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: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
22 Description: client could not reconnect to OST because of an active request.
23 Details : The client now retries to reconnect to the same server, if a
24 connect request failed with EBUSY or -EAGAIN. Test case 35b
25 conf-sanity.sh has been created.
29 Description: Don't allow make backward step on assignin osc next id.
30 Details : race between allocation next id and ll_sync thread can be cause
31 of set wrong osc next id and can be kill valid ost objects.
35 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
39 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
43 Description: Error handling in osc_statfs_interpret() has been improved.
44 Details : Check in osc_statfs_interpret() for EBADR.
48 Description: Do not send statfs() requests to OSTs disabled by administrator.
49 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
51 Severity : enhancement
53 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
55 Severity : enhancement
57 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
59 Severity : enhancement
61 Description: Enhancement for lfs(1) command to use numeric uid/gid.
65 Description: Improved cleanup procedure for conf-sanity test_46a.
69 Description: Do not start bulk transfer, if client tried to reconnect.
70 Details : Do not start bulk transfer, if client tried to reconnect,
71 but failed due to active rpcs. Check for exp_abort_active_req
76 Description: Prevent inconsistences between linux and lustre mount structures.
77 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
78 Make the sleep interruptible.
80 Severity : enhancement
82 Description: provide server to client comms path
83 Details : server to client communications path via new LDLM_SET_INFO rpc
87 Description: truncate starts GFP_FS allocation under transaction causing deadlock
88 Details : ldiskfs_truncate calls grab_cache_page which may start page
89 allocation under an open transaction. This may lead to
90 calling prune_icache with consequent lustre reentrance.
94 Description: send correctly lsm on open replay
95 Details : MDS is trust to LSM size on replay open, but client can set wrong size
100 Description: enable client interface failover
101 Details : When a child reconnects from another NID, properly update export
102 nid hash position and ldlm reverse import.
104 Severity : enhancement
106 Description: Communicate OST degraded/readonly state via statfs to MDS
107 Details : Flags in the statfs returned from OSTs indicate whether the
108 OST is in a degraded RAID state, or if the filesystem has
109 turned read-only after a filesystem error is detected.
114 Description: don't panic if EPROTO was hit when reading symlink
115 Details : correctly handling request reference in error cases.
117 Severity : enhancement
119 Description: Add LustreNetLink, a kernel-userspace communcation path.
121 Severity : enhancement
123 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
128 Description: access to llog context before init.
129 Details : move handling CATALOGS file at osc layer and forbid access to llog
133 Frequency : rare, only when client is writing to OST on local node
135 Description: Avoid deadlock for local client writes
136 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
137 memory freeing context. This allows OST threads to set the
138 PF_MEMALLOC flag on task structures in order to allocate memory
139 from reserved pools and complete IO.
140 Use GFP_HIGHUSER for OST allocations for non-local client writes,
141 so that the OST threads generate memory pressure and allow
142 inactive pages to be reclaimed.
144 Severity : enhancement
146 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
151 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
152 Details : move ldlm namespace creation in setup phase to avoid grab
153 _lprocfs_lock with cli_sem held.
157 Description: Temporarily disable grant shrink.
158 Details : Disable the feature for debugging.
161 Frequency : only when formatting test filesystems
163 Description: Unable to run several mkfs.lustre on loop devices at the same time
164 Details : mkfs.lustre returns error 256 on the concurrent loop devices
165 formatting. The solution is to proper handle the error.
167 Severity : enhancement
169 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
171 Severity : enhancement
173 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
178 Description: resolve race between obd_disconnect and class_disconnect_exports
179 Details : if obd_disconnect will be called to already disconnected export he
180 forget release one reference and osc module can't unloaded.
182 Severity : enhancement
184 Description: Allow tuning service thread via /proc
185 Details : For each service a new
186 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
187 created that can be used to set min/max thread counts, and get the
188 current number of running threads.
193 Descriptoin: don't return error if have partially created objects for file.
194 Details : lov_update_create_set uses set->set_success as index for created
195 objects, so if some requests will be failed, they will have hole
196 at end of array and qos_shrink_lsm for allocate correct lsm.
198 Severity : enhancement
200 Description: Update OFED support to 1.4.1
205 Description: Reduce small size read RPC
206 Details : Set read-ahead limit for every file and only do read-ahead when
207 available read-ahead pages are bigger than 1M to avoid small size
210 Severity : enhancement
212 Description: move AT tunable parameters for more consistent usage
213 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
215 Severity : enhancement
217 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
218 Details : allow skip disconnected ost for send statfs request and hide error
224 Description: fix racy locking of mballoc block bitmaps causing BUG
225 Details : The locking of the mballoc buddy bitmap and the in-memory
226 block bitmap was using two different spin locks in some
227 cases. This made it possible to incorrectly access the
228 mballoc bitmap while another process was modifying it,
229 causing a sanity assertion to fail. While no on-disk corruption
230 was reported, there was some risk of this happening.
233 Frequency : rare, on llog test 6
235 Descriptoin: don't allow connect to already connected import
236 Details : allowing connect to already connected import is hide connecting problem.
241 Descriptoin: allow kill process which wait statahead result
242 Details : for some reasons 'ls' can stick in waiting result from statahead,
243 in this case need way for kill this process.
245 Severity : enhancement
247 Description: Add state history info file, enhance import info file
248 Details : Track import connection state changes in a new osc/mdc proc file;
249 add overview-type data to the osc/mdc import proc file.
251 Severity : enhancement
253 Description: MDS create should not wait for statfs RPC while holding DLM lock.
255 Severity : enhancement
257 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
262 Descriptoin: OST grant too much space to client even there are not enough space.
263 Details : Client will shrink its grant cache to OST if there are no write
264 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
265 this grant cache if there are already not enough avaible space
266 (left_space < total_clients * 32M).
269 Frequency : start MDS on uncleanly shutdowned MDS device
271 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
272 Details : stay in waiting mds<>ost recovery finished produce random bugs
273 due race between two ll_sync thread for one lov target. send
274 ACTIVATE event only if connect realy finished and import have
278 Frequency : rare, connect and disconnect target at same time
280 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
281 Details : don't call obd_disconnect under lov_lock. this long time
282 operation and can block ptlrpcd which answer to connect request.
287 Descriptoin: don't lose wakeup for imp_recovery_waitq
288 Details : recover_import_no_retry or invalidate_import and import_close can
289 both sleep on imp_recovery_waitq, but we was send only one wakeup
293 Frequency : always with long access acl
295 Descriptoin: mds can't pack reply with long acl.
296 Details : mds don't control size of acl but they limited by reint/getattr
299 Severity : enhancement
301 Description: Update to SLES10 kernel-2.6.16.60-0.33.
303 Severity : enhancement
305 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
308 Frequency : start MDS on uncleanly shutdowned MDS device
310 Descriptoin: aborting recovery hang on MDS
311 Details : don't throttle destroy RPCs for the MDT.
314 Frequency : on remount
316 Description: external journal device not working after the remount
317 Details : clear dev_rdonly flag for external journal devices in
323 Description: shutdown vs evict race
324 Details : client_disconnect_export vs connect request race.
325 if client will evicted at this time - we start invalidate
326 thread without referece to import and import can be freed
330 Frequency : rare, need acl's on inode.
332 Description: client can't handle ost additional correctly
333 Details : if ost was added after client connected to mds client can have
334 hit lnet_try_match_md ... to big messages to wide striped files.
335 in this case need teach client to handle config events about add
336 lov target and update client max ea size at that event.
338 Severity : enhancement
340 Description: Changelogs
341 Details : Changelogs are a lightweight mechanism to track filesystem
342 metadata and namespace changes. The changelog is recorded
343 permanently on the MDTs, and is periodically "consumed" / purged
344 when records are no longer needed.
346 Severity : enhancement
348 Description: compact fld format with extents
349 Details : Store range of seq rather than every seq in FLD. Seq
350 controller update FLD rather than clients. In Case of CMD, mdt0
351 has FLD, all other metadata server act as non persistent proxy
352 for FLD queries and cache fld entries in fld cache.
357 Description: don't skip ost target if they assigned to file
358 Details : Drop slow OSCs if we can, but not for requested start idx.
359 This means "if OSC is slow and it is not the requested
360 start OST, then it can be skipped, otherwise skip it only
361 if it is inactive/recovering/out-of-space.
365 Description: more cleanup in mds_lov
366 Details : not send LOV EA under replay, we can't know about they size at this
367 time. Don't allow client connect to mds before any ost connected,
368 for avoid problems with LOV EA size and returning EIO to client.
370 Severity : enhancement
372 Description: Interoperability at server side (Disk interoperability)
374 Severity : enhancement
376 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
378 Severity : enhancement
380 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
382 Severity : enhancement
384 Description: New client IO stack (CLIO).
386 Severity : enhancement
388 Description: Commit on sharing. Eliminate inter-client dependencies between
389 uncommitted transactions by doing transaction commits.
390 Thereby clients may recovery independently.
393 Frequency : Create a symlink file with a very long name
395 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
396 Details : If there is no extra space in the request for early cancels,
397 ldlm_req_handles_avail() returns 0 instead of a negative value.
399 Severity : enhancement
401 Description: Add /proc entry for import status
402 Details : The mdc, osc, and mgc import directories now have
403 an import directory that contains useful import data for debugging
406 Severity : enhancement
408 Description: Re-disable certain /proc logging
409 Details : Enable and disable client's offset_stats, extents_stats and
410 extents_stats_per_process stats logging on the fly.
413 Frequency : Only on FC kernels 2.6.22+
415 Description: oops in statahead
416 Details : Do not drop reference count for the dentry from VFS when lookup,
417 VFS will do that by itself.
419 Severity : enhancement
421 Description: Generic /proc file permissions
422 Details : Set /Proc file permissions in a more generic way to enable non-
423 root users operate on some /proc files.
427 Description: Hitting mdc_commit_close() ASSERTION
428 Details : Properly handle request reference release in
429 ll_release_openhandle().
433 Description: quota recovery deadlock during mds failover
434 Details : This patch includes att18982, att18236, att18237 in bz14840.
436 1. fix osts hang when mds does failover with quotaon
437 2. prevent watchdog storm when osts threads wait for the
442 Frequency : only patchless client
443 Description: add workaround for race between add/remove dentry from hash
445 Severity : enhancement
447 Description: Allow OST glimpses to return PW locks
451 Description: LBUG when llog conf file is full
452 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
456 Description: Prevent import from entering FULL state when server in recovery
460 Description: service mount cannot take device name with ":"
461 Details : Only when device name contains ":/" will mount treat it as
467 Description: replace ptlrpcd with the statahead thread to interpret the async
468 statahead RPC callback
472 Frequency : on recovery
473 Description: I/O failures after umount during fail back
474 Details : if client reconnected to restarted server we need join to recovery
475 instead of find server handler is changed and process self
476 eviction with cancel all locks.
478 Severity : enhancement
480 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
482 Severity : enhancement
484 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
486 Severity : enhancement
488 Description: Upcall on Lustre log has been dumped
489 Details : Allow for a user mode script to be called once a Lustre log has
490 been dumped. It passes the filename of the dumped log to the
491 script, the location of the script can be specified via
492 /proc/sys/lnet/debug_log_upcall.
497 Description: avoid idr_remove called for id which is not allocated.
498 Details : Move assigment s_dev for clustered nfs to end of initialization,
499 to avoid problem with error handling.
504 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
505 Details : When connection is reused this not moved from CONN_UNUSED_HASH
506 into CONN_USED_HASH and this prodice warning when put connection
507 again in unused hash.
510 Severity : enhancement
512 Description: File striping can now be set to use an arbitrary pool of OSTs.
514 Severity : enhancement
516 Description: Export bytes_read/bytes_write count on OSC/OST.
520 Description: Early reply size mismatch, MGC loses connection
521 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
522 the connect flags are properly negotiated.
527 Description: quotas are not honored with O_DIRECT
528 Details : all writes with the flag O_DIRECT will use grants which leads to
529 this problem. Now using OBD_BRW_SYNC to guard this.
533 Description: add quota statistics
534 Details : 1. sort out quota proc entries and proc code.
535 2. add quota statistics
537 Severity : enhancement
539 Description: enable quota support for HEAD.
543 Description: Properly propagate oinfo flags from lov to osc for statfs
544 Details : restore missing copy oi_flags to lov requests.
546 Severity : enhancement
548 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
549 : l_getgroups(8), lst(8), routerstat(8)
550 Details : included man pages for llobdstat(8), llstat(8),
551 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
553 Severity : enhancement
555 Description: configure's --enable-quota should check the
556 : kernel .config for CONFIG_QUOTA
557 Details : configure is terminated if --enable-quota is passed but
558 : no quota support is in kernel
562 Description: Remove portals compatibility
563 Details : Remove portals compatibility, not interoperable with releases
568 Description: Resolve device initialization race
569 Details : Prevent proc handler from accessing devices added to the
570 obd_devs array but yet be intialized.
572 Severity : enhancement
574 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
576 Severity : enhancement
578 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
583 Description: Using wrong pointer in osc_brw_prep_request
584 Details : Access to array[-1] can produce panic if kernel compiled with
585 CONFIG_PAGE_ALLOC enabled
589 Description: Client runs out of low memory
590 Details : Consider only lowmem when counting initial number of llap pages
594 Description: *optional* service tags registration
595 Details : if the "service tags" package is installed on a Lustre node
596 When the filesystem is mounted, a local-node service tag will
597 be created. See http://inventory.sun.com/ for more information
598 about the Service Tags asset management system.
602 Description: Kernel BUG tries to release flock
603 Details : Lustre does not destroy flock lock before last reference goes
604 away. So always drop flock locks when client is evicted and
605 perform unlock regardless of successfulness of speaking to MDS.
609 Description: add refcount protection for osc callbacks, avoid panic on shutdown
613 Description: sanity test 65a fails if stripecount of -1 is set
614 Details : handle -1 striping on filesystem in ll_dirstripe_verify
619 Description: ASSERTION(CheckWriteback(page,cmd)) failed
620 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
623 Severity : enhancement
625 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
629 Description: do not process already freed flock
630 Details : flock can possibly be freed by another thread before it reaches
631 to ldlm_flock_completion_ast.
635 Description: LBUG during stress test
636 Details : Need properly lock accesses the flock deadlock detection list.
640 Description: oops in page fault handler
641 Details : kernel page fault handler can return two special 'pages' in error
642 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
646 Description: timeout with invalidate import.
647 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
648 should be handled by ptlrpcd. This produce long age waiting and
649 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
651 Severity : enhancement
653 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
657 Description: enable MGS and MDT services start separately
658 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
659 a co-located MGS without starting the MGS, which is a complement
660 to 'nosvc' mount option.
664 Frequency : after recovery
665 Description: precreate to many object's after del orphan.
666 Details : del orphan st in oscc last_id == next_id and this triger growing
667 count of precreated objects. Set flag LOW to skip increase count
668 of precreated objects.
672 Frequency : rare, on clear nid stats
673 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
674 Details : when clean nid stats sometimes try destroy live entry,
675 and this produce panic in free.
679 Description: Stack overflow during MDS log replay
680 ease stack pressure by using a thread dealing llog_process.
684 Description: wait until IO finished before start new when do lock cancel.
685 Details : VM protocol want old IO finished before start new, in this case
686 need wait until PG_writeback is cleared until check dirty flag and
687 call writepages in lock cancel callback.
689 Severity : enhancement
691 Description: using special macro for print time and cleanup in includes.
695 Description: mds_mfd_close() ASSERTION(rc == 0)
696 Details : In mds_mfd_close(), we need protect inode's writecount change
697 within its orphan write semaphore to prevent possible races.
701 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
705 Frequency : rare, on shutdown ost
706 Description: don't hit live lock with umount ost.
707 Details : shrink_dcache_parent can be in long loop with destroy dentries,
708 use shrink_dcache_sb instead.
712 Description: don't panic with use echo client
713 Details : echo client pass NULL as client nid pointer and this produce null
718 Description: fix build on ppc32
719 Details : compile code with -m64 flag produce wrong object file for ppc32.
723 Description: add message levels for liblustreapi
727 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
728 Details : Move the 'good_osts' check before the 'total_bavail' check. This
729 will result in an -EAGAIN and in the exit call path we call
730 alloc_rr() which will with increasing aggressiveness attempt to
731 aquire precreated objects on the minimum number of required OSCs.
735 Description: Use old size assignment to avoid deadlock
736 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
737 the scheduling while holding a spinlock. We do not need locking
738 for size in ll_update_inode() because size is only updated from
739 the MDS for directories or files without objects, so there is no
740 other place to do the update, and concurrent access to such inodes
741 are protected by the inode lock.
745 Description: resolve "_IOWR redefined" build error on SLES10
749 Description: dump the memory debugging after all modules are unloaded to
750 suppress false negative in conf_sanity test 39
752 Severity : enhancement
754 Description: build kernel-ib packages for OFED 1.3 in our release cycle
759 Description: fix SLES kernel versioning
760 Details : the kernel version for our SLES 10 kernel did not include a "-"
761 before the "smp" at the end. while this was not a problem in
762 general, it did mean that software trying to use the kernel
763 version to try to detect a vendor specific kernel would fail.
764 this was most evident by the OFED build scripts.
768 Description: Don't update lov_desc members until making sure they are valid
769 Details : When updating lov_desc members via proc fs, need fix their
770 validities before doing the real update.
774 Description: don't put request into delay list while invalidate in flight.
775 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
776 import in flight. this produce timeout for invalidate and sometimes
777 can cause stale data.
781 Frequency : on ppc only
782 Description: not convert ost objects for directory because it's not exist.
783 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
787 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
789 Severity : enhancement
791 Description: Update to support 2.6.22.14 vanilla kernel.
795 Frequency : rare, on recovery
796 Description: read procfs can produce deadlock in some situation
797 Details : Holding lprocfs lock which send rpc can produce block for destroy
798 obd objects and this also block reconnect with -EALREADY. This
799 isn't fix all lprocfs bugs - but make it rare.
801 Severity : enhancement
803 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
806 Frequency : frequent on X2 node
808 Description: mdc_set_open_replay_data LBUG
809 Details : Set replay data for requests that are eligible for replay.
813 Description: lustre_mgs: operation 101 on unconnected MGS
814 Details : When MGC is disconnected from MGS long enough, MGS will evict the
815 MGC, and late on MGC cannot successfully connect to MGS and a lot
816 of the error messages complaining that MGS is not connected.
820 Frequency : on network error
821 Description: panic with double free request if network error
822 Details : mdc_finish_enqueue is finish request if any network error ocuring,
823 but it's true only for synchronus enqueue, for async enqueue
824 (via ptlrpcd) this incorrect and ptlrpcd want finish request
827 Severity : enhancement
829 Description: client-side metadata stat-ahead during readdir(directory readahead)
830 Details : perform client-side metadata stat-ahead when the client detects
831 readdir and sequential stat of dir entries therein
834 Frequency : on start mds
836 Description: Implement get_info(last_id) in obdfilter.
839 Frequency : occasional
841 Description: Correctly check stale fid, not start epoch if ost not support SOM
842 Details : open with flag O_CREATE need set old fid in op_fid3 because
843 op_fid2 was overwritten with new generated fid, but mds can answer
844 with one of these two fids and both is not stale. Setattr
845 incorrectly started an epoch and assume will be called
846 done_writing, but without SOM done_writing ever being called.
849 Frequency : rare, depends on device drivers and load
851 Description: MDS or OSS nodes crash due to stack overflow
852 Details : Code changes in 1.8.0 increased the stack usage of some functions.
853 In some cases, in conjunction with device drivers that use a lot
854 of stack the MDS (or possibly OSS) service threads could overflow
855 the stack. One change which was identified to consume additional
856 stack has been reworked to avoid the extra stack usage.
859 Frequency : occasional
861 Description: Do not fail import if osc_interpret_create gets -EAGAIN
862 Details : If osc_interpret_create got -EAGAIN it immediately exits and
863 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
864 oscc_has_objects and see OSC has no objests and call
865 oscc_internal_create to resend create request.
867 Severity : enhancement
869 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
871 Severity : enhancement
873 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
876 Frequency : very rare
878 Description: Fix replay if there is an un-replied request and open
879 Details : In some cases, older replay request will revert the
880 mcd->mcd_last_xid on MDS which is used to record the client's
883 Severity : enhancement
885 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
887 Severity : enhancement
889 Description: Add rhel5 support to HEAD.
891 Serverity : enhancement
893 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
898 Description: Don't allow skipping OSTs if index has been specified.
899 Details : Don't allow skipping OSTs if index has been specified, make
900 locking in internal create lots better.
904 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
905 Details : access to bitfield in structure is always rounded to long
906 and this produce problem with not atomic change any bit.
910 Description: Lustre make rpms failed.
911 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
912 builds Lustre rpms from tarball.
915 Frequency : rare on shutdown ost
917 Description: If llog cancel was not send before clean_exports phase, this can
918 produce deadlock in llog code.
919 Details : If llog thread has last reference to obd and call class_import_put
920 this produce deadlock because llog_cleanup_commit_master wait when
921 last llog_commit_thread exited, but this never success because was
922 called from llog_commit_thread.
926 Description: allow userland application know is lost one of stripes.
927 Details : fill lvb_blocks with error code on ost and return it to
928 application if error flag found.
932 Description: NULL lov_tgts causing MDS oops
933 Details : more safe checks for NULL lov_tgts for avoid oops.
935 Severity : enhancement
937 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
941 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
942 Details : If an OST is down the MDS will hang indefinitely in
943 obd_statfs_rqset() waiting for the statfs data. While for
944 MDS QOS usage of statfs, it should not stuck in waiting.
946 Severity : enhancement
948 Description: remote_acl support
949 Details : Support ACL-based permission check for remote user.
950 Support setfacl/getfacl for remote user with the utils
951 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
952 the system "{s,g}etfacl" utils.
954 Severity : enhancement
956 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
958 Severity : enhancement
960 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
964 Description: make mgs_setparam() handle fsname containing dash
965 Details : fsname containing a dash does not work with lctl conf_param
967 Severity : enhancement
969 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
971 Severity : enhancement
973 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
977 Frequency : rare, at shutdown
978 Description: access already free / zero obd_namespace.
979 Details : if client_disconnect_export was called without force flag set,
980 and exist connect request in flight, this can produce access to
981 NULL pointer (or already free pointer) when connect_interpret
982 store ocd flags in obd_namespace.
986 Frequency : only at startup
987 Description: not alloc memory with spinlock held.
988 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
989 if any spinlock held.
991 Severity : enhancement
993 Description: make lustre randomly fail allocating memory
994 Details : Make lustre randomly failed allocating memory for testing purpose.
996 Severity : enhancement
998 Description: lost problems with lov objid file
999 Details : Fixes some scability and access to not inited memory problems
1000 in work with lov objdid file.
1005 Description: lfs find does not continue on file error
1006 Details : Continue other files processing when a file/dir is absent.
1010 Description: Inconsistent usage of lustre_pack_reply()
1011 Details : Standardize the usage of lustre_pack_reply() such that it
1012 always generate a CERROR on failure.
1015 Frequency : occasional
1017 Description: MDS hang or stay in waiting lock
1018 Details : If client receive lock with CBPENDING flag ldlm need send lock
1019 cancel as separate rpc, to avoid situation when cancel request
1020 can't processed due all i/o threads stay in wait lock.
1024 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1028 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1032 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1036 Description: Update to RHEL4 latest kernel.
1038 Severity : enhancement
1040 Description: Build SLES10 patchless client fails
1041 Details : The configure was broken by run ./configure with
1042 --with-linux-obj=.... argument for patchless client. When the
1043 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1044 can't search header adequately. Use absolute path such as
1045 -I($LINUX)/include instead.
1049 Description: interrupt oig_wait produce painc on resend.
1050 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1051 private set, and this produce situation when rq_ptlrpcd_data not
1052 copyed to new allocated request and triggered LBUG on assert
1053 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1054 ptlrpcd set if request is changed and can be safety changed to use
1059 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1060 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1061 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1062 these macros and replace them with their difinitions instead.
1066 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1067 Details : Update which_patch & target file for SLES10 latest kernel.
1070 Description: Remove client patches from SLES 10 kernel.
1071 Details : This causes SLES 10 clients to behave as patchless clients
1072 even on a Lustre-patched (server) kernel.
1074 Severity : enhancement
1076 Description: self-adjustable client's lru lists
1077 Details : use adaptive algorithm for managing client cached locks lru
1078 lists according to current server load, other client's work
1079 pattern, memory activities, etc. Both, server and client
1080 side namespaces provide number of proc tunables for controlling
1083 Severity : enhancement
1085 Description: light-weight GSS support
1086 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1087 and reduce performance overhead.
1089 Severity : enhancement
1091 Description: Linux keyring support
1092 Details : Support using service of Linux keyring for Lustre GSS internal
1093 context refresh/cache mechanism.
1097 Description: Fix errors in lfs documentation
1098 Details : Fixes man pages
1102 Description: don't use GFP_* in generic Lustre code.
1103 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1107 Description: obdclass is limited by single OBD_ALLOC(idarray)
1108 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1112 Description: warnings with build patchless client with vanila 2.6.19 and up
1113 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1114 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1118 Frequency : Only for SLES
1119 Description: sanity_quota fail test_1
1120 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1121 file, which makes TSTID[2] inunique.
1125 Frequency : Only for SLES9
1126 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1127 Details : Update target/ChangeLog/which_patch .
1131 Frequency : Only for test_10 in sanity-quota.sh
1132 Description: a bug in quota test code
1133 Details : checking a test flag in a wrong place
1137 Frequency : Only for test_13 in sanity-quota.sh
1138 Description: a bug in quota test script
1139 Details : shouldn't check overfull allocated quota
1143 Description: jbd statistics
1144 Details : Port older jbd statistics patch for sles10
1146 Severity : enhancement
1148 Description: Add printing inode info into message about error in writepage.
1153 Description: reply_lock_interpret crash due to race with it and lock cancel.
1154 Details : Do not replay locks that are being cancelled. Do not reference
1155 locks by their address during replay, just by their handle.
1159 Frequency : When flocks are used.
1160 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1161 Details : Flock locks might destroy just granted lock if it could be merged
1162 with another existing flock, this is done in completion handler,
1163 so teach ldlm_cli_enquque_fini that this is a valid case for
1169 Description: Oops in read and write path when failing to allocate lock.
1170 Details : Check if lock allocation failed and return error back.
1174 Description: lstripe command fails for valid OST index
1175 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1176 instead of lov->desc.ld_active_tgt_count.
1180 Description: sanity.sh failed test 103
1181 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1186 Description: df doesn't work properly if diskfs blocksize != 4K
1187 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1191 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1192 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1194 Severity : enhancement
1196 Description: Adaptive timeouts
1197 Details : RPC timeouts adapt to changing server load and network
1198 conditions to reduce resend attempts and improve recovery time.
1202 Description: llapi_file_create() does not allow some changes
1203 Details : add llapi_file_open() that allows specifying the mode and
1204 open flags, and also returns an open file handle.
1208 Description: merge and cleanup kernel patches.
1209 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1213 Description: Add journal checksum support.(Kernel part)
1214 Details : The journal checksum feature adds two new flags i.e
1215 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1216 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1217 indicates that the commit block contains the checksum for
1218 the blocks described by the descriptor blocks. Now commit
1219 record can be sent to disk without waiting for descriptor
1220 blocks to be written to disk. This behavior is controlled
1221 using JBD2_FEATURE_ASYNC_COMMIT flag.
1225 Description: OSS needs mutliple precreate threads
1226 Details : Add ability to start more than one create thread per OSS.
1230 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1231 Details : execute lfs setstripe on client
1235 Description: mds_obd_create error creating tmp object
1236 Details : When the user sets quota on root, llog will be affected and can't
1237 create files and write files.
1240 Frequency : Always on ia64 patchless client, and possibly others.
1242 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1243 Details : This allows the patchless client to be loaded on architectures
1244 without this export.
1248 Description: RedHat Update kernel for RHEL5
1249 Details : Add the kernel config file for RHEL5.
1253 Description: RedHat Update kernel for RHEL5
1254 Details : Modify the kernel config file more closer RHEL5.
1258 Description: Build failure against Centos5 (RHEL5)
1259 Details : Define PAGE_SIZE when it isn't present.
1263 Description: test_7 on sanity-quota.sh can't be used on separate machines
1264 Details : add facet to handle it
1268 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1269 Details : with NFS, the anon dentry's parent was set to itself in
1270 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1271 corresponding dentry other than use rec->ur_name.
1273 Severity : enhancement
1275 Description: lfs setstripe enhancement
1276 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1280 Description: enable data checksumming by default
1281 Details : enable checksum by default, allow --disable-checksum
1282 configure option and "-o nochecksum" mount option. Checksums
1283 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1287 Description: lustre support for RHEL5
1288 Details : Add support for RHEL5.
1292 Description: Client eviction due to failover config
1293 Details : after a connection loss, the lustre client should attempt to
1294 reconnect to the last active server first before trying the
1295 other potential connections.
1299 Description: when mds and osts use different quota unit(32bit and 64bit),
1300 quota will be released repeatly.
1301 Details : void sending multiple quota reqs to mds, which will keep the
1302 status between the reqs.
1306 Description: osts not allocated evenly to files
1307 Details : change the condition to increase offset_idx
1310 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1312 Description: Data corruption for OSTs that are formatted larger than 2TB
1314 Details : When generating the bio request for lustre file writes the
1315 sector number would overflow a temporary variable before being
1316 used for the IO. The data reads correctly from Lustre (which
1317 will overflow in a similar manner) but other file data or
1318 filesystem metadata may be corrupted in some cases.
1322 Description: Tune the kernel for good SCSI performance.
1323 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1324 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1329 Description: rewrite ext2-derived code in obdclass/uuid.c
1330 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1331 from scratch preserving functionality.
1335 Description: "lfs find -obd UUID" prints directories
1336 Details : "lfs find -obd UUID" will return all directory names instead
1337 of just file names. It is incorrect because the directories
1338 do not reside on the OSTs.
1342 Description: use i_size_read and i_size_write in 2.6 port
1343 Details : replace inode->i_size access with i_size_read/write()
1346 Frequency : when removing large files
1348 Description: scheduling issue during removal of large Lustre files
1349 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1350 It causes scheduling issues when removing large files (17TB in the
1354 Frequency : only with liblustre clients on XT3
1356 Description: evictions taking too long
1357 Details : allow llrd to evict clients directly on OSTs
1360 Frequency : only on ppc
1362 Description: /proc/fs/lustre/devices broken on ppc
1363 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1364 We should make sure the type of 'index' is loff_t and then cast
1365 explicitly as needed below. Do not assign an explicitly cast
1369 Frequency : only for rhel5
1371 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1372 Details : Modify the target file & which_kernel.
1374 Severity : enhancement
1376 Description: omit set fsid for export NFS
1377 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1378 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1382 Frequency : Always, for kernels after 2.6.16
1383 Description: Fix warning idr_remove called for id=.. which is not allocated.
1384 Details : Last kernels save old s_dev before kill super and not allow
1385 to restore from callback - restore it before call kill_anon_super.
1389 Description: buffer overruns could theoretically occur
1390 Details : llapi_semantic_traverse() modifies the "path" argument by
1391 appending values to the end of the origin string, and a
1392 overrun may occur. Adding buffer overrun check in liblustreapi.
1396 Description: Fix error on 'ls .' at the top of the Lustre mount.
1397 Details : Don't revalidate dentry if it is a root dentry.
1401 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1402 Details : Modify vm-tunables-rhel4.patch.
1406 Description: Kernel config for 2.6.18-vanilla.
1407 Details : Modify targets/2.6-vanilla.target.in.
1408 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1409 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1410 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1411 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1415 Description: improve handling recoverable errors
1416 Details : if request processig with error which can be recoverable on server
1417 request should be resend, otherwise page released from cache and
1422 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1423 Details : Modify target file & which_patch.
1424 A flaw was found in the IA32 system call emulation provided
1425 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1426 value could be stored in the %RAX register, which could trigger an
1427 out-of-bounds system call table access. An untrusted local user
1428 could exploit this flaw to run code in the kernel
1429 (ie a root privilege escalation). (CVE-2007-4573).
1433 Description: O_DIRECT bypasses client statistics.
1434 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1435 read_ahead_stats were updated. copied the stats section from
1436 osc_send_oap_rpc() into async_internal().
1440 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1444 Description: change order of libsysio includes
1445 Details : '#include sysio.h' should always come before '#include xtio.h'
1449 Description: handle "serious error: objid * already exists" more gracefully
1450 Details : If LAST_ID value on disk is smaller than the objects existing in
1451 the O/0/d* directories, it indicates disk corruption and causes an
1452 LBUG(). If the object is 0-length, then we should use the existing
1453 object. This will help to avoid a full fsck in most cases.
1455 Severity : enhancement
1457 Description: adapt the lustre_config script to support the upgrade case
1458 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1459 server targets to 1.6 in parallel.
1463 Description: To avoid grant space > avaible space when the disk is almost
1464 full. Without this patch you might see the error "grant XXXX >
1465 available" or some LBUG about grant, when the disk is almost
1467 Details : In filter_check_grant, for non_grant cache write, we should
1468 check the left space by if (*left > ungranted + bytes), instead
1469 of (*left > ungranted), because only we are sure the left space
1470 is enough for another "bytes", then the ungrant space should be
1471 increase. In client, we should update cl_avail_grant only there
1472 is OBD_MD_FLGRANT in the reply.
1476 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1477 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1478 described in CVE-2007-4573. This problem could allow untrusted
1479 local users to gain root access.
1482 Frequency : when using O_DIRECT and quotas
1484 Description: Incorrect file ownership on O_DIRECT output files
1485 Details : block usage reported by 'lfs quota' does not take into account
1486 files that have been written with O_DIRECT.
1491 Description: touch file failed when fs is not full
1492 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1493 otherwise we can get ENOSP while fs is not full.
1497 Description: parallel lock callbacks
1498 Details : Instead of sending blocking and completion callbacks as separated
1499 requests, adding them to a set and sending in parallel.
1502 Frequency : only for Cray XT3
1503 Bugzilla : 12829/13455
1504 Description: Changing primary group doesn't change the group lustre assigns to
1506 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1507 overridden with the primary group provided by the group upcall,
1508 whereas the supplied fsgid can be trusted if it is in the list of
1509 supplementary groups returned by the group upcall.
1511 Severity : enhancement
1513 Description: Allow masking D_WARNING, D_ERROR messages from console
1514 Details : Console messages can now be disabled via lnet.printk.
1518 Description: User code with malformed file open parameter crashes client node
1519 Details : Before packing join_file req, all the related reference should be
1520 checked carefully in case some malformed flags cause fake
1521 join_file req on client.
1525 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1526 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1527 mode is equal to the granted mode. However, it should also check
1528 that LDLM_FL_CP_REQD is not set.
1532 Description: Heavy nfs access might result in deadlocks
1533 Details : After ELC code landed, it is now improper to enqueue any mds
1534 locks under och_sem, because enqueue might want to decide to
1535 cancel open locks for same inode we are holding och_sem for.
1539 Description: Client eviction while running blogbench
1540 Details : A lot of unlink operations with concurrent I/O can lead to a
1541 deadlock causing evictions. To address the problem, the number of
1542 oustanding OST_DESTROY requests is now throttled to
1543 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1544 callbacks are processed in priority.
1548 Description: enable ACLs on MDS by default
1549 Details : ACLs must be enabled on MDS by default.
1552 Frequency : PPC/PPC64 only
1554 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1555 Details : fix duplicated definitions between asm-ppc64/types.h and
1556 lustre_types.h on PPC.
1559 Frequency : PPC/PPC64 only
1561 Description: asm-ppc/segment.h does not exist
1562 Details : fix compile issue on PPC.
1566 Description: better handle error messages in extents code
1569 Frequency : RHEL4 only
1571 Description: mkfs is very slow on IA64/RHEL4
1572 Details : A performance regression has been discovered in the MPT Fusion
1573 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1574 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1575 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1577 Severity : enhancement
1579 Description: SNMP support enhancement
1580 Details : Adding total number of sampled request for an MDS node in snmp
1583 Severity : enhancement
1585 Description: Optimize ldlm waiting list processing for PR extent locks
1586 Details : When processing waiting list for read extent lock and meeting read
1587 lock that is same or wider to it that is not contended, skip
1588 processing rest of the list and immediatelly return current
1589 status of conflictness, since we are guaranteed there are no
1590 conflicting locks in the rest of the list.
1594 Description: Time out and refuse to reconnect
1595 Details : When the failover node is the primary node, it is possible
1596 to have two identical connections in imp_conn_list. We must
1597 compare not conn's pointers but NIDs, otherwise we can defeat
1598 connection throttling.
1602 Description: port llog fixes from b1_6 into HEAD
1603 Details : Port llog reference couting and some llog cleanups from b1_6
1604 (bug 10800) into HEAD, for protect from panic and access to already
1605 free llog structures.
1609 Description: Detect stride IO mode in read-ahead
1610 Details : When a client does stride read, read-ahead should detect that and
1611 read-ahead pages according to the detected stride pattern.
1615 Description: data checksumming impacts single node performance
1616 Details : add support for several checksum algorithm. Currently, only CRC32
1617 and Adler-32 are supported. The checksum type can be changed on
1618 the fly via /proc/fs/lustre/osc/*/checksum_type.
1622 Description: use adler32 for page checksums
1623 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1628 Description: build for x2 fails
1629 Details : fix compile issue on Cray systems.
1633 Description: Properly match for duplicate locks
1634 Details : Due to different lock order from skiplists code, we need to
1635 traverse entire list for now
1638 Frequency : only on PPC/SLES10
1640 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1641 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1642 be defined. Add a hack in mkfs_lustre.c to work around this compile
1647 Description: LASSERT on MDS when client holding flock lock dies
1648 Details : ldlm pool logic depends on number of granted locks equal to
1649 number of released locks which is not true for flock locks, so
1650 just exclude such locks from consideration.
1654 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1655 Details : Use fsfilt_sync() for both the whole filesystem sync and
1656 individual file sync to eliminate dangerous inode locking
1657 with I_LOCK that can lead to a deadlock.
1661 Description: performance in 1.6.3
1662 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1663 sure that 1MB requests are not fragmented by the block layer.
1665 Severity : enhancement
1667 Description: organize the server-side client stats on per-nid basis
1668 Details : Change the structure of stats under obdfilter and mds to
1676 The "uuid"s file would list the uuids of _active_ exports.
1677 And the clear entry is to clear all stats and stale nids.
1679 Severity : enhancement
1681 Description: eliminate client locks in face of contention
1682 Details : file contention detection and lockless i/o implementation
1683 for contended files.
1687 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1690 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1692 Description: MDS may lose file striping (and hence file data) in some cases
1693 Details : If there are additional extended attributes stored on the MDS,
1694 in particular ACLs, SELinux, or user attributes (if user_xattr
1695 is specified for the client mount options) then there is a risk
1696 of attribute loss. Additionally, the Lustre file striping
1697 needs to be larger than default (e.g. striped over all OSTs),
1698 and an additional attribute must be stored initially in the
1699 inode and then increase in size enough to be moved to the
1700 external attribute block (e.g. ACL growing in size) for file
1705 Description: skiplist implementation simplification
1706 Details : skiplists are used to group compatible locks on granted list
1707 that was implemented as tracking first and last lock of each
1708 lock group the patch changes that to using doubly linked lists
1712 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1713 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1714 because LASSERT on both the data supplied by a client, and the
1715 data on disk is dangerous and incorrect.
1717 Severity : enhancement
1719 Description: Slow truncate/writes to huge files at high offsets.
1720 Details : Directly associate cached pages to lock that protect those pages,
1721 this allows us to quickly find what pages to write and remove
1722 once lock callback is received.
1726 Description: more ldlm soft lockups
1727 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1728 starve other threads from the resource lock for a long time in
1729 case of long waiting queue, so change the debug level from
1730 D_OTHER to the less frequently used D_INFO.
1732 Severity : enhancement
1734 Description: add -gid, -group, -uid, -user options to lfs find
1738 Description: Hung threads in invalidate_inode_pages2_range
1739 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1740 once one is completed. As a result, some RPCs are stuck in the
1741 queue and are never sent.
1745 Description: filter threads hungs on waiting journal commit
1746 Details : Cleanup filter group llog code, then only filter group llog will
1747 be only created in the MDS/OST syncing process.
1751 Description: Procfs and llog threads access destoryed import sometimes.
1752 Details : Sync the import destoryed process with procfs and llog threads by
1753 the import refcount and semaphore.
1755 Severity : enhancement
1757 Description: openlock cache of b1_6 port to HEAD
1762 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1763 Details : If insertion of an extent fails, then discard the inode
1764 preallocation and free data blocks else it can lead to duplicate
1769 Description: don't always update ctime in ext3_xattr_set_handle()
1770 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1771 In some cases the ctime should not be updated, for example for
1772 2.0->1.8 compatibility it is necessary to delete an xattr and it
1773 should not update the ctime.
1777 Bugzilla : 15713/16362
1778 Description: Assertion in iopen_connect_dentry in 1.6.3
1779 Details : looking up an inode via iopen with the wrong generation number can
1780 populate the dcache with a disconneced dentry while the inode
1781 number is in the process of being reallocated. This causes an
1782 assertion failure in iopen since the inode's dentry list contains
1783 both a connected and disconnected dentry.
1787 Description: assertion failure in ldlm_handle2lock()
1788 Details : fix a race between class_handle_unhash() and class_handle2object()
1789 introduced in lustre 1.6.5 by bug 13622.
1794 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1795 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1800 Description: oops in sd_iostats_seq_show()
1801 Details : unloading/reloading the scsi low level driver triggers a kernel
1802 bug when trying to access the sd iostat file.
1807 Description: Kernel panics during QLogic driver reload
1808 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1809 patch, causing memory corruption.
1814 Description: journal_dev option does not work in b1_6
1815 Details : pass mount option during pre-mount.
1817 Severity : enhancement
1819 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1820 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1821 extent information of a file. It can be used to map logical blocks
1822 in a file to physical blocks in the block device.
1826 Description: LDLM soft lockups - improvement
1827 Details : It is be possible to send the lock handle along with each read
1828 or write request because the client is already doing a lock match
1829 itself so there isn't any reason the OST should have to re-do that
1833 Frequency : only on Cray X2
1835 Description: X2 build failures
1836 Details : fix build failures on Cray X2.
1840 Description: xid & resent requests
1841 Details : Initialize RPC XID from clock at startup (randomly if clock is
1844 Severity : enhancement
1846 Description: Add lustre_start utility to start or stop multiple Lustre servers
1851 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1852 Details : In case of memory pressure, list_del() can be called twice on
1853 req->rq_history_list, causing a kernel oops.
1857 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1858 Details : fix stack overflow in the distributed lock manager by defering
1859 export eviction after a failed AST to the elt thread instead of
1860 handling it in the dlm interpret routine.
1864 Description: Convert some comments to new format.
1865 Details : Update documenting comments to match doxygen conventions.
1869 Description: Grammar fixes.
1870 Details : A couple of trivial spelling fixes.
1874 Description: OSD_COUNTERS-mandatory
1875 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1876 to simplify the code.
1880 Description: simplify cmm_device freeing logic.
1881 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1885 Description: Add lockdep support to dt_object_operations locking interface.
1886 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1887 parameter indicating lock ordering. Update mdd code to use new
1892 Description: Introduce failloc constants for lockless IO tests.
1893 Details : Add two new failloc constants to test lockless IO. Only one of
1894 them in implemented---another is checked in yet to be landed
1899 Description: Add lockdep support for inode mutex.
1900 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1901 in the situations where Lustre has to lock more than one inode
1906 Description: Add optional invariants checking support.
1907 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1908 configure switch. This macro is to be used for consistency and
1909 sanity checks that are too expensive to be left in `production'
1914 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1915 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1920 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1921 Details : Use spin_lock_nested() in (the only) situation where more than
1922 one ldlm_lock is locked simultaneously. Also, fix possible
1923 dead-lock in ldlm_lock_change_resource() by enforcing particular
1928 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1929 Details : Instead of specifying each ldlm_lock call-back through separate
1930 parameter, wrap them into struct ldlm_callback_suite.
1934 Description: Kill join_lru obd method and its callers.
1935 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1936 in memory---a requirement implemented through ->o_join_lru() obd
1937 method in HEAD. Remove this method and its users.
1941 Description: Add asynchronous ldlm ENQUEUE completion handler.
1942 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1943 case that stock ldlm_completion_ast() cannot handle as it waits
1944 until lock is granted. Introduce new ldlm_completion_ast_async()
1945 for this. Also comment ldlm_completion_ast().
1949 Description: ldlm_error <-> errno conversion.
1950 Details : Add functions to map (rather arbitrary) between LDLM error codes
1951 and standard errno values. CLIO needs this to prevent LDLM specific
1952 constants from escaping ldlm and osc.
1956 Description: Kill unused ldlm_handle2lock_ns() function.
1957 Details : Kill unused ldlm_handle2lock_ns() function.
1961 Description: Add lu_ref support to ldlm_lock
1962 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1963 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1964 and ldlm_resource. LDLM interface has to be changed, because code that
1965 releases a reference on a lock, has to "know" what reference this is.
1966 In the most frequent case
1968 lock = ldlm_handle2lock(handle);
1970 LDLM_LOCK_PUT(lock);
1972 no changes are required. When any other reference (received _not_ from
1973 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1974 instead of LDLM_LOCK_PUT().
1976 Arguably, changes are pervasive, and interface requires some discipline
1977 for proper use. On the other hand, it was very instrumental in finding
1978 a few leaked lock references.
1982 Description: Add ldlm_lock_addref_try().
1983 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1984 attempts to addref a lock that might be being canceled
1989 Description: Add ldlm_weigh_callback().
1990 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1991 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1992 hard-coded `number of pages' logic.
1996 Description: Add lockdep annotations to llog code.
1997 Details : Use appropriately tagged _nested() locking calls in the places
1998 where llog takes more than one ->lgh_lock lock.
2002 Description: Add loi_kms_set().
2003 Details : Wrap kms updates into a helper function.
2007 Description: Constify instances of struct lsm_operations.
2008 Details : Constify instances of struct lsm_operations.
2012 Description: lu_conf support.
2013 Details : On a server, a file system object is uniquely identified
2014 by a fid, which is sufficient to locate and load all object
2015 state (inode). On a client, on the other hand, more data are
2016 necessary instantiate an object. Change lu_object_find() and
2017 friends to take additional `lu_conf' argument describing object.
2018 Typically this includes layout information.
2022 Description: lu_context fixes.
2023 Details : Introduce new lu_context functions that are needed on the client
2024 side, where some system threads (ptlrpcd) are shared by multiple
2025 modules, and so cannot be stopped during module shutdown.
2029 Description: Add start and stop methods to lu_device_type_operations.
2030 Details : Introduce two new methods in lu_device_type_operations, that are
2031 invoked when first instance of a given type is created and last one
2032 is destroyed respectively. This is need by CLIO.
2036 Description: Add lu_ref support to struct lu_device.
2037 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2038 track leaked references.
2042 Description: Introduce lu_kmem_descr.
2043 Details : lu_kmem_descr and its companion interface allow to create
2044 and destroy a number of kmem caches at once.
2048 Description: Fix lu_object finalization race.
2049 Details : Fix a race between lu_object_find() finding an object and its
2050 concurrent finalization. This race is (most likely) not possible
2051 on the server, but might happen on the client.
2055 Description: Introduce lu_ref interface.
2056 Details : lu_ref is a debugging module allowing to track references to
2057 a given object. It is quite cpu expensive, and has to be
2058 explicitly enabled with --enable-lu_ref. See usage description
2063 Description: Factor lu_site procfs stats into a separate function.
2064 Details : Separate lu_site stats printing code into a separate function
2065 that can be reused on a client.
2069 Description: Constify instances of struct {lu,dt,md}_device_operations.
2070 Details : Constify instances of struct {lu,dt,md}_device_operations.
2074 Description: Introduce struct md_site and move meta-data specific parts of
2075 struct lu_site here.
2076 Details : Move md-specific fields out of struct lu_site into special struct
2077 md_site, so that lu_site can be used on a client.
2081 Description: Kill mdd_lov_destroy().
2082 Details : Remove unused mdd code.
2086 Description: Add st_block checking to multistat.c.
2087 Details : Add st_block checking to multistat.c.
2091 Description: Add lu_ref support to struct obd_device.
2092 Details : Add lu_ref tracking to obd_device.
2096 Description: Kill obd_set_fail_loc().
2097 Details : Remove unused code.
2101 Description: Add special type for ptlrpc_request interpret functions.
2102 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2103 there. Actual usage will be in CLIO.
2107 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2108 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2109 doesn't work with lockdep.
2113 Description: Add rwv.c test program.
2114 Details : New testing program exercising readv(2) and writev(2) (Qian).
2118 Description: Add sendfile.c test program.
2119 Details : New testing program exercising sendfile(2) (Jay).
2123 Description: Ratelimit a message that can be very frequent.
2124 Details : Ratelimit a memory allocation failure message that can
2129 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2130 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2134 Description: Liblustre build fixes.
2135 Details : Liblustre build fixes.
2139 Description: libcfs: add cfs_{need,cond}_resched() interface.
2140 Details : libcfs: add cfs_{need,cond}_resched() definition and
2141 implementations for Linux, NT, and liblustre.
2143 Severity : enhancement
2145 Description: More exported tunables for mballoc
2146 Details : Add support for tunable preallocation window and new tunables for
2147 large/small requests
2151 Description: Detect corruption of block bitmap and checking for preallocations
2152 Details : Checks validity of on-disk block bitmap. Also it does better
2153 checking of number of applied preallocations. When corruption is
2154 found, it turns filesystem readonly to prevent further corruptions.
2158 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2159 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2160 window to 0. If it does hit cache miss, and read-pattern is still
2161 stride-io mode, does not reset the stride window, but also does not
2162 increase the stride window length in this case.
2166 Frequency : only for big-endian servers
2167 Description: Check if system is big-endian while mounting fs with extents feature
2168 Details : Mounting a filesystem with extents feature will fail on big-endian
2169 systems since ext3-based ldiskfs is not supported on big-endian
2170 systems. This can be over-riden with "bigendian_extents" mount option.
2172 Severity : enhancement
2174 Description: The root squash functionality
2175 Details : A security feature, which is to prevent users from being able
2176 to mount lustre on their desktop, run as root, and delete
2177 all of the files in the filesystem. The goal is accomplished by
2178 remapping user id (UID) and group id (GID) of the root user to
2179 a UID and GID specified by the system administartor via Lustre
2180 configuration management server (MGS). The functionality also
2181 allows to specify sets of clients for which the remapping does
2186 Description: Excessive recovery window
2187 Details : With AT enabled, the recovery window can be excessively long (6000+
2188 seconds). To address this problem, we no longer use
2189 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2190 timeout no longer depends on the service time, it is set to
2191 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2192 time via pb_service_time.
2196 Description: Watchdog triggered on MDS failover
2197 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2198 the OSTs know that the MDS "UUID" can be reused for the same export
2199 from a different NID, so we do not need to wait for the export to be
2203 Frequency : rare, only if using MMP with Linux RAID
2205 Description: MMP doesn't work with Linux RAID
2206 Details : While using HA for Lustre servers with Linux RAID, it is possible
2207 that MMP will not detect multiple mounts. To make this work we
2208 need to unplug the device queue in RAID when the MMP block is being
2209 written. Also while reading the MMP block, we should read it from
2210 disk and not the cached one.
2212 Severity : enhancement
2214 Description: open file using fid
2215 Details : A file can be opened using just its fid, like
2216 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2219 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2222 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2223 Details : add proper locking for creation of jbd_4k slab cache
2227 Description: MMP check in ext3_remount() fails without displaying any error
2228 Details : When multiple mount protection fails during remount, proper error
2231 Severity : enhancement
2233 Description: Allow stripe size to be up to 4G-64k
2234 Details : Fix math logic to allow large stripe sizes.
2238 Description: add check for >8TB ldiskfs filesystems
2239 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2240 Don't allow >8TB ldiskfs filesystems to be mounted without
2241 force_over_8tb mount option
2246 Descriptoin: OSS complains "too many blocks" and crashes
2247 Details : In filter_commitrw_write():
2248 1. lquota_chkquota();
2249 2. obdo_from_inode();
2250 3. lquota_pending_commit();
2251 "2." has changed uid and gid in this case, then caused
2252 lquota_chkquota() and lquota_pending_commit() to use
2253 different uid and gid. That is the root of the bug.
2257 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2258 Details : Detect and print out the 'stripe count = -1' case properly.
2260 --------------------------------------------------------------------------------
2262 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2264 * Support for kernels:
2265 2.6.5-7.283 (SLES 9),
2266 2.6.9-55.EL (RHEL 4),
2267 2.6.16.46-0.14 (SLES 10),
2268 2.6.18.8 vanilla (kernel.org)
2269 * Client support for unpatched kernels:
2270 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2271 2.6.16 - 2.6.22 vanilla (kernel.org)
2272 * Due to recently discovered recovery problems, we do not recommend
2273 using patchless RHEL 4 clients with this or any earlier release.
2274 * Recommended e2fsprogs version: 1.39.cfs8
2275 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2276 * Starting with this release, the ldiskfs backing filesystem required
2277 by Lustre is now in its own package, lustre-ldiskfs. This package
2278 should be installed. It is versioned separately from Lustre and
2279 may be released separately in future.
2283 Description: block reactivating mgc import until all deactivates complete
2284 Details : Fix race when failing back MDT/MGS to itself (testing)
2287 Frequency : at statup only
2289 Description: mds_lov_synchronize race leads to various problems
2290 Details : simultaneous MDT->OST connections at startup can cause the
2291 sync to abort, leaving the OSC in a bad state.
2293 Severity : enhancement
2295 Description: add optional extra BUILD_VERSION info
2296 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2297 to override the lustre version.
2300 Frequency : 2.6.18 servers only
2302 Description: ll_kern_mount() doesn't release the module reference
2303 Details : The ldiskfs module reference count never drops down to 0
2304 because ll_kern_mount() doesn't release the module reference.
2309 Description: server LBUG when using old ost_num_threads parameter
2310 Details : Accept the old ost_num_threads parameter but warn that it
2311 is deprecated, and fix an off-by-one error that caused an LBUG.
2316 Description: Transient SCSI error results in persistent IO issue
2317 Details : iobuf->dr_error is not reinitialized to 0 between two
2321 Frequency : sometimes when underlying device returns I/O errors
2323 Description: OSTs not going read-only during write failures
2324 Details : OSTs are not remounted read-only when the journal commit threads
2325 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2326 instead of the ext3 wrappers.
2330 Description: poor connect scaling with increasing client count
2331 Details : Don't run filter_grant_sanity_check for more than 100 exports
2332 to improve scaling for large numbers of clients.
2335 Frequency : SLES10 only
2337 Description: sanity-quota.sh quotacheck failed: rc = -22
2338 Details : Quotas cannot be enabled on SLES10.
2341 Frequency : liblustre clients only
2343 Description: getdirentries does not give error when run on compute nodes
2344 Details : getdirentries does not fail when the size specified as an argument
2345 is too small to contain at least one entry
2347 Severity : enhancement
2349 Description: Add LNET router traceability for debug purposes
2350 Details : If a checksum failure occurs with a router as part of the
2351 IO path, the NID of the last router that forwarded the bulk data
2352 is printed so it can be identified.
2357 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2358 Details : Due to a race condition, liblustre clients were occasionally
2359 evicted incorrectly.
2361 Severity : enhancement
2363 Description: lfs setstripe use optional parameters instead of postional
2366 Severity : enhancement
2368 Description: Nanosecond timestamp support for ldiskfs
2369 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2370 resolution timestamps. There is not yet support for this at
2371 the Lustre filesystem level.
2374 Frequency : during server recovery
2376 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2377 Details : request with rq_no_resend flag not awake l_wait_event if they get
2381 Frequency : nfs export on patchless client
2383 Description: connectathon hang when test nfs export over patchless client
2384 Details : Disconnected dentry cannot be found with lookup, so we do not need
2385 to unhash it or make it invalid
2388 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2389 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2390 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2395 Description: open req refcounting wrong on reconnect
2396 Details : If reconnect happened between getting open reply from server and
2397 call to mdc_set_replay_data in ll_file_open, we will schedule
2398 replay for unreferenced request that we are about to free.
2399 Subsequent close will crash in variety of ways.
2400 Check that request is still eligible for replay in
2401 mdc_set_replay_data().
2406 Description: disable writes to filesystem when reading health_check file
2407 Details : the default for reading the health_check proc file has changed
2408 to NOT do a journal transaction and write to disk, because this
2409 can cause reads of the /proc file to hang and block HA state
2410 checking on a healthy but otherwise heavily loaded system. It
2411 is possible to return to the previous behaviour during configure
2412 with --enable-health-write.
2414 Severity : enhancement
2416 Description: 64-bit inode version
2417 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2418 to the inode. This will be required for version-based recovery.
2423 Description: MDS fails to start if a duplicate client export is detected
2424 Details : in some rare cases it was possible for a client to connect to
2425 an MDS multiple times. Upon recovery the MDS would detect this
2426 and fail during startup. Handle this more gracefully.
2428 Severity : enhancement
2430 Description: Add -o localflock option to simulate old noflock
2432 Details : This will achieve local-only flock/fcntl locks
2438 Description: log_commit_thread vs filter_destroy race leads to crash
2439 Details : Take import reference before releasing llog record semaphore
2444 Description: Wrong request locking in request set processing
2445 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2446 imp_delayed_list, in this place should be used imp_lock.
2449 Frequency : when reconnection
2451 Description: Grant Leak when osc reconnect to OST
2452 Details : When osc reconnect ost, OST(filter) should check whether it
2453 should grant more space to client by comparing fed_grant and
2454 cl_avail_grant, and return the granted space to client instead
2455 of "new granted" space, because client will call osc_init_grant
2456 to update the client grant space info.
2459 Frequency : when client reconnect to OST
2461 Description: Grant Leak when osc do resend and replay bulk write
2462 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2463 bulk write request, because the grant info while be sync between
2464 OSC and OST when reconnect, and we should ignore the grant info
2465 these of resend/replay write req.
2470 Description: Grant space more than avaiable left space sometimes.
2471 Details : When then OST is about to be full, if two bulk writing from
2472 different clients came to OST. Accord the avaliable space of the
2473 OST, the first req should be permitted, and the second one
2474 should be denied by ENOSPC. But if the seconde arrived before
2475 the first one is commited. The OST might wrongly permit second
2476 writing, which will cause grant space > avaiable space.
2479 Frequency : when client is evicted
2481 Description: Grant might be wrongly erased when osc is evicted by OST
2482 Details : when the import is evicted by server, it will fork another
2483 thread ptlrpc_invalidate_import_thread to invalidate the
2484 import, where the grant will be set to 0. While the original
2485 thread will update the grant it got when connecting. So if
2486 the former happened latter, the grant will be wrongly errased
2487 because of this race.
2492 Description: Checking Stale with correct fid
2493 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2494 to check whether it is stale, because sometimes, we want the
2495 enqueue happened anyway, and op_data.fid2 will not be initialized.
2497 Severity : enhancement
2499 Description: update patchless client
2500 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2503 Frequency : only with 2.4 kernel
2505 Description: random memory corruption
2506 Details : size of struct ll_inode_info is to big for union inode.u and this
2507 can be cause of random memory corruption.
2512 Description: Memory leak in recovery
2513 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2514 It should also check obd_fail before fsfilt_start, otherwise if
2515 fsfilt_start return -EROFS,(failover mds during mds recovery).
2516 then the req will return with repmsg->transno = 0 and rc = EROFS.
2517 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2518 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2519 be freed no matter whether fsfilt_commit success or not.
2522 Frequency : only with huge count clients
2524 Description: Prevents from taking the superblock lock in llap_from_page for
2526 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2527 for llap_from_page call in ll_removepage prevents from taking the
2528 superblock lock for a soon died page.
2533 Description: Not check open intent error before release open handle
2534 Details : in some rare cases, the open intent error is not checked before
2535 release open handle, which may cause
2536 ASSERTION(open_req->rq_transno != 0), because it tries to release
2537 the failed open handle.
2542 Description: Set cat log bitmap only after create log success.
2543 Details : in some rare cases, the cat log bitmap is set too early. and it
2544 should be set only after create log success.
2548 Description: Accessing a block bevice can re-enable I/O when Lustre is
2549 tearing down a device.
2550 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2555 Description: service threads may hog cpus when there are a lot of requests
2556 Details : Insert cond_resched to give other threads a chance to use some CPU
2561 Description: the cat log was not initialized in recovery
2562 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2563 unlink log on mds will not be initialized until mds post
2564 recovery. And also in mds post recovery, the unlink log will
2565 initialization will be done asynchronausly, so there will be race
2566 between add unlink log and unlink log initialization.
2570 Description: brw_stats were being printed incorrectly
2571 Details : brw_stats were being printed as log2 but all of them were not
2572 recorded as log2. Also remove some code duplication arising from
2573 filter_tally_{read,write}.
2577 Frequency : rare, only in recovery.
2578 Description: ASSERTION(req->rq_type != LI_POISON) failed
2579 Details : imp_lock should be held while iterating over imp_sending_list for
2580 prevent destroy request after get timeout in ptlrpc_queue_wait.
2584 Description: replay-single.sh test 52 fails
2585 Details : A lock's skiplist need to be cleanup when it being unlinked
2586 from its resource list.
2590 Description: Short directio read returns full requested size rather than
2592 Details : Direct I/O operations should return actual amount of bytes
2593 transferred rather than requested size.
2595 Severity : enhancement
2597 Description: metadata RPC reduction (e.g. for rm performance)
2598 Details : decrease the amount of synchronous RPC between clients and servers
2599 by canceling conflicing lock before the operation on the client
2600 and packing thier handles into the main operation RPC to server.
2602 Severity : enhancement
2604 Description: Async OSC create to avoid the blocking unnecessarily.
2605 Details : If a OST has no remain object, system will block on the creating
2606 when need to create a new object on this OST. Now, ways use
2607 pre-created objects when available, instead of blocking on an
2608 empty osc while others are not empty. If we must block, we block
2609 for the shortest possible period of time.
2613 Description: improve handling recoverable errors
2614 Details : if request processig with error which can be recoverable on server
2615 request should be resend, otherwise page released from cache and
2618 Severity : enhancement
2620 Description: refine locking for avoid write wrong info into lov_objid file
2621 Details : fix possible races with add new target and write/update data in
2624 --------------------------------------------------------------------------------
2626 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2631 Frequency : on some architectures
2633 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2634 Details : Uninitialized flags sometimes cause configuration commands to
2638 Frequency : patchless clients only
2640 Description: missing __iget() symbol export
2641 Details : The __iget() symbol export is missing. To avoid the need for
2642 this on patchless clients the deathrow inode reaper is turned
2643 off, and we depend on the VM to clean up old inodes. This
2644 dependency was during via the fix for bug 12181.
2646 --------------------------------------------------------------------------------
2648 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2650 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2651 INTEROPERATE with older versions automatically. In many cases a
2652 special upgrade step is needed. Please read the
2653 user documentation before upgrading any part of a live system.
2654 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2655 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2656 * WARNING: Lustre configuration and startup changes are required with
2657 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2661 Severity : enhancement
2663 Description: MountConf
2664 Details : Lustre configuration is now managed via mkfs and mount
2665 commands instead of lmc and lconf. New obd types (MGS, MGC)
2666 are added for dynamic configuration management. See
2667 https://mail.clusterfs.com/wikis/lustre/MountConf for
2670 Severity : enhancement
2672 Description: dynamic OST addition
2673 Details : OSTs can now be added to a live filesystem
2675 Severity : enhancement
2677 Description: startup order invariance
2678 Details : MDTs and OSTs can be started in any order. Clients only
2679 require the MDT to complete startup.
2681 Severity : enhancement
2683 Description: parallel, asynchronous orphan cleanup
2684 Details : orphan cleanup is now performed in separate threads for each
2685 OST, allowing parallel non-blocking operation.
2687 Severity : enhancement
2689 Description: optimized stripe assignment
2690 Details : stripe assignments are now made based on ost space available,
2691 ost previous usage, and OSS previous usage, in order to try
2692 to optimize storage space and networking resources.
2694 Severity : enhancement
2696 Description: Permanently set tunables
2697 Details : All writable /proc/fs/lustre tunables can now be permanently
2698 set on a per-server basis, at mkfs time or on a live system.
2700 Severity : enhancement
2702 Description: Lustre message v2
2703 Details : Add lustre message format v2.
2705 Severity : enhancement
2707 Description: client OST exclusion list
2708 Details : Clients can be started with a list of OSTs that should be
2709 declared "inactive" for known non-responsive OSTs.
2713 Description: SPEC SFS validation failure on NFS v2 over lustre.
2714 Details : Changes the blocksize for regular files to be 2x RPC size,
2715 and not depend on stripe size.
2717 Severity : enhancement
2719 Description: Multiple MD RPCs in flight.
2720 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2721 To avoid overly-overloading MDS, introduce a limit on number of
2722 MDS RPCs in flight for a single client and add /proc controls
2723 to adjust this limit.
2725 Severity : enhancement
2727 Description: client read/write statistics
2728 Details : Add client read/write call usage stats for performance
2729 analysis of user processes.
2730 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2731 file access. extents_stats shows chunk size distribution.
2732 extents_stats_per_process show chunk size distribution per
2735 Severity : enhancement
2737 Description: mds statistics
2738 Details : Add detailed mds operations statistics in
2739 /proc/fs/lustre/mds/*/stats.
2743 Description: Failure of copying files with lustre special EAs.
2744 Details : Client side always return success for setxattr call for lustre
2745 special xattr (currently only "trusted.lov").
2750 Description: Refcount LNET uuids
2751 Details : The global LNET uuid list grew linearly with every startup;
2752 refcount repeated list entries instead of always adding to
2755 Severity : enhancement
2757 Description: Dynamic service threads
2758 Details : Within a small range, start extra service threads
2759 automatically when the request queue builds up.
2762 Frequency : mixed-endian client/server environments
2764 Description: mixed-endian crashes
2765 Details : The new msg_v2 system had some failures in mixed-endian
2768 Severity : enhancement
2770 Description: Easy OST removal
2771 Details : OSTs can be permanently deactivated with e.g. 'lctl
2772 conf_param lustre-OST0001.osc.active=0'
2774 Severity : enhancement
2776 Description: MGS proc entries
2777 Details : Added basic proc entries for the MGS showing what filesystems
2780 Severity : enhancement
2782 Description: provide MGS failover
2783 Details : Added config lock reacquisition after MGS server failover.
2785 Severity : enhancement
2787 Description: add Linux 2.4 support
2788 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2792 Description: a large application tries to do I/O to the same resource and dies
2793 in the middle of it.
2794 Details : Check the req->rq_arrival time after the call to
2795 ost_brw_lock_get(), but before we do anything about
2796 processing it & sending the BULK transfer request. This
2797 should help move old stale pending locks off the queue as
2798 quickly as obd_timeout.
2801 Frequency : when an incorrect nid is specified during startup
2803 Description: ptlrpc connect to non-existant node causes kernel crash
2804 Details : LNET can't be re-entered from an event callback, which
2805 happened when we expire a message after the export has been
2806 cleaned up. Instead, hand the zombie cleanup off to another
2809 Severity : enhancement
2811 Description: plain/inodebits lock performance improvement
2812 Details : Grouping plain/inodebits in granted list by their request modes
2813 and bits policy, thus improving the performance of search through
2817 Frequency : only if OST filesystem is corrupted
2819 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2820 Details : for a short time RPCs with bulk IO are in the replay list,
2821 but replay of bulk IOs is unimplemented. If the OST filesystem
2822 is corrupted due to disk cache incoherency and then replay is
2823 started it is possible to trip an assertion. Avoid putting
2824 committed RPCs into the replay list at all to avoid this issue.
2827 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2829 Description: System hang on startup
2830 Details : This bug allowed the liblustre (e.g. catamount) client to
2831 return to the app before handling all startup RPCs. This
2832 could leave the node unresponsive to lustre network traffic
2833 and manifested as a server ptllnd timeout.
2835 Severity : enhancement
2837 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2838 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2839 boolean to control whether to print peer debug info when a
2840 client's RPC times out.
2843 Frequency : only for kernels with patches from Lustre below 1.4.3
2845 Description: Remove old rdonly API
2846 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2849 Frequency : only for devices with external journals
2851 Description: Set external device read-only also
2852 Details : During a commanded failover stop, we set the disk device
2853 read-only while the server shuts down. We now also set any
2854 external journal device read-only at the same time.
2857 Frequency : when upgrading from 1.4 while trying to change parameters
2859 Description: The wrong (new) MDC name was used when setting parameters for
2860 upgraded MDT's. Also allows changing of OSC (and MDC)
2861 parameters if --writeconf is specified at tunefs upgrade time.
2864 Frequency : when setting specific ost indicies
2866 Description: QOS code breaks on skipped indicies
2867 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2868 created with --index need not be sequential.
2870 Severity : enhancement
2872 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2873 Details : The uninit_groups feature works in conjunction with the kernel
2874 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2875 up the pass1 processing of e2fsck. This is a read-only feature
2876 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2877 mount filesystems that have had this feature enabled.
2879 Severity : enhancement
2881 Description: Improve multi-block allocation algorithm to avoid fragmentation
2882 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2883 allocation locality and avoid filesystem fragmentation.
2885 ------------------------------------------------------------------------------
2887 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2889 * Support for kernels:
2890 2.4.21-47.0.1.EL (RHEL 3)
2891 2.6.5-7.283 (SLES 9)
2892 2.6.9-42.0.10.EL (RHEL 4)
2893 2.6.12.6 vanilla (kernel.org)
2894 2.6.16.27-0.9 (SLES 10)
2895 * Recommended e2fsprogs version: 1.39.cfs5
2897 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2901 Frequency : occasional, depends on client load and configuration
2902 Bugzilla : 12181, 12203
2903 Description: data loss for recently-modified files
2905 Details : In some cases it is possible that recently written or created
2906 files may not be written to disk in a timely manner (this should
2907 normally be within 30s unless client IO load is very high).
2908 The problem appears as zero-length files or files that are a
2909 multiple of 1MB in size after a client crash or client eviction
2910 that are missing data at the end of the file.
2912 This problem is more likely to be hit on clients where files are
2913 repeatedly created and unlinked in the same directory, clients
2914 have a large amount of RAM, have many CPUs, the filesystem has
2915 many OSTs, the clients are rebooted frequently, and/or the files
2916 are not accessed by other nodes after being written.
2918 The presence of the problem can be detected by looking at
2919 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2920 smaller than the second (nr_unused) then dirty files will not
2921 be flushed automatically to disk. "sync; sleep 10" should be
2922 run several times on the node before unmounting it to update
2923 Lustre (this is also safe to run on nodes without this problem).
2925 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2926 that can cause this same problem, so customers using that kernel
2927 also need to update the kernel in addition to Lustre. In order
2928 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2930 It is normal that files written just before a client crash (less
2931 than 30s) may not yet have been flushed to disk, even for local
2935 Frequency : frequent on thin XT3 nodes
2937 Description: UUID collision on thin XT3 Linux nodes
2938 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2939 randomly, since we relied on an insufficiently-seeded PRNG.
2944 Description: OSS hangs after "All ost request buffers busy"
2945 Details : A deadlock between quota and journal operations caused OSS
2946 hangs after printing "All ost request buffers busy."
2949 Frequency : always on liblustre builds
2951 Description: Cleanup compiler warnings on liblustre
2954 Frequency : always on liblustre builds on XT3
2956 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2960 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2961 > 32000 subdirectories
2962 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2963 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2964 subdirectory count crosses 32000. This will aid e2fsck to
2965 correctly handle more than 32000 subdirectories.
2968 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2970 Description: System hang on startup
2971 Details : This bug allowed the liblustre (e.g. catamount) client to
2972 return to the app before handling all startup RPCs. This
2973 could leave the node unresponsive to lustre network traffic
2974 and manifested as a server ptllnd timeout.
2976 Severity : enhancement
2978 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2979 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2980 boolean to control whether to print peer debug info when a
2981 client's RPC times out.
2986 Description: make O_SYNC working on 2.6 kernels
2987 Details : 2.6 kernels use different method for mark pages for write,
2988 so need add a code to lustre for O_SYNC work.
2993 Description: Failure to close file and release space on NFS
2994 Details : Put inode details into lock acquired in ll_intent_file_open.
2995 Use mdc_intent_lock in ll_intent_open to properly
2996 detect all kind of errors unhandled by mdc_enqueue.
3001 Description: proc file read during shutdown sometimes raced obd removal,
3003 Details : Add lock to prevent obd access after proc file removal.
3006 Frequency : Only for files larger than 4GB on 32-bit clients.
3008 Description: improperly doing page alignment of locks
3009 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3010 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3015 Description: RPCs being resent when they shouldn't be
3016 Details : Some RPCs that should not be resent are being resent. This
3017 can cause inconsistencies in the RPC state machine. Do not
3018 resend such requests.
3021 Frequency : rare, only with NFS export
3023 Description: Crash on NFS re-export node
3024 Details : under very unusual load conditions an assertion is hit in
3025 ll_intent_file_open()
3028 Frequency : only if OST filesystem is corrupted
3030 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3031 Details : for a short time RPCs with bulk IO are in the replay list,
3032 but replay of bulk IOs is unimplemented. If the OST filesystem
3033 is corrupted due to disk cache incoherency and then replay is
3034 started it is possible to trip an assertion. Avoid putting
3035 committed RPCs into the replay list at all to avoid this issue.
3040 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3041 Details : Large single O_DIRECT read and write calls can fail to allocate
3042 a sufficiently large buffer to process the request. In case of
3043 allocation failure the allocation is retried with a smaller
3044 buffer and broken into smaller requests.
3046 Severity : enhancement
3048 Description: Add -o localflock option to simulate old noflock behaviour.
3049 Details : This will achieve local-only flock/fcntl locks coherentness.
3054 Description: versioning check is incomplete
3055 Details : Checking the version difference of client vs. server, report
3056 error if the gap is too big.
3061 Description: add support PG_writeback bit
3062 Details : add support for PG_writeback bit for Lustre, for more carefull
3063 work with page cache in 2.6 kernel. This also fix some deadlocks
3064 and remove hack for work O_SYNC with 2.6 kernel.
3066 Severity : enhancement
3068 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3069 Details : The uninit_groups feature works in conjunction with the kernel
3070 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3071 up the pass1 processing of e2fsck. This is a read-only feature
3072 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3073 mount filesystems that have had this feature enabled.
3075 Severity : enhancement
3077 Description: Improve multi-block allocation algorithm to avoid fragmentation
3078 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3079 allocation locality and avoid filesystem fragmentation.
3081 ------------------------------------------------------------------------------
3083 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3085 * Support for kernels:
3086 2.6.9-42.0.3.EL (RHEL 4)
3087 2.6.5-7.276 (SLES 9)
3088 2.4.21-47.0.1.EL (RHEL 3)
3089 2.6.12.6 vanilla (kernel.org)
3090 2.6.16.21-0.8 (SLES10)
3091 * Recommended e2fsprogs version: 1.39.cfs2-0
3093 * The backwards-compatible /proc/sys/portals symlink has been removed
3094 in this release. Before upgrading, please ensure that you change
3095 any configuration scripts or /etc/sysctl.conf files that access
3096 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3097 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3098 in advance of the upgrade on any system running Lustre 1.4.6 or
3099 newer, since /proc/sys/lnet was added in that version.
3100 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3104 Frequency : only when quota is used
3106 Description: avoid scanning export list for quota master
3107 Details : Change the algorithms to avoid scanning export list in order
3108 to improve the efficiency.
3111 Frequency : MDS failover only, very rarely
3113 Description: "went back in time" messages on mds failover
3114 Details : The greatest transno may be lost when the current operation
3115 finishes with an error (transno==0) and the client's last_rcvd
3116 record is over-written. Save the greatest transno in the
3117 mds_last_transno for this case.
3120 Frequency : always for specific kernels and striping counts
3122 Description: client may get "Matching packet too big" without ACL support
3123 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3124 when trying to access files in certain configurations. The
3125 clients should in fact be denied when mounting because they do
3126 not understand ACLs.
3129 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3131 Description: many clients connecting with IO in progress causes connect timeouts
3132 Details : Avoid synchronous journal commits to avoid delays caused by many
3133 clients connecting/disconnecting when bulk IO is in progress.
3134 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3135 OST_IO_PORTAL to avoid delays behind potentially many pending
3139 Frequency : occasionally with multiple writers to a single file
3141 Description: shared writes to file may result in wrong size reported by stat()
3142 Details : Allow growing of kms when extent lock is cancelled
3145 Frequency : always with random mmap IO to multi-striped file
3147 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3148 Details : Only if the hole is at the end of OST object so that kms is too
3149 small. Fix is to increase kms accordingly in ll_nopage.
3152 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3154 Description: writes to a missing object would leak memory on the OST
3155 Details : If there is an inconsistency between the MDS and OST filesystems,
3156 such that the MDS references an object that doesn't exist, writes
3157 to that object will leak memory due to incorrect cleanup in the
3158 error handling path, eventually running out of memory on the OST.
3163 Description: Creating too long symlink causes lustre errors
3164 Details : Check symlink and name lengths before sending requests to MDS.
3167 Frequency : only if flock is enabled (not on by default)
3169 Description: posix locks not released on fd closure on 2.6.9+
3170 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3171 kernels, this caused such locks not to be released on fd close and
3172 then assertions on fs unmount about still used locks.
3175 Frequency : MDS failover only, very rarely
3177 Description: clients may get ASSERTION(granted_lock != NULL)
3178 Details : When request was taking a long time, and a client was resending
3179 a getattr by name lock request. The were multiple lock requests
3180 with the same client lock handle and
3181 mds_getattr_name->fixup_handle_for_resent_request found one of the
3182 lock handles but later failed with ASSERTION(granted_lock != NULL).
3187 Description: handle->h_buffer_credits > 0, assertion failure
3188 Details : h_buffer_credits is zero after truncate, causing assertion
3189 failure. This patch extends the transaction or creates a new
3193 Frequency : NFS re-export or patchless client
3194 Bugzilla : 11179, 10796
3195 Description: Crash on NFS re-export node (__d_move)
3196 Details : We do not want to hash the dentry if we don't have a lock.
3197 But if this dentry is later used in d_move, we'd hit uninitialised
3198 list head d_hash, so we just do this to init d_hash field but
3199 leave dentry unhashed.
3202 Frequency : NFS re-export or patchless client
3204 Description: NFS exports has problem with symbolic link
3205 Details : lustre client didn't properly install dentry when re-exported
3206 to NFS or running patchless client.
3209 Frequency : NFS re-export or patchless client
3211 Description: Various nfs/patchless fixes.
3212 Details : fixes reuse disconected alias for lookup process - this fixes
3213 warning "find_exported_dentry: npd != pd",
3214 fix permission error with open files at nfs.
3215 fix apply umask when do revalidate.
3218 Frequency : occasional
3220 Description: Crash on NFS re-export node
3221 Details : calling clear_page() on the wrong pointer triggered oops in
3222 generic_mapping_read().
3225 Frequency : rarely, using O_DIRECT IO
3227 Description: unaligned directio crashes client with LASSERT
3228 Details : check for unaligned buffers before trying any requests.
3231 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3233 Description: stale data returned from RAID cache
3234 Details : If only a small amount of IO is done to the RAID device before
3235 reading it again it is possible to get stale data from the RAID
3236 cache instead of reading it from disk.
3239 Frequency : always for sles10 kernel
3241 Description: sles10 support
3242 Details : ll_follow_link: compile fixes and using of nd_set_link
3243 under newer kernels.
3246 Frequency : depends on arch, kernel and compiler version, always on sles10
3249 Description: recursive or deep enough symlinks cause stack overflow
3250 Details : getting rid of large stack-allocated variable in
3254 Frequency : depends on hardware
3256 Description: lustre write performance loss in the SLES10 kernel
3257 Details : the performance loss is caused by using of write barriers in the
3258 ext3 code. The SLES10 kernel turns barrier support on by
3259 default. The fix is to undo that change for ldiskfs.
3261 ------------------------------------------------------------------------------
3263 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3265 * Support for kernels:
3266 2.6.9-42.0.3EL (RHEL 4)
3267 2.6.5-7.276 (SLES 9)
3268 2.4.21-47.0.1.EL (RHEL 3)
3269 2.6.12.6 vanilla (kernel.org)
3273 Frequency : quota enabled and large files being deleted
3275 Description: releasing more than 4GB of quota at once hangs OST
3276 Details : If a user deletes more than 4GB of files on a single OST it
3277 will cause the OST to spin in an infinite loop. Release
3278 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3283 Description: statfs data retrieved from /proc may be stale or zero
3284 Details : When reading per-device statfs data from /proc, in the
3285 {kbytes,files}_{total,free,avail} files, it may appear
3286 as zero or be out of date.
3289 Frequency : systems with MD RAID1 external journal devices
3291 Description: lconf's call to blkid is confused by RAID1 journal devices
3292 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3293 of returning all block devices that match the journal UUID.
3296 Frequency : always, for aggregate stripe size over 4GB
3298 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3299 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3300 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3303 Frequency : always if "lfs find" used on a local file/directory
3305 Description: "lfs find" segfaults if used on a local file/directory
3306 Details : The case where a directory component was not specified wasn't
3307 handled correctly. Handle this properly.
3310 Frequency : always on ppc64
3312 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3313 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3314 perform endianness conversion.
3317 Frequency : rarely (truncate to non-zero file size after write under load)
3318 Bugzilla : 10730, 10687
3319 Description: Files padded with zeros to next 4K multiple
3320 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3321 it is possible that files that are truncated to a non-zero size
3322 immediately after being written are filled with zero bytes beyond
3323 the truncated size. No file data is lost.
3325 Severity : enhancement
3327 Description: Allow recovery/failover for liblustre clients.
3328 Details : liblustre clients were unaware of failover configurations until
3331 Severity : enhancement
3333 Description: user file locks should fail when not mounting with flock option
3334 Details : Set up an error-returning stub in ll_file_operations.lock field
3335 to prevent incorrect behaviour when client is mounted without
3336 flock option. Also, set up properly f_op->flock field for
3340 Frequency : always on ia64
3342 Description: "lfs df" loops on printing out MDS statfs information
3343 Details : The obd_ioctl_data was not initialized and in some systems
3344 this caused a failure during the ioctl that did not return
3345 an error. Initialize the struct and return an error on failure.
3348 Frequency : SLES 9 only
3350 Description: Error of copying files with lustre special EAs as root
3351 Details : Client side always return success for setxattr call for lustre
3352 special xattr (currently only "trusted.lov").
3355 Frequency : rarely on clusters with both ia64+i386 clients
3357 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3358 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3359 clients are doing concurrent writes to the same file it is
3360 possible that the ia64 clients may LASSERT because the OST
3361 extent locks are not PAGE_SIZE aligned. Ensure that grown
3362 locks are always aligned on the request boundary.
3365 Frequency : specific use, occasional
3367 Description: Overwriting in use executable truncates on-disk binary image
3368 Details : If one node attempts to overwrite an executable in use by
3369 another node, we now correctly return ETXTBSY instead of
3370 truncating the file.
3372 Severity : enhancement
3374 Description: Async OSC create to avoid the blocking unnecessarily.
3375 Details : If a OST has no remain object, system will block on the creating
3376 when need to create a new object on this OST. Now, ways use
3377 pre-created objects when available, instead of blocking on an
3378 empty osc while others are not empty. If we must block, we block
3379 for the shortest possible period of time.
3384 Description: chmod on Lustre root is propagated to other clients
3385 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3386 invalid by the follow_mount time.
3389 Frequency : liblustre clients only
3391 Description: Race in 'instant cancel' lock handling could lead to such locks
3392 never to be granted in case of SMP MDS
3393 Details : Do not destroy not yet granted but cbpending locks in
3397 Frequency : replay/resend of open
3399 Description: non null lock assetion failure in mds_intent_policy
3400 Details : Trying to replay/resend lockless open requests resulted in
3401 mds_open() returning 0 with no lock. Now it sets a flag if
3402 a lock is going to be returned.
3404 Severity : enhancement
3406 Description: Checksum enhancements
3407 Details : New checksum enhancements allow for resending RPCs that failed
3410 Severity : enhancement
3412 Description: Tunables on number of dirty pages in cacche
3413 Details : Allow to set limit on number of dirty pages cached.
3418 Description: client crash on unmount - lock still has references
3419 Details : In some error handling cases it was possible to leak a lock
3420 reference on a client while accessing a file. This was not
3421 harmful to the client during operation, but would cause the
3422 client to crash when the filesystem is unmounted.
3425 Frequency : specific case, rare
3427 Description: ETXTBSY on mds though file not in use by client
3428 Details : ETXTBSY is no longer incorrectly returned when attempting to
3429 chmod or chown a directory that the user previously tried to
3430 execute or a currently-executing binary.
3433 Frequency : extremely rare except on liblustre-based clients
3435 Description: Lustre space not freed when files are deleted
3436 Details : Clean up open-unlinked files after client eviction. Previously
3437 the unlink was skipped and the files remained as orphans.
3442 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3443 Details : In some cases it was possible to send a blocking callback to
3444 a client doing a glimpse, even though that client didn't get
3445 a lock granted. When the glimpse lock is cancelled on the OST
3446 the freed lock is left on the waiting list and corrupted the list.
3449 Frequency : all core dumps
3451 Description: Broke core dumps to lustre
3452 Details : Negative dentry may be unhashed if parent does not have UPDATE
3453 lock, but some callers, e.g. do_coredump, expect dentry to be
3454 hashed after successful create, hash it in ll_create_it.
3456 ------------------------------------------------------------------------------
3458 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3460 * Support for kernels:
3461 2.6.9-42.0.2.EL (RHEL 4)
3462 2.6.5-7.276 (SLES 9)
3463 2.4.21-40.EL (RHEL 3)
3464 2.6.12.6 vanilla (kernel.org)
3468 Frequency : always on RHEL 3
3470 Description: Number of open files grows over time
3471 Details : The number of open files grows over time, whether or not
3472 Lustre is started. This was due to a filp leak introduced
3473 by one of our kernel patches.
3475 ------------------------------------------------------------------------------
3477 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3479 * Support for kernels:
3480 2.6.9-42.EL (RHEL 4)
3481 2.6.5-7.276 (SLES 9)
3482 2.4.21-40.EL (RHEL 3)
3483 2.6.12.6 vanilla (kernel.org)
3488 Bugzilla : 5719, 9635, 9792, 9684
3489 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3490 Details : If a server is under heavy load and cannot reply to new
3491 connection requests before the client resends the (re)connect,
3492 the connection handling code can behave badly if two service
3493 threads are concurrently handing separate (re)connections from
3494 the same client. Add better locking to the connection handling
3495 code, and ensure that only a single connection will be processed
3496 for a given client UUID, even if the lock is dropped.
3498 Severity : enhancement
3500 Description: add TCP zero-copy support to kernel
3501 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3502 sends if the hardware supports scatter-gather and checksumming.
3503 This allows socklnd to do client-write and server-read more
3504 efficiently and reduce CPU utilization from skbuf copying.
3507 Frequency : only if NFS exporting from client
3509 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3510 Details : If setting attributes on a file created under NFS that had
3511 never been opened it would be possible to oops the client
3512 if the file had no objects.
3515 Frequency : always for liblustre
3517 Description: liblustre client does MDS+OSTs setattr RPC for each write
3518 Details : When doing a write from a liblustre client, the client
3519 incorrectly issued an RPC to the MDS and each OST the file was
3520 striped over in order to update the timestamps. When writing
3521 with small chunks and many clients this could overwhelm the MDS
3522 with RPCs. In all cases it would slow down the write because
3523 these RPCs are unnecessary.
3525 Severity : enhancement
3527 Description: allow number of MDS service threads to be changed at module load
3528 Details : It is now possible to change the number of MDS service threads
3529 running. Adding "options mds mds_num_threads={N}" to the MDS's
3530 /etc/modprobe.conf will set the number of threads for the next
3531 time Lustre is restarted (assuming the "mds" module is also
3532 reloaded at that time). The default number of threads will
3533 stay the same, 32 for most systems.
3538 Description: OST crash if filesystem is unformatted or corrupt
3539 Details : If an OST is started on a device that has never been formatted
3540 or if the filesystem is corrupt and cannot even mount then the
3541 error handling cleanup routines would dereference a NULL pointer.
3546 Description: NULL pointer deref in llap_from_page.
3547 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3548 incorrect) mapping if the page was truncated/reclaimed while it was
3549 searched for. Check for this condition and skip such pages when
3550 doing readahead. Introduce extra check to llap_from_page() to
3551 verify page->mapping->host is non-NULL (so page is not anonymous).
3554 Frequency : Sometimes when using sys_sendfile
3556 Description: "page not covered by a lock" warnings from ll_readpage
3557 Details : sendfile called ll_readpage without right page locks present.
3558 Now we introduced ll_file_sendfile that does necessary locking
3559 around call to generic_file_sendfile() much like we do in
3563 Frequency : with certain MDS communication failures at client mount time
3565 Description: NULL pointer deref after failed client mount
3566 Details : a client connection request may delayed by the network layer
3567 and not be sent until after the PTLRPC layer has timed out the
3568 request. If the client fails the mount immediately it will try
3569 to clean up before the network times out the request. Add a
3570 reference from the request import to the obd device and delay
3571 the cleanup until the network drops the request.
3574 Frequency : occasionally during client (re)connect
3576 Description: assertion failure during client (re)connect
3577 Details : processing a client connection request may be delayed by the
3578 client or server longer than the client connect timeout. This
3579 causes the client to resend the connection request. If the
3580 original connection request is replied in this interval, the
3581 client may trip an assertion failure in ptlrpc_connect_interpret()
3582 which thought it would be the only running connect process.
3585 Frequency : only with obd_echo servers and clients that are rebooted
3587 Description: kernel BUG accessing uninitialized data structure
3588 Details : When running an obd_echo server it did not start the ping_evictor
3589 thread, and when a client was evicted an uninitialized data
3590 structure was accessed. Start the ping_evictor in the RPC
3591 service startup instead of the OBD startup.
3593 Severity : enhancement
3594 Bugzilla : 10193 (patchless)
3595 Description: Remove dependency on various unexported kernel interfaces.
3596 Details : No longer need reparent_to_init, exit_mm, exit_files,
3597 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3600 Frequency : rare (only users of deprecated and unsupported LDAP config)
3602 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3603 Details : LDAP apparently contains 'lustreName' attributes instead of
3604 'name'. A simple remapping of the name is sufficient.
3607 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3609 Description: waiting_locks_callback trips kernel BUG if client is evicted
3610 Details : Running with the dump_on_timeout debug flag turned on makes
3611 it possible that the waiting_locks_callback() can try to dump
3612 the Lustre kernel debug logs from an interrupt handler. Defer
3613 this log dumping to the expired_lock_main() thread.
3615 Severity : enhancement
3617 Description: Support NFS exporting on 2.6 kernels.
3618 Details : Implement non-rawops metadata methods for NFS server to use without
3619 changing NFS server code.
3622 Frequency : very rare (synthetic metadata workload only)
3624 Description: two racing renames might cause an MDS thread to deadlock
3625 Details : Running the "racer" program may cause one MDS thread to rename
3626 a file from being the source of a rename to being the target of
3627 a rename at exactly the same time that another thread is doing
3628 so, and the second thread has already enqueued these locks after
3629 doing a lookup of the target and is trying to relock them in
3630 order. Ensure that we don't try to re-lock the same resource.
3633 Frequency : only very large systems with liblustre clients
3635 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3636 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3637 all OSTs in parallel. This allows the eviction of stale liblustre
3638 clients to proceed much faster than if they were done in series,
3639 and also offers similar improvements for other set_info RPCs.
3644 Description: excessive CPU usage during initial read phase on client
3645 Details : During the initial read phase on a client, it would agressively
3646 retry readahead on the file, consuming too much CPU and impacting
3647 performance (since 1.4.5.8). Improve the readahead algorithm
3648 to avoid this, and also improve some other common cases (read
3649 of small files in particular, where "small" is files smaller than
3650 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3655 Description: MDS crash when receiving packet with unknown intent.
3656 Details : Do not LBUG in unknown intent case, just return -EFAULT
3658 Severity : enhancement
3659 Bugzilla : 9293, 9385
3660 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3661 Details : Do not serialize getattr (non-intent version) and statfs.
3664 Frequency : occasional, when OST network is overloaded/intermittent
3666 Description: client evicted by OST after bulk IO timeout
3667 Details : If a client sends a bulk IO request (read or write) the OST
3668 may evict the client if it is unresposive to its data GET/PUT
3669 request. This is incorrect if the network is overloaded (takes
3670 too long to transfer the RPC data) or dropped the OST GET/PUT
3671 request. There is no need to evict the client at all, since
3672 the pinger and/or lock callbacks will handle this, and the
3673 client can restart the bulk request.
3676 Frequency : Always when mmapping file with no objects
3678 Description: client crashes when mmapping file with no objects
3679 Details : Check that we actually have objects in a file before doing any
3680 operations on objects in ll_vm_open, ll_vm_close and
3686 Description: Request leak when working with deleted CWD
3687 Details : Introduce advanced request refcount tracking for requests
3688 referenced from lustre intent.
3690 Severity : Enhancement
3692 Description: Cache open file handles on client.
3693 Details : MDS now will return special lock along with openhandle, if
3694 requested and client is allowed to hold openhandle, even if unused,
3695 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3696 closing files for every read/write openration.
3698 Severity : Enhancement
3700 Description: Cache open negative dentries on client when possible.
3701 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3702 negative dentries on lock revocation.
3707 Description: Remounting a client read-only wasn't possible with a zconf mount
3708 Details : It wasn't possible to remount a client read-only with llmount.
3710 Severity : enhancement
3711 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3712 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3713 MPICH is included to add Lustre-specific ADIO interfaces.
3714 This is based closely on the UFS ADIO layer and only differs
3715 in file creation, in order to allow the OST striping to be set.
3716 This is user-contributed code and not supported by CFS.
3721 Description: extended inode attributes (immutable, append-only) work improperly
3722 when 2.4 and 2.6 kernels are used on client/server or vice versa
3723 Details : Introduce kernel-independent values for these flags.
3725 Severity : enhancement
3728 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3729 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3730 granularity. Now, Fractional MB granularity is be supported,
3731 this is very useful for low memory system.
3733 Severity : enhancement
3735 Description: Getattr by fid
3736 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3737 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3738 have locks and avoids getattr rpc requests for it on every stat.
3741 Frequency : Always, for filesystems larger than 2TB
3743 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3744 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3745 filesystems larger than 2TB to be mounted with mballoc enabled.
3748 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3750 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3751 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3752 filesystems larger than 2TB is mounted then the kernel will
3753 silently corrupt the start of the filesystem. CONFIG_LBD is
3754 enabled for all CFS-supported kernels, but the possibility of
3755 this happening with a modified kernel config exists.
3757 Severity : enhancement
3759 Description: add client O_DIRECT support for 2.6 kernels
3760 Details : It is now possible to do O_DIRECT reads and writes to files
3761 in the Lustre client mountpoint on 2.6 kernel clients.
3763 Severity : enhancement
3765 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3766 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3767 parallel, not waiting for response from every OST before sending
3768 a rpc to the next OST.
3773 Description: setattr vs write race when updating file timestamps
3774 Details : Client processes that update a file timestamp into the past
3775 right after writing to the file (e.g. tar) it is possible that
3776 the updated file modification time can be reset to the current
3777 time due to a race between processing the setattr and write RPC.
3779 Severity : enhancement
3781 Description: Bring 'lfs find' closer in line with regular Linux find.
3782 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3783 -print0 options and obtains all the needed info through the lustre
3786 Severity : enhancement
3788 Description: support up to 1024 configured devices on one node
3789 Details : change obd_dev array from statically allocated to dynamically
3790 allocated structs as they are first used to reduce memory usage
3795 Description: Flush dirty partially truncated pages during truncate
3796 Details : Immediatelly flush partially truncated pages in filter_setattr,
3797 this way we completely avoid having any pages in page cache on OST
3798 and can retire ugly workarounds during writes to flush such pages.
3803 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3804 Details : protect inode from truncation within vfs_unlink() context
3805 just take a reference before calling vfs_unlink() and release it
3806 when parent's i_sem is free.
3811 Description: last_id value checked outside lock on OST caused LASSERT failure
3812 Details : If there were multiple MDS->OST object precreate requests in
3813 flight, it was possible that the OST's last object id was checked
3814 outside a lock and incorrectly tripped an assertion. Move checks
3815 inside locks, and discard old precreate requests.
3818 Frequency : always, if extents are used on OSTs
3820 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3821 Details : OSTs using the extents format would not zero the high 16 bits of
3822 the index physical block number. This is not a problem for any
3823 OST filesystems smaller than 16TB, and no kernels support ext3
3824 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3825 new/modified files) and can be fixed for existing filesystems
3826 with e2fsprogs-1.39-cfs1.
3831 Description: import connection selection may be incorrect if timer wraps
3832 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3833 import connections to be ignored if the 32-bit jiffies counter
3834 wraps. Use a 64-bit jiffies counter.
3837 Frequency : very large clusters immediately after boot
3839 Description: LNET request buffers exhausted under heavy short-term load
3840 Details : If a large number of client requests are generated on a service
3841 that has previously never seen so many requests it is possible
3842 that the request buffer growth cannot keep up with the spike in
3843 demand. Instead of dropping incoming requests, they are held in
3844 the LND until the RPC service can accept more requests.
3847 Frequency : Sometimes during replay
3849 Description: Assertion failure in ll_local_open after replay.
3850 Details : If replay happened on an open request reply before we were able
3851 to set replay handler, reply will become not swabbed tripping the
3852 assertion in ll_local_open. Now we set the handler right after
3853 recognising of open request
3856 Frequency : very rare
3858 Description: kernel reports "badness in vsnprintf"
3859 Details : Reading from the "recovery_status" /proc file in small chunks
3860 may cause a negative length in lprocfs_obd_rd_recovery_status()
3861 call to vsnprintf() (which is otherwise harmless). Exit early
3862 if there is no more space in the output buffer.
3864 Severity : enhancement
3866 Description: clear OBD RPC statistics by writing to them
3867 Details : It is now possible to clear the OBD RPC statistics by writing
3868 to the "stats" file.
3873 Description: Client mtime is not the same on different clients after utimes
3874 Details : In some cases, the client was using the utimes() syscall on
3875 a file cached on another node. The clients now validate the
3876 ctime from the MDS + OSTs to determine which one is right.
3881 Description: Inability to activate failout mode
3882 Details : lconf script incorrectly assumed that in pythong string's numeric
3883 value is used in comparisons.
3886 Frequency : always with multiple stripes per file
3888 Description: Inefficient object allocation for mutli-stripe files
3889 Details : When selecting which OSTs to stripe files over, for files with
3890 a stripe count that divides evenly into the number of OSTs,
3891 the MDS is always picking the same starting OST for each file.
3892 Return the OST selection heuristic to the original design.
3897 Description: mount failures may take full timeout to return an error
3898 Details : Under some heavy load conditions it is possible that a
3899 failed mount can wait for the full obd_timeout interval,
3900 possibly several minutes, before reporting an error.
3901 Instead return an error as soon as the status is known.
3903 Frequency : quota enabled and large files being deleted
3905 Description: releasing more than 4GB of quota at once hangs OST
3906 Details : If a user deletes more than 4GB of files on a single OST it
3907 will cause the OST to spin in an infinite loop. Release
3908 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3913 Description: statfs data retrieved from /proc may be stale or zero
3914 Details : When reading per-device statfs data from /proc, in the
3915 {kbytes,files}_{total,free,avail} files, it may appear
3916 as zero or be out of date.
3919 Frequency : systems with MD RAID1 external journal devices
3921 Description: lconf's call to blkid is confused by RAID1 journal devices
3922 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3923 of returning all block devices that match the journal UUID.
3926 Frequency : always, for aggregate stripe size over 4GB
3928 Description: assertion fails when trying to use 4GB stripe size
3929 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3930 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3933 Frequency : always on ppc64
3935 Description: the first write on an ext3 filesystem with mballoc got stuck
3936 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3937 perform endianness conversion.
3939 ------------------------------------------------------------------------------
3941 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3943 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3944 INTEROPERATE with older versions automatically. Please read the
3945 user documentation before upgrading any part of a live system.
3946 * WARNING: Lustre networking configuration changes are required with
3947 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3950 * Support for kernels:
3951 2.6.9-22.0.2.EL (RHEL 4)
3952 2.6.5-7.244 (SLES 9)
3953 2.6.12.6 vanilla (kernel.org)
3956 Severity : enhancement
3957 Bugzilla : 7981/8208
3958 Description: Introduced Lustre Networking (LNET)
3959 Details : LNET is new networking infrastructure for Lustre, it includes
3960 a reorganized network configuration mode (see the user
3961 documentation for full details) as well as support for routing
3962 between different network fabrics. Lustre Networking Devices
3963 (LNDs) for the supported network fabrics have also been
3964 created for this new infrastructure.
3966 Severity : enhancement
3967 Description: Introduced Access control lists
3968 Details : clients can set ACLs on files and directories in order to have
3969 more fine-grained permissions than the standard Unix UGO+RWX.
3970 The MDS must be started with the "-o acl" mount option.
3972 Severity : enhancement
3973 Description: Introduced filesystem quotas
3974 Details : Administrators may now establish per-user quotas on the
3977 Severity : enhancement
3979 Description: Configuration change for the XT3
3980 The PTLLND is now used to run Lustre over Portals on the XT3
3981 The configure option(s) --with-cray-portals are no longer used.
3982 Rather --with-portals=<path-to-portals-includes> is used to
3983 enable building on the XT3. In addition to enable XT3 specific
3984 features the option --enable-cray-xt3 must be used.
3989 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3990 Details : A race between allocating a new llcd and re-getting the llcd_lock
3991 allowed another thread to grab newly-allocated llcd.
3993 Severity : enhancement
3995 Description: 2.6 OST async journal commit and locking fix to improve performance
3996 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3997 2.6 kernels are now async as they already were in 2.4 kernels so
3998 that they can commit concurrently with the network bulk transfer.
3999 For block-allocated files the filter allocation semaphore is held
4000 to avoid filesystem fragmentation during allocation. BKL lock
4001 removed for 2.6 xattr operations where it is no longer needed.
4006 Description: lconf incorrectly determined whether two IP networks could talk
4007 Details : In some more complicated routing and multiple-network
4008 configurations, lconf will avoid trying to make a network
4009 connection to a disjoint part of the IP space. It was doing the
4010 math incorrectly for one set of cases.
4015 Description: Fix for potential infinite loop processing records in an llog.
4016 Details : If an llog record is corrupted/zeroed, it is possible to loop
4017 forever in llog_process(). Validate the llog record length
4018 and skip the remainder of the block on error.
4021 Frequency : occasional (liblustre only)
4023 Description: liblustre could not open files whose last component is a symlink
4024 Details : sysio_path_walk() would incorrectly pass the open intent to
4025 intermediate path components.
4028 Frequency : rare (liblustre only with non-standard tuning)
4029 Bugzilla : 7201 (7350)
4030 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4031 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4032 after it has been released, only do this for referenced locks
4034 Severity : enhancement
4036 Description: specifying an (invalid) directory default stripe_size of -1
4037 would reset the directory default striping
4038 Details : stripe_size -1 was used internally to signal directory stripe
4039 removal, now use "all default" to signal dir stripe removal
4040 as a directory striping of "all default" is not useful
4043 Frequency : common for large clusters running liblustre clients
4045 Description: doing an ls when liblustre clients are running is slow
4046 Details : sending a glimpse AST to a liblustre client waits for every AST
4047 to time out, as liblustre clients will not respond. Since they
4048 cannot cache data we refresh the OST lock LVB from disk instead.
4050 Severity : enhancement
4052 Description: doing an ls at the same time as file IO can be slow
4053 Details : enqueue and other "small" requests can be blocked behind many
4054 large IO requests. Create a new OST IO portal for non-IO
4055 requests so they can be processed faster.
4058 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4060 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4061 created file with mode 000
4062 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4063 allow an ftruncate() as a non-root user to a file with mode 000.
4064 Lustre now respects this flag to disable mode checks when
4065 truncating a file owned by the user
4068 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4071 Description: Revoking locks from clients that went dead or catatonic might take
4073 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4074 cancellation of such locks instant on servers without waiting for
4075 any reply from clients. Clients drops these locks when cancel
4076 notification from server is received without replying.
4079 Frequency : liblustre-only, when liblustre client dies or becomes busy
4081 Description: Doing ls on Linux clients can take a long time with active
4083 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4084 granting such locks to it in the first place if possible. Locks
4085 are taken by proxy on the OST during the read or write and
4086 dropped immediately afterward. Add connect flags handling, do
4087 not grant locks to liblustre clients for glimpse ASTs.
4089 Severity : enhancement
4091 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4092 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4093 sequential reads and behaves badly with applications doing random
4094 reads. Improve it by reading ahead at least read region, and
4095 avoiding excessive large RPC for small reads.
4097 Severity : enhancement
4099 Description: Creating more than 1000 files for a single job may cause a load
4100 imbalance on the OSTs if there are also a large number of OSTs.
4101 Details : qos_prep_create() uses an OST index reseed value that is an
4102 even multiple of the number of available OSTs so that if the
4103 reseed happens in the middle of the object allocation it will
4104 still utilize the OSTs as uniformly as possible.
4109 Description: OST or MDS may oops in ping_evictor_main()
4110 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4111 but doesn't restart at beginning of obd_exports_timed list
4114 Severity : enhancement
4116 Description: improve by-nid export eviction on the MDS and OST
4117 Details : allow multiple exports with the same NID to be evicted at one
4118 time without re-searching the exports list.
4121 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4123 Description: MDS may oops in groups_free()
4124 Details : in rare race conditions a newly allocated group_info struct is
4125 freed again, and this can be NULL. The 2.4 compatibility code
4126 for groups_free() checked for a NULL pointer, but 2.6 did not.
4129 Frequency : common for liblustre clients doing little filesystem IO
4130 Bugzilla : 9352, 7313
4131 Description: server may evict liblustre clients accessing contended locks
4132 Details : if a client is granted a lock or receives a completion AST
4133 with a blocking AST already set it would not reply to the AST
4134 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4137 Frequency : lfs setstripe, only systems with more than 160 OSTs
4139 Description: unable to set striping with a starting offset beyond OST 160
4140 Details : llapi_create_file() incorrectly limited the starting stripe
4141 index to the maximum single-file stripe count.
4144 Frequency : LDAP users only
4146 Description: lconf did not handle in-kernel recovery with LDAP properly
4147 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4149 Severity : enhancement
4151 Description: bind OST threads to NUMA nodes to improve performance
4152 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4153 node and do their allocations there to localize memory access
4155 Severity : enhancement
4157 Description: llmount can determine client NID directly from Myrinet (GM)
4158 Details : the client NID code from gmnalnid was moved directly into
4159 llmount, removing the need to use this or specifying the
4160 client NID explicitly when mounting GM clients with zeroconf
4163 Frequency : if client is started with down MDS
4165 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4166 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4167 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4168 ptlrpc_import_delayed_req() because we didn't check if the
4169 request was interrupted, and we also didn't break out of the
4170 event loop if there was no timeout
4175 Description: data loss during non-page-aligned writes to a single file from
4176 both multiple nodes and multiple threads on one node at same time
4177 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4178 inconsistency led to false short-reads, that were cached and later
4179 used by ->prepare_write() to fill in partially written page,
4180 leading to data loss.
4183 Frequency : always, if lconf --abort_recovery used
4185 Description: lconf --abort_recovery fails with 'Operation not supported'
4186 Details : lconf was attempting to abort recovery on the MDT device and not
4189 Severity : enhancement
4191 Description: remove cleanup logs
4192 Details : replace lconf-generated cleanup logs with lustre internal
4193 cleanup routines. Eliminates the need for client-cleanup and
4196 Severity : enhancement
4198 Description: add support for EAs (user and system) on lustre filesystems
4199 Details : it is now possible to store extended attributes in the Lustre
4200 client filesystem, and with the user_xattr mount option it
4201 is possible to allow users to store EAs on their files also
4203 Severity : enhancement
4205 Description: Add possibility (config option) to show minimal available OST free
4207 Details : When compiled with --enable-mindf configure option, statfs(2)
4208 (and so, df) will return least minimal free space available from
4209 all OSTs as amount of free space on FS, instead of summary of
4210 free spaces of all OSTs.
4212 Severity : enhancement
4214 Description: do not expand extent locks acquired on OST-side
4215 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4216 by server: they are not cached anyway.
4219 Frequency : when mmap is used/binaries executed from Lustre
4221 Description: Unmmap pages before throwing them away from read cache.
4222 Details : llap_shrink cache now attempts to unmap pages before discarding
4223 them (if unmapping failed - do not discard). SLES9 kernel has
4224 extra checks that trigger if this unmapping is not done first.
4229 Description: lconf didn't resolve symlinks before checking to see whether a
4230 given mountpoint was already in use
4233 Frequency : when migrating failover services
4234 Bugzilla : 6395, 9514
4235 Description: When migrating a subset of services from a node (e.g. failback
4236 from a failover service node) the remaining services would
4237 time out and evict clients.
4238 Details : lconf --force (implied by --failover) sets the global obd_timeout
4239 to 5 seconds in order to quickly disconnect, but this caused
4240 other RPCs to time out too quickly. Do not change the global
4241 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4243 Severity : enhancement
4244 Frequency : if MDS is started with down OST
4245 Bugzilla : 9439,5706
4246 Description: Allow startup/shutdown of an MDS without depending on the
4247 availability of the OSTs.
4248 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4249 Add appropriate locking and lov-osc refcounts for safe
4250 cleaning. Add osc abort_inflight calls in case the
4251 synchronize never started.
4254 Frequency : occasional (Cray XT3 only)
4256 Description: root not authorized to access files in CRAY_PORTALS environment
4257 Details : The client process capabilities were not honoured on the MDS in
4258 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4259 been accessed by an authorized user then root was able to access
4260 the file on the local client also. The root user capabilities
4261 are now allowed on the MDS, as this environment has secure UID.
4264 Frequency : occasional
4266 Description: ldiskfs "too long searching" message happens too often
4267 Details : A debugging message (otherwise harmless) prints too often on
4268 the OST console. This has been reduced to only happen when
4269 there are fragmentation problems on the filesystem.
4274 Description: Division by zero in statfs when all OSCs are inactive
4275 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4276 lov_statfs divides by value returned by lov_get_stripecnt().
4280 Bugzilla : 9489, 3273
4281 Description: First write from each client to each OST was only 4kB in size,
4282 to initialize client writeback cache, which caused sub-optimal
4283 RPCs and poor layout on disk for the first writen file.
4284 Details : Clients now request an initial cache grant at (re)connect time
4285 and so that they can start streaming writes to the cache right
4286 away and always do full-sized RPCs if there is enough data.
4287 If the OST is rebooted the client also re-establishes its grant
4288 so that client cached writes will be honoured under the grant.
4293 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4294 Details : Now I/O RPCs go to different portal number and (presumably) fast
4295 lock requests (and glimses) and other RPCs get their own service
4296 threads pool that should be able to service those RPCs
4299 Severity : enhancement
4301 Description: Ability to exchange lustre version between client and servers and
4302 issue warnings at client side if client is too old. Also for
4303 liblustre clients there is ability to refuse connection of too old
4305 Details : New 'version' field is added to connect data structure that is
4306 filled with version info. That info is later checked by server and
4310 Frequency : rare, liblustre only.
4311 Bugzilla : 9296, 9581
4312 Description: Two simultaneous writes from liblustre at offset within same page
4313 might proceed at the same time overwriting eachother with stale
4315 Details : I/O lock withing llu_file_prwv was released too early, before data
4316 actually was hitting the wire. Extended lock-holding time until
4317 server acknowledges receiving data.
4320 Frequency : extremely rare. Never observed in practice.
4322 Description: avoid generating lustre_handle cookie of 0.
4323 Details : class_handle_hash() generates handle cookies by incrementing
4324 global counter, and can hit 0 occasionaly (this is unlikely, but
4325 not impossible, because initial value of cookie counter is
4326 selected randonly). Value of 0 is used as a sentinel meaning
4327 "unassigned handle" --- avoid it. Also coalesce two critical
4328 sections in this function into one.
4330 Severity : enhancement
4332 Description: allow liblustre clients to delegate truncate locking to OST
4333 Details : To avoid overhead of locking, liblustre client instructs OST to
4334 take extent lock in ost_punch() on client's behalf. New connection
4335 flag is added to handle backward compatibility.
4337 Severity : enhancement
4338 Bugzilla : 4928, 7341, 9758
4339 Description: allow number of OST service threads to be specified
4340 Details : a module parameter allows the number of OST service threads
4341 to be specified via "options ost ost_num_threads={N}" in the
4342 OSS's /etc/modules.conf or /etc/modprobe.conf.
4346 Bugzilla : 6146, 9635, 9895
4347 Description: servers crash with bad pointer in target_handle_connect()
4348 Details : In rare cases when a client is reconnecting it was possible that
4349 the connection request was the last reference for that export.
4350 We would temporarily drop the export reference and get a new
4351 one, but this may have been the last reference and the export
4352 was just destroyed. Get new reference before dropping old one.
4354 Severity : enhancement
4355 Frequency : if client is started with failover MDS
4357 Description: Allow multiple MDS hostnames in the mount command
4358 Details : Try to read the configuration from all specified MDS
4359 hostnames during a client mount in case the "primary"
4362 Severity : enhancement
4364 Description: Stop sending data to evicted clients as soon as possible.
4365 Details : Check if the client we are about to send or are sending data to
4366 was evicted already. (Check is done every second of waiting,
4367 for which l_wait_event interface was extended to allow checking
4368 of exit condition at specified intervals).
4371 Frequency : rare, normally only when NFS exporting is done from client
4373 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4375 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4378 Frequency : always, if srand() is called before liblustre initialization
4380 Description: Liblustre uses system PRNG disturbing its usage by user application
4381 Details : Introduce internal to lustre fast and high-quality PRNG for
4382 lustre usage and make liblustre and some other places in generic
4383 lustre code to use it.
4385 Severity : enhancement
4386 Bugzilla : 9477, 9557, 9870
4387 Description: Verify that the MDS configuration logs are updated when xml is
4388 Details : Check if the .xml configuration logs are newer than the config
4389 logs stored on the MDS and report an error if this is the case.
4390 Request --write-conf, or allow starting with --old_conf.
4392 Severity : enhancement
4394 Description: Handle symlinks in the path when checking if Lustre is mounted.
4395 Details : Resolve intermediate symlinks when checking if a client has
4396 mounted a filesystem to avoid duplicate client mounts.
4401 Description: lconf can hit an error exception but still return success.
4402 Details : The lconf command catches the Command error exception at the top
4403 level script context and will exit with the associated exit
4404 status, but doesn't ensure that this exit status is non-zero.
4409 Description: failure of ptlrpc thread startup can cause oops
4410 Details : Starting a ptlrpc service thread can fail if there are a large
4411 number of threads or the server memory is very fragmented.
4412 Handle this without oopsing.
4415 Frequency : always, only if liblustre and non-default acceptor port was used
4417 Description: liblustre cannot connect to servers with non-default acceptor port
4418 Details : tcpnal_set_default_params() was not called and was therefore
4419 ignoring the environment varaible TCPNAL_PORT, as well as other
4420 TCPNAL_ environment variables
4425 Description: two objects could be created on the same OST for a single file
4426 Details : If an OST is down, in some cases it was possible to create two
4427 objects on a single OST for a single file. No problems other
4428 than potential performance impact and spurious error messages.
4432 Bugzilla : 5681, 9562
4433 Description: Client may oops in ll_unhash_aliases
4434 Details : Client dcache may become inconsistent in race condition.
4435 In some cases "getcwd" can fail if the current directory is
4441 Description: Inode refcounting problems in NFS export code
4442 Details : link_raw functions used to call d_instantiate without obtaining
4443 extra inode reference first.
4447 Bugzilla : 9942, 9903
4448 Description: Referencing freed requests leading to crash, memleaks with NFS.
4449 Details : We used to require that call to ll_revalidate_it was always
4450 followed by ll_lookup_it. Also with revalidate_special() it is
4451 possible to call ll_revalidate_it() twice for the same dentry
4452 even if first occurence returned success. This fix changes semantic
4453 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4454 reference on a request referred from the intent.
4455 ll_intent_release() then releases such a request.
4458 Frequency : rare, normally benchmark loads only
4460 Description: unlinked inodes were kept in memory on the client
4461 Details : If a client is repeatedly creating and unlinking files it
4462 can accumulate a lot of stale inodes in the inode slab cache.
4463 If there is no other client load running this can cause the
4464 client node to run out of memory. Instead flush old inodes
4465 from client cache that have the same inode number as a new inode.
4468 Frequency : SLES9 2.6.5 kernel and long filenames only
4469 Bugzilla : 9969, 10379
4470 Description: utime reports stale NFS file handle
4471 Details : SLES9 uses out-of-dentry names in some cases, which confused
4472 the lustre dentry revalidation. Change it to always use the
4476 Frequency : rare, unless heavy write-truncate concurrency is continuous
4477 Bugzilla : 4180, 6984, 7171, 9963, 9331
4478 Description: OST becomes very slow and/or deadlocked during object unlink
4479 Details : filter_destroy() was holding onto the parent directory lock
4480 while truncating+unlinking objects. For very large objects this
4481 may block other threads for a long time and slow overall OST
4482 responsiveness. It may also be possible to get a lock ordering
4483 deadlock in this case, or run out of journal credits because of
4484 the combined truncate+unlink. Solution is to do object truncate
4485 first in one transaction without parent lock, and then do the
4486 final unlink in a new transaction with the parent lock. This
4487 reduces the lock hold time dramatically.
4490 Frequency : rare, 2.4 kernels only
4492 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4493 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4494 before unmounting their filesystem in order to avoid deadlock.
4495 The kernel_locked() function in 2.4 kernels only checks whether
4496 the kernel lock is held, not whether it is this process that is
4497 holding it as 2.6 kernels do.
4502 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4503 Details : The client ptlrpc code may be trying to reconnect to a down
4504 server before a previous connection attempt has timed out.
4505 Increase the reconnect interval to be longer than the connection
4506 timeout interval to avoid sending duplicate connections to
4510 Frequency : echo_client brw_test command
4512 Description: fix echo_client to work with OST preallocated code
4513 Details : OST preallocation code (5137) didn't take echo_client IO path
4514 into account: echo_client calls filter methods outside of any
4515 OST thread and, hence, there is no per-thread preallocated
4516 pages and buffers to use. Solution: hijack pga pages for IO. As
4517 a byproduct, this avoids unnecessary data copying.
4521 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4522 Description: Client can oops in mdc_commit_close() after open replay
4523 Details : It was possible for the MDS to return an open request with no
4524 transaction number in mds_finish_transno() if the client was
4525 evicted, but without actually returning an error. Clients
4526 would later try to replay that open and may trip an assertion
4527 Simplify the client close codepath, and always return an error
4528 from the MDS in case the open is not successful.
4531 Frequency : rare, 2.6 OSTs only
4533 Description: OST may deadlock under high load on fragmented files
4534 Details : If there was a heavy load and highly-fragmented OST filesystems
4535 it was possible to have all the OST threads deadlock waiting on
4536 allocation of biovecs, because the biovecs were not released
4537 until the entire RPC IO was completed. Instead, release biovecs
4538 as soon as they are complete to ensure forward IO progress.
4540 Severity : enhancement
4542 Description: Support for specifying external journal device at mount
4543 Details : If an OST or MDS device is formatted with an external journal
4544 device, this device major/minor is stored in the ext3 superblock
4545 and may not be valid for failover. Allow detecting and
4546 specifying the external journal at mount time.
4551 Description: Mounting an MDS with pending unlinked files may cause oops
4552 Details : target_finish_recovery() calls mds_postrecov() which returned
4553 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4554 considers this an error and immediately begins cleaning up the
4555 lov, just after starting the mds_lov process
4557 Severity : enhancement
4559 Description: Implement 'lfs df' to report actual free space on per-OST basis
4560 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4561 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4562 usage of objects; '-h' to report in human readable format.
4564 ------------------------------------------------------------------------------
4566 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4573 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4574 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4575 at startup during orphan unlinking. Instead, skip buddy update
4576 before setup, as it will be regenerated after recovery is complete.
4581 Description: If an OST is inactive, its locks might reference stale inodes.
4582 Details : lov_change_cbdata() must iterate over all namespaces, even if
4583 they are inactive to clear inode references from the lock.
4585 Severity : enhancement
4586 Frequency : occasional, if non-standard max_dirty_mb used
4588 Description: Client will block write RPCs if not enough grant
4589 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4590 then the client will block writes while waiting for another RPC
4591 to complete instead of consuming its dirty limit. With change
4592 we get improved performance when max_dirty_mb is small.
4594 Severity : enhancement
4595 Bugzilla : 3389, 6253
4596 Description: Add support for supplementary groups on the MDS.
4597 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4598 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4599 lookups for user supplementary groups into a cache.
4603 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4604 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4605 for liblustre O_EXCL is enforced by MDS.
4609 Description: readdir never returns NULL in liblustre.
4610 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4611 dentry in struct dirent.
4615 Frequency : liblustre only, depends on application IO pattern
4616 Description: liblustre clients evicted if not contacting servers
4617 Details : Don't put liblustre clients into the ping_evictor list, so
4618 they will not be evicted by the pinger ever.
4620 Severity : enhancement
4622 Description: Add ability to evict clients by NID from MDS.
4623 Details : By echoing "nid:$NID" string into
4624 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4625 $NID would be instantly evicted from this MDS and from all active
4626 OSTs connected to it.
4630 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4631 Details : lookup_it_finish() used to query file size from OSTs that was not
4636 Description: service threads change working directory to that of init
4637 Details : Starting lustre service threads may pin the working directory
4638 of the parent thread, making that filesystem busy. Threads
4639 now change to the working directory of init to avoid this.
4643 Frequency : during shutdown only
4644 Description: shutdown with a failed MDS or OST can cause unmount to hang
4645 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4650 Frequency : 2.6 only
4651 Description: chmod/chown may include an extra supplementary group
4652 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4653 supplementary group and if none is specified this is used.
4656 Bugzilla : 5479 (6816)
4658 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4659 Details : If lookup is in progress on a file that is unlinked we might try
4660 to revalidate the inode and fail in revalidate after lookup is
4661 complete and ll_file_open() enqueues the open again but
4662 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4665 Frequency : always, if lconf --abort_recovery used
4667 Description: lconf --abort_recovery fails with 'Operation not supported'
4668 Details : lconf was attempting to abort recovery on the MDT device and not
4671 ------------------------------------------------------------------------------
4673 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4678 Frequency : rare (only unsupported configurations with a node running as an
4680 Bugzilla : 6514, 5137
4681 Description: Mounting a Lustre file system on a node running as an OST could
4683 Details : OSTs now preallocates memory needed to write out data at
4684 startup, instead of when needed, to avoid having to
4685 allocate memory in possibly low memory situations.
4686 Specifically, if the file system is mounted on on OST,
4687 memory pressure could force it to try to write out data,
4688 which it needed to allocate memory to do. Due to the low
4689 memory, it would be unable to do so and the node would
4690 become unresponsive.
4692 Severity : enhancement
4694 Description: Addition of lconf --service command line option
4695 Details : lconf now accepts a '--service <arg>' option, which is
4696 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4698 Severity : enhancement
4700 Description: Failover mode is now the default for OSTs.
4701 Details : By default, OSTs will now run in failover mode. To return to
4702 the old behaviour, add '--failout' to the lmc line for OSTs.
4704 Severity : enhancement
4706 Description: Health checks are now provided for MDS and OSTs
4707 Details : Additional detailed health check information on MSD and OSTs
4708 is now provided through the procfs health_check value.
4711 Frequency : occasional, depends on IO load
4713 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4714 after numerous create/delete cycles
4715 Details : The ext3 inode allocation policy would not allocate new inodes
4716 very well on the OSTs because there are no new directories
4717 being created. Instead we look for groups with free space if
4718 the parent directories are nearly full.
4723 Description: Network or server problems during mount may cause partially
4724 mounted clients instead of returning an error.
4725 Details : The config llog parsing code may overwrite the error return
4726 code during mount error handling, returning success instead
4732 Description: MDS can fail to allocate large reply buffers
4733 Details : After long uptimes the MDS can fail to allocate large reply
4734 buffers (e.g. zconf client mount config records) due to memory
4735 fragmentation or consumption by the buffer cache. Preallocate
4736 some large reply buffers so that these replies can be sent even
4737 under memory pressure.
4741 Frequency : rare (liblustre)
4742 Description: fsx running with liblustre complained that using truncate() to
4743 extend the file doesn't work. This patch corrects that issue.
4744 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4745 ATTR_SIZE and lsm use in llu_setattr_raw.
4749 Frequency : rare, only 2.6 kernels
4750 Description: Unusual file access patterns on the MDS may result in inode
4751 data being lost in very rare circumstances.
4752 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4753 code in the 2.6 kernel caused the inode and/or EA data not to
4754 be read from disk, causing single-file corruption.
4758 Frequency : rare, only 2.6 filesystems using extents
4759 Description: Heavy concurrent write and delete load may cause data corruption.
4760 Details : It was possible under high-load situations to have an extent
4761 metadata block in the block device cache from a just-unlinked
4762 file overwrite a newly-allocated data block. We now unmap any
4763 metadata buffers that alias just-allocated data blocks.
4767 Frequency : filesystems with default stripe_count larger than 77
4768 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4769 Details : lconf specifies an inode size of 4096 bytes when the default
4770 stripe_count is larger than 77. This conflicts with the default
4771 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4772 this case to avoid pinning too much memory for large EAs.
4774 ------------------------------------------------------------------------------
4776 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4781 Frequency : rare (extremely heavy IO load with hundreds of clients)
4783 Description: Client is evicted, gets IO error writing to file
4784 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4785 caused clients to be evicted for AST timeouts. The fixes in
4786 bug 5192 mean we no longer need to have such short AST timeouts
4787 so ldlm_timeout has been increased.
4790 Frequency : occasional during --force or --failover shutdown under load
4791 Bugzilla : 5949, 4834
4792 Description: Server oops/LBUG if stopped with --force or --failover under load
4793 Details : a collection of import/export refcount and cleanup ordering
4794 issues fixed for safer force cleanup
4797 Frequency : only filesystems larger than 120 OSTs
4798 Bugzilla : 5990, 6223
4799 Description: lfs getstripe would oops on a very large filesystem
4800 Details : lov_getconfig used kfree on vmalloc'd memory
4803 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4804 Bugzilla : 6242, 6243
4805 Description: reading from files that had been truncated to a non-zero size
4806 but never opened returned no data
4807 Details : ll_file_read() reads zeros from no-object files to EOF
4812 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4813 Details : There's little harm in aborting MDS/OSS recovery and letting it
4814 try again, so I removed the LASSERT and return an error instead.
4816 Severity : enhancement
4818 Description: New debugging infrastructure for tracking down data corruption
4819 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4820 (b) cover more of the client-side code path, and (c) try to narrow
4821 down where problems occurred
4825 Bugzilla : 3819, 4364, 4397, 6313
4826 Description: Racing close and eviction MDS could cause assertion in mds_close
4827 Details : It was possible to get multiple mfd references during close and
4828 client eviction, leading to one thread referencing a freed mfd.
4830 Severity: : enhancement
4831 Bugzilla : 3262, 6359
4832 Description: Attempts to reconnect to servers are now more aggressive.
4833 Details : This builds on the enhanced upcall-less recovery that was added
4834 in 1.4.2. When trying to reconnect to servers, clients will
4835 now try each server in the failover group every 10 seconds. By
4836 default, clients would previously try one server every 25 seconds.
4841 Description: After recovery, certain operations trigger a failed
4842 assertion on a client.
4843 Details : Failing over an mds, using lconf -d --failover, while a
4844 client was doing a readdir() call would cause the client to
4845 LBUG after recovery completed and the readdir() was resent.
4847 Severity : enhancement
4849 Description: Default groups are now added by lconf
4850 Details : You can now run lconf --group <servicename> without having to
4851 manually add groups with lmc.
4854 Frequency : occasional
4856 Description: Nodes with an elan id of 0 trigger a failed assertion
4859 Frequency : always when accessing e.g. tty/console device nodes
4861 Description: tty and some other devices nodes cannot be used on lustre
4862 Details : file's private_data field is used by device data and lustre
4863 values in there got lost. New field was added to struct file to
4864 store fs-specific private data.
4867 Frequency : when exporting Lustre via NFS
4869 Description: NFSD failed occasionally when looking up a path component
4870 Details : NFSD is looking up ".." which was broken in ext3 directories
4871 that had grown large enough to become hashed.
4874 Frequency : Clusters with multiple interfaces not on the same subnet
4876 Description: Nodes will repeatedly try to reconnect to an interface which it
4877 cannot reach and report an error to the log.
4878 Details : Extra peer list entries will be created by lconf with some peers
4879 unreachable. lconf now validates the peer before adding it.
4882 Frequency : Only if a default stripe is set on the filesystem root.
4884 Description: Setting a default stripe on the filesystem root prevented the
4885 filesystem from being remounted.
4886 Details : The client was sending extra request flags in the root getattr
4887 request and did not allocate a reply buffer for the dir EA.
4890 Frequency : occasional, higher if lots of files are accessed by one client
4891 Bugzilla : 6159, 6097
4892 Description: Client trips assertion regarding lsm mismatch/magic
4893 Details : While revalidating inodes the VFS looks up inodes with ifind()
4894 and in rare cases can find an inode that is being freed.
4895 The ll_test_inode() code will free the lsm during ifind()
4896 when it finds an existing inode and then the VFS later attaches
4897 this free lsm to a new inode.
4901 Bugzilla : 6422, 7030
4902 Description: MDS deadlock between mkdir and client eviction
4903 Details : Creating a new file via mkdir or mknod (starting a transaction
4904 and getting the ns lock) can deadlock with client eviction
4905 (gets ns lock and trying to finish a synchronous transaction).
4908 Frequency : occasional
4909 Description: While starting a server, the fsfilt_ext3 module could not be
4911 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4912 kernels. Previously, lconf would still use the ext3 name
4913 when trying to load modules. Now, it will correctly use
4914 ext3 on 2.4 and ldiskfs on 2.6.
4916 Severity : enhancement
4917 Description: The default stripe count has been changed to 1
4918 Details : The interpretation of the default stripe count (0, to lfs
4919 or lmc) has been changed to mean striping across a single
4920 OST, rather than all available. For general usage we have
4921 found a stripe count of 1 or 2 works best.
4923 Severity : enhancement
4924 Description: Add support for compiling against Cray portals.
4925 Details : Conditional compiling for some areas that are different
4929 Frequency : occasional
4930 Bugzilla : 6409, 6834
4931 Description: Creating files with an explicit stripe count may lead to
4932 a failed assertion on the MDS
4933 Details : If some OSTs are full or unavailable, creating files may
4934 trigger a failed assertion on the MDS. Now, Lustre will
4935 try to use other servers or return an error to the
4939 Frequency : occasional
4941 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4942 caused lock timeout during readahead (since 1.4.2).
4943 Details : Processes doing readahead might match a lock that hasn't been
4944 granted yet if there are overlapping and conflicting lock
4945 requests. The readahead process waits on ungranted lock
4946 (original lock is CBPENDING), while OST waits for that process
4947 to cancel CBPENDING read lock and eventually evicts client.
4949 Severity : enhancement
4951 Description: Initial enabling of flock support for clients
4952 Details : Implements fcntl advisory locking and file status functions.
4953 This feature is provided as an optional mount flag (default
4954 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4955 locking are implemented yet, and those that are are not guaranteed
4956 to be completely correct in production environments.
4957 mount -t lustre -o [flock|noflock] ...
4960 Frequency : occasional
4962 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4963 assertion in the ext3 JBD (journaling) layer.
4964 Details : The b_committed_data struct is protected by the big kernel lock
4965 in 2.4 kernels, serializing journal_commit_transaction() and
4966 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4967 access to this struct. In 2.6 kernels there is finer grained
4968 locking to improve SMP performance of the JBD layer.
4972 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4977 Description: merge and cleanup kernel patches.
4980 -----------------------------------------------------------------------------
4982 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4984 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4985 versions of Lustre. Please update all servers and clients to
4986 version 1.4.2 or later at the same time. You must also run
4987 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4988 to update the configuration logs.
4990 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4991 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4992 - swab llog records of type '0' so we get proper header size/idx (5861)
4993 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4994 - fix rename of one directory over another leaking an inode (5953)
4995 - avoid SetPageDirty on 2.6 (5981)
4996 - don't re-add just-being-destroyed locks to the waiting list (5653)
4997 - when creating new directories, inherit the parent's custom
4998 striping settings if present parent (3048)
4999 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5000 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5001 - don't hold client locks on temporary worklist from l_lru (5666)
5002 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5003 - automatically evict dead clients (5921)
5004 - Update file size properly in create+truncate+fstat case (6196)
5005 - Do not unhash mountpoint dentries, do not allow removal of
5007 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5008 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5009 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5010 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5011 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5012 - Correctly handle reads of files with no objects (6243)
5013 - lctl recover will also mark a device active if deactivate used (5933)
5015 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5016 - allow --write-conf on an MDS with different nettype than client (5619)
5017 - don't write config llogs to MDS for mounts not from that MDS (5617)
5018 - lconf should create multiple TCP connections from a client (5201)
5019 - init scripts are now turned off by default; run chkconfig --on
5020 lustre and chkconfig --on lustrefs to use them
5021 - upcalls are no longer needed for clients to recover to failover
5023 - add --abort-recovery option to lconf to abort recovery on device
5025 - add support for an arbitrary number of OSTs (3026)
5026 - Quota support protocol changes.
5027 - forward compatibility changes to wire structs (6007)
5028 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5029 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5030 - improved "lustre status" script
5031 - initialize blocksize for non-regular files (6062)
5032 - added --disable-server and --disable-client configure options (5782)
5033 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5034 - Vanilla 2.4.29 support
5035 - increase maximum number of obd devices to 520 (6242)
5036 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5037 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5038 SLES 9 2.6.5 kernels
5039 - limit stripes per file to 160 (the maximum EA size) (6093)
5041 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5044 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5045 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5046 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5047 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5048 - small ext3 extents cleanups and fixes (5733)
5049 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5050 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5051 - avoid needless client->OST connect, fix handle mismatch (5317)
5052 - fix DLM error path that led to out-of-sync client, long delays (5779)
5053 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5054 - fix several locking issues related to i_size (5492,5624,5654,5672)
5055 - don't move pending lock onto export if it is already evicted (5683)
5056 - fix kernel oops when creating .foo in unlinked directory (5548)
5057 - fix deadlock in obdfilter statistics vs. object create (5811)
5058 - use time_{before,after} to avoid timer jiffies wrap (5882)
5059 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5060 - Do not leak request if server was not able to process it (5154)
5061 - If mds_open unable to find parent dir, make that negative lookup(5154)
5062 - don't create new directories with extent-mapping (5909, 5936)
5064 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5065 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5066 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5067 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5068 - support for RHEL4 kernel on i686 (5773)
5069 - provide error messages when incompatible logs are encountered (5898)
5071 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5072 * version 1.4.0.10 (1.4.1 release candidate 1)
5074 - don't keep a lock reference when lock is not granted (4238)
5075 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5076 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5077 - drop import inflight refcount on signal_completed_replay error (5255)
5078 - unlock page after async write error during send (3677)
5079 - handle missing objects in filter_preprw_read properly (5265)
5080 - no transno return for symlink open, don't save no-trasno open (3440)
5081 - don't try to complete elan receive that already failed (4012)
5082 - free RPC server reply state on error (5406)
5083 - clean up thread from ptlrpc_start_thread() on error (5160)
5084 - readahead could read extra page into cache that wasn't ejected (5388)
5085 - prevent races in class_attach/setup/cleanup/detach (5260)
5086 - don't dereference de->d_inode after l_dput of de (5458)
5087 - use "int" for stripe value returned from lock_to_stripe (5544)
5088 - mballoc allocation and error-checking fixes in 2.6 (5504)
5089 - block device patches to fix I/O request sizes in 2.6 (5482)
5090 - look up hostnames for IB nals (5602)
5091 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5092 - don't start multiple acceptors for the same port (5277)
5093 - fix incorrect LASSERT in mds_getattr_name (5635)
5094 - export a proc file for general "ping" checking (5628)
5095 - fix "lfs check" to not block when the MDS is down (5628)
5097 - service request history (4965)
5098 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5099 - create an "evict_client" /proc entry on OSTs, like the MDS has
5100 - fix mount usage message, return errors per mount(8) (5168)
5101 - change grep [] to grep "[]" in tests so they work in more UMLs
5102 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5103 - remove ancient LOV_MAGIC_V0 EA support (5047)
5104 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5105 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5106 - allow readahead allocations to fail when low on memory (5383)
5107 - mmap locking landed again, after considerable improvement (2828)
5108 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5110 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5113 - send OST transaction number in read/write reply to free req (4966)
5114 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5115 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5116 - the watchdog thread now runs as interruptible (5246)
5117 - flock/lockf fixes (but it's still disabled, pending 5135)
5118 - don't use EXT3 constants in llite code (5094)
5119 - memory shortage at startup could cause assertion (5176)
5121 - reorganization of lov code
5122 - single portals codebase
5124 - add extents/mballoc support (5025)
5125 - direct I/O reads in the obdfilter (4048)
5126 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5128 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5131 - send OST transaction number in read/write reply to free req (4966)
5132 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5133 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5134 - the watchdog thread now runs as interruptible (5246)
5135 - handle missing objects in filter_preprw_read properly (5265)
5136 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5137 - drop import inflight refcount on signal_completed_replay error (5255)
5138 - unlock page after async write error during send (3677)
5139 - return original error code on reconstructed replies (3761)
5140 - no transno return for symlink open, don't save no-trasno open (3440)
5142 - add pid to ldlm debugging output (4922)
5143 - bump the watchdog timeouts -- we can't handle 30sec yet
5144 - extra debugging for orphan dentry/inode bug (5259)
5146 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5149 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5150 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5151 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5152 - fix NULL dereference in /proc/sys/portals/routes (4827)
5153 - allow failed mdc_close() operations to be interrupted (4561)
5154 - stop precreate on OST before MDS would time out on it (4778)
5155 - don't send partial-page writes before EOF from client (4410)
5156 - discard client grant for sub-page writes on large-page clients (4520)
5157 - don't free dentries not owned by NFS code, check generation (4806)
5158 - fix lsm leak if mds_create_objects() fails (4801)
5159 - limit debug_daemon file size, always print CERROR messages (4789)
5160 - use transno after validating reply (3892)
5161 - process timed out requests if import state changes (3754)
5162 - update mtime on OST during writes, return in glimpse (4829)
5163 - add mkfsoptions to LDAP (4679)
5164 - use ->max_readahead method instead of zapping global ra (5039)
5165 - don't interrupt __l_wait_event() during strace
5167 - add software watchdogs to catch hung threads quickly (4941)
5168 - make lustrefs init script start after nfs is mounted
5169 - fix CWARN/ERROR duplication (4930)
5170 - return async write errors to application if possible (2248)
5171 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5172 - print NAL number in %x format (4645)
5173 - update barely-supported suse-2.4.21-171 series (4842)
5174 - support for sles 9 %post scripts
5175 - support for building 2.6 kernel-source packages
5176 - support for sles km_* packages
5178 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5181 - ignore -ENOENT errors in osc_destroy (3639)
5182 - notify osc create thread that OSC is being cleaned up (4600)
5183 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5184 - reconstruct ost_handle() like mds_handle() (4657)
5185 - create a new thread to do import eviction to avoid deadlock (3969)
5186 - let lconf resolve symlinked-to devices (4629)
5187 - don't unlink "objects" from directory with default EA (4554)
5188 - hold socknal file ref over connect in case target is down (4394)
5189 - allow more than 32000 subdirectories in a single directory (3244)
5190 - fix blocks count for O_DIRECT writes (3751)
5191 - OST returns ENOSPC from object create when no space left (4539)
5192 - don't send truncate RPC if file size isn't changing (4410)
5193 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5194 - bind to privileged port in socknal and tcpnal (3689)
5196 - rate limit CERROR/CWARN console message to avoid overload (4519)
5197 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5198 - basic mmap support (3918)
5199 - kernel patch series update from b1_4 (4711)
5201 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5204 - avoid crash during MDS cleanup with OST shut down (2775)
5205 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5206 - don't use bad inodes on the MDS (3744)
5207 - dynamic object preallocation to improve recovery speed (4236)
5208 - don't hold spinlock over lock dumping or change debug flags (4401)
5209 - don't zero obd_dev when it is force cleaned (3651)
5210 - print grants to console if they go negative (4431)
5211 - "lctl deactivate" will stop automatic recovery attempts (3406)
5212 - look for existing locks in ldlm_handle_enqueue() (3764)
5213 - don't resolve lock handle twice in recovery avoiding race (4401)
5214 - revalidate should check working dir is a directory (4134)
5216 - don't always mark "slow" obdfilter messages as errors (4418)
5218 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5221 - don't close LustreDB during write_conf until it is done (3860)
5222 - fix typo in lconf for_each_profile (3821)
5223 - allow dumping logs from multiple threads at one time (3820)
5224 - don't allow multiple threads in OSC recovery (3812)
5225 - fix debug_size parameters (3864)
5226 - fix mds_postrecov to initialize import for llog ctxt (3121)
5227 - replace config semaphore with spinlock (3306)
5228 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5229 - don't allow enqueue to complete on a destroyed export (3822)
5230 - down write_lock before checking llog header bitmap (3825)
5231 - recover from lock replay timeout (3764)
5232 - up llog sem before sending rpc (3652)
5233 - reduce ns lock hold times when setting kms (3267)
5234 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5235 - fix NULL deref and obd_dev leak on setup error (3312)
5236 - replace some LBUG about llog ops with error handling (3841)
5237 - don't match INVALID dentries from d_lookup and spin (3784)
5238 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5239 - fix invalid assertion in ptlrpc_set_wait (3880)
5241 - add libwrap support for the TCP acceptor (3996)
5242 - add /proc/sys/portals/routes for non-root route listing (3994)
5243 - allow setting MDS UUID in .xml (2580)
5244 - print the stack of a process that LBUGs (4228)
5246 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5249 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5250 - make sure to unset replay flag from failed open requests (3440)
5251 - if default stripe count is 0, use OST count for inode size (3636)
5252 - update parent mtime/ctime on client for create/unlink (2611)
5253 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5254 - free recovery state on server during a forced cleanup (3571)
5255 - unregister_reply for resent reqs (3063)
5256 - loop back devices mounting and status check on 2.6 (3563)
5257 - fix resource-creation race that can provoke i_size == 0 (3513)
5258 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5259 - more debugging for page-accounting assertion (3746)
5260 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5261 - avoid deadlock after precreation failure (3758)
5262 - fix race and lock order deadlock in orphan handling (3450, 3750)
5263 - add validity checks when grabbing inodes from l_ast_data (3599)
5265 - add /proc/.../recovery_status to obdfilter (3428)
5266 - lightweight CDEBUG infrastructure, debug daemon (3668)
5267 - change default OSC RPC parameters to be better on small clusters
5268 - turn off OST read cache for files smaller than 32MB
5269 - install man pages and include them in rpms (3100)
5270 - add new init script for (un)mounting lustre filesystems (2593)
5271 - run chkconfig in %post for init scripts (3701)
5272 - drop scimac NAL (unmaintained)
5274 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5277 - clean kiobufs before and after use (3485)
5278 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5279 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5280 - change init script to fail more clearly if not run as root (1528)
5281 - allow clients to reconnect during replay (1742)
5282 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5283 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5284 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5285 - fix cond_resched (3554)
5286 - only evict unfinished clients after recovery (3515)
5287 - allow bulk resend, prevent data loss (3570)
5288 - dynamic ptlrpc request buffer allocation (2102)
5289 - don't allow unlinking open directory if it isn't empty (2904)
5290 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5291 - remove extraneous obd dereference causing LASSERT failure (3334)
5292 - don't use get_cycles() when creating temp. files on the mds (3156)
5293 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5294 - handle EEXIST for set-stripe, set proper directory name (3336)
5296 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5297 - fix ksocknal_fmb_callback() error messages (2918)
5299 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5302 - don't copy lvb into (possibly NULL) reply on error (2983)
5303 - don't deref dentry after dput, don't free lvb on error (2922)
5304 - use the kms to determine writeback rpc length (2947)
5305 - increment oti_logcookies when osc is inactive (2948)
5306 - update client's i_blocks count via lvb messages (2543)
5307 - handle intent open/close of special files properly (1557)
5308 - mount MDS with errors=remount-ro, like obdfilter (2009)
5309 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5310 - don't use cancelling-locks' kms values (2947)
5311 - use highest lock extent for kms, not last one (2925)
5312 - don't dereference ERR_PTR() dentry in error handling path (3107)
5313 - fix thread race in portals_debug_dumplog() (3122)
5314 - create lprocfs device entries at setup instead of at attach (1519)
5315 - common AST error handler, don't evict client on completion race (3145)
5316 - zero nameidata in detach_mnt in 2.6 (3118)
5317 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5318 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5319 - initialize RPC timeout timer earlier for 2.6 (3219)
5320 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5321 - print nal/nid for unknown nid (3258)
5322 - additional checks for oscc recovery before doing precreate (3284)
5323 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5324 - don't crash in mdc_close for bad permissions on open (3285)
5325 - zero i_rdev for non-device files (3147)
5326 - clear page->private before handing to FS, better assertion (3119)
5327 - tune the read pipeline (3236)
5328 - fix incorrect decref of invalidated dentry (2350)
5329 - provide read-ahead stats and refine rpc in flight stats (3328)
5330 - don't hold journal transaction open across create RPC (3313)
5331 - update atime on MDS at close time (3265)
5332 - close LDAP connection when recovering to avoid server load (3315)
5333 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5334 - don't leak open file on MDS after open resend (3325)
5335 - serialize filter_precreate and filter_destroy_precreated (3329)
5336 - loop device shouldn't call sync_dev() for nul device (3092)
5337 - clear page cache after eviction (2766)
5338 - resynchronize MDS->OST in background (2824)
5339 - refuse to mount the same filesystem twice on same mountpoint (3394)
5340 - allow llmount to create routes for mounting behind routers (3320)
5341 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5342 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5343 - fix rare problem with rename on htree directories (3417)
5345 - allow default OST striping configuration per directory (1414)
5346 - fix compilation for qswnal for 2.6 kernels (3125)
5347 - increase maximum number of MDS request buffers for large systems
5348 - change liblustreapi to be useful for external progs like lfsck (3098)
5349 - increase local configuration timeout for slow disks (3353)
5350 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5352 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5355 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5356 - don't overwrite extent policy data in reply if lock was blocked (2901)
5357 - drop filter export grants atomically with removal from device (2663)
5358 - del obd_self_export from work_list in class_disconnect_exports (2908)
5359 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5360 - swab reply message in mdc_close, other PPC fixes (2464)
5361 - fix destroying of named logs (2325)
5362 - overwrite old logs when running lconf --write_conf (2264)
5363 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5364 - fix race in target_handle_connect (2898)
5365 - mds_reint_create() should take same inode create lock (2926)
5366 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5367 - don't close files for self_export to avoid uninitialized obd (2936)
5368 - allow MDS with the same name as client node (2939)
5369 - hold dentry reference for closed log files for unlink (2325)
5370 - reserve space for all logs during transactions (2059)
5371 - don't evict page beyond end of stripe extent (2925)
5372 - don't oops on a deleted current working directory (2399)
5373 - handle hard links to targets without a parent properly (2517)
5374 - don't dereference NULL lock when racing during eviction (2867)
5375 - don't grow lock extents when lots of conflicting locks (2919)
5377 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5380 - account for cache space usage on clients to avoid data loss (974)
5381 - lfsck support in lustre kernel code (2349)
5382 - reduce journal credits needed for BRW writes (2370)
5383 - orphan handling to avoid losing space on client/server crashes
5384 - ptlrpcd can be blocked, stopping ALL progress (2477)
5385 - use lock value blocks to assist in proper KMS, faster stat (1021)
5386 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5387 - recovery for initial connections (2355)
5388 - fixes for mds_cleanup_orphans (1934)
5389 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5390 - block all file creations until orphan recovery completes (1901)
5391 - client remove rq_connection from request struct (2423)
5392 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5393 - recovery timer race (2670)
5394 - mdc_close recovey bug (2532)
5395 - ptlrpc cleanup bug (2710)
5396 - mds timeout on local locks (2588)
5397 - namespace lock held during RPCs (2431)
5398 - handle interrupted sync write properly (2503)
5399 - don't try to handle a message that hasn't been replied to (2699)
5400 - client assert failure during cleanup after abort recovery (2701)
5401 - leak mdc device after failed mount (2712)
5402 - ptlrpc_check_set allows timedout requests to complete (2714)
5403 - wait for inflight reqs when ptlrpcd finishes (2710)
5404 - make sure unregistered services are removed from the srv_list
5405 - reset bulk XID's when resending them (caught by 1138 test)
5406 - unregister_bulk after timeout
5407 - fix lconf error (2694)
5408 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5409 - readahead locks pages, leaves pending causing memory pressure (2673)
5410 - increase OST request buffers to 4096 on large machines (2729)
5411 - fix up permission of existing directories in simple_mkdir (2661)
5412 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5413 - don't assign transno to errored transactions (2742)
5414 - don't delete objects on OST if given a bogus objid from MDS (2751)
5415 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5416 - if rq_no_resend, then timeout request after recovery (2432)
5417 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5418 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5419 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5420 - MDS recovery shouldn't delete objects that it has given out (2730)
5421 - if enqueue arrives after completion, don't clobber LVB (2819)
5422 - don't unlock pages twice when trigger_group_io returns error (2814)
5423 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5424 - don't write pages to disk if there was an error (1450)
5425 - don't ping imports that have recovery disabled (2676)
5426 - take buffered bytes into account when balancing socknal conn (2817)
5427 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5428 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5429 - remove little-endian swabbing of llog records (1987)
5430 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5431 - retry reposting request buffers if they fail (1191)
5432 - grow extent at grant time to avoid granting a revoked lock (2809)
5433 - lock revoke doesn't evict page if covered by a second lock (2765)
5434 - disable VM readahead to avoid reading outside lock extents (2805)
5436 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5437 - updated kernel patches for hp-2.4.20 kernel (2681)
5439 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5442 - fix truncated write corruption (2366)
5443 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5445 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5446 - ensure len(uuid) < 37 in lmc (1171)
5447 - fix ia64 OOPS in llog_test (2255)
5448 - zero end of page at obdfilter for partial page writes (2648)
5449 - don't leave stale dentries around after renames (bug 2428)
5450 - fix timeouts when evicting a client with a single lock held (2642)
5451 - set deadline for the initial HELLO message to drain (2634)
5452 - print out dotted-quad IP addresses in the socknal (2302)
5454 - additional debugging for MDS client eviction problem (2443)
5455 - fix mkfsoptions support for osts (2603, 2604)
5457 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5460 - add series for the vanilla 2.6.0 kernel
5461 - add series for the vanilla 2.4.24 kernel
5462 - add series for a cray x86/64 UL kernel drop
5463 - fix xattr patches for the vanilla 2.4.19 series
5465 - generate true UUIDs in lmc (1171)
5466 - have portals stack dumping break in UML (2466)
5467 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5468 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5469 - wake callback waiting threads on client eviction (2460)
5470 - Add --ptldebug and --subsystem to lmc (1719)
5471 - update assertion to allow safe interrupt allocation
5472 - set rq_no_resend for cancel requests (2432)
5473 - recalculate ptlrpcd timeout after resend (2494)
5474 - call vfs_rmdir when removing pending directories (2368)
5475 - fix renaming a file to itself (2429)
5476 - lmc creates a default one-stripe lov (2454)
5477 - expand procfs space to handle large clusters (2326)
5478 - increase UML stack to avoid overflow
5479 - update lconf's list of debug and subsystem masks
5480 - fix lfs find --obd (2510)
5481 - /proc tunable for disabling filter read caching (2591)
5482 - stop rpm packages from altering slapd.conf (2301)
5483 - disable nagle in the socknal under 0conf (2578)
5484 - choose mds inode size based on stripe count (2572)
5485 - fix kernel-source rpm problems (2516)
5487 - add --disable-doc to avoid pdf generation (2421)
5488 - update documentation, tests, type-os, comments
5489 - avoid format warnings on ia64
5490 - remove the TOE NAL
5491 - tiny code cleanups by removing unused fields
5493 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5496 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5497 - remove the most blatant lies from BUILDING file (2371)
5498 - change default debug level to reasonable production setting
5499 - reduce client side cache size to reduce cache flush time
5500 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5501 - make TCP zerocopy and pinger support enabled by default (2476)
5502 - sync writes completed after process exits caused crashes (2319)
5503 - maintain correct mount count on the MDS (2356)
5504 - backout 1557, because 2316 wasn't really fixed
5505 - better file I/O statistics gathering in /proc
5506 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5507 - another kernel patch to fix zero-copy TCP function export
5508 - don't take duplicate lock when processing re-sent getattr (2420)
5509 - lctl uses obd_self_export instead of creating new conn (2353)
5510 - MDS/OST recovery case which requires object creation asserted (2425)
5511 - move lfs from /usr/sbin to /usr/bin in packages
5512 - fix race between mds_client_add and mds_client_free (2417)
5513 - use kmalloc instead of slabs in portals (2430)
5514 - don't create duplicate records when a failover MDS is present (2442)
5515 - remove unnecessary mount age check (2332)
5516 - don't remove directory inodes from locks prematurely (2451)
5517 - don't break if MDS service name is the same as hostname (2103)
5518 - fix races in client write RPC generation when cache full (2482)
5520 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5523 - remove now-unused request->rq_obd (278)
5524 - if an allocation fails, print out how much memory we've used (1933)
5525 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5526 - add the "configurable stack size" patch to most series files (1256)
5527 - ability to write large log records, for 100+ OST configs (2306)
5528 - fix NULL deref when filter_prep fails (2314)
5529 - fix operator precedence error in filter_sync
5530 - dynamic allocation of socknal TX descriptors (2315)
5531 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5532 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5533 - ensure that CWARN messages in Portals always get to the syslog
5534 - __init/__exit are not for prototype decls (ldlm_init/exit)
5535 - x86-64 compile warning fixes
5536 - fix gateway LMC keyword conflict (2318)
5537 - fix MDS lock inversions in getattr/reint paths (1844)
5538 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5539 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5540 - DLM race condition prevented some lock evictions (2328)
5541 - ENOMEM detection and retry on socknal sends (2230)
5542 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5543 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5544 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5545 - better error messages when a client is rejected during recovery (1505)
5546 - avoid cancelling locks which were never granted, after failure (2330)
5547 - fix i_sem/journal inversion in mds_client_add (2333)
5548 - fix truncate/getattr lock cycle deadlock (2334)
5549 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5550 - fix two rare exit paths which could leak an l_lock() ref (2321)
5551 - fencepost error in MDS/OST orphan recovery (2226)
5552 - make log record alignment 8 bytes (1988)
5553 - lstripe now fails when requested offset > ost_count (2237)
5554 - ensure that all kernel series have a complete list.h (1607)
5555 - fix crashes in special-file operations (2316)
5556 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5558 - allow configurable automake binary, for testing new versions
5559 - small update to the lfs documentation
5561 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5563 * fix negative export reference count in fsfilt_sync (2312)
5565 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5566 * release candidate 0.9.1
5568 - orphans are moved into the PENDING directory for possible recovery
5569 - replayed opens now open by fid for orphan/rename safety (1042)
5570 - last close of an orphan inode generates a transno (683)
5571 - chdir() and mount() now pin the directory entry (1020)
5572 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5573 - discard very old requests without processing them (1502)
5574 - remove some common, well-understood CERRORs (1505)
5575 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5576 - clear "grant" flags in OST replies until OST grant code lands (1644)
5577 - fix read performance by not clobbering i_blksize on client (1598)
5578 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5579 - make LRU size a /proc tunable, clears locks when reduced (707)
5580 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5581 - server locks take references on exports now (1558)
5582 - build fixes for 2.4.20-rh trees (1663)
5583 - return an error from lov_create if all OSCs are inactive (1751)
5584 - fix import levels when a reconnect happens without a timeout (1597)
5585 - exit early from mds_open if we get a lookup error (1749)
5586 - partial page read at EOF wouldn't wait for disk before sending (1642)
5587 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5588 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5589 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5590 - write/truncate lock inversion (1639)
5591 - Don't auto-load obdclass, portals modules during cleanup (1495)
5592 - fix timestamps from jumping to "now" (1763)
5593 - extra journal assertions (1648)
5594 - add an extra multiunlink test (1771)
5595 - fix read_record/write_record API (1776)
5596 - fix leak of offset_extent, possible incorrect i_size later (1772)
5597 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5598 - Debugging for the kqswnal_get_idle_tx problems (1820)
5599 - Allow recovery to be attempted multiple times (1536)
5600 - Write out MDS last_rcvd file after it is first created (1600)
5601 - Fix tx_descriptor leak in failed transmit situations (1827)
5602 - ext3 journaling fixes for assertion failure after IO error (1871)
5603 - class_export_put() on freed export after completion AST error (1896)
5604 - Fix revalidate looping in VFS (1322)
5605 - Don't access a freed export during MDS_REINT timeout (1521)
5606 - Add open-unlink recovery support on the MDS (1673,1764)
5607 - Return an error if no MDS data was read from last_rcvd (1946)
5608 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5609 - Don't setup a disk device that doesn't match exported UUID (317)
5610 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5611 - avoid committing NULL handle in force close
5612 - local.sh is now a one-stripe LOV configuration
5613 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5614 - fix ext3 htree duplicate directory entry corruption (1516)
5615 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5616 - update to new LOV EA format (2097)
5617 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5619 2003-06-15 Phil Schwan <phil@clusterfs.com>
5622 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5623 - per-import/export recovery handling (958, 931, 959)
5624 - multiple last-rcvd slots, for serving multiple FSes (949)
5625 - connections are again shared between multiple imp/exports (963, 964)
5626 - "umount -f" would hang if any requests needed to be sent (393, 978)
5627 - avoid pinning large req buffer by copying for queued messages (989)
5628 - add "uuid" to "lctl device" command to help upcalls (991)
5629 - "open" RPCs with transnos would confuse recovery counters (1037)
5630 - do proper endian conversion of all wire messages (288, 340, 891)
5631 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5632 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5633 - fix ptlrpc_connection leak in target_handle_connect (1174)
5634 - fix import refcounting bug in OST and MDS cleanup (1134)
5635 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5636 - fix very unlikely obd_types race condition (501)
5637 - remove osc_open hack for echo_client (1187)
5638 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5639 - a failure in read_inode2 leads to deadlock (1139)
5640 - cancel ack-locks as soon as transaction is committed (1072)
5641 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5642 - make sure to commitrw after any preprw to avoid deadlock (1162)
5643 - failing to execute a file in a lustre FS would lock inode (1203)
5644 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5645 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5646 - obd_uuid2tgt would walk off the end of the list (1255)
5647 - on IA64 the osc would give portals incorrect bulk size (1258)
5648 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5649 - fix lock inversion caused by new llite matching code (1282)
5650 - limit the number of dirty pages on a client to 10MB (1286)
5651 - timed out locks were not being corrected cancelled (1289)
5652 - fix O_DIRECT above 4GB on IA-32 (1292)
5653 * major user-visible changes
5654 - fail out/fail over policy now controlled by the upcall (993)
5656 - add OBD_PING to check server availability and failure (954)
5657 - lustre messages are now sent in sending host order (288, 340, 891)
5658 - add eadatalen to MDS getattr reply (340)
5659 - OST read replies may contain second buffer, with per-page status (593)
5661 2003-03-11 Phil Schwan <phil@clusterfs.com>
5664 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5665 - failed open()s could cause deadlock; fixed (867, 869)
5666 - stop cancelling OST locks when files are closed (481)
5667 - overlapping XID spaces caused network corruption (851, 853)
5668 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5669 - setattr_raw added, to do single-RPC, server-side setattrs
5670 - lmc/lconf syntax change for OST UUIDs
5671 - fix crashy race condition between ptlrpc_free_req and osc_close
5672 - don't use request in mdc_enqueue if we hit a timeout (889)
5673 - don't set the inode i_size for regular files from the MDS (896)
5674 - handle out of order completion AST (842)
5675 - don't LBUG if a lock request times out after receiving AST (913)
5676 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5677 - if a bad lock AST arrives, send an error instead of dropping entirely
5678 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5679 - fix leak in bulk IO when only partially completed (899, 900, 926)
5680 - fix O_DIRECT for ia64 (55)
5681 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5682 - C-z after timeout could hang a process forever; fixed (977)
5684 - client-side I/O cache (678, 924, 929, 941, 970)
5686 - READPAGE and SETATTRs which don't take server-side locks get
5689 2003-02-11 Phil Schwan <phil@clusterfs.com>
5692 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5693 - Fix fsfilt_extN_readpage() to read a full page of directory
5694 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5695 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5696 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5697 - Most intent operations were removed, in favour of a new RPC mode
5698 that does a single RPC to the server and bypasses most of the VFS
5699 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5700 - Aggressively cancel local locks on DLM servers
5701 - mds_reint_unlink sends EA to the client if it's the last nlink.
5702 client uses that EA to unlink OST objects.
5703 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5704 - recursive symlinks were fixed (439)
5705 - fixed NULL deref in DEBUG_REQ
5706 - filter_update_lastobjid no longer calls sync, which annoyed extN
5707 - fixed multi-client small-writes to a single file problem (445)
5708 - fixed mtime updates during file writes (607)
5709 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5710 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5711 - fixed Linux OST size reporting problem (444, 656)
5712 - OST now updates object mtime with writes or setattr (607, 619)
5713 - client verifies file size before zeroing page past EOF (445)
5714 - OST now writes last allocated objid to disk with allocation (108)
5715 - LOV on echo now works (409)
5717 - mds_reint_unlink sends a new buffer, with the EA included. this
5718 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5719 set if a regular file was being unlinked, and it was the last link
5720 - use PtlGet from the target for bulk writes (315)
5721 - OST now updates object mtime with writes or setattr (607, 619)
5722 - LDLM now has a grant-time callback to revalidate locked items, if
5724 - Many MDS operations were reorganized to combat race conditions
5726 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5727 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5729 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5732 - Fully reactivate OST imports after reconnection (512, others)
5733 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5734 - More graceful error handling for truncating on dead OST (515)
5735 - Don't error out unless we're actually accessing dead stripes (474)
5736 - Fix garbage sizes when stripes are missing (410)
5737 - LRU counters were broken, causing constant lock purge (433, 432)
5738 - garbage on read from stripes with failed OSTs (441)
5739 - mark OSCs as active before reconnecting during recovery (438)
5740 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5741 - lfind did not preserve OST order in output (443)
5742 - symlinks cause hung clients, incorrect data (439)
5743 - stop dereferencing request after dropping refcount (457)
5744 - don't LASSERT(spin_is_locked) on non-SMP (455)
5745 - fixes for many rename() bugs
5746 - fstat didn't correctly synchronize attributes (399)
5747 - server must handle lock cancellation during blocking AST prep (487)
5748 - bulk descriptors were free()d too soon (511)
5749 - fix paths in lconf, which would load incorrect modules (451, 507)
5750 - fix confusing lconf 'host not found' error message (386)
5751 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5752 - fix race condition in mdc_blocking_ast() for inode access (526)
5753 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5754 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5755 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5756 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5757 - fix lstripe in several ways (kernel side) (527)
5758 - fix request leak in ldlm_cli_enqueue (262)
5759 - incorrect OSC was marked inactive after OST failure
5760 - call mds_fs_cleanup before unmounting filesystem (524)
5761 - fix races between taking ns_lock and ldlm_lock_change_resource
5762 - fix races updating LOV export open file list
5763 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5764 - fix recovery NULL deref in ldlm_cli_cancel_unused
5765 - fix some DLM races by using new hash table for lock handles (419)
5766 - permit the client to specify desired inodes, at replay
5767 - duplicate requests when we queue them for replay reintegration
5768 - fix last_rcvd offset calculation
5769 - sync after each recovered transaction, so we always make progress
5770 - never, not always, ERESTART requests without transnos
5771 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5773 - skip replay if the MDS says that the client is already connected
5774 - don't check for a recovery-enabled export to match lctl's UUID
5775 - don't INC_USE_COUNT for phantom exports
5776 - don't crash when cleaning up phantom exports (567)
5777 - don't double-finish or set replay data for errored mdc_open requests
5778 - abort requests when they time out, so we don't get old replies
5779 - send/receive replies for AST messages again
5780 - if the client says that it doesn't have the lock, cancel it on the
5782 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5783 mark it as destroyed, it will all work out when decref is called
5784 - fix module use counts (22, 581)
5786 - ASTs now expect a reply (server cancels lock on error reply)
5788 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5791 - fix many simultaneous client startup (392)
5792 - fix dentry->d_it clobbering
5793 - credentials weren't being shipped for readdir/getattr operations
5794 - remove invalid assertions triggered during some concurrent MD
5796 - proper Lustre versions added (336, 389)
5797 - fix memory leak for create error case (398)
5798 - fix LOV locking bug that would get cli/srv out of sync
5799 - fix echo client over LOV (409)
5800 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5801 - fix double-O_EXCL intent crash (424)
5802 - avoid sending multiple lock CANCELs (352)
5804 - MDS can do multi-client recovery (modulo bugs in new code)
5806 - many updates, edits, cleanups
5808 2002-11-18 Phil Schwan <phil@clusterfs.com>
5811 - fix null d_it dereference (346)
5812 - fix full OST/dbench hang (333)
5813 - fix permission problem with file removal (286)
5814 - fix removal of OSCs from LOV when they fail
5815 - fix NULL deref during bulk timeout (214)
5816 - fix problems related to multiple filesystems on one MDS (241)
5817 - fixed serious subtle metadata locking bugs
5818 - free locks on clients when inodes are removed due to memory
5820 - fix inode pointer in lock data (285)
5821 - partial support for multiple MDS on a single host (241)
5822 - data locks weren't cancelled at clear_inode time (290, 311)
5823 - intent locks could lead to unbounded lock growth (205)
5824 - added a maximum lock count, an LRU list, and a flusher
5825 - fix multiple rename (365)
5826 - properly abstracted the echo client
5827 - OSC locked 1 byte too many; fixed
5828 - rewrote brw callback code:
5829 - fixed recovery bugs related to LOVs (306)
5830 - fixed too-many-pages-in-one-write crash (191)
5831 - fixed (again) crash in sync_io_timeout (214)
5832 - probably fixed callback-related race (385)
5834 - Add capability to MDS protocol
5835 - LDLM cancellations and callbacks on different portals
5837 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5840 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5841 - defer open object destruction to close time (601981, 138)
5842 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5843 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5844 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5845 - don't call obd_setattr in ll_file_release for destroyed objects
5846 * protocol change to lustre_msg: move |version| and add |flags|
5847 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5848 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5849 * added replay of create, unlink, link and rename operations during
5850 MDS failover; recovery should be much more robust now
5851 * remove failed OSCs from LOVs (only lov_create uses this so far)
5852 * the lustre-HOWTO was brought (more) up to date (582544)
5854 2002-10-23 Phil Schwan <phil@clusterfs.com>
5857 - in-use dentries weren't being reused properly (617851)
5858 - prevent multiple LDLM setup (599178)
5859 - fix LOV size calculations for truncate (617853)
5860 - fix client handling of MDS intent errors (POSIX)
5861 - fix permission bug in lovstripe.c test (624321)
5862 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5863 - truncate past end of file could corrupt data
5864 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5865 - a race in recovery could return ETIMEDOUT to apps (623947)
5866 - building outside the source directory was fixed
5867 * the lustre-HOWTO was brought (more) up to date (582544)
5868 * major progress was made on recovery functionality
5870 2002-10-10 Phil Schwan <phil@clusterfs.com>
5873 - recovery deadlock fix
5874 - rm -rf causes LBUG fix (617817)
5875 - file open by multiple tasks fix (618962)
5876 - directory permissions bugs (602707 and 620007)
5877 - journal_stop fixed with locking (611313)
5878 - O_APPEND failures resolved (618273, perhaps 614459)
5879 - lconf PATH fix (619770)
5880 - IA64 build fix (621450)
5881 - RPC buffer sizes scale with amount of memory
5883 2002-10-01 Phil Schwan <phil@clusterfs.com>
5886 - locks would be cancelled without throwing away data pages,
5887 resulting in inconsistent data (605627)
5888 - inode attributes were not always being refreshed (605627, 612449)
5889 - lconf now continues to cleanup after lctl reports an error
5890 - MDS now enforces user permissions (602707)
5891 - lprocfs cleanup fixed, but not yet enabled (614157)
5892 - fixed infinite server hang, should a client not respond to an AST
5893 - avoid going into recovery if user calls readlink() with a buffer
5894 that's too small (613941)
5895 - AST RPCs no longer require replies (614867) -- this may be changed
5896 - don't crash server if client sends an IOV that's too big (611336)
5897 - fixed lock conversion deadlock (611892)
5898 - fixed the following of symlinks (614622)
5899 * recovery: the server can remove locks from a client that dies, other
5900 clients can make progress
5901 * more extN patch fixes
5902 * compile-time configurable ptlrpc buffer allocations
5904 - collaborative read cache document
5905 - Lustre Lite Performance CDR document-in-progress
5907 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5910 - fix typo in patch-2.4.18
5912 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5915 - clear ptlrpc request each time in handle_incoming_request()
5916 - unlink of files now destroys the object on the OST
5918 2002-09-19 Peter Braam <braam@clusterfs.com>
5920 * add hard link support
5921 * change obdfile creation method
5922 * kernel patch changed
5924 2002-09-19 Peter Braam <braam@clusterfs.com>
5927 - stack overflow bug in extN fixed
5929 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5931 * documentation updates
5932 - add man pages for config tools
5933 - update tests/README to describe testing with new config tools
5934 - finish metadata API descriptions
5935 * bug fixes and cleanups
5936 - statfs workaround for 16TB limit
5937 - LOV stripe allocation improved, can stripe on subset of OSTs
5938 - LOV file size/IO offset was wrong for files > 4GB in size
5939 - object EA data was being dropped, caused files to be unreadable
5940 - memory overflow with non-LOV OST caused memory corruption
5941 - fixed regression tests to work with new config tools, obdfilter
5942 - fixed bug when directory size became larger than 1 block
5943 - fixed bug (for single client case) when PWD was deleted
5944 - invalidate local directory pages when doing intent-based ops
5945 - avoid LDLM oops when lock callback contained bad data
5947 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5949 * documentation updates
5950 * bug fixes and cleanups
5951 - configuration tools
5954 - 64-bit compile warnings
5955 - 64-bit internal statfs data
5957 * test_brw on persistent OST devices
5959 * lprocfs (disabled)
5961 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5963 * documentation updates
5964 * bug fixes and cleanups
5965 * configuration tools
5967 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5970 * many small fixes to 0_5_4
5971 * io/network handling
5972 * thinkos in MDS operations
5974 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5977 * crucial basic fixes to 0.5.3
5978 * IOR, Iozone work over Elan
5981 2002-08-07 Phil Schwan <phil@clusterfs.com>
5982 * version 0_5_3, our first alpha
5983 * we use the new Portals iovs
5984 * documentation updates
5985 * bug fixes and cleanups
5986 * small changes in the DLM wire protocol
5988 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5989 * version 0_5_1 with some initial stability,
5990 * locking on MD and file I/O.
5991 * documentation updates
5992 * several bug fixes since 0.5.0
5993 * small changes in wire protocol
5995 2002-07-18 Phil Schwan <phil@clusterfs.com>
5997 * delivered as Lustre Light Alpha
5998 * fixed a crash after handling invalid MDS requests
5999 * fixed directory pages for architectures with non-4k pages sizes
6001 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6002 * release version v0_4_4
6003 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6005 * Fixes a number of interruption problems with OST operations.
6006 * Update documentation for portals header changes
6007 * Move all wire protocol structs/defines to lustre_idl.h
6008 * Fixes symlink length bug.
6009 * Add tcpdump to repository.
6011 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6012 * release version v0_4_3
6013 * Fixes statfs for inodes on extN.
6014 * Fixes bug in runtests which would delete /etc/hosts.
6015 * Use 64-bit object IDs wherever possible (not into VFS though)
6016 Remove ost_get_info, which is unused by lustre, and out of date.
6018 2002-07-03 Peter Braam <braam@clusterfs.com>
6019 * release version v0_4_2 Fixes a lookup error (type not passed)
6020 * move forward to head of Portals
6021 * move forward to latest Lustre kernel
6023 2002-06-25 Peter Braam <braam@clusterfs.com>
6024 * release version v0_4_1. Hopefully stable on single node use.