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.
18 Description: don't panic if EPROTO was hit when reading symlink
19 Details : correctly handling request reference in error cases.
21 Severity : enhancement
23 Description: Add LustreNetLink, a kernel-userspace communcation path.
25 Severity : enhancement
27 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
32 Description: access to llog context before init.
33 Details : move handling CATALOGS file at osc layer and forbid access to llog
38 Description: Avoid deadlock for local client writes
39 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
40 memory freeing context. This allows OST threads to set the
41 PF_MEMALLOC flag on task structures in order to allocate memory
42 from reserved pools and complete IO.
43 Use GFP_HIGHUSER for OST allocations for non-local client writes,
44 so that the OST threads generate memory pressure and allow
45 inactive pages to be reclaimed.
47 Severity : enhancement
49 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
54 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
55 Details : move ldlm namespace creation in setup phase to avoid grab
56 _lprocfs_lock with cli_sem held.
60 Description: Temporarily disable grant shrink.
61 Details : Disable the feature for debugging.
65 Description: Unable to run several mkfs.lustre on loop devices at the same
67 Details : mkfs.lustre returns error 256 on the concurrent loop devices
68 formatting. The solution is to proper handle the error.
70 Severity : enhancement
72 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
74 Severity : enhancement
76 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
80 Description: resolve race between obd_disconnect and class_disconnect_exports
81 Details : if obd_disconnect will be called to already disconnected export he
82 forget release one reference and osc module can't unloaded.
84 Severity : enhancement
86 Description: Allow tuning service thread via /proc
87 Details : For each service a new
88 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
89 created that can be used to set min/max thread counts, and get the
90 current number of running threads.
94 Descriptoin: don't return error if have particaly created objects for file.
95 Details : lov_update_create_set uses set->set_success as index for created objects,
96 so if some requests will be failed, they will have hole at end of
97 array and we can use qos_shrink_lsm for allocate correct lsm.
99 Severity : enhancement
101 Description: Update OFED support to 1.4.1
105 Description: Reduce small size read RPC
106 Details : Set read-ahead limite for every file and only do read-ahead when
107 available read-ahead pages are bigger than 1M to avoid small size
110 Severity : enhancement
112 Description: move AT tunable parameters for more consistent usage
113 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
115 Severity : enhancement
117 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
118 Details : allow skip disconnected ost for send statfs request and hide error
124 Description: fix racy locking of mballoc block bitmaps causing BUG
125 Details : The locking of the mballoc buddy bitmap and the in-memory
126 block bitmap was using two different spin locks in some
127 cases. This made it possible to incorrectly access the
128 mballoc bitmap while another process was modifying it,
129 causing a sanity assertion to fail. While no on-disk corruption
130 was reported, there was some risk of this happening.
133 Frequency : rare, on llog test 6
135 Descriptoin: don't allow connect to already connected import
136 Details : allowing connect to already connected import is hide connecting problem.
141 Descriptoin: allow kill process which wait statahead result
142 Details : for some reasons 'ls' can stick in waiting result from statahead,
143 in this case need way for kill this process.
145 Severity : enhancement
147 Description: Add state history info file, enhance import info file
148 Details : Track import connection state changes in a new osc/mdc proc file;
149 add overview-type data to the osc/mdc import proc file.
151 Severity : enhancement
153 Description: MDS create should not wait for statfs RPC while holding DLM lock.
155 Severity : enhancement
157 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
162 Descriptoin: OST grant too much space to client even there are not enough space.
163 Details : Client will shrink its grant cache to OST if there are no write
164 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
165 this grant cache if there are already not enough avaible space
166 (left_space < total_clients * 32M).
169 Frequency : start MDS on uncleanly shutdowned MDS device
171 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
172 Details : stay in waiting mds<>ost recovery finished produce random bugs
173 due race between two ll_sync thread for one lov target. send
174 ACTIVATE event only if connect realy finished and import have
178 Frequency : rare, connect and disconnect target at same time
180 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
181 Details : don't call obd_disconnect under lov_lock. this long time
182 operation and can block ptlrpcd which answer to connect request.
187 Descriptoin: don't lose wakeup for imp_recovery_waitq
188 Details : recover_import_no_retry or invalidate_import and import_close can
189 both sleep on imp_recovery_waitq, but we was send only one wakeup
193 Frequency : always with long access acl
195 Descriptoin: mds can't pack reply with long acl.
196 Details : mds don't control size of acl but they limited by reint/getattr
199 Severity : enhancement
201 Description: Update to SLES10 kernel-2.6.16.60-0.33.
203 Severity : enhancement
205 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
208 Frequency : start MDS on uncleanly shutdowned MDS device
210 Descriptoin: aborting recovery hang on MDS
211 Details : don't throttle destroy RPCs for the MDT.
214 Frequency : on remount
216 Description: external journal device not working after the remount
217 Details : clear dev_rdonly flag for external journal devices in
223 Description: shutdown vs evict race
224 Details : client_disconnect_export vs connect request race.
225 if client will evicted at this time - we start invalidate
226 thread without referece to import and import can be freed
230 Frequency : rare, need acl's on inode.
232 Description: client can't handle ost additional correctly
233 Details : if ost was added after client connected to mds client can have
234 hit lnet_try_match_md ... to big messages to wide striped files.
235 in this case need teach client to handle config events about add
236 lov target and update client max ea size at that event.
238 Severity : enhancement
240 Description: Changelogs
241 Details : Changelogs are a lightweight mechanism to track filesystem
242 metadata and namespace changes. The changelog is recorded
243 permanently on the MDTs, and is periodically "consumed" / purged
244 when records are no longer needed.
246 Severity : enhancement
248 Description: compact fld format with extents
249 Details : Store range of seq rather than every seq in FLD. Seq
250 controller update FLD rather than clients. In Case of CMD, mdt0
251 has FLD, all other metadata server act as non persistent proxy
252 for FLD queries and cache fld entries in fld cache.
257 Description: don't skip ost target if they assigned to file
258 Details : Drop slow OSCs if we can, but not for requested start idx.
259 This means "if OSC is slow and it is not the requested
260 start OST, then it can be skipped, otherwise skip it only
261 if it is inactive/recovering/out-of-space.
265 Description: more cleanup in mds_lov
266 Details : not send LOV EA under replay, we can't know about they size at this
267 time. Don't allow client connect to mds before any ost connected,
268 for avoid problems with LOV EA size and returning EIO to client.
270 Severity : enhancement
272 Description: Interoperability at server side (Disk interoperability)
274 Severity : enhancement
276 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
278 Severity : enhancement
280 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
282 Severity : enhancement
284 Description: New client IO stack (CLIO).
286 Severity : enhancement
288 Description: Commit on sharing. Eliminate inter-client dependencies between
289 uncommitted transactions by doing transaction commits.
290 Thereby clients may recovery independently.
293 Frequency : Create a symlink file with a very long name
295 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
296 Details : If there is no extra space in the request for early cancels,
297 ldlm_req_handles_avail() returns 0 instead of a negative value.
299 Severity : enhancement
301 Description: Add /proc entry for import status
302 Details : The mdc, osc, and mgc import directories now have
303 an import directory that contains useful import data for debugging
306 Severity : enhancement
308 Description: Re-disable certain /proc logging
309 Details : Enable and disable client's offset_stats, extents_stats and
310 extents_stats_per_process stats logging on the fly.
313 Frequency : Only on FC kernels 2.6.22+
315 Description: oops in statahead
316 Details : Do not drop reference count for the dentry from VFS when lookup,
317 VFS will do that by itself.
319 Severity : enhancement
321 Description: Generic /proc file permissions
322 Details : Set /Proc file permissions in a more generic way to enable non-
323 root users operate on some /proc files.
327 Description: Hitting mdc_commit_close() ASSERTION
328 Details : Properly handle request reference release in
329 ll_release_openhandle().
333 Description: quota recovery deadlock during mds failover
334 Details : This patch includes att18982, att18236, att18237 in bz14840.
336 1. fix osts hang when mds does failover with quotaon
337 2. prevent watchdog storm when osts threads wait for the
342 Frequency : only patchless client
343 Description: add workaround for race between add/remove dentry from hash
345 Severity : enhancement
347 Description: Allow OST glimpses to return PW locks
351 Description: LBUG when llog conf file is full
352 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
356 Description: Prevent import from entering FULL state when server in recovery
360 Description: service mount cannot take device name with ":"
361 Details : Only when device name contains ":/" will mount treat it as
367 Description: replace ptlrpcd with the statahead thread to interpret the async
368 statahead RPC callback
372 Frequency : on recovery
373 Description: I/O failures after umount during fail back
374 Details : if client reconnected to restarted server we need join to recovery
375 instead of find server handler is changed and process self
376 eviction with cancel all locks.
378 Severity : enhancement
380 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
382 Severity : enhancement
384 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
386 Severity : enhancement
388 Description: Upcall on Lustre log has been dumped
389 Details : Allow for a user mode script to be called once a Lustre log has
390 been dumped. It passes the filename of the dumped log to the
391 script, the location of the script can be specified via
392 /proc/sys/lnet/debug_log_upcall.
397 Description: avoid idr_remove called for id which is not allocated.
398 Details : Move assigment s_dev for clustered nfs to end of initialization,
399 to avoid problem with error handling.
404 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
405 Details : When connection is reused this not moved from CONN_UNUSED_HASH
406 into CONN_USED_HASH and this prodice warning when put connection
407 again in unused hash.
410 Severity : enhancement
412 Description: File striping can now be set to use an arbitrary pool of OSTs.
414 Severity : enhancement
416 Description: Export bytes_read/bytes_write count on OSC/OST.
420 Description: Early reply size mismatch, MGC loses connection
421 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
422 the connect flags are properly negotiated.
427 Description: quotas are not honored with O_DIRECT
428 Details : all writes with the flag O_DIRECT will use grants which leads to
429 this problem. Now using OBD_BRW_SYNC to guard this.
433 Description: add quota statistics
434 Details : 1. sort out quota proc entries and proc code.
435 2. add quota statistics
437 Severity : enhancement
439 Description: enable quota support for HEAD.
443 Description: Properly propagate oinfo flags from lov to osc for statfs
444 Details : restore missing copy oi_flags to lov requests.
446 Severity : enhancement
448 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
449 : l_getgroups(8), lst(8), routerstat(8)
450 Details : included man pages for llobdstat(8), llstat(8),
451 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
453 Severity : enhancement
455 Description: configure's --enable-quota should check the
456 : kernel .config for CONFIG_QUOTA
457 Details : configure is terminated if --enable-quota is passed but
458 : no quota support is in kernel
462 Description: Remove portals compatibility
463 Details : Remove portals compatibility, not interoperable with releases
468 Description: Resolve device initialization race
469 Details : Prevent proc handler from accessing devices added to the
470 obd_devs array but yet be intialized.
472 Severity : enhancement
474 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
476 Severity : enhancement
478 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
483 Description: Using wrong pointer in osc_brw_prep_request
484 Details : Access to array[-1] can produce panic if kernel compiled with
485 CONFIG_PAGE_ALLOC enabled
489 Description: Client runs out of low memory
490 Details : Consider only lowmem when counting initial number of llap pages
494 Description: *optional* service tags registration
495 Details : if the "service tags" package is installed on a Lustre node
496 When the filesystem is mounted, a local-node service tag will
497 be created. See http://inventory.sun.com/ for more information
498 about the Service Tags asset management system.
502 Description: Kernel BUG tries to release flock
503 Details : Lustre does not destroy flock lock before last reference goes
504 away. So always drop flock locks when client is evicted and
505 perform unlock regardless of successfulness of speaking to MDS.
509 Description: add refcount protection for osc callbacks, avoid panic on shutdown
513 Description: sanity test 65a fails if stripecount of -1 is set
514 Details : handle -1 striping on filesystem in ll_dirstripe_verify
519 Description: ASSERTION(CheckWriteback(page,cmd)) failed
520 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
523 Severity : enhancement
525 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
529 Description: do not process already freed flock
530 Details : flock can possibly be freed by another thread before it reaches
531 to ldlm_flock_completion_ast.
535 Description: LBUG during stress test
536 Details : Need properly lock accesses the flock deadlock detection list.
540 Description: oops in page fault handler
541 Details : kernel page fault handler can return two special 'pages' in error
542 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
546 Description: timeout with invalidate import.
547 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
548 should be handled by ptlrpcd. This produce long age waiting and
549 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
551 Severity : enhancement
553 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
557 Description: enable MGS and MDT services start separately
558 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
559 a co-located MGS without starting the MGS, which is a complement
560 to 'nosvc' mount option.
564 Frequency : after recovery
565 Description: precreate to many object's after del orphan.
566 Details : del orphan st in oscc last_id == next_id and this triger growing
567 count of precreated objects. Set flag LOW to skip increase count
568 of precreated objects.
572 Frequency : rare, on clear nid stats
573 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
574 Details : when clean nid stats sometimes try destroy live entry,
575 and this produce panic in free.
579 Description: Stack overflow during MDS log replay
580 ease stack pressure by using a thread dealing llog_process.
584 Description: wait until IO finished before start new when do lock cancel.
585 Details : VM protocol want old IO finished before start new, in this case
586 need wait until PG_writeback is cleared until check dirty flag and
587 call writepages in lock cancel callback.
589 Severity : enhancement
591 Description: using special macro for print time and cleanup in includes.
595 Description: mds_mfd_close() ASSERTION(rc == 0)
596 Details : In mds_mfd_close(), we need protect inode's writecount change
597 within its orphan write semaphore to prevent possible races.
601 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
605 Frequency : rare, on shutdown ost
606 Description: don't hit live lock with umount ost.
607 Details : shrink_dcache_parent can be in long loop with destroy dentries,
608 use shrink_dcache_sb instead.
612 Description: don't panic with use echo client
613 Details : echo client pass NULL as client nid pointer and this produce null
618 Description: fix build on ppc32
619 Details : compile code with -m64 flag produce wrong object file for ppc32.
623 Description: add message levels for liblustreapi
627 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
628 Details : Move the 'good_osts' check before the 'total_bavail' check. This
629 will result in an -EAGAIN and in the exit call path we call
630 alloc_rr() which will with increasing aggressiveness attempt to
631 aquire precreated objects on the minimum number of required OSCs.
635 Description: Use old size assignment to avoid deadlock
636 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
637 the scheduling while holding a spinlock. We do not need locking
638 for size in ll_update_inode() because size is only updated from
639 the MDS for directories or files without objects, so there is no
640 other place to do the update, and concurrent access to such inodes
641 are protected by the inode lock.
645 Description: resolve "_IOWR redefined" build error on SLES10
649 Description: dump the memory debugging after all modules are unloaded to
650 suppress false negative in conf_sanity test 39
652 Severity : enhancement
654 Description: build kernel-ib packages for OFED 1.3 in our release cycle
659 Description: fix SLES kernel versioning
660 Details : the kernel version for our SLES 10 kernel did not include a "-"
661 before the "smp" at the end. while this was not a problem in
662 general, it did mean that software trying to use the kernel
663 version to try to detect a vendor specific kernel would fail.
664 this was most evident by the OFED build scripts.
668 Description: Don't update lov_desc members until making sure they are valid
669 Details : When updating lov_desc members via proc fs, need fix their
670 validities before doing the real update.
674 Description: don't put request into delay list while invalidate in flight.
675 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
676 import in flight. this produce timeout for invalidate and sometimes
677 can cause stale data.
681 Frequency : on ppc only
682 Description: not convert ost objects for directory because it's not exist.
683 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
687 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
689 Severity : enhancement
691 Description: Update to support 2.6.22.14 vanilla kernel.
695 Frequency : rare, on recovery
696 Description: read procfs can produce deadlock in some situation
697 Details : Holding lprocfs lock which send rpc can produce block for destroy
698 obd objects and this also block reconnect with -EALREADY. This
699 isn't fix all lprocfs bugs - but make it rare.
701 Severity : enhancement
703 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
706 Frequency : frequent on X2 node
708 Description: mdc_set_open_replay_data LBUG
709 Details : Set replay data for requests that are eligible for replay.
713 Description: lustre_mgs: operation 101 on unconnected MGS
714 Details : When MGC is disconnected from MGS long enough, MGS will evict the
715 MGC, and late on MGC cannot successfully connect to MGS and a lot
716 of the error messages complaining that MGS is not connected.
720 Frequency : on network error
721 Description: panic with double free request if network error
722 Details : mdc_finish_enqueue is finish request if any network error ocuring,
723 but it's true only for synchronus enqueue, for async enqueue
724 (via ptlrpcd) this incorrect and ptlrpcd want finish request
727 Severity : enhancement
729 Description: client-side metadata stat-ahead during readdir(directory readahead)
730 Details : perform client-side metadata stat-ahead when the client detects
731 readdir and sequential stat of dir entries therein
734 Frequency : on start mds
736 Description: Implement get_info(last_id) in obdfilter.
739 Frequency : occasional
741 Description: Correctly check stale fid, not start epoch if ost not support SOM
742 Details : open with flag O_CREATE need set old fid in op_fid3 because
743 op_fid2 was overwritten with new generated fid, but mds can answer
744 with one of these two fids and both is not stale. Setattr
745 incorrectly started an epoch and assume will be called
746 done_writing, but without SOM done_writing ever being called.
749 Frequency : rare, depends on device drivers and load
751 Description: MDS or OSS nodes crash due to stack overflow
752 Details : Code changes in 1.8.0 increased the stack usage of some functions.
753 In some cases, in conjunction with device drivers that use a lot
754 of stack the MDS (or possibly OSS) service threads could overflow
755 the stack. One change which was identified to consume additional
756 stack has been reworked to avoid the extra stack usage.
759 Frequency : occasional
761 Description: Do not fail import if osc_interpret_create gets -EAGAIN
762 Details : If osc_interpret_create got -EAGAIN it immediately exits and
763 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
764 oscc_has_objects and see OSC has no objests and call
765 oscc_internal_create to resend create request.
767 Severity : enhancement
769 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
771 Severity : enhancement
773 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
776 Frequency : very rare
778 Description: Fix replay if there is an un-replied request and open
779 Details : In some cases, older replay request will revert the
780 mcd->mcd_last_xid on MDS which is used to record the client's
783 Severity : enhancement
785 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
787 Severity : enhancement
789 Description: Add rhel5 support to HEAD.
791 Serverity : enhancement
793 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
798 Description: Don't allow skipping OSTs if index has been specified.
799 Details : Don't allow skipping OSTs if index has been specified, make
800 locking in internal create lots better.
804 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
805 Details : access to bitfield in structure is always rounded to long
806 and this produce problem with not atomic change any bit.
810 Description: Lustre make rpms failed.
811 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
812 builds Lustre rpms from tarball.
815 Frequency : rare on shutdown ost
817 Description: If llog cancel was not send before clean_exports phase, this can
818 produce deadlock in llog code.
819 Details : If llog thread has last reference to obd and call class_import_put
820 this produce deadlock because llog_cleanup_commit_master wait when
821 last llog_commit_thread exited, but this never success because was
822 called from llog_commit_thread.
826 Description: allow userland application know is lost one of stripes.
827 Details : fill lvb_blocks with error code on ost and return it to
828 application if error flag found.
832 Description: NULL lov_tgts causing MDS oops
833 Details : more safe checks for NULL lov_tgts for avoid oops.
835 Severity : enhancement
837 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
841 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
842 Details : If an OST is down the MDS will hang indefinitely in
843 obd_statfs_rqset() waiting for the statfs data. While for
844 MDS QOS usage of statfs, it should not stuck in waiting.
846 Severity : enhancement
848 Description: remote_acl support
849 Details : Support ACL-based permission check for remote user.
850 Support setfacl/getfacl for remote user with the utils
851 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
852 the system "{s,g}etfacl" utils.
854 Severity : enhancement
856 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
858 Severity : enhancement
860 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
864 Description: make mgs_setparam() handle fsname containing dash
865 Details : fsname containing a dash does not work with lctl conf_param
867 Severity : enhancement
869 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
871 Severity : enhancement
873 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
877 Frequency : rare, at shutdown
878 Description: access already free / zero obd_namespace.
879 Details : if client_disconnect_export was called without force flag set,
880 and exist connect request in flight, this can produce access to
881 NULL pointer (or already free pointer) when connect_interpret
882 store ocd flags in obd_namespace.
886 Frequency : only at startup
887 Description: not alloc memory with spinlock held.
888 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
889 if any spinlock held.
891 Severity : enhancement
893 Description: make lustre randomly fail allocating memory
894 Details : Make lustre randomly failed allocating memory for testing purpose.
896 Severity : enhancement
898 Description: lost problems with lov objid file
899 Details : Fixes some scability and access to not inited memory problems
900 in work with lov objdid file.
905 Description: lfs find does not continue on file error
906 Details : Continue other files processing when a file/dir is absent.
910 Description: Inconsistent usage of lustre_pack_reply()
911 Details : Standardize the usage of lustre_pack_reply() such that it
912 always generate a CERROR on failure.
915 Frequency : occasional
917 Description: MDS hang or stay in waiting lock
918 Details : If client receive lock with CBPENDING flag ldlm need send lock
919 cancel as separate rpc, to avoid situation when cancel request
920 can't processed due all i/o threads stay in wait lock.
924 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
928 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
932 Description: Update to SLES9 kernel-2.6.5-7.287.3.
936 Description: Update to RHEL4 latest kernel.
938 Severity : enhancement
940 Description: Build SLES10 patchless client fails
941 Details : The configure was broken by run ./configure with
942 --with-linux-obj=.... argument for patchless client. When the
943 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
944 can't search header adequately. Use absolute path such as
945 -I($LINUX)/include instead.
949 Description: interrupt oig_wait produce painc on resend.
950 Details : brw_redo_request can be used for resend requests from ptlrpcd and
951 private set, and this produce situation when rq_ptlrpcd_data not
952 copyed to new allocated request and triggered LBUG on assert
953 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
954 ptlrpcd set if request is changed and can be safety changed to use
959 Description: LASSERT_{REQ,REP}SWAB macros are buggy
960 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
961 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
962 these macros and replace them with their difinitions instead.
966 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
967 Details : Update which_patch & target file for SLES10 latest kernel.
970 Description: Remove client patches from SLES 10 kernel.
971 Details : This causes SLES 10 clients to behave as patchless clients
972 even on a Lustre-patched (server) kernel.
974 Severity : enhancement
976 Description: self-adjustable client's lru lists
977 Details : use adaptive algorithm for managing client cached locks lru
978 lists according to current server load, other client's work
979 pattern, memory activities, etc. Both, server and client
980 side namespaces provide number of proc tunables for controlling
983 Severity : enhancement
985 Description: light-weight GSS support
986 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
987 and reduce performance overhead.
989 Severity : enhancement
991 Description: Linux keyring support
992 Details : Support using service of Linux keyring for Lustre GSS internal
993 context refresh/cache mechanism.
997 Description: Fix errors in lfs documentation
998 Details : Fixes man pages
1002 Description: don't use GFP_* in generic Lustre code.
1003 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1007 Description: obdclass is limited by single OBD_ALLOC(idarray)
1008 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1012 Description: warnings with build patchless client with vanila 2.6.19 and up
1013 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1014 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1018 Frequency : Only for SLES
1019 Description: sanity_quota fail test_1
1020 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1021 file, which makes TSTID[2] inunique.
1025 Frequency : Only for SLES9
1026 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1027 Details : Update target/ChangeLog/which_patch .
1031 Frequency : Only for test_10 in sanity-quota.sh
1032 Description: a bug in quota test code
1033 Details : checking a test flag in a wrong place
1037 Frequency : Only for test_13 in sanity-quota.sh
1038 Description: a bug in quota test script
1039 Details : shouldn't check overfull allocated quota
1043 Description: jbd statistics
1044 Details : Port older jbd statistics patch for sles10
1046 Severity : enhancement
1048 Description: Add printing inode info into message about error in writepage.
1053 Description: reply_lock_interpret crash due to race with it and lock cancel.
1054 Details : Do not replay locks that are being cancelled. Do not reference
1055 locks by their address during replay, just by their handle.
1059 Frequency : When flocks are used.
1060 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1061 Details : Flock locks might destroy just granted lock if it could be merged
1062 with another existing flock, this is done in completion handler,
1063 so teach ldlm_cli_enquque_fini that this is a valid case for
1069 Description: Oops in read and write path when failing to allocate lock.
1070 Details : Check if lock allocation failed and return error back.
1074 Description: lstripe command fails for valid OST index
1075 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1076 instead of lov->desc.ld_active_tgt_count.
1080 Description: sanity.sh failed test 103
1081 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1086 Description: df doesn't work properly if diskfs blocksize != 4K
1087 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1091 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1092 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1094 Severity : enhancement
1096 Description: Adaptive timeouts
1097 Details : RPC timeouts adapt to changing server load and network
1098 conditions to reduce resend attempts and improve recovery time.
1102 Description: llapi_file_create() does not allow some changes
1103 Details : add llapi_file_open() that allows specifying the mode and
1104 open flags, and also returns an open file handle.
1108 Description: merge and cleanup kernel patches.
1109 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1113 Description: Add journal checksum support.(Kernel part)
1114 Details : The journal checksum feature adds two new flags i.e
1115 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1116 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1117 indicates that the commit block contains the checksum for
1118 the blocks described by the descriptor blocks. Now commit
1119 record can be sent to disk without waiting for descriptor
1120 blocks to be written to disk. This behavior is controlled
1121 using JBD2_FEATURE_ASYNC_COMMIT flag.
1125 Description: OSS needs mutliple precreate threads
1126 Details : Add ability to start more than one create thread per OSS.
1130 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1131 Details : execute lfs setstripe on client
1135 Description: mds_obd_create error creating tmp object
1136 Details : When the user sets quota on root, llog will be affected and can't
1137 create files and write files.
1140 Frequency : Always on ia64 patchless client, and possibly others.
1142 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1143 Details : This allows the patchless client to be loaded on architectures
1144 without this export.
1148 Description: RedHat Update kernel for RHEL5
1149 Details : Add the kernel config file for RHEL5.
1153 Description: RedHat Update kernel for RHEL5
1154 Details : Modify the kernel config file more closer RHEL5.
1158 Description: Build failure against Centos5 (RHEL5)
1159 Details : Define PAGE_SIZE when it isn't present.
1163 Description: test_7 on sanity-quota.sh can't be used on separate machines
1164 Details : add facet to handle it
1168 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1169 Details : with NFS, the anon dentry's parent was set to itself in
1170 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1171 corresponding dentry other than use rec->ur_name.
1173 Severity : enhancement
1175 Description: lfs setstripe enhancement
1176 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1180 Description: enable data checksumming by default
1181 Details : enable checksum by default, allow --disable-checksum
1182 configure option and "-o nochecksum" mount option. Checksums
1183 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1187 Description: lustre support for RHEL5
1188 Details : Add support for RHEL5.
1192 Description: Client eviction due to failover config
1193 Details : after a connection loss, the lustre client should attempt to
1194 reconnect to the last active server first before trying the
1195 other potential connections.
1199 Description: when mds and osts use different quota unit(32bit and 64bit),
1200 quota will be released repeatly.
1201 Details : void sending multiple quota reqs to mds, which will keep the
1202 status between the reqs.
1206 Description: osts not allocated evenly to files
1207 Details : change the condition to increase offset_idx
1210 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1212 Description: Data corruption for OSTs that are formatted larger than 2TB
1214 Details : When generating the bio request for lustre file writes the
1215 sector number would overflow a temporary variable before being
1216 used for the IO. The data reads correctly from Lustre (which
1217 will overflow in a similar manner) but other file data or
1218 filesystem metadata may be corrupted in some cases.
1222 Description: Tune the kernel for good SCSI performance.
1223 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1224 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1229 Description: rewrite ext2-derived code in obdclass/uuid.c
1230 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1231 from scratch preserving functionality.
1235 Description: "lfs find -obd UUID" prints directories
1236 Details : "lfs find -obd UUID" will return all directory names instead
1237 of just file names. It is incorrect because the directories
1238 do not reside on the OSTs.
1242 Description: use i_size_read and i_size_write in 2.6 port
1243 Details : replace inode->i_size access with i_size_read/write()
1246 Frequency : when removing large files
1248 Description: scheduling issue during removal of large Lustre files
1249 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1250 It causes scheduling issues when removing large files (17TB in the
1254 Frequency : only with liblustre clients on XT3
1256 Description: evictions taking too long
1257 Details : allow llrd to evict clients directly on OSTs
1260 Frequency : only on ppc
1262 Description: /proc/fs/lustre/devices broken on ppc
1263 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1264 We should make sure the type of 'index' is loff_t and then cast
1265 explicitly as needed below. Do not assign an explicitly cast
1269 Frequency : only for rhel5
1271 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1272 Details : Modify the target file & which_kernel.
1274 Severity : enhancement
1276 Description: omit set fsid for export NFS
1277 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1278 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1282 Frequency : Always, for kernels after 2.6.16
1283 Description: Fix warning idr_remove called for id=.. which is not allocated.
1284 Details : Last kernels save old s_dev before kill super and not allow
1285 to restore from callback - restore it before call kill_anon_super.
1289 Description: buffer overruns could theoretically occur
1290 Details : llapi_semantic_traverse() modifies the "path" argument by
1291 appending values to the end of the origin string, and a
1292 overrun may occur. Adding buffer overrun check in liblustreapi.
1296 Description: Fix error on 'ls .' at the top of the Lustre mount.
1297 Details : Don't revalidate dentry if it is a root dentry.
1301 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1302 Details : Modify vm-tunables-rhel4.patch.
1306 Description: Kernel config for 2.6.18-vanilla.
1307 Details : Modify targets/2.6-vanilla.target.in.
1308 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1309 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1310 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1311 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1315 Description: improve handling recoverable errors
1316 Details : if request processig with error which can be recoverable on server
1317 request should be resend, otherwise page released from cache and
1322 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1323 Details : Modify target file & which_patch.
1324 A flaw was found in the IA32 system call emulation provided
1325 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1326 value could be stored in the %RAX register, which could trigger an
1327 out-of-bounds system call table access. An untrusted local user
1328 could exploit this flaw to run code in the kernel
1329 (ie a root privilege escalation). (CVE-2007-4573).
1333 Description: O_DIRECT bypasses client statistics.
1334 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1335 read_ahead_stats were updated. copied the stats section from
1336 osc_send_oap_rpc() into async_internal().
1340 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1344 Description: change order of libsysio includes
1345 Details : '#include sysio.h' should always come before '#include xtio.h'
1349 Description: handle "serious error: objid * already exists" more gracefully
1350 Details : If LAST_ID value on disk is smaller than the objects existing in
1351 the O/0/d* directories, it indicates disk corruption and causes an
1352 LBUG(). If the object is 0-length, then we should use the existing
1353 object. This will help to avoid a full fsck in most cases.
1355 Severity : enhancement
1357 Description: adapt the lustre_config script to support the upgrade case
1358 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1359 server targets to 1.6 in parallel.
1363 Description: To avoid grant space > avaible space when the disk is almost
1364 full. Without this patch you might see the error "grant XXXX >
1365 available" or some LBUG about grant, when the disk is almost
1367 Details : In filter_check_grant, for non_grant cache write, we should
1368 check the left space by if (*left > ungranted + bytes), instead
1369 of (*left > ungranted), because only we are sure the left space
1370 is enough for another "bytes", then the ungrant space should be
1371 increase. In client, we should update cl_avail_grant only there
1372 is OBD_MD_FLGRANT in the reply.
1376 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1377 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1378 described in CVE-2007-4573. This problem could allow untrusted
1379 local users to gain root access.
1382 Frequency : when using O_DIRECT and quotas
1384 Description: Incorrect file ownership on O_DIRECT output files
1385 Details : block usage reported by 'lfs quota' does not take into account
1386 files that have been written with O_DIRECT.
1391 Description: touch file failed when fs is not full
1392 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1393 otherwise we can get ENOSP while fs is not full.
1397 Description: parallel lock callbacks
1398 Details : Instead of sending blocking and completion callbacks as separated
1399 requests, adding them to a set and sending in parallel.
1402 Frequency : only for Cray XT3
1403 Bugzilla : 12829/13455
1404 Description: Changing primary group doesn't change the group lustre assigns to
1406 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1407 overridden with the primary group provided by the group upcall,
1408 whereas the supplied fsgid can be trusted if it is in the list of
1409 supplementary groups returned by the group upcall.
1411 Severity : enhancement
1413 Description: Allow masking D_WARNING, D_ERROR messages from console
1414 Details : Console messages can now be disabled via lnet.printk.
1418 Description: User code with malformed file open parameter crashes client node
1419 Details : Before packing join_file req, all the related reference should be
1420 checked carefully in case some malformed flags cause fake
1421 join_file req on client.
1425 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1426 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1427 mode is equal to the granted mode. However, it should also check
1428 that LDLM_FL_CP_REQD is not set.
1432 Description: Heavy nfs access might result in deadlocks
1433 Details : After ELC code landed, it is now improper to enqueue any mds
1434 locks under och_sem, because enqueue might want to decide to
1435 cancel open locks for same inode we are holding och_sem for.
1439 Description: Client eviction while running blogbench
1440 Details : A lot of unlink operations with concurrent I/O can lead to a
1441 deadlock causing evictions. To address the problem, the number of
1442 oustanding OST_DESTROY requests is now throttled to
1443 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1444 callbacks are processed in priority.
1448 Description: enable ACLs on MDS by default
1449 Details : ACLs must be enabled on MDS by default.
1452 Frequency : PPC/PPC64 only
1454 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1455 Details : fix duplicated definitions between asm-ppc64/types.h and
1456 lustre_types.h on PPC.
1459 Frequency : PPC/PPC64 only
1461 Description: asm-ppc/segment.h does not exist
1462 Details : fix compile issue on PPC.
1466 Description: better handle error messages in extents code
1469 Frequency : RHEL4 only
1471 Description: mkfs is very slow on IA64/RHEL4
1472 Details : A performance regression has been discovered in the MPT Fusion
1473 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1474 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1475 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1477 Severity : enhancement
1479 Description: SNMP support enhancement
1480 Details : Adding total number of sampled request for an MDS node in snmp
1483 Severity : enhancement
1485 Description: Optimize ldlm waiting list processing for PR extent locks
1486 Details : When processing waiting list for read extent lock and meeting read
1487 lock that is same or wider to it that is not contended, skip
1488 processing rest of the list and immediatelly return current
1489 status of conflictness, since we are guaranteed there are no
1490 conflicting locks in the rest of the list.
1494 Description: Time out and refuse to reconnect
1495 Details : When the failover node is the primary node, it is possible
1496 to have two identical connections in imp_conn_list. We must
1497 compare not conn's pointers but NIDs, otherwise we can defeat
1498 connection throttling.
1502 Description: port llog fixes from b1_6 into HEAD
1503 Details : Port llog reference couting and some llog cleanups from b1_6
1504 (bug 10800) into HEAD, for protect from panic and access to already
1505 free llog structures.
1509 Description: Detect stride IO mode in read-ahead
1510 Details : When a client does stride read, read-ahead should detect that and
1511 read-ahead pages according to the detected stride pattern.
1515 Description: data checksumming impacts single node performance
1516 Details : add support for several checksum algorithm. Currently, only CRC32
1517 and Adler-32 are supported. The checksum type can be changed on
1518 the fly via /proc/fs/lustre/osc/*/checksum_type.
1522 Description: use adler32 for page checksums
1523 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1528 Description: build for x2 fails
1529 Details : fix compile issue on Cray systems.
1533 Description: Properly match for duplicate locks
1534 Details : Due to different lock order from skiplists code, we need to
1535 traverse entire list for now
1538 Frequency : only on PPC/SLES10
1540 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1541 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1542 be defined. Add a hack in mkfs_lustre.c to work around this compile
1547 Description: LASSERT on MDS when client holding flock lock dies
1548 Details : ldlm pool logic depends on number of granted locks equal to
1549 number of released locks which is not true for flock locks, so
1550 just exclude such locks from consideration.
1554 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1555 Details : Use fsfilt_sync() for both the whole filesystem sync and
1556 individual file sync to eliminate dangerous inode locking
1557 with I_LOCK that can lead to a deadlock.
1561 Description: performance in 1.6.3
1562 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1563 sure that 1MB requests are not fragmented by the block layer.
1565 Severity : enhancement
1567 Description: organize the server-side client stats on per-nid basis
1568 Details : Change the structure of stats under obdfilter and mds to
1576 The "uuid"s file would list the uuids of _active_ exports.
1577 And the clear entry is to clear all stats and stale nids.
1579 Severity : enhancement
1581 Description: eliminate client locks in face of contention
1582 Details : file contention detection and lockless i/o implementation
1583 for contended files.
1587 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1590 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1592 Description: MDS may lose file striping (and hence file data) in some cases
1593 Details : If there are additional extended attributes stored on the MDS,
1594 in particular ACLs, SELinux, or user attributes (if user_xattr
1595 is specified for the client mount options) then there is a risk
1596 of attribute loss. Additionally, the Lustre file striping
1597 needs to be larger than default (e.g. striped over all OSTs),
1598 and an additional attribute must be stored initially in the
1599 inode and then increase in size enough to be moved to the
1600 external attribute block (e.g. ACL growing in size) for file
1605 Description: skiplist implementation simplification
1606 Details : skiplists are used to group compatible locks on granted list
1607 that was implemented as tracking first and last lock of each
1608 lock group the patch changes that to using doubly linked lists
1612 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1613 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1614 because LASSERT on both the data supplied by a client, and the
1615 data on disk is dangerous and incorrect.
1617 Severity : enhancement
1619 Description: Slow truncate/writes to huge files at high offsets.
1620 Details : Directly associate cached pages to lock that protect those pages,
1621 this allows us to quickly find what pages to write and remove
1622 once lock callback is received.
1626 Description: more ldlm soft lockups
1627 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1628 starve other threads from the resource lock for a long time in
1629 case of long waiting queue, so change the debug level from
1630 D_OTHER to the less frequently used D_INFO.
1632 Severity : enhancement
1634 Description: add -gid, -group, -uid, -user options to lfs find
1638 Description: Hung threads in invalidate_inode_pages2_range
1639 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1640 once one is completed. As a result, some RPCs are stuck in the
1641 queue and are never sent.
1645 Description: filter threads hungs on waiting journal commit
1646 Details : Cleanup filter group llog code, then only filter group llog will
1647 be only created in the MDS/OST syncing process.
1651 Description: Procfs and llog threads access destoryed import sometimes.
1652 Details : Sync the import destoryed process with procfs and llog threads by
1653 the import refcount and semaphore.
1655 Severity : enhancement
1657 Description: openlock cache of b1_6 port to HEAD
1662 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1663 Details : If insertion of an extent fails, then discard the inode
1664 preallocation and free data blocks else it can lead to duplicate
1669 Description: don't always update ctime in ext3_xattr_set_handle()
1670 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1671 In some cases the ctime should not be updated, for example for
1672 2.0->1.8 compatibility it is necessary to delete an xattr and it
1673 should not update the ctime.
1677 Bugzilla : 15713/16362
1678 Description: Assertion in iopen_connect_dentry in 1.6.3
1679 Details : looking up an inode via iopen with the wrong generation number can
1680 populate the dcache with a disconneced dentry while the inode
1681 number is in the process of being reallocated. This causes an
1682 assertion failure in iopen since the inode's dentry list contains
1683 both a connected and disconnected dentry.
1687 Description: assertion failure in ldlm_handle2lock()
1688 Details : fix a race between class_handle_unhash() and class_handle2object()
1689 introduced in lustre 1.6.5 by bug 13622.
1694 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1695 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1700 Description: oops in sd_iostats_seq_show()
1701 Details : unloading/reloading the scsi low level driver triggers a kernel
1702 bug when trying to access the sd iostat file.
1707 Description: Kernel panics during QLogic driver reload
1708 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1709 patch, causing memory corruption.
1714 Description: journal_dev option does not work in b1_6
1715 Details : pass mount option during pre-mount.
1717 Severity : enhancement
1719 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1720 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1721 extent information of a file. It can be used to map logical blocks
1722 in a file to physical blocks in the block device.
1726 Description: LDLM soft lockups - improvement
1727 Details : It is be possible to send the lock handle along with each read
1728 or write request because the client is already doing a lock match
1729 itself so there isn't any reason the OST should have to re-do that
1733 Frequency : only on Cray X2
1735 Description: X2 build failures
1736 Details : fix build failures on Cray X2.
1740 Description: xid & resent requests
1741 Details : Initialize RPC XID from clock at startup (randomly if clock is
1744 Severity : enhancement
1746 Description: Add lustre_start utility to start or stop multiple Lustre servers
1751 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1752 Details : In case of memory pressure, list_del() can be called twice on
1753 req->rq_history_list, causing a kernel oops.
1757 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1758 Details : fix stack overflow in the distributed lock manager by defering
1759 export eviction after a failed AST to the elt thread instead of
1760 handling it in the dlm interpret routine.
1764 Description: Convert some comments to new format.
1765 Details : Update documenting comments to match doxygen conventions.
1769 Description: Grammar fixes.
1770 Details : A couple of trivial spelling fixes.
1774 Description: OSD_COUNTERS-mandatory
1775 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1776 to simplify the code.
1780 Description: simplify cmm_device freeing logic.
1781 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1785 Description: Add lockdep support to dt_object_operations locking interface.
1786 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1787 parameter indicating lock ordering. Update mdd code to use new
1792 Description: Introduce failloc constants for lockless IO tests.
1793 Details : Add two new failloc constants to test lockless IO. Only one of
1794 them in implemented---another is checked in yet to be landed
1799 Description: Add lockdep support for inode mutex.
1800 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1801 in the situations where Lustre has to lock more than one inode
1806 Description: Add optional invariants checking support.
1807 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1808 configure switch. This macro is to be used for consistency and
1809 sanity checks that are too expensive to be left in `production'
1814 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1815 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1820 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1821 Details : Use spin_lock_nested() in (the only) situation where more than
1822 one ldlm_lock is locked simultaneously. Also, fix possible
1823 dead-lock in ldlm_lock_change_resource() by enforcing particular
1828 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1829 Details : Instead of specifying each ldlm_lock call-back through separate
1830 parameter, wrap them into struct ldlm_callback_suite.
1834 Description: Kill join_lru obd method and its callers.
1835 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1836 in memory---a requirement implemented through ->o_join_lru() obd
1837 method in HEAD. Remove this method and its users.
1841 Description: Add asynchronous ldlm ENQUEUE completion handler.
1842 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1843 case that stock ldlm_completion_ast() cannot handle as it waits
1844 until lock is granted. Introduce new ldlm_completion_ast_async()
1845 for this. Also comment ldlm_completion_ast().
1849 Description: ldlm_error <-> errno conversion.
1850 Details : Add functions to map (rather arbitrary) between LDLM error codes
1851 and standard errno values. CLIO needs this to prevent LDLM specific
1852 constants from escaping ldlm and osc.
1856 Description: Kill unused ldlm_handle2lock_ns() function.
1857 Details : Kill unused ldlm_handle2lock_ns() function.
1861 Description: Add lu_ref support to ldlm_lock
1862 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1863 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1864 and ldlm_resource. LDLM interface has to be changed, because code that
1865 releases a reference on a lock, has to "know" what reference this is.
1866 In the most frequent case
1868 lock = ldlm_handle2lock(handle);
1870 LDLM_LOCK_PUT(lock);
1872 no changes are required. When any other reference (received _not_ from
1873 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1874 instead of LDLM_LOCK_PUT().
1876 Arguably, changes are pervasive, and interface requires some discipline
1877 for proper use. On the other hand, it was very instrumental in finding
1878 a few leaked lock references.
1882 Description: Add ldlm_lock_addref_try().
1883 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1884 attempts to addref a lock that might be being canceled
1889 Description: Add ldlm_weigh_callback().
1890 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1891 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1892 hard-coded `number of pages' logic.
1896 Description: Add lockdep annotations to llog code.
1897 Details : Use appropriately tagged _nested() locking calls in the places
1898 where llog takes more than one ->lgh_lock lock.
1902 Description: Add loi_kms_set().
1903 Details : Wrap kms updates into a helper function.
1907 Description: Constify instances of struct lsm_operations.
1908 Details : Constify instances of struct lsm_operations.
1912 Description: lu_conf support.
1913 Details : On a server, a file system object is uniquely identified
1914 by a fid, which is sufficient to locate and load all object
1915 state (inode). On a client, on the other hand, more data are
1916 necessary instantiate an object. Change lu_object_find() and
1917 friends to take additional `lu_conf' argument describing object.
1918 Typically this includes layout information.
1922 Description: lu_context fixes.
1923 Details : Introduce new lu_context functions that are needed on the client
1924 side, where some system threads (ptlrpcd) are shared by multiple
1925 modules, and so cannot be stopped during module shutdown.
1929 Description: Add start and stop methods to lu_device_type_operations.
1930 Details : Introduce two new methods in lu_device_type_operations, that are
1931 invoked when first instance of a given type is created and last one
1932 is destroyed respectively. This is need by CLIO.
1936 Description: Add lu_ref support to struct lu_device.
1937 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1938 track leaked references.
1942 Description: Introduce lu_kmem_descr.
1943 Details : lu_kmem_descr and its companion interface allow to create
1944 and destroy a number of kmem caches at once.
1948 Description: Fix lu_object finalization race.
1949 Details : Fix a race between lu_object_find() finding an object and its
1950 concurrent finalization. This race is (most likely) not possible
1951 on the server, but might happen on the client.
1955 Description: Introduce lu_ref interface.
1956 Details : lu_ref is a debugging module allowing to track references to
1957 a given object. It is quite cpu expensive, and has to be
1958 explicitly enabled with --enable-lu_ref. See usage description
1963 Description: Factor lu_site procfs stats into a separate function.
1964 Details : Separate lu_site stats printing code into a separate function
1965 that can be reused on a client.
1969 Description: Constify instances of struct {lu,dt,md}_device_operations.
1970 Details : Constify instances of struct {lu,dt,md}_device_operations.
1974 Description: Introduce struct md_site and move meta-data specific parts of
1975 struct lu_site here.
1976 Details : Move md-specific fields out of struct lu_site into special struct
1977 md_site, so that lu_site can be used on a client.
1981 Description: Kill mdd_lov_destroy().
1982 Details : Remove unused mdd code.
1986 Description: Add st_block checking to multistat.c.
1987 Details : Add st_block checking to multistat.c.
1991 Description: Add lu_ref support to struct obd_device.
1992 Details : Add lu_ref tracking to obd_device.
1996 Description: Kill obd_set_fail_loc().
1997 Details : Remove unused code.
2001 Description: Add special type for ptlrpc_request interpret functions.
2002 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2003 there. Actual usage will be in CLIO.
2007 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2008 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2009 doesn't work with lockdep.
2013 Description: Add rwv.c test program.
2014 Details : New testing program exercising readv(2) and writev(2) (Qian).
2018 Description: Add sendfile.c test program.
2019 Details : New testing program exercising sendfile(2) (Jay).
2023 Description: Ratelimit a message that can be very frequent.
2024 Details : Ratelimit a memory allocation failure message that can
2029 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2030 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2034 Description: Liblustre build fixes.
2035 Details : Liblustre build fixes.
2039 Description: libcfs: add cfs_{need,cond}_resched() interface.
2040 Details : libcfs: add cfs_{need,cond}_resched() definition and
2041 implementations for Linux, NT, and liblustre.
2043 Severity : enhancement
2045 Description: More exported tunables for mballoc
2046 Details : Add support for tunable preallocation window and new tunables for
2047 large/small requests
2051 Description: Detect corruption of block bitmap and checking for preallocations
2052 Details : Checks validity of on-disk block bitmap. Also it does better
2053 checking of number of applied preallocations. When corruption is
2054 found, it turns filesystem readonly to prevent further corruptions.
2058 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2059 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2060 window to 0. If it does hit cache miss, and read-pattern is still
2061 stride-io mode, does not reset the stride window, but also does not
2062 increase the stride window length in this case.
2066 Frequency : only for big-endian servers
2067 Description: Check if system is big-endian while mounting fs with extents feature
2068 Details : Mounting a filesystem with extents feature will fail on big-endian
2069 systems since ext3-based ldiskfs is not supported on big-endian
2070 systems. This can be over-riden with "bigendian_extents" mount option.
2072 Severity : enhancement
2074 Description: The root squash functionality
2075 Details : A security feature, which is to prevent users from being able
2076 to mount lustre on their desktop, run as root, and delete
2077 all of the files in the filesystem. The goal is accomplished by
2078 remapping user id (UID) and group id (GID) of the root user to
2079 a UID and GID specified by the system administartor via Lustre
2080 configuration management server (MGS). The functionality also
2081 allows to specify sets of clients for which the remapping does
2086 Description: Excessive recovery window
2087 Details : With AT enabled, the recovery window can be excessively long (6000+
2088 seconds). To address this problem, we no longer use
2089 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2090 timeout no longer depends on the service time, it is set to
2091 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2092 time via pb_service_time.
2096 Description: Watchdog triggered on MDS failover
2097 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2098 the OSTs know that the MDS "UUID" can be reused for the same export
2099 from a different NID, so we do not need to wait for the export to be
2103 Frequency : rare, only if using MMP with Linux RAID
2105 Description: MMP doesn't work with Linux RAID
2106 Details : While using HA for Lustre servers with Linux RAID, it is possible
2107 that MMP will not detect multiple mounts. To make this work we
2108 need to unplug the device queue in RAID when the MMP block is being
2109 written. Also while reading the MMP block, we should read it from
2110 disk and not the cached one.
2112 Severity : enhancement
2114 Description: open file using fid
2115 Details : A file can be opened using just its fid, like
2116 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2119 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2122 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2123 Details : add proper locking for creation of jbd_4k slab cache
2127 Description: MMP check in ext3_remount() fails without displaying any error
2128 Details : When multiple mount protection fails during remount, proper error
2131 Severity : enhancement
2133 Description: Allow stripe size to be up to 4G-64k
2134 Details : Fix math logic to allow large stripe sizes.
2138 Description: add check for >8TB ldiskfs filesystems
2139 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2140 Don't allow >8TB ldiskfs filesystems to be mounted without
2141 force_over_8tb mount option
2146 Descriptoin: OSS complains "too many blocks" and crashes
2147 Details : In filter_commitrw_write():
2148 1. lquota_chkquota();
2149 2. obdo_from_inode();
2150 3. lquota_pending_commit();
2151 "2." has changed uid and gid in this case, then caused
2152 lquota_chkquota() and lquota_pending_commit() to use
2153 different uid and gid. That is the root of the bug.
2155 --------------------------------------------------------------------------------
2157 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2159 * Support for kernels:
2160 2.6.5-7.283 (SLES 9),
2161 2.6.9-55.EL (RHEL 4),
2162 2.6.16.46-0.14 (SLES 10),
2163 2.6.18.8 vanilla (kernel.org)
2164 * Client support for unpatched kernels:
2165 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2166 2.6.16 - 2.6.22 vanilla (kernel.org)
2167 * Due to recently discovered recovery problems, we do not recommend
2168 using patchless RHEL 4 clients with this or any earlier release.
2169 * Recommended e2fsprogs version: 1.39.cfs8
2170 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2171 * Starting with this release, the ldiskfs backing filesystem required
2172 by Lustre is now in its own package, lustre-ldiskfs. This package
2173 should be installed. It is versioned separately from Lustre and
2174 may be released separately in future.
2178 Description: block reactivating mgc import until all deactivates complete
2179 Details : Fix race when failing back MDT/MGS to itself (testing)
2182 Frequency : at statup only
2184 Description: mds_lov_synchronize race leads to various problems
2185 Details : simultaneous MDT->OST connections at startup can cause the
2186 sync to abort, leaving the OSC in a bad state.
2188 Severity : enhancement
2190 Description: add optional extra BUILD_VERSION info
2191 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2192 to override the lustre version.
2195 Frequency : 2.6.18 servers only
2197 Description: ll_kern_mount() doesn't release the module reference
2198 Details : The ldiskfs module reference count never drops down to 0
2199 because ll_kern_mount() doesn't release the module reference.
2204 Description: server LBUG when using old ost_num_threads parameter
2205 Details : Accept the old ost_num_threads parameter but warn that it
2206 is deprecated, and fix an off-by-one error that caused an LBUG.
2211 Description: Transient SCSI error results in persistent IO issue
2212 Details : iobuf->dr_error is not reinitialized to 0 between two
2216 Frequency : sometimes when underlying device returns I/O errors
2218 Description: OSTs not going read-only during write failures
2219 Details : OSTs are not remounted read-only when the journal commit threads
2220 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2221 instead of the ext3 wrappers.
2225 Description: poor connect scaling with increasing client count
2226 Details : Don't run filter_grant_sanity_check for more than 100 exports
2227 to improve scaling for large numbers of clients.
2230 Frequency : SLES10 only
2232 Description: sanity-quota.sh quotacheck failed: rc = -22
2233 Details : Quotas cannot be enabled on SLES10.
2236 Frequency : liblustre clients only
2238 Description: getdirentries does not give error when run on compute nodes
2239 Details : getdirentries does not fail when the size specified as an argument
2240 is too small to contain at least one entry
2242 Severity : enhancement
2244 Description: Add LNET router traceability for debug purposes
2245 Details : If a checksum failure occurs with a router as part of the
2246 IO path, the NID of the last router that forwarded the bulk data
2247 is printed so it can be identified.
2252 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2253 Details : Due to a race condition, liblustre clients were occasionally
2254 evicted incorrectly.
2256 Severity : enhancement
2258 Description: lfs setstripe use optional parameters instead of postional
2261 Severity : enhancement
2263 Description: Nanosecond timestamp support for ldiskfs
2264 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2265 resolution timestamps. There is not yet support for this at
2266 the Lustre filesystem level.
2269 Frequency : during server recovery
2271 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2272 Details : request with rq_no_resend flag not awake l_wait_event if they get
2276 Frequency : nfs export on patchless client
2278 Description: connectathon hang when test nfs export over patchless client
2279 Details : Disconnected dentry cannot be found with lookup, so we do not need
2280 to unhash it or make it invalid
2283 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2284 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2285 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2290 Description: open req refcounting wrong on reconnect
2291 Details : If reconnect happened between getting open reply from server and
2292 call to mdc_set_replay_data in ll_file_open, we will schedule
2293 replay for unreferenced request that we are about to free.
2294 Subsequent close will crash in variety of ways.
2295 Check that request is still eligible for replay in
2296 mdc_set_replay_data().
2301 Description: disable writes to filesystem when reading health_check file
2302 Details : the default for reading the health_check proc file has changed
2303 to NOT do a journal transaction and write to disk, because this
2304 can cause reads of the /proc file to hang and block HA state
2305 checking on a healthy but otherwise heavily loaded system. It
2306 is possible to return to the previous behaviour during configure
2307 with --enable-health-write.
2309 Severity : enhancement
2311 Description: 64-bit inode version
2312 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2313 to the inode. This will be required for version-based recovery.
2318 Description: MDS fails to start if a duplicate client export is detected
2319 Details : in some rare cases it was possible for a client to connect to
2320 an MDS multiple times. Upon recovery the MDS would detect this
2321 and fail during startup. Handle this more gracefully.
2323 Severity : enhancement
2325 Description: Add -o localflock option to simulate old noflock
2327 Details : This will achieve local-only flock/fcntl locks
2333 Description: log_commit_thread vs filter_destroy race leads to crash
2334 Details : Take import reference before releasing llog record semaphore
2339 Description: Wrong request locking in request set processing
2340 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2341 imp_delayed_list, in this place should be used imp_lock.
2344 Frequency : when reconnection
2346 Description: Grant Leak when osc reconnect to OST
2347 Details : When osc reconnect ost, OST(filter) should check whether it
2348 should grant more space to client by comparing fed_grant and
2349 cl_avail_grant, and return the granted space to client instead
2350 of "new granted" space, because client will call osc_init_grant
2351 to update the client grant space info.
2354 Frequency : when client reconnect to OST
2356 Description: Grant Leak when osc do resend and replay bulk write
2357 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2358 bulk write request, because the grant info while be sync between
2359 OSC and OST when reconnect, and we should ignore the grant info
2360 these of resend/replay write req.
2365 Description: Grant space more than avaiable left space sometimes.
2366 Details : When then OST is about to be full, if two bulk writing from
2367 different clients came to OST. Accord the avaliable space of the
2368 OST, the first req should be permitted, and the second one
2369 should be denied by ENOSPC. But if the seconde arrived before
2370 the first one is commited. The OST might wrongly permit second
2371 writing, which will cause grant space > avaiable space.
2374 Frequency : when client is evicted
2376 Description: Grant might be wrongly erased when osc is evicted by OST
2377 Details : when the import is evicted by server, it will fork another
2378 thread ptlrpc_invalidate_import_thread to invalidate the
2379 import, where the grant will be set to 0. While the original
2380 thread will update the grant it got when connecting. So if
2381 the former happened latter, the grant will be wrongly errased
2382 because of this race.
2387 Description: Checking Stale with correct fid
2388 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2389 to check whether it is stale, because sometimes, we want the
2390 enqueue happened anyway, and op_data.fid2 will not be initialized.
2392 Severity : enhancement
2394 Description: update patchless client
2395 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2398 Frequency : only with 2.4 kernel
2400 Description: random memory corruption
2401 Details : size of struct ll_inode_info is to big for union inode.u and this
2402 can be cause of random memory corruption.
2407 Description: Memory leak in recovery
2408 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2409 It should also check obd_fail before fsfilt_start, otherwise if
2410 fsfilt_start return -EROFS,(failover mds during mds recovery).
2411 then the req will return with repmsg->transno = 0 and rc = EROFS.
2412 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2413 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2414 be freed no matter whether fsfilt_commit success or not.
2417 Frequency : only with huge count clients
2419 Description: Prevents from taking the superblock lock in llap_from_page for
2421 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2422 for llap_from_page call in ll_removepage prevents from taking the
2423 superblock lock for a soon died page.
2428 Description: Not check open intent error before release open handle
2429 Details : in some rare cases, the open intent error is not checked before
2430 release open handle, which may cause
2431 ASSERTION(open_req->rq_transno != 0), because it tries to release
2432 the failed open handle.
2437 Description: Set cat log bitmap only after create log success.
2438 Details : in some rare cases, the cat log bitmap is set too early. and it
2439 should be set only after create log success.
2443 Description: Accessing a block bevice can re-enable I/O when Lustre is
2444 tearing down a device.
2445 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2450 Description: service threads may hog cpus when there are a lot of requests
2451 Details : Insert cond_resched to give other threads a chance to use some CPU
2456 Description: the cat log was not initialized in recovery
2457 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2458 unlink log on mds will not be initialized until mds post
2459 recovery. And also in mds post recovery, the unlink log will
2460 initialization will be done asynchronausly, so there will be race
2461 between add unlink log and unlink log initialization.
2465 Description: brw_stats were being printed incorrectly
2466 Details : brw_stats were being printed as log2 but all of them were not
2467 recorded as log2. Also remove some code duplication arising from
2468 filter_tally_{read,write}.
2472 Frequency : rare, only in recovery.
2473 Description: ASSERTION(req->rq_type != LI_POISON) failed
2474 Details : imp_lock should be held while iterating over imp_sending_list for
2475 prevent destroy request after get timeout in ptlrpc_queue_wait.
2479 Description: replay-single.sh test 52 fails
2480 Details : A lock's skiplist need to be cleanup when it being unlinked
2481 from its resource list.
2485 Description: Short directio read returns full requested size rather than
2487 Details : Direct I/O operations should return actual amount of bytes
2488 transferred rather than requested size.
2490 Severity : enhancement
2492 Description: metadata RPC reduction (e.g. for rm performance)
2493 Details : decrease the amount of synchronous RPC between clients and servers
2494 by canceling conflicing lock before the operation on the client
2495 and packing thier handles into the main operation RPC to server.
2497 Severity : enhancement
2499 Description: Async OSC create to avoid the blocking unnecessarily.
2500 Details : If a OST has no remain object, system will block on the creating
2501 when need to create a new object on this OST. Now, ways use
2502 pre-created objects when available, instead of blocking on an
2503 empty osc while others are not empty. If we must block, we block
2504 for the shortest possible period of time.
2508 Description: improve handling recoverable errors
2509 Details : if request processig with error which can be recoverable on server
2510 request should be resend, otherwise page released from cache and
2513 Severity : enhancement
2515 Description: refine locking for avoid write wrong info into lov_objid file
2516 Details : fix possible races with add new target and write/update data in
2519 --------------------------------------------------------------------------------
2521 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2526 Frequency : on some architectures
2528 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2529 Details : Uninitialized flags sometimes cause configuration commands to
2533 Frequency : patchless clients only
2535 Description: missing __iget() symbol export
2536 Details : The __iget() symbol export is missing. To avoid the need for
2537 this on patchless clients the deathrow inode reaper is turned
2538 off, and we depend on the VM to clean up old inodes. This
2539 dependency was during via the fix for bug 12181.
2541 --------------------------------------------------------------------------------
2543 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2545 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2546 INTEROPERATE with older versions automatically. In many cases a
2547 special upgrade step is needed. Please read the
2548 user documentation before upgrading any part of a live system.
2549 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2550 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2551 * WARNING: Lustre configuration and startup changes are required with
2552 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2556 Severity : enhancement
2558 Description: MountConf
2559 Details : Lustre configuration is now managed via mkfs and mount
2560 commands instead of lmc and lconf. New obd types (MGS, MGC)
2561 are added for dynamic configuration management. See
2562 https://mail.clusterfs.com/wikis/lustre/MountConf for
2565 Severity : enhancement
2567 Description: dynamic OST addition
2568 Details : OSTs can now be added to a live filesystem
2570 Severity : enhancement
2572 Description: startup order invariance
2573 Details : MDTs and OSTs can be started in any order. Clients only
2574 require the MDT to complete startup.
2576 Severity : enhancement
2578 Description: parallel, asynchronous orphan cleanup
2579 Details : orphan cleanup is now performed in separate threads for each
2580 OST, allowing parallel non-blocking operation.
2582 Severity : enhancement
2584 Description: optimized stripe assignment
2585 Details : stripe assignments are now made based on ost space available,
2586 ost previous usage, and OSS previous usage, in order to try
2587 to optimize storage space and networking resources.
2589 Severity : enhancement
2591 Description: Permanently set tunables
2592 Details : All writable /proc/fs/lustre tunables can now be permanently
2593 set on a per-server basis, at mkfs time or on a live system.
2595 Severity : enhancement
2597 Description: Lustre message v2
2598 Details : Add lustre message format v2.
2600 Severity : enhancement
2602 Description: client OST exclusion list
2603 Details : Clients can be started with a list of OSTs that should be
2604 declared "inactive" for known non-responsive OSTs.
2608 Description: SPEC SFS validation failure on NFS v2 over lustre.
2609 Details : Changes the blocksize for regular files to be 2x RPC size,
2610 and not depend on stripe size.
2612 Severity : enhancement
2614 Description: Multiple MD RPCs in flight.
2615 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2616 To avoid overly-overloading MDS, introduce a limit on number of
2617 MDS RPCs in flight for a single client and add /proc controls
2618 to adjust this limit.
2620 Severity : enhancement
2622 Description: client read/write statistics
2623 Details : Add client read/write call usage stats for performance
2624 analysis of user processes.
2625 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2626 file access. extents_stats shows chunk size distribution.
2627 extents_stats_per_process show chunk size distribution per
2630 Severity : enhancement
2632 Description: mds statistics
2633 Details : Add detailed mds operations statistics in
2634 /proc/fs/lustre/mds/*/stats.
2638 Description: Failure of copying files with lustre special EAs.
2639 Details : Client side always return success for setxattr call for lustre
2640 special xattr (currently only "trusted.lov").
2645 Description: Refcount LNET uuids
2646 Details : The global LNET uuid list grew linearly with every startup;
2647 refcount repeated list entries instead of always adding to
2650 Severity : enhancement
2652 Description: Dynamic service threads
2653 Details : Within a small range, start extra service threads
2654 automatically when the request queue builds up.
2657 Frequency : mixed-endian client/server environments
2659 Description: mixed-endian crashes
2660 Details : The new msg_v2 system had some failures in mixed-endian
2663 Severity : enhancement
2665 Description: Easy OST removal
2666 Details : OSTs can be permanently deactivated with e.g. 'lctl
2667 conf_param lustre-OST0001.osc.active=0'
2669 Severity : enhancement
2671 Description: MGS proc entries
2672 Details : Added basic proc entries for the MGS showing what filesystems
2675 Severity : enhancement
2677 Description: provide MGS failover
2678 Details : Added config lock reacquisition after MGS server failover.
2680 Severity : enhancement
2682 Description: add Linux 2.4 support
2683 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2687 Description: a large application tries to do I/O to the same resource and dies
2688 in the middle of it.
2689 Details : Check the req->rq_arrival time after the call to
2690 ost_brw_lock_get(), but before we do anything about
2691 processing it & sending the BULK transfer request. This
2692 should help move old stale pending locks off the queue as
2693 quickly as obd_timeout.
2696 Frequency : when an incorrect nid is specified during startup
2698 Description: ptlrpc connect to non-existant node causes kernel crash
2699 Details : LNET can't be re-entered from an event callback, which
2700 happened when we expire a message after the export has been
2701 cleaned up. Instead, hand the zombie cleanup off to another
2704 Severity : enhancement
2706 Description: plain/inodebits lock performance improvement
2707 Details : Grouping plain/inodebits in granted list by their request modes
2708 and bits policy, thus improving the performance of search through
2712 Frequency : only if OST filesystem is corrupted
2714 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2715 Details : for a short time RPCs with bulk IO are in the replay list,
2716 but replay of bulk IOs is unimplemented. If the OST filesystem
2717 is corrupted due to disk cache incoherency and then replay is
2718 started it is possible to trip an assertion. Avoid putting
2719 committed RPCs into the replay list at all to avoid this issue.
2722 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2724 Description: System hang on startup
2725 Details : This bug allowed the liblustre (e.g. catamount) client to
2726 return to the app before handling all startup RPCs. This
2727 could leave the node unresponsive to lustre network traffic
2728 and manifested as a server ptllnd timeout.
2730 Severity : enhancement
2732 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2733 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2734 boolean to control whether to print peer debug info when a
2735 client's RPC times out.
2738 Frequency : only for kernels with patches from Lustre below 1.4.3
2740 Description: Remove old rdonly API
2741 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2744 Frequency : only for devices with external journals
2746 Description: Set external device read-only also
2747 Details : During a commanded failover stop, we set the disk device
2748 read-only while the server shuts down. We now also set any
2749 external journal device read-only at the same time.
2752 Frequency : when upgrading from 1.4 while trying to change parameters
2754 Description: The wrong (new) MDC name was used when setting parameters for
2755 upgraded MDT's. Also allows changing of OSC (and MDC)
2756 parameters if --writeconf is specified at tunefs upgrade time.
2759 Frequency : when setting specific ost indicies
2761 Description: QOS code breaks on skipped indicies
2762 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2763 created with --index need not be sequential.
2765 Severity : enhancement
2767 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2768 Details : The uninit_groups feature works in conjunction with the kernel
2769 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2770 up the pass1 processing of e2fsck. This is a read-only feature
2771 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2772 mount filesystems that have had this feature enabled.
2774 Severity : enhancement
2776 Description: Improve multi-block allocation algorithm to avoid fragmentation
2777 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2778 allocation locality and avoid filesystem fragmentation.
2780 ------------------------------------------------------------------------------
2782 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2784 * Support for kernels:
2785 2.4.21-47.0.1.EL (RHEL 3)
2786 2.6.5-7.283 (SLES 9)
2787 2.6.9-42.0.10.EL (RHEL 4)
2788 2.6.12.6 vanilla (kernel.org)
2789 2.6.16.27-0.9 (SLES 10)
2790 * Recommended e2fsprogs version: 1.39.cfs5
2792 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2796 Frequency : occasional, depends on client load and configuration
2797 Bugzilla : 12181, 12203
2798 Description: data loss for recently-modified files
2800 Details : In some cases it is possible that recently written or created
2801 files may not be written to disk in a timely manner (this should
2802 normally be within 30s unless client IO load is very high).
2803 The problem appears as zero-length files or files that are a
2804 multiple of 1MB in size after a client crash or client eviction
2805 that are missing data at the end of the file.
2807 This problem is more likely to be hit on clients where files are
2808 repeatedly created and unlinked in the same directory, clients
2809 have a large amount of RAM, have many CPUs, the filesystem has
2810 many OSTs, the clients are rebooted frequently, and/or the files
2811 are not accessed by other nodes after being written.
2813 The presence of the problem can be detected by looking at
2814 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2815 smaller than the second (nr_unused) then dirty files will not
2816 be flushed automatically to disk. "sync; sleep 10" should be
2817 run several times on the node before unmounting it to update
2818 Lustre (this is also safe to run on nodes without this problem).
2820 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2821 that can cause this same problem, so customers using that kernel
2822 also need to update the kernel in addition to Lustre. In order
2823 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2825 It is normal that files written just before a client crash (less
2826 than 30s) may not yet have been flushed to disk, even for local
2830 Frequency : frequent on thin XT3 nodes
2832 Description: UUID collision on thin XT3 Linux nodes
2833 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2834 randomly, since we relied on an insufficiently-seeded PRNG.
2839 Description: OSS hangs after "All ost request buffers busy"
2840 Details : A deadlock between quota and journal operations caused OSS
2841 hangs after printing "All ost request buffers busy."
2844 Frequency : always on liblustre builds
2846 Description: Cleanup compiler warnings on liblustre
2849 Frequency : always on liblustre builds on XT3
2851 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2855 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2856 > 32000 subdirectories
2857 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2858 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2859 subdirectory count crosses 32000. This will aid e2fsck to
2860 correctly handle more than 32000 subdirectories.
2863 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2865 Description: System hang on startup
2866 Details : This bug allowed the liblustre (e.g. catamount) client to
2867 return to the app before handling all startup RPCs. This
2868 could leave the node unresponsive to lustre network traffic
2869 and manifested as a server ptllnd timeout.
2871 Severity : enhancement
2873 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2874 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2875 boolean to control whether to print peer debug info when a
2876 client's RPC times out.
2881 Description: make O_SYNC working on 2.6 kernels
2882 Details : 2.6 kernels use different method for mark pages for write,
2883 so need add a code to lustre for O_SYNC work.
2888 Description: Failure to close file and release space on NFS
2889 Details : Put inode details into lock acquired in ll_intent_file_open.
2890 Use mdc_intent_lock in ll_intent_open to properly
2891 detect all kind of errors unhandled by mdc_enqueue.
2896 Description: proc file read during shutdown sometimes raced obd removal,
2898 Details : Add lock to prevent obd access after proc file removal.
2901 Frequency : Only for files larger than 4GB on 32-bit clients.
2903 Description: improperly doing page alignment of locks
2904 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2905 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2910 Description: RPCs being resent when they shouldn't be
2911 Details : Some RPCs that should not be resent are being resent. This
2912 can cause inconsistencies in the RPC state machine. Do not
2913 resend such requests.
2916 Frequency : rare, only with NFS export
2918 Description: Crash on NFS re-export node
2919 Details : under very unusual load conditions an assertion is hit in
2920 ll_intent_file_open()
2923 Frequency : only if OST filesystem is corrupted
2925 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2926 Details : for a short time RPCs with bulk IO are in the replay list,
2927 but replay of bulk IOs is unimplemented. If the OST filesystem
2928 is corrupted due to disk cache incoherency and then replay is
2929 started it is possible to trip an assertion. Avoid putting
2930 committed RPCs into the replay list at all to avoid this issue.
2935 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2936 Details : Large single O_DIRECT read and write calls can fail to allocate
2937 a sufficiently large buffer to process the request. In case of
2938 allocation failure the allocation is retried with a smaller
2939 buffer and broken into smaller requests.
2941 Severity : enhancement
2943 Description: Add -o localflock option to simulate old noflock behaviour.
2944 Details : This will achieve local-only flock/fcntl locks coherentness.
2949 Description: versioning check is incomplete
2950 Details : Checking the version difference of client vs. server, report
2951 error if the gap is too big.
2956 Description: add support PG_writeback bit
2957 Details : add support for PG_writeback bit for Lustre, for more carefull
2958 work with page cache in 2.6 kernel. This also fix some deadlocks
2959 and remove hack for work O_SYNC with 2.6 kernel.
2961 Severity : enhancement
2963 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2964 Details : The uninit_groups feature works in conjunction with the kernel
2965 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2966 up the pass1 processing of e2fsck. This is a read-only feature
2967 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2968 mount filesystems that have had this feature enabled.
2970 Severity : enhancement
2972 Description: Improve multi-block allocation algorithm to avoid fragmentation
2973 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2974 allocation locality and avoid filesystem fragmentation.
2976 ------------------------------------------------------------------------------
2978 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2980 * Support for kernels:
2981 2.6.9-42.0.3.EL (RHEL 4)
2982 2.6.5-7.276 (SLES 9)
2983 2.4.21-47.0.1.EL (RHEL 3)
2984 2.6.12.6 vanilla (kernel.org)
2985 2.6.16.21-0.8 (SLES10)
2986 * Recommended e2fsprogs version: 1.39.cfs2-0
2988 * The backwards-compatible /proc/sys/portals symlink has been removed
2989 in this release. Before upgrading, please ensure that you change
2990 any configuration scripts or /etc/sysctl.conf files that access
2991 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2992 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2993 in advance of the upgrade on any system running Lustre 1.4.6 or
2994 newer, since /proc/sys/lnet was added in that version.
2995 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2999 Frequency : only when quota is used
3001 Description: avoid scanning export list for quota master
3002 Details : Change the algorithms to avoid scanning export list in order
3003 to improve the efficiency.
3006 Frequency : MDS failover only, very rarely
3008 Description: "went back in time" messages on mds failover
3009 Details : The greatest transno may be lost when the current operation
3010 finishes with an error (transno==0) and the client's last_rcvd
3011 record is over-written. Save the greatest transno in the
3012 mds_last_transno for this case.
3015 Frequency : always for specific kernels and striping counts
3017 Description: client may get "Matching packet too big" without ACL support
3018 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3019 when trying to access files in certain configurations. The
3020 clients should in fact be denied when mounting because they do
3021 not understand ACLs.
3024 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3026 Description: many clients connecting with IO in progress causes connect timeouts
3027 Details : Avoid synchronous journal commits to avoid delays caused by many
3028 clients connecting/disconnecting when bulk IO is in progress.
3029 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3030 OST_IO_PORTAL to avoid delays behind potentially many pending
3034 Frequency : occasionally with multiple writers to a single file
3036 Description: shared writes to file may result in wrong size reported by stat()
3037 Details : Allow growing of kms when extent lock is cancelled
3040 Frequency : always with random mmap IO to multi-striped file
3042 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3043 Details : Only if the hole is at the end of OST object so that kms is too
3044 small. Fix is to increase kms accordingly in ll_nopage.
3047 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3049 Description: writes to a missing object would leak memory on the OST
3050 Details : If there is an inconsistency between the MDS and OST filesystems,
3051 such that the MDS references an object that doesn't exist, writes
3052 to that object will leak memory due to incorrect cleanup in the
3053 error handling path, eventually running out of memory on the OST.
3058 Description: Creating too long symlink causes lustre errors
3059 Details : Check symlink and name lengths before sending requests to MDS.
3062 Frequency : only if flock is enabled (not on by default)
3064 Description: posix locks not released on fd closure on 2.6.9+
3065 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3066 kernels, this caused such locks not to be released on fd close and
3067 then assertions on fs unmount about still used locks.
3070 Frequency : MDS failover only, very rarely
3072 Description: clients may get ASSERTION(granted_lock != NULL)
3073 Details : When request was taking a long time, and a client was resending
3074 a getattr by name lock request. The were multiple lock requests
3075 with the same client lock handle and
3076 mds_getattr_name->fixup_handle_for_resent_request found one of the
3077 lock handles but later failed with ASSERTION(granted_lock != NULL).
3082 Description: handle->h_buffer_credits > 0, assertion failure
3083 Details : h_buffer_credits is zero after truncate, causing assertion
3084 failure. This patch extends the transaction or creates a new
3088 Frequency : NFS re-export or patchless client
3089 Bugzilla : 11179, 10796
3090 Description: Crash on NFS re-export node (__d_move)
3091 Details : We do not want to hash the dentry if we don't have a lock.
3092 But if this dentry is later used in d_move, we'd hit uninitialised
3093 list head d_hash, so we just do this to init d_hash field but
3094 leave dentry unhashed.
3097 Frequency : NFS re-export or patchless client
3099 Description: NFS exports has problem with symbolic link
3100 Details : lustre client didn't properly install dentry when re-exported
3101 to NFS or running patchless client.
3104 Frequency : NFS re-export or patchless client
3106 Description: Various nfs/patchless fixes.
3107 Details : fixes reuse disconected alias for lookup process - this fixes
3108 warning "find_exported_dentry: npd != pd",
3109 fix permission error with open files at nfs.
3110 fix apply umask when do revalidate.
3113 Frequency : occasional
3115 Description: Crash on NFS re-export node
3116 Details : calling clear_page() on the wrong pointer triggered oops in
3117 generic_mapping_read().
3120 Frequency : rarely, using O_DIRECT IO
3122 Description: unaligned directio crashes client with LASSERT
3123 Details : check for unaligned buffers before trying any requests.
3126 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3128 Description: stale data returned from RAID cache
3129 Details : If only a small amount of IO is done to the RAID device before
3130 reading it again it is possible to get stale data from the RAID
3131 cache instead of reading it from disk.
3134 Frequency : always for sles10 kernel
3136 Description: sles10 support
3137 Details : ll_follow_link: compile fixes and using of nd_set_link
3138 under newer kernels.
3141 Frequency : depends on arch, kernel and compiler version, always on sles10
3144 Description: recursive or deep enough symlinks cause stack overflow
3145 Details : getting rid of large stack-allocated variable in
3149 Frequency : depends on hardware
3151 Description: lustre write performance loss in the SLES10 kernel
3152 Details : the performance loss is caused by using of write barriers in the
3153 ext3 code. The SLES10 kernel turns barrier support on by
3154 default. The fix is to undo that change for ldiskfs.
3156 ------------------------------------------------------------------------------
3158 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3160 * Support for kernels:
3161 2.6.9-42.0.3EL (RHEL 4)
3162 2.6.5-7.276 (SLES 9)
3163 2.4.21-47.0.1.EL (RHEL 3)
3164 2.6.12.6 vanilla (kernel.org)
3168 Frequency : quota enabled and large files being deleted
3170 Description: releasing more than 4GB of quota at once hangs OST
3171 Details : If a user deletes more than 4GB of files on a single OST it
3172 will cause the OST to spin in an infinite loop. Release
3173 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3178 Description: statfs data retrieved from /proc may be stale or zero
3179 Details : When reading per-device statfs data from /proc, in the
3180 {kbytes,files}_{total,free,avail} files, it may appear
3181 as zero or be out of date.
3184 Frequency : systems with MD RAID1 external journal devices
3186 Description: lconf's call to blkid is confused by RAID1 journal devices
3187 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3188 of returning all block devices that match the journal UUID.
3191 Frequency : always, for aggregate stripe size over 4GB
3193 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3194 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3195 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3198 Frequency : always if "lfs find" used on a local file/directory
3200 Description: "lfs find" segfaults if used on a local file/directory
3201 Details : The case where a directory component was not specified wasn't
3202 handled correctly. Handle this properly.
3205 Frequency : always on ppc64
3207 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3208 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3209 perform endianness conversion.
3212 Frequency : rarely (truncate to non-zero file size after write under load)
3213 Bugzilla : 10730, 10687
3214 Description: Files padded with zeros to next 4K multiple
3215 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3216 it is possible that files that are truncated to a non-zero size
3217 immediately after being written are filled with zero bytes beyond
3218 the truncated size. No file data is lost.
3220 Severity : enhancement
3222 Description: Allow recovery/failover for liblustre clients.
3223 Details : liblustre clients were unaware of failover configurations until
3226 Severity : enhancement
3228 Description: user file locks should fail when not mounting with flock option
3229 Details : Set up an error-returning stub in ll_file_operations.lock field
3230 to prevent incorrect behaviour when client is mounted without
3231 flock option. Also, set up properly f_op->flock field for
3235 Frequency : always on ia64
3237 Description: "lfs df" loops on printing out MDS statfs information
3238 Details : The obd_ioctl_data was not initialized and in some systems
3239 this caused a failure during the ioctl that did not return
3240 an error. Initialize the struct and return an error on failure.
3243 Frequency : SLES 9 only
3245 Description: Error of copying files with lustre special EAs as root
3246 Details : Client side always return success for setxattr call for lustre
3247 special xattr (currently only "trusted.lov").
3250 Frequency : rarely on clusters with both ia64+i386 clients
3252 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3253 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3254 clients are doing concurrent writes to the same file it is
3255 possible that the ia64 clients may LASSERT because the OST
3256 extent locks are not PAGE_SIZE aligned. Ensure that grown
3257 locks are always aligned on the request boundary.
3260 Frequency : specific use, occasional
3262 Description: Overwriting in use executable truncates on-disk binary image
3263 Details : If one node attempts to overwrite an executable in use by
3264 another node, we now correctly return ETXTBSY instead of
3265 truncating the file.
3267 Severity : enhancement
3269 Description: Async OSC create to avoid the blocking unnecessarily.
3270 Details : If a OST has no remain object, system will block on the creating
3271 when need to create a new object on this OST. Now, ways use
3272 pre-created objects when available, instead of blocking on an
3273 empty osc while others are not empty. If we must block, we block
3274 for the shortest possible period of time.
3279 Description: chmod on Lustre root is propagated to other clients
3280 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3281 invalid by the follow_mount time.
3284 Frequency : liblustre clients only
3286 Description: Race in 'instant cancel' lock handling could lead to such locks
3287 never to be granted in case of SMP MDS
3288 Details : Do not destroy not yet granted but cbpending locks in
3292 Frequency : replay/resend of open
3294 Description: non null lock assetion failure in mds_intent_policy
3295 Details : Trying to replay/resend lockless open requests resulted in
3296 mds_open() returning 0 with no lock. Now it sets a flag if
3297 a lock is going to be returned.
3299 Severity : enhancement
3301 Description: Checksum enhancements
3302 Details : New checksum enhancements allow for resending RPCs that failed
3305 Severity : enhancement
3307 Description: Tunables on number of dirty pages in cacche
3308 Details : Allow to set limit on number of dirty pages cached.
3313 Description: client crash on unmount - lock still has references
3314 Details : In some error handling cases it was possible to leak a lock
3315 reference on a client while accessing a file. This was not
3316 harmful to the client during operation, but would cause the
3317 client to crash when the filesystem is unmounted.
3320 Frequency : specific case, rare
3322 Description: ETXTBSY on mds though file not in use by client
3323 Details : ETXTBSY is no longer incorrectly returned when attempting to
3324 chmod or chown a directory that the user previously tried to
3325 execute or a currently-executing binary.
3328 Frequency : extremely rare except on liblustre-based clients
3330 Description: Lustre space not freed when files are deleted
3331 Details : Clean up open-unlinked files after client eviction. Previously
3332 the unlink was skipped and the files remained as orphans.
3337 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3338 Details : In some cases it was possible to send a blocking callback to
3339 a client doing a glimpse, even though that client didn't get
3340 a lock granted. When the glimpse lock is cancelled on the OST
3341 the freed lock is left on the waiting list and corrupted the list.
3344 Frequency : all core dumps
3346 Description: Broke core dumps to lustre
3347 Details : Negative dentry may be unhashed if parent does not have UPDATE
3348 lock, but some callers, e.g. do_coredump, expect dentry to be
3349 hashed after successful create, hash it in ll_create_it.
3351 ------------------------------------------------------------------------------
3353 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3355 * Support for kernels:
3356 2.6.9-42.0.2.EL (RHEL 4)
3357 2.6.5-7.276 (SLES 9)
3358 2.4.21-40.EL (RHEL 3)
3359 2.6.12.6 vanilla (kernel.org)
3363 Frequency : always on RHEL 3
3365 Description: Number of open files grows over time
3366 Details : The number of open files grows over time, whether or not
3367 Lustre is started. This was due to a filp leak introduced
3368 by one of our kernel patches.
3370 ------------------------------------------------------------------------------
3372 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3374 * Support for kernels:
3375 2.6.9-42.EL (RHEL 4)
3376 2.6.5-7.276 (SLES 9)
3377 2.4.21-40.EL (RHEL 3)
3378 2.6.12.6 vanilla (kernel.org)
3383 Bugzilla : 5719, 9635, 9792, 9684
3384 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3385 Details : If a server is under heavy load and cannot reply to new
3386 connection requests before the client resends the (re)connect,
3387 the connection handling code can behave badly if two service
3388 threads are concurrently handing separate (re)connections from
3389 the same client. Add better locking to the connection handling
3390 code, and ensure that only a single connection will be processed
3391 for a given client UUID, even if the lock is dropped.
3393 Severity : enhancement
3395 Description: add TCP zero-copy support to kernel
3396 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3397 sends if the hardware supports scatter-gather and checksumming.
3398 This allows socklnd to do client-write and server-read more
3399 efficiently and reduce CPU utilization from skbuf copying.
3402 Frequency : only if NFS exporting from client
3404 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3405 Details : If setting attributes on a file created under NFS that had
3406 never been opened it would be possible to oops the client
3407 if the file had no objects.
3410 Frequency : always for liblustre
3412 Description: liblustre client does MDS+OSTs setattr RPC for each write
3413 Details : When doing a write from a liblustre client, the client
3414 incorrectly issued an RPC to the MDS and each OST the file was
3415 striped over in order to update the timestamps. When writing
3416 with small chunks and many clients this could overwhelm the MDS
3417 with RPCs. In all cases it would slow down the write because
3418 these RPCs are unnecessary.
3420 Severity : enhancement
3422 Description: allow number of MDS service threads to be changed at module load
3423 Details : It is now possible to change the number of MDS service threads
3424 running. Adding "options mds mds_num_threads={N}" to the MDS's
3425 /etc/modprobe.conf will set the number of threads for the next
3426 time Lustre is restarted (assuming the "mds" module is also
3427 reloaded at that time). The default number of threads will
3428 stay the same, 32 for most systems.
3433 Description: OST crash if filesystem is unformatted or corrupt
3434 Details : If an OST is started on a device that has never been formatted
3435 or if the filesystem is corrupt and cannot even mount then the
3436 error handling cleanup routines would dereference a NULL pointer.
3441 Description: NULL pointer deref in llap_from_page.
3442 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3443 incorrect) mapping if the page was truncated/reclaimed while it was
3444 searched for. Check for this condition and skip such pages when
3445 doing readahead. Introduce extra check to llap_from_page() to
3446 verify page->mapping->host is non-NULL (so page is not anonymous).
3449 Frequency : Sometimes when using sys_sendfile
3451 Description: "page not covered by a lock" warnings from ll_readpage
3452 Details : sendfile called ll_readpage without right page locks present.
3453 Now we introduced ll_file_sendfile that does necessary locking
3454 around call to generic_file_sendfile() much like we do in
3458 Frequency : with certain MDS communication failures at client mount time
3460 Description: NULL pointer deref after failed client mount
3461 Details : a client connection request may delayed by the network layer
3462 and not be sent until after the PTLRPC layer has timed out the
3463 request. If the client fails the mount immediately it will try
3464 to clean up before the network times out the request. Add a
3465 reference from the request import to the obd device and delay
3466 the cleanup until the network drops the request.
3469 Frequency : occasionally during client (re)connect
3471 Description: assertion failure during client (re)connect
3472 Details : processing a client connection request may be delayed by the
3473 client or server longer than the client connect timeout. This
3474 causes the client to resend the connection request. If the
3475 original connection request is replied in this interval, the
3476 client may trip an assertion failure in ptlrpc_connect_interpret()
3477 which thought it would be the only running connect process.
3480 Frequency : only with obd_echo servers and clients that are rebooted
3482 Description: kernel BUG accessing uninitialized data structure
3483 Details : When running an obd_echo server it did not start the ping_evictor
3484 thread, and when a client was evicted an uninitialized data
3485 structure was accessed. Start the ping_evictor in the RPC
3486 service startup instead of the OBD startup.
3488 Severity : enhancement
3489 Bugzilla : 10193 (patchless)
3490 Description: Remove dependency on various unexported kernel interfaces.
3491 Details : No longer need reparent_to_init, exit_mm, exit_files,
3492 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3495 Frequency : rare (only users of deprecated and unsupported LDAP config)
3497 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3498 Details : LDAP apparently contains 'lustreName' attributes instead of
3499 'name'. A simple remapping of the name is sufficient.
3502 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3504 Description: waiting_locks_callback trips kernel BUG if client is evicted
3505 Details : Running with the dump_on_timeout debug flag turned on makes
3506 it possible that the waiting_locks_callback() can try to dump
3507 the Lustre kernel debug logs from an interrupt handler. Defer
3508 this log dumping to the expired_lock_main() thread.
3510 Severity : enhancement
3512 Description: Support NFS exporting on 2.6 kernels.
3513 Details : Implement non-rawops metadata methods for NFS server to use without
3514 changing NFS server code.
3517 Frequency : very rare (synthetic metadata workload only)
3519 Description: two racing renames might cause an MDS thread to deadlock
3520 Details : Running the "racer" program may cause one MDS thread to rename
3521 a file from being the source of a rename to being the target of
3522 a rename at exactly the same time that another thread is doing
3523 so, and the second thread has already enqueued these locks after
3524 doing a lookup of the target and is trying to relock them in
3525 order. Ensure that we don't try to re-lock the same resource.
3528 Frequency : only very large systems with liblustre clients
3530 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3531 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3532 all OSTs in parallel. This allows the eviction of stale liblustre
3533 clients to proceed much faster than if they were done in series,
3534 and also offers similar improvements for other set_info RPCs.
3539 Description: excessive CPU usage during initial read phase on client
3540 Details : During the initial read phase on a client, it would agressively
3541 retry readahead on the file, consuming too much CPU and impacting
3542 performance (since 1.4.5.8). Improve the readahead algorithm
3543 to avoid this, and also improve some other common cases (read
3544 of small files in particular, where "small" is files smaller than
3545 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3550 Description: MDS crash when receiving packet with unknown intent.
3551 Details : Do not LBUG in unknown intent case, just return -EFAULT
3553 Severity : enhancement
3554 Bugzilla : 9293, 9385
3555 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3556 Details : Do not serialize getattr (non-intent version) and statfs.
3559 Frequency : occasional, when OST network is overloaded/intermittent
3561 Description: client evicted by OST after bulk IO timeout
3562 Details : If a client sends a bulk IO request (read or write) the OST
3563 may evict the client if it is unresposive to its data GET/PUT
3564 request. This is incorrect if the network is overloaded (takes
3565 too long to transfer the RPC data) or dropped the OST GET/PUT
3566 request. There is no need to evict the client at all, since
3567 the pinger and/or lock callbacks will handle this, and the
3568 client can restart the bulk request.
3571 Frequency : Always when mmapping file with no objects
3573 Description: client crashes when mmapping file with no objects
3574 Details : Check that we actually have objects in a file before doing any
3575 operations on objects in ll_vm_open, ll_vm_close and
3581 Description: Request leak when working with deleted CWD
3582 Details : Introduce advanced request refcount tracking for requests
3583 referenced from lustre intent.
3585 Severity : Enhancement
3587 Description: Cache open file handles on client.
3588 Details : MDS now will return special lock along with openhandle, if
3589 requested and client is allowed to hold openhandle, even if unused,
3590 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3591 closing files for every read/write openration.
3593 Severity : Enhancement
3595 Description: Cache open negative dentries on client when possible.
3596 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3597 negative dentries on lock revocation.
3602 Description: Remounting a client read-only wasn't possible with a zconf mount
3603 Details : It wasn't possible to remount a client read-only with llmount.
3605 Severity : enhancement
3606 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3607 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3608 MPICH is included to add Lustre-specific ADIO interfaces.
3609 This is based closely on the UFS ADIO layer and only differs
3610 in file creation, in order to allow the OST striping to be set.
3611 This is user-contributed code and not supported by CFS.
3616 Description: extended inode attributes (immutable, append-only) work improperly
3617 when 2.4 and 2.6 kernels are used on client/server or vice versa
3618 Details : Introduce kernel-independent values for these flags.
3620 Severity : enhancement
3623 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3624 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3625 granularity. Now, Fractional MB granularity is be supported,
3626 this is very useful for low memory system.
3628 Severity : enhancement
3630 Description: Getattr by fid
3631 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3632 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3633 have locks and avoids getattr rpc requests for it on every stat.
3636 Frequency : Always, for filesystems larger than 2TB
3638 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3639 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3640 filesystems larger than 2TB to be mounted with mballoc enabled.
3643 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3645 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3646 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3647 filesystems larger than 2TB is mounted then the kernel will
3648 silently corrupt the start of the filesystem. CONFIG_LBD is
3649 enabled for all CFS-supported kernels, but the possibility of
3650 this happening with a modified kernel config exists.
3652 Severity : enhancement
3654 Description: add client O_DIRECT support for 2.6 kernels
3655 Details : It is now possible to do O_DIRECT reads and writes to files
3656 in the Lustre client mountpoint on 2.6 kernel clients.
3658 Severity : enhancement
3660 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3661 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3662 parallel, not waiting for response from every OST before sending
3663 a rpc to the next OST.
3668 Description: setattr vs write race when updating file timestamps
3669 Details : Client processes that update a file timestamp into the past
3670 right after writing to the file (e.g. tar) it is possible that
3671 the updated file modification time can be reset to the current
3672 time due to a race between processing the setattr and write RPC.
3674 Severity : enhancement
3676 Description: Bring 'lfs find' closer in line with regular Linux find.
3677 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3678 -print0 options and obtains all the needed info through the lustre
3681 Severity : enhancement
3683 Description: support up to 1024 configured devices on one node
3684 Details : change obd_dev array from statically allocated to dynamically
3685 allocated structs as they are first used to reduce memory usage
3690 Description: Flush dirty partially truncated pages during truncate
3691 Details : Immediatelly flush partially truncated pages in filter_setattr,
3692 this way we completely avoid having any pages in page cache on OST
3693 and can retire ugly workarounds during writes to flush such pages.
3698 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3699 Details : protect inode from truncation within vfs_unlink() context
3700 just take a reference before calling vfs_unlink() and release it
3701 when parent's i_sem is free.
3706 Description: last_id value checked outside lock on OST caused LASSERT failure
3707 Details : If there were multiple MDS->OST object precreate requests in
3708 flight, it was possible that the OST's last object id was checked
3709 outside a lock and incorrectly tripped an assertion. Move checks
3710 inside locks, and discard old precreate requests.
3713 Frequency : always, if extents are used on OSTs
3715 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3716 Details : OSTs using the extents format would not zero the high 16 bits of
3717 the index physical block number. This is not a problem for any
3718 OST filesystems smaller than 16TB, and no kernels support ext3
3719 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3720 new/modified files) and can be fixed for existing filesystems
3721 with e2fsprogs-1.39-cfs1.
3726 Description: import connection selection may be incorrect if timer wraps
3727 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3728 import connections to be ignored if the 32-bit jiffies counter
3729 wraps. Use a 64-bit jiffies counter.
3732 Frequency : very large clusters immediately after boot
3734 Description: LNET request buffers exhausted under heavy short-term load
3735 Details : If a large number of client requests are generated on a service
3736 that has previously never seen so many requests it is possible
3737 that the request buffer growth cannot keep up with the spike in
3738 demand. Instead of dropping incoming requests, they are held in
3739 the LND until the RPC service can accept more requests.
3742 Frequency : Sometimes during replay
3744 Description: Assertion failure in ll_local_open after replay.
3745 Details : If replay happened on an open request reply before we were able
3746 to set replay handler, reply will become not swabbed tripping the
3747 assertion in ll_local_open. Now we set the handler right after
3748 recognising of open request
3751 Frequency : very rare
3753 Description: kernel reports "badness in vsnprintf"
3754 Details : Reading from the "recovery_status" /proc file in small chunks
3755 may cause a negative length in lprocfs_obd_rd_recovery_status()
3756 call to vsnprintf() (which is otherwise harmless). Exit early
3757 if there is no more space in the output buffer.
3759 Severity : enhancement
3761 Description: clear OBD RPC statistics by writing to them
3762 Details : It is now possible to clear the OBD RPC statistics by writing
3763 to the "stats" file.
3768 Description: Client mtime is not the same on different clients after utimes
3769 Details : In some cases, the client was using the utimes() syscall on
3770 a file cached on another node. The clients now validate the
3771 ctime from the MDS + OSTs to determine which one is right.
3776 Description: Inability to activate failout mode
3777 Details : lconf script incorrectly assumed that in pythong string's numeric
3778 value is used in comparisons.
3781 Frequency : always with multiple stripes per file
3783 Description: Inefficient object allocation for mutli-stripe files
3784 Details : When selecting which OSTs to stripe files over, for files with
3785 a stripe count that divides evenly into the number of OSTs,
3786 the MDS is always picking the same starting OST for each file.
3787 Return the OST selection heuristic to the original design.
3792 Description: mount failures may take full timeout to return an error
3793 Details : Under some heavy load conditions it is possible that a
3794 failed mount can wait for the full obd_timeout interval,
3795 possibly several minutes, before reporting an error.
3796 Instead return an error as soon as the status is known.
3798 Frequency : quota enabled and large files being deleted
3800 Description: releasing more than 4GB of quota at once hangs OST
3801 Details : If a user deletes more than 4GB of files on a single OST it
3802 will cause the OST to spin in an infinite loop. Release
3803 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3808 Description: statfs data retrieved from /proc may be stale or zero
3809 Details : When reading per-device statfs data from /proc, in the
3810 {kbytes,files}_{total,free,avail} files, it may appear
3811 as zero or be out of date.
3814 Frequency : systems with MD RAID1 external journal devices
3816 Description: lconf's call to blkid is confused by RAID1 journal devices
3817 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3818 of returning all block devices that match the journal UUID.
3821 Frequency : always, for aggregate stripe size over 4GB
3823 Description: assertion fails when trying to use 4GB stripe size
3824 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3825 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3828 Frequency : always on ppc64
3830 Description: the first write on an ext3 filesystem with mballoc got stuck
3831 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3832 perform endianness conversion.
3834 ------------------------------------------------------------------------------
3836 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3838 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3839 INTEROPERATE with older versions automatically. Please read the
3840 user documentation before upgrading any part of a live system.
3841 * WARNING: Lustre networking configuration changes are required with
3842 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3845 * Support for kernels:
3846 2.6.9-22.0.2.EL (RHEL 4)
3847 2.6.5-7.244 (SLES 9)
3848 2.6.12.6 vanilla (kernel.org)
3851 Severity : enhancement
3852 Bugzilla : 7981/8208
3853 Description: Introduced Lustre Networking (LNET)
3854 Details : LNET is new networking infrastructure for Lustre, it includes
3855 a reorganized network configuration mode (see the user
3856 documentation for full details) as well as support for routing
3857 between different network fabrics. Lustre Networking Devices
3858 (LNDs) for the supported network fabrics have also been
3859 created for this new infrastructure.
3861 Severity : enhancement
3862 Description: Introduced Access control lists
3863 Details : clients can set ACLs on files and directories in order to have
3864 more fine-grained permissions than the standard Unix UGO+RWX.
3865 The MDS must be started with the "-o acl" mount option.
3867 Severity : enhancement
3868 Description: Introduced filesystem quotas
3869 Details : Administrators may now establish per-user quotas on the
3872 Severity : enhancement
3874 Description: Configuration change for the XT3
3875 The PTLLND is now used to run Lustre over Portals on the XT3
3876 The configure option(s) --with-cray-portals are no longer used.
3877 Rather --with-portals=<path-to-portals-includes> is used to
3878 enable building on the XT3. In addition to enable XT3 specific
3879 features the option --enable-cray-xt3 must be used.
3884 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3885 Details : A race between allocating a new llcd and re-getting the llcd_lock
3886 allowed another thread to grab newly-allocated llcd.
3888 Severity : enhancement
3890 Description: 2.6 OST async journal commit and locking fix to improve performance
3891 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3892 2.6 kernels are now async as they already were in 2.4 kernels so
3893 that they can commit concurrently with the network bulk transfer.
3894 For block-allocated files the filter allocation semaphore is held
3895 to avoid filesystem fragmentation during allocation. BKL lock
3896 removed for 2.6 xattr operations where it is no longer needed.
3901 Description: lconf incorrectly determined whether two IP networks could talk
3902 Details : In some more complicated routing and multiple-network
3903 configurations, lconf will avoid trying to make a network
3904 connection to a disjoint part of the IP space. It was doing the
3905 math incorrectly for one set of cases.
3910 Description: Fix for potential infinite loop processing records in an llog.
3911 Details : If an llog record is corrupted/zeroed, it is possible to loop
3912 forever in llog_process(). Validate the llog record length
3913 and skip the remainder of the block on error.
3916 Frequency : occasional (liblustre only)
3918 Description: liblustre could not open files whose last component is a symlink
3919 Details : sysio_path_walk() would incorrectly pass the open intent to
3920 intermediate path components.
3923 Frequency : rare (liblustre only with non-standard tuning)
3924 Bugzilla : 7201 (7350)
3925 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3926 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3927 after it has been released, only do this for referenced locks
3929 Severity : enhancement
3931 Description: specifying an (invalid) directory default stripe_size of -1
3932 would reset the directory default striping
3933 Details : stripe_size -1 was used internally to signal directory stripe
3934 removal, now use "all default" to signal dir stripe removal
3935 as a directory striping of "all default" is not useful
3938 Frequency : common for large clusters running liblustre clients
3940 Description: doing an ls when liblustre clients are running is slow
3941 Details : sending a glimpse AST to a liblustre client waits for every AST
3942 to time out, as liblustre clients will not respond. Since they
3943 cannot cache data we refresh the OST lock LVB from disk instead.
3945 Severity : enhancement
3947 Description: doing an ls at the same time as file IO can be slow
3948 Details : enqueue and other "small" requests can be blocked behind many
3949 large IO requests. Create a new OST IO portal for non-IO
3950 requests so they can be processed faster.
3953 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3955 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3956 created file with mode 000
3957 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3958 allow an ftruncate() as a non-root user to a file with mode 000.
3959 Lustre now respects this flag to disable mode checks when
3960 truncating a file owned by the user
3963 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3966 Description: Revoking locks from clients that went dead or catatonic might take
3968 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3969 cancellation of such locks instant on servers without waiting for
3970 any reply from clients. Clients drops these locks when cancel
3971 notification from server is received without replying.
3974 Frequency : liblustre-only, when liblustre client dies or becomes busy
3976 Description: Doing ls on Linux clients can take a long time with active
3978 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3979 granting such locks to it in the first place if possible. Locks
3980 are taken by proxy on the OST during the read or write and
3981 dropped immediately afterward. Add connect flags handling, do
3982 not grant locks to liblustre clients for glimpse ASTs.
3984 Severity : enhancement
3986 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3987 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3988 sequential reads and behaves badly with applications doing random
3989 reads. Improve it by reading ahead at least read region, and
3990 avoiding excessive large RPC for small reads.
3992 Severity : enhancement
3994 Description: Creating more than 1000 files for a single job may cause a load
3995 imbalance on the OSTs if there are also a large number of OSTs.
3996 Details : qos_prep_create() uses an OST index reseed value that is an
3997 even multiple of the number of available OSTs so that if the
3998 reseed happens in the middle of the object allocation it will
3999 still utilize the OSTs as uniformly as possible.
4004 Description: OST or MDS may oops in ping_evictor_main()
4005 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4006 but doesn't restart at beginning of obd_exports_timed list
4009 Severity : enhancement
4011 Description: improve by-nid export eviction on the MDS and OST
4012 Details : allow multiple exports with the same NID to be evicted at one
4013 time without re-searching the exports list.
4016 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4018 Description: MDS may oops in groups_free()
4019 Details : in rare race conditions a newly allocated group_info struct is
4020 freed again, and this can be NULL. The 2.4 compatibility code
4021 for groups_free() checked for a NULL pointer, but 2.6 did not.
4024 Frequency : common for liblustre clients doing little filesystem IO
4025 Bugzilla : 9352, 7313
4026 Description: server may evict liblustre clients accessing contended locks
4027 Details : if a client is granted a lock or receives a completion AST
4028 with a blocking AST already set it would not reply to the AST
4029 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4032 Frequency : lfs setstripe, only systems with more than 160 OSTs
4034 Description: unable to set striping with a starting offset beyond OST 160
4035 Details : llapi_create_file() incorrectly limited the starting stripe
4036 index to the maximum single-file stripe count.
4039 Frequency : LDAP users only
4041 Description: lconf did not handle in-kernel recovery with LDAP properly
4042 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4044 Severity : enhancement
4046 Description: bind OST threads to NUMA nodes to improve performance
4047 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4048 node and do their allocations there to localize memory access
4050 Severity : enhancement
4052 Description: llmount can determine client NID directly from Myrinet (GM)
4053 Details : the client NID code from gmnalnid was moved directly into
4054 llmount, removing the need to use this or specifying the
4055 client NID explicitly when mounting GM clients with zeroconf
4058 Frequency : if client is started with down MDS
4060 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4061 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4062 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4063 ptlrpc_import_delayed_req() because we didn't check if the
4064 request was interrupted, and we also didn't break out of the
4065 event loop if there was no timeout
4070 Description: data loss during non-page-aligned writes to a single file from
4071 both multiple nodes and multiple threads on one node at same time
4072 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4073 inconsistency led to false short-reads, that were cached and later
4074 used by ->prepare_write() to fill in partially written page,
4075 leading to data loss.
4078 Frequency : always, if lconf --abort_recovery used
4080 Description: lconf --abort_recovery fails with 'Operation not supported'
4081 Details : lconf was attempting to abort recovery on the MDT device and not
4084 Severity : enhancement
4086 Description: remove cleanup logs
4087 Details : replace lconf-generated cleanup logs with lustre internal
4088 cleanup routines. Eliminates the need for client-cleanup and
4091 Severity : enhancement
4093 Description: add support for EAs (user and system) on lustre filesystems
4094 Details : it is now possible to store extended attributes in the Lustre
4095 client filesystem, and with the user_xattr mount option it
4096 is possible to allow users to store EAs on their files also
4098 Severity : enhancement
4100 Description: Add possibility (config option) to show minimal available OST free
4102 Details : When compiled with --enable-mindf configure option, statfs(2)
4103 (and so, df) will return least minimal free space available from
4104 all OSTs as amount of free space on FS, instead of summary of
4105 free spaces of all OSTs.
4107 Severity : enhancement
4109 Description: do not expand extent locks acquired on OST-side
4110 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4111 by server: they are not cached anyway.
4114 Frequency : when mmap is used/binaries executed from Lustre
4116 Description: Unmmap pages before throwing them away from read cache.
4117 Details : llap_shrink cache now attempts to unmap pages before discarding
4118 them (if unmapping failed - do not discard). SLES9 kernel has
4119 extra checks that trigger if this unmapping is not done first.
4124 Description: lconf didn't resolve symlinks before checking to see whether a
4125 given mountpoint was already in use
4128 Frequency : when migrating failover services
4129 Bugzilla : 6395, 9514
4130 Description: When migrating a subset of services from a node (e.g. failback
4131 from a failover service node) the remaining services would
4132 time out and evict clients.
4133 Details : lconf --force (implied by --failover) sets the global obd_timeout
4134 to 5 seconds in order to quickly disconnect, but this caused
4135 other RPCs to time out too quickly. Do not change the global
4136 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4138 Severity : enhancement
4139 Frequency : if MDS is started with down OST
4140 Bugzilla : 9439,5706
4141 Description: Allow startup/shutdown of an MDS without depending on the
4142 availability of the OSTs.
4143 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4144 Add appropriate locking and lov-osc refcounts for safe
4145 cleaning. Add osc abort_inflight calls in case the
4146 synchronize never started.
4149 Frequency : occasional (Cray XT3 only)
4151 Description: root not authorized to access files in CRAY_PORTALS environment
4152 Details : The client process capabilities were not honoured on the MDS in
4153 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4154 been accessed by an authorized user then root was able to access
4155 the file on the local client also. The root user capabilities
4156 are now allowed on the MDS, as this environment has secure UID.
4159 Frequency : occasional
4161 Description: ldiskfs "too long searching" message happens too often
4162 Details : A debugging message (otherwise harmless) prints too often on
4163 the OST console. This has been reduced to only happen when
4164 there are fragmentation problems on the filesystem.
4169 Description: Division by zero in statfs when all OSCs are inactive
4170 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4171 lov_statfs divides by value returned by lov_get_stripecnt().
4175 Bugzilla : 9489, 3273
4176 Description: First write from each client to each OST was only 4kB in size,
4177 to initialize client writeback cache, which caused sub-optimal
4178 RPCs and poor layout on disk for the first writen file.
4179 Details : Clients now request an initial cache grant at (re)connect time
4180 and so that they can start streaming writes to the cache right
4181 away and always do full-sized RPCs if there is enough data.
4182 If the OST is rebooted the client also re-establishes its grant
4183 so that client cached writes will be honoured under the grant.
4188 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4189 Details : Now I/O RPCs go to different portal number and (presumably) fast
4190 lock requests (and glimses) and other RPCs get their own service
4191 threads pool that should be able to service those RPCs
4194 Severity : enhancement
4196 Description: Ability to exchange lustre version between client and servers and
4197 issue warnings at client side if client is too old. Also for
4198 liblustre clients there is ability to refuse connection of too old
4200 Details : New 'version' field is added to connect data structure that is
4201 filled with version info. That info is later checked by server and
4205 Frequency : rare, liblustre only.
4206 Bugzilla : 9296, 9581
4207 Description: Two simultaneous writes from liblustre at offset within same page
4208 might proceed at the same time overwriting eachother with stale
4210 Details : I/O lock withing llu_file_prwv was released too early, before data
4211 actually was hitting the wire. Extended lock-holding time until
4212 server acknowledges receiving data.
4215 Frequency : extremely rare. Never observed in practice.
4217 Description: avoid generating lustre_handle cookie of 0.
4218 Details : class_handle_hash() generates handle cookies by incrementing
4219 global counter, and can hit 0 occasionaly (this is unlikely, but
4220 not impossible, because initial value of cookie counter is
4221 selected randonly). Value of 0 is used as a sentinel meaning
4222 "unassigned handle" --- avoid it. Also coalesce two critical
4223 sections in this function into one.
4225 Severity : enhancement
4227 Description: allow liblustre clients to delegate truncate locking to OST
4228 Details : To avoid overhead of locking, liblustre client instructs OST to
4229 take extent lock in ost_punch() on client's behalf. New connection
4230 flag is added to handle backward compatibility.
4232 Severity : enhancement
4233 Bugzilla : 4928, 7341, 9758
4234 Description: allow number of OST service threads to be specified
4235 Details : a module parameter allows the number of OST service threads
4236 to be specified via "options ost ost_num_threads={N}" in the
4237 OSS's /etc/modules.conf or /etc/modprobe.conf.
4241 Bugzilla : 6146, 9635, 9895
4242 Description: servers crash with bad pointer in target_handle_connect()
4243 Details : In rare cases when a client is reconnecting it was possible that
4244 the connection request was the last reference for that export.
4245 We would temporarily drop the export reference and get a new
4246 one, but this may have been the last reference and the export
4247 was just destroyed. Get new reference before dropping old one.
4249 Severity : enhancement
4250 Frequency : if client is started with failover MDS
4252 Description: Allow multiple MDS hostnames in the mount command
4253 Details : Try to read the configuration from all specified MDS
4254 hostnames during a client mount in case the "primary"
4257 Severity : enhancement
4259 Description: Stop sending data to evicted clients as soon as possible.
4260 Details : Check if the client we are about to send or are sending data to
4261 was evicted already. (Check is done every second of waiting,
4262 for which l_wait_event interface was extended to allow checking
4263 of exit condition at specified intervals).
4266 Frequency : rare, normally only when NFS exporting is done from client
4268 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4270 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4273 Frequency : always, if srand() is called before liblustre initialization
4275 Description: Liblustre uses system PRNG disturbing its usage by user application
4276 Details : Introduce internal to lustre fast and high-quality PRNG for
4277 lustre usage and make liblustre and some other places in generic
4278 lustre code to use it.
4280 Severity : enhancement
4281 Bugzilla : 9477, 9557, 9870
4282 Description: Verify that the MDS configuration logs are updated when xml is
4283 Details : Check if the .xml configuration logs are newer than the config
4284 logs stored on the MDS and report an error if this is the case.
4285 Request --write-conf, or allow starting with --old_conf.
4287 Severity : enhancement
4289 Description: Handle symlinks in the path when checking if Lustre is mounted.
4290 Details : Resolve intermediate symlinks when checking if a client has
4291 mounted a filesystem to avoid duplicate client mounts.
4296 Description: lconf can hit an error exception but still return success.
4297 Details : The lconf command catches the Command error exception at the top
4298 level script context and will exit with the associated exit
4299 status, but doesn't ensure that this exit status is non-zero.
4304 Description: failure of ptlrpc thread startup can cause oops
4305 Details : Starting a ptlrpc service thread can fail if there are a large
4306 number of threads or the server memory is very fragmented.
4307 Handle this without oopsing.
4310 Frequency : always, only if liblustre and non-default acceptor port was used
4312 Description: liblustre cannot connect to servers with non-default acceptor port
4313 Details : tcpnal_set_default_params() was not called and was therefore
4314 ignoring the environment varaible TCPNAL_PORT, as well as other
4315 TCPNAL_ environment variables
4320 Description: two objects could be created on the same OST for a single file
4321 Details : If an OST is down, in some cases it was possible to create two
4322 objects on a single OST for a single file. No problems other
4323 than potential performance impact and spurious error messages.
4327 Bugzilla : 5681, 9562
4328 Description: Client may oops in ll_unhash_aliases
4329 Details : Client dcache may become inconsistent in race condition.
4330 In some cases "getcwd" can fail if the current directory is
4336 Description: Inode refcounting problems in NFS export code
4337 Details : link_raw functions used to call d_instantiate without obtaining
4338 extra inode reference first.
4342 Bugzilla : 9942, 9903
4343 Description: Referencing freed requests leading to crash, memleaks with NFS.
4344 Details : We used to require that call to ll_revalidate_it was always
4345 followed by ll_lookup_it. Also with revalidate_special() it is
4346 possible to call ll_revalidate_it() twice for the same dentry
4347 even if first occurence returned success. This fix changes semantic
4348 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4349 reference on a request referred from the intent.
4350 ll_intent_release() then releases such a request.
4353 Frequency : rare, normally benchmark loads only
4355 Description: unlinked inodes were kept in memory on the client
4356 Details : If a client is repeatedly creating and unlinking files it
4357 can accumulate a lot of stale inodes in the inode slab cache.
4358 If there is no other client load running this can cause the
4359 client node to run out of memory. Instead flush old inodes
4360 from client cache that have the same inode number as a new inode.
4363 Frequency : SLES9 2.6.5 kernel and long filenames only
4364 Bugzilla : 9969, 10379
4365 Description: utime reports stale NFS file handle
4366 Details : SLES9 uses out-of-dentry names in some cases, which confused
4367 the lustre dentry revalidation. Change it to always use the
4371 Frequency : rare, unless heavy write-truncate concurrency is continuous
4372 Bugzilla : 4180, 6984, 7171, 9963, 9331
4373 Description: OST becomes very slow and/or deadlocked during object unlink
4374 Details : filter_destroy() was holding onto the parent directory lock
4375 while truncating+unlinking objects. For very large objects this
4376 may block other threads for a long time and slow overall OST
4377 responsiveness. It may also be possible to get a lock ordering
4378 deadlock in this case, or run out of journal credits because of
4379 the combined truncate+unlink. Solution is to do object truncate
4380 first in one transaction without parent lock, and then do the
4381 final unlink in a new transaction with the parent lock. This
4382 reduces the lock hold time dramatically.
4385 Frequency : rare, 2.4 kernels only
4387 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4388 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4389 before unmounting their filesystem in order to avoid deadlock.
4390 The kernel_locked() function in 2.4 kernels only checks whether
4391 the kernel lock is held, not whether it is this process that is
4392 holding it as 2.6 kernels do.
4397 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4398 Details : The client ptlrpc code may be trying to reconnect to a down
4399 server before a previous connection attempt has timed out.
4400 Increase the reconnect interval to be longer than the connection
4401 timeout interval to avoid sending duplicate connections to
4405 Frequency : echo_client brw_test command
4407 Description: fix echo_client to work with OST preallocated code
4408 Details : OST preallocation code (5137) didn't take echo_client IO path
4409 into account: echo_client calls filter methods outside of any
4410 OST thread and, hence, there is no per-thread preallocated
4411 pages and buffers to use. Solution: hijack pga pages for IO. As
4412 a byproduct, this avoids unnecessary data copying.
4416 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4417 Description: Client can oops in mdc_commit_close() after open replay
4418 Details : It was possible for the MDS to return an open request with no
4419 transaction number in mds_finish_transno() if the client was
4420 evicted, but without actually returning an error. Clients
4421 would later try to replay that open and may trip an assertion
4422 Simplify the client close codepath, and always return an error
4423 from the MDS in case the open is not successful.
4426 Frequency : rare, 2.6 OSTs only
4428 Description: OST may deadlock under high load on fragmented files
4429 Details : If there was a heavy load and highly-fragmented OST filesystems
4430 it was possible to have all the OST threads deadlock waiting on
4431 allocation of biovecs, because the biovecs were not released
4432 until the entire RPC IO was completed. Instead, release biovecs
4433 as soon as they are complete to ensure forward IO progress.
4435 Severity : enhancement
4437 Description: Support for specifying external journal device at mount
4438 Details : If an OST or MDS device is formatted with an external journal
4439 device, this device major/minor is stored in the ext3 superblock
4440 and may not be valid for failover. Allow detecting and
4441 specifying the external journal at mount time.
4446 Description: Mounting an MDS with pending unlinked files may cause oops
4447 Details : target_finish_recovery() calls mds_postrecov() which returned
4448 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4449 considers this an error and immediately begins cleaning up the
4450 lov, just after starting the mds_lov process
4452 Severity : enhancement
4454 Description: Implement 'lfs df' to report actual free space on per-OST basis
4455 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4456 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4457 usage of objects; '-h' to report in human readable format.
4459 ------------------------------------------------------------------------------
4461 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4468 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4469 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4470 at startup during orphan unlinking. Instead, skip buddy update
4471 before setup, as it will be regenerated after recovery is complete.
4476 Description: If an OST is inactive, its locks might reference stale inodes.
4477 Details : lov_change_cbdata() must iterate over all namespaces, even if
4478 they are inactive to clear inode references from the lock.
4480 Severity : enhancement
4481 Frequency : occasional, if non-standard max_dirty_mb used
4483 Description: Client will block write RPCs if not enough grant
4484 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4485 then the client will block writes while waiting for another RPC
4486 to complete instead of consuming its dirty limit. With change
4487 we get improved performance when max_dirty_mb is small.
4489 Severity : enhancement
4490 Bugzilla : 3389, 6253
4491 Description: Add support for supplementary groups on the MDS.
4492 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4493 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4494 lookups for user supplementary groups into a cache.
4498 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4499 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4500 for liblustre O_EXCL is enforced by MDS.
4504 Description: readdir never returns NULL in liblustre.
4505 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4506 dentry in struct dirent.
4510 Frequency : liblustre only, depends on application IO pattern
4511 Description: liblustre clients evicted if not contacting servers
4512 Details : Don't put liblustre clients into the ping_evictor list, so
4513 they will not be evicted by the pinger ever.
4515 Severity : enhancement
4517 Description: Add ability to evict clients by NID from MDS.
4518 Details : By echoing "nid:$NID" string into
4519 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4520 $NID would be instantly evicted from this MDS and from all active
4521 OSTs connected to it.
4525 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4526 Details : lookup_it_finish() used to query file size from OSTs that was not
4531 Description: service threads change working directory to that of init
4532 Details : Starting lustre service threads may pin the working directory
4533 of the parent thread, making that filesystem busy. Threads
4534 now change to the working directory of init to avoid this.
4538 Frequency : during shutdown only
4539 Description: shutdown with a failed MDS or OST can cause unmount to hang
4540 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4545 Frequency : 2.6 only
4546 Description: chmod/chown may include an extra supplementary group
4547 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4548 supplementary group and if none is specified this is used.
4551 Bugzilla : 5479 (6816)
4553 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4554 Details : If lookup is in progress on a file that is unlinked we might try
4555 to revalidate the inode and fail in revalidate after lookup is
4556 complete and ll_file_open() enqueues the open again but
4557 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4560 Frequency : always, if lconf --abort_recovery used
4562 Description: lconf --abort_recovery fails with 'Operation not supported'
4563 Details : lconf was attempting to abort recovery on the MDT device and not
4566 ------------------------------------------------------------------------------
4568 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4573 Frequency : rare (only unsupported configurations with a node running as an
4575 Bugzilla : 6514, 5137
4576 Description: Mounting a Lustre file system on a node running as an OST could
4578 Details : OSTs now preallocates memory needed to write out data at
4579 startup, instead of when needed, to avoid having to
4580 allocate memory in possibly low memory situations.
4581 Specifically, if the file system is mounted on on OST,
4582 memory pressure could force it to try to write out data,
4583 which it needed to allocate memory to do. Due to the low
4584 memory, it would be unable to do so and the node would
4585 become unresponsive.
4587 Severity : enhancement
4589 Description: Addition of lconf --service command line option
4590 Details : lconf now accepts a '--service <arg>' option, which is
4591 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4593 Severity : enhancement
4595 Description: Failover mode is now the default for OSTs.
4596 Details : By default, OSTs will now run in failover mode. To return to
4597 the old behaviour, add '--failout' to the lmc line for OSTs.
4599 Severity : enhancement
4601 Description: Health checks are now provided for MDS and OSTs
4602 Details : Additional detailed health check information on MSD and OSTs
4603 is now provided through the procfs health_check value.
4606 Frequency : occasional, depends on IO load
4608 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4609 after numerous create/delete cycles
4610 Details : The ext3 inode allocation policy would not allocate new inodes
4611 very well on the OSTs because there are no new directories
4612 being created. Instead we look for groups with free space if
4613 the parent directories are nearly full.
4618 Description: Network or server problems during mount may cause partially
4619 mounted clients instead of returning an error.
4620 Details : The config llog parsing code may overwrite the error return
4621 code during mount error handling, returning success instead
4627 Description: MDS can fail to allocate large reply buffers
4628 Details : After long uptimes the MDS can fail to allocate large reply
4629 buffers (e.g. zconf client mount config records) due to memory
4630 fragmentation or consumption by the buffer cache. Preallocate
4631 some large reply buffers so that these replies can be sent even
4632 under memory pressure.
4636 Frequency : rare (liblustre)
4637 Description: fsx running with liblustre complained that using truncate() to
4638 extend the file doesn't work. This patch corrects that issue.
4639 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4640 ATTR_SIZE and lsm use in llu_setattr_raw.
4644 Frequency : rare, only 2.6 kernels
4645 Description: Unusual file access patterns on the MDS may result in inode
4646 data being lost in very rare circumstances.
4647 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4648 code in the 2.6 kernel caused the inode and/or EA data not to
4649 be read from disk, causing single-file corruption.
4653 Frequency : rare, only 2.6 filesystems using extents
4654 Description: Heavy concurrent write and delete load may cause data corruption.
4655 Details : It was possible under high-load situations to have an extent
4656 metadata block in the block device cache from a just-unlinked
4657 file overwrite a newly-allocated data block. We now unmap any
4658 metadata buffers that alias just-allocated data blocks.
4662 Frequency : filesystems with default stripe_count larger than 77
4663 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4664 Details : lconf specifies an inode size of 4096 bytes when the default
4665 stripe_count is larger than 77. This conflicts with the default
4666 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4667 this case to avoid pinning too much memory for large EAs.
4669 ------------------------------------------------------------------------------
4671 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4676 Frequency : rare (extremely heavy IO load with hundreds of clients)
4678 Description: Client is evicted, gets IO error writing to file
4679 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4680 caused clients to be evicted for AST timeouts. The fixes in
4681 bug 5192 mean we no longer need to have such short AST timeouts
4682 so ldlm_timeout has been increased.
4685 Frequency : occasional during --force or --failover shutdown under load
4686 Bugzilla : 5949, 4834
4687 Description: Server oops/LBUG if stopped with --force or --failover under load
4688 Details : a collection of import/export refcount and cleanup ordering
4689 issues fixed for safer force cleanup
4692 Frequency : only filesystems larger than 120 OSTs
4693 Bugzilla : 5990, 6223
4694 Description: lfs getstripe would oops on a very large filesystem
4695 Details : lov_getconfig used kfree on vmalloc'd memory
4698 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4699 Bugzilla : 6242, 6243
4700 Description: reading from files that had been truncated to a non-zero size
4701 but never opened returned no data
4702 Details : ll_file_read() reads zeros from no-object files to EOF
4707 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4708 Details : There's little harm in aborting MDS/OSS recovery and letting it
4709 try again, so I removed the LASSERT and return an error instead.
4711 Severity : enhancement
4713 Description: New debugging infrastructure for tracking down data corruption
4714 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4715 (b) cover more of the client-side code path, and (c) try to narrow
4716 down where problems occurred
4720 Bugzilla : 3819, 4364, 4397, 6313
4721 Description: Racing close and eviction MDS could cause assertion in mds_close
4722 Details : It was possible to get multiple mfd references during close and
4723 client eviction, leading to one thread referencing a freed mfd.
4725 Severity: : enhancement
4726 Bugzilla : 3262, 6359
4727 Description: Attempts to reconnect to servers are now more aggressive.
4728 Details : This builds on the enhanced upcall-less recovery that was added
4729 in 1.4.2. When trying to reconnect to servers, clients will
4730 now try each server in the failover group every 10 seconds. By
4731 default, clients would previously try one server every 25 seconds.
4736 Description: After recovery, certain operations trigger a failed
4737 assertion on a client.
4738 Details : Failing over an mds, using lconf -d --failover, while a
4739 client was doing a readdir() call would cause the client to
4740 LBUG after recovery completed and the readdir() was resent.
4742 Severity : enhancement
4744 Description: Default groups are now added by lconf
4745 Details : You can now run lconf --group <servicename> without having to
4746 manually add groups with lmc.
4749 Frequency : occasional
4751 Description: Nodes with an elan id of 0 trigger a failed assertion
4754 Frequency : always when accessing e.g. tty/console device nodes
4756 Description: tty and some other devices nodes cannot be used on lustre
4757 Details : file's private_data field is used by device data and lustre
4758 values in there got lost. New field was added to struct file to
4759 store fs-specific private data.
4762 Frequency : when exporting Lustre via NFS
4764 Description: NFSD failed occasionally when looking up a path component
4765 Details : NFSD is looking up ".." which was broken in ext3 directories
4766 that had grown large enough to become hashed.
4769 Frequency : Clusters with multiple interfaces not on the same subnet
4771 Description: Nodes will repeatedly try to reconnect to an interface which it
4772 cannot reach and report an error to the log.
4773 Details : Extra peer list entries will be created by lconf with some peers
4774 unreachable. lconf now validates the peer before adding it.
4777 Frequency : Only if a default stripe is set on the filesystem root.
4779 Description: Setting a default stripe on the filesystem root prevented the
4780 filesystem from being remounted.
4781 Details : The client was sending extra request flags in the root getattr
4782 request and did not allocate a reply buffer for the dir EA.
4785 Frequency : occasional, higher if lots of files are accessed by one client
4786 Bugzilla : 6159, 6097
4787 Description: Client trips assertion regarding lsm mismatch/magic
4788 Details : While revalidating inodes the VFS looks up inodes with ifind()
4789 and in rare cases can find an inode that is being freed.
4790 The ll_test_inode() code will free the lsm during ifind()
4791 when it finds an existing inode and then the VFS later attaches
4792 this free lsm to a new inode.
4796 Bugzilla : 6422, 7030
4797 Description: MDS deadlock between mkdir and client eviction
4798 Details : Creating a new file via mkdir or mknod (starting a transaction
4799 and getting the ns lock) can deadlock with client eviction
4800 (gets ns lock and trying to finish a synchronous transaction).
4803 Frequency : occasional
4804 Description: While starting a server, the fsfilt_ext3 module could not be
4806 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4807 kernels. Previously, lconf would still use the ext3 name
4808 when trying to load modules. Now, it will correctly use
4809 ext3 on 2.4 and ldiskfs on 2.6.
4811 Severity : enhancement
4812 Description: The default stripe count has been changed to 1
4813 Details : The interpretation of the default stripe count (0, to lfs
4814 or lmc) has been changed to mean striping across a single
4815 OST, rather than all available. For general usage we have
4816 found a stripe count of 1 or 2 works best.
4818 Severity : enhancement
4819 Description: Add support for compiling against Cray portals.
4820 Details : Conditional compiling for some areas that are different
4824 Frequency : occasional
4825 Bugzilla : 6409, 6834
4826 Description: Creating files with an explicit stripe count may lead to
4827 a failed assertion on the MDS
4828 Details : If some OSTs are full or unavailable, creating files may
4829 trigger a failed assertion on the MDS. Now, Lustre will
4830 try to use other servers or return an error to the
4834 Frequency : occasional
4836 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4837 caused lock timeout during readahead (since 1.4.2).
4838 Details : Processes doing readahead might match a lock that hasn't been
4839 granted yet if there are overlapping and conflicting lock
4840 requests. The readahead process waits on ungranted lock
4841 (original lock is CBPENDING), while OST waits for that process
4842 to cancel CBPENDING read lock and eventually evicts client.
4844 Severity : enhancement
4846 Description: Initial enabling of flock support for clients
4847 Details : Implements fcntl advisory locking and file status functions.
4848 This feature is provided as an optional mount flag (default
4849 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4850 locking are implemented yet, and those that are are not guaranteed
4851 to be completely correct in production environments.
4852 mount -t lustre -o [flock|noflock] ...
4855 Frequency : occasional
4857 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4858 assertion in the ext3 JBD (journaling) layer.
4859 Details : The b_committed_data struct is protected by the big kernel lock
4860 in 2.4 kernels, serializing journal_commit_transaction() and
4861 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4862 access to this struct. In 2.6 kernels there is finer grained
4863 locking to improve SMP performance of the JBD layer.
4867 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4872 Description: merge and cleanup kernel patches.
4875 -----------------------------------------------------------------------------
4877 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4879 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4880 versions of Lustre. Please update all servers and clients to
4881 version 1.4.2 or later at the same time. You must also run
4882 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4883 to update the configuration logs.
4885 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4886 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4887 - swab llog records of type '0' so we get proper header size/idx (5861)
4888 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4889 - fix rename of one directory over another leaking an inode (5953)
4890 - avoid SetPageDirty on 2.6 (5981)
4891 - don't re-add just-being-destroyed locks to the waiting list (5653)
4892 - when creating new directories, inherit the parent's custom
4893 striping settings if present parent (3048)
4894 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4895 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4896 - don't hold client locks on temporary worklist from l_lru (5666)
4897 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4898 - automatically evict dead clients (5921)
4899 - Update file size properly in create+truncate+fstat case (6196)
4900 - Do not unhash mountpoint dentries, do not allow removal of
4902 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4903 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4904 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4905 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4906 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4907 - Correctly handle reads of files with no objects (6243)
4908 - lctl recover will also mark a device active if deactivate used (5933)
4910 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4911 - allow --write-conf on an MDS with different nettype than client (5619)
4912 - don't write config llogs to MDS for mounts not from that MDS (5617)
4913 - lconf should create multiple TCP connections from a client (5201)
4914 - init scripts are now turned off by default; run chkconfig --on
4915 lustre and chkconfig --on lustrefs to use them
4916 - upcalls are no longer needed for clients to recover to failover
4918 - add --abort-recovery option to lconf to abort recovery on device
4920 - add support for an arbitrary number of OSTs (3026)
4921 - Quota support protocol changes.
4922 - forward compatibility changes to wire structs (6007)
4923 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4924 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4925 - improved "lustre status" script
4926 - initialize blocksize for non-regular files (6062)
4927 - added --disable-server and --disable-client configure options (5782)
4928 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4929 - Vanilla 2.4.29 support
4930 - increase maximum number of obd devices to 520 (6242)
4931 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4932 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4933 SLES 9 2.6.5 kernels
4934 - limit stripes per file to 160 (the maximum EA size) (6093)
4936 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4939 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4940 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4941 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4942 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4943 - small ext3 extents cleanups and fixes (5733)
4944 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4945 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4946 - avoid needless client->OST connect, fix handle mismatch (5317)
4947 - fix DLM error path that led to out-of-sync client, long delays (5779)
4948 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4949 - fix several locking issues related to i_size (5492,5624,5654,5672)
4950 - don't move pending lock onto export if it is already evicted (5683)
4951 - fix kernel oops when creating .foo in unlinked directory (5548)
4952 - fix deadlock in obdfilter statistics vs. object create (5811)
4953 - use time_{before,after} to avoid timer jiffies wrap (5882)
4954 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4955 - Do not leak request if server was not able to process it (5154)
4956 - If mds_open unable to find parent dir, make that negative lookup(5154)
4957 - don't create new directories with extent-mapping (5909, 5936)
4959 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4960 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4961 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4962 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4963 - support for RHEL4 kernel on i686 (5773)
4964 - provide error messages when incompatible logs are encountered (5898)
4966 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4967 * version 1.4.0.10 (1.4.1 release candidate 1)
4969 - don't keep a lock reference when lock is not granted (4238)
4970 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4971 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4972 - drop import inflight refcount on signal_completed_replay error (5255)
4973 - unlock page after async write error during send (3677)
4974 - handle missing objects in filter_preprw_read properly (5265)
4975 - no transno return for symlink open, don't save no-trasno open (3440)
4976 - don't try to complete elan receive that already failed (4012)
4977 - free RPC server reply state on error (5406)
4978 - clean up thread from ptlrpc_start_thread() on error (5160)
4979 - readahead could read extra page into cache that wasn't ejected (5388)
4980 - prevent races in class_attach/setup/cleanup/detach (5260)
4981 - don't dereference de->d_inode after l_dput of de (5458)
4982 - use "int" for stripe value returned from lock_to_stripe (5544)
4983 - mballoc allocation and error-checking fixes in 2.6 (5504)
4984 - block device patches to fix I/O request sizes in 2.6 (5482)
4985 - look up hostnames for IB nals (5602)
4986 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4987 - don't start multiple acceptors for the same port (5277)
4988 - fix incorrect LASSERT in mds_getattr_name (5635)
4989 - export a proc file for general "ping" checking (5628)
4990 - fix "lfs check" to not block when the MDS is down (5628)
4992 - service request history (4965)
4993 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4994 - create an "evict_client" /proc entry on OSTs, like the MDS has
4995 - fix mount usage message, return errors per mount(8) (5168)
4996 - change grep [] to grep "[]" in tests so they work in more UMLs
4997 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4998 - remove ancient LOV_MAGIC_V0 EA support (5047)
4999 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5000 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5001 - allow readahead allocations to fail when low on memory (5383)
5002 - mmap locking landed again, after considerable improvement (2828)
5003 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5005 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5008 - send OST transaction number in read/write reply to free req (4966)
5009 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5010 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5011 - the watchdog thread now runs as interruptible (5246)
5012 - flock/lockf fixes (but it's still disabled, pending 5135)
5013 - don't use EXT3 constants in llite code (5094)
5014 - memory shortage at startup could cause assertion (5176)
5016 - reorganization of lov code
5017 - single portals codebase
5019 - add extents/mballoc support (5025)
5020 - direct I/O reads in the obdfilter (4048)
5021 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5023 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5026 - send OST transaction number in read/write reply to free req (4966)
5027 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5028 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5029 - the watchdog thread now runs as interruptible (5246)
5030 - handle missing objects in filter_preprw_read properly (5265)
5031 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5032 - drop import inflight refcount on signal_completed_replay error (5255)
5033 - unlock page after async write error during send (3677)
5034 - return original error code on reconstructed replies (3761)
5035 - no transno return for symlink open, don't save no-trasno open (3440)
5037 - add pid to ldlm debugging output (4922)
5038 - bump the watchdog timeouts -- we can't handle 30sec yet
5039 - extra debugging for orphan dentry/inode bug (5259)
5041 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5044 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5045 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5046 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5047 - fix NULL dereference in /proc/sys/portals/routes (4827)
5048 - allow failed mdc_close() operations to be interrupted (4561)
5049 - stop precreate on OST before MDS would time out on it (4778)
5050 - don't send partial-page writes before EOF from client (4410)
5051 - discard client grant for sub-page writes on large-page clients (4520)
5052 - don't free dentries not owned by NFS code, check generation (4806)
5053 - fix lsm leak if mds_create_objects() fails (4801)
5054 - limit debug_daemon file size, always print CERROR messages (4789)
5055 - use transno after validating reply (3892)
5056 - process timed out requests if import state changes (3754)
5057 - update mtime on OST during writes, return in glimpse (4829)
5058 - add mkfsoptions to LDAP (4679)
5059 - use ->max_readahead method instead of zapping global ra (5039)
5060 - don't interrupt __l_wait_event() during strace
5062 - add software watchdogs to catch hung threads quickly (4941)
5063 - make lustrefs init script start after nfs is mounted
5064 - fix CWARN/ERROR duplication (4930)
5065 - return async write errors to application if possible (2248)
5066 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5067 - print NAL number in %x format (4645)
5068 - update barely-supported suse-2.4.21-171 series (4842)
5069 - support for sles 9 %post scripts
5070 - support for building 2.6 kernel-source packages
5071 - support for sles km_* packages
5073 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5076 - ignore -ENOENT errors in osc_destroy (3639)
5077 - notify osc create thread that OSC is being cleaned up (4600)
5078 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5079 - reconstruct ost_handle() like mds_handle() (4657)
5080 - create a new thread to do import eviction to avoid deadlock (3969)
5081 - let lconf resolve symlinked-to devices (4629)
5082 - don't unlink "objects" from directory with default EA (4554)
5083 - hold socknal file ref over connect in case target is down (4394)
5084 - allow more than 32000 subdirectories in a single directory (3244)
5085 - fix blocks count for O_DIRECT writes (3751)
5086 - OST returns ENOSPC from object create when no space left (4539)
5087 - don't send truncate RPC if file size isn't changing (4410)
5088 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5089 - bind to privileged port in socknal and tcpnal (3689)
5091 - rate limit CERROR/CWARN console message to avoid overload (4519)
5092 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5093 - basic mmap support (3918)
5094 - kernel patch series update from b1_4 (4711)
5096 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5099 - avoid crash during MDS cleanup with OST shut down (2775)
5100 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5101 - don't use bad inodes on the MDS (3744)
5102 - dynamic object preallocation to improve recovery speed (4236)
5103 - don't hold spinlock over lock dumping or change debug flags (4401)
5104 - don't zero obd_dev when it is force cleaned (3651)
5105 - print grants to console if they go negative (4431)
5106 - "lctl deactivate" will stop automatic recovery attempts (3406)
5107 - look for existing locks in ldlm_handle_enqueue() (3764)
5108 - don't resolve lock handle twice in recovery avoiding race (4401)
5109 - revalidate should check working dir is a directory (4134)
5111 - don't always mark "slow" obdfilter messages as errors (4418)
5113 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5116 - don't close LustreDB during write_conf until it is done (3860)
5117 - fix typo in lconf for_each_profile (3821)
5118 - allow dumping logs from multiple threads at one time (3820)
5119 - don't allow multiple threads in OSC recovery (3812)
5120 - fix debug_size parameters (3864)
5121 - fix mds_postrecov to initialize import for llog ctxt (3121)
5122 - replace config semaphore with spinlock (3306)
5123 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5124 - don't allow enqueue to complete on a destroyed export (3822)
5125 - down write_lock before checking llog header bitmap (3825)
5126 - recover from lock replay timeout (3764)
5127 - up llog sem before sending rpc (3652)
5128 - reduce ns lock hold times when setting kms (3267)
5129 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5130 - fix NULL deref and obd_dev leak on setup error (3312)
5131 - replace some LBUG about llog ops with error handling (3841)
5132 - don't match INVALID dentries from d_lookup and spin (3784)
5133 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5134 - fix invalid assertion in ptlrpc_set_wait (3880)
5136 - add libwrap support for the TCP acceptor (3996)
5137 - add /proc/sys/portals/routes for non-root route listing (3994)
5138 - allow setting MDS UUID in .xml (2580)
5139 - print the stack of a process that LBUGs (4228)
5141 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5144 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5145 - make sure to unset replay flag from failed open requests (3440)
5146 - if default stripe count is 0, use OST count for inode size (3636)
5147 - update parent mtime/ctime on client for create/unlink (2611)
5148 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5149 - free recovery state on server during a forced cleanup (3571)
5150 - unregister_reply for resent reqs (3063)
5151 - loop back devices mounting and status check on 2.6 (3563)
5152 - fix resource-creation race that can provoke i_size == 0 (3513)
5153 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5154 - more debugging for page-accounting assertion (3746)
5155 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5156 - avoid deadlock after precreation failure (3758)
5157 - fix race and lock order deadlock in orphan handling (3450, 3750)
5158 - add validity checks when grabbing inodes from l_ast_data (3599)
5160 - add /proc/.../recovery_status to obdfilter (3428)
5161 - lightweight CDEBUG infrastructure, debug daemon (3668)
5162 - change default OSC RPC parameters to be better on small clusters
5163 - turn off OST read cache for files smaller than 32MB
5164 - install man pages and include them in rpms (3100)
5165 - add new init script for (un)mounting lustre filesystems (2593)
5166 - run chkconfig in %post for init scripts (3701)
5167 - drop scimac NAL (unmaintained)
5169 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5172 - clean kiobufs before and after use (3485)
5173 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5174 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5175 - change init script to fail more clearly if not run as root (1528)
5176 - allow clients to reconnect during replay (1742)
5177 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5178 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5179 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5180 - fix cond_resched (3554)
5181 - only evict unfinished clients after recovery (3515)
5182 - allow bulk resend, prevent data loss (3570)
5183 - dynamic ptlrpc request buffer allocation (2102)
5184 - don't allow unlinking open directory if it isn't empty (2904)
5185 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5186 - remove extraneous obd dereference causing LASSERT failure (3334)
5187 - don't use get_cycles() when creating temp. files on the mds (3156)
5188 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5189 - handle EEXIST for set-stripe, set proper directory name (3336)
5191 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5192 - fix ksocknal_fmb_callback() error messages (2918)
5194 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5197 - don't copy lvb into (possibly NULL) reply on error (2983)
5198 - don't deref dentry after dput, don't free lvb on error (2922)
5199 - use the kms to determine writeback rpc length (2947)
5200 - increment oti_logcookies when osc is inactive (2948)
5201 - update client's i_blocks count via lvb messages (2543)
5202 - handle intent open/close of special files properly (1557)
5203 - mount MDS with errors=remount-ro, like obdfilter (2009)
5204 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5205 - don't use cancelling-locks' kms values (2947)
5206 - use highest lock extent for kms, not last one (2925)
5207 - don't dereference ERR_PTR() dentry in error handling path (3107)
5208 - fix thread race in portals_debug_dumplog() (3122)
5209 - create lprocfs device entries at setup instead of at attach (1519)
5210 - common AST error handler, don't evict client on completion race (3145)
5211 - zero nameidata in detach_mnt in 2.6 (3118)
5212 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5213 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5214 - initialize RPC timeout timer earlier for 2.6 (3219)
5215 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5216 - print nal/nid for unknown nid (3258)
5217 - additional checks for oscc recovery before doing precreate (3284)
5218 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5219 - don't crash in mdc_close for bad permissions on open (3285)
5220 - zero i_rdev for non-device files (3147)
5221 - clear page->private before handing to FS, better assertion (3119)
5222 - tune the read pipeline (3236)
5223 - fix incorrect decref of invalidated dentry (2350)
5224 - provide read-ahead stats and refine rpc in flight stats (3328)
5225 - don't hold journal transaction open across create RPC (3313)
5226 - update atime on MDS at close time (3265)
5227 - close LDAP connection when recovering to avoid server load (3315)
5228 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5229 - don't leak open file on MDS after open resend (3325)
5230 - serialize filter_precreate and filter_destroy_precreated (3329)
5231 - loop device shouldn't call sync_dev() for nul device (3092)
5232 - clear page cache after eviction (2766)
5233 - resynchronize MDS->OST in background (2824)
5234 - refuse to mount the same filesystem twice on same mountpoint (3394)
5235 - allow llmount to create routes for mounting behind routers (3320)
5236 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5237 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5238 - fix rare problem with rename on htree directories (3417)
5240 - allow default OST striping configuration per directory (1414)
5241 - fix compilation for qswnal for 2.6 kernels (3125)
5242 - increase maximum number of MDS request buffers for large systems
5243 - change liblustreapi to be useful for external progs like lfsck (3098)
5244 - increase local configuration timeout for slow disks (3353)
5245 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5247 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5250 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5251 - don't overwrite extent policy data in reply if lock was blocked (2901)
5252 - drop filter export grants atomically with removal from device (2663)
5253 - del obd_self_export from work_list in class_disconnect_exports (2908)
5254 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5255 - swab reply message in mdc_close, other PPC fixes (2464)
5256 - fix destroying of named logs (2325)
5257 - overwrite old logs when running lconf --write_conf (2264)
5258 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5259 - fix race in target_handle_connect (2898)
5260 - mds_reint_create() should take same inode create lock (2926)
5261 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5262 - don't close files for self_export to avoid uninitialized obd (2936)
5263 - allow MDS with the same name as client node (2939)
5264 - hold dentry reference for closed log files for unlink (2325)
5265 - reserve space for all logs during transactions (2059)
5266 - don't evict page beyond end of stripe extent (2925)
5267 - don't oops on a deleted current working directory (2399)
5268 - handle hard links to targets without a parent properly (2517)
5269 - don't dereference NULL lock when racing during eviction (2867)
5270 - don't grow lock extents when lots of conflicting locks (2919)
5272 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5275 - account for cache space usage on clients to avoid data loss (974)
5276 - lfsck support in lustre kernel code (2349)
5277 - reduce journal credits needed for BRW writes (2370)
5278 - orphan handling to avoid losing space on client/server crashes
5279 - ptlrpcd can be blocked, stopping ALL progress (2477)
5280 - use lock value blocks to assist in proper KMS, faster stat (1021)
5281 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5282 - recovery for initial connections (2355)
5283 - fixes for mds_cleanup_orphans (1934)
5284 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5285 - block all file creations until orphan recovery completes (1901)
5286 - client remove rq_connection from request struct (2423)
5287 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5288 - recovery timer race (2670)
5289 - mdc_close recovey bug (2532)
5290 - ptlrpc cleanup bug (2710)
5291 - mds timeout on local locks (2588)
5292 - namespace lock held during RPCs (2431)
5293 - handle interrupted sync write properly (2503)
5294 - don't try to handle a message that hasn't been replied to (2699)
5295 - client assert failure during cleanup after abort recovery (2701)
5296 - leak mdc device after failed mount (2712)
5297 - ptlrpc_check_set allows timedout requests to complete (2714)
5298 - wait for inflight reqs when ptlrpcd finishes (2710)
5299 - make sure unregistered services are removed from the srv_list
5300 - reset bulk XID's when resending them (caught by 1138 test)
5301 - unregister_bulk after timeout
5302 - fix lconf error (2694)
5303 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5304 - readahead locks pages, leaves pending causing memory pressure (2673)
5305 - increase OST request buffers to 4096 on large machines (2729)
5306 - fix up permission of existing directories in simple_mkdir (2661)
5307 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5308 - don't assign transno to errored transactions (2742)
5309 - don't delete objects on OST if given a bogus objid from MDS (2751)
5310 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5311 - if rq_no_resend, then timeout request after recovery (2432)
5312 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5313 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5314 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5315 - MDS recovery shouldn't delete objects that it has given out (2730)
5316 - if enqueue arrives after completion, don't clobber LVB (2819)
5317 - don't unlock pages twice when trigger_group_io returns error (2814)
5318 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5319 - don't write pages to disk if there was an error (1450)
5320 - don't ping imports that have recovery disabled (2676)
5321 - take buffered bytes into account when balancing socknal conn (2817)
5322 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5323 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5324 - remove little-endian swabbing of llog records (1987)
5325 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5326 - retry reposting request buffers if they fail (1191)
5327 - grow extent at grant time to avoid granting a revoked lock (2809)
5328 - lock revoke doesn't evict page if covered by a second lock (2765)
5329 - disable VM readahead to avoid reading outside lock extents (2805)
5331 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5332 - updated kernel patches for hp-2.4.20 kernel (2681)
5334 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5337 - fix truncated write corruption (2366)
5338 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5340 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5341 - ensure len(uuid) < 37 in lmc (1171)
5342 - fix ia64 OOPS in llog_test (2255)
5343 - zero end of page at obdfilter for partial page writes (2648)
5344 - don't leave stale dentries around after renames (bug 2428)
5345 - fix timeouts when evicting a client with a single lock held (2642)
5346 - set deadline for the initial HELLO message to drain (2634)
5347 - print out dotted-quad IP addresses in the socknal (2302)
5349 - additional debugging for MDS client eviction problem (2443)
5350 - fix mkfsoptions support for osts (2603, 2604)
5352 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5355 - add series for the vanilla 2.6.0 kernel
5356 - add series for the vanilla 2.4.24 kernel
5357 - add series for a cray x86/64 UL kernel drop
5358 - fix xattr patches for the vanilla 2.4.19 series
5360 - generate true UUIDs in lmc (1171)
5361 - have portals stack dumping break in UML (2466)
5362 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5363 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5364 - wake callback waiting threads on client eviction (2460)
5365 - Add --ptldebug and --subsystem to lmc (1719)
5366 - update assertion to allow safe interrupt allocation
5367 - set rq_no_resend for cancel requests (2432)
5368 - recalculate ptlrpcd timeout after resend (2494)
5369 - call vfs_rmdir when removing pending directories (2368)
5370 - fix renaming a file to itself (2429)
5371 - lmc creates a default one-stripe lov (2454)
5372 - expand procfs space to handle large clusters (2326)
5373 - increase UML stack to avoid overflow
5374 - update lconf's list of debug and subsystem masks
5375 - fix lfs find --obd (2510)
5376 - /proc tunable for disabling filter read caching (2591)
5377 - stop rpm packages from altering slapd.conf (2301)
5378 - disable nagle in the socknal under 0conf (2578)
5379 - choose mds inode size based on stripe count (2572)
5380 - fix kernel-source rpm problems (2516)
5382 - add --disable-doc to avoid pdf generation (2421)
5383 - update documentation, tests, type-os, comments
5384 - avoid format warnings on ia64
5385 - remove the TOE NAL
5386 - tiny code cleanups by removing unused fields
5388 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5391 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5392 - remove the most blatant lies from BUILDING file (2371)
5393 - change default debug level to reasonable production setting
5394 - reduce client side cache size to reduce cache flush time
5395 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5396 - make TCP zerocopy and pinger support enabled by default (2476)
5397 - sync writes completed after process exits caused crashes (2319)
5398 - maintain correct mount count on the MDS (2356)
5399 - backout 1557, because 2316 wasn't really fixed
5400 - better file I/O statistics gathering in /proc
5401 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5402 - another kernel patch to fix zero-copy TCP function export
5403 - don't take duplicate lock when processing re-sent getattr (2420)
5404 - lctl uses obd_self_export instead of creating new conn (2353)
5405 - MDS/OST recovery case which requires object creation asserted (2425)
5406 - move lfs from /usr/sbin to /usr/bin in packages
5407 - fix race between mds_client_add and mds_client_free (2417)
5408 - use kmalloc instead of slabs in portals (2430)
5409 - don't create duplicate records when a failover MDS is present (2442)
5410 - remove unnecessary mount age check (2332)
5411 - don't remove directory inodes from locks prematurely (2451)
5412 - don't break if MDS service name is the same as hostname (2103)
5413 - fix races in client write RPC generation when cache full (2482)
5415 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5418 - remove now-unused request->rq_obd (278)
5419 - if an allocation fails, print out how much memory we've used (1933)
5420 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5421 - add the "configurable stack size" patch to most series files (1256)
5422 - ability to write large log records, for 100+ OST configs (2306)
5423 - fix NULL deref when filter_prep fails (2314)
5424 - fix operator precedence error in filter_sync
5425 - dynamic allocation of socknal TX descriptors (2315)
5426 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5427 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5428 - ensure that CWARN messages in Portals always get to the syslog
5429 - __init/__exit are not for prototype decls (ldlm_init/exit)
5430 - x86-64 compile warning fixes
5431 - fix gateway LMC keyword conflict (2318)
5432 - fix MDS lock inversions in getattr/reint paths (1844)
5433 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5434 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5435 - DLM race condition prevented some lock evictions (2328)
5436 - ENOMEM detection and retry on socknal sends (2230)
5437 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5438 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5439 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5440 - better error messages when a client is rejected during recovery (1505)
5441 - avoid cancelling locks which were never granted, after failure (2330)
5442 - fix i_sem/journal inversion in mds_client_add (2333)
5443 - fix truncate/getattr lock cycle deadlock (2334)
5444 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5445 - fix two rare exit paths which could leak an l_lock() ref (2321)
5446 - fencepost error in MDS/OST orphan recovery (2226)
5447 - make log record alignment 8 bytes (1988)
5448 - lstripe now fails when requested offset > ost_count (2237)
5449 - ensure that all kernel series have a complete list.h (1607)
5450 - fix crashes in special-file operations (2316)
5451 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5453 - allow configurable automake binary, for testing new versions
5454 - small update to the lfs documentation
5456 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5458 * fix negative export reference count in fsfilt_sync (2312)
5460 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5461 * release candidate 0.9.1
5463 - orphans are moved into the PENDING directory for possible recovery
5464 - replayed opens now open by fid for orphan/rename safety (1042)
5465 - last close of an orphan inode generates a transno (683)
5466 - chdir() and mount() now pin the directory entry (1020)
5467 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5468 - discard very old requests without processing them (1502)
5469 - remove some common, well-understood CERRORs (1505)
5470 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5471 - clear "grant" flags in OST replies until OST grant code lands (1644)
5472 - fix read performance by not clobbering i_blksize on client (1598)
5473 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5474 - make LRU size a /proc tunable, clears locks when reduced (707)
5475 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5476 - server locks take references on exports now (1558)
5477 - build fixes for 2.4.20-rh trees (1663)
5478 - return an error from lov_create if all OSCs are inactive (1751)
5479 - fix import levels when a reconnect happens without a timeout (1597)
5480 - exit early from mds_open if we get a lookup error (1749)
5481 - partial page read at EOF wouldn't wait for disk before sending (1642)
5482 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5483 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5484 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5485 - write/truncate lock inversion (1639)
5486 - Don't auto-load obdclass, portals modules during cleanup (1495)
5487 - fix timestamps from jumping to "now" (1763)
5488 - extra journal assertions (1648)
5489 - add an extra multiunlink test (1771)
5490 - fix read_record/write_record API (1776)
5491 - fix leak of offset_extent, possible incorrect i_size later (1772)
5492 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5493 - Debugging for the kqswnal_get_idle_tx problems (1820)
5494 - Allow recovery to be attempted multiple times (1536)
5495 - Write out MDS last_rcvd file after it is first created (1600)
5496 - Fix tx_descriptor leak in failed transmit situations (1827)
5497 - ext3 journaling fixes for assertion failure after IO error (1871)
5498 - class_export_put() on freed export after completion AST error (1896)
5499 - Fix revalidate looping in VFS (1322)
5500 - Don't access a freed export during MDS_REINT timeout (1521)
5501 - Add open-unlink recovery support on the MDS (1673,1764)
5502 - Return an error if no MDS data was read from last_rcvd (1946)
5503 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5504 - Don't setup a disk device that doesn't match exported UUID (317)
5505 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5506 - avoid committing NULL handle in force close
5507 - local.sh is now a one-stripe LOV configuration
5508 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5509 - fix ext3 htree duplicate directory entry corruption (1516)
5510 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5511 - update to new LOV EA format (2097)
5512 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5514 2003-06-15 Phil Schwan <phil@clusterfs.com>
5517 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5518 - per-import/export recovery handling (958, 931, 959)
5519 - multiple last-rcvd slots, for serving multiple FSes (949)
5520 - connections are again shared between multiple imp/exports (963, 964)
5521 - "umount -f" would hang if any requests needed to be sent (393, 978)
5522 - avoid pinning large req buffer by copying for queued messages (989)
5523 - add "uuid" to "lctl device" command to help upcalls (991)
5524 - "open" RPCs with transnos would confuse recovery counters (1037)
5525 - do proper endian conversion of all wire messages (288, 340, 891)
5526 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5527 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5528 - fix ptlrpc_connection leak in target_handle_connect (1174)
5529 - fix import refcounting bug in OST and MDS cleanup (1134)
5530 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5531 - fix very unlikely obd_types race condition (501)
5532 - remove osc_open hack for echo_client (1187)
5533 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5534 - a failure in read_inode2 leads to deadlock (1139)
5535 - cancel ack-locks as soon as transaction is committed (1072)
5536 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5537 - make sure to commitrw after any preprw to avoid deadlock (1162)
5538 - failing to execute a file in a lustre FS would lock inode (1203)
5539 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5540 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5541 - obd_uuid2tgt would walk off the end of the list (1255)
5542 - on IA64 the osc would give portals incorrect bulk size (1258)
5543 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5544 - fix lock inversion caused by new llite matching code (1282)
5545 - limit the number of dirty pages on a client to 10MB (1286)
5546 - timed out locks were not being corrected cancelled (1289)
5547 - fix O_DIRECT above 4GB on IA-32 (1292)
5548 * major user-visible changes
5549 - fail out/fail over policy now controlled by the upcall (993)
5551 - add OBD_PING to check server availability and failure (954)
5552 - lustre messages are now sent in sending host order (288, 340, 891)
5553 - add eadatalen to MDS getattr reply (340)
5554 - OST read replies may contain second buffer, with per-page status (593)
5556 2003-03-11 Phil Schwan <phil@clusterfs.com>
5559 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5560 - failed open()s could cause deadlock; fixed (867, 869)
5561 - stop cancelling OST locks when files are closed (481)
5562 - overlapping XID spaces caused network corruption (851, 853)
5563 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5564 - setattr_raw added, to do single-RPC, server-side setattrs
5565 - lmc/lconf syntax change for OST UUIDs
5566 - fix crashy race condition between ptlrpc_free_req and osc_close
5567 - don't use request in mdc_enqueue if we hit a timeout (889)
5568 - don't set the inode i_size for regular files from the MDS (896)
5569 - handle out of order completion AST (842)
5570 - don't LBUG if a lock request times out after receiving AST (913)
5571 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5572 - if a bad lock AST arrives, send an error instead of dropping entirely
5573 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5574 - fix leak in bulk IO when only partially completed (899, 900, 926)
5575 - fix O_DIRECT for ia64 (55)
5576 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5577 - C-z after timeout could hang a process forever; fixed (977)
5579 - client-side I/O cache (678, 924, 929, 941, 970)
5581 - READPAGE and SETATTRs which don't take server-side locks get
5584 2003-02-11 Phil Schwan <phil@clusterfs.com>
5587 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5588 - Fix fsfilt_extN_readpage() to read a full page of directory
5589 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5590 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5591 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5592 - Most intent operations were removed, in favour of a new RPC mode
5593 that does a single RPC to the server and bypasses most of the VFS
5594 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5595 - Aggressively cancel local locks on DLM servers
5596 - mds_reint_unlink sends EA to the client if it's the last nlink.
5597 client uses that EA to unlink OST objects.
5598 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5599 - recursive symlinks were fixed (439)
5600 - fixed NULL deref in DEBUG_REQ
5601 - filter_update_lastobjid no longer calls sync, which annoyed extN
5602 - fixed multi-client small-writes to a single file problem (445)
5603 - fixed mtime updates during file writes (607)
5604 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5605 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5606 - fixed Linux OST size reporting problem (444, 656)
5607 - OST now updates object mtime with writes or setattr (607, 619)
5608 - client verifies file size before zeroing page past EOF (445)
5609 - OST now writes last allocated objid to disk with allocation (108)
5610 - LOV on echo now works (409)
5612 - mds_reint_unlink sends a new buffer, with the EA included. this
5613 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5614 set if a regular file was being unlinked, and it was the last link
5615 - use PtlGet from the target for bulk writes (315)
5616 - OST now updates object mtime with writes or setattr (607, 619)
5617 - LDLM now has a grant-time callback to revalidate locked items, if
5619 - Many MDS operations were reorganized to combat race conditions
5621 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5622 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5624 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5627 - Fully reactivate OST imports after reconnection (512, others)
5628 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5629 - More graceful error handling for truncating on dead OST (515)
5630 - Don't error out unless we're actually accessing dead stripes (474)
5631 - Fix garbage sizes when stripes are missing (410)
5632 - LRU counters were broken, causing constant lock purge (433, 432)
5633 - garbage on read from stripes with failed OSTs (441)
5634 - mark OSCs as active before reconnecting during recovery (438)
5635 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5636 - lfind did not preserve OST order in output (443)
5637 - symlinks cause hung clients, incorrect data (439)
5638 - stop dereferencing request after dropping refcount (457)
5639 - don't LASSERT(spin_is_locked) on non-SMP (455)
5640 - fixes for many rename() bugs
5641 - fstat didn't correctly synchronize attributes (399)
5642 - server must handle lock cancellation during blocking AST prep (487)
5643 - bulk descriptors were free()d too soon (511)
5644 - fix paths in lconf, which would load incorrect modules (451, 507)
5645 - fix confusing lconf 'host not found' error message (386)
5646 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5647 - fix race condition in mdc_blocking_ast() for inode access (526)
5648 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5649 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5650 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5651 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5652 - fix lstripe in several ways (kernel side) (527)
5653 - fix request leak in ldlm_cli_enqueue (262)
5654 - incorrect OSC was marked inactive after OST failure
5655 - call mds_fs_cleanup before unmounting filesystem (524)
5656 - fix races between taking ns_lock and ldlm_lock_change_resource
5657 - fix races updating LOV export open file list
5658 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5659 - fix recovery NULL deref in ldlm_cli_cancel_unused
5660 - fix some DLM races by using new hash table for lock handles (419)
5661 - permit the client to specify desired inodes, at replay
5662 - duplicate requests when we queue them for replay reintegration
5663 - fix last_rcvd offset calculation
5664 - sync after each recovered transaction, so we always make progress
5665 - never, not always, ERESTART requests without transnos
5666 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5668 - skip replay if the MDS says that the client is already connected
5669 - don't check for a recovery-enabled export to match lctl's UUID
5670 - don't INC_USE_COUNT for phantom exports
5671 - don't crash when cleaning up phantom exports (567)
5672 - don't double-finish or set replay data for errored mdc_open requests
5673 - abort requests when they time out, so we don't get old replies
5674 - send/receive replies for AST messages again
5675 - if the client says that it doesn't have the lock, cancel it on the
5677 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5678 mark it as destroyed, it will all work out when decref is called
5679 - fix module use counts (22, 581)
5681 - ASTs now expect a reply (server cancels lock on error reply)
5683 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5686 - fix many simultaneous client startup (392)
5687 - fix dentry->d_it clobbering
5688 - credentials weren't being shipped for readdir/getattr operations
5689 - remove invalid assertions triggered during some concurrent MD
5691 - proper Lustre versions added (336, 389)
5692 - fix memory leak for create error case (398)
5693 - fix LOV locking bug that would get cli/srv out of sync
5694 - fix echo client over LOV (409)
5695 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5696 - fix double-O_EXCL intent crash (424)
5697 - avoid sending multiple lock CANCELs (352)
5699 - MDS can do multi-client recovery (modulo bugs in new code)
5701 - many updates, edits, cleanups
5703 2002-11-18 Phil Schwan <phil@clusterfs.com>
5706 - fix null d_it dereference (346)
5707 - fix full OST/dbench hang (333)
5708 - fix permission problem with file removal (286)
5709 - fix removal of OSCs from LOV when they fail
5710 - fix NULL deref during bulk timeout (214)
5711 - fix problems related to multiple filesystems on one MDS (241)
5712 - fixed serious subtle metadata locking bugs
5713 - free locks on clients when inodes are removed due to memory
5715 - fix inode pointer in lock data (285)
5716 - partial support for multiple MDS on a single host (241)
5717 - data locks weren't cancelled at clear_inode time (290, 311)
5718 - intent locks could lead to unbounded lock growth (205)
5719 - added a maximum lock count, an LRU list, and a flusher
5720 - fix multiple rename (365)
5721 - properly abstracted the echo client
5722 - OSC locked 1 byte too many; fixed
5723 - rewrote brw callback code:
5724 - fixed recovery bugs related to LOVs (306)
5725 - fixed too-many-pages-in-one-write crash (191)
5726 - fixed (again) crash in sync_io_timeout (214)
5727 - probably fixed callback-related race (385)
5729 - Add capability to MDS protocol
5730 - LDLM cancellations and callbacks on different portals
5732 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5735 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5736 - defer open object destruction to close time (601981, 138)
5737 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5738 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5739 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5740 - don't call obd_setattr in ll_file_release for destroyed objects
5741 * protocol change to lustre_msg: move |version| and add |flags|
5742 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5743 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5744 * added replay of create, unlink, link and rename operations during
5745 MDS failover; recovery should be much more robust now
5746 * remove failed OSCs from LOVs (only lov_create uses this so far)
5747 * the lustre-HOWTO was brought (more) up to date (582544)
5749 2002-10-23 Phil Schwan <phil@clusterfs.com>
5752 - in-use dentries weren't being reused properly (617851)
5753 - prevent multiple LDLM setup (599178)
5754 - fix LOV size calculations for truncate (617853)
5755 - fix client handling of MDS intent errors (POSIX)
5756 - fix permission bug in lovstripe.c test (624321)
5757 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5758 - truncate past end of file could corrupt data
5759 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5760 - a race in recovery could return ETIMEDOUT to apps (623947)
5761 - building outside the source directory was fixed
5762 * the lustre-HOWTO was brought (more) up to date (582544)
5763 * major progress was made on recovery functionality
5765 2002-10-10 Phil Schwan <phil@clusterfs.com>
5768 - recovery deadlock fix
5769 - rm -rf causes LBUG fix (617817)
5770 - file open by multiple tasks fix (618962)
5771 - directory permissions bugs (602707 and 620007)
5772 - journal_stop fixed with locking (611313)
5773 - O_APPEND failures resolved (618273, perhaps 614459)
5774 - lconf PATH fix (619770)
5775 - IA64 build fix (621450)
5776 - RPC buffer sizes scale with amount of memory
5778 2002-10-01 Phil Schwan <phil@clusterfs.com>
5781 - locks would be cancelled without throwing away data pages,
5782 resulting in inconsistent data (605627)
5783 - inode attributes were not always being refreshed (605627, 612449)
5784 - lconf now continues to cleanup after lctl reports an error
5785 - MDS now enforces user permissions (602707)
5786 - lprocfs cleanup fixed, but not yet enabled (614157)
5787 - fixed infinite server hang, should a client not respond to an AST
5788 - avoid going into recovery if user calls readlink() with a buffer
5789 that's too small (613941)
5790 - AST RPCs no longer require replies (614867) -- this may be changed
5791 - don't crash server if client sends an IOV that's too big (611336)
5792 - fixed lock conversion deadlock (611892)
5793 - fixed the following of symlinks (614622)
5794 * recovery: the server can remove locks from a client that dies, other
5795 clients can make progress
5796 * more extN patch fixes
5797 * compile-time configurable ptlrpc buffer allocations
5799 - collaborative read cache document
5800 - Lustre Lite Performance CDR document-in-progress
5802 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5805 - fix typo in patch-2.4.18
5807 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5810 - clear ptlrpc request each time in handle_incoming_request()
5811 - unlink of files now destroys the object on the OST
5813 2002-09-19 Peter Braam <braam@clusterfs.com>
5815 * add hard link support
5816 * change obdfile creation method
5817 * kernel patch changed
5819 2002-09-19 Peter Braam <braam@clusterfs.com>
5822 - stack overflow bug in extN fixed
5824 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5826 * documentation updates
5827 - add man pages for config tools
5828 - update tests/README to describe testing with new config tools
5829 - finish metadata API descriptions
5830 * bug fixes and cleanups
5831 - statfs workaround for 16TB limit
5832 - LOV stripe allocation improved, can stripe on subset of OSTs
5833 - LOV file size/IO offset was wrong for files > 4GB in size
5834 - object EA data was being dropped, caused files to be unreadable
5835 - memory overflow with non-LOV OST caused memory corruption
5836 - fixed regression tests to work with new config tools, obdfilter
5837 - fixed bug when directory size became larger than 1 block
5838 - fixed bug (for single client case) when PWD was deleted
5839 - invalidate local directory pages when doing intent-based ops
5840 - avoid LDLM oops when lock callback contained bad data
5842 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5844 * documentation updates
5845 * bug fixes and cleanups
5846 - configuration tools
5849 - 64-bit compile warnings
5850 - 64-bit internal statfs data
5852 * test_brw on persistent OST devices
5854 * lprocfs (disabled)
5856 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5858 * documentation updates
5859 * bug fixes and cleanups
5860 * configuration tools
5862 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5865 * many small fixes to 0_5_4
5866 * io/network handling
5867 * thinkos in MDS operations
5869 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5872 * crucial basic fixes to 0.5.3
5873 * IOR, Iozone work over Elan
5876 2002-08-07 Phil Schwan <phil@clusterfs.com>
5877 * version 0_5_3, our first alpha
5878 * we use the new Portals iovs
5879 * documentation updates
5880 * bug fixes and cleanups
5881 * small changes in the DLM wire protocol
5883 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5884 * version 0_5_1 with some initial stability,
5885 * locking on MD and file I/O.
5886 * documentation updates
5887 * several bug fixes since 0.5.0
5888 * small changes in wire protocol
5890 2002-07-18 Phil Schwan <phil@clusterfs.com>
5892 * delivered as Lustre Light Alpha
5893 * fixed a crash after handling invalid MDS requests
5894 * fixed directory pages for architectures with non-4k pages sizes
5896 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5897 * release version v0_4_4
5898 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5900 * Fixes a number of interruption problems with OST operations.
5901 * Update documentation for portals header changes
5902 * Move all wire protocol structs/defines to lustre_idl.h
5903 * Fixes symlink length bug.
5904 * Add tcpdump to repository.
5906 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5907 * release version v0_4_3
5908 * Fixes statfs for inodes on extN.
5909 * Fixes bug in runtests which would delete /etc/hosts.
5910 * Use 64-bit object IDs wherever possible (not into VFS though)
5911 Remove ost_get_info, which is unused by lustre, and out of date.
5913 2002-07-03 Peter Braam <braam@clusterfs.com>
5914 * release version v0_4_2 Fixes a lookup error (type not passed)
5915 * move forward to head of Portals
5916 * move forward to latest Lustre kernel
5918 2002-06-25 Peter Braam <braam@clusterfs.com>
5919 * release version v0_4_1. Hopefully stable on single node use.