1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.39.3 (SLES 10),
5 2.6.18-128.1.14.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.41.6.sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
14 * File join has been disabled in this release, refer to Bugzilla 16929.
18 Description: Error handling in osc_statfs_interpret() has been improved.
19 Details : Check in osc_statfs_interpret() for EBADR.
23 Description: Do not send statfs() requests to OSTs disabled by administrator.
24 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
26 Severity : enhancement
28 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
30 Severity : enhancement
32 Description: Enhancement for lfs(1) command to use numeric uid/gid.
36 Description: Improved cleanup procedure for conf-sanity test_46a.
40 Description: Do not start bulk transfer, if client tried to reconnect.
41 Details : Do not start bulk transfer, if client tried to reconnect,
42 but failed due to active rpcs. Check for exp_abort_active_req
47 Description: Prevent inconsistences between linux and lustre mount structures.
48 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
49 Make the sleep interruptible.
51 Severity : enhancement
53 Description: provide server to client comms path
54 Details : server to client communications path via new LDLM_SET_INFO rpc
58 Description: truncate starts GFP_FS allocation under transaction causing deadlock
59 Details : ldiskfs_truncate calls grab_cache_page which may start page
60 allocation under an open transaction. This may lead to
61 calling prune_icache with consequent lustre reentrance.
65 Description: send correctly lsm on open replay
66 Details : MDS is trust to LSM size on replay open, but client can set wrong size
71 Description: enable client interface failover
72 Details : When a child reconnects from another NID, properly update export
73 nid hash position and ldlm reverse import.
75 Severity : enhancement
77 Description: Communicate OST degraded/readonly state via statfs to MDS
78 Details : Flags in the statfs returned from OSTs indicate whether the
79 OST is in a degraded RAID state, or if the filesystem has
80 turned read-only after a filesystem error is detected.
85 Description: don't panic if EPROTO was hit when reading symlink
86 Details : correctly handling request reference in error cases.
88 Severity : enhancement
90 Description: Add LustreNetLink, a kernel-userspace communcation path.
92 Severity : enhancement
94 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
99 Description: access to llog context before init.
100 Details : move handling CATALOGS file at osc layer and forbid access to llog
104 Frequency : rare, only when client is writing to OST on local node
106 Description: Avoid deadlock for local client writes
107 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
108 memory freeing context. This allows OST threads to set the
109 PF_MEMALLOC flag on task structures in order to allocate memory
110 from reserved pools and complete IO.
111 Use GFP_HIGHUSER for OST allocations for non-local client writes,
112 so that the OST threads generate memory pressure and allow
113 inactive pages to be reclaimed.
115 Severity : enhancement
117 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
122 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
123 Details : move ldlm namespace creation in setup phase to avoid grab
124 _lprocfs_lock with cli_sem held.
128 Description: Temporarily disable grant shrink.
129 Details : Disable the feature for debugging.
132 Frequency : only when formatting test filesystems
134 Description: Unable to run several mkfs.lustre on loop devices at the same time
135 Details : mkfs.lustre returns error 256 on the concurrent loop devices
136 formatting. The solution is to proper handle the error.
138 Severity : enhancement
140 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
142 Severity : enhancement
144 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
149 Description: resolve race between obd_disconnect and class_disconnect_exports
150 Details : if obd_disconnect will be called to already disconnected export he
151 forget release one reference and osc module can't unloaded.
153 Severity : enhancement
155 Description: Allow tuning service thread via /proc
156 Details : For each service a new
157 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
158 created that can be used to set min/max thread counts, and get the
159 current number of running threads.
164 Descriptoin: don't return error if have partially created objects for file.
165 Details : lov_update_create_set uses set->set_success as index for created
166 objects, so if some requests will be failed, they will have hole
167 at end of array and qos_shrink_lsm for allocate correct lsm.
169 Severity : enhancement
171 Description: Update OFED support to 1.4.1
176 Description: Reduce small size read RPC
177 Details : Set read-ahead limit for every file and only do read-ahead when
178 available read-ahead pages are bigger than 1M to avoid small size
181 Severity : enhancement
183 Description: move AT tunable parameters for more consistent usage
184 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
186 Severity : enhancement
188 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
189 Details : allow skip disconnected ost for send statfs request and hide error
195 Description: fix racy locking of mballoc block bitmaps causing BUG
196 Details : The locking of the mballoc buddy bitmap and the in-memory
197 block bitmap was using two different spin locks in some
198 cases. This made it possible to incorrectly access the
199 mballoc bitmap while another process was modifying it,
200 causing a sanity assertion to fail. While no on-disk corruption
201 was reported, there was some risk of this happening.
204 Frequency : rare, on llog test 6
206 Descriptoin: don't allow connect to already connected import
207 Details : allowing connect to already connected import is hide connecting problem.
212 Descriptoin: allow kill process which wait statahead result
213 Details : for some reasons 'ls' can stick in waiting result from statahead,
214 in this case need way for kill this process.
216 Severity : enhancement
218 Description: Add state history info file, enhance import info file
219 Details : Track import connection state changes in a new osc/mdc proc file;
220 add overview-type data to the osc/mdc import proc file.
222 Severity : enhancement
224 Description: MDS create should not wait for statfs RPC while holding DLM lock.
226 Severity : enhancement
228 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
233 Descriptoin: OST grant too much space to client even there are not enough space.
234 Details : Client will shrink its grant cache to OST if there are no write
235 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
236 this grant cache if there are already not enough avaible space
237 (left_space < total_clients * 32M).
240 Frequency : start MDS on uncleanly shutdowned MDS device
242 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
243 Details : stay in waiting mds<>ost recovery finished produce random bugs
244 due race between two ll_sync thread for one lov target. send
245 ACTIVATE event only if connect realy finished and import have
249 Frequency : rare, connect and disconnect target at same time
251 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
252 Details : don't call obd_disconnect under lov_lock. this long time
253 operation and can block ptlrpcd which answer to connect request.
258 Descriptoin: don't lose wakeup for imp_recovery_waitq
259 Details : recover_import_no_retry or invalidate_import and import_close can
260 both sleep on imp_recovery_waitq, but we was send only one wakeup
264 Frequency : always with long access acl
266 Descriptoin: mds can't pack reply with long acl.
267 Details : mds don't control size of acl but they limited by reint/getattr
270 Severity : enhancement
272 Description: Update to SLES10 kernel-2.6.16.60-0.33.
274 Severity : enhancement
276 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
279 Frequency : start MDS on uncleanly shutdowned MDS device
281 Descriptoin: aborting recovery hang on MDS
282 Details : don't throttle destroy RPCs for the MDT.
285 Frequency : on remount
287 Description: external journal device not working after the remount
288 Details : clear dev_rdonly flag for external journal devices in
294 Description: shutdown vs evict race
295 Details : client_disconnect_export vs connect request race.
296 if client will evicted at this time - we start invalidate
297 thread without referece to import and import can be freed
301 Frequency : rare, need acl's on inode.
303 Description: client can't handle ost additional correctly
304 Details : if ost was added after client connected to mds client can have
305 hit lnet_try_match_md ... to big messages to wide striped files.
306 in this case need teach client to handle config events about add
307 lov target and update client max ea size at that event.
309 Severity : enhancement
311 Description: Changelogs
312 Details : Changelogs are a lightweight mechanism to track filesystem
313 metadata and namespace changes. The changelog is recorded
314 permanently on the MDTs, and is periodically "consumed" / purged
315 when records are no longer needed.
317 Severity : enhancement
319 Description: compact fld format with extents
320 Details : Store range of seq rather than every seq in FLD. Seq
321 controller update FLD rather than clients. In Case of CMD, mdt0
322 has FLD, all other metadata server act as non persistent proxy
323 for FLD queries and cache fld entries in fld cache.
328 Description: don't skip ost target if they assigned to file
329 Details : Drop slow OSCs if we can, but not for requested start idx.
330 This means "if OSC is slow and it is not the requested
331 start OST, then it can be skipped, otherwise skip it only
332 if it is inactive/recovering/out-of-space.
336 Description: more cleanup in mds_lov
337 Details : not send LOV EA under replay, we can't know about they size at this
338 time. Don't allow client connect to mds before any ost connected,
339 for avoid problems with LOV EA size and returning EIO to client.
341 Severity : enhancement
343 Description: Interoperability at server side (Disk interoperability)
345 Severity : enhancement
347 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
349 Severity : enhancement
351 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
353 Severity : enhancement
355 Description: New client IO stack (CLIO).
357 Severity : enhancement
359 Description: Commit on sharing. Eliminate inter-client dependencies between
360 uncommitted transactions by doing transaction commits.
361 Thereby clients may recovery independently.
364 Frequency : Create a symlink file with a very long name
366 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
367 Details : If there is no extra space in the request for early cancels,
368 ldlm_req_handles_avail() returns 0 instead of a negative value.
370 Severity : enhancement
372 Description: Add /proc entry for import status
373 Details : The mdc, osc, and mgc import directories now have
374 an import directory that contains useful import data for debugging
377 Severity : enhancement
379 Description: Re-disable certain /proc logging
380 Details : Enable and disable client's offset_stats, extents_stats and
381 extents_stats_per_process stats logging on the fly.
384 Frequency : Only on FC kernels 2.6.22+
386 Description: oops in statahead
387 Details : Do not drop reference count for the dentry from VFS when lookup,
388 VFS will do that by itself.
390 Severity : enhancement
392 Description: Generic /proc file permissions
393 Details : Set /Proc file permissions in a more generic way to enable non-
394 root users operate on some /proc files.
398 Description: Hitting mdc_commit_close() ASSERTION
399 Details : Properly handle request reference release in
400 ll_release_openhandle().
404 Description: quota recovery deadlock during mds failover
405 Details : This patch includes att18982, att18236, att18237 in bz14840.
407 1. fix osts hang when mds does failover with quotaon
408 2. prevent watchdog storm when osts threads wait for the
413 Frequency : only patchless client
414 Description: add workaround for race between add/remove dentry from hash
416 Severity : enhancement
418 Description: Allow OST glimpses to return PW locks
422 Description: LBUG when llog conf file is full
423 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
427 Description: Prevent import from entering FULL state when server in recovery
431 Description: service mount cannot take device name with ":"
432 Details : Only when device name contains ":/" will mount treat it as
438 Description: replace ptlrpcd with the statahead thread to interpret the async
439 statahead RPC callback
443 Frequency : on recovery
444 Description: I/O failures after umount during fail back
445 Details : if client reconnected to restarted server we need join to recovery
446 instead of find server handler is changed and process self
447 eviction with cancel all locks.
449 Severity : enhancement
451 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
453 Severity : enhancement
455 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
457 Severity : enhancement
459 Description: Upcall on Lustre log has been dumped
460 Details : Allow for a user mode script to be called once a Lustre log has
461 been dumped. It passes the filename of the dumped log to the
462 script, the location of the script can be specified via
463 /proc/sys/lnet/debug_log_upcall.
468 Description: avoid idr_remove called for id which is not allocated.
469 Details : Move assigment s_dev for clustered nfs to end of initialization,
470 to avoid problem with error handling.
475 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
476 Details : When connection is reused this not moved from CONN_UNUSED_HASH
477 into CONN_USED_HASH and this prodice warning when put connection
478 again in unused hash.
481 Severity : enhancement
483 Description: File striping can now be set to use an arbitrary pool of OSTs.
485 Severity : enhancement
487 Description: Export bytes_read/bytes_write count on OSC/OST.
491 Description: Early reply size mismatch, MGC loses connection
492 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
493 the connect flags are properly negotiated.
498 Description: quotas are not honored with O_DIRECT
499 Details : all writes with the flag O_DIRECT will use grants which leads to
500 this problem. Now using OBD_BRW_SYNC to guard this.
504 Description: add quota statistics
505 Details : 1. sort out quota proc entries and proc code.
506 2. add quota statistics
508 Severity : enhancement
510 Description: enable quota support for HEAD.
514 Description: Properly propagate oinfo flags from lov to osc for statfs
515 Details : restore missing copy oi_flags to lov requests.
517 Severity : enhancement
519 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
520 : l_getgroups(8), lst(8), routerstat(8)
521 Details : included man pages for llobdstat(8), llstat(8),
522 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
524 Severity : enhancement
526 Description: configure's --enable-quota should check the
527 : kernel .config for CONFIG_QUOTA
528 Details : configure is terminated if --enable-quota is passed but
529 : no quota support is in kernel
533 Description: Remove portals compatibility
534 Details : Remove portals compatibility, not interoperable with releases
539 Description: Resolve device initialization race
540 Details : Prevent proc handler from accessing devices added to the
541 obd_devs array but yet be intialized.
543 Severity : enhancement
545 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
547 Severity : enhancement
549 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
554 Description: Using wrong pointer in osc_brw_prep_request
555 Details : Access to array[-1] can produce panic if kernel compiled with
556 CONFIG_PAGE_ALLOC enabled
560 Description: Client runs out of low memory
561 Details : Consider only lowmem when counting initial number of llap pages
565 Description: *optional* service tags registration
566 Details : if the "service tags" package is installed on a Lustre node
567 When the filesystem is mounted, a local-node service tag will
568 be created. See http://inventory.sun.com/ for more information
569 about the Service Tags asset management system.
573 Description: Kernel BUG tries to release flock
574 Details : Lustre does not destroy flock lock before last reference goes
575 away. So always drop flock locks when client is evicted and
576 perform unlock regardless of successfulness of speaking to MDS.
580 Description: add refcount protection for osc callbacks, avoid panic on shutdown
584 Description: sanity test 65a fails if stripecount of -1 is set
585 Details : handle -1 striping on filesystem in ll_dirstripe_verify
590 Description: ASSERTION(CheckWriteback(page,cmd)) failed
591 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
594 Severity : enhancement
596 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
600 Description: do not process already freed flock
601 Details : flock can possibly be freed by another thread before it reaches
602 to ldlm_flock_completion_ast.
606 Description: LBUG during stress test
607 Details : Need properly lock accesses the flock deadlock detection list.
611 Description: oops in page fault handler
612 Details : kernel page fault handler can return two special 'pages' in error
613 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
617 Description: timeout with invalidate import.
618 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
619 should be handled by ptlrpcd. This produce long age waiting and
620 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
622 Severity : enhancement
624 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
628 Description: enable MGS and MDT services start separately
629 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
630 a co-located MGS without starting the MGS, which is a complement
631 to 'nosvc' mount option.
635 Frequency : after recovery
636 Description: precreate to many object's after del orphan.
637 Details : del orphan st in oscc last_id == next_id and this triger growing
638 count of precreated objects. Set flag LOW to skip increase count
639 of precreated objects.
643 Frequency : rare, on clear nid stats
644 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
645 Details : when clean nid stats sometimes try destroy live entry,
646 and this produce panic in free.
650 Description: Stack overflow during MDS log replay
651 ease stack pressure by using a thread dealing llog_process.
655 Description: wait until IO finished before start new when do lock cancel.
656 Details : VM protocol want old IO finished before start new, in this case
657 need wait until PG_writeback is cleared until check dirty flag and
658 call writepages in lock cancel callback.
660 Severity : enhancement
662 Description: using special macro for print time and cleanup in includes.
666 Description: mds_mfd_close() ASSERTION(rc == 0)
667 Details : In mds_mfd_close(), we need protect inode's writecount change
668 within its orphan write semaphore to prevent possible races.
672 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
676 Frequency : rare, on shutdown ost
677 Description: don't hit live lock with umount ost.
678 Details : shrink_dcache_parent can be in long loop with destroy dentries,
679 use shrink_dcache_sb instead.
683 Description: don't panic with use echo client
684 Details : echo client pass NULL as client nid pointer and this produce null
689 Description: fix build on ppc32
690 Details : compile code with -m64 flag produce wrong object file for ppc32.
694 Description: add message levels for liblustreapi
698 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
699 Details : Move the 'good_osts' check before the 'total_bavail' check. This
700 will result in an -EAGAIN and in the exit call path we call
701 alloc_rr() which will with increasing aggressiveness attempt to
702 aquire precreated objects on the minimum number of required OSCs.
706 Description: Use old size assignment to avoid deadlock
707 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
708 the scheduling while holding a spinlock. We do not need locking
709 for size in ll_update_inode() because size is only updated from
710 the MDS for directories or files without objects, so there is no
711 other place to do the update, and concurrent access to such inodes
712 are protected by the inode lock.
716 Description: resolve "_IOWR redefined" build error on SLES10
720 Description: dump the memory debugging after all modules are unloaded to
721 suppress false negative in conf_sanity test 39
723 Severity : enhancement
725 Description: build kernel-ib packages for OFED 1.3 in our release cycle
730 Description: fix SLES kernel versioning
731 Details : the kernel version for our SLES 10 kernel did not include a "-"
732 before the "smp" at the end. while this was not a problem in
733 general, it did mean that software trying to use the kernel
734 version to try to detect a vendor specific kernel would fail.
735 this was most evident by the OFED build scripts.
739 Description: Don't update lov_desc members until making sure they are valid
740 Details : When updating lov_desc members via proc fs, need fix their
741 validities before doing the real update.
745 Description: don't put request into delay list while invalidate in flight.
746 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
747 import in flight. this produce timeout for invalidate and sometimes
748 can cause stale data.
752 Frequency : on ppc only
753 Description: not convert ost objects for directory because it's not exist.
754 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
758 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
760 Severity : enhancement
762 Description: Update to support 2.6.22.14 vanilla kernel.
766 Frequency : rare, on recovery
767 Description: read procfs can produce deadlock in some situation
768 Details : Holding lprocfs lock which send rpc can produce block for destroy
769 obd objects and this also block reconnect with -EALREADY. This
770 isn't fix all lprocfs bugs - but make it rare.
772 Severity : enhancement
774 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
777 Frequency : frequent on X2 node
779 Description: mdc_set_open_replay_data LBUG
780 Details : Set replay data for requests that are eligible for replay.
784 Description: lustre_mgs: operation 101 on unconnected MGS
785 Details : When MGC is disconnected from MGS long enough, MGS will evict the
786 MGC, and late on MGC cannot successfully connect to MGS and a lot
787 of the error messages complaining that MGS is not connected.
791 Frequency : on network error
792 Description: panic with double free request if network error
793 Details : mdc_finish_enqueue is finish request if any network error ocuring,
794 but it's true only for synchronus enqueue, for async enqueue
795 (via ptlrpcd) this incorrect and ptlrpcd want finish request
798 Severity : enhancement
800 Description: client-side metadata stat-ahead during readdir(directory readahead)
801 Details : perform client-side metadata stat-ahead when the client detects
802 readdir and sequential stat of dir entries therein
805 Frequency : on start mds
807 Description: Implement get_info(last_id) in obdfilter.
810 Frequency : occasional
812 Description: Correctly check stale fid, not start epoch if ost not support SOM
813 Details : open with flag O_CREATE need set old fid in op_fid3 because
814 op_fid2 was overwritten with new generated fid, but mds can answer
815 with one of these two fids and both is not stale. Setattr
816 incorrectly started an epoch and assume will be called
817 done_writing, but without SOM done_writing ever being called.
820 Frequency : rare, depends on device drivers and load
822 Description: MDS or OSS nodes crash due to stack overflow
823 Details : Code changes in 1.8.0 increased the stack usage of some functions.
824 In some cases, in conjunction with device drivers that use a lot
825 of stack the MDS (or possibly OSS) service threads could overflow
826 the stack. One change which was identified to consume additional
827 stack has been reworked to avoid the extra stack usage.
830 Frequency : occasional
832 Description: Do not fail import if osc_interpret_create gets -EAGAIN
833 Details : If osc_interpret_create got -EAGAIN it immediately exits and
834 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
835 oscc_has_objects and see OSC has no objests and call
836 oscc_internal_create to resend create request.
838 Severity : enhancement
840 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
842 Severity : enhancement
844 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
847 Frequency : very rare
849 Description: Fix replay if there is an un-replied request and open
850 Details : In some cases, older replay request will revert the
851 mcd->mcd_last_xid on MDS which is used to record the client's
854 Severity : enhancement
856 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
858 Severity : enhancement
860 Description: Add rhel5 support to HEAD.
862 Serverity : enhancement
864 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
869 Description: Don't allow skipping OSTs if index has been specified.
870 Details : Don't allow skipping OSTs if index has been specified, make
871 locking in internal create lots better.
875 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
876 Details : access to bitfield in structure is always rounded to long
877 and this produce problem with not atomic change any bit.
881 Description: Lustre make rpms failed.
882 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
883 builds Lustre rpms from tarball.
886 Frequency : rare on shutdown ost
888 Description: If llog cancel was not send before clean_exports phase, this can
889 produce deadlock in llog code.
890 Details : If llog thread has last reference to obd and call class_import_put
891 this produce deadlock because llog_cleanup_commit_master wait when
892 last llog_commit_thread exited, but this never success because was
893 called from llog_commit_thread.
897 Description: allow userland application know is lost one of stripes.
898 Details : fill lvb_blocks with error code on ost and return it to
899 application if error flag found.
903 Description: NULL lov_tgts causing MDS oops
904 Details : more safe checks for NULL lov_tgts for avoid oops.
906 Severity : enhancement
908 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
912 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
913 Details : If an OST is down the MDS will hang indefinitely in
914 obd_statfs_rqset() waiting for the statfs data. While for
915 MDS QOS usage of statfs, it should not stuck in waiting.
917 Severity : enhancement
919 Description: remote_acl support
920 Details : Support ACL-based permission check for remote user.
921 Support setfacl/getfacl for remote user with the utils
922 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
923 the system "{s,g}etfacl" utils.
925 Severity : enhancement
927 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
929 Severity : enhancement
931 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
935 Description: make mgs_setparam() handle fsname containing dash
936 Details : fsname containing a dash does not work with lctl conf_param
938 Severity : enhancement
940 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
942 Severity : enhancement
944 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
948 Frequency : rare, at shutdown
949 Description: access already free / zero obd_namespace.
950 Details : if client_disconnect_export was called without force flag set,
951 and exist connect request in flight, this can produce access to
952 NULL pointer (or already free pointer) when connect_interpret
953 store ocd flags in obd_namespace.
957 Frequency : only at startup
958 Description: not alloc memory with spinlock held.
959 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
960 if any spinlock held.
962 Severity : enhancement
964 Description: make lustre randomly fail allocating memory
965 Details : Make lustre randomly failed allocating memory for testing purpose.
967 Severity : enhancement
969 Description: lost problems with lov objid file
970 Details : Fixes some scability and access to not inited memory problems
971 in work with lov objdid file.
976 Description: lfs find does not continue on file error
977 Details : Continue other files processing when a file/dir is absent.
981 Description: Inconsistent usage of lustre_pack_reply()
982 Details : Standardize the usage of lustre_pack_reply() such that it
983 always generate a CERROR on failure.
986 Frequency : occasional
988 Description: MDS hang or stay in waiting lock
989 Details : If client receive lock with CBPENDING flag ldlm need send lock
990 cancel as separate rpc, to avoid situation when cancel request
991 can't processed due all i/o threads stay in wait lock.
995 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
999 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1003 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1007 Description: Update to RHEL4 latest kernel.
1009 Severity : enhancement
1011 Description: Build SLES10 patchless client fails
1012 Details : The configure was broken by run ./configure with
1013 --with-linux-obj=.... argument for patchless client. When the
1014 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1015 can't search header adequately. Use absolute path such as
1016 -I($LINUX)/include instead.
1020 Description: interrupt oig_wait produce painc on resend.
1021 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1022 private set, and this produce situation when rq_ptlrpcd_data not
1023 copyed to new allocated request and triggered LBUG on assert
1024 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1025 ptlrpcd set if request is changed and can be safety changed to use
1030 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1031 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1032 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1033 these macros and replace them with their difinitions instead.
1037 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1038 Details : Update which_patch & target file for SLES10 latest kernel.
1041 Description: Remove client patches from SLES 10 kernel.
1042 Details : This causes SLES 10 clients to behave as patchless clients
1043 even on a Lustre-patched (server) kernel.
1045 Severity : enhancement
1047 Description: self-adjustable client's lru lists
1048 Details : use adaptive algorithm for managing client cached locks lru
1049 lists according to current server load, other client's work
1050 pattern, memory activities, etc. Both, server and client
1051 side namespaces provide number of proc tunables for controlling
1054 Severity : enhancement
1056 Description: light-weight GSS support
1057 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1058 and reduce performance overhead.
1060 Severity : enhancement
1062 Description: Linux keyring support
1063 Details : Support using service of Linux keyring for Lustre GSS internal
1064 context refresh/cache mechanism.
1068 Description: Fix errors in lfs documentation
1069 Details : Fixes man pages
1073 Description: don't use GFP_* in generic Lustre code.
1074 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1078 Description: obdclass is limited by single OBD_ALLOC(idarray)
1079 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1083 Description: warnings with build patchless client with vanila 2.6.19 and up
1084 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1085 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1089 Frequency : Only for SLES
1090 Description: sanity_quota fail test_1
1091 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1092 file, which makes TSTID[2] inunique.
1096 Frequency : Only for SLES9
1097 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1098 Details : Update target/ChangeLog/which_patch .
1102 Frequency : Only for test_10 in sanity-quota.sh
1103 Description: a bug in quota test code
1104 Details : checking a test flag in a wrong place
1108 Frequency : Only for test_13 in sanity-quota.sh
1109 Description: a bug in quota test script
1110 Details : shouldn't check overfull allocated quota
1114 Description: jbd statistics
1115 Details : Port older jbd statistics patch for sles10
1117 Severity : enhancement
1119 Description: Add printing inode info into message about error in writepage.
1124 Description: reply_lock_interpret crash due to race with it and lock cancel.
1125 Details : Do not replay locks that are being cancelled. Do not reference
1126 locks by their address during replay, just by their handle.
1130 Frequency : When flocks are used.
1131 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1132 Details : Flock locks might destroy just granted lock if it could be merged
1133 with another existing flock, this is done in completion handler,
1134 so teach ldlm_cli_enquque_fini that this is a valid case for
1140 Description: Oops in read and write path when failing to allocate lock.
1141 Details : Check if lock allocation failed and return error back.
1145 Description: lstripe command fails for valid OST index
1146 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1147 instead of lov->desc.ld_active_tgt_count.
1151 Description: sanity.sh failed test 103
1152 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1157 Description: df doesn't work properly if diskfs blocksize != 4K
1158 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1162 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1163 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1165 Severity : enhancement
1167 Description: Adaptive timeouts
1168 Details : RPC timeouts adapt to changing server load and network
1169 conditions to reduce resend attempts and improve recovery time.
1173 Description: llapi_file_create() does not allow some changes
1174 Details : add llapi_file_open() that allows specifying the mode and
1175 open flags, and also returns an open file handle.
1179 Description: merge and cleanup kernel patches.
1180 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1184 Description: Add journal checksum support.(Kernel part)
1185 Details : The journal checksum feature adds two new flags i.e
1186 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1187 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1188 indicates that the commit block contains the checksum for
1189 the blocks described by the descriptor blocks. Now commit
1190 record can be sent to disk without waiting for descriptor
1191 blocks to be written to disk. This behavior is controlled
1192 using JBD2_FEATURE_ASYNC_COMMIT flag.
1196 Description: OSS needs mutliple precreate threads
1197 Details : Add ability to start more than one create thread per OSS.
1201 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1202 Details : execute lfs setstripe on client
1206 Description: mds_obd_create error creating tmp object
1207 Details : When the user sets quota on root, llog will be affected and can't
1208 create files and write files.
1211 Frequency : Always on ia64 patchless client, and possibly others.
1213 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1214 Details : This allows the patchless client to be loaded on architectures
1215 without this export.
1219 Description: RedHat Update kernel for RHEL5
1220 Details : Add the kernel config file for RHEL5.
1224 Description: RedHat Update kernel for RHEL5
1225 Details : Modify the kernel config file more closer RHEL5.
1229 Description: Build failure against Centos5 (RHEL5)
1230 Details : Define PAGE_SIZE when it isn't present.
1234 Description: test_7 on sanity-quota.sh can't be used on separate machines
1235 Details : add facet to handle it
1239 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1240 Details : with NFS, the anon dentry's parent was set to itself in
1241 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1242 corresponding dentry other than use rec->ur_name.
1244 Severity : enhancement
1246 Description: lfs setstripe enhancement
1247 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1251 Description: enable data checksumming by default
1252 Details : enable checksum by default, allow --disable-checksum
1253 configure option and "-o nochecksum" mount option. Checksums
1254 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1258 Description: lustre support for RHEL5
1259 Details : Add support for RHEL5.
1263 Description: Client eviction due to failover config
1264 Details : after a connection loss, the lustre client should attempt to
1265 reconnect to the last active server first before trying the
1266 other potential connections.
1270 Description: when mds and osts use different quota unit(32bit and 64bit),
1271 quota will be released repeatly.
1272 Details : void sending multiple quota reqs to mds, which will keep the
1273 status between the reqs.
1277 Description: osts not allocated evenly to files
1278 Details : change the condition to increase offset_idx
1281 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1283 Description: Data corruption for OSTs that are formatted larger than 2TB
1285 Details : When generating the bio request for lustre file writes the
1286 sector number would overflow a temporary variable before being
1287 used for the IO. The data reads correctly from Lustre (which
1288 will overflow in a similar manner) but other file data or
1289 filesystem metadata may be corrupted in some cases.
1293 Description: Tune the kernel for good SCSI performance.
1294 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1295 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1300 Description: rewrite ext2-derived code in obdclass/uuid.c
1301 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1302 from scratch preserving functionality.
1306 Description: "lfs find -obd UUID" prints directories
1307 Details : "lfs find -obd UUID" will return all directory names instead
1308 of just file names. It is incorrect because the directories
1309 do not reside on the OSTs.
1313 Description: use i_size_read and i_size_write in 2.6 port
1314 Details : replace inode->i_size access with i_size_read/write()
1317 Frequency : when removing large files
1319 Description: scheduling issue during removal of large Lustre files
1320 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1321 It causes scheduling issues when removing large files (17TB in the
1325 Frequency : only with liblustre clients on XT3
1327 Description: evictions taking too long
1328 Details : allow llrd to evict clients directly on OSTs
1331 Frequency : only on ppc
1333 Description: /proc/fs/lustre/devices broken on ppc
1334 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1335 We should make sure the type of 'index' is loff_t and then cast
1336 explicitly as needed below. Do not assign an explicitly cast
1340 Frequency : only for rhel5
1342 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1343 Details : Modify the target file & which_kernel.
1345 Severity : enhancement
1347 Description: omit set fsid for export NFS
1348 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1349 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1353 Frequency : Always, for kernels after 2.6.16
1354 Description: Fix warning idr_remove called for id=.. which is not allocated.
1355 Details : Last kernels save old s_dev before kill super and not allow
1356 to restore from callback - restore it before call kill_anon_super.
1360 Description: buffer overruns could theoretically occur
1361 Details : llapi_semantic_traverse() modifies the "path" argument by
1362 appending values to the end of the origin string, and a
1363 overrun may occur. Adding buffer overrun check in liblustreapi.
1367 Description: Fix error on 'ls .' at the top of the Lustre mount.
1368 Details : Don't revalidate dentry if it is a root dentry.
1372 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1373 Details : Modify vm-tunables-rhel4.patch.
1377 Description: Kernel config for 2.6.18-vanilla.
1378 Details : Modify targets/2.6-vanilla.target.in.
1379 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1380 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1381 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1382 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1386 Description: improve handling recoverable errors
1387 Details : if request processig with error which can be recoverable on server
1388 request should be resend, otherwise page released from cache and
1393 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1394 Details : Modify target file & which_patch.
1395 A flaw was found in the IA32 system call emulation provided
1396 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1397 value could be stored in the %RAX register, which could trigger an
1398 out-of-bounds system call table access. An untrusted local user
1399 could exploit this flaw to run code in the kernel
1400 (ie a root privilege escalation). (CVE-2007-4573).
1404 Description: O_DIRECT bypasses client statistics.
1405 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1406 read_ahead_stats were updated. copied the stats section from
1407 osc_send_oap_rpc() into async_internal().
1411 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1415 Description: change order of libsysio includes
1416 Details : '#include sysio.h' should always come before '#include xtio.h'
1420 Description: handle "serious error: objid * already exists" more gracefully
1421 Details : If LAST_ID value on disk is smaller than the objects existing in
1422 the O/0/d* directories, it indicates disk corruption and causes an
1423 LBUG(). If the object is 0-length, then we should use the existing
1424 object. This will help to avoid a full fsck in most cases.
1426 Severity : enhancement
1428 Description: adapt the lustre_config script to support the upgrade case
1429 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1430 server targets to 1.6 in parallel.
1434 Description: To avoid grant space > avaible space when the disk is almost
1435 full. Without this patch you might see the error "grant XXXX >
1436 available" or some LBUG about grant, when the disk is almost
1438 Details : In filter_check_grant, for non_grant cache write, we should
1439 check the left space by if (*left > ungranted + bytes), instead
1440 of (*left > ungranted), because only we are sure the left space
1441 is enough for another "bytes", then the ungrant space should be
1442 increase. In client, we should update cl_avail_grant only there
1443 is OBD_MD_FLGRANT in the reply.
1447 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1448 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1449 described in CVE-2007-4573. This problem could allow untrusted
1450 local users to gain root access.
1453 Frequency : when using O_DIRECT and quotas
1455 Description: Incorrect file ownership on O_DIRECT output files
1456 Details : block usage reported by 'lfs quota' does not take into account
1457 files that have been written with O_DIRECT.
1462 Description: touch file failed when fs is not full
1463 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1464 otherwise we can get ENOSP while fs is not full.
1468 Description: parallel lock callbacks
1469 Details : Instead of sending blocking and completion callbacks as separated
1470 requests, adding them to a set and sending in parallel.
1473 Frequency : only for Cray XT3
1474 Bugzilla : 12829/13455
1475 Description: Changing primary group doesn't change the group lustre assigns to
1477 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1478 overridden with the primary group provided by the group upcall,
1479 whereas the supplied fsgid can be trusted if it is in the list of
1480 supplementary groups returned by the group upcall.
1482 Severity : enhancement
1484 Description: Allow masking D_WARNING, D_ERROR messages from console
1485 Details : Console messages can now be disabled via lnet.printk.
1489 Description: User code with malformed file open parameter crashes client node
1490 Details : Before packing join_file req, all the related reference should be
1491 checked carefully in case some malformed flags cause fake
1492 join_file req on client.
1496 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1497 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1498 mode is equal to the granted mode. However, it should also check
1499 that LDLM_FL_CP_REQD is not set.
1503 Description: Heavy nfs access might result in deadlocks
1504 Details : After ELC code landed, it is now improper to enqueue any mds
1505 locks under och_sem, because enqueue might want to decide to
1506 cancel open locks for same inode we are holding och_sem for.
1510 Description: Client eviction while running blogbench
1511 Details : A lot of unlink operations with concurrent I/O can lead to a
1512 deadlock causing evictions. To address the problem, the number of
1513 oustanding OST_DESTROY requests is now throttled to
1514 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1515 callbacks are processed in priority.
1519 Description: enable ACLs on MDS by default
1520 Details : ACLs must be enabled on MDS by default.
1523 Frequency : PPC/PPC64 only
1525 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1526 Details : fix duplicated definitions between asm-ppc64/types.h and
1527 lustre_types.h on PPC.
1530 Frequency : PPC/PPC64 only
1532 Description: asm-ppc/segment.h does not exist
1533 Details : fix compile issue on PPC.
1537 Description: better handle error messages in extents code
1540 Frequency : RHEL4 only
1542 Description: mkfs is very slow on IA64/RHEL4
1543 Details : A performance regression has been discovered in the MPT Fusion
1544 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1545 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1546 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1548 Severity : enhancement
1550 Description: SNMP support enhancement
1551 Details : Adding total number of sampled request for an MDS node in snmp
1554 Severity : enhancement
1556 Description: Optimize ldlm waiting list processing for PR extent locks
1557 Details : When processing waiting list for read extent lock and meeting read
1558 lock that is same or wider to it that is not contended, skip
1559 processing rest of the list and immediatelly return current
1560 status of conflictness, since we are guaranteed there are no
1561 conflicting locks in the rest of the list.
1565 Description: Time out and refuse to reconnect
1566 Details : When the failover node is the primary node, it is possible
1567 to have two identical connections in imp_conn_list. We must
1568 compare not conn's pointers but NIDs, otherwise we can defeat
1569 connection throttling.
1573 Description: port llog fixes from b1_6 into HEAD
1574 Details : Port llog reference couting and some llog cleanups from b1_6
1575 (bug 10800) into HEAD, for protect from panic and access to already
1576 free llog structures.
1580 Description: Detect stride IO mode in read-ahead
1581 Details : When a client does stride read, read-ahead should detect that and
1582 read-ahead pages according to the detected stride pattern.
1586 Description: data checksumming impacts single node performance
1587 Details : add support for several checksum algorithm. Currently, only CRC32
1588 and Adler-32 are supported. The checksum type can be changed on
1589 the fly via /proc/fs/lustre/osc/*/checksum_type.
1593 Description: use adler32 for page checksums
1594 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1599 Description: build for x2 fails
1600 Details : fix compile issue on Cray systems.
1604 Description: Properly match for duplicate locks
1605 Details : Due to different lock order from skiplists code, we need to
1606 traverse entire list for now
1609 Frequency : only on PPC/SLES10
1611 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1612 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1613 be defined. Add a hack in mkfs_lustre.c to work around this compile
1618 Description: LASSERT on MDS when client holding flock lock dies
1619 Details : ldlm pool logic depends on number of granted locks equal to
1620 number of released locks which is not true for flock locks, so
1621 just exclude such locks from consideration.
1625 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1626 Details : Use fsfilt_sync() for both the whole filesystem sync and
1627 individual file sync to eliminate dangerous inode locking
1628 with I_LOCK that can lead to a deadlock.
1632 Description: performance in 1.6.3
1633 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1634 sure that 1MB requests are not fragmented by the block layer.
1636 Severity : enhancement
1638 Description: organize the server-side client stats on per-nid basis
1639 Details : Change the structure of stats under obdfilter and mds to
1647 The "uuid"s file would list the uuids of _active_ exports.
1648 And the clear entry is to clear all stats and stale nids.
1650 Severity : enhancement
1652 Description: eliminate client locks in face of contention
1653 Details : file contention detection and lockless i/o implementation
1654 for contended files.
1658 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1661 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1663 Description: MDS may lose file striping (and hence file data) in some cases
1664 Details : If there are additional extended attributes stored on the MDS,
1665 in particular ACLs, SELinux, or user attributes (if user_xattr
1666 is specified for the client mount options) then there is a risk
1667 of attribute loss. Additionally, the Lustre file striping
1668 needs to be larger than default (e.g. striped over all OSTs),
1669 and an additional attribute must be stored initially in the
1670 inode and then increase in size enough to be moved to the
1671 external attribute block (e.g. ACL growing in size) for file
1676 Description: skiplist implementation simplification
1677 Details : skiplists are used to group compatible locks on granted list
1678 that was implemented as tracking first and last lock of each
1679 lock group the patch changes that to using doubly linked lists
1683 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1684 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1685 because LASSERT on both the data supplied by a client, and the
1686 data on disk is dangerous and incorrect.
1688 Severity : enhancement
1690 Description: Slow truncate/writes to huge files at high offsets.
1691 Details : Directly associate cached pages to lock that protect those pages,
1692 this allows us to quickly find what pages to write and remove
1693 once lock callback is received.
1697 Description: more ldlm soft lockups
1698 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1699 starve other threads from the resource lock for a long time in
1700 case of long waiting queue, so change the debug level from
1701 D_OTHER to the less frequently used D_INFO.
1703 Severity : enhancement
1705 Description: add -gid, -group, -uid, -user options to lfs find
1709 Description: Hung threads in invalidate_inode_pages2_range
1710 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1711 once one is completed. As a result, some RPCs are stuck in the
1712 queue and are never sent.
1716 Description: filter threads hungs on waiting journal commit
1717 Details : Cleanup filter group llog code, then only filter group llog will
1718 be only created in the MDS/OST syncing process.
1722 Description: Procfs and llog threads access destoryed import sometimes.
1723 Details : Sync the import destoryed process with procfs and llog threads by
1724 the import refcount and semaphore.
1726 Severity : enhancement
1728 Description: openlock cache of b1_6 port to HEAD
1733 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1734 Details : If insertion of an extent fails, then discard the inode
1735 preallocation and free data blocks else it can lead to duplicate
1740 Description: don't always update ctime in ext3_xattr_set_handle()
1741 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1742 In some cases the ctime should not be updated, for example for
1743 2.0->1.8 compatibility it is necessary to delete an xattr and it
1744 should not update the ctime.
1748 Bugzilla : 15713/16362
1749 Description: Assertion in iopen_connect_dentry in 1.6.3
1750 Details : looking up an inode via iopen with the wrong generation number can
1751 populate the dcache with a disconneced dentry while the inode
1752 number is in the process of being reallocated. This causes an
1753 assertion failure in iopen since the inode's dentry list contains
1754 both a connected and disconnected dentry.
1758 Description: assertion failure in ldlm_handle2lock()
1759 Details : fix a race between class_handle_unhash() and class_handle2object()
1760 introduced in lustre 1.6.5 by bug 13622.
1765 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1766 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1771 Description: oops in sd_iostats_seq_show()
1772 Details : unloading/reloading the scsi low level driver triggers a kernel
1773 bug when trying to access the sd iostat file.
1778 Description: Kernel panics during QLogic driver reload
1779 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1780 patch, causing memory corruption.
1785 Description: journal_dev option does not work in b1_6
1786 Details : pass mount option during pre-mount.
1788 Severity : enhancement
1790 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1791 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1792 extent information of a file. It can be used to map logical blocks
1793 in a file to physical blocks in the block device.
1797 Description: LDLM soft lockups - improvement
1798 Details : It is be possible to send the lock handle along with each read
1799 or write request because the client is already doing a lock match
1800 itself so there isn't any reason the OST should have to re-do that
1804 Frequency : only on Cray X2
1806 Description: X2 build failures
1807 Details : fix build failures on Cray X2.
1811 Description: xid & resent requests
1812 Details : Initialize RPC XID from clock at startup (randomly if clock is
1815 Severity : enhancement
1817 Description: Add lustre_start utility to start or stop multiple Lustre servers
1822 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1823 Details : In case of memory pressure, list_del() can be called twice on
1824 req->rq_history_list, causing a kernel oops.
1828 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1829 Details : fix stack overflow in the distributed lock manager by defering
1830 export eviction after a failed AST to the elt thread instead of
1831 handling it in the dlm interpret routine.
1835 Description: Convert some comments to new format.
1836 Details : Update documenting comments to match doxygen conventions.
1840 Description: Grammar fixes.
1841 Details : A couple of trivial spelling fixes.
1845 Description: OSD_COUNTERS-mandatory
1846 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1847 to simplify the code.
1851 Description: simplify cmm_device freeing logic.
1852 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1856 Description: Add lockdep support to dt_object_operations locking interface.
1857 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1858 parameter indicating lock ordering. Update mdd code to use new
1863 Description: Introduce failloc constants for lockless IO tests.
1864 Details : Add two new failloc constants to test lockless IO. Only one of
1865 them in implemented---another is checked in yet to be landed
1870 Description: Add lockdep support for inode mutex.
1871 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1872 in the situations where Lustre has to lock more than one inode
1877 Description: Add optional invariants checking support.
1878 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1879 configure switch. This macro is to be used for consistency and
1880 sanity checks that are too expensive to be left in `production'
1885 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1886 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1891 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1892 Details : Use spin_lock_nested() in (the only) situation where more than
1893 one ldlm_lock is locked simultaneously. Also, fix possible
1894 dead-lock in ldlm_lock_change_resource() by enforcing particular
1899 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1900 Details : Instead of specifying each ldlm_lock call-back through separate
1901 parameter, wrap them into struct ldlm_callback_suite.
1905 Description: Kill join_lru obd method and its callers.
1906 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1907 in memory---a requirement implemented through ->o_join_lru() obd
1908 method in HEAD. Remove this method and its users.
1912 Description: Add asynchronous ldlm ENQUEUE completion handler.
1913 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1914 case that stock ldlm_completion_ast() cannot handle as it waits
1915 until lock is granted. Introduce new ldlm_completion_ast_async()
1916 for this. Also comment ldlm_completion_ast().
1920 Description: ldlm_error <-> errno conversion.
1921 Details : Add functions to map (rather arbitrary) between LDLM error codes
1922 and standard errno values. CLIO needs this to prevent LDLM specific
1923 constants from escaping ldlm and osc.
1927 Description: Kill unused ldlm_handle2lock_ns() function.
1928 Details : Kill unused ldlm_handle2lock_ns() function.
1932 Description: Add lu_ref support to ldlm_lock
1933 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1934 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1935 and ldlm_resource. LDLM interface has to be changed, because code that
1936 releases a reference on a lock, has to "know" what reference this is.
1937 In the most frequent case
1939 lock = ldlm_handle2lock(handle);
1941 LDLM_LOCK_PUT(lock);
1943 no changes are required. When any other reference (received _not_ from
1944 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1945 instead of LDLM_LOCK_PUT().
1947 Arguably, changes are pervasive, and interface requires some discipline
1948 for proper use. On the other hand, it was very instrumental in finding
1949 a few leaked lock references.
1953 Description: Add ldlm_lock_addref_try().
1954 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1955 attempts to addref a lock that might be being canceled
1960 Description: Add ldlm_weigh_callback().
1961 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1962 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1963 hard-coded `number of pages' logic.
1967 Description: Add lockdep annotations to llog code.
1968 Details : Use appropriately tagged _nested() locking calls in the places
1969 where llog takes more than one ->lgh_lock lock.
1973 Description: Add loi_kms_set().
1974 Details : Wrap kms updates into a helper function.
1978 Description: Constify instances of struct lsm_operations.
1979 Details : Constify instances of struct lsm_operations.
1983 Description: lu_conf support.
1984 Details : On a server, a file system object is uniquely identified
1985 by a fid, which is sufficient to locate and load all object
1986 state (inode). On a client, on the other hand, more data are
1987 necessary instantiate an object. Change lu_object_find() and
1988 friends to take additional `lu_conf' argument describing object.
1989 Typically this includes layout information.
1993 Description: lu_context fixes.
1994 Details : Introduce new lu_context functions that are needed on the client
1995 side, where some system threads (ptlrpcd) are shared by multiple
1996 modules, and so cannot be stopped during module shutdown.
2000 Description: Add start and stop methods to lu_device_type_operations.
2001 Details : Introduce two new methods in lu_device_type_operations, that are
2002 invoked when first instance of a given type is created and last one
2003 is destroyed respectively. This is need by CLIO.
2007 Description: Add lu_ref support to struct lu_device.
2008 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2009 track leaked references.
2013 Description: Introduce lu_kmem_descr.
2014 Details : lu_kmem_descr and its companion interface allow to create
2015 and destroy a number of kmem caches at once.
2019 Description: Fix lu_object finalization race.
2020 Details : Fix a race between lu_object_find() finding an object and its
2021 concurrent finalization. This race is (most likely) not possible
2022 on the server, but might happen on the client.
2026 Description: Introduce lu_ref interface.
2027 Details : lu_ref is a debugging module allowing to track references to
2028 a given object. It is quite cpu expensive, and has to be
2029 explicitly enabled with --enable-lu_ref. See usage description
2034 Description: Factor lu_site procfs stats into a separate function.
2035 Details : Separate lu_site stats printing code into a separate function
2036 that can be reused on a client.
2040 Description: Constify instances of struct {lu,dt,md}_device_operations.
2041 Details : Constify instances of struct {lu,dt,md}_device_operations.
2045 Description: Introduce struct md_site and move meta-data specific parts of
2046 struct lu_site here.
2047 Details : Move md-specific fields out of struct lu_site into special struct
2048 md_site, so that lu_site can be used on a client.
2052 Description: Kill mdd_lov_destroy().
2053 Details : Remove unused mdd code.
2057 Description: Add st_block checking to multistat.c.
2058 Details : Add st_block checking to multistat.c.
2062 Description: Add lu_ref support to struct obd_device.
2063 Details : Add lu_ref tracking to obd_device.
2067 Description: Kill obd_set_fail_loc().
2068 Details : Remove unused code.
2072 Description: Add special type for ptlrpc_request interpret functions.
2073 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2074 there. Actual usage will be in CLIO.
2078 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2079 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2080 doesn't work with lockdep.
2084 Description: Add rwv.c test program.
2085 Details : New testing program exercising readv(2) and writev(2) (Qian).
2089 Description: Add sendfile.c test program.
2090 Details : New testing program exercising sendfile(2) (Jay).
2094 Description: Ratelimit a message that can be very frequent.
2095 Details : Ratelimit a memory allocation failure message that can
2100 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2101 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2105 Description: Liblustre build fixes.
2106 Details : Liblustre build fixes.
2110 Description: libcfs: add cfs_{need,cond}_resched() interface.
2111 Details : libcfs: add cfs_{need,cond}_resched() definition and
2112 implementations for Linux, NT, and liblustre.
2114 Severity : enhancement
2116 Description: More exported tunables for mballoc
2117 Details : Add support for tunable preallocation window and new tunables for
2118 large/small requests
2122 Description: Detect corruption of block bitmap and checking for preallocations
2123 Details : Checks validity of on-disk block bitmap. Also it does better
2124 checking of number of applied preallocations. When corruption is
2125 found, it turns filesystem readonly to prevent further corruptions.
2129 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2130 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2131 window to 0. If it does hit cache miss, and read-pattern is still
2132 stride-io mode, does not reset the stride window, but also does not
2133 increase the stride window length in this case.
2137 Frequency : only for big-endian servers
2138 Description: Check if system is big-endian while mounting fs with extents feature
2139 Details : Mounting a filesystem with extents feature will fail on big-endian
2140 systems since ext3-based ldiskfs is not supported on big-endian
2141 systems. This can be over-riden with "bigendian_extents" mount option.
2143 Severity : enhancement
2145 Description: The root squash functionality
2146 Details : A security feature, which is to prevent users from being able
2147 to mount lustre on their desktop, run as root, and delete
2148 all of the files in the filesystem. The goal is accomplished by
2149 remapping user id (UID) and group id (GID) of the root user to
2150 a UID and GID specified by the system administartor via Lustre
2151 configuration management server (MGS). The functionality also
2152 allows to specify sets of clients for which the remapping does
2157 Description: Excessive recovery window
2158 Details : With AT enabled, the recovery window can be excessively long (6000+
2159 seconds). To address this problem, we no longer use
2160 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2161 timeout no longer depends on the service time, it is set to
2162 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2163 time via pb_service_time.
2167 Description: Watchdog triggered on MDS failover
2168 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2169 the OSTs know that the MDS "UUID" can be reused for the same export
2170 from a different NID, so we do not need to wait for the export to be
2174 Frequency : rare, only if using MMP with Linux RAID
2176 Description: MMP doesn't work with Linux RAID
2177 Details : While using HA for Lustre servers with Linux RAID, it is possible
2178 that MMP will not detect multiple mounts. To make this work we
2179 need to unplug the device queue in RAID when the MMP block is being
2180 written. Also while reading the MMP block, we should read it from
2181 disk and not the cached one.
2183 Severity : enhancement
2185 Description: open file using fid
2186 Details : A file can be opened using just its fid, like
2187 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2190 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2193 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2194 Details : add proper locking for creation of jbd_4k slab cache
2198 Description: MMP check in ext3_remount() fails without displaying any error
2199 Details : When multiple mount protection fails during remount, proper error
2202 Severity : enhancement
2204 Description: Allow stripe size to be up to 4G-64k
2205 Details : Fix math logic to allow large stripe sizes.
2209 Description: add check for >8TB ldiskfs filesystems
2210 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2211 Don't allow >8TB ldiskfs filesystems to be mounted without
2212 force_over_8tb mount option
2217 Descriptoin: OSS complains "too many blocks" and crashes
2218 Details : In filter_commitrw_write():
2219 1. lquota_chkquota();
2220 2. obdo_from_inode();
2221 3. lquota_pending_commit();
2222 "2." has changed uid and gid in this case, then caused
2223 lquota_chkquota() and lquota_pending_commit() to use
2224 different uid and gid. That is the root of the bug.
2228 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2229 Details : Detect and print out the 'stripe count = -1' case properly.
2231 --------------------------------------------------------------------------------
2233 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2235 * Support for kernels:
2236 2.6.5-7.283 (SLES 9),
2237 2.6.9-55.EL (RHEL 4),
2238 2.6.16.46-0.14 (SLES 10),
2239 2.6.18.8 vanilla (kernel.org)
2240 * Client support for unpatched kernels:
2241 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2242 2.6.16 - 2.6.22 vanilla (kernel.org)
2243 * Due to recently discovered recovery problems, we do not recommend
2244 using patchless RHEL 4 clients with this or any earlier release.
2245 * Recommended e2fsprogs version: 1.39.cfs8
2246 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2247 * Starting with this release, the ldiskfs backing filesystem required
2248 by Lustre is now in its own package, lustre-ldiskfs. This package
2249 should be installed. It is versioned separately from Lustre and
2250 may be released separately in future.
2254 Description: block reactivating mgc import until all deactivates complete
2255 Details : Fix race when failing back MDT/MGS to itself (testing)
2258 Frequency : at statup only
2260 Description: mds_lov_synchronize race leads to various problems
2261 Details : simultaneous MDT->OST connections at startup can cause the
2262 sync to abort, leaving the OSC in a bad state.
2264 Severity : enhancement
2266 Description: add optional extra BUILD_VERSION info
2267 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2268 to override the lustre version.
2271 Frequency : 2.6.18 servers only
2273 Description: ll_kern_mount() doesn't release the module reference
2274 Details : The ldiskfs module reference count never drops down to 0
2275 because ll_kern_mount() doesn't release the module reference.
2280 Description: server LBUG when using old ost_num_threads parameter
2281 Details : Accept the old ost_num_threads parameter but warn that it
2282 is deprecated, and fix an off-by-one error that caused an LBUG.
2287 Description: Transient SCSI error results in persistent IO issue
2288 Details : iobuf->dr_error is not reinitialized to 0 between two
2292 Frequency : sometimes when underlying device returns I/O errors
2294 Description: OSTs not going read-only during write failures
2295 Details : OSTs are not remounted read-only when the journal commit threads
2296 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2297 instead of the ext3 wrappers.
2301 Description: poor connect scaling with increasing client count
2302 Details : Don't run filter_grant_sanity_check for more than 100 exports
2303 to improve scaling for large numbers of clients.
2306 Frequency : SLES10 only
2308 Description: sanity-quota.sh quotacheck failed: rc = -22
2309 Details : Quotas cannot be enabled on SLES10.
2312 Frequency : liblustre clients only
2314 Description: getdirentries does not give error when run on compute nodes
2315 Details : getdirentries does not fail when the size specified as an argument
2316 is too small to contain at least one entry
2318 Severity : enhancement
2320 Description: Add LNET router traceability for debug purposes
2321 Details : If a checksum failure occurs with a router as part of the
2322 IO path, the NID of the last router that forwarded the bulk data
2323 is printed so it can be identified.
2328 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2329 Details : Due to a race condition, liblustre clients were occasionally
2330 evicted incorrectly.
2332 Severity : enhancement
2334 Description: lfs setstripe use optional parameters instead of postional
2337 Severity : enhancement
2339 Description: Nanosecond timestamp support for ldiskfs
2340 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2341 resolution timestamps. There is not yet support for this at
2342 the Lustre filesystem level.
2345 Frequency : during server recovery
2347 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2348 Details : request with rq_no_resend flag not awake l_wait_event if they get
2352 Frequency : nfs export on patchless client
2354 Description: connectathon hang when test nfs export over patchless client
2355 Details : Disconnected dentry cannot be found with lookup, so we do not need
2356 to unhash it or make it invalid
2359 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2360 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2361 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2366 Description: open req refcounting wrong on reconnect
2367 Details : If reconnect happened between getting open reply from server and
2368 call to mdc_set_replay_data in ll_file_open, we will schedule
2369 replay for unreferenced request that we are about to free.
2370 Subsequent close will crash in variety of ways.
2371 Check that request is still eligible for replay in
2372 mdc_set_replay_data().
2377 Description: disable writes to filesystem when reading health_check file
2378 Details : the default for reading the health_check proc file has changed
2379 to NOT do a journal transaction and write to disk, because this
2380 can cause reads of the /proc file to hang and block HA state
2381 checking on a healthy but otherwise heavily loaded system. It
2382 is possible to return to the previous behaviour during configure
2383 with --enable-health-write.
2385 Severity : enhancement
2387 Description: 64-bit inode version
2388 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2389 to the inode. This will be required for version-based recovery.
2394 Description: MDS fails to start if a duplicate client export is detected
2395 Details : in some rare cases it was possible for a client to connect to
2396 an MDS multiple times. Upon recovery the MDS would detect this
2397 and fail during startup. Handle this more gracefully.
2399 Severity : enhancement
2401 Description: Add -o localflock option to simulate old noflock
2403 Details : This will achieve local-only flock/fcntl locks
2409 Description: log_commit_thread vs filter_destroy race leads to crash
2410 Details : Take import reference before releasing llog record semaphore
2415 Description: Wrong request locking in request set processing
2416 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2417 imp_delayed_list, in this place should be used imp_lock.
2420 Frequency : when reconnection
2422 Description: Grant Leak when osc reconnect to OST
2423 Details : When osc reconnect ost, OST(filter) should check whether it
2424 should grant more space to client by comparing fed_grant and
2425 cl_avail_grant, and return the granted space to client instead
2426 of "new granted" space, because client will call osc_init_grant
2427 to update the client grant space info.
2430 Frequency : when client reconnect to OST
2432 Description: Grant Leak when osc do resend and replay bulk write
2433 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2434 bulk write request, because the grant info while be sync between
2435 OSC and OST when reconnect, and we should ignore the grant info
2436 these of resend/replay write req.
2441 Description: Grant space more than avaiable left space sometimes.
2442 Details : When then OST is about to be full, if two bulk writing from
2443 different clients came to OST. Accord the avaliable space of the
2444 OST, the first req should be permitted, and the second one
2445 should be denied by ENOSPC. But if the seconde arrived before
2446 the first one is commited. The OST might wrongly permit second
2447 writing, which will cause grant space > avaiable space.
2450 Frequency : when client is evicted
2452 Description: Grant might be wrongly erased when osc is evicted by OST
2453 Details : when the import is evicted by server, it will fork another
2454 thread ptlrpc_invalidate_import_thread to invalidate the
2455 import, where the grant will be set to 0. While the original
2456 thread will update the grant it got when connecting. So if
2457 the former happened latter, the grant will be wrongly errased
2458 because of this race.
2463 Description: Checking Stale with correct fid
2464 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2465 to check whether it is stale, because sometimes, we want the
2466 enqueue happened anyway, and op_data.fid2 will not be initialized.
2468 Severity : enhancement
2470 Description: update patchless client
2471 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2474 Frequency : only with 2.4 kernel
2476 Description: random memory corruption
2477 Details : size of struct ll_inode_info is to big for union inode.u and this
2478 can be cause of random memory corruption.
2483 Description: Memory leak in recovery
2484 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2485 It should also check obd_fail before fsfilt_start, otherwise if
2486 fsfilt_start return -EROFS,(failover mds during mds recovery).
2487 then the req will return with repmsg->transno = 0 and rc = EROFS.
2488 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2489 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2490 be freed no matter whether fsfilt_commit success or not.
2493 Frequency : only with huge count clients
2495 Description: Prevents from taking the superblock lock in llap_from_page for
2497 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2498 for llap_from_page call in ll_removepage prevents from taking the
2499 superblock lock for a soon died page.
2504 Description: Not check open intent error before release open handle
2505 Details : in some rare cases, the open intent error is not checked before
2506 release open handle, which may cause
2507 ASSERTION(open_req->rq_transno != 0), because it tries to release
2508 the failed open handle.
2513 Description: Set cat log bitmap only after create log success.
2514 Details : in some rare cases, the cat log bitmap is set too early. and it
2515 should be set only after create log success.
2519 Description: Accessing a block bevice can re-enable I/O when Lustre is
2520 tearing down a device.
2521 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2526 Description: service threads may hog cpus when there are a lot of requests
2527 Details : Insert cond_resched to give other threads a chance to use some CPU
2532 Description: the cat log was not initialized in recovery
2533 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2534 unlink log on mds will not be initialized until mds post
2535 recovery. And also in mds post recovery, the unlink log will
2536 initialization will be done asynchronausly, so there will be race
2537 between add unlink log and unlink log initialization.
2541 Description: brw_stats were being printed incorrectly
2542 Details : brw_stats were being printed as log2 but all of them were not
2543 recorded as log2. Also remove some code duplication arising from
2544 filter_tally_{read,write}.
2548 Frequency : rare, only in recovery.
2549 Description: ASSERTION(req->rq_type != LI_POISON) failed
2550 Details : imp_lock should be held while iterating over imp_sending_list for
2551 prevent destroy request after get timeout in ptlrpc_queue_wait.
2555 Description: replay-single.sh test 52 fails
2556 Details : A lock's skiplist need to be cleanup when it being unlinked
2557 from its resource list.
2561 Description: Short directio read returns full requested size rather than
2563 Details : Direct I/O operations should return actual amount of bytes
2564 transferred rather than requested size.
2566 Severity : enhancement
2568 Description: metadata RPC reduction (e.g. for rm performance)
2569 Details : decrease the amount of synchronous RPC between clients and servers
2570 by canceling conflicing lock before the operation on the client
2571 and packing thier handles into the main operation RPC to server.
2573 Severity : enhancement
2575 Description: Async OSC create to avoid the blocking unnecessarily.
2576 Details : If a OST has no remain object, system will block on the creating
2577 when need to create a new object on this OST. Now, ways use
2578 pre-created objects when available, instead of blocking on an
2579 empty osc while others are not empty. If we must block, we block
2580 for the shortest possible period of time.
2584 Description: improve handling recoverable errors
2585 Details : if request processig with error which can be recoverable on server
2586 request should be resend, otherwise page released from cache and
2589 Severity : enhancement
2591 Description: refine locking for avoid write wrong info into lov_objid file
2592 Details : fix possible races with add new target and write/update data in
2595 --------------------------------------------------------------------------------
2597 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2602 Frequency : on some architectures
2604 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2605 Details : Uninitialized flags sometimes cause configuration commands to
2609 Frequency : patchless clients only
2611 Description: missing __iget() symbol export
2612 Details : The __iget() symbol export is missing. To avoid the need for
2613 this on patchless clients the deathrow inode reaper is turned
2614 off, and we depend on the VM to clean up old inodes. This
2615 dependency was during via the fix for bug 12181.
2617 --------------------------------------------------------------------------------
2619 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2621 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2622 INTEROPERATE with older versions automatically. In many cases a
2623 special upgrade step is needed. Please read the
2624 user documentation before upgrading any part of a live system.
2625 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2626 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2627 * WARNING: Lustre configuration and startup changes are required with
2628 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2632 Severity : enhancement
2634 Description: MountConf
2635 Details : Lustre configuration is now managed via mkfs and mount
2636 commands instead of lmc and lconf. New obd types (MGS, MGC)
2637 are added for dynamic configuration management. See
2638 https://mail.clusterfs.com/wikis/lustre/MountConf for
2641 Severity : enhancement
2643 Description: dynamic OST addition
2644 Details : OSTs can now be added to a live filesystem
2646 Severity : enhancement
2648 Description: startup order invariance
2649 Details : MDTs and OSTs can be started in any order. Clients only
2650 require the MDT to complete startup.
2652 Severity : enhancement
2654 Description: parallel, asynchronous orphan cleanup
2655 Details : orphan cleanup is now performed in separate threads for each
2656 OST, allowing parallel non-blocking operation.
2658 Severity : enhancement
2660 Description: optimized stripe assignment
2661 Details : stripe assignments are now made based on ost space available,
2662 ost previous usage, and OSS previous usage, in order to try
2663 to optimize storage space and networking resources.
2665 Severity : enhancement
2667 Description: Permanently set tunables
2668 Details : All writable /proc/fs/lustre tunables can now be permanently
2669 set on a per-server basis, at mkfs time or on a live system.
2671 Severity : enhancement
2673 Description: Lustre message v2
2674 Details : Add lustre message format v2.
2676 Severity : enhancement
2678 Description: client OST exclusion list
2679 Details : Clients can be started with a list of OSTs that should be
2680 declared "inactive" for known non-responsive OSTs.
2684 Description: SPEC SFS validation failure on NFS v2 over lustre.
2685 Details : Changes the blocksize for regular files to be 2x RPC size,
2686 and not depend on stripe size.
2688 Severity : enhancement
2690 Description: Multiple MD RPCs in flight.
2691 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2692 To avoid overly-overloading MDS, introduce a limit on number of
2693 MDS RPCs in flight for a single client and add /proc controls
2694 to adjust this limit.
2696 Severity : enhancement
2698 Description: client read/write statistics
2699 Details : Add client read/write call usage stats for performance
2700 analysis of user processes.
2701 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2702 file access. extents_stats shows chunk size distribution.
2703 extents_stats_per_process show chunk size distribution per
2706 Severity : enhancement
2708 Description: mds statistics
2709 Details : Add detailed mds operations statistics in
2710 /proc/fs/lustre/mds/*/stats.
2714 Description: Failure of copying files with lustre special EAs.
2715 Details : Client side always return success for setxattr call for lustre
2716 special xattr (currently only "trusted.lov").
2721 Description: Refcount LNET uuids
2722 Details : The global LNET uuid list grew linearly with every startup;
2723 refcount repeated list entries instead of always adding to
2726 Severity : enhancement
2728 Description: Dynamic service threads
2729 Details : Within a small range, start extra service threads
2730 automatically when the request queue builds up.
2733 Frequency : mixed-endian client/server environments
2735 Description: mixed-endian crashes
2736 Details : The new msg_v2 system had some failures in mixed-endian
2739 Severity : enhancement
2741 Description: Easy OST removal
2742 Details : OSTs can be permanently deactivated with e.g. 'lctl
2743 conf_param lustre-OST0001.osc.active=0'
2745 Severity : enhancement
2747 Description: MGS proc entries
2748 Details : Added basic proc entries for the MGS showing what filesystems
2751 Severity : enhancement
2753 Description: provide MGS failover
2754 Details : Added config lock reacquisition after MGS server failover.
2756 Severity : enhancement
2758 Description: add Linux 2.4 support
2759 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2763 Description: a large application tries to do I/O to the same resource and dies
2764 in the middle of it.
2765 Details : Check the req->rq_arrival time after the call to
2766 ost_brw_lock_get(), but before we do anything about
2767 processing it & sending the BULK transfer request. This
2768 should help move old stale pending locks off the queue as
2769 quickly as obd_timeout.
2772 Frequency : when an incorrect nid is specified during startup
2774 Description: ptlrpc connect to non-existant node causes kernel crash
2775 Details : LNET can't be re-entered from an event callback, which
2776 happened when we expire a message after the export has been
2777 cleaned up. Instead, hand the zombie cleanup off to another
2780 Severity : enhancement
2782 Description: plain/inodebits lock performance improvement
2783 Details : Grouping plain/inodebits in granted list by their request modes
2784 and bits policy, thus improving the performance of search through
2788 Frequency : only if OST filesystem is corrupted
2790 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2791 Details : for a short time RPCs with bulk IO are in the replay list,
2792 but replay of bulk IOs is unimplemented. If the OST filesystem
2793 is corrupted due to disk cache incoherency and then replay is
2794 started it is possible to trip an assertion. Avoid putting
2795 committed RPCs into the replay list at all to avoid this issue.
2798 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2800 Description: System hang on startup
2801 Details : This bug allowed the liblustre (e.g. catamount) client to
2802 return to the app before handling all startup RPCs. This
2803 could leave the node unresponsive to lustre network traffic
2804 and manifested as a server ptllnd timeout.
2806 Severity : enhancement
2808 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2809 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2810 boolean to control whether to print peer debug info when a
2811 client's RPC times out.
2814 Frequency : only for kernels with patches from Lustre below 1.4.3
2816 Description: Remove old rdonly API
2817 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2820 Frequency : only for devices with external journals
2822 Description: Set external device read-only also
2823 Details : During a commanded failover stop, we set the disk device
2824 read-only while the server shuts down. We now also set any
2825 external journal device read-only at the same time.
2828 Frequency : when upgrading from 1.4 while trying to change parameters
2830 Description: The wrong (new) MDC name was used when setting parameters for
2831 upgraded MDT's. Also allows changing of OSC (and MDC)
2832 parameters if --writeconf is specified at tunefs upgrade time.
2835 Frequency : when setting specific ost indicies
2837 Description: QOS code breaks on skipped indicies
2838 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2839 created with --index need not be sequential.
2841 Severity : enhancement
2843 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2844 Details : The uninit_groups feature works in conjunction with the kernel
2845 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2846 up the pass1 processing of e2fsck. This is a read-only feature
2847 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2848 mount filesystems that have had this feature enabled.
2850 Severity : enhancement
2852 Description: Improve multi-block allocation algorithm to avoid fragmentation
2853 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2854 allocation locality and avoid filesystem fragmentation.
2856 ------------------------------------------------------------------------------
2858 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2860 * Support for kernels:
2861 2.4.21-47.0.1.EL (RHEL 3)
2862 2.6.5-7.283 (SLES 9)
2863 2.6.9-42.0.10.EL (RHEL 4)
2864 2.6.12.6 vanilla (kernel.org)
2865 2.6.16.27-0.9 (SLES 10)
2866 * Recommended e2fsprogs version: 1.39.cfs5
2868 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2872 Frequency : occasional, depends on client load and configuration
2873 Bugzilla : 12181, 12203
2874 Description: data loss for recently-modified files
2876 Details : In some cases it is possible that recently written or created
2877 files may not be written to disk in a timely manner (this should
2878 normally be within 30s unless client IO load is very high).
2879 The problem appears as zero-length files or files that are a
2880 multiple of 1MB in size after a client crash or client eviction
2881 that are missing data at the end of the file.
2883 This problem is more likely to be hit on clients where files are
2884 repeatedly created and unlinked in the same directory, clients
2885 have a large amount of RAM, have many CPUs, the filesystem has
2886 many OSTs, the clients are rebooted frequently, and/or the files
2887 are not accessed by other nodes after being written.
2889 The presence of the problem can be detected by looking at
2890 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2891 smaller than the second (nr_unused) then dirty files will not
2892 be flushed automatically to disk. "sync; sleep 10" should be
2893 run several times on the node before unmounting it to update
2894 Lustre (this is also safe to run on nodes without this problem).
2896 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2897 that can cause this same problem, so customers using that kernel
2898 also need to update the kernel in addition to Lustre. In order
2899 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2901 It is normal that files written just before a client crash (less
2902 than 30s) may not yet have been flushed to disk, even for local
2906 Frequency : frequent on thin XT3 nodes
2908 Description: UUID collision on thin XT3 Linux nodes
2909 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2910 randomly, since we relied on an insufficiently-seeded PRNG.
2915 Description: OSS hangs after "All ost request buffers busy"
2916 Details : A deadlock between quota and journal operations caused OSS
2917 hangs after printing "All ost request buffers busy."
2920 Frequency : always on liblustre builds
2922 Description: Cleanup compiler warnings on liblustre
2925 Frequency : always on liblustre builds on XT3
2927 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2931 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2932 > 32000 subdirectories
2933 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2934 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2935 subdirectory count crosses 32000. This will aid e2fsck to
2936 correctly handle more than 32000 subdirectories.
2939 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2941 Description: System hang on startup
2942 Details : This bug allowed the liblustre (e.g. catamount) client to
2943 return to the app before handling all startup RPCs. This
2944 could leave the node unresponsive to lustre network traffic
2945 and manifested as a server ptllnd timeout.
2947 Severity : enhancement
2949 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2950 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2951 boolean to control whether to print peer debug info when a
2952 client's RPC times out.
2957 Description: make O_SYNC working on 2.6 kernels
2958 Details : 2.6 kernels use different method for mark pages for write,
2959 so need add a code to lustre for O_SYNC work.
2964 Description: Failure to close file and release space on NFS
2965 Details : Put inode details into lock acquired in ll_intent_file_open.
2966 Use mdc_intent_lock in ll_intent_open to properly
2967 detect all kind of errors unhandled by mdc_enqueue.
2972 Description: proc file read during shutdown sometimes raced obd removal,
2974 Details : Add lock to prevent obd access after proc file removal.
2977 Frequency : Only for files larger than 4GB on 32-bit clients.
2979 Description: improperly doing page alignment of locks
2980 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2981 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2986 Description: RPCs being resent when they shouldn't be
2987 Details : Some RPCs that should not be resent are being resent. This
2988 can cause inconsistencies in the RPC state machine. Do not
2989 resend such requests.
2992 Frequency : rare, only with NFS export
2994 Description: Crash on NFS re-export node
2995 Details : under very unusual load conditions an assertion is hit in
2996 ll_intent_file_open()
2999 Frequency : only if OST filesystem is corrupted
3001 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3002 Details : for a short time RPCs with bulk IO are in the replay list,
3003 but replay of bulk IOs is unimplemented. If the OST filesystem
3004 is corrupted due to disk cache incoherency and then replay is
3005 started it is possible to trip an assertion. Avoid putting
3006 committed RPCs into the replay list at all to avoid this issue.
3011 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3012 Details : Large single O_DIRECT read and write calls can fail to allocate
3013 a sufficiently large buffer to process the request. In case of
3014 allocation failure the allocation is retried with a smaller
3015 buffer and broken into smaller requests.
3017 Severity : enhancement
3019 Description: Add -o localflock option to simulate old noflock behaviour.
3020 Details : This will achieve local-only flock/fcntl locks coherentness.
3025 Description: versioning check is incomplete
3026 Details : Checking the version difference of client vs. server, report
3027 error if the gap is too big.
3032 Description: add support PG_writeback bit
3033 Details : add support for PG_writeback bit for Lustre, for more carefull
3034 work with page cache in 2.6 kernel. This also fix some deadlocks
3035 and remove hack for work O_SYNC with 2.6 kernel.
3037 Severity : enhancement
3039 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3040 Details : The uninit_groups feature works in conjunction with the kernel
3041 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3042 up the pass1 processing of e2fsck. This is a read-only feature
3043 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3044 mount filesystems that have had this feature enabled.
3046 Severity : enhancement
3048 Description: Improve multi-block allocation algorithm to avoid fragmentation
3049 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3050 allocation locality and avoid filesystem fragmentation.
3052 ------------------------------------------------------------------------------
3054 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3056 * Support for kernels:
3057 2.6.9-42.0.3.EL (RHEL 4)
3058 2.6.5-7.276 (SLES 9)
3059 2.4.21-47.0.1.EL (RHEL 3)
3060 2.6.12.6 vanilla (kernel.org)
3061 2.6.16.21-0.8 (SLES10)
3062 * Recommended e2fsprogs version: 1.39.cfs2-0
3064 * The backwards-compatible /proc/sys/portals symlink has been removed
3065 in this release. Before upgrading, please ensure that you change
3066 any configuration scripts or /etc/sysctl.conf files that access
3067 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3068 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3069 in advance of the upgrade on any system running Lustre 1.4.6 or
3070 newer, since /proc/sys/lnet was added in that version.
3071 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3075 Frequency : only when quota is used
3077 Description: avoid scanning export list for quota master
3078 Details : Change the algorithms to avoid scanning export list in order
3079 to improve the efficiency.
3082 Frequency : MDS failover only, very rarely
3084 Description: "went back in time" messages on mds failover
3085 Details : The greatest transno may be lost when the current operation
3086 finishes with an error (transno==0) and the client's last_rcvd
3087 record is over-written. Save the greatest transno in the
3088 mds_last_transno for this case.
3091 Frequency : always for specific kernels and striping counts
3093 Description: client may get "Matching packet too big" without ACL support
3094 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3095 when trying to access files in certain configurations. The
3096 clients should in fact be denied when mounting because they do
3097 not understand ACLs.
3100 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3102 Description: many clients connecting with IO in progress causes connect timeouts
3103 Details : Avoid synchronous journal commits to avoid delays caused by many
3104 clients connecting/disconnecting when bulk IO is in progress.
3105 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3106 OST_IO_PORTAL to avoid delays behind potentially many pending
3110 Frequency : occasionally with multiple writers to a single file
3112 Description: shared writes to file may result in wrong size reported by stat()
3113 Details : Allow growing of kms when extent lock is cancelled
3116 Frequency : always with random mmap IO to multi-striped file
3118 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3119 Details : Only if the hole is at the end of OST object so that kms is too
3120 small. Fix is to increase kms accordingly in ll_nopage.
3123 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3125 Description: writes to a missing object would leak memory on the OST
3126 Details : If there is an inconsistency between the MDS and OST filesystems,
3127 such that the MDS references an object that doesn't exist, writes
3128 to that object will leak memory due to incorrect cleanup in the
3129 error handling path, eventually running out of memory on the OST.
3134 Description: Creating too long symlink causes lustre errors
3135 Details : Check symlink and name lengths before sending requests to MDS.
3138 Frequency : only if flock is enabled (not on by default)
3140 Description: posix locks not released on fd closure on 2.6.9+
3141 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3142 kernels, this caused such locks not to be released on fd close and
3143 then assertions on fs unmount about still used locks.
3146 Frequency : MDS failover only, very rarely
3148 Description: clients may get ASSERTION(granted_lock != NULL)
3149 Details : When request was taking a long time, and a client was resending
3150 a getattr by name lock request. The were multiple lock requests
3151 with the same client lock handle and
3152 mds_getattr_name->fixup_handle_for_resent_request found one of the
3153 lock handles but later failed with ASSERTION(granted_lock != NULL).
3158 Description: handle->h_buffer_credits > 0, assertion failure
3159 Details : h_buffer_credits is zero after truncate, causing assertion
3160 failure. This patch extends the transaction or creates a new
3164 Frequency : NFS re-export or patchless client
3165 Bugzilla : 11179, 10796
3166 Description: Crash on NFS re-export node (__d_move)
3167 Details : We do not want to hash the dentry if we don't have a lock.
3168 But if this dentry is later used in d_move, we'd hit uninitialised
3169 list head d_hash, so we just do this to init d_hash field but
3170 leave dentry unhashed.
3173 Frequency : NFS re-export or patchless client
3175 Description: NFS exports has problem with symbolic link
3176 Details : lustre client didn't properly install dentry when re-exported
3177 to NFS or running patchless client.
3180 Frequency : NFS re-export or patchless client
3182 Description: Various nfs/patchless fixes.
3183 Details : fixes reuse disconected alias for lookup process - this fixes
3184 warning "find_exported_dentry: npd != pd",
3185 fix permission error with open files at nfs.
3186 fix apply umask when do revalidate.
3189 Frequency : occasional
3191 Description: Crash on NFS re-export node
3192 Details : calling clear_page() on the wrong pointer triggered oops in
3193 generic_mapping_read().
3196 Frequency : rarely, using O_DIRECT IO
3198 Description: unaligned directio crashes client with LASSERT
3199 Details : check for unaligned buffers before trying any requests.
3202 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3204 Description: stale data returned from RAID cache
3205 Details : If only a small amount of IO is done to the RAID device before
3206 reading it again it is possible to get stale data from the RAID
3207 cache instead of reading it from disk.
3210 Frequency : always for sles10 kernel
3212 Description: sles10 support
3213 Details : ll_follow_link: compile fixes and using of nd_set_link
3214 under newer kernels.
3217 Frequency : depends on arch, kernel and compiler version, always on sles10
3220 Description: recursive or deep enough symlinks cause stack overflow
3221 Details : getting rid of large stack-allocated variable in
3225 Frequency : depends on hardware
3227 Description: lustre write performance loss in the SLES10 kernel
3228 Details : the performance loss is caused by using of write barriers in the
3229 ext3 code. The SLES10 kernel turns barrier support on by
3230 default. The fix is to undo that change for ldiskfs.
3232 ------------------------------------------------------------------------------
3234 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3236 * Support for kernels:
3237 2.6.9-42.0.3EL (RHEL 4)
3238 2.6.5-7.276 (SLES 9)
3239 2.4.21-47.0.1.EL (RHEL 3)
3240 2.6.12.6 vanilla (kernel.org)
3244 Frequency : quota enabled and large files being deleted
3246 Description: releasing more than 4GB of quota at once hangs OST
3247 Details : If a user deletes more than 4GB of files on a single OST it
3248 will cause the OST to spin in an infinite loop. Release
3249 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3254 Description: statfs data retrieved from /proc may be stale or zero
3255 Details : When reading per-device statfs data from /proc, in the
3256 {kbytes,files}_{total,free,avail} files, it may appear
3257 as zero or be out of date.
3260 Frequency : systems with MD RAID1 external journal devices
3262 Description: lconf's call to blkid is confused by RAID1 journal devices
3263 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3264 of returning all block devices that match the journal UUID.
3267 Frequency : always, for aggregate stripe size over 4GB
3269 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3270 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3271 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3274 Frequency : always if "lfs find" used on a local file/directory
3276 Description: "lfs find" segfaults if used on a local file/directory
3277 Details : The case where a directory component was not specified wasn't
3278 handled correctly. Handle this properly.
3281 Frequency : always on ppc64
3283 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3284 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3285 perform endianness conversion.
3288 Frequency : rarely (truncate to non-zero file size after write under load)
3289 Bugzilla : 10730, 10687
3290 Description: Files padded with zeros to next 4K multiple
3291 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3292 it is possible that files that are truncated to a non-zero size
3293 immediately after being written are filled with zero bytes beyond
3294 the truncated size. No file data is lost.
3296 Severity : enhancement
3298 Description: Allow recovery/failover for liblustre clients.
3299 Details : liblustre clients were unaware of failover configurations until
3302 Severity : enhancement
3304 Description: user file locks should fail when not mounting with flock option
3305 Details : Set up an error-returning stub in ll_file_operations.lock field
3306 to prevent incorrect behaviour when client is mounted without
3307 flock option. Also, set up properly f_op->flock field for
3311 Frequency : always on ia64
3313 Description: "lfs df" loops on printing out MDS statfs information
3314 Details : The obd_ioctl_data was not initialized and in some systems
3315 this caused a failure during the ioctl that did not return
3316 an error. Initialize the struct and return an error on failure.
3319 Frequency : SLES 9 only
3321 Description: Error of copying files with lustre special EAs as root
3322 Details : Client side always return success for setxattr call for lustre
3323 special xattr (currently only "trusted.lov").
3326 Frequency : rarely on clusters with both ia64+i386 clients
3328 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3329 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3330 clients are doing concurrent writes to the same file it is
3331 possible that the ia64 clients may LASSERT because the OST
3332 extent locks are not PAGE_SIZE aligned. Ensure that grown
3333 locks are always aligned on the request boundary.
3336 Frequency : specific use, occasional
3338 Description: Overwriting in use executable truncates on-disk binary image
3339 Details : If one node attempts to overwrite an executable in use by
3340 another node, we now correctly return ETXTBSY instead of
3341 truncating the file.
3343 Severity : enhancement
3345 Description: Async OSC create to avoid the blocking unnecessarily.
3346 Details : If a OST has no remain object, system will block on the creating
3347 when need to create a new object on this OST. Now, ways use
3348 pre-created objects when available, instead of blocking on an
3349 empty osc while others are not empty. If we must block, we block
3350 for the shortest possible period of time.
3355 Description: chmod on Lustre root is propagated to other clients
3356 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3357 invalid by the follow_mount time.
3360 Frequency : liblustre clients only
3362 Description: Race in 'instant cancel' lock handling could lead to such locks
3363 never to be granted in case of SMP MDS
3364 Details : Do not destroy not yet granted but cbpending locks in
3368 Frequency : replay/resend of open
3370 Description: non null lock assetion failure in mds_intent_policy
3371 Details : Trying to replay/resend lockless open requests resulted in
3372 mds_open() returning 0 with no lock. Now it sets a flag if
3373 a lock is going to be returned.
3375 Severity : enhancement
3377 Description: Checksum enhancements
3378 Details : New checksum enhancements allow for resending RPCs that failed
3381 Severity : enhancement
3383 Description: Tunables on number of dirty pages in cacche
3384 Details : Allow to set limit on number of dirty pages cached.
3389 Description: client crash on unmount - lock still has references
3390 Details : In some error handling cases it was possible to leak a lock
3391 reference on a client while accessing a file. This was not
3392 harmful to the client during operation, but would cause the
3393 client to crash when the filesystem is unmounted.
3396 Frequency : specific case, rare
3398 Description: ETXTBSY on mds though file not in use by client
3399 Details : ETXTBSY is no longer incorrectly returned when attempting to
3400 chmod or chown a directory that the user previously tried to
3401 execute or a currently-executing binary.
3404 Frequency : extremely rare except on liblustre-based clients
3406 Description: Lustre space not freed when files are deleted
3407 Details : Clean up open-unlinked files after client eviction. Previously
3408 the unlink was skipped and the files remained as orphans.
3413 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3414 Details : In some cases it was possible to send a blocking callback to
3415 a client doing a glimpse, even though that client didn't get
3416 a lock granted. When the glimpse lock is cancelled on the OST
3417 the freed lock is left on the waiting list and corrupted the list.
3420 Frequency : all core dumps
3422 Description: Broke core dumps to lustre
3423 Details : Negative dentry may be unhashed if parent does not have UPDATE
3424 lock, but some callers, e.g. do_coredump, expect dentry to be
3425 hashed after successful create, hash it in ll_create_it.
3427 ------------------------------------------------------------------------------
3429 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3431 * Support for kernels:
3432 2.6.9-42.0.2.EL (RHEL 4)
3433 2.6.5-7.276 (SLES 9)
3434 2.4.21-40.EL (RHEL 3)
3435 2.6.12.6 vanilla (kernel.org)
3439 Frequency : always on RHEL 3
3441 Description: Number of open files grows over time
3442 Details : The number of open files grows over time, whether or not
3443 Lustre is started. This was due to a filp leak introduced
3444 by one of our kernel patches.
3446 ------------------------------------------------------------------------------
3448 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3450 * Support for kernels:
3451 2.6.9-42.EL (RHEL 4)
3452 2.6.5-7.276 (SLES 9)
3453 2.4.21-40.EL (RHEL 3)
3454 2.6.12.6 vanilla (kernel.org)
3459 Bugzilla : 5719, 9635, 9792, 9684
3460 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3461 Details : If a server is under heavy load and cannot reply to new
3462 connection requests before the client resends the (re)connect,
3463 the connection handling code can behave badly if two service
3464 threads are concurrently handing separate (re)connections from
3465 the same client. Add better locking to the connection handling
3466 code, and ensure that only a single connection will be processed
3467 for a given client UUID, even if the lock is dropped.
3469 Severity : enhancement
3471 Description: add TCP zero-copy support to kernel
3472 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3473 sends if the hardware supports scatter-gather and checksumming.
3474 This allows socklnd to do client-write and server-read more
3475 efficiently and reduce CPU utilization from skbuf copying.
3478 Frequency : only if NFS exporting from client
3480 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3481 Details : If setting attributes on a file created under NFS that had
3482 never been opened it would be possible to oops the client
3483 if the file had no objects.
3486 Frequency : always for liblustre
3488 Description: liblustre client does MDS+OSTs setattr RPC for each write
3489 Details : When doing a write from a liblustre client, the client
3490 incorrectly issued an RPC to the MDS and each OST the file was
3491 striped over in order to update the timestamps. When writing
3492 with small chunks and many clients this could overwhelm the MDS
3493 with RPCs. In all cases it would slow down the write because
3494 these RPCs are unnecessary.
3496 Severity : enhancement
3498 Description: allow number of MDS service threads to be changed at module load
3499 Details : It is now possible to change the number of MDS service threads
3500 running. Adding "options mds mds_num_threads={N}" to the MDS's
3501 /etc/modprobe.conf will set the number of threads for the next
3502 time Lustre is restarted (assuming the "mds" module is also
3503 reloaded at that time). The default number of threads will
3504 stay the same, 32 for most systems.
3509 Description: OST crash if filesystem is unformatted or corrupt
3510 Details : If an OST is started on a device that has never been formatted
3511 or if the filesystem is corrupt and cannot even mount then the
3512 error handling cleanup routines would dereference a NULL pointer.
3517 Description: NULL pointer deref in llap_from_page.
3518 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3519 incorrect) mapping if the page was truncated/reclaimed while it was
3520 searched for. Check for this condition and skip such pages when
3521 doing readahead. Introduce extra check to llap_from_page() to
3522 verify page->mapping->host is non-NULL (so page is not anonymous).
3525 Frequency : Sometimes when using sys_sendfile
3527 Description: "page not covered by a lock" warnings from ll_readpage
3528 Details : sendfile called ll_readpage without right page locks present.
3529 Now we introduced ll_file_sendfile that does necessary locking
3530 around call to generic_file_sendfile() much like we do in
3534 Frequency : with certain MDS communication failures at client mount time
3536 Description: NULL pointer deref after failed client mount
3537 Details : a client connection request may delayed by the network layer
3538 and not be sent until after the PTLRPC layer has timed out the
3539 request. If the client fails the mount immediately it will try
3540 to clean up before the network times out the request. Add a
3541 reference from the request import to the obd device and delay
3542 the cleanup until the network drops the request.
3545 Frequency : occasionally during client (re)connect
3547 Description: assertion failure during client (re)connect
3548 Details : processing a client connection request may be delayed by the
3549 client or server longer than the client connect timeout. This
3550 causes the client to resend the connection request. If the
3551 original connection request is replied in this interval, the
3552 client may trip an assertion failure in ptlrpc_connect_interpret()
3553 which thought it would be the only running connect process.
3556 Frequency : only with obd_echo servers and clients that are rebooted
3558 Description: kernel BUG accessing uninitialized data structure
3559 Details : When running an obd_echo server it did not start the ping_evictor
3560 thread, and when a client was evicted an uninitialized data
3561 structure was accessed. Start the ping_evictor in the RPC
3562 service startup instead of the OBD startup.
3564 Severity : enhancement
3565 Bugzilla : 10193 (patchless)
3566 Description: Remove dependency on various unexported kernel interfaces.
3567 Details : No longer need reparent_to_init, exit_mm, exit_files,
3568 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3571 Frequency : rare (only users of deprecated and unsupported LDAP config)
3573 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3574 Details : LDAP apparently contains 'lustreName' attributes instead of
3575 'name'. A simple remapping of the name is sufficient.
3578 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3580 Description: waiting_locks_callback trips kernel BUG if client is evicted
3581 Details : Running with the dump_on_timeout debug flag turned on makes
3582 it possible that the waiting_locks_callback() can try to dump
3583 the Lustre kernel debug logs from an interrupt handler. Defer
3584 this log dumping to the expired_lock_main() thread.
3586 Severity : enhancement
3588 Description: Support NFS exporting on 2.6 kernels.
3589 Details : Implement non-rawops metadata methods for NFS server to use without
3590 changing NFS server code.
3593 Frequency : very rare (synthetic metadata workload only)
3595 Description: two racing renames might cause an MDS thread to deadlock
3596 Details : Running the "racer" program may cause one MDS thread to rename
3597 a file from being the source of a rename to being the target of
3598 a rename at exactly the same time that another thread is doing
3599 so, and the second thread has already enqueued these locks after
3600 doing a lookup of the target and is trying to relock them in
3601 order. Ensure that we don't try to re-lock the same resource.
3604 Frequency : only very large systems with liblustre clients
3606 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3607 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3608 all OSTs in parallel. This allows the eviction of stale liblustre
3609 clients to proceed much faster than if they were done in series,
3610 and also offers similar improvements for other set_info RPCs.
3615 Description: excessive CPU usage during initial read phase on client
3616 Details : During the initial read phase on a client, it would agressively
3617 retry readahead on the file, consuming too much CPU and impacting
3618 performance (since 1.4.5.8). Improve the readahead algorithm
3619 to avoid this, and also improve some other common cases (read
3620 of small files in particular, where "small" is files smaller than
3621 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3626 Description: MDS crash when receiving packet with unknown intent.
3627 Details : Do not LBUG in unknown intent case, just return -EFAULT
3629 Severity : enhancement
3630 Bugzilla : 9293, 9385
3631 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3632 Details : Do not serialize getattr (non-intent version) and statfs.
3635 Frequency : occasional, when OST network is overloaded/intermittent
3637 Description: client evicted by OST after bulk IO timeout
3638 Details : If a client sends a bulk IO request (read or write) the OST
3639 may evict the client if it is unresposive to its data GET/PUT
3640 request. This is incorrect if the network is overloaded (takes
3641 too long to transfer the RPC data) or dropped the OST GET/PUT
3642 request. There is no need to evict the client at all, since
3643 the pinger and/or lock callbacks will handle this, and the
3644 client can restart the bulk request.
3647 Frequency : Always when mmapping file with no objects
3649 Description: client crashes when mmapping file with no objects
3650 Details : Check that we actually have objects in a file before doing any
3651 operations on objects in ll_vm_open, ll_vm_close and
3657 Description: Request leak when working with deleted CWD
3658 Details : Introduce advanced request refcount tracking for requests
3659 referenced from lustre intent.
3661 Severity : Enhancement
3663 Description: Cache open file handles on client.
3664 Details : MDS now will return special lock along with openhandle, if
3665 requested and client is allowed to hold openhandle, even if unused,
3666 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3667 closing files for every read/write openration.
3669 Severity : Enhancement
3671 Description: Cache open negative dentries on client when possible.
3672 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3673 negative dentries on lock revocation.
3678 Description: Remounting a client read-only wasn't possible with a zconf mount
3679 Details : It wasn't possible to remount a client read-only with llmount.
3681 Severity : enhancement
3682 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3683 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3684 MPICH is included to add Lustre-specific ADIO interfaces.
3685 This is based closely on the UFS ADIO layer and only differs
3686 in file creation, in order to allow the OST striping to be set.
3687 This is user-contributed code and not supported by CFS.
3692 Description: extended inode attributes (immutable, append-only) work improperly
3693 when 2.4 and 2.6 kernels are used on client/server or vice versa
3694 Details : Introduce kernel-independent values for these flags.
3696 Severity : enhancement
3699 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3700 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3701 granularity. Now, Fractional MB granularity is be supported,
3702 this is very useful for low memory system.
3704 Severity : enhancement
3706 Description: Getattr by fid
3707 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3708 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3709 have locks and avoids getattr rpc requests for it on every stat.
3712 Frequency : Always, for filesystems larger than 2TB
3714 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3715 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3716 filesystems larger than 2TB to be mounted with mballoc enabled.
3719 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3721 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3722 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3723 filesystems larger than 2TB is mounted then the kernel will
3724 silently corrupt the start of the filesystem. CONFIG_LBD is
3725 enabled for all CFS-supported kernels, but the possibility of
3726 this happening with a modified kernel config exists.
3728 Severity : enhancement
3730 Description: add client O_DIRECT support for 2.6 kernels
3731 Details : It is now possible to do O_DIRECT reads and writes to files
3732 in the Lustre client mountpoint on 2.6 kernel clients.
3734 Severity : enhancement
3736 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3737 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3738 parallel, not waiting for response from every OST before sending
3739 a rpc to the next OST.
3744 Description: setattr vs write race when updating file timestamps
3745 Details : Client processes that update a file timestamp into the past
3746 right after writing to the file (e.g. tar) it is possible that
3747 the updated file modification time can be reset to the current
3748 time due to a race between processing the setattr and write RPC.
3750 Severity : enhancement
3752 Description: Bring 'lfs find' closer in line with regular Linux find.
3753 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3754 -print0 options and obtains all the needed info through the lustre
3757 Severity : enhancement
3759 Description: support up to 1024 configured devices on one node
3760 Details : change obd_dev array from statically allocated to dynamically
3761 allocated structs as they are first used to reduce memory usage
3766 Description: Flush dirty partially truncated pages during truncate
3767 Details : Immediatelly flush partially truncated pages in filter_setattr,
3768 this way we completely avoid having any pages in page cache on OST
3769 and can retire ugly workarounds during writes to flush such pages.
3774 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3775 Details : protect inode from truncation within vfs_unlink() context
3776 just take a reference before calling vfs_unlink() and release it
3777 when parent's i_sem is free.
3782 Description: last_id value checked outside lock on OST caused LASSERT failure
3783 Details : If there were multiple MDS->OST object precreate requests in
3784 flight, it was possible that the OST's last object id was checked
3785 outside a lock and incorrectly tripped an assertion. Move checks
3786 inside locks, and discard old precreate requests.
3789 Frequency : always, if extents are used on OSTs
3791 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3792 Details : OSTs using the extents format would not zero the high 16 bits of
3793 the index physical block number. This is not a problem for any
3794 OST filesystems smaller than 16TB, and no kernels support ext3
3795 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3796 new/modified files) and can be fixed for existing filesystems
3797 with e2fsprogs-1.39-cfs1.
3802 Description: import connection selection may be incorrect if timer wraps
3803 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3804 import connections to be ignored if the 32-bit jiffies counter
3805 wraps. Use a 64-bit jiffies counter.
3808 Frequency : very large clusters immediately after boot
3810 Description: LNET request buffers exhausted under heavy short-term load
3811 Details : If a large number of client requests are generated on a service
3812 that has previously never seen so many requests it is possible
3813 that the request buffer growth cannot keep up with the spike in
3814 demand. Instead of dropping incoming requests, they are held in
3815 the LND until the RPC service can accept more requests.
3818 Frequency : Sometimes during replay
3820 Description: Assertion failure in ll_local_open after replay.
3821 Details : If replay happened on an open request reply before we were able
3822 to set replay handler, reply will become not swabbed tripping the
3823 assertion in ll_local_open. Now we set the handler right after
3824 recognising of open request
3827 Frequency : very rare
3829 Description: kernel reports "badness in vsnprintf"
3830 Details : Reading from the "recovery_status" /proc file in small chunks
3831 may cause a negative length in lprocfs_obd_rd_recovery_status()
3832 call to vsnprintf() (which is otherwise harmless). Exit early
3833 if there is no more space in the output buffer.
3835 Severity : enhancement
3837 Description: clear OBD RPC statistics by writing to them
3838 Details : It is now possible to clear the OBD RPC statistics by writing
3839 to the "stats" file.
3844 Description: Client mtime is not the same on different clients after utimes
3845 Details : In some cases, the client was using the utimes() syscall on
3846 a file cached on another node. The clients now validate the
3847 ctime from the MDS + OSTs to determine which one is right.
3852 Description: Inability to activate failout mode
3853 Details : lconf script incorrectly assumed that in pythong string's numeric
3854 value is used in comparisons.
3857 Frequency : always with multiple stripes per file
3859 Description: Inefficient object allocation for mutli-stripe files
3860 Details : When selecting which OSTs to stripe files over, for files with
3861 a stripe count that divides evenly into the number of OSTs,
3862 the MDS is always picking the same starting OST for each file.
3863 Return the OST selection heuristic to the original design.
3868 Description: mount failures may take full timeout to return an error
3869 Details : Under some heavy load conditions it is possible that a
3870 failed mount can wait for the full obd_timeout interval,
3871 possibly several minutes, before reporting an error.
3872 Instead return an error as soon as the status is known.
3874 Frequency : quota enabled and large files being deleted
3876 Description: releasing more than 4GB of quota at once hangs OST
3877 Details : If a user deletes more than 4GB of files on a single OST it
3878 will cause the OST to spin in an infinite loop. Release
3879 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3884 Description: statfs data retrieved from /proc may be stale or zero
3885 Details : When reading per-device statfs data from /proc, in the
3886 {kbytes,files}_{total,free,avail} files, it may appear
3887 as zero or be out of date.
3890 Frequency : systems with MD RAID1 external journal devices
3892 Description: lconf's call to blkid is confused by RAID1 journal devices
3893 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3894 of returning all block devices that match the journal UUID.
3897 Frequency : always, for aggregate stripe size over 4GB
3899 Description: assertion fails when trying to use 4GB stripe size
3900 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3901 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3904 Frequency : always on ppc64
3906 Description: the first write on an ext3 filesystem with mballoc got stuck
3907 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3908 perform endianness conversion.
3910 ------------------------------------------------------------------------------
3912 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3914 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3915 INTEROPERATE with older versions automatically. Please read the
3916 user documentation before upgrading any part of a live system.
3917 * WARNING: Lustre networking configuration changes are required with
3918 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3921 * Support for kernels:
3922 2.6.9-22.0.2.EL (RHEL 4)
3923 2.6.5-7.244 (SLES 9)
3924 2.6.12.6 vanilla (kernel.org)
3927 Severity : enhancement
3928 Bugzilla : 7981/8208
3929 Description: Introduced Lustre Networking (LNET)
3930 Details : LNET is new networking infrastructure for Lustre, it includes
3931 a reorganized network configuration mode (see the user
3932 documentation for full details) as well as support for routing
3933 between different network fabrics. Lustre Networking Devices
3934 (LNDs) for the supported network fabrics have also been
3935 created for this new infrastructure.
3937 Severity : enhancement
3938 Description: Introduced Access control lists
3939 Details : clients can set ACLs on files and directories in order to have
3940 more fine-grained permissions than the standard Unix UGO+RWX.
3941 The MDS must be started with the "-o acl" mount option.
3943 Severity : enhancement
3944 Description: Introduced filesystem quotas
3945 Details : Administrators may now establish per-user quotas on the
3948 Severity : enhancement
3950 Description: Configuration change for the XT3
3951 The PTLLND is now used to run Lustre over Portals on the XT3
3952 The configure option(s) --with-cray-portals are no longer used.
3953 Rather --with-portals=<path-to-portals-includes> is used to
3954 enable building on the XT3. In addition to enable XT3 specific
3955 features the option --enable-cray-xt3 must be used.
3960 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3961 Details : A race between allocating a new llcd and re-getting the llcd_lock
3962 allowed another thread to grab newly-allocated llcd.
3964 Severity : enhancement
3966 Description: 2.6 OST async journal commit and locking fix to improve performance
3967 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3968 2.6 kernels are now async as they already were in 2.4 kernels so
3969 that they can commit concurrently with the network bulk transfer.
3970 For block-allocated files the filter allocation semaphore is held
3971 to avoid filesystem fragmentation during allocation. BKL lock
3972 removed for 2.6 xattr operations where it is no longer needed.
3977 Description: lconf incorrectly determined whether two IP networks could talk
3978 Details : In some more complicated routing and multiple-network
3979 configurations, lconf will avoid trying to make a network
3980 connection to a disjoint part of the IP space. It was doing the
3981 math incorrectly for one set of cases.
3986 Description: Fix for potential infinite loop processing records in an llog.
3987 Details : If an llog record is corrupted/zeroed, it is possible to loop
3988 forever in llog_process(). Validate the llog record length
3989 and skip the remainder of the block on error.
3992 Frequency : occasional (liblustre only)
3994 Description: liblustre could not open files whose last component is a symlink
3995 Details : sysio_path_walk() would incorrectly pass the open intent to
3996 intermediate path components.
3999 Frequency : rare (liblustre only with non-standard tuning)
4000 Bugzilla : 7201 (7350)
4001 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4002 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4003 after it has been released, only do this for referenced locks
4005 Severity : enhancement
4007 Description: specifying an (invalid) directory default stripe_size of -1
4008 would reset the directory default striping
4009 Details : stripe_size -1 was used internally to signal directory stripe
4010 removal, now use "all default" to signal dir stripe removal
4011 as a directory striping of "all default" is not useful
4014 Frequency : common for large clusters running liblustre clients
4016 Description: doing an ls when liblustre clients are running is slow
4017 Details : sending a glimpse AST to a liblustre client waits for every AST
4018 to time out, as liblustre clients will not respond. Since they
4019 cannot cache data we refresh the OST lock LVB from disk instead.
4021 Severity : enhancement
4023 Description: doing an ls at the same time as file IO can be slow
4024 Details : enqueue and other "small" requests can be blocked behind many
4025 large IO requests. Create a new OST IO portal for non-IO
4026 requests so they can be processed faster.
4029 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4031 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4032 created file with mode 000
4033 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4034 allow an ftruncate() as a non-root user to a file with mode 000.
4035 Lustre now respects this flag to disable mode checks when
4036 truncating a file owned by the user
4039 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4042 Description: Revoking locks from clients that went dead or catatonic might take
4044 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4045 cancellation of such locks instant on servers without waiting for
4046 any reply from clients. Clients drops these locks when cancel
4047 notification from server is received without replying.
4050 Frequency : liblustre-only, when liblustre client dies or becomes busy
4052 Description: Doing ls on Linux clients can take a long time with active
4054 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4055 granting such locks to it in the first place if possible. Locks
4056 are taken by proxy on the OST during the read or write and
4057 dropped immediately afterward. Add connect flags handling, do
4058 not grant locks to liblustre clients for glimpse ASTs.
4060 Severity : enhancement
4062 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4063 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4064 sequential reads and behaves badly with applications doing random
4065 reads. Improve it by reading ahead at least read region, and
4066 avoiding excessive large RPC for small reads.
4068 Severity : enhancement
4070 Description: Creating more than 1000 files for a single job may cause a load
4071 imbalance on the OSTs if there are also a large number of OSTs.
4072 Details : qos_prep_create() uses an OST index reseed value that is an
4073 even multiple of the number of available OSTs so that if the
4074 reseed happens in the middle of the object allocation it will
4075 still utilize the OSTs as uniformly as possible.
4080 Description: OST or MDS may oops in ping_evictor_main()
4081 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4082 but doesn't restart at beginning of obd_exports_timed list
4085 Severity : enhancement
4087 Description: improve by-nid export eviction on the MDS and OST
4088 Details : allow multiple exports with the same NID to be evicted at one
4089 time without re-searching the exports list.
4092 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4094 Description: MDS may oops in groups_free()
4095 Details : in rare race conditions a newly allocated group_info struct is
4096 freed again, and this can be NULL. The 2.4 compatibility code
4097 for groups_free() checked for a NULL pointer, but 2.6 did not.
4100 Frequency : common for liblustre clients doing little filesystem IO
4101 Bugzilla : 9352, 7313
4102 Description: server may evict liblustre clients accessing contended locks
4103 Details : if a client is granted a lock or receives a completion AST
4104 with a blocking AST already set it would not reply to the AST
4105 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4108 Frequency : lfs setstripe, only systems with more than 160 OSTs
4110 Description: unable to set striping with a starting offset beyond OST 160
4111 Details : llapi_create_file() incorrectly limited the starting stripe
4112 index to the maximum single-file stripe count.
4115 Frequency : LDAP users only
4117 Description: lconf did not handle in-kernel recovery with LDAP properly
4118 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4120 Severity : enhancement
4122 Description: bind OST threads to NUMA nodes to improve performance
4123 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4124 node and do their allocations there to localize memory access
4126 Severity : enhancement
4128 Description: llmount can determine client NID directly from Myrinet (GM)
4129 Details : the client NID code from gmnalnid was moved directly into
4130 llmount, removing the need to use this or specifying the
4131 client NID explicitly when mounting GM clients with zeroconf
4134 Frequency : if client is started with down MDS
4136 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4137 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4138 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4139 ptlrpc_import_delayed_req() because we didn't check if the
4140 request was interrupted, and we also didn't break out of the
4141 event loop if there was no timeout
4146 Description: data loss during non-page-aligned writes to a single file from
4147 both multiple nodes and multiple threads on one node at same time
4148 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4149 inconsistency led to false short-reads, that were cached and later
4150 used by ->prepare_write() to fill in partially written page,
4151 leading to data loss.
4154 Frequency : always, if lconf --abort_recovery used
4156 Description: lconf --abort_recovery fails with 'Operation not supported'
4157 Details : lconf was attempting to abort recovery on the MDT device and not
4160 Severity : enhancement
4162 Description: remove cleanup logs
4163 Details : replace lconf-generated cleanup logs with lustre internal
4164 cleanup routines. Eliminates the need for client-cleanup and
4167 Severity : enhancement
4169 Description: add support for EAs (user and system) on lustre filesystems
4170 Details : it is now possible to store extended attributes in the Lustre
4171 client filesystem, and with the user_xattr mount option it
4172 is possible to allow users to store EAs on their files also
4174 Severity : enhancement
4176 Description: Add possibility (config option) to show minimal available OST free
4178 Details : When compiled with --enable-mindf configure option, statfs(2)
4179 (and so, df) will return least minimal free space available from
4180 all OSTs as amount of free space on FS, instead of summary of
4181 free spaces of all OSTs.
4183 Severity : enhancement
4185 Description: do not expand extent locks acquired on OST-side
4186 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4187 by server: they are not cached anyway.
4190 Frequency : when mmap is used/binaries executed from Lustre
4192 Description: Unmmap pages before throwing them away from read cache.
4193 Details : llap_shrink cache now attempts to unmap pages before discarding
4194 them (if unmapping failed - do not discard). SLES9 kernel has
4195 extra checks that trigger if this unmapping is not done first.
4200 Description: lconf didn't resolve symlinks before checking to see whether a
4201 given mountpoint was already in use
4204 Frequency : when migrating failover services
4205 Bugzilla : 6395, 9514
4206 Description: When migrating a subset of services from a node (e.g. failback
4207 from a failover service node) the remaining services would
4208 time out and evict clients.
4209 Details : lconf --force (implied by --failover) sets the global obd_timeout
4210 to 5 seconds in order to quickly disconnect, but this caused
4211 other RPCs to time out too quickly. Do not change the global
4212 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4214 Severity : enhancement
4215 Frequency : if MDS is started with down OST
4216 Bugzilla : 9439,5706
4217 Description: Allow startup/shutdown of an MDS without depending on the
4218 availability of the OSTs.
4219 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4220 Add appropriate locking and lov-osc refcounts for safe
4221 cleaning. Add osc abort_inflight calls in case the
4222 synchronize never started.
4225 Frequency : occasional (Cray XT3 only)
4227 Description: root not authorized to access files in CRAY_PORTALS environment
4228 Details : The client process capabilities were not honoured on the MDS in
4229 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4230 been accessed by an authorized user then root was able to access
4231 the file on the local client also. The root user capabilities
4232 are now allowed on the MDS, as this environment has secure UID.
4235 Frequency : occasional
4237 Description: ldiskfs "too long searching" message happens too often
4238 Details : A debugging message (otherwise harmless) prints too often on
4239 the OST console. This has been reduced to only happen when
4240 there are fragmentation problems on the filesystem.
4245 Description: Division by zero in statfs when all OSCs are inactive
4246 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4247 lov_statfs divides by value returned by lov_get_stripecnt().
4251 Bugzilla : 9489, 3273
4252 Description: First write from each client to each OST was only 4kB in size,
4253 to initialize client writeback cache, which caused sub-optimal
4254 RPCs and poor layout on disk for the first writen file.
4255 Details : Clients now request an initial cache grant at (re)connect time
4256 and so that they can start streaming writes to the cache right
4257 away and always do full-sized RPCs if there is enough data.
4258 If the OST is rebooted the client also re-establishes its grant
4259 so that client cached writes will be honoured under the grant.
4264 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4265 Details : Now I/O RPCs go to different portal number and (presumably) fast
4266 lock requests (and glimses) and other RPCs get their own service
4267 threads pool that should be able to service those RPCs
4270 Severity : enhancement
4272 Description: Ability to exchange lustre version between client and servers and
4273 issue warnings at client side if client is too old. Also for
4274 liblustre clients there is ability to refuse connection of too old
4276 Details : New 'version' field is added to connect data structure that is
4277 filled with version info. That info is later checked by server and
4281 Frequency : rare, liblustre only.
4282 Bugzilla : 9296, 9581
4283 Description: Two simultaneous writes from liblustre at offset within same page
4284 might proceed at the same time overwriting eachother with stale
4286 Details : I/O lock withing llu_file_prwv was released too early, before data
4287 actually was hitting the wire. Extended lock-holding time until
4288 server acknowledges receiving data.
4291 Frequency : extremely rare. Never observed in practice.
4293 Description: avoid generating lustre_handle cookie of 0.
4294 Details : class_handle_hash() generates handle cookies by incrementing
4295 global counter, and can hit 0 occasionaly (this is unlikely, but
4296 not impossible, because initial value of cookie counter is
4297 selected randonly). Value of 0 is used as a sentinel meaning
4298 "unassigned handle" --- avoid it. Also coalesce two critical
4299 sections in this function into one.
4301 Severity : enhancement
4303 Description: allow liblustre clients to delegate truncate locking to OST
4304 Details : To avoid overhead of locking, liblustre client instructs OST to
4305 take extent lock in ost_punch() on client's behalf. New connection
4306 flag is added to handle backward compatibility.
4308 Severity : enhancement
4309 Bugzilla : 4928, 7341, 9758
4310 Description: allow number of OST service threads to be specified
4311 Details : a module parameter allows the number of OST service threads
4312 to be specified via "options ost ost_num_threads={N}" in the
4313 OSS's /etc/modules.conf or /etc/modprobe.conf.
4317 Bugzilla : 6146, 9635, 9895
4318 Description: servers crash with bad pointer in target_handle_connect()
4319 Details : In rare cases when a client is reconnecting it was possible that
4320 the connection request was the last reference for that export.
4321 We would temporarily drop the export reference and get a new
4322 one, but this may have been the last reference and the export
4323 was just destroyed. Get new reference before dropping old one.
4325 Severity : enhancement
4326 Frequency : if client is started with failover MDS
4328 Description: Allow multiple MDS hostnames in the mount command
4329 Details : Try to read the configuration from all specified MDS
4330 hostnames during a client mount in case the "primary"
4333 Severity : enhancement
4335 Description: Stop sending data to evicted clients as soon as possible.
4336 Details : Check if the client we are about to send or are sending data to
4337 was evicted already. (Check is done every second of waiting,
4338 for which l_wait_event interface was extended to allow checking
4339 of exit condition at specified intervals).
4342 Frequency : rare, normally only when NFS exporting is done from client
4344 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4346 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4349 Frequency : always, if srand() is called before liblustre initialization
4351 Description: Liblustre uses system PRNG disturbing its usage by user application
4352 Details : Introduce internal to lustre fast and high-quality PRNG for
4353 lustre usage and make liblustre and some other places in generic
4354 lustre code to use it.
4356 Severity : enhancement
4357 Bugzilla : 9477, 9557, 9870
4358 Description: Verify that the MDS configuration logs are updated when xml is
4359 Details : Check if the .xml configuration logs are newer than the config
4360 logs stored on the MDS and report an error if this is the case.
4361 Request --write-conf, or allow starting with --old_conf.
4363 Severity : enhancement
4365 Description: Handle symlinks in the path when checking if Lustre is mounted.
4366 Details : Resolve intermediate symlinks when checking if a client has
4367 mounted a filesystem to avoid duplicate client mounts.
4372 Description: lconf can hit an error exception but still return success.
4373 Details : The lconf command catches the Command error exception at the top
4374 level script context and will exit with the associated exit
4375 status, but doesn't ensure that this exit status is non-zero.
4380 Description: failure of ptlrpc thread startup can cause oops
4381 Details : Starting a ptlrpc service thread can fail if there are a large
4382 number of threads or the server memory is very fragmented.
4383 Handle this without oopsing.
4386 Frequency : always, only if liblustre and non-default acceptor port was used
4388 Description: liblustre cannot connect to servers with non-default acceptor port
4389 Details : tcpnal_set_default_params() was not called and was therefore
4390 ignoring the environment varaible TCPNAL_PORT, as well as other
4391 TCPNAL_ environment variables
4396 Description: two objects could be created on the same OST for a single file
4397 Details : If an OST is down, in some cases it was possible to create two
4398 objects on a single OST for a single file. No problems other
4399 than potential performance impact and spurious error messages.
4403 Bugzilla : 5681, 9562
4404 Description: Client may oops in ll_unhash_aliases
4405 Details : Client dcache may become inconsistent in race condition.
4406 In some cases "getcwd" can fail if the current directory is
4412 Description: Inode refcounting problems in NFS export code
4413 Details : link_raw functions used to call d_instantiate without obtaining
4414 extra inode reference first.
4418 Bugzilla : 9942, 9903
4419 Description: Referencing freed requests leading to crash, memleaks with NFS.
4420 Details : We used to require that call to ll_revalidate_it was always
4421 followed by ll_lookup_it. Also with revalidate_special() it is
4422 possible to call ll_revalidate_it() twice for the same dentry
4423 even if first occurence returned success. This fix changes semantic
4424 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4425 reference on a request referred from the intent.
4426 ll_intent_release() then releases such a request.
4429 Frequency : rare, normally benchmark loads only
4431 Description: unlinked inodes were kept in memory on the client
4432 Details : If a client is repeatedly creating and unlinking files it
4433 can accumulate a lot of stale inodes in the inode slab cache.
4434 If there is no other client load running this can cause the
4435 client node to run out of memory. Instead flush old inodes
4436 from client cache that have the same inode number as a new inode.
4439 Frequency : SLES9 2.6.5 kernel and long filenames only
4440 Bugzilla : 9969, 10379
4441 Description: utime reports stale NFS file handle
4442 Details : SLES9 uses out-of-dentry names in some cases, which confused
4443 the lustre dentry revalidation. Change it to always use the
4447 Frequency : rare, unless heavy write-truncate concurrency is continuous
4448 Bugzilla : 4180, 6984, 7171, 9963, 9331
4449 Description: OST becomes very slow and/or deadlocked during object unlink
4450 Details : filter_destroy() was holding onto the parent directory lock
4451 while truncating+unlinking objects. For very large objects this
4452 may block other threads for a long time and slow overall OST
4453 responsiveness. It may also be possible to get a lock ordering
4454 deadlock in this case, or run out of journal credits because of
4455 the combined truncate+unlink. Solution is to do object truncate
4456 first in one transaction without parent lock, and then do the
4457 final unlink in a new transaction with the parent lock. This
4458 reduces the lock hold time dramatically.
4461 Frequency : rare, 2.4 kernels only
4463 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4464 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4465 before unmounting their filesystem in order to avoid deadlock.
4466 The kernel_locked() function in 2.4 kernels only checks whether
4467 the kernel lock is held, not whether it is this process that is
4468 holding it as 2.6 kernels do.
4473 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4474 Details : The client ptlrpc code may be trying to reconnect to a down
4475 server before a previous connection attempt has timed out.
4476 Increase the reconnect interval to be longer than the connection
4477 timeout interval to avoid sending duplicate connections to
4481 Frequency : echo_client brw_test command
4483 Description: fix echo_client to work with OST preallocated code
4484 Details : OST preallocation code (5137) didn't take echo_client IO path
4485 into account: echo_client calls filter methods outside of any
4486 OST thread and, hence, there is no per-thread preallocated
4487 pages and buffers to use. Solution: hijack pga pages for IO. As
4488 a byproduct, this avoids unnecessary data copying.
4492 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4493 Description: Client can oops in mdc_commit_close() after open replay
4494 Details : It was possible for the MDS to return an open request with no
4495 transaction number in mds_finish_transno() if the client was
4496 evicted, but without actually returning an error. Clients
4497 would later try to replay that open and may trip an assertion
4498 Simplify the client close codepath, and always return an error
4499 from the MDS in case the open is not successful.
4502 Frequency : rare, 2.6 OSTs only
4504 Description: OST may deadlock under high load on fragmented files
4505 Details : If there was a heavy load and highly-fragmented OST filesystems
4506 it was possible to have all the OST threads deadlock waiting on
4507 allocation of biovecs, because the biovecs were not released
4508 until the entire RPC IO was completed. Instead, release biovecs
4509 as soon as they are complete to ensure forward IO progress.
4511 Severity : enhancement
4513 Description: Support for specifying external journal device at mount
4514 Details : If an OST or MDS device is formatted with an external journal
4515 device, this device major/minor is stored in the ext3 superblock
4516 and may not be valid for failover. Allow detecting and
4517 specifying the external journal at mount time.
4522 Description: Mounting an MDS with pending unlinked files may cause oops
4523 Details : target_finish_recovery() calls mds_postrecov() which returned
4524 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4525 considers this an error and immediately begins cleaning up the
4526 lov, just after starting the mds_lov process
4528 Severity : enhancement
4530 Description: Implement 'lfs df' to report actual free space on per-OST basis
4531 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4532 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4533 usage of objects; '-h' to report in human readable format.
4535 ------------------------------------------------------------------------------
4537 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4544 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4545 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4546 at startup during orphan unlinking. Instead, skip buddy update
4547 before setup, as it will be regenerated after recovery is complete.
4552 Description: If an OST is inactive, its locks might reference stale inodes.
4553 Details : lov_change_cbdata() must iterate over all namespaces, even if
4554 they are inactive to clear inode references from the lock.
4556 Severity : enhancement
4557 Frequency : occasional, if non-standard max_dirty_mb used
4559 Description: Client will block write RPCs if not enough grant
4560 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4561 then the client will block writes while waiting for another RPC
4562 to complete instead of consuming its dirty limit. With change
4563 we get improved performance when max_dirty_mb is small.
4565 Severity : enhancement
4566 Bugzilla : 3389, 6253
4567 Description: Add support for supplementary groups on the MDS.
4568 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4569 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4570 lookups for user supplementary groups into a cache.
4574 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4575 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4576 for liblustre O_EXCL is enforced by MDS.
4580 Description: readdir never returns NULL in liblustre.
4581 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4582 dentry in struct dirent.
4586 Frequency : liblustre only, depends on application IO pattern
4587 Description: liblustre clients evicted if not contacting servers
4588 Details : Don't put liblustre clients into the ping_evictor list, so
4589 they will not be evicted by the pinger ever.
4591 Severity : enhancement
4593 Description: Add ability to evict clients by NID from MDS.
4594 Details : By echoing "nid:$NID" string into
4595 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4596 $NID would be instantly evicted from this MDS and from all active
4597 OSTs connected to it.
4601 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4602 Details : lookup_it_finish() used to query file size from OSTs that was not
4607 Description: service threads change working directory to that of init
4608 Details : Starting lustre service threads may pin the working directory
4609 of the parent thread, making that filesystem busy. Threads
4610 now change to the working directory of init to avoid this.
4614 Frequency : during shutdown only
4615 Description: shutdown with a failed MDS or OST can cause unmount to hang
4616 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4621 Frequency : 2.6 only
4622 Description: chmod/chown may include an extra supplementary group
4623 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4624 supplementary group and if none is specified this is used.
4627 Bugzilla : 5479 (6816)
4629 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4630 Details : If lookup is in progress on a file that is unlinked we might try
4631 to revalidate the inode and fail in revalidate after lookup is
4632 complete and ll_file_open() enqueues the open again but
4633 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4636 Frequency : always, if lconf --abort_recovery used
4638 Description: lconf --abort_recovery fails with 'Operation not supported'
4639 Details : lconf was attempting to abort recovery on the MDT device and not
4642 ------------------------------------------------------------------------------
4644 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4649 Frequency : rare (only unsupported configurations with a node running as an
4651 Bugzilla : 6514, 5137
4652 Description: Mounting a Lustre file system on a node running as an OST could
4654 Details : OSTs now preallocates memory needed to write out data at
4655 startup, instead of when needed, to avoid having to
4656 allocate memory in possibly low memory situations.
4657 Specifically, if the file system is mounted on on OST,
4658 memory pressure could force it to try to write out data,
4659 which it needed to allocate memory to do. Due to the low
4660 memory, it would be unable to do so and the node would
4661 become unresponsive.
4663 Severity : enhancement
4665 Description: Addition of lconf --service command line option
4666 Details : lconf now accepts a '--service <arg>' option, which is
4667 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4669 Severity : enhancement
4671 Description: Failover mode is now the default for OSTs.
4672 Details : By default, OSTs will now run in failover mode. To return to
4673 the old behaviour, add '--failout' to the lmc line for OSTs.
4675 Severity : enhancement
4677 Description: Health checks are now provided for MDS and OSTs
4678 Details : Additional detailed health check information on MSD and OSTs
4679 is now provided through the procfs health_check value.
4682 Frequency : occasional, depends on IO load
4684 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4685 after numerous create/delete cycles
4686 Details : The ext3 inode allocation policy would not allocate new inodes
4687 very well on the OSTs because there are no new directories
4688 being created. Instead we look for groups with free space if
4689 the parent directories are nearly full.
4694 Description: Network or server problems during mount may cause partially
4695 mounted clients instead of returning an error.
4696 Details : The config llog parsing code may overwrite the error return
4697 code during mount error handling, returning success instead
4703 Description: MDS can fail to allocate large reply buffers
4704 Details : After long uptimes the MDS can fail to allocate large reply
4705 buffers (e.g. zconf client mount config records) due to memory
4706 fragmentation or consumption by the buffer cache. Preallocate
4707 some large reply buffers so that these replies can be sent even
4708 under memory pressure.
4712 Frequency : rare (liblustre)
4713 Description: fsx running with liblustre complained that using truncate() to
4714 extend the file doesn't work. This patch corrects that issue.
4715 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4716 ATTR_SIZE and lsm use in llu_setattr_raw.
4720 Frequency : rare, only 2.6 kernels
4721 Description: Unusual file access patterns on the MDS may result in inode
4722 data being lost in very rare circumstances.
4723 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4724 code in the 2.6 kernel caused the inode and/or EA data not to
4725 be read from disk, causing single-file corruption.
4729 Frequency : rare, only 2.6 filesystems using extents
4730 Description: Heavy concurrent write and delete load may cause data corruption.
4731 Details : It was possible under high-load situations to have an extent
4732 metadata block in the block device cache from a just-unlinked
4733 file overwrite a newly-allocated data block. We now unmap any
4734 metadata buffers that alias just-allocated data blocks.
4738 Frequency : filesystems with default stripe_count larger than 77
4739 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4740 Details : lconf specifies an inode size of 4096 bytes when the default
4741 stripe_count is larger than 77. This conflicts with the default
4742 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4743 this case to avoid pinning too much memory for large EAs.
4745 ------------------------------------------------------------------------------
4747 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4752 Frequency : rare (extremely heavy IO load with hundreds of clients)
4754 Description: Client is evicted, gets IO error writing to file
4755 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4756 caused clients to be evicted for AST timeouts. The fixes in
4757 bug 5192 mean we no longer need to have such short AST timeouts
4758 so ldlm_timeout has been increased.
4761 Frequency : occasional during --force or --failover shutdown under load
4762 Bugzilla : 5949, 4834
4763 Description: Server oops/LBUG if stopped with --force or --failover under load
4764 Details : a collection of import/export refcount and cleanup ordering
4765 issues fixed for safer force cleanup
4768 Frequency : only filesystems larger than 120 OSTs
4769 Bugzilla : 5990, 6223
4770 Description: lfs getstripe would oops on a very large filesystem
4771 Details : lov_getconfig used kfree on vmalloc'd memory
4774 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4775 Bugzilla : 6242, 6243
4776 Description: reading from files that had been truncated to a non-zero size
4777 but never opened returned no data
4778 Details : ll_file_read() reads zeros from no-object files to EOF
4783 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4784 Details : There's little harm in aborting MDS/OSS recovery and letting it
4785 try again, so I removed the LASSERT and return an error instead.
4787 Severity : enhancement
4789 Description: New debugging infrastructure for tracking down data corruption
4790 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4791 (b) cover more of the client-side code path, and (c) try to narrow
4792 down where problems occurred
4796 Bugzilla : 3819, 4364, 4397, 6313
4797 Description: Racing close and eviction MDS could cause assertion in mds_close
4798 Details : It was possible to get multiple mfd references during close and
4799 client eviction, leading to one thread referencing a freed mfd.
4801 Severity: : enhancement
4802 Bugzilla : 3262, 6359
4803 Description: Attempts to reconnect to servers are now more aggressive.
4804 Details : This builds on the enhanced upcall-less recovery that was added
4805 in 1.4.2. When trying to reconnect to servers, clients will
4806 now try each server in the failover group every 10 seconds. By
4807 default, clients would previously try one server every 25 seconds.
4812 Description: After recovery, certain operations trigger a failed
4813 assertion on a client.
4814 Details : Failing over an mds, using lconf -d --failover, while a
4815 client was doing a readdir() call would cause the client to
4816 LBUG after recovery completed and the readdir() was resent.
4818 Severity : enhancement
4820 Description: Default groups are now added by lconf
4821 Details : You can now run lconf --group <servicename> without having to
4822 manually add groups with lmc.
4825 Frequency : occasional
4827 Description: Nodes with an elan id of 0 trigger a failed assertion
4830 Frequency : always when accessing e.g. tty/console device nodes
4832 Description: tty and some other devices nodes cannot be used on lustre
4833 Details : file's private_data field is used by device data and lustre
4834 values in there got lost. New field was added to struct file to
4835 store fs-specific private data.
4838 Frequency : when exporting Lustre via NFS
4840 Description: NFSD failed occasionally when looking up a path component
4841 Details : NFSD is looking up ".." which was broken in ext3 directories
4842 that had grown large enough to become hashed.
4845 Frequency : Clusters with multiple interfaces not on the same subnet
4847 Description: Nodes will repeatedly try to reconnect to an interface which it
4848 cannot reach and report an error to the log.
4849 Details : Extra peer list entries will be created by lconf with some peers
4850 unreachable. lconf now validates the peer before adding it.
4853 Frequency : Only if a default stripe is set on the filesystem root.
4855 Description: Setting a default stripe on the filesystem root prevented the
4856 filesystem from being remounted.
4857 Details : The client was sending extra request flags in the root getattr
4858 request and did not allocate a reply buffer for the dir EA.
4861 Frequency : occasional, higher if lots of files are accessed by one client
4862 Bugzilla : 6159, 6097
4863 Description: Client trips assertion regarding lsm mismatch/magic
4864 Details : While revalidating inodes the VFS looks up inodes with ifind()
4865 and in rare cases can find an inode that is being freed.
4866 The ll_test_inode() code will free the lsm during ifind()
4867 when it finds an existing inode and then the VFS later attaches
4868 this free lsm to a new inode.
4872 Bugzilla : 6422, 7030
4873 Description: MDS deadlock between mkdir and client eviction
4874 Details : Creating a new file via mkdir or mknod (starting a transaction
4875 and getting the ns lock) can deadlock with client eviction
4876 (gets ns lock and trying to finish a synchronous transaction).
4879 Frequency : occasional
4880 Description: While starting a server, the fsfilt_ext3 module could not be
4882 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4883 kernels. Previously, lconf would still use the ext3 name
4884 when trying to load modules. Now, it will correctly use
4885 ext3 on 2.4 and ldiskfs on 2.6.
4887 Severity : enhancement
4888 Description: The default stripe count has been changed to 1
4889 Details : The interpretation of the default stripe count (0, to lfs
4890 or lmc) has been changed to mean striping across a single
4891 OST, rather than all available. For general usage we have
4892 found a stripe count of 1 or 2 works best.
4894 Severity : enhancement
4895 Description: Add support for compiling against Cray portals.
4896 Details : Conditional compiling for some areas that are different
4900 Frequency : occasional
4901 Bugzilla : 6409, 6834
4902 Description: Creating files with an explicit stripe count may lead to
4903 a failed assertion on the MDS
4904 Details : If some OSTs are full or unavailable, creating files may
4905 trigger a failed assertion on the MDS. Now, Lustre will
4906 try to use other servers or return an error to the
4910 Frequency : occasional
4912 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4913 caused lock timeout during readahead (since 1.4.2).
4914 Details : Processes doing readahead might match a lock that hasn't been
4915 granted yet if there are overlapping and conflicting lock
4916 requests. The readahead process waits on ungranted lock
4917 (original lock is CBPENDING), while OST waits for that process
4918 to cancel CBPENDING read lock and eventually evicts client.
4920 Severity : enhancement
4922 Description: Initial enabling of flock support for clients
4923 Details : Implements fcntl advisory locking and file status functions.
4924 This feature is provided as an optional mount flag (default
4925 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4926 locking are implemented yet, and those that are are not guaranteed
4927 to be completely correct in production environments.
4928 mount -t lustre -o [flock|noflock] ...
4931 Frequency : occasional
4933 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4934 assertion in the ext3 JBD (journaling) layer.
4935 Details : The b_committed_data struct is protected by the big kernel lock
4936 in 2.4 kernels, serializing journal_commit_transaction() and
4937 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4938 access to this struct. In 2.6 kernels there is finer grained
4939 locking to improve SMP performance of the JBD layer.
4943 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4948 Description: merge and cleanup kernel patches.
4951 -----------------------------------------------------------------------------
4953 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4955 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4956 versions of Lustre. Please update all servers and clients to
4957 version 1.4.2 or later at the same time. You must also run
4958 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4959 to update the configuration logs.
4961 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4962 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4963 - swab llog records of type '0' so we get proper header size/idx (5861)
4964 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4965 - fix rename of one directory over another leaking an inode (5953)
4966 - avoid SetPageDirty on 2.6 (5981)
4967 - don't re-add just-being-destroyed locks to the waiting list (5653)
4968 - when creating new directories, inherit the parent's custom
4969 striping settings if present parent (3048)
4970 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4971 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4972 - don't hold client locks on temporary worklist from l_lru (5666)
4973 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4974 - automatically evict dead clients (5921)
4975 - Update file size properly in create+truncate+fstat case (6196)
4976 - Do not unhash mountpoint dentries, do not allow removal of
4978 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4979 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4980 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4981 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4982 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4983 - Correctly handle reads of files with no objects (6243)
4984 - lctl recover will also mark a device active if deactivate used (5933)
4986 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4987 - allow --write-conf on an MDS with different nettype than client (5619)
4988 - don't write config llogs to MDS for mounts not from that MDS (5617)
4989 - lconf should create multiple TCP connections from a client (5201)
4990 - init scripts are now turned off by default; run chkconfig --on
4991 lustre and chkconfig --on lustrefs to use them
4992 - upcalls are no longer needed for clients to recover to failover
4994 - add --abort-recovery option to lconf to abort recovery on device
4996 - add support for an arbitrary number of OSTs (3026)
4997 - Quota support protocol changes.
4998 - forward compatibility changes to wire structs (6007)
4999 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5000 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5001 - improved "lustre status" script
5002 - initialize blocksize for non-regular files (6062)
5003 - added --disable-server and --disable-client configure options (5782)
5004 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5005 - Vanilla 2.4.29 support
5006 - increase maximum number of obd devices to 520 (6242)
5007 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5008 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5009 SLES 9 2.6.5 kernels
5010 - limit stripes per file to 160 (the maximum EA size) (6093)
5012 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5015 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5016 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5017 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5018 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5019 - small ext3 extents cleanups and fixes (5733)
5020 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5021 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5022 - avoid needless client->OST connect, fix handle mismatch (5317)
5023 - fix DLM error path that led to out-of-sync client, long delays (5779)
5024 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5025 - fix several locking issues related to i_size (5492,5624,5654,5672)
5026 - don't move pending lock onto export if it is already evicted (5683)
5027 - fix kernel oops when creating .foo in unlinked directory (5548)
5028 - fix deadlock in obdfilter statistics vs. object create (5811)
5029 - use time_{before,after} to avoid timer jiffies wrap (5882)
5030 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5031 - Do not leak request if server was not able to process it (5154)
5032 - If mds_open unable to find parent dir, make that negative lookup(5154)
5033 - don't create new directories with extent-mapping (5909, 5936)
5035 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5036 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5037 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5038 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5039 - support for RHEL4 kernel on i686 (5773)
5040 - provide error messages when incompatible logs are encountered (5898)
5042 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5043 * version 1.4.0.10 (1.4.1 release candidate 1)
5045 - don't keep a lock reference when lock is not granted (4238)
5046 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5047 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5048 - drop import inflight refcount on signal_completed_replay error (5255)
5049 - unlock page after async write error during send (3677)
5050 - handle missing objects in filter_preprw_read properly (5265)
5051 - no transno return for symlink open, don't save no-trasno open (3440)
5052 - don't try to complete elan receive that already failed (4012)
5053 - free RPC server reply state on error (5406)
5054 - clean up thread from ptlrpc_start_thread() on error (5160)
5055 - readahead could read extra page into cache that wasn't ejected (5388)
5056 - prevent races in class_attach/setup/cleanup/detach (5260)
5057 - don't dereference de->d_inode after l_dput of de (5458)
5058 - use "int" for stripe value returned from lock_to_stripe (5544)
5059 - mballoc allocation and error-checking fixes in 2.6 (5504)
5060 - block device patches to fix I/O request sizes in 2.6 (5482)
5061 - look up hostnames for IB nals (5602)
5062 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5063 - don't start multiple acceptors for the same port (5277)
5064 - fix incorrect LASSERT in mds_getattr_name (5635)
5065 - export a proc file for general "ping" checking (5628)
5066 - fix "lfs check" to not block when the MDS is down (5628)
5068 - service request history (4965)
5069 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5070 - create an "evict_client" /proc entry on OSTs, like the MDS has
5071 - fix mount usage message, return errors per mount(8) (5168)
5072 - change grep [] to grep "[]" in tests so they work in more UMLs
5073 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5074 - remove ancient LOV_MAGIC_V0 EA support (5047)
5075 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5076 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5077 - allow readahead allocations to fail when low on memory (5383)
5078 - mmap locking landed again, after considerable improvement (2828)
5079 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5081 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5084 - send OST transaction number in read/write reply to free req (4966)
5085 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5086 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5087 - the watchdog thread now runs as interruptible (5246)
5088 - flock/lockf fixes (but it's still disabled, pending 5135)
5089 - don't use EXT3 constants in llite code (5094)
5090 - memory shortage at startup could cause assertion (5176)
5092 - reorganization of lov code
5093 - single portals codebase
5095 - add extents/mballoc support (5025)
5096 - direct I/O reads in the obdfilter (4048)
5097 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5099 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5102 - send OST transaction number in read/write reply to free req (4966)
5103 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5104 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5105 - the watchdog thread now runs as interruptible (5246)
5106 - handle missing objects in filter_preprw_read properly (5265)
5107 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5108 - drop import inflight refcount on signal_completed_replay error (5255)
5109 - unlock page after async write error during send (3677)
5110 - return original error code on reconstructed replies (3761)
5111 - no transno return for symlink open, don't save no-trasno open (3440)
5113 - add pid to ldlm debugging output (4922)
5114 - bump the watchdog timeouts -- we can't handle 30sec yet
5115 - extra debugging for orphan dentry/inode bug (5259)
5117 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5120 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5121 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5122 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5123 - fix NULL dereference in /proc/sys/portals/routes (4827)
5124 - allow failed mdc_close() operations to be interrupted (4561)
5125 - stop precreate on OST before MDS would time out on it (4778)
5126 - don't send partial-page writes before EOF from client (4410)
5127 - discard client grant for sub-page writes on large-page clients (4520)
5128 - don't free dentries not owned by NFS code, check generation (4806)
5129 - fix lsm leak if mds_create_objects() fails (4801)
5130 - limit debug_daemon file size, always print CERROR messages (4789)
5131 - use transno after validating reply (3892)
5132 - process timed out requests if import state changes (3754)
5133 - update mtime on OST during writes, return in glimpse (4829)
5134 - add mkfsoptions to LDAP (4679)
5135 - use ->max_readahead method instead of zapping global ra (5039)
5136 - don't interrupt __l_wait_event() during strace
5138 - add software watchdogs to catch hung threads quickly (4941)
5139 - make lustrefs init script start after nfs is mounted
5140 - fix CWARN/ERROR duplication (4930)
5141 - return async write errors to application if possible (2248)
5142 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5143 - print NAL number in %x format (4645)
5144 - update barely-supported suse-2.4.21-171 series (4842)
5145 - support for sles 9 %post scripts
5146 - support for building 2.6 kernel-source packages
5147 - support for sles km_* packages
5149 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5152 - ignore -ENOENT errors in osc_destroy (3639)
5153 - notify osc create thread that OSC is being cleaned up (4600)
5154 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5155 - reconstruct ost_handle() like mds_handle() (4657)
5156 - create a new thread to do import eviction to avoid deadlock (3969)
5157 - let lconf resolve symlinked-to devices (4629)
5158 - don't unlink "objects" from directory with default EA (4554)
5159 - hold socknal file ref over connect in case target is down (4394)
5160 - allow more than 32000 subdirectories in a single directory (3244)
5161 - fix blocks count for O_DIRECT writes (3751)
5162 - OST returns ENOSPC from object create when no space left (4539)
5163 - don't send truncate RPC if file size isn't changing (4410)
5164 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5165 - bind to privileged port in socknal and tcpnal (3689)
5167 - rate limit CERROR/CWARN console message to avoid overload (4519)
5168 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5169 - basic mmap support (3918)
5170 - kernel patch series update from b1_4 (4711)
5172 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5175 - avoid crash during MDS cleanup with OST shut down (2775)
5176 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5177 - don't use bad inodes on the MDS (3744)
5178 - dynamic object preallocation to improve recovery speed (4236)
5179 - don't hold spinlock over lock dumping or change debug flags (4401)
5180 - don't zero obd_dev when it is force cleaned (3651)
5181 - print grants to console if they go negative (4431)
5182 - "lctl deactivate" will stop automatic recovery attempts (3406)
5183 - look for existing locks in ldlm_handle_enqueue() (3764)
5184 - don't resolve lock handle twice in recovery avoiding race (4401)
5185 - revalidate should check working dir is a directory (4134)
5187 - don't always mark "slow" obdfilter messages as errors (4418)
5189 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5192 - don't close LustreDB during write_conf until it is done (3860)
5193 - fix typo in lconf for_each_profile (3821)
5194 - allow dumping logs from multiple threads at one time (3820)
5195 - don't allow multiple threads in OSC recovery (3812)
5196 - fix debug_size parameters (3864)
5197 - fix mds_postrecov to initialize import for llog ctxt (3121)
5198 - replace config semaphore with spinlock (3306)
5199 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5200 - don't allow enqueue to complete on a destroyed export (3822)
5201 - down write_lock before checking llog header bitmap (3825)
5202 - recover from lock replay timeout (3764)
5203 - up llog sem before sending rpc (3652)
5204 - reduce ns lock hold times when setting kms (3267)
5205 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5206 - fix NULL deref and obd_dev leak on setup error (3312)
5207 - replace some LBUG about llog ops with error handling (3841)
5208 - don't match INVALID dentries from d_lookup and spin (3784)
5209 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5210 - fix invalid assertion in ptlrpc_set_wait (3880)
5212 - add libwrap support for the TCP acceptor (3996)
5213 - add /proc/sys/portals/routes for non-root route listing (3994)
5214 - allow setting MDS UUID in .xml (2580)
5215 - print the stack of a process that LBUGs (4228)
5217 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5220 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5221 - make sure to unset replay flag from failed open requests (3440)
5222 - if default stripe count is 0, use OST count for inode size (3636)
5223 - update parent mtime/ctime on client for create/unlink (2611)
5224 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5225 - free recovery state on server during a forced cleanup (3571)
5226 - unregister_reply for resent reqs (3063)
5227 - loop back devices mounting and status check on 2.6 (3563)
5228 - fix resource-creation race that can provoke i_size == 0 (3513)
5229 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5230 - more debugging for page-accounting assertion (3746)
5231 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5232 - avoid deadlock after precreation failure (3758)
5233 - fix race and lock order deadlock in orphan handling (3450, 3750)
5234 - add validity checks when grabbing inodes from l_ast_data (3599)
5236 - add /proc/.../recovery_status to obdfilter (3428)
5237 - lightweight CDEBUG infrastructure, debug daemon (3668)
5238 - change default OSC RPC parameters to be better on small clusters
5239 - turn off OST read cache for files smaller than 32MB
5240 - install man pages and include them in rpms (3100)
5241 - add new init script for (un)mounting lustre filesystems (2593)
5242 - run chkconfig in %post for init scripts (3701)
5243 - drop scimac NAL (unmaintained)
5245 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5248 - clean kiobufs before and after use (3485)
5249 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5250 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5251 - change init script to fail more clearly if not run as root (1528)
5252 - allow clients to reconnect during replay (1742)
5253 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5254 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5255 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5256 - fix cond_resched (3554)
5257 - only evict unfinished clients after recovery (3515)
5258 - allow bulk resend, prevent data loss (3570)
5259 - dynamic ptlrpc request buffer allocation (2102)
5260 - don't allow unlinking open directory if it isn't empty (2904)
5261 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5262 - remove extraneous obd dereference causing LASSERT failure (3334)
5263 - don't use get_cycles() when creating temp. files on the mds (3156)
5264 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5265 - handle EEXIST for set-stripe, set proper directory name (3336)
5267 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5268 - fix ksocknal_fmb_callback() error messages (2918)
5270 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5273 - don't copy lvb into (possibly NULL) reply on error (2983)
5274 - don't deref dentry after dput, don't free lvb on error (2922)
5275 - use the kms to determine writeback rpc length (2947)
5276 - increment oti_logcookies when osc is inactive (2948)
5277 - update client's i_blocks count via lvb messages (2543)
5278 - handle intent open/close of special files properly (1557)
5279 - mount MDS with errors=remount-ro, like obdfilter (2009)
5280 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5281 - don't use cancelling-locks' kms values (2947)
5282 - use highest lock extent for kms, not last one (2925)
5283 - don't dereference ERR_PTR() dentry in error handling path (3107)
5284 - fix thread race in portals_debug_dumplog() (3122)
5285 - create lprocfs device entries at setup instead of at attach (1519)
5286 - common AST error handler, don't evict client on completion race (3145)
5287 - zero nameidata in detach_mnt in 2.6 (3118)
5288 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5289 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5290 - initialize RPC timeout timer earlier for 2.6 (3219)
5291 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5292 - print nal/nid for unknown nid (3258)
5293 - additional checks for oscc recovery before doing precreate (3284)
5294 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5295 - don't crash in mdc_close for bad permissions on open (3285)
5296 - zero i_rdev for non-device files (3147)
5297 - clear page->private before handing to FS, better assertion (3119)
5298 - tune the read pipeline (3236)
5299 - fix incorrect decref of invalidated dentry (2350)
5300 - provide read-ahead stats and refine rpc in flight stats (3328)
5301 - don't hold journal transaction open across create RPC (3313)
5302 - update atime on MDS at close time (3265)
5303 - close LDAP connection when recovering to avoid server load (3315)
5304 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5305 - don't leak open file on MDS after open resend (3325)
5306 - serialize filter_precreate and filter_destroy_precreated (3329)
5307 - loop device shouldn't call sync_dev() for nul device (3092)
5308 - clear page cache after eviction (2766)
5309 - resynchronize MDS->OST in background (2824)
5310 - refuse to mount the same filesystem twice on same mountpoint (3394)
5311 - allow llmount to create routes for mounting behind routers (3320)
5312 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5313 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5314 - fix rare problem with rename on htree directories (3417)
5316 - allow default OST striping configuration per directory (1414)
5317 - fix compilation for qswnal for 2.6 kernels (3125)
5318 - increase maximum number of MDS request buffers for large systems
5319 - change liblustreapi to be useful for external progs like lfsck (3098)
5320 - increase local configuration timeout for slow disks (3353)
5321 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5323 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5326 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5327 - don't overwrite extent policy data in reply if lock was blocked (2901)
5328 - drop filter export grants atomically with removal from device (2663)
5329 - del obd_self_export from work_list in class_disconnect_exports (2908)
5330 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5331 - swab reply message in mdc_close, other PPC fixes (2464)
5332 - fix destroying of named logs (2325)
5333 - overwrite old logs when running lconf --write_conf (2264)
5334 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5335 - fix race in target_handle_connect (2898)
5336 - mds_reint_create() should take same inode create lock (2926)
5337 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5338 - don't close files for self_export to avoid uninitialized obd (2936)
5339 - allow MDS with the same name as client node (2939)
5340 - hold dentry reference for closed log files for unlink (2325)
5341 - reserve space for all logs during transactions (2059)
5342 - don't evict page beyond end of stripe extent (2925)
5343 - don't oops on a deleted current working directory (2399)
5344 - handle hard links to targets without a parent properly (2517)
5345 - don't dereference NULL lock when racing during eviction (2867)
5346 - don't grow lock extents when lots of conflicting locks (2919)
5348 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5351 - account for cache space usage on clients to avoid data loss (974)
5352 - lfsck support in lustre kernel code (2349)
5353 - reduce journal credits needed for BRW writes (2370)
5354 - orphan handling to avoid losing space on client/server crashes
5355 - ptlrpcd can be blocked, stopping ALL progress (2477)
5356 - use lock value blocks to assist in proper KMS, faster stat (1021)
5357 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5358 - recovery for initial connections (2355)
5359 - fixes for mds_cleanup_orphans (1934)
5360 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5361 - block all file creations until orphan recovery completes (1901)
5362 - client remove rq_connection from request struct (2423)
5363 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5364 - recovery timer race (2670)
5365 - mdc_close recovey bug (2532)
5366 - ptlrpc cleanup bug (2710)
5367 - mds timeout on local locks (2588)
5368 - namespace lock held during RPCs (2431)
5369 - handle interrupted sync write properly (2503)
5370 - don't try to handle a message that hasn't been replied to (2699)
5371 - client assert failure during cleanup after abort recovery (2701)
5372 - leak mdc device after failed mount (2712)
5373 - ptlrpc_check_set allows timedout requests to complete (2714)
5374 - wait for inflight reqs when ptlrpcd finishes (2710)
5375 - make sure unregistered services are removed from the srv_list
5376 - reset bulk XID's when resending them (caught by 1138 test)
5377 - unregister_bulk after timeout
5378 - fix lconf error (2694)
5379 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5380 - readahead locks pages, leaves pending causing memory pressure (2673)
5381 - increase OST request buffers to 4096 on large machines (2729)
5382 - fix up permission of existing directories in simple_mkdir (2661)
5383 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5384 - don't assign transno to errored transactions (2742)
5385 - don't delete objects on OST if given a bogus objid from MDS (2751)
5386 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5387 - if rq_no_resend, then timeout request after recovery (2432)
5388 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5389 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5390 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5391 - MDS recovery shouldn't delete objects that it has given out (2730)
5392 - if enqueue arrives after completion, don't clobber LVB (2819)
5393 - don't unlock pages twice when trigger_group_io returns error (2814)
5394 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5395 - don't write pages to disk if there was an error (1450)
5396 - don't ping imports that have recovery disabled (2676)
5397 - take buffered bytes into account when balancing socknal conn (2817)
5398 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5399 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5400 - remove little-endian swabbing of llog records (1987)
5401 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5402 - retry reposting request buffers if they fail (1191)
5403 - grow extent at grant time to avoid granting a revoked lock (2809)
5404 - lock revoke doesn't evict page if covered by a second lock (2765)
5405 - disable VM readahead to avoid reading outside lock extents (2805)
5407 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5408 - updated kernel patches for hp-2.4.20 kernel (2681)
5410 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5413 - fix truncated write corruption (2366)
5414 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5416 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5417 - ensure len(uuid) < 37 in lmc (1171)
5418 - fix ia64 OOPS in llog_test (2255)
5419 - zero end of page at obdfilter for partial page writes (2648)
5420 - don't leave stale dentries around after renames (bug 2428)
5421 - fix timeouts when evicting a client with a single lock held (2642)
5422 - set deadline for the initial HELLO message to drain (2634)
5423 - print out dotted-quad IP addresses in the socknal (2302)
5425 - additional debugging for MDS client eviction problem (2443)
5426 - fix mkfsoptions support for osts (2603, 2604)
5428 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5431 - add series for the vanilla 2.6.0 kernel
5432 - add series for the vanilla 2.4.24 kernel
5433 - add series for a cray x86/64 UL kernel drop
5434 - fix xattr patches for the vanilla 2.4.19 series
5436 - generate true UUIDs in lmc (1171)
5437 - have portals stack dumping break in UML (2466)
5438 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5439 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5440 - wake callback waiting threads on client eviction (2460)
5441 - Add --ptldebug and --subsystem to lmc (1719)
5442 - update assertion to allow safe interrupt allocation
5443 - set rq_no_resend for cancel requests (2432)
5444 - recalculate ptlrpcd timeout after resend (2494)
5445 - call vfs_rmdir when removing pending directories (2368)
5446 - fix renaming a file to itself (2429)
5447 - lmc creates a default one-stripe lov (2454)
5448 - expand procfs space to handle large clusters (2326)
5449 - increase UML stack to avoid overflow
5450 - update lconf's list of debug and subsystem masks
5451 - fix lfs find --obd (2510)
5452 - /proc tunable for disabling filter read caching (2591)
5453 - stop rpm packages from altering slapd.conf (2301)
5454 - disable nagle in the socknal under 0conf (2578)
5455 - choose mds inode size based on stripe count (2572)
5456 - fix kernel-source rpm problems (2516)
5458 - add --disable-doc to avoid pdf generation (2421)
5459 - update documentation, tests, type-os, comments
5460 - avoid format warnings on ia64
5461 - remove the TOE NAL
5462 - tiny code cleanups by removing unused fields
5464 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5467 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5468 - remove the most blatant lies from BUILDING file (2371)
5469 - change default debug level to reasonable production setting
5470 - reduce client side cache size to reduce cache flush time
5471 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5472 - make TCP zerocopy and pinger support enabled by default (2476)
5473 - sync writes completed after process exits caused crashes (2319)
5474 - maintain correct mount count on the MDS (2356)
5475 - backout 1557, because 2316 wasn't really fixed
5476 - better file I/O statistics gathering in /proc
5477 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5478 - another kernel patch to fix zero-copy TCP function export
5479 - don't take duplicate lock when processing re-sent getattr (2420)
5480 - lctl uses obd_self_export instead of creating new conn (2353)
5481 - MDS/OST recovery case which requires object creation asserted (2425)
5482 - move lfs from /usr/sbin to /usr/bin in packages
5483 - fix race between mds_client_add and mds_client_free (2417)
5484 - use kmalloc instead of slabs in portals (2430)
5485 - don't create duplicate records when a failover MDS is present (2442)
5486 - remove unnecessary mount age check (2332)
5487 - don't remove directory inodes from locks prematurely (2451)
5488 - don't break if MDS service name is the same as hostname (2103)
5489 - fix races in client write RPC generation when cache full (2482)
5491 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5494 - remove now-unused request->rq_obd (278)
5495 - if an allocation fails, print out how much memory we've used (1933)
5496 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5497 - add the "configurable stack size" patch to most series files (1256)
5498 - ability to write large log records, for 100+ OST configs (2306)
5499 - fix NULL deref when filter_prep fails (2314)
5500 - fix operator precedence error in filter_sync
5501 - dynamic allocation of socknal TX descriptors (2315)
5502 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5503 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5504 - ensure that CWARN messages in Portals always get to the syslog
5505 - __init/__exit are not for prototype decls (ldlm_init/exit)
5506 - x86-64 compile warning fixes
5507 - fix gateway LMC keyword conflict (2318)
5508 - fix MDS lock inversions in getattr/reint paths (1844)
5509 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5510 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5511 - DLM race condition prevented some lock evictions (2328)
5512 - ENOMEM detection and retry on socknal sends (2230)
5513 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5514 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5515 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5516 - better error messages when a client is rejected during recovery (1505)
5517 - avoid cancelling locks which were never granted, after failure (2330)
5518 - fix i_sem/journal inversion in mds_client_add (2333)
5519 - fix truncate/getattr lock cycle deadlock (2334)
5520 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5521 - fix two rare exit paths which could leak an l_lock() ref (2321)
5522 - fencepost error in MDS/OST orphan recovery (2226)
5523 - make log record alignment 8 bytes (1988)
5524 - lstripe now fails when requested offset > ost_count (2237)
5525 - ensure that all kernel series have a complete list.h (1607)
5526 - fix crashes in special-file operations (2316)
5527 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5529 - allow configurable automake binary, for testing new versions
5530 - small update to the lfs documentation
5532 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5534 * fix negative export reference count in fsfilt_sync (2312)
5536 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5537 * release candidate 0.9.1
5539 - orphans are moved into the PENDING directory for possible recovery
5540 - replayed opens now open by fid for orphan/rename safety (1042)
5541 - last close of an orphan inode generates a transno (683)
5542 - chdir() and mount() now pin the directory entry (1020)
5543 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5544 - discard very old requests without processing them (1502)
5545 - remove some common, well-understood CERRORs (1505)
5546 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5547 - clear "grant" flags in OST replies until OST grant code lands (1644)
5548 - fix read performance by not clobbering i_blksize on client (1598)
5549 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5550 - make LRU size a /proc tunable, clears locks when reduced (707)
5551 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5552 - server locks take references on exports now (1558)
5553 - build fixes for 2.4.20-rh trees (1663)
5554 - return an error from lov_create if all OSCs are inactive (1751)
5555 - fix import levels when a reconnect happens without a timeout (1597)
5556 - exit early from mds_open if we get a lookup error (1749)
5557 - partial page read at EOF wouldn't wait for disk before sending (1642)
5558 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5559 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5560 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5561 - write/truncate lock inversion (1639)
5562 - Don't auto-load obdclass, portals modules during cleanup (1495)
5563 - fix timestamps from jumping to "now" (1763)
5564 - extra journal assertions (1648)
5565 - add an extra multiunlink test (1771)
5566 - fix read_record/write_record API (1776)
5567 - fix leak of offset_extent, possible incorrect i_size later (1772)
5568 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5569 - Debugging for the kqswnal_get_idle_tx problems (1820)
5570 - Allow recovery to be attempted multiple times (1536)
5571 - Write out MDS last_rcvd file after it is first created (1600)
5572 - Fix tx_descriptor leak in failed transmit situations (1827)
5573 - ext3 journaling fixes for assertion failure after IO error (1871)
5574 - class_export_put() on freed export after completion AST error (1896)
5575 - Fix revalidate looping in VFS (1322)
5576 - Don't access a freed export during MDS_REINT timeout (1521)
5577 - Add open-unlink recovery support on the MDS (1673,1764)
5578 - Return an error if no MDS data was read from last_rcvd (1946)
5579 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5580 - Don't setup a disk device that doesn't match exported UUID (317)
5581 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5582 - avoid committing NULL handle in force close
5583 - local.sh is now a one-stripe LOV configuration
5584 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5585 - fix ext3 htree duplicate directory entry corruption (1516)
5586 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5587 - update to new LOV EA format (2097)
5588 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5590 2003-06-15 Phil Schwan <phil@clusterfs.com>
5593 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5594 - per-import/export recovery handling (958, 931, 959)
5595 - multiple last-rcvd slots, for serving multiple FSes (949)
5596 - connections are again shared between multiple imp/exports (963, 964)
5597 - "umount -f" would hang if any requests needed to be sent (393, 978)
5598 - avoid pinning large req buffer by copying for queued messages (989)
5599 - add "uuid" to "lctl device" command to help upcalls (991)
5600 - "open" RPCs with transnos would confuse recovery counters (1037)
5601 - do proper endian conversion of all wire messages (288, 340, 891)
5602 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5603 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5604 - fix ptlrpc_connection leak in target_handle_connect (1174)
5605 - fix import refcounting bug in OST and MDS cleanup (1134)
5606 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5607 - fix very unlikely obd_types race condition (501)
5608 - remove osc_open hack for echo_client (1187)
5609 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5610 - a failure in read_inode2 leads to deadlock (1139)
5611 - cancel ack-locks as soon as transaction is committed (1072)
5612 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5613 - make sure to commitrw after any preprw to avoid deadlock (1162)
5614 - failing to execute a file in a lustre FS would lock inode (1203)
5615 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5616 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5617 - obd_uuid2tgt would walk off the end of the list (1255)
5618 - on IA64 the osc would give portals incorrect bulk size (1258)
5619 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5620 - fix lock inversion caused by new llite matching code (1282)
5621 - limit the number of dirty pages on a client to 10MB (1286)
5622 - timed out locks were not being corrected cancelled (1289)
5623 - fix O_DIRECT above 4GB on IA-32 (1292)
5624 * major user-visible changes
5625 - fail out/fail over policy now controlled by the upcall (993)
5627 - add OBD_PING to check server availability and failure (954)
5628 - lustre messages are now sent in sending host order (288, 340, 891)
5629 - add eadatalen to MDS getattr reply (340)
5630 - OST read replies may contain second buffer, with per-page status (593)
5632 2003-03-11 Phil Schwan <phil@clusterfs.com>
5635 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5636 - failed open()s could cause deadlock; fixed (867, 869)
5637 - stop cancelling OST locks when files are closed (481)
5638 - overlapping XID spaces caused network corruption (851, 853)
5639 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5640 - setattr_raw added, to do single-RPC, server-side setattrs
5641 - lmc/lconf syntax change for OST UUIDs
5642 - fix crashy race condition between ptlrpc_free_req and osc_close
5643 - don't use request in mdc_enqueue if we hit a timeout (889)
5644 - don't set the inode i_size for regular files from the MDS (896)
5645 - handle out of order completion AST (842)
5646 - don't LBUG if a lock request times out after receiving AST (913)
5647 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5648 - if a bad lock AST arrives, send an error instead of dropping entirely
5649 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5650 - fix leak in bulk IO when only partially completed (899, 900, 926)
5651 - fix O_DIRECT for ia64 (55)
5652 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5653 - C-z after timeout could hang a process forever; fixed (977)
5655 - client-side I/O cache (678, 924, 929, 941, 970)
5657 - READPAGE and SETATTRs which don't take server-side locks get
5660 2003-02-11 Phil Schwan <phil@clusterfs.com>
5663 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5664 - Fix fsfilt_extN_readpage() to read a full page of directory
5665 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5666 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5667 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5668 - Most intent operations were removed, in favour of a new RPC mode
5669 that does a single RPC to the server and bypasses most of the VFS
5670 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5671 - Aggressively cancel local locks on DLM servers
5672 - mds_reint_unlink sends EA to the client if it's the last nlink.
5673 client uses that EA to unlink OST objects.
5674 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5675 - recursive symlinks were fixed (439)
5676 - fixed NULL deref in DEBUG_REQ
5677 - filter_update_lastobjid no longer calls sync, which annoyed extN
5678 - fixed multi-client small-writes to a single file problem (445)
5679 - fixed mtime updates during file writes (607)
5680 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5681 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5682 - fixed Linux OST size reporting problem (444, 656)
5683 - OST now updates object mtime with writes or setattr (607, 619)
5684 - client verifies file size before zeroing page past EOF (445)
5685 - OST now writes last allocated objid to disk with allocation (108)
5686 - LOV on echo now works (409)
5688 - mds_reint_unlink sends a new buffer, with the EA included. this
5689 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5690 set if a regular file was being unlinked, and it was the last link
5691 - use PtlGet from the target for bulk writes (315)
5692 - OST now updates object mtime with writes or setattr (607, 619)
5693 - LDLM now has a grant-time callback to revalidate locked items, if
5695 - Many MDS operations were reorganized to combat race conditions
5697 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5698 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5700 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5703 - Fully reactivate OST imports after reconnection (512, others)
5704 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5705 - More graceful error handling for truncating on dead OST (515)
5706 - Don't error out unless we're actually accessing dead stripes (474)
5707 - Fix garbage sizes when stripes are missing (410)
5708 - LRU counters were broken, causing constant lock purge (433, 432)
5709 - garbage on read from stripes with failed OSTs (441)
5710 - mark OSCs as active before reconnecting during recovery (438)
5711 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5712 - lfind did not preserve OST order in output (443)
5713 - symlinks cause hung clients, incorrect data (439)
5714 - stop dereferencing request after dropping refcount (457)
5715 - don't LASSERT(spin_is_locked) on non-SMP (455)
5716 - fixes for many rename() bugs
5717 - fstat didn't correctly synchronize attributes (399)
5718 - server must handle lock cancellation during blocking AST prep (487)
5719 - bulk descriptors were free()d too soon (511)
5720 - fix paths in lconf, which would load incorrect modules (451, 507)
5721 - fix confusing lconf 'host not found' error message (386)
5722 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5723 - fix race condition in mdc_blocking_ast() for inode access (526)
5724 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5725 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5726 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5727 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5728 - fix lstripe in several ways (kernel side) (527)
5729 - fix request leak in ldlm_cli_enqueue (262)
5730 - incorrect OSC was marked inactive after OST failure
5731 - call mds_fs_cleanup before unmounting filesystem (524)
5732 - fix races between taking ns_lock and ldlm_lock_change_resource
5733 - fix races updating LOV export open file list
5734 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5735 - fix recovery NULL deref in ldlm_cli_cancel_unused
5736 - fix some DLM races by using new hash table for lock handles (419)
5737 - permit the client to specify desired inodes, at replay
5738 - duplicate requests when we queue them for replay reintegration
5739 - fix last_rcvd offset calculation
5740 - sync after each recovered transaction, so we always make progress
5741 - never, not always, ERESTART requests without transnos
5742 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5744 - skip replay if the MDS says that the client is already connected
5745 - don't check for a recovery-enabled export to match lctl's UUID
5746 - don't INC_USE_COUNT for phantom exports
5747 - don't crash when cleaning up phantom exports (567)
5748 - don't double-finish or set replay data for errored mdc_open requests
5749 - abort requests when they time out, so we don't get old replies
5750 - send/receive replies for AST messages again
5751 - if the client says that it doesn't have the lock, cancel it on the
5753 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5754 mark it as destroyed, it will all work out when decref is called
5755 - fix module use counts (22, 581)
5757 - ASTs now expect a reply (server cancels lock on error reply)
5759 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5762 - fix many simultaneous client startup (392)
5763 - fix dentry->d_it clobbering
5764 - credentials weren't being shipped for readdir/getattr operations
5765 - remove invalid assertions triggered during some concurrent MD
5767 - proper Lustre versions added (336, 389)
5768 - fix memory leak for create error case (398)
5769 - fix LOV locking bug that would get cli/srv out of sync
5770 - fix echo client over LOV (409)
5771 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5772 - fix double-O_EXCL intent crash (424)
5773 - avoid sending multiple lock CANCELs (352)
5775 - MDS can do multi-client recovery (modulo bugs in new code)
5777 - many updates, edits, cleanups
5779 2002-11-18 Phil Schwan <phil@clusterfs.com>
5782 - fix null d_it dereference (346)
5783 - fix full OST/dbench hang (333)
5784 - fix permission problem with file removal (286)
5785 - fix removal of OSCs from LOV when they fail
5786 - fix NULL deref during bulk timeout (214)
5787 - fix problems related to multiple filesystems on one MDS (241)
5788 - fixed serious subtle metadata locking bugs
5789 - free locks on clients when inodes are removed due to memory
5791 - fix inode pointer in lock data (285)
5792 - partial support for multiple MDS on a single host (241)
5793 - data locks weren't cancelled at clear_inode time (290, 311)
5794 - intent locks could lead to unbounded lock growth (205)
5795 - added a maximum lock count, an LRU list, and a flusher
5796 - fix multiple rename (365)
5797 - properly abstracted the echo client
5798 - OSC locked 1 byte too many; fixed
5799 - rewrote brw callback code:
5800 - fixed recovery bugs related to LOVs (306)
5801 - fixed too-many-pages-in-one-write crash (191)
5802 - fixed (again) crash in sync_io_timeout (214)
5803 - probably fixed callback-related race (385)
5805 - Add capability to MDS protocol
5806 - LDLM cancellations and callbacks on different portals
5808 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5811 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5812 - defer open object destruction to close time (601981, 138)
5813 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5814 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5815 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5816 - don't call obd_setattr in ll_file_release for destroyed objects
5817 * protocol change to lustre_msg: move |version| and add |flags|
5818 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5819 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5820 * added replay of create, unlink, link and rename operations during
5821 MDS failover; recovery should be much more robust now
5822 * remove failed OSCs from LOVs (only lov_create uses this so far)
5823 * the lustre-HOWTO was brought (more) up to date (582544)
5825 2002-10-23 Phil Schwan <phil@clusterfs.com>
5828 - in-use dentries weren't being reused properly (617851)
5829 - prevent multiple LDLM setup (599178)
5830 - fix LOV size calculations for truncate (617853)
5831 - fix client handling of MDS intent errors (POSIX)
5832 - fix permission bug in lovstripe.c test (624321)
5833 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5834 - truncate past end of file could corrupt data
5835 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5836 - a race in recovery could return ETIMEDOUT to apps (623947)
5837 - building outside the source directory was fixed
5838 * the lustre-HOWTO was brought (more) up to date (582544)
5839 * major progress was made on recovery functionality
5841 2002-10-10 Phil Schwan <phil@clusterfs.com>
5844 - recovery deadlock fix
5845 - rm -rf causes LBUG fix (617817)
5846 - file open by multiple tasks fix (618962)
5847 - directory permissions bugs (602707 and 620007)
5848 - journal_stop fixed with locking (611313)
5849 - O_APPEND failures resolved (618273, perhaps 614459)
5850 - lconf PATH fix (619770)
5851 - IA64 build fix (621450)
5852 - RPC buffer sizes scale with amount of memory
5854 2002-10-01 Phil Schwan <phil@clusterfs.com>
5857 - locks would be cancelled without throwing away data pages,
5858 resulting in inconsistent data (605627)
5859 - inode attributes were not always being refreshed (605627, 612449)
5860 - lconf now continues to cleanup after lctl reports an error
5861 - MDS now enforces user permissions (602707)
5862 - lprocfs cleanup fixed, but not yet enabled (614157)
5863 - fixed infinite server hang, should a client not respond to an AST
5864 - avoid going into recovery if user calls readlink() with a buffer
5865 that's too small (613941)
5866 - AST RPCs no longer require replies (614867) -- this may be changed
5867 - don't crash server if client sends an IOV that's too big (611336)
5868 - fixed lock conversion deadlock (611892)
5869 - fixed the following of symlinks (614622)
5870 * recovery: the server can remove locks from a client that dies, other
5871 clients can make progress
5872 * more extN patch fixes
5873 * compile-time configurable ptlrpc buffer allocations
5875 - collaborative read cache document
5876 - Lustre Lite Performance CDR document-in-progress
5878 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5881 - fix typo in patch-2.4.18
5883 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5886 - clear ptlrpc request each time in handle_incoming_request()
5887 - unlink of files now destroys the object on the OST
5889 2002-09-19 Peter Braam <braam@clusterfs.com>
5891 * add hard link support
5892 * change obdfile creation method
5893 * kernel patch changed
5895 2002-09-19 Peter Braam <braam@clusterfs.com>
5898 - stack overflow bug in extN fixed
5900 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5902 * documentation updates
5903 - add man pages for config tools
5904 - update tests/README to describe testing with new config tools
5905 - finish metadata API descriptions
5906 * bug fixes and cleanups
5907 - statfs workaround for 16TB limit
5908 - LOV stripe allocation improved, can stripe on subset of OSTs
5909 - LOV file size/IO offset was wrong for files > 4GB in size
5910 - object EA data was being dropped, caused files to be unreadable
5911 - memory overflow with non-LOV OST caused memory corruption
5912 - fixed regression tests to work with new config tools, obdfilter
5913 - fixed bug when directory size became larger than 1 block
5914 - fixed bug (for single client case) when PWD was deleted
5915 - invalidate local directory pages when doing intent-based ops
5916 - avoid LDLM oops when lock callback contained bad data
5918 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5920 * documentation updates
5921 * bug fixes and cleanups
5922 - configuration tools
5925 - 64-bit compile warnings
5926 - 64-bit internal statfs data
5928 * test_brw on persistent OST devices
5930 * lprocfs (disabled)
5932 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5934 * documentation updates
5935 * bug fixes and cleanups
5936 * configuration tools
5938 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5941 * many small fixes to 0_5_4
5942 * io/network handling
5943 * thinkos in MDS operations
5945 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5948 * crucial basic fixes to 0.5.3
5949 * IOR, Iozone work over Elan
5952 2002-08-07 Phil Schwan <phil@clusterfs.com>
5953 * version 0_5_3, our first alpha
5954 * we use the new Portals iovs
5955 * documentation updates
5956 * bug fixes and cleanups
5957 * small changes in the DLM wire protocol
5959 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5960 * version 0_5_1 with some initial stability,
5961 * locking on MD and file I/O.
5962 * documentation updates
5963 * several bug fixes since 0.5.0
5964 * small changes in wire protocol
5966 2002-07-18 Phil Schwan <phil@clusterfs.com>
5968 * delivered as Lustre Light Alpha
5969 * fixed a crash after handling invalid MDS requests
5970 * fixed directory pages for architectures with non-4k pages sizes
5972 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5973 * release version v0_4_4
5974 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5976 * Fixes a number of interruption problems with OST operations.
5977 * Update documentation for portals header changes
5978 * Move all wire protocol structs/defines to lustre_idl.h
5979 * Fixes symlink length bug.
5980 * Add tcpdump to repository.
5982 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5983 * release version v0_4_3
5984 * Fixes statfs for inodes on extN.
5985 * Fixes bug in runtests which would delete /etc/hosts.
5986 * Use 64-bit object IDs wherever possible (not into VFS though)
5987 Remove ost_get_info, which is unused by lustre, and out of date.
5989 2002-07-03 Peter Braam <braam@clusterfs.com>
5990 * release version v0_4_2 Fixes a lookup error (type not passed)
5991 * move forward to head of Portals
5992 * move forward to latest Lustre kernel
5994 2002-06-25 Peter Braam <braam@clusterfs.com>
5995 * release version v0_4_1. Hopefully stable on single node use.