1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.39.3 (SLES 10),
5 2.6.18-128.7.1.el5 (RHEL 5),
6 2.6.18-128.7.1.el5 (OEL 5),
7 2.6.22.14 vanilla (kernel.org).
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.16 - 2.6.21 vanilla (kernel.org)
11 * Recommended e2fsprogs version: 1.41.6.sun1
12 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
13 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
14 removed cwd "./" (refer to Bugzilla 14399).
15 * File join has been disabled in this release, refer to Bugzilla 16929.
17 Severity : enhancement
19 Description: Adjust locks' extents on their first enqueue, so that at the time
20 they get granted, there is no need for another pass through the
21 queues since they are already shaped into the proper forms.
25 Description: Fix in ptlrpc_expire_one_request() to print the signed time
28 Severity : enhancement
30 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
33 Severity : enhancement
35 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
37 Severity : enhancement
39 Description: correctly handle big reply message.
40 Details : send LNet event if reply is bigger then buffer and adjust this buffer
45 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
47 Severity : enhancement
49 Description: Add support for OEL5.
53 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
57 Description: client could not reconnect to OST because of an active request.
58 Details : The client now retries to reconnect to the same server, if a
59 connect request failed with EBUSY or -EAGAIN. Test case 35b
60 conf-sanity.sh has been created.
64 Description: Don't allow make backward step on assignin osc next id.
65 Details : race between allocation next id and ll_sync thread can be cause
66 of set wrong osc next id and can be kill valid ost objects.
70 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
74 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
78 Description: Error handling in osc_statfs_interpret() has been improved.
79 Details : Check in osc_statfs_interpret() for EBADR.
83 Description: Do not send statfs() requests to OSTs disabled by administrator.
84 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
86 Severity : enhancement
88 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
90 Severity : enhancement
92 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
94 Severity : enhancement
96 Description: Enhancement for lfs(1) command to use numeric uid/gid.
100 Description: Improved cleanup procedure for conf-sanity test_46a.
104 Description: Do not start bulk transfer, if client tried to reconnect.
105 Details : Do not start bulk transfer, if client tried to reconnect,
106 but failed due to active rpcs. Check for exp_abort_active_req
107 has been implemented.
111 Description: Prevent inconsistences between linux and lustre mount structures.
112 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
113 Make the sleep interruptible.
115 Severity : enhancement
117 Description: provide server to client comms path
118 Details : server to client communications path via new LDLM_SET_INFO rpc
122 Description: truncate starts GFP_FS allocation under transaction causing deadlock
123 Details : ldiskfs_truncate calls grab_cache_page which may start page
124 allocation under an open transaction. This may lead to
125 calling prune_icache with consequent lustre reentrance.
129 Description: send correctly lsm on open replay
130 Details : MDS is trust to LSM size on replay open, but client can set wrong size
135 Description: enable client interface failover
136 Details : When a child reconnects from another NID, properly update export
137 nid hash position and ldlm reverse import.
139 Severity : enhancement
141 Description: Communicate OST degraded/readonly state via statfs to MDS
142 Details : Flags in the statfs returned from OSTs indicate whether the
143 OST is in a degraded RAID state, or if the filesystem has
144 turned read-only after a filesystem error is detected.
149 Description: don't panic if EPROTO was hit when reading symlink
150 Details : correctly handling request reference in error cases.
152 Severity : enhancement
154 Description: Add LustreNetLink, a kernel-userspace communcation path.
156 Severity : enhancement
158 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
163 Description: access to llog context before init.
164 Details : move handling CATALOGS file at osc layer and forbid access to llog
168 Frequency : rare, only when client is writing to OST on local node
170 Description: Avoid deadlock for local client writes
171 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
172 memory freeing context. This allows OST threads to set the
173 PF_MEMALLOC flag on task structures in order to allocate memory
174 from reserved pools and complete IO.
175 Use GFP_HIGHUSER for OST allocations for non-local client writes,
176 so that the OST threads generate memory pressure and allow
177 inactive pages to be reclaimed.
179 Severity : enhancement
181 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
186 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
187 Details : move ldlm namespace creation in setup phase to avoid grab
188 _lprocfs_lock with cli_sem held.
192 Description: Temporarily disable grant shrink.
193 Details : Disable the feature for debugging.
196 Frequency : only when formatting test filesystems
198 Description: Unable to run several mkfs.lustre on loop devices at the same time
199 Details : mkfs.lustre returns error 256 on the concurrent loop devices
200 formatting. The solution is to proper handle the error.
202 Severity : enhancement
204 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
206 Severity : enhancement
208 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
213 Description: resolve race between obd_disconnect and class_disconnect_exports
214 Details : if obd_disconnect will be called to already disconnected export he
215 forget release one reference and osc module can't unloaded.
217 Severity : enhancement
219 Description: Allow tuning service thread via /proc
220 Details : For each service a new
221 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
222 created that can be used to set min/max thread counts, and get the
223 current number of running threads.
228 Descriptoin: don't return error if have partially created objects for file.
229 Details : lov_update_create_set uses set->set_success as index for created
230 objects, so if some requests will be failed, they will have hole
231 at end of array and qos_shrink_lsm for allocate correct lsm.
233 Severity : enhancement
235 Description: Update OFED support to 1.4.1
240 Description: Reduce small size read RPC
241 Details : Set read-ahead limit for every file and only do read-ahead when
242 available read-ahead pages are bigger than 1M to avoid small size
245 Severity : enhancement
247 Description: move AT tunable parameters for more consistent usage
248 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
250 Severity : enhancement
252 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
253 Details : allow skip disconnected ost for send statfs request and hide error
259 Description: fix racy locking of mballoc block bitmaps causing BUG
260 Details : The locking of the mballoc buddy bitmap and the in-memory
261 block bitmap was using two different spin locks in some
262 cases. This made it possible to incorrectly access the
263 mballoc bitmap while another process was modifying it,
264 causing a sanity assertion to fail. While no on-disk corruption
265 was reported, there was some risk of this happening.
268 Frequency : rare, on llog test 6
270 Descriptoin: don't allow connect to already connected import
271 Details : allowing connect to already connected import is hide connecting problem.
276 Descriptoin: allow kill process which wait statahead result
277 Details : for some reasons 'ls' can stick in waiting result from statahead,
278 in this case need way for kill this process.
280 Severity : enhancement
282 Description: Add state history info file, enhance import info file
283 Details : Track import connection state changes in a new osc/mdc proc file;
284 add overview-type data to the osc/mdc import proc file.
286 Severity : enhancement
288 Description: MDS create should not wait for statfs RPC while holding DLM lock.
290 Severity : enhancement
292 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
297 Descriptoin: OST grant too much space to client even there are not enough space.
298 Details : Client will shrink its grant cache to OST if there are no write
299 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
300 this grant cache if there are already not enough avaible space
301 (left_space < total_clients * 32M).
304 Frequency : start MDS on uncleanly shutdowned MDS device
306 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
307 Details : stay in waiting mds<>ost recovery finished produce random bugs
308 due race between two ll_sync thread for one lov target. send
309 ACTIVATE event only if connect realy finished and import have
313 Frequency : rare, connect and disconnect target at same time
315 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
316 Details : don't call obd_disconnect under lov_lock. this long time
317 operation and can block ptlrpcd which answer to connect request.
322 Descriptoin: don't lose wakeup for imp_recovery_waitq
323 Details : recover_import_no_retry or invalidate_import and import_close can
324 both sleep on imp_recovery_waitq, but we was send only one wakeup
328 Frequency : always with long access acl
330 Descriptoin: mds can't pack reply with long acl.
331 Details : mds don't control size of acl but they limited by reint/getattr
334 Severity : enhancement
336 Description: Update to SLES10 kernel-2.6.16.60-0.33.
338 Severity : enhancement
340 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
343 Frequency : start MDS on uncleanly shutdowned MDS device
345 Descriptoin: aborting recovery hang on MDS
346 Details : don't throttle destroy RPCs for the MDT.
349 Frequency : on remount
351 Description: external journal device not working after the remount
352 Details : clear dev_rdonly flag for external journal devices in
358 Description: shutdown vs evict race
359 Details : client_disconnect_export vs connect request race.
360 if client will evicted at this time - we start invalidate
361 thread without referece to import and import can be freed
365 Frequency : rare, need acl's on inode.
367 Description: client can't handle ost additional correctly
368 Details : if ost was added after client connected to mds client can have
369 hit lnet_try_match_md ... to big messages to wide striped files.
370 in this case need teach client to handle config events about add
371 lov target and update client max ea size at that event.
373 Severity : enhancement
375 Description: Changelogs
376 Details : Changelogs are a lightweight mechanism to track filesystem
377 metadata and namespace changes. The changelog is recorded
378 permanently on the MDTs, and is periodically "consumed" / purged
379 when records are no longer needed.
381 Severity : enhancement
383 Description: compact fld format with extents
384 Details : Store range of seq rather than every seq in FLD. Seq
385 controller update FLD rather than clients. In Case of CMD, mdt0
386 has FLD, all other metadata server act as non persistent proxy
387 for FLD queries and cache fld entries in fld cache.
392 Description: don't skip ost target if they assigned to file
393 Details : Drop slow OSCs if we can, but not for requested start idx.
394 This means "if OSC is slow and it is not the requested
395 start OST, then it can be skipped, otherwise skip it only
396 if it is inactive/recovering/out-of-space.
400 Description: more cleanup in mds_lov
401 Details : not send LOV EA under replay, we can't know about they size at this
402 time. Don't allow client connect to mds before any ost connected,
403 for avoid problems with LOV EA size and returning EIO to client.
405 Severity : enhancement
407 Description: Interoperability at server side (Disk interoperability)
409 Severity : enhancement
411 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
413 Severity : enhancement
415 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
417 Severity : enhancement
419 Description: New client IO stack (CLIO).
421 Severity : enhancement
423 Description: Commit on sharing. Eliminate inter-client dependencies between
424 uncommitted transactions by doing transaction commits.
425 Thereby clients may recovery independently.
428 Frequency : Create a symlink file with a very long name
430 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
431 Details : If there is no extra space in the request for early cancels,
432 ldlm_req_handles_avail() returns 0 instead of a negative value.
434 Severity : enhancement
436 Description: Add /proc entry for import status
437 Details : The mdc, osc, and mgc import directories now have
438 an import directory that contains useful import data for debugging
441 Severity : enhancement
443 Description: Re-disable certain /proc logging
444 Details : Enable and disable client's offset_stats, extents_stats and
445 extents_stats_per_process stats logging on the fly.
448 Frequency : Only on FC kernels 2.6.22+
450 Description: oops in statahead
451 Details : Do not drop reference count for the dentry from VFS when lookup,
452 VFS will do that by itself.
454 Severity : enhancement
456 Description: Generic /proc file permissions
457 Details : Set /Proc file permissions in a more generic way to enable non-
458 root users operate on some /proc files.
462 Description: Hitting mdc_commit_close() ASSERTION
463 Details : Properly handle request reference release in
464 ll_release_openhandle().
468 Description: quota recovery deadlock during mds failover
469 Details : This patch includes att18982, att18236, att18237 in bz14840.
471 1. fix osts hang when mds does failover with quotaon
472 2. prevent watchdog storm when osts threads wait for the
477 Frequency : only patchless client
478 Description: add workaround for race between add/remove dentry from hash
480 Severity : enhancement
482 Description: Allow OST glimpses to return PW locks
486 Description: LBUG when llog conf file is full
487 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
491 Description: Prevent import from entering FULL state when server in recovery
495 Description: service mount cannot take device name with ":"
496 Details : Only when device name contains ":/" will mount treat it as
502 Description: replace ptlrpcd with the statahead thread to interpret the async
503 statahead RPC callback
507 Frequency : on recovery
508 Description: I/O failures after umount during fail back
509 Details : if client reconnected to restarted server we need join to recovery
510 instead of find server handler is changed and process self
511 eviction with cancel all locks.
513 Severity : enhancement
515 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
517 Severity : enhancement
519 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
521 Severity : enhancement
523 Description: Upcall on Lustre log has been dumped
524 Details : Allow for a user mode script to be called once a Lustre log has
525 been dumped. It passes the filename of the dumped log to the
526 script, the location of the script can be specified via
527 /proc/sys/lnet/debug_log_upcall.
532 Description: avoid idr_remove called for id which is not allocated.
533 Details : Move assigment s_dev for clustered nfs to end of initialization,
534 to avoid problem with error handling.
539 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
540 Details : When connection is reused this not moved from CONN_UNUSED_HASH
541 into CONN_USED_HASH and this prodice warning when put connection
542 again in unused hash.
545 Severity : enhancement
547 Description: File striping can now be set to use an arbitrary pool of OSTs.
549 Severity : enhancement
551 Description: Export bytes_read/bytes_write count on OSC/OST.
555 Description: Early reply size mismatch, MGC loses connection
556 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
557 the connect flags are properly negotiated.
562 Description: quotas are not honored with O_DIRECT
563 Details : all writes with the flag O_DIRECT will use grants which leads to
564 this problem. Now using OBD_BRW_SYNC to guard this.
568 Description: add quota statistics
569 Details : 1. sort out quota proc entries and proc code.
570 2. add quota statistics
572 Severity : enhancement
574 Description: enable quota support for HEAD.
578 Description: Properly propagate oinfo flags from lov to osc for statfs
579 Details : restore missing copy oi_flags to lov requests.
581 Severity : enhancement
583 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
584 : l_getgroups(8), lst(8), routerstat(8)
585 Details : included man pages for llobdstat(8), llstat(8),
586 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
588 Severity : enhancement
590 Description: configure's --enable-quota should check the
591 : kernel .config for CONFIG_QUOTA
592 Details : configure is terminated if --enable-quota is passed but
593 : no quota support is in kernel
597 Description: Remove portals compatibility
598 Details : Remove portals compatibility, not interoperable with releases
603 Description: Resolve device initialization race
604 Details : Prevent proc handler from accessing devices added to the
605 obd_devs array but yet be intialized.
607 Severity : enhancement
609 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
611 Severity : enhancement
613 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
618 Description: Using wrong pointer in osc_brw_prep_request
619 Details : Access to array[-1] can produce panic if kernel compiled with
620 CONFIG_PAGE_ALLOC enabled
624 Description: Client runs out of low memory
625 Details : Consider only lowmem when counting initial number of llap pages
629 Description: *optional* service tags registration
630 Details : if the "service tags" package is installed on a Lustre node
631 When the filesystem is mounted, a local-node service tag will
632 be created. See http://inventory.sun.com/ for more information
633 about the Service Tags asset management system.
637 Description: Kernel BUG tries to release flock
638 Details : Lustre does not destroy flock lock before last reference goes
639 away. So always drop flock locks when client is evicted and
640 perform unlock regardless of successfulness of speaking to MDS.
644 Description: add refcount protection for osc callbacks, avoid panic on shutdown
648 Description: sanity test 65a fails if stripecount of -1 is set
649 Details : handle -1 striping on filesystem in ll_dirstripe_verify
654 Description: ASSERTION(CheckWriteback(page,cmd)) failed
655 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
658 Severity : enhancement
660 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
664 Description: do not process already freed flock
665 Details : flock can possibly be freed by another thread before it reaches
666 to ldlm_flock_completion_ast.
670 Description: LBUG during stress test
671 Details : Need properly lock accesses the flock deadlock detection list.
675 Description: oops in page fault handler
676 Details : kernel page fault handler can return two special 'pages' in error
677 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
681 Description: timeout with invalidate import.
682 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
683 should be handled by ptlrpcd. This produce long age waiting and
684 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
686 Severity : enhancement
688 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
692 Description: enable MGS and MDT services start separately
693 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
694 a co-located MGS without starting the MGS, which is a complement
695 to 'nosvc' mount option.
699 Frequency : after recovery
700 Description: precreate to many object's after del orphan.
701 Details : del orphan st in oscc last_id == next_id and this triger growing
702 count of precreated objects. Set flag LOW to skip increase count
703 of precreated objects.
707 Frequency : rare, on clear nid stats
708 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
709 Details : when clean nid stats sometimes try destroy live entry,
710 and this produce panic in free.
714 Description: Stack overflow during MDS log replay
715 ease stack pressure by using a thread dealing llog_process.
719 Description: wait until IO finished before start new when do lock cancel.
720 Details : VM protocol want old IO finished before start new, in this case
721 need wait until PG_writeback is cleared until check dirty flag and
722 call writepages in lock cancel callback.
724 Severity : enhancement
726 Description: using special macro for print time and cleanup in includes.
730 Description: mds_mfd_close() ASSERTION(rc == 0)
731 Details : In mds_mfd_close(), we need protect inode's writecount change
732 within its orphan write semaphore to prevent possible races.
736 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
740 Frequency : rare, on shutdown ost
741 Description: don't hit live lock with umount ost.
742 Details : shrink_dcache_parent can be in long loop with destroy dentries,
743 use shrink_dcache_sb instead.
747 Description: don't panic with use echo client
748 Details : echo client pass NULL as client nid pointer and this produce null
753 Description: fix build on ppc32
754 Details : compile code with -m64 flag produce wrong object file for ppc32.
758 Description: add message levels for liblustreapi
762 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
763 Details : Move the 'good_osts' check before the 'total_bavail' check. This
764 will result in an -EAGAIN and in the exit call path we call
765 alloc_rr() which will with increasing aggressiveness attempt to
766 aquire precreated objects on the minimum number of required OSCs.
770 Description: Use old size assignment to avoid deadlock
771 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
772 the scheduling while holding a spinlock. We do not need locking
773 for size in ll_update_inode() because size is only updated from
774 the MDS for directories or files without objects, so there is no
775 other place to do the update, and concurrent access to such inodes
776 are protected by the inode lock.
780 Description: resolve "_IOWR redefined" build error on SLES10
784 Description: dump the memory debugging after all modules are unloaded to
785 suppress false negative in conf_sanity test 39
787 Severity : enhancement
789 Description: build kernel-ib packages for OFED 1.3 in our release cycle
794 Description: fix SLES kernel versioning
795 Details : the kernel version for our SLES 10 kernel did not include a "-"
796 before the "smp" at the end. while this was not a problem in
797 general, it did mean that software trying to use the kernel
798 version to try to detect a vendor specific kernel would fail.
799 this was most evident by the OFED build scripts.
803 Description: Don't update lov_desc members until making sure they are valid
804 Details : When updating lov_desc members via proc fs, need fix their
805 validities before doing the real update.
809 Description: don't put request into delay list while invalidate in flight.
810 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
811 import in flight. this produce timeout for invalidate and sometimes
812 can cause stale data.
816 Frequency : on ppc only
817 Description: not convert ost objects for directory because it's not exist.
818 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
822 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
824 Severity : enhancement
826 Description: Update to support 2.6.22.14 vanilla kernel.
830 Frequency : rare, on recovery
831 Description: read procfs can produce deadlock in some situation
832 Details : Holding lprocfs lock which send rpc can produce block for destroy
833 obd objects and this also block reconnect with -EALREADY. This
834 isn't fix all lprocfs bugs - but make it rare.
836 Severity : enhancement
838 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
841 Frequency : frequent on X2 node
843 Description: mdc_set_open_replay_data LBUG
844 Details : Set replay data for requests that are eligible for replay.
848 Description: lustre_mgs: operation 101 on unconnected MGS
849 Details : When MGC is disconnected from MGS long enough, MGS will evict the
850 MGC, and late on MGC cannot successfully connect to MGS and a lot
851 of the error messages complaining that MGS is not connected.
855 Frequency : on network error
856 Description: panic with double free request if network error
857 Details : mdc_finish_enqueue is finish request if any network error ocuring,
858 but it's true only for synchronus enqueue, for async enqueue
859 (via ptlrpcd) this incorrect and ptlrpcd want finish request
862 Severity : enhancement
864 Description: client-side metadata stat-ahead during readdir(directory readahead)
865 Details : perform client-side metadata stat-ahead when the client detects
866 readdir and sequential stat of dir entries therein
869 Frequency : on start mds
871 Description: Implement get_info(last_id) in obdfilter.
874 Frequency : occasional
876 Description: Correctly check stale fid, not start epoch if ost not support SOM
877 Details : open with flag O_CREATE need set old fid in op_fid3 because
878 op_fid2 was overwritten with new generated fid, but mds can answer
879 with one of these two fids and both is not stale. Setattr
880 incorrectly started an epoch and assume will be called
881 done_writing, but without SOM done_writing ever being called.
884 Frequency : rare, depends on device drivers and load
886 Description: MDS or OSS nodes crash due to stack overflow
887 Details : Code changes in 1.8.0 increased the stack usage of some functions.
888 In some cases, in conjunction with device drivers that use a lot
889 of stack the MDS (or possibly OSS) service threads could overflow
890 the stack. One change which was identified to consume additional
891 stack has been reworked to avoid the extra stack usage.
894 Frequency : occasional
896 Description: Do not fail import if osc_interpret_create gets -EAGAIN
897 Details : If osc_interpret_create got -EAGAIN it immediately exits and
898 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
899 oscc_has_objects and see OSC has no objests and call
900 oscc_internal_create to resend create request.
902 Severity : enhancement
904 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
906 Severity : enhancement
908 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
911 Frequency : very rare
913 Description: Fix replay if there is an un-replied request and open
914 Details : In some cases, older replay request will revert the
915 mcd->mcd_last_xid on MDS which is used to record the client's
918 Severity : enhancement
920 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
922 Severity : enhancement
924 Description: Add rhel5 support to HEAD.
926 Serverity : enhancement
928 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
933 Description: Don't allow skipping OSTs if index has been specified.
934 Details : Don't allow skipping OSTs if index has been specified, make
935 locking in internal create lots better.
939 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
940 Details : access to bitfield in structure is always rounded to long
941 and this produce problem with not atomic change any bit.
945 Description: Lustre make rpms failed.
946 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
947 builds Lustre rpms from tarball.
950 Frequency : rare on shutdown ost
952 Description: If llog cancel was not send before clean_exports phase, this can
953 produce deadlock in llog code.
954 Details : If llog thread has last reference to obd and call class_import_put
955 this produce deadlock because llog_cleanup_commit_master wait when
956 last llog_commit_thread exited, but this never success because was
957 called from llog_commit_thread.
961 Description: allow userland application know is lost one of stripes.
962 Details : fill lvb_blocks with error code on ost and return it to
963 application if error flag found.
967 Description: NULL lov_tgts causing MDS oops
968 Details : more safe checks for NULL lov_tgts for avoid oops.
970 Severity : enhancement
972 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
976 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
977 Details : If an OST is down the MDS will hang indefinitely in
978 obd_statfs_rqset() waiting for the statfs data. While for
979 MDS QOS usage of statfs, it should not stuck in waiting.
981 Severity : enhancement
983 Description: remote_acl support
984 Details : Support ACL-based permission check for remote user.
985 Support setfacl/getfacl for remote user with the utils
986 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
987 the system "{s,g}etfacl" utils.
989 Severity : enhancement
991 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
993 Severity : enhancement
995 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
999 Description: make mgs_setparam() handle fsname containing dash
1000 Details : fsname containing a dash does not work with lctl conf_param
1002 Severity : enhancement
1004 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1006 Severity : enhancement
1008 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1012 Frequency : rare, at shutdown
1013 Description: access already free / zero obd_namespace.
1014 Details : if client_disconnect_export was called without force flag set,
1015 and exist connect request in flight, this can produce access to
1016 NULL pointer (or already free pointer) when connect_interpret
1017 store ocd flags in obd_namespace.
1021 Frequency : only at startup
1022 Description: not alloc memory with spinlock held.
1023 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1024 if any spinlock held.
1026 Severity : enhancement
1028 Description: make lustre randomly fail allocating memory
1029 Details : Make lustre randomly failed allocating memory for testing purpose.
1031 Severity : enhancement
1033 Description: lost problems with lov objid file
1034 Details : Fixes some scability and access to not inited memory problems
1035 in work with lov objdid file.
1040 Description: lfs find does not continue on file error
1041 Details : Continue other files processing when a file/dir is absent.
1045 Description: Inconsistent usage of lustre_pack_reply()
1046 Details : Standardize the usage of lustre_pack_reply() such that it
1047 always generate a CERROR on failure.
1050 Frequency : occasional
1052 Description: MDS hang or stay in waiting lock
1053 Details : If client receive lock with CBPENDING flag ldlm need send lock
1054 cancel as separate rpc, to avoid situation when cancel request
1055 can't processed due all i/o threads stay in wait lock.
1059 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1063 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1067 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1071 Description: Update to RHEL4 latest kernel.
1073 Severity : enhancement
1075 Description: Build SLES10 patchless client fails
1076 Details : The configure was broken by run ./configure with
1077 --with-linux-obj=.... argument for patchless client. When the
1078 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1079 can't search header adequately. Use absolute path such as
1080 -I($LINUX)/include instead.
1084 Description: interrupt oig_wait produce painc on resend.
1085 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1086 private set, and this produce situation when rq_ptlrpcd_data not
1087 copyed to new allocated request and triggered LBUG on assert
1088 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1089 ptlrpcd set if request is changed and can be safety changed to use
1094 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1095 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1096 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1097 these macros and replace them with their difinitions instead.
1101 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1102 Details : Update which_patch & target file for SLES10 latest kernel.
1105 Description: Remove client patches from SLES 10 kernel.
1106 Details : This causes SLES 10 clients to behave as patchless clients
1107 even on a Lustre-patched (server) kernel.
1109 Severity : enhancement
1111 Description: self-adjustable client's lru lists
1112 Details : use adaptive algorithm for managing client cached locks lru
1113 lists according to current server load, other client's work
1114 pattern, memory activities, etc. Both, server and client
1115 side namespaces provide number of proc tunables for controlling
1118 Severity : enhancement
1120 Description: light-weight GSS support
1121 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1122 and reduce performance overhead.
1124 Severity : enhancement
1126 Description: Linux keyring support
1127 Details : Support using service of Linux keyring for Lustre GSS internal
1128 context refresh/cache mechanism.
1132 Description: Fix errors in lfs documentation
1133 Details : Fixes man pages
1137 Description: don't use GFP_* in generic Lustre code.
1138 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1142 Description: obdclass is limited by single OBD_ALLOC(idarray)
1143 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1147 Description: warnings with build patchless client with vanila 2.6.19 and up
1148 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1149 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1153 Frequency : Only for SLES
1154 Description: sanity_quota fail test_1
1155 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1156 file, which makes TSTID[2] inunique.
1160 Frequency : Only for SLES9
1161 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1162 Details : Update target/ChangeLog/which_patch .
1166 Frequency : Only for test_10 in sanity-quota.sh
1167 Description: a bug in quota test code
1168 Details : checking a test flag in a wrong place
1172 Frequency : Only for test_13 in sanity-quota.sh
1173 Description: a bug in quota test script
1174 Details : shouldn't check overfull allocated quota
1178 Description: jbd statistics
1179 Details : Port older jbd statistics patch for sles10
1181 Severity : enhancement
1183 Description: Add printing inode info into message about error in writepage.
1188 Description: reply_lock_interpret crash due to race with it and lock cancel.
1189 Details : Do not replay locks that are being cancelled. Do not reference
1190 locks by their address during replay, just by their handle.
1194 Frequency : When flocks are used.
1195 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1196 Details : Flock locks might destroy just granted lock if it could be merged
1197 with another existing flock, this is done in completion handler,
1198 so teach ldlm_cli_enquque_fini that this is a valid case for
1204 Description: Oops in read and write path when failing to allocate lock.
1205 Details : Check if lock allocation failed and return error back.
1209 Description: lstripe command fails for valid OST index
1210 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1211 instead of lov->desc.ld_active_tgt_count.
1215 Description: sanity.sh failed test 103
1216 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1221 Description: df doesn't work properly if diskfs blocksize != 4K
1222 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1226 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1227 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1229 Severity : enhancement
1231 Description: Adaptive timeouts
1232 Details : RPC timeouts adapt to changing server load and network
1233 conditions to reduce resend attempts and improve recovery time.
1237 Description: llapi_file_create() does not allow some changes
1238 Details : add llapi_file_open() that allows specifying the mode and
1239 open flags, and also returns an open file handle.
1243 Description: merge and cleanup kernel patches.
1244 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1248 Description: Add journal checksum support.(Kernel part)
1249 Details : The journal checksum feature adds two new flags i.e
1250 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1251 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1252 indicates that the commit block contains the checksum for
1253 the blocks described by the descriptor blocks. Now commit
1254 record can be sent to disk without waiting for descriptor
1255 blocks to be written to disk. This behavior is controlled
1256 using JBD2_FEATURE_ASYNC_COMMIT flag.
1260 Description: OSS needs mutliple precreate threads
1261 Details : Add ability to start more than one create thread per OSS.
1265 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1266 Details : execute lfs setstripe on client
1270 Description: mds_obd_create error creating tmp object
1271 Details : When the user sets quota on root, llog will be affected and can't
1272 create files and write files.
1275 Frequency : Always on ia64 patchless client, and possibly others.
1277 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1278 Details : This allows the patchless client to be loaded on architectures
1279 without this export.
1283 Description: RedHat Update kernel for RHEL5
1284 Details : Add the kernel config file for RHEL5.
1288 Description: RedHat Update kernel for RHEL5
1289 Details : Modify the kernel config file more closer RHEL5.
1293 Description: Build failure against Centos5 (RHEL5)
1294 Details : Define PAGE_SIZE when it isn't present.
1298 Description: test_7 on sanity-quota.sh can't be used on separate machines
1299 Details : add facet to handle it
1303 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1304 Details : with NFS, the anon dentry's parent was set to itself in
1305 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1306 corresponding dentry other than use rec->ur_name.
1308 Severity : enhancement
1310 Description: lfs setstripe enhancement
1311 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1315 Description: enable data checksumming by default
1316 Details : enable checksum by default, allow --disable-checksum
1317 configure option and "-o nochecksum" mount option. Checksums
1318 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1322 Description: lustre support for RHEL5
1323 Details : Add support for RHEL5.
1327 Description: Client eviction due to failover config
1328 Details : after a connection loss, the lustre client should attempt to
1329 reconnect to the last active server first before trying the
1330 other potential connections.
1334 Description: when mds and osts use different quota unit(32bit and 64bit),
1335 quota will be released repeatly.
1336 Details : void sending multiple quota reqs to mds, which will keep the
1337 status between the reqs.
1341 Description: osts not allocated evenly to files
1342 Details : change the condition to increase offset_idx
1345 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1347 Description: Data corruption for OSTs that are formatted larger than 2TB
1349 Details : When generating the bio request for lustre file writes the
1350 sector number would overflow a temporary variable before being
1351 used for the IO. The data reads correctly from Lustre (which
1352 will overflow in a similar manner) but other file data or
1353 filesystem metadata may be corrupted in some cases.
1357 Description: Tune the kernel for good SCSI performance.
1358 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1359 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1364 Description: rewrite ext2-derived code in obdclass/uuid.c
1365 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1366 from scratch preserving functionality.
1370 Description: "lfs find -obd UUID" prints directories
1371 Details : "lfs find -obd UUID" will return all directory names instead
1372 of just file names. It is incorrect because the directories
1373 do not reside on the OSTs.
1377 Description: use i_size_read and i_size_write in 2.6 port
1378 Details : replace inode->i_size access with i_size_read/write()
1381 Frequency : when removing large files
1383 Description: scheduling issue during removal of large Lustre files
1384 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1385 It causes scheduling issues when removing large files (17TB in the
1389 Frequency : only with liblustre clients on XT3
1391 Description: evictions taking too long
1392 Details : allow llrd to evict clients directly on OSTs
1395 Frequency : only on ppc
1397 Description: /proc/fs/lustre/devices broken on ppc
1398 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1399 We should make sure the type of 'index' is loff_t and then cast
1400 explicitly as needed below. Do not assign an explicitly cast
1404 Frequency : only for rhel5
1406 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1407 Details : Modify the target file & which_kernel.
1409 Severity : enhancement
1411 Description: omit set fsid for export NFS
1412 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1413 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1417 Frequency : Always, for kernels after 2.6.16
1418 Description: Fix warning idr_remove called for id=.. which is not allocated.
1419 Details : Last kernels save old s_dev before kill super and not allow
1420 to restore from callback - restore it before call kill_anon_super.
1424 Description: buffer overruns could theoretically occur
1425 Details : llapi_semantic_traverse() modifies the "path" argument by
1426 appending values to the end of the origin string, and a
1427 overrun may occur. Adding buffer overrun check in liblustreapi.
1431 Description: Fix error on 'ls .' at the top of the Lustre mount.
1432 Details : Don't revalidate dentry if it is a root dentry.
1436 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1437 Details : Modify vm-tunables-rhel4.patch.
1441 Description: Kernel config for 2.6.18-vanilla.
1442 Details : Modify targets/2.6-vanilla.target.in.
1443 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1444 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1445 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1446 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1450 Description: improve handling recoverable errors
1451 Details : if request processig with error which can be recoverable on server
1452 request should be resend, otherwise page released from cache and
1457 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1458 Details : Modify target file & which_patch.
1459 A flaw was found in the IA32 system call emulation provided
1460 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1461 value could be stored in the %RAX register, which could trigger an
1462 out-of-bounds system call table access. An untrusted local user
1463 could exploit this flaw to run code in the kernel
1464 (ie a root privilege escalation). (CVE-2007-4573).
1468 Description: O_DIRECT bypasses client statistics.
1469 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1470 read_ahead_stats were updated. copied the stats section from
1471 osc_send_oap_rpc() into async_internal().
1475 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1479 Description: change order of libsysio includes
1480 Details : '#include sysio.h' should always come before '#include xtio.h'
1484 Description: handle "serious error: objid * already exists" more gracefully
1485 Details : If LAST_ID value on disk is smaller than the objects existing in
1486 the O/0/d* directories, it indicates disk corruption and causes an
1487 LBUG(). If the object is 0-length, then we should use the existing
1488 object. This will help to avoid a full fsck in most cases.
1490 Severity : enhancement
1492 Description: adapt the lustre_config script to support the upgrade case
1493 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1494 server targets to 1.6 in parallel.
1498 Description: To avoid grant space > avaible space when the disk is almost
1499 full. Without this patch you might see the error "grant XXXX >
1500 available" or some LBUG about grant, when the disk is almost
1502 Details : In filter_check_grant, for non_grant cache write, we should
1503 check the left space by if (*left > ungranted + bytes), instead
1504 of (*left > ungranted), because only we are sure the left space
1505 is enough for another "bytes", then the ungrant space should be
1506 increase. In client, we should update cl_avail_grant only there
1507 is OBD_MD_FLGRANT in the reply.
1511 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1512 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1513 described in CVE-2007-4573. This problem could allow untrusted
1514 local users to gain root access.
1517 Frequency : when using O_DIRECT and quotas
1519 Description: Incorrect file ownership on O_DIRECT output files
1520 Details : block usage reported by 'lfs quota' does not take into account
1521 files that have been written with O_DIRECT.
1526 Description: touch file failed when fs is not full
1527 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1528 otherwise we can get ENOSP while fs is not full.
1532 Description: parallel lock callbacks
1533 Details : Instead of sending blocking and completion callbacks as separated
1534 requests, adding them to a set and sending in parallel.
1537 Frequency : only for Cray XT3
1538 Bugzilla : 12829/13455
1539 Description: Changing primary group doesn't change the group lustre assigns to
1541 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1542 overridden with the primary group provided by the group upcall,
1543 whereas the supplied fsgid can be trusted if it is in the list of
1544 supplementary groups returned by the group upcall.
1546 Severity : enhancement
1548 Description: Allow masking D_WARNING, D_ERROR messages from console
1549 Details : Console messages can now be disabled via lnet.printk.
1553 Description: User code with malformed file open parameter crashes client node
1554 Details : Before packing join_file req, all the related reference should be
1555 checked carefully in case some malformed flags cause fake
1556 join_file req on client.
1560 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1561 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1562 mode is equal to the granted mode. However, it should also check
1563 that LDLM_FL_CP_REQD is not set.
1567 Description: Heavy nfs access might result in deadlocks
1568 Details : After ELC code landed, it is now improper to enqueue any mds
1569 locks under och_sem, because enqueue might want to decide to
1570 cancel open locks for same inode we are holding och_sem for.
1574 Description: Client eviction while running blogbench
1575 Details : A lot of unlink operations with concurrent I/O can lead to a
1576 deadlock causing evictions. To address the problem, the number of
1577 oustanding OST_DESTROY requests is now throttled to
1578 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1579 callbacks are processed in priority.
1583 Description: enable ACLs on MDS by default
1584 Details : ACLs must be enabled on MDS by default.
1587 Frequency : PPC/PPC64 only
1589 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1590 Details : fix duplicated definitions between asm-ppc64/types.h and
1591 lustre_types.h on PPC.
1594 Frequency : PPC/PPC64 only
1596 Description: asm-ppc/segment.h does not exist
1597 Details : fix compile issue on PPC.
1601 Description: better handle error messages in extents code
1604 Frequency : RHEL4 only
1606 Description: mkfs is very slow on IA64/RHEL4
1607 Details : A performance regression has been discovered in the MPT Fusion
1608 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1609 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1610 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1612 Severity : enhancement
1614 Description: SNMP support enhancement
1615 Details : Adding total number of sampled request for an MDS node in snmp
1618 Severity : enhancement
1620 Description: Optimize ldlm waiting list processing for PR extent locks
1621 Details : When processing waiting list for read extent lock and meeting read
1622 lock that is same or wider to it that is not contended, skip
1623 processing rest of the list and immediatelly return current
1624 status of conflictness, since we are guaranteed there are no
1625 conflicting locks in the rest of the list.
1629 Description: Time out and refuse to reconnect
1630 Details : When the failover node is the primary node, it is possible
1631 to have two identical connections in imp_conn_list. We must
1632 compare not conn's pointers but NIDs, otherwise we can defeat
1633 connection throttling.
1637 Description: port llog fixes from b1_6 into HEAD
1638 Details : Port llog reference couting and some llog cleanups from b1_6
1639 (bug 10800) into HEAD, for protect from panic and access to already
1640 free llog structures.
1644 Description: Detect stride IO mode in read-ahead
1645 Details : When a client does stride read, read-ahead should detect that and
1646 read-ahead pages according to the detected stride pattern.
1650 Description: data checksumming impacts single node performance
1651 Details : add support for several checksum algorithm. Currently, only CRC32
1652 and Adler-32 are supported. The checksum type can be changed on
1653 the fly via /proc/fs/lustre/osc/*/checksum_type.
1657 Description: use adler32 for page checksums
1658 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1663 Description: build for x2 fails
1664 Details : fix compile issue on Cray systems.
1668 Description: Properly match for duplicate locks
1669 Details : Due to different lock order from skiplists code, we need to
1670 traverse entire list for now
1673 Frequency : only on PPC/SLES10
1675 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1676 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1677 be defined. Add a hack in mkfs_lustre.c to work around this compile
1682 Description: LASSERT on MDS when client holding flock lock dies
1683 Details : ldlm pool logic depends on number of granted locks equal to
1684 number of released locks which is not true for flock locks, so
1685 just exclude such locks from consideration.
1689 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1690 Details : Use fsfilt_sync() for both the whole filesystem sync and
1691 individual file sync to eliminate dangerous inode locking
1692 with I_LOCK that can lead to a deadlock.
1696 Description: performance in 1.6.3
1697 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1698 sure that 1MB requests are not fragmented by the block layer.
1700 Severity : enhancement
1702 Description: organize the server-side client stats on per-nid basis
1703 Details : Change the structure of stats under obdfilter and mds to
1711 The "uuid"s file would list the uuids of _active_ exports.
1712 And the clear entry is to clear all stats and stale nids.
1714 Severity : enhancement
1716 Description: eliminate client locks in face of contention
1717 Details : file contention detection and lockless i/o implementation
1718 for contended files.
1722 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1725 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1727 Description: MDS may lose file striping (and hence file data) in some cases
1728 Details : If there are additional extended attributes stored on the MDS,
1729 in particular ACLs, SELinux, or user attributes (if user_xattr
1730 is specified for the client mount options) then there is a risk
1731 of attribute loss. Additionally, the Lustre file striping
1732 needs to be larger than default (e.g. striped over all OSTs),
1733 and an additional attribute must be stored initially in the
1734 inode and then increase in size enough to be moved to the
1735 external attribute block (e.g. ACL growing in size) for file
1740 Description: skiplist implementation simplification
1741 Details : skiplists are used to group compatible locks on granted list
1742 that was implemented as tracking first and last lock of each
1743 lock group the patch changes that to using doubly linked lists
1747 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1748 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1749 because LASSERT on both the data supplied by a client, and the
1750 data on disk is dangerous and incorrect.
1752 Severity : enhancement
1754 Description: Slow truncate/writes to huge files at high offsets.
1755 Details : Directly associate cached pages to lock that protect those pages,
1756 this allows us to quickly find what pages to write and remove
1757 once lock callback is received.
1761 Description: more ldlm soft lockups
1762 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1763 starve other threads from the resource lock for a long time in
1764 case of long waiting queue, so change the debug level from
1765 D_OTHER to the less frequently used D_INFO.
1767 Severity : enhancement
1769 Description: add -gid, -group, -uid, -user options to lfs find
1773 Description: Hung threads in invalidate_inode_pages2_range
1774 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1775 once one is completed. As a result, some RPCs are stuck in the
1776 queue and are never sent.
1780 Description: filter threads hungs on waiting journal commit
1781 Details : Cleanup filter group llog code, then only filter group llog will
1782 be only created in the MDS/OST syncing process.
1786 Description: Procfs and llog threads access destoryed import sometimes.
1787 Details : Sync the import destoryed process with procfs and llog threads by
1788 the import refcount and semaphore.
1790 Severity : enhancement
1792 Description: openlock cache of b1_6 port to HEAD
1797 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1798 Details : If insertion of an extent fails, then discard the inode
1799 preallocation and free data blocks else it can lead to duplicate
1804 Description: don't always update ctime in ext3_xattr_set_handle()
1805 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1806 In some cases the ctime should not be updated, for example for
1807 2.0->1.8 compatibility it is necessary to delete an xattr and it
1808 should not update the ctime.
1812 Bugzilla : 15713/16362
1813 Description: Assertion in iopen_connect_dentry in 1.6.3
1814 Details : looking up an inode via iopen with the wrong generation number can
1815 populate the dcache with a disconneced dentry while the inode
1816 number is in the process of being reallocated. This causes an
1817 assertion failure in iopen since the inode's dentry list contains
1818 both a connected and disconnected dentry.
1822 Description: assertion failure in ldlm_handle2lock()
1823 Details : fix a race between class_handle_unhash() and class_handle2object()
1824 introduced in lustre 1.6.5 by bug 13622.
1829 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1830 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1835 Description: oops in sd_iostats_seq_show()
1836 Details : unloading/reloading the scsi low level driver triggers a kernel
1837 bug when trying to access the sd iostat file.
1842 Description: Kernel panics during QLogic driver reload
1843 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1844 patch, causing memory corruption.
1849 Description: journal_dev option does not work in b1_6
1850 Details : pass mount option during pre-mount.
1852 Severity : enhancement
1854 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1855 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1856 extent information of a file. It can be used to map logical blocks
1857 in a file to physical blocks in the block device.
1861 Description: LDLM soft lockups - improvement
1862 Details : It is be possible to send the lock handle along with each read
1863 or write request because the client is already doing a lock match
1864 itself so there isn't any reason the OST should have to re-do that
1868 Frequency : only on Cray X2
1870 Description: X2 build failures
1871 Details : fix build failures on Cray X2.
1875 Description: xid & resent requests
1876 Details : Initialize RPC XID from clock at startup (randomly if clock is
1879 Severity : enhancement
1881 Description: Add lustre_start utility to start or stop multiple Lustre servers
1886 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1887 Details : In case of memory pressure, list_del() can be called twice on
1888 req->rq_history_list, causing a kernel oops.
1892 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1893 Details : fix stack overflow in the distributed lock manager by defering
1894 export eviction after a failed AST to the elt thread instead of
1895 handling it in the dlm interpret routine.
1899 Description: Convert some comments to new format.
1900 Details : Update documenting comments to match doxygen conventions.
1904 Description: Grammar fixes.
1905 Details : A couple of trivial spelling fixes.
1909 Description: OSD_COUNTERS-mandatory
1910 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1911 to simplify the code.
1915 Description: simplify cmm_device freeing logic.
1916 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1920 Description: Add lockdep support to dt_object_operations locking interface.
1921 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1922 parameter indicating lock ordering. Update mdd code to use new
1927 Description: Introduce failloc constants for lockless IO tests.
1928 Details : Add two new failloc constants to test lockless IO. Only one of
1929 them in implemented---another is checked in yet to be landed
1934 Description: Add lockdep support for inode mutex.
1935 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1936 in the situations where Lustre has to lock more than one inode
1941 Description: Add optional invariants checking support.
1942 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1943 configure switch. This macro is to be used for consistency and
1944 sanity checks that are too expensive to be left in `production'
1949 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1950 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1955 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1956 Details : Use spin_lock_nested() in (the only) situation where more than
1957 one ldlm_lock is locked simultaneously. Also, fix possible
1958 dead-lock in ldlm_lock_change_resource() by enforcing particular
1963 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1964 Details : Instead of specifying each ldlm_lock call-back through separate
1965 parameter, wrap them into struct ldlm_callback_suite.
1969 Description: Kill join_lru obd method and its callers.
1970 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1971 in memory---a requirement implemented through ->o_join_lru() obd
1972 method in HEAD. Remove this method and its users.
1976 Description: Add asynchronous ldlm ENQUEUE completion handler.
1977 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1978 case that stock ldlm_completion_ast() cannot handle as it waits
1979 until lock is granted. Introduce new ldlm_completion_ast_async()
1980 for this. Also comment ldlm_completion_ast().
1984 Description: ldlm_error <-> errno conversion.
1985 Details : Add functions to map (rather arbitrary) between LDLM error codes
1986 and standard errno values. CLIO needs this to prevent LDLM specific
1987 constants from escaping ldlm and osc.
1991 Description: Kill unused ldlm_handle2lock_ns() function.
1992 Details : Kill unused ldlm_handle2lock_ns() function.
1996 Description: Add lu_ref support to ldlm_lock
1997 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1998 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1999 and ldlm_resource. LDLM interface has to be changed, because code that
2000 releases a reference on a lock, has to "know" what reference this is.
2001 In the most frequent case
2003 lock = ldlm_handle2lock(handle);
2005 LDLM_LOCK_PUT(lock);
2007 no changes are required. When any other reference (received _not_ from
2008 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2009 instead of LDLM_LOCK_PUT().
2011 Arguably, changes are pervasive, and interface requires some discipline
2012 for proper use. On the other hand, it was very instrumental in finding
2013 a few leaked lock references.
2017 Description: Add ldlm_lock_addref_try().
2018 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2019 attempts to addref a lock that might be being canceled
2024 Description: Add ldlm_weigh_callback().
2025 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2026 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2027 hard-coded `number of pages' logic.
2031 Description: Add lockdep annotations to llog code.
2032 Details : Use appropriately tagged _nested() locking calls in the places
2033 where llog takes more than one ->lgh_lock lock.
2037 Description: Add loi_kms_set().
2038 Details : Wrap kms updates into a helper function.
2042 Description: Constify instances of struct lsm_operations.
2043 Details : Constify instances of struct lsm_operations.
2047 Description: lu_conf support.
2048 Details : On a server, a file system object is uniquely identified
2049 by a fid, which is sufficient to locate and load all object
2050 state (inode). On a client, on the other hand, more data are
2051 necessary instantiate an object. Change lu_object_find() and
2052 friends to take additional `lu_conf' argument describing object.
2053 Typically this includes layout information.
2057 Description: lu_context fixes.
2058 Details : Introduce new lu_context functions that are needed on the client
2059 side, where some system threads (ptlrpcd) are shared by multiple
2060 modules, and so cannot be stopped during module shutdown.
2064 Description: Add start and stop methods to lu_device_type_operations.
2065 Details : Introduce two new methods in lu_device_type_operations, that are
2066 invoked when first instance of a given type is created and last one
2067 is destroyed respectively. This is need by CLIO.
2071 Description: Add lu_ref support to struct lu_device.
2072 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2073 track leaked references.
2077 Description: Introduce lu_kmem_descr.
2078 Details : lu_kmem_descr and its companion interface allow to create
2079 and destroy a number of kmem caches at once.
2083 Description: Fix lu_object finalization race.
2084 Details : Fix a race between lu_object_find() finding an object and its
2085 concurrent finalization. This race is (most likely) not possible
2086 on the server, but might happen on the client.
2090 Description: Introduce lu_ref interface.
2091 Details : lu_ref is a debugging module allowing to track references to
2092 a given object. It is quite cpu expensive, and has to be
2093 explicitly enabled with --enable-lu_ref. See usage description
2098 Description: Factor lu_site procfs stats into a separate function.
2099 Details : Separate lu_site stats printing code into a separate function
2100 that can be reused on a client.
2104 Description: Constify instances of struct {lu,dt,md}_device_operations.
2105 Details : Constify instances of struct {lu,dt,md}_device_operations.
2109 Description: Introduce struct md_site and move meta-data specific parts of
2110 struct lu_site here.
2111 Details : Move md-specific fields out of struct lu_site into special struct
2112 md_site, so that lu_site can be used on a client.
2116 Description: Kill mdd_lov_destroy().
2117 Details : Remove unused mdd code.
2121 Description: Add st_block checking to multistat.c.
2122 Details : Add st_block checking to multistat.c.
2126 Description: Add lu_ref support to struct obd_device.
2127 Details : Add lu_ref tracking to obd_device.
2131 Description: Kill obd_set_fail_loc().
2132 Details : Remove unused code.
2136 Description: Add special type for ptlrpc_request interpret functions.
2137 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2138 there. Actual usage will be in CLIO.
2142 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2143 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2144 doesn't work with lockdep.
2148 Description: Add rwv.c test program.
2149 Details : New testing program exercising readv(2) and writev(2) (Qian).
2153 Description: Add sendfile.c test program.
2154 Details : New testing program exercising sendfile(2) (Jay).
2158 Description: Ratelimit a message that can be very frequent.
2159 Details : Ratelimit a memory allocation failure message that can
2164 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2165 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2169 Description: Liblustre build fixes.
2170 Details : Liblustre build fixes.
2174 Description: libcfs: add cfs_{need,cond}_resched() interface.
2175 Details : libcfs: add cfs_{need,cond}_resched() definition and
2176 implementations for Linux, NT, and liblustre.
2178 Severity : enhancement
2180 Description: More exported tunables for mballoc
2181 Details : Add support for tunable preallocation window and new tunables for
2182 large/small requests
2186 Description: Detect corruption of block bitmap and checking for preallocations
2187 Details : Checks validity of on-disk block bitmap. Also it does better
2188 checking of number of applied preallocations. When corruption is
2189 found, it turns filesystem readonly to prevent further corruptions.
2193 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2194 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2195 window to 0. If it does hit cache miss, and read-pattern is still
2196 stride-io mode, does not reset the stride window, but also does not
2197 increase the stride window length in this case.
2201 Frequency : only for big-endian servers
2202 Description: Check if system is big-endian while mounting fs with extents feature
2203 Details : Mounting a filesystem with extents feature will fail on big-endian
2204 systems since ext3-based ldiskfs is not supported on big-endian
2205 systems. This can be over-riden with "bigendian_extents" mount option.
2207 Severity : enhancement
2209 Description: The root squash functionality
2210 Details : A security feature, which is to prevent users from being able
2211 to mount lustre on their desktop, run as root, and delete
2212 all of the files in the filesystem. The goal is accomplished by
2213 remapping user id (UID) and group id (GID) of the root user to
2214 a UID and GID specified by the system administartor via Lustre
2215 configuration management server (MGS). The functionality also
2216 allows to specify sets of clients for which the remapping does
2221 Description: Excessive recovery window
2222 Details : With AT enabled, the recovery window can be excessively long (6000+
2223 seconds). To address this problem, we no longer use
2224 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2225 timeout no longer depends on the service time, it is set to
2226 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2227 time via pb_service_time.
2231 Description: Watchdog triggered on MDS failover
2232 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2233 the OSTs know that the MDS "UUID" can be reused for the same export
2234 from a different NID, so we do not need to wait for the export to be
2238 Frequency : rare, only if using MMP with Linux RAID
2240 Description: MMP doesn't work with Linux RAID
2241 Details : While using HA for Lustre servers with Linux RAID, it is possible
2242 that MMP will not detect multiple mounts. To make this work we
2243 need to unplug the device queue in RAID when the MMP block is being
2244 written. Also while reading the MMP block, we should read it from
2245 disk and not the cached one.
2247 Severity : enhancement
2249 Description: open file using fid
2250 Details : A file can be opened using just its fid, like
2251 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2254 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2257 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2258 Details : add proper locking for creation of jbd_4k slab cache
2262 Description: MMP check in ext3_remount() fails without displaying any error
2263 Details : When multiple mount protection fails during remount, proper error
2266 Severity : enhancement
2268 Description: Allow stripe size to be up to 4G-64k
2269 Details : Fix math logic to allow large stripe sizes.
2273 Description: add check for >8TB ldiskfs filesystems
2274 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2275 Don't allow >8TB ldiskfs filesystems to be mounted without
2276 force_over_8tb mount option
2281 Descriptoin: OSS complains "too many blocks" and crashes
2282 Details : In filter_commitrw_write():
2283 1. lquota_chkquota();
2284 2. obdo_from_inode();
2285 3. lquota_pending_commit();
2286 "2." has changed uid and gid in this case, then caused
2287 lquota_chkquota() and lquota_pending_commit() to use
2288 different uid and gid. That is the root of the bug.
2292 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2293 Details : Detect and print out the 'stripe count = -1' case properly.
2295 --------------------------------------------------------------------------------
2297 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2299 * Support for kernels:
2300 2.6.5-7.283 (SLES 9),
2301 2.6.9-55.EL (RHEL 4),
2302 2.6.16.46-0.14 (SLES 10),
2303 2.6.18.8 vanilla (kernel.org)
2304 * Client support for unpatched kernels:
2305 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2306 2.6.16 - 2.6.22 vanilla (kernel.org)
2307 * Due to recently discovered recovery problems, we do not recommend
2308 using patchless RHEL 4 clients with this or any earlier release.
2309 * Recommended e2fsprogs version: 1.39.cfs8
2310 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2311 * Starting with this release, the ldiskfs backing filesystem required
2312 by Lustre is now in its own package, lustre-ldiskfs. This package
2313 should be installed. It is versioned separately from Lustre and
2314 may be released separately in future.
2318 Description: block reactivating mgc import until all deactivates complete
2319 Details : Fix race when failing back MDT/MGS to itself (testing)
2322 Frequency : at statup only
2324 Description: mds_lov_synchronize race leads to various problems
2325 Details : simultaneous MDT->OST connections at startup can cause the
2326 sync to abort, leaving the OSC in a bad state.
2328 Severity : enhancement
2330 Description: add optional extra BUILD_VERSION info
2331 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2332 to override the lustre version.
2335 Frequency : 2.6.18 servers only
2337 Description: ll_kern_mount() doesn't release the module reference
2338 Details : The ldiskfs module reference count never drops down to 0
2339 because ll_kern_mount() doesn't release the module reference.
2344 Description: server LBUG when using old ost_num_threads parameter
2345 Details : Accept the old ost_num_threads parameter but warn that it
2346 is deprecated, and fix an off-by-one error that caused an LBUG.
2351 Description: Transient SCSI error results in persistent IO issue
2352 Details : iobuf->dr_error is not reinitialized to 0 between two
2356 Frequency : sometimes when underlying device returns I/O errors
2358 Description: OSTs not going read-only during write failures
2359 Details : OSTs are not remounted read-only when the journal commit threads
2360 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2361 instead of the ext3 wrappers.
2365 Description: poor connect scaling with increasing client count
2366 Details : Don't run filter_grant_sanity_check for more than 100 exports
2367 to improve scaling for large numbers of clients.
2370 Frequency : SLES10 only
2372 Description: sanity-quota.sh quotacheck failed: rc = -22
2373 Details : Quotas cannot be enabled on SLES10.
2376 Frequency : liblustre clients only
2378 Description: getdirentries does not give error when run on compute nodes
2379 Details : getdirentries does not fail when the size specified as an argument
2380 is too small to contain at least one entry
2382 Severity : enhancement
2384 Description: Add LNET router traceability for debug purposes
2385 Details : If a checksum failure occurs with a router as part of the
2386 IO path, the NID of the last router that forwarded the bulk data
2387 is printed so it can be identified.
2392 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2393 Details : Due to a race condition, liblustre clients were occasionally
2394 evicted incorrectly.
2396 Severity : enhancement
2398 Description: lfs setstripe use optional parameters instead of postional
2401 Severity : enhancement
2403 Description: Nanosecond timestamp support for ldiskfs
2404 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2405 resolution timestamps. There is not yet support for this at
2406 the Lustre filesystem level.
2409 Frequency : during server recovery
2411 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2412 Details : request with rq_no_resend flag not awake l_wait_event if they get
2416 Frequency : nfs export on patchless client
2418 Description: connectathon hang when test nfs export over patchless client
2419 Details : Disconnected dentry cannot be found with lookup, so we do not need
2420 to unhash it or make it invalid
2423 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2424 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2425 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2430 Description: open req refcounting wrong on reconnect
2431 Details : If reconnect happened between getting open reply from server and
2432 call to mdc_set_replay_data in ll_file_open, we will schedule
2433 replay for unreferenced request that we are about to free.
2434 Subsequent close will crash in variety of ways.
2435 Check that request is still eligible for replay in
2436 mdc_set_replay_data().
2441 Description: disable writes to filesystem when reading health_check file
2442 Details : the default for reading the health_check proc file has changed
2443 to NOT do a journal transaction and write to disk, because this
2444 can cause reads of the /proc file to hang and block HA state
2445 checking on a healthy but otherwise heavily loaded system. It
2446 is possible to return to the previous behaviour during configure
2447 with --enable-health-write.
2449 Severity : enhancement
2451 Description: 64-bit inode version
2452 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2453 to the inode. This will be required for version-based recovery.
2458 Description: MDS fails to start if a duplicate client export is detected
2459 Details : in some rare cases it was possible for a client to connect to
2460 an MDS multiple times. Upon recovery the MDS would detect this
2461 and fail during startup. Handle this more gracefully.
2463 Severity : enhancement
2465 Description: Add -o localflock option to simulate old noflock
2467 Details : This will achieve local-only flock/fcntl locks
2473 Description: log_commit_thread vs filter_destroy race leads to crash
2474 Details : Take import reference before releasing llog record semaphore
2479 Description: Wrong request locking in request set processing
2480 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2481 imp_delayed_list, in this place should be used imp_lock.
2484 Frequency : when reconnection
2486 Description: Grant Leak when osc reconnect to OST
2487 Details : When osc reconnect ost, OST(filter) should check whether it
2488 should grant more space to client by comparing fed_grant and
2489 cl_avail_grant, and return the granted space to client instead
2490 of "new granted" space, because client will call osc_init_grant
2491 to update the client grant space info.
2494 Frequency : when client reconnect to OST
2496 Description: Grant Leak when osc do resend and replay bulk write
2497 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2498 bulk write request, because the grant info while be sync between
2499 OSC and OST when reconnect, and we should ignore the grant info
2500 these of resend/replay write req.
2505 Description: Grant space more than avaiable left space sometimes.
2506 Details : When then OST is about to be full, if two bulk writing from
2507 different clients came to OST. Accord the avaliable space of the
2508 OST, the first req should be permitted, and the second one
2509 should be denied by ENOSPC. But if the seconde arrived before
2510 the first one is commited. The OST might wrongly permit second
2511 writing, which will cause grant space > avaiable space.
2514 Frequency : when client is evicted
2516 Description: Grant might be wrongly erased when osc is evicted by OST
2517 Details : when the import is evicted by server, it will fork another
2518 thread ptlrpc_invalidate_import_thread to invalidate the
2519 import, where the grant will be set to 0. While the original
2520 thread will update the grant it got when connecting. So if
2521 the former happened latter, the grant will be wrongly errased
2522 because of this race.
2527 Description: Checking Stale with correct fid
2528 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2529 to check whether it is stale, because sometimes, we want the
2530 enqueue happened anyway, and op_data.fid2 will not be initialized.
2532 Severity : enhancement
2534 Description: update patchless client
2535 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2538 Frequency : only with 2.4 kernel
2540 Description: random memory corruption
2541 Details : size of struct ll_inode_info is to big for union inode.u and this
2542 can be cause of random memory corruption.
2547 Description: Memory leak in recovery
2548 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2549 It should also check obd_fail before fsfilt_start, otherwise if
2550 fsfilt_start return -EROFS,(failover mds during mds recovery).
2551 then the req will return with repmsg->transno = 0 and rc = EROFS.
2552 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2553 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2554 be freed no matter whether fsfilt_commit success or not.
2557 Frequency : only with huge count clients
2559 Description: Prevents from taking the superblock lock in llap_from_page for
2561 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2562 for llap_from_page call in ll_removepage prevents from taking the
2563 superblock lock for a soon died page.
2568 Description: Not check open intent error before release open handle
2569 Details : in some rare cases, the open intent error is not checked before
2570 release open handle, which may cause
2571 ASSERTION(open_req->rq_transno != 0), because it tries to release
2572 the failed open handle.
2577 Description: Set cat log bitmap only after create log success.
2578 Details : in some rare cases, the cat log bitmap is set too early. and it
2579 should be set only after create log success.
2583 Description: Accessing a block bevice can re-enable I/O when Lustre is
2584 tearing down a device.
2585 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2590 Description: service threads may hog cpus when there are a lot of requests
2591 Details : Insert cond_resched to give other threads a chance to use some CPU
2596 Description: the cat log was not initialized in recovery
2597 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2598 unlink log on mds will not be initialized until mds post
2599 recovery. And also in mds post recovery, the unlink log will
2600 initialization will be done asynchronausly, so there will be race
2601 between add unlink log and unlink log initialization.
2605 Description: brw_stats were being printed incorrectly
2606 Details : brw_stats were being printed as log2 but all of them were not
2607 recorded as log2. Also remove some code duplication arising from
2608 filter_tally_{read,write}.
2612 Frequency : rare, only in recovery.
2613 Description: ASSERTION(req->rq_type != LI_POISON) failed
2614 Details : imp_lock should be held while iterating over imp_sending_list for
2615 prevent destroy request after get timeout in ptlrpc_queue_wait.
2619 Description: replay-single.sh test 52 fails
2620 Details : A lock's skiplist need to be cleanup when it being unlinked
2621 from its resource list.
2625 Description: Short directio read returns full requested size rather than
2627 Details : Direct I/O operations should return actual amount of bytes
2628 transferred rather than requested size.
2630 Severity : enhancement
2632 Description: metadata RPC reduction (e.g. for rm performance)
2633 Details : decrease the amount of synchronous RPC between clients and servers
2634 by canceling conflicing lock before the operation on the client
2635 and packing thier handles into the main operation RPC to server.
2637 Severity : enhancement
2639 Description: Async OSC create to avoid the blocking unnecessarily.
2640 Details : If a OST has no remain object, system will block on the creating
2641 when need to create a new object on this OST. Now, ways use
2642 pre-created objects when available, instead of blocking on an
2643 empty osc while others are not empty. If we must block, we block
2644 for the shortest possible period of time.
2648 Description: improve handling recoverable errors
2649 Details : if request processig with error which can be recoverable on server
2650 request should be resend, otherwise page released from cache and
2653 Severity : enhancement
2655 Description: refine locking for avoid write wrong info into lov_objid file
2656 Details : fix possible races with add new target and write/update data in
2659 --------------------------------------------------------------------------------
2661 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2666 Frequency : on some architectures
2668 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2669 Details : Uninitialized flags sometimes cause configuration commands to
2673 Frequency : patchless clients only
2675 Description: missing __iget() symbol export
2676 Details : The __iget() symbol export is missing. To avoid the need for
2677 this on patchless clients the deathrow inode reaper is turned
2678 off, and we depend on the VM to clean up old inodes. This
2679 dependency was during via the fix for bug 12181.
2681 --------------------------------------------------------------------------------
2683 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2685 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2686 INTEROPERATE with older versions automatically. In many cases a
2687 special upgrade step is needed. Please read the
2688 user documentation before upgrading any part of a live system.
2689 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2690 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2691 * WARNING: Lustre configuration and startup changes are required with
2692 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2696 Severity : enhancement
2698 Description: MountConf
2699 Details : Lustre configuration is now managed via mkfs and mount
2700 commands instead of lmc and lconf. New obd types (MGS, MGC)
2701 are added for dynamic configuration management. See
2702 https://mail.clusterfs.com/wikis/lustre/MountConf for
2705 Severity : enhancement
2707 Description: dynamic OST addition
2708 Details : OSTs can now be added to a live filesystem
2710 Severity : enhancement
2712 Description: startup order invariance
2713 Details : MDTs and OSTs can be started in any order. Clients only
2714 require the MDT to complete startup.
2716 Severity : enhancement
2718 Description: parallel, asynchronous orphan cleanup
2719 Details : orphan cleanup is now performed in separate threads for each
2720 OST, allowing parallel non-blocking operation.
2722 Severity : enhancement
2724 Description: optimized stripe assignment
2725 Details : stripe assignments are now made based on ost space available,
2726 ost previous usage, and OSS previous usage, in order to try
2727 to optimize storage space and networking resources.
2729 Severity : enhancement
2731 Description: Permanently set tunables
2732 Details : All writable /proc/fs/lustre tunables can now be permanently
2733 set on a per-server basis, at mkfs time or on a live system.
2735 Severity : enhancement
2737 Description: Lustre message v2
2738 Details : Add lustre message format v2.
2740 Severity : enhancement
2742 Description: client OST exclusion list
2743 Details : Clients can be started with a list of OSTs that should be
2744 declared "inactive" for known non-responsive OSTs.
2748 Description: SPEC SFS validation failure on NFS v2 over lustre.
2749 Details : Changes the blocksize for regular files to be 2x RPC size,
2750 and not depend on stripe size.
2752 Severity : enhancement
2754 Description: Multiple MD RPCs in flight.
2755 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2756 To avoid overly-overloading MDS, introduce a limit on number of
2757 MDS RPCs in flight for a single client and add /proc controls
2758 to adjust this limit.
2760 Severity : enhancement
2762 Description: client read/write statistics
2763 Details : Add client read/write call usage stats for performance
2764 analysis of user processes.
2765 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2766 file access. extents_stats shows chunk size distribution.
2767 extents_stats_per_process show chunk size distribution per
2770 Severity : enhancement
2772 Description: mds statistics
2773 Details : Add detailed mds operations statistics in
2774 /proc/fs/lustre/mds/*/stats.
2778 Description: Failure of copying files with lustre special EAs.
2779 Details : Client side always return success for setxattr call for lustre
2780 special xattr (currently only "trusted.lov").
2785 Description: Refcount LNET uuids
2786 Details : The global LNET uuid list grew linearly with every startup;
2787 refcount repeated list entries instead of always adding to
2790 Severity : enhancement
2792 Description: Dynamic service threads
2793 Details : Within a small range, start extra service threads
2794 automatically when the request queue builds up.
2797 Frequency : mixed-endian client/server environments
2799 Description: mixed-endian crashes
2800 Details : The new msg_v2 system had some failures in mixed-endian
2803 Severity : enhancement
2805 Description: Easy OST removal
2806 Details : OSTs can be permanently deactivated with e.g. 'lctl
2807 conf_param lustre-OST0001.osc.active=0'
2809 Severity : enhancement
2811 Description: MGS proc entries
2812 Details : Added basic proc entries for the MGS showing what filesystems
2815 Severity : enhancement
2817 Description: provide MGS failover
2818 Details : Added config lock reacquisition after MGS server failover.
2820 Severity : enhancement
2822 Description: add Linux 2.4 support
2823 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2827 Description: a large application tries to do I/O to the same resource and dies
2828 in the middle of it.
2829 Details : Check the req->rq_arrival time after the call to
2830 ost_brw_lock_get(), but before we do anything about
2831 processing it & sending the BULK transfer request. This
2832 should help move old stale pending locks off the queue as
2833 quickly as obd_timeout.
2836 Frequency : when an incorrect nid is specified during startup
2838 Description: ptlrpc connect to non-existant node causes kernel crash
2839 Details : LNET can't be re-entered from an event callback, which
2840 happened when we expire a message after the export has been
2841 cleaned up. Instead, hand the zombie cleanup off to another
2844 Severity : enhancement
2846 Description: plain/inodebits lock performance improvement
2847 Details : Grouping plain/inodebits in granted list by their request modes
2848 and bits policy, thus improving the performance of search through
2852 Frequency : only if OST filesystem is corrupted
2854 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2855 Details : for a short time RPCs with bulk IO are in the replay list,
2856 but replay of bulk IOs is unimplemented. If the OST filesystem
2857 is corrupted due to disk cache incoherency and then replay is
2858 started it is possible to trip an assertion. Avoid putting
2859 committed RPCs into the replay list at all to avoid this issue.
2862 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2864 Description: System hang on startup
2865 Details : This bug allowed the liblustre (e.g. catamount) client to
2866 return to the app before handling all startup RPCs. This
2867 could leave the node unresponsive to lustre network traffic
2868 and manifested as a server ptllnd timeout.
2870 Severity : enhancement
2872 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2873 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2874 boolean to control whether to print peer debug info when a
2875 client's RPC times out.
2878 Frequency : only for kernels with patches from Lustre below 1.4.3
2880 Description: Remove old rdonly API
2881 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2884 Frequency : only for devices with external journals
2886 Description: Set external device read-only also
2887 Details : During a commanded failover stop, we set the disk device
2888 read-only while the server shuts down. We now also set any
2889 external journal device read-only at the same time.
2892 Frequency : when upgrading from 1.4 while trying to change parameters
2894 Description: The wrong (new) MDC name was used when setting parameters for
2895 upgraded MDT's. Also allows changing of OSC (and MDC)
2896 parameters if --writeconf is specified at tunefs upgrade time.
2899 Frequency : when setting specific ost indicies
2901 Description: QOS code breaks on skipped indicies
2902 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2903 created with --index need not be sequential.
2905 Severity : enhancement
2907 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2908 Details : The uninit_groups feature works in conjunction with the kernel
2909 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2910 up the pass1 processing of e2fsck. This is a read-only feature
2911 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2912 mount filesystems that have had this feature enabled.
2914 Severity : enhancement
2916 Description: Improve multi-block allocation algorithm to avoid fragmentation
2917 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2918 allocation locality and avoid filesystem fragmentation.
2920 ------------------------------------------------------------------------------
2922 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2924 * Support for kernels:
2925 2.4.21-47.0.1.EL (RHEL 3)
2926 2.6.5-7.283 (SLES 9)
2927 2.6.9-42.0.10.EL (RHEL 4)
2928 2.6.12.6 vanilla (kernel.org)
2929 2.6.16.27-0.9 (SLES 10)
2930 * Recommended e2fsprogs version: 1.39.cfs5
2932 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2936 Frequency : occasional, depends on client load and configuration
2937 Bugzilla : 12181, 12203
2938 Description: data loss for recently-modified files
2940 Details : In some cases it is possible that recently written or created
2941 files may not be written to disk in a timely manner (this should
2942 normally be within 30s unless client IO load is very high).
2943 The problem appears as zero-length files or files that are a
2944 multiple of 1MB in size after a client crash or client eviction
2945 that are missing data at the end of the file.
2947 This problem is more likely to be hit on clients where files are
2948 repeatedly created and unlinked in the same directory, clients
2949 have a large amount of RAM, have many CPUs, the filesystem has
2950 many OSTs, the clients are rebooted frequently, and/or the files
2951 are not accessed by other nodes after being written.
2953 The presence of the problem can be detected by looking at
2954 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2955 smaller than the second (nr_unused) then dirty files will not
2956 be flushed automatically to disk. "sync; sleep 10" should be
2957 run several times on the node before unmounting it to update
2958 Lustre (this is also safe to run on nodes without this problem).
2960 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2961 that can cause this same problem, so customers using that kernel
2962 also need to update the kernel in addition to Lustre. In order
2963 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2965 It is normal that files written just before a client crash (less
2966 than 30s) may not yet have been flushed to disk, even for local
2970 Frequency : frequent on thin XT3 nodes
2972 Description: UUID collision on thin XT3 Linux nodes
2973 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2974 randomly, since we relied on an insufficiently-seeded PRNG.
2979 Description: OSS hangs after "All ost request buffers busy"
2980 Details : A deadlock between quota and journal operations caused OSS
2981 hangs after printing "All ost request buffers busy."
2984 Frequency : always on liblustre builds
2986 Description: Cleanup compiler warnings on liblustre
2989 Frequency : always on liblustre builds on XT3
2991 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2995 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2996 > 32000 subdirectories
2997 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2998 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2999 subdirectory count crosses 32000. This will aid e2fsck to
3000 correctly handle more than 32000 subdirectories.
3003 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3005 Description: System hang on startup
3006 Details : This bug allowed the liblustre (e.g. catamount) client to
3007 return to the app before handling all startup RPCs. This
3008 could leave the node unresponsive to lustre network traffic
3009 and manifested as a server ptllnd timeout.
3011 Severity : enhancement
3013 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3014 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3015 boolean to control whether to print peer debug info when a
3016 client's RPC times out.
3021 Description: make O_SYNC working on 2.6 kernels
3022 Details : 2.6 kernels use different method for mark pages for write,
3023 so need add a code to lustre for O_SYNC work.
3028 Description: Failure to close file and release space on NFS
3029 Details : Put inode details into lock acquired in ll_intent_file_open.
3030 Use mdc_intent_lock in ll_intent_open to properly
3031 detect all kind of errors unhandled by mdc_enqueue.
3036 Description: proc file read during shutdown sometimes raced obd removal,
3038 Details : Add lock to prevent obd access after proc file removal.
3041 Frequency : Only for files larger than 4GB on 32-bit clients.
3043 Description: improperly doing page alignment of locks
3044 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3045 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3050 Description: RPCs being resent when they shouldn't be
3051 Details : Some RPCs that should not be resent are being resent. This
3052 can cause inconsistencies in the RPC state machine. Do not
3053 resend such requests.
3056 Frequency : rare, only with NFS export
3058 Description: Crash on NFS re-export node
3059 Details : under very unusual load conditions an assertion is hit in
3060 ll_intent_file_open()
3063 Frequency : only if OST filesystem is corrupted
3065 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3066 Details : for a short time RPCs with bulk IO are in the replay list,
3067 but replay of bulk IOs is unimplemented. If the OST filesystem
3068 is corrupted due to disk cache incoherency and then replay is
3069 started it is possible to trip an assertion. Avoid putting
3070 committed RPCs into the replay list at all to avoid this issue.
3075 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3076 Details : Large single O_DIRECT read and write calls can fail to allocate
3077 a sufficiently large buffer to process the request. In case of
3078 allocation failure the allocation is retried with a smaller
3079 buffer and broken into smaller requests.
3081 Severity : enhancement
3083 Description: Add -o localflock option to simulate old noflock behaviour.
3084 Details : This will achieve local-only flock/fcntl locks coherentness.
3089 Description: versioning check is incomplete
3090 Details : Checking the version difference of client vs. server, report
3091 error if the gap is too big.
3096 Description: add support PG_writeback bit
3097 Details : add support for PG_writeback bit for Lustre, for more carefull
3098 work with page cache in 2.6 kernel. This also fix some deadlocks
3099 and remove hack for work O_SYNC with 2.6 kernel.
3101 Severity : enhancement
3103 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3104 Details : The uninit_groups feature works in conjunction with the kernel
3105 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3106 up the pass1 processing of e2fsck. This is a read-only feature
3107 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3108 mount filesystems that have had this feature enabled.
3110 Severity : enhancement
3112 Description: Improve multi-block allocation algorithm to avoid fragmentation
3113 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3114 allocation locality and avoid filesystem fragmentation.
3116 ------------------------------------------------------------------------------
3118 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3120 * Support for kernels:
3121 2.6.9-42.0.3.EL (RHEL 4)
3122 2.6.5-7.276 (SLES 9)
3123 2.4.21-47.0.1.EL (RHEL 3)
3124 2.6.12.6 vanilla (kernel.org)
3125 2.6.16.21-0.8 (SLES10)
3126 * Recommended e2fsprogs version: 1.39.cfs2-0
3128 * The backwards-compatible /proc/sys/portals symlink has been removed
3129 in this release. Before upgrading, please ensure that you change
3130 any configuration scripts or /etc/sysctl.conf files that access
3131 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3132 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3133 in advance of the upgrade on any system running Lustre 1.4.6 or
3134 newer, since /proc/sys/lnet was added in that version.
3135 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3139 Frequency : only when quota is used
3141 Description: avoid scanning export list for quota master
3142 Details : Change the algorithms to avoid scanning export list in order
3143 to improve the efficiency.
3146 Frequency : MDS failover only, very rarely
3148 Description: "went back in time" messages on mds failover
3149 Details : The greatest transno may be lost when the current operation
3150 finishes with an error (transno==0) and the client's last_rcvd
3151 record is over-written. Save the greatest transno in the
3152 mds_last_transno for this case.
3155 Frequency : always for specific kernels and striping counts
3157 Description: client may get "Matching packet too big" without ACL support
3158 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3159 when trying to access files in certain configurations. The
3160 clients should in fact be denied when mounting because they do
3161 not understand ACLs.
3164 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3166 Description: many clients connecting with IO in progress causes connect timeouts
3167 Details : Avoid synchronous journal commits to avoid delays caused by many
3168 clients connecting/disconnecting when bulk IO is in progress.
3169 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3170 OST_IO_PORTAL to avoid delays behind potentially many pending
3174 Frequency : occasionally with multiple writers to a single file
3176 Description: shared writes to file may result in wrong size reported by stat()
3177 Details : Allow growing of kms when extent lock is cancelled
3180 Frequency : always with random mmap IO to multi-striped file
3182 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3183 Details : Only if the hole is at the end of OST object so that kms is too
3184 small. Fix is to increase kms accordingly in ll_nopage.
3187 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3189 Description: writes to a missing object would leak memory on the OST
3190 Details : If there is an inconsistency between the MDS and OST filesystems,
3191 such that the MDS references an object that doesn't exist, writes
3192 to that object will leak memory due to incorrect cleanup in the
3193 error handling path, eventually running out of memory on the OST.
3198 Description: Creating too long symlink causes lustre errors
3199 Details : Check symlink and name lengths before sending requests to MDS.
3202 Frequency : only if flock is enabled (not on by default)
3204 Description: posix locks not released on fd closure on 2.6.9+
3205 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3206 kernels, this caused such locks not to be released on fd close and
3207 then assertions on fs unmount about still used locks.
3210 Frequency : MDS failover only, very rarely
3212 Description: clients may get ASSERTION(granted_lock != NULL)
3213 Details : When request was taking a long time, and a client was resending
3214 a getattr by name lock request. The were multiple lock requests
3215 with the same client lock handle and
3216 mds_getattr_name->fixup_handle_for_resent_request found one of the
3217 lock handles but later failed with ASSERTION(granted_lock != NULL).
3222 Description: handle->h_buffer_credits > 0, assertion failure
3223 Details : h_buffer_credits is zero after truncate, causing assertion
3224 failure. This patch extends the transaction or creates a new
3228 Frequency : NFS re-export or patchless client
3229 Bugzilla : 11179, 10796
3230 Description: Crash on NFS re-export node (__d_move)
3231 Details : We do not want to hash the dentry if we don't have a lock.
3232 But if this dentry is later used in d_move, we'd hit uninitialised
3233 list head d_hash, so we just do this to init d_hash field but
3234 leave dentry unhashed.
3237 Frequency : NFS re-export or patchless client
3239 Description: NFS exports has problem with symbolic link
3240 Details : lustre client didn't properly install dentry when re-exported
3241 to NFS or running patchless client.
3244 Frequency : NFS re-export or patchless client
3246 Description: Various nfs/patchless fixes.
3247 Details : fixes reuse disconected alias for lookup process - this fixes
3248 warning "find_exported_dentry: npd != pd",
3249 fix permission error with open files at nfs.
3250 fix apply umask when do revalidate.
3253 Frequency : occasional
3255 Description: Crash on NFS re-export node
3256 Details : calling clear_page() on the wrong pointer triggered oops in
3257 generic_mapping_read().
3260 Frequency : rarely, using O_DIRECT IO
3262 Description: unaligned directio crashes client with LASSERT
3263 Details : check for unaligned buffers before trying any requests.
3266 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3268 Description: stale data returned from RAID cache
3269 Details : If only a small amount of IO is done to the RAID device before
3270 reading it again it is possible to get stale data from the RAID
3271 cache instead of reading it from disk.
3274 Frequency : always for sles10 kernel
3276 Description: sles10 support
3277 Details : ll_follow_link: compile fixes and using of nd_set_link
3278 under newer kernels.
3281 Frequency : depends on arch, kernel and compiler version, always on sles10
3284 Description: recursive or deep enough symlinks cause stack overflow
3285 Details : getting rid of large stack-allocated variable in
3289 Frequency : depends on hardware
3291 Description: lustre write performance loss in the SLES10 kernel
3292 Details : the performance loss is caused by using of write barriers in the
3293 ext3 code. The SLES10 kernel turns barrier support on by
3294 default. The fix is to undo that change for ldiskfs.
3296 ------------------------------------------------------------------------------
3298 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3300 * Support for kernels:
3301 2.6.9-42.0.3EL (RHEL 4)
3302 2.6.5-7.276 (SLES 9)
3303 2.4.21-47.0.1.EL (RHEL 3)
3304 2.6.12.6 vanilla (kernel.org)
3308 Frequency : quota enabled and large files being deleted
3310 Description: releasing more than 4GB of quota at once hangs OST
3311 Details : If a user deletes more than 4GB of files on a single OST it
3312 will cause the OST to spin in an infinite loop. Release
3313 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3318 Description: statfs data retrieved from /proc may be stale or zero
3319 Details : When reading per-device statfs data from /proc, in the
3320 {kbytes,files}_{total,free,avail} files, it may appear
3321 as zero or be out of date.
3324 Frequency : systems with MD RAID1 external journal devices
3326 Description: lconf's call to blkid is confused by RAID1 journal devices
3327 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3328 of returning all block devices that match the journal UUID.
3331 Frequency : always, for aggregate stripe size over 4GB
3333 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3334 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3335 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3338 Frequency : always if "lfs find" used on a local file/directory
3340 Description: "lfs find" segfaults if used on a local file/directory
3341 Details : The case where a directory component was not specified wasn't
3342 handled correctly. Handle this properly.
3345 Frequency : always on ppc64
3347 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3348 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3349 perform endianness conversion.
3352 Frequency : rarely (truncate to non-zero file size after write under load)
3353 Bugzilla : 10730, 10687
3354 Description: Files padded with zeros to next 4K multiple
3355 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3356 it is possible that files that are truncated to a non-zero size
3357 immediately after being written are filled with zero bytes beyond
3358 the truncated size. No file data is lost.
3360 Severity : enhancement
3362 Description: Allow recovery/failover for liblustre clients.
3363 Details : liblustre clients were unaware of failover configurations until
3366 Severity : enhancement
3368 Description: user file locks should fail when not mounting with flock option
3369 Details : Set up an error-returning stub in ll_file_operations.lock field
3370 to prevent incorrect behaviour when client is mounted without
3371 flock option. Also, set up properly f_op->flock field for
3375 Frequency : always on ia64
3377 Description: "lfs df" loops on printing out MDS statfs information
3378 Details : The obd_ioctl_data was not initialized and in some systems
3379 this caused a failure during the ioctl that did not return
3380 an error. Initialize the struct and return an error on failure.
3383 Frequency : SLES 9 only
3385 Description: Error of copying files with lustre special EAs as root
3386 Details : Client side always return success for setxattr call for lustre
3387 special xattr (currently only "trusted.lov").
3390 Frequency : rarely on clusters with both ia64+i386 clients
3392 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3393 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3394 clients are doing concurrent writes to the same file it is
3395 possible that the ia64 clients may LASSERT because the OST
3396 extent locks are not PAGE_SIZE aligned. Ensure that grown
3397 locks are always aligned on the request boundary.
3400 Frequency : specific use, occasional
3402 Description: Overwriting in use executable truncates on-disk binary image
3403 Details : If one node attempts to overwrite an executable in use by
3404 another node, we now correctly return ETXTBSY instead of
3405 truncating the file.
3407 Severity : enhancement
3409 Description: Async OSC create to avoid the blocking unnecessarily.
3410 Details : If a OST has no remain object, system will block on the creating
3411 when need to create a new object on this OST. Now, ways use
3412 pre-created objects when available, instead of blocking on an
3413 empty osc while others are not empty. If we must block, we block
3414 for the shortest possible period of time.
3419 Description: chmod on Lustre root is propagated to other clients
3420 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3421 invalid by the follow_mount time.
3424 Frequency : liblustre clients only
3426 Description: Race in 'instant cancel' lock handling could lead to such locks
3427 never to be granted in case of SMP MDS
3428 Details : Do not destroy not yet granted but cbpending locks in
3432 Frequency : replay/resend of open
3434 Description: non null lock assetion failure in mds_intent_policy
3435 Details : Trying to replay/resend lockless open requests resulted in
3436 mds_open() returning 0 with no lock. Now it sets a flag if
3437 a lock is going to be returned.
3439 Severity : enhancement
3441 Description: Checksum enhancements
3442 Details : New checksum enhancements allow for resending RPCs that failed
3445 Severity : enhancement
3447 Description: Tunables on number of dirty pages in cacche
3448 Details : Allow to set limit on number of dirty pages cached.
3453 Description: client crash on unmount - lock still has references
3454 Details : In some error handling cases it was possible to leak a lock
3455 reference on a client while accessing a file. This was not
3456 harmful to the client during operation, but would cause the
3457 client to crash when the filesystem is unmounted.
3460 Frequency : specific case, rare
3462 Description: ETXTBSY on mds though file not in use by client
3463 Details : ETXTBSY is no longer incorrectly returned when attempting to
3464 chmod or chown a directory that the user previously tried to
3465 execute or a currently-executing binary.
3468 Frequency : extremely rare except on liblustre-based clients
3470 Description: Lustre space not freed when files are deleted
3471 Details : Clean up open-unlinked files after client eviction. Previously
3472 the unlink was skipped and the files remained as orphans.
3477 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3478 Details : In some cases it was possible to send a blocking callback to
3479 a client doing a glimpse, even though that client didn't get
3480 a lock granted. When the glimpse lock is cancelled on the OST
3481 the freed lock is left on the waiting list and corrupted the list.
3484 Frequency : all core dumps
3486 Description: Broke core dumps to lustre
3487 Details : Negative dentry may be unhashed if parent does not have UPDATE
3488 lock, but some callers, e.g. do_coredump, expect dentry to be
3489 hashed after successful create, hash it in ll_create_it.
3491 ------------------------------------------------------------------------------
3493 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3495 * Support for kernels:
3496 2.6.9-42.0.2.EL (RHEL 4)
3497 2.6.5-7.276 (SLES 9)
3498 2.4.21-40.EL (RHEL 3)
3499 2.6.12.6 vanilla (kernel.org)
3503 Frequency : always on RHEL 3
3505 Description: Number of open files grows over time
3506 Details : The number of open files grows over time, whether or not
3507 Lustre is started. This was due to a filp leak introduced
3508 by one of our kernel patches.
3510 ------------------------------------------------------------------------------
3512 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3514 * Support for kernels:
3515 2.6.9-42.EL (RHEL 4)
3516 2.6.5-7.276 (SLES 9)
3517 2.4.21-40.EL (RHEL 3)
3518 2.6.12.6 vanilla (kernel.org)
3523 Bugzilla : 5719, 9635, 9792, 9684
3524 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3525 Details : If a server is under heavy load and cannot reply to new
3526 connection requests before the client resends the (re)connect,
3527 the connection handling code can behave badly if two service
3528 threads are concurrently handing separate (re)connections from
3529 the same client. Add better locking to the connection handling
3530 code, and ensure that only a single connection will be processed
3531 for a given client UUID, even if the lock is dropped.
3533 Severity : enhancement
3535 Description: add TCP zero-copy support to kernel
3536 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3537 sends if the hardware supports scatter-gather and checksumming.
3538 This allows socklnd to do client-write and server-read more
3539 efficiently and reduce CPU utilization from skbuf copying.
3542 Frequency : only if NFS exporting from client
3544 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3545 Details : If setting attributes on a file created under NFS that had
3546 never been opened it would be possible to oops the client
3547 if the file had no objects.
3550 Frequency : always for liblustre
3552 Description: liblustre client does MDS+OSTs setattr RPC for each write
3553 Details : When doing a write from a liblustre client, the client
3554 incorrectly issued an RPC to the MDS and each OST the file was
3555 striped over in order to update the timestamps. When writing
3556 with small chunks and many clients this could overwhelm the MDS
3557 with RPCs. In all cases it would slow down the write because
3558 these RPCs are unnecessary.
3560 Severity : enhancement
3562 Description: allow number of MDS service threads to be changed at module load
3563 Details : It is now possible to change the number of MDS service threads
3564 running. Adding "options mds mds_num_threads={N}" to the MDS's
3565 /etc/modprobe.conf will set the number of threads for the next
3566 time Lustre is restarted (assuming the "mds" module is also
3567 reloaded at that time). The default number of threads will
3568 stay the same, 32 for most systems.
3573 Description: OST crash if filesystem is unformatted or corrupt
3574 Details : If an OST is started on a device that has never been formatted
3575 or if the filesystem is corrupt and cannot even mount then the
3576 error handling cleanup routines would dereference a NULL pointer.
3581 Description: NULL pointer deref in llap_from_page.
3582 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3583 incorrect) mapping if the page was truncated/reclaimed while it was
3584 searched for. Check for this condition and skip such pages when
3585 doing readahead. Introduce extra check to llap_from_page() to
3586 verify page->mapping->host is non-NULL (so page is not anonymous).
3589 Frequency : Sometimes when using sys_sendfile
3591 Description: "page not covered by a lock" warnings from ll_readpage
3592 Details : sendfile called ll_readpage without right page locks present.
3593 Now we introduced ll_file_sendfile that does necessary locking
3594 around call to generic_file_sendfile() much like we do in
3598 Frequency : with certain MDS communication failures at client mount time
3600 Description: NULL pointer deref after failed client mount
3601 Details : a client connection request may delayed by the network layer
3602 and not be sent until after the PTLRPC layer has timed out the
3603 request. If the client fails the mount immediately it will try
3604 to clean up before the network times out the request. Add a
3605 reference from the request import to the obd device and delay
3606 the cleanup until the network drops the request.
3609 Frequency : occasionally during client (re)connect
3611 Description: assertion failure during client (re)connect
3612 Details : processing a client connection request may be delayed by the
3613 client or server longer than the client connect timeout. This
3614 causes the client to resend the connection request. If the
3615 original connection request is replied in this interval, the
3616 client may trip an assertion failure in ptlrpc_connect_interpret()
3617 which thought it would be the only running connect process.
3620 Frequency : only with obd_echo servers and clients that are rebooted
3622 Description: kernel BUG accessing uninitialized data structure
3623 Details : When running an obd_echo server it did not start the ping_evictor
3624 thread, and when a client was evicted an uninitialized data
3625 structure was accessed. Start the ping_evictor in the RPC
3626 service startup instead of the OBD startup.
3628 Severity : enhancement
3629 Bugzilla : 10193 (patchless)
3630 Description: Remove dependency on various unexported kernel interfaces.
3631 Details : No longer need reparent_to_init, exit_mm, exit_files,
3632 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3635 Frequency : rare (only users of deprecated and unsupported LDAP config)
3637 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3638 Details : LDAP apparently contains 'lustreName' attributes instead of
3639 'name'. A simple remapping of the name is sufficient.
3642 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3644 Description: waiting_locks_callback trips kernel BUG if client is evicted
3645 Details : Running with the dump_on_timeout debug flag turned on makes
3646 it possible that the waiting_locks_callback() can try to dump
3647 the Lustre kernel debug logs from an interrupt handler. Defer
3648 this log dumping to the expired_lock_main() thread.
3650 Severity : enhancement
3652 Description: Support NFS exporting on 2.6 kernels.
3653 Details : Implement non-rawops metadata methods for NFS server to use without
3654 changing NFS server code.
3657 Frequency : very rare (synthetic metadata workload only)
3659 Description: two racing renames might cause an MDS thread to deadlock
3660 Details : Running the "racer" program may cause one MDS thread to rename
3661 a file from being the source of a rename to being the target of
3662 a rename at exactly the same time that another thread is doing
3663 so, and the second thread has already enqueued these locks after
3664 doing a lookup of the target and is trying to relock them in
3665 order. Ensure that we don't try to re-lock the same resource.
3668 Frequency : only very large systems with liblustre clients
3670 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3671 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3672 all OSTs in parallel. This allows the eviction of stale liblustre
3673 clients to proceed much faster than if they were done in series,
3674 and also offers similar improvements for other set_info RPCs.
3679 Description: excessive CPU usage during initial read phase on client
3680 Details : During the initial read phase on a client, it would agressively
3681 retry readahead on the file, consuming too much CPU and impacting
3682 performance (since 1.4.5.8). Improve the readahead algorithm
3683 to avoid this, and also improve some other common cases (read
3684 of small files in particular, where "small" is files smaller than
3685 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3690 Description: MDS crash when receiving packet with unknown intent.
3691 Details : Do not LBUG in unknown intent case, just return -EFAULT
3693 Severity : enhancement
3694 Bugzilla : 9293, 9385
3695 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3696 Details : Do not serialize getattr (non-intent version) and statfs.
3699 Frequency : occasional, when OST network is overloaded/intermittent
3701 Description: client evicted by OST after bulk IO timeout
3702 Details : If a client sends a bulk IO request (read or write) the OST
3703 may evict the client if it is unresposive to its data GET/PUT
3704 request. This is incorrect if the network is overloaded (takes
3705 too long to transfer the RPC data) or dropped the OST GET/PUT
3706 request. There is no need to evict the client at all, since
3707 the pinger and/or lock callbacks will handle this, and the
3708 client can restart the bulk request.
3711 Frequency : Always when mmapping file with no objects
3713 Description: client crashes when mmapping file with no objects
3714 Details : Check that we actually have objects in a file before doing any
3715 operations on objects in ll_vm_open, ll_vm_close and
3721 Description: Request leak when working with deleted CWD
3722 Details : Introduce advanced request refcount tracking for requests
3723 referenced from lustre intent.
3725 Severity : Enhancement
3727 Description: Cache open file handles on client.
3728 Details : MDS now will return special lock along with openhandle, if
3729 requested and client is allowed to hold openhandle, even if unused,
3730 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3731 closing files for every read/write openration.
3733 Severity : Enhancement
3735 Description: Cache open negative dentries on client when possible.
3736 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3737 negative dentries on lock revocation.
3742 Description: Remounting a client read-only wasn't possible with a zconf mount
3743 Details : It wasn't possible to remount a client read-only with llmount.
3745 Severity : enhancement
3746 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3747 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3748 MPICH is included to add Lustre-specific ADIO interfaces.
3749 This is based closely on the UFS ADIO layer and only differs
3750 in file creation, in order to allow the OST striping to be set.
3751 This is user-contributed code and not supported by CFS.
3756 Description: extended inode attributes (immutable, append-only) work improperly
3757 when 2.4 and 2.6 kernels are used on client/server or vice versa
3758 Details : Introduce kernel-independent values for these flags.
3760 Severity : enhancement
3763 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3764 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3765 granularity. Now, Fractional MB granularity is be supported,
3766 this is very useful for low memory system.
3768 Severity : enhancement
3770 Description: Getattr by fid
3771 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3772 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3773 have locks and avoids getattr rpc requests for it on every stat.
3776 Frequency : Always, for filesystems larger than 2TB
3778 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3779 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3780 filesystems larger than 2TB to be mounted with mballoc enabled.
3783 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3785 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3786 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3787 filesystems larger than 2TB is mounted then the kernel will
3788 silently corrupt the start of the filesystem. CONFIG_LBD is
3789 enabled for all CFS-supported kernels, but the possibility of
3790 this happening with a modified kernel config exists.
3792 Severity : enhancement
3794 Description: add client O_DIRECT support for 2.6 kernels
3795 Details : It is now possible to do O_DIRECT reads and writes to files
3796 in the Lustre client mountpoint on 2.6 kernel clients.
3798 Severity : enhancement
3800 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3801 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3802 parallel, not waiting for response from every OST before sending
3803 a rpc to the next OST.
3808 Description: setattr vs write race when updating file timestamps
3809 Details : Client processes that update a file timestamp into the past
3810 right after writing to the file (e.g. tar) it is possible that
3811 the updated file modification time can be reset to the current
3812 time due to a race between processing the setattr and write RPC.
3814 Severity : enhancement
3816 Description: Bring 'lfs find' closer in line with regular Linux find.
3817 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3818 -print0 options and obtains all the needed info through the lustre
3821 Severity : enhancement
3823 Description: support up to 1024 configured devices on one node
3824 Details : change obd_dev array from statically allocated to dynamically
3825 allocated structs as they are first used to reduce memory usage
3830 Description: Flush dirty partially truncated pages during truncate
3831 Details : Immediatelly flush partially truncated pages in filter_setattr,
3832 this way we completely avoid having any pages in page cache on OST
3833 and can retire ugly workarounds during writes to flush such pages.
3838 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3839 Details : protect inode from truncation within vfs_unlink() context
3840 just take a reference before calling vfs_unlink() and release it
3841 when parent's i_sem is free.
3846 Description: last_id value checked outside lock on OST caused LASSERT failure
3847 Details : If there were multiple MDS->OST object precreate requests in
3848 flight, it was possible that the OST's last object id was checked
3849 outside a lock and incorrectly tripped an assertion. Move checks
3850 inside locks, and discard old precreate requests.
3853 Frequency : always, if extents are used on OSTs
3855 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3856 Details : OSTs using the extents format would not zero the high 16 bits of
3857 the index physical block number. This is not a problem for any
3858 OST filesystems smaller than 16TB, and no kernels support ext3
3859 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3860 new/modified files) and can be fixed for existing filesystems
3861 with e2fsprogs-1.39-cfs1.
3866 Description: import connection selection may be incorrect if timer wraps
3867 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3868 import connections to be ignored if the 32-bit jiffies counter
3869 wraps. Use a 64-bit jiffies counter.
3872 Frequency : very large clusters immediately after boot
3874 Description: LNET request buffers exhausted under heavy short-term load
3875 Details : If a large number of client requests are generated on a service
3876 that has previously never seen so many requests it is possible
3877 that the request buffer growth cannot keep up with the spike in
3878 demand. Instead of dropping incoming requests, they are held in
3879 the LND until the RPC service can accept more requests.
3882 Frequency : Sometimes during replay
3884 Description: Assertion failure in ll_local_open after replay.
3885 Details : If replay happened on an open request reply before we were able
3886 to set replay handler, reply will become not swabbed tripping the
3887 assertion in ll_local_open. Now we set the handler right after
3888 recognising of open request
3891 Frequency : very rare
3893 Description: kernel reports "badness in vsnprintf"
3894 Details : Reading from the "recovery_status" /proc file in small chunks
3895 may cause a negative length in lprocfs_obd_rd_recovery_status()
3896 call to vsnprintf() (which is otherwise harmless). Exit early
3897 if there is no more space in the output buffer.
3899 Severity : enhancement
3901 Description: clear OBD RPC statistics by writing to them
3902 Details : It is now possible to clear the OBD RPC statistics by writing
3903 to the "stats" file.
3908 Description: Client mtime is not the same on different clients after utimes
3909 Details : In some cases, the client was using the utimes() syscall on
3910 a file cached on another node. The clients now validate the
3911 ctime from the MDS + OSTs to determine which one is right.
3916 Description: Inability to activate failout mode
3917 Details : lconf script incorrectly assumed that in pythong string's numeric
3918 value is used in comparisons.
3921 Frequency : always with multiple stripes per file
3923 Description: Inefficient object allocation for mutli-stripe files
3924 Details : When selecting which OSTs to stripe files over, for files with
3925 a stripe count that divides evenly into the number of OSTs,
3926 the MDS is always picking the same starting OST for each file.
3927 Return the OST selection heuristic to the original design.
3932 Description: mount failures may take full timeout to return an error
3933 Details : Under some heavy load conditions it is possible that a
3934 failed mount can wait for the full obd_timeout interval,
3935 possibly several minutes, before reporting an error.
3936 Instead return an error as soon as the status is known.
3938 Frequency : quota enabled and large files being deleted
3940 Description: releasing more than 4GB of quota at once hangs OST
3941 Details : If a user deletes more than 4GB of files on a single OST it
3942 will cause the OST to spin in an infinite loop. Release
3943 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3948 Description: statfs data retrieved from /proc may be stale or zero
3949 Details : When reading per-device statfs data from /proc, in the
3950 {kbytes,files}_{total,free,avail} files, it may appear
3951 as zero or be out of date.
3954 Frequency : systems with MD RAID1 external journal devices
3956 Description: lconf's call to blkid is confused by RAID1 journal devices
3957 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3958 of returning all block devices that match the journal UUID.
3961 Frequency : always, for aggregate stripe size over 4GB
3963 Description: assertion fails when trying to use 4GB stripe size
3964 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3965 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3968 Frequency : always on ppc64
3970 Description: the first write on an ext3 filesystem with mballoc got stuck
3971 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3972 perform endianness conversion.
3974 ------------------------------------------------------------------------------
3976 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3978 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3979 INTEROPERATE with older versions automatically. Please read the
3980 user documentation before upgrading any part of a live system.
3981 * WARNING: Lustre networking configuration changes are required with
3982 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3985 * Support for kernels:
3986 2.6.9-22.0.2.EL (RHEL 4)
3987 2.6.5-7.244 (SLES 9)
3988 2.6.12.6 vanilla (kernel.org)
3991 Severity : enhancement
3992 Bugzilla : 7981/8208
3993 Description: Introduced Lustre Networking (LNET)
3994 Details : LNET is new networking infrastructure for Lustre, it includes
3995 a reorganized network configuration mode (see the user
3996 documentation for full details) as well as support for routing
3997 between different network fabrics. Lustre Networking Devices
3998 (LNDs) for the supported network fabrics have also been
3999 created for this new infrastructure.
4001 Severity : enhancement
4002 Description: Introduced Access control lists
4003 Details : clients can set ACLs on files and directories in order to have
4004 more fine-grained permissions than the standard Unix UGO+RWX.
4005 The MDS must be started with the "-o acl" mount option.
4007 Severity : enhancement
4008 Description: Introduced filesystem quotas
4009 Details : Administrators may now establish per-user quotas on the
4012 Severity : enhancement
4014 Description: Configuration change for the XT3
4015 The PTLLND is now used to run Lustre over Portals on the XT3
4016 The configure option(s) --with-cray-portals are no longer used.
4017 Rather --with-portals=<path-to-portals-includes> is used to
4018 enable building on the XT3. In addition to enable XT3 specific
4019 features the option --enable-cray-xt3 must be used.
4024 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4025 Details : A race between allocating a new llcd and re-getting the llcd_lock
4026 allowed another thread to grab newly-allocated llcd.
4028 Severity : enhancement
4030 Description: 2.6 OST async journal commit and locking fix to improve performance
4031 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4032 2.6 kernels are now async as they already were in 2.4 kernels so
4033 that they can commit concurrently with the network bulk transfer.
4034 For block-allocated files the filter allocation semaphore is held
4035 to avoid filesystem fragmentation during allocation. BKL lock
4036 removed for 2.6 xattr operations where it is no longer needed.
4041 Description: lconf incorrectly determined whether two IP networks could talk
4042 Details : In some more complicated routing and multiple-network
4043 configurations, lconf will avoid trying to make a network
4044 connection to a disjoint part of the IP space. It was doing the
4045 math incorrectly for one set of cases.
4050 Description: Fix for potential infinite loop processing records in an llog.
4051 Details : If an llog record is corrupted/zeroed, it is possible to loop
4052 forever in llog_process(). Validate the llog record length
4053 and skip the remainder of the block on error.
4056 Frequency : occasional (liblustre only)
4058 Description: liblustre could not open files whose last component is a symlink
4059 Details : sysio_path_walk() would incorrectly pass the open intent to
4060 intermediate path components.
4063 Frequency : rare (liblustre only with non-standard tuning)
4064 Bugzilla : 7201 (7350)
4065 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4066 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4067 after it has been released, only do this for referenced locks
4069 Severity : enhancement
4071 Description: specifying an (invalid) directory default stripe_size of -1
4072 would reset the directory default striping
4073 Details : stripe_size -1 was used internally to signal directory stripe
4074 removal, now use "all default" to signal dir stripe removal
4075 as a directory striping of "all default" is not useful
4078 Frequency : common for large clusters running liblustre clients
4080 Description: doing an ls when liblustre clients are running is slow
4081 Details : sending a glimpse AST to a liblustre client waits for every AST
4082 to time out, as liblustre clients will not respond. Since they
4083 cannot cache data we refresh the OST lock LVB from disk instead.
4085 Severity : enhancement
4087 Description: doing an ls at the same time as file IO can be slow
4088 Details : enqueue and other "small" requests can be blocked behind many
4089 large IO requests. Create a new OST IO portal for non-IO
4090 requests so they can be processed faster.
4093 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4095 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4096 created file with mode 000
4097 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4098 allow an ftruncate() as a non-root user to a file with mode 000.
4099 Lustre now respects this flag to disable mode checks when
4100 truncating a file owned by the user
4103 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4106 Description: Revoking locks from clients that went dead or catatonic might take
4108 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4109 cancellation of such locks instant on servers without waiting for
4110 any reply from clients. Clients drops these locks when cancel
4111 notification from server is received without replying.
4114 Frequency : liblustre-only, when liblustre client dies or becomes busy
4116 Description: Doing ls on Linux clients can take a long time with active
4118 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4119 granting such locks to it in the first place if possible. Locks
4120 are taken by proxy on the OST during the read or write and
4121 dropped immediately afterward. Add connect flags handling, do
4122 not grant locks to liblustre clients for glimpse ASTs.
4124 Severity : enhancement
4126 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4127 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4128 sequential reads and behaves badly with applications doing random
4129 reads. Improve it by reading ahead at least read region, and
4130 avoiding excessive large RPC for small reads.
4132 Severity : enhancement
4134 Description: Creating more than 1000 files for a single job may cause a load
4135 imbalance on the OSTs if there are also a large number of OSTs.
4136 Details : qos_prep_create() uses an OST index reseed value that is an
4137 even multiple of the number of available OSTs so that if the
4138 reseed happens in the middle of the object allocation it will
4139 still utilize the OSTs as uniformly as possible.
4144 Description: OST or MDS may oops in ping_evictor_main()
4145 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4146 but doesn't restart at beginning of obd_exports_timed list
4149 Severity : enhancement
4151 Description: improve by-nid export eviction on the MDS and OST
4152 Details : allow multiple exports with the same NID to be evicted at one
4153 time without re-searching the exports list.
4156 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4158 Description: MDS may oops in groups_free()
4159 Details : in rare race conditions a newly allocated group_info struct is
4160 freed again, and this can be NULL. The 2.4 compatibility code
4161 for groups_free() checked for a NULL pointer, but 2.6 did not.
4164 Frequency : common for liblustre clients doing little filesystem IO
4165 Bugzilla : 9352, 7313
4166 Description: server may evict liblustre clients accessing contended locks
4167 Details : if a client is granted a lock or receives a completion AST
4168 with a blocking AST already set it would not reply to the AST
4169 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4172 Frequency : lfs setstripe, only systems with more than 160 OSTs
4174 Description: unable to set striping with a starting offset beyond OST 160
4175 Details : llapi_create_file() incorrectly limited the starting stripe
4176 index to the maximum single-file stripe count.
4179 Frequency : LDAP users only
4181 Description: lconf did not handle in-kernel recovery with LDAP properly
4182 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4184 Severity : enhancement
4186 Description: bind OST threads to NUMA nodes to improve performance
4187 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4188 node and do their allocations there to localize memory access
4190 Severity : enhancement
4192 Description: llmount can determine client NID directly from Myrinet (GM)
4193 Details : the client NID code from gmnalnid was moved directly into
4194 llmount, removing the need to use this or specifying the
4195 client NID explicitly when mounting GM clients with zeroconf
4198 Frequency : if client is started with down MDS
4200 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4201 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4202 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4203 ptlrpc_import_delayed_req() because we didn't check if the
4204 request was interrupted, and we also didn't break out of the
4205 event loop if there was no timeout
4210 Description: data loss during non-page-aligned writes to a single file from
4211 both multiple nodes and multiple threads on one node at same time
4212 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4213 inconsistency led to false short-reads, that were cached and later
4214 used by ->prepare_write() to fill in partially written page,
4215 leading to data loss.
4218 Frequency : always, if lconf --abort_recovery used
4220 Description: lconf --abort_recovery fails with 'Operation not supported'
4221 Details : lconf was attempting to abort recovery on the MDT device and not
4224 Severity : enhancement
4226 Description: remove cleanup logs
4227 Details : replace lconf-generated cleanup logs with lustre internal
4228 cleanup routines. Eliminates the need for client-cleanup and
4231 Severity : enhancement
4233 Description: add support for EAs (user and system) on lustre filesystems
4234 Details : it is now possible to store extended attributes in the Lustre
4235 client filesystem, and with the user_xattr mount option it
4236 is possible to allow users to store EAs on their files also
4238 Severity : enhancement
4240 Description: Add possibility (config option) to show minimal available OST free
4242 Details : When compiled with --enable-mindf configure option, statfs(2)
4243 (and so, df) will return least minimal free space available from
4244 all OSTs as amount of free space on FS, instead of summary of
4245 free spaces of all OSTs.
4247 Severity : enhancement
4249 Description: do not expand extent locks acquired on OST-side
4250 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4251 by server: they are not cached anyway.
4254 Frequency : when mmap is used/binaries executed from Lustre
4256 Description: Unmmap pages before throwing them away from read cache.
4257 Details : llap_shrink cache now attempts to unmap pages before discarding
4258 them (if unmapping failed - do not discard). SLES9 kernel has
4259 extra checks that trigger if this unmapping is not done first.
4264 Description: lconf didn't resolve symlinks before checking to see whether a
4265 given mountpoint was already in use
4268 Frequency : when migrating failover services
4269 Bugzilla : 6395, 9514
4270 Description: When migrating a subset of services from a node (e.g. failback
4271 from a failover service node) the remaining services would
4272 time out and evict clients.
4273 Details : lconf --force (implied by --failover) sets the global obd_timeout
4274 to 5 seconds in order to quickly disconnect, but this caused
4275 other RPCs to time out too quickly. Do not change the global
4276 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4278 Severity : enhancement
4279 Frequency : if MDS is started with down OST
4280 Bugzilla : 9439,5706
4281 Description: Allow startup/shutdown of an MDS without depending on the
4282 availability of the OSTs.
4283 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4284 Add appropriate locking and lov-osc refcounts for safe
4285 cleaning. Add osc abort_inflight calls in case the
4286 synchronize never started.
4289 Frequency : occasional (Cray XT3 only)
4291 Description: root not authorized to access files in CRAY_PORTALS environment
4292 Details : The client process capabilities were not honoured on the MDS in
4293 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4294 been accessed by an authorized user then root was able to access
4295 the file on the local client also. The root user capabilities
4296 are now allowed on the MDS, as this environment has secure UID.
4299 Frequency : occasional
4301 Description: ldiskfs "too long searching" message happens too often
4302 Details : A debugging message (otherwise harmless) prints too often on
4303 the OST console. This has been reduced to only happen when
4304 there are fragmentation problems on the filesystem.
4309 Description: Division by zero in statfs when all OSCs are inactive
4310 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4311 lov_statfs divides by value returned by lov_get_stripecnt().
4315 Bugzilla : 9489, 3273
4316 Description: First write from each client to each OST was only 4kB in size,
4317 to initialize client writeback cache, which caused sub-optimal
4318 RPCs and poor layout on disk for the first writen file.
4319 Details : Clients now request an initial cache grant at (re)connect time
4320 and so that they can start streaming writes to the cache right
4321 away and always do full-sized RPCs if there is enough data.
4322 If the OST is rebooted the client also re-establishes its grant
4323 so that client cached writes will be honoured under the grant.
4328 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4329 Details : Now I/O RPCs go to different portal number and (presumably) fast
4330 lock requests (and glimses) and other RPCs get their own service
4331 threads pool that should be able to service those RPCs
4334 Severity : enhancement
4336 Description: Ability to exchange lustre version between client and servers and
4337 issue warnings at client side if client is too old. Also for
4338 liblustre clients there is ability to refuse connection of too old
4340 Details : New 'version' field is added to connect data structure that is
4341 filled with version info. That info is later checked by server and
4345 Frequency : rare, liblustre only.
4346 Bugzilla : 9296, 9581
4347 Description: Two simultaneous writes from liblustre at offset within same page
4348 might proceed at the same time overwriting eachother with stale
4350 Details : I/O lock withing llu_file_prwv was released too early, before data
4351 actually was hitting the wire. Extended lock-holding time until
4352 server acknowledges receiving data.
4355 Frequency : extremely rare. Never observed in practice.
4357 Description: avoid generating lustre_handle cookie of 0.
4358 Details : class_handle_hash() generates handle cookies by incrementing
4359 global counter, and can hit 0 occasionaly (this is unlikely, but
4360 not impossible, because initial value of cookie counter is
4361 selected randonly). Value of 0 is used as a sentinel meaning
4362 "unassigned handle" --- avoid it. Also coalesce two critical
4363 sections in this function into one.
4365 Severity : enhancement
4367 Description: allow liblustre clients to delegate truncate locking to OST
4368 Details : To avoid overhead of locking, liblustre client instructs OST to
4369 take extent lock in ost_punch() on client's behalf. New connection
4370 flag is added to handle backward compatibility.
4372 Severity : enhancement
4373 Bugzilla : 4928, 7341, 9758
4374 Description: allow number of OST service threads to be specified
4375 Details : a module parameter allows the number of OST service threads
4376 to be specified via "options ost ost_num_threads={N}" in the
4377 OSS's /etc/modules.conf or /etc/modprobe.conf.
4381 Bugzilla : 6146, 9635, 9895
4382 Description: servers crash with bad pointer in target_handle_connect()
4383 Details : In rare cases when a client is reconnecting it was possible that
4384 the connection request was the last reference for that export.
4385 We would temporarily drop the export reference and get a new
4386 one, but this may have been the last reference and the export
4387 was just destroyed. Get new reference before dropping old one.
4389 Severity : enhancement
4390 Frequency : if client is started with failover MDS
4392 Description: Allow multiple MDS hostnames in the mount command
4393 Details : Try to read the configuration from all specified MDS
4394 hostnames during a client mount in case the "primary"
4397 Severity : enhancement
4399 Description: Stop sending data to evicted clients as soon as possible.
4400 Details : Check if the client we are about to send or are sending data to
4401 was evicted already. (Check is done every second of waiting,
4402 for which l_wait_event interface was extended to allow checking
4403 of exit condition at specified intervals).
4406 Frequency : rare, normally only when NFS exporting is done from client
4408 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4410 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4413 Frequency : always, if srand() is called before liblustre initialization
4415 Description: Liblustre uses system PRNG disturbing its usage by user application
4416 Details : Introduce internal to lustre fast and high-quality PRNG for
4417 lustre usage and make liblustre and some other places in generic
4418 lustre code to use it.
4420 Severity : enhancement
4421 Bugzilla : 9477, 9557, 9870
4422 Description: Verify that the MDS configuration logs are updated when xml is
4423 Details : Check if the .xml configuration logs are newer than the config
4424 logs stored on the MDS and report an error if this is the case.
4425 Request --write-conf, or allow starting with --old_conf.
4427 Severity : enhancement
4429 Description: Handle symlinks in the path when checking if Lustre is mounted.
4430 Details : Resolve intermediate symlinks when checking if a client has
4431 mounted a filesystem to avoid duplicate client mounts.
4436 Description: lconf can hit an error exception but still return success.
4437 Details : The lconf command catches the Command error exception at the top
4438 level script context and will exit with the associated exit
4439 status, but doesn't ensure that this exit status is non-zero.
4444 Description: failure of ptlrpc thread startup can cause oops
4445 Details : Starting a ptlrpc service thread can fail if there are a large
4446 number of threads or the server memory is very fragmented.
4447 Handle this without oopsing.
4450 Frequency : always, only if liblustre and non-default acceptor port was used
4452 Description: liblustre cannot connect to servers with non-default acceptor port
4453 Details : tcpnal_set_default_params() was not called and was therefore
4454 ignoring the environment varaible TCPNAL_PORT, as well as other
4455 TCPNAL_ environment variables
4460 Description: two objects could be created on the same OST for a single file
4461 Details : If an OST is down, in some cases it was possible to create two
4462 objects on a single OST for a single file. No problems other
4463 than potential performance impact and spurious error messages.
4467 Bugzilla : 5681, 9562
4468 Description: Client may oops in ll_unhash_aliases
4469 Details : Client dcache may become inconsistent in race condition.
4470 In some cases "getcwd" can fail if the current directory is
4476 Description: Inode refcounting problems in NFS export code
4477 Details : link_raw functions used to call d_instantiate without obtaining
4478 extra inode reference first.
4482 Bugzilla : 9942, 9903
4483 Description: Referencing freed requests leading to crash, memleaks with NFS.
4484 Details : We used to require that call to ll_revalidate_it was always
4485 followed by ll_lookup_it. Also with revalidate_special() it is
4486 possible to call ll_revalidate_it() twice for the same dentry
4487 even if first occurence returned success. This fix changes semantic
4488 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4489 reference on a request referred from the intent.
4490 ll_intent_release() then releases such a request.
4493 Frequency : rare, normally benchmark loads only
4495 Description: unlinked inodes were kept in memory on the client
4496 Details : If a client is repeatedly creating and unlinking files it
4497 can accumulate a lot of stale inodes in the inode slab cache.
4498 If there is no other client load running this can cause the
4499 client node to run out of memory. Instead flush old inodes
4500 from client cache that have the same inode number as a new inode.
4503 Frequency : SLES9 2.6.5 kernel and long filenames only
4504 Bugzilla : 9969, 10379
4505 Description: utime reports stale NFS file handle
4506 Details : SLES9 uses out-of-dentry names in some cases, which confused
4507 the lustre dentry revalidation. Change it to always use the
4511 Frequency : rare, unless heavy write-truncate concurrency is continuous
4512 Bugzilla : 4180, 6984, 7171, 9963, 9331
4513 Description: OST becomes very slow and/or deadlocked during object unlink
4514 Details : filter_destroy() was holding onto the parent directory lock
4515 while truncating+unlinking objects. For very large objects this
4516 may block other threads for a long time and slow overall OST
4517 responsiveness. It may also be possible to get a lock ordering
4518 deadlock in this case, or run out of journal credits because of
4519 the combined truncate+unlink. Solution is to do object truncate
4520 first in one transaction without parent lock, and then do the
4521 final unlink in a new transaction with the parent lock. This
4522 reduces the lock hold time dramatically.
4525 Frequency : rare, 2.4 kernels only
4527 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4528 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4529 before unmounting their filesystem in order to avoid deadlock.
4530 The kernel_locked() function in 2.4 kernels only checks whether
4531 the kernel lock is held, not whether it is this process that is
4532 holding it as 2.6 kernels do.
4537 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4538 Details : The client ptlrpc code may be trying to reconnect to a down
4539 server before a previous connection attempt has timed out.
4540 Increase the reconnect interval to be longer than the connection
4541 timeout interval to avoid sending duplicate connections to
4545 Frequency : echo_client brw_test command
4547 Description: fix echo_client to work with OST preallocated code
4548 Details : OST preallocation code (5137) didn't take echo_client IO path
4549 into account: echo_client calls filter methods outside of any
4550 OST thread and, hence, there is no per-thread preallocated
4551 pages and buffers to use. Solution: hijack pga pages for IO. As
4552 a byproduct, this avoids unnecessary data copying.
4556 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4557 Description: Client can oops in mdc_commit_close() after open replay
4558 Details : It was possible for the MDS to return an open request with no
4559 transaction number in mds_finish_transno() if the client was
4560 evicted, but without actually returning an error. Clients
4561 would later try to replay that open and may trip an assertion
4562 Simplify the client close codepath, and always return an error
4563 from the MDS in case the open is not successful.
4566 Frequency : rare, 2.6 OSTs only
4568 Description: OST may deadlock under high load on fragmented files
4569 Details : If there was a heavy load and highly-fragmented OST filesystems
4570 it was possible to have all the OST threads deadlock waiting on
4571 allocation of biovecs, because the biovecs were not released
4572 until the entire RPC IO was completed. Instead, release biovecs
4573 as soon as they are complete to ensure forward IO progress.
4575 Severity : enhancement
4577 Description: Support for specifying external journal device at mount
4578 Details : If an OST or MDS device is formatted with an external journal
4579 device, this device major/minor is stored in the ext3 superblock
4580 and may not be valid for failover. Allow detecting and
4581 specifying the external journal at mount time.
4586 Description: Mounting an MDS with pending unlinked files may cause oops
4587 Details : target_finish_recovery() calls mds_postrecov() which returned
4588 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4589 considers this an error and immediately begins cleaning up the
4590 lov, just after starting the mds_lov process
4592 Severity : enhancement
4594 Description: Implement 'lfs df' to report actual free space on per-OST basis
4595 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4596 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4597 usage of objects; '-h' to report in human readable format.
4599 ------------------------------------------------------------------------------
4601 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4608 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4609 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4610 at startup during orphan unlinking. Instead, skip buddy update
4611 before setup, as it will be regenerated after recovery is complete.
4616 Description: If an OST is inactive, its locks might reference stale inodes.
4617 Details : lov_change_cbdata() must iterate over all namespaces, even if
4618 they are inactive to clear inode references from the lock.
4620 Severity : enhancement
4621 Frequency : occasional, if non-standard max_dirty_mb used
4623 Description: Client will block write RPCs if not enough grant
4624 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4625 then the client will block writes while waiting for another RPC
4626 to complete instead of consuming its dirty limit. With change
4627 we get improved performance when max_dirty_mb is small.
4629 Severity : enhancement
4630 Bugzilla : 3389, 6253
4631 Description: Add support for supplementary groups on the MDS.
4632 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4633 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4634 lookups for user supplementary groups into a cache.
4638 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4639 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4640 for liblustre O_EXCL is enforced by MDS.
4644 Description: readdir never returns NULL in liblustre.
4645 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4646 dentry in struct dirent.
4650 Frequency : liblustre only, depends on application IO pattern
4651 Description: liblustre clients evicted if not contacting servers
4652 Details : Don't put liblustre clients into the ping_evictor list, so
4653 they will not be evicted by the pinger ever.
4655 Severity : enhancement
4657 Description: Add ability to evict clients by NID from MDS.
4658 Details : By echoing "nid:$NID" string into
4659 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4660 $NID would be instantly evicted from this MDS and from all active
4661 OSTs connected to it.
4665 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4666 Details : lookup_it_finish() used to query file size from OSTs that was not
4671 Description: service threads change working directory to that of init
4672 Details : Starting lustre service threads may pin the working directory
4673 of the parent thread, making that filesystem busy. Threads
4674 now change to the working directory of init to avoid this.
4678 Frequency : during shutdown only
4679 Description: shutdown with a failed MDS or OST can cause unmount to hang
4680 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4685 Frequency : 2.6 only
4686 Description: chmod/chown may include an extra supplementary group
4687 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4688 supplementary group and if none is specified this is used.
4691 Bugzilla : 5479 (6816)
4693 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4694 Details : If lookup is in progress on a file that is unlinked we might try
4695 to revalidate the inode and fail in revalidate after lookup is
4696 complete and ll_file_open() enqueues the open again but
4697 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4700 Frequency : always, if lconf --abort_recovery used
4702 Description: lconf --abort_recovery fails with 'Operation not supported'
4703 Details : lconf was attempting to abort recovery on the MDT device and not
4706 ------------------------------------------------------------------------------
4708 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4713 Frequency : rare (only unsupported configurations with a node running as an
4715 Bugzilla : 6514, 5137
4716 Description: Mounting a Lustre file system on a node running as an OST could
4718 Details : OSTs now preallocates memory needed to write out data at
4719 startup, instead of when needed, to avoid having to
4720 allocate memory in possibly low memory situations.
4721 Specifically, if the file system is mounted on on OST,
4722 memory pressure could force it to try to write out data,
4723 which it needed to allocate memory to do. Due to the low
4724 memory, it would be unable to do so and the node would
4725 become unresponsive.
4727 Severity : enhancement
4729 Description: Addition of lconf --service command line option
4730 Details : lconf now accepts a '--service <arg>' option, which is
4731 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4733 Severity : enhancement
4735 Description: Failover mode is now the default for OSTs.
4736 Details : By default, OSTs will now run in failover mode. To return to
4737 the old behaviour, add '--failout' to the lmc line for OSTs.
4739 Severity : enhancement
4741 Description: Health checks are now provided for MDS and OSTs
4742 Details : Additional detailed health check information on MSD and OSTs
4743 is now provided through the procfs health_check value.
4746 Frequency : occasional, depends on IO load
4748 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4749 after numerous create/delete cycles
4750 Details : The ext3 inode allocation policy would not allocate new inodes
4751 very well on the OSTs because there are no new directories
4752 being created. Instead we look for groups with free space if
4753 the parent directories are nearly full.
4758 Description: Network or server problems during mount may cause partially
4759 mounted clients instead of returning an error.
4760 Details : The config llog parsing code may overwrite the error return
4761 code during mount error handling, returning success instead
4767 Description: MDS can fail to allocate large reply buffers
4768 Details : After long uptimes the MDS can fail to allocate large reply
4769 buffers (e.g. zconf client mount config records) due to memory
4770 fragmentation or consumption by the buffer cache. Preallocate
4771 some large reply buffers so that these replies can be sent even
4772 under memory pressure.
4776 Frequency : rare (liblustre)
4777 Description: fsx running with liblustre complained that using truncate() to
4778 extend the file doesn't work. This patch corrects that issue.
4779 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4780 ATTR_SIZE and lsm use in llu_setattr_raw.
4784 Frequency : rare, only 2.6 kernels
4785 Description: Unusual file access patterns on the MDS may result in inode
4786 data being lost in very rare circumstances.
4787 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4788 code in the 2.6 kernel caused the inode and/or EA data not to
4789 be read from disk, causing single-file corruption.
4793 Frequency : rare, only 2.6 filesystems using extents
4794 Description: Heavy concurrent write and delete load may cause data corruption.
4795 Details : It was possible under high-load situations to have an extent
4796 metadata block in the block device cache from a just-unlinked
4797 file overwrite a newly-allocated data block. We now unmap any
4798 metadata buffers that alias just-allocated data blocks.
4802 Frequency : filesystems with default stripe_count larger than 77
4803 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4804 Details : lconf specifies an inode size of 4096 bytes when the default
4805 stripe_count is larger than 77. This conflicts with the default
4806 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4807 this case to avoid pinning too much memory for large EAs.
4809 ------------------------------------------------------------------------------
4811 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4816 Frequency : rare (extremely heavy IO load with hundreds of clients)
4818 Description: Client is evicted, gets IO error writing to file
4819 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4820 caused clients to be evicted for AST timeouts. The fixes in
4821 bug 5192 mean we no longer need to have such short AST timeouts
4822 so ldlm_timeout has been increased.
4825 Frequency : occasional during --force or --failover shutdown under load
4826 Bugzilla : 5949, 4834
4827 Description: Server oops/LBUG if stopped with --force or --failover under load
4828 Details : a collection of import/export refcount and cleanup ordering
4829 issues fixed for safer force cleanup
4832 Frequency : only filesystems larger than 120 OSTs
4833 Bugzilla : 5990, 6223
4834 Description: lfs getstripe would oops on a very large filesystem
4835 Details : lov_getconfig used kfree on vmalloc'd memory
4838 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4839 Bugzilla : 6242, 6243
4840 Description: reading from files that had been truncated to a non-zero size
4841 but never opened returned no data
4842 Details : ll_file_read() reads zeros from no-object files to EOF
4847 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4848 Details : There's little harm in aborting MDS/OSS recovery and letting it
4849 try again, so I removed the LASSERT and return an error instead.
4851 Severity : enhancement
4853 Description: New debugging infrastructure for tracking down data corruption
4854 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4855 (b) cover more of the client-side code path, and (c) try to narrow
4856 down where problems occurred
4860 Bugzilla : 3819, 4364, 4397, 6313
4861 Description: Racing close and eviction MDS could cause assertion in mds_close
4862 Details : It was possible to get multiple mfd references during close and
4863 client eviction, leading to one thread referencing a freed mfd.
4865 Severity: : enhancement
4866 Bugzilla : 3262, 6359
4867 Description: Attempts to reconnect to servers are now more aggressive.
4868 Details : This builds on the enhanced upcall-less recovery that was added
4869 in 1.4.2. When trying to reconnect to servers, clients will
4870 now try each server in the failover group every 10 seconds. By
4871 default, clients would previously try one server every 25 seconds.
4876 Description: After recovery, certain operations trigger a failed
4877 assertion on a client.
4878 Details : Failing over an mds, using lconf -d --failover, while a
4879 client was doing a readdir() call would cause the client to
4880 LBUG after recovery completed and the readdir() was resent.
4882 Severity : enhancement
4884 Description: Default groups are now added by lconf
4885 Details : You can now run lconf --group <servicename> without having to
4886 manually add groups with lmc.
4889 Frequency : occasional
4891 Description: Nodes with an elan id of 0 trigger a failed assertion
4894 Frequency : always when accessing e.g. tty/console device nodes
4896 Description: tty and some other devices nodes cannot be used on lustre
4897 Details : file's private_data field is used by device data and lustre
4898 values in there got lost. New field was added to struct file to
4899 store fs-specific private data.
4902 Frequency : when exporting Lustre via NFS
4904 Description: NFSD failed occasionally when looking up a path component
4905 Details : NFSD is looking up ".." which was broken in ext3 directories
4906 that had grown large enough to become hashed.
4909 Frequency : Clusters with multiple interfaces not on the same subnet
4911 Description: Nodes will repeatedly try to reconnect to an interface which it
4912 cannot reach and report an error to the log.
4913 Details : Extra peer list entries will be created by lconf with some peers
4914 unreachable. lconf now validates the peer before adding it.
4917 Frequency : Only if a default stripe is set on the filesystem root.
4919 Description: Setting a default stripe on the filesystem root prevented the
4920 filesystem from being remounted.
4921 Details : The client was sending extra request flags in the root getattr
4922 request and did not allocate a reply buffer for the dir EA.
4925 Frequency : occasional, higher if lots of files are accessed by one client
4926 Bugzilla : 6159, 6097
4927 Description: Client trips assertion regarding lsm mismatch/magic
4928 Details : While revalidating inodes the VFS looks up inodes with ifind()
4929 and in rare cases can find an inode that is being freed.
4930 The ll_test_inode() code will free the lsm during ifind()
4931 when it finds an existing inode and then the VFS later attaches
4932 this free lsm to a new inode.
4936 Bugzilla : 6422, 7030
4937 Description: MDS deadlock between mkdir and client eviction
4938 Details : Creating a new file via mkdir or mknod (starting a transaction
4939 and getting the ns lock) can deadlock with client eviction
4940 (gets ns lock and trying to finish a synchronous transaction).
4943 Frequency : occasional
4944 Description: While starting a server, the fsfilt_ext3 module could not be
4946 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4947 kernels. Previously, lconf would still use the ext3 name
4948 when trying to load modules. Now, it will correctly use
4949 ext3 on 2.4 and ldiskfs on 2.6.
4951 Severity : enhancement
4952 Description: The default stripe count has been changed to 1
4953 Details : The interpretation of the default stripe count (0, to lfs
4954 or lmc) has been changed to mean striping across a single
4955 OST, rather than all available. For general usage we have
4956 found a stripe count of 1 or 2 works best.
4958 Severity : enhancement
4959 Description: Add support for compiling against Cray portals.
4960 Details : Conditional compiling for some areas that are different
4964 Frequency : occasional
4965 Bugzilla : 6409, 6834
4966 Description: Creating files with an explicit stripe count may lead to
4967 a failed assertion on the MDS
4968 Details : If some OSTs are full or unavailable, creating files may
4969 trigger a failed assertion on the MDS. Now, Lustre will
4970 try to use other servers or return an error to the
4974 Frequency : occasional
4976 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4977 caused lock timeout during readahead (since 1.4.2).
4978 Details : Processes doing readahead might match a lock that hasn't been
4979 granted yet if there are overlapping and conflicting lock
4980 requests. The readahead process waits on ungranted lock
4981 (original lock is CBPENDING), while OST waits for that process
4982 to cancel CBPENDING read lock and eventually evicts client.
4984 Severity : enhancement
4986 Description: Initial enabling of flock support for clients
4987 Details : Implements fcntl advisory locking and file status functions.
4988 This feature is provided as an optional mount flag (default
4989 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4990 locking are implemented yet, and those that are are not guaranteed
4991 to be completely correct in production environments.
4992 mount -t lustre -o [flock|noflock] ...
4995 Frequency : occasional
4997 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4998 assertion in the ext3 JBD (journaling) layer.
4999 Details : The b_committed_data struct is protected by the big kernel lock
5000 in 2.4 kernels, serializing journal_commit_transaction() and
5001 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5002 access to this struct. In 2.6 kernels there is finer grained
5003 locking to improve SMP performance of the JBD layer.
5007 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5012 Description: merge and cleanup kernel patches.
5015 -----------------------------------------------------------------------------
5017 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5019 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5020 versions of Lustre. Please update all servers and clients to
5021 version 1.4.2 or later at the same time. You must also run
5022 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5023 to update the configuration logs.
5025 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5026 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5027 - swab llog records of type '0' so we get proper header size/idx (5861)
5028 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5029 - fix rename of one directory over another leaking an inode (5953)
5030 - avoid SetPageDirty on 2.6 (5981)
5031 - don't re-add just-being-destroyed locks to the waiting list (5653)
5032 - when creating new directories, inherit the parent's custom
5033 striping settings if present parent (3048)
5034 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5035 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5036 - don't hold client locks on temporary worklist from l_lru (5666)
5037 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5038 - automatically evict dead clients (5921)
5039 - Update file size properly in create+truncate+fstat case (6196)
5040 - Do not unhash mountpoint dentries, do not allow removal of
5042 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5043 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5044 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5045 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5046 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5047 - Correctly handle reads of files with no objects (6243)
5048 - lctl recover will also mark a device active if deactivate used (5933)
5050 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5051 - allow --write-conf on an MDS with different nettype than client (5619)
5052 - don't write config llogs to MDS for mounts not from that MDS (5617)
5053 - lconf should create multiple TCP connections from a client (5201)
5054 - init scripts are now turned off by default; run chkconfig --on
5055 lustre and chkconfig --on lustrefs to use them
5056 - upcalls are no longer needed for clients to recover to failover
5058 - add --abort-recovery option to lconf to abort recovery on device
5060 - add support for an arbitrary number of OSTs (3026)
5061 - Quota support protocol changes.
5062 - forward compatibility changes to wire structs (6007)
5063 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5064 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5065 - improved "lustre status" script
5066 - initialize blocksize for non-regular files (6062)
5067 - added --disable-server and --disable-client configure options (5782)
5068 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5069 - Vanilla 2.4.29 support
5070 - increase maximum number of obd devices to 520 (6242)
5071 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5072 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5073 SLES 9 2.6.5 kernels
5074 - limit stripes per file to 160 (the maximum EA size) (6093)
5076 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5079 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5080 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5081 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5082 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5083 - small ext3 extents cleanups and fixes (5733)
5084 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5085 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5086 - avoid needless client->OST connect, fix handle mismatch (5317)
5087 - fix DLM error path that led to out-of-sync client, long delays (5779)
5088 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5089 - fix several locking issues related to i_size (5492,5624,5654,5672)
5090 - don't move pending lock onto export if it is already evicted (5683)
5091 - fix kernel oops when creating .foo in unlinked directory (5548)
5092 - fix deadlock in obdfilter statistics vs. object create (5811)
5093 - use time_{before,after} to avoid timer jiffies wrap (5882)
5094 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5095 - Do not leak request if server was not able to process it (5154)
5096 - If mds_open unable to find parent dir, make that negative lookup(5154)
5097 - don't create new directories with extent-mapping (5909, 5936)
5099 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5100 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5101 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5102 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5103 - support for RHEL4 kernel on i686 (5773)
5104 - provide error messages when incompatible logs are encountered (5898)
5106 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5107 * version 1.4.0.10 (1.4.1 release candidate 1)
5109 - don't keep a lock reference when lock is not granted (4238)
5110 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5111 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5112 - drop import inflight refcount on signal_completed_replay error (5255)
5113 - unlock page after async write error during send (3677)
5114 - handle missing objects in filter_preprw_read properly (5265)
5115 - no transno return for symlink open, don't save no-trasno open (3440)
5116 - don't try to complete elan receive that already failed (4012)
5117 - free RPC server reply state on error (5406)
5118 - clean up thread from ptlrpc_start_thread() on error (5160)
5119 - readahead could read extra page into cache that wasn't ejected (5388)
5120 - prevent races in class_attach/setup/cleanup/detach (5260)
5121 - don't dereference de->d_inode after l_dput of de (5458)
5122 - use "int" for stripe value returned from lock_to_stripe (5544)
5123 - mballoc allocation and error-checking fixes in 2.6 (5504)
5124 - block device patches to fix I/O request sizes in 2.6 (5482)
5125 - look up hostnames for IB nals (5602)
5126 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5127 - don't start multiple acceptors for the same port (5277)
5128 - fix incorrect LASSERT in mds_getattr_name (5635)
5129 - export a proc file for general "ping" checking (5628)
5130 - fix "lfs check" to not block when the MDS is down (5628)
5132 - service request history (4965)
5133 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5134 - create an "evict_client" /proc entry on OSTs, like the MDS has
5135 - fix mount usage message, return errors per mount(8) (5168)
5136 - change grep [] to grep "[]" in tests so they work in more UMLs
5137 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5138 - remove ancient LOV_MAGIC_V0 EA support (5047)
5139 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5140 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5141 - allow readahead allocations to fail when low on memory (5383)
5142 - mmap locking landed again, after considerable improvement (2828)
5143 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5145 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5148 - send OST transaction number in read/write reply to free req (4966)
5149 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5150 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5151 - the watchdog thread now runs as interruptible (5246)
5152 - flock/lockf fixes (but it's still disabled, pending 5135)
5153 - don't use EXT3 constants in llite code (5094)
5154 - memory shortage at startup could cause assertion (5176)
5156 - reorganization of lov code
5157 - single portals codebase
5159 - add extents/mballoc support (5025)
5160 - direct I/O reads in the obdfilter (4048)
5161 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5163 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5166 - send OST transaction number in read/write reply to free req (4966)
5167 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5168 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5169 - the watchdog thread now runs as interruptible (5246)
5170 - handle missing objects in filter_preprw_read properly (5265)
5171 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5172 - drop import inflight refcount on signal_completed_replay error (5255)
5173 - unlock page after async write error during send (3677)
5174 - return original error code on reconstructed replies (3761)
5175 - no transno return for symlink open, don't save no-trasno open (3440)
5177 - add pid to ldlm debugging output (4922)
5178 - bump the watchdog timeouts -- we can't handle 30sec yet
5179 - extra debugging for orphan dentry/inode bug (5259)
5181 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5184 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5185 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5186 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5187 - fix NULL dereference in /proc/sys/portals/routes (4827)
5188 - allow failed mdc_close() operations to be interrupted (4561)
5189 - stop precreate on OST before MDS would time out on it (4778)
5190 - don't send partial-page writes before EOF from client (4410)
5191 - discard client grant for sub-page writes on large-page clients (4520)
5192 - don't free dentries not owned by NFS code, check generation (4806)
5193 - fix lsm leak if mds_create_objects() fails (4801)
5194 - limit debug_daemon file size, always print CERROR messages (4789)
5195 - use transno after validating reply (3892)
5196 - process timed out requests if import state changes (3754)
5197 - update mtime on OST during writes, return in glimpse (4829)
5198 - add mkfsoptions to LDAP (4679)
5199 - use ->max_readahead method instead of zapping global ra (5039)
5200 - don't interrupt __l_wait_event() during strace
5202 - add software watchdogs to catch hung threads quickly (4941)
5203 - make lustrefs init script start after nfs is mounted
5204 - fix CWARN/ERROR duplication (4930)
5205 - return async write errors to application if possible (2248)
5206 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5207 - print NAL number in %x format (4645)
5208 - update barely-supported suse-2.4.21-171 series (4842)
5209 - support for sles 9 %post scripts
5210 - support for building 2.6 kernel-source packages
5211 - support for sles km_* packages
5213 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5216 - ignore -ENOENT errors in osc_destroy (3639)
5217 - notify osc create thread that OSC is being cleaned up (4600)
5218 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5219 - reconstruct ost_handle() like mds_handle() (4657)
5220 - create a new thread to do import eviction to avoid deadlock (3969)
5221 - let lconf resolve symlinked-to devices (4629)
5222 - don't unlink "objects" from directory with default EA (4554)
5223 - hold socknal file ref over connect in case target is down (4394)
5224 - allow more than 32000 subdirectories in a single directory (3244)
5225 - fix blocks count for O_DIRECT writes (3751)
5226 - OST returns ENOSPC from object create when no space left (4539)
5227 - don't send truncate RPC if file size isn't changing (4410)
5228 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5229 - bind to privileged port in socknal and tcpnal (3689)
5231 - rate limit CERROR/CWARN console message to avoid overload (4519)
5232 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5233 - basic mmap support (3918)
5234 - kernel patch series update from b1_4 (4711)
5236 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5239 - avoid crash during MDS cleanup with OST shut down (2775)
5240 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5241 - don't use bad inodes on the MDS (3744)
5242 - dynamic object preallocation to improve recovery speed (4236)
5243 - don't hold spinlock over lock dumping or change debug flags (4401)
5244 - don't zero obd_dev when it is force cleaned (3651)
5245 - print grants to console if they go negative (4431)
5246 - "lctl deactivate" will stop automatic recovery attempts (3406)
5247 - look for existing locks in ldlm_handle_enqueue() (3764)
5248 - don't resolve lock handle twice in recovery avoiding race (4401)
5249 - revalidate should check working dir is a directory (4134)
5251 - don't always mark "slow" obdfilter messages as errors (4418)
5253 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5256 - don't close LustreDB during write_conf until it is done (3860)
5257 - fix typo in lconf for_each_profile (3821)
5258 - allow dumping logs from multiple threads at one time (3820)
5259 - don't allow multiple threads in OSC recovery (3812)
5260 - fix debug_size parameters (3864)
5261 - fix mds_postrecov to initialize import for llog ctxt (3121)
5262 - replace config semaphore with spinlock (3306)
5263 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5264 - don't allow enqueue to complete on a destroyed export (3822)
5265 - down write_lock before checking llog header bitmap (3825)
5266 - recover from lock replay timeout (3764)
5267 - up llog sem before sending rpc (3652)
5268 - reduce ns lock hold times when setting kms (3267)
5269 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5270 - fix NULL deref and obd_dev leak on setup error (3312)
5271 - replace some LBUG about llog ops with error handling (3841)
5272 - don't match INVALID dentries from d_lookup and spin (3784)
5273 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5274 - fix invalid assertion in ptlrpc_set_wait (3880)
5276 - add libwrap support for the TCP acceptor (3996)
5277 - add /proc/sys/portals/routes for non-root route listing (3994)
5278 - allow setting MDS UUID in .xml (2580)
5279 - print the stack of a process that LBUGs (4228)
5281 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5284 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5285 - make sure to unset replay flag from failed open requests (3440)
5286 - if default stripe count is 0, use OST count for inode size (3636)
5287 - update parent mtime/ctime on client for create/unlink (2611)
5288 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5289 - free recovery state on server during a forced cleanup (3571)
5290 - unregister_reply for resent reqs (3063)
5291 - loop back devices mounting and status check on 2.6 (3563)
5292 - fix resource-creation race that can provoke i_size == 0 (3513)
5293 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5294 - more debugging for page-accounting assertion (3746)
5295 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5296 - avoid deadlock after precreation failure (3758)
5297 - fix race and lock order deadlock in orphan handling (3450, 3750)
5298 - add validity checks when grabbing inodes from l_ast_data (3599)
5300 - add /proc/.../recovery_status to obdfilter (3428)
5301 - lightweight CDEBUG infrastructure, debug daemon (3668)
5302 - change default OSC RPC parameters to be better on small clusters
5303 - turn off OST read cache for files smaller than 32MB
5304 - install man pages and include them in rpms (3100)
5305 - add new init script for (un)mounting lustre filesystems (2593)
5306 - run chkconfig in %post for init scripts (3701)
5307 - drop scimac NAL (unmaintained)
5309 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5312 - clean kiobufs before and after use (3485)
5313 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5314 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5315 - change init script to fail more clearly if not run as root (1528)
5316 - allow clients to reconnect during replay (1742)
5317 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5318 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5319 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5320 - fix cond_resched (3554)
5321 - only evict unfinished clients after recovery (3515)
5322 - allow bulk resend, prevent data loss (3570)
5323 - dynamic ptlrpc request buffer allocation (2102)
5324 - don't allow unlinking open directory if it isn't empty (2904)
5325 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5326 - remove extraneous obd dereference causing LASSERT failure (3334)
5327 - don't use get_cycles() when creating temp. files on the mds (3156)
5328 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5329 - handle EEXIST for set-stripe, set proper directory name (3336)
5331 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5332 - fix ksocknal_fmb_callback() error messages (2918)
5334 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5337 - don't copy lvb into (possibly NULL) reply on error (2983)
5338 - don't deref dentry after dput, don't free lvb on error (2922)
5339 - use the kms to determine writeback rpc length (2947)
5340 - increment oti_logcookies when osc is inactive (2948)
5341 - update client's i_blocks count via lvb messages (2543)
5342 - handle intent open/close of special files properly (1557)
5343 - mount MDS with errors=remount-ro, like obdfilter (2009)
5344 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5345 - don't use cancelling-locks' kms values (2947)
5346 - use highest lock extent for kms, not last one (2925)
5347 - don't dereference ERR_PTR() dentry in error handling path (3107)
5348 - fix thread race in portals_debug_dumplog() (3122)
5349 - create lprocfs device entries at setup instead of at attach (1519)
5350 - common AST error handler, don't evict client on completion race (3145)
5351 - zero nameidata in detach_mnt in 2.6 (3118)
5352 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5353 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5354 - initialize RPC timeout timer earlier for 2.6 (3219)
5355 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5356 - print nal/nid for unknown nid (3258)
5357 - additional checks for oscc recovery before doing precreate (3284)
5358 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5359 - don't crash in mdc_close for bad permissions on open (3285)
5360 - zero i_rdev for non-device files (3147)
5361 - clear page->private before handing to FS, better assertion (3119)
5362 - tune the read pipeline (3236)
5363 - fix incorrect decref of invalidated dentry (2350)
5364 - provide read-ahead stats and refine rpc in flight stats (3328)
5365 - don't hold journal transaction open across create RPC (3313)
5366 - update atime on MDS at close time (3265)
5367 - close LDAP connection when recovering to avoid server load (3315)
5368 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5369 - don't leak open file on MDS after open resend (3325)
5370 - serialize filter_precreate and filter_destroy_precreated (3329)
5371 - loop device shouldn't call sync_dev() for nul device (3092)
5372 - clear page cache after eviction (2766)
5373 - resynchronize MDS->OST in background (2824)
5374 - refuse to mount the same filesystem twice on same mountpoint (3394)
5375 - allow llmount to create routes for mounting behind routers (3320)
5376 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5377 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5378 - fix rare problem with rename on htree directories (3417)
5380 - allow default OST striping configuration per directory (1414)
5381 - fix compilation for qswnal for 2.6 kernels (3125)
5382 - increase maximum number of MDS request buffers for large systems
5383 - change liblustreapi to be useful for external progs like lfsck (3098)
5384 - increase local configuration timeout for slow disks (3353)
5385 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5387 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5390 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5391 - don't overwrite extent policy data in reply if lock was blocked (2901)
5392 - drop filter export grants atomically with removal from device (2663)
5393 - del obd_self_export from work_list in class_disconnect_exports (2908)
5394 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5395 - swab reply message in mdc_close, other PPC fixes (2464)
5396 - fix destroying of named logs (2325)
5397 - overwrite old logs when running lconf --write_conf (2264)
5398 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5399 - fix race in target_handle_connect (2898)
5400 - mds_reint_create() should take same inode create lock (2926)
5401 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5402 - don't close files for self_export to avoid uninitialized obd (2936)
5403 - allow MDS with the same name as client node (2939)
5404 - hold dentry reference for closed log files for unlink (2325)
5405 - reserve space for all logs during transactions (2059)
5406 - don't evict page beyond end of stripe extent (2925)
5407 - don't oops on a deleted current working directory (2399)
5408 - handle hard links to targets without a parent properly (2517)
5409 - don't dereference NULL lock when racing during eviction (2867)
5410 - don't grow lock extents when lots of conflicting locks (2919)
5412 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5415 - account for cache space usage on clients to avoid data loss (974)
5416 - lfsck support in lustre kernel code (2349)
5417 - reduce journal credits needed for BRW writes (2370)
5418 - orphan handling to avoid losing space on client/server crashes
5419 - ptlrpcd can be blocked, stopping ALL progress (2477)
5420 - use lock value blocks to assist in proper KMS, faster stat (1021)
5421 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5422 - recovery for initial connections (2355)
5423 - fixes for mds_cleanup_orphans (1934)
5424 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5425 - block all file creations until orphan recovery completes (1901)
5426 - client remove rq_connection from request struct (2423)
5427 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5428 - recovery timer race (2670)
5429 - mdc_close recovey bug (2532)
5430 - ptlrpc cleanup bug (2710)
5431 - mds timeout on local locks (2588)
5432 - namespace lock held during RPCs (2431)
5433 - handle interrupted sync write properly (2503)
5434 - don't try to handle a message that hasn't been replied to (2699)
5435 - client assert failure during cleanup after abort recovery (2701)
5436 - leak mdc device after failed mount (2712)
5437 - ptlrpc_check_set allows timedout requests to complete (2714)
5438 - wait for inflight reqs when ptlrpcd finishes (2710)
5439 - make sure unregistered services are removed from the srv_list
5440 - reset bulk XID's when resending them (caught by 1138 test)
5441 - unregister_bulk after timeout
5442 - fix lconf error (2694)
5443 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5444 - readahead locks pages, leaves pending causing memory pressure (2673)
5445 - increase OST request buffers to 4096 on large machines (2729)
5446 - fix up permission of existing directories in simple_mkdir (2661)
5447 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5448 - don't assign transno to errored transactions (2742)
5449 - don't delete objects on OST if given a bogus objid from MDS (2751)
5450 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5451 - if rq_no_resend, then timeout request after recovery (2432)
5452 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5453 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5454 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5455 - MDS recovery shouldn't delete objects that it has given out (2730)
5456 - if enqueue arrives after completion, don't clobber LVB (2819)
5457 - don't unlock pages twice when trigger_group_io returns error (2814)
5458 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5459 - don't write pages to disk if there was an error (1450)
5460 - don't ping imports that have recovery disabled (2676)
5461 - take buffered bytes into account when balancing socknal conn (2817)
5462 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5463 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5464 - remove little-endian swabbing of llog records (1987)
5465 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5466 - retry reposting request buffers if they fail (1191)
5467 - grow extent at grant time to avoid granting a revoked lock (2809)
5468 - lock revoke doesn't evict page if covered by a second lock (2765)
5469 - disable VM readahead to avoid reading outside lock extents (2805)
5471 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5472 - updated kernel patches for hp-2.4.20 kernel (2681)
5474 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5477 - fix truncated write corruption (2366)
5478 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5480 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5481 - ensure len(uuid) < 37 in lmc (1171)
5482 - fix ia64 OOPS in llog_test (2255)
5483 - zero end of page at obdfilter for partial page writes (2648)
5484 - don't leave stale dentries around after renames (bug 2428)
5485 - fix timeouts when evicting a client with a single lock held (2642)
5486 - set deadline for the initial HELLO message to drain (2634)
5487 - print out dotted-quad IP addresses in the socknal (2302)
5489 - additional debugging for MDS client eviction problem (2443)
5490 - fix mkfsoptions support for osts (2603, 2604)
5492 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5495 - add series for the vanilla 2.6.0 kernel
5496 - add series for the vanilla 2.4.24 kernel
5497 - add series for a cray x86/64 UL kernel drop
5498 - fix xattr patches for the vanilla 2.4.19 series
5500 - generate true UUIDs in lmc (1171)
5501 - have portals stack dumping break in UML (2466)
5502 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5503 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5504 - wake callback waiting threads on client eviction (2460)
5505 - Add --ptldebug and --subsystem to lmc (1719)
5506 - update assertion to allow safe interrupt allocation
5507 - set rq_no_resend for cancel requests (2432)
5508 - recalculate ptlrpcd timeout after resend (2494)
5509 - call vfs_rmdir when removing pending directories (2368)
5510 - fix renaming a file to itself (2429)
5511 - lmc creates a default one-stripe lov (2454)
5512 - expand procfs space to handle large clusters (2326)
5513 - increase UML stack to avoid overflow
5514 - update lconf's list of debug and subsystem masks
5515 - fix lfs find --obd (2510)
5516 - /proc tunable for disabling filter read caching (2591)
5517 - stop rpm packages from altering slapd.conf (2301)
5518 - disable nagle in the socknal under 0conf (2578)
5519 - choose mds inode size based on stripe count (2572)
5520 - fix kernel-source rpm problems (2516)
5522 - add --disable-doc to avoid pdf generation (2421)
5523 - update documentation, tests, type-os, comments
5524 - avoid format warnings on ia64
5525 - remove the TOE NAL
5526 - tiny code cleanups by removing unused fields
5528 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5531 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5532 - remove the most blatant lies from BUILDING file (2371)
5533 - change default debug level to reasonable production setting
5534 - reduce client side cache size to reduce cache flush time
5535 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5536 - make TCP zerocopy and pinger support enabled by default (2476)
5537 - sync writes completed after process exits caused crashes (2319)
5538 - maintain correct mount count on the MDS (2356)
5539 - backout 1557, because 2316 wasn't really fixed
5540 - better file I/O statistics gathering in /proc
5541 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5542 - another kernel patch to fix zero-copy TCP function export
5543 - don't take duplicate lock when processing re-sent getattr (2420)
5544 - lctl uses obd_self_export instead of creating new conn (2353)
5545 - MDS/OST recovery case which requires object creation asserted (2425)
5546 - move lfs from /usr/sbin to /usr/bin in packages
5547 - fix race between mds_client_add and mds_client_free (2417)
5548 - use kmalloc instead of slabs in portals (2430)
5549 - don't create duplicate records when a failover MDS is present (2442)
5550 - remove unnecessary mount age check (2332)
5551 - don't remove directory inodes from locks prematurely (2451)
5552 - don't break if MDS service name is the same as hostname (2103)
5553 - fix races in client write RPC generation when cache full (2482)
5555 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5558 - remove now-unused request->rq_obd (278)
5559 - if an allocation fails, print out how much memory we've used (1933)
5560 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5561 - add the "configurable stack size" patch to most series files (1256)
5562 - ability to write large log records, for 100+ OST configs (2306)
5563 - fix NULL deref when filter_prep fails (2314)
5564 - fix operator precedence error in filter_sync
5565 - dynamic allocation of socknal TX descriptors (2315)
5566 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5567 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5568 - ensure that CWARN messages in Portals always get to the syslog
5569 - __init/__exit are not for prototype decls (ldlm_init/exit)
5570 - x86-64 compile warning fixes
5571 - fix gateway LMC keyword conflict (2318)
5572 - fix MDS lock inversions in getattr/reint paths (1844)
5573 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5574 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5575 - DLM race condition prevented some lock evictions (2328)
5576 - ENOMEM detection and retry on socknal sends (2230)
5577 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5578 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5579 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5580 - better error messages when a client is rejected during recovery (1505)
5581 - avoid cancelling locks which were never granted, after failure (2330)
5582 - fix i_sem/journal inversion in mds_client_add (2333)
5583 - fix truncate/getattr lock cycle deadlock (2334)
5584 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5585 - fix two rare exit paths which could leak an l_lock() ref (2321)
5586 - fencepost error in MDS/OST orphan recovery (2226)
5587 - make log record alignment 8 bytes (1988)
5588 - lstripe now fails when requested offset > ost_count (2237)
5589 - ensure that all kernel series have a complete list.h (1607)
5590 - fix crashes in special-file operations (2316)
5591 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5593 - allow configurable automake binary, for testing new versions
5594 - small update to the lfs documentation
5596 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5598 * fix negative export reference count in fsfilt_sync (2312)
5600 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5601 * release candidate 0.9.1
5603 - orphans are moved into the PENDING directory for possible recovery
5604 - replayed opens now open by fid for orphan/rename safety (1042)
5605 - last close of an orphan inode generates a transno (683)
5606 - chdir() and mount() now pin the directory entry (1020)
5607 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5608 - discard very old requests without processing them (1502)
5609 - remove some common, well-understood CERRORs (1505)
5610 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5611 - clear "grant" flags in OST replies until OST grant code lands (1644)
5612 - fix read performance by not clobbering i_blksize on client (1598)
5613 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5614 - make LRU size a /proc tunable, clears locks when reduced (707)
5615 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5616 - server locks take references on exports now (1558)
5617 - build fixes for 2.4.20-rh trees (1663)
5618 - return an error from lov_create if all OSCs are inactive (1751)
5619 - fix import levels when a reconnect happens without a timeout (1597)
5620 - exit early from mds_open if we get a lookup error (1749)
5621 - partial page read at EOF wouldn't wait for disk before sending (1642)
5622 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5623 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5624 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5625 - write/truncate lock inversion (1639)
5626 - Don't auto-load obdclass, portals modules during cleanup (1495)
5627 - fix timestamps from jumping to "now" (1763)
5628 - extra journal assertions (1648)
5629 - add an extra multiunlink test (1771)
5630 - fix read_record/write_record API (1776)
5631 - fix leak of offset_extent, possible incorrect i_size later (1772)
5632 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5633 - Debugging for the kqswnal_get_idle_tx problems (1820)
5634 - Allow recovery to be attempted multiple times (1536)
5635 - Write out MDS last_rcvd file after it is first created (1600)
5636 - Fix tx_descriptor leak in failed transmit situations (1827)
5637 - ext3 journaling fixes for assertion failure after IO error (1871)
5638 - class_export_put() on freed export after completion AST error (1896)
5639 - Fix revalidate looping in VFS (1322)
5640 - Don't access a freed export during MDS_REINT timeout (1521)
5641 - Add open-unlink recovery support on the MDS (1673,1764)
5642 - Return an error if no MDS data was read from last_rcvd (1946)
5643 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5644 - Don't setup a disk device that doesn't match exported UUID (317)
5645 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5646 - avoid committing NULL handle in force close
5647 - local.sh is now a one-stripe LOV configuration
5648 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5649 - fix ext3 htree duplicate directory entry corruption (1516)
5650 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5651 - update to new LOV EA format (2097)
5652 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5654 2003-06-15 Phil Schwan <phil@clusterfs.com>
5657 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5658 - per-import/export recovery handling (958, 931, 959)
5659 - multiple last-rcvd slots, for serving multiple FSes (949)
5660 - connections are again shared between multiple imp/exports (963, 964)
5661 - "umount -f" would hang if any requests needed to be sent (393, 978)
5662 - avoid pinning large req buffer by copying for queued messages (989)
5663 - add "uuid" to "lctl device" command to help upcalls (991)
5664 - "open" RPCs with transnos would confuse recovery counters (1037)
5665 - do proper endian conversion of all wire messages (288, 340, 891)
5666 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5667 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5668 - fix ptlrpc_connection leak in target_handle_connect (1174)
5669 - fix import refcounting bug in OST and MDS cleanup (1134)
5670 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5671 - fix very unlikely obd_types race condition (501)
5672 - remove osc_open hack for echo_client (1187)
5673 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5674 - a failure in read_inode2 leads to deadlock (1139)
5675 - cancel ack-locks as soon as transaction is committed (1072)
5676 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5677 - make sure to commitrw after any preprw to avoid deadlock (1162)
5678 - failing to execute a file in a lustre FS would lock inode (1203)
5679 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5680 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5681 - obd_uuid2tgt would walk off the end of the list (1255)
5682 - on IA64 the osc would give portals incorrect bulk size (1258)
5683 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5684 - fix lock inversion caused by new llite matching code (1282)
5685 - limit the number of dirty pages on a client to 10MB (1286)
5686 - timed out locks were not being corrected cancelled (1289)
5687 - fix O_DIRECT above 4GB on IA-32 (1292)
5688 * major user-visible changes
5689 - fail out/fail over policy now controlled by the upcall (993)
5691 - add OBD_PING to check server availability and failure (954)
5692 - lustre messages are now sent in sending host order (288, 340, 891)
5693 - add eadatalen to MDS getattr reply (340)
5694 - OST read replies may contain second buffer, with per-page status (593)
5696 2003-03-11 Phil Schwan <phil@clusterfs.com>
5699 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5700 - failed open()s could cause deadlock; fixed (867, 869)
5701 - stop cancelling OST locks when files are closed (481)
5702 - overlapping XID spaces caused network corruption (851, 853)
5703 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5704 - setattr_raw added, to do single-RPC, server-side setattrs
5705 - lmc/lconf syntax change for OST UUIDs
5706 - fix crashy race condition between ptlrpc_free_req and osc_close
5707 - don't use request in mdc_enqueue if we hit a timeout (889)
5708 - don't set the inode i_size for regular files from the MDS (896)
5709 - handle out of order completion AST (842)
5710 - don't LBUG if a lock request times out after receiving AST (913)
5711 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5712 - if a bad lock AST arrives, send an error instead of dropping entirely
5713 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5714 - fix leak in bulk IO when only partially completed (899, 900, 926)
5715 - fix O_DIRECT for ia64 (55)
5716 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5717 - C-z after timeout could hang a process forever; fixed (977)
5719 - client-side I/O cache (678, 924, 929, 941, 970)
5721 - READPAGE and SETATTRs which don't take server-side locks get
5724 2003-02-11 Phil Schwan <phil@clusterfs.com>
5727 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5728 - Fix fsfilt_extN_readpage() to read a full page of directory
5729 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5730 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5731 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5732 - Most intent operations were removed, in favour of a new RPC mode
5733 that does a single RPC to the server and bypasses most of the VFS
5734 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5735 - Aggressively cancel local locks on DLM servers
5736 - mds_reint_unlink sends EA to the client if it's the last nlink.
5737 client uses that EA to unlink OST objects.
5738 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5739 - recursive symlinks were fixed (439)
5740 - fixed NULL deref in DEBUG_REQ
5741 - filter_update_lastobjid no longer calls sync, which annoyed extN
5742 - fixed multi-client small-writes to a single file problem (445)
5743 - fixed mtime updates during file writes (607)
5744 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5745 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5746 - fixed Linux OST size reporting problem (444, 656)
5747 - OST now updates object mtime with writes or setattr (607, 619)
5748 - client verifies file size before zeroing page past EOF (445)
5749 - OST now writes last allocated objid to disk with allocation (108)
5750 - LOV on echo now works (409)
5752 - mds_reint_unlink sends a new buffer, with the EA included. this
5753 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5754 set if a regular file was being unlinked, and it was the last link
5755 - use PtlGet from the target for bulk writes (315)
5756 - OST now updates object mtime with writes or setattr (607, 619)
5757 - LDLM now has a grant-time callback to revalidate locked items, if
5759 - Many MDS operations were reorganized to combat race conditions
5761 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5762 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5764 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5767 - Fully reactivate OST imports after reconnection (512, others)
5768 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5769 - More graceful error handling for truncating on dead OST (515)
5770 - Don't error out unless we're actually accessing dead stripes (474)
5771 - Fix garbage sizes when stripes are missing (410)
5772 - LRU counters were broken, causing constant lock purge (433, 432)
5773 - garbage on read from stripes with failed OSTs (441)
5774 - mark OSCs as active before reconnecting during recovery (438)
5775 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5776 - lfind did not preserve OST order in output (443)
5777 - symlinks cause hung clients, incorrect data (439)
5778 - stop dereferencing request after dropping refcount (457)
5779 - don't LASSERT(spin_is_locked) on non-SMP (455)
5780 - fixes for many rename() bugs
5781 - fstat didn't correctly synchronize attributes (399)
5782 - server must handle lock cancellation during blocking AST prep (487)
5783 - bulk descriptors were free()d too soon (511)
5784 - fix paths in lconf, which would load incorrect modules (451, 507)
5785 - fix confusing lconf 'host not found' error message (386)
5786 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5787 - fix race condition in mdc_blocking_ast() for inode access (526)
5788 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5789 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5790 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5791 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5792 - fix lstripe in several ways (kernel side) (527)
5793 - fix request leak in ldlm_cli_enqueue (262)
5794 - incorrect OSC was marked inactive after OST failure
5795 - call mds_fs_cleanup before unmounting filesystem (524)
5796 - fix races between taking ns_lock and ldlm_lock_change_resource
5797 - fix races updating LOV export open file list
5798 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5799 - fix recovery NULL deref in ldlm_cli_cancel_unused
5800 - fix some DLM races by using new hash table for lock handles (419)
5801 - permit the client to specify desired inodes, at replay
5802 - duplicate requests when we queue them for replay reintegration
5803 - fix last_rcvd offset calculation
5804 - sync after each recovered transaction, so we always make progress
5805 - never, not always, ERESTART requests without transnos
5806 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5808 - skip replay if the MDS says that the client is already connected
5809 - don't check for a recovery-enabled export to match lctl's UUID
5810 - don't INC_USE_COUNT for phantom exports
5811 - don't crash when cleaning up phantom exports (567)
5812 - don't double-finish or set replay data for errored mdc_open requests
5813 - abort requests when they time out, so we don't get old replies
5814 - send/receive replies for AST messages again
5815 - if the client says that it doesn't have the lock, cancel it on the
5817 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5818 mark it as destroyed, it will all work out when decref is called
5819 - fix module use counts (22, 581)
5821 - ASTs now expect a reply (server cancels lock on error reply)
5823 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5826 - fix many simultaneous client startup (392)
5827 - fix dentry->d_it clobbering
5828 - credentials weren't being shipped for readdir/getattr operations
5829 - remove invalid assertions triggered during some concurrent MD
5831 - proper Lustre versions added (336, 389)
5832 - fix memory leak for create error case (398)
5833 - fix LOV locking bug that would get cli/srv out of sync
5834 - fix echo client over LOV (409)
5835 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5836 - fix double-O_EXCL intent crash (424)
5837 - avoid sending multiple lock CANCELs (352)
5839 - MDS can do multi-client recovery (modulo bugs in new code)
5841 - many updates, edits, cleanups
5843 2002-11-18 Phil Schwan <phil@clusterfs.com>
5846 - fix null d_it dereference (346)
5847 - fix full OST/dbench hang (333)
5848 - fix permission problem with file removal (286)
5849 - fix removal of OSCs from LOV when they fail
5850 - fix NULL deref during bulk timeout (214)
5851 - fix problems related to multiple filesystems on one MDS (241)
5852 - fixed serious subtle metadata locking bugs
5853 - free locks on clients when inodes are removed due to memory
5855 - fix inode pointer in lock data (285)
5856 - partial support for multiple MDS on a single host (241)
5857 - data locks weren't cancelled at clear_inode time (290, 311)
5858 - intent locks could lead to unbounded lock growth (205)
5859 - added a maximum lock count, an LRU list, and a flusher
5860 - fix multiple rename (365)
5861 - properly abstracted the echo client
5862 - OSC locked 1 byte too many; fixed
5863 - rewrote brw callback code:
5864 - fixed recovery bugs related to LOVs (306)
5865 - fixed too-many-pages-in-one-write crash (191)
5866 - fixed (again) crash in sync_io_timeout (214)
5867 - probably fixed callback-related race (385)
5869 - Add capability to MDS protocol
5870 - LDLM cancellations and callbacks on different portals
5872 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5875 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5876 - defer open object destruction to close time (601981, 138)
5877 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5878 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5879 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5880 - don't call obd_setattr in ll_file_release for destroyed objects
5881 * protocol change to lustre_msg: move |version| and add |flags|
5882 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5883 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5884 * added replay of create, unlink, link and rename operations during
5885 MDS failover; recovery should be much more robust now
5886 * remove failed OSCs from LOVs (only lov_create uses this so far)
5887 * the lustre-HOWTO was brought (more) up to date (582544)
5889 2002-10-23 Phil Schwan <phil@clusterfs.com>
5892 - in-use dentries weren't being reused properly (617851)
5893 - prevent multiple LDLM setup (599178)
5894 - fix LOV size calculations for truncate (617853)
5895 - fix client handling of MDS intent errors (POSIX)
5896 - fix permission bug in lovstripe.c test (624321)
5897 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5898 - truncate past end of file could corrupt data
5899 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5900 - a race in recovery could return ETIMEDOUT to apps (623947)
5901 - building outside the source directory was fixed
5902 * the lustre-HOWTO was brought (more) up to date (582544)
5903 * major progress was made on recovery functionality
5905 2002-10-10 Phil Schwan <phil@clusterfs.com>
5908 - recovery deadlock fix
5909 - rm -rf causes LBUG fix (617817)
5910 - file open by multiple tasks fix (618962)
5911 - directory permissions bugs (602707 and 620007)
5912 - journal_stop fixed with locking (611313)
5913 - O_APPEND failures resolved (618273, perhaps 614459)
5914 - lconf PATH fix (619770)
5915 - IA64 build fix (621450)
5916 - RPC buffer sizes scale with amount of memory
5918 2002-10-01 Phil Schwan <phil@clusterfs.com>
5921 - locks would be cancelled without throwing away data pages,
5922 resulting in inconsistent data (605627)
5923 - inode attributes were not always being refreshed (605627, 612449)
5924 - lconf now continues to cleanup after lctl reports an error
5925 - MDS now enforces user permissions (602707)
5926 - lprocfs cleanup fixed, but not yet enabled (614157)
5927 - fixed infinite server hang, should a client not respond to an AST
5928 - avoid going into recovery if user calls readlink() with a buffer
5929 that's too small (613941)
5930 - AST RPCs no longer require replies (614867) -- this may be changed
5931 - don't crash server if client sends an IOV that's too big (611336)
5932 - fixed lock conversion deadlock (611892)
5933 - fixed the following of symlinks (614622)
5934 * recovery: the server can remove locks from a client that dies, other
5935 clients can make progress
5936 * more extN patch fixes
5937 * compile-time configurable ptlrpc buffer allocations
5939 - collaborative read cache document
5940 - Lustre Lite Performance CDR document-in-progress
5942 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5945 - fix typo in patch-2.4.18
5947 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5950 - clear ptlrpc request each time in handle_incoming_request()
5951 - unlink of files now destroys the object on the OST
5953 2002-09-19 Peter Braam <braam@clusterfs.com>
5955 * add hard link support
5956 * change obdfile creation method
5957 * kernel patch changed
5959 2002-09-19 Peter Braam <braam@clusterfs.com>
5962 - stack overflow bug in extN fixed
5964 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5966 * documentation updates
5967 - add man pages for config tools
5968 - update tests/README to describe testing with new config tools
5969 - finish metadata API descriptions
5970 * bug fixes and cleanups
5971 - statfs workaround for 16TB limit
5972 - LOV stripe allocation improved, can stripe on subset of OSTs
5973 - LOV file size/IO offset was wrong for files > 4GB in size
5974 - object EA data was being dropped, caused files to be unreadable
5975 - memory overflow with non-LOV OST caused memory corruption
5976 - fixed regression tests to work with new config tools, obdfilter
5977 - fixed bug when directory size became larger than 1 block
5978 - fixed bug (for single client case) when PWD was deleted
5979 - invalidate local directory pages when doing intent-based ops
5980 - avoid LDLM oops when lock callback contained bad data
5982 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5984 * documentation updates
5985 * bug fixes and cleanups
5986 - configuration tools
5989 - 64-bit compile warnings
5990 - 64-bit internal statfs data
5992 * test_brw on persistent OST devices
5994 * lprocfs (disabled)
5996 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5998 * documentation updates
5999 * bug fixes and cleanups
6000 * configuration tools
6002 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6005 * many small fixes to 0_5_4
6006 * io/network handling
6007 * thinkos in MDS operations
6009 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6012 * crucial basic fixes to 0.5.3
6013 * IOR, Iozone work over Elan
6016 2002-08-07 Phil Schwan <phil@clusterfs.com>
6017 * version 0_5_3, our first alpha
6018 * we use the new Portals iovs
6019 * documentation updates
6020 * bug fixes and cleanups
6021 * small changes in the DLM wire protocol
6023 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6024 * version 0_5_1 with some initial stability,
6025 * locking on MD and file I/O.
6026 * documentation updates
6027 * several bug fixes since 0.5.0
6028 * small changes in wire protocol
6030 2002-07-18 Phil Schwan <phil@clusterfs.com>
6032 * delivered as Lustre Light Alpha
6033 * fixed a crash after handling invalid MDS requests
6034 * fixed directory pages for architectures with non-4k pages sizes
6036 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6037 * release version v0_4_4
6038 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6040 * Fixes a number of interruption problems with OST operations.
6041 * Update documentation for portals header changes
6042 * Move all wire protocol structs/defines to lustre_idl.h
6043 * Fixes symlink length bug.
6044 * Add tcpdump to repository.
6046 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6047 * release version v0_4_3
6048 * Fixes statfs for inodes on extN.
6049 * Fixes bug in runtests which would delete /etc/hosts.
6050 * Use 64-bit object IDs wherever possible (not into VFS though)
6051 Remove ost_get_info, which is unused by lustre, and out of date.
6053 2002-07-03 Peter Braam <braam@clusterfs.com>
6054 * release version v0_4_2 Fixes a lookup error (type not passed)
6055 * move forward to head of Portals
6056 * move forward to latest Lustre kernel
6058 2002-06-25 Peter Braam <braam@clusterfs.com>
6059 * release version v0_4_1. Hopefully stable on single node use.