params() was not called and was therefore
4344 ignoring the environment varaible TCPNAL_PORT, as well as other
4345 TCPNAL_ environment variables
4350 Description: two objects could be created on the same OST for a single file
4351 Details : If an OST is down, in some cases it was possible to create two
4352 objects on a single OST for a single file. No problems other
4353 than potential performance impact and spurious error messages.
4357 Bugzilla : 5681, 9562
4358 Description: Client may oops in ll_unhash_aliases
4359 Details : Client dcache may become inconsistent in race condition.
4360 In some cases "getcwd" can fail if the current directory is
4366 Description: Inode refcounting problems in NFS export code
4367 Details : link_raw functions used to call d_instantiate without obtaining
4368 extra inode reference first.
4372 Bugzilla : 9942, 9903
4373 Description: Referencing freed requests leading to crash, memleaks with NFS.
4374 Details : We used to require that call to ll_revalidate_it was always
4375 followed by ll_lookup_it. Also with revalidate_special() it is
4376 possible to call ll_revalidate_it() twice for the same dentry
4377 even if first occurence returned success. This fix changes semantic
4378 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4379 reference on a request referred from the intent.
4380 ll_intent_release() then releases such a request.
4383 Frequency : rare, normally benchmark loads only
4385 Description: unlinked inodes were kept in memory on the client
4386 Details : If a client is repeatedly creating and unlinking files it
4387 can accumulate a lot of stale inodes in the inode slab cache.
4388 If there is no other client load running this can cause the
4389 client node to run out of memory. Instead flush old inodes
4390 from client cache that have the same inode number as a new inode.
4393 Frequency : SLES9 2.6.5 kernel and long filenames only
4394 Bugzilla : 9969, 10379
4395 Description: utime reports stale NFS file handle
4396 Details : SLES9 uses out-of-dentry names in some cases, which confused
4397 the lustre dentry revalidation. Change it to always use the
4401 Frequency : rare, unless heavy write-truncate concurrency is continuous
4402 Bugzilla : 4180, 6984, 7171, 9963, 9331
4403 Description: OST becomes very slow and/or deadlocked during object unlink
4404 Details : filter_destroy() was holding onto the parent directory lock
4405 while truncating+unlinking objects. For very large objects this
4406 may block other threads for a long time and slow overall OST
4407 responsiveness. It may also be possible to get a lock ordering
4408 deadlock in this case, or run out of journal credits because of
4409 the combined truncate+unlink. Solution is to do object truncate
4410 first in one transaction without parent lock, and then do the
4411 final unlink in a new transaction with the parent lock. This
4412 reduces the lock hold time dramatically.
4415 Frequency : rare, 2.4 kernels only
4417 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4418 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4419 before unmounting their filesystem in order to avoid deadlock.
4420 The kernel_locked() function in 2.4 kernels only checks whether
4421 the kernel lock is held, not whether it is this process that is
4422 holding it as 2.6 kernels do.
4427 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4428 Details : The client ptlrpc code may be trying to reconnect to a down
4429 server before a previous connection attempt has timed out.
4430 Increase the reconnect interval to be longer than the connection
4431 timeout interval to avoid sending duplicate connections to
4435 Frequency : echo_client brw_test command
4437 Description: fix echo_client to work with OST preallocated code
4438 Details : OST preallocation code (5137) didn't take echo_client IO path
4439 into account: echo_client calls filter methods outside of any
4440 OST thread and, hence, there is no per-thread preallocated
4441 pages and buffers to use. Solution: hijack pga pages for IO. As
4442 a byproduct, this avoids unnecessary data copying.
4446 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4447 Description: Client can oops in mdc_commit_close() after open replay
4448 Details : It was possible for the MDS to return an open request with no
4449 transaction number in mds_finish_transno() if the client was
4450 evicted, but without actually returning an error. Clients
4451 would later try to replay that open and may trip an assertion
4452 Simplify the client close codepath, and always return an error
4453 from the MDS in case the open is not successful.
4456 Frequency : rare, 2.6 OSTs only
4458 Description: OST may deadlock under high load on fragmented files
4459 Details : If there was a heavy load and highly-fragmented OST filesystems
4460 it was possible to have all the OST threads deadlock waiting on
4461 allocation of biovecs, because the biovecs were not released
4462 until the entire RPC IO was completed. Instead, release biovecs
4463 as soon as they are complete to ensure forward IO progress.
4465 Severity : enhancement
4467 Description: Support for specifying external journal device at mount
4468 Details : If an OST or MDS device is formatted with an external journal
4469 device, this device major/minor is stored in the ext3 superblock
4470 and may not be valid for failover. Allow detecting and
4471 specifying the external journal at mount time.
4476 Description: Mounting an MDS with pending unlinked files may cause oops
4477 Details : target_finish_recovery() calls mds_postrecov() which returned
4478 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4479 considers this an error and immediately begins cleaning up the
4480 lov, just after starting the mds_lov process
4482 Severity : enhancement
4484 Description: Implement 'lfs df' to report actual free space on per-OST basis
4485 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4486 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4487 usage of objects; '-h' to report in human readable format.
4489 ------------------------------------------------------------------------------
4491 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4498 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4499 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4500 at startup during orphan unlinking. Instead, skip buddy update
4501 before setup, as it will be regenerated after recovery is complete.
4506 Description: If an OST is inactive, its locks might reference stale inodes.
4507 Details : lov_change_cbdata() must iterate over all namespaces, even if
4508 they are inactive to clear inode references from the lock.
4510 Severity : enhancement
4511 Frequency : occasional, if non-standard max_dirty_mb used
4513 Description: Client will block write RPCs if not enough grant
4514 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4515 then the client will block writes while waiting for another RPC
4516 to complete instead of consuming its dirty limit. With change
4517 we get improved performance when max_dirty_mb is small.
4519 Severity : enhancement
4520 Bugzilla : 3389, 6253
4521 Description: Add support for supplementary groups on the MDS.
4522 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4523 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4524 lookups for user supplementary groups into a cache.
4528 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4529 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4530 for liblustre O_EXCL is enforced by MDS.
4534 Description: readdir never returns NULL in liblustre.
4535 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4536 dentry in struct dirent.
4540 Frequency : liblustre only, depends on application IO pattern
4541 Description: liblustre clients evicted if not contacting servers
4542 Details : Don't put liblustre clients into the ping_evictor list, so
4543 they will not be evicted by the pinger ever.
4545 Severity : enhancement
4547 Description: Add ability to evict clients by NID from MDS.
4548 Details : By echoing "nid:$NID" string into
4549 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4550 $NID would be instantly evicted from this MDS and from all active
4551 OSTs connected to it.
4555 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4556 Details : lookup_it_finish() used to query file size from OSTs that was not
4561 Description: service threads change working directory to that of init
4562 Details : Starting lustre service threads may pin the working directory
4563 of the parent thread, making that filesystem busy. Threads
4564 now change to the working directory of init to avoid this.
4568 Frequency : during shutdown only
4569 Description: shutdown with a failed MDS or OST can cause unmount to hang
4570 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4575 Frequency : 2.6 only
4576 Description: chmod/chown may include an extra supplementary group
4577 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4578 supplementary group and if none is specified this is used.
4581 Bugzilla : 5479 (6816)
4583 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4584 Details : If lookup is in progress on a file that is unlinked we might try
4585 to revalidate the inode and fail in revalidate after lookup is
4586 complete and ll_file_open() enqueues the open again but
4587 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4590 Frequency : always, if lconf --abort_recovery used
4592 Description: lconf --abort_recovery fails with 'Operation not supported'
4593 Details : lconf was attempting to abort recovery on the MDT device and not
4596 ------------------------------------------------------------------------------
4598 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4603 Frequency : rare (only unsupported configurations with a node running as an
4605 Bugzilla : 6514, 5137
4606 Description: Mounting a Lustre file system on a node running as an OST could
4608 Details : OSTs now preallocates memory needed to write out data at
4609 startup, instead of when needed, to avoid having to
4610 allocate memory in possibly low memory situations.
4611 Specifically, if the file system is mounted on on OST,
4612 memory pressure could force it to try to write out data,
4613 which it needed to allocate memory to do. Due to the low
4614 memory, it would be unable to do so and the node would
4615 become unresponsive.
4617 Severity : enhancement
4619 Description: Addition of lconf --service command line option
4620 Details : lconf now accepts a '--service <arg>' option, which is
4621 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4623 Severity : enhancement
4625 Description: Failover mode is now the default for OSTs.
4626 Details : By default, OSTs will now run in failover mode. To return to
4627 the old behaviour, add '--failout' to the lmc line for OSTs.
4629 Severity : enhancement
4631 Description: Health checks are now provided for MDS and OSTs
4632 Details : Additional detailed health check information on MSD and OSTs
4633 is now provided through the procfs health_check value.
4636 Frequency : occasional, depends on IO load
4638 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4639 after numerous create/delete cycles
4640 Details : The ext3 inode allocation policy would not allocate new inodes
4641 very well on the OSTs because there are no new directories
4642 being created. Instead we look for groups with free space if
4643 the parent directories are nearly full.
4648 Description: Network or server problems during mount may cause partially
4649 mounted clients instead of returning an error.
4650 Details : The config llog parsing code may overwrite the error return
4651 code during mount error handling, returning success instead
4657 Description: MDS can fail to allocate large reply buffers
4658 Details : After long uptimes the MDS can fail to allocate large reply
4659 buffers (e.g. zconf client mount config records) due to memory
4660 fragmentation or consumption by the buffer cache. Preallocate
4661 some large reply buffers so that these replies can be sent even
4662 under memory pressure.
4666 Frequency : rare (liblustre)
4667 Description: fsx running with liblustre complained that using truncate() to
4668 extend the file doesn't work. This patch corrects that issue.
4669 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4670 ATTR_SIZE and lsm use in llu_setattr_raw.
4674 Frequency : rare, only 2.6 kernels
4675 Description: Unusual file access patterns on the MDS may result in inode
4676 data being lost in very rare circumstances.
4677 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4678 code in the 2.6 kernel caused the inode and/or EA data not to
4679 be read from disk, causing single-file corruption.
4683 Frequency : rare, only 2.6 filesystems using extents
4684 Description: Heavy concurrent write and delete load may cause data corruption.
4685 Details : It was possible under high-load situations to have an extent
4686 metadata block in the block device cache from a just-unlinked
4687 file overwrite a newly-allocated data block. We now unmap any
4688 metadata buffers that alias just-allocated data blocks.
4692 Frequency : filesystems with default stripe_count larger than 77
4693 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4694 Details : lconf specifies an inode size of 4096 bytes when the default
4695 stripe_count is larger than 77. This conflicts with the default
4696 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4697 this case to avoid pinning too much memory for large EAs.
4699 ------------------------------------------------------------------------------
4701 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4706 Frequency : rare (extremely heavy IO load with hundreds of clients)
4708 Description: Client is evicted, gets IO error writing to file
4709 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4710 caused clients to be evicted for AST timeouts. The fixes in
4711 bug 5192 mean we no longer need to have such short AST timeouts
4712 so ldlm_timeout has been increased.
4715 Frequency : occasional during --force or --failover shutdown under load
4716 Bugzilla : 5949, 4834
4717 Description: Server oops/LBUG if stopped with --force or --failover under load
4718 Details : a collection of import/export refcount and cleanup ordering
4719 issues fixed for safer force cleanup
4722 Frequency : only filesystems larger than 120 OSTs
4723 Bugzilla : 5990, 6223
4724 Description: lfs getstripe would oops on a very large filesystem
4725 Details : lov_getconfig used kfree on vmalloc'd memory
4728 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4729 Bugzilla : 6242, 6243
4730 Description: reading from files that had been truncated to a non-zero size
4731 but never opened returned no data
4732 Details : ll_file_read() reads zeros from no-object files to EOF
4737 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4738 Details : There's little harm in aborting MDS/OSS recovery and letting it
4739 try again, so I removed the LASSERT and return an error instead.
4741 Severity : enhancement
4743 Description: New debugging infrastructure for tracking down data corruption
4744 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4745 (b) cover more of the client-side code path, and (c) try to narrow
4746 down where problems occurred
4750 Bugzilla : 3819, 4364, 4397, 6313
4751 Description: Racing close and eviction MDS could cause assertion in mds_close
4752 Details : It was possible to get multiple mfd references during close and
4753 client eviction, leading to one thread referencing a freed mfd.
4755 Severity: : enhancement
4756 Bugzilla : 3262, 6359
4757 Description: Attempts to reconnect to servers are now more aggressive.
4758 Details : This builds on the enhanced upcall-less recovery that was added
4759 in 1.4.2. When trying to reconnect to servers, clients will
4760 now try each server in the failover group every 10 seconds. By
4761 default, clients would previously try one server every 25 seconds.
4766 Description: After recovery, certain operations trigger a failed
4767 assertion on a client.
4768 Details : Failing over an mds, using lconf -d --failover, while a
4769 client was doing a readdir() call would cause the client to
4770 LBUG after recovery completed and the readdir() was resent.
4772 Severity : enhancement
4774 Description: Default groups are now added by lconf
4775 Details : You can now run lconf --group <servicename> without having to
4776 manually add groups with lmc.
4779 Frequency : occasional
4781 Description: Nodes with an elan id of 0 trigger a failed assertion
4784 Frequency : always when accessing e.g. tty/console device nodes
4786 Description: tty and some other devices nodes cannot be used on lustre
4787 Details : file's private_data field is used by device data and lustre
4788 values in there got lost. New field was added to struct file to
4789 store fs-specific private data.
4792 Frequency : when exporting Lustre via NFS
4794 Description: NFSD failed occasionally when looking up a path component
4795 Details : NFSD is looking up ".." which was broken in ext3 directories
4796 that had grown large enough to become hashed.
4799 Frequency : Clusters with multiple interfaces not on the same subnet
4801 Description: Nodes will repeatedly try to reconnect to an interface which it
4802 cannot reach and report an error to the log.
4803 Details : Extra peer list entries will be created by lconf with some peers
4804 unreachable. lconf now validates the peer before adding it.
4807 Frequency : Only if a default stripe is set on the filesystem root.
4809 Description: Setting a default stripe on the filesystem root prevented the
4810 filesystem from being remounted.
4811 Details : The client was sending extra request flags in the root getattr
4812 request and did not allocate a reply buffer for the dir EA.
4815 Frequency : occasional, higher if lots of files are accessed by one client
4816 Bugzilla : 6159, 6097
4817 Description: Client trips assertion regarding lsm mismatch/magic
4818 Details : While revalidating inodes the VFS looks up inodes with ifind()
4819 and in rare cases can find an inode that is being freed.
4820 The ll_test_inode() code will free the lsm during ifind()
4821 when it finds an existing inode and then the VFS later attaches
4822 this free lsm to a new inode.
4826 Bugzilla : 6422, 7030
4827 Description: MDS deadlock between mkdir and client eviction
4828 Details : Creating a new file via mkdir or mknod (starting a transaction
4829 and getting the ns lock) can deadlock with client eviction
4830 (gets ns lock and trying to finish a synchronous transaction).
4833 Frequency : occasional
4834 Description: While starting a server, the fsfilt_ext3 module could not be
4836 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4837 kernels. Previously, lconf would still use the ext3 name
4838 when trying to load modules. Now, it will correctly use
4839 ext3 on 2.4 and ldiskfs on 2.6.
4841 Severity : enhancement
4842 Description: The default stripe count has been changed to 1
4843 Details : The interpretation of the default stripe count (0, to lfs
4844 or lmc) has been changed to mean striping across a single
4845 OST, rather than all available. For general usage we have
4846 found a stripe count of 1 or 2 works best.
4848 Severity : enhancement
4849 Description: Add support for compiling against Cray portals.
4850 Details : Conditional compiling for some areas that are different
4854 Frequency : occasional
4855 Bugzilla : 6409, 6834
4856 Description: Creating files with an explicit stripe count may lead to
4857 a failed assertion on the MDS
4858 Details : If some OSTs are full or unavailable, creating files may
4859 trigger a failed assertion on the MDS. Now, Lustre will
4860 try to use other servers or return an error to the
4864 Frequency : occasional
4866 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4867 caused lock timeout during readahead (since 1.4.2).
4868 Details : Processes doing readahead might match a lock that hasn't been
4869 granted yet if there are overlapping and conflicting lock
4870 requests. The readahead process waits on ungranted lock
4871 (original lock is CBPENDING), while OST waits for that process
4872 to cancel CBPENDING read lock and eventually evicts client.
4874 Severity : enhancement
4876 Description: Initial enabling of flock support for clients
4877 Details : Implements fcntl advisory locking and file status functions.
4878 This feature is provided as an optional mount flag (default
4879 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4880 locking are implemented yet, and those that are are not guaranteed
4881 to be completely correct in production environments.
4882 mount -t lustre -o [flock|noflock] ...
4885 Frequency : occasional
4887 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4888 assertion in the ext3 JBD (journaling) layer.
4889 Details : The b_committed_data struct is protected by the big kernel lock
4890 in 2.4 kernels, serializing journal_commit_transaction() and
4891 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4892 access to this struct. In 2.6 kernels there is finer grained
4893 locking to improve SMP performance of the JBD layer.
4897 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4902 Description: merge and cleanup kernel patches.
4905 -----------------------------------------------------------------------------
4907 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4909 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4910 versions of Lustre. Please update all servers and clients to
4911 version 1.4.2 or later at the same time. You must also run
4912 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4913 to update the configuration logs.
4915 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4916 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4917 - swab llog records of type '0' so we get proper header size/idx (5861)
4918 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4919 - fix rename of one directory over another leaking an inode (5953)
4920 - avoid SetPageDirty on 2.6 (5981)
4921 - don't re-add just-being-destroyed locks to the waiting list (5653)
4922 - when creating new directories, inherit the parent's custom
4923 striping settings if present parent (3048)
4924 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4925 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4926 - don't hold client locks on temporary worklist from l_lru (5666)
4927 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4928 - automatically evict dead clients (5921)
4929 - Update file size properly in create+truncate+fstat case (6196)
4930 - Do not unhash mountpoint dentries, do not allow removal of
4932 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4933 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4934 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4935 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4936 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4937 - Correctly handle reads of files with no objects (6243)
4938 - lctl recover will also mark a device active if deactivate used (5933)
4940 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4941 - allow --write-conf on an MDS with different nettype than client (5619)
4942 - don't write config llogs to MDS for mounts not from that MDS (5617)
4943 - lconf should create multiple TCP connections from a client (5201)
4944 - init scripts are now turned off by default; run chkconfig --on
4945 lustre and chkconfig --on lustrefs to use them
4946 - upcalls are no longer needed for clients to recover to failover
4948 - add --abort-recovery option to lconf to abort recovery on device
4950 - add support for an arbitrary number of OSTs (3026)
4951 - Quota support protocol changes.
4952 - forward compatibility changes to wire structs (6007)
4953 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4954 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4955 - improved "lustre status" script
4956 - initialize blocksize for non-regular files (6062)
4957 - added --disable-server and --disable-client configure options (5782)
4958 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4959 - Vanilla 2.4.29 support
4960 - increase maximum number of obd devices to 520 (6242)
4961 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4962 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4963 SLES 9 2.6.5 kernels
4964 - limit stripes per file to 160 (the maximum EA size) (6093)
4966 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4969 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4970 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4971 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4972 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4973 - small ext3 extents cleanups and fixes (5733)
4974 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4975 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4976 - avoid needless client->OST connect, fix handle mismatch (5317)
4977 - fix DLM error path that led to out-of-sync client, long delays (5779)
4978 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4979 - fix several locking issues related to i_size (5492,5624,5654,5672)
4980 - don't move pending lock onto export if it is already evicted (5683)
4981 - fix kernel oops when creating .foo in unlinked directory (5548)
4982 - fix deadlock in obdfilter statistics vs. object create (5811)
4983 - use time_{before,after} to avoid timer jiffies wrap (5882)
4984 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4985 - Do not leak request if server was not able to process it (5154)
4986 - If mds_open unable to find parent dir, make that negative lookup(5154)
4987 - don't create new directories with extent-mapping (5909, 5936)
4989 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4990 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4991 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4992 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4993 - support for RHEL4 kernel on i686 (5773)
4994 - provide error messages when incompatible logs are encountered (5898)
4996 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4997 * version 1.4.0.10 (1.4.1 release candidate 1)
4999 - don't keep a lock reference when lock is not granted (4238)
5000 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5001 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5002 - drop import inflight refcount on signal_completed_replay error (5255)
5003 - unlock page after async write error during send (3677)
5004 - handle missing objects in filter_preprw_read properly (5265)
5005 - no transno return for symlink open, don't save no-trasno open (3440)
5006 - don't try to complete elan receive that already failed (4012)
5007 - free RPC server reply state on error (5406)
5008 - clean up thread from ptlrpc_start_thread() on error (5160)
5009 - readahead could read extra page into cache that wasn't ejected (5388)
5010 - prevent races in class_attach/setup/cleanup/detach (5260)
5011 - don't dereference de->d_inode after l_dput of de (5458)
5012 - use "int" for stripe value returned from lock_to_stripe (5544)
5013 - mballoc allocation and error-checking fixes in 2.6 (5504)
5014 - block device patches to fix I/O request sizes in 2.6 (5482)
5015 - look up hostnames for IB nals (5602)
5016 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5017 - don't start multiple acceptors for the same port (5277)
5018 - fix incorrect LASSERT in mds_getattr_name (5635)
5019 - export a proc file for general "ping" checking (5628)
5020 - fix "lfs check" to not block when the MDS is down (5628)
5022 - service request history (4965)
5023 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5024 - create an "evict_client" /proc entry on OSTs, like the MDS has
5025 - fix mount usage message, return errors per mount(8) (5168)
5026 - change grep [] to grep "[]" in tests so they work in more UMLs
5027 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5028 - remove ancient LOV_MAGIC_V0 EA support (5047)
5029 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5030 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5031 - allow readahead allocations to fail when low on memory (5383)
5032 - mmap locking landed again, after considerable improvement (2828)
5033 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5035 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5038 - send OST transaction number in read/write reply to free req (4966)
5039 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5040 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5041 - the watchdog thread now runs as interruptible (5246)
5042 - flock/lockf fixes (but it's still disabled, pending 5135)
5043 - don't use EXT3 constants in llite code (5094)
5044 - memory shortage at startup could cause assertion (5176)
5046 - reorganization of lov code
5047 - single portals codebase
5049 - add extents/mballoc support (5025)
5050 - direct I/O reads in the obdfilter (4048)
5051 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5053 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5056 - send OST transaction number in read/write reply to free req (4966)
5057 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5058 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5059 - the watchdog thread now runs as interruptible (5246)
5060 - handle missing objects in filter_preprw_read properly (5265)
5061 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5062 - drop import inflight refcount on signal_completed_replay error (5255)
5063 - unlock page after async write error during send (3677)
5064 - return original error code on reconstructed replies (3761)
5065 - no transno return for symlink open, don't save no-trasno open (3440)
5067 - add pid to ldlm debugging output (4922)
5068 - bump the watchdog timeouts -- we can't handle 30sec yet
5069 - extra debugging for orphan dentry/inode bug (5259)
5071 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5074 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5075 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5076 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5077 - fix NULL dereference in /proc/sys/portals/routes (4827)
5078 - allow failed mdc_close() operations to be interrupted (4561)
5079 - stop precreate on OST before MDS would time out on it (4778)
5080 - don't send partial-page writes before EOF from client (4410)
5081 - discard client grant for sub-page writes on large-page clients (4520)
5082 - don't free dentries not owned by NFS code, check generation (4806)
5083 - fix lsm leak if mds_create_objects() fails (4801)
5084 - limit debug_daemon file size, always print CERROR messages (4789)
5085 - use transno after validating reply (3892)
5086 - process timed out requests if import state changes (3754)
5087 - update mtime on OST during writes, return in glimpse (4829)
5088 - add mkfsoptions to LDAP (4679)
5089 - use ->max_readahead method instead of zapping global ra (5039)
5090 - don't interrupt __l_wait_event() during strace
5092 - add software watchdogs to catch hung threads quickly (4941)
5093 - make lustrefs init script start after nfs is mounted
5094 - fix CWARN/ERROR duplication (4930)
5095 - return async write errors to application if possible (2248)
5096 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5097 - print NAL number in %x format (4645)
5098 - update barely-supported suse-2.4.21-171 series (4842)
5099 - support for sles 9 %post scripts
5100 - support for building 2.6 kernel-source packages
5101 - support for sles km_* packages
5103 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5106 - ignore -ENOENT errors in osc_destroy (3639)
5107 - notify osc create thread that OSC is being cleaned up (4600)
5108 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5109 - reconstruct ost_handle() like mds_handle() (4657)
5110 - create a new thread to do import eviction to avoid deadlock (3969)
5111 - let lconf resolve symlinked-to devices (4629)
5112 - don't unlink "objects" from directory with default EA (4554)
5113 - hold socknal file ref over connect in case target is down (4394)
5114 - allow more than 32000 subdirectories in a single directory (3244)
5115 - fix blocks count for O_DIRECT writes (3751)
5116 - OST returns ENOSPC from object create when no space left (4539)
5117 - don't send truncate RPC if file size isn't changing (4410)
5118 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5119 - bind to privileged port in socknal and tcpnal (3689)
5121 - rate limit CERROR/CWARN console message to avoid overload (4519)
5122 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5123 - basic mmap support (3918)
5124 - kernel patch series update from b1_4 (4711)
5126 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5129 - avoid crash during MDS cleanup with OST shut down (2775)
5130 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5131 - don't use bad inodes on the MDS (3744)
5132 - dynamic object preallocation to improve recovery speed (4236)
5133 - don't hold spinlock over lock dumping or change debug flags (4401)
5134 - don't zero obd_dev when it is force cleaned (3651)
5135 - print grants to console if they go negative (4431)
5136 - "lctl deactivate" will stop automatic recovery attempts (3406)
5137 - look for existing locks in ldlm_handle_enqueue() (3764)
5138 - don't resolve lock handle twice in recovery avoiding race (4401)
5139 - revalidate should check working dir is a directory (4134)
5141 - don't always mark "slow" obdfilter messages as errors (4418)
5143 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5146 - don't close LustreDB during write_conf until it is done (3860)
5147 - fix typo in lconf for_each_profile (3821)
5148 - allow dumping logs from multiple threads at one time (3820)
5149 - don't allow multiple threads in OSC recovery (3812)
5150 - fix debug_size parameters (3864)
5151 - fix mds_postrecov to initialize import for llog ctxt (3121)
5152 - replace config semaphore with spinlock (3306)
5153 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5154 - don't allow enqueue to complete on a destroyed export (3822)
5155 - down write_lock before checking llog header bitmap (3825)
5156 - recover from lock replay timeout (3764)
5157 - up llog sem before sending rpc (3652)
5158 - reduce ns lock hold times when setting kms (3267)
5159 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5160 - fix NULL deref and obd_dev leak on setup error (3312)
5161 - replace some LBUG about llog ops with error handling (3841)
5162 - don't match INVALID dentries from d_lookup and spin (3784)
5163 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5164 - fix invalid assertion in ptlrpc_set_wait (3880)
5166 - add libwrap support for the TCP acceptor (3996)
5167 - add /proc/sys/portals/routes for non-root route listing (3994)
5168 - allow setting MDS UUID in .xml (2580)
5169 - print the stack of a process that LBUGs (4228)
5171 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5174 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5175 - make sure to unset replay flag from failed open requests (3440)
5176 - if default stripe count is 0, use OST count for inode size (3636)
5177 - update parent mtime/ctime on client for create/unlink (2611)
5178 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5179 - free recovery state on server during a forced cleanup (3571)
5180 - unregister_reply for resent reqs (3063)
5181 - loop back devices mounting and status check on 2.6 (3563)
5182 - fix resource-creation race that can provoke i_size == 0 (3513)
5183 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5184 - more debugging for page-accounting assertion (3746)
5185 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5186 - avoid deadlock after precreation failure (3758)
5187 - fix race and lock order deadlock in orphan handling (3450, 3750)
5188 - add validity checks when grabbing inodes from l_ast_data (3599)
5190 - add /proc/.../recovery_status to obdfilter (3428)
5191 - lightweight CDEBUG infrastructure, debug daemon (3668)
5192 - change default OSC RPC parameters to be better on small clusters
5193 - turn off OST read cache for files smaller than 32MB
5194 - install man pages and include them in rpms (3100)
5195 - add new init script for (un)mounting lustre filesystems (2593)
5196 - run chkconfig in %post for init scripts (3701)
5197 - drop scimac NAL (unmaintained)
5199 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5202 - clean kiobufs before and after use (3485)
5203 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5204 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5205 - change init script to fail more clearly if not run as root (1528)
5206 - allow clients to reconnect during replay (1742)
5207 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5208 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5209 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5210 - fix cond_resched (3554)
5211 - only evict unfinished clients after recovery (3515)
5212 - allow bulk resend, prevent data loss (3570)
5213 - dynamic ptlrpc request buffer allocation (2102)
5214 - don't allow unlinking open directory if it isn't empty (2904)
5215 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5216 - remove extraneous obd dereference causing LASSERT failure (3334)
5217 - don't use get_cycles() when creating temp. files on the mds (3156)
5218 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5219 - handle EEXIST for set-stripe, set proper directory name (3336)
5221 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5222 - fix ksocknal_fmb_callback() error messages (2918)
5224 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5227 - don't copy lvb into (possibly NULL) reply on error (2983)
5228 - don't deref dentry after dput, don't free lvb on error (2922)
5229 - use the kms to determine writeback rpc length (2947)
5230 - increment oti_logcookies when osc is inactive (2948)
5231 - update client's i_blocks count via lvb messages (2543)
5232 - handle intent open/close of special files properly (1557)
5233 - mount MDS with errors=remount-ro, like obdfilter (2009)
5234 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5235 - don't use cancelling-locks' kms values (2947)
5236 - use highest lock extent for kms, not last one (2925)
5237 - don't dereference ERR_PTR() dentry in error handling path (3107)
5238 - fix thread race in portals_debug_dumplog() (3122)
5239 - create lprocfs device entries at setup instead of at attach (1519)
5240 - common AST error handler, don't evict client on completion race (3145)
5241 - zero nameidata in detach_mnt in 2.6 (3118)
5242 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5243 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5244 - initialize RPC timeout timer earlier for 2.6 (3219)
5245 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5246 - print nal/nid for unknown nid (3258)
5247 - additional checks for oscc recovery before doing precreate (3284)
5248 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5249 - don't crash in mdc_close for bad permissions on open (3285)
5250 - zero i_rdev for non-device files (3147)
5251 - clear page->private before handing to FS, better assertion (3119)
5252 - tune the read pipeline (3236)
5253 - fix incorrect decref of invalidated dentry (2350)
5254 - provide read-ahead stats and refine rpc in flight stats (3328)
5255 - don't hold journal transaction open across create RPC (3313)
5256 - update atime on MDS at close time (3265)
5257 - close LDAP connection when recovering to avoid server load (3315)
5258 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5259 - don't leak open file on MDS after open resend (3325)
5260 - serialize filter_precreate and filter_destroy_precreated (3329)
5261 - loop device shouldn't call sync_dev() for nul device (3092)
5262 - clear page cache after eviction (2766)
5263 - resynchronize MDS->OST in background (2824)
5264 - refuse to mount the same filesystem twice on same mountpoint (3394)
5265 - allow llmount to create routes for mounting behind routers (3320)
5266 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5267 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5268 - fix rare problem with rename on htree directories (3417)
5270 - allow default OST striping configuration per directory (1414)
5271 - fix compilation for qswnal for 2.6 kernels (3125)
5272 - increase maximum number of MDS request buffers for large systems
5273 - change liblustreapi to be useful for external progs like lfsck (3098)
5274 - increase local configuration timeout for slow disks (3353)
5275 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5277 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5280 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5281 - don't overwrite extent policy data in reply if lock was blocked (2901)
5282 - drop filter export grants atomically with removal from device (2663)
5283 - del obd_self_export from work_list in class_disconnect_exports (2908)
5284 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5285 - swab reply message in mdc_close, other PPC fixes (2464)
5286 - fix destroying of named logs (2325)
5287 - overwrite old logs when running lconf --write_conf (2264)
5288 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5289 - fix race in target_handle_connect (2898)
5290 - mds_reint_create() should take same inode create lock (2926)
5291 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5292 - don't close files for self_export to avoid uninitialized obd (2936)
5293 - allow MDS with the same name as client node (2939)
5294 - hold dentry reference for closed log files for unlink (2325)
5295 - reserve space for all logs during transactions (2059)
5296 - don't evict page beyond end of stripe extent (2925)
5297 - don't oops on a deleted current working directory (2399)
5298 - handle hard links to targets without a parent properly (2517)
5299 - don't dereference NULL lock when racing during eviction (2867)
5300 - don't grow lock extents when lots of conflicting locks (2919)
5302 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5305 - account for cache space usage on clients to avoid data loss (974)
5306 - lfsck support in lustre kernel code (2349)
5307 - reduce journal credits needed for BRW writes (2370)
5308 - orphan handling to avoid losing space on client/server crashes
5309 - ptlrpcd can be blocked, stopping ALL progress (2477)
5310 - use lock value blocks to assist in proper KMS, faster stat (1021)
5311 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5312 - recovery for initial connections (2355)
5313 - fixes for mds_cleanup_orphans (1934)
5314 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5315 - block all file creations until orphan recovery completes (1901)
5316 - client remove rq_connection from request struct (2423)
5317 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5318 - recovery timer race (2670)
5319 - mdc_close recovey bug (2532)
5320 - ptlrpc cleanup bug (2710)
5321 - mds timeout on local locks (2588)
5322 - namespace lock held during RPCs (2431)
5323 - handle interrupted sync write properly (2503)
5324 - don't try to handle a message that hasn't been replied to (2699)
5325 - client assert failure during cleanup after abort recovery (2701)
5326 - leak mdc device after failed mount (2712)
5327 - ptlrpc_check_set allows timedout requests to complete (2714)
5328 - wait for inflight reqs when ptlrpcd finishes (2710)
5329 - make sure unregistered services are removed from the srv_list
5330 - reset bulk XID's when resending them (caught by 1138 test)
5331 - unregister_bulk after timeout
5332 - fix lconf error (2694)
5333 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5334 - readahead locks pages, leaves pending causing memory pressure (2673)
5335 - increase OST request buffers to 4096 on large machines (2729)
5336 - fix up permission of existing directories in simple_mkdir (2661)
5337 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5338 - don't assign transno to errored transactions (2742)
5339 - don't delete objects on OST if given a bogus objid from MDS (2751)
5340 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5341 - if rq_no_resend, then timeout request after recovery (2432)
5342 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5343 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5344 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5345 - MDS recovery shouldn't delete objects that it has given out (2730)
5346 - if enqueue arrives after completion, don't clobber LVB (2819)
5347 - don't unlock pages twice when trigger_group_io returns error (2814)
5348 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5349 - don't write pages to disk if there was an error (1450)
5350 - don't ping imports that have recovery disabled (2676)
5351 - take buffered bytes into account when balancing socknal conn (2817)
5352 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5353 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5354 - remove little-endian swabbing of llog records (1987)
5355 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5356 - retry reposting request buffers if they fail (1191)
5357 - grow extent at grant time to avoid granting a revoked lock (2809)
5358 - lock revoke doesn't evict page if covered by a second lock (2765)
5359 - disable VM readahead to avoid reading outside lock extents (2805)
5361 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5362 - updated kernel patches for hp-2.4.20 kernel (2681)
5364 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5367 - fix truncated write corruption (2366)
5368 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5370 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5371 - ensure len(uuid) < 37 in lmc (1171)
5372 - fix ia64 OOPS in llog_test (2255)
5373 - zero end of page at obdfilter for partial page writes (2648)
5374 - don't leave stale dentries around after renames (bug 2428)
5375 - fix timeouts when evicting a client with a single lock held (2642)
5376 - set deadline for the initial HELLO message to drain (2634)
5377 - print out dotted-quad IP addresses in the socknal (2302)
5379 - additional debugging for MDS client eviction problem (2443)
5380 - fix mkfsoptions support for osts (2603, 2604)
5382 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5385 - add series for the vanilla 2.6.0 kernel
5386 - add series for the vanilla 2.4.24 kernel
5387 - add series for a cray x86/64 UL kernel drop
5388 - fix xattr patches for the vanilla 2.4.19 series
5390 - generate true UUIDs in lmc (1171)
5391 - have portals stack dumping break in UML (2466)
5392 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5393 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5394 - wake callback waiting threads on client eviction (2460)
5395 - Add --ptldebug and --subsystem to lmc (1719)
5396 - update assertion to allow safe interrupt allocation
5397 - set rq_no_resend for cancel requests (2432)
5398 - recalculate ptlrpcd timeout after resend (2494)
5399 - call vfs_rmdir when removing pending directories (2368)
5400 - fix renaming a file to itself (2429)
5401 - lmc creates a default one-stripe lov (2454)
5402 - expand procfs space to handle large clusters (2326)
5403 - increase UML stack to avoid overflow
5404 - update lconf's list of debug and subsystem masks
5405 - fix lfs find --obd (2510)
5406 - /proc tunable for disabling filter read caching (2591)
5407 - stop rpm packages from altering slapd.conf (2301)
5408 - disable nagle in the socknal under 0conf (2578)
5409 - choose mds inode size based on stripe count (2572)
5410 - fix kernel-source rpm problems (2516)
5412 - add --disable-doc to avoid pdf generation (2421)
5413 - update documentation, tests, type-os, comments
5414 - avoid format warnings on ia64
5415 - remove the TOE NAL
5416 - tiny code cleanups by removing unused fields
5418 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5421 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5422 - remove the most blatant lies from BUILDING file (2371)
5423 - change default debug level to reasonable production setting
5424 - reduce client side cache size to reduce cache flush time
5425 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5426 - make TCP zerocopy and pinger support enabled by default (2476)
5427 - sync writes completed after process exits caused crashes (2319)
5428 - maintain correct mount count on the MDS (2356)
5429 - backout 1557, because 2316 wasn't really fixed
5430 - better file I/O statistics gathering in /proc
5431 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5432 - another kernel patch to fix zero-copy TCP function export
5433 - don't take duplicate lock when processing re-sent getattr (2420)
5434 - lctl uses obd_self_export instead of creating new conn (2353)
5435 - MDS/OST recovery case which requires object creation asserted (2425)
5436 - move lfs from /usr/sbin to /usr/bin in packages
5437 - fix race between mds_client_add and mds_client_free (2417)
5438 - use kmalloc instead of slabs in portals (2430)
5439 - don't create duplicate records when a failover MDS is present (2442)
5440 - remove unnecessary mount age check (2332)
5441 - don't remove directory inodes from locks prematurely (2451)
5442 - don't break if MDS service name is the same as hostname (2103)
5443 - fix races in client write RPC generation when cache full (2482)
5445 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5448 - remove now-unused request->rq_obd (278)
5449 - if an allocation fails, print out how much memory we've used (1933)
5450 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5451 - add the "configurable stack size" patch to most series files (1256)
5452 - ability to write large log records, for 100+ OST configs (2306)
5453 - fix NULL deref when filter_prep fails (2314)
5454 - fix operator precedence error in filter_sync
5455 - dynamic allocation of socknal TX descriptors (2315)
5456 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5457 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5458 - ensure that CWARN messages in Portals always get to the syslog
5459 - __init/__exit are not for prototype decls (ldlm_init/exit)
5460 - x86-64 compile warning fixes
5461 - fix gateway LMC keyword conflict (2318)
5462 - fix MDS lock inversions in getattr/reint paths (1844)
5463 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5464 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5465 - DLM race condition prevented some lock evictions (2328)
5466 - ENOMEM detection and retry on socknal sends (2230)
5467 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5468 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5469 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5470 - better error messages when a client is rejected during recovery (1505)
5471 - avoid cancelling locks which were never granted, after failure (2330)
5472 - fix i_sem/journal inversion in mds_client_add (2333)
5473 - fix truncate/getattr lock cycle deadlock (2334)
5474 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5475 - fix two rare exit paths which could leak an l_lock() ref (2321)
5476 - fencepost error in MDS/OST orphan recovery (2226)
5477 - make log record alignment 8 bytes (1988)
5478 - lstripe now fails when requested offset > ost_count (2237)
5479 - ensure that all kernel series have a complete list.h (1607)
5480 - fix crashes in special-file operations (2316)
5481 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5483 - allow configurable automake binary, for testing new versions
5484 - small update to the lfs documentation
5486 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5488 * fix negative export reference count in fsfilt_sync (2312)
5490 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5491 * release candidate 0.9.1
5493 - orphans are moved into the PENDING directory for possible recovery
5494 - replayed opens now open by fid for orphan/rename safety (1042)
5495 - last close of an orphan inode generates a transno (683)
5496 - chdir() and mount() now pin the directory entry (1020)
5497 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5498 - discard very old requests without processing them (1502)
5499 - remove some common, well-understood CERRORs (1505)
5500 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5501 - clear "grant" flags in OST replies until OST grant code lands (1644)
5502 - fix read performance by not clobbering i_blksize on client (1598)
5503 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5504 - make LRU size a /proc tunable, clears locks when reduced (707)
5505 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5506 - server locks take references on exports now (1558)
5507 - build fixes for 2.4.20-rh trees (1663)
5508 - return an error from lov_create if all OSCs are inactive (1751)
5509 - fix import levels when a reconnect happens without a timeout (1597)
5510 - exit early from mds_open if we get a lookup error (1749)
5511 - partial page read at EOF wouldn't wait for disk before sending (1642)
5512 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5513 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5514 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5515 - write/truncate lock inversion (1639)
5516 - Don't auto-load obdclass, portals modules during cleanup (1495)
5517 - fix timestamps from jumping to "now" (1763)
5518 - extra journal assertions (1648)
5519 - add an extra multiunlink test (1771)
5520 - fix read_record/write_record API (1776)
5521 - fix leak of offset_extent, possible incorrect i_size later (1772)
5522 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5523 - Debugging for the kqswnal_get_idle_tx problems (1820)
5524 - Allow recovery to be attempted multiple times (1536)
5525 - Write out MDS last_rcvd file after it is first created (1600)
5526 - Fix tx_descriptor leak in failed transmit situations (1827)
5527 - ext3 journaling fixes for assertion failure after IO error (1871)
5528 - class_export_put() on freed export after completion AST error (1896)
5529 - Fix revalidate looping in VFS (1322)
5530 - Don't access a freed export during MDS_REINT timeout (1521)
5531 - Add open-unlink recovery support on the MDS (1673,1764)
5532 - Return an error if no MDS data was read from last_rcvd (1946)
5533 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5534 - Don't setup a disk device that doesn't match exported UUID (317)
5535 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5536 - avoid committing NULL handle in force close
5537 - local.sh is now a one-stripe LOV configuration
5538 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5539 - fix ext3 htree duplicate directory entry corruption (1516)
5540 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5541 - update to new LOV EA format (2097)
5542 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5544 2003-06-15 Phil Schwan <phil@clusterfs.com>
5547 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5548 - per-import/export recovery handling (958, 931, 959)
5549 - multiple last-rcvd slots, for serving multiple FSes (949)
5550 - connections are again shared between multiple imp/exports (963, 964)
5551 - "umount -f" would hang if any requests needed to be sent (393, 978)
5552 - avoid pinning large req buffer by copying for queued messages (989)
5553 - add "uuid" to "lctl device" command to help upcalls (991)
5554 - "open" RPCs with transnos would confuse recovery counters (1037)
5555 - do proper endian conversion of all wire messages (288, 340, 891)
5556 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5557 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5558 - fix ptlrpc_connection leak in target_handle_connect (1174)
5559 - fix import refcounting bug in OST and MDS cleanup (1134)
5560 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5561 - fix very unlikely obd_types race condition (501)
5562 - remove osc_open hack for echo_client (1187)
5563 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5564 - a failure in read_inode2 leads to deadlock (1139)
5565 - cancel ack-locks as soon as transaction is committed (1072)
5566 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5567 - make sure to commitrw after any preprw to avoid deadlock (1162)
5568 - failing to execute a file in a lustre FS would lock inode (1203)
5569 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5570 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5571 - obd_uuid2tgt would walk off the end of the list (1255)
5572 - on IA64 the osc would give portals incorrect bulk size (1258)
5573 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5574 - fix lock inversion caused by new llite matching code (1282)
5575 - limit the number of dirty pages on a client to 10MB (1286)
5576 - timed out locks were not being corrected cancelled (1289)
5577 - fix O_DIRECT above 4GB on IA-32 (1292)
5578 * major user-visible changes
5579 - fail out/fail over policy now controlled by the upcall (993)
5581 - add OBD_PING to check server availability and failure (954)
5582 - lustre messages are now sent in sending host order (288, 340, 891)
5583 - add eadatalen to MDS getattr reply (340)
5584 - OST read replies may contain second buffer, with per-page status (593)
5586 2003-03-11 Phil Schwan <phil@clusterfs.com>
5589 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5590 - failed open()s could cause deadlock; fixed (867, 869)
5591 - stop cancelling OST locks when files are closed (481)
5592 - overlapping XID spaces caused network corruption (851, 853)
5593 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5594 - setattr_raw added, to do single-RPC, server-side setattrs
5595 - lmc/lconf syntax change for OST UUIDs
5596 - fix crashy race condition between ptlrpc_free_req and osc_close
5597 - don't use request in mdc_enqueue if we hit a timeout (889)
5598 - don't set the inode i_size for regular files from the MDS (896)
5599 - handle out of order completion AST (842)
5600 - don't LBUG if a lock request times out after receiving AST (913)
5601 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5602 - if a bad lock AST arrives, send an error instead of dropping entirely
5603 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5604 - fix leak in bulk IO when only partially completed (899, 900, 926)
5605 - fix O_DIRECT for ia64 (55)
5606 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5607 - C-z after timeout could hang a process forever; fixed (977)
5609 - client-side I/O cache (678, 924, 929, 941, 970)
5611 - READPAGE and SETATTRs which don't take server-side locks get
5614 2003-02-11 Phil Schwan <phil@clusterfs.com>
5617 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5618 - Fix fsfilt_extN_readpage() to read a full page of directory
5619 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5620 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5621 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5622 - Most intent operations were removed, in favour of a new RPC mode
5623 that does a single RPC to the server and bypasses most of the VFS
5624 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5625 - Aggressively cancel local locks on DLM servers
5626 - mds_reint_unlink sends EA to the client if it's the last nlink.
5627 client uses that EA to unlink OST objects.
5628 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5629 - recursive symlinks were fixed (439)
5630 - fixed NULL deref in DEBUG_REQ
5631 - filter_update_lastobjid no longer calls sync, which annoyed extN
5632 - fixed multi-client small-writes to a single file problem (445)
5633 - fixed mtime updates during file writes (607)
5634 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5635 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5636 - fixed Linux OST size reporting problem (444, 656)
5637 - OST now updates object mtime with writes or setattr (607, 619)
5638 - client verifies file size before zeroing page past EOF (445)
5639 - OST now writes last allocated objid to disk with allocation (108)
5640 - LOV on echo now works (409)
5642 - mds_reint_unlink sends a new buffer, with the EA included. this
5643 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5644 set if a regular file was being unlinked, and it was the last link
5645 - use PtlGet from the target for bulk writes (315)
5646 - OST now updates object mtime with writes or setattr (607, 619)
5647 - LDLM now has a grant-time callback to revalidate locked items, if
5649 - Many MDS operations were reorganized to combat race conditions
5651 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5652 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5654 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5657 - Fully reactivate OST imports after reconnection (512, others)
5658 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5659 - More graceful error handling for truncating on dead OST (515)
5660 - Don't error out unless we're actually accessing dead stripes (474)
5661 - Fix garbage sizes when stripes are missing (410)
5662 - LRU counters were broken, causing constant lock purge (433, 432)
5663 - garbage on read from stripes with failed OSTs (441)
5664 - mark OSCs as active before reconnecting during recovery (438)
5665 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5666 - lfind did not preserve OST order in output (443)
5667 - symlinks cause hung clients, incorrect data (439)
5668 - stop dereferencing request after dropping refcount (457)
5669 - don't LASSERT(spin_is_locked) on non-SMP (455)
5670 - fixes for many rename() bugs
5671 - fstat didn't correctly synchronize attributes (399)
5672 - server must handle lock cancellation during blocking AST prep (487)
5673 - bulk descriptors were free()d too soon (511)
5674 - fix paths in lconf, which would load incorrect modules (451, 507)
5675 - fix confusing lconf 'host not found' error message (386)
5676 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5677 - fix race condition in mdc_blocking_ast() for inode access (526)
5678 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5679 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5680 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5681 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5682 - fix lstripe in several ways (kernel side) (527)
5683 - fix request leak in ldlm_cli_enqueue (262)
5684 - incorrect OSC was marked inactive after OST failure
5685 - call mds_fs_cleanup before unmounting filesystem (524)
5686 - fix races between taking ns_lock and ldlm_lock_change_resource
5687 - fix races updating LOV export open file list
5688 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5689 - fix recovery NULL deref in ldlm_cli_cancel_unused
5690 - fix some DLM races by using new hash table for lock handles (419)
5691 - permit the client to specify desired inodes, at replay
5692 - duplicate requests when we queue them for replay reintegration
5693 - fix last_rcvd offset calculation
5694 - sync after each recovered transaction, so we always make progress
5695 - never, not always, ERESTART requests without transnos
5696 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5698 - skip replay if the MDS says that the client is already connected
5699 - don't check for a recovery-enabled export to match lctl's UUID
5700 - don't INC_USE_COUNT for phantom exports
5701 - don't crash when cleaning up phantom exports (567)
5702 - don't double-finish or set replay data for errored mdc_open requests
5703 - abort requests when they time out, so we don't get old replies
5704 - send/receive replies for AST messages again
5705 - if the client says that it doesn't have the lock, cancel it on the
5707 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5708 mark it as destroyed, it will all work out when decref is called
5709 - fix module use counts (22, 581)
5711 - ASTs now expect a reply (server cancels lock on error reply)
5713 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5716 - fix many simultaneous client startup (392)
5717 - fix dentry->d_it clobbering
5718 - credentials weren't being shipped for readdir/getattr operations
5719 - remove invalid assertions triggered during some concurrent MD
5721 - proper Lustre versions added (336, 389)
5722 - fix memory leak for create error case (398)
5723 - fix LOV locking bug that would get cli/srv out of sync
5724 - fix echo client over LOV (409)
5725 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5726 - fix double-O_EXCL intent crash (424)
5727 - avoid sending multiple lock CANCELs (352)
5729 - MDS can do multi-client recovery (modulo bugs in new code)
5731 - many updates, edits, cleanups
5733 2002-11-18 Phil Schwan <phil@clusterfs.com>
5736 - fix null d_it dereference (346)
5737 - fix full OST/dbench hang (333)
5738 - fix permission problem with file removal (286)
5739 - fix removal of OSCs from LOV when they fail
5740 - fix NULL deref during bulk timeout (214)
5741 - fix problems related to multiple filesystems on one MDS (241)
5742 - fixed serious subtle metadata locking bugs
5743 - free locks on clients when inodes are removed due to memory
5745 - fix inode pointer in lock data (285)
5746 - partial support for multiple MDS on a single host (241)
5747 - data locks weren't cancelled at clear_inode time (290, 311)
5748 - intent locks could lead to unbounded lock growth (205)
5749 - added a maximum lock count, an LRU list, and a flusher
5750 - fix multiple rename (365)
5751 - properly abstracted the echo client
5752 - OSC locked 1 byte too many; fixed
5753 - rewrote brw callback code:
5754 - fixed recovery bugs related to LOVs (306)
5755 - fixed too-many-pages-in-one-write crash (191)
5756 - fixed (again) crash in sync_io_timeout (214)
5757 - probably fixed callback-related race (385)
5759 - Add capability to MDS protocol
5760 - LDLM cancellations and callbacks on different portals
5762 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5765 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5766 - defer open object destruction to close time (601981, 138)
5767 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5768 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5769 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5770 - don't call obd_setattr in ll_file_release for destroyed objects
5771 * protocol change to lustre_msg: move |version| and add |flags|
5772 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5773 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5774 * added replay of create, unlink, link and rename operations during
5775 MDS failover; recovery should be much more robust now
5776 * remove failed OSCs from LOVs (only lov_create uses this so far)
5777 * the lustre-HOWTO was brought (more) up to date (582544)
5779 2002-10-23 Phil Schwan <phil@clusterfs.com>
5782 - in-use dentries weren't being reused properly (617851)
5783 - prevent multiple LDLM setup (599178)
5784 - fix LOV size calculations for truncate (617853)
5785 - fix client handling of MDS intent errors (POSIX)
5786 - fix permission bug in lovstripe.c test (624321)
5787 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5788 - truncate past end of file could corrupt data
5789 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5790 - a race in recovery could return ETIMEDOUT to apps (623947)
5791 - building outside the source directory was fixed
5792 * the lustre-HOWTO was brought (more) up to date (582544)
5793 * major progress was made on recovery functionality
5795 2002-10-10 Phil Schwan <phil@clusterfs.com>
5798 - recovery deadlock fix
5799 - rm -rf causes LBUG fix (617817)
5800 - file open by multiple tasks fix (618962)
5801 - directory permissions bugs (602707 and 620007)
5802 - journal_stop fixed with locking (611313)
5803 - O_APPEND failures resolved (618273, perhaps 614459)
5804 - lconf PATH fix (619770)
5805 - IA64 build fix (621450)
5806 - RPC buffer sizes scale with amount of memory
5808 2002-10-01 Phil Schwan <phil@clusterfs.com>
5811 - locks would be cancelled without throwing away data pages,
5812 resulting in inconsistent data (605627)
5813 - inode attributes were not always being refreshed (605627, 612449)
5814 - lconf now continues to cleanup after lctl reports an error
5815 - MDS now enforces user permissions (602707)
5816 - lprocfs cleanup fixed, but not yet enabled (614157)
5817 - fixed infinite server hang, should a client not respond to an AST
5818 - avoid going into recovery if user calls readlink() with a buffer
5819 that's too small (613941)
5820 - AST RPCs no longer require replies (614867) -- this may be changed
5821 - don't crash server if client sends an IOV that's too big (611336)
5822 - fixed lock conversion deadlock (611892)
5823 - fixed the following of symlinks (614622)
5824 * recovery: the server can remove locks from a client that dies, other
5825 clients can make progress
5826 * more extN patch fixes
5827 * compile-time configurable ptlrpc buffer allocations
5829 - collaborative read cache document
5830 - Lustre Lite Performance CDR document-in-progress
5832 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5835 - fix typo in patch-2.4.18
5837 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5840 - clear ptlrpc request each time in handle_incoming_request()
5841 - unlink of files now destroys the object on the OST
5843 2002-09-19 Peter Braam <braam@clusterfs.com>
5845 * add hard link support
5846 * change obdfile creation method
5847 * kernel patch changed
5849 2002-09-19 Peter Braam <braam@clusterfs.com>
5852 - stack overflow bug in extN fixed
5854 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5856 * documentation updates
5857 - add man pages for config tools
5858 - update tests/README to describe testing with new config tools
5859 - finish metadata API descriptions
5860 * bug fixes and cleanups
5861 - statfs workaround for 16TB limit
5862 - LOV stripe allocation improved, can stripe on subset of OSTs
5863 - LOV file size/IO offset was wrong for files > 4GB in size
5864 - object EA data was being dropped, caused files to be unreadable
5865 - memory overflow with non-LOV OST caused memory corruption
5866 - fixed regression tests to work with new config tools, obdfilter
5867 - fixed bug when directory size became larger than 1 block
5868 - fixed bug (for single client case) when PWD was deleted
5869 - invalidate local directory pages when doing intent-based ops
5870 - avoid LDLM oops when lock callback contained bad data
5872 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5874 * documentation updates
5875 * bug fixes and cleanups
5876 - configuration tools
5879 - 64-bit compile warnings
5880 - 64-bit internal statfs data
5882 * test_brw on persistent OST devices
5884 * lprocfs (disabled)
5886 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5888 * documentation updates
5889 * bug fixes and cleanups
5890 * configuration tools
5892 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5895 * many small fixes to 0_5_4
5896 * io/network handling
5897 * thinkos in MDS operations
5899 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5902 * crucial basic fixes to 0.5.3
5903 * IOR, Iozone work over Elan
5906 2002-08-07 Phil Schwan <phil@clusterfs.com>
5907 * version 0_5_3, our first alpha
5908 * we use the new Portals iovs
5909 * documentation updates
5910 * bug fixes and cleanups
5911 * small changes in the DLM wire protocol
5913 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5914 * version 0_5_1 with some initial stability,
5915 * locking on MD and file I/O.
5916 * documentation updates
5917 * several bug fixes since 0.5.0
5918 * small changes in wire protocol
5920 2002-07-18 Phil Schwan <phil@clusterfs.com>
5922 * delivered as Lustre Light Alpha
5923 * fixed a crash after handling invalid MDS requests
5924 * fixed directory pages for architectures with non-4k pages sizes
5926 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5927 * release version v0_4_4
5928 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5930 * Fixes a number of interruption problems with OST operations.
5931 * Update documentation for portals header changes
5932 * Move all wire protocol structs/defines to lustre_idl.h
5933 * Fixes symlink length bug.
5934 * Add tcpdump to repository.
5936 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5937 * release version v0_4_3
5938 * Fixes statfs for inodes on extN.
5939 * Fixes bug in runtests which would delete /etc/hosts.
5940 * Use 64-bit object IDs wherever possible (not into VFS though)
5941 Remove ost_get_info, which is unused by lustre, and out of date.
5943 2002-07-03 Peter Braam <braam@clusterfs.com>
5944 * release version v0_4_2 Fixes a lookup error (type not passed)
5945 * move forward to head of Portals
5946 * move forward to latest Lustre kernel
5948 2002-06-25 Peter Braam <braam@clusterfs.com>
5949 * release version v0_4_1. Hopefully stable on single node use.