1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.39.3 (SLES 10),
5 2.6.18-128.7.1.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.41.6.sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
14 * File join has been disabled in this release, refer to Bugzilla 16929.
18 Description: client could not reconnect to OST because of an active request.
19 Details : The client now retries to reconnect to the same server, if a
20 connect request failed with EBUSY or -EAGAIN. Test case 35b
21 conf-sanity.sh has been created.
25 Description: Don't allow make backward step on assignin osc next id.
26 Details : race between allocation next id and ll_sync thread can be cause
27 of set wrong osc next id and can be kill valid ost objects.
31 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
35 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
39 Description: Error handling in osc_statfs_interpret() has been improved.
40 Details : Check in osc_statfs_interpret() for EBADR.
44 Description: Do not send statfs() requests to OSTs disabled by administrator.
45 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
47 Severity : enhancement
49 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
51 Severity : enhancement
53 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
55 Severity : enhancement
57 Description: Enhancement for lfs(1) command to use numeric uid/gid.
61 Description: Improved cleanup procedure for conf-sanity test_46a.
65 Description: Do not start bulk transfer, if client tried to reconnect.
66 Details : Do not start bulk transfer, if client tried to reconnect,
67 but failed due to active rpcs. Check for exp_abort_active_req
72 Description: Prevent inconsistences between linux and lustre mount structures.
73 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
74 Make the sleep interruptible.
76 Severity : enhancement
78 Description: provide server to client comms path
79 Details : server to client communications path via new LDLM_SET_INFO rpc
83 Description: truncate starts GFP_FS allocation under transaction causing deadlock
84 Details : ldiskfs_truncate calls grab_cache_page which may start page
85 allocation under an open transaction. This may lead to
86 calling prune_icache with consequent lustre reentrance.
90 Description: send correctly lsm on open replay
91 Details : MDS is trust to LSM size on replay open, but client can set wrong size
96 Description: enable client interface failover
97 Details : When a child reconnects from another NID, properly update export
98 nid hash position and ldlm reverse import.
100 Severity : enhancement
102 Description: Communicate OST degraded/readonly state via statfs to MDS
103 Details : Flags in the statfs returned from OSTs indicate whether the
104 OST is in a degraded RAID state, or if the filesystem has
105 turned read-only after a filesystem error is detected.
110 Description: don't panic if EPROTO was hit when reading symlink
111 Details : correctly handling request reference in error cases.
113 Severity : enhancement
115 Description: Add LustreNetLink, a kernel-userspace communcation path.
117 Severity : enhancement
119 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
124 Description: access to llog context before init.
125 Details : move handling CATALOGS file at osc layer and forbid access to llog
129 Frequency : rare, only when client is writing to OST on local node
131 Description: Avoid deadlock for local client writes
132 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
133 memory freeing context. This allows OST threads to set the
134 PF_MEMALLOC flag on task structures in order to allocate memory
135 from reserved pools and complete IO.
136 Use GFP_HIGHUSER for OST allocations for non-local client writes,
137 so that the OST threads generate memory pressure and allow
138 inactive pages to be reclaimed.
140 Severity : enhancement
142 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
147 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
148 Details : move ldlm namespace creation in setup phase to avoid grab
149 _lprocfs_lock with cli_sem held.
153 Description: Temporarily disable grant shrink.
154 Details : Disable the feature for debugging.
157 Frequency : only when formatting test filesystems
159 Description: Unable to run several mkfs.lustre on loop devices at the same time
160 Details : mkfs.lustre returns error 256 on the concurrent loop devices
161 formatting. The solution is to proper handle the error.
163 Severity : enhancement
165 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
167 Severity : enhancement
169 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
174 Description: resolve race between obd_disconnect and class_disconnect_exports
175 Details : if obd_disconnect will be called to already disconnected export he
176 forget release one reference and osc module can't unloaded.
178 Severity : enhancement
180 Description: Allow tuning service thread via /proc
181 Details : For each service a new
182 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
183 created that can be used to set min/max thread counts, and get the
184 current number of running threads.
189 Descriptoin: don't return error if have partially created objects for file.
190 Details : lov_update_create_set uses set->set_success as index for created
191 objects, so if some requests will be failed, they will have hole
192 at end of array and qos_shrink_lsm for allocate correct lsm.
194 Severity : enhancement
196 Description: Update OFED support to 1.4.1
201 Description: Reduce small size read RPC
202 Details : Set read-ahead limit for every file and only do read-ahead when
203 available read-ahead pages are bigger than 1M to avoid small size
206 Severity : enhancement
208 Description: move AT tunable parameters for more consistent usage
209 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
211 Severity : enhancement
213 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
214 Details : allow skip disconnected ost for send statfs request and hide error
220 Description: fix racy locking of mballoc block bitmaps causing BUG
221 Details : The locking of the mballoc buddy bitmap and the in-memory
222 block bitmap was using two different spin locks in some
223 cases. This made it possible to incorrectly access the
224 mballoc bitmap while another process was modifying it,
225 causing a sanity assertion to fail. While no on-disk corruption
226 was reported, there was some risk of this happening.
229 Frequency : rare, on llog test 6
231 Descriptoin: don't allow connect to already connected import
232 Details : allowing connect to already connected import is hide connecting problem.
237 Descriptoin: allow kill process which wait statahead result
238 Details : for some reasons 'ls' can stick in waiting result from statahead,
239 in this case need way for kill this process.
241 Severity : enhancement
243 Description: Add state history info file, enhance import info file
244 Details : Track import connection state changes in a new osc/mdc proc file;
245 add overview-type data to the osc/mdc import proc file.
247 Severity : enhancement
249 Description: MDS create should not wait for statfs RPC while holding DLM lock.
251 Severity : enhancement
253 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
258 Descriptoin: OST grant too much space to client even there are not enough space.
259 Details : Client will shrink its grant cache to OST if there are no write
260 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
261 this grant cache if there are already not enough avaible space
262 (left_space < total_clients * 32M).
265 Frequency : start MDS on uncleanly shutdowned MDS device
267 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
268 Details : stay in waiting mds<>ost recovery finished produce random bugs
269 due race between two ll_sync thread for one lov target. send
270 ACTIVATE event only if connect realy finished and import have
274 Frequency : rare, connect and disconnect target at same time
276 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
277 Details : don't call obd_disconnect under lov_lock. this long time
278 operation and can block ptlrpcd which answer to connect request.
283 Descriptoin: don't lose wakeup for imp_recovery_waitq
284 Details : recover_import_no_retry or invalidate_import and import_close can
285 both sleep on imp_recovery_waitq, but we was send only one wakeup
289 Frequency : always with long access acl
291 Descriptoin: mds can't pack reply with long acl.
292 Details : mds don't control size of acl but they limited by reint/getattr
295 Severity : enhancement
297 Description: Update to SLES10 kernel-2.6.16.60-0.33.
299 Severity : enhancement
301 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
304 Frequency : start MDS on uncleanly shutdowned MDS device
306 Descriptoin: aborting recovery hang on MDS
307 Details : don't throttle destroy RPCs for the MDT.
310 Frequency : on remount
312 Description: external journal device not working after the remount
313 Details : clear dev_rdonly flag for external journal devices in
319 Description: shutdown vs evict race
320 Details : client_disconnect_export vs connect request race.
321 if client will evicted at this time - we start invalidate
322 thread without referece to import and import can be freed
326 Frequency : rare, need acl's on inode.
328 Description: client can't handle ost additional correctly
329 Details : if ost was added after client connected to mds client can have
330 hit lnet_try_match_md ... to big messages to wide striped files.
331 in this case need teach client to handle config events about add
332 lov target and update client max ea size at that event.
334 Severity : enhancement
336 Description: Changelogs
337 Details : Changelogs are a lightweight mechanism to track filesystem
338 metadata and namespace changes. The changelog is recorded
339 permanently on the MDTs, and is periodically "consumed" / purged
340 when records are no longer needed.
342 Severity : enhancement
344 Description: compact fld format with extents
345 Details : Store range of seq rather than every seq in FLD. Seq
346 controller update FLD rather than clients. In Case of CMD, mdt0
347 has FLD, all other metadata server act as non persistent proxy
348 for FLD queries and cache fld entries in fld cache.
353 Description: don't skip ost target if they assigned to file
354 Details : Drop slow OSCs if we can, but not for requested start idx.
355 This means "if OSC is slow and it is not the requested
356 start OST, then it can be skipped, otherwise skip it only
357 if it is inactive/recovering/out-of-space.
361 Description: more cleanup in mds_lov
362 Details : not send LOV EA under replay, we can't know about they size at this
363 time. Don't allow client connect to mds before any ost connected,
364 for avoid problems with LOV EA size and returning EIO to client.
366 Severity : enhancement
368 Description: Interoperability at server side (Disk interoperability)
370 Severity : enhancement
372 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
374 Severity : enhancement
376 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
378 Severity : enhancement
380 Description: New client IO stack (CLIO).
382 Severity : enhancement
384 Description: Commit on sharing. Eliminate inter-client dependencies between
385 uncommitted transactions by doing transaction commits.
386 Thereby clients may recovery independently.
389 Frequency : Create a symlink file with a very long name
391 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
392 Details : If there is no extra space in the request for early cancels,
393 ldlm_req_handles_avail() returns 0 instead of a negative value.
395 Severity : enhancement
397 Description: Add /proc entry for import status
398 Details : The mdc, osc, and mgc import directories now have
399 an import directory that contains useful import data for debugging
402 Severity : enhancement
404 Description: Re-disable certain /proc logging
405 Details : Enable and disable client's offset_stats, extents_stats and
406 extents_stats_per_process stats logging on the fly.
409 Frequency : Only on FC kernels 2.6.22+
411 Description: oops in statahead
412 Details : Do not drop reference count for the dentry from VFS when lookup,
413 VFS will do that by itself.
415 Severity : enhancement
417 Description: Generic /proc file permissions
418 Details : Set /Proc file permissions in a more generic way to enable non-
419 root users operate on some /proc files.
423 Description: Hitting mdc_commit_close() ASSERTION
424 Details : Properly handle request reference release in
425 ll_release_openhandle().
429 Description: quota recovery deadlock during mds failover
430 Details : This patch includes att18982, att18236, att18237 in bz14840.
432 1. fix osts hang when mds does failover with quotaon
433 2. prevent watchdog storm when osts threads wait for the
438 Frequency : only patchless client
439 Description: add workaround for race between add/remove dentry from hash
441 Severity : enhancement
443 Description: Allow OST glimpses to return PW locks
447 Description: LBUG when llog conf file is full
448 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
452 Description: Prevent import from entering FULL state when server in recovery
456 Description: service mount cannot take device name with ":"
457 Details : Only when device name contains ":/" will mount treat it as
463 Description: replace ptlrpcd with the statahead thread to interpret the async
464 statahead RPC callback
468 Frequency : on recovery
469 Description: I/O failures after umount during fail back
470 Details : if client reconnected to restarted server we need join to recovery
471 instead of find server handler is changed and process self
472 eviction with cancel all locks.
474 Severity : enhancement
476 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
478 Severity : enhancement
480 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
482 Severity : enhancement
484 Description: Upcall on Lustre log has been dumped
485 Details : Allow for a user mode script to be called once a Lustre log has
486 been dumped. It passes the filename of the dumped log to the
487 script, the location of the script can be specified via
488 /proc/sys/lnet/debug_log_upcall.
493 Description: avoid idr_remove called for id which is not allocated.
494 Details : Move assigment s_dev for clustered nfs to end of initialization,
495 to avoid problem with error handling.
500 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
501 Details : When connection is reused this not moved from CONN_UNUSED_HASH
502 into CONN_USED_HASH and this prodice warning when put connection
503 again in unused hash.
506 Severity : enhancement
508 Description: File striping can now be set to use an arbitrary pool of OSTs.
510 Severity : enhancement
512 Description: Export bytes_read/bytes_write count on OSC/OST.
516 Description: Early reply size mismatch, MGC loses connection
517 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
518 the connect flags are properly negotiated.
523 Description: quotas are not honored with O_DIRECT
524 Details : all writes with the flag O_DIRECT will use grants which leads to
525 this problem. Now using OBD_BRW_SYNC to guard this.
529 Description: add quota statistics
530 Details : 1. sort out quota proc entries and proc code.
531 2. add quota statistics
533 Severity : enhancement
535 Description: enable quota support for HEAD.
539 Description: Properly propagate oinfo flags from lov to osc for statfs
540 Details : restore missing copy oi_flags to lov requests.
542 Severity : enhancement
544 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
545 : l_getgroups(8), lst(8), routerstat(8)
546 Details : included man pages for llobdstat(8), llstat(8),
547 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
549 Severity : enhancement
551 Description: configure's --enable-quota should check the
552 : kernel .config for CONFIG_QUOTA
553 Details : configure is terminated if --enable-quota is passed but
554 : no quota support is in kernel
558 Description: Remove portals compatibility
559 Details : Remove portals compatibility, not interoperable with releases
564 Description: Resolve device initialization race
565 Details : Prevent proc handler from accessing devices added to the
566 obd_devs array but yet be intialized.
568 Severity : enhancement
570 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
572 Severity : enhancement
574 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
579 Description: Using wrong pointer in osc_brw_prep_request
580 Details : Access to array[-1] can produce panic if kernel compiled with
581 CONFIG_PAGE_ALLOC enabled
585 Description: Client runs out of low memory
586 Details : Consider only lowmem when counting initial number of llap pages
590 Description: *optional* service tags registration
591 Details : if the "service tags" package is installed on a Lustre node
592 When the filesystem is mounted, a local-node service tag will
593 be created. See http://inventory.sun.com/ for more information
594 about the Service Tags asset management system.
598 Description: Kernel BUG tries to release flock
599 Details : Lustre does not destroy flock lock before last reference goes
600 away. So always drop flock locks when client is evicted and
601 perform unlock regardless of successfulness of speaking to MDS.
605 Description: add refcount protection for osc callbacks, avoid panic on shutdown
609 Description: sanity test 65a fails if stripecount of -1 is set
610 Details : handle -1 striping on filesystem in ll_dirstripe_verify
615 Description: ASSERTION(CheckWriteback(page,cmd)) failed
616 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
619 Severity : enhancement
621 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
625 Description: do not process already freed flock
626 Details : flock can possibly be freed by another thread before it reaches
627 to ldlm_flock_completion_ast.
631 Description: LBUG during stress test
632 Details : Need properly lock accesses the flock deadlock detection list.
636 Description: oops in page fault handler
637 Details : kernel page fault handler can return two special 'pages' in error
638 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
642 Description: timeout with invalidate import.
643 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
644 should be handled by ptlrpcd. This produce long age waiting and
645 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
647 Severity : enhancement
649 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
653 Description: enable MGS and MDT services start separately
654 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
655 a co-located MGS without starting the MGS, which is a complement
656 to 'nosvc' mount option.
660 Frequency : after recovery
661 Description: precreate to many object's after del orphan.
662 Details : del orphan st in oscc last_id == next_id and this triger growing
663 count of precreated objects. Set flag LOW to skip increase count
664 of precreated objects.
668 Frequency : rare, on clear nid stats
669 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
670 Details : when clean nid stats sometimes try destroy live entry,
671 and this produce panic in free.
675 Description: Stack overflow during MDS log replay
676 ease stack pressure by using a thread dealing llog_process.
680 Description: wait until IO finished before start new when do lock cancel.
681 Details : VM protocol want old IO finished before start new, in this case
682 need wait until PG_writeback is cleared until check dirty flag and
683 call writepages in lock cancel callback.
685 Severity : enhancement
687 Description: using special macro for print time and cleanup in includes.
691 Description: mds_mfd_close() ASSERTION(rc == 0)
692 Details : In mds_mfd_close(), we need protect inode's writecount change
693 within its orphan write semaphore to prevent possible races.
697 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
701 Frequency : rare, on shutdown ost
702 Description: don't hit live lock with umount ost.
703 Details : shrink_dcache_parent can be in long loop with destroy dentries,
704 use shrink_dcache_sb instead.
708 Description: don't panic with use echo client
709 Details : echo client pass NULL as client nid pointer and this produce null
714 Description: fix build on ppc32
715 Details : compile code with -m64 flag produce wrong object file for ppc32.
719 Description: add message levels for liblustreapi
723 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
724 Details : Move the 'good_osts' check before the 'total_bavail' check. This
725 will result in an -EAGAIN and in the exit call path we call
726 alloc_rr() which will with increasing aggressiveness attempt to
727 aquire precreated objects on the minimum number of required OSCs.
731 Description: Use old size assignment to avoid deadlock
732 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
733 the scheduling while holding a spinlock. We do not need locking
734 for size in ll_update_inode() because size is only updated from
735 the MDS for directories or files without objects, so there is no
736 other place to do the update, and concurrent access to such inodes
737 are protected by the inode lock.
741 Description: resolve "_IOWR redefined" build error on SLES10
745 Description: dump the memory debugging after all modules are unloaded to
746 suppress false negative in conf_sanity test 39
748 Severity : enhancement
750 Description: build kernel-ib packages for OFED 1.3 in our release cycle
755 Description: fix SLES kernel versioning
756 Details : the kernel version for our SLES 10 kernel did not include a "-"
757 before the "smp" at the end. while this was not a problem in
758 general, it did mean that software trying to use the kernel
759 version to try to detect a vendor specific kernel would fail.
760 this was most evident by the OFED build scripts.
764 Description: Don't update lov_desc members until making sure they are valid
765 Details : When updating lov_desc members via proc fs, need fix their
766 validities before doing the real update.
770 Description: don't put request into delay list while invalidate in flight.
771 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
772 import in flight. this produce timeout for invalidate and sometimes
773 can cause stale data.
777 Frequency : on ppc only
778 Description: not convert ost objects for directory because it's not exist.
779 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
783 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
785 Severity : enhancement
787 Description: Update to support 2.6.22.14 vanilla kernel.
791 Frequency : rare, on recovery
792 Description: read procfs can produce deadlock in some situation
793 Details : Holding lprocfs lock which send rpc can produce block for destroy
794 obd objects and this also block reconnect with -EALREADY. This
795 isn't fix all lprocfs bugs - but make it rare.
797 Severity : enhancement
799 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
802 Frequency : frequent on X2 node
804 Description: mdc_set_open_replay_data LBUG
805 Details : Set replay data for requests that are eligible for replay.
809 Description: lustre_mgs: operation 101 on unconnected MGS
810 Details : When MGC is disconnected from MGS long enough, MGS will evict the
811 MGC, and late on MGC cannot successfully connect to MGS and a lot
812 of the error messages complaining that MGS is not connected.
816 Frequency : on network error
817 Description: panic with double free request if network error
818 Details : mdc_finish_enqueue is finish request if any network error ocuring,
819 but it's true only for synchronus enqueue, for async enqueue
820 (via ptlrpcd) this incorrect and ptlrpcd want finish request
823 Severity : enhancement
825 Description: client-side metadata stat-ahead during readdir(directory readahead)
826 Details : perform client-side metadata stat-ahead when the client detects
827 readdir and sequential stat of dir entries therein
830 Frequency : on start mds
832 Description: Implement get_info(last_id) in obdfilter.
835 Frequency : occasional
837 Description: Correctly check stale fid, not start epoch if ost not support SOM
838 Details : open with flag O_CREATE need set old fid in op_fid3 because
839 op_fid2 was overwritten with new generated fid, but mds can answer
840 with one of these two fids and both is not stale. Setattr
841 incorrectly started an epoch and assume will be called
842 done_writing, but without SOM done_writing ever being called.
845 Frequency : rare, depends on device drivers and load
847 Description: MDS or OSS nodes crash due to stack overflow
848 Details : Code changes in 1.8.0 increased the stack usage of some functions.
849 In some cases, in conjunction with device drivers that use a lot
850 of stack the MDS (or possibly OSS) service threads could overflow
851 the stack. One change which was identified to consume additional
852 stack has been reworked to avoid the extra stack usage.
855 Frequency : occasional
857 Description: Do not fail import if osc_interpret_create gets -EAGAIN
858 Details : If osc_interpret_create got -EAGAIN it immediately exits and
859 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
860 oscc_has_objects and see OSC has no objests and call
861 oscc_internal_create to resend create request.
863 Severity : enhancement
865 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
867 Severity : enhancement
869 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
872 Frequency : very rare
874 Description: Fix replay if there is an un-replied request and open
875 Details : In some cases, older replay request will revert the
876 mcd->mcd_last_xid on MDS which is used to record the client's
879 Severity : enhancement
881 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
883 Severity : enhancement
885 Description: Add rhel5 support to HEAD.
887 Serverity : enhancement
889 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
894 Description: Don't allow skipping OSTs if index has been specified.
895 Details : Don't allow skipping OSTs if index has been specified, make
896 locking in internal create lots better.
900 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
901 Details : access to bitfield in structure is always rounded to long
902 and this produce problem with not atomic change any bit.
906 Description: Lustre make rpms failed.
907 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
908 builds Lustre rpms from tarball.
911 Frequency : rare on shutdown ost
913 Description: If llog cancel was not send before clean_exports phase, this can
914 produce deadlock in llog code.
915 Details : If llog thread has last reference to obd and call class_import_put
916 this produce deadlock because llog_cleanup_commit_master wait when
917 last llog_commit_thread exited, but this never success because was
918 called from llog_commit_thread.
922 Description: allow userland application know is lost one of stripes.
923 Details : fill lvb_blocks with error code on ost and return it to
924 application if error flag found.
928 Description: NULL lov_tgts causing MDS oops
929 Details : more safe checks for NULL lov_tgts for avoid oops.
931 Severity : enhancement
933 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
937 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
938 Details : If an OST is down the MDS will hang indefinitely in
939 obd_statfs_rqset() waiting for the statfs data. While for
940 MDS QOS usage of statfs, it should not stuck in waiting.
942 Severity : enhancement
944 Description: remote_acl support
945 Details : Support ACL-based permission check for remote user.
946 Support setfacl/getfacl for remote user with the utils
947 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
948 the system "{s,g}etfacl" utils.
950 Severity : enhancement
952 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
954 Severity : enhancement
956 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
960 Description: make mgs_setparam() handle fsname containing dash
961 Details : fsname containing a dash does not work with lctl conf_param
963 Severity : enhancement
965 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
967 Severity : enhancement
969 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
973 Frequency : rare, at shutdown
974 Description: access already free / zero obd_namespace.
975 Details : if client_disconnect_export was called without force flag set,
976 and exist connect request in flight, this can produce access to
977 NULL pointer (or already free pointer) when connect_interpret
978 store ocd flags in obd_namespace.
982 Frequency : only at startup
983 Description: not alloc memory with spinlock held.
984 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
985 if any spinlock held.
987 Severity : enhancement
989 Description: make lustre randomly fail allocating memory
990 Details : Make lustre randomly failed allocating memory for testing purpose.
992 Severity : enhancement
994 Description: lost problems with lov objid file
995 Details : Fixes some scability and access to not inited memory problems
996 in work with lov objdid file.
1001 Description: lfs find does not continue on file error
1002 Details : Continue other files processing when a file/dir is absent.
1006 Description: Inconsistent usage of lustre_pack_reply()
1007 Details : Standardize the usage of lustre_pack_reply() such that it
1008 always generate a CERROR on failure.
1011 Frequency : occasional
1013 Description: MDS hang or stay in waiting lock
1014 Details : If client receive lock with CBPENDING flag ldlm need send lock
1015 cancel as separate rpc, to avoid situation when cancel request
1016 can't processed due all i/o threads stay in wait lock.
1020 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1024 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1028 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1032 Description: Update to RHEL4 latest kernel.
1034 Severity : enhancement
1036 Description: Build SLES10 patchless client fails
1037 Details : The configure was broken by run ./configure with
1038 --with-linux-obj=.... argument for patchless client. When the
1039 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1040 can't search header adequately. Use absolute path such as
1041 -I($LINUX)/include instead.
1045 Description: interrupt oig_wait produce painc on resend.
1046 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1047 private set, and this produce situation when rq_ptlrpcd_data not
1048 copyed to new allocated request and triggered LBUG on assert
1049 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1050 ptlrpcd set if request is changed and can be safety changed to use
1055 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1056 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1057 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1058 these macros and replace them with their difinitions instead.
1062 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1063 Details : Update which_patch & target file for SLES10 latest kernel.
1066 Description: Remove client patches from SLES 10 kernel.
1067 Details : This causes SLES 10 clients to behave as patchless clients
1068 even on a Lustre-patched (server) kernel.
1070 Severity : enhancement
1072 Description: self-adjustable client's lru lists
1073 Details : use adaptive algorithm for managing client cached locks lru
1074 lists according to current server load, other client's work
1075 pattern, memory activities, etc. Both, server and client
1076 side namespaces provide number of proc tunables for controlling
1079 Severity : enhancement
1081 Description: light-weight GSS support
1082 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1083 and reduce performance overhead.
1085 Severity : enhancement
1087 Description: Linux keyring support
1088 Details : Support using service of Linux keyring for Lustre GSS internal
1089 context refresh/cache mechanism.
1093 Description: Fix errors in lfs documentation
1094 Details : Fixes man pages
1098 Description: don't use GFP_* in generic Lustre code.
1099 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1103 Description: obdclass is limited by single OBD_ALLOC(idarray)
1104 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1108 Description: warnings with build patchless client with vanila 2.6.19 and up
1109 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1110 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1114 Frequency : Only for SLES
1115 Description: sanity_quota fail test_1
1116 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1117 file, which makes TSTID[2] inunique.
1121 Frequency : Only for SLES9
1122 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1123 Details : Update target/ChangeLog/which_patch .
1127 Frequency : Only for test_10 in sanity-quota.sh
1128 Description: a bug in quota test code
1129 Details : checking a test flag in a wrong place
1133 Frequency : Only for test_13 in sanity-quota.sh
1134 Description: a bug in quota test script
1135 Details : shouldn't check overfull allocated quota
1139 Description: jbd statistics
1140 Details : Port older jbd statistics patch for sles10
1142 Severity : enhancement
1144 Description: Add printing inode info into message about error in writepage.
1149 Description: reply_lock_interpret crash due to race with it and lock cancel.
1150 Details : Do not replay locks that are being cancelled. Do not reference
1151 locks by their address during replay, just by their handle.
1155 Frequency : When flocks are used.
1156 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1157 Details : Flock locks might destroy just granted lock if it could be merged
1158 with another existing flock, this is done in completion handler,
1159 so teach ldlm_cli_enquque_fini that this is a valid case for
1165 Description: Oops in read and write path when failing to allocate lock.
1166 Details : Check if lock allocation failed and return error back.
1170 Description: lstripe command fails for valid OST index
1171 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1172 instead of lov->desc.ld_active_tgt_count.
1176 Description: sanity.sh failed test 103
1177 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1182 Description: df doesn't work properly if diskfs blocksize != 4K
1183 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1187 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1188 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1190 Severity : enhancement
1192 Description: Adaptive timeouts
1193 Details : RPC timeouts adapt to changing server load and network
1194 conditions to reduce resend attempts and improve recovery time.
1198 Description: llapi_file_create() does not allow some changes
1199 Details : add llapi_file_open() that allows specifying the mode and
1200 open flags, and also returns an open file handle.
1204 Description: merge and cleanup kernel patches.
1205 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1209 Description: Add journal checksum support.(Kernel part)
1210 Details : The journal checksum feature adds two new flags i.e
1211 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1212 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1213 indicates that the commit block contains the checksum for
1214 the blocks described by the descriptor blocks. Now commit
1215 record can be sent to disk without waiting for descriptor
1216 blocks to be written to disk. This behavior is controlled
1217 using JBD2_FEATURE_ASYNC_COMMIT flag.
1221 Description: OSS needs mutliple precreate threads
1222 Details : Add ability to start more than one create thread per OSS.
1226 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1227 Details : execute lfs setstripe on client
1231 Description: mds_obd_create error creating tmp object
1232 Details : When the user sets quota on root, llog will be affected and can't
1233 create files and write files.
1236 Frequency : Always on ia64 patchless client, and possibly others.
1238 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1239 Details : This allows the patchless client to be loaded on architectures
1240 without this export.
1244 Description: RedHat Update kernel for RHEL5
1245 Details : Add the kernel config file for RHEL5.
1249 Description: RedHat Update kernel for RHEL5
1250 Details : Modify the kernel config file more closer RHEL5.
1254 Description: Build failure against Centos5 (RHEL5)
1255 Details : Define PAGE_SIZE when it isn't present.
1259 Description: test_7 on sanity-quota.sh can't be used on separate machines
1260 Details : add facet to handle it
1264 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1265 Details : with NFS, the anon dentry's parent was set to itself in
1266 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1267 corresponding dentry other than use rec->ur_name.
1269 Severity : enhancement
1271 Description: lfs setstripe enhancement
1272 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1276 Description: enable data checksumming by default
1277 Details : enable checksum by default, allow --disable-checksum
1278 configure option and "-o nochecksum" mount option. Checksums
1279 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1283 Description: lustre support for RHEL5
1284 Details : Add support for RHEL5.
1288 Description: Client eviction due to failover config
1289 Details : after a connection loss, the lustre client should attempt to
1290 reconnect to the last active server first before trying the
1291 other potential connections.
1295 Description: when mds and osts use different quota unit(32bit and 64bit),
1296 quota will be released repeatly.
1297 Details : void sending multiple quota reqs to mds, which will keep the
1298 status between the reqs.
1302 Description: osts not allocated evenly to files
1303 Details : change the condition to increase offset_idx
1306 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1308 Description: Data corruption for OSTs that are formatted larger than 2TB
1310 Details : When generating the bio request for lustre file writes the
1311 sector number would overflow a temporary variable before being
1312 used for the IO. The data reads correctly from Lustre (which
1313 will overflow in a similar manner) but other file data or
1314 filesystem metadata may be corrupted in some cases.
1318 Description: Tune the kernel for good SCSI performance.
1319 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1320 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1325 Description: rewrite ext2-derived code in obdclass/uuid.c
1326 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1327 from scratch preserving functionality.
1331 Description: "lfs find -obd UUID" prints directories
1332 Details : "lfs find -obd UUID" will return all directory names instead
1333 of just file names. It is incorrect because the directories
1334 do not reside on the OSTs.
1338 Description: use i_size_read and i_size_write in 2.6 port
1339 Details : replace inode->i_size access with i_size_read/write()
1342 Frequency : when removing large files
1344 Description: scheduling issue during removal of large Lustre files
1345 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1346 It causes scheduling issues when removing large files (17TB in the
1350 Frequency : only with liblustre clients on XT3
1352 Description: evictions taking too long
1353 Details : allow llrd to evict clients directly on OSTs
1356 Frequency : only on ppc
1358 Description: /proc/fs/lustre/devices broken on ppc
1359 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1360 We should make sure the type of 'index' is loff_t and then cast
1361 explicitly as needed below. Do not assign an explicitly cast
1365 Frequency : only for rhel5
1367 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1368 Details : Modify the target file & which_kernel.
1370 Severity : enhancement
1372 Description: omit set fsid for export NFS
1373 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1374 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1378 Frequency : Always, for kernels after 2.6.16
1379 Description: Fix warning idr_remove called for id=.. which is not allocated.
1380 Details : Last kernels save old s_dev before kill super and not allow
1381 to restore from callback - restore it before call kill_anon_super.
1385 Description: buffer overruns could theoretically occur
1386 Details : llapi_semantic_traverse() modifies the "path" argument by
1387 appending values to the end of the origin string, and a
1388 overrun may occur. Adding buffer overrun check in liblustreapi.
1392 Description: Fix error on 'ls .' at the top of the Lustre mount.
1393 Details : Don't revalidate dentry if it is a root dentry.
1397 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1398 Details : Modify vm-tunables-rhel4.patch.
1402 Description: Kernel config for 2.6.18-vanilla.
1403 Details : Modify targets/2.6-vanilla.target.in.
1404 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1405 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1406 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1407 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1411 Description: improve handling recoverable errors
1412 Details : if request processig with error which can be recoverable on server
1413 request should be resend, otherwise page released from cache and
1418 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1419 Details : Modify target file & which_patch.
1420 A flaw was found in the IA32 system call emulation provided
1421 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1422 value could be stored in the %RAX register, which could trigger an
1423 out-of-bounds system call table access. An untrusted local user
1424 could exploit this flaw to run code in the kernel
1425 (ie a root privilege escalation). (CVE-2007-4573).
1429 Description: O_DIRECT bypasses client statistics.
1430 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1431 read_ahead_stats were updated. copied the stats section from
1432 osc_send_oap_rpc() into async_internal().
1436 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1440 Description: change order of libsysio includes
1441 Details : '#include sysio.h' should always come before '#include xtio.h'
1445 Description: handle "serious error: objid * already exists" more gracefully
1446 Details : If LAST_ID value on disk is smaller than the objects existing in
1447 the O/0/d* directories, it indicates disk corruption and causes an
1448 LBUG(). If the object is 0-length, then we should use the existing
1449 object. This will help to avoid a full fsck in most cases.
1451 Severity : enhancement
1453 Description: adapt the lustre_config script to support the upgrade case
1454 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1455 server targets to 1.6 in parallel.
1459 Description: To avoid grant space > avaible space when the disk is almost
1460 full. Without this patch you might see the error "grant XXXX >
1461 available" or some LBUG about grant, when the disk is almost
1463 Details : In filter_check_grant, for non_grant cache write, we should
1464 check the left space by if (*left > ungranted + bytes), instead
1465 of (*left > ungranted), because only we are sure the left space
1466 is enough for another "bytes", then the ungrant space should be
1467 increase. In client, we should update cl_avail_grant only there
1468 is OBD_MD_FLGRANT in the reply.
1472 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1473 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1474 described in CVE-2007-4573. This problem could allow untrusted
1475 local users to gain root access.
1478 Frequency : when using O_DIRECT and quotas
1480 Description: Incorrect file ownership on O_DIRECT output files
1481 Details : block usage reported by 'lfs quota' does not take into account
1482 files that have been written with O_DIRECT.
1487 Description: touch file failed when fs is not full
1488 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1489 otherwise we can get ENOSP while fs is not full.
1493 Description: parallel lock callbacks
1494 Details : Instead of sending blocking and completion callbacks as separated
1495 requests, adding them to a set and sending in parallel.
1498 Frequency : only for Cray XT3
1499 Bugzilla : 12829/13455
1500 Description: Changing primary group doesn't change the group lustre assigns to
1502 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1503 overridden with the primary group provided by the group upcall,
1504 whereas the supplied fsgid can be trusted if it is in the list of
1505 supplementary groups returned by the group upcall.
1507 Severity : enhancement
1509 Description: Allow masking D_WARNING, D_ERROR messages from console
1510 Details : Console messages can now be disabled via lnet.printk.
1514 Description: User code with malformed file open parameter crashes client node
1515 Details : Before packing join_file req, all the related reference should be
1516 checked carefully in case some malformed flags cause fake
1517 join_file req on client.
1521 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1522 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1523 mode is equal to the granted mode. However, it should also check
1524 that LDLM_FL_CP_REQD is not set.
1528 Description: Heavy nfs access might result in deadlocks
1529 Details : After ELC code landed, it is now improper to enqueue any mds
1530 locks under och_sem, because enqueue might want to decide to
1531 cancel open locks for same inode we are holding och_sem for.
1535 Description: Client eviction while running blogbench
1536 Details : A lot of unlink operations with concurrent I/O can lead to a
1537 deadlock causing evictions. To address the problem, the number of
1538 oustanding OST_DESTROY requests is now throttled to
1539 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1540 callbacks are processed in priority.
1544 Description: enable ACLs on MDS by default
1545 Details : ACLs must be enabled on MDS by default.
1548 Frequency : PPC/PPC64 only
1550 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1551 Details : fix duplicated definitions between asm-ppc64/types.h and
1552 lustre_types.h on PPC.
1555 Frequency : PPC/PPC64 only
1557 Description: asm-ppc/segment.h does not exist
1558 Details : fix compile issue on PPC.
1562 Description: better handle error messages in extents code
1565 Frequency : RHEL4 only
1567 Description: mkfs is very slow on IA64/RHEL4
1568 Details : A performance regression has been discovered in the MPT Fusion
1569 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1570 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1571 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1573 Severity : enhancement
1575 Description: SNMP support enhancement
1576 Details : Adding total number of sampled request for an MDS node in snmp
1579 Severity : enhancement
1581 Description: Optimize ldlm waiting list processing for PR extent locks
1582 Details : When processing waiting list for read extent lock and meeting read
1583 lock that is same or wider to it that is not contended, skip
1584 processing rest of the list and immediatelly return current
1585 status of conflictness, since we are guaranteed there are no
1586 conflicting locks in the rest of the list.
1590 Description: Time out and refuse to reconnect
1591 Details : When the failover node is the primary node, it is possible
1592 to have two identical connections in imp_conn_list. We must
1593 compare not conn's pointers but NIDs, otherwise we can defeat
1594 connection throttling.
1598 Description: port llog fixes from b1_6 into HEAD
1599 Details : Port llog reference couting and some llog cleanups from b1_6
1600 (bug 10800) into HEAD, for protect from panic and access to already
1601 free llog structures.
1605 Description: Detect stride IO mode in read-ahead
1606 Details : When a client does stride read, read-ahead should detect that and
1607 read-ahead pages according to the detected stride pattern.
1611 Description: data checksumming impacts single node performance
1612 Details : add support for several checksum algorithm. Currently, only CRC32
1613 and Adler-32 are supported. The checksum type can be changed on
1614 the fly via /proc/fs/lustre/osc/*/checksum_type.
1618 Description: use adler32 for page checksums
1619 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1624 Description: build for x2 fails
1625 Details : fix compile issue on Cray systems.
1629 Description: Properly match for duplicate locks
1630 Details : Due to different lock order from skiplists code, we need to
1631 traverse entire list for now
1634 Frequency : only on PPC/SLES10
1636 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1637 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1638 be defined. Add a hack in mkfs_lustre.c to work around this compile
1643 Description: LASSERT on MDS when client holding flock lock dies
1644 Details : ldlm pool logic depends on number of granted locks equal to
1645 number of released locks which is not true for flock locks, so
1646 just exclude such locks from consideration.
1650 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1651 Details : Use fsfilt_sync() for both the whole filesystem sync and
1652 individual file sync to eliminate dangerous inode locking
1653 with I_LOCK that can lead to a deadlock.
1657 Description: performance in 1.6.3
1658 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1659 sure that 1MB requests are not fragmented by the block layer.
1661 Severity : enhancement
1663 Description: organize the server-side client stats on per-nid basis
1664 Details : Change the structure of stats under obdfilter and mds to
1672 The "uuid"s file would list the uuids of _active_ exports.
1673 And the clear entry is to clear all stats and stale nids.
1675 Severity : enhancement
1677 Description: eliminate client locks in face of contention
1678 Details : file contention detection and lockless i/o implementation
1679 for contended files.
1683 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1686 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1688 Description: MDS may lose file striping (and hence file data) in some cases
1689 Details : If there are additional extended attributes stored on the MDS,
1690 in particular ACLs, SELinux, or user attributes (if user_xattr
1691 is specified for the client mount options) then there is a risk
1692 of attribute loss. Additionally, the Lustre file striping
1693 needs to be larger than default (e.g. striped over all OSTs),
1694 and an additional attribute must be stored initially in the
1695 inode and then increase in size enough to be moved to the
1696 external attribute block (e.g. ACL growing in size) for file
1701 Description: skiplist implementation simplification
1702 Details : skiplists are used to group compatible locks on granted list
1703 that was implemented as tracking first and last lock of each
1704 lock group the patch changes that to using doubly linked lists
1708 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1709 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1710 because LASSERT on both the data supplied by a client, and the
1711 data on disk is dangerous and incorrect.
1713 Severity : enhancement
1715 Description: Slow truncate/writes to huge files at high offsets.
1716 Details : Directly associate cached pages to lock that protect those pages,
1717 this allows us to quickly find what pages to write and remove
1718 once lock callback is received.
1722 Description: more ldlm soft lockups
1723 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1724 starve other threads from the resource lock for a long time in
1725 case of long waiting queue, so change the debug level from
1726 D_OTHER to the less frequently used D_INFO.
1728 Severity : enhancement
1730 Description: add -gid, -group, -uid, -user options to lfs find
1734 Description: Hung threads in invalidate_inode_pages2_range
1735 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1736 once one is completed. As a result, some RPCs are stuck in the
1737 queue and are never sent.
1741 Description: filter threads hungs on waiting journal commit
1742 Details : Cleanup filter group llog code, then only filter group llog will
1743 be only created in the MDS/OST syncing process.
1747 Description: Procfs and llog threads access destoryed import sometimes.
1748 Details : Sync the import destoryed process with procfs and llog threads by
1749 the import refcount and semaphore.
1751 Severity : enhancement
1753 Description: openlock cache of b1_6 port to HEAD
1758 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1759 Details : If insertion of an extent fails, then discard the inode
1760 preallocation and free data blocks else it can lead to duplicate
1765 Description: don't always update ctime in ext3_xattr_set_handle()
1766 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1767 In some cases the ctime should not be updated, for example for
1768 2.0->1.8 compatibility it is necessary to delete an xattr and it
1769 should not update the ctime.
1773 Bugzilla : 15713/16362
1774 Description: Assertion in iopen_connect_dentry in 1.6.3
1775 Details : looking up an inode via iopen with the wrong generation number can
1776 populate the dcache with a disconneced dentry while the inode
1777 number is in the process of being reallocated. This causes an
1778 assertion failure in iopen since the inode's dentry list contains
1779 both a connected and disconnected dentry.
1783 Description: assertion failure in ldlm_handle2lock()
1784 Details : fix a race between class_handle_unhash() and class_handle2object()
1785 introduced in lustre 1.6.5 by bug 13622.
1790 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1791 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1796 Description: oops in sd_iostats_seq_show()
1797 Details : unloading/reloading the scsi low level driver triggers a kernel
1798 bug when trying to access the sd iostat file.
1803 Description: Kernel panics during QLogic driver reload
1804 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1805 patch, causing memory corruption.
1810 Description: journal_dev option does not work in b1_6
1811 Details : pass mount option during pre-mount.
1813 Severity : enhancement
1815 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1816 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1817 extent information of a file. It can be used to map logical blocks
1818 in a file to physical blocks in the block device.
1822 Description: LDLM soft lockups - improvement
1823 Details : It is be possible to send the lock handle along with each read
1824 or write request because the client is already doing a lock match
1825 itself so there isn't any reason the OST should have to re-do that
1829 Frequency : only on Cray X2
1831 Description: X2 build failures
1832 Details : fix build failures on Cray X2.
1836 Description: xid & resent requests
1837 Details : Initialize RPC XID from clock at startup (randomly if clock is
1840 Severity : enhancement
1842 Description: Add lustre_start utility to start or stop multiple Lustre servers
1847 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1848 Details : In case of memory pressure, list_del() can be called twice on
1849 req->rq_history_list, causing a kernel oops.
1853 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1854 Details : fix stack overflow in the distributed lock manager by defering
1855 export eviction after a failed AST to the elt thread instead of
1856 handling it in the dlm interpret routine.
1860 Description: Convert some comments to new format.
1861 Details : Update documenting comments to match doxygen conventions.
1865 Description: Grammar fixes.
1866 Details : A couple of trivial spelling fixes.
1870 Description: OSD_COUNTERS-mandatory
1871 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1872 to simplify the code.
1876 Description: simplify cmm_device freeing logic.
1877 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1881 Description: Add lockdep support to dt_object_operations locking interface.
1882 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1883 parameter indicating lock ordering. Update mdd code to use new
1888 Description: Introduce failloc constants for lockless IO tests.
1889 Details : Add two new failloc constants to test lockless IO. Only one of
1890 them in implemented---another is checked in yet to be landed
1895 Description: Add lockdep support for inode mutex.
1896 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1897 in the situations where Lustre has to lock more than one inode
1902 Description: Add optional invariants checking support.
1903 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1904 configure switch. This macro is to be used for consistency and
1905 sanity checks that are too expensive to be left in `production'
1910 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1911 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1916 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1917 Details : Use spin_lock_nested() in (the only) situation where more than
1918 one ldlm_lock is locked simultaneously. Also, fix possible
1919 dead-lock in ldlm_lock_change_resource() by enforcing particular
1924 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1925 Details : Instead of specifying each ldlm_lock call-back through separate
1926 parameter, wrap them into struct ldlm_callback_suite.
1930 Description: Kill join_lru obd method and its callers.
1931 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1932 in memory---a requirement implemented through ->o_join_lru() obd
1933 method in HEAD. Remove this method and its users.
1937 Description: Add asynchronous ldlm ENQUEUE completion handler.
1938 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1939 case that stock ldlm_completion_ast() cannot handle as it waits
1940 until lock is granted. Introduce new ldlm_completion_ast_async()
1941 for this. Also comment ldlm_completion_ast().
1945 Description: ldlm_error <-> errno conversion.
1946 Details : Add functions to map (rather arbitrary) between LDLM error codes
1947 and standard errno values. CLIO needs this to prevent LDLM specific
1948 constants from escaping ldlm and osc.
1952 Description: Kill unused ldlm_handle2lock_ns() function.
1953 Details : Kill unused ldlm_handle2lock_ns() function.
1957 Description: Add lu_ref support to ldlm_lock
1958 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1959 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1960 and ldlm_resource. LDLM interface has to be changed, because code that
1961 releases a reference on a lock, has to "know" what reference this is.
1962 In the most frequent case
1964 lock = ldlm_handle2lock(handle);
1966 LDLM_LOCK_PUT(lock);
1968 no changes are required. When any other reference (received _not_ from
1969 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1970 instead of LDLM_LOCK_PUT().
1972 Arguably, changes are pervasive, and interface requires some discipline
1973 for proper use. On the other hand, it was very instrumental in finding
1974 a few leaked lock references.
1978 Description: Add ldlm_lock_addref_try().
1979 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1980 attempts to addref a lock that might be being canceled
1985 Description: Add ldlm_weigh_callback().
1986 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1987 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1988 hard-coded `number of pages' logic.
1992 Description: Add lockdep annotations to llog code.
1993 Details : Use appropriately tagged _nested() locking calls in the places
1994 where llog takes more than one ->lgh_lock lock.
1998 Description: Add loi_kms_set().
1999 Details : Wrap kms updates into a helper function.
2003 Description: Constify instances of struct lsm_operations.
2004 Details : Constify instances of struct lsm_operations.
2008 Description: lu_conf support.
2009 Details : On a server, a file system object is uniquely identified
2010 by a fid, which is sufficient to locate and load all object
2011 state (inode). On a client, on the other hand, more data are
2012 necessary instantiate an object. Change lu_object_find() and
2013 friends to take additional `lu_conf' argument describing object.
2014 Typically this includes layout information.
2018 Description: lu_context fixes.
2019 Details : Introduce new lu_context functions that are needed on the client
2020 side, where some system threads (ptlrpcd) are shared by multiple
2021 modules, and so cannot be stopped during module shutdown.
2025 Description: Add start and stop methods to lu_device_type_operations.
2026 Details : Introduce two new methods in lu_device_type_operations, that are
2027 invoked when first instance of a given type is created and last one
2028 is destroyed respectively. This is need by CLIO.
2032 Description: Add lu_ref support to struct lu_device.
2033 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2034 track leaked references.
2038 Description: Introduce lu_kmem_descr.
2039 Details : lu_kmem_descr and its companion interface allow to create
2040 and destroy a number of kmem caches at once.
2044 Description: Fix lu_object finalization race.
2045 Details : Fix a race between lu_object_find() finding an object and its
2046 concurrent finalization. This race is (most likely) not possible
2047 on the server, but might happen on the client.
2051 Description: Introduce lu_ref interface.
2052 Details : lu_ref is a debugging module allowing to track references to
2053 a given object. It is quite cpu expensive, and has to be
2054 explicitly enabled with --enable-lu_ref. See usage description
2059 Description: Factor lu_site procfs stats into a separate function.
2060 Details : Separate lu_site stats printing code into a separate function
2061 that can be reused on a client.
2065 Description: Constify instances of struct {lu,dt,md}_device_operations.
2066 Details : Constify instances of struct {lu,dt,md}_device_operations.
2070 Description: Introduce struct md_site and move meta-data specific parts of
2071 struct lu_site here.
2072 Details : Move md-specific fields out of struct lu_site into special struct
2073 md_site, so that lu_site can be used on a client.
2077 Description: Kill mdd_lov_destroy().
2078 Details : Remove unused mdd code.
2082 Description: Add st_block checking to multistat.c.
2083 Details : Add st_block checking to multistat.c.
2087 Description: Add lu_ref support to struct obd_device.
2088 Details : Add lu_ref tracking to obd_device.
2092 Description: Kill obd_set_fail_loc().
2093 Details : Remove unused code.
2097 Description: Add special type for ptlrpc_request interpret functions.
2098 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2099 there. Actual usage will be in CLIO.
2103 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2104 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2105 doesn't work with lockdep.
2109 Description: Add rwv.c test program.
2110 Details : New testing program exercising readv(2) and writev(2) (Qian).
2114 Description: Add sendfile.c test program.
2115 Details : New testing program exercising sendfile(2) (Jay).
2119 Description: Ratelimit a message that can be very frequent.
2120 Details : Ratelimit a memory allocation failure message that can
2125 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2126 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2130 Description: Liblustre build fixes.
2131 Details : Liblustre build fixes.
2135 Description: libcfs: add cfs_{need,cond}_resched() interface.
2136 Details : libcfs: add cfs_{need,cond}_resched() definition and
2137 implementations for Linux, NT, and liblustre.
2139 Severity : enhancement
2141 Description: More exported tunables for mballoc
2142 Details : Add support for tunable preallocation window and new tunables for
2143 large/small requests
2147 Description: Detect corruption of block bitmap and checking for preallocations
2148 Details : Checks validity of on-disk block bitmap. Also it does better
2149 checking of number of applied preallocations. When corruption is
2150 found, it turns filesystem readonly to prevent further corruptions.
2154 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2155 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2156 window to 0. If it does hit cache miss, and read-pattern is still
2157 stride-io mode, does not reset the stride window, but also does not
2158 increase the stride window length in this case.
2162 Frequency : only for big-endian servers
2163 Description: Check if system is big-endian while mounting fs with extents feature
2164 Details : Mounting a filesystem with extents feature will fail on big-endian
2165 systems since ext3-based ldiskfs is not supported on big-endian
2166 systems. This can be over-riden with "bigendian_extents" mount option.
2168 Severity : enhancement
2170 Description: The root squash functionality
2171 Details : A security feature, which is to prevent users from being able
2172 to mount lustre on their desktop, run as root, and delete
2173 all of the files in the filesystem. The goal is accomplished by
2174 remapping user id (UID) and group id (GID) of the root user to
2175 a UID and GID specified by the system administartor via Lustre
2176 configuration management server (MGS). The functionality also
2177 allows to specify sets of clients for which the remapping does
2182 Description: Excessive recovery window
2183 Details : With AT enabled, the recovery window can be excessively long (6000+
2184 seconds). To address this problem, we no longer use
2185 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2186 timeout no longer depends on the service time, it is set to
2187 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2188 time via pb_service_time.
2192 Description: Watchdog triggered on MDS failover
2193 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2194 the OSTs know that the MDS "UUID" can be reused for the same export
2195 from a different NID, so we do not need to wait for the export to be
2199 Frequency : rare, only if using MMP with Linux RAID
2201 Description: MMP doesn't work with Linux RAID
2202 Details : While using HA for Lustre servers with Linux RAID, it is possible
2203 that MMP will not detect multiple mounts. To make this work we
2204 need to unplug the device queue in RAID when the MMP block is being
2205 written. Also while reading the MMP block, we should read it from
2206 disk and not the cached one.
2208 Severity : enhancement
2210 Description: open file using fid
2211 Details : A file can be opened using just its fid, like
2212 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2215 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2218 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2219 Details : add proper locking for creation of jbd_4k slab cache
2223 Description: MMP check in ext3_remount() fails without displaying any error
2224 Details : When multiple mount protection fails during remount, proper error
2227 Severity : enhancement
2229 Description: Allow stripe size to be up to 4G-64k
2230 Details : Fix math logic to allow large stripe sizes.
2234 Description: add check for >8TB ldiskfs filesystems
2235 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2236 Don't allow >8TB ldiskfs filesystems to be mounted without
2237 force_over_8tb mount option
2242 Descriptoin: OSS complains "too many blocks" and crashes
2243 Details : In filter_commitrw_write():
2244 1. lquota_chkquota();
2245 2. obdo_from_inode();
2246 3. lquota_pending_commit();
2247 "2." has changed uid and gid in this case, then caused
2248 lquota_chkquota() and lquota_pending_commit() to use
2249 different uid and gid. That is the root of the bug.
2253 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2254 Details : Detect and print out the 'stripe count = -1' case properly.
2256 --------------------------------------------------------------------------------
2258 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2260 * Support for kernels:
2261 2.6.5-7.283 (SLES 9),
2262 2.6.9-55.EL (RHEL 4),
2263 2.6.16.46-0.14 (SLES 10),
2264 2.6.18.8 vanilla (kernel.org)
2265 * Client support for unpatched kernels:
2266 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2267 2.6.16 - 2.6.22 vanilla (kernel.org)
2268 * Due to recently discovered recovery problems, we do not recommend
2269 using patchless RHEL 4 clients with this or any earlier release.
2270 * Recommended e2fsprogs version: 1.39.cfs8
2271 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2272 * Starting with this release, the ldiskfs backing filesystem required
2273 by Lustre is now in its own package, lustre-ldiskfs. This package
2274 should be installed. It is versioned separately from Lustre and
2275 may be released separately in future.
2279 Description: block reactivating mgc import until all deactivates complete
2280 Details : Fix race when failing back MDT/MGS to itself (testing)
2283 Frequency : at statup only
2285 Description: mds_lov_synchronize race leads to various problems
2286 Details : simultaneous MDT->OST connections at startup can cause the
2287 sync to abort, leaving the OSC in a bad state.
2289 Severity : enhancement
2291 Description: add optional extra BUILD_VERSION info
2292 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2293 to override the lustre version.
2296 Frequency : 2.6.18 servers only
2298 Description: ll_kern_mount() doesn't release the module reference
2299 Details : The ldiskfs module reference count never drops down to 0
2300 because ll_kern_mount() doesn't release the module reference.
2305 Description: server LBUG when using old ost_num_threads parameter
2306 Details : Accept the old ost_num_threads parameter but warn that it
2307 is deprecated, and fix an off-by-one error that caused an LBUG.
2312 Description: Transient SCSI error results in persistent IO issue
2313 Details : iobuf->dr_error is not reinitialized to 0 between two
2317 Frequency : sometimes when underlying device returns I/O errors
2319 Description: OSTs not going read-only during write failures
2320 Details : OSTs are not remounted read-only when the journal commit threads
2321 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2322 instead of the ext3 wrappers.
2326 Description: poor connect scaling with increasing client count
2327 Details : Don't run filter_grant_sanity_check for more than 100 exports
2328 to improve scaling for large numbers of clients.
2331 Frequency : SLES10 only
2333 Description: sanity-quota.sh quotacheck failed: rc = -22
2334 Details : Quotas cannot be enabled on SLES10.
2337 Frequency : liblustre clients only
2339 Description: getdirentries does not give error when run on compute nodes
2340 Details : getdirentries does not fail when the size specified as an argument
2341 is too small to contain at least one entry
2343 Severity : enhancement
2345 Description: Add LNET router traceability for debug purposes
2346 Details : If a checksum failure occurs with a router as part of the
2347 IO path, the NID of the last router that forwarded the bulk data
2348 is printed so it can be identified.
2353 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2354 Details : Due to a race condition, liblustre clients were occasionally
2355 evicted incorrectly.
2357 Severity : enhancement
2359 Description: lfs setstripe use optional parameters instead of postional
2362 Severity : enhancement
2364 Description: Nanosecond timestamp support for ldiskfs
2365 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2366 resolution timestamps. There is not yet support for this at
2367 the Lustre filesystem level.
2370 Frequency : during server recovery
2372 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2373 Details : request with rq_no_resend flag not awake l_wait_event if they get
2377 Frequency : nfs export on patchless client
2379 Description: connectathon hang when test nfs export over patchless client
2380 Details : Disconnected dentry cannot be found with lookup, so we do not need
2381 to unhash it or make it invalid
2384 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2385 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2386 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2391 Description: open req refcounting wrong on reconnect
2392 Details : If reconnect happened between getting open reply from server and
2393 call to mdc_set_replay_data in ll_file_open, we will schedule
2394 replay for unreferenced request that we are about to free.
2395 Subsequent close will crash in variety of ways.
2396 Check that request is still eligible for replay in
2397 mdc_set_replay_data().
2402 Description: disable writes to filesystem when reading health_check file
2403 Details : the default for reading the health_check proc file has changed
2404 to NOT do a journal transaction and write to disk, because this
2405 can cause reads of the /proc file to hang and block HA state
2406 checking on a healthy but otherwise heavily loaded system. It
2407 is possible to return to the previous behaviour during configure
2408 with --enable-health-write.
2410 Severity : enhancement
2412 Description: 64-bit inode version
2413 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2414 to the inode. This will be required for version-based recovery.
2419 Description: MDS fails to start if a duplicate client export is detected
2420 Details : in some rare cases it was possible for a client to connect to
2421 an MDS multiple times. Upon recovery the MDS would detect this
2422 and fail during startup. Handle this more gracefully.
2424 Severity : enhancement
2426 Description: Add -o localflock option to simulate old noflock
2428 Details : This will achieve local-only flock/fcntl locks
2434 Description: log_commit_thread vs filter_destroy race leads to crash
2435 Details : Take import reference before releasing llog record semaphore
2440 Description: Wrong request locking in request set processing
2441 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2442 imp_delayed_list, in this place should be used imp_lock.
2445 Frequency : when reconnection
2447 Description: Grant Leak when osc reconnect to OST
2448 Details : When osc reconnect ost, OST(filter) should check whether it
2449 should grant more space to client by comparing fed_grant and
2450 cl_avail_grant, and return the granted space to client instead
2451 of "new granted" space, because client will call osc_init_grant
2452 to update the client grant space info.
2455 Frequency : when client reconnect to OST
2457 Description: Grant Leak when osc do resend and replay bulk write
2458 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2459 bulk write request, because the grant info while be sync between
2460 OSC and OST when reconnect, and we should ignore the grant info
2461 these of resend/replay write req.
2466 Description: Grant space more than avaiable left space sometimes.
2467 Details : When then OST is about to be full, if two bulk writing from
2468 different clients came to OST. Accord the avaliable space of the
2469 OST, the first req should be permitted, and the second one
2470 should be denied by ENOSPC. But if the seconde arrived before
2471 the first one is commited. The OST might wrongly permit second
2472 writing, which will cause grant space > avaiable space.
2475 Frequency : when client is evicted
2477 Description: Grant might be wrongly erased when osc is evicted by OST
2478 Details : when the import is evicted by server, it will fork another
2479 thread ptlrpc_invalidate_import_thread to invalidate the
2480 import, where the grant will be set to 0. While the original
2481 thread will update the grant it got when connecting. So if
2482 the former happened latter, the grant will be wrongly errased
2483 because of this race.
2488 Description: Checking Stale with correct fid
2489 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2490 to check whether it is stale, because sometimes, we want the
2491 enqueue happened anyway, and op_data.fid2 will not be initialized.
2493 Severity : enhancement
2495 Description: update patchless client
2496 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2499 Frequency : only with 2.4 kernel
2501 Description: random memory corruption
2502 Details : size of struct ll_inode_info is to big for union inode.u and this
2503 can be cause of random memory corruption.
2508 Description: Memory leak in recovery
2509 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2510 It should also check obd_fail before fsfilt_start, otherwise if
2511 fsfilt_start return -EROFS,(failover mds during mds recovery).
2512 then the req will return with repmsg->transno = 0 and rc = EROFS.
2513 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2514 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2515 be freed no matter whether fsfilt_commit success or not.
2518 Frequency : only with huge count clients
2520 Description: Prevents from taking the superblock lock in llap_from_page for
2522 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2523 for llap_from_page call in ll_removepage prevents from taking the
2524 superblock lock for a soon died page.
2529 Description: Not check open intent error before release open handle
2530 Details : in some rare cases, the open intent error is not checked before
2531 release open handle, which may cause
2532 ASSERTION(open_req->rq_transno != 0), because it tries to release
2533 the failed open handle.
2538 Description: Set cat log bitmap only after create log success.
2539 Details : in some rare cases, the cat log bitmap is set too early. and it
2540 should be set only after create log success.
2544 Description: Accessing a block bevice can re-enable I/O when Lustre is
2545 tearing down a device.
2546 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2551 Description: service threads may hog cpus when there are a lot of requests
2552 Details : Insert cond_resched to give other threads a chance to use some CPU
2557 Description: the cat log was not initialized in recovery
2558 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2559 unlink log on mds will not be initialized until mds post
2560 recovery. And also in mds post recovery, the unlink log will
2561 initialization will be done asynchronausly, so there will be race
2562 between add unlink log and unlink log initialization.
2566 Description: brw_stats were being printed incorrectly
2567 Details : brw_stats were being printed as log2 but all of them were not
2568 recorded as log2. Also remove some code duplication arising from
2569 filter_tally_{read,write}.
2573 Frequency : rare, only in recovery.
2574 Description: ASSERTION(req->rq_type != LI_POISON) failed
2575 Details : imp_lock should be held while iterating over imp_sending_list for
2576 prevent destroy request after get timeout in ptlrpc_queue_wait.
2580 Description: replay-single.sh test 52 fails
2581 Details : A lock's skiplist need to be cleanup when it being unlinked
2582 from its resource list.
2586 Description: Short directio read returns full requested size rather than
2588 Details : Direct I/O operations should return actual amount of bytes
2589 transferred rather than requested size.
2591 Severity : enhancement
2593 Description: metadata RPC reduction (e.g. for rm performance)
2594 Details : decrease the amount of synchronous RPC between clients and servers
2595 by canceling conflicing lock before the operation on the client
2596 and packing thier handles into the main operation RPC to server.
2598 Severity : enhancement
2600 Description: Async OSC create to avoid the blocking unnecessarily.
2601 Details : If a OST has no remain object, system will block on the creating
2602 when need to create a new object on this OST. Now, ways use
2603 pre-created objects when available, instead of blocking on an
2604 empty osc while others are not empty. If we must block, we block
2605 for the shortest possible period of time.
2609 Description: improve handling recoverable errors
2610 Details : if request processig with error which can be recoverable on server
2611 request should be resend, otherwise page released from cache and
2614 Severity : enhancement
2616 Description: refine locking for avoid write wrong info into lov_objid file
2617 Details : fix possible races with add new target and write/update data in
2620 --------------------------------------------------------------------------------
2622 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2627 Frequency : on some architectures
2629 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2630 Details : Uninitialized flags sometimes cause configuration commands to
2634 Frequency : patchless clients only
2636 Description: missing __iget() symbol export
2637 Details : The __iget() symbol export is missing. To avoid the need for
2638 this on patchless clients the deathrow inode reaper is turned
2639 off, and we depend on the VM to clean up old inodes. This
2640 dependency was during via the fix for bug 12181.
2642 --------------------------------------------------------------------------------
2644 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2646 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2647 INTEROPERATE with older versions automatically. In many cases a
2648 special upgrade step is needed. Please read the
2649 user documentation before upgrading any part of a live system.
2650 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2651 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2652 * WARNING: Lustre configuration and startup changes are required with
2653 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2657 Severity : enhancement
2659 Description: MountConf
2660 Details : Lustre configuration is now managed via mkfs and mount
2661 commands instead of lmc and lconf. New obd types (MGS, MGC)
2662 are added for dynamic configuration management. See
2663 https://mail.clusterfs.com/wikis/lustre/MountConf for
2666 Severity : enhancement
2668 Description: dynamic OST addition
2669 Details : OSTs can now be added to a live filesystem
2671 Severity : enhancement
2673 Description: startup order invariance
2674 Details : MDTs and OSTs can be started in any order. Clients only
2675 require the MDT to complete startup.
2677 Severity : enhancement
2679 Description: parallel, asynchronous orphan cleanup
2680 Details : orphan cleanup is now performed in separate threads for each
2681 OST, allowing parallel non-blocking operation.
2683 Severity : enhancement
2685 Description: optimized stripe assignment
2686 Details : stripe assignments are now made based on ost space available,
2687 ost previous usage, and OSS previous usage, in order to try
2688 to optimize storage space and networking resources.
2690 Severity : enhancement
2692 Description: Permanently set tunables
2693 Details : All writable /proc/fs/lustre tunables can now be permanently
2694 set on a per-server basis, at mkfs time or on a live system.
2696 Severity : enhancement
2698 Description: Lustre message v2
2699 Details : Add lustre message format v2.
2701 Severity : enhancement
2703 Description: client OST exclusion list
2704 Details : Clients can be started with a list of OSTs that should be
2705 declared "inactive" for known non-responsive OSTs.
2709 Description: SPEC SFS validation failure on NFS v2 over lustre.
2710 Details : Changes the blocksize for regular files to be 2x RPC size,
2711 and not depend on stripe size.
2713 Severity : enhancement
2715 Description: Multiple MD RPCs in flight.
2716 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2717 To avoid overly-overloading MDS, introduce a limit on number of
2718 MDS RPCs in flight for a single client and add /proc controls
2719 to adjust this limit.
2721 Severity : enhancement
2723 Description: client read/write statistics
2724 Details : Add client read/write call usage stats for performance
2725 analysis of user processes.
2726 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2727 file access. extents_stats shows chunk size distribution.
2728 extents_stats_per_process show chunk size distribution per
2731 Severity : enhancement
2733 Description: mds statistics
2734 Details : Add detailed mds operations statistics in
2735 /proc/fs/lustre/mds/*/stats.
2739 Description: Failure of copying files with lustre special EAs.
2740 Details : Client side always return success for setxattr call for lustre
2741 special xattr (currently only "trusted.lov").
2746 Description: Refcount LNET uuids
2747 Details : The global LNET uuid list grew linearly with every startup;
2748 refcount repeated list entries instead of always adding to
2751 Severity : enhancement
2753 Description: Dynamic service threads
2754 Details : Within a small range, start extra service threads
2755 automatically when the request queue builds up.
2758 Frequency : mixed-endian client/server environments
2760 Description: mixed-endian crashes
2761 Details : The new msg_v2 system had some failures in mixed-endian
2764 Severity : enhancement
2766 Description: Easy OST removal
2767 Details : OSTs can be permanently deactivated with e.g. 'lctl
2768 conf_param lustre-OST0001.osc.active=0'
2770 Severity : enhancement
2772 Description: MGS proc entries
2773 Details : Added basic proc entries for the MGS showing what filesystems
2776 Severity : enhancement
2778 Description: provide MGS failover
2779 Details : Added config lock reacquisition after MGS server failover.
2781 Severity : enhancement
2783 Description: add Linux 2.4 support
2784 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2788 Description: a large application tries to do I/O to the same resource and dies
2789 in the middle of it.
2790 Details : Check the req->rq_arrival time after the call to
2791 ost_brw_lock_get(), but before we do anything about
2792 processing it & sending the BULK transfer request. This
2793 should help move old stale pending locks off the queue as
2794 quickly as obd_timeout.
2797 Frequency : when an incorrect nid is specified during startup
2799 Description: ptlrpc connect to non-existant node causes kernel crash
2800 Details : LNET can't be re-entered from an event callback, which
2801 happened when we expire a message after the export has been
2802 cleaned up. Instead, hand the zombie cleanup off to another
2805 Severity : enhancement
2807 Description: plain/inodebits lock performance improvement
2808 Details : Grouping plain/inodebits in granted list by their request modes
2809 and bits policy, thus improving the performance of search through
2813 Frequency : only if OST filesystem is corrupted
2815 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2816 Details : for a short time RPCs with bulk IO are in the replay list,
2817 but replay of bulk IOs is unimplemented. If the OST filesystem
2818 is corrupted due to disk cache incoherency and then replay is
2819 started it is possible to trip an assertion. Avoid putting
2820 committed RPCs into the replay list at all to avoid this issue.
2823 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2825 Description: System hang on startup
2826 Details : This bug allowed the liblustre (e.g. catamount) client to
2827 return to the app before handling all startup RPCs. This
2828 could leave the node unresponsive to lustre network traffic
2829 and manifested as a server ptllnd timeout.
2831 Severity : enhancement
2833 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2834 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2835 boolean to control whether to print peer debug info when a
2836 client's RPC times out.
2839 Frequency : only for kernels with patches from Lustre below 1.4.3
2841 Description: Remove old rdonly API
2842 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2845 Frequency : only for devices with external journals
2847 Description: Set external device read-only also
2848 Details : During a commanded failover stop, we set the disk device
2849 read-only while the server shuts down. We now also set any
2850 external journal device read-only at the same time.
2853 Frequency : when upgrading from 1.4 while trying to change parameters
2855 Description: The wrong (new) MDC name was used when setting parameters for
2856 upgraded MDT's. Also allows changing of OSC (and MDC)
2857 parameters if --writeconf is specified at tunefs upgrade time.
2860 Frequency : when setting specific ost indicies
2862 Description: QOS code breaks on skipped indicies
2863 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2864 created with --index need not be sequential.
2866 Severity : enhancement
2868 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2869 Details : The uninit_groups feature works in conjunction with the kernel
2870 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2871 up the pass1 processing of e2fsck. This is a read-only feature
2872 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2873 mount filesystems that have had this feature enabled.
2875 Severity : enhancement
2877 Description: Improve multi-block allocation algorithm to avoid fragmentation
2878 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2879 allocation locality and avoid filesystem fragmentation.
2881 ------------------------------------------------------------------------------
2883 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2885 * Support for kernels:
2886 2.4.21-47.0.1.EL (RHEL 3)
2887 2.6.5-7.283 (SLES 9)
2888 2.6.9-42.0.10.EL (RHEL 4)
2889 2.6.12.6 vanilla (kernel.org)
2890 2.6.16.27-0.9 (SLES 10)
2891 * Recommended e2fsprogs version: 1.39.cfs5
2893 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2897 Frequency : occasional, depends on client load and configuration
2898 Bugzilla : 12181, 12203
2899 Description: data loss for recently-modified files
2901 Details : In some cases it is possible that recently written or created
2902 files may not be written to disk in a timely manner (this should
2903 normally be within 30s unless client IO load is very high).
2904 The problem appears as zero-length files or files that are a
2905 multiple of 1MB in size after a client crash or client eviction
2906 that are missing data at the end of the file.
2908 This problem is more likely to be hit on clients where files are
2909 repeatedly created and unlinked in the same directory, clients
2910 have a large amount of RAM, have many CPUs, the filesystem has
2911 many OSTs, the clients are rebooted frequently, and/or the files
2912 are not accessed by other nodes after being written.
2914 The presence of the problem can be detected by looking at
2915 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2916 smaller than the second (nr_unused) then dirty files will not
2917 be flushed automatically to disk. "sync; sleep 10" should be
2918 run several times on the node before unmounting it to update
2919 Lustre (this is also safe to run on nodes without this problem).
2921 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2922 that can cause this same problem, so customers using that kernel
2923 also need to update the kernel in addition to Lustre. In order
2924 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2926 It is normal that files written just before a client crash (less
2927 than 30s) may not yet have been flushed to disk, even for local
2931 Frequency : frequent on thin XT3 nodes
2933 Description: UUID collision on thin XT3 Linux nodes
2934 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2935 randomly, since we relied on an insufficiently-seeded PRNG.
2940 Description: OSS hangs after "All ost request buffers busy"
2941 Details : A deadlock between quota and journal operations caused OSS
2942 hangs after printing "All ost request buffers busy."
2945 Frequency : always on liblustre builds
2947 Description: Cleanup compiler warnings on liblustre
2950 Frequency : always on liblustre builds on XT3
2952 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2956 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2957 > 32000 subdirectories
2958 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2959 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2960 subdirectory count crosses 32000. This will aid e2fsck to
2961 correctly handle more than 32000 subdirectories.
2964 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2966 Description: System hang on startup
2967 Details : This bug allowed the liblustre (e.g. catamount) client to
2968 return to the app before handling all startup RPCs. This
2969 could leave the node unresponsive to lustre network traffic
2970 and manifested as a server ptllnd timeout.
2972 Severity : enhancement
2974 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2975 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2976 boolean to control whether to print peer debug info when a
2977 client's RPC times out.
2982 Description: make O_SYNC working on 2.6 kernels
2983 Details : 2.6 kernels use different method for mark pages for write,
2984 so need add a code to lustre for O_SYNC work.
2989 Description: Failure to close file and release space on NFS
2990 Details : Put inode details into lock acquired in ll_intent_file_open.
2991 Use mdc_intent_lock in ll_intent_open to properly
2992 detect all kind of errors unhandled by mdc_enqueue.
2997 Description: proc file read during shutdown sometimes raced obd removal,
2999 Details : Add lock to prevent obd access after proc file removal.
3002 Frequency : Only for files larger than 4GB on 32-bit clients.
3004 Description: improperly doing page alignment of locks
3005 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3006 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3011 Description: RPCs being resent when they shouldn't be
3012 Details : Some RPCs that should not be resent are being resent. This
3013 can cause inconsistencies in the RPC state machine. Do not
3014 resend such requests.
3017 Frequency : rare, only with NFS export
3019 Description: Crash on NFS re-export node
3020 Details : under very unusual load conditions an assertion is hit in
3021 ll_intent_file_open()
3024 Frequency : only if OST filesystem is corrupted
3026 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3027 Details : for a short time RPCs with bulk IO are in the replay list,
3028 but replay of bulk IOs is unimplemented. If the OST filesystem
3029 is corrupted due to disk cache incoherency and then replay is
3030 started it is possible to trip an assertion. Avoid putting
3031 committed RPCs into the replay list at all to avoid this issue.
3036 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3037 Details : Large single O_DIRECT read and write calls can fail to allocate
3038 a sufficiently large buffer to process the request. In case of
3039 allocation failure the allocation is retried with a smaller
3040 buffer and broken into smaller requests.
3042 Severity : enhancement
3044 Description: Add -o localflock option to simulate old noflock behaviour.
3045 Details : This will achieve local-only flock/fcntl locks coherentness.
3050 Description: versioning check is incomplete
3051 Details : Checking the version difference of client vs. server, report
3052 error if the gap is too big.
3057 Description: add support PG_writeback bit
3058 Details : add support for PG_writeback bit for Lustre, for more carefull
3059 work with page cache in 2.6 kernel. This also fix some deadlocks
3060 and remove hack for work O_SYNC with 2.6 kernel.
3062 Severity : enhancement
3064 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3065 Details : The uninit_groups feature works in conjunction with the kernel
3066 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3067 up the pass1 processing of e2fsck. This is a read-only feature
3068 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3069 mount filesystems that have had this feature enabled.
3071 Severity : enhancement
3073 Description: Improve multi-block allocation algorithm to avoid fragmentation
3074 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3075 allocation locality and avoid filesystem fragmentation.
3077 ------------------------------------------------------------------------------
3079 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3081 * Support for kernels:
3082 2.6.9-42.0.3.EL (RHEL 4)
3083 2.6.5-7.276 (SLES 9)
3084 2.4.21-47.0.1.EL (RHEL 3)
3085 2.6.12.6 vanilla (kernel.org)
3086 2.6.16.21-0.8 (SLES10)
3087 * Recommended e2fsprogs version: 1.39.cfs2-0
3089 * The backwards-compatible /proc/sys/portals symlink has been removed
3090 in this release. Before upgrading, please ensure that you change
3091 any configuration scripts or /etc/sysctl.conf files that access
3092 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3093 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3094 in advance of the upgrade on any system running Lustre 1.4.6 or
3095 newer, since /proc/sys/lnet was added in that version.
3096 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3100 Frequency : only when quota is used
3102 Description: avoid scanning export list for quota master
3103 Details : Change the algorithms to avoid scanning export list in order
3104 to improve the efficiency.
3107 Frequency : MDS failover only, very rarely
3109 Description: "went back in time" messages on mds failover
3110 Details : The greatest transno may be lost when the current operation
3111 finishes with an error (transno==0) and the client's last_rcvd
3112 record is over-written. Save the greatest transno in the
3113 mds_last_transno for this case.
3116 Frequency : always for specific kernels and striping counts
3118 Description: client may get "Matching packet too big" without ACL support
3119 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3120 when trying to access files in certain configurations. The
3121 clients should in fact be denied when mounting because they do
3122 not understand ACLs.
3125 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3127 Description: many clients connecting with IO in progress causes connect timeouts
3128 Details : Avoid synchronous journal commits to avoid delays caused by many
3129 clients connecting/disconnecting when bulk IO is in progress.
3130 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3131 OST_IO_PORTAL to avoid delays behind potentially many pending
3135 Frequency : occasionally with multiple writers to a single file
3137 Description: shared writes to file may result in wrong size reported by stat()
3138 Details : Allow growing of kms when extent lock is cancelled
3141 Frequency : always with random mmap IO to multi-striped file
3143 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3144 Details : Only if the hole is at the end of OST object so that kms is too
3145 small. Fix is to increase kms accordingly in ll_nopage.
3148 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3150 Description: writes to a missing object would leak memory on the OST
3151 Details : If there is an inconsistency between the MDS and OST filesystems,
3152 such that the MDS references an object that doesn't exist, writes
3153 to that object will leak memory due to incorrect cleanup in the
3154 error handling path, eventually running out of memory on the OST.
3159 Description: Creating too long symlink causes lustre errors
3160 Details : Check symlink and name lengths before sending requests to MDS.
3163 Frequency : only if flock is enabled (not on by default)
3165 Description: posix locks not released on fd closure on 2.6.9+
3166 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3167 kernels, this caused such locks not to be released on fd close and
3168 then assertions on fs unmount about still used locks.
3171 Frequency : MDS failover only, very rarely
3173 Description: clients may get ASSERTION(granted_lock != NULL)
3174 Details : When request was taking a long time, and a client was resending
3175 a getattr by name lock request. The were multiple lock requests
3176 with the same client lock handle and
3177 mds_getattr_name->fixup_handle_for_resent_request found one of the
3178 lock handles but later failed with ASSERTION(granted_lock != NULL).
3183 Description: handle->h_buffer_credits > 0, assertion failure
3184 Details : h_buffer_credits is zero after truncate, causing assertion
3185 failure. This patch extends the transaction or creates a new
3189 Frequency : NFS re-export or patchless client
3190 Bugzilla : 11179, 10796
3191 Description: Crash on NFS re-export node (__d_move)
3192 Details : We do not want to hash the dentry if we don't have a lock.
3193 But if this dentry is later used in d_move, we'd hit uninitialised
3194 list head d_hash, so we just do this to init d_hash field but
3195 leave dentry unhashed.
3198 Frequency : NFS re-export or patchless client
3200 Description: NFS exports has problem with symbolic link
3201 Details : lustre client didn't properly install dentry when re-exported
3202 to NFS or running patchless client.
3205 Frequency : NFS re-export or patchless client
3207 Description: Various nfs/patchless fixes.
3208 Details : fixes reuse disconected alias for lookup process - this fixes
3209 warning "find_exported_dentry: npd != pd",
3210 fix permission error with open files at nfs.
3211 fix apply umask when do revalidate.
3214 Frequency : occasional
3216 Description: Crash on NFS re-export node
3217 Details : calling clear_page() on the wrong pointer triggered oops in
3218 generic_mapping_read().
3221 Frequency : rarely, using O_DIRECT IO
3223 Description: unaligned directio crashes client with LASSERT
3224 Details : check for unaligned buffers before trying any requests.
3227 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3229 Description: stale data returned from RAID cache
3230 Details : If only a small amount of IO is done to the RAID device before
3231 reading it again it is possible to get stale data from the RAID
3232 cache instead of reading it from disk.
3235 Frequency : always for sles10 kernel
3237 Description: sles10 support
3238 Details : ll_follow_link: compile fixes and using of nd_set_link
3239 under newer kernels.
3242 Frequency : depends on arch, kernel and compiler version, always on sles10
3245 Description: recursive or deep enough symlinks cause stack overflow
3246 Details : getting rid of large stack-allocated variable in
3250 Frequency : depends on hardware
3252 Description: lustre write performance loss in the SLES10 kernel
3253 Details : the performance loss is caused by using of write barriers in the
3254 ext3 code. The SLES10 kernel turns barrier support on by
3255 default. The fix is to undo that change for ldiskfs.
3257 ------------------------------------------------------------------------------
3259 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3261 * Support for kernels:
3262 2.6.9-42.0.3EL (RHEL 4)
3263 2.6.5-7.276 (SLES 9)
3264 2.4.21-47.0.1.EL (RHEL 3)
3265 2.6.12.6 vanilla (kernel.org)
3269 Frequency : quota enabled and large files being deleted
3271 Description: releasing more than 4GB of quota at once hangs OST
3272 Details : If a user deletes more than 4GB of files on a single OST it
3273 will cause the OST to spin in an infinite loop. Release
3274 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3279 Description: statfs data retrieved from /proc may be stale or zero
3280 Details : When reading per-device statfs data from /proc, in the
3281 {kbytes,files}_{total,free,avail} files, it may appear
3282 as zero or be out of date.
3285 Frequency : systems with MD RAID1 external journal devices
3287 Description: lconf's call to blkid is confused by RAID1 journal devices
3288 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3289 of returning all block devices that match the journal UUID.
3292 Frequency : always, for aggregate stripe size over 4GB
3294 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3295 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3296 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3299 Frequency : always if "lfs find" used on a local file/directory
3301 Description: "lfs find" segfaults if used on a local file/directory
3302 Details : The case where a directory component was not specified wasn't
3303 handled correctly. Handle this properly.
3306 Frequency : always on ppc64
3308 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3309 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3310 perform endianness conversion.
3313 Frequency : rarely (truncate to non-zero file size after write under load)
3314 Bugzilla : 10730, 10687
3315 Description: Files padded with zeros to next 4K multiple
3316 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3317 it is possible that files that are truncated to a non-zero size
3318 immediately after being written are filled with zero bytes beyond
3319 the truncated size. No file data is lost.
3321 Severity : enhancement
3323 Description: Allow recovery/failover for liblustre clients.
3324 Details : liblustre clients were unaware of failover configurations until
3327 Severity : enhancement
3329 Description: user file locks should fail when not mounting with flock option
3330 Details : Set up an error-returning stub in ll_file_operations.lock field
3331 to prevent incorrect behaviour when client is mounted without
3332 flock option. Also, set up properly f_op->flock field for
3336 Frequency : always on ia64
3338 Description: "lfs df" loops on printing out MDS statfs information
3339 Details : The obd_ioctl_data was not initialized and in some systems
3340 this caused a failure during the ioctl that did not return
3341 an error. Initialize the struct and return an error on failure.
3344 Frequency : SLES 9 only
3346 Description: Error of copying files with lustre special EAs as root
3347 Details : Client side always return success for setxattr call for lustre
3348 special xattr (currently only "trusted.lov").
3351 Frequency : rarely on clusters with both ia64+i386 clients
3353 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3354 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3355 clients are doing concurrent writes to the same file it is
3356 possible that the ia64 clients may LASSERT because the OST
3357 extent locks are not PAGE_SIZE aligned. Ensure that grown
3358 locks are always aligned on the request boundary.
3361 Frequency : specific use, occasional
3363 Description: Overwriting in use executable truncates on-disk binary image
3364 Details : If one node attempts to overwrite an executable in use by
3365 another node, we now correctly return ETXTBSY instead of
3366 truncating the file.
3368 Severity : enhancement
3370 Description: Async OSC create to avoid the blocking unnecessarily.
3371 Details : If a OST has no remain object, system will block on the creating
3372 when need to create a new object on this OST. Now, ways use
3373 pre-created objects when available, instead of blocking on an
3374 empty osc while others are not empty. If we must block, we block
3375 for the shortest possible period of time.
3380 Description: chmod on Lustre root is propagated to other clients
3381 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3382 invalid by the follow_mount time.
3385 Frequency : liblustre clients only
3387 Description: Race in 'instant cancel' lock handling could lead to such locks
3388 never to be granted in case of SMP MDS
3389 Details : Do not destroy not yet granted but cbpending locks in
3393 Frequency : replay/resend of open
3395 Description: non null lock assetion failure in mds_intent_policy
3396 Details : Trying to replay/resend lockless open requests resulted in
3397 mds_open() returning 0 with no lock. Now it sets a flag if
3398 a lock is going to be returned.
3400 Severity : enhancement
3402 Description: Checksum enhancements
3403 Details : New checksum enhancements allow for resending RPCs that failed
3406 Severity : enhancement
3408 Description: Tunables on number of dirty pages in cacche
3409 Details : Allow to set limit on number of dirty pages cached.
3414 Description: client crash on unmount - lock still has references
3415 Details : In some error handling cases it was possible to leak a lock
3416 reference on a client while accessing a file. This was not
3417 harmful to the client during operation, but would cause the
3418 client to crash when the filesystem is unmounted.
3421 Frequency : specific case, rare
3423 Description: ETXTBSY on mds though file not in use by client
3424 Details : ETXTBSY is no longer incorrectly returned when attempting to
3425 chmod or chown a directory that the user previously tried to
3426 execute or a currently-executing binary.
3429 Frequency : extremely rare except on liblustre-based clients
3431 Description: Lustre space not freed when files are deleted
3432 Details : Clean up open-unlinked files after client eviction. Previously
3433 the unlink was skipped and the files remained as orphans.
3438 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3439 Details : In some cases it was possible to send a blocking callback to
3440 a client doing a glimpse, even though that client didn't get
3441 a lock granted. When the glimpse lock is cancelled on the OST
3442 the freed lock is left on the waiting list and corrupted the list.
3445 Frequency : all core dumps
3447 Description: Broke core dumps to lustre
3448 Details : Negative dentry may be unhashed if parent does not have UPDATE
3449 lock, but some callers, e.g. do_coredump, expect dentry to be
3450 hashed after successful create, hash it in ll_create_it.
3452 ------------------------------------------------------------------------------
3454 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3456 * Support for kernels:
3457 2.6.9-42.0.2.EL (RHEL 4)
3458 2.6.5-7.276 (SLES 9)
3459 2.4.21-40.EL (RHEL 3)
3460 2.6.12.6 vanilla (kernel.org)
3464 Frequency : always on RHEL 3
3466 Description: Number of open files grows over time
3467 Details : The number of open files grows over time, whether or not
3468 Lustre is started. This was due to a filp leak introduced
3469 by one of our kernel patches.
3471 ------------------------------------------------------------------------------
3473 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3475 * Support for kernels:
3476 2.6.9-42.EL (RHEL 4)
3477 2.6.5-7.276 (SLES 9)
3478 2.4.21-40.EL (RHEL 3)
3479 2.6.12.6 vanilla (kernel.org)
3484 Bugzilla : 5719, 9635, 9792, 9684
3485 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3486 Details : If a server is under heavy load and cannot reply to new
3487 connection requests before the client resends the (re)connect,
3488 the connection handling code can behave badly if two service
3489 threads are concurrently handing separate (re)connections from
3490 the same client. Add better locking to the connection handling
3491 code, and ensure that only a single connection will be processed
3492 for a given client UUID, even if the lock is dropped.
3494 Severity : enhancement
3496 Description: add TCP zero-copy support to kernel
3497 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3498 sends if the hardware supports scatter-gather and checksumming.
3499 This allows socklnd to do client-write and server-read more
3500 efficiently and reduce CPU utilization from skbuf copying.
3503 Frequency : only if NFS exporting from client
3505 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3506 Details : If setting attributes on a file created under NFS that had
3507 never been opened it would be possible to oops the client
3508 if the file had no objects.
3511 Frequency : always for liblustre
3513 Description: liblustre client does MDS+OSTs setattr RPC for each write
3514 Details : When doing a write from a liblustre client, the client
3515 incorrectly issued an RPC to the MDS and each OST the file was
3516 striped over in order to update the timestamps. When writing
3517 with small chunks and many clients this could overwhelm the MDS
3518 with RPCs. In all cases it would slow down the write because
3519 these RPCs are unnecessary.
3521 Severity : enhancement
3523 Description: allow number of MDS service threads to be changed at module load
3524 Details : It is now possible to change the number of MDS service threads
3525 running. Adding "options mds mds_num_threads={N}" to the MDS's
3526 /etc/modprobe.conf will set the number of threads for the next
3527 time Lustre is restarted (assuming the "mds" module is also
3528 reloaded at that time). The default number of threads will
3529 stay the same, 32 for most systems.
3534 Description: OST crash if filesystem is unformatted or corrupt
3535 Details : If an OST is started on a device that has never been formatted
3536 or if the filesystem is corrupt and cannot even mount then the
3537 error handling cleanup routines would dereference a NULL pointer.
3542 Description: NULL pointer deref in llap_from_page.
3543 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3544 incorrect) mapping if the page was truncated/reclaimed while it was
3545 searched for. Check for this condition and skip such pages when
3546 doing readahead. Introduce extra check to llap_from_page() to
3547 verify page->mapping->host is non-NULL (so page is not anonymous).
3550 Frequency : Sometimes when using sys_sendfile
3552 Description: "page not covered by a lock" warnings from ll_readpage
3553 Details : sendfile called ll_readpage without right page locks present.
3554 Now we introduced ll_file_sendfile that does necessary locking
3555 around call to generic_file_sendfile() much like we do in
3559 Frequency : with certain MDS communication failures at client mount time
3561 Description: NULL pointer deref after failed client mount
3562 Details : a client connection request may delayed by the network layer
3563 and not be sent until after the PTLRPC layer has timed out the
3564 request. If the client fails the mount immediately it will try
3565 to clean up before the network times out the request. Add a
3566 reference from the request import to the obd device and delay
3567 the cleanup until the network drops the request.
3570 Frequency : occasionally during client (re)connect
3572 Description: assertion failure during client (re)connect
3573 Details : processing a client connection request may be delayed by the
3574 client or server longer than the client connect timeout. This
3575 causes the client to resend the connection request. If the
3576 original connection request is replied in this interval, the
3577 client may trip an assertion failure in ptlrpc_connect_interpret()
3578 which thought it would be the only running connect process.
3581 Frequency : only with obd_echo servers and clients that are rebooted
3583 Description: kernel BUG accessing uninitialized data structure
3584 Details : When running an obd_echo server it did not start the ping_evictor
3585 thread, and when a client was evicted an uninitialized data
3586 structure was accessed. Start the ping_evictor in the RPC
3587 service startup instead of the OBD startup.
3589 Severity : enhancement
3590 Bugzilla : 10193 (patchless)
3591 Description: Remove dependency on various unexported kernel interfaces.
3592 Details : No longer need reparent_to_init, exit_mm, exit_files,
3593 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3596 Frequency : rare (only users of deprecated and unsupported LDAP config)
3598 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3599 Details : LDAP apparently contains 'lustreName' attributes instead of
3600 'name'. A simple remapping of the name is sufficient.
3603 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3605 Description: waiting_locks_callback trips kernel BUG if client is evicted
3606 Details : Running with the dump_on_timeout debug flag turned on makes
3607 it possible that the waiting_locks_callback() can try to dump
3608 the Lustre kernel debug logs from an interrupt handler. Defer
3609 this log dumping to the expired_lock_main() thread.
3611 Severity : enhancement
3613 Description: Support NFS exporting on 2.6 kernels.
3614 Details : Implement non-rawops metadata methods for NFS server to use without
3615 changing NFS server code.
3618 Frequency : very rare (synthetic metadata workload only)
3620 Description: two racing renames might cause an MDS thread to deadlock
3621 Details : Running the "racer" program may cause one MDS thread to rename
3622 a file from being the source of a rename to being the target of
3623 a rename at exactly the same time that another thread is doing
3624 so, and the second thread has already enqueued these locks after
3625 doing a lookup of the target and is trying to relock them in
3626 order. Ensure that we don't try to re-lock the same resource.
3629 Frequency : only very large systems with liblustre clients
3631 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3632 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3633 all OSTs in parallel. This allows the eviction of stale liblustre
3634 clients to proceed much faster than if they were done in series,
3635 and also offers similar improvements for other set_info RPCs.
3640 Description: excessive CPU usage during initial read phase on client
3641 Details : During the initial read phase on a client, it would agressively
3642 retry readahead on the file, consuming too much CPU and impacting
3643 performance (since 1.4.5.8). Improve the readahead algorithm
3644 to avoid this, and also improve some other common cases (read
3645 of small files in particular, where "small" is files smaller than
3646 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3651 Description: MDS crash when receiving packet with unknown intent.
3652 Details : Do not LBUG in unknown intent case, just return -EFAULT
3654 Severity : enhancement
3655 Bugzilla : 9293, 9385
3656 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3657 Details : Do not serialize getattr (non-intent version) and statfs.
3660 Frequency : occasional, when OST network is overloaded/intermittent
3662 Description: client evicted by OST after bulk IO timeout
3663 Details : If a client sends a bulk IO request (read or write) the OST
3664 may evict the client if it is unresposive to its data GET/PUT
3665 request. This is incorrect if the network is overloaded (takes
3666 too long to transfer the RPC data) or dropped the OST GET/PUT
3667 request. There is no need to evict the client at all, since
3668 the pinger and/or lock callbacks will handle this, and the
3669 client can restart the bulk request.
3672 Frequency : Always when mmapping file with no objects
3674 Description: client crashes when mmapping file with no objects
3675 Details : Check that we actually have objects in a file before doing any
3676 operations on objects in ll_vm_open, ll_vm_close and
3682 Description: Request leak when working with deleted CWD
3683 Details : Introduce advanced request refcount tracking for requests
3684 referenced from lustre intent.
3686 Severity : Enhancement
3688 Description: Cache open file handles on client.
3689 Details : MDS now will return special lock along with openhandle, if
3690 requested and client is allowed to hold openhandle, even if unused,
3691 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3692 closing files for every read/write openration.
3694 Severity : Enhancement
3696 Description: Cache open negative dentries on client when possible.
3697 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3698 negative dentries on lock revocation.
3703 Description: Remounting a client read-only wasn't possible with a zconf mount
3704 Details : It wasn't possible to remount a client read-only with llmount.
3706 Severity : enhancement
3707 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3708 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3709 MPICH is included to add Lustre-specific ADIO interfaces.
3710 This is based closely on the UFS ADIO layer and only differs
3711 in file creation, in order to allow the OST striping to be set.
3712 This is user-contributed code and not supported by CFS.
3717 Description: extended inode attributes (immutable, append-only) work improperly
3718 when 2.4 and 2.6 kernels are used on client/server or vice versa
3719 Details : Introduce kernel-independent values for these flags.
3721 Severity : enhancement
3724 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3725 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3726 granularity. Now, Fractional MB granularity is be supported,
3727 this is very useful for low memory system.
3729 Severity : enhancement
3731 Description: Getattr by fid
3732 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3733 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3734 have locks and avoids getattr rpc requests for it on every stat.
3737 Frequency : Always, for filesystems larger than 2TB
3739 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3740 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3741 filesystems larger than 2TB to be mounted with mballoc enabled.
3744 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3746 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3747 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3748 filesystems larger than 2TB is mounted then the kernel will
3749 silently corrupt the start of the filesystem. CONFIG_LBD is
3750 enabled for all CFS-supported kernels, but the possibility of
3751 this happening with a modified kernel config exists.
3753 Severity : enhancement
3755 Description: add client O_DIRECT support for 2.6 kernels
3756 Details : It is now possible to do O_DIRECT reads and writes to files
3757 in the Lustre client mountpoint on 2.6 kernel clients.
3759 Severity : enhancement
3761 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3762 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3763 parallel, not waiting for response from every OST before sending
3764 a rpc to the next OST.
3769 Description: setattr vs write race when updating file timestamps
3770 Details : Client processes that update a file timestamp into the past
3771 right after writing to the file (e.g. tar) it is possible that
3772 the updated file modification time can be reset to the current
3773 time due to a race between processing the setattr and write RPC.
3775 Severity : enhancement
3777 Description: Bring 'lfs find' closer in line with regular Linux find.
3778 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3779 -print0 options and obtains all the needed info through the lustre
3782 Severity : enhancement
3784 Description: support up to 1024 configured devices on one node
3785 Details : change obd_dev array from statically allocated to dynamically
3786 allocated structs as they are first used to reduce memory usage
3791 Description: Flush dirty partially truncated pages during truncate
3792 Details : Immediatelly flush partially truncated pages in filter_setattr,
3793 this way we completely avoid having any pages in page cache on OST
3794 and can retire ugly workarounds during writes to flush such pages.
3799 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3800 Details : protect inode from truncation within vfs_unlink() context
3801 just take a reference before calling vfs_unlink() and release it
3802 when parent's i_sem is free.
3807 Description: last_id value checked outside lock on OST caused LASSERT failure
3808 Details : If there were multiple MDS->OST object precreate requests in
3809 flight, it was possible that the OST's last object id was checked
3810 outside a lock and incorrectly tripped an assertion. Move checks
3811 inside locks, and discard old precreate requests.
3814 Frequency : always, if extents are used on OSTs
3816 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3817 Details : OSTs using the extents format would not zero the high 16 bits of
3818 the index physical block number. This is not a problem for any
3819 OST filesystems smaller than 16TB, and no kernels support ext3
3820 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3821 new/modified files) and can be fixed for existing filesystems
3822 with e2fsprogs-1.39-cfs1.
3827 Description: import connection selection may be incorrect if timer wraps
3828 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3829 import connections to be ignored if the 32-bit jiffies counter
3830 wraps. Use a 64-bit jiffies counter.
3833 Frequency : very large clusters immediately after boot
3835 Description: LNET request buffers exhausted under heavy short-term load
3836 Details : If a large number of client requests are generated on a service
3837 that has previously never seen so many requests it is possible
3838 that the request buffer growth cannot keep up with the spike in
3839 demand. Instead of dropping incoming requests, they are held in
3840 the LND until the RPC service can accept more requests.
3843 Frequency : Sometimes during replay
3845 Description: Assertion failure in ll_local_open after replay.
3846 Details : If replay happened on an open request reply before we were able
3847 to set replay handler, reply will become not swabbed tripping the
3848 assertion in ll_local_open. Now we set the handler right after
3849 recognising of open request
3852 Frequency : very rare
3854 Description: kernel reports "badness in vsnprintf"
3855 Details : Reading from the "recovery_status" /proc file in small chunks
3856 may cause a negative length in lprocfs_obd_rd_recovery_status()
3857 call to vsnprintf() (which is otherwise harmless). Exit early
3858 if there is no more space in the output buffer.
3860 Severity : enhancement
3862 Description: clear OBD RPC statistics by writing to them
3863 Details : It is now possible to clear the OBD RPC statistics by writing
3864 to the "stats" file.
3869 Description: Client mtime is not the same on different clients after utimes
3870 Details : In some cases, the client was using the utimes() syscall on
3871 a file cached on another node. The clients now validate the
3872 ctime from the MDS + OSTs to determine which one is right.
3877 Description: Inability to activate failout mode
3878 Details : lconf script incorrectly assumed that in pythong string's numeric
3879 value is used in comparisons.
3882 Frequency : always with multiple stripes per file
3884 Description: Inefficient object allocation for mutli-stripe files
3885 Details : When selecting which OSTs to stripe files over, for files with
3886 a stripe count that divides evenly into the number of OSTs,
3887 the MDS is always picking the same starting OST for each file.
3888 Return the OST selection heuristic to the original design.
3893 Description: mount failures may take full timeout to return an error
3894 Details : Under some heavy load conditions it is possible that a
3895 failed mount can wait for the full obd_timeout interval,
3896 possibly several minutes, before reporting an error.
3897 Instead return an error as soon as the status is known.
3899 Frequency : quota enabled and large files being deleted
3901 Description: releasing more than 4GB of quota at once hangs OST
3902 Details : If a user deletes more than 4GB of files on a single OST it
3903 will cause the OST to spin in an infinite loop. Release
3904 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3909 Description: statfs data retrieved from /proc may be stale or zero
3910 Details : When reading per-device statfs data from /proc, in the
3911 {kbytes,files}_{total,free,avail} files, it may appear
3912 as zero or be out of date.
3915 Frequency : systems with MD RAID1 external journal devices
3917 Description: lconf's call to blkid is confused by RAID1 journal devices
3918 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3919 of returning all block devices that match the journal UUID.
3922 Frequency : always, for aggregate stripe size over 4GB
3924 Description: assertion fails when trying to use 4GB stripe size
3925 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3926 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3929 Frequency : always on ppc64
3931 Description: the first write on an ext3 filesystem with mballoc got stuck
3932 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3933 perform endianness conversion.
3935 ------------------------------------------------------------------------------
3937 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3939 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3940 INTEROPERATE with older versions automatically. Please read the
3941 user documentation before upgrading any part of a live system.
3942 * WARNING: Lustre networking configuration changes are required with
3943 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3946 * Support for kernels:
3947 2.6.9-22.0.2.EL (RHEL 4)
3948 2.6.5-7.244 (SLES 9)
3949 2.6.12.6 vanilla (kernel.org)
3952 Severity : enhancement
3953 Bugzilla : 7981/8208
3954 Description: Introduced Lustre Networking (LNET)
3955 Details : LNET is new networking infrastructure for Lustre, it includes
3956 a reorganized network configuration mode (see the user
3957 documentation for full details) as well as support for routing
3958 between different network fabrics. Lustre Networking Devices
3959 (LNDs) for the supported network fabrics have also been
3960 created for this new infrastructure.
3962 Severity : enhancement
3963 Description: Introduced Access control lists
3964 Details : clients can set ACLs on files and directories in order to have
3965 more fine-grained permissions than the standard Unix UGO+RWX.
3966 The MDS must be started with the "-o acl" mount option.
3968 Severity : enhancement
3969 Description: Introduced filesystem quotas
3970 Details : Administrators may now establish per-user quotas on the
3973 Severity : enhancement
3975 Description: Configuration change for the XT3
3976 The PTLLND is now used to run Lustre over Portals on the XT3
3977 The configure option(s) --with-cray-portals are no longer used.
3978 Rather --with-portals=<path-to-portals-includes> is used to
3979 enable building on the XT3. In addition to enable XT3 specific
3980 features the option --enable-cray-xt3 must be used.
3985 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3986 Details : A race between allocating a new llcd and re-getting the llcd_lock
3987 allowed another thread to grab newly-allocated llcd.
3989 Severity : enhancement
3991 Description: 2.6 OST async journal commit and locking fix to improve performance
3992 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3993 2.6 kernels are now async as they already were in 2.4 kernels so
3994 that they can commit concurrently with the network bulk transfer.
3995 For block-allocated files the filter allocation semaphore is held
3996 to avoid filesystem fragmentation during allocation. BKL lock
3997 removed for 2.6 xattr operations where it is no longer needed.
4002 Description: lconf incorrectly determined whether two IP networks could talk
4003 Details : In some more complicated routing and multiple-network
4004 configurations, lconf will avoid trying to make a network
4005 connection to a disjoint part of the IP space. It was doing the
4006 math incorrectly for one set of cases.
4011 Description: Fix for potential infinite loop processing records in an llog.
4012 Details : If an llog record is corrupted/zeroed, it is possible to loop
4013 forever in llog_process(). Validate the llog record length
4014 and skip the remainder of the block on error.
4017 Frequency : occasional (liblustre only)
4019 Description: liblustre could not open files whose last component is a symlink
4020 Details : sysio_path_walk() would incorrectly pass the open intent to
4021 intermediate path components.
4024 Frequency : rare (liblustre only with non-standard tuning)
4025 Bugzilla : 7201 (7350)
4026 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4027 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4028 after it has been released, only do this for referenced locks
4030 Severity : enhancement
4032 Description: specifying an (invalid) directory default stripe_size of -1
4033 would reset the directory default striping
4034 Details : stripe_size -1 was used internally to signal directory stripe
4035 removal, now use "all default" to signal dir stripe removal
4036 as a directory striping of "all default" is not useful
4039 Frequency : common for large clusters running liblustre clients
4041 Description: doing an ls when liblustre clients are running is slow
4042 Details : sending a glimpse AST to a liblustre client waits for every AST
4043 to time out, as liblustre clients will not respond. Since they
4044 cannot cache data we refresh the OST lock LVB from disk instead.
4046 Severity : enhancement
4048 Description: doing an ls at the same time as file IO can be slow
4049 Details : enqueue and other "small" requests can be blocked behind many
4050 large IO requests. Create a new OST IO portal for non-IO
4051 requests so they can be processed faster.
4054 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4056 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4057 created file with mode 000
4058 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4059 allow an ftruncate() as a non-root user to a file with mode 000.
4060 Lustre now respects this flag to disable mode checks when
4061 truncating a file owned by the user
4064 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4067 Description: Revoking locks from clients that went dead or catatonic might take
4069 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4070 cancellation of such locks instant on servers without waiting for
4071 any reply from clients. Clients drops these locks when cancel
4072 notification from server is received without replying.
4075 Frequency : liblustre-only, when liblustre client dies or becomes busy
4077 Description: Doing ls on Linux clients can take a long time with active
4079 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4080 granting such locks to it in the first place if possible. Locks
4081 are taken by proxy on the OST during the read or write and
4082 dropped immediately afterward. Add connect flags handling, do
4083 not grant locks to liblustre clients for glimpse ASTs.
4085 Severity : enhancement
4087 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4088 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4089 sequential reads and behaves badly with applications doing random
4090 reads. Improve it by reading ahead at least read region, and
4091 avoiding excessive large RPC for small reads.
4093 Severity : enhancement
4095 Description: Creating more than 1000 files for a single job may cause a load
4096 imbalance on the OSTs if there are also a large number of OSTs.
4097 Details : qos_prep_create() uses an OST index reseed value that is an
4098 even multiple of the number of available OSTs so that if the
4099 reseed happens in the middle of the object allocation it will
4100 still utilize the OSTs as uniformly as possible.
4105 Description: OST or MDS may oops in ping_evictor_main()
4106 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4107 but doesn't restart at beginning of obd_exports_timed list
4110 Severity : enhancement
4112 Description: improve by-nid export eviction on the MDS and OST
4113 Details : allow multiple exports with the same NID to be evicted at one
4114 time without re-searching the exports list.
4117 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4119 Description: MDS may oops in groups_free()
4120 Details : in rare race conditions a newly allocated group_info struct is
4121 freed again, and this can be NULL. The 2.4 compatibility code
4122 for groups_free() checked for a NULL pointer, but 2.6 did not.
4125 Frequency : common for liblustre clients doing little filesystem IO
4126 Bugzilla : 9352, 7313
4127 Description: server may evict liblustre clients accessing contended locks
4128 Details : if a client is granted a lock or receives a completion AST
4129 with a blocking AST already set it would not reply to the AST
4130 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4133 Frequency : lfs setstripe, only systems with more than 160 OSTs
4135 Description: unable to set striping with a starting offset beyond OST 160
4136 Details : llapi_create_file() incorrectly limited the starting stripe
4137 index to the maximum single-file stripe count.
4140 Frequency : LDAP users only
4142 Description: lconf did not handle in-kernel recovery with LDAP properly
4143 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4145 Severity : enhancement
4147 Description: bind OST threads to NUMA nodes to improve performance
4148 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4149 node and do their allocations there to localize memory access
4151 Severity : enhancement
4153 Description: llmount can determine client NID directly from Myrinet (GM)
4154 Details : the client NID code from gmnalnid was moved directly into
4155 llmount, removing the need to use this or specifying the
4156 client NID explicitly when mounting GM clients with zeroconf
4159 Frequency : if client is started with down MDS
4161 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4162 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4163 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4164 ptlrpc_import_delayed_req() because we didn't check if the
4165 request was interrupted, and we also didn't break out of the
4166 event loop if there was no timeout
4171 Description: data loss during non-page-aligned writes to a single file from
4172 both multiple nodes and multiple threads on one node at same time
4173 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4174 inconsistency led to false short-reads, that were cached and later
4175 used by ->prepare_write() to fill in partially written page,
4176 leading to data loss.
4179 Frequency : always, if lconf --abort_recovery used
4181 Description: lconf --abort_recovery fails with 'Operation not supported'
4182 Details : lconf was attempting to abort recovery on the MDT device and not
4185 Severity : enhancement
4187 Description: remove cleanup logs
4188 Details : replace lconf-generated cleanup logs with lustre internal
4189 cleanup routines. Eliminates the need for client-cleanup and
4192 Severity : enhancement
4194 Description: add support for EAs (user and system) on lustre filesystems
4195 Details : it is now possible to store extended attributes in the Lustre
4196 client filesystem, and with the user_xattr mount option it
4197 is possible to allow users to store EAs on their files also
4199 Severity : enhancement
4201 Description: Add possibility (config option) to show minimal available OST free
4203 Details : When compiled with --enable-mindf configure option, statfs(2)
4204 (and so, df) will return least minimal free space available from
4205 all OSTs as amount of free space on FS, instead of summary of
4206 free spaces of all OSTs.
4208 Severity : enhancement
4210 Description: do not expand extent locks acquired on OST-side
4211 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4212 by server: they are not cached anyway.
4215 Frequency : when mmap is used/binaries executed from Lustre
4217 Description: Unmmap pages before throwing them away from read cache.
4218 Details : llap_shrink cache now attempts to unmap pages before discarding
4219 them (if unmapping failed - do not discard). SLES9 kernel has
4220 extra checks that trigger if this unmapping is not done first.
4225 Description: lconf didn't resolve symlinks before checking to see whether a
4226 given mountpoint was already in use
4229 Frequency : when migrating failover services
4230 Bugzilla : 6395, 9514
4231 Description: When migrating a subset of services from a node (e.g. failback
4232 from a failover service node) the remaining services would
4233 time out and evict clients.
4234 Details : lconf --force (implied by --failover) sets the global obd_timeout
4235 to 5 seconds in order to quickly disconnect, but this caused
4236 other RPCs to time out too quickly. Do not change the global
4237 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4239 Severity : enhancement
4240 Frequency : if MDS is started with down OST
4241 Bugzilla : 9439,5706
4242 Description: Allow startup/shutdown of an MDS without depending on the
4243 availability of the OSTs.
4244 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4245 Add appropriate locking and lov-osc refcounts for safe
4246 cleaning. Add osc abort_inflight calls in case the
4247 synchronize never started.
4250 Frequency : occasional (Cray XT3 only)
4252 Description: root not authorized to access files in CRAY_PORTALS environment
4253 Details : The client process capabilities were not honoured on the MDS in
4254 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4255 been accessed by an authorized user then root was able to access
4256 the file on the local client also. The root user capabilities
4257 are now allowed on the MDS, as this environment has secure UID.
4260 Frequency : occasional
4262 Description: ldiskfs "too long searching" message happens too often
4263 Details : A debugging message (otherwise harmless) prints too often on
4264 the OST console. This has been reduced to only happen when
4265 there are fragmentation problems on the filesystem.
4270 Description: Division by zero in statfs when all OSCs are inactive
4271 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4272 lov_statfs divides by value returned by lov_get_stripecnt().
4276 Bugzilla : 9489, 3273
4277 Description: First write from each client to each OST was only 4kB in size,
4278 to initialize client writeback cache, which caused sub-optimal
4279 RPCs and poor layout on disk for the first writen file.
4280 Details : Clients now request an initial cache grant at (re)connect time
4281 and so that they can start streaming writes to the cache right
4282 away and always do full-sized RPCs if there is enough data.
4283 If the OST is rebooted the client also re-establishes its grant
4284 so that client cached writes will be honoured under the grant.
4289 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4290 Details : Now I/O RPCs go to different portal number and (presumably) fast
4291 lock requests (and glimses) and other RPCs get their own service
4292 threads pool that should be able to service those RPCs
4295 Severity : enhancement
4297 Description: Ability to exchange lustre version between client and servers and
4298 issue warnings at client side if client is too old. Also for
4299 liblustre clients there is ability to refuse connection of too old
4301 Details : New 'version' field is added to connect data structure that is
4302 filled with version info. That info is later checked by server and
4306 Frequency : rare, liblustre only.
4307 Bugzilla : 9296, 9581
4308 Description: Two simultaneous writes from liblustre at offset within same page
4309 might proceed at the same time overwriting eachother with stale
4311 Details : I/O lock withing llu_file_prwv was released too early, before data
4312 actually was hitting the wire. Extended lock-holding time until
4313 server acknowledges receiving data.
4316 Frequency : extremely rare. Never observed in practice.
4318 Description: avoid generating lustre_handle cookie of 0.
4319 Details : class_handle_hash() generates handle cookies by incrementing
4320 global counter, and can hit 0 occasionaly (this is unlikely, but
4321 not impossible, because initial value of cookie counter is
4322 selected randonly). Value of 0 is used as a sentinel meaning
4323 "unassigned handle" --- avoid it. Also coalesce two critical
4324 sections in this function into one.
4326 Severity : enhancement
4328 Description: allow liblustre clients to delegate truncate locking to OST
4329 Details : To avoid overhead of locking, liblustre client instructs OST to
4330 take extent lock in ost_punch() on client's behalf. New connection
4331 flag is added to handle backward compatibility.
4333 Severity : enhancement
4334 Bugzilla : 4928, 7341, 9758
4335 Description: allow number of OST service threads to be specified
4336 Details : a module parameter allows the number of OST service threads
4337 to be specified via "options ost ost_num_threads={N}" in the
4338 OSS's /etc/modules.conf or /etc/modprobe.conf.
4342 Bugzilla : 6146, 9635, 9895
4343 Description: servers crash with bad pointer in target_handle_connect()
4344 Details : In rare cases when a client is reconnecting it was possible that
4345 the connection request was the last reference for that export.
4346 We would temporarily drop the export reference and get a new
4347 one, but this may have been the last reference and the export
4348 was just destroyed. Get new reference before dropping old one.
4350 Severity : enhancement
4351 Frequency : if client is started with failover MDS
4353 Description: Allow multiple MDS hostnames in the mount command
4354 Details : Try to read the configuration from all specified MDS
4355 hostnames during a client mount in case the "primary"
4358 Severity : enhancement
4360 Description: Stop sending data to evicted clients as soon as possible.
4361 Details : Check if the client we are about to send or are sending data to
4362 was evicted already. (Check is done every second of waiting,
4363 for which l_wait_event interface was extended to allow checking
4364 of exit condition at specified intervals).
4367 Frequency : rare, normally only when NFS exporting is done from client
4369 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4371 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4374 Frequency : always, if srand() is called before liblustre initialization
4376 Description: Liblustre uses system PRNG disturbing its usage by user application
4377 Details : Introduce internal to lustre fast and high-quality PRNG for
4378 lustre usage and make liblustre and some other places in generic
4379 lustre code to use it.
4381 Severity : enhancement
4382 Bugzilla : 9477, 9557, 9870
4383 Description: Verify that the MDS configuration logs are updated when xml is
4384 Details : Check if the .xml configuration logs are newer than the config
4385 logs stored on the MDS and report an error if this is the case.
4386 Request --write-conf, or allow starting with --old_conf.
4388 Severity : enhancement
4390 Description: Handle symlinks in the path when checking if Lustre is mounted.
4391 Details : Resolve intermediate symlinks when checking if a client has
4392 mounted a filesystem to avoid duplicate client mounts.
4397 Description: lconf can hit an error exception but still return success.
4398 Details : The lconf command catches the Command error exception at the top
4399 level script context and will exit with the associated exit
4400 status, but doesn't ensure that this exit status is non-zero.
4405 Description: failure of ptlrpc thread startup can cause oops
4406 Details : Starting a ptlrpc service thread can fail if there are a large
4407 number of threads or the server memory is very fragmented.
4408 Handle this without oopsing.
4411 Frequency : always, only if liblustre and non-default acceptor port was used
4413 Description: liblustre cannot connect to servers with non-default acceptor port
4414 Details : tcpnal_set_default_params() was not called and was therefore
4415 ignoring the environment varaible TCPNAL_PORT, as well as other
4416 TCPNAL_ environment variables
4421 Description: two objects could be created on the same OST for a single file
4422 Details : If an OST is down, in some cases it was possible to create two
4423 objects on a single OST for a single file. No problems other
4424 than potential performance impact and spurious error messages.
4428 Bugzilla : 5681, 9562
4429 Description: Client may oops in ll_unhash_aliases
4430 Details : Client dcache may become inconsistent in race condition.
4431 In some cases "getcwd" can fail if the current directory is
4437 Description: Inode refcounting problems in NFS export code
4438 Details : link_raw functions used to call d_instantiate without obtaining
4439 extra inode reference first.
4443 Bugzilla : 9942, 9903
4444 Description: Referencing freed requests leading to crash, memleaks with NFS.
4445 Details : We used to require that call to ll_revalidate_it was always
4446 followed by ll_lookup_it. Also with revalidate_special() it is
4447 possible to call ll_revalidate_it() twice for the same dentry
4448 even if first occurence returned success. This fix changes semantic
4449 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4450 reference on a request referred from the intent.
4451 ll_intent_release() then releases such a request.
4454 Frequency : rare, normally benchmark loads only
4456 Description: unlinked inodes were kept in memory on the client
4457 Details : If a client is repeatedly creating and unlinking files it
4458 can accumulate a lot of stale inodes in the inode slab cache.
4459 If there is no other client load running this can cause the
4460 client node to run out of memory. Instead flush old inodes
4461 from client cache that have the same inode number as a new inode.
4464 Frequency : SLES9 2.6.5 kernel and long filenames only
4465 Bugzilla : 9969, 10379
4466 Description: utime reports stale NFS file handle
4467 Details : SLES9 uses out-of-dentry names in some cases, which confused
4468 the lustre dentry revalidation. Change it to always use the
4472 Frequency : rare, unless heavy write-truncate concurrency is continuous
4473 Bugzilla : 4180, 6984, 7171, 9963, 9331
4474 Description: OST becomes very slow and/or deadlocked during object unlink
4475 Details : filter_destroy() was holding onto the parent directory lock
4476 while truncating+unlinking objects. For very large objects this
4477 may block other threads for a long time and slow overall OST
4478 responsiveness. It may also be possible to get a lock ordering
4479 deadlock in this case, or run out of journal credits because of
4480 the combined truncate+unlink. Solution is to do object truncate
4481 first in one transaction without parent lock, and then do the
4482 final unlink in a new transaction with the parent lock. This
4483 reduces the lock hold time dramatically.
4486 Frequency : rare, 2.4 kernels only
4488 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4489 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4490 before unmounting their filesystem in order to avoid deadlock.
4491 The kernel_locked() function in 2.4 kernels only checks whether
4492 the kernel lock is held, not whether it is this process that is
4493 holding it as 2.6 kernels do.
4498 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4499 Details : The client ptlrpc code may be trying to reconnect to a down
4500 server before a previous connection attempt has timed out.
4501 Increase the reconnect interval to be longer than the connection
4502 timeout interval to avoid sending duplicate connections to
4506 Frequency : echo_client brw_test command
4508 Description: fix echo_client to work with OST preallocated code
4509 Details : OST preallocation code (5137) didn't take echo_client IO path
4510 into account: echo_client calls filter methods outside of any
4511 OST thread and, hence, there is no per-thread preallocated
4512 pages and buffers to use. Solution: hijack pga pages for IO. As
4513 a byproduct, this avoids unnecessary data copying.
4517 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4518 Description: Client can oops in mdc_commit_close() after open replay
4519 Details : It was possible for the MDS to return an open request with no
4520 transaction number in mds_finish_transno() if the client was
4521 evicted, but without actually returning an error. Clients
4522 would later try to replay that open and may trip an assertion
4523 Simplify the client close codepath, and always return an error
4524 from the MDS in case the open is not successful.
4527 Frequency : rare, 2.6 OSTs only
4529 Description: OST may deadlock under high load on fragmented files
4530 Details : If there was a heavy load and highly-fragmented OST filesystems
4531 it was possible to have all the OST threads deadlock waiting on
4532 allocation of biovecs, because the biovecs were not released
4533 until the entire RPC IO was completed. Instead, release biovecs
4534 as soon as they are complete to ensure forward IO progress.
4536 Severity : enhancement
4538 Description: Support for specifying external journal device at mount
4539 Details : If an OST or MDS device is formatted with an external journal
4540 device, this device major/minor is stored in the ext3 superblock
4541 and may not be valid for failover. Allow detecting and
4542 specifying the external journal at mount time.
4547 Description: Mounting an MDS with pending unlinked files may cause oops
4548 Details : target_finish_recovery() calls mds_postrecov() which returned
4549 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4550 considers this an error and immediately begins cleaning up the
4551 lov, just after starting the mds_lov process
4553 Severity : enhancement
4555 Description: Implement 'lfs df' to report actual free space on per-OST basis
4556 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4557 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4558 usage of objects; '-h' to report in human readable format.
4560 ------------------------------------------------------------------------------
4562 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4569 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4570 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4571 at startup during orphan unlinking. Instead, skip buddy update
4572 before setup, as it will be regenerated after recovery is complete.
4577 Description: If an OST is inactive, its locks might reference stale inodes.
4578 Details : lov_change_cbdata() must iterate over all namespaces, even if
4579 they are inactive to clear inode references from the lock.
4581 Severity : enhancement
4582 Frequency : occasional, if non-standard max_dirty_mb used
4584 Description: Client will block write RPCs if not enough grant
4585 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4586 then the client will block writes while waiting for another RPC
4587 to complete instead of consuming its dirty limit. With change
4588 we get improved performance when max_dirty_mb is small.
4590 Severity : enhancement
4591 Bugzilla : 3389, 6253
4592 Description: Add support for supplementary groups on the MDS.
4593 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4594 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4595 lookups for user supplementary groups into a cache.
4599 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4600 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4601 for liblustre O_EXCL is enforced by MDS.
4605 Description: readdir never returns NULL in liblustre.
4606 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4607 dentry in struct dirent.
4611 Frequency : liblustre only, depends on application IO pattern
4612 Description: liblustre clients evicted if not contacting servers
4613 Details : Don't put liblustre clients into the ping_evictor list, so
4614 they will not be evicted by the pinger ever.
4616 Severity : enhancement
4618 Description: Add ability to evict clients by NID from MDS.
4619 Details : By echoing "nid:$NID" string into
4620 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4621 $NID would be instantly evicted from this MDS and from all active
4622 OSTs connected to it.
4626 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4627 Details : lookup_it_finish() used to query file size from OSTs that was not
4632 Description: service threads change working directory to that of init
4633 Details : Starting lustre service threads may pin the working directory
4634 of the parent thread, making that filesystem busy. Threads
4635 now change to the working directory of init to avoid this.
4639 Frequency : during shutdown only
4640 Description: shutdown with a failed MDS or OST can cause unmount to hang
4641 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4646 Frequency : 2.6 only
4647 Description: chmod/chown may include an extra supplementary group
4648 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4649 supplementary group and if none is specified this is used.
4652 Bugzilla : 5479 (6816)
4654 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4655 Details : If lookup is in progress on a file that is unlinked we might try
4656 to revalidate the inode and fail in revalidate after lookup is
4657 complete and ll_file_open() enqueues the open again but
4658 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4661 Frequency : always, if lconf --abort_recovery used
4663 Description: lconf --abort_recovery fails with 'Operation not supported'
4664 Details : lconf was attempting to abort recovery on the MDT device and not
4667 ------------------------------------------------------------------------------
4669 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4674 Frequency : rare (only unsupported configurations with a node running as an
4676 Bugzilla : 6514, 5137
4677 Description: Mounting a Lustre file system on a node running as an OST could
4679 Details : OSTs now preallocates memory needed to write out data at
4680 startup, instead of when needed, to avoid having to
4681 allocate memory in possibly low memory situations.
4682 Specifically, if the file system is mounted on on OST,
4683 memory pressure could force it to try to write out data,
4684 which it needed to allocate memory to do. Due to the low
4685 memory, it would be unable to do so and the node would
4686 become unresponsive.
4688 Severity : enhancement
4690 Description: Addition of lconf --service command line option
4691 Details : lconf now accepts a '--service <arg>' option, which is
4692 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4694 Severity : enhancement
4696 Description: Failover mode is now the default for OSTs.
4697 Details : By default, OSTs will now run in failover mode. To return to
4698 the old behaviour, add '--failout' to the lmc line for OSTs.
4700 Severity : enhancement
4702 Description: Health checks are now provided for MDS and OSTs
4703 Details : Additional detailed health check information on MSD and OSTs
4704 is now provided through the procfs health_check value.
4707 Frequency : occasional, depends on IO load
4709 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4710 after numerous create/delete cycles
4711 Details : The ext3 inode allocation policy would not allocate new inodes
4712 very well on the OSTs because there are no new directories
4713 being created. Instead we look for groups with free space if
4714 the parent directories are nearly full.
4719 Description: Network or server problems during mount may cause partially
4720 mounted clients instead of returning an error.
4721 Details : The config llog parsing code may overwrite the error return
4722 code during mount error handling, returning success instead
4728 Description: MDS can fail to allocate large reply buffers
4729 Details : After long uptimes the MDS can fail to allocate large reply
4730 buffers (e.g. zconf client mount config records) due to memory
4731 fragmentation or consumption by the buffer cache. Preallocate
4732 some large reply buffers so that these replies can be sent even
4733 under memory pressure.
4737 Frequency : rare (liblustre)
4738 Description: fsx running with liblustre complained that using truncate() to
4739 extend the file doesn't work. This patch corrects that issue.
4740 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4741 ATTR_SIZE and lsm use in llu_setattr_raw.
4745 Frequency : rare, only 2.6 kernels
4746 Description: Unusual file access patterns on the MDS may result in inode
4747 data being lost in very rare circumstances.
4748 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4749 code in the 2.6 kernel caused the inode and/or EA data not to
4750 be read from disk, causing single-file corruption.
4754 Frequency : rare, only 2.6 filesystems using extents
4755 Description: Heavy concurrent write and delete load may cause data corruption.
4756 Details : It was possible under high-load situations to have an extent
4757 metadata block in the block device cache from a just-unlinked
4758 file overwrite a newly-allocated data block. We now unmap any
4759 metadata buffers that alias just-allocated data blocks.
4763 Frequency : filesystems with default stripe_count larger than 77
4764 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4765 Details : lconf specifies an inode size of 4096 bytes when the default
4766 stripe_count is larger than 77. This conflicts with the default
4767 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4768 this case to avoid pinning too much memory for large EAs.
4770 ------------------------------------------------------------------------------
4772 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4777 Frequency : rare (extremely heavy IO load with hundreds of clients)
4779 Description: Client is evicted, gets IO error writing to file
4780 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4781 caused clients to be evicted for AST timeouts. The fixes in
4782 bug 5192 mean we no longer need to have such short AST timeouts
4783 so ldlm_timeout has been increased.
4786 Frequency : occasional during --force or --failover shutdown under load
4787 Bugzilla : 5949, 4834
4788 Description: Server oops/LBUG if stopped with --force or --failover under load
4789 Details : a collection of import/export refcount and cleanup ordering
4790 issues fixed for safer force cleanup
4793 Frequency : only filesystems larger than 120 OSTs
4794 Bugzilla : 5990, 6223
4795 Description: lfs getstripe would oops on a very large filesystem
4796 Details : lov_getconfig used kfree on vmalloc'd memory
4799 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4800 Bugzilla : 6242, 6243
4801 Description: reading from files that had been truncated to a non-zero size
4802 but never opened returned no data
4803 Details : ll_file_read() reads zeros from no-object files to EOF
4808 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4809 Details : There's little harm in aborting MDS/OSS recovery and letting it
4810 try again, so I removed the LASSERT and return an error instead.
4812 Severity : enhancement
4814 Description: New debugging infrastructure for tracking down data corruption
4815 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4816 (b) cover more of the client-side code path, and (c) try to narrow
4817 down where problems occurred
4821 Bugzilla : 3819, 4364, 4397, 6313
4822 Description: Racing close and eviction MDS could cause assertion in mds_close
4823 Details : It was possible to get multiple mfd references during close and
4824 client eviction, leading to one thread referencing a freed mfd.
4826 Severity: : enhancement
4827 Bugzilla : 3262, 6359
4828 Description: Attempts to reconnect to servers are now more aggressive.
4829 Details : This builds on the enhanced upcall-less recovery that was added
4830 in 1.4.2. When trying to reconnect to servers, clients will
4831 now try each server in the failover group every 10 seconds. By
4832 default, clients would previously try one server every 25 seconds.
4837 Description: After recovery, certain operations trigger a failed
4838 assertion on a client.
4839 Details : Failing over an mds, using lconf -d --failover, while a
4840 client was doing a readdir() call would cause the client to
4841 LBUG after recovery completed and the readdir() was resent.
4843 Severity : enhancement
4845 Description: Default groups are now added by lconf
4846 Details : You can now run lconf --group <servicename> without having to
4847 manually add groups with lmc.
4850 Frequency : occasional
4852 Description: Nodes with an elan id of 0 trigger a failed assertion
4855 Frequency : always when accessing e.g. tty/console device nodes
4857 Description: tty and some other devices nodes cannot be used on lustre
4858 Details : file's private_data field is used by device data and lustre
4859 values in there got lost. New field was added to struct file to
4860 store fs-specific private data.
4863 Frequency : when exporting Lustre via NFS
4865 Description: NFSD failed occasionally when looking up a path component
4866 Details : NFSD is looking up ".." which was broken in ext3 directories
4867 that had grown large enough to become hashed.
4870 Frequency : Clusters with multiple interfaces not on the same subnet
4872 Description: Nodes will repeatedly try to reconnect to an interface which it
4873 cannot reach and report an error to the log.
4874 Details : Extra peer list entries will be created by lconf with some peers
4875 unreachable. lconf now validates the peer before adding it.
4878 Frequency : Only if a default stripe is set on the filesystem root.
4880 Description: Setting a default stripe on the filesystem root prevented the
4881 filesystem from being remounted.
4882 Details : The client was sending extra request flags in the root getattr
4883 request and did not allocate a reply buffer for the dir EA.
4886 Frequency : occasional, higher if lots of files are accessed by one client
4887 Bugzilla : 6159, 6097
4888 Description: Client trips assertion regarding lsm mismatch/magic
4889 Details : While revalidating inodes the VFS looks up inodes with ifind()
4890 and in rare cases can find an inode that is being freed.
4891 The ll_test_inode() code will free the lsm during ifind()
4892 when it finds an existing inode and then the VFS later attaches
4893 this free lsm to a new inode.
4897 Bugzilla : 6422, 7030
4898 Description: MDS deadlock between mkdir and client eviction
4899 Details : Creating a new file via mkdir or mknod (starting a transaction
4900 and getting the ns lock) can deadlock with client eviction
4901 (gets ns lock and trying to finish a synchronous transaction).
4904 Frequency : occasional
4905 Description: While starting a server, the fsfilt_ext3 module could not be
4907 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4908 kernels. Previously, lconf would still use the ext3 name
4909 when trying to load modules. Now, it will correctly use
4910 ext3 on 2.4 and ldiskfs on 2.6.
4912 Severity : enhancement
4913 Description: The default stripe count has been changed to 1
4914 Details : The interpretation of the default stripe count (0, to lfs
4915 or lmc) has been changed to mean striping across a single
4916 OST, rather than all available. For general usage we have
4917 found a stripe count of 1 or 2 works best.
4919 Severity : enhancement
4920 Description: Add support for compiling against Cray portals.
4921 Details : Conditional compiling for some areas that are different
4925 Frequency : occasional
4926 Bugzilla : 6409, 6834
4927 Description: Creating files with an explicit stripe count may lead to
4928 a failed assertion on the MDS
4929 Details : If some OSTs are full or unavailable, creating files may
4930 trigger a failed assertion on the MDS. Now, Lustre will
4931 try to use other servers or return an error to the
4935 Frequency : occasional
4937 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4938 caused lock timeout during readahead (since 1.4.2).
4939 Details : Processes doing readahead might match a lock that hasn't been
4940 granted yet if there are overlapping and conflicting lock
4941 requests. The readahead process waits on ungranted lock
4942 (original lock is CBPENDING), while OST waits for that process
4943 to cancel CBPENDING read lock and eventually evicts client.
4945 Severity : enhancement
4947 Description: Initial enabling of flock support for clients
4948 Details : Implements fcntl advisory locking and file status functions.
4949 This feature is provided as an optional mount flag (default
4950 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4951 locking are implemented yet, and those that are are not guaranteed
4952 to be completely correct in production environments.
4953 mount -t lustre -o [flock|noflock] ...
4956 Frequency : occasional
4958 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4959 assertion in the ext3 JBD (journaling) layer.
4960 Details : The b_committed_data struct is protected by the big kernel lock
4961 in 2.4 kernels, serializing journal_commit_transaction() and
4962 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4963 access to this struct. In 2.6 kernels there is finer grained
4964 locking to improve SMP performance of the JBD layer.
4968 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4973 Description: merge and cleanup kernel patches.
4976 -----------------------------------------------------------------------------
4978 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4980 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4981 versions of Lustre. Please update all servers and clients to
4982 version 1.4.2 or later at the same time. You must also run
4983 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4984 to update the configuration logs.
4986 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4987 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4988 - swab llog records of type '0' so we get proper header size/idx (5861)
4989 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4990 - fix rename of one directory over another leaking an inode (5953)
4991 - avoid SetPageDirty on 2.6 (5981)
4992 - don't re-add just-being-destroyed locks to the waiting list (5653)
4993 - when creating new directories, inherit the parent's custom
4994 striping settings if present parent (3048)
4995 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4996 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4997 - don't hold client locks on temporary worklist from l_lru (5666)
4998 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4999 - automatically evict dead clients (5921)
5000 - Update file size properly in create+truncate+fstat case (6196)
5001 - Do not unhash mountpoint dentries, do not allow removal of
5003 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5004 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5005 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5006 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5007 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5008 - Correctly handle reads of files with no objects (6243)
5009 - lctl recover will also mark a device active if deactivate used (5933)
5011 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5012 - allow --write-conf on an MDS with different nettype than client (5619)
5013 - don't write config llogs to MDS for mounts not from that MDS (5617)
5014 - lconf should create multiple TCP connections from a client (5201)
5015 - init scripts are now turned off by default; run chkconfig --on
5016 lustre and chkconfig --on lustrefs to use them
5017 - upcalls are no longer needed for clients to recover to failover
5019 - add --abort-recovery option to lconf to abort recovery on device
5021 - add support for an arbitrary number of OSTs (3026)
5022 - Quota support protocol changes.
5023 - forward compatibility changes to wire structs (6007)
5024 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5025 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5026 - improved "lustre status" script
5027 - initialize blocksize for non-regular files (6062)
5028 - added --disable-server and --disable-client configure options (5782)
5029 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5030 - Vanilla 2.4.29 support
5031 - increase maximum number of obd devices to 520 (6242)
5032 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5033 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5034 SLES 9 2.6.5 kernels
5035 - limit stripes per file to 160 (the maximum EA size) (6093)
5037 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5040 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5041 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5042 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5043 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5044 - small ext3 extents cleanups and fixes (5733)
5045 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5046 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5047 - avoid needless client->OST connect, fix handle mismatch (5317)
5048 - fix DLM error path that led to out-of-sync client, long delays (5779)
5049 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5050 - fix several locking issues related to i_size (5492,5624,5654,5672)
5051 - don't move pending lock onto export if it is already evicted (5683)
5052 - fix kernel oops when creating .foo in unlinked directory (5548)
5053 - fix deadlock in obdfilter statistics vs. object create (5811)
5054 - use time_{before,after} to avoid timer jiffies wrap (5882)
5055 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5056 - Do not leak request if server was not able to process it (5154)
5057 - If mds_open unable to find parent dir, make that negative lookup(5154)
5058 - don't create new directories with extent-mapping (5909, 5936)
5060 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5061 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5062 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5063 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5064 - support for RHEL4 kernel on i686 (5773)
5065 - provide error messages when incompatible logs are encountered (5898)
5067 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5068 * version 1.4.0.10 (1.4.1 release candidate 1)
5070 - don't keep a lock reference when lock is not granted (4238)
5071 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5072 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5073 - drop import inflight refcount on signal_completed_replay error (5255)
5074 - unlock page after async write error during send (3677)
5075 - handle missing objects in filter_preprw_read properly (5265)
5076 - no transno return for symlink open, don't save no-trasno open (3440)
5077 - don't try to complete elan receive that already failed (4012)
5078 - free RPC server reply state on error (5406)
5079 - clean up thread from ptlrpc_start_thread() on error (5160)
5080 - readahead could read extra page into cache that wasn't ejected (5388)
5081 - prevent races in class_attach/setup/cleanup/detach (5260)
5082 - don't dereference de->d_inode after l_dput of de (5458)
5083 - use "int" for stripe value returned from lock_to_stripe (5544)
5084 - mballoc allocation and error-checking fixes in 2.6 (5504)
5085 - block device patches to fix I/O request sizes in 2.6 (5482)
5086 - look up hostnames for IB nals (5602)
5087 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5088 - don't start multiple acceptors for the same port (5277)
5089 - fix incorrect LASSERT in mds_getattr_name (5635)
5090 - export a proc file for general "ping" checking (5628)
5091 - fix "lfs check" to not block when the MDS is down (5628)
5093 - service request history (4965)
5094 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5095 - create an "evict_client" /proc entry on OSTs, like the MDS has
5096 - fix mount usage message, return errors per mount(8) (5168)
5097 - change grep [] to grep "[]" in tests so they work in more UMLs
5098 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5099 - remove ancient LOV_MAGIC_V0 EA support (5047)
5100 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5101 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5102 - allow readahead allocations to fail when low on memory (5383)
5103 - mmap locking landed again, after considerable improvement (2828)
5104 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5106 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5109 - send OST transaction number in read/write reply to free req (4966)
5110 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5111 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5112 - the watchdog thread now runs as interruptible (5246)
5113 - flock/lockf fixes (but it's still disabled, pending 5135)
5114 - don't use EXT3 constants in llite code (5094)
5115 - memory shortage at startup could cause assertion (5176)
5117 - reorganization of lov code
5118 - single portals codebase
5120 - add extents/mballoc support (5025)
5121 - direct I/O reads in the obdfilter (4048)
5122 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5124 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5127 - send OST transaction number in read/write reply to free req (4966)
5128 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5129 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5130 - the watchdog thread now runs as interruptible (5246)
5131 - handle missing objects in filter_preprw_read properly (5265)
5132 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5133 - drop import inflight refcount on signal_completed_replay error (5255)
5134 - unlock page after async write error during send (3677)
5135 - return original error code on reconstructed replies (3761)
5136 - no transno return for symlink open, don't save no-trasno open (3440)
5138 - add pid to ldlm debugging output (4922)
5139 - bump the watchdog timeouts -- we can't handle 30sec yet
5140 - extra debugging for orphan dentry/inode bug (5259)
5142 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5145 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5146 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5147 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5148 - fix NULL dereference in /proc/sys/portals/routes (4827)
5149 - allow failed mdc_close() operations to be interrupted (4561)
5150 - stop precreate on OST before MDS would time out on it (4778)
5151 - don't send partial-page writes before EOF from client (4410)
5152 - discard client grant for sub-page writes on large-page clients (4520)
5153 - don't free dentries not owned by NFS code, check generation (4806)
5154 - fix lsm leak if mds_create_objects() fails (4801)
5155 - limit debug_daemon file size, always print CERROR messages (4789)
5156 - use transno after validating reply (3892)
5157 - process timed out requests if import state changes (3754)
5158 - update mtime on OST during writes, return in glimpse (4829)
5159 - add mkfsoptions to LDAP (4679)
5160 - use ->max_readahead method instead of zapping global ra (5039)
5161 - don't interrupt __l_wait_event() during strace
5163 - add software watchdogs to catch hung threads quickly (4941)
5164 - make lustrefs init script start after nfs is mounted
5165 - fix CWARN/ERROR duplication (4930)
5166 - return async write errors to application if possible (2248)
5167 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5168 - print NAL number in %x format (4645)
5169 - update barely-supported suse-2.4.21-171 series (4842)
5170 - support for sles 9 %post scripts
5171 - support for building 2.6 kernel-source packages
5172 - support for sles km_* packages
5174 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5177 - ignore -ENOENT errors in osc_destroy (3639)
5178 - notify osc create thread that OSC is being cleaned up (4600)
5179 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5180 - reconstruct ost_handle() like mds_handle() (4657)
5181 - create a new thread to do import eviction to avoid deadlock (3969)
5182 - let lconf resolve symlinked-to devices (4629)
5183 - don't unlink "objects" from directory with default EA (4554)
5184 - hold socknal file ref over connect in case target is down (4394)
5185 - allow more than 32000 subdirectories in a single directory (3244)
5186 - fix blocks count for O_DIRECT writes (3751)
5187 - OST returns ENOSPC from object create when no space left (4539)
5188 - don't send truncate RPC if file size isn't changing (4410)
5189 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5190 - bind to privileged port in socknal and tcpnal (3689)
5192 - rate limit CERROR/CWARN console message to avoid overload (4519)
5193 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5194 - basic mmap support (3918)
5195 - kernel patch series update from b1_4 (4711)
5197 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5200 - avoid crash during MDS cleanup with OST shut down (2775)
5201 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5202 - don't use bad inodes on the MDS (3744)
5203 - dynamic object preallocation to improve recovery speed (4236)
5204 - don't hold spinlock over lock dumping or change debug flags (4401)
5205 - don't zero obd_dev when it is force cleaned (3651)
5206 - print grants to console if they go negative (4431)
5207 - "lctl deactivate" will stop automatic recovery attempts (3406)
5208 - look for existing locks in ldlm_handle_enqueue() (3764)
5209 - don't resolve lock handle twice in recovery avoiding race (4401)
5210 - revalidate should check working dir is a directory (4134)
5212 - don't always mark "slow" obdfilter messages as errors (4418)
5214 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5217 - don't close LustreDB during write_conf until it is done (3860)
5218 - fix typo in lconf for_each_profile (3821)
5219 - allow dumping logs from multiple threads at one time (3820)
5220 - don't allow multiple threads in OSC recovery (3812)
5221 - fix debug_size parameters (3864)
5222 - fix mds_postrecov to initialize import for llog ctxt (3121)
5223 - replace config semaphore with spinlock (3306)
5224 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5225 - don't allow enqueue to complete on a destroyed export (3822)
5226 - down write_lock before checking llog header bitmap (3825)
5227 - recover from lock replay timeout (3764)
5228 - up llog sem before sending rpc (3652)
5229 - reduce ns lock hold times when setting kms (3267)
5230 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5231 - fix NULL deref and obd_dev leak on setup error (3312)
5232 - replace some LBUG about llog ops with error handling (3841)
5233 - don't match INVALID dentries from d_lookup and spin (3784)
5234 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5235 - fix invalid assertion in ptlrpc_set_wait (3880)
5237 - add libwrap support for the TCP acceptor (3996)
5238 - add /proc/sys/portals/routes for non-root route listing (3994)
5239 - allow setting MDS UUID in .xml (2580)
5240 - print the stack of a process that LBUGs (4228)
5242 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5245 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5246 - make sure to unset replay flag from failed open requests (3440)
5247 - if default stripe count is 0, use OST count for inode size (3636)
5248 - update parent mtime/ctime on client for create/unlink (2611)
5249 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5250 - free recovery state on server during a forced cleanup (3571)
5251 - unregister_reply for resent reqs (3063)
5252 - loop back devices mounting and status check on 2.6 (3563)
5253 - fix resource-creation race that can provoke i_size == 0 (3513)
5254 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5255 - more debugging for page-accounting assertion (3746)
5256 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5257 - avoid deadlock after precreation failure (3758)
5258 - fix race and lock order deadlock in orphan handling (3450, 3750)
5259 - add validity checks when grabbing inodes from l_ast_data (3599)
5261 - add /proc/.../recovery_status to obdfilter (3428)
5262 - lightweight CDEBUG infrastructure, debug daemon (3668)
5263 - change default OSC RPC parameters to be better on small clusters
5264 - turn off OST read cache for files smaller than 32MB
5265 - install man pages and include them in rpms (3100)
5266 - add new init script for (un)mounting lustre filesystems (2593)
5267 - run chkconfig in %post for init scripts (3701)
5268 - drop scimac NAL (unmaintained)
5270 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5273 - clean kiobufs before and after use (3485)
5274 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5275 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5276 - change init script to fail more clearly if not run as root (1528)
5277 - allow clients to reconnect during replay (1742)
5278 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5279 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5280 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5281 - fix cond_resched (3554)
5282 - only evict unfinished clients after recovery (3515)
5283 - allow bulk resend, prevent data loss (3570)
5284 - dynamic ptlrpc request buffer allocation (2102)
5285 - don't allow unlinking open directory if it isn't empty (2904)
5286 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5287 - remove extraneous obd dereference causing LASSERT failure (3334)
5288 - don't use get_cycles() when creating temp. files on the mds (3156)
5289 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5290 - handle EEXIST for set-stripe, set proper directory name (3336)
5292 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5293 - fix ksocknal_fmb_callback() error messages (2918)
5295 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5298 - don't copy lvb into (possibly NULL) reply on error (2983)
5299 - don't deref dentry after dput, don't free lvb on error (2922)
5300 - use the kms to determine writeback rpc length (2947)
5301 - increment oti_logcookies when osc is inactive (2948)
5302 - update client's i_blocks count via lvb messages (2543)
5303 - handle intent open/close of special files properly (1557)
5304 - mount MDS with errors=remount-ro, like obdfilter (2009)
5305 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5306 - don't use cancelling-locks' kms values (2947)
5307 - use highest lock extent for kms, not last one (2925)
5308 - don't dereference ERR_PTR() dentry in error handling path (3107)
5309 - fix thread race in portals_debug_dumplog() (3122)
5310 - create lprocfs device entries at setup instead of at attach (1519)
5311 - common AST error handler, don't evict client on completion race (3145)
5312 - zero nameidata in detach_mnt in 2.6 (3118)
5313 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5314 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5315 - initialize RPC timeout timer earlier for 2.6 (3219)
5316 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5317 - print nal/nid for unknown nid (3258)
5318 - additional checks for oscc recovery before doing precreate (3284)
5319 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5320 - don't crash in mdc_close for bad permissions on open (3285)
5321 - zero i_rdev for non-device files (3147)
5322 - clear page->private before handing to FS, better assertion (3119)
5323 - tune the read pipeline (3236)
5324 - fix incorrect decref of invalidated dentry (2350)
5325 - provide read-ahead stats and refine rpc in flight stats (3328)
5326 - don't hold journal transaction open across create RPC (3313)
5327 - update atime on MDS at close time (3265)
5328 - close LDAP connection when recovering to avoid server load (3315)
5329 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5330 - don't leak open file on MDS after open resend (3325)
5331 - serialize filter_precreate and filter_destroy_precreated (3329)
5332 - loop device shouldn't call sync_dev() for nul device (3092)
5333 - clear page cache after eviction (2766)
5334 - resynchronize MDS->OST in background (2824)
5335 - refuse to mount the same filesystem twice on same mountpoint (3394)
5336 - allow llmount to create routes for mounting behind routers (3320)
5337 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5338 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5339 - fix rare problem with rename on htree directories (3417)
5341 - allow default OST striping configuration per directory (1414)
5342 - fix compilation for qswnal for 2.6 kernels (3125)
5343 - increase maximum number of MDS request buffers for large systems
5344 - change liblustreapi to be useful for external progs like lfsck (3098)
5345 - increase local configuration timeout for slow disks (3353)
5346 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5348 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5351 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5352 - don't overwrite extent policy data in reply if lock was blocked (2901)
5353 - drop filter export grants atomically with removal from device (2663)
5354 - del obd_self_export from work_list in class_disconnect_exports (2908)
5355 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5356 - swab reply message in mdc_close, other PPC fixes (2464)
5357 - fix destroying of named logs (2325)
5358 - overwrite old logs when running lconf --write_conf (2264)
5359 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5360 - fix race in target_handle_connect (2898)
5361 - mds_reint_create() should take same inode create lock (2926)
5362 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5363 - don't close files for self_export to avoid uninitialized obd (2936)
5364 - allow MDS with the same name as client node (2939)
5365 - hold dentry reference for closed log files for unlink (2325)
5366 - reserve space for all logs during transactions (2059)
5367 - don't evict page beyond end of stripe extent (2925)
5368 - don't oops on a deleted current working directory (2399)
5369 - handle hard links to targets without a parent properly (2517)
5370 - don't dereference NULL lock when racing during eviction (2867)
5371 - don't grow lock extents when lots of conflicting locks (2919)
5373 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5376 - account for cache space usage on clients to avoid data loss (974)
5377 - lfsck support in lustre kernel code (2349)
5378 - reduce journal credits needed for BRW writes (2370)
5379 - orphan handling to avoid losing space on client/server crashes
5380 - ptlrpcd can be blocked, stopping ALL progress (2477)
5381 - use lock value blocks to assist in proper KMS, faster stat (1021)
5382 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5383 - recovery for initial connections (2355)
5384 - fixes for mds_cleanup_orphans (1934)
5385 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5386 - block all file creations until orphan recovery completes (1901)
5387 - client remove rq_connection from request struct (2423)
5388 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5389 - recovery timer race (2670)
5390 - mdc_close recovey bug (2532)
5391 - ptlrpc cleanup bug (2710)
5392 - mds timeout on local locks (2588)
5393 - namespace lock held during RPCs (2431)
5394 - handle interrupted sync write properly (2503)
5395 - don't try to handle a message that hasn't been replied to (2699)
5396 - client assert failure during cleanup after abort recovery (2701)
5397 - leak mdc device after failed mount (2712)
5398 - ptlrpc_check_set allows timedout requests to complete (2714)
5399 - wait for inflight reqs when ptlrpcd finishes (2710)
5400 - make sure unregistered services are removed from the srv_list
5401 - reset bulk XID's when resending them (caught by 1138 test)
5402 - unregister_bulk after timeout
5403 - fix lconf error (2694)
5404 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5405 - readahead locks pages, leaves pending causing memory pressure (2673)
5406 - increase OST request buffers to 4096 on large machines (2729)
5407 - fix up permission of existing directories in simple_mkdir (2661)
5408 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5409 - don't assign transno to errored transactions (2742)
5410 - don't delete objects on OST if given a bogus objid from MDS (2751)
5411 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5412 - if rq_no_resend, then timeout request after recovery (2432)
5413 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5414 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5415 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5416 - MDS recovery shouldn't delete objects that it has given out (2730)
5417 - if enqueue arrives after completion, don't clobber LVB (2819)
5418 - don't unlock pages twice when trigger_group_io returns error (2814)
5419 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5420 - don't write pages to disk if there was an error (1450)
5421 - don't ping imports that have recovery disabled (2676)
5422 - take buffered bytes into account when balancing socknal conn (2817)
5423 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5424 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5425 - remove little-endian swabbing of llog records (1987)
5426 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5427 - retry reposting request buffers if they fail (1191)
5428 - grow extent at grant time to avoid granting a revoked lock (2809)
5429 - lock revoke doesn't evict page if covered by a second lock (2765)
5430 - disable VM readahead to avoid reading outside lock extents (2805)
5432 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5433 - updated kernel patches for hp-2.4.20 kernel (2681)
5435 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5438 - fix truncated write corruption (2366)
5439 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5441 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5442 - ensure len(uuid) < 37 in lmc (1171)
5443 - fix ia64 OOPS in llog_test (2255)
5444 - zero end of page at obdfilter for partial page writes (2648)
5445 - don't leave stale dentries around after renames (bug 2428)
5446 - fix timeouts when evicting a client with a single lock held (2642)
5447 - set deadline for the initial HELLO message to drain (2634)
5448 - print out dotted-quad IP addresses in the socknal (2302)
5450 - additional debugging for MDS client eviction problem (2443)
5451 - fix mkfsoptions support for osts (2603, 2604)
5453 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5456 - add series for the vanilla 2.6.0 kernel
5457 - add series for the vanilla 2.4.24 kernel
5458 - add series for a cray x86/64 UL kernel drop
5459 - fix xattr patches for the vanilla 2.4.19 series
5461 - generate true UUIDs in lmc (1171)
5462 - have portals stack dumping break in UML (2466)
5463 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5464 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5465 - wake callback waiting threads on client eviction (2460)
5466 - Add --ptldebug and --subsystem to lmc (1719)
5467 - update assertion to allow safe interrupt allocation
5468 - set rq_no_resend for cancel requests (2432)
5469 - recalculate ptlrpcd timeout after resend (2494)
5470 - call vfs_rmdir when removing pending directories (2368)
5471 - fix renaming a file to itself (2429)
5472 - lmc creates a default one-stripe lov (2454)
5473 - expand procfs space to handle large clusters (2326)
5474 - increase UML stack to avoid overflow
5475 - update lconf's list of debug and subsystem masks
5476 - fix lfs find --obd (2510)
5477 - /proc tunable for disabling filter read caching (2591)
5478 - stop rpm packages from altering slapd.conf (2301)
5479 - disable nagle in the socknal under 0conf (2578)
5480 - choose mds inode size based on stripe count (2572)
5481 - fix kernel-source rpm problems (2516)
5483 - add --disable-doc to avoid pdf generation (2421)
5484 - update documentation, tests, type-os, comments
5485 - avoid format warnings on ia64
5486 - remove the TOE NAL
5487 - tiny code cleanups by removing unused fields
5489 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5492 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5493 - remove the most blatant lies from BUILDING file (2371)
5494 - change default debug level to reasonable production setting
5495 - reduce client side cache size to reduce cache flush time
5496 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5497 - make TCP zerocopy and pinger support enabled by default (2476)
5498 - sync writes completed after process exits caused crashes (2319)
5499 - maintain correct mount count on the MDS (2356)
5500 - backout 1557, because 2316 wasn't really fixed
5501 - better file I/O statistics gathering in /proc
5502 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5503 - another kernel patch to fix zero-copy TCP function export
5504 - don't take duplicate lock when processing re-sent getattr (2420)
5505 - lctl uses obd_self_export instead of creating new conn (2353)
5506 - MDS/OST recovery case which requires object creation asserted (2425)
5507 - move lfs from /usr/sbin to /usr/bin in packages
5508 - fix race between mds_client_add and mds_client_free (2417)
5509 - use kmalloc instead of slabs in portals (2430)
5510 - don't create duplicate records when a failover MDS is present (2442)
5511 - remove unnecessary mount age check (2332)
5512 - don't remove directory inodes from locks prematurely (2451)
5513 - don't break if MDS service name is the same as hostname (2103)
5514 - fix races in client write RPC generation when cache full (2482)
5516 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5519 - remove now-unused request->rq_obd (278)
5520 - if an allocation fails, print out how much memory we've used (1933)
5521 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5522 - add the "configurable stack size" patch to most series files (1256)
5523 - ability to write large log records, for 100+ OST configs (2306)
5524 - fix NULL deref when filter_prep fails (2314)
5525 - fix operator precedence error in filter_sync
5526 - dynamic allocation of socknal TX descriptors (2315)
5527 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5528 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5529 - ensure that CWARN messages in Portals always get to the syslog
5530 - __init/__exit are not for prototype decls (ldlm_init/exit)
5531 - x86-64 compile warning fixes
5532 - fix gateway LMC keyword conflict (2318)
5533 - fix MDS lock inversions in getattr/reint paths (1844)
5534 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5535 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5536 - DLM race condition prevented some lock evictions (2328)
5537 - ENOMEM detection and retry on socknal sends (2230)
5538 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5539 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5540 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5541 - better error messages when a client is rejected during recovery (1505)
5542 - avoid cancelling locks which were never granted, after failure (2330)
5543 - fix i_sem/journal inversion in mds_client_add (2333)
5544 - fix truncate/getattr lock cycle deadlock (2334)
5545 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5546 - fix two rare exit paths which could leak an l_lock() ref (2321)
5547 - fencepost error in MDS/OST orphan recovery (2226)
5548 - make log record alignment 8 bytes (1988)
5549 - lstripe now fails when requested offset > ost_count (2237)
5550 - ensure that all kernel series have a complete list.h (1607)
5551 - fix crashes in special-file operations (2316)
5552 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5554 - allow configurable automake binary, for testing new versions
5555 - small update to the lfs documentation
5557 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5559 * fix negative export reference count in fsfilt_sync (2312)
5561 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5562 * release candidate 0.9.1
5564 - orphans are moved into the PENDING directory for possible recovery
5565 - replayed opens now open by fid for orphan/rename safety (1042)
5566 - last close of an orphan inode generates a transno (683)
5567 - chdir() and mount() now pin the directory entry (1020)
5568 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5569 - discard very old requests without processing them (1502)
5570 - remove some common, well-understood CERRORs (1505)
5571 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5572 - clear "grant" flags in OST replies until OST grant code lands (1644)
5573 - fix read performance by not clobbering i_blksize on client (1598)
5574 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5575 - make LRU size a /proc tunable, clears locks when reduced (707)
5576 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5577 - server locks take references on exports now (1558)
5578 - build fixes for 2.4.20-rh trees (1663)
5579 - return an error from lov_create if all OSCs are inactive (1751)
5580 - fix import levels when a reconnect happens without a timeout (1597)
5581 - exit early from mds_open if we get a lookup error (1749)
5582 - partial page read at EOF wouldn't wait for disk before sending (1642)
5583 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5584 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5585 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5586 - write/truncate lock inversion (1639)
5587 - Don't auto-load obdclass, portals modules during cleanup (1495)
5588 - fix timestamps from jumping to "now" (1763)
5589 - extra journal assertions (1648)
5590 - add an extra multiunlink test (1771)
5591 - fix read_record/write_record API (1776)
5592 - fix leak of offset_extent, possible incorrect i_size later (1772)
5593 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5594 - Debugging for the kqswnal_get_idle_tx problems (1820)
5595 - Allow recovery to be attempted multiple times (1536)
5596 - Write out MDS last_rcvd file after it is first created (1600)
5597 - Fix tx_descriptor leak in failed transmit situations (1827)
5598 - ext3 journaling fixes for assertion failure after IO error (1871)
5599 - class_export_put() on freed export after completion AST error (1896)
5600 - Fix revalidate looping in VFS (1322)
5601 - Don't access a freed export during MDS_REINT timeout (1521)
5602 - Add open-unlink recovery support on the MDS (1673,1764)
5603 - Return an error if no MDS data was read from last_rcvd (1946)
5604 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5605 - Don't setup a disk device that doesn't match exported UUID (317)
5606 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5607 - avoid committing NULL handle in force close
5608 - local.sh is now a one-stripe LOV configuration
5609 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5610 - fix ext3 htree duplicate directory entry corruption (1516)
5611 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5612 - update to new LOV EA format (2097)
5613 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5615 2003-06-15 Phil Schwan <phil@clusterfs.com>
5618 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5619 - per-import/export recovery handling (958, 931, 959)
5620 - multiple last-rcvd slots, for serving multiple FSes (949)
5621 - connections are again shared between multiple imp/exports (963, 964)
5622 - "umount -f" would hang if any requests needed to be sent (393, 978)
5623 - avoid pinning large req buffer by copying for queued messages (989)
5624 - add "uuid" to "lctl device" command to help upcalls (991)
5625 - "open" RPCs with transnos would confuse recovery counters (1037)
5626 - do proper endian conversion of all wire messages (288, 340, 891)
5627 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5628 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5629 - fix ptlrpc_connection leak in target_handle_connect (1174)
5630 - fix import refcounting bug in OST and MDS cleanup (1134)
5631 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5632 - fix very unlikely obd_types race condition (501)
5633 - remove osc_open hack for echo_client (1187)
5634 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5635 - a failure in read_inode2 leads to deadlock (1139)
5636 - cancel ack-locks as soon as transaction is committed (1072)
5637 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5638 - make sure to commitrw after any preprw to avoid deadlock (1162)
5639 - failing to execute a file in a lustre FS would lock inode (1203)
5640 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5641 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5642 - obd_uuid2tgt would walk off the end of the list (1255)
5643 - on IA64 the osc would give portals incorrect bulk size (1258)
5644 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5645 - fix lock inversion caused by new llite matching code (1282)
5646 - limit the number of dirty pages on a client to 10MB (1286)
5647 - timed out locks were not being corrected cancelled (1289)
5648 - fix O_DIRECT above 4GB on IA-32 (1292)
5649 * major user-visible changes
5650 - fail out/fail over policy now controlled by the upcall (993)
5652 - add OBD_PING to check server availability and failure (954)
5653 - lustre messages are now sent in sending host order (288, 340, 891)
5654 - add eadatalen to MDS getattr reply (340)
5655 - OST read replies may contain second buffer, with per-page status (593)
5657 2003-03-11 Phil Schwan <phil@clusterfs.com>
5660 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5661 - failed open()s could cause deadlock; fixed (867, 869)
5662 - stop cancelling OST locks when files are closed (481)
5663 - overlapping XID spaces caused network corruption (851, 853)
5664 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5665 - setattr_raw added, to do single-RPC, server-side setattrs
5666 - lmc/lconf syntax change for OST UUIDs
5667 - fix crashy race condition between ptlrpc_free_req and osc_close
5668 - don't use request in mdc_enqueue if we hit a timeout (889)
5669 - don't set the inode i_size for regular files from the MDS (896)
5670 - handle out of order completion AST (842)
5671 - don't LBUG if a lock request times out after receiving AST (913)
5672 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5673 - if a bad lock AST arrives, send an error instead of dropping entirely
5674 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5675 - fix leak in bulk IO when only partially completed (899, 900, 926)
5676 - fix O_DIRECT for ia64 (55)
5677 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5678 - C-z after timeout could hang a process forever; fixed (977)
5680 - client-side I/O cache (678, 924, 929, 941, 970)
5682 - READPAGE and SETATTRs which don't take server-side locks get
5685 2003-02-11 Phil Schwan <phil@clusterfs.com>
5688 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5689 - Fix fsfilt_extN_readpage() to read a full page of directory
5690 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5691 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5692 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5693 - Most intent operations were removed, in favour of a new RPC mode
5694 that does a single RPC to the server and bypasses most of the VFS
5695 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5696 - Aggressively cancel local locks on DLM servers
5697 - mds_reint_unlink sends EA to the client if it's the last nlink.
5698 client uses that EA to unlink OST objects.
5699 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5700 - recursive symlinks were fixed (439)
5701 - fixed NULL deref in DEBUG_REQ
5702 - filter_update_lastobjid no longer calls sync, which annoyed extN
5703 - fixed multi-client small-writes to a single file problem (445)
5704 - fixed mtime updates during file writes (607)
5705 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5706 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5707 - fixed Linux OST size reporting problem (444, 656)
5708 - OST now updates object mtime with writes or setattr (607, 619)
5709 - client verifies file size before zeroing page past EOF (445)
5710 - OST now writes last allocated objid to disk with allocation (108)
5711 - LOV on echo now works (409)
5713 - mds_reint_unlink sends a new buffer, with the EA included. this
5714 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5715 set if a regular file was being unlinked, and it was the last link
5716 - use PtlGet from the target for bulk writes (315)
5717 - OST now updates object mtime with writes or setattr (607, 619)
5718 - LDLM now has a grant-time callback to revalidate locked items, if
5720 - Many MDS operations were reorganized to combat race conditions
5722 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5723 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5725 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5728 - Fully reactivate OST imports after reconnection (512, others)
5729 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5730 - More graceful error handling for truncating on dead OST (515)
5731 - Don't error out unless we're actually accessing dead stripes (474)
5732 - Fix garbage sizes when stripes are missing (410)
5733 - LRU counters were broken, causing constant lock purge (433, 432)
5734 - garbage on read from stripes with failed OSTs (441)
5735 - mark OSCs as active before reconnecting during recovery (438)
5736 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5737 - lfind did not preserve OST order in output (443)
5738 - symlinks cause hung clients, incorrect data (439)
5739 - stop dereferencing request after dropping refcount (457)
5740 - don't LASSERT(spin_is_locked) on non-SMP (455)
5741 - fixes for many rename() bugs
5742 - fstat didn't correctly synchronize attributes (399)
5743 - server must handle lock cancellation during blocking AST prep (487)
5744 - bulk descriptors were free()d too soon (511)
5745 - fix paths in lconf, which would load incorrect modules (451, 507)
5746 - fix confusing lconf 'host not found' error message (386)
5747 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5748 - fix race condition in mdc_blocking_ast() for inode access (526)
5749 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5750 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5751 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5752 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5753 - fix lstripe in several ways (kernel side) (527)
5754 - fix request leak in ldlm_cli_enqueue (262)
5755 - incorrect OSC was marked inactive after OST failure
5756 - call mds_fs_cleanup before unmounting filesystem (524)
5757 - fix races between taking ns_lock and ldlm_lock_change_resource
5758 - fix races updating LOV export open file list
5759 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5760 - fix recovery NULL deref in ldlm_cli_cancel_unused
5761 - fix some DLM races by using new hash table for lock handles (419)
5762 - permit the client to specify desired inodes, at replay
5763 - duplicate requests when we queue them for replay reintegration
5764 - fix last_rcvd offset calculation
5765 - sync after each recovered transaction, so we always make progress
5766 - never, not always, ERESTART requests without transnos
5767 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5769 - skip replay if the MDS says that the client is already connected
5770 - don't check for a recovery-enabled export to match lctl's UUID
5771 - don't INC_USE_COUNT for phantom exports
5772 - don't crash when cleaning up phantom exports (567)
5773 - don't double-finish or set replay data for errored mdc_open requests
5774 - abort requests when they time out, so we don't get old replies
5775 - send/receive replies for AST messages again
5776 - if the client says that it doesn't have the lock, cancel it on the
5778 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5779 mark it as destroyed, it will all work out when decref is called
5780 - fix module use counts (22, 581)
5782 - ASTs now expect a reply (server cancels lock on error reply)
5784 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5787 - fix many simultaneous client startup (392)
5788 - fix dentry->d_it clobbering
5789 - credentials weren't being shipped for readdir/getattr operations
5790 - remove invalid assertions triggered during some concurrent MD
5792 - proper Lustre versions added (336, 389)
5793 - fix memory leak for create error case (398)
5794 - fix LOV locking bug that would get cli/srv out of sync
5795 - fix echo client over LOV (409)
5796 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5797 - fix double-O_EXCL intent crash (424)
5798 - avoid sending multiple lock CANCELs (352)
5800 - MDS can do multi-client recovery (modulo bugs in new code)
5802 - many updates, edits, cleanups
5804 2002-11-18 Phil Schwan <phil@clusterfs.com>
5807 - fix null d_it dereference (346)
5808 - fix full OST/dbench hang (333)
5809 - fix permission problem with file removal (286)
5810 - fix removal of OSCs from LOV when they fail
5811 - fix NULL deref during bulk timeout (214)
5812 - fix problems related to multiple filesystems on one MDS (241)
5813 - fixed serious subtle metadata locking bugs
5814 - free locks on clients when inodes are removed due to memory
5816 - fix inode pointer in lock data (285)
5817 - partial support for multiple MDS on a single host (241)
5818 - data locks weren't cancelled at clear_inode time (290, 311)
5819 - intent locks could lead to unbounded lock growth (205)
5820 - added a maximum lock count, an LRU list, and a flusher
5821 - fix multiple rename (365)
5822 - properly abstracted the echo client
5823 - OSC locked 1 byte too many; fixed
5824 - rewrote brw callback code:
5825 - fixed recovery bugs related to LOVs (306)
5826 - fixed too-many-pages-in-one-write crash (191)
5827 - fixed (again) crash in sync_io_timeout (214)
5828 - probably fixed callback-related race (385)
5830 - Add capability to MDS protocol
5831 - LDLM cancellations and callbacks on different portals
5833 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5836 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5837 - defer open object destruction to close time (601981, 138)
5838 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5839 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5840 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5841 - don't call obd_setattr in ll_file_release for destroyed objects
5842 * protocol change to lustre_msg: move |version| and add |flags|
5843 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5844 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5845 * added replay of create, unlink, link and rename operations during
5846 MDS failover; recovery should be much more robust now
5847 * remove failed OSCs from LOVs (only lov_create uses this so far)
5848 * the lustre-HOWTO was brought (more) up to date (582544)
5850 2002-10-23 Phil Schwan <phil@clusterfs.com>
5853 - in-use dentries weren't being reused properly (617851)
5854 - prevent multiple LDLM setup (599178)
5855 - fix LOV size calculations for truncate (617853)
5856 - fix client handling of MDS intent errors (POSIX)
5857 - fix permission bug in lovstripe.c test (624321)
5858 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5859 - truncate past end of file could corrupt data
5860 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5861 - a race in recovery could return ETIMEDOUT to apps (623947)
5862 - building outside the source directory was fixed
5863 * the lustre-HOWTO was brought (more) up to date (582544)
5864 * major progress was made on recovery functionality
5866 2002-10-10 Phil Schwan <phil@clusterfs.com>
5869 - recovery deadlock fix
5870 - rm -rf causes LBUG fix (617817)
5871 - file open by multiple tasks fix (618962)
5872 - directory permissions bugs (602707 and 620007)
5873 - journal_stop fixed with locking (611313)
5874 - O_APPEND failures resolved (618273, perhaps 614459)
5875 - lconf PATH fix (619770)
5876 - IA64 build fix (621450)
5877 - RPC buffer sizes scale with amount of memory
5879 2002-10-01 Phil Schwan <phil@clusterfs.com>
5882 - locks would be cancelled without throwing away data pages,
5883 resulting in inconsistent data (605627)
5884 - inode attributes were not always being refreshed (605627, 612449)
5885 - lconf now continues to cleanup after lctl reports an error
5886 - MDS now enforces user permissions (602707)
5887 - lprocfs cleanup fixed, but not yet enabled (614157)
5888 - fixed infinite server hang, should a client not respond to an AST
5889 - avoid going into recovery if user calls readlink() with a buffer
5890 that's too small (613941)
5891 - AST RPCs no longer require replies (614867) -- this may be changed
5892 - don't crash server if client sends an IOV that's too big (611336)
5893 - fixed lock conversion deadlock (611892)
5894 - fixed the following of symlinks (614622)
5895 * recovery: the server can remove locks from a client that dies, other
5896 clients can make progress
5897 * more extN patch fixes
5898 * compile-time configurable ptlrpc buffer allocations
5900 - collaborative read cache document
5901 - Lustre Lite Performance CDR document-in-progress
5903 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5906 - fix typo in patch-2.4.18
5908 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5911 - clear ptlrpc request each time in handle_incoming_request()
5912 - unlink of files now destroys the object on the OST
5914 2002-09-19 Peter Braam <braam@clusterfs.com>
5916 * add hard link support
5917 * change obdfile creation method
5918 * kernel patch changed
5920 2002-09-19 Peter Braam <braam@clusterfs.com>
5923 - stack overflow bug in extN fixed
5925 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5927 * documentation updates
5928 - add man pages for config tools
5929 - update tests/README to describe testing with new config tools
5930 - finish metadata API descriptions
5931 * bug fixes and cleanups
5932 - statfs workaround for 16TB limit
5933 - LOV stripe allocation improved, can stripe on subset of OSTs
5934 - LOV file size/IO offset was wrong for files > 4GB in size
5935 - object EA data was being dropped, caused files to be unreadable
5936 - memory overflow with non-LOV OST caused memory corruption
5937 - fixed regression tests to work with new config tools, obdfilter
5938 - fixed bug when directory size became larger than 1 block
5939 - fixed bug (for single client case) when PWD was deleted
5940 - invalidate local directory pages when doing intent-based ops
5941 - avoid LDLM oops when lock callback contained bad data
5943 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5945 * documentation updates
5946 * bug fixes and cleanups
5947 - configuration tools
5950 - 64-bit compile warnings
5951 - 64-bit internal statfs data
5953 * test_brw on persistent OST devices
5955 * lprocfs (disabled)
5957 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5959 * documentation updates
5960 * bug fixes and cleanups
5961 * configuration tools
5963 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5966 * many small fixes to 0_5_4
5967 * io/network handling
5968 * thinkos in MDS operations
5970 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5973 * crucial basic fixes to 0.5.3
5974 * IOR, Iozone work over Elan
5977 2002-08-07 Phil Schwan <phil@clusterfs.com>
5978 * version 0_5_3, our first alpha
5979 * we use the new Portals iovs
5980 * documentation updates
5981 * bug fixes and cleanups
5982 * small changes in the DLM wire protocol
5984 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5985 * version 0_5_1 with some initial stability,
5986 * locking on MD and file I/O.
5987 * documentation updates
5988 * several bug fixes since 0.5.0
5989 * small changes in wire protocol
5991 2002-07-18 Phil Schwan <phil@clusterfs.com>
5993 * delivered as Lustre Light Alpha
5994 * fixed a crash after handling invalid MDS requests
5995 * fixed directory pages for architectures with non-4k pages sizes
5997 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5998 * release version v0_4_4
5999 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6001 * Fixes a number of interruption problems with OST operations.
6002 * Update documentation for portals header changes
6003 * Move all wire protocol structs/defines to lustre_idl.h
6004 * Fixes symlink length bug.
6005 * Add tcpdump to repository.
6007 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6008 * release version v0_4_3
6009 * Fixes statfs for inodes on extN.
6010 * Fixes bug in runtests which would delete /etc/hosts.
6011 * Use 64-bit object IDs wherever possible (not into VFS though)
6012 Remove ost_get_info, which is unused by lustre, and out of date.
6014 2002-07-03 Peter Braam <braam@clusterfs.com>
6015 * release version v0_4_2 Fixes a lookup error (type not passed)
6016 * move forward to head of Portals
6017 * move forward to latest Lustre kernel
6019 2002-06-25 Peter Braam <braam@clusterfs.com>
6020 * release version v0_4_1. Hopefully stable on single node use.