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: enable client interface failover
19 Details : When a child reconnects from another NID, properly update export
20 nid hash position and ldlm reverse import.
22 Severity : enhancement
24 Description: Communicate OST degraded/readonly state via statfs to MDS
25 Details : Flags in the statfs returned from OSTs indicate whether the
26 OST is in a degraded RAID state, or if the filesystem has
27 turned read-only after a filesystem error is detected.
32 Description: don't panic if EPROTO was hit when reading symlink
33 Details : correctly handling request reference in error cases.
35 Severity : enhancement
37 Description: Add LustreNetLink, a kernel-userspace communcation path.
39 Severity : enhancement
41 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
46 Description: access to llog context before init.
47 Details : move handling CATALOGS file at osc layer and forbid access to llog
51 Frequency : rare, only when client is writing to OST on local node
53 Description: Avoid deadlock for local client writes
54 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
55 memory freeing context. This allows OST threads to set the
56 PF_MEMALLOC flag on task structures in order to allocate memory
57 from reserved pools and complete IO.
58 Use GFP_HIGHUSER for OST allocations for non-local client writes,
59 so that the OST threads generate memory pressure and allow
60 inactive pages to be reclaimed.
62 Severity : enhancement
64 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
69 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
70 Details : move ldlm namespace creation in setup phase to avoid grab
71 _lprocfs_lock with cli_sem held.
75 Description: Temporarily disable grant shrink.
76 Details : Disable the feature for debugging.
79 Frequency : only when formatting test filesystems
81 Description: Unable to run several mkfs.lustre on loop devices at the same time
82 Details : mkfs.lustre returns error 256 on the concurrent loop devices
83 formatting. The solution is to proper handle the error.
85 Severity : enhancement
87 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
89 Severity : enhancement
91 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
96 Description: resolve race between obd_disconnect and class_disconnect_exports
97 Details : if obd_disconnect will be called to already disconnected export he
98 forget release one reference and osc module can't unloaded.
100 Severity : enhancement
102 Description: Allow tuning service thread via /proc
103 Details : For each service a new
104 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
105 created that can be used to set min/max thread counts, and get the
106 current number of running threads.
111 Descriptoin: don't return error if have partially created objects for file.
112 Details : lov_update_create_set uses set->set_success as index for created
113 objects, so if some requests will be failed, they will have hole
114 at end of array and qos_shrink_lsm for allocate correct lsm.
116 Severity : enhancement
118 Description: Update OFED support to 1.4.1
123 Description: Reduce small size read RPC
124 Details : Set read-ahead limit for every file and only do read-ahead when
125 available read-ahead pages are bigger than 1M to avoid small size
128 Severity : enhancement
130 Description: move AT tunable parameters for more consistent usage
131 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
133 Severity : enhancement
135 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
136 Details : allow skip disconnected ost for send statfs request and hide error
142 Description: fix racy locking of mballoc block bitmaps causing BUG
143 Details : The locking of the mballoc buddy bitmap and the in-memory
144 block bitmap was using two different spin locks in some
145 cases. This made it possible to incorrectly access the
146 mballoc bitmap while another process was modifying it,
147 causing a sanity assertion to fail. While no on-disk corruption
148 was reported, there was some risk of this happening.
151 Frequency : rare, on llog test 6
153 Descriptoin: don't allow connect to already connected import
154 Details : allowing connect to already connected import is hide connecting problem.
159 Descriptoin: allow kill process which wait statahead result
160 Details : for some reasons 'ls' can stick in waiting result from statahead,
161 in this case need way for kill this process.
163 Severity : enhancement
165 Description: Add state history info file, enhance import info file
166 Details : Track import connection state changes in a new osc/mdc proc file;
167 add overview-type data to the osc/mdc import proc file.
169 Severity : enhancement
171 Description: MDS create should not wait for statfs RPC while holding DLM lock.
173 Severity : enhancement
175 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
180 Descriptoin: OST grant too much space to client even there are not enough space.
181 Details : Client will shrink its grant cache to OST if there are no write
182 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
183 this grant cache if there are already not enough avaible space
184 (left_space < total_clients * 32M).
187 Frequency : start MDS on uncleanly shutdowned MDS device
189 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
190 Details : stay in waiting mds<>ost recovery finished produce random bugs
191 due race between two ll_sync thread for one lov target. send
192 ACTIVATE event only if connect realy finished and import have
196 Frequency : rare, connect and disconnect target at same time
198 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
199 Details : don't call obd_disconnect under lov_lock. this long time
200 operation and can block ptlrpcd which answer to connect request.
205 Descriptoin: don't lose wakeup for imp_recovery_waitq
206 Details : recover_import_no_retry or invalidate_import and import_close can
207 both sleep on imp_recovery_waitq, but we was send only one wakeup
211 Frequency : always with long access acl
213 Descriptoin: mds can't pack reply with long acl.
214 Details : mds don't control size of acl but they limited by reint/getattr
217 Severity : enhancement
219 Description: Update to SLES10 kernel-2.6.16.60-0.33.
221 Severity : enhancement
223 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
226 Frequency : start MDS on uncleanly shutdowned MDS device
228 Descriptoin: aborting recovery hang on MDS
229 Details : don't throttle destroy RPCs for the MDT.
232 Frequency : on remount
234 Description: external journal device not working after the remount
235 Details : clear dev_rdonly flag for external journal devices in
241 Description: shutdown vs evict race
242 Details : client_disconnect_export vs connect request race.
243 if client will evicted at this time - we start invalidate
244 thread without referece to import and import can be freed
248 Frequency : rare, need acl's on inode.
250 Description: client can't handle ost additional correctly
251 Details : if ost was added after client connected to mds client can have
252 hit lnet_try_match_md ... to big messages to wide striped files.
253 in this case need teach client to handle config events about add
254 lov target and update client max ea size at that event.
256 Severity : enhancement
258 Description: Changelogs
259 Details : Changelogs are a lightweight mechanism to track filesystem
260 metadata and namespace changes. The changelog is recorded
261 permanently on the MDTs, and is periodically "consumed" / purged
262 when records are no longer needed.
264 Severity : enhancement
266 Description: compact fld format with extents
267 Details : Store range of seq rather than every seq in FLD. Seq
268 controller update FLD rather than clients. In Case of CMD, mdt0
269 has FLD, all other metadata server act as non persistent proxy
270 for FLD queries and cache fld entries in fld cache.
275 Description: don't skip ost target if they assigned to file
276 Details : Drop slow OSCs if we can, but not for requested start idx.
277 This means "if OSC is slow and it is not the requested
278 start OST, then it can be skipped, otherwise skip it only
279 if it is inactive/recovering/out-of-space.
283 Description: more cleanup in mds_lov
284 Details : not send LOV EA under replay, we can't know about they size at this
285 time. Don't allow client connect to mds before any ost connected,
286 for avoid problems with LOV EA size and returning EIO to client.
288 Severity : enhancement
290 Description: Interoperability at server side (Disk interoperability)
292 Severity : enhancement
294 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
296 Severity : enhancement
298 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
300 Severity : enhancement
302 Description: New client IO stack (CLIO).
304 Severity : enhancement
306 Description: Commit on sharing. Eliminate inter-client dependencies between
307 uncommitted transactions by doing transaction commits.
308 Thereby clients may recovery independently.
311 Frequency : Create a symlink file with a very long name
313 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
314 Details : If there is no extra space in the request for early cancels,
315 ldlm_req_handles_avail() returns 0 instead of a negative value.
317 Severity : enhancement
319 Description: Add /proc entry for import status
320 Details : The mdc, osc, and mgc import directories now have
321 an import directory that contains useful import data for debugging
324 Severity : enhancement
326 Description: Re-disable certain /proc logging
327 Details : Enable and disable client's offset_stats, extents_stats and
328 extents_stats_per_process stats logging on the fly.
331 Frequency : Only on FC kernels 2.6.22+
333 Description: oops in statahead
334 Details : Do not drop reference count for the dentry from VFS when lookup,
335 VFS will do that by itself.
337 Severity : enhancement
339 Description: Generic /proc file permissions
340 Details : Set /Proc file permissions in a more generic way to enable non-
341 root users operate on some /proc files.
345 Description: Hitting mdc_commit_close() ASSERTION
346 Details : Properly handle request reference release in
347 ll_release_openhandle().
351 Description: quota recovery deadlock during mds failover
352 Details : This patch includes att18982, att18236, att18237 in bz14840.
354 1. fix osts hang when mds does failover with quotaon
355 2. prevent watchdog storm when osts threads wait for the
360 Frequency : only patchless client
361 Description: add workaround for race between add/remove dentry from hash
363 Severity : enhancement
365 Description: Allow OST glimpses to return PW locks
369 Description: LBUG when llog conf file is full
370 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
374 Description: Prevent import from entering FULL state when server in recovery
378 Description: service mount cannot take device name with ":"
379 Details : Only when device name contains ":/" will mount treat it as
385 Description: replace ptlrpcd with the statahead thread to interpret the async
386 statahead RPC callback
390 Frequency : on recovery
391 Description: I/O failures after umount during fail back
392 Details : if client reconnected to restarted server we need join to recovery
393 instead of find server handler is changed and process self
394 eviction with cancel all locks.
396 Severity : enhancement
398 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
400 Severity : enhancement
402 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
404 Severity : enhancement
406 Description: Upcall on Lustre log has been dumped
407 Details : Allow for a user mode script to be called once a Lustre log has
408 been dumped. It passes the filename of the dumped log to the
409 script, the location of the script can be specified via
410 /proc/sys/lnet/debug_log_upcall.
415 Description: avoid idr_remove called for id which is not allocated.
416 Details : Move assigment s_dev for clustered nfs to end of initialization,
417 to avoid problem with error handling.
422 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
423 Details : When connection is reused this not moved from CONN_UNUSED_HASH
424 into CONN_USED_HASH and this prodice warning when put connection
425 again in unused hash.
428 Severity : enhancement
430 Description: File striping can now be set to use an arbitrary pool of OSTs.
432 Severity : enhancement
434 Description: Export bytes_read/bytes_write count on OSC/OST.
438 Description: Early reply size mismatch, MGC loses connection
439 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
440 the connect flags are properly negotiated.
445 Description: quotas are not honored with O_DIRECT
446 Details : all writes with the flag O_DIRECT will use grants which leads to
447 this problem. Now using OBD_BRW_SYNC to guard this.
451 Description: add quota statistics
452 Details : 1. sort out quota proc entries and proc code.
453 2. add quota statistics
455 Severity : enhancement
457 Description: enable quota support for HEAD.
461 Description: Properly propagate oinfo flags from lov to osc for statfs
462 Details : restore missing copy oi_flags to lov requests.
464 Severity : enhancement
466 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
467 : l_getgroups(8), lst(8), routerstat(8)
468 Details : included man pages for llobdstat(8), llstat(8),
469 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
471 Severity : enhancement
473 Description: configure's --enable-quota should check the
474 : kernel .config for CONFIG_QUOTA
475 Details : configure is terminated if --enable-quota is passed but
476 : no quota support is in kernel
480 Description: Remove portals compatibility
481 Details : Remove portals compatibility, not interoperable with releases
486 Description: Resolve device initialization race
487 Details : Prevent proc handler from accessing devices added to the
488 obd_devs array but yet be intialized.
490 Severity : enhancement
492 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
494 Severity : enhancement
496 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
501 Description: Using wrong pointer in osc_brw_prep_request
502 Details : Access to array[-1] can produce panic if kernel compiled with
503 CONFIG_PAGE_ALLOC enabled
507 Description: Client runs out of low memory
508 Details : Consider only lowmem when counting initial number of llap pages
512 Description: *optional* service tags registration
513 Details : if the "service tags" package is installed on a Lustre node
514 When the filesystem is mounted, a local-node service tag will
515 be created. See http://inventory.sun.com/ for more information
516 about the Service Tags asset management system.
520 Description: Kernel BUG tries to release flock
521 Details : Lustre does not destroy flock lock before last reference goes
522 away. So always drop flock locks when client is evicted and
523 perform unlock regardless of successfulness of speaking to MDS.
527 Description: add refcount protection for osc callbacks, avoid panic on shutdown
531 Description: sanity test 65a fails if stripecount of -1 is set
532 Details : handle -1 striping on filesystem in ll_dirstripe_verify
537 Description: ASSERTION(CheckWriteback(page,cmd)) failed
538 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
541 Severity : enhancement
543 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
547 Description: do not process already freed flock
548 Details : flock can possibly be freed by another thread before it reaches
549 to ldlm_flock_completion_ast.
553 Description: LBUG during stress test
554 Details : Need properly lock accesses the flock deadlock detection list.
558 Description: oops in page fault handler
559 Details : kernel page fault handler can return two special 'pages' in error
560 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
564 Description: timeout with invalidate import.
565 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
566 should be handled by ptlrpcd. This produce long age waiting and
567 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
569 Severity : enhancement
571 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
575 Description: enable MGS and MDT services start separately
576 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
577 a co-located MGS without starting the MGS, which is a complement
578 to 'nosvc' mount option.
582 Frequency : after recovery
583 Description: precreate to many object's after del orphan.
584 Details : del orphan st in oscc last_id == next_id and this triger growing
585 count of precreated objects. Set flag LOW to skip increase count
586 of precreated objects.
590 Frequency : rare, on clear nid stats
591 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
592 Details : when clean nid stats sometimes try destroy live entry,
593 and this produce panic in free.
597 Description: Stack overflow during MDS log replay
598 ease stack pressure by using a thread dealing llog_process.
602 Description: wait until IO finished before start new when do lock cancel.
603 Details : VM protocol want old IO finished before start new, in this case
604 need wait until PG_writeback is cleared until check dirty flag and
605 call writepages in lock cancel callback.
607 Severity : enhancement
609 Description: using special macro for print time and cleanup in includes.
613 Description: mds_mfd_close() ASSERTION(rc == 0)
614 Details : In mds_mfd_close(), we need protect inode's writecount change
615 within its orphan write semaphore to prevent possible races.
619 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
623 Frequency : rare, on shutdown ost
624 Description: don't hit live lock with umount ost.
625 Details : shrink_dcache_parent can be in long loop with destroy dentries,
626 use shrink_dcache_sb instead.
630 Description: don't panic with use echo client
631 Details : echo client pass NULL as client nid pointer and this produce null
636 Description: fix build on ppc32
637 Details : compile code with -m64 flag produce wrong object file for ppc32.
641 Description: add message levels for liblustreapi
645 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
646 Details : Move the 'good_osts' check before the 'total_bavail' check. This
647 will result in an -EAGAIN and in the exit call path we call
648 alloc_rr() which will with increasing aggressiveness attempt to
649 aquire precreated objects on the minimum number of required OSCs.
653 Description: Use old size assignment to avoid deadlock
654 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
655 the scheduling while holding a spinlock. We do not need locking
656 for size in ll_update_inode() because size is only updated from
657 the MDS for directories or files without objects, so there is no
658 other place to do the update, and concurrent access to such inodes
659 are protected by the inode lock.
663 Description: resolve "_IOWR redefined" build error on SLES10
667 Description: dump the memory debugging after all modules are unloaded to
668 suppress false negative in conf_sanity test 39
670 Severity : enhancement
672 Description: build kernel-ib packages for OFED 1.3 in our release cycle
677 Description: fix SLES kernel versioning
678 Details : the kernel version for our SLES 10 kernel did not include a "-"
679 before the "smp" at the end. while this was not a problem in
680 general, it did mean that software trying to use the kernel
681 version to try to detect a vendor specific kernel would fail.
682 this was most evident by the OFED build scripts.
686 Description: Don't update lov_desc members until making sure they are valid
687 Details : When updating lov_desc members via proc fs, need fix their
688 validities before doing the real update.
692 Description: don't put request into delay list while invalidate in flight.
693 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
694 import in flight. this produce timeout for invalidate and sometimes
695 can cause stale data.
699 Frequency : on ppc only
700 Description: not convert ost objects for directory because it's not exist.
701 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
705 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
707 Severity : enhancement
709 Description: Update to support 2.6.22.14 vanilla kernel.
713 Frequency : rare, on recovery
714 Description: read procfs can produce deadlock in some situation
715 Details : Holding lprocfs lock which send rpc can produce block for destroy
716 obd objects and this also block reconnect with -EALREADY. This
717 isn't fix all lprocfs bugs - but make it rare.
719 Severity : enhancement
721 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
724 Frequency : frequent on X2 node
726 Description: mdc_set_open_replay_data LBUG
727 Details : Set replay data for requests that are eligible for replay.
731 Description: lustre_mgs: operation 101 on unconnected MGS
732 Details : When MGC is disconnected from MGS long enough, MGS will evict the
733 MGC, and late on MGC cannot successfully connect to MGS and a lot
734 of the error messages complaining that MGS is not connected.
738 Frequency : on network error
739 Description: panic with double free request if network error
740 Details : mdc_finish_enqueue is finish request if any network error ocuring,
741 but it's true only for synchronus enqueue, for async enqueue
742 (via ptlrpcd) this incorrect and ptlrpcd want finish request
745 Severity : enhancement
747 Description: client-side metadata stat-ahead during readdir(directory readahead)
748 Details : perform client-side metadata stat-ahead when the client detects
749 readdir and sequential stat of dir entries therein
752 Frequency : on start mds
754 Description: Implement get_info(last_id) in obdfilter.
757 Frequency : occasional
759 Description: Correctly check stale fid, not start epoch if ost not support SOM
760 Details : open with flag O_CREATE need set old fid in op_fid3 because
761 op_fid2 was overwritten with new generated fid, but mds can answer
762 with one of these two fids and both is not stale. Setattr
763 incorrectly started an epoch and assume will be called
764 done_writing, but without SOM done_writing ever being called.
767 Frequency : rare, depends on device drivers and load
769 Description: MDS or OSS nodes crash due to stack overflow
770 Details : Code changes in 1.8.0 increased the stack usage of some functions.
771 In some cases, in conjunction with device drivers that use a lot
772 of stack the MDS (or possibly OSS) service threads could overflow
773 the stack. One change which was identified to consume additional
774 stack has been reworked to avoid the extra stack usage.
777 Frequency : occasional
779 Description: Do not fail import if osc_interpret_create gets -EAGAIN
780 Details : If osc_interpret_create got -EAGAIN it immediately exits and
781 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
782 oscc_has_objects and see OSC has no objests and call
783 oscc_internal_create to resend create request.
785 Severity : enhancement
787 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
789 Severity : enhancement
791 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
794 Frequency : very rare
796 Description: Fix replay if there is an un-replied request and open
797 Details : In some cases, older replay request will revert the
798 mcd->mcd_last_xid on MDS which is used to record the client's
801 Severity : enhancement
803 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
805 Severity : enhancement
807 Description: Add rhel5 support to HEAD.
809 Serverity : enhancement
811 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
816 Description: Don't allow skipping OSTs if index has been specified.
817 Details : Don't allow skipping OSTs if index has been specified, make
818 locking in internal create lots better.
822 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
823 Details : access to bitfield in structure is always rounded to long
824 and this produce problem with not atomic change any bit.
828 Description: Lustre make rpms failed.
829 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
830 builds Lustre rpms from tarball.
833 Frequency : rare on shutdown ost
835 Description: If llog cancel was not send before clean_exports phase, this can
836 produce deadlock in llog code.
837 Details : If llog thread has last reference to obd and call class_import_put
838 this produce deadlock because llog_cleanup_commit_master wait when
839 last llog_commit_thread exited, but this never success because was
840 called from llog_commit_thread.
844 Description: allow userland application know is lost one of stripes.
845 Details : fill lvb_blocks with error code on ost and return it to
846 application if error flag found.
850 Description: NULL lov_tgts causing MDS oops
851 Details : more safe checks for NULL lov_tgts for avoid oops.
853 Severity : enhancement
855 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
859 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
860 Details : If an OST is down the MDS will hang indefinitely in
861 obd_statfs_rqset() waiting for the statfs data. While for
862 MDS QOS usage of statfs, it should not stuck in waiting.
864 Severity : enhancement
866 Description: remote_acl support
867 Details : Support ACL-based permission check for remote user.
868 Support setfacl/getfacl for remote user with the utils
869 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
870 the system "{s,g}etfacl" utils.
872 Severity : enhancement
874 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
876 Severity : enhancement
878 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
882 Description: make mgs_setparam() handle fsname containing dash
883 Details : fsname containing a dash does not work with lctl conf_param
885 Severity : enhancement
887 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
889 Severity : enhancement
891 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
895 Frequency : rare, at shutdown
896 Description: access already free / zero obd_namespace.
897 Details : if client_disconnect_export was called without force flag set,
898 and exist connect request in flight, this can produce access to
899 NULL pointer (or already free pointer) when connect_interpret
900 store ocd flags in obd_namespace.
904 Frequency : only at startup
905 Description: not alloc memory with spinlock held.
906 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
907 if any spinlock held.
909 Severity : enhancement
911 Description: make lustre randomly fail allocating memory
912 Details : Make lustre randomly failed allocating memory for testing purpose.
914 Severity : enhancement
916 Description: lost problems with lov objid file
917 Details : Fixes some scability and access to not inited memory problems
918 in work with lov objdid file.
923 Description: lfs find does not continue on file error
924 Details : Continue other files processing when a file/dir is absent.
928 Description: Inconsistent usage of lustre_pack_reply()
929 Details : Standardize the usage of lustre_pack_reply() such that it
930 always generate a CERROR on failure.
933 Frequency : occasional
935 Description: MDS hang or stay in waiting lock
936 Details : If client receive lock with CBPENDING flag ldlm need send lock
937 cancel as separate rpc, to avoid situation when cancel request
938 can't processed due all i/o threads stay in wait lock.
942 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
946 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
950 Description: Update to SLES9 kernel-2.6.5-7.287.3.
954 Description: Update to RHEL4 latest kernel.
956 Severity : enhancement
958 Description: Build SLES10 patchless client fails
959 Details : The configure was broken by run ./configure with
960 --with-linux-obj=.... argument for patchless client. When the
961 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
962 can't search header adequately. Use absolute path such as
963 -I($LINUX)/include instead.
967 Description: interrupt oig_wait produce painc on resend.
968 Details : brw_redo_request can be used for resend requests from ptlrpcd and
969 private set, and this produce situation when rq_ptlrpcd_data not
970 copyed to new allocated request and triggered LBUG on assert
971 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
972 ptlrpcd set if request is changed and can be safety changed to use
977 Description: LASSERT_{REQ,REP}SWAB macros are buggy
978 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
979 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
980 these macros and replace them with their difinitions instead.
984 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
985 Details : Update which_patch & target file for SLES10 latest kernel.
988 Description: Remove client patches from SLES 10 kernel.
989 Details : This causes SLES 10 clients to behave as patchless clients
990 even on a Lustre-patched (server) kernel.
992 Severity : enhancement
994 Description: self-adjustable client's lru lists
995 Details : use adaptive algorithm for managing client cached locks lru
996 lists according to current server load, other client's work
997 pattern, memory activities, etc. Both, server and client
998 side namespaces provide number of proc tunables for controlling
1001 Severity : enhancement
1003 Description: light-weight GSS support
1004 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1005 and reduce performance overhead.
1007 Severity : enhancement
1009 Description: Linux keyring support
1010 Details : Support using service of Linux keyring for Lustre GSS internal
1011 context refresh/cache mechanism.
1015 Description: Fix errors in lfs documentation
1016 Details : Fixes man pages
1020 Description: don't use GFP_* in generic Lustre code.
1021 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1025 Description: obdclass is limited by single OBD_ALLOC(idarray)
1026 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1030 Description: warnings with build patchless client with vanila 2.6.19 and up
1031 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1032 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1036 Frequency : Only for SLES
1037 Description: sanity_quota fail test_1
1038 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1039 file, which makes TSTID[2] inunique.
1043 Frequency : Only for SLES9
1044 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1045 Details : Update target/ChangeLog/which_patch .
1049 Frequency : Only for test_10 in sanity-quota.sh
1050 Description: a bug in quota test code
1051 Details : checking a test flag in a wrong place
1055 Frequency : Only for test_13 in sanity-quota.sh
1056 Description: a bug in quota test script
1057 Details : shouldn't check overfull allocated quota
1061 Description: jbd statistics
1062 Details : Port older jbd statistics patch for sles10
1064 Severity : enhancement
1066 Description: Add printing inode info into message about error in writepage.
1071 Description: reply_lock_interpret crash due to race with it and lock cancel.
1072 Details : Do not replay locks that are being cancelled. Do not reference
1073 locks by their address during replay, just by their handle.
1077 Frequency : When flocks are used.
1078 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1079 Details : Flock locks might destroy just granted lock if it could be merged
1080 with another existing flock, this is done in completion handler,
1081 so teach ldlm_cli_enquque_fini that this is a valid case for
1087 Description: Oops in read and write path when failing to allocate lock.
1088 Details : Check if lock allocation failed and return error back.
1092 Description: lstripe command fails for valid OST index
1093 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1094 instead of lov->desc.ld_active_tgt_count.
1098 Description: sanity.sh failed test 103
1099 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1104 Description: df doesn't work properly if diskfs blocksize != 4K
1105 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1109 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1110 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1112 Severity : enhancement
1114 Description: Adaptive timeouts
1115 Details : RPC timeouts adapt to changing server load and network
1116 conditions to reduce resend attempts and improve recovery time.
1120 Description: llapi_file_create() does not allow some changes
1121 Details : add llapi_file_open() that allows specifying the mode and
1122 open flags, and also returns an open file handle.
1126 Description: merge and cleanup kernel patches.
1127 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1131 Description: Add journal checksum support.(Kernel part)
1132 Details : The journal checksum feature adds two new flags i.e
1133 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1134 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1135 indicates that the commit block contains the checksum for
1136 the blocks described by the descriptor blocks. Now commit
1137 record can be sent to disk without waiting for descriptor
1138 blocks to be written to disk. This behavior is controlled
1139 using JBD2_FEATURE_ASYNC_COMMIT flag.
1143 Description: OSS needs mutliple precreate threads
1144 Details : Add ability to start more than one create thread per OSS.
1148 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1149 Details : execute lfs setstripe on client
1153 Description: mds_obd_create error creating tmp object
1154 Details : When the user sets quota on root, llog will be affected and can't
1155 create files and write files.
1158 Frequency : Always on ia64 patchless client, and possibly others.
1160 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1161 Details : This allows the patchless client to be loaded on architectures
1162 without this export.
1166 Description: RedHat Update kernel for RHEL5
1167 Details : Add the kernel config file for RHEL5.
1171 Description: RedHat Update kernel for RHEL5
1172 Details : Modify the kernel config file more closer RHEL5.
1176 Description: Build failure against Centos5 (RHEL5)
1177 Details : Define PAGE_SIZE when it isn't present.
1181 Description: test_7 on sanity-quota.sh can't be used on separate machines
1182 Details : add facet to handle it
1186 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1187 Details : with NFS, the anon dentry's parent was set to itself in
1188 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1189 corresponding dentry other than use rec->ur_name.
1191 Severity : enhancement
1193 Description: lfs setstripe enhancement
1194 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1198 Description: enable data checksumming by default
1199 Details : enable checksum by default, allow --disable-checksum
1200 configure option and "-o nochecksum" mount option. Checksums
1201 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1205 Description: lustre support for RHEL5
1206 Details : Add support for RHEL5.
1210 Description: Client eviction due to failover config
1211 Details : after a connection loss, the lustre client should attempt to
1212 reconnect to the last active server first before trying the
1213 other potential connections.
1217 Description: when mds and osts use different quota unit(32bit and 64bit),
1218 quota will be released repeatly.
1219 Details : void sending multiple quota reqs to mds, which will keep the
1220 status between the reqs.
1224 Description: osts not allocated evenly to files
1225 Details : change the condition to increase offset_idx
1228 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1230 Description: Data corruption for OSTs that are formatted larger than 2TB
1232 Details : When generating the bio request for lustre file writes the
1233 sector number would overflow a temporary variable before being
1234 used for the IO. The data reads correctly from Lustre (which
1235 will overflow in a similar manner) but other file data or
1236 filesystem metadata may be corrupted in some cases.
1240 Description: Tune the kernel for good SCSI performance.
1241 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1242 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1247 Description: rewrite ext2-derived code in obdclass/uuid.c
1248 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1249 from scratch preserving functionality.
1253 Description: "lfs find -obd UUID" prints directories
1254 Details : "lfs find -obd UUID" will return all directory names instead
1255 of just file names. It is incorrect because the directories
1256 do not reside on the OSTs.
1260 Description: use i_size_read and i_size_write in 2.6 port
1261 Details : replace inode->i_size access with i_size_read/write()
1264 Frequency : when removing large files
1266 Description: scheduling issue during removal of large Lustre files
1267 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1268 It causes scheduling issues when removing large files (17TB in the
1272 Frequency : only with liblustre clients on XT3
1274 Description: evictions taking too long
1275 Details : allow llrd to evict clients directly on OSTs
1278 Frequency : only on ppc
1280 Description: /proc/fs/lustre/devices broken on ppc
1281 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1282 We should make sure the type of 'index' is loff_t and then cast
1283 explicitly as needed below. Do not assign an explicitly cast
1287 Frequency : only for rhel5
1289 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1290 Details : Modify the target file & which_kernel.
1292 Severity : enhancement
1294 Description: omit set fsid for export NFS
1295 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1296 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1300 Frequency : Always, for kernels after 2.6.16
1301 Description: Fix warning idr_remove called for id=.. which is not allocated.
1302 Details : Last kernels save old s_dev before kill super and not allow
1303 to restore from callback - restore it before call kill_anon_super.
1307 Description: buffer overruns could theoretically occur
1308 Details : llapi_semantic_traverse() modifies the "path" argument by
1309 appending values to the end of the origin string, and a
1310 overrun may occur. Adding buffer overrun check in liblustreapi.
1314 Description: Fix error on 'ls .' at the top of the Lustre mount.
1315 Details : Don't revalidate dentry if it is a root dentry.
1319 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1320 Details : Modify vm-tunables-rhel4.patch.
1324 Description: Kernel config for 2.6.18-vanilla.
1325 Details : Modify targets/2.6-vanilla.target.in.
1326 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1327 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1328 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1329 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1333 Description: improve handling recoverable errors
1334 Details : if request processig with error which can be recoverable on server
1335 request should be resend, otherwise page released from cache and
1340 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1341 Details : Modify target file & which_patch.
1342 A flaw was found in the IA32 system call emulation provided
1343 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1344 value could be stored in the %RAX register, which could trigger an
1345 out-of-bounds system call table access. An untrusted local user
1346 could exploit this flaw to run code in the kernel
1347 (ie a root privilege escalation). (CVE-2007-4573).
1351 Description: O_DIRECT bypasses client statistics.
1352 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1353 read_ahead_stats were updated. copied the stats section from
1354 osc_send_oap_rpc() into async_internal().
1358 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1362 Description: change order of libsysio includes
1363 Details : '#include sysio.h' should always come before '#include xtio.h'
1367 Description: handle "serious error: objid * already exists" more gracefully
1368 Details : If LAST_ID value on disk is smaller than the objects existing in
1369 the O/0/d* directories, it indicates disk corruption and causes an
1370 LBUG(). If the object is 0-length, then we should use the existing
1371 object. This will help to avoid a full fsck in most cases.
1373 Severity : enhancement
1375 Description: adapt the lustre_config script to support the upgrade case
1376 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1377 server targets to 1.6 in parallel.
1381 Description: To avoid grant space > avaible space when the disk is almost
1382 full. Without this patch you might see the error "grant XXXX >
1383 available" or some LBUG about grant, when the disk is almost
1385 Details : In filter_check_grant, for non_grant cache write, we should
1386 check the left space by if (*left > ungranted + bytes), instead
1387 of (*left > ungranted), because only we are sure the left space
1388 is enough for another "bytes", then the ungrant space should be
1389 increase. In client, we should update cl_avail_grant only there
1390 is OBD_MD_FLGRANT in the reply.
1394 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1395 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1396 described in CVE-2007-4573. This problem could allow untrusted
1397 local users to gain root access.
1400 Frequency : when using O_DIRECT and quotas
1402 Description: Incorrect file ownership on O_DIRECT output files
1403 Details : block usage reported by 'lfs quota' does not take into account
1404 files that have been written with O_DIRECT.
1409 Description: touch file failed when fs is not full
1410 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1411 otherwise we can get ENOSP while fs is not full.
1415 Description: parallel lock callbacks
1416 Details : Instead of sending blocking and completion callbacks as separated
1417 requests, adding them to a set and sending in parallel.
1420 Frequency : only for Cray XT3
1421 Bugzilla : 12829/13455
1422 Description: Changing primary group doesn't change the group lustre assigns to
1424 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1425 overridden with the primary group provided by the group upcall,
1426 whereas the supplied fsgid can be trusted if it is in the list of
1427 supplementary groups returned by the group upcall.
1429 Severity : enhancement
1431 Description: Allow masking D_WARNING, D_ERROR messages from console
1432 Details : Console messages can now be disabled via lnet.printk.
1436 Description: User code with malformed file open parameter crashes client node
1437 Details : Before packing join_file req, all the related reference should be
1438 checked carefully in case some malformed flags cause fake
1439 join_file req on client.
1443 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1444 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1445 mode is equal to the granted mode. However, it should also check
1446 that LDLM_FL_CP_REQD is not set.
1450 Description: Heavy nfs access might result in deadlocks
1451 Details : After ELC code landed, it is now improper to enqueue any mds
1452 locks under och_sem, because enqueue might want to decide to
1453 cancel open locks for same inode we are holding och_sem for.
1457 Description: Client eviction while running blogbench
1458 Details : A lot of unlink operations with concurrent I/O can lead to a
1459 deadlock causing evictions. To address the problem, the number of
1460 oustanding OST_DESTROY requests is now throttled to
1461 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1462 callbacks are processed in priority.
1466 Description: enable ACLs on MDS by default
1467 Details : ACLs must be enabled on MDS by default.
1470 Frequency : PPC/PPC64 only
1472 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1473 Details : fix duplicated definitions between asm-ppc64/types.h and
1474 lustre_types.h on PPC.
1477 Frequency : PPC/PPC64 only
1479 Description: asm-ppc/segment.h does not exist
1480 Details : fix compile issue on PPC.
1484 Description: better handle error messages in extents code
1487 Frequency : RHEL4 only
1489 Description: mkfs is very slow on IA64/RHEL4
1490 Details : A performance regression has been discovered in the MPT Fusion
1491 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1492 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1493 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1495 Severity : enhancement
1497 Description: SNMP support enhancement
1498 Details : Adding total number of sampled request for an MDS node in snmp
1501 Severity : enhancement
1503 Description: Optimize ldlm waiting list processing for PR extent locks
1504 Details : When processing waiting list for read extent lock and meeting read
1505 lock that is same or wider to it that is not contended, skip
1506 processing rest of the list and immediatelly return current
1507 status of conflictness, since we are guaranteed there are no
1508 conflicting locks in the rest of the list.
1512 Description: Time out and refuse to reconnect
1513 Details : When the failover node is the primary node, it is possible
1514 to have two identical connections in imp_conn_list. We must
1515 compare not conn's pointers but NIDs, otherwise we can defeat
1516 connection throttling.
1520 Description: port llog fixes from b1_6 into HEAD
1521 Details : Port llog reference couting and some llog cleanups from b1_6
1522 (bug 10800) into HEAD, for protect from panic and access to already
1523 free llog structures.
1527 Description: Detect stride IO mode in read-ahead
1528 Details : When a client does stride read, read-ahead should detect that and
1529 read-ahead pages according to the detected stride pattern.
1533 Description: data checksumming impacts single node performance
1534 Details : add support for several checksum algorithm. Currently, only CRC32
1535 and Adler-32 are supported. The checksum type can be changed on
1536 the fly via /proc/fs/lustre/osc/*/checksum_type.
1540 Description: use adler32 for page checksums
1541 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1546 Description: build for x2 fails
1547 Details : fix compile issue on Cray systems.
1551 Description: Properly match for duplicate locks
1552 Details : Due to different lock order from skiplists code, we need to
1553 traverse entire list for now
1556 Frequency : only on PPC/SLES10
1558 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1559 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1560 be defined. Add a hack in mkfs_lustre.c to work around this compile
1565 Description: LASSERT on MDS when client holding flock lock dies
1566 Details : ldlm pool logic depends on number of granted locks equal to
1567 number of released locks which is not true for flock locks, so
1568 just exclude such locks from consideration.
1572 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1573 Details : Use fsfilt_sync() for both the whole filesystem sync and
1574 individual file sync to eliminate dangerous inode locking
1575 with I_LOCK that can lead to a deadlock.
1579 Description: performance in 1.6.3
1580 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1581 sure that 1MB requests are not fragmented by the block layer.
1583 Severity : enhancement
1585 Description: organize the server-side client stats on per-nid basis
1586 Details : Change the structure of stats under obdfilter and mds to
1594 The "uuid"s file would list the uuids of _active_ exports.
1595 And the clear entry is to clear all stats and stale nids.
1597 Severity : enhancement
1599 Description: eliminate client locks in face of contention
1600 Details : file contention detection and lockless i/o implementation
1601 for contended files.
1605 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1608 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1610 Description: MDS may lose file striping (and hence file data) in some cases
1611 Details : If there are additional extended attributes stored on the MDS,
1612 in particular ACLs, SELinux, or user attributes (if user_xattr
1613 is specified for the client mount options) then there is a risk
1614 of attribute loss. Additionally, the Lustre file striping
1615 needs to be larger than default (e.g. striped over all OSTs),
1616 and an additional attribute must be stored initially in the
1617 inode and then increase in size enough to be moved to the
1618 external attribute block (e.g. ACL growing in size) for file
1623 Description: skiplist implementation simplification
1624 Details : skiplists are used to group compatible locks on granted list
1625 that was implemented as tracking first and last lock of each
1626 lock group the patch changes that to using doubly linked lists
1630 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1631 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1632 because LASSERT on both the data supplied by a client, and the
1633 data on disk is dangerous and incorrect.
1635 Severity : enhancement
1637 Description: Slow truncate/writes to huge files at high offsets.
1638 Details : Directly associate cached pages to lock that protect those pages,
1639 this allows us to quickly find what pages to write and remove
1640 once lock callback is received.
1644 Description: more ldlm soft lockups
1645 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1646 starve other threads from the resource lock for a long time in
1647 case of long waiting queue, so change the debug level from
1648 D_OTHER to the less frequently used D_INFO.
1650 Severity : enhancement
1652 Description: add -gid, -group, -uid, -user options to lfs find
1656 Description: Hung threads in invalidate_inode_pages2_range
1657 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1658 once one is completed. As a result, some RPCs are stuck in the
1659 queue and are never sent.
1663 Description: filter threads hungs on waiting journal commit
1664 Details : Cleanup filter group llog code, then only filter group llog will
1665 be only created in the MDS/OST syncing process.
1669 Description: Procfs and llog threads access destoryed import sometimes.
1670 Details : Sync the import destoryed process with procfs and llog threads by
1671 the import refcount and semaphore.
1673 Severity : enhancement
1675 Description: openlock cache of b1_6 port to HEAD
1680 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1681 Details : If insertion of an extent fails, then discard the inode
1682 preallocation and free data blocks else it can lead to duplicate
1687 Description: don't always update ctime in ext3_xattr_set_handle()
1688 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1689 In some cases the ctime should not be updated, for example for
1690 2.0->1.8 compatibility it is necessary to delete an xattr and it
1691 should not update the ctime.
1695 Bugzilla : 15713/16362
1696 Description: Assertion in iopen_connect_dentry in 1.6.3
1697 Details : looking up an inode via iopen with the wrong generation number can
1698 populate the dcache with a disconneced dentry while the inode
1699 number is in the process of being reallocated. This causes an
1700 assertion failure in iopen since the inode's dentry list contains
1701 both a connected and disconnected dentry.
1705 Description: assertion failure in ldlm_handle2lock()
1706 Details : fix a race between class_handle_unhash() and class_handle2object()
1707 introduced in lustre 1.6.5 by bug 13622.
1712 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1713 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1718 Description: oops in sd_iostats_seq_show()
1719 Details : unloading/reloading the scsi low level driver triggers a kernel
1720 bug when trying to access the sd iostat file.
1725 Description: Kernel panics during QLogic driver reload
1726 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1727 patch, causing memory corruption.
1732 Description: journal_dev option does not work in b1_6
1733 Details : pass mount option during pre-mount.
1735 Severity : enhancement
1737 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1738 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1739 extent information of a file. It can be used to map logical blocks
1740 in a file to physical blocks in the block device.
1744 Description: LDLM soft lockups - improvement
1745 Details : It is be possible to send the lock handle along with each read
1746 or write request because the client is already doing a lock match
1747 itself so there isn't any reason the OST should have to re-do that
1751 Frequency : only on Cray X2
1753 Description: X2 build failures
1754 Details : fix build failures on Cray X2.
1758 Description: xid & resent requests
1759 Details : Initialize RPC XID from clock at startup (randomly if clock is
1762 Severity : enhancement
1764 Description: Add lustre_start utility to start or stop multiple Lustre servers
1769 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1770 Details : In case of memory pressure, list_del() can be called twice on
1771 req->rq_history_list, causing a kernel oops.
1775 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1776 Details : fix stack overflow in the distributed lock manager by defering
1777 export eviction after a failed AST to the elt thread instead of
1778 handling it in the dlm interpret routine.
1782 Description: Convert some comments to new format.
1783 Details : Update documenting comments to match doxygen conventions.
1787 Description: Grammar fixes.
1788 Details : A couple of trivial spelling fixes.
1792 Description: OSD_COUNTERS-mandatory
1793 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1794 to simplify the code.
1798 Description: simplify cmm_device freeing logic.
1799 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1803 Description: Add lockdep support to dt_object_operations locking interface.
1804 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1805 parameter indicating lock ordering. Update mdd code to use new
1810 Description: Introduce failloc constants for lockless IO tests.
1811 Details : Add two new failloc constants to test lockless IO. Only one of
1812 them in implemented---another is checked in yet to be landed
1817 Description: Add lockdep support for inode mutex.
1818 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1819 in the situations where Lustre has to lock more than one inode
1824 Description: Add optional invariants checking support.
1825 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1826 configure switch. This macro is to be used for consistency and
1827 sanity checks that are too expensive to be left in `production'
1832 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1833 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1838 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1839 Details : Use spin_lock_nested() in (the only) situation where more than
1840 one ldlm_lock is locked simultaneously. Also, fix possible
1841 dead-lock in ldlm_lock_change_resource() by enforcing particular
1846 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1847 Details : Instead of specifying each ldlm_lock call-back through separate
1848 parameter, wrap them into struct ldlm_callback_suite.
1852 Description: Kill join_lru obd method and its callers.
1853 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1854 in memory---a requirement implemented through ->o_join_lru() obd
1855 method in HEAD. Remove this method and its users.
1859 Description: Add asynchronous ldlm ENQUEUE completion handler.
1860 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1861 case that stock ldlm_completion_ast() cannot handle as it waits
1862 until lock is granted. Introduce new ldlm_completion_ast_async()
1863 for this. Also comment ldlm_completion_ast().
1867 Description: ldlm_error <-> errno conversion.
1868 Details : Add functions to map (rather arbitrary) between LDLM error codes
1869 and standard errno values. CLIO needs this to prevent LDLM specific
1870 constants from escaping ldlm and osc.
1874 Description: Kill unused ldlm_handle2lock_ns() function.
1875 Details : Kill unused ldlm_handle2lock_ns() function.
1879 Description: Add lu_ref support to ldlm_lock
1880 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1881 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1882 and ldlm_resource. LDLM interface has to be changed, because code that
1883 releases a reference on a lock, has to "know" what reference this is.
1884 In the most frequent case
1886 lock = ldlm_handle2lock(handle);
1888 LDLM_LOCK_PUT(lock);
1890 no changes are required. When any other reference (received _not_ from
1891 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1892 instead of LDLM_LOCK_PUT().
1894 Arguably, changes are pervasive, and interface requires some discipline
1895 for proper use. On the other hand, it was very instrumental in finding
1896 a few leaked lock references.
1900 Description: Add ldlm_lock_addref_try().
1901 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1902 attempts to addref a lock that might be being canceled
1907 Description: Add ldlm_weigh_callback().
1908 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1909 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1910 hard-coded `number of pages' logic.
1914 Description: Add lockdep annotations to llog code.
1915 Details : Use appropriately tagged _nested() locking calls in the places
1916 where llog takes more than one ->lgh_lock lock.
1920 Description: Add loi_kms_set().
1921 Details : Wrap kms updates into a helper function.
1925 Description: Constify instances of struct lsm_operations.
1926 Details : Constify instances of struct lsm_operations.
1930 Description: lu_conf support.
1931 Details : On a server, a file system object is uniquely identified
1932 by a fid, which is sufficient to locate and load all object
1933 state (inode). On a client, on the other hand, more data are
1934 necessary instantiate an object. Change lu_object_find() and
1935 friends to take additional `lu_conf' argument describing object.
1936 Typically this includes layout information.
1940 Description: lu_context fixes.
1941 Details : Introduce new lu_context functions that are needed on the client
1942 side, where some system threads (ptlrpcd) are shared by multiple
1943 modules, and so cannot be stopped during module shutdown.
1947 Description: Add start and stop methods to lu_device_type_operations.
1948 Details : Introduce two new methods in lu_device_type_operations, that are
1949 invoked when first instance of a given type is created and last one
1950 is destroyed respectively. This is need by CLIO.
1954 Description: Add lu_ref support to struct lu_device.
1955 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1956 track leaked references.
1960 Description: Introduce lu_kmem_descr.
1961 Details : lu_kmem_descr and its companion interface allow to create
1962 and destroy a number of kmem caches at once.
1966 Description: Fix lu_object finalization race.
1967 Details : Fix a race between lu_object_find() finding an object and its
1968 concurrent finalization. This race is (most likely) not possible
1969 on the server, but might happen on the client.
1973 Description: Introduce lu_ref interface.
1974 Details : lu_ref is a debugging module allowing to track references to
1975 a given object. It is quite cpu expensive, and has to be
1976 explicitly enabled with --enable-lu_ref. See usage description
1981 Description: Factor lu_site procfs stats into a separate function.
1982 Details : Separate lu_site stats printing code into a separate function
1983 that can be reused on a client.
1987 Description: Constify instances of struct {lu,dt,md}_device_operations.
1988 Details : Constify instances of struct {lu,dt,md}_device_operations.
1992 Description: Introduce struct md_site and move meta-data specific parts of
1993 struct lu_site here.
1994 Details : Move md-specific fields out of struct lu_site into special struct
1995 md_site, so that lu_site can be used on a client.
1999 Description: Kill mdd_lov_destroy().
2000 Details : Remove unused mdd code.
2004 Description: Add st_block checking to multistat.c.
2005 Details : Add st_block checking to multistat.c.
2009 Description: Add lu_ref support to struct obd_device.
2010 Details : Add lu_ref tracking to obd_device.
2014 Description: Kill obd_set_fail_loc().
2015 Details : Remove unused code.
2019 Description: Add special type for ptlrpc_request interpret functions.
2020 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2021 there. Actual usage will be in CLIO.
2025 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2026 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2027 doesn't work with lockdep.
2031 Description: Add rwv.c test program.
2032 Details : New testing program exercising readv(2) and writev(2) (Qian).
2036 Description: Add sendfile.c test program.
2037 Details : New testing program exercising sendfile(2) (Jay).
2041 Description: Ratelimit a message that can be very frequent.
2042 Details : Ratelimit a memory allocation failure message that can
2047 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2048 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2052 Description: Liblustre build fixes.
2053 Details : Liblustre build fixes.
2057 Description: libcfs: add cfs_{need,cond}_resched() interface.
2058 Details : libcfs: add cfs_{need,cond}_resched() definition and
2059 implementations for Linux, NT, and liblustre.
2061 Severity : enhancement
2063 Description: More exported tunables for mballoc
2064 Details : Add support for tunable preallocation window and new tunables for
2065 large/small requests
2069 Description: Detect corruption of block bitmap and checking for preallocations
2070 Details : Checks validity of on-disk block bitmap. Also it does better
2071 checking of number of applied preallocations. When corruption is
2072 found, it turns filesystem readonly to prevent further corruptions.
2076 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2077 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2078 window to 0. If it does hit cache miss, and read-pattern is still
2079 stride-io mode, does not reset the stride window, but also does not
2080 increase the stride window length in this case.
2084 Frequency : only for big-endian servers
2085 Description: Check if system is big-endian while mounting fs with extents feature
2086 Details : Mounting a filesystem with extents feature will fail on big-endian
2087 systems since ext3-based ldiskfs is not supported on big-endian
2088 systems. This can be over-riden with "bigendian_extents" mount option.
2090 Severity : enhancement
2092 Description: The root squash functionality
2093 Details : A security feature, which is to prevent users from being able
2094 to mount lustre on their desktop, run as root, and delete
2095 all of the files in the filesystem. The goal is accomplished by
2096 remapping user id (UID) and group id (GID) of the root user to
2097 a UID and GID specified by the system administartor via Lustre
2098 configuration management server (MGS). The functionality also
2099 allows to specify sets of clients for which the remapping does
2104 Description: Excessive recovery window
2105 Details : With AT enabled, the recovery window can be excessively long (6000+
2106 seconds). To address this problem, we no longer use
2107 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2108 timeout no longer depends on the service time, it is set to
2109 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2110 time via pb_service_time.
2114 Description: Watchdog triggered on MDS failover
2115 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2116 the OSTs know that the MDS "UUID" can be reused for the same export
2117 from a different NID, so we do not need to wait for the export to be
2121 Frequency : rare, only if using MMP with Linux RAID
2123 Description: MMP doesn't work with Linux RAID
2124 Details : While using HA for Lustre servers with Linux RAID, it is possible
2125 that MMP will not detect multiple mounts. To make this work we
2126 need to unplug the device queue in RAID when the MMP block is being
2127 written. Also while reading the MMP block, we should read it from
2128 disk and not the cached one.
2130 Severity : enhancement
2132 Description: open file using fid
2133 Details : A file can be opened using just its fid, like
2134 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2137 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2140 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2141 Details : add proper locking for creation of jbd_4k slab cache
2145 Description: MMP check in ext3_remount() fails without displaying any error
2146 Details : When multiple mount protection fails during remount, proper error
2149 Severity : enhancement
2151 Description: Allow stripe size to be up to 4G-64k
2152 Details : Fix math logic to allow large stripe sizes.
2156 Description: add check for >8TB ldiskfs filesystems
2157 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2158 Don't allow >8TB ldiskfs filesystems to be mounted without
2159 force_over_8tb mount option
2164 Descriptoin: OSS complains "too many blocks" and crashes
2165 Details : In filter_commitrw_write():
2166 1. lquota_chkquota();
2167 2. obdo_from_inode();
2168 3. lquota_pending_commit();
2169 "2." has changed uid and gid in this case, then caused
2170 lquota_chkquota() and lquota_pending_commit() to use
2171 different uid and gid. That is the root of the bug.
2173 --------------------------------------------------------------------------------
2175 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2177 * Support for kernels:
2178 2.6.5-7.283 (SLES 9),
2179 2.6.9-55.EL (RHEL 4),
2180 2.6.16.46-0.14 (SLES 10),
2181 2.6.18.8 vanilla (kernel.org)
2182 * Client support for unpatched kernels:
2183 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2184 2.6.16 - 2.6.22 vanilla (kernel.org)
2185 * Due to recently discovered recovery problems, we do not recommend
2186 using patchless RHEL 4 clients with this or any earlier release.
2187 * Recommended e2fsprogs version: 1.39.cfs8
2188 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2189 * Starting with this release, the ldiskfs backing filesystem required
2190 by Lustre is now in its own package, lustre-ldiskfs. This package
2191 should be installed. It is versioned separately from Lustre and
2192 may be released separately in future.
2196 Description: block reactivating mgc import until all deactivates complete
2197 Details : Fix race when failing back MDT/MGS to itself (testing)
2200 Frequency : at statup only
2202 Description: mds_lov_synchronize race leads to various problems
2203 Details : simultaneous MDT->OST connections at startup can cause the
2204 sync to abort, leaving the OSC in a bad state.
2206 Severity : enhancement
2208 Description: add optional extra BUILD_VERSION info
2209 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2210 to override the lustre version.
2213 Frequency : 2.6.18 servers only
2215 Description: ll_kern_mount() doesn't release the module reference
2216 Details : The ldiskfs module reference count never drops down to 0
2217 because ll_kern_mount() doesn't release the module reference.
2222 Description: server LBUG when using old ost_num_threads parameter
2223 Details : Accept the old ost_num_threads parameter but warn that it
2224 is deprecated, and fix an off-by-one error that caused an LBUG.
2229 Description: Transient SCSI error results in persistent IO issue
2230 Details : iobuf->dr_error is not reinitialized to 0 between two
2234 Frequency : sometimes when underlying device returns I/O errors
2236 Description: OSTs not going read-only during write failures
2237 Details : OSTs are not remounted read-only when the journal commit threads
2238 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2239 instead of the ext3 wrappers.
2243 Description: poor connect scaling with increasing client count
2244 Details : Don't run filter_grant_sanity_check for more than 100 exports
2245 to improve scaling for large numbers of clients.
2248 Frequency : SLES10 only
2250 Description: sanity-quota.sh quotacheck failed: rc = -22
2251 Details : Quotas cannot be enabled on SLES10.
2254 Frequency : liblustre clients only
2256 Description: getdirentries does not give error when run on compute nodes
2257 Details : getdirentries does not fail when the size specified as an argument
2258 is too small to contain at least one entry
2260 Severity : enhancement
2262 Description: Add LNET router traceability for debug purposes
2263 Details : If a checksum failure occurs with a router as part of the
2264 IO path, the NID of the last router that forwarded the bulk data
2265 is printed so it can be identified.
2270 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2271 Details : Due to a race condition, liblustre clients were occasionally
2272 evicted incorrectly.
2274 Severity : enhancement
2276 Description: lfs setstripe use optional parameters instead of postional
2279 Severity : enhancement
2281 Description: Nanosecond timestamp support for ldiskfs
2282 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2283 resolution timestamps. There is not yet support for this at
2284 the Lustre filesystem level.
2287 Frequency : during server recovery
2289 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2290 Details : request with rq_no_resend flag not awake l_wait_event if they get
2294 Frequency : nfs export on patchless client
2296 Description: connectathon hang when test nfs export over patchless client
2297 Details : Disconnected dentry cannot be found with lookup, so we do not need
2298 to unhash it or make it invalid
2301 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2302 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2303 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2308 Description: open req refcounting wrong on reconnect
2309 Details : If reconnect happened between getting open reply from server and
2310 call to mdc_set_replay_data in ll_file_open, we will schedule
2311 replay for unreferenced request that we are about to free.
2312 Subsequent close will crash in variety of ways.
2313 Check that request is still eligible for replay in
2314 mdc_set_replay_data().
2319 Description: disable writes to filesystem when reading health_check file
2320 Details : the default for reading the health_check proc file has changed
2321 to NOT do a journal transaction and write to disk, because this
2322 can cause reads of the /proc file to hang and block HA state
2323 checking on a healthy but otherwise heavily loaded system. It
2324 is possible to return to the previous behaviour during configure
2325 with --enable-health-write.
2327 Severity : enhancement
2329 Description: 64-bit inode version
2330 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2331 to the inode. This will be required for version-based recovery.
2336 Description: MDS fails to start if a duplicate client export is detected
2337 Details : in some rare cases it was possible for a client to connect to
2338 an MDS multiple times. Upon recovery the MDS would detect this
2339 and fail during startup. Handle this more gracefully.
2341 Severity : enhancement
2343 Description: Add -o localflock option to simulate old noflock
2345 Details : This will achieve local-only flock/fcntl locks
2351 Description: log_commit_thread vs filter_destroy race leads to crash
2352 Details : Take import reference before releasing llog record semaphore
2357 Description: Wrong request locking in request set processing
2358 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2359 imp_delayed_list, in this place should be used imp_lock.
2362 Frequency : when reconnection
2364 Description: Grant Leak when osc reconnect to OST
2365 Details : When osc reconnect ost, OST(filter) should check whether it
2366 should grant more space to client by comparing fed_grant and
2367 cl_avail_grant, and return the granted space to client instead
2368 of "new granted" space, because client will call osc_init_grant
2369 to update the client grant space info.
2372 Frequency : when client reconnect to OST
2374 Description: Grant Leak when osc do resend and replay bulk write
2375 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2376 bulk write request, because the grant info while be sync between
2377 OSC and OST when reconnect, and we should ignore the grant info
2378 these of resend/replay write req.
2383 Description: Grant space more than avaiable left space sometimes.
2384 Details : When then OST is about to be full, if two bulk writing from
2385 different clients came to OST. Accord the avaliable space of the
2386 OST, the first req should be permitted, and the second one
2387 should be denied by ENOSPC. But if the seconde arrived before
2388 the first one is commited. The OST might wrongly permit second
2389 writing, which will cause grant space > avaiable space.
2392 Frequency : when client is evicted
2394 Description: Grant might be wrongly erased when osc is evicted by OST
2395 Details : when the import is evicted by server, it will fork another
2396 thread ptlrpc_invalidate_import_thread to invalidate the
2397 import, where the grant will be set to 0. While the original
2398 thread will update the grant it got when connecting. So if
2399 the former happened latter, the grant will be wrongly errased
2400 because of this race.
2405 Description: Checking Stale with correct fid
2406 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2407 to check whether it is stale, because sometimes, we want the
2408 enqueue happened anyway, and op_data.fid2 will not be initialized.
2410 Severity : enhancement
2412 Description: update patchless client
2413 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2416 Frequency : only with 2.4 kernel
2418 Description: random memory corruption
2419 Details : size of struct ll_inode_info is to big for union inode.u and this
2420 can be cause of random memory corruption.
2425 Description: Memory leak in recovery
2426 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2427 It should also check obd_fail before fsfilt_start, otherwise if
2428 fsfilt_start return -EROFS,(failover mds during mds recovery).
2429 then the req will return with repmsg->transno = 0 and rc = EROFS.
2430 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2431 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2432 be freed no matter whether fsfilt_commit success or not.
2435 Frequency : only with huge count clients
2437 Description: Prevents from taking the superblock lock in llap_from_page for
2439 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2440 for llap_from_page call in ll_removepage prevents from taking the
2441 superblock lock for a soon died page.
2446 Description: Not check open intent error before release open handle
2447 Details : in some rare cases, the open intent error is not checked before
2448 release open handle, which may cause
2449 ASSERTION(open_req->rq_transno != 0), because it tries to release
2450 the failed open handle.
2455 Description: Set cat log bitmap only after create log success.
2456 Details : in some rare cases, the cat log bitmap is set too early. and it
2457 should be set only after create log success.
2461 Description: Accessing a block bevice can re-enable I/O when Lustre is
2462 tearing down a device.
2463 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2468 Description: service threads may hog cpus when there are a lot of requests
2469 Details : Insert cond_resched to give other threads a chance to use some CPU
2474 Description: the cat log was not initialized in recovery
2475 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2476 unlink log on mds will not be initialized until mds post
2477 recovery. And also in mds post recovery, the unlink log will
2478 initialization will be done asynchronausly, so there will be race
2479 between add unlink log and unlink log initialization.
2483 Description: brw_stats were being printed incorrectly
2484 Details : brw_stats were being printed as log2 but all of them were not
2485 recorded as log2. Also remove some code duplication arising from
2486 filter_tally_{read,write}.
2490 Frequency : rare, only in recovery.
2491 Description: ASSERTION(req->rq_type != LI_POISON) failed
2492 Details : imp_lock should be held while iterating over imp_sending_list for
2493 prevent destroy request after get timeout in ptlrpc_queue_wait.
2497 Description: replay-single.sh test 52 fails
2498 Details : A lock's skiplist need to be cleanup when it being unlinked
2499 from its resource list.
2503 Description: Short directio read returns full requested size rather than
2505 Details : Direct I/O operations should return actual amount of bytes
2506 transferred rather than requested size.
2508 Severity : enhancement
2510 Description: metadata RPC reduction (e.g. for rm performance)
2511 Details : decrease the amount of synchronous RPC between clients and servers
2512 by canceling conflicing lock before the operation on the client
2513 and packing thier handles into the main operation RPC to server.
2515 Severity : enhancement
2517 Description: Async OSC create to avoid the blocking unnecessarily.
2518 Details : If a OST has no remain object, system will block on the creating
2519 when need to create a new object on this OST. Now, ways use
2520 pre-created objects when available, instead of blocking on an
2521 empty osc while others are not empty. If we must block, we block
2522 for the shortest possible period of time.
2526 Description: improve handling recoverable errors
2527 Details : if request processig with error which can be recoverable on server
2528 request should be resend, otherwise page released from cache and
2531 Severity : enhancement
2533 Description: refine locking for avoid write wrong info into lov_objid file
2534 Details : fix possible races with add new target and write/update data in
2537 --------------------------------------------------------------------------------
2539 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2544 Frequency : on some architectures
2546 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2547 Details : Uninitialized flags sometimes cause configuration commands to
2551 Frequency : patchless clients only
2553 Description: missing __iget() symbol export
2554 Details : The __iget() symbol export is missing. To avoid the need for
2555 this on patchless clients the deathrow inode reaper is turned
2556 off, and we depend on the VM to clean up old inodes. This
2557 dependency was during via the fix for bug 12181.
2559 --------------------------------------------------------------------------------
2561 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2563 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2564 INTEROPERATE with older versions automatically. In many cases a
2565 special upgrade step is needed. Please read the
2566 user documentation before upgrading any part of a live system.
2567 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2568 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2569 * WARNING: Lustre configuration and startup changes are required with
2570 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2574 Severity : enhancement
2576 Description: MountConf
2577 Details : Lustre configuration is now managed via mkfs and mount
2578 commands instead of lmc and lconf. New obd types (MGS, MGC)
2579 are added for dynamic configuration management. See
2580 https://mail.clusterfs.com/wikis/lustre/MountConf for
2583 Severity : enhancement
2585 Description: dynamic OST addition
2586 Details : OSTs can now be added to a live filesystem
2588 Severity : enhancement
2590 Description: startup order invariance
2591 Details : MDTs and OSTs can be started in any order. Clients only
2592 require the MDT to complete startup.
2594 Severity : enhancement
2596 Description: parallel, asynchronous orphan cleanup
2597 Details : orphan cleanup is now performed in separate threads for each
2598 OST, allowing parallel non-blocking operation.
2600 Severity : enhancement
2602 Description: optimized stripe assignment
2603 Details : stripe assignments are now made based on ost space available,
2604 ost previous usage, and OSS previous usage, in order to try
2605 to optimize storage space and networking resources.
2607 Severity : enhancement
2609 Description: Permanently set tunables
2610 Details : All writable /proc/fs/lustre tunables can now be permanently
2611 set on a per-server basis, at mkfs time or on a live system.
2613 Severity : enhancement
2615 Description: Lustre message v2
2616 Details : Add lustre message format v2.
2618 Severity : enhancement
2620 Description: client OST exclusion list
2621 Details : Clients can be started with a list of OSTs that should be
2622 declared "inactive" for known non-responsive OSTs.
2626 Description: SPEC SFS validation failure on NFS v2 over lustre.
2627 Details : Changes the blocksize for regular files to be 2x RPC size,
2628 and not depend on stripe size.
2630 Severity : enhancement
2632 Description: Multiple MD RPCs in flight.
2633 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2634 To avoid overly-overloading MDS, introduce a limit on number of
2635 MDS RPCs in flight for a single client and add /proc controls
2636 to adjust this limit.
2638 Severity : enhancement
2640 Description: client read/write statistics
2641 Details : Add client read/write call usage stats for performance
2642 analysis of user processes.
2643 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2644 file access. extents_stats shows chunk size distribution.
2645 extents_stats_per_process show chunk size distribution per
2648 Severity : enhancement
2650 Description: mds statistics
2651 Details : Add detailed mds operations statistics in
2652 /proc/fs/lustre/mds/*/stats.
2656 Description: Failure of copying files with lustre special EAs.
2657 Details : Client side always return success for setxattr call for lustre
2658 special xattr (currently only "trusted.lov").
2663 Description: Refcount LNET uuids
2664 Details : The global LNET uuid list grew linearly with every startup;
2665 refcount repeated list entries instead of always adding to
2668 Severity : enhancement
2670 Description: Dynamic service threads
2671 Details : Within a small range, start extra service threads
2672 automatically when the request queue builds up.
2675 Frequency : mixed-endian client/server environments
2677 Description: mixed-endian crashes
2678 Details : The new msg_v2 system had some failures in mixed-endian
2681 Severity : enhancement
2683 Description: Easy OST removal
2684 Details : OSTs can be permanently deactivated with e.g. 'lctl
2685 conf_param lustre-OST0001.osc.active=0'
2687 Severity : enhancement
2689 Description: MGS proc entries
2690 Details : Added basic proc entries for the MGS showing what filesystems
2693 Severity : enhancement
2695 Description: provide MGS failover
2696 Details : Added config lock reacquisition after MGS server failover.
2698 Severity : enhancement
2700 Description: add Linux 2.4 support
2701 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2705 Description: a large application tries to do I/O to the same resource and dies
2706 in the middle of it.
2707 Details : Check the req->rq_arrival time after the call to
2708 ost_brw_lock_get(), but before we do anything about
2709 processing it & sending the BULK transfer request. This
2710 should help move old stale pending locks off the queue as
2711 quickly as obd_timeout.
2714 Frequency : when an incorrect nid is specified during startup
2716 Description: ptlrpc connect to non-existant node causes kernel crash
2717 Details : LNET can't be re-entered from an event callback, which
2718 happened when we expire a message after the export has been
2719 cleaned up. Instead, hand the zombie cleanup off to another
2722 Severity : enhancement
2724 Description: plain/inodebits lock performance improvement
2725 Details : Grouping plain/inodebits in granted list by their request modes
2726 and bits policy, thus improving the performance of search through
2730 Frequency : only if OST filesystem is corrupted
2732 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2733 Details : for a short time RPCs with bulk IO are in the replay list,
2734 but replay of bulk IOs is unimplemented. If the OST filesystem
2735 is corrupted due to disk cache incoherency and then replay is
2736 started it is possible to trip an assertion. Avoid putting
2737 committed RPCs into the replay list at all to avoid this issue.
2740 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2742 Description: System hang on startup
2743 Details : This bug allowed the liblustre (e.g. catamount) client to
2744 return to the app before handling all startup RPCs. This
2745 could leave the node unresponsive to lustre network traffic
2746 and manifested as a server ptllnd timeout.
2748 Severity : enhancement
2750 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2751 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2752 boolean to control whether to print peer debug info when a
2753 client's RPC times out.
2756 Frequency : only for kernels with patches from Lustre below 1.4.3
2758 Description: Remove old rdonly API
2759 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2762 Frequency : only for devices with external journals
2764 Description: Set external device read-only also
2765 Details : During a commanded failover stop, we set the disk device
2766 read-only while the server shuts down. We now also set any
2767 external journal device read-only at the same time.
2770 Frequency : when upgrading from 1.4 while trying to change parameters
2772 Description: The wrong (new) MDC name was used when setting parameters for
2773 upgraded MDT's. Also allows changing of OSC (and MDC)
2774 parameters if --writeconf is specified at tunefs upgrade time.
2777 Frequency : when setting specific ost indicies
2779 Description: QOS code breaks on skipped indicies
2780 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2781 created with --index need not be sequential.
2783 Severity : enhancement
2785 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2786 Details : The uninit_groups feature works in conjunction with the kernel
2787 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2788 up the pass1 processing of e2fsck. This is a read-only feature
2789 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2790 mount filesystems that have had this feature enabled.
2792 Severity : enhancement
2794 Description: Improve multi-block allocation algorithm to avoid fragmentation
2795 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2796 allocation locality and avoid filesystem fragmentation.
2798 ------------------------------------------------------------------------------
2800 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2802 * Support for kernels:
2803 2.4.21-47.0.1.EL (RHEL 3)
2804 2.6.5-7.283 (SLES 9)
2805 2.6.9-42.0.10.EL (RHEL 4)
2806 2.6.12.6 vanilla (kernel.org)
2807 2.6.16.27-0.9 (SLES 10)
2808 * Recommended e2fsprogs version: 1.39.cfs5
2810 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2814 Frequency : occasional, depends on client load and configuration
2815 Bugzilla : 12181, 12203
2816 Description: data loss for recently-modified files
2818 Details : In some cases it is possible that recently written or created
2819 files may not be written to disk in a timely manner (this should
2820 normally be within 30s unless client IO load is very high).
2821 The problem appears as zero-length files or files that are a
2822 multiple of 1MB in size after a client crash or client eviction
2823 that are missing data at the end of the file.
2825 This problem is more likely to be hit on clients where files are
2826 repeatedly created and unlinked in the same directory, clients
2827 have a large amount of RAM, have many CPUs, the filesystem has
2828 many OSTs, the clients are rebooted frequently, and/or the files
2829 are not accessed by other nodes after being written.
2831 The presence of the problem can be detected by looking at
2832 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2833 smaller than the second (nr_unused) then dirty files will not
2834 be flushed automatically to disk. "sync; sleep 10" should be
2835 run several times on the node before unmounting it to update
2836 Lustre (this is also safe to run on nodes without this problem).
2838 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2839 that can cause this same problem, so customers using that kernel
2840 also need to update the kernel in addition to Lustre. In order
2841 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2843 It is normal that files written just before a client crash (less
2844 than 30s) may not yet have been flushed to disk, even for local
2848 Frequency : frequent on thin XT3 nodes
2850 Description: UUID collision on thin XT3 Linux nodes
2851 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2852 randomly, since we relied on an insufficiently-seeded PRNG.
2857 Description: OSS hangs after "All ost request buffers busy"
2858 Details : A deadlock between quota and journal operations caused OSS
2859 hangs after printing "All ost request buffers busy."
2862 Frequency : always on liblustre builds
2864 Description: Cleanup compiler warnings on liblustre
2867 Frequency : always on liblustre builds on XT3
2869 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2873 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2874 > 32000 subdirectories
2875 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2876 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2877 subdirectory count crosses 32000. This will aid e2fsck to
2878 correctly handle more than 32000 subdirectories.
2881 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2883 Description: System hang on startup
2884 Details : This bug allowed the liblustre (e.g. catamount) client to
2885 return to the app before handling all startup RPCs. This
2886 could leave the node unresponsive to lustre network traffic
2887 and manifested as a server ptllnd timeout.
2889 Severity : enhancement
2891 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2892 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2893 boolean to control whether to print peer debug info when a
2894 client's RPC times out.
2899 Description: make O_SYNC working on 2.6 kernels
2900 Details : 2.6 kernels use different method for mark pages for write,
2901 so need add a code to lustre for O_SYNC work.
2906 Description: Failure to close file and release space on NFS
2907 Details : Put inode details into lock acquired in ll_intent_file_open.
2908 Use mdc_intent_lock in ll_intent_open to properly
2909 detect all kind of errors unhandled by mdc_enqueue.
2914 Description: proc file read during shutdown sometimes raced obd removal,
2916 Details : Add lock to prevent obd access after proc file removal.
2919 Frequency : Only for files larger than 4GB on 32-bit clients.
2921 Description: improperly doing page alignment of locks
2922 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2923 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2928 Description: RPCs being resent when they shouldn't be
2929 Details : Some RPCs that should not be resent are being resent. This
2930 can cause inconsistencies in the RPC state machine. Do not
2931 resend such requests.
2934 Frequency : rare, only with NFS export
2936 Description: Crash on NFS re-export node
2937 Details : under very unusual load conditions an assertion is hit in
2938 ll_intent_file_open()
2941 Frequency : only if OST filesystem is corrupted
2943 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2944 Details : for a short time RPCs with bulk IO are in the replay list,
2945 but replay of bulk IOs is unimplemented. If the OST filesystem
2946 is corrupted due to disk cache incoherency and then replay is
2947 started it is possible to trip an assertion. Avoid putting
2948 committed RPCs into the replay list at all to avoid this issue.
2953 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2954 Details : Large single O_DIRECT read and write calls can fail to allocate
2955 a sufficiently large buffer to process the request. In case of
2956 allocation failure the allocation is retried with a smaller
2957 buffer and broken into smaller requests.
2959 Severity : enhancement
2961 Description: Add -o localflock option to simulate old noflock behaviour.
2962 Details : This will achieve local-only flock/fcntl locks coherentness.
2967 Description: versioning check is incomplete
2968 Details : Checking the version difference of client vs. server, report
2969 error if the gap is too big.
2974 Description: add support PG_writeback bit
2975 Details : add support for PG_writeback bit for Lustre, for more carefull
2976 work with page cache in 2.6 kernel. This also fix some deadlocks
2977 and remove hack for work O_SYNC with 2.6 kernel.
2979 Severity : enhancement
2981 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2982 Details : The uninit_groups feature works in conjunction with the kernel
2983 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2984 up the pass1 processing of e2fsck. This is a read-only feature
2985 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2986 mount filesystems that have had this feature enabled.
2988 Severity : enhancement
2990 Description: Improve multi-block allocation algorithm to avoid fragmentation
2991 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2992 allocation locality and avoid filesystem fragmentation.
2994 ------------------------------------------------------------------------------
2996 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2998 * Support for kernels:
2999 2.6.9-42.0.3.EL (RHEL 4)
3000 2.6.5-7.276 (SLES 9)
3001 2.4.21-47.0.1.EL (RHEL 3)
3002 2.6.12.6 vanilla (kernel.org)
3003 2.6.16.21-0.8 (SLES10)
3004 * Recommended e2fsprogs version: 1.39.cfs2-0
3006 * The backwards-compatible /proc/sys/portals symlink has been removed
3007 in this release. Before upgrading, please ensure that you change
3008 any configuration scripts or /etc/sysctl.conf files that access
3009 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3010 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3011 in advance of the upgrade on any system running Lustre 1.4.6 or
3012 newer, since /proc/sys/lnet was added in that version.
3013 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3017 Frequency : only when quota is used
3019 Description: avoid scanning export list for quota master
3020 Details : Change the algorithms to avoid scanning export list in order
3021 to improve the efficiency.
3024 Frequency : MDS failover only, very rarely
3026 Description: "went back in time" messages on mds failover
3027 Details : The greatest transno may be lost when the current operation
3028 finishes with an error (transno==0) and the client's last_rcvd
3029 record is over-written. Save the greatest transno in the
3030 mds_last_transno for this case.
3033 Frequency : always for specific kernels and striping counts
3035 Description: client may get "Matching packet too big" without ACL support
3036 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3037 when trying to access files in certain configurations. The
3038 clients should in fact be denied when mounting because they do
3039 not understand ACLs.
3042 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3044 Description: many clients connecting with IO in progress causes connect timeouts
3045 Details : Avoid synchronous journal commits to avoid delays caused by many
3046 clients connecting/disconnecting when bulk IO is in progress.
3047 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3048 OST_IO_PORTAL to avoid delays behind potentially many pending
3052 Frequency : occasionally with multiple writers to a single file
3054 Description: shared writes to file may result in wrong size reported by stat()
3055 Details : Allow growing of kms when extent lock is cancelled
3058 Frequency : always with random mmap IO to multi-striped file
3060 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3061 Details : Only if the hole is at the end of OST object so that kms is too
3062 small. Fix is to increase kms accordingly in ll_nopage.
3065 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3067 Description: writes to a missing object would leak memory on the OST
3068 Details : If there is an inconsistency between the MDS and OST filesystems,
3069 such that the MDS references an object that doesn't exist, writes
3070 to that object will leak memory due to incorrect cleanup in the
3071 error handling path, eventually running out of memory on the OST.
3076 Description: Creating too long symlink causes lustre errors
3077 Details : Check symlink and name lengths before sending requests to MDS.
3080 Frequency : only if flock is enabled (not on by default)
3082 Description: posix locks not released on fd closure on 2.6.9+
3083 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3084 kernels, this caused such locks not to be released on fd close and
3085 then assertions on fs unmount about still used locks.
3088 Frequency : MDS failover only, very rarely
3090 Description: clients may get ASSERTION(granted_lock != NULL)
3091 Details : When request was taking a long time, and a client was resending
3092 a getattr by name lock request. The were multiple lock requests
3093 with the same client lock handle and
3094 mds_getattr_name->fixup_handle_for_resent_request found one of the
3095 lock handles but later failed with ASSERTION(granted_lock != NULL).
3100 Description: handle->h_buffer_credits > 0, assertion failure
3101 Details : h_buffer_credits is zero after truncate, causing assertion
3102 failure. This patch extends the transaction or creates a new
3106 Frequency : NFS re-export or patchless client
3107 Bugzilla : 11179, 10796
3108 Description: Crash on NFS re-export node (__d_move)
3109 Details : We do not want to hash the dentry if we don't have a lock.
3110 But if this dentry is later used in d_move, we'd hit uninitialised
3111 list head d_hash, so we just do this to init d_hash field but
3112 leave dentry unhashed.
3115 Frequency : NFS re-export or patchless client
3117 Description: NFS exports has problem with symbolic link
3118 Details : lustre client didn't properly install dentry when re-exported
3119 to NFS or running patchless client.
3122 Frequency : NFS re-export or patchless client
3124 Description: Various nfs/patchless fixes.
3125 Details : fixes reuse disconected alias for lookup process - this fixes
3126 warning "find_exported_dentry: npd != pd",
3127 fix permission error with open files at nfs.
3128 fix apply umask when do revalidate.
3131 Frequency : occasional
3133 Description: Crash on NFS re-export node
3134 Details : calling clear_page() on the wrong pointer triggered oops in
3135 generic_mapping_read().
3138 Frequency : rarely, using O_DIRECT IO
3140 Description: unaligned directio crashes client with LASSERT
3141 Details : check for unaligned buffers before trying any requests.
3144 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3146 Description: stale data returned from RAID cache
3147 Details : If only a small amount of IO is done to the RAID device before
3148 reading it again it is possible to get stale data from the RAID
3149 cache instead of reading it from disk.
3152 Frequency : always for sles10 kernel
3154 Description: sles10 support
3155 Details : ll_follow_link: compile fixes and using of nd_set_link
3156 under newer kernels.
3159 Frequency : depends on arch, kernel and compiler version, always on sles10
3162 Description: recursive or deep enough symlinks cause stack overflow
3163 Details : getting rid of large stack-allocated variable in
3167 Frequency : depends on hardware
3169 Description: lustre write performance loss in the SLES10 kernel
3170 Details : the performance loss is caused by using of write barriers in the
3171 ext3 code. The SLES10 kernel turns barrier support on by
3172 default. The fix is to undo that change for ldiskfs.
3174 ------------------------------------------------------------------------------
3176 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3178 * Support for kernels:
3179 2.6.9-42.0.3EL (RHEL 4)
3180 2.6.5-7.276 (SLES 9)
3181 2.4.21-47.0.1.EL (RHEL 3)
3182 2.6.12.6 vanilla (kernel.org)
3186 Frequency : quota enabled and large files being deleted
3188 Description: releasing more than 4GB of quota at once hangs OST
3189 Details : If a user deletes more than 4GB of files on a single OST it
3190 will cause the OST to spin in an infinite loop. Release
3191 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3196 Description: statfs data retrieved from /proc may be stale or zero
3197 Details : When reading per-device statfs data from /proc, in the
3198 {kbytes,files}_{total,free,avail} files, it may appear
3199 as zero or be out of date.
3202 Frequency : systems with MD RAID1 external journal devices
3204 Description: lconf's call to blkid is confused by RAID1 journal devices
3205 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3206 of returning all block devices that match the journal UUID.
3209 Frequency : always, for aggregate stripe size over 4GB
3211 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3212 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3213 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3216 Frequency : always if "lfs find" used on a local file/directory
3218 Description: "lfs find" segfaults if used on a local file/directory
3219 Details : The case where a directory component was not specified wasn't
3220 handled correctly. Handle this properly.
3223 Frequency : always on ppc64
3225 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3226 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3227 perform endianness conversion.
3230 Frequency : rarely (truncate to non-zero file size after write under load)
3231 Bugzilla : 10730, 10687
3232 Description: Files padded with zeros to next 4K multiple
3233 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3234 it is possible that files that are truncated to a non-zero size
3235 immediately after being written are filled with zero bytes beyond
3236 the truncated size. No file data is lost.
3238 Severity : enhancement
3240 Description: Allow recovery/failover for liblustre clients.
3241 Details : liblustre clients were unaware of failover configurations until
3244 Severity : enhancement
3246 Description: user file locks should fail when not mounting with flock option
3247 Details : Set up an error-returning stub in ll_file_operations.lock field
3248 to prevent incorrect behaviour when client is mounted without
3249 flock option. Also, set up properly f_op->flock field for
3253 Frequency : always on ia64
3255 Description: "lfs df" loops on printing out MDS statfs information
3256 Details : The obd_ioctl_data was not initialized and in some systems
3257 this caused a failure during the ioctl that did not return
3258 an error. Initialize the struct and return an error on failure.
3261 Frequency : SLES 9 only
3263 Description: Error of copying files with lustre special EAs as root
3264 Details : Client side always return success for setxattr call for lustre
3265 special xattr (currently only "trusted.lov").
3268 Frequency : rarely on clusters with both ia64+i386 clients
3270 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3271 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3272 clients are doing concurrent writes to the same file it is
3273 possible that the ia64 clients may LASSERT because the OST
3274 extent locks are not PAGE_SIZE aligned. Ensure that grown
3275 locks are always aligned on the request boundary.
3278 Frequency : specific use, occasional
3280 Description: Overwriting in use executable truncates on-disk binary image
3281 Details : If one node attempts to overwrite an executable in use by
3282 another node, we now correctly return ETXTBSY instead of
3283 truncating the file.
3285 Severity : enhancement
3287 Description: Async OSC create to avoid the blocking unnecessarily.
3288 Details : If a OST has no remain object, system will block on the creating
3289 when need to create a new object on this OST. Now, ways use
3290 pre-created objects when available, instead of blocking on an
3291 empty osc while others are not empty. If we must block, we block
3292 for the shortest possible period of time.
3297 Description: chmod on Lustre root is propagated to other clients
3298 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3299 invalid by the follow_mount time.
3302 Frequency : liblustre clients only
3304 Description: Race in 'instant cancel' lock handling could lead to such locks
3305 never to be granted in case of SMP MDS
3306 Details : Do not destroy not yet granted but cbpending locks in
3310 Frequency : replay/resend of open
3312 Description: non null lock assetion failure in mds_intent_policy
3313 Details : Trying to replay/resend lockless open requests resulted in
3314 mds_open() returning 0 with no lock. Now it sets a flag if
3315 a lock is going to be returned.
3317 Severity : enhancement
3319 Description: Checksum enhancements
3320 Details : New checksum enhancements allow for resending RPCs that failed
3323 Severity : enhancement
3325 Description: Tunables on number of dirty pages in cacche
3326 Details : Allow to set limit on number of dirty pages cached.
3331 Description: client crash on unmount - lock still has references
3332 Details : In some error handling cases it was possible to leak a lock
3333 reference on a client while accessing a file. This was not
3334 harmful to the client during operation, but would cause the
3335 client to crash when the filesystem is unmounted.
3338 Frequency : specific case, rare
3340 Description: ETXTBSY on mds though file not in use by client
3341 Details : ETXTBSY is no longer incorrectly returned when attempting to
3342 chmod or chown a directory that the user previously tried to
3343 execute or a currently-executing binary.
3346 Frequency : extremely rare except on liblustre-based clients
3348 Description: Lustre space not freed when files are deleted
3349 Details : Clean up open-unlinked files after client eviction. Previously
3350 the unlink was skipped and the files remained as orphans.
3355 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3356 Details : In some cases it was possible to send a blocking callback to
3357 a client doing a glimpse, even though that client didn't get
3358 a lock granted. When the glimpse lock is cancelled on the OST
3359 the freed lock is left on the waiting list and corrupted the list.
3362 Frequency : all core dumps
3364 Description: Broke core dumps to lustre
3365 Details : Negative dentry may be unhashed if parent does not have UPDATE
3366 lock, but some callers, e.g. do_coredump, expect dentry to be
3367 hashed after successful create, hash it in ll_create_it.
3369 ------------------------------------------------------------------------------
3371 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3373 * Support for kernels:
3374 2.6.9-42.0.2.EL (RHEL 4)
3375 2.6.5-7.276 (SLES 9)
3376 2.4.21-40.EL (RHEL 3)
3377 2.6.12.6 vanilla (kernel.org)
3381 Frequency : always on RHEL 3
3383 Description: Number of open files grows over time
3384 Details : The number of open files grows over time, whether or not
3385 Lustre is started. This was due to a filp leak introduced
3386 by one of our kernel patches.
3388 ------------------------------------------------------------------------------
3390 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3392 * Support for kernels:
3393 2.6.9-42.EL (RHEL 4)
3394 2.6.5-7.276 (SLES 9)
3395 2.4.21-40.EL (RHEL 3)
3396 2.6.12.6 vanilla (kernel.org)
3401 Bugzilla : 5719, 9635, 9792, 9684
3402 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3403 Details : If a server is under heavy load and cannot reply to new
3404 connection requests before the client resends the (re)connect,
3405 the connection handling code can behave badly if two service
3406 threads are concurrently handing separate (re)connections from
3407 the same client. Add better locking to the connection handling
3408 code, and ensure that only a single connection will be processed
3409 for a given client UUID, even if the lock is dropped.
3411 Severity : enhancement
3413 Description: add TCP zero-copy support to kernel
3414 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3415 sends if the hardware supports scatter-gather and checksumming.
3416 This allows socklnd to do client-write and server-read more
3417 efficiently and reduce CPU utilization from skbuf copying.
3420 Frequency : only if NFS exporting from client
3422 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3423 Details : If setting attributes on a file created under NFS that had
3424 never been opened it would be possible to oops the client
3425 if the file had no objects.
3428 Frequency : always for liblustre
3430 Description: liblustre client does MDS+OSTs setattr RPC for each write
3431 Details : When doing a write from a liblustre client, the client
3432 incorrectly issued an RPC to the MDS and each OST the file was
3433 striped over in order to update the timestamps. When writing
3434 with small chunks and many clients this could overwhelm the MDS
3435 with RPCs. In all cases it would slow down the write because
3436 these RPCs are unnecessary.
3438 Severity : enhancement
3440 Description: allow number of MDS service threads to be changed at module load
3441 Details : It is now possible to change the number of MDS service threads
3442 running. Adding "options mds mds_num_threads={N}" to the MDS's
3443 /etc/modprobe.conf will set the number of threads for the next
3444 time Lustre is restarted (assuming the "mds" module is also
3445 reloaded at that time). The default number of threads will
3446 stay the same, 32 for most systems.
3451 Description: OST crash if filesystem is unformatted or corrupt
3452 Details : If an OST is started on a device that has never been formatted
3453 or if the filesystem is corrupt and cannot even mount then the
3454 error handling cleanup routines would dereference a NULL pointer.
3459 Description: NULL pointer deref in llap_from_page.
3460 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3461 incorrect) mapping if the page was truncated/reclaimed while it was
3462 searched for. Check for this condition and skip such pages when
3463 doing readahead. Introduce extra check to llap_from_page() to
3464 verify page->mapping->host is non-NULL (so page is not anonymous).
3467 Frequency : Sometimes when using sys_sendfile
3469 Description: "page not covered by a lock" warnings from ll_readpage
3470 Details : sendfile called ll_readpage without right page locks present.
3471 Now we introduced ll_file_sendfile that does necessary locking
3472 around call to generic_file_sendfile() much like we do in
3476 Frequency : with certain MDS communication failures at client mount time
3478 Description: NULL pointer deref after failed client mount
3479 Details : a client connection request may delayed by the network layer
3480 and not be sent until after the PTLRPC layer has timed out the
3481 request. If the client fails the mount immediately it will try
3482 to clean up before the network times out the request. Add a
3483 reference from the request import to the obd device and delay
3484 the cleanup until the network drops the request.
3487 Frequency : occasionally during client (re)connect
3489 Description: assertion failure during client (re)connect
3490 Details : processing a client connection request may be delayed by the
3491 client or server longer than the client connect timeout. This
3492 causes the client to resend the connection request. If the
3493 original connection request is replied in this interval, the
3494 client may trip an assertion failure in ptlrpc_connect_interpret()
3495 which thought it would be the only running connect process.
3498 Frequency : only with obd_echo servers and clients that are rebooted
3500 Description: kernel BUG accessing uninitialized data structure
3501 Details : When running an obd_echo server it did not start the ping_evictor
3502 thread, and when a client was evicted an uninitialized data
3503 structure was accessed. Start the ping_evictor in the RPC
3504 service startup instead of the OBD startup.
3506 Severity : enhancement
3507 Bugzilla : 10193 (patchless)
3508 Description: Remove dependency on various unexported kernel interfaces.
3509 Details : No longer need reparent_to_init, exit_mm, exit_files,
3510 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3513 Frequency : rare (only users of deprecated and unsupported LDAP config)
3515 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3516 Details : LDAP apparently contains 'lustreName' attributes instead of
3517 'name'. A simple remapping of the name is sufficient.
3520 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3522 Description: waiting_locks_callback trips kernel BUG if client is evicted
3523 Details : Running with the dump_on_timeout debug flag turned on makes
3524 it possible that the waiting_locks_callback() can try to dump
3525 the Lustre kernel debug logs from an interrupt handler. Defer
3526 this log dumping to the expired_lock_main() thread.
3528 Severity : enhancement
3530 Description: Support NFS exporting on 2.6 kernels.
3531 Details : Implement non-rawops metadata methods for NFS server to use without
3532 changing NFS server code.
3535 Frequency : very rare (synthetic metadata workload only)
3537 Description: two racing renames might cause an MDS thread to deadlock
3538 Details : Running the "racer" program may cause one MDS thread to rename
3539 a file from being the source of a rename to being the target of
3540 a rename at exactly the same time that another thread is doing
3541 so, and the second thread has already enqueued these locks after
3542 doing a lookup of the target and is trying to relock them in
3543 order. Ensure that we don't try to re-lock the same resource.
3546 Frequency : only very large systems with liblustre clients
3548 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3549 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3550 all OSTs in parallel. This allows the eviction of stale liblustre
3551 clients to proceed much faster than if they were done in series,
3552 and also offers similar improvements for other set_info RPCs.
3557 Description: excessive CPU usage during initial read phase on client
3558 Details : During the initial read phase on a client, it would agressively
3559 retry readahead on the file, consuming too much CPU and impacting
3560 performance (since 1.4.5.8). Improve the readahead algorithm
3561 to avoid this, and also improve some other common cases (read
3562 of small files in particular, where "small" is files smaller than
3563 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3568 Description: MDS crash when receiving packet with unknown intent.
3569 Details : Do not LBUG in unknown intent case, just return -EFAULT
3571 Severity : enhancement
3572 Bugzilla : 9293, 9385
3573 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3574 Details : Do not serialize getattr (non-intent version) and statfs.
3577 Frequency : occasional, when OST network is overloaded/intermittent
3579 Description: client evicted by OST after bulk IO timeout
3580 Details : If a client sends a bulk IO request (read or write) the OST
3581 may evict the client if it is unresposive to its data GET/PUT
3582 request. This is incorrect if the network is overloaded (takes
3583 too long to transfer the RPC data) or dropped the OST GET/PUT
3584 request. There is no need to evict the client at all, since
3585 the pinger and/or lock callbacks will handle this, and the
3586 client can restart the bulk request.
3589 Frequency : Always when mmapping file with no objects
3591 Description: client crashes when mmapping file with no objects
3592 Details : Check that we actually have objects in a file before doing any
3593 operations on objects in ll_vm_open, ll_vm_close and
3599 Description: Request leak when working with deleted CWD
3600 Details : Introduce advanced request refcount tracking for requests
3601 referenced from lustre intent.
3603 Severity : Enhancement
3605 Description: Cache open file handles on client.
3606 Details : MDS now will return special lock along with openhandle, if
3607 requested and client is allowed to hold openhandle, even if unused,
3608 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3609 closing files for every read/write openration.
3611 Severity : Enhancement
3613 Description: Cache open negative dentries on client when possible.
3614 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3615 negative dentries on lock revocation.
3620 Description: Remounting a client read-only wasn't possible with a zconf mount
3621 Details : It wasn't possible to remount a client read-only with llmount.
3623 Severity : enhancement
3624 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3625 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3626 MPICH is included to add Lustre-specific ADIO interfaces.
3627 This is based closely on the UFS ADIO layer and only differs
3628 in file creation, in order to allow the OST striping to be set.
3629 This is user-contributed code and not supported by CFS.
3634 Description: extended inode attributes (immutable, append-only) work improperly
3635 when 2.4 and 2.6 kernels are used on client/server or vice versa
3636 Details : Introduce kernel-independent values for these flags.
3638 Severity : enhancement
3641 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3642 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3643 granularity. Now, Fractional MB granularity is be supported,
3644 this is very useful for low memory system.
3646 Severity : enhancement
3648 Description: Getattr by fid
3649 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3650 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3651 have locks and avoids getattr rpc requests for it on every stat.
3654 Frequency : Always, for filesystems larger than 2TB
3656 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3657 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3658 filesystems larger than 2TB to be mounted with mballoc enabled.
3661 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3663 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3664 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3665 filesystems larger than 2TB is mounted then the kernel will
3666 silently corrupt the start of the filesystem. CONFIG_LBD is
3667 enabled for all CFS-supported kernels, but the possibility of
3668 this happening with a modified kernel config exists.
3670 Severity : enhancement
3672 Description: add client O_DIRECT support for 2.6 kernels
3673 Details : It is now possible to do O_DIRECT reads and writes to files
3674 in the Lustre client mountpoint on 2.6 kernel clients.
3676 Severity : enhancement
3678 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3679 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3680 parallel, not waiting for response from every OST before sending
3681 a rpc to the next OST.
3686 Description: setattr vs write race when updating file timestamps
3687 Details : Client processes that update a file timestamp into the past
3688 right after writing to the file (e.g. tar) it is possible that
3689 the updated file modification time can be reset to the current
3690 time due to a race between processing the setattr and write RPC.
3692 Severity : enhancement
3694 Description: Bring 'lfs find' closer in line with regular Linux find.
3695 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3696 -print0 options and obtains all the needed info through the lustre
3699 Severity : enhancement
3701 Description: support up to 1024 configured devices on one node
3702 Details : change obd_dev array from statically allocated to dynamically
3703 allocated structs as they are first used to reduce memory usage
3708 Description: Flush dirty partially truncated pages during truncate
3709 Details : Immediatelly flush partially truncated pages in filter_setattr,
3710 this way we completely avoid having any pages in page cache on OST
3711 and can retire ugly workarounds during writes to flush such pages.
3716 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3717 Details : protect inode from truncation within vfs_unlink() context
3718 just take a reference before calling vfs_unlink() and release it
3719 when parent's i_sem is free.
3724 Description: last_id value checked outside lock on OST caused LASSERT failure
3725 Details : If there were multiple MDS->OST object precreate requests in
3726 flight, it was possible that the OST's last object id was checked
3727 outside a lock and incorrectly tripped an assertion. Move checks
3728 inside locks, and discard old precreate requests.
3731 Frequency : always, if extents are used on OSTs
3733 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3734 Details : OSTs using the extents format would not zero the high 16 bits of
3735 the index physical block number. This is not a problem for any
3736 OST filesystems smaller than 16TB, and no kernels support ext3
3737 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3738 new/modified files) and can be fixed for existing filesystems
3739 with e2fsprogs-1.39-cfs1.
3744 Description: import connection selection may be incorrect if timer wraps
3745 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3746 import connections to be ignored if the 32-bit jiffies counter
3747 wraps. Use a 64-bit jiffies counter.
3750 Frequency : very large clusters immediately after boot
3752 Description: LNET request buffers exhausted under heavy short-term load
3753 Details : If a large number of client requests are generated on a service
3754 that has previously never seen so many requests it is possible
3755 that the request buffer growth cannot keep up with the spike in
3756 demand. Instead of dropping incoming requests, they are held in
3757 the LND until the RPC service can accept more requests.
3760 Frequency : Sometimes during replay
3762 Description: Assertion failure in ll_local_open after replay.
3763 Details : If replay happened on an open request reply before we were able
3764 to set replay handler, reply will become not swabbed tripping the
3765 assertion in ll_local_open. Now we set the handler right after
3766 recognising of open request
3769 Frequency : very rare
3771 Description: kernel reports "badness in vsnprintf"
3772 Details : Reading from the "recovery_status" /proc file in small chunks
3773 may cause a negative length in lprocfs_obd_rd_recovery_status()
3774 call to vsnprintf() (which is otherwise harmless). Exit early
3775 if there is no more space in the output buffer.
3777 Severity : enhancement
3779 Description: clear OBD RPC statistics by writing to them
3780 Details : It is now possible to clear the OBD RPC statistics by writing
3781 to the "stats" file.
3786 Description: Client mtime is not the same on different clients after utimes
3787 Details : In some cases, the client was using the utimes() syscall on
3788 a file cached on another node. The clients now validate the
3789 ctime from the MDS + OSTs to determine which one is right.
3794 Description: Inability to activate failout mode
3795 Details : lconf script incorrectly assumed that in pythong string's numeric
3796 value is used in comparisons.
3799 Frequency : always with multiple stripes per file
3801 Description: Inefficient object allocation for mutli-stripe files
3802 Details : When selecting which OSTs to stripe files over, for files with
3803 a stripe count that divides evenly into the number of OSTs,
3804 the MDS is always picking the same starting OST for each file.
3805 Return the OST selection heuristic to the original design.
3810 Description: mount failures may take full timeout to return an error
3811 Details : Under some heavy load conditions it is possible that a
3812 failed mount can wait for the full obd_timeout interval,
3813 possibly several minutes, before reporting an error.
3814 Instead return an error as soon as the status is known.
3816 Frequency : quota enabled and large files being deleted
3818 Description: releasing more than 4GB of quota at once hangs OST
3819 Details : If a user deletes more than 4GB of files on a single OST it
3820 will cause the OST to spin in an infinite loop. Release
3821 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3826 Description: statfs data retrieved from /proc may be stale or zero
3827 Details : When reading per-device statfs data from /proc, in the
3828 {kbytes,files}_{total,free,avail} files, it may appear
3829 as zero or be out of date.
3832 Frequency : systems with MD RAID1 external journal devices
3834 Description: lconf's call to blkid is confused by RAID1 journal devices
3835 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3836 of returning all block devices that match the journal UUID.
3839 Frequency : always, for aggregate stripe size over 4GB
3841 Description: assertion fails when trying to use 4GB stripe size
3842 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3843 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3846 Frequency : always on ppc64
3848 Description: the first write on an ext3 filesystem with mballoc got stuck
3849 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3850 perform endianness conversion.
3852 ------------------------------------------------------------------------------
3854 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3856 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3857 INTEROPERATE with older versions automatically. Please read the
3858 user documentation before upgrading any part of a live system.
3859 * WARNING: Lustre networking configuration changes are required with
3860 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3863 * Support for kernels:
3864 2.6.9-22.0.2.EL (RHEL 4)
3865 2.6.5-7.244 (SLES 9)
3866 2.6.12.6 vanilla (kernel.org)
3869 Severity : enhancement
3870 Bugzilla : 7981/8208
3871 Description: Introduced Lustre Networking (LNET)
3872 Details : LNET is new networking infrastructure for Lustre, it includes
3873 a reorganized network configuration mode (see the user
3874 documentation for full details) as well as support for routing
3875 between different network fabrics. Lustre Networking Devices
3876 (LNDs) for the supported network fabrics have also been
3877 created for this new infrastructure.
3879 Severity : enhancement
3880 Description: Introduced Access control lists
3881 Details : clients can set ACLs on files and directories in order to have
3882 more fine-grained permissions than the standard Unix UGO+RWX.
3883 The MDS must be started with the "-o acl" mount option.
3885 Severity : enhancement
3886 Description: Introduced filesystem quotas
3887 Details : Administrators may now establish per-user quotas on the
3890 Severity : enhancement
3892 Description: Configuration change for the XT3
3893 The PTLLND is now used to run Lustre over Portals on the XT3
3894 The configure option(s) --with-cray-portals are no longer used.
3895 Rather --with-portals=<path-to-portals-includes> is used to
3896 enable building on the XT3. In addition to enable XT3 specific
3897 features the option --enable-cray-xt3 must be used.
3902 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3903 Details : A race between allocating a new llcd and re-getting the llcd_lock
3904 allowed another thread to grab newly-allocated llcd.
3906 Severity : enhancement
3908 Description: 2.6 OST async journal commit and locking fix to improve performance
3909 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3910 2.6 kernels are now async as they already were in 2.4 kernels so
3911 that they can commit concurrently with the network bulk transfer.
3912 For block-allocated files the filter allocation semaphore is held
3913 to avoid filesystem fragmentation during allocation. BKL lock
3914 removed for 2.6 xattr operations where it is no longer needed.
3919 Description: lconf incorrectly determined whether two IP networks could talk
3920 Details : In some more complicated routing and multiple-network
3921 configurations, lconf will avoid trying to make a network
3922 connection to a disjoint part of the IP space. It was doing the
3923 math incorrectly for one set of cases.
3928 Description: Fix for potential infinite loop processing records in an llog.
3929 Details : If an llog record is corrupted/zeroed, it is possible to loop
3930 forever in llog_process(). Validate the llog record length
3931 and skip the remainder of the block on error.
3934 Frequency : occasional (liblustre only)
3936 Description: liblustre could not open files whose last component is a symlink
3937 Details : sysio_path_walk() would incorrectly pass the open intent to
3938 intermediate path components.
3941 Frequency : rare (liblustre only with non-standard tuning)
3942 Bugzilla : 7201 (7350)
3943 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3944 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3945 after it has been released, only do this for referenced locks
3947 Severity : enhancement
3949 Description: specifying an (invalid) directory default stripe_size of -1
3950 would reset the directory default striping
3951 Details : stripe_size -1 was used internally to signal directory stripe
3952 removal, now use "all default" to signal dir stripe removal
3953 as a directory striping of "all default" is not useful
3956 Frequency : common for large clusters running liblustre clients
3958 Description: doing an ls when liblustre clients are running is slow
3959 Details : sending a glimpse AST to a liblustre client waits for every AST
3960 to time out, as liblustre clients will not respond. Since they
3961 cannot cache data we refresh the OST lock LVB from disk instead.
3963 Severity : enhancement
3965 Description: doing an ls at the same time as file IO can be slow
3966 Details : enqueue and other "small" requests can be blocked behind many
3967 large IO requests. Create a new OST IO portal for non-IO
3968 requests so they can be processed faster.
3971 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3973 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3974 created file with mode 000
3975 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3976 allow an ftruncate() as a non-root user to a file with mode 000.
3977 Lustre now respects this flag to disable mode checks when
3978 truncating a file owned by the user
3981 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3984 Description: Revoking locks from clients that went dead or catatonic might take
3986 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3987 cancellation of such locks instant on servers without waiting for
3988 any reply from clients. Clients drops these locks when cancel
3989 notification from server is received without replying.
3992 Frequency : liblustre-only, when liblustre client dies or becomes busy
3994 Description: Doing ls on Linux clients can take a long time with active
3996 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3997 granting such locks to it in the first place if possible. Locks
3998 are taken by proxy on the OST during the read or write and
3999 dropped immediately afterward. Add connect flags handling, do
4000 not grant locks to liblustre clients for glimpse ASTs.
4002 Severity : enhancement
4004 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4005 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4006 sequential reads and behaves badly with applications doing random
4007 reads. Improve it by reading ahead at least read region, and
4008 avoiding excessive large RPC for small reads.
4010 Severity : enhancement
4012 Description: Creating more than 1000 files for a single job may cause a load
4013 imbalance on the OSTs if there are also a large number of OSTs.
4014 Details : qos_prep_create() uses an OST index reseed value that is an
4015 even multiple of the number of available OSTs so that if the
4016 reseed happens in the middle of the object allocation it will
4017 still utilize the OSTs as uniformly as possible.
4022 Description: OST or MDS may oops in ping_evictor_main()
4023 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4024 but doesn't restart at beginning of obd_exports_timed list
4027 Severity : enhancement
4029 Description: improve by-nid export eviction on the MDS and OST
4030 Details : allow multiple exports with the same NID to be evicted at one
4031 time without re-searching the exports list.
4034 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4036 Description: MDS may oops in groups_free()
4037 Details : in rare race conditions a newly allocated group_info struct is
4038 freed again, and this can be NULL. The 2.4 compatibility code
4039 for groups_free() checked for a NULL pointer, but 2.6 did not.
4042 Frequency : common for liblustre clients doing little filesystem IO
4043 Bugzilla : 9352, 7313
4044 Description: server may evict liblustre clients accessing contended locks
4045 Details : if a client is granted a lock or receives a completion AST
4046 with a blocking AST already set it would not reply to the AST
4047 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4050 Frequency : lfs setstripe, only systems with more than 160 OSTs
4052 Description: unable to set striping with a starting offset beyond OST 160
4053 Details : llapi_create_file() incorrectly limited the starting stripe
4054 index to the maximum single-file stripe count.
4057 Frequency : LDAP users only
4059 Description: lconf did not handle in-kernel recovery with LDAP properly
4060 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4062 Severity : enhancement
4064 Description: bind OST threads to NUMA nodes to improve performance
4065 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4066 node and do their allocations there to localize memory access
4068 Severity : enhancement
4070 Description: llmount can determine client NID directly from Myrinet (GM)
4071 Details : the client NID code from gmnalnid was moved directly into
4072 llmount, removing the need to use this or specifying the
4073 client NID explicitly when mounting GM clients with zeroconf
4076 Frequency : if client is started with down MDS
4078 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4079 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4080 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4081 ptlrpc_import_delayed_req() because we didn't check if the
4082 request was interrupted, and we also didn't break out of the
4083 event loop if there was no timeout
4088 Description: data loss during non-page-aligned writes to a single file from
4089 both multiple nodes and multiple threads on one node at same time
4090 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4091 inconsistency led to false short-reads, that were cached and later
4092 used by ->prepare_write() to fill in partially written page,
4093 leading to data loss.
4096 Frequency : always, if lconf --abort_recovery used
4098 Description: lconf --abort_recovery fails with 'Operation not supported'
4099 Details : lconf was attempting to abort recovery on the MDT device and not
4102 Severity : enhancement
4104 Description: remove cleanup logs
4105 Details : replace lconf-generated cleanup logs with lustre internal
4106 cleanup routines. Eliminates the need for client-cleanup and
4109 Severity : enhancement
4111 Description: add support for EAs (user and system) on lustre filesystems
4112 Details : it is now possible to store extended attributes in the Lustre
4113 client filesystem, and with the user_xattr mount option it
4114 is possible to allow users to store EAs on their files also
4116 Severity : enhancement
4118 Description: Add possibility (config option) to show minimal available OST free
4120 Details : When compiled with --enable-mindf configure option, statfs(2)
4121 (and so, df) will return least minimal free space available from
4122 all OSTs as amount of free space on FS, instead of summary of
4123 free spaces of all OSTs.
4125 Severity : enhancement
4127 Description: do not expand extent locks acquired on OST-side
4128 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4129 by server: they are not cached anyway.
4132 Frequency : when mmap is used/binaries executed from Lustre
4134 Description: Unmmap pages before throwing them away from read cache.
4135 Details : llap_shrink cache now attempts to unmap pages before discarding
4136 them (if unmapping failed - do not discard). SLES9 kernel has
4137 extra checks that trigger if this unmapping is not done first.
4142 Description: lconf didn't resolve symlinks before checking to see whether a
4143 given mountpoint was already in use
4146 Frequency : when migrating failover services
4147 Bugzilla : 6395, 9514
4148 Description: When migrating a subset of services from a node (e.g. failback
4149 from a failover service node) the remaining services would
4150 time out and evict clients.
4151 Details : lconf --force (implied by --failover) sets the global obd_timeout
4152 to 5 seconds in order to quickly disconnect, but this caused
4153 other RPCs to time out too quickly. Do not change the global
4154 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4156 Severity : enhancement
4157 Frequency : if MDS is started with down OST
4158 Bugzilla : 9439,5706
4159 Description: Allow startup/shutdown of an MDS without depending on the
4160 availability of the OSTs.
4161 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4162 Add appropriate locking and lov-osc refcounts for safe
4163 cleaning. Add osc abort_inflight calls in case the
4164 synchronize never started.
4167 Frequency : occasional (Cray XT3 only)
4169 Description: root not authorized to access files in CRAY_PORTALS environment
4170 Details : The client process capabilities were not honoured on the MDS in
4171 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4172 been accessed by an authorized user then root was able to access
4173 the file on the local client also. The root user capabilities
4174 are now allowed on the MDS, as this environment has secure UID.
4177 Frequency : occasional
4179 Description: ldiskfs "too long searching" message happens too often
4180 Details : A debugging message (otherwise harmless) prints too often on
4181 the OST console. This has been reduced to only happen when
4182 there are fragmentation problems on the filesystem.
4187 Description: Division by zero in statfs when all OSCs are inactive
4188 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4189 lov_statfs divides by value returned by lov_get_stripecnt().
4193 Bugzilla : 9489, 3273
4194 Description: First write from each client to each OST was only 4kB in size,
4195 to initialize client writeback cache, which caused sub-optimal
4196 RPCs and poor layout on disk for the first writen file.
4197 Details : Clients now request an initial cache grant at (re)connect time
4198 and so that they can start streaming writes to the cache right
4199 away and always do full-sized RPCs if there is enough data.
4200 If the OST is rebooted the client also re-establishes its grant
4201 so that client cached writes will be honoured under the grant.
4206 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4207 Details : Now I/O RPCs go to different portal number and (presumably) fast
4208 lock requests (and glimses) and other RPCs get their own service
4209 threads pool that should be able to service those RPCs
4212 Severity : enhancement
4214 Description: Ability to exchange lustre version between client and servers and
4215 issue warnings at client side if client is too old. Also for
4216 liblustre clients there is ability to refuse connection of too old
4218 Details : New 'version' field is added to connect data structure that is
4219 filled with version info. That info is later checked by server and
4223 Frequency : rare, liblustre only.
4224 Bugzilla : 9296, 9581
4225 Description: Two simultaneous writes from liblustre at offset within same page
4226 might proceed at the same time overwriting eachother with stale
4228 Details : I/O lock withing llu_file_prwv was released too early, before data
4229 actually was hitting the wire. Extended lock-holding time until
4230 server acknowledges receiving data.
4233 Frequency : extremely rare. Never observed in practice.
4235 Description: avoid generating lustre_handle cookie of 0.
4236 Details : class_handle_hash() generates handle cookies by incrementing
4237 global counter, and can hit 0 occasionaly (this is unlikely, but
4238 not impossible, because initial value of cookie counter is
4239 selected randonly). Value of 0 is used as a sentinel meaning
4240 "unassigned handle" --- avoid it. Also coalesce two critical
4241 sections in this function into one.
4243 Severity : enhancement
4245 Description: allow liblustre clients to delegate truncate locking to OST
4246 Details : To avoid overhead of locking, liblustre client instructs OST to
4247 take extent lock in ost_punch() on client's behalf. New connection
4248 flag is added to handle backward compatibility.
4250 Severity : enhancement
4251 Bugzilla : 4928, 7341, 9758
4252 Description: allow number of OST service threads to be specified
4253 Details : a module parameter allows the number of OST service threads
4254 to be specified via "options ost ost_num_threads={N}" in the
4255 OSS's /etc/modules.conf or /etc/modprobe.conf.
4259 Bugzilla : 6146, 9635, 9895
4260 Description: servers crash with bad pointer in target_handle_connect()
4261 Details : In rare cases when a client is reconnecting it was possible that
4262 the connection request was the last reference for that export.
4263 We would temporarily drop the export reference and get a new
4264 one, but this may have been the last reference and the export
4265 was just destroyed. Get new reference before dropping old one.
4267 Severity : enhancement
4268 Frequency : if client is started with failover MDS
4270 Description: Allow multiple MDS hostnames in the mount command
4271 Details : Try to read the configuration from all specified MDS
4272 hostnames during a client mount in case the "primary"
4275 Severity : enhancement
4277 Description: Stop sending data to evicted clients as soon as possible.
4278 Details : Check if the client we are about to send or are sending data to
4279 was evicted already. (Check is done every second of waiting,
4280 for which l_wait_event interface was extended to allow checking
4281 of exit condition at specified intervals).
4284 Frequency : rare, normally only when NFS exporting is done from client
4286 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4288 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4291 Frequency : always, if srand() is called before liblustre initialization
4293 Description: Liblustre uses system PRNG disturbing its usage by user application
4294 Details : Introduce internal to lustre fast and high-quality PRNG for
4295 lustre usage and make liblustre and some other places in generic
4296 lustre code to use it.
4298 Severity : enhancement
4299 Bugzilla : 9477, 9557, 9870
4300 Description: Verify that the MDS configuration logs are updated when xml is
4301 Details : Check if the .xml configuration logs are newer than the config
4302 logs stored on the MDS and report an error if this is the case.
4303 Request --write-conf, or allow starting with --old_conf.
4305 Severity : enhancement
4307 Description: Handle symlinks in the path when checking if Lustre is mounted.
4308 Details : Resolve intermediate symlinks when checking if a client has
4309 mounted a filesystem to avoid duplicate client mounts.
4314 Description: lconf can hit an error exception but still return success.
4315 Details : The lconf command catches the Command error exception at the top
4316 level script context and will exit with the associated exit
4317 status, but doesn't ensure that this exit status is non-zero.
4322 Description: failure of ptlrpc thread startup can cause oops
4323 Details : Starting a ptlrpc service thread can fail if there are a large
4324 number of threads or the server memory is very fragmented.
4325 Handle this without oopsing.
4328 Frequency : always, only if liblustre and non-default acceptor port was used
4330 Description: liblustre cannot connect to servers with non-default acceptor port
4331 Details : tcpnal_set_default_params() was not called and was therefore
4332 ignoring the environment varaible TCPNAL_PORT, as well as other
4333 TCPNAL_ environment variables
4338 Description: two objects could be created on the same OST for a single file
4339 Details : If an OST is down, in some cases it was possible to create two
4340 objects on a single OST for a single file. No problems other
4341 than potential performance impact and spurious error messages.
4345 Bugzilla : 5681, 9562
4346 Description: Client may oops in ll_unhash_aliases
4347 Details : Client dcache may become inconsistent in race condition.
4348 In some cases "getcwd" can fail if the current directory is
4354 Description: Inode refcounting problems in NFS export code
4355 Details : link_raw functions used to call d_instantiate without obtaining
4356 extra inode reference first.
4360 Bugzilla : 9942, 9903
4361 Description: Referencing freed requests leading to crash, memleaks with NFS.
4362 Details : We used to require that call to ll_revalidate_it was always
4363 followed by ll_lookup_it. Also with revalidate_special() it is
4364 possible to call ll_revalidate_it() twice for the same dentry
4365 even if first occurence returned success. This fix changes semantic
4366 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4367 reference on a request referred from the intent.
4368 ll_intent_release() then releases such a request.
4371 Frequency : rare, normally benchmark loads only
4373 Description: unlinked inodes were kept in memory on the client
4374 Details : If a client is repeatedly creating and unlinking files it
4375 can accumulate a lot of stale inodes in the inode slab cache.
4376 If there is no other client load running this can cause the
4377 client node to run out of memory. Instead flush old inodes
4378 from client cache that have the same inode number as a new inode.
4381 Frequency : SLES9 2.6.5 kernel and long filenames only
4382 Bugzilla : 9969, 10379
4383 Description: utime reports stale NFS file handle
4384 Details : SLES9 uses out-of-dentry names in some cases, which confused
4385 the lustre dentry revalidation. Change it to always use the
4389 Frequency : rare, unless heavy write-truncate concurrency is continuous
4390 Bugzilla : 4180, 6984, 7171, 9963, 9331
4391 Description: OST becomes very slow and/or deadlocked during object unlink
4392 Details : filter_destroy() was holding onto the parent directory lock
4393 while truncating+unlinking objects. For very large objects this
4394 may block other threads for a long time and slow overall OST
4395 responsiveness. It may also be possible to get a lock ordering
4396 deadlock in this case, or run out of journal credits because of
4397 the combined truncate+unlink. Solution is to do object truncate
4398 first in one transaction without parent lock, and then do the
4399 final unlink in a new transaction with the parent lock. This
4400 reduces the lock hold time dramatically.
4403 Frequency : rare, 2.4 kernels only
4405 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4406 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4407 before unmounting their filesystem in order to avoid deadlock.
4408 The kernel_locked() function in 2.4 kernels only checks whether
4409 the kernel lock is held, not whether it is this process that is
4410 holding it as 2.6 kernels do.
4415 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4416 Details : The client ptlrpc code may be trying to reconnect to a down
4417 server before a previous connection attempt has timed out.
4418 Increase the reconnect interval to be longer than the connection
4419 timeout interval to avoid sending duplicate connections to
4423 Frequency : echo_client brw_test command
4425 Description: fix echo_client to work with OST preallocated code
4426 Details : OST preallocation code (5137) didn't take echo_client IO path
4427 into account: echo_client calls filter methods outside of any
4428 OST thread and, hence, there is no per-thread preallocated
4429 pages and buffers to use. Solution: hijack pga pages for IO. As
4430 a byproduct, this avoids unnecessary data copying.
4434 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4435 Description: Client can oops in mdc_commit_close() after open replay
4436 Details : It was possible for the MDS to return an open request with no
4437 transaction number in mds_finish_transno() if the client was
4438 evicted, but without actually returning an error. Clients
4439 would later try to replay that open and may trip an assertion
4440 Simplify the client close codepath, and always return an error
4441 from the MDS in case the open is not successful.
4444 Frequency : rare, 2.6 OSTs only
4446 Description: OST may deadlock under high load on fragmented files
4447 Details : If there was a heavy load and highly-fragmented OST filesystems
4448 it was possible to have all the OST threads deadlock waiting on
4449 allocation of biovecs, because the biovecs were not released
4450 until the entire RPC IO was completed. Instead, release biovecs
4451 as soon as they are complete to ensure forward IO progress.
4453 Severity : enhancement
4455 Description: Support for specifying external journal device at mount
4456 Details : If an OST or MDS device is formatted with an external journal
4457 device, this device major/minor is stored in the ext3 superblock
4458 and may not be valid for failover. Allow detecting and
4459 specifying the external journal at mount time.
4464 Description: Mounting an MDS with pending unlinked files may cause oops
4465 Details : target_finish_recovery() calls mds_postrecov() which returned
4466 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4467 considers this an error and immediately begins cleaning up the
4468 lov, just after starting the mds_lov process
4470 Severity : enhancement
4472 Description: Implement 'lfs df' to report actual free space on per-OST basis
4473 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4474 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4475 usage of objects; '-h' to report in human readable format.
4477 ------------------------------------------------------------------------------
4479 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4486 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4487 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4488 at startup during orphan unlinking. Instead, skip buddy update
4489 before setup, as it will be regenerated after recovery is complete.
4494 Description: If an OST is inactive, its locks might reference stale inodes.
4495 Details : lov_change_cbdata() must iterate over all namespaces, even if
4496 they are inactive to clear inode references from the lock.
4498 Severity : enhancement
4499 Frequency : occasional, if non-standard max_dirty_mb used
4501 Description: Client will block write RPCs if not enough grant
4502 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4503 then the client will block writes while waiting for another RPC
4504 to complete instead of consuming its dirty limit. With change
4505 we get improved performance when max_dirty_mb is small.
4507 Severity : enhancement
4508 Bugzilla : 3389, 6253
4509 Description: Add support for supplementary groups on the MDS.
4510 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4511 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4512 lookups for user supplementary groups into a cache.
4516 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4517 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4518 for liblustre O_EXCL is enforced by MDS.
4522 Description: readdir never returns NULL in liblustre.
4523 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4524 dentry in struct dirent.
4528 Frequency : liblustre only, depends on application IO pattern
4529 Description: liblustre clients evicted if not contacting servers
4530 Details : Don't put liblustre clients into the ping_evictor list, so
4531 they will not be evicted by the pinger ever.
4533 Severity : enhancement
4535 Description: Add ability to evict clients by NID from MDS.
4536 Details : By echoing "nid:$NID" string into
4537 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4538 $NID would be instantly evicted from this MDS and from all active
4539 OSTs connected to it.
4543 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4544 Details : lookup_it_finish() used to query file size from OSTs that was not
4549 Description: service threads change working directory to that of init
4550 Details : Starting lustre service threads may pin the working directory
4551 of the parent thread, making that filesystem busy. Threads
4552 now change to the working directory of init to avoid this.
4556 Frequency : during shutdown only
4557 Description: shutdown with a failed MDS or OST can cause unmount to hang
4558 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4563 Frequency : 2.6 only
4564 Description: chmod/chown may include an extra supplementary group
4565 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4566 supplementary group and if none is specified this is used.
4569 Bugzilla : 5479 (6816)
4571 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4572 Details : If lookup is in progress on a file that is unlinked we might try
4573 to revalidate the inode and fail in revalidate after lookup is
4574 complete and ll_file_open() enqueues the open again but
4575 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4578 Frequency : always, if lconf --abort_recovery used
4580 Description: lconf --abort_recovery fails with 'Operation not supported'
4581 Details : lconf was attempting to abort recovery on the MDT device and not
4584 ------------------------------------------------------------------------------
4586 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4591 Frequency : rare (only unsupported configurations with a node running as an
4593 Bugzilla : 6514, 5137
4594 Description: Mounting a Lustre file system on a node running as an OST could
4596 Details : OSTs now preallocates memory needed to write out data at
4597 startup, instead of when needed, to avoid having to
4598 allocate memory in possibly low memory situations.
4599 Specifically, if the file system is mounted on on OST,
4600 memory pressure could force it to try to write out data,
4601 which it needed to allocate memory to do. Due to the low
4602 memory, it would be unable to do so and the node would
4603 become unresponsive.
4605 Severity : enhancement
4607 Description: Addition of lconf --service command line option
4608 Details : lconf now accepts a '--service <arg>' option, which is
4609 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4611 Severity : enhancement
4613 Description: Failover mode is now the default for OSTs.
4614 Details : By default, OSTs will now run in failover mode. To return to
4615 the old behaviour, add '--failout' to the lmc line for OSTs.
4617 Severity : enhancement
4619 Description: Health checks are now provided for MDS and OSTs
4620 Details : Additional detailed health check information on MSD and OSTs
4621 is now provided through the procfs health_check value.
4624 Frequency : occasional, depends on IO load
4626 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4627 after numerous create/delete cycles
4628 Details : The ext3 inode allocation policy would not allocate new inodes
4629 very well on the OSTs because there are no new directories
4630 being created. Instead we look for groups with free space if
4631 the parent directories are nearly full.
4636 Description: Network or server problems during mount may cause partially
4637 mounted clients instead of returning an error.
4638 Details : The config llog parsing code may overwrite the error return
4639 code during mount error handling, returning success instead
4645 Description: MDS can fail to allocate large reply buffers
4646 Details : After long uptimes the MDS can fail to allocate large reply
4647 buffers (e.g. zconf client mount config records) due to memory
4648 fragmentation or consumption by the buffer cache. Preallocate
4649 some large reply buffers so that these replies can be sent even
4650 under memory pressure.
4654 Frequency : rare (liblustre)
4655 Description: fsx running with liblustre complained that using truncate() to
4656 extend the file doesn't work. This patch corrects that issue.
4657 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4658 ATTR_SIZE and lsm use in llu_setattr_raw.
4662 Frequency : rare, only 2.6 kernels
4663 Description: Unusual file access patterns on the MDS may result in inode
4664 data being lost in very rare circumstances.
4665 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4666 code in the 2.6 kernel caused the inode and/or EA data not to
4667 be read from disk, causing single-file corruption.
4671 Frequency : rare, only 2.6 filesystems using extents
4672 Description: Heavy concurrent write and delete load may cause data corruption.
4673 Details : It was possible under high-load situations to have an extent
4674 metadata block in the block device cache from a just-unlinked
4675 file overwrite a newly-allocated data block. We now unmap any
4676 metadata buffers that alias just-allocated data blocks.
4680 Frequency : filesystems with default stripe_count larger than 77
4681 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4682 Details : lconf specifies an inode size of 4096 bytes when the default
4683 stripe_count is larger than 77. This conflicts with the default
4684 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4685 this case to avoid pinning too much memory for large EAs.
4687 ------------------------------------------------------------------------------
4689 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4694 Frequency : rare (extremely heavy IO load with hundreds of clients)
4696 Description: Client is evicted, gets IO error writing to file
4697 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4698 caused clients to be evicted for AST timeouts. The fixes in
4699 bug 5192 mean we no longer need to have such short AST timeouts
4700 so ldlm_timeout has been increased.
4703 Frequency : occasional during --force or --failover shutdown under load
4704 Bugzilla : 5949, 4834
4705 Description: Server oops/LBUG if stopped with --force or --failover under load
4706 Details : a collection of import/export refcount and cleanup ordering
4707 issues fixed for safer force cleanup
4710 Frequency : only filesystems larger than 120 OSTs
4711 Bugzilla : 5990, 6223
4712 Description: lfs getstripe would oops on a very large filesystem
4713 Details : lov_getconfig used kfree on vmalloc'd memory
4716 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4717 Bugzilla : 6242, 6243
4718 Description: reading from files that had been truncated to a non-zero size
4719 but never opened returned no data
4720 Details : ll_file_read() reads zeros from no-object files to EOF
4725 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4726 Details : There's little harm in aborting MDS/OSS recovery and letting it
4727 try again, so I removed the LASSERT and return an error instead.
4729 Severity : enhancement
4731 Description: New debugging infrastructure for tracking down data corruption
4732 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4733 (b) cover more of the client-side code path, and (c) try to narrow
4734 down where problems occurred
4738 Bugzilla : 3819, 4364, 4397, 6313
4739 Description: Racing close and eviction MDS could cause assertion in mds_close
4740 Details : It was possible to get multiple mfd references during close and
4741 client eviction, leading to one thread referencing a freed mfd.
4743 Severity: : enhancement
4744 Bugzilla : 3262, 6359
4745 Description: Attempts to reconnect to servers are now more aggressive.
4746 Details : This builds on the enhanced upcall-less recovery that was added
4747 in 1.4.2. When trying to reconnect to servers, clients will
4748 now try each server in the failover group every 10 seconds. By
4749 default, clients would previously try one server every 25 seconds.
4754 Description: After recovery, certain operations trigger a failed
4755 assertion on a client.
4756 Details : Failing over an mds, using lconf -d --failover, while a
4757 client was doing a readdir() call would cause the client to
4758 LBUG after recovery completed and the readdir() was resent.
4760 Severity : enhancement
4762 Description: Default groups are now added by lconf
4763 Details : You can now run lconf --group <servicename> without having to
4764 manually add groups with lmc.
4767 Frequency : occasional
4769 Description: Nodes with an elan id of 0 trigger a failed assertion
4772 Frequency : always when accessing e.g. tty/console device nodes
4774 Description: tty and some other devices nodes cannot be used on lustre
4775 Details : file's private_data field is used by device data and lustre
4776 values in there got lost. New field was added to struct file to
4777 store fs-specific private data.
4780 Frequency : when exporting Lustre via NFS
4782 Description: NFSD failed occasionally when looking up a path component
4783 Details : NFSD is looking up ".." which was broken in ext3 directories
4784 that had grown large enough to become hashed.
4787 Frequency : Clusters with multiple interfaces not on the same subnet
4789 Description: Nodes will repeatedly try to reconnect to an interface which it
4790 cannot reach and report an error to the log.
4791 Details : Extra peer list entries will be created by lconf with some peers
4792 unreachable. lconf now validates the peer before adding it.
4795 Frequency : Only if a default stripe is set on the filesystem root.
4797 Description: Setting a default stripe on the filesystem root prevented the
4798 filesystem from being remounted.
4799 Details : The client was sending extra request flags in the root getattr
4800 request and did not allocate a reply buffer for the dir EA.
4803 Frequency : occasional, higher if lots of files are accessed by one client
4804 Bugzilla : 6159, 6097
4805 Description: Client trips assertion regarding lsm mismatch/magic
4806 Details : While revalidating inodes the VFS looks up inodes with ifind()
4807 and in rare cases can find an inode that is being freed.
4808 The ll_test_inode() code will free the lsm during ifind()
4809 when it finds an existing inode and then the VFS later attaches
4810 this free lsm to a new inode.
4814 Bugzilla : 6422, 7030
4815 Description: MDS deadlock between mkdir and client eviction
4816 Details : Creating a new file via mkdir or mknod (starting a transaction
4817 and getting the ns lock) can deadlock with client eviction
4818 (gets ns lock and trying to finish a synchronous transaction).
4821 Frequency : occasional
4822 Description: While starting a server, the fsfilt_ext3 module could not be
4824 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4825 kernels. Previously, lconf would still use the ext3 name
4826 when trying to load modules. Now, it will correctly use
4827 ext3 on 2.4 and ldiskfs on 2.6.
4829 Severity : enhancement
4830 Description: The default stripe count has been changed to 1
4831 Details : The interpretation of the default stripe count (0, to lfs
4832 or lmc) has been changed to mean striping across a single
4833 OST, rather than all available. For general usage we have
4834 found a stripe count of 1 or 2 works best.
4836 Severity : enhancement
4837 Description: Add support for compiling against Cray portals.
4838 Details : Conditional compiling for some areas that are different
4842 Frequency : occasional
4843 Bugzilla : 6409, 6834
4844 Description: Creating files with an explicit stripe count may lead to
4845 a failed assertion on the MDS
4846 Details : If some OSTs are full or unavailable, creating files may
4847 trigger a failed assertion on the MDS. Now, Lustre will
4848 try to use other servers or return an error to the
4852 Frequency : occasional
4854 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4855 caused lock timeout during readahead (since 1.4.2).
4856 Details : Processes doing readahead might match a lock that hasn't been
4857 granted yet if there are overlapping and conflicting lock
4858 requests. The readahead process waits on ungranted lock
4859 (original lock is CBPENDING), while OST waits for that process
4860 to cancel CBPENDING read lock and eventually evicts client.
4862 Severity : enhancement
4864 Description: Initial enabling of flock support for clients
4865 Details : Implements fcntl advisory locking and file status functions.
4866 This feature is provided as an optional mount flag (default
4867 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4868 locking are implemented yet, and those that are are not guaranteed
4869 to be completely correct in production environments.
4870 mount -t lustre -o [flock|noflock] ...
4873 Frequency : occasional
4875 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4876 assertion in the ext3 JBD (journaling) layer.
4877 Details : The b_committed_data struct is protected by the big kernel lock
4878 in 2.4 kernels, serializing journal_commit_transaction() and
4879 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4880 access to this struct. In 2.6 kernels there is finer grained
4881 locking to improve SMP performance of the JBD layer.
4885 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4890 Description: merge and cleanup kernel patches.
4893 -----------------------------------------------------------------------------
4895 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4897 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4898 versions of Lustre. Please update all servers and clients to
4899 version 1.4.2 or later at the same time. You must also run
4900 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4901 to update the configuration logs.
4903 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4904 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4905 - swab llog records of type '0' so we get proper header size/idx (5861)
4906 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4907 - fix rename of one directory over another leaking an inode (5953)
4908 - avoid SetPageDirty on 2.6 (5981)
4909 - don't re-add just-being-destroyed locks to the waiting list (5653)
4910 - when creating new directories, inherit the parent's custom
4911 striping settings if present parent (3048)
4912 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4913 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4914 - don't hold client locks on temporary worklist from l_lru (5666)
4915 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4916 - automatically evict dead clients (5921)
4917 - Update file size properly in create+truncate+fstat case (6196)
4918 - Do not unhash mountpoint dentries, do not allow removal of
4920 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4921 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4922 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4923 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4924 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4925 - Correctly handle reads of files with no objects (6243)
4926 - lctl recover will also mark a device active if deactivate used (5933)
4928 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4929 - allow --write-conf on an MDS with different nettype than client (5619)
4930 - don't write config llogs to MDS for mounts not from that MDS (5617)
4931 - lconf should create multiple TCP connections from a client (5201)
4932 - init scripts are now turned off by default; run chkconfig --on
4933 lustre and chkconfig --on lustrefs to use them
4934 - upcalls are no longer needed for clients to recover to failover
4936 - add --abort-recovery option to lconf to abort recovery on device
4938 - add support for an arbitrary number of OSTs (3026)
4939 - Quota support protocol changes.
4940 - forward compatibility changes to wire structs (6007)
4941 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4942 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4943 - improved "lustre status" script
4944 - initialize blocksize for non-regular files (6062)
4945 - added --disable-server and --disable-client configure options (5782)
4946 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4947 - Vanilla 2.4.29 support
4948 - increase maximum number of obd devices to 520 (6242)
4949 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4950 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4951 SLES 9 2.6.5 kernels
4952 - limit stripes per file to 160 (the maximum EA size) (6093)
4954 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4957 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4958 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4959 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4960 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4961 - small ext3 extents cleanups and fixes (5733)
4962 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4963 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4964 - avoid needless client->OST connect, fix handle mismatch (5317)
4965 - fix DLM error path that led to out-of-sync client, long delays (5779)
4966 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4967 - fix several locking issues related to i_size (5492,5624,5654,5672)
4968 - don't move pending lock onto export if it is already evicted (5683)
4969 - fix kernel oops when creating .foo in unlinked directory (5548)
4970 - fix deadlock in obdfilter statistics vs. object create (5811)
4971 - use time_{before,after} to avoid timer jiffies wrap (5882)
4972 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4973 - Do not leak request if server was not able to process it (5154)
4974 - If mds_open unable to find parent dir, make that negative lookup(5154)
4975 - don't create new directories with extent-mapping (5909, 5936)
4977 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4978 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4979 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4980 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4981 - support for RHEL4 kernel on i686 (5773)
4982 - provide error messages when incompatible logs are encountered (5898)
4984 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4985 * version 1.4.0.10 (1.4.1 release candidate 1)
4987 - don't keep a lock reference when lock is not granted (4238)
4988 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4989 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4990 - drop import inflight refcount on signal_completed_replay error (5255)
4991 - unlock page after async write error during send (3677)
4992 - handle missing objects in filter_preprw_read properly (5265)
4993 - no transno return for symlink open, don't save no-trasno open (3440)
4994 - don't try to complete elan receive that already failed (4012)
4995 - free RPC server reply state on error (5406)
4996 - clean up thread from ptlrpc_start_thread() on error (5160)
4997 - readahead could read extra page into cache that wasn't ejected (5388)
4998 - prevent races in class_attach/setup/cleanup/detach (5260)
4999 - don't dereference de->d_inode after l_dput of de (5458)
5000 - use "int" for stripe value returned from lock_to_stripe (5544)
5001 - mballoc allocation and error-checking fixes in 2.6 (5504)
5002 - block device patches to fix I/O request sizes in 2.6 (5482)
5003 - look up hostnames for IB nals (5602)
5004 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5005 - don't start multiple acceptors for the same port (5277)
5006 - fix incorrect LASSERT in mds_getattr_name (5635)
5007 - export a proc file for general "ping" checking (5628)
5008 - fix "lfs check" to not block when the MDS is down (5628)
5010 - service request history (4965)
5011 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5012 - create an "evict_client" /proc entry on OSTs, like the MDS has
5013 - fix mount usage message, return errors per mount(8) (5168)
5014 - change grep [] to grep "[]" in tests so they work in more UMLs
5015 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5016 - remove ancient LOV_MAGIC_V0 EA support (5047)
5017 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5018 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5019 - allow readahead allocations to fail when low on memory (5383)
5020 - mmap locking landed again, after considerable improvement (2828)
5021 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5023 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5026 - send OST transaction number in read/write reply to free req (4966)
5027 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5028 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5029 - the watchdog thread now runs as interruptible (5246)
5030 - flock/lockf fixes (but it's still disabled, pending 5135)
5031 - don't use EXT3 constants in llite code (5094)
5032 - memory shortage at startup could cause assertion (5176)
5034 - reorganization of lov code
5035 - single portals codebase
5037 - add extents/mballoc support (5025)
5038 - direct I/O reads in the obdfilter (4048)
5039 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5041 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5044 - send OST transaction number in read/write reply to free req (4966)
5045 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5046 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5047 - the watchdog thread now runs as interruptible (5246)
5048 - handle missing objects in filter_preprw_read properly (5265)
5049 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5050 - drop import inflight refcount on signal_completed_replay error (5255)
5051 - unlock page after async write error during send (3677)
5052 - return original error code on reconstructed replies (3761)
5053 - no transno return for symlink open, don't save no-trasno open (3440)
5055 - add pid to ldlm debugging output (4922)
5056 - bump the watchdog timeouts -- we can't handle 30sec yet
5057 - extra debugging for orphan dentry/inode bug (5259)
5059 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5062 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5063 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5064 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5065 - fix NULL dereference in /proc/sys/portals/routes (4827)
5066 - allow failed mdc_close() operations to be interrupted (4561)
5067 - stop precreate on OST before MDS would time out on it (4778)
5068 - don't send partial-page writes before EOF from client (4410)
5069 - discard client grant for sub-page writes on large-page clients (4520)
5070 - don't free dentries not owned by NFS code, check generation (4806)
5071 - fix lsm leak if mds_create_objects() fails (4801)
5072 - limit debug_daemon file size, always print CERROR messages (4789)
5073 - use transno after validating reply (3892)
5074 - process timed out requests if import state changes (3754)
5075 - update mtime on OST during writes, return in glimpse (4829)
5076 - add mkfsoptions to LDAP (4679)
5077 - use ->max_readahead method instead of zapping global ra (5039)
5078 - don't interrupt __l_wait_event() during strace
5080 - add software watchdogs to catch hung threads quickly (4941)
5081 - make lustrefs init script start after nfs is mounted
5082 - fix CWARN/ERROR duplication (4930)
5083 - return async write errors to application if possible (2248)
5084 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5085 - print NAL number in %x format (4645)
5086 - update barely-supported suse-2.4.21-171 series (4842)
5087 - support for sles 9 %post scripts
5088 - support for building 2.6 kernel-source packages
5089 - support for sles km_* packages
5091 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5094 - ignore -ENOENT errors in osc_destroy (3639)
5095 - notify osc create thread that OSC is being cleaned up (4600)
5096 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5097 - reconstruct ost_handle() like mds_handle() (4657)
5098 - create a new thread to do import eviction to avoid deadlock (3969)
5099 - let lconf resolve symlinked-to devices (4629)
5100 - don't unlink "objects" from directory with default EA (4554)
5101 - hold socknal file ref over connect in case target is down (4394)
5102 - allow more than 32000 subdirectories in a single directory (3244)
5103 - fix blocks count for O_DIRECT writes (3751)
5104 - OST returns ENOSPC from object create when no space left (4539)
5105 - don't send truncate RPC if file size isn't changing (4410)
5106 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5107 - bind to privileged port in socknal and tcpnal (3689)
5109 - rate limit CERROR/CWARN console message to avoid overload (4519)
5110 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5111 - basic mmap support (3918)
5112 - kernel patch series update from b1_4 (4711)
5114 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5117 - avoid crash during MDS cleanup with OST shut down (2775)
5118 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5119 - don't use bad inodes on the MDS (3744)
5120 - dynamic object preallocation to improve recovery speed (4236)
5121 - don't hold spinlock over lock dumping or change debug flags (4401)
5122 - don't zero obd_dev when it is force cleaned (3651)
5123 - print grants to console if they go negative (4431)
5124 - "lctl deactivate" will stop automatic recovery attempts (3406)
5125 - look for existing locks in ldlm_handle_enqueue() (3764)
5126 - don't resolve lock handle twice in recovery avoiding race (4401)
5127 - revalidate should check working dir is a directory (4134)
5129 - don't always mark "slow" obdfilter messages as errors (4418)
5131 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5134 - don't close LustreDB during write_conf until it is done (3860)
5135 - fix typo in lconf for_each_profile (3821)
5136 - allow dumping logs from multiple threads at one time (3820)
5137 - don't allow multiple threads in OSC recovery (3812)
5138 - fix debug_size parameters (3864)
5139 - fix mds_postrecov to initialize import for llog ctxt (3121)
5140 - replace config semaphore with spinlock (3306)
5141 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5142 - don't allow enqueue to complete on a destroyed export (3822)
5143 - down write_lock before checking llog header bitmap (3825)
5144 - recover from lock replay timeout (3764)
5145 - up llog sem before sending rpc (3652)
5146 - reduce ns lock hold times when setting kms (3267)
5147 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5148 - fix NULL deref and obd_dev leak on setup error (3312)
5149 - replace some LBUG about llog ops with error handling (3841)
5150 - don't match INVALID dentries from d_lookup and spin (3784)
5151 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5152 - fix invalid assertion in ptlrpc_set_wait (3880)
5154 - add libwrap support for the TCP acceptor (3996)
5155 - add /proc/sys/portals/routes for non-root route listing (3994)
5156 - allow setting MDS UUID in .xml (2580)
5157 - print the stack of a process that LBUGs (4228)
5159 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5162 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5163 - make sure to unset replay flag from failed open requests (3440)
5164 - if default stripe count is 0, use OST count for inode size (3636)
5165 - update parent mtime/ctime on client for create/unlink (2611)
5166 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5167 - free recovery state on server during a forced cleanup (3571)
5168 - unregister_reply for resent reqs (3063)
5169 - loop back devices mounting and status check on 2.6 (3563)
5170 - fix resource-creation race that can provoke i_size == 0 (3513)
5171 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5172 - more debugging for page-accounting assertion (3746)
5173 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5174 - avoid deadlock after precreation failure (3758)
5175 - fix race and lock order deadlock in orphan handling (3450, 3750)
5176 - add validity checks when grabbing inodes from l_ast_data (3599)
5178 - add /proc/.../recovery_status to obdfilter (3428)
5179 - lightweight CDEBUG infrastructure, debug daemon (3668)
5180 - change default OSC RPC parameters to be better on small clusters
5181 - turn off OST read cache for files smaller than 32MB
5182 - install man pages and include them in rpms (3100)
5183 - add new init script for (un)mounting lustre filesystems (2593)
5184 - run chkconfig in %post for init scripts (3701)
5185 - drop scimac NAL (unmaintained)
5187 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5190 - clean kiobufs before and after use (3485)
5191 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5192 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5193 - change init script to fail more clearly if not run as root (1528)
5194 - allow clients to reconnect during replay (1742)
5195 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5196 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5197 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5198 - fix cond_resched (3554)
5199 - only evict unfinished clients after recovery (3515)
5200 - allow bulk resend, prevent data loss (3570)
5201 - dynamic ptlrpc request buffer allocation (2102)
5202 - don't allow unlinking open directory if it isn't empty (2904)
5203 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5204 - remove extraneous obd dereference causing LASSERT failure (3334)
5205 - don't use get_cycles() when creating temp. files on the mds (3156)
5206 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5207 - handle EEXIST for set-stripe, set proper directory name (3336)
5209 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5210 - fix ksocknal_fmb_callback() error messages (2918)
5212 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5215 - don't copy lvb into (possibly NULL) reply on error (2983)
5216 - don't deref dentry after dput, don't free lvb on error (2922)
5217 - use the kms to determine writeback rpc length (2947)
5218 - increment oti_logcookies when osc is inactive (2948)
5219 - update client's i_blocks count via lvb messages (2543)
5220 - handle intent open/close of special files properly (1557)
5221 - mount MDS with errors=remount-ro, like obdfilter (2009)
5222 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5223 - don't use cancelling-locks' kms values (2947)
5224 - use highest lock extent for kms, not last one (2925)
5225 - don't dereference ERR_PTR() dentry in error handling path (3107)
5226 - fix thread race in portals_debug_dumplog() (3122)
5227 - create lprocfs device entries at setup instead of at attach (1519)
5228 - common AST error handler, don't evict client on completion race (3145)
5229 - zero nameidata in detach_mnt in 2.6 (3118)
5230 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5231 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5232 - initialize RPC timeout timer earlier for 2.6 (3219)
5233 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5234 - print nal/nid for unknown nid (3258)
5235 - additional checks for oscc recovery before doing precreate (3284)
5236 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5237 - don't crash in mdc_close for bad permissions on open (3285)
5238 - zero i_rdev for non-device files (3147)
5239 - clear page->private before handing to FS, better assertion (3119)
5240 - tune the read pipeline (3236)
5241 - fix incorrect decref of invalidated dentry (2350)
5242 - provide read-ahead stats and refine rpc in flight stats (3328)
5243 - don't hold journal transaction open across create RPC (3313)
5244 - update atime on MDS at close time (3265)
5245 - close LDAP connection when recovering to avoid server load (3315)
5246 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5247 - don't leak open file on MDS after open resend (3325)
5248 - serialize filter_precreate and filter_destroy_precreated (3329)
5249 - loop device shouldn't call sync_dev() for nul device (3092)
5250 - clear page cache after eviction (2766)
5251 - resynchronize MDS->OST in background (2824)
5252 - refuse to mount the same filesystem twice on same mountpoint (3394)
5253 - allow llmount to create routes for mounting behind routers (3320)
5254 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5255 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5256 - fix rare problem with rename on htree directories (3417)
5258 - allow default OST striping configuration per directory (1414)
5259 - fix compilation for qswnal for 2.6 kernels (3125)
5260 - increase maximum number of MDS request buffers for large systems
5261 - change liblustreapi to be useful for external progs like lfsck (3098)
5262 - increase local configuration timeout for slow disks (3353)
5263 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5265 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5268 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5269 - don't overwrite extent policy data in reply if lock was blocked (2901)
5270 - drop filter export grants atomically with removal from device (2663)
5271 - del obd_self_export from work_list in class_disconnect_exports (2908)
5272 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5273 - swab reply message in mdc_close, other PPC fixes (2464)
5274 - fix destroying of named logs (2325)
5275 - overwrite old logs when running lconf --write_conf (2264)
5276 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5277 - fix race in target_handle_connect (2898)
5278 - mds_reint_create() should take same inode create lock (2926)
5279 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5280 - don't close files for self_export to avoid uninitialized obd (2936)
5281 - allow MDS with the same name as client node (2939)
5282 - hold dentry reference for closed log files for unlink (2325)
5283 - reserve space for all logs during transactions (2059)
5284 - don't evict page beyond end of stripe extent (2925)
5285 - don't oops on a deleted current working directory (2399)
5286 - handle hard links to targets without a parent properly (2517)
5287 - don't dereference NULL lock when racing during eviction (2867)
5288 - don't grow lock extents when lots of conflicting locks (2919)
5290 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5293 - account for cache space usage on clients to avoid data loss (974)
5294 - lfsck support in lustre kernel code (2349)
5295 - reduce journal credits needed for BRW writes (2370)
5296 - orphan handling to avoid losing space on client/server crashes
5297 - ptlrpcd can be blocked, stopping ALL progress (2477)
5298 - use lock value blocks to assist in proper KMS, faster stat (1021)
5299 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5300 - recovery for initial connections (2355)
5301 - fixes for mds_cleanup_orphans (1934)
5302 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5303 - block all file creations until orphan recovery completes (1901)
5304 - client remove rq_connection from request struct (2423)
5305 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5306 - recovery timer race (2670)
5307 - mdc_close recovey bug (2532)
5308 - ptlrpc cleanup bug (2710)
5309 - mds timeout on local locks (2588)
5310 - namespace lock held during RPCs (2431)
5311 - handle interrupted sync write properly (2503)
5312 - don't try to handle a message that hasn't been replied to (2699)
5313 - client assert failure during cleanup after abort recovery (2701)
5314 - leak mdc device after failed mount (2712)
5315 - ptlrpc_check_set allows timedout requests to complete (2714)
5316 - wait for inflight reqs when ptlrpcd finishes (2710)
5317 - make sure unregistered services are removed from the srv_list
5318 - reset bulk XID's when resending them (caught by 1138 test)
5319 - unregister_bulk after timeout
5320 - fix lconf error (2694)
5321 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5322 - readahead locks pages, leaves pending causing memory pressure (2673)
5323 - increase OST request buffers to 4096 on large machines (2729)
5324 - fix up permission of existing directories in simple_mkdir (2661)
5325 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5326 - don't assign transno to errored transactions (2742)
5327 - don't delete objects on OST if given a bogus objid from MDS (2751)
5328 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5329 - if rq_no_resend, then timeout request after recovery (2432)
5330 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5331 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5332 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5333 - MDS recovery shouldn't delete objects that it has given out (2730)
5334 - if enqueue arrives after completion, don't clobber LVB (2819)
5335 - don't unlock pages twice when trigger_group_io returns error (2814)
5336 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5337 - don't write pages to disk if there was an error (1450)
5338 - don't ping imports that have recovery disabled (2676)
5339 - take buffered bytes into account when balancing socknal conn (2817)
5340 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5341 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5342 - remove little-endian swabbing of llog records (1987)
5343 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5344 - retry reposting request buffers if they fail (1191)
5345 - grow extent at grant time to avoid granting a revoked lock (2809)
5346 - lock revoke doesn't evict page if covered by a second lock (2765)
5347 - disable VM readahead to avoid reading outside lock extents (2805)
5349 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5350 - updated kernel patches for hp-2.4.20 kernel (2681)
5352 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5355 - fix truncated write corruption (2366)
5356 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5358 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5359 - ensure len(uuid) < 37 in lmc (1171)
5360 - fix ia64 OOPS in llog_test (2255)
5361 - zero end of page at obdfilter for partial page writes (2648)
5362 - don't leave stale dentries around after renames (bug 2428)
5363 - fix timeouts when evicting a client with a single lock held (2642)
5364 - set deadline for the initial HELLO message to drain (2634)
5365 - print out dotted-quad IP addresses in the socknal (2302)
5367 - additional debugging for MDS client eviction problem (2443)
5368 - fix mkfsoptions support for osts (2603, 2604)
5370 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5373 - add series for the vanilla 2.6.0 kernel
5374 - add series for the vanilla 2.4.24 kernel
5375 - add series for a cray x86/64 UL kernel drop
5376 - fix xattr patches for the vanilla 2.4.19 series
5378 - generate true UUIDs in lmc (1171)
5379 - have portals stack dumping break in UML (2466)
5380 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5381 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5382 - wake callback waiting threads on client eviction (2460)
5383 - Add --ptldebug and --subsystem to lmc (1719)
5384 - update assertion to allow safe interrupt allocation
5385 - set rq_no_resend for cancel requests (2432)
5386 - recalculate ptlrpcd timeout after resend (2494)
5387 - call vfs_rmdir when removing pending directories (2368)
5388 - fix renaming a file to itself (2429)
5389 - lmc creates a default one-stripe lov (2454)
5390 - expand procfs space to handle large clusters (2326)
5391 - increase UML stack to avoid overflow
5392 - update lconf's list of debug and subsystem masks
5393 - fix lfs find --obd (2510)
5394 - /proc tunable for disabling filter read caching (2591)
5395 - stop rpm packages from altering slapd.conf (2301)
5396 - disable nagle in the socknal under 0conf (2578)
5397 - choose mds inode size based on stripe count (2572)
5398 - fix kernel-source rpm problems (2516)
5400 - add --disable-doc to avoid pdf generation (2421)
5401 - update documentation, tests, type-os, comments
5402 - avoid format warnings on ia64
5403 - remove the TOE NAL
5404 - tiny code cleanups by removing unused fields
5406 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5409 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5410 - remove the most blatant lies from BUILDING file (2371)
5411 - change default debug level to reasonable production setting
5412 - reduce client side cache size to reduce cache flush time
5413 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5414 - make TCP zerocopy and pinger support enabled by default (2476)
5415 - sync writes completed after process exits caused crashes (2319)
5416 - maintain correct mount count on the MDS (2356)
5417 - backout 1557, because 2316 wasn't really fixed
5418 - better file I/O statistics gathering in /proc
5419 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5420 - another kernel patch to fix zero-copy TCP function export
5421 - don't take duplicate lock when processing re-sent getattr (2420)
5422 - lctl uses obd_self_export instead of creating new conn (2353)
5423 - MDS/OST recovery case which requires object creation asserted (2425)
5424 - move lfs from /usr/sbin to /usr/bin in packages
5425 - fix race between mds_client_add and mds_client_free (2417)
5426 - use kmalloc instead of slabs in portals (2430)
5427 - don't create duplicate records when a failover MDS is present (2442)
5428 - remove unnecessary mount age check (2332)
5429 - don't remove directory inodes from locks prematurely (2451)
5430 - don't break if MDS service name is the same as hostname (2103)
5431 - fix races in client write RPC generation when cache full (2482)
5433 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5436 - remove now-unused request->rq_obd (278)
5437 - if an allocation fails, print out how much memory we've used (1933)
5438 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5439 - add the "configurable stack size" patch to most series files (1256)
5440 - ability to write large log records, for 100+ OST configs (2306)
5441 - fix NULL deref when filter_prep fails (2314)
5442 - fix operator precedence error in filter_sync
5443 - dynamic allocation of socknal TX descriptors (2315)
5444 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5445 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5446 - ensure that CWARN messages in Portals always get to the syslog
5447 - __init/__exit are not for prototype decls (ldlm_init/exit)
5448 - x86-64 compile warning fixes
5449 - fix gateway LMC keyword conflict (2318)
5450 - fix MDS lock inversions in getattr/reint paths (1844)
5451 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5452 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5453 - DLM race condition prevented some lock evictions (2328)
5454 - ENOMEM detection and retry on socknal sends (2230)
5455 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5456 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5457 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5458 - better error messages when a client is rejected during recovery (1505)
5459 - avoid cancelling locks which were never granted, after failure (2330)
5460 - fix i_sem/journal inversion in mds_client_add (2333)
5461 - fix truncate/getattr lock cycle deadlock (2334)
5462 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5463 - fix two rare exit paths which could leak an l_lock() ref (2321)
5464 - fencepost error in MDS/OST orphan recovery (2226)
5465 - make log record alignment 8 bytes (1988)
5466 - lstripe now fails when requested offset > ost_count (2237)
5467 - ensure that all kernel series have a complete list.h (1607)
5468 - fix crashes in special-file operations (2316)
5469 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5471 - allow configurable automake binary, for testing new versions
5472 - small update to the lfs documentation
5474 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5476 * fix negative export reference count in fsfilt_sync (2312)
5478 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5479 * release candidate 0.9.1
5481 - orphans are moved into the PENDING directory for possible recovery
5482 - replayed opens now open by fid for orphan/rename safety (1042)
5483 - last close of an orphan inode generates a transno (683)
5484 - chdir() and mount() now pin the directory entry (1020)
5485 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5486 - discard very old requests without processing them (1502)
5487 - remove some common, well-understood CERRORs (1505)
5488 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5489 - clear "grant" flags in OST replies until OST grant code lands (1644)
5490 - fix read performance by not clobbering i_blksize on client (1598)
5491 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5492 - make LRU size a /proc tunable, clears locks when reduced (707)
5493 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5494 - server locks take references on exports now (1558)
5495 - build fixes for 2.4.20-rh trees (1663)
5496 - return an error from lov_create if all OSCs are inactive (1751)
5497 - fix import levels when a reconnect happens without a timeout (1597)
5498 - exit early from mds_open if we get a lookup error (1749)
5499 - partial page read at EOF wouldn't wait for disk before sending (1642)
5500 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5501 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5502 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5503 - write/truncate lock inversion (1639)
5504 - Don't auto-load obdclass, portals modules during cleanup (1495)
5505 - fix timestamps from jumping to "now" (1763)
5506 - extra journal assertions (1648)
5507 - add an extra multiunlink test (1771)
5508 - fix read_record/write_record API (1776)
5509 - fix leak of offset_extent, possible incorrect i_size later (1772)
5510 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5511 - Debugging for the kqswnal_get_idle_tx problems (1820)
5512 - Allow recovery to be attempted multiple times (1536)
5513 - Write out MDS last_rcvd file after it is first created (1600)
5514 - Fix tx_descriptor leak in failed transmit situations (1827)
5515 - ext3 journaling fixes for assertion failure after IO error (1871)
5516 - class_export_put() on freed export after completion AST error (1896)
5517 - Fix revalidate looping in VFS (1322)
5518 - Don't access a freed export during MDS_REINT timeout (1521)
5519 - Add open-unlink recovery support on the MDS (1673,1764)
5520 - Return an error if no MDS data was read from last_rcvd (1946)
5521 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5522 - Don't setup a disk device that doesn't match exported UUID (317)
5523 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5524 - avoid committing NULL handle in force close
5525 - local.sh is now a one-stripe LOV configuration
5526 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5527 - fix ext3 htree duplicate directory entry corruption (1516)
5528 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5529 - update to new LOV EA format (2097)
5530 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5532 2003-06-15 Phil Schwan <phil@clusterfs.com>
5535 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5536 - per-import/export recovery handling (958, 931, 959)
5537 - multiple last-rcvd slots, for serving multiple FSes (949)
5538 - connections are again shared between multiple imp/exports (963, 964)
5539 - "umount -f" would hang if any requests needed to be sent (393, 978)
5540 - avoid pinning large req buffer by copying for queued messages (989)
5541 - add "uuid" to "lctl device" command to help upcalls (991)
5542 - "open" RPCs with transnos would confuse recovery counters (1037)
5543 - do proper endian conversion of all wire messages (288, 340, 891)
5544 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5545 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5546 - fix ptlrpc_connection leak in target_handle_connect (1174)
5547 - fix import refcounting bug in OST and MDS cleanup (1134)
5548 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5549 - fix very unlikely obd_types race condition (501)
5550 - remove osc_open hack for echo_client (1187)
5551 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5552 - a failure in read_inode2 leads to deadlock (1139)
5553 - cancel ack-locks as soon as transaction is committed (1072)
5554 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5555 - make sure to commitrw after any preprw to avoid deadlock (1162)
5556 - failing to execute a file in a lustre FS would lock inode (1203)
5557 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5558 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5559 - obd_uuid2tgt would walk off the end of the list (1255)
5560 - on IA64 the osc would give portals incorrect bulk size (1258)
5561 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5562 - fix lock inversion caused by new llite matching code (1282)
5563 - limit the number of dirty pages on a client to 10MB (1286)
5564 - timed out locks were not being corrected cancelled (1289)
5565 - fix O_DIRECT above 4GB on IA-32 (1292)
5566 * major user-visible changes
5567 - fail out/fail over policy now controlled by the upcall (993)
5569 - add OBD_PING to check server availability and failure (954)
5570 - lustre messages are now sent in sending host order (288, 340, 891)
5571 - add eadatalen to MDS getattr reply (340)
5572 - OST read replies may contain second buffer, with per-page status (593)
5574 2003-03-11 Phil Schwan <phil@clusterfs.com>
5577 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5578 - failed open()s could cause deadlock; fixed (867, 869)
5579 - stop cancelling OST locks when files are closed (481)
5580 - overlapping XID spaces caused network corruption (851, 853)
5581 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5582 - setattr_raw added, to do single-RPC, server-side setattrs
5583 - lmc/lconf syntax change for OST UUIDs
5584 - fix crashy race condition between ptlrpc_free_req and osc_close
5585 - don't use request in mdc_enqueue if we hit a timeout (889)
5586 - don't set the inode i_size for regular files from the MDS (896)
5587 - handle out of order completion AST (842)
5588 - don't LBUG if a lock request times out after receiving AST (913)
5589 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5590 - if a bad lock AST arrives, send an error instead of dropping entirely
5591 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5592 - fix leak in bulk IO when only partially completed (899, 900, 926)
5593 - fix O_DIRECT for ia64 (55)
5594 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5595 - C-z after timeout could hang a process forever; fixed (977)
5597 - client-side I/O cache (678, 924, 929, 941, 970)
5599 - READPAGE and SETATTRs which don't take server-side locks get
5602 2003-02-11 Phil Schwan <phil@clusterfs.com>
5605 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5606 - Fix fsfilt_extN_readpage() to read a full page of directory
5607 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5608 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5609 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5610 - Most intent operations were removed, in favour of a new RPC mode
5611 that does a single RPC to the server and bypasses most of the VFS
5612 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5613 - Aggressively cancel local locks on DLM servers
5614 - mds_reint_unlink sends EA to the client if it's the last nlink.
5615 client uses that EA to unlink OST objects.
5616 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5617 - recursive symlinks were fixed (439)
5618 - fixed NULL deref in DEBUG_REQ
5619 - filter_update_lastobjid no longer calls sync, which annoyed extN
5620 - fixed multi-client small-writes to a single file problem (445)
5621 - fixed mtime updates during file writes (607)
5622 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5623 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5624 - fixed Linux OST size reporting problem (444, 656)
5625 - OST now updates object mtime with writes or setattr (607, 619)
5626 - client verifies file size before zeroing page past EOF (445)
5627 - OST now writes last allocated objid to disk with allocation (108)
5628 - LOV on echo now works (409)
5630 - mds_reint_unlink sends a new buffer, with the EA included. this
5631 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5632 set if a regular file was being unlinked, and it was the last link
5633 - use PtlGet from the target for bulk writes (315)
5634 - OST now updates object mtime with writes or setattr (607, 619)
5635 - LDLM now has a grant-time callback to revalidate locked items, if
5637 - Many MDS operations were reorganized to combat race conditions
5639 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5640 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5642 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5645 - Fully reactivate OST imports after reconnection (512, others)
5646 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5647 - More graceful error handling for truncating on dead OST (515)
5648 - Don't error out unless we're actually accessing dead stripes (474)
5649 - Fix garbage sizes when stripes are missing (410)
5650 - LRU counters were broken, causing constant lock purge (433, 432)
5651 - garbage on read from stripes with failed OSTs (441)
5652 - mark OSCs as active before reconnecting during recovery (438)
5653 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5654 - lfind did not preserve OST order in output (443)
5655 - symlinks cause hung clients, incorrect data (439)
5656 - stop dereferencing request after dropping refcount (457)
5657 - don't LASSERT(spin_is_locked) on non-SMP (455)
5658 - fixes for many rename() bugs
5659 - fstat didn't correctly synchronize attributes (399)
5660 - server must handle lock cancellation during blocking AST prep (487)
5661 - bulk descriptors were free()d too soon (511)
5662 - fix paths in lconf, which would load incorrect modules (451, 507)
5663 - fix confusing lconf 'host not found' error message (386)
5664 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5665 - fix race condition in mdc_blocking_ast() for inode access (526)
5666 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5667 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5668 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5669 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5670 - fix lstripe in several ways (kernel side) (527)
5671 - fix request leak in ldlm_cli_enqueue (262)
5672 - incorrect OSC was marked inactive after OST failure
5673 - call mds_fs_cleanup before unmounting filesystem (524)
5674 - fix races between taking ns_lock and ldlm_lock_change_resource
5675 - fix races updating LOV export open file list
5676 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5677 - fix recovery NULL deref in ldlm_cli_cancel_unused
5678 - fix some DLM races by using new hash table for lock handles (419)
5679 - permit the client to specify desired inodes, at replay
5680 - duplicate requests when we queue them for replay reintegration
5681 - fix last_rcvd offset calculation
5682 - sync after each recovered transaction, so we always make progress
5683 - never, not always, ERESTART requests without transnos
5684 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5686 - skip replay if the MDS says that the client is already connected
5687 - don't check for a recovery-enabled export to match lctl's UUID
5688 - don't INC_USE_COUNT for phantom exports
5689 - don't crash when cleaning up phantom exports (567)
5690 - don't double-finish or set replay data for errored mdc_open requests
5691 - abort requests when they time out, so we don't get old replies
5692 - send/receive replies for AST messages again
5693 - if the client says that it doesn't have the lock, cancel it on the
5695 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5696 mark it as destroyed, it will all work out when decref is called
5697 - fix module use counts (22, 581)
5699 - ASTs now expect a reply (server cancels lock on error reply)
5701 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5704 - fix many simultaneous client startup (392)
5705 - fix dentry->d_it clobbering
5706 - credentials weren't being shipped for readdir/getattr operations
5707 - remove invalid assertions triggered during some concurrent MD
5709 - proper Lustre versions added (336, 389)
5710 - fix memory leak for create error case (398)
5711 - fix LOV locking bug that would get cli/srv out of sync
5712 - fix echo client over LOV (409)
5713 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5714 - fix double-O_EXCL intent crash (424)
5715 - avoid sending multiple lock CANCELs (352)
5717 - MDS can do multi-client recovery (modulo bugs in new code)
5719 - many updates, edits, cleanups
5721 2002-11-18 Phil Schwan <phil@clusterfs.com>
5724 - fix null d_it dereference (346)
5725 - fix full OST/dbench hang (333)
5726 - fix permission problem with file removal (286)
5727 - fix removal of OSCs from LOV when they fail
5728 - fix NULL deref during bulk timeout (214)
5729 - fix problems related to multiple filesystems on one MDS (241)
5730 - fixed serious subtle metadata locking bugs
5731 - free locks on clients when inodes are removed due to memory
5733 - fix inode pointer in lock data (285)
5734 - partial support for multiple MDS on a single host (241)
5735 - data locks weren't cancelled at clear_inode time (290, 311)
5736 - intent locks could lead to unbounded lock growth (205)
5737 - added a maximum lock count, an LRU list, and a flusher
5738 - fix multiple rename (365)
5739 - properly abstracted the echo client
5740 - OSC locked 1 byte too many; fixed
5741 - rewrote brw callback code:
5742 - fixed recovery bugs related to LOVs (306)
5743 - fixed too-many-pages-in-one-write crash (191)
5744 - fixed (again) crash in sync_io_timeout (214)
5745 - probably fixed callback-related race (385)
5747 - Add capability to MDS protocol
5748 - LDLM cancellations and callbacks on different portals
5750 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5753 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5754 - defer open object destruction to close time (601981, 138)
5755 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5756 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5757 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5758 - don't call obd_setattr in ll_file_release for destroyed objects
5759 * protocol change to lustre_msg: move |version| and add |flags|
5760 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5761 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5762 * added replay of create, unlink, link and rename operations during
5763 MDS failover; recovery should be much more robust now
5764 * remove failed OSCs from LOVs (only lov_create uses this so far)
5765 * the lustre-HOWTO was brought (more) up to date (582544)
5767 2002-10-23 Phil Schwan <phil@clusterfs.com>
5770 - in-use dentries weren't being reused properly (617851)
5771 - prevent multiple LDLM setup (599178)
5772 - fix LOV size calculations for truncate (617853)
5773 - fix client handling of MDS intent errors (POSIX)
5774 - fix permission bug in lovstripe.c test (624321)
5775 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5776 - truncate past end of file could corrupt data
5777 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5778 - a race in recovery could return ETIMEDOUT to apps (623947)
5779 - building outside the source directory was fixed
5780 * the lustre-HOWTO was brought (more) up to date (582544)
5781 * major progress was made on recovery functionality
5783 2002-10-10 Phil Schwan <phil@clusterfs.com>
5786 - recovery deadlock fix
5787 - rm -rf causes LBUG fix (617817)
5788 - file open by multiple tasks fix (618962)
5789 - directory permissions bugs (602707 and 620007)
5790 - journal_stop fixed with locking (611313)
5791 - O_APPEND failures resolved (618273, perhaps 614459)
5792 - lconf PATH fix (619770)
5793 - IA64 build fix (621450)
5794 - RPC buffer sizes scale with amount of memory
5796 2002-10-01 Phil Schwan <phil@clusterfs.com>
5799 - locks would be cancelled without throwing away data pages,
5800 resulting in inconsistent data (605627)
5801 - inode attributes were not always being refreshed (605627, 612449)
5802 - lconf now continues to cleanup after lctl reports an error
5803 - MDS now enforces user permissions (602707)
5804 - lprocfs cleanup fixed, but not yet enabled (614157)
5805 - fixed infinite server hang, should a client not respond to an AST
5806 - avoid going into recovery if user calls readlink() with a buffer
5807 that's too small (613941)
5808 - AST RPCs no longer require replies (614867) -- this may be changed
5809 - don't crash server if client sends an IOV that's too big (611336)
5810 - fixed lock conversion deadlock (611892)
5811 - fixed the following of symlinks (614622)
5812 * recovery: the server can remove locks from a client that dies, other
5813 clients can make progress
5814 * more extN patch fixes
5815 * compile-time configurable ptlrpc buffer allocations
5817 - collaborative read cache document
5818 - Lustre Lite Performance CDR document-in-progress
5820 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5823 - fix typo in patch-2.4.18
5825 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5828 - clear ptlrpc request each time in handle_incoming_request()
5829 - unlink of files now destroys the object on the OST
5831 2002-09-19 Peter Braam <braam@clusterfs.com>
5833 * add hard link support
5834 * change obdfile creation method
5835 * kernel patch changed
5837 2002-09-19 Peter Braam <braam@clusterfs.com>
5840 - stack overflow bug in extN fixed
5842 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5844 * documentation updates
5845 - add man pages for config tools
5846 - update tests/README to describe testing with new config tools
5847 - finish metadata API descriptions
5848 * bug fixes and cleanups
5849 - statfs workaround for 16TB limit
5850 - LOV stripe allocation improved, can stripe on subset of OSTs
5851 - LOV file size/IO offset was wrong for files > 4GB in size
5852 - object EA data was being dropped, caused files to be unreadable
5853 - memory overflow with non-LOV OST caused memory corruption
5854 - fixed regression tests to work with new config tools, obdfilter
5855 - fixed bug when directory size became larger than 1 block
5856 - fixed bug (for single client case) when PWD was deleted
5857 - invalidate local directory pages when doing intent-based ops
5858 - avoid LDLM oops when lock callback contained bad data
5860 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5862 * documentation updates
5863 * bug fixes and cleanups
5864 - configuration tools
5867 - 64-bit compile warnings
5868 - 64-bit internal statfs data
5870 * test_brw on persistent OST devices
5872 * lprocfs (disabled)
5874 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5876 * documentation updates
5877 * bug fixes and cleanups
5878 * configuration tools
5880 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5883 * many small fixes to 0_5_4
5884 * io/network handling
5885 * thinkos in MDS operations
5887 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5890 * crucial basic fixes to 0.5.3
5891 * IOR, Iozone work over Elan
5894 2002-08-07 Phil Schwan <phil@clusterfs.com>
5895 * version 0_5_3, our first alpha
5896 * we use the new Portals iovs
5897 * documentation updates
5898 * bug fixes and cleanups
5899 * small changes in the DLM wire protocol
5901 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5902 * version 0_5_1 with some initial stability,
5903 * locking on MD and file I/O.
5904 * documentation updates
5905 * several bug fixes since 0.5.0
5906 * small changes in wire protocol
5908 2002-07-18 Phil Schwan <phil@clusterfs.com>
5910 * delivered as Lustre Light Alpha
5911 * fixed a crash after handling invalid MDS requests
5912 * fixed directory pages for architectures with non-4k pages sizes
5914 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5915 * release version v0_4_4
5916 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5918 * Fixes a number of interruption problems with OST operations.
5919 * Update documentation for portals header changes
5920 * Move all wire protocol structs/defines to lustre_idl.h
5921 * Fixes symlink length bug.
5922 * Add tcpdump to repository.
5924 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5925 * release version v0_4_3
5926 * Fixes statfs for inodes on extN.
5927 * Fixes bug in runtests which would delete /etc/hosts.
5928 * Use 64-bit object IDs wherever possible (not into VFS though)
5929 Remove ost_get_info, which is unused by lustre, and out of date.
5931 2002-07-03 Peter Braam <braam@clusterfs.com>
5932 * release version v0_4_2 Fixes a lookup error (type not passed)
5933 * move forward to head of Portals
5934 * move forward to latest Lustre kernel
5936 2002-06-25 Peter Braam <braam@clusterfs.com>
5937 * release version v0_4_1. Hopefully stable on single node use.