1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.37 (SLES 10),
5 2.6.18-128.1.6.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.
16 Severity : enhancement
18 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
20 Severity : enhancement
22 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
26 Description: resolve race between obd_disconnect and class_disconnect_exports
27 Details : if obd_disconnect will be called to already disconnected export he
28 forget release one reference and osc module can't unloaded.
30 Severity : enhancement
32 Description: Allow tuning service thread via /proc
33 Details : For each service a new
34 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
35 created that can be used to set min/max thread counts, and get the
36 current number of running threads.
40 Descriptoin: don't return error if have particaly created objects for file.
41 Details : lov_update_create_set uses set->set_success as index for created objects,
42 so if some requests will be failed, they will have hole at end of
43 array and we can use qos_shrink_lsm for allocate correct lsm.
45 Severity : enhancement
47 Description: Update OFED support to 1.4.1
51 Description: Reduce small size read RPC
52 Details : Set read-ahead limite for every file and only do read-ahead when
53 available read-ahead pages are bigger than 1M to avoid small size
56 Severity : enhancement
58 Description: move AT tunable parameters for more consistent usage
59 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
61 Severity : enhancement
63 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
64 Details : allow skip disconnected ost for send statfs request and hide error
70 Description: fix racy locking of mballoc block bitmaps causing BUG
71 Details : The locking of the mballoc buddy bitmap and the in-memory
72 block bitmap was using two different spin locks in some
73 cases. This made it possible to incorrectly access the
74 mballoc bitmap while another process was modifying it,
75 causing a sanity assertion to fail. While no on-disk corruption
76 was reported, there was some risk of this happening.
79 Frequency : rare, on llog test 6
81 Descriptoin: don't allow connect to already connected import
82 Details : allowing connect to already connected import is hide connecting problem.
87 Descriptoin: allow kill process which wait statahead result
88 Details : for some reasons 'ls' can stick in waiting result from statahead,
89 in this case need way for kill this process.
91 Severity : enhancement
93 Description: Add state history info file, enhance import info file
94 Details : Track import connection state changes in a new osc/mdc proc file;
95 add overview-type data to the osc/mdc import proc file.
97 Severity : enhancement
99 Description: MDS create should not wait for statfs RPC while holding DLM lock.
101 Severity : enhancement
103 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
108 Descriptoin: OST grant too much space to client even there are not enough space.
109 Details : Client will shrink its grant cache to OST if there are no write
110 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
111 this grant cache if there are already not enough avaible space
112 (left_space < total_clients * 32M).
115 Frequency : start MDS on uncleanly shutdowned MDS device
117 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
118 Details : stay in waiting mds<>ost recovery finished produce random bugs
119 due race between two ll_sync thread for one lov target. send
120 ACTIVATE event only if connect realy finished and import have
124 Frequency : rare, connect and disconnect target at same time
126 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
127 Details : don't call obd_disconnect under lov_lock. this long time
128 operation and can block ptlrpcd which answer to connect request.
133 Descriptoin: don't lose wakeup for imp_recovery_waitq
134 Details : recover_import_no_retry or invalidate_import and import_close can
135 both sleep on imp_recovery_waitq, but we was send only one wakeup
139 Frequency : always with long access acl
141 Descriptoin: mds can't pack reply with long acl.
142 Details : mds don't control size of acl but they limited by reint/getattr
145 Severity : enhancement
147 Description: Update to SLES10 kernel-2.6.16.60-0.33.
149 Severity : enhancement
151 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
154 Frequency : start MDS on uncleanly shutdowned MDS device
156 Descriptoin: aborting recovery hang on MDS
157 Details : don't throttle destroy RPCs for the MDT.
160 Frequency : on remount
162 Description: external journal device not working after the remount
163 Details : clear dev_rdonly flag for external journal devices in
169 Description: shutdown vs evict race
170 Details : client_disconnect_export vs connect request race.
171 if client will evicted at this time - we start invalidate
172 thread without referece to import and import can be freed
176 Frequency : rare, need acl's on inode.
178 Description: client can't handle ost additional correctly
179 Details : if ost was added after client connected to mds client can have
180 hit lnet_try_match_md ... to big messages to wide striped files.
181 in this case need teach client to handle config events about add
182 lov target and update client max ea size at that event.
184 Severity : enhancement
186 Description: Changelogs
187 Details : Changelogs are a lightweight mechanism to track filesystem
188 metadata and namespace changes. The changelog is recorded
189 permanently on the MDTs, and is periodically "consumed" / purged
190 when records are no longer needed.
192 Severity : enhancement
194 Description: compact fld format with extents
195 Details : Store range of seq rather than every seq in FLD. Seq
196 controller update FLD rather than clients. In Case of CMD, mdt0
197 has FLD, all other metadata server act as non persistent proxy
198 for FLD queries and cache fld entries in fld cache.
203 Description: don't skip ost target if they assigned to file
204 Details : Drop slow OSCs if we can, but not for requested start idx.
205 This means "if OSC is slow and it is not the requested
206 start OST, then it can be skipped, otherwise skip it only
207 if it is inactive/recovering/out-of-space.
211 Description: more cleanup in mds_lov
212 Details : not send LOV EA under replay, we can't know about they size at this
213 time. Don't allow client connect to mds before any ost connected,
214 for avoid problems with LOV EA size and returning EIO to client.
216 Severity : enhancement
218 Description: Interoperability at server side (Disk interoperability)
220 Severity : enhancement
222 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
224 Severity : enhancement
226 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
228 Severity : enhancement
230 Description: New client IO stack (CLIO).
232 Severity : enhancement
234 Description: Commit on sharing. Eliminate inter-client dependencies between
235 uncommitted transactions by doing transaction commits.
236 Thereby clients may recovery independently.
239 Frequency : Create a symlink file with a very long name
241 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
242 Details : If there is no extra space in the request for early cancels,
243 ldlm_req_handles_avail() returns 0 instead of a negative value.
245 Severity : enhancement
247 Description: Add /proc entry for import status
248 Details : The mdc, osc, and mgc import directories now have
249 an import directory that contains useful import data for debugging
252 Severity : enhancement
254 Description: Re-disable certain /proc logging
255 Details : Enable and disable client's offset_stats, extents_stats and
256 extents_stats_per_process stats logging on the fly.
259 Frequency : Only on FC kernels 2.6.22+
261 Description: oops in statahead
262 Details : Do not drop reference count for the dentry from VFS when lookup,
263 VFS will do that by itself.
265 Severity : enhancement
267 Description: Generic /proc file permissions
268 Details : Set /Proc file permissions in a more generic way to enable non-
269 root users operate on some /proc files.
273 Description: Hitting mdc_commit_close() ASSERTION
274 Details : Properly handle request reference release in
275 ll_release_openhandle().
279 Description: quota recovery deadlock during mds failover
280 Details : This patch includes att18982, att18236, att18237 in bz14840.
282 1. fix osts hang when mds does failover with quotaon
283 2. prevent watchdog storm when osts threads wait for the
288 Frequency : only patchless client
289 Description: add workaround for race between add/remove dentry from hash
291 Severity : enhancement
293 Description: Allow OST glimpses to return PW locks
297 Description: LBUG when llog conf file is full
298 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
302 Description: Prevent import from entering FULL state when server in recovery
306 Description: service mount cannot take device name with ":"
307 Details : Only when device name contains ":/" will mount treat it as
313 Description: replace ptlrpcd with the statahead thread to interpret the async
314 statahead RPC callback
318 Frequency : on recovery
319 Description: I/O failures after umount during fail back
320 Details : if client reconnected to restarted server we need join to recovery
321 instead of find server handler is changed and process self
322 eviction with cancel all locks.
324 Severity : enhancement
326 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
328 Severity : enhancement
330 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
332 Severity : enhancement
334 Description: Upcall on Lustre log has been dumped
335 Details : Allow for a user mode script to be called once a Lustre log has
336 been dumped. It passes the filename of the dumped log to the
337 script, the location of the script can be specified via
338 /proc/sys/lnet/debug_log_upcall.
343 Description: avoid idr_remove called for id which is not allocated.
344 Details : Move assigment s_dev for clustered nfs to end of initialization,
345 to avoid problem with error handling.
350 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
351 Details : When connection is reused this not moved from CONN_UNUSED_HASH
352 into CONN_USED_HASH and this prodice warning when put connection
353 again in unused hash.
356 Severity : enhancement
358 Description: File striping can now be set to use an arbitrary pool of OSTs.
360 Severity : enhancement
362 Description: Export bytes_read/bytes_write count on OSC/OST.
366 Description: Early reply size mismatch, MGC loses connection
367 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
368 the connect flags are properly negotiated.
373 Description: quotas are not honored with O_DIRECT
374 Details : all writes with the flag O_DIRECT will use grants which leads to
375 this problem. Now using OBD_BRW_SYNC to guard this.
379 Description: add quota statistics
380 Details : 1. sort out quota proc entries and proc code.
381 2. add quota statistics
383 Severity : enhancement
385 Description: enable quota support for HEAD.
389 Description: Properly propagate oinfo flags from lov to osc for statfs
390 Details : restore missing copy oi_flags to lov requests.
392 Severity : enhancement
394 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
395 : l_getgroups(8), lst(8), routerstat(8)
396 Details : included man pages for llobdstat(8), llstat(8),
397 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
399 Severity : enhancement
401 Description: configure's --enable-quota should check the
402 : kernel .config for CONFIG_QUOTA
403 Details : configure is terminated if --enable-quota is passed but
404 : no quota support is in kernel
408 Description: Remove portals compatibility
409 Details : Remove portals compatibility, not interoperable with releases
414 Description: Resolve device initialization race
415 Details : Prevent proc handler from accessing devices added to the
416 obd_devs array but yet be intialized.
418 Severity : enhancement
420 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
422 Severity : enhancement
424 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
429 Description: Using wrong pointer in osc_brw_prep_request
430 Details : Access to array[-1] can produce panic if kernel compiled with
431 CONFIG_PAGE_ALLOC enabled
435 Description: Client runs out of low memory
436 Details : Consider only lowmem when counting initial number of llap pages
440 Description: *optional* service tags registration
441 Details : if the "service tags" package is installed on a Lustre node
442 When the filesystem is mounted, a local-node service tag will
443 be created. See http://inventory.sun.com/ for more information
444 about the Service Tags asset management system.
448 Description: Kernel BUG tries to release flock
449 Details : Lustre does not destroy flock lock before last reference goes
450 away. So always drop flock locks when client is evicted and
451 perform unlock regardless of successfulness of speaking to MDS.
455 Description: add refcount protection for osc callbacks, avoid panic on shutdown
459 Description: sanity test 65a fails if stripecount of -1 is set
460 Details : handle -1 striping on filesystem in ll_dirstripe_verify
465 Description: ASSERTION(CheckWriteback(page,cmd)) failed
466 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
469 Severity : enhancement
471 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
475 Description: do not process already freed flock
476 Details : flock can possibly be freed by another thread before it reaches
477 to ldlm_flock_completion_ast.
481 Description: LBUG during stress test
482 Details : Need properly lock accesses the flock deadlock detection list.
486 Description: oops in page fault handler
487 Details : kernel page fault handler can return two special 'pages' in error
488 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
492 Description: timeout with invalidate import.
493 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
494 should be handled by ptlrpcd. This produce long age waiting and
495 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
497 Severity : enhancement
499 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
503 Description: enable MGS and MDT services start separately
504 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
505 a co-located MGS without starting the MGS, which is a complement
506 to 'nosvc' mount option.
510 Frequency : after recovery
511 Description: precreate to many object's after del orphan.
512 Details : del orphan st in oscc last_id == next_id and this triger growing
513 count of precreated objects. Set flag LOW to skip increase count
514 of precreated objects.
518 Frequency : rare, on clear nid stats
519 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
520 Details : when clean nid stats sometimes try destroy live entry,
521 and this produce panic in free.
525 Description: Stack overflow during MDS log replay
526 ease stack pressure by using a thread dealing llog_process.
530 Description: wait until IO finished before start new when do lock cancel.
531 Details : VM protocol want old IO finished before start new, in this case
532 need wait until PG_writeback is cleared until check dirty flag and
533 call writepages in lock cancel callback.
535 Severity : enhancement
537 Description: using special macro for print time and cleanup in includes.
541 Description: mds_mfd_close() ASSERTION(rc == 0)
542 Details : In mds_mfd_close(), we need protect inode's writecount change
543 within its orphan write semaphore to prevent possible races.
547 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
551 Frequency : rare, on shutdown ost
552 Description: don't hit live lock with umount ost.
553 Details : shrink_dcache_parent can be in long loop with destroy dentries,
554 use shrink_dcache_sb instead.
558 Description: don't panic with use echo client
559 Details : echo client pass NULL as client nid pointer and this produce null
564 Description: fix build on ppc32
565 Details : compile code with -m64 flag produce wrong object file for ppc32.
569 Description: add message levels for liblustreapi
573 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
574 Details : Move the 'good_osts' check before the 'total_bavail' check. This
575 will result in an -EAGAIN and in the exit call path we call
576 alloc_rr() which will with increasing aggressiveness attempt to
577 aquire precreated objects on the minimum number of required OSCs.
581 Description: Use old size assignment to avoid deadlock
582 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
583 the scheduling while holding a spinlock. We do not need locking
584 for size in ll_update_inode() because size is only updated from
585 the MDS for directories or files without objects, so there is no
586 other place to do the update, and concurrent access to such inodes
587 are protected by the inode lock.
591 Description: resolve "_IOWR redefined" build error on SLES10
595 Description: dump the memory debugging after all modules are unloaded to
596 suppress false negative in conf_sanity test 39
598 Severity : enhancement
600 Description: build kernel-ib packages for OFED 1.3 in our release cycle
605 Description: fix SLES kernel versioning
606 Details : the kernel version for our SLES 10 kernel did not include a "-"
607 before the "smp" at the end. while this was not a problem in
608 general, it did mean that software trying to use the kernel
609 version to try to detect a vendor specific kernel would fail.
610 this was most evident by the OFED build scripts.
614 Description: Don't update lov_desc members until making sure they are valid
615 Details : When updating lov_desc members via proc fs, need fix their
616 validities before doing the real update.
620 Description: don't put request into delay list while invalidate in flight.
621 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
622 import in flight. this produce timeout for invalidate and sometimes
623 can cause stale data.
627 Frequency : on ppc only
628 Description: not convert ost objects for directory because it's not exist.
629 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
633 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
635 Severity : enhancement
637 Description: Update to support 2.6.22.14 vanilla kernel.
641 Frequency : rare, on recovery
642 Description: read procfs can produce deadlock in some situation
643 Details : Holding lprocfs lock which send rpc can produce block for destroy
644 obd objects and this also block reconnect with -EALREADY. This
645 isn't fix all lprocfs bugs - but make it rare.
647 Severity : enhancement
649 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
652 Frequency : frequent on X2 node
654 Description: mdc_set_open_replay_data LBUG
655 Details : Set replay data for requests that are eligible for replay.
659 Description: lustre_mgs: operation 101 on unconnected MGS
660 Details : When MGC is disconnected from MGS long enough, MGS will evict the
661 MGC, and late on MGC cannot successfully connect to MGS and a lot
662 of the error messages complaining that MGS is not connected.
666 Frequency : on network error
667 Description: panic with double free request if network error
668 Details : mdc_finish_enqueue is finish request if any network error ocuring,
669 but it's true only for synchronus enqueue, for async enqueue
670 (via ptlrpcd) this incorrect and ptlrpcd want finish request
673 Severity : enhancement
675 Description: client-side metadata stat-ahead during readdir(directory readahead)
676 Details : perform client-side metadata stat-ahead when the client detects
677 readdir and sequential stat of dir entries therein
680 Frequency : on start mds
682 Description: Implement get_info(last_id) in obdfilter.
685 Frequency : occasional
687 Description: Correctly check stale fid, not start epoch if ost not support SOM
688 Details : open with flag O_CREATE need set old fid in op_fid3 because
689 op_fid2 was overwritten with new generated fid, but mds can answer
690 with one of these two fids and both is not stale. Setattr
691 incorrectly started an epoch and assume will be called
692 done_writing, but without SOM done_writing ever being called.
695 Frequency : rare, depends on device drivers and load
697 Description: MDS or OSS nodes crash due to stack overflow
698 Details : Code changes in 1.8.0 increased the stack usage of some functions.
699 In some cases, in conjunction with device drivers that use a lot
700 of stack the MDS (or possibly OSS) service threads could overflow
701 the stack. One change which was identified to consume additional
702 stack has been reworked to avoid the extra stack usage.
705 Frequency : occasional
707 Description: Do not fail import if osc_interpret_create gets -EAGAIN
708 Details : If osc_interpret_create got -EAGAIN it immediately exits and
709 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
710 oscc_has_objects and see OSC has no objests and call
711 oscc_internal_create to resend create request.
713 Severity : enhancement
715 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
717 Severity : enhancement
719 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
722 Frequency : very rare
724 Description: Fix replay if there is an un-replied request and open
725 Details : In some cases, older replay request will revert the
726 mcd->mcd_last_xid on MDS which is used to record the client's
729 Severity : enhancement
731 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
733 Severity : enhancement
735 Description: Add rhel5 support to HEAD.
737 Serverity : enhancement
739 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
744 Description: Don't allow skipping OSTs if index has been specified.
745 Details : Don't allow skipping OSTs if index has been specified, make
746 locking in internal create lots better.
750 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
751 Details : access to bitfield in structure is always rounded to long
752 and this produce problem with not atomic change any bit.
756 Description: Lustre make rpms failed.
757 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
758 builds Lustre rpms from tarball.
761 Frequency : rare on shutdown ost
763 Description: If llog cancel was not send before clean_exports phase, this can
764 produce deadlock in llog code.
765 Details : If llog thread has last reference to obd and call class_import_put
766 this produce deadlock because llog_cleanup_commit_master wait when
767 last llog_commit_thread exited, but this never success because was
768 called from llog_commit_thread.
772 Description: allow userland application know is lost one of stripes.
773 Details : fill lvb_blocks with error code on ost and return it to
774 application if error flag found.
778 Description: NULL lov_tgts causing MDS oops
779 Details : more safe checks for NULL lov_tgts for avoid oops.
781 Severity : enhancement
783 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
787 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
788 Details : If an OST is down the MDS will hang indefinitely in
789 obd_statfs_rqset() waiting for the statfs data. While for
790 MDS QOS usage of statfs, it should not stuck in waiting.
792 Severity : enhancement
794 Description: remote_acl support
795 Details : Support ACL-based permission check for remote user.
796 Support setfacl/getfacl for remote user with the utils
797 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
798 the system "{s,g}etfacl" utils.
800 Severity : enhancement
802 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
804 Severity : enhancement
806 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
810 Description: make mgs_setparam() handle fsname containing dash
811 Details : fsname containing a dash does not work with lctl conf_param
813 Severity : enhancement
815 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
817 Severity : enhancement
819 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
823 Frequency : rare, at shutdown
824 Description: access already free / zero obd_namespace.
825 Details : if client_disconnect_export was called without force flag set,
826 and exist connect request in flight, this can produce access to
827 NULL pointer (or already free pointer) when connect_interpret
828 store ocd flags in obd_namespace.
832 Frequency : only at startup
833 Description: not alloc memory with spinlock held.
834 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
835 if any spinlock held.
837 Severity : enhancement
839 Description: make lustre randomly fail allocating memory
840 Details : Make lustre randomly failed allocating memory for testing purpose.
842 Severity : enhancement
844 Description: lost problems with lov objid file
845 Details : Fixes some scability and access to not inited memory problems
846 in work with lov objdid file.
851 Description: lfs find does not continue on file error
852 Details : Continue other files processing when a file/dir is absent.
856 Description: Inconsistent usage of lustre_pack_reply()
857 Details : Standardize the usage of lustre_pack_reply() such that it
858 always generate a CERROR on failure.
861 Frequency : occasional
863 Description: MDS hang or stay in waiting lock
864 Details : If client receive lock with CBPENDING flag ldlm need send lock
865 cancel as separate rpc, to avoid situation when cancel request
866 can't processed due all i/o threads stay in wait lock.
870 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
874 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
878 Description: Update to SLES9 kernel-2.6.5-7.287.3.
882 Description: Update to RHEL4 latest kernel.
884 Severity : enhancement
886 Description: Build SLES10 patchless client fails
887 Details : The configure was broken by run ./configure with
888 --with-linux-obj=.... argument for patchless client. When the
889 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
890 can't search header adequately. Use absolute path such as
891 -I($LINUX)/include instead.
895 Description: interrupt oig_wait produce painc on resend.
896 Details : brw_redo_request can be used for resend requests from ptlrpcd and
897 private set, and this produce situation when rq_ptlrpcd_data not
898 copyed to new allocated request and triggered LBUG on assert
899 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
900 ptlrpcd set if request is changed and can be safety changed to use
905 Description: LASSERT_{REQ,REP}SWAB macros are buggy
906 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
907 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
908 these macros and replace them with their difinitions instead.
912 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
913 Details : Update which_patch & target file for SLES10 latest kernel.
916 Description: Remove client patches from SLES 10 kernel.
917 Details : This causes SLES 10 clients to behave as patchless clients
918 even on a Lustre-patched (server) kernel.
920 Severity : enhancement
922 Description: self-adjustable client's lru lists
923 Details : use adaptive algorithm for managing client cached locks lru
924 lists according to current server load, other client's work
925 pattern, memory activities, etc. Both, server and client
926 side namespaces provide number of proc tunables for controlling
929 Severity : enhancement
931 Description: light-weight GSS support
932 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
933 and reduce performance overhead.
935 Severity : enhancement
937 Description: Linux keyring support
938 Details : Support using service of Linux keyring for Lustre GSS internal
939 context refresh/cache mechanism.
943 Description: Fix errors in lfs documentation
944 Details : Fixes man pages
948 Description: don't use GFP_* in generic Lustre code.
949 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
953 Description: obdclass is limited by single OBD_ALLOC(idarray)
954 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
958 Description: warnings with build patchless client with vanila 2.6.19 and up
959 Details : change old ctl_table style and replace ctl_table/ctl_table_header
960 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
964 Frequency : Only for SLES
965 Description: sanity_quota fail test_1
966 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
967 file, which makes TSTID[2] inunique.
971 Frequency : Only for SLES9
972 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
973 Details : Update target/ChangeLog/which_patch .
977 Frequency : Only for test_10 in sanity-quota.sh
978 Description: a bug in quota test code
979 Details : checking a test flag in a wrong place
983 Frequency : Only for test_13 in sanity-quota.sh
984 Description: a bug in quota test script
985 Details : shouldn't check overfull allocated quota
989 Description: jbd statistics
990 Details : Port older jbd statistics patch for sles10
992 Severity : enhancement
994 Description: Add printing inode info into message about error in writepage.
999 Description: reply_lock_interpret crash due to race with it and lock cancel.
1000 Details : Do not replay locks that are being cancelled. Do not reference
1001 locks by their address during replay, just by their handle.
1005 Frequency : When flocks are used.
1006 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1007 Details : Flock locks might destroy just granted lock if it could be merged
1008 with another existing flock, this is done in completion handler,
1009 so teach ldlm_cli_enquque_fini that this is a valid case for
1015 Description: Oops in read and write path when failing to allocate lock.
1016 Details : Check if lock allocation failed and return error back.
1020 Description: lstripe command fails for valid OST index
1021 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1022 instead of lov->desc.ld_active_tgt_count.
1026 Description: sanity.sh failed test 103
1027 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1032 Description: df doesn't work properly if diskfs blocksize != 4K
1033 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1037 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1038 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1040 Severity : enhancement
1042 Description: Adaptive timeouts
1043 Details : RPC timeouts adapt to changing server load and network
1044 conditions to reduce resend attempts and improve recovery time.
1048 Description: llapi_file_create() does not allow some changes
1049 Details : add llapi_file_open() that allows specifying the mode and
1050 open flags, and also returns an open file handle.
1054 Description: merge and cleanup kernel patches.
1055 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1059 Description: Add journal checksum support.(Kernel part)
1060 Details : The journal checksum feature adds two new flags i.e
1061 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1062 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1063 indicates that the commit block contains the checksum for
1064 the blocks described by the descriptor blocks. Now commit
1065 record can be sent to disk without waiting for descriptor
1066 blocks to be written to disk. This behavior is controlled
1067 using JBD2_FEATURE_ASYNC_COMMIT flag.
1071 Description: OSS needs mutliple precreate threads
1072 Details : Add ability to start more than one create thread per OSS.
1076 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1077 Details : execute lfs setstripe on client
1081 Description: mds_obd_create error creating tmp object
1082 Details : When the user sets quota on root, llog will be affected and can't
1083 create files and write files.
1086 Frequency : Always on ia64 patchless client, and possibly others.
1088 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1089 Details : This allows the patchless client to be loaded on architectures
1090 without this export.
1094 Description: RedHat Update kernel for RHEL5
1095 Details : Add the kernel config file for RHEL5.
1099 Description: RedHat Update kernel for RHEL5
1100 Details : Modify the kernel config file more closer RHEL5.
1104 Description: Build failure against Centos5 (RHEL5)
1105 Details : Define PAGE_SIZE when it isn't present.
1109 Description: test_7 on sanity-quota.sh can't be used on separate machines
1110 Details : add facet to handle it
1114 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1115 Details : with NFS, the anon dentry's parent was set to itself in
1116 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1117 corresponding dentry other than use rec->ur_name.
1119 Severity : enhancement
1121 Description: lfs setstripe enhancement
1122 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1126 Description: enable data checksumming by default
1127 Details : enable checksum by default, allow --disable-checksum
1128 configure option and "-o nochecksum" mount option. Checksums
1129 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1133 Description: lustre support for RHEL5
1134 Details : Add support for RHEL5.
1138 Description: Client eviction due to failover config
1139 Details : after a connection loss, the lustre client should attempt to
1140 reconnect to the last active server first before trying the
1141 other potential connections.
1145 Description: when mds and osts use different quota unit(32bit and 64bit),
1146 quota will be released repeatly.
1147 Details : void sending multiple quota reqs to mds, which will keep the
1148 status between the reqs.
1152 Description: osts not allocated evenly to files
1153 Details : change the condition to increase offset_idx
1156 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1158 Description: Data corruption for OSTs that are formatted larger than 2TB
1160 Details : When generating the bio request for lustre file writes the
1161 sector number would overflow a temporary variable before being
1162 used for the IO. The data reads correctly from Lustre (which
1163 will overflow in a similar manner) but other file data or
1164 filesystem metadata may be corrupted in some cases.
1168 Description: Tune the kernel for good SCSI performance.
1169 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1170 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1175 Description: rewrite ext2-derived code in obdclass/uuid.c
1176 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1177 from scratch preserving functionality.
1181 Description: "lfs find -obd UUID" prints directories
1182 Details : "lfs find -obd UUID" will return all directory names instead
1183 of just file names. It is incorrect because the directories
1184 do not reside on the OSTs.
1188 Description: use i_size_read and i_size_write in 2.6 port
1189 Details : replace inode->i_size access with i_size_read/write()
1192 Frequency : when removing large files
1194 Description: scheduling issue during removal of large Lustre files
1195 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1196 It causes scheduling issues when removing large files (17TB in the
1200 Frequency : only with liblustre clients on XT3
1202 Description: evictions taking too long
1203 Details : allow llrd to evict clients directly on OSTs
1206 Frequency : only on ppc
1208 Description: /proc/fs/lustre/devices broken on ppc
1209 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1210 We should make sure the type of 'index' is loff_t and then cast
1211 explicitly as needed below. Do not assign an explicitly cast
1215 Frequency : only for rhel5
1217 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1218 Details : Modify the target file & which_kernel.
1220 Severity : enhancement
1222 Description: omit set fsid for export NFS
1223 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1224 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1228 Frequency : Always, for kernels after 2.6.16
1229 Description: Fix warning idr_remove called for id=.. which is not allocated.
1230 Details : Last kernels save old s_dev before kill super and not allow
1231 to restore from callback - restore it before call kill_anon_super.
1235 Description: buffer overruns could theoretically occur
1236 Details : llapi_semantic_traverse() modifies the "path" argument by
1237 appending values to the end of the origin string, and a
1238 overrun may occur. Adding buffer overrun check in liblustreapi.
1242 Description: Fix error on 'ls .' at the top of the Lustre mount.
1243 Details : Don't revalidate dentry if it is a root dentry.
1247 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1248 Details : Modify vm-tunables-rhel4.patch.
1252 Description: Kernel config for 2.6.18-vanilla.
1253 Details : Modify targets/2.6-vanilla.target.in.
1254 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1255 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1256 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1257 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1261 Description: improve handling recoverable errors
1262 Details : if request processig with error which can be recoverable on server
1263 request should be resend, otherwise page released from cache and
1268 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1269 Details : Modify target file & which_patch.
1270 A flaw was found in the IA32 system call emulation provided
1271 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1272 value could be stored in the %RAX register, which could trigger an
1273 out-of-bounds system call table access. An untrusted local user
1274 could exploit this flaw to run code in the kernel
1275 (ie a root privilege escalation). (CVE-2007-4573).
1279 Description: O_DIRECT bypasses client statistics.
1280 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1281 read_ahead_stats were updated. copied the stats section from
1282 osc_send_oap_rpc() into async_internal().
1286 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1290 Description: change order of libsysio includes
1291 Details : '#include sysio.h' should always come before '#include xtio.h'
1295 Description: handle "serious error: objid * already exists" more gracefully
1296 Details : If LAST_ID value on disk is smaller than the objects existing in
1297 the O/0/d* directories, it indicates disk corruption and causes an
1298 LBUG(). If the object is 0-length, then we should use the existing
1299 object. This will help to avoid a full fsck in most cases.
1301 Severity : enhancement
1303 Description: adapt the lustre_config script to support the upgrade case
1304 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1305 server targets to 1.6 in parallel.
1309 Description: To avoid grant space > avaible space when the disk is almost
1310 full. Without this patch you might see the error "grant XXXX >
1311 available" or some LBUG about grant, when the disk is almost
1313 Details : In filter_check_grant, for non_grant cache write, we should
1314 check the left space by if (*left > ungranted + bytes), instead
1315 of (*left > ungranted), because only we are sure the left space
1316 is enough for another "bytes", then the ungrant space should be
1317 increase. In client, we should update cl_avail_grant only there
1318 is OBD_MD_FLGRANT in the reply.
1322 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1323 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1324 described in CVE-2007-4573. This problem could allow untrusted
1325 local users to gain root access.
1328 Frequency : when using O_DIRECT and quotas
1330 Description: Incorrect file ownership on O_DIRECT output files
1331 Details : block usage reported by 'lfs quota' does not take into account
1332 files that have been written with O_DIRECT.
1337 Description: touch file failed when fs is not full
1338 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1339 otherwise we can get ENOSP while fs is not full.
1343 Description: parallel lock callbacks
1344 Details : Instead of sending blocking and completion callbacks as separated
1345 requests, adding them to a set and sending in parallel.
1348 Frequency : only for Cray XT3
1349 Bugzilla : 12829/13455
1350 Description: Changing primary group doesn't change the group lustre assigns to
1352 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1353 overridden with the primary group provided by the group upcall,
1354 whereas the supplied fsgid can be trusted if it is in the list of
1355 supplementary groups returned by the group upcall.
1357 Severity : enhancement
1359 Description: Allow masking D_WARNING, D_ERROR messages from console
1360 Details : Console messages can now be disabled via lnet.printk.
1364 Description: User code with malformed file open parameter crashes client node
1365 Details : Before packing join_file req, all the related reference should be
1366 checked carefully in case some malformed flags cause fake
1367 join_file req on client.
1371 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1372 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1373 mode is equal to the granted mode. However, it should also check
1374 that LDLM_FL_CP_REQD is not set.
1378 Description: Heavy nfs access might result in deadlocks
1379 Details : After ELC code landed, it is now improper to enqueue any mds
1380 locks under och_sem, because enqueue might want to decide to
1381 cancel open locks for same inode we are holding och_sem for.
1385 Description: Client eviction while running blogbench
1386 Details : A lot of unlink operations with concurrent I/O can lead to a
1387 deadlock causing evictions. To address the problem, the number of
1388 oustanding OST_DESTROY requests is now throttled to
1389 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1390 callbacks are processed in priority.
1394 Description: enable ACLs on MDS by default
1395 Details : ACLs must be enabled on MDS by default.
1398 Frequency : PPC/PPC64 only
1400 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1401 Details : fix duplicated definitions between asm-ppc64/types.h and
1402 lustre_types.h on PPC.
1405 Frequency : PPC/PPC64 only
1407 Description: asm-ppc/segment.h does not exist
1408 Details : fix compile issue on PPC.
1412 Description: better handle error messages in extents code
1415 Frequency : RHEL4 only
1417 Description: mkfs is very slow on IA64/RHEL4
1418 Details : A performance regression has been discovered in the MPT Fusion
1419 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1420 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1421 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1423 Severity : enhancement
1425 Description: SNMP support enhancement
1426 Details : Adding total number of sampled request for an MDS node in snmp
1429 Severity : enhancement
1431 Description: Optimize ldlm waiting list processing for PR extent locks
1432 Details : When processing waiting list for read extent lock and meeting read
1433 lock that is same or wider to it that is not contended, skip
1434 processing rest of the list and immediatelly return current
1435 status of conflictness, since we are guaranteed there are no
1436 conflicting locks in the rest of the list.
1440 Description: Time out and refuse to reconnect
1441 Details : When the failover node is the primary node, it is possible
1442 to have two identical connections in imp_conn_list. We must
1443 compare not conn's pointers but NIDs, otherwise we can defeat
1444 connection throttling.
1448 Description: port llog fixes from b1_6 into HEAD
1449 Details : Port llog reference couting and some llog cleanups from b1_6
1450 (bug 10800) into HEAD, for protect from panic and access to already
1451 free llog structures.
1455 Description: Detect stride IO mode in read-ahead
1456 Details : When a client does stride read, read-ahead should detect that and
1457 read-ahead pages according to the detected stride pattern.
1461 Description: data checksumming impacts single node performance
1462 Details : add support for several checksum algorithm. Currently, only CRC32
1463 and Adler-32 are supported. The checksum type can be changed on
1464 the fly via /proc/fs/lustre/osc/*/checksum_type.
1468 Description: use adler32 for page checksums
1469 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1474 Description: build for x2 fails
1475 Details : fix compile issue on Cray systems.
1479 Description: Properly match for duplicate locks
1480 Details : Due to different lock order from skiplists code, we need to
1481 traverse entire list for now
1484 Frequency : only on PPC/SLES10
1486 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1487 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1488 be defined. Add a hack in mkfs_lustre.c to work around this compile
1493 Description: LASSERT on MDS when client holding flock lock dies
1494 Details : ldlm pool logic depends on number of granted locks equal to
1495 number of released locks which is not true for flock locks, so
1496 just exclude such locks from consideration.
1500 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1501 Details : Use fsfilt_sync() for both the whole filesystem sync and
1502 individual file sync to eliminate dangerous inode locking
1503 with I_LOCK that can lead to a deadlock.
1507 Description: performance in 1.6.3
1508 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1509 sure that 1MB requests are not fragmented by the block layer.
1511 Severity : enhancement
1513 Description: organize the server-side client stats on per-nid basis
1514 Details : Change the structure of stats under obdfilter and mds to
1522 The "uuid"s file would list the uuids of _active_ exports.
1523 And the clear entry is to clear all stats and stale nids.
1525 Severity : enhancement
1527 Description: eliminate client locks in face of contention
1528 Details : file contention detection and lockless i/o implementation
1529 for contended files.
1533 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1536 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1538 Description: MDS may lose file striping (and hence file data) in some cases
1539 Details : If there are additional extended attributes stored on the MDS,
1540 in particular ACLs, SELinux, or user attributes (if user_xattr
1541 is specified for the client mount options) then there is a risk
1542 of attribute loss. Additionally, the Lustre file striping
1543 needs to be larger than default (e.g. striped over all OSTs),
1544 and an additional attribute must be stored initially in the
1545 inode and then increase in size enough to be moved to the
1546 external attribute block (e.g. ACL growing in size) for file
1551 Description: skiplist implementation simplification
1552 Details : skiplists are used to group compatible locks on granted list
1553 that was implemented as tracking first and last lock of each
1554 lock group the patch changes that to using doubly linked lists
1558 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1559 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1560 because LASSERT on both the data supplied by a client, and the
1561 data on disk is dangerous and incorrect.
1563 Severity : enhancement
1565 Description: Slow truncate/writes to huge files at high offsets.
1566 Details : Directly associate cached pages to lock that protect those pages,
1567 this allows us to quickly find what pages to write and remove
1568 once lock callback is received.
1572 Description: more ldlm soft lockups
1573 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1574 starve other threads from the resource lock for a long time in
1575 case of long waiting queue, so change the debug level from
1576 D_OTHER to the less frequently used D_INFO.
1578 Severity : enhancement
1580 Description: add -gid, -group, -uid, -user options to lfs find
1584 Description: Hung threads in invalidate_inode_pages2_range
1585 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1586 once one is completed. As a result, some RPCs are stuck in the
1587 queue and are never sent.
1591 Description: filter threads hungs on waiting journal commit
1592 Details : Cleanup filter group llog code, then only filter group llog will
1593 be only created in the MDS/OST syncing process.
1597 Description: Procfs and llog threads access destoryed import sometimes.
1598 Details : Sync the import destoryed process with procfs and llog threads by
1599 the import refcount and semaphore.
1601 Severity : enhancement
1603 Description: openlock cache of b1_6 port to HEAD
1608 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1609 Details : If insertion of an extent fails, then discard the inode
1610 preallocation and free data blocks else it can lead to duplicate
1615 Description: don't always update ctime in ext3_xattr_set_handle()
1616 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1617 In some cases the ctime should not be updated, for example for
1618 2.0->1.8 compatibility it is necessary to delete an xattr and it
1619 should not update the ctime.
1623 Bugzilla : 15713/16362
1624 Description: Assertion in iopen_connect_dentry in 1.6.3
1625 Details : looking up an inode via iopen with the wrong generation number can
1626 populate the dcache with a disconneced dentry while the inode
1627 number is in the process of being reallocated. This causes an
1628 assertion failure in iopen since the inode's dentry list contains
1629 both a connected and disconnected dentry.
1633 Description: assertion failure in ldlm_handle2lock()
1634 Details : fix a race between class_handle_unhash() and class_handle2object()
1635 introduced in lustre 1.6.5 by bug 13622.
1640 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1641 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1646 Description: oops in sd_iostats_seq_show()
1647 Details : unloading/reloading the scsi low level driver triggers a kernel
1648 bug when trying to access the sd iostat file.
1653 Description: Kernel panics during QLogic driver reload
1654 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1655 patch, causing memory corruption.
1660 Description: journal_dev option does not work in b1_6
1661 Details : pass mount option during pre-mount.
1663 Severity : enhancement
1665 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1666 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1667 extent information of a file. It can be used to map logical blocks
1668 in a file to physical blocks in the block device.
1672 Description: LDLM soft lockups - improvement
1673 Details : It is be possible to send the lock handle along with each read
1674 or write request because the client is already doing a lock match
1675 itself so there isn't any reason the OST should have to re-do that
1679 Frequency : only on Cray X2
1681 Description: X2 build failures
1682 Details : fix build failures on Cray X2.
1686 Description: xid & resent requests
1687 Details : Initialize RPC XID from clock at startup (randomly if clock is
1690 Severity : enhancement
1692 Description: Add lustre_start utility to start or stop multiple Lustre servers
1697 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1698 Details : In case of memory pressure, list_del() can be called twice on
1699 req->rq_history_list, causing a kernel oops.
1703 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1704 Details : fix stack overflow in the distributed lock manager by defering
1705 export eviction after a failed AST to the elt thread instead of
1706 handling it in the dlm interpret routine.
1710 Description: Convert some comments to new format.
1711 Details : Update documenting comments to match doxygen conventions.
1715 Description: Grammar fixes.
1716 Details : A couple of trivial spelling fixes.
1720 Description: OSD_COUNTERS-mandatory
1721 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1722 to simplify the code.
1726 Description: simplify cmm_device freeing logic.
1727 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1731 Description: Add lockdep support to dt_object_operations locking interface.
1732 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1733 parameter indicating lock ordering. Update mdd code to use new
1738 Description: Introduce failloc constants for lockless IO tests.
1739 Details : Add two new failloc constants to test lockless IO. Only one of
1740 them in implemented---another is checked in yet to be landed
1745 Description: Add lockdep support for inode mutex.
1746 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1747 in the situations where Lustre has to lock more than one inode
1752 Description: Add optional invariants checking support.
1753 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1754 configure switch. This macro is to be used for consistency and
1755 sanity checks that are too expensive to be left in `production'
1760 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1761 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1766 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1767 Details : Use spin_lock_nested() in (the only) situation where more than
1768 one ldlm_lock is locked simultaneously. Also, fix possible
1769 dead-lock in ldlm_lock_change_resource() by enforcing particular
1774 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1775 Details : Instead of specifying each ldlm_lock call-back through separate
1776 parameter, wrap them into struct ldlm_callback_suite.
1780 Description: Kill join_lru obd method and its callers.
1781 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1782 in memory---a requirement implemented through ->o_join_lru() obd
1783 method in HEAD. Remove this method and its users.
1787 Description: Add asynchronous ldlm ENQUEUE completion handler.
1788 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1789 case that stock ldlm_completion_ast() cannot handle as it waits
1790 until lock is granted. Introduce new ldlm_completion_ast_async()
1791 for this. Also comment ldlm_completion_ast().
1795 Description: ldlm_error <-> errno conversion.
1796 Details : Add functions to map (rather arbitrary) between LDLM error codes
1797 and standard errno values. CLIO needs this to prevent LDLM specific
1798 constants from escaping ldlm and osc.
1802 Description: Kill unused ldlm_handle2lock_ns() function.
1803 Details : Kill unused ldlm_handle2lock_ns() function.
1807 Description: Add lu_ref support to ldlm_lock
1808 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1809 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1810 and ldlm_resource. LDLM interface has to be changed, because code that
1811 releases a reference on a lock, has to "know" what reference this is.
1812 In the most frequent case
1814 lock = ldlm_handle2lock(handle);
1816 LDLM_LOCK_PUT(lock);
1818 no changes are required. When any other reference (received _not_ from
1819 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1820 instead of LDLM_LOCK_PUT().
1822 Arguably, changes are pervasive, and interface requires some discipline
1823 for proper use. On the other hand, it was very instrumental in finding
1824 a few leaked lock references.
1828 Description: Add ldlm_lock_addref_try().
1829 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1830 attempts to addref a lock that might be being canceled
1835 Description: Add ldlm_weigh_callback().
1836 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1837 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1838 hard-coded `number of pages' logic.
1842 Description: Add lockdep annotations to llog code.
1843 Details : Use appropriately tagged _nested() locking calls in the places
1844 where llog takes more than one ->lgh_lock lock.
1848 Description: Add loi_kms_set().
1849 Details : Wrap kms updates into a helper function.
1853 Description: Constify instances of struct lsm_operations.
1854 Details : Constify instances of struct lsm_operations.
1858 Description: lu_conf support.
1859 Details : On a server, a file system object is uniquely identified
1860 by a fid, which is sufficient to locate and load all object
1861 state (inode). On a client, on the other hand, more data are
1862 necessary instantiate an object. Change lu_object_find() and
1863 friends to take additional `lu_conf' argument describing object.
1864 Typically this includes layout information.
1868 Description: lu_context fixes.
1869 Details : Introduce new lu_context functions that are needed on the client
1870 side, where some system threads (ptlrpcd) are shared by multiple
1871 modules, and so cannot be stopped during module shutdown.
1875 Description: Add start and stop methods to lu_device_type_operations.
1876 Details : Introduce two new methods in lu_device_type_operations, that are
1877 invoked when first instance of a given type is created and last one
1878 is destroyed respectively. This is need by CLIO.
1882 Description: Add lu_ref support to struct lu_device.
1883 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1884 track leaked references.
1888 Description: Introduce lu_kmem_descr.
1889 Details : lu_kmem_descr and its companion interface allow to create
1890 and destroy a number of kmem caches at once.
1894 Description: Fix lu_object finalization race.
1895 Details : Fix a race between lu_object_find() finding an object and its
1896 concurrent finalization. This race is (most likely) not possible
1897 on the server, but might happen on the client.
1901 Description: Introduce lu_ref interface.
1902 Details : lu_ref is a debugging module allowing to track references to
1903 a given object. It is quite cpu expensive, and has to be
1904 explicitly enabled with --enable-lu_ref. See usage description
1909 Description: Factor lu_site procfs stats into a separate function.
1910 Details : Separate lu_site stats printing code into a separate function
1911 that can be reused on a client.
1915 Description: Constify instances of struct {lu,dt,md}_device_operations.
1916 Details : Constify instances of struct {lu,dt,md}_device_operations.
1920 Description: Introduce struct md_site and move meta-data specific parts of
1921 struct lu_site here.
1922 Details : Move md-specific fields out of struct lu_site into special struct
1923 md_site, so that lu_site can be used on a client.
1927 Description: Kill mdd_lov_destroy().
1928 Details : Remove unused mdd code.
1932 Description: Add st_block checking to multistat.c.
1933 Details : Add st_block checking to multistat.c.
1937 Description: Add lu_ref support to struct obd_device.
1938 Details : Add lu_ref tracking to obd_device.
1942 Description: Kill obd_set_fail_loc().
1943 Details : Remove unused code.
1947 Description: Add special type for ptlrpc_request interpret functions.
1948 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1949 there. Actual usage will be in CLIO.
1953 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1954 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1955 doesn't work with lockdep.
1959 Description: Add rwv.c test program.
1960 Details : New testing program exercising readv(2) and writev(2) (Qian).
1964 Description: Add sendfile.c test program.
1965 Details : New testing program exercising sendfile(2) (Jay).
1969 Description: Ratelimit a message that can be very frequent.
1970 Details : Ratelimit a memory allocation failure message that can
1975 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1976 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1980 Description: Liblustre build fixes.
1981 Details : Liblustre build fixes.
1985 Description: libcfs: add cfs_{need,cond}_resched() interface.
1986 Details : libcfs: add cfs_{need,cond}_resched() definition and
1987 implementations for Linux, NT, and liblustre.
1989 Severity : enhancement
1991 Description: More exported tunables for mballoc
1992 Details : Add support for tunable preallocation window and new tunables for
1993 large/small requests
1997 Description: Detect corruption of block bitmap and checking for preallocations
1998 Details : Checks validity of on-disk block bitmap. Also it does better
1999 checking of number of applied preallocations. When corruption is
2000 found, it turns filesystem readonly to prevent further corruptions.
2004 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2005 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2006 window to 0. If it does hit cache miss, and read-pattern is still
2007 stride-io mode, does not reset the stride window, but also does not
2008 increase the stride window length in this case.
2012 Frequency : only for big-endian servers
2013 Description: Check if system is big-endian while mounting fs with extents feature
2014 Details : Mounting a filesystem with extents feature will fail on big-endian
2015 systems since ext3-based ldiskfs is not supported on big-endian
2016 systems. This can be over-riden with "bigendian_extents" mount option.
2018 Severity : enhancement
2020 Description: The root squash functionality
2021 Details : A security feature, which is to prevent users from being able
2022 to mount lustre on their desktop, run as root, and delete
2023 all of the files in the filesystem. The goal is accomplished by
2024 remapping user id (UID) and group id (GID) of the root user to
2025 a UID and GID specified by the system administartor via Lustre
2026 configuration management server (MGS). The functionality also
2027 allows to specify sets of clients for which the remapping does
2032 Description: Excessive recovery window
2033 Details : With AT enabled, the recovery window can be excessively long (6000+
2034 seconds). To address this problem, we no longer use
2035 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2036 timeout no longer depends on the service time, it is set to
2037 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2038 time via pb_service_time.
2042 Description: Watchdog triggered on MDS failover
2043 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2044 the OSTs know that the MDS "UUID" can be reused for the same export
2045 from a different NID, so we do not need to wait for the export to be
2049 Frequency : rare, only if using MMP with Linux RAID
2051 Description: MMP doesn't work with Linux RAID
2052 Details : While using HA for Lustre servers with Linux RAID, it is possible
2053 that MMP will not detect multiple mounts. To make this work we
2054 need to unplug the device queue in RAID when the MMP block is being
2055 written. Also while reading the MMP block, we should read it from
2056 disk and not the cached one.
2058 Severity : enhancement
2060 Description: open file using fid
2061 Details : A file can be opened using just its fid, like
2062 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2065 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2068 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2069 Details : add proper locking for creation of jbd_4k slab cache
2073 Description: MMP check in ext3_remount() fails without displaying any error
2074 Details : When multiple mount protection fails during remount, proper error
2077 Severity : enhancement
2079 Description: Allow stripe size to be up to 4G-64k
2080 Details : Fix math logic to allow large stripe sizes.
2083 --------------------------------------------------------------------------------
2085 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2087 * Support for kernels:
2088 2.6.5-7.283 (SLES 9),
2089 2.6.9-55.EL (RHEL 4),
2090 2.6.16.46-0.14 (SLES 10),
2091 2.6.18.8 vanilla (kernel.org)
2092 * Client support for unpatched kernels:
2093 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2094 2.6.16 - 2.6.22 vanilla (kernel.org)
2095 * Due to recently discovered recovery problems, we do not recommend
2096 using patchless RHEL 4 clients with this or any earlier release.
2097 * Recommended e2fsprogs version: 1.39.cfs8
2098 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2099 * Starting with this release, the ldiskfs backing filesystem required
2100 by Lustre is now in its own package, lustre-ldiskfs. This package
2101 should be installed. It is versioned separately from Lustre and
2102 may be released separately in future.
2106 Description: block reactivating mgc import until all deactivates complete
2107 Details : Fix race when failing back MDT/MGS to itself (testing)
2110 Frequency : at statup only
2112 Description: mds_lov_synchronize race leads to various problems
2113 Details : simultaneous MDT->OST connections at startup can cause the
2114 sync to abort, leaving the OSC in a bad state.
2116 Severity : enhancement
2118 Description: add optional extra BUILD_VERSION info
2119 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2120 to override the lustre version.
2123 Frequency : 2.6.18 servers only
2125 Description: ll_kern_mount() doesn't release the module reference
2126 Details : The ldiskfs module reference count never drops down to 0
2127 because ll_kern_mount() doesn't release the module reference.
2132 Description: server LBUG when using old ost_num_threads parameter
2133 Details : Accept the old ost_num_threads parameter but warn that it
2134 is deprecated, and fix an off-by-one error that caused an LBUG.
2139 Description: Transient SCSI error results in persistent IO issue
2140 Details : iobuf->dr_error is not reinitialized to 0 between two
2144 Frequency : sometimes when underlying device returns I/O errors
2146 Description: OSTs not going read-only during write failures
2147 Details : OSTs are not remounted read-only when the journal commit threads
2148 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2149 instead of the ext3 wrappers.
2153 Description: poor connect scaling with increasing client count
2154 Details : Don't run filter_grant_sanity_check for more than 100 exports
2155 to improve scaling for large numbers of clients.
2158 Frequency : SLES10 only
2160 Description: sanity-quota.sh quotacheck failed: rc = -22
2161 Details : Quotas cannot be enabled on SLES10.
2164 Frequency : liblustre clients only
2166 Description: getdirentries does not give error when run on compute nodes
2167 Details : getdirentries does not fail when the size specified as an argument
2168 is too small to contain at least one entry
2170 Severity : enhancement
2172 Description: Add LNET router traceability for debug purposes
2173 Details : If a checksum failure occurs with a router as part of the
2174 IO path, the NID of the last router that forwarded the bulk data
2175 is printed so it can be identified.
2180 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2181 Details : Due to a race condition, liblustre clients were occasionally
2182 evicted incorrectly.
2184 Severity : enhancement
2186 Description: lfs setstripe use optional parameters instead of postional
2189 Severity : enhancement
2191 Description: Nanosecond timestamp support for ldiskfs
2192 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2193 resolution timestamps. There is not yet support for this at
2194 the Lustre filesystem level.
2197 Frequency : during server recovery
2199 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2200 Details : request with rq_no_resend flag not awake l_wait_event if they get
2204 Frequency : nfs export on patchless client
2206 Description: connectathon hang when test nfs export over patchless client
2207 Details : Disconnected dentry cannot be found with lookup, so we do not need
2208 to unhash it or make it invalid
2211 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2212 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2213 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2218 Description: open req refcounting wrong on reconnect
2219 Details : If reconnect happened between getting open reply from server and
2220 call to mdc_set_replay_data in ll_file_open, we will schedule
2221 replay for unreferenced request that we are about to free.
2222 Subsequent close will crash in variety of ways.
2223 Check that request is still eligible for replay in
2224 mdc_set_replay_data().
2229 Description: disable writes to filesystem when reading health_check file
2230 Details : the default for reading the health_check proc file has changed
2231 to NOT do a journal transaction and write to disk, because this
2232 can cause reads of the /proc file to hang and block HA state
2233 checking on a healthy but otherwise heavily loaded system. It
2234 is possible to return to the previous behaviour during configure
2235 with --enable-health-write.
2237 Severity : enhancement
2239 Description: 64-bit inode version
2240 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2241 to the inode. This will be required for version-based recovery.
2246 Description: MDS fails to start if a duplicate client export is detected
2247 Details : in some rare cases it was possible for a client to connect to
2248 an MDS multiple times. Upon recovery the MDS would detect this
2249 and fail during startup. Handle this more gracefully.
2251 Severity : enhancement
2253 Description: Add -o localflock option to simulate old noflock
2255 Details : This will achieve local-only flock/fcntl locks
2261 Description: log_commit_thread vs filter_destroy race leads to crash
2262 Details : Take import reference before releasing llog record semaphore
2267 Description: Wrong request locking in request set processing
2268 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2269 imp_delayed_list, in this place should be used imp_lock.
2272 Frequency : when reconnection
2274 Description: Grant Leak when osc reconnect to OST
2275 Details : When osc reconnect ost, OST(filter) should check whether it
2276 should grant more space to client by comparing fed_grant and
2277 cl_avail_grant, and return the granted space to client instead
2278 of "new granted" space, because client will call osc_init_grant
2279 to update the client grant space info.
2282 Frequency : when client reconnect to OST
2284 Description: Grant Leak when osc do resend and replay bulk write
2285 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2286 bulk write request, because the grant info while be sync between
2287 OSC and OST when reconnect, and we should ignore the grant info
2288 these of resend/replay write req.
2293 Description: Grant space more than avaiable left space sometimes.
2294 Details : When then OST is about to be full, if two bulk writing from
2295 different clients came to OST. Accord the avaliable space of the
2296 OST, the first req should be permitted, and the second one
2297 should be denied by ENOSPC. But if the seconde arrived before
2298 the first one is commited. The OST might wrongly permit second
2299 writing, which will cause grant space > avaiable space.
2302 Frequency : when client is evicted
2304 Description: Grant might be wrongly erased when osc is evicted by OST
2305 Details : when the import is evicted by server, it will fork another
2306 thread ptlrpc_invalidate_import_thread to invalidate the
2307 import, where the grant will be set to 0. While the original
2308 thread will update the grant it got when connecting. So if
2309 the former happened latter, the grant will be wrongly errased
2310 because of this race.
2315 Description: Checking Stale with correct fid
2316 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2317 to check whether it is stale, because sometimes, we want the
2318 enqueue happened anyway, and op_data.fid2 will not be initialized.
2320 Severity : enhancement
2322 Description: update patchless client
2323 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2326 Frequency : only with 2.4 kernel
2328 Description: random memory corruption
2329 Details : size of struct ll_inode_info is to big for union inode.u and this
2330 can be cause of random memory corruption.
2335 Description: Memory leak in recovery
2336 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2337 It should also check obd_fail before fsfilt_start, otherwise if
2338 fsfilt_start return -EROFS,(failover mds during mds recovery).
2339 then the req will return with repmsg->transno = 0 and rc = EROFS.
2340 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2341 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2342 be freed no matter whether fsfilt_commit success or not.
2345 Frequency : only with huge count clients
2347 Description: Prevents from taking the superblock lock in llap_from_page for
2349 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2350 for llap_from_page call in ll_removepage prevents from taking the
2351 superblock lock for a soon died page.
2356 Description: Not check open intent error before release open handle
2357 Details : in some rare cases, the open intent error is not checked before
2358 release open handle, which may cause
2359 ASSERTION(open_req->rq_transno != 0), because it tries to release
2360 the failed open handle.
2365 Description: Set cat log bitmap only after create log success.
2366 Details : in some rare cases, the cat log bitmap is set too early. and it
2367 should be set only after create log success.
2371 Description: Accessing a block bevice can re-enable I/O when Lustre is
2372 tearing down a device.
2373 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2378 Description: service threads may hog cpus when there are a lot of requests
2379 Details : Insert cond_resched to give other threads a chance to use some CPU
2384 Description: the cat log was not initialized in recovery
2385 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2386 unlink log on mds will not be initialized until mds post
2387 recovery. And also in mds post recovery, the unlink log will
2388 initialization will be done asynchronausly, so there will be race
2389 between add unlink log and unlink log initialization.
2393 Description: brw_stats were being printed incorrectly
2394 Details : brw_stats were being printed as log2 but all of them were not
2395 recorded as log2. Also remove some code duplication arising from
2396 filter_tally_{read,write}.
2400 Frequency : rare, only in recovery.
2401 Description: ASSERTION(req->rq_type != LI_POISON) failed
2402 Details : imp_lock should be held while iterating over imp_sending_list for
2403 prevent destroy request after get timeout in ptlrpc_queue_wait.
2407 Description: replay-single.sh test 52 fails
2408 Details : A lock's skiplist need to be cleanup when it being unlinked
2409 from its resource list.
2413 Description: Short directio read returns full requested size rather than
2415 Details : Direct I/O operations should return actual amount of bytes
2416 transferred rather than requested size.
2418 Severity : enhancement
2420 Description: metadata RPC reduction (e.g. for rm performance)
2421 Details : decrease the amount of synchronous RPC between clients and servers
2422 by canceling conflicing lock before the operation on the client
2423 and packing thier handles into the main operation RPC to server.
2425 Severity : enhancement
2427 Description: Async OSC create to avoid the blocking unnecessarily.
2428 Details : If a OST has no remain object, system will block on the creating
2429 when need to create a new object on this OST. Now, ways use
2430 pre-created objects when available, instead of blocking on an
2431 empty osc while others are not empty. If we must block, we block
2432 for the shortest possible period of time.
2436 Description: improve handling recoverable errors
2437 Details : if request processig with error which can be recoverable on server
2438 request should be resend, otherwise page released from cache and
2441 Severity : enhancement
2443 Description: refine locking for avoid write wrong info into lov_objid file
2444 Details : fix possible races with add new target and write/update data in
2447 --------------------------------------------------------------------------------
2449 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2454 Frequency : on some architectures
2456 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2457 Details : Uninitialized flags sometimes cause configuration commands to
2461 Frequency : patchless clients only
2463 Description: missing __iget() symbol export
2464 Details : The __iget() symbol export is missing. To avoid the need for
2465 this on patchless clients the deathrow inode reaper is turned
2466 off, and we depend on the VM to clean up old inodes. This
2467 dependency was during via the fix for bug 12181.
2469 --------------------------------------------------------------------------------
2471 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2473 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2474 INTEROPERATE with older versions automatically. In many cases a
2475 special upgrade step is needed. Please read the
2476 user documentation before upgrading any part of a live system.
2477 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2478 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2479 * WARNING: Lustre configuration and startup changes are required with
2480 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2484 Severity : enhancement
2486 Description: MountConf
2487 Details : Lustre configuration is now managed via mkfs and mount
2488 commands instead of lmc and lconf. New obd types (MGS, MGC)
2489 are added for dynamic configuration management. See
2490 https://mail.clusterfs.com/wikis/lustre/MountConf for
2493 Severity : enhancement
2495 Description: dynamic OST addition
2496 Details : OSTs can now be added to a live filesystem
2498 Severity : enhancement
2500 Description: startup order invariance
2501 Details : MDTs and OSTs can be started in any order. Clients only
2502 require the MDT to complete startup.
2504 Severity : enhancement
2506 Description: parallel, asynchronous orphan cleanup
2507 Details : orphan cleanup is now performed in separate threads for each
2508 OST, allowing parallel non-blocking operation.
2510 Severity : enhancement
2512 Description: optimized stripe assignment
2513 Details : stripe assignments are now made based on ost space available,
2514 ost previous usage, and OSS previous usage, in order to try
2515 to optimize storage space and networking resources.
2517 Severity : enhancement
2519 Description: Permanently set tunables
2520 Details : All writable /proc/fs/lustre tunables can now be permanently
2521 set on a per-server basis, at mkfs time or on a live system.
2523 Severity : enhancement
2525 Description: Lustre message v2
2526 Details : Add lustre message format v2.
2528 Severity : enhancement
2530 Description: client OST exclusion list
2531 Details : Clients can be started with a list of OSTs that should be
2532 declared "inactive" for known non-responsive OSTs.
2536 Description: SPEC SFS validation failure on NFS v2 over lustre.
2537 Details : Changes the blocksize for regular files to be 2x RPC size,
2538 and not depend on stripe size.
2540 Severity : enhancement
2542 Description: Multiple MD RPCs in flight.
2543 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2544 To avoid overly-overloading MDS, introduce a limit on number of
2545 MDS RPCs in flight for a single client and add /proc controls
2546 to adjust this limit.
2548 Severity : enhancement
2550 Description: client read/write statistics
2551 Details : Add client read/write call usage stats for performance
2552 analysis of user processes.
2553 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2554 file access. extents_stats shows chunk size distribution.
2555 extents_stats_per_process show chunk size distribution per
2558 Severity : enhancement
2560 Description: mds statistics
2561 Details : Add detailed mds operations statistics in
2562 /proc/fs/lustre/mds/*/stats.
2566 Description: Failure of copying files with lustre special EAs.
2567 Details : Client side always return success for setxattr call for lustre
2568 special xattr (currently only "trusted.lov").
2573 Description: Refcount LNET uuids
2574 Details : The global LNET uuid list grew linearly with every startup;
2575 refcount repeated list entries instead of always adding to
2578 Severity : enhancement
2580 Description: Dynamic service threads
2581 Details : Within a small range, start extra service threads
2582 automatically when the request queue builds up.
2585 Frequency : mixed-endian client/server environments
2587 Description: mixed-endian crashes
2588 Details : The new msg_v2 system had some failures in mixed-endian
2591 Severity : enhancement
2593 Description: Easy OST removal
2594 Details : OSTs can be permanently deactivated with e.g. 'lctl
2595 conf_param lustre-OST0001.osc.active=0'
2597 Severity : enhancement
2599 Description: MGS proc entries
2600 Details : Added basic proc entries for the MGS showing what filesystems
2603 Severity : enhancement
2605 Description: provide MGS failover
2606 Details : Added config lock reacquisition after MGS server failover.
2608 Severity : enhancement
2610 Description: add Linux 2.4 support
2611 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2615 Description: a large application tries to do I/O to the same resource and dies
2616 in the middle of it.
2617 Details : Check the req->rq_arrival time after the call to
2618 ost_brw_lock_get(), but before we do anything about
2619 processing it & sending the BULK transfer request. This
2620 should help move old stale pending locks off the queue as
2621 quickly as obd_timeout.
2624 Frequency : when an incorrect nid is specified during startup
2626 Description: ptlrpc connect to non-existant node causes kernel crash
2627 Details : LNET can't be re-entered from an event callback, which
2628 happened when we expire a message after the export has been
2629 cleaned up. Instead, hand the zombie cleanup off to another
2632 Severity : enhancement
2634 Description: plain/inodebits lock performance improvement
2635 Details : Grouping plain/inodebits in granted list by their request modes
2636 and bits policy, thus improving the performance of search through
2640 Frequency : only if OST filesystem is corrupted
2642 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2643 Details : for a short time RPCs with bulk IO are in the replay list,
2644 but replay of bulk IOs is unimplemented. If the OST filesystem
2645 is corrupted due to disk cache incoherency and then replay is
2646 started it is possible to trip an assertion. Avoid putting
2647 committed RPCs into the replay list at all to avoid this issue.
2650 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2652 Description: System hang on startup
2653 Details : This bug allowed the liblustre (e.g. catamount) client to
2654 return to the app before handling all startup RPCs. This
2655 could leave the node unresponsive to lustre network traffic
2656 and manifested as a server ptllnd timeout.
2658 Severity : enhancement
2660 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2661 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2662 boolean to control whether to print peer debug info when a
2663 client's RPC times out.
2666 Frequency : only for kernels with patches from Lustre below 1.4.3
2668 Description: Remove old rdonly API
2669 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2672 Frequency : only for devices with external journals
2674 Description: Set external device read-only also
2675 Details : During a commanded failover stop, we set the disk device
2676 read-only while the server shuts down. We now also set any
2677 external journal device read-only at the same time.
2680 Frequency : when upgrading from 1.4 while trying to change parameters
2682 Description: The wrong (new) MDC name was used when setting parameters for
2683 upgraded MDT's. Also allows changing of OSC (and MDC)
2684 parameters if --writeconf is specified at tunefs upgrade time.
2687 Frequency : when setting specific ost indicies
2689 Description: QOS code breaks on skipped indicies
2690 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2691 created with --index need not be sequential.
2693 Severity : enhancement
2695 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2696 Details : The uninit_groups feature works in conjunction with the kernel
2697 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2698 up the pass1 processing of e2fsck. This is a read-only feature
2699 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2700 mount filesystems that have had this feature enabled.
2702 Severity : enhancement
2704 Description: Improve multi-block allocation algorithm to avoid fragmentation
2705 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2706 allocation locality and avoid filesystem fragmentation.
2708 ------------------------------------------------------------------------------
2710 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2712 * Support for kernels:
2713 2.4.21-47.0.1.EL (RHEL 3)
2714 2.6.5-7.283 (SLES 9)
2715 2.6.9-42.0.10.EL (RHEL 4)
2716 2.6.12.6 vanilla (kernel.org)
2717 2.6.16.27-0.9 (SLES 10)
2718 * Recommended e2fsprogs version: 1.39.cfs5
2720 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2724 Frequency : occasional, depends on client load and configuration
2725 Bugzilla : 12181, 12203
2726 Description: data loss for recently-modified files
2728 Details : In some cases it is possible that recently written or created
2729 files may not be written to disk in a timely manner (this should
2730 normally be within 30s unless client IO load is very high).
2731 The problem appears as zero-length files or files that are a
2732 multiple of 1MB in size after a client crash or client eviction
2733 that are missing data at the end of the file.
2735 This problem is more likely to be hit on clients where files are
2736 repeatedly created and unlinked in the same directory, clients
2737 have a large amount of RAM, have many CPUs, the filesystem has
2738 many OSTs, the clients are rebooted frequently, and/or the files
2739 are not accessed by other nodes after being written.
2741 The presence of the problem can be detected by looking at
2742 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2743 smaller than the second (nr_unused) then dirty files will not
2744 be flushed automatically to disk. "sync; sleep 10" should be
2745 run several times on the node before unmounting it to update
2746 Lustre (this is also safe to run on nodes without this problem).
2748 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2749 that can cause this same problem, so customers using that kernel
2750 also need to update the kernel in addition to Lustre. In order
2751 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2753 It is normal that files written just before a client crash (less
2754 than 30s) may not yet have been flushed to disk, even for local
2758 Frequency : frequent on thin XT3 nodes
2760 Description: UUID collision on thin XT3 Linux nodes
2761 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2762 randomly, since we relied on an insufficiently-seeded PRNG.
2767 Description: OSS hangs after "All ost request buffers busy"
2768 Details : A deadlock between quota and journal operations caused OSS
2769 hangs after printing "All ost request buffers busy."
2772 Frequency : always on liblustre builds
2774 Description: Cleanup compiler warnings on liblustre
2777 Frequency : always on liblustre builds on XT3
2779 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2783 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2784 > 32000 subdirectories
2785 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2786 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2787 subdirectory count crosses 32000. This will aid e2fsck to
2788 correctly handle more than 32000 subdirectories.
2791 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2793 Description: System hang on startup
2794 Details : This bug allowed the liblustre (e.g. catamount) client to
2795 return to the app before handling all startup RPCs. This
2796 could leave the node unresponsive to lustre network traffic
2797 and manifested as a server ptllnd timeout.
2799 Severity : enhancement
2801 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2802 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2803 boolean to control whether to print peer debug info when a
2804 client's RPC times out.
2809 Description: make O_SYNC working on 2.6 kernels
2810 Details : 2.6 kernels use different method for mark pages for write,
2811 so need add a code to lustre for O_SYNC work.
2816 Description: Failure to close file and release space on NFS
2817 Details : Put inode details into lock acquired in ll_intent_file_open.
2818 Use mdc_intent_lock in ll_intent_open to properly
2819 detect all kind of errors unhandled by mdc_enqueue.
2824 Description: proc file read during shutdown sometimes raced obd removal,
2826 Details : Add lock to prevent obd access after proc file removal.
2829 Frequency : Only for files larger than 4GB on 32-bit clients.
2831 Description: improperly doing page alignment of locks
2832 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2833 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2838 Description: RPCs being resent when they shouldn't be
2839 Details : Some RPCs that should not be resent are being resent. This
2840 can cause inconsistencies in the RPC state machine. Do not
2841 resend such requests.
2844 Frequency : rare, only with NFS export
2846 Description: Crash on NFS re-export node
2847 Details : under very unusual load conditions an assertion is hit in
2848 ll_intent_file_open()
2851 Frequency : only if OST filesystem is corrupted
2853 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2854 Details : for a short time RPCs with bulk IO are in the replay list,
2855 but replay of bulk IOs is unimplemented. If the OST filesystem
2856 is corrupted due to disk cache incoherency and then replay is
2857 started it is possible to trip an assertion. Avoid putting
2858 committed RPCs into the replay list at all to avoid this issue.
2863 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2864 Details : Large single O_DIRECT read and write calls can fail to allocate
2865 a sufficiently large buffer to process the request. In case of
2866 allocation failure the allocation is retried with a smaller
2867 buffer and broken into smaller requests.
2869 Severity : enhancement
2871 Description: Add -o localflock option to simulate old noflock behaviour.
2872 Details : This will achieve local-only flock/fcntl locks coherentness.
2877 Description: versioning check is incomplete
2878 Details : Checking the version difference of client vs. server, report
2879 error if the gap is too big.
2884 Description: add support PG_writeback bit
2885 Details : add support for PG_writeback bit for Lustre, for more carefull
2886 work with page cache in 2.6 kernel. This also fix some deadlocks
2887 and remove hack for work O_SYNC with 2.6 kernel.
2889 Severity : enhancement
2891 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2892 Details : The uninit_groups feature works in conjunction with the kernel
2893 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2894 up the pass1 processing of e2fsck. This is a read-only feature
2895 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2896 mount filesystems that have had this feature enabled.
2898 Severity : enhancement
2900 Description: Improve multi-block allocation algorithm to avoid fragmentation
2901 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2902 allocation locality and avoid filesystem fragmentation.
2904 ------------------------------------------------------------------------------
2906 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2908 * Support for kernels:
2909 2.6.9-42.0.3.EL (RHEL 4)
2910 2.6.5-7.276 (SLES 9)
2911 2.4.21-47.0.1.EL (RHEL 3)
2912 2.6.12.6 vanilla (kernel.org)
2913 2.6.16.21-0.8 (SLES10)
2914 * Recommended e2fsprogs version: 1.39.cfs2-0
2916 * The backwards-compatible /proc/sys/portals symlink has been removed
2917 in this release. Before upgrading, please ensure that you change
2918 any configuration scripts or /etc/sysctl.conf files that access
2919 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2920 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2921 in advance of the upgrade on any system running Lustre 1.4.6 or
2922 newer, since /proc/sys/lnet was added in that version.
2923 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2927 Frequency : only when quota is used
2929 Description: avoid scanning export list for quota master
2930 Details : Change the algorithms to avoid scanning export list in order
2931 to improve the efficiency.
2934 Frequency : MDS failover only, very rarely
2936 Description: "went back in time" messages on mds failover
2937 Details : The greatest transno may be lost when the current operation
2938 finishes with an error (transno==0) and the client's last_rcvd
2939 record is over-written. Save the greatest transno in the
2940 mds_last_transno for this case.
2943 Frequency : always for specific kernels and striping counts
2945 Description: client may get "Matching packet too big" without ACL support
2946 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2947 when trying to access files in certain configurations. The
2948 clients should in fact be denied when mounting because they do
2949 not understand ACLs.
2952 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2954 Description: many clients connecting with IO in progress causes connect timeouts
2955 Details : Avoid synchronous journal commits to avoid delays caused by many
2956 clients connecting/disconnecting when bulk IO is in progress.
2957 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2958 OST_IO_PORTAL to avoid delays behind potentially many pending
2962 Frequency : occasionally with multiple writers to a single file
2964 Description: shared writes to file may result in wrong size reported by stat()
2965 Details : Allow growing of kms when extent lock is cancelled
2968 Frequency : always with random mmap IO to multi-striped file
2970 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2971 Details : Only if the hole is at the end of OST object so that kms is too
2972 small. Fix is to increase kms accordingly in ll_nopage.
2975 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2977 Description: writes to a missing object would leak memory on the OST
2978 Details : If there is an inconsistency between the MDS and OST filesystems,
2979 such that the MDS references an object that doesn't exist, writes
2980 to that object will leak memory due to incorrect cleanup in the
2981 error handling path, eventually running out of memory on the OST.
2986 Description: Creating too long symlink causes lustre errors
2987 Details : Check symlink and name lengths before sending requests to MDS.
2990 Frequency : only if flock is enabled (not on by default)
2992 Description: posix locks not released on fd closure on 2.6.9+
2993 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2994 kernels, this caused such locks not to be released on fd close and
2995 then assertions on fs unmount about still used locks.
2998 Frequency : MDS failover only, very rarely
3000 Description: clients may get ASSERTION(granted_lock != NULL)
3001 Details : When request was taking a long time, and a client was resending
3002 a getattr by name lock request. The were multiple lock requests
3003 with the same client lock handle and
3004 mds_getattr_name->fixup_handle_for_resent_request found one of the
3005 lock handles but later failed with ASSERTION(granted_lock != NULL).
3010 Description: handle->h_buffer_credits > 0, assertion failure
3011 Details : h_buffer_credits is zero after truncate, causing assertion
3012 failure. This patch extends the transaction or creates a new
3016 Frequency : NFS re-export or patchless client
3017 Bugzilla : 11179, 10796
3018 Description: Crash on NFS re-export node (__d_move)
3019 Details : We do not want to hash the dentry if we don't have a lock.
3020 But if this dentry is later used in d_move, we'd hit uninitialised
3021 list head d_hash, so we just do this to init d_hash field but
3022 leave dentry unhashed.
3025 Frequency : NFS re-export or patchless client
3027 Description: NFS exports has problem with symbolic link
3028 Details : lustre client didn't properly install dentry when re-exported
3029 to NFS or running patchless client.
3032 Frequency : NFS re-export or patchless client
3034 Description: Various nfs/patchless fixes.
3035 Details : fixes reuse disconected alias for lookup process - this fixes
3036 warning "find_exported_dentry: npd != pd",
3037 fix permission error with open files at nfs.
3038 fix apply umask when do revalidate.
3041 Frequency : occasional
3043 Description: Crash on NFS re-export node
3044 Details : calling clear_page() on the wrong pointer triggered oops in
3045 generic_mapping_read().
3048 Frequency : rarely, using O_DIRECT IO
3050 Description: unaligned directio crashes client with LASSERT
3051 Details : check for unaligned buffers before trying any requests.
3054 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3056 Description: stale data returned from RAID cache
3057 Details : If only a small amount of IO is done to the RAID device before
3058 reading it again it is possible to get stale data from the RAID
3059 cache instead of reading it from disk.
3062 Frequency : always for sles10 kernel
3064 Description: sles10 support
3065 Details : ll_follow_link: compile fixes and using of nd_set_link
3066 under newer kernels.
3069 Frequency : depends on arch, kernel and compiler version, always on sles10
3072 Description: recursive or deep enough symlinks cause stack overflow
3073 Details : getting rid of large stack-allocated variable in
3077 Frequency : depends on hardware
3079 Description: lustre write performance loss in the SLES10 kernel
3080 Details : the performance loss is caused by using of write barriers in the
3081 ext3 code. The SLES10 kernel turns barrier support on by
3082 default. The fix is to undo that change for ldiskfs.
3084 ------------------------------------------------------------------------------
3086 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3088 * Support for kernels:
3089 2.6.9-42.0.3EL (RHEL 4)
3090 2.6.5-7.276 (SLES 9)
3091 2.4.21-47.0.1.EL (RHEL 3)
3092 2.6.12.6 vanilla (kernel.org)
3096 Frequency : quota enabled and large files being deleted
3098 Description: releasing more than 4GB of quota at once hangs OST
3099 Details : If a user deletes more than 4GB of files on a single OST it
3100 will cause the OST to spin in an infinite loop. Release
3101 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3106 Description: statfs data retrieved from /proc may be stale or zero
3107 Details : When reading per-device statfs data from /proc, in the
3108 {kbytes,files}_{total,free,avail} files, it may appear
3109 as zero or be out of date.
3112 Frequency : systems with MD RAID1 external journal devices
3114 Description: lconf's call to blkid is confused by RAID1 journal devices
3115 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3116 of returning all block devices that match the journal UUID.
3119 Frequency : always, for aggregate stripe size over 4GB
3121 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3122 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3123 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3126 Frequency : always if "lfs find" used on a local file/directory
3128 Description: "lfs find" segfaults if used on a local file/directory
3129 Details : The case where a directory component was not specified wasn't
3130 handled correctly. Handle this properly.
3133 Frequency : always on ppc64
3135 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3136 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3137 perform endianness conversion.
3140 Frequency : rarely (truncate to non-zero file size after write under load)
3141 Bugzilla : 10730, 10687
3142 Description: Files padded with zeros to next 4K multiple
3143 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3144 it is possible that files that are truncated to a non-zero size
3145 immediately after being written are filled with zero bytes beyond
3146 the truncated size. No file data is lost.
3148 Severity : enhancement
3150 Description: Allow recovery/failover for liblustre clients.
3151 Details : liblustre clients were unaware of failover configurations until
3154 Severity : enhancement
3156 Description: user file locks should fail when not mounting with flock option
3157 Details : Set up an error-returning stub in ll_file_operations.lock field
3158 to prevent incorrect behaviour when client is mounted without
3159 flock option. Also, set up properly f_op->flock field for
3163 Frequency : always on ia64
3165 Description: "lfs df" loops on printing out MDS statfs information
3166 Details : The obd_ioctl_data was not initialized and in some systems
3167 this caused a failure during the ioctl that did not return
3168 an error. Initialize the struct and return an error on failure.
3171 Frequency : SLES 9 only
3173 Description: Error of copying files with lustre special EAs as root
3174 Details : Client side always return success for setxattr call for lustre
3175 special xattr (currently only "trusted.lov").
3178 Frequency : rarely on clusters with both ia64+i386 clients
3180 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3181 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3182 clients are doing concurrent writes to the same file it is
3183 possible that the ia64 clients may LASSERT because the OST
3184 extent locks are not PAGE_SIZE aligned. Ensure that grown
3185 locks are always aligned on the request boundary.
3188 Frequency : specific use, occasional
3190 Description: Overwriting in use executable truncates on-disk binary image
3191 Details : If one node attempts to overwrite an executable in use by
3192 another node, we now correctly return ETXTBSY instead of
3193 truncating the file.
3195 Severity : enhancement
3197 Description: Async OSC create to avoid the blocking unnecessarily.
3198 Details : If a OST has no remain object, system will block on the creating
3199 when need to create a new object on this OST. Now, ways use
3200 pre-created objects when available, instead of blocking on an
3201 empty osc while others are not empty. If we must block, we block
3202 for the shortest possible period of time.
3207 Description: chmod on Lustre root is propagated to other clients
3208 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3209 invalid by the follow_mount time.
3212 Frequency : liblustre clients only
3214 Description: Race in 'instant cancel' lock handling could lead to such locks
3215 never to be granted in case of SMP MDS
3216 Details : Do not destroy not yet granted but cbpending locks in
3220 Frequency : replay/resend of open
3222 Description: non null lock assetion failure in mds_intent_policy
3223 Details : Trying to replay/resend lockless open requests resulted in
3224 mds_open() returning 0 with no lock. Now it sets a flag if
3225 a lock is going to be returned.
3227 Severity : enhancement
3229 Description: Checksum enhancements
3230 Details : New checksum enhancements allow for resending RPCs that failed
3233 Severity : enhancement
3235 Description: Tunables on number of dirty pages in cacche
3236 Details : Allow to set limit on number of dirty pages cached.
3241 Description: client crash on unmount - lock still has references
3242 Details : In some error handling cases it was possible to leak a lock
3243 reference on a client while accessing a file. This was not
3244 harmful to the client during operation, but would cause the
3245 client to crash when the filesystem is unmounted.
3248 Frequency : specific case, rare
3250 Description: ETXTBSY on mds though file not in use by client
3251 Details : ETXTBSY is no longer incorrectly returned when attempting to
3252 chmod or chown a directory that the user previously tried to
3253 execute or a currently-executing binary.
3256 Frequency : extremely rare except on liblustre-based clients
3258 Description: Lustre space not freed when files are deleted
3259 Details : Clean up open-unlinked files after client eviction. Previously
3260 the unlink was skipped and the files remained as orphans.
3265 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3266 Details : In some cases it was possible to send a blocking callback to
3267 a client doing a glimpse, even though that client didn't get
3268 a lock granted. When the glimpse lock is cancelled on the OST
3269 the freed lock is left on the waiting list and corrupted the list.
3272 Frequency : all core dumps
3274 Description: Broke core dumps to lustre
3275 Details : Negative dentry may be unhashed if parent does not have UPDATE
3276 lock, but some callers, e.g. do_coredump, expect dentry to be
3277 hashed after successful create, hash it in ll_create_it.
3279 ------------------------------------------------------------------------------
3281 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3283 * Support for kernels:
3284 2.6.9-42.0.2.EL (RHEL 4)
3285 2.6.5-7.276 (SLES 9)
3286 2.4.21-40.EL (RHEL 3)
3287 2.6.12.6 vanilla (kernel.org)
3291 Frequency : always on RHEL 3
3293 Description: Number of open files grows over time
3294 Details : The number of open files grows over time, whether or not
3295 Lustre is started. This was due to a filp leak introduced
3296 by one of our kernel patches.
3298 ------------------------------------------------------------------------------
3300 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3302 * Support for kernels:
3303 2.6.9-42.EL (RHEL 4)
3304 2.6.5-7.276 (SLES 9)
3305 2.4.21-40.EL (RHEL 3)
3306 2.6.12.6 vanilla (kernel.org)
3311 Bugzilla : 5719, 9635, 9792, 9684
3312 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3313 Details : If a server is under heavy load and cannot reply to new
3314 connection requests before the client resends the (re)connect,
3315 the connection handling code can behave badly if two service
3316 threads are concurrently handing separate (re)connections from
3317 the same client. Add better locking to the connection handling
3318 code, and ensure that only a single connection will be processed
3319 for a given client UUID, even if the lock is dropped.
3321 Severity : enhancement
3323 Description: add TCP zero-copy support to kernel
3324 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3325 sends if the hardware supports scatter-gather and checksumming.
3326 This allows socklnd to do client-write and server-read more
3327 efficiently and reduce CPU utilization from skbuf copying.
3330 Frequency : only if NFS exporting from client
3332 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3333 Details : If setting attributes on a file created under NFS that had
3334 never been opened it would be possible to oops the client
3335 if the file had no objects.
3338 Frequency : always for liblustre
3340 Description: liblustre client does MDS+OSTs setattr RPC for each write
3341 Details : When doing a write from a liblustre client, the client
3342 incorrectly issued an RPC to the MDS and each OST the file was
3343 striped over in order to update the timestamps. When writing
3344 with small chunks and many clients this could overwhelm the MDS
3345 with RPCs. In all cases it would slow down the write because
3346 these RPCs are unnecessary.
3348 Severity : enhancement
3350 Description: allow number of MDS service threads to be changed at module load
3351 Details : It is now possible to change the number of MDS service threads
3352 running. Adding "options mds mds_num_threads={N}" to the MDS's
3353 /etc/modprobe.conf will set the number of threads for the next
3354 time Lustre is restarted (assuming the "mds" module is also
3355 reloaded at that time). The default number of threads will
3356 stay the same, 32 for most systems.
3361 Description: OST crash if filesystem is unformatted or corrupt
3362 Details : If an OST is started on a device that has never been formatted
3363 or if the filesystem is corrupt and cannot even mount then the
3364 error handling cleanup routines would dereference a NULL pointer.
3369 Description: NULL pointer deref in llap_from_page.
3370 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3371 incorrect) mapping if the page was truncated/reclaimed while it was
3372 searched for. Check for this condition and skip such pages when
3373 doing readahead. Introduce extra check to llap_from_page() to
3374 verify page->mapping->host is non-NULL (so page is not anonymous).
3377 Frequency : Sometimes when using sys_sendfile
3379 Description: "page not covered by a lock" warnings from ll_readpage
3380 Details : sendfile called ll_readpage without right page locks present.
3381 Now we introduced ll_file_sendfile that does necessary locking
3382 around call to generic_file_sendfile() much like we do in
3386 Frequency : with certain MDS communication failures at client mount time
3388 Description: NULL pointer deref after failed client mount
3389 Details : a client connection request may delayed by the network layer
3390 and not be sent until after the PTLRPC layer has timed out the
3391 request. If the client fails the mount immediately it will try
3392 to clean up before the network times out the request. Add a
3393 reference from the request import to the obd device and delay
3394 the cleanup until the network drops the request.
3397 Frequency : occasionally during client (re)connect
3399 Description: assertion failure during client (re)connect
3400 Details : processing a client connection request may be delayed by the
3401 client or server longer than the client connect timeout. This
3402 causes the client to resend the connection request. If the
3403 original connection request is replied in this interval, the
3404 client may trip an assertion failure in ptlrpc_connect_interpret()
3405 which thought it would be the only running connect process.
3408 Frequency : only with obd_echo servers and clients that are rebooted
3410 Description: kernel BUG accessing uninitialized data structure
3411 Details : When running an obd_echo server it did not start the ping_evictor
3412 thread, and when a client was evicted an uninitialized data
3413 structure was accessed. Start the ping_evictor in the RPC
3414 service startup instead of the OBD startup.
3416 Severity : enhancement
3417 Bugzilla : 10193 (patchless)
3418 Description: Remove dependency on various unexported kernel interfaces.
3419 Details : No longer need reparent_to_init, exit_mm, exit_files,
3420 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3423 Frequency : rare (only users of deprecated and unsupported LDAP config)
3425 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3426 Details : LDAP apparently contains 'lustreName' attributes instead of
3427 'name'. A simple remapping of the name is sufficient.
3430 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3432 Description: waiting_locks_callback trips kernel BUG if client is evicted
3433 Details : Running with the dump_on_timeout debug flag turned on makes
3434 it possible that the waiting_locks_callback() can try to dump
3435 the Lustre kernel debug logs from an interrupt handler. Defer
3436 this log dumping to the expired_lock_main() thread.
3438 Severity : enhancement
3440 Description: Support NFS exporting on 2.6 kernels.
3441 Details : Implement non-rawops metadata methods for NFS server to use without
3442 changing NFS server code.
3445 Frequency : very rare (synthetic metadata workload only)
3447 Description: two racing renames might cause an MDS thread to deadlock
3448 Details : Running the "racer" program may cause one MDS thread to rename
3449 a file from being the source of a rename to being the target of
3450 a rename at exactly the same time that another thread is doing
3451 so, and the second thread has already enqueued these locks after
3452 doing a lookup of the target and is trying to relock them in
3453 order. Ensure that we don't try to re-lock the same resource.
3456 Frequency : only very large systems with liblustre clients
3458 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3459 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3460 all OSTs in parallel. This allows the eviction of stale liblustre
3461 clients to proceed much faster than if they were done in series,
3462 and also offers similar improvements for other set_info RPCs.
3467 Description: excessive CPU usage during initial read phase on client
3468 Details : During the initial read phase on a client, it would agressively
3469 retry readahead on the file, consuming too much CPU and impacting
3470 performance (since 1.4.5.8). Improve the readahead algorithm
3471 to avoid this, and also improve some other common cases (read
3472 of small files in particular, where "small" is files smaller than
3473 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3478 Description: MDS crash when receiving packet with unknown intent.
3479 Details : Do not LBUG in unknown intent case, just return -EFAULT
3481 Severity : enhancement
3482 Bugzilla : 9293, 9385
3483 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3484 Details : Do not serialize getattr (non-intent version) and statfs.
3487 Frequency : occasional, when OST network is overloaded/intermittent
3489 Description: client evicted by OST after bulk IO timeout
3490 Details : If a client sends a bulk IO request (read or write) the OST
3491 may evict the client if it is unresposive to its data GET/PUT
3492 request. This is incorrect if the network is overloaded (takes
3493 too long to transfer the RPC data) or dropped the OST GET/PUT
3494 request. There is no need to evict the client at all, since
3495 the pinger and/or lock callbacks will handle this, and the
3496 client can restart the bulk request.
3499 Frequency : Always when mmapping file with no objects
3501 Description: client crashes when mmapping file with no objects
3502 Details : Check that we actually have objects in a file before doing any
3503 operations on objects in ll_vm_open, ll_vm_close and
3509 Description: Request leak when working with deleted CWD
3510 Details : Introduce advanced request refcount tracking for requests
3511 referenced from lustre intent.
3513 Severity : Enhancement
3515 Description: Cache open file handles on client.
3516 Details : MDS now will return special lock along with openhandle, if
3517 requested and client is allowed to hold openhandle, even if unused,
3518 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3519 closing files for every read/write openration.
3521 Severity : Enhancement
3523 Description: Cache open negative dentries on client when possible.
3524 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3525 negative dentries on lock revocation.
3530 Description: Remounting a client read-only wasn't possible with a zconf mount
3531 Details : It wasn't possible to remount a client read-only with llmount.
3533 Severity : enhancement
3534 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3535 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3536 MPICH is included to add Lustre-specific ADIO interfaces.
3537 This is based closely on the UFS ADIO layer and only differs
3538 in file creation, in order to allow the OST striping to be set.
3539 This is user-contributed code and not supported by CFS.
3544 Description: extended inode attributes (immutable, append-only) work improperly
3545 when 2.4 and 2.6 kernels are used on client/server or vice versa
3546 Details : Introduce kernel-independent values for these flags.
3548 Severity : enhancement
3551 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3552 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3553 granularity. Now, Fractional MB granularity is be supported,
3554 this is very useful for low memory system.
3556 Severity : enhancement
3558 Description: Getattr by fid
3559 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3560 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3561 have locks and avoids getattr rpc requests for it on every stat.
3564 Frequency : Always, for filesystems larger than 2TB
3566 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3567 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3568 filesystems larger than 2TB to be mounted with mballoc enabled.
3571 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3573 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3574 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3575 filesystems larger than 2TB is mounted then the kernel will
3576 silently corrupt the start of the filesystem. CONFIG_LBD is
3577 enabled for all CFS-supported kernels, but the possibility of
3578 this happening with a modified kernel config exists.
3580 Severity : enhancement
3582 Description: add client O_DIRECT support for 2.6 kernels
3583 Details : It is now possible to do O_DIRECT reads and writes to files
3584 in the Lustre client mountpoint on 2.6 kernel clients.
3586 Severity : enhancement
3588 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3589 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3590 parallel, not waiting for response from every OST before sending
3591 a rpc to the next OST.
3596 Description: setattr vs write race when updating file timestamps
3597 Details : Client processes that update a file timestamp into the past
3598 right after writing to the file (e.g. tar) it is possible that
3599 the updated file modification time can be reset to the current
3600 time due to a race between processing the setattr and write RPC.
3602 Severity : enhancement
3604 Description: Bring 'lfs find' closer in line with regular Linux find.
3605 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3606 -print0 options and obtains all the needed info through the lustre
3609 Severity : enhancement
3611 Description: support up to 1024 configured devices on one node
3612 Details : change obd_dev array from statically allocated to dynamically
3613 allocated structs as they are first used to reduce memory usage
3618 Description: Flush dirty partially truncated pages during truncate
3619 Details : Immediatelly flush partially truncated pages in filter_setattr,
3620 this way we completely avoid having any pages in page cache on OST
3621 and can retire ugly workarounds during writes to flush such pages.
3626 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3627 Details : protect inode from truncation within vfs_unlink() context
3628 just take a reference before calling vfs_unlink() and release it
3629 when parent's i_sem is free.
3634 Description: last_id value checked outside lock on OST caused LASSERT failure
3635 Details : If there were multiple MDS->OST object precreate requests in
3636 flight, it was possible that the OST's last object id was checked
3637 outside a lock and incorrectly tripped an assertion. Move checks
3638 inside locks, and discard old precreate requests.
3641 Frequency : always, if extents are used on OSTs
3643 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3644 Details : OSTs using the extents format would not zero the high 16 bits of
3645 the index physical block number. This is not a problem for any
3646 OST filesystems smaller than 16TB, and no kernels support ext3
3647 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3648 new/modified files) and can be fixed for existing filesystems
3649 with e2fsprogs-1.39-cfs1.
3654 Description: import connection selection may be incorrect if timer wraps
3655 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3656 import connections to be ignored if the 32-bit jiffies counter
3657 wraps. Use a 64-bit jiffies counter.
3660 Frequency : very large clusters immediately after boot
3662 Description: LNET request buffers exhausted under heavy short-term load
3663 Details : If a large number of client requests are generated on a service
3664 that has previously never seen so many requests it is possible
3665 that the request buffer growth cannot keep up with the spike in
3666 demand. Instead of dropping incoming requests, they are held in
3667 the LND until the RPC service can accept more requests.
3670 Frequency : Sometimes during replay
3672 Description: Assertion failure in ll_local_open after replay.
3673 Details : If replay happened on an open request reply before we were able
3674 to set replay handler, reply will become not swabbed tripping the
3675 assertion in ll_local_open. Now we set the handler right after
3676 recognising of open request
3679 Frequency : very rare
3681 Description: kernel reports "badness in vsnprintf"
3682 Details : Reading from the "recovery_status" /proc file in small chunks
3683 may cause a negative length in lprocfs_obd_rd_recovery_status()
3684 call to vsnprintf() (which is otherwise harmless). Exit early
3685 if there is no more space in the output buffer.
3687 Severity : enhancement
3689 Description: clear OBD RPC statistics by writing to them
3690 Details : It is now possible to clear the OBD RPC statistics by writing
3691 to the "stats" file.
3696 Description: Client mtime is not the same on different clients after utimes
3697 Details : In some cases, the client was using the utimes() syscall on
3698 a file cached on another node. The clients now validate the
3699 ctime from the MDS + OSTs to determine which one is right.
3704 Description: Inability to activate failout mode
3705 Details : lconf script incorrectly assumed that in pythong string's numeric
3706 value is used in comparisons.
3709 Frequency : always with multiple stripes per file
3711 Description: Inefficient object allocation for mutli-stripe files
3712 Details : When selecting which OSTs to stripe files over, for files with
3713 a stripe count that divides evenly into the number of OSTs,
3714 the MDS is always picking the same starting OST for each file.
3715 Return the OST selection heuristic to the original design.
3720 Description: mount failures may take full timeout to return an error
3721 Details : Under some heavy load conditions it is possible that a
3722 failed mount can wait for the full obd_timeout interval,
3723 possibly several minutes, before reporting an error.
3724 Instead return an error as soon as the status is known.
3726 Frequency : quota enabled and large files being deleted
3728 Description: releasing more than 4GB of quota at once hangs OST
3729 Details : If a user deletes more than 4GB of files on a single OST it
3730 will cause the OST to spin in an infinite loop. Release
3731 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3736 Description: statfs data retrieved from /proc may be stale or zero
3737 Details : When reading per-device statfs data from /proc, in the
3738 {kbytes,files}_{total,free,avail} files, it may appear
3739 as zero or be out of date.
3742 Frequency : systems with MD RAID1 external journal devices
3744 Description: lconf's call to blkid is confused by RAID1 journal devices
3745 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3746 of returning all block devices that match the journal UUID.
3749 Frequency : always, for aggregate stripe size over 4GB
3751 Description: assertion fails when trying to use 4GB stripe size
3752 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3753 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3756 Frequency : always on ppc64
3758 Description: the first write on an ext3 filesystem with mballoc got stuck
3759 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3760 perform endianness conversion.
3762 ------------------------------------------------------------------------------
3764 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3766 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3767 INTEROPERATE with older versions automatically. Please read the
3768 user documentation before upgrading any part of a live system.
3769 * WARNING: Lustre networking configuration changes are required with
3770 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3773 * Support for kernels:
3774 2.6.9-22.0.2.EL (RHEL 4)
3775 2.6.5-7.244 (SLES 9)
3776 2.6.12.6 vanilla (kernel.org)
3779 Severity : enhancement
3780 Bugzilla : 7981/8208
3781 Description: Introduced Lustre Networking (LNET)
3782 Details : LNET is new networking infrastructure for Lustre, it includes
3783 a reorganized network configuration mode (see the user
3784 documentation for full details) as well as support for routing
3785 between different network fabrics. Lustre Networking Devices
3786 (LNDs) for the supported network fabrics have also been
3787 created for this new infrastructure.
3789 Severity : enhancement
3790 Description: Introduced Access control lists
3791 Details : clients can set ACLs on files and directories in order to have
3792 more fine-grained permissions than the standard Unix UGO+RWX.
3793 The MDS must be started with the "-o acl" mount option.
3795 Severity : enhancement
3796 Description: Introduced filesystem quotas
3797 Details : Administrators may now establish per-user quotas on the
3800 Severity : enhancement
3802 Description: Configuration change for the XT3
3803 The PTLLND is now used to run Lustre over Portals on the XT3
3804 The configure option(s) --with-cray-portals are no longer used.
3805 Rather --with-portals=<path-to-portals-includes> is used to
3806 enable building on the XT3. In addition to enable XT3 specific
3807 features the option --enable-cray-xt3 must be used.
3812 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3813 Details : A race between allocating a new llcd and re-getting the llcd_lock
3814 allowed another thread to grab newly-allocated llcd.
3816 Severity : enhancement
3818 Description: 2.6 OST async journal commit and locking fix to improve performance
3819 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3820 2.6 kernels are now async as they already were in 2.4 kernels so
3821 that they can commit concurrently with the network bulk transfer.
3822 For block-allocated files the filter allocation semaphore is held
3823 to avoid filesystem fragmentation during allocation. BKL lock
3824 removed for 2.6 xattr operations where it is no longer needed.
3829 Description: lconf incorrectly determined whether two IP networks could talk
3830 Details : In some more complicated routing and multiple-network
3831 configurations, lconf will avoid trying to make a network
3832 connection to a disjoint part of the IP space. It was doing the
3833 math incorrectly for one set of cases.
3838 Description: Fix for potential infinite loop processing records in an llog.
3839 Details : If an llog record is corrupted/zeroed, it is possible to loop
3840 forever in llog_process(). Validate the llog record length
3841 and skip the remainder of the block on error.
3844 Frequency : occasional (liblustre only)
3846 Description: liblustre could not open files whose last component is a symlink
3847 Details : sysio_path_walk() would incorrectly pass the open intent to
3848 intermediate path components.
3851 Frequency : rare (liblustre only with non-standard tuning)
3852 Bugzilla : 7201 (7350)
3853 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3854 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3855 after it has been released, only do this for referenced locks
3857 Severity : enhancement
3859 Description: specifying an (invalid) directory default stripe_size of -1
3860 would reset the directory default striping
3861 Details : stripe_size -1 was used internally to signal directory stripe
3862 removal, now use "all default" to signal dir stripe removal
3863 as a directory striping of "all default" is not useful
3866 Frequency : common for large clusters running liblustre clients
3868 Description: doing an ls when liblustre clients are running is slow
3869 Details : sending a glimpse AST to a liblustre client waits for every AST
3870 to time out, as liblustre clients will not respond. Since they
3871 cannot cache data we refresh the OST lock LVB from disk instead.
3873 Severity : enhancement
3875 Description: doing an ls at the same time as file IO can be slow
3876 Details : enqueue and other "small" requests can be blocked behind many
3877 large IO requests. Create a new OST IO portal for non-IO
3878 requests so they can be processed faster.
3881 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3883 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3884 created file with mode 000
3885 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3886 allow an ftruncate() as a non-root user to a file with mode 000.
3887 Lustre now respects this flag to disable mode checks when
3888 truncating a file owned by the user
3891 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3894 Description: Revoking locks from clients that went dead or catatonic might take
3896 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3897 cancellation of such locks instant on servers without waiting for
3898 any reply from clients. Clients drops these locks when cancel
3899 notification from server is received without replying.
3902 Frequency : liblustre-only, when liblustre client dies or becomes busy
3904 Description: Doing ls on Linux clients can take a long time with active
3906 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3907 granting such locks to it in the first place if possible. Locks
3908 are taken by proxy on the OST during the read or write and
3909 dropped immediately afterward. Add connect flags handling, do
3910 not grant locks to liblustre clients for glimpse ASTs.
3912 Severity : enhancement
3914 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3915 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3916 sequential reads and behaves badly with applications doing random
3917 reads. Improve it by reading ahead at least read region, and
3918 avoiding excessive large RPC for small reads.
3920 Severity : enhancement
3922 Description: Creating more than 1000 files for a single job may cause a load
3923 imbalance on the OSTs if there are also a large number of OSTs.
3924 Details : qos_prep_create() uses an OST index reseed value that is an
3925 even multiple of the number of available OSTs so that if the
3926 reseed happens in the middle of the object allocation it will
3927 still utilize the OSTs as uniformly as possible.
3932 Description: OST or MDS may oops in ping_evictor_main()
3933 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3934 but doesn't restart at beginning of obd_exports_timed list
3937 Severity : enhancement
3939 Description: improve by-nid export eviction on the MDS and OST
3940 Details : allow multiple exports with the same NID to be evicted at one
3941 time without re-searching the exports list.
3944 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3946 Description: MDS may oops in groups_free()
3947 Details : in rare race conditions a newly allocated group_info struct is
3948 freed again, and this can be NULL. The 2.4 compatibility code
3949 for groups_free() checked for a NULL pointer, but 2.6 did not.
3952 Frequency : common for liblustre clients doing little filesystem IO
3953 Bugzilla : 9352, 7313
3954 Description: server may evict liblustre clients accessing contended locks
3955 Details : if a client is granted a lock or receives a completion AST
3956 with a blocking AST already set it would not reply to the AST
3957 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3960 Frequency : lfs setstripe, only systems with more than 160 OSTs
3962 Description: unable to set striping with a starting offset beyond OST 160
3963 Details : llapi_create_file() incorrectly limited the starting stripe
3964 index to the maximum single-file stripe count.
3967 Frequency : LDAP users only
3969 Description: lconf did not handle in-kernel recovery with LDAP properly
3970 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3972 Severity : enhancement
3974 Description: bind OST threads to NUMA nodes to improve performance
3975 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3976 node and do their allocations there to localize memory access
3978 Severity : enhancement
3980 Description: llmount can determine client NID directly from Myrinet (GM)
3981 Details : the client NID code from gmnalnid was moved directly into
3982 llmount, removing the need to use this or specifying the
3983 client NID explicitly when mounting GM clients with zeroconf
3986 Frequency : if client is started with down MDS
3988 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3989 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3990 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3991 ptlrpc_import_delayed_req() because we didn't check if the
3992 request was interrupted, and we also didn't break out of the
3993 event loop if there was no timeout
3998 Description: data loss during non-page-aligned writes to a single file from
3999 both multiple nodes and multiple threads on one node at same time
4000 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4001 inconsistency led to false short-reads, that were cached and later
4002 used by ->prepare_write() to fill in partially written page,
4003 leading to data loss.
4006 Frequency : always, if lconf --abort_recovery used
4008 Description: lconf --abort_recovery fails with 'Operation not supported'
4009 Details : lconf was attempting to abort recovery on the MDT device and not
4012 Severity : enhancement
4014 Description: remove cleanup logs
4015 Details : replace lconf-generated cleanup logs with lustre internal
4016 cleanup routines. Eliminates the need for client-cleanup and
4019 Severity : enhancement
4021 Description: add support for EAs (user and system) on lustre filesystems
4022 Details : it is now possible to store extended attributes in the Lustre
4023 client filesystem, and with the user_xattr mount option it
4024 is possible to allow users to store EAs on their files also
4026 Severity : enhancement
4028 Description: Add possibility (config option) to show minimal available OST free
4030 Details : When compiled with --enable-mindf configure option, statfs(2)
4031 (and so, df) will return least minimal free space available from
4032 all OSTs as amount of free space on FS, instead of summary of
4033 free spaces of all OSTs.
4035 Severity : enhancement
4037 Description: do not expand extent locks acquired on OST-side
4038 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4039 by server: they are not cached anyway.
4042 Frequency : when mmap is used/binaries executed from Lustre
4044 Description: Unmmap pages before throwing them away from read cache.
4045 Details : llap_shrink cache now attempts to unmap pages before discarding
4046 them (if unmapping failed - do not discard). SLES9 kernel has
4047 extra checks that trigger if this unmapping is not done first.
4052 Description: lconf didn't resolve symlinks before checking to see whether a
4053 given mountpoint was already in use
4056 Frequency : when migrating failover services
4057 Bugzilla : 6395, 9514
4058 Description: When migrating a subset of services from a node (e.g. failback
4059 from a failover service node) the remaining services would
4060 time out and evict clients.
4061 Details : lconf --force (implied by --failover) sets the global obd_timeout
4062 to 5 seconds in order to quickly disconnect, but this caused
4063 other RPCs to time out too quickly. Do not change the global
4064 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4066 Severity : enhancement
4067 Frequency : if MDS is started with down OST
4068 Bugzilla : 9439,5706
4069 Description: Allow startup/shutdown of an MDS without depending on the
4070 availability of the OSTs.
4071 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4072 Add appropriate locking and lov-osc refcounts for safe
4073 cleaning. Add osc abort_inflight calls in case the
4074 synchronize never started.
4077 Frequency : occasional (Cray XT3 only)
4079 Description: root not authorized to access files in CRAY_PORTALS environment
4080 Details : The client process capabilities were not honoured on the MDS in
4081 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4082 been accessed by an authorized user then root was able to access
4083 the file on the local client also. The root user capabilities
4084 are now allowed on the MDS, as this environment has secure UID.
4087 Frequency : occasional
4089 Description: ldiskfs "too long searching" message happens too often
4090 Details : A debugging message (otherwise harmless) prints too often on
4091 the OST console. This has been reduced to only happen when
4092 there are fragmentation problems on the filesystem.
4097 Description: Division by zero in statfs when all OSCs are inactive
4098 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4099 lov_statfs divides by value returned by lov_get_stripecnt().
4103 Bugzilla : 9489, 3273
4104 Description: First write from each client to each OST was only 4kB in size,
4105 to initialize client writeback cache, which caused sub-optimal
4106 RPCs and poor layout on disk for the first writen file.
4107 Details : Clients now request an initial cache grant at (re)connect time
4108 and so that they can start streaming writes to the cache right
4109 away and always do full-sized RPCs if there is enough data.
4110 If the OST is rebooted the client also re-establishes its grant
4111 so that client cached writes will be honoured under the grant.
4116 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4117 Details : Now I/O RPCs go to different portal number and (presumably) fast
4118 lock requests (and glimses) and other RPCs get their own service
4119 threads pool that should be able to service those RPCs
4122 Severity : enhancement
4124 Description: Ability to exchange lustre version between client and servers and
4125 issue warnings at client side if client is too old. Also for
4126 liblustre clients there is ability to refuse connection of too old
4128 Details : New 'version' field is added to connect data structure that is
4129 filled with version info. That info is later checked by server and
4133 Frequency : rare, liblustre only.
4134 Bugzilla : 9296, 9581
4135 Description: Two simultaneous writes from liblustre at offset within same page
4136 might proceed at the same time overwriting eachother with stale
4138 Details : I/O lock withing llu_file_prwv was released too early, before data
4139 actually was hitting the wire. Extended lock-holding time until
4140 server acknowledges receiving data.
4143 Frequency : extremely rare. Never observed in practice.
4145 Description: avoid generating lustre_handle cookie of 0.
4146 Details : class_handle_hash() generates handle cookies by incrementing
4147 global counter, and can hit 0 occasionaly (this is unlikely, but
4148 not impossible, because initial value of cookie counter is
4149 selected randonly). Value of 0 is used as a sentinel meaning
4150 "unassigned handle" --- avoid it. Also coalesce two critical
4151 sections in this function into one.
4153 Severity : enhancement
4155 Description: allow liblustre clients to delegate truncate locking to OST
4156 Details : To avoid overhead of locking, liblustre client instructs OST to
4157 take extent lock in ost_punch() on client's behalf. New connection
4158 flag is added to handle backward compatibility.
4160 Severity : enhancement
4161 Bugzilla : 4928, 7341, 9758
4162 Description: allow number of OST service threads to be specified
4163 Details : a module parameter allows the number of OST service threads
4164 to be specified via "options ost ost_num_threads={N}" in the
4165 OSS's /etc/modules.conf or /etc/modprobe.conf.
4169 Bugzilla : 6146, 9635, 9895
4170 Description: servers crash with bad pointer in target_handle_connect()
4171 Details : In rare cases when a client is reconnecting it was possible that
4172 the connection request was the last reference for that export.
4173 We would temporarily drop the export reference and get a new
4174 one, but this may have been the last reference and the export
4175 was just destroyed. Get new reference before dropping old one.
4177 Severity : enhancement
4178 Frequency : if client is started with failover MDS
4180 Description: Allow multiple MDS hostnames in the mount command
4181 Details : Try to read the configuration from all specified MDS
4182 hostnames during a client mount in case the "primary"
4185 Severity : enhancement
4187 Description: Stop sending data to evicted clients as soon as possible.
4188 Details : Check if the client we are about to send or are sending data to
4189 was evicted already. (Check is done every second of waiting,
4190 for which l_wait_event interface was extended to allow checking
4191 of exit condition at specified intervals).
4194 Frequency : rare, normally only when NFS exporting is done from client
4196 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4198 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4201 Frequency : always, if srand() is called before liblustre initialization
4203 Description: Liblustre uses system PRNG disturbing its usage by user application
4204 Details : Introduce internal to lustre fast and high-quality PRNG for
4205 lustre usage and make liblustre and some other places in generic
4206 lustre code to use it.
4208 Severity : enhancement
4209 Bugzilla : 9477, 9557, 9870
4210 Description: Verify that the MDS configuration logs are updated when xml is
4211 Details : Check if the .xml configuration logs are newer than the config
4212 logs stored on the MDS and report an error if this is the case.
4213 Request --write-conf, or allow starting with --old_conf.
4215 Severity : enhancement
4217 Description: Handle symlinks in the path when checking if Lustre is mounted.
4218 Details : Resolve intermediate symlinks when checking if a client has
4219 mounted a filesystem to avoid duplicate client mounts.
4224 Description: lconf can hit an error exception but still return success.
4225 Details : The lconf command catches the Command error exception at the top
4226 level script context and will exit with the associated exit
4227 status, but doesn't ensure that this exit status is non-zero.
4232 Description: failure of ptlrpc thread startup can cause oops
4233 Details : Starting a ptlrpc service thread can fail if there are a large
4234 number of threads or the server memory is very fragmented.
4235 Handle this without oopsing.
4238 Frequency : always, only if liblustre and non-default acceptor port was used
4240 Description: liblustre cannot connect to servers with non-default acceptor port
4241 Details : tcpnal_set_default_params() was not called and was therefore
4242 ignoring the environment varaible TCPNAL_PORT, as well as other
4243 TCPNAL_ environment variables
4248 Description: two objects could be created on the same OST for a single file
4249 Details : If an OST is down, in some cases it was possible to create two
4250 objects on a single OST for a single file. No problems other
4251 than potential performance impact and spurious error messages.
4255 Bugzilla : 5681, 9562
4256 Description: Client may oops in ll_unhash_aliases
4257 Details : Client dcache may become inconsistent in race condition.
4258 In some cases "getcwd" can fail if the current directory is
4264 Description: Inode refcounting problems in NFS export code
4265 Details : link_raw functions used to call d_instantiate without obtaining
4266 extra inode reference first.
4270 Bugzilla : 9942, 9903
4271 Description: Referencing freed requests leading to crash, memleaks with NFS.
4272 Details : We used to require that call to ll_revalidate_it was always
4273 followed by ll_lookup_it. Also with revalidate_special() it is
4274 possible to call ll_revalidate_it() twice for the same dentry
4275 even if first occurence returned success. This fix changes semantic
4276 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4277 reference on a request referred from the intent.
4278 ll_intent_release() then releases such a request.
4281 Frequency : rare, normally benchmark loads only
4283 Description: unlinked inodes were kept in memory on the client
4284 Details : If a client is repeatedly creating and unlinking files it
4285 can accumulate a lot of stale inodes in the inode slab cache.
4286 If there is no other client load running this can cause the
4287 client node to run out of memory. Instead flush old inodes
4288 from client cache that have the same inode number as a new inode.
4291 Frequency : SLES9 2.6.5 kernel and long filenames only
4292 Bugzilla : 9969, 10379
4293 Description: utime reports stale NFS file handle
4294 Details : SLES9 uses out-of-dentry names in some cases, which confused
4295 the lustre dentry revalidation. Change it to always use the
4299 Frequency : rare, unless heavy write-truncate concurrency is continuous
4300 Bugzilla : 4180, 6984, 7171, 9963, 9331
4301 Description: OST becomes very slow and/or deadlocked during object unlink
4302 Details : filter_destroy() was holding onto the parent directory lock
4303 while truncating+unlinking objects. For very large objects this
4304 may block other threads for a long time and slow overall OST
4305 responsiveness. It may also be possible to get a lock ordering
4306 deadlock in this case, or run out of journal credits because of
4307 the combined truncate+unlink. Solution is to do object truncate
4308 first in one transaction without parent lock, and then do the
4309 final unlink in a new transaction with the parent lock. This
4310 reduces the lock hold time dramatically.
4313 Frequency : rare, 2.4 kernels only
4315 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4316 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4317 before unmounting their filesystem in order to avoid deadlock.
4318 The kernel_locked() function in 2.4 kernels only checks whether
4319 the kernel lock is held, not whether it is this process that is
4320 holding it as 2.6 kernels do.
4325 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4326 Details : The client ptlrpc code may be trying to reconnect to a down
4327 server before a previous connection attempt has timed out.
4328 Increase the reconnect interval to be longer than the connection
4329 timeout interval to avoid sending duplicate connections to
4333 Frequency : echo_client brw_test command
4335 Description: fix echo_client to work with OST preallocated code
4336 Details : OST preallocation code (5137) didn't take echo_client IO path
4337 into account: echo_client calls filter methods outside of any
4338 OST thread and, hence, there is no per-thread preallocated
4339 pages and buffers to use. Solution: hijack pga pages for IO. As
4340 a byproduct, this avoids unnecessary data copying.
4344 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4345 Description: Client can oops in mdc_commit_close() after open replay
4346 Details : It was possible for the MDS to return an open request with no
4347 transaction number in mds_finish_transno() if the client was
4348 evicted, but without actually returning an error. Clients
4349 would later try to replay that open and may trip an assertion
4350 Simplify the client close codepath, and always return an error
4351 from the MDS in case the open is not successful.
4354 Frequency : rare, 2.6 OSTs only
4356 Description: OST may deadlock under high load on fragmented files
4357 Details : If there was a heavy load and highly-fragmented OST filesystems
4358 it was possible to have all the OST threads deadlock waiting on
4359 allocation of biovecs, because the biovecs were not released
4360 until the entire RPC IO was completed. Instead, release biovecs
4361 as soon as they are complete to ensure forward IO progress.
4363 Severity : enhancement
4365 Description: Support for specifying external journal device at mount
4366 Details : If an OST or MDS device is formatted with an external journal
4367 device, this device major/minor is stored in the ext3 superblock
4368 and may not be valid for failover. Allow detecting and
4369 specifying the external journal at mount time.
4374 Description: Mounting an MDS with pending unlinked files may cause oops
4375 Details : target_finish_recovery() calls mds_postrecov() which returned
4376 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4377 considers this an error and immediately begins cleaning up the
4378 lov, just after starting the mds_lov process
4380 Severity : enhancement
4382 Description: Implement 'lfs df' to report actual free space on per-OST basis
4383 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4384 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4385 usage of objects; '-h' to report in human readable format.
4387 ------------------------------------------------------------------------------
4389 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4396 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4397 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4398 at startup during orphan unlinking. Instead, skip buddy update
4399 before setup, as it will be regenerated after recovery is complete.
4404 Description: If an OST is inactive, its locks might reference stale inodes.
4405 Details : lov_change_cbdata() must iterate over all namespaces, even if
4406 they are inactive to clear inode references from the lock.
4408 Severity : enhancement
4409 Frequency : occasional, if non-standard max_dirty_mb used
4411 Description: Client will block write RPCs if not enough grant
4412 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4413 then the client will block writes while waiting for another RPC
4414 to complete instead of consuming its dirty limit. With change
4415 we get improved performance when max_dirty_mb is small.
4417 Severity : enhancement
4418 Bugzilla : 3389, 6253
4419 Description: Add support for supplementary groups on the MDS.
4420 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4421 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4422 lookups for user supplementary groups into a cache.
4426 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4427 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4428 for liblustre O_EXCL is enforced by MDS.
4432 Description: readdir never returns NULL in liblustre.
4433 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4434 dentry in struct dirent.
4438 Frequency : liblustre only, depends on application IO pattern
4439 Description: liblustre clients evicted if not contacting servers
4440 Details : Don't put liblustre clients into the ping_evictor list, so
4441 they will not be evicted by the pinger ever.
4443 Severity : enhancement
4445 Description: Add ability to evict clients by NID from MDS.
4446 Details : By echoing "nid:$NID" string into
4447 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4448 $NID would be instantly evicted from this MDS and from all active
4449 OSTs connected to it.
4453 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4454 Details : lookup_it_finish() used to query file size from OSTs that was not
4459 Description: service threads change working directory to that of init
4460 Details : Starting lustre service threads may pin the working directory
4461 of the parent thread, making that filesystem busy. Threads
4462 now change to the working directory of init to avoid this.
4466 Frequency : during shutdown only
4467 Description: shutdown with a failed MDS or OST can cause unmount to hang
4468 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4473 Frequency : 2.6 only
4474 Description: chmod/chown may include an extra supplementary group
4475 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4476 supplementary group and if none is specified this is used.
4479 Bugzilla : 5479 (6816)
4481 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4482 Details : If lookup is in progress on a file that is unlinked we might try
4483 to revalidate the inode and fail in revalidate after lookup is
4484 complete and ll_file_open() enqueues the open again but
4485 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4488 Frequency : always, if lconf --abort_recovery used
4490 Description: lconf --abort_recovery fails with 'Operation not supported'
4491 Details : lconf was attempting to abort recovery on the MDT device and not
4494 ------------------------------------------------------------------------------
4496 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4501 Frequency : rare (only unsupported configurations with a node running as an
4503 Bugzilla : 6514, 5137
4504 Description: Mounting a Lustre file system on a node running as an OST could
4506 Details : OSTs now preallocates memory needed to write out data at
4507 startup, instead of when needed, to avoid having to
4508 allocate memory in possibly low memory situations.
4509 Specifically, if the file system is mounted on on OST,
4510 memory pressure could force it to try to write out data,
4511 which it needed to allocate memory to do. Due to the low
4512 memory, it would be unable to do so and the node would
4513 become unresponsive.
4515 Severity : enhancement
4517 Description: Addition of lconf --service command line option
4518 Details : lconf now accepts a '--service <arg>' option, which is
4519 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4521 Severity : enhancement
4523 Description: Failover mode is now the default for OSTs.
4524 Details : By default, OSTs will now run in failover mode. To return to
4525 the old behaviour, add '--failout' to the lmc line for OSTs.
4527 Severity : enhancement
4529 Description: Health checks are now provided for MDS and OSTs
4530 Details : Additional detailed health check information on MSD and OSTs
4531 is now provided through the procfs health_check value.
4534 Frequency : occasional, depends on IO load
4536 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4537 after numerous create/delete cycles
4538 Details : The ext3 inode allocation policy would not allocate new inodes
4539 very well on the OSTs because there are no new directories
4540 being created. Instead we look for groups with free space if
4541 the parent directories are nearly full.
4546 Description: Network or server problems during mount may cause partially
4547 mounted clients instead of returning an error.
4548 Details : The config llog parsing code may overwrite the error return
4549 code during mount error handling, returning success instead
4555 Description: MDS can fail to allocate large reply buffers
4556 Details : After long uptimes the MDS can fail to allocate large reply
4557 buffers (e.g. zconf client mount config records) due to memory
4558 fragmentation or consumption by the buffer cache. Preallocate
4559 some large reply buffers so that these replies can be sent even
4560 under memory pressure.
4564 Frequency : rare (liblustre)
4565 Description: fsx running with liblustre complained that using truncate() to
4566 extend the file doesn't work. This patch corrects that issue.
4567 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4568 ATTR_SIZE and lsm use in llu_setattr_raw.
4572 Frequency : rare, only 2.6 kernels
4573 Description: Unusual file access patterns on the MDS may result in inode
4574 data being lost in very rare circumstances.
4575 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4576 code in the 2.6 kernel caused the inode and/or EA data not to
4577 be read from disk, causing single-file corruption.
4581 Frequency : rare, only 2.6 filesystems using extents
4582 Description: Heavy concurrent write and delete load may cause data corruption.
4583 Details : It was possible under high-load situations to have an extent
4584 metadata block in the block device cache from a just-unlinked
4585 file overwrite a newly-allocated data block. We now unmap any
4586 metadata buffers that alias just-allocated data blocks.
4590 Frequency : filesystems with default stripe_count larger than 77
4591 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4592 Details : lconf specifies an inode size of 4096 bytes when the default
4593 stripe_count is larger than 77. This conflicts with the default
4594 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4595 this case to avoid pinning too much memory for large EAs.
4597 ------------------------------------------------------------------------------
4599 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4604 Frequency : rare (extremely heavy IO load with hundreds of clients)
4606 Description: Client is evicted, gets IO error writing to file
4607 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4608 caused clients to be evicted for AST timeouts. The fixes in
4609 bug 5192 mean we no longer need to have such short AST timeouts
4610 so ldlm_timeout has been increased.
4613 Frequency : occasional during --force or --failover shutdown under load
4614 Bugzilla : 5949, 4834
4615 Description: Server oops/LBUG if stopped with --force or --failover under load
4616 Details : a collection of import/export refcount and cleanup ordering
4617 issues fixed for safer force cleanup
4620 Frequency : only filesystems larger than 120 OSTs
4621 Bugzilla : 5990, 6223
4622 Description: lfs getstripe would oops on a very large filesystem
4623 Details : lov_getconfig used kfree on vmalloc'd memory
4626 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4627 Bugzilla : 6242, 6243
4628 Description: reading from files that had been truncated to a non-zero size
4629 but never opened returned no data
4630 Details : ll_file_read() reads zeros from no-object files to EOF
4635 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4636 Details : There's little harm in aborting MDS/OSS recovery and letting it
4637 try again, so I removed the LASSERT and return an error instead.
4639 Severity : enhancement
4641 Description: New debugging infrastructure for tracking down data corruption
4642 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4643 (b) cover more of the client-side code path, and (c) try to narrow
4644 down where problems occurred
4648 Bugzilla : 3819, 4364, 4397, 6313
4649 Description: Racing close and eviction MDS could cause assertion in mds_close
4650 Details : It was possible to get multiple mfd references during close and
4651 client eviction, leading to one thread referencing a freed mfd.
4653 Severity: : enhancement
4654 Bugzilla : 3262, 6359
4655 Description: Attempts to reconnect to servers are now more aggressive.
4656 Details : This builds on the enhanced upcall-less recovery that was added
4657 in 1.4.2. When trying to reconnect to servers, clients will
4658 now try each server in the failover group every 10 seconds. By
4659 default, clients would previously try one server every 25 seconds.
4664 Description: After recovery, certain operations trigger a failed
4665 assertion on a client.
4666 Details : Failing over an mds, using lconf -d --failover, while a
4667 client was doing a readdir() call would cause the client to
4668 LBUG after recovery completed and the readdir() was resent.
4670 Severity : enhancement
4672 Description: Default groups are now added by lconf
4673 Details : You can now run lconf --group <servicename> without having to
4674 manually add groups with lmc.
4677 Frequency : occasional
4679 Description: Nodes with an elan id of 0 trigger a failed assertion
4682 Frequency : always when accessing e.g. tty/console device nodes
4684 Description: tty and some other devices nodes cannot be used on lustre
4685 Details : file's private_data field is used by device data and lustre
4686 values in there got lost. New field was added to struct file to
4687 store fs-specific private data.
4690 Frequency : when exporting Lustre via NFS
4692 Description: NFSD failed occasionally when looking up a path component
4693 Details : NFSD is looking up ".." which was broken in ext3 directories
4694 that had grown large enough to become hashed.
4697 Frequency : Clusters with multiple interfaces not on the same subnet
4699 Description: Nodes will repeatedly try to reconnect to an interface which it
4700 cannot reach and report an error to the log.
4701 Details : Extra peer list entries will be created by lconf with some peers
4702 unreachable. lconf now validates the peer before adding it.
4705 Frequency : Only if a default stripe is set on the filesystem root.
4707 Description: Setting a default stripe on the filesystem root prevented the
4708 filesystem from being remounted.
4709 Details : The client was sending extra request flags in the root getattr
4710 request and did not allocate a reply buffer for the dir EA.
4713 Frequency : occasional, higher if lots of files are accessed by one client
4714 Bugzilla : 6159, 6097
4715 Description: Client trips assertion regarding lsm mismatch/magic
4716 Details : While revalidating inodes the VFS looks up inodes with ifind()
4717 and in rare cases can find an inode that is being freed.
4718 The ll_test_inode() code will free the lsm during ifind()
4719 when it finds an existing inode and then the VFS later attaches
4720 this free lsm to a new inode.
4724 Bugzilla : 6422, 7030
4725 Description: MDS deadlock between mkdir and client eviction
4726 Details : Creating a new file via mkdir or mknod (starting a transaction
4727 and getting the ns lock) can deadlock with client eviction
4728 (gets ns lock and trying to finish a synchronous transaction).
4731 Frequency : occasional
4732 Description: While starting a server, the fsfilt_ext3 module could not be
4734 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4735 kernels. Previously, lconf would still use the ext3 name
4736 when trying to load modules. Now, it will correctly use
4737 ext3 on 2.4 and ldiskfs on 2.6.
4739 Severity : enhancement
4740 Description: The default stripe count has been changed to 1
4741 Details : The interpretation of the default stripe count (0, to lfs
4742 or lmc) has been changed to mean striping across a single
4743 OST, rather than all available. For general usage we have
4744 found a stripe count of 1 or 2 works best.
4746 Severity : enhancement
4747 Description: Add support for compiling against Cray portals.
4748 Details : Conditional compiling for some areas that are different
4752 Frequency : occasional
4753 Bugzilla : 6409, 6834
4754 Description: Creating files with an explicit stripe count may lead to
4755 a failed assertion on the MDS
4756 Details : If some OSTs are full or unavailable, creating files may
4757 trigger a failed assertion on the MDS. Now, Lustre will
4758 try to use other servers or return an error to the
4762 Frequency : occasional
4764 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4765 caused lock timeout during readahead (since 1.4.2).
4766 Details : Processes doing readahead might match a lock that hasn't been
4767 granted yet if there are overlapping and conflicting lock
4768 requests. The readahead process waits on ungranted lock
4769 (original lock is CBPENDING), while OST waits for that process
4770 to cancel CBPENDING read lock and eventually evicts client.
4772 Severity : enhancement
4774 Description: Initial enabling of flock support for clients
4775 Details : Implements fcntl advisory locking and file status functions.
4776 This feature is provided as an optional mount flag (default
4777 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4778 locking are implemented yet, and those that are are not guaranteed
4779 to be completely correct in production environments.
4780 mount -t lustre -o [flock|noflock] ...
4783 Frequency : occasional
4785 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4786 assertion in the ext3 JBD (journaling) layer.
4787 Details : The b_committed_data struct is protected by the big kernel lock
4788 in 2.4 kernels, serializing journal_commit_transaction() and
4789 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4790 access to this struct. In 2.6 kernels there is finer grained
4791 locking to improve SMP performance of the JBD layer.
4795 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4800 Description: merge and cleanup kernel patches.
4803 -----------------------------------------------------------------------------
4805 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4807 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4808 versions of Lustre. Please update all servers and clients to
4809 version 1.4.2 or later at the same time. You must also run
4810 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4811 to update the configuration logs.
4813 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4814 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4815 - swab llog records of type '0' so we get proper header size/idx (5861)
4816 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4817 - fix rename of one directory over another leaking an inode (5953)
4818 - avoid SetPageDirty on 2.6 (5981)
4819 - don't re-add just-being-destroyed locks to the waiting list (5653)
4820 - when creating new directories, inherit the parent's custom
4821 striping settings if present parent (3048)
4822 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4823 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4824 - don't hold client locks on temporary worklist from l_lru (5666)
4825 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4826 - automatically evict dead clients (5921)
4827 - Update file size properly in create+truncate+fstat case (6196)
4828 - Do not unhash mountpoint dentries, do not allow removal of
4830 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4831 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4832 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4833 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4834 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4835 - Correctly handle reads of files with no objects (6243)
4836 - lctl recover will also mark a device active if deactivate used (5933)
4838 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4839 - allow --write-conf on an MDS with different nettype than client (5619)
4840 - don't write config llogs to MDS for mounts not from that MDS (5617)
4841 - lconf should create multiple TCP connections from a client (5201)
4842 - init scripts are now turned off by default; run chkconfig --on
4843 lustre and chkconfig --on lustrefs to use them
4844 - upcalls are no longer needed for clients to recover to failover
4846 - add --abort-recovery option to lconf to abort recovery on device
4848 - add support for an arbitrary number of OSTs (3026)
4849 - Quota support protocol changes.
4850 - forward compatibility changes to wire structs (6007)
4851 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4852 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4853 - improved "lustre status" script
4854 - initialize blocksize for non-regular files (6062)
4855 - added --disable-server and --disable-client configure options (5782)
4856 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4857 - Vanilla 2.4.29 support
4858 - increase maximum number of obd devices to 520 (6242)
4859 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4860 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4861 SLES 9 2.6.5 kernels
4862 - limit stripes per file to 160 (the maximum EA size) (6093)
4864 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4867 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4868 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4869 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4870 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4871 - small ext3 extents cleanups and fixes (5733)
4872 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4873 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4874 - avoid needless client->OST connect, fix handle mismatch (5317)
4875 - fix DLM error path that led to out-of-sync client, long delays (5779)
4876 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4877 - fix several locking issues related to i_size (5492,5624,5654,5672)
4878 - don't move pending lock onto export if it is already evicted (5683)
4879 - fix kernel oops when creating .foo in unlinked directory (5548)
4880 - fix deadlock in obdfilter statistics vs. object create (5811)
4881 - use time_{before,after} to avoid timer jiffies wrap (5882)
4882 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4883 - Do not leak request if server was not able to process it (5154)
4884 - If mds_open unable to find parent dir, make that negative lookup(5154)
4885 - don't create new directories with extent-mapping (5909, 5936)
4887 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4888 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4889 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4890 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4891 - support for RHEL4 kernel on i686 (5773)
4892 - provide error messages when incompatible logs are encountered (5898)
4894 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4895 * version 1.4.0.10 (1.4.1 release candidate 1)
4897 - don't keep a lock reference when lock is not granted (4238)
4898 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4899 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4900 - drop import inflight refcount on signal_completed_replay error (5255)
4901 - unlock page after async write error during send (3677)
4902 - handle missing objects in filter_preprw_read properly (5265)
4903 - no transno return for symlink open, don't save no-trasno open (3440)
4904 - don't try to complete elan receive that already failed (4012)
4905 - free RPC server reply state on error (5406)
4906 - clean up thread from ptlrpc_start_thread() on error (5160)
4907 - readahead could read extra page into cache that wasn't ejected (5388)
4908 - prevent races in class_attach/setup/cleanup/detach (5260)
4909 - don't dereference de->d_inode after l_dput of de (5458)
4910 - use "int" for stripe value returned from lock_to_stripe (5544)
4911 - mballoc allocation and error-checking fixes in 2.6 (5504)
4912 - block device patches to fix I/O request sizes in 2.6 (5482)
4913 - look up hostnames for IB nals (5602)
4914 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4915 - don't start multiple acceptors for the same port (5277)
4916 - fix incorrect LASSERT in mds_getattr_name (5635)
4917 - export a proc file for general "ping" checking (5628)
4918 - fix "lfs check" to not block when the MDS is down (5628)
4920 - service request history (4965)
4921 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4922 - create an "evict_client" /proc entry on OSTs, like the MDS has
4923 - fix mount usage message, return errors per mount(8) (5168)
4924 - change grep [] to grep "[]" in tests so they work in more UMLs
4925 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4926 - remove ancient LOV_MAGIC_V0 EA support (5047)
4927 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4928 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4929 - allow readahead allocations to fail when low on memory (5383)
4930 - mmap locking landed again, after considerable improvement (2828)
4931 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4933 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4936 - send OST transaction number in read/write reply to free req (4966)
4937 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4938 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4939 - the watchdog thread now runs as interruptible (5246)
4940 - flock/lockf fixes (but it's still disabled, pending 5135)
4941 - don't use EXT3 constants in llite code (5094)
4942 - memory shortage at startup could cause assertion (5176)
4944 - reorganization of lov code
4945 - single portals codebase
4947 - add extents/mballoc support (5025)
4948 - direct I/O reads in the obdfilter (4048)
4949 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4951 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4954 - send OST transaction number in read/write reply to free req (4966)
4955 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4956 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4957 - the watchdog thread now runs as interruptible (5246)
4958 - handle missing objects in filter_preprw_read properly (5265)
4959 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4960 - drop import inflight refcount on signal_completed_replay error (5255)
4961 - unlock page after async write error during send (3677)
4962 - return original error code on reconstructed replies (3761)
4963 - no transno return for symlink open, don't save no-trasno open (3440)
4965 - add pid to ldlm debugging output (4922)
4966 - bump the watchdog timeouts -- we can't handle 30sec yet
4967 - extra debugging for orphan dentry/inode bug (5259)
4969 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4972 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4973 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4974 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4975 - fix NULL dereference in /proc/sys/portals/routes (4827)
4976 - allow failed mdc_close() operations to be interrupted (4561)
4977 - stop precreate on OST before MDS would time out on it (4778)
4978 - don't send partial-page writes before EOF from client (4410)
4979 - discard client grant for sub-page writes on large-page clients (4520)
4980 - don't free dentries not owned by NFS code, check generation (4806)
4981 - fix lsm leak if mds_create_objects() fails (4801)
4982 - limit debug_daemon file size, always print CERROR messages (4789)
4983 - use transno after validating reply (3892)
4984 - process timed out requests if import state changes (3754)
4985 - update mtime on OST during writes, return in glimpse (4829)
4986 - add mkfsoptions to LDAP (4679)
4987 - use ->max_readahead method instead of zapping global ra (5039)
4988 - don't interrupt __l_wait_event() during strace
4990 - add software watchdogs to catch hung threads quickly (4941)
4991 - make lustrefs init script start after nfs is mounted
4992 - fix CWARN/ERROR duplication (4930)
4993 - return async write errors to application if possible (2248)
4994 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4995 - print NAL number in %x format (4645)
4996 - update barely-supported suse-2.4.21-171 series (4842)
4997 - support for sles 9 %post scripts
4998 - support for building 2.6 kernel-source packages
4999 - support for sles km_* packages
5001 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5004 - ignore -ENOENT errors in osc_destroy (3639)
5005 - notify osc create thread that OSC is being cleaned up (4600)
5006 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5007 - reconstruct ost_handle() like mds_handle() (4657)
5008 - create a new thread to do import eviction to avoid deadlock (3969)
5009 - let lconf resolve symlinked-to devices (4629)
5010 - don't unlink "objects" from directory with default EA (4554)
5011 - hold socknal file ref over connect in case target is down (4394)
5012 - allow more than 32000 subdirectories in a single directory (3244)
5013 - fix blocks count for O_DIRECT writes (3751)
5014 - OST returns ENOSPC from object create when no space left (4539)
5015 - don't send truncate RPC if file size isn't changing (4410)
5016 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5017 - bind to privileged port in socknal and tcpnal (3689)
5019 - rate limit CERROR/CWARN console message to avoid overload (4519)
5020 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5021 - basic mmap support (3918)
5022 - kernel patch series update from b1_4 (4711)
5024 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5027 - avoid crash during MDS cleanup with OST shut down (2775)
5028 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5029 - don't use bad inodes on the MDS (3744)
5030 - dynamic object preallocation to improve recovery speed (4236)
5031 - don't hold spinlock over lock dumping or change debug flags (4401)
5032 - don't zero obd_dev when it is force cleaned (3651)
5033 - print grants to console if they go negative (4431)
5034 - "lctl deactivate" will stop automatic recovery attempts (3406)
5035 - look for existing locks in ldlm_handle_enqueue() (3764)
5036 - don't resolve lock handle twice in recovery avoiding race (4401)
5037 - revalidate should check working dir is a directory (4134)
5039 - don't always mark "slow" obdfilter messages as errors (4418)
5041 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5044 - don't close LustreDB during write_conf until it is done (3860)
5045 - fix typo in lconf for_each_profile (3821)
5046 - allow dumping logs from multiple threads at one time (3820)
5047 - don't allow multiple threads in OSC recovery (3812)
5048 - fix debug_size parameters (3864)
5049 - fix mds_postrecov to initialize import for llog ctxt (3121)
5050 - replace config semaphore with spinlock (3306)
5051 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5052 - don't allow enqueue to complete on a destroyed export (3822)
5053 - down write_lock before checking llog header bitmap (3825)
5054 - recover from lock replay timeout (3764)
5055 - up llog sem before sending rpc (3652)
5056 - reduce ns lock hold times when setting kms (3267)
5057 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5058 - fix NULL deref and obd_dev leak on setup error (3312)
5059 - replace some LBUG about llog ops with error handling (3841)
5060 - don't match INVALID dentries from d_lookup and spin (3784)
5061 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5062 - fix invalid assertion in ptlrpc_set_wait (3880)
5064 - add libwrap support for the TCP acceptor (3996)
5065 - add /proc/sys/portals/routes for non-root route listing (3994)
5066 - allow setting MDS UUID in .xml (2580)
5067 - print the stack of a process that LBUGs (4228)
5069 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5072 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5073 - make sure to unset replay flag from failed open requests (3440)
5074 - if default stripe count is 0, use OST count for inode size (3636)
5075 - update parent mtime/ctime on client for create/unlink (2611)
5076 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5077 - free recovery state on server during a forced cleanup (3571)
5078 - unregister_reply for resent reqs (3063)
5079 - loop back devices mounting and status check on 2.6 (3563)
5080 - fix resource-creation race that can provoke i_size == 0 (3513)
5081 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5082 - more debugging for page-accounting assertion (3746)
5083 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5084 - avoid deadlock after precreation failure (3758)
5085 - fix race and lock order deadlock in orphan handling (3450, 3750)
5086 - add validity checks when grabbing inodes from l_ast_data (3599)
5088 - add /proc/.../recovery_status to obdfilter (3428)
5089 - lightweight CDEBUG infrastructure, debug daemon (3668)
5090 - change default OSC RPC parameters to be better on small clusters
5091 - turn off OST read cache for files smaller than 32MB
5092 - install man pages and include them in rpms (3100)
5093 - add new init script for (un)mounting lustre filesystems (2593)
5094 - run chkconfig in %post for init scripts (3701)
5095 - drop scimac NAL (unmaintained)
5097 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5100 - clean kiobufs before and after use (3485)
5101 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5102 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5103 - change init script to fail more clearly if not run as root (1528)
5104 - allow clients to reconnect during replay (1742)
5105 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5106 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5107 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5108 - fix cond_resched (3554)
5109 - only evict unfinished clients after recovery (3515)
5110 - allow bulk resend, prevent data loss (3570)
5111 - dynamic ptlrpc request buffer allocation (2102)
5112 - don't allow unlinking open directory if it isn't empty (2904)
5113 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5114 - remove extraneous obd dereference causing LASSERT failure (3334)
5115 - don't use get_cycles() when creating temp. files on the mds (3156)
5116 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5117 - handle EEXIST for set-stripe, set proper directory name (3336)
5119 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5120 - fix ksocknal_fmb_callback() error messages (2918)
5122 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5125 - don't copy lvb into (possibly NULL) reply on error (2983)
5126 - don't deref dentry after dput, don't free lvb on error (2922)
5127 - use the kms to determine writeback rpc length (2947)
5128 - increment oti_logcookies when osc is inactive (2948)
5129 - update client's i_blocks count via lvb messages (2543)
5130 - handle intent open/close of special files properly (1557)
5131 - mount MDS with errors=remount-ro, like obdfilter (2009)
5132 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5133 - don't use cancelling-locks' kms values (2947)
5134 - use highest lock extent for kms, not last one (2925)
5135 - don't dereference ERR_PTR() dentry in error handling path (3107)
5136 - fix thread race in portals_debug_dumplog() (3122)
5137 - create lprocfs device entries at setup instead of at attach (1519)
5138 - common AST error handler, don't evict client on completion race (3145)
5139 - zero nameidata in detach_mnt in 2.6 (3118)
5140 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5141 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5142 - initialize RPC timeout timer earlier for 2.6 (3219)
5143 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5144 - print nal/nid for unknown nid (3258)
5145 - additional checks for oscc recovery before doing precreate (3284)
5146 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5147 - don't crash in mdc_close for bad permissions on open (3285)
5148 - zero i_rdev for non-device files (3147)
5149 - clear page->private before handing to FS, better assertion (3119)
5150 - tune the read pipeline (3236)
5151 - fix incorrect decref of invalidated dentry (2350)
5152 - provide read-ahead stats and refine rpc in flight stats (3328)
5153 - don't hold journal transaction open across create RPC (3313)
5154 - update atime on MDS at close time (3265)
5155 - close LDAP connection when recovering to avoid server load (3315)
5156 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5157 - don't leak open file on MDS after open resend (3325)
5158 - serialize filter_precreate and filter_destroy_precreated (3329)
5159 - loop device shouldn't call sync_dev() for nul device (3092)
5160 - clear page cache after eviction (2766)
5161 - resynchronize MDS->OST in background (2824)
5162 - refuse to mount the same filesystem twice on same mountpoint (3394)
5163 - allow llmount to create routes for mounting behind routers (3320)
5164 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5165 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5166 - fix rare problem with rename on htree directories (3417)
5168 - allow default OST striping configuration per directory (1414)
5169 - fix compilation for qswnal for 2.6 kernels (3125)
5170 - increase maximum number of MDS request buffers for large systems
5171 - change liblustreapi to be useful for external progs like lfsck (3098)
5172 - increase local configuration timeout for slow disks (3353)
5173 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5175 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5178 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5179 - don't overwrite extent policy data in reply if lock was blocked (2901)
5180 - drop filter export grants atomically with removal from device (2663)
5181 - del obd_self_export from work_list in class_disconnect_exports (2908)
5182 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5183 - swab reply message in mdc_close, other PPC fixes (2464)
5184 - fix destroying of named logs (2325)
5185 - overwrite old logs when running lconf --write_conf (2264)
5186 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5187 - fix race in target_handle_connect (2898)
5188 - mds_reint_create() should take same inode create lock (2926)
5189 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5190 - don't close files for self_export to avoid uninitialized obd (2936)
5191 - allow MDS with the same name as client node (2939)
5192 - hold dentry reference for closed log files for unlink (2325)
5193 - reserve space for all logs during transactions (2059)
5194 - don't evict page beyond end of stripe extent (2925)
5195 - don't oops on a deleted current working directory (2399)
5196 - handle hard links to targets without a parent properly (2517)
5197 - don't dereference NULL lock when racing during eviction (2867)
5198 - don't grow lock extents when lots of conflicting locks (2919)
5200 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5203 - account for cache space usage on clients to avoid data loss (974)
5204 - lfsck support in lustre kernel code (2349)
5205 - reduce journal credits needed for BRW writes (2370)
5206 - orphan handling to avoid losing space on client/server crashes
5207 - ptlrpcd can be blocked, stopping ALL progress (2477)
5208 - use lock value blocks to assist in proper KMS, faster stat (1021)
5209 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5210 - recovery for initial connections (2355)
5211 - fixes for mds_cleanup_orphans (1934)
5212 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5213 - block all file creations until orphan recovery completes (1901)
5214 - client remove rq_connection from request struct (2423)
5215 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5216 - recovery timer race (2670)
5217 - mdc_close recovey bug (2532)
5218 - ptlrpc cleanup bug (2710)
5219 - mds timeout on local locks (2588)
5220 - namespace lock held during RPCs (2431)
5221 - handle interrupted sync write properly (2503)
5222 - don't try to handle a message that hasn't been replied to (2699)
5223 - client assert failure during cleanup after abort recovery (2701)
5224 - leak mdc device after failed mount (2712)
5225 - ptlrpc_check_set allows timedout requests to complete (2714)
5226 - wait for inflight reqs when ptlrpcd finishes (2710)
5227 - make sure unregistered services are removed from the srv_list
5228 - reset bulk XID's when resending them (caught by 1138 test)
5229 - unregister_bulk after timeout
5230 - fix lconf error (2694)
5231 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5232 - readahead locks pages, leaves pending causing memory pressure (2673)
5233 - increase OST request buffers to 4096 on large machines (2729)
5234 - fix up permission of existing directories in simple_mkdir (2661)
5235 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5236 - don't assign transno to errored transactions (2742)
5237 - don't delete objects on OST if given a bogus objid from MDS (2751)
5238 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5239 - if rq_no_resend, then timeout request after recovery (2432)
5240 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5241 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5242 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5243 - MDS recovery shouldn't delete objects that it has given out (2730)
5244 - if enqueue arrives after completion, don't clobber LVB (2819)
5245 - don't unlock pages twice when trigger_group_io returns error (2814)
5246 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5247 - don't write pages to disk if there was an error (1450)
5248 - don't ping imports that have recovery disabled (2676)
5249 - take buffered bytes into account when balancing socknal conn (2817)
5250 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5251 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5252 - remove little-endian swabbing of llog records (1987)
5253 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5254 - retry reposting request buffers if they fail (1191)
5255 - grow extent at grant time to avoid granting a revoked lock (2809)
5256 - lock revoke doesn't evict page if covered by a second lock (2765)
5257 - disable VM readahead to avoid reading outside lock extents (2805)
5259 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5260 - updated kernel patches for hp-2.4.20 kernel (2681)
5262 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5265 - fix truncated write corruption (2366)
5266 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5268 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5269 - ensure len(uuid) < 37 in lmc (1171)
5270 - fix ia64 OOPS in llog_test (2255)
5271 - zero end of page at obdfilter for partial page writes (2648)
5272 - don't leave stale dentries around after renames (bug 2428)
5273 - fix timeouts when evicting a client with a single lock held (2642)
5274 - set deadline for the initial HELLO message to drain (2634)
5275 - print out dotted-quad IP addresses in the socknal (2302)
5277 - additional debugging for MDS client eviction problem (2443)
5278 - fix mkfsoptions support for osts (2603, 2604)
5280 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5283 - add series for the vanilla 2.6.0 kernel
5284 - add series for the vanilla 2.4.24 kernel
5285 - add series for a cray x86/64 UL kernel drop
5286 - fix xattr patches for the vanilla 2.4.19 series
5288 - generate true UUIDs in lmc (1171)
5289 - have portals stack dumping break in UML (2466)
5290 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5291 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5292 - wake callback waiting threads on client eviction (2460)
5293 - Add --ptldebug and --subsystem to lmc (1719)
5294 - update assertion to allow safe interrupt allocation
5295 - set rq_no_resend for cancel requests (2432)
5296 - recalculate ptlrpcd timeout after resend (2494)
5297 - call vfs_rmdir when removing pending directories (2368)
5298 - fix renaming a file to itself (2429)
5299 - lmc creates a default one-stripe lov (2454)
5300 - expand procfs space to handle large clusters (2326)
5301 - increase UML stack to avoid overflow
5302 - update lconf's list of debug and subsystem masks
5303 - fix lfs find --obd (2510)
5304 - /proc tunable for disabling filter read caching (2591)
5305 - stop rpm packages from altering slapd.conf (2301)
5306 - disable nagle in the socknal under 0conf (2578)
5307 - choose mds inode size based on stripe count (2572)
5308 - fix kernel-source rpm problems (2516)
5310 - add --disable-doc to avoid pdf generation (2421)
5311 - update documentation, tests, type-os, comments
5312 - avoid format warnings on ia64
5313 - remove the TOE NAL
5314 - tiny code cleanups by removing unused fields
5316 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5319 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5320 - remove the most blatant lies from BUILDING file (2371)
5321 - change default debug level to reasonable production setting
5322 - reduce client side cache size to reduce cache flush time
5323 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5324 - make TCP zerocopy and pinger support enabled by default (2476)
5325 - sync writes completed after process exits caused crashes (2319)
5326 - maintain correct mount count on the MDS (2356)
5327 - backout 1557, because 2316 wasn't really fixed
5328 - better file I/O statistics gathering in /proc
5329 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5330 - another kernel patch to fix zero-copy TCP function export
5331 - don't take duplicate lock when processing re-sent getattr (2420)
5332 - lctl uses obd_self_export instead of creating new conn (2353)
5333 - MDS/OST recovery case which requires object creation asserted (2425)
5334 - move lfs from /usr/sbin to /usr/bin in packages
5335 - fix race between mds_client_add and mds_client_free (2417)
5336 - use kmalloc instead of slabs in portals (2430)
5337 - don't create duplicate records when a failover MDS is present (2442)
5338 - remove unnecessary mount age check (2332)
5339 - don't remove directory inodes from locks prematurely (2451)
5340 - don't break if MDS service name is the same as hostname (2103)
5341 - fix races in client write RPC generation when cache full (2482)
5343 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5346 - remove now-unused request->rq_obd (278)
5347 - if an allocation fails, print out how much memory we've used (1933)
5348 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5349 - add the "configurable stack size" patch to most series files (1256)
5350 - ability to write large log records, for 100+ OST configs (2306)
5351 - fix NULL deref when filter_prep fails (2314)
5352 - fix operator precedence error in filter_sync
5353 - dynamic allocation of socknal TX descriptors (2315)
5354 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5355 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5356 - ensure that CWARN messages in Portals always get to the syslog
5357 - __init/__exit are not for prototype decls (ldlm_init/exit)
5358 - x86-64 compile warning fixes
5359 - fix gateway LMC keyword conflict (2318)
5360 - fix MDS lock inversions in getattr/reint paths (1844)
5361 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5362 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5363 - DLM race condition prevented some lock evictions (2328)
5364 - ENOMEM detection and retry on socknal sends (2230)
5365 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5366 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5367 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5368 - better error messages when a client is rejected during recovery (1505)
5369 - avoid cancelling locks which were never granted, after failure (2330)
5370 - fix i_sem/journal inversion in mds_client_add (2333)
5371 - fix truncate/getattr lock cycle deadlock (2334)
5372 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5373 - fix two rare exit paths which could leak an l_lock() ref (2321)
5374 - fencepost error in MDS/OST orphan recovery (2226)
5375 - make log record alignment 8 bytes (1988)
5376 - lstripe now fails when requested offset > ost_count (2237)
5377 - ensure that all kernel series have a complete list.h (1607)
5378 - fix crashes in special-file operations (2316)
5379 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5381 - allow configurable automake binary, for testing new versions
5382 - small update to the lfs documentation
5384 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5386 * fix negative export reference count in fsfilt_sync (2312)
5388 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5389 * release candidate 0.9.1
5391 - orphans are moved into the PENDING directory for possible recovery
5392 - replayed opens now open by fid for orphan/rename safety (1042)
5393 - last close of an orphan inode generates a transno (683)
5394 - chdir() and mount() now pin the directory entry (1020)
5395 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5396 - discard very old requests without processing them (1502)
5397 - remove some common, well-understood CERRORs (1505)
5398 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5399 - clear "grant" flags in OST replies until OST grant code lands (1644)
5400 - fix read performance by not clobbering i_blksize on client (1598)
5401 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5402 - make LRU size a /proc tunable, clears locks when reduced (707)
5403 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5404 - server locks take references on exports now (1558)
5405 - build fixes for 2.4.20-rh trees (1663)
5406 - return an error from lov_create if all OSCs are inactive (1751)
5407 - fix import levels when a reconnect happens without a timeout (1597)
5408 - exit early from mds_open if we get a lookup error (1749)
5409 - partial page read at EOF wouldn't wait for disk before sending (1642)
5410 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5411 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5412 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5413 - write/truncate lock inversion (1639)
5414 - Don't auto-load obdclass, portals modules during cleanup (1495)
5415 - fix timestamps from jumping to "now" (1763)
5416 - extra journal assertions (1648)
5417 - add an extra multiunlink test (1771)
5418 - fix read_record/write_record API (1776)
5419 - fix leak of offset_extent, possible incorrect i_size later (1772)
5420 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5421 - Debugging for the kqswnal_get_idle_tx problems (1820)
5422 - Allow recovery to be attempted multiple times (1536)
5423 - Write out MDS last_rcvd file after it is first created (1600)
5424 - Fix tx_descriptor leak in failed transmit situations (1827)
5425 - ext3 journaling fixes for assertion failure after IO error (1871)
5426 - class_export_put() on freed export after completion AST error (1896)
5427 - Fix revalidate looping in VFS (1322)
5428 - Don't access a freed export during MDS_REINT timeout (1521)
5429 - Add open-unlink recovery support on the MDS (1673,1764)
5430 - Return an error if no MDS data was read from last_rcvd (1946)
5431 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5432 - Don't setup a disk device that doesn't match exported UUID (317)
5433 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5434 - avoid committing NULL handle in force close
5435 - local.sh is now a one-stripe LOV configuration
5436 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5437 - fix ext3 htree duplicate directory entry corruption (1516)
5438 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5439 - update to new LOV EA format (2097)
5440 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5442 2003-06-15 Phil Schwan <phil@clusterfs.com>
5445 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5446 - per-import/export recovery handling (958, 931, 959)
5447 - multiple last-rcvd slots, for serving multiple FSes (949)
5448 - connections are again shared between multiple imp/exports (963, 964)
5449 - "umount -f" would hang if any requests needed to be sent (393, 978)
5450 - avoid pinning large req buffer by copying for queued messages (989)
5451 - add "uuid" to "lctl device" command to help upcalls (991)
5452 - "open" RPCs with transnos would confuse recovery counters (1037)
5453 - do proper endian conversion of all wire messages (288, 340, 891)
5454 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5455 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5456 - fix ptlrpc_connection leak in target_handle_connect (1174)
5457 - fix import refcounting bug in OST and MDS cleanup (1134)
5458 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5459 - fix very unlikely obd_types race condition (501)
5460 - remove osc_open hack for echo_client (1187)
5461 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5462 - a failure in read_inode2 leads to deadlock (1139)
5463 - cancel ack-locks as soon as transaction is committed (1072)
5464 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5465 - make sure to commitrw after any preprw to avoid deadlock (1162)
5466 - failing to execute a file in a lustre FS would lock inode (1203)
5467 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5468 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5469 - obd_uuid2tgt would walk off the end of the list (1255)
5470 - on IA64 the osc would give portals incorrect bulk size (1258)
5471 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5472 - fix lock inversion caused by new llite matching code (1282)
5473 - limit the number of dirty pages on a client to 10MB (1286)
5474 - timed out locks were not being corrected cancelled (1289)
5475 - fix O_DIRECT above 4GB on IA-32 (1292)
5476 * major user-visible changes
5477 - fail out/fail over policy now controlled by the upcall (993)
5479 - add OBD_PING to check server availability and failure (954)
5480 - lustre messages are now sent in sending host order (288, 340, 891)
5481 - add eadatalen to MDS getattr reply (340)
5482 - OST read replies may contain second buffer, with per-page status (593)
5484 2003-03-11 Phil Schwan <phil@clusterfs.com>
5487 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5488 - failed open()s could cause deadlock; fixed (867, 869)
5489 - stop cancelling OST locks when files are closed (481)
5490 - overlapping XID spaces caused network corruption (851, 853)
5491 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5492 - setattr_raw added, to do single-RPC, server-side setattrs
5493 - lmc/lconf syntax change for OST UUIDs
5494 - fix crashy race condition between ptlrpc_free_req and osc_close
5495 - don't use request in mdc_enqueue if we hit a timeout (889)
5496 - don't set the inode i_size for regular files from the MDS (896)
5497 - handle out of order completion AST (842)
5498 - don't LBUG if a lock request times out after receiving AST (913)
5499 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5500 - if a bad lock AST arrives, send an error instead of dropping entirely
5501 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5502 - fix leak in bulk IO when only partially completed (899, 900, 926)
5503 - fix O_DIRECT for ia64 (55)
5504 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5505 - C-z after timeout could hang a process forever; fixed (977)
5507 - client-side I/O cache (678, 924, 929, 941, 970)
5509 - READPAGE and SETATTRs which don't take server-side locks get
5512 2003-02-11 Phil Schwan <phil@clusterfs.com>
5515 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5516 - Fix fsfilt_extN_readpage() to read a full page of directory
5517 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5518 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5519 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5520 - Most intent operations were removed, in favour of a new RPC mode
5521 that does a single RPC to the server and bypasses most of the VFS
5522 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5523 - Aggressively cancel local locks on DLM servers
5524 - mds_reint_unlink sends EA to the client if it's the last nlink.
5525 client uses that EA to unlink OST objects.
5526 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5527 - recursive symlinks were fixed (439)
5528 - fixed NULL deref in DEBUG_REQ
5529 - filter_update_lastobjid no longer calls sync, which annoyed extN
5530 - fixed multi-client small-writes to a single file problem (445)
5531 - fixed mtime updates during file writes (607)
5532 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5533 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5534 - fixed Linux OST size reporting problem (444, 656)
5535 - OST now updates object mtime with writes or setattr (607, 619)
5536 - client verifies file size before zeroing page past EOF (445)
5537 - OST now writes last allocated objid to disk with allocation (108)
5538 - LOV on echo now works (409)
5540 - mds_reint_unlink sends a new buffer, with the EA included. this
5541 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5542 set if a regular file was being unlinked, and it was the last link
5543 - use PtlGet from the target for bulk writes (315)
5544 - OST now updates object mtime with writes or setattr (607, 619)
5545 - LDLM now has a grant-time callback to revalidate locked items, if
5547 - Many MDS operations were reorganized to combat race conditions
5549 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5550 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5552 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5555 - Fully reactivate OST imports after reconnection (512, others)
5556 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5557 - More graceful error handling for truncating on dead OST (515)
5558 - Don't error out unless we're actually accessing dead stripes (474)
5559 - Fix garbage sizes when stripes are missing (410)
5560 - LRU counters were broken, causing constant lock purge (433, 432)
5561 - garbage on read from stripes with failed OSTs (441)
5562 - mark OSCs as active before reconnecting during recovery (438)
5563 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5564 - lfind did not preserve OST order in output (443)
5565 - symlinks cause hung clients, incorrect data (439)
5566 - stop dereferencing request after dropping refcount (457)
5567 - don't LASSERT(spin_is_locked) on non-SMP (455)
5568 - fixes for many rename() bugs
5569 - fstat didn't correctly synchronize attributes (399)
5570 - server must handle lock cancellation during blocking AST prep (487)
5571 - bulk descriptors were free()d too soon (511)
5572 - fix paths in lconf, which would load incorrect modules (451, 507)
5573 - fix confusing lconf 'host not found' error message (386)
5574 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5575 - fix race condition in mdc_blocking_ast() for inode access (526)
5576 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5577 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5578 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5579 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5580 - fix lstripe in several ways (kernel side) (527)
5581 - fix request leak in ldlm_cli_enqueue (262)
5582 - incorrect OSC was marked inactive after OST failure
5583 - call mds_fs_cleanup before unmounting filesystem (524)
5584 - fix races between taking ns_lock and ldlm_lock_change_resource
5585 - fix races updating LOV export open file list
5586 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5587 - fix recovery NULL deref in ldlm_cli_cancel_unused
5588 - fix some DLM races by using new hash table for lock handles (419)
5589 - permit the client to specify desired inodes, at replay
5590 - duplicate requests when we queue them for replay reintegration
5591 - fix last_rcvd offset calculation
5592 - sync after each recovered transaction, so we always make progress
5593 - never, not always, ERESTART requests without transnos
5594 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5596 - skip replay if the MDS says that the client is already connected
5597 - don't check for a recovery-enabled export to match lctl's UUID
5598 - don't INC_USE_COUNT for phantom exports
5599 - don't crash when cleaning up phantom exports (567)
5600 - don't double-finish or set replay data for errored mdc_open requests
5601 - abort requests when they time out, so we don't get old replies
5602 - send/receive replies for AST messages again
5603 - if the client says that it doesn't have the lock, cancel it on the
5605 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5606 mark it as destroyed, it will all work out when decref is called
5607 - fix module use counts (22, 581)
5609 - ASTs now expect a reply (server cancels lock on error reply)
5611 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5614 - fix many simultaneous client startup (392)
5615 - fix dentry->d_it clobbering
5616 - credentials weren't being shipped for readdir/getattr operations
5617 - remove invalid assertions triggered during some concurrent MD
5619 - proper Lustre versions added (336, 389)
5620 - fix memory leak for create error case (398)
5621 - fix LOV locking bug that would get cli/srv out of sync
5622 - fix echo client over LOV (409)
5623 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5624 - fix double-O_EXCL intent crash (424)
5625 - avoid sending multiple lock CANCELs (352)
5627 - MDS can do multi-client recovery (modulo bugs in new code)
5629 - many updates, edits, cleanups
5631 2002-11-18 Phil Schwan <phil@clusterfs.com>
5634 - fix null d_it dereference (346)
5635 - fix full OST/dbench hang (333)
5636 - fix permission problem with file removal (286)
5637 - fix removal of OSCs from LOV when they fail
5638 - fix NULL deref during bulk timeout (214)
5639 - fix problems related to multiple filesystems on one MDS (241)
5640 - fixed serious subtle metadata locking bugs
5641 - free locks on clients when inodes are removed due to memory
5643 - fix inode pointer in lock data (285)
5644 - partial support for multiple MDS on a single host (241)
5645 - data locks weren't cancelled at clear_inode time (290, 311)
5646 - intent locks could lead to unbounded lock growth (205)
5647 - added a maximum lock count, an LRU list, and a flusher
5648 - fix multiple rename (365)
5649 - properly abstracted the echo client
5650 - OSC locked 1 byte too many; fixed
5651 - rewrote brw callback code:
5652 - fixed recovery bugs related to LOVs (306)
5653 - fixed too-many-pages-in-one-write crash (191)
5654 - fixed (again) crash in sync_io_timeout (214)
5655 - probably fixed callback-related race (385)
5657 - Add capability to MDS protocol
5658 - LDLM cancellations and callbacks on different portals
5660 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5663 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5664 - defer open object destruction to close time (601981, 138)
5665 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5666 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5667 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5668 - don't call obd_setattr in ll_file_release for destroyed objects
5669 * protocol change to lustre_msg: move |version| and add |flags|
5670 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5671 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5672 * added replay of create, unlink, link and rename operations during
5673 MDS failover; recovery should be much more robust now
5674 * remove failed OSCs from LOVs (only lov_create uses this so far)
5675 * the lustre-HOWTO was brought (more) up to date (582544)
5677 2002-10-23 Phil Schwan <phil@clusterfs.com>
5680 - in-use dentries weren't being reused properly (617851)
5681 - prevent multiple LDLM setup (599178)
5682 - fix LOV size calculations for truncate (617853)
5683 - fix client handling of MDS intent errors (POSIX)
5684 - fix permission bug in lovstripe.c test (624321)
5685 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5686 - truncate past end of file could corrupt data
5687 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5688 - a race in recovery could return ETIMEDOUT to apps (623947)
5689 - building outside the source directory was fixed
5690 * the lustre-HOWTO was brought (more) up to date (582544)
5691 * major progress was made on recovery functionality
5693 2002-10-10 Phil Schwan <phil@clusterfs.com>
5696 - recovery deadlock fix
5697 - rm -rf causes LBUG fix (617817)
5698 - file open by multiple tasks fix (618962)
5699 - directory permissions bugs (602707 and 620007)
5700 - journal_stop fixed with locking (611313)
5701 - O_APPEND failures resolved (618273, perhaps 614459)
5702 - lconf PATH fix (619770)
5703 - IA64 build fix (621450)
5704 - RPC buffer sizes scale with amount of memory
5706 2002-10-01 Phil Schwan <phil@clusterfs.com>
5709 - locks would be cancelled without throwing away data pages,
5710 resulting in inconsistent data (605627)
5711 - inode attributes were not always being refreshed (605627, 612449)
5712 - lconf now continues to cleanup after lctl reports an error
5713 - MDS now enforces user permissions (602707)
5714 - lprocfs cleanup fixed, but not yet enabled (614157)
5715 - fixed infinite server hang, should a client not respond to an AST
5716 - avoid going into recovery if user calls readlink() with a buffer
5717 that's too small (613941)
5718 - AST RPCs no longer require replies (614867) -- this may be changed
5719 - don't crash server if client sends an IOV that's too big (611336)
5720 - fixed lock conversion deadlock (611892)
5721 - fixed the following of symlinks (614622)
5722 * recovery: the server can remove locks from a client that dies, other
5723 clients can make progress
5724 * more extN patch fixes
5725 * compile-time configurable ptlrpc buffer allocations
5727 - collaborative read cache document
5728 - Lustre Lite Performance CDR document-in-progress
5730 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5733 - fix typo in patch-2.4.18
5735 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5738 - clear ptlrpc request each time in handle_incoming_request()
5739 - unlink of files now destroys the object on the OST
5741 2002-09-19 Peter Braam <braam@clusterfs.com>
5743 * add hard link support
5744 * change obdfile creation method
5745 * kernel patch changed
5747 2002-09-19 Peter Braam <braam@clusterfs.com>
5750 - stack overflow bug in extN fixed
5752 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5754 * documentation updates
5755 - add man pages for config tools
5756 - update tests/README to describe testing with new config tools
5757 - finish metadata API descriptions
5758 * bug fixes and cleanups
5759 - statfs workaround for 16TB limit
5760 - LOV stripe allocation improved, can stripe on subset of OSTs
5761 - LOV file size/IO offset was wrong for files > 4GB in size
5762 - object EA data was being dropped, caused files to be unreadable
5763 - memory overflow with non-LOV OST caused memory corruption
5764 - fixed regression tests to work with new config tools, obdfilter
5765 - fixed bug when directory size became larger than 1 block
5766 - fixed bug (for single client case) when PWD was deleted
5767 - invalidate local directory pages when doing intent-based ops
5768 - avoid LDLM oops when lock callback contained bad data
5770 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5772 * documentation updates
5773 * bug fixes and cleanups
5774 - configuration tools
5777 - 64-bit compile warnings
5778 - 64-bit internal statfs data
5780 * test_brw on persistent OST devices
5782 * lprocfs (disabled)
5784 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5786 * documentation updates
5787 * bug fixes and cleanups
5788 * configuration tools
5790 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5793 * many small fixes to 0_5_4
5794 * io/network handling
5795 * thinkos in MDS operations
5797 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5800 * crucial basic fixes to 0.5.3
5801 * IOR, Iozone work over Elan
5804 2002-08-07 Phil Schwan <phil@clusterfs.com>
5805 * version 0_5_3, our first alpha
5806 * we use the new Portals iovs
5807 * documentation updates
5808 * bug fixes and cleanups
5809 * small changes in the DLM wire protocol
5811 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5812 * version 0_5_1 with some initial stability,
5813 * locking on MD and file I/O.
5814 * documentation updates
5815 * several bug fixes since 0.5.0
5816 * small changes in wire protocol
5818 2002-07-18 Phil Schwan <phil@clusterfs.com>
5820 * delivered as Lustre Light Alpha
5821 * fixed a crash after handling invalid MDS requests
5822 * fixed directory pages for architectures with non-4k pages sizes
5824 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5825 * release version v0_4_4
5826 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5828 * Fixes a number of interruption problems with OST operations.
5829 * Update documentation for portals header changes
5830 * Move all wire protocol structs/defines to lustre_idl.h
5831 * Fixes symlink length bug.
5832 * Add tcpdump to repository.
5834 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5835 * release version v0_4_3
5836 * Fixes statfs for inodes on extN.
5837 * Fixes bug in runtests which would delete /etc/hosts.
5838 * Use 64-bit object IDs wherever possible (not into VFS though)
5839 Remove ost_get_info, which is unused by lustre, and out of date.
5841 2002-07-03 Peter Braam <braam@clusterfs.com>
5842 * release version v0_4_2 Fixes a lookup error (type not passed)
5843 * move forward to head of Portals
5844 * move forward to latest Lustre kernel
5846 2002-06-25 Peter Braam <braam@clusterfs.com>
5847 * release version v0_4_1. Hopefully stable on single node use.