1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.39.3 (SLES 10),
5 2.6.18-128.1.14.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.41.6.sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
14 * File join has been disabled in this release, refer to Bugzilla 16929.
16 Severity : enhancement
18 Description: Communicate OST degraded/readonly state via statfs to MDS
19 Details : Flags in the statfs returned from OSTs indicate whether the
20 OST is in a degraded RAID state, or if the filesystem has
21 turned read-only after a filesystem error is detected.
26 Description: don't panic if EPROTO was hit when reading symlink
27 Details : correctly handling request reference in error cases.
29 Severity : enhancement
31 Description: Add LustreNetLink, a kernel-userspace communcation path.
33 Severity : enhancement
35 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
40 Description: access to llog context before init.
41 Details : move handling CATALOGS file at osc layer and forbid access to llog
45 Frequency : rare, only when client is writing to OST on local node
47 Description: Avoid deadlock for local client writes
48 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
49 memory freeing context. This allows OST threads to set the
50 PF_MEMALLOC flag on task structures in order to allocate memory
51 from reserved pools and complete IO.
52 Use GFP_HIGHUSER for OST allocations for non-local client writes,
53 so that the OST threads generate memory pressure and allow
54 inactive pages to be reclaimed.
56 Severity : enhancement
58 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
63 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
64 Details : move ldlm namespace creation in setup phase to avoid grab
65 _lprocfs_lock with cli_sem held.
69 Description: Temporarily disable grant shrink.
70 Details : Disable the feature for debugging.
73 Frequency : only when formatting test filesystems
75 Description: Unable to run several mkfs.lustre on loop devices at the same time
76 Details : mkfs.lustre returns error 256 on the concurrent loop devices
77 formatting. The solution is to proper handle the error.
79 Severity : enhancement
81 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
83 Severity : enhancement
85 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
90 Description: resolve race between obd_disconnect and class_disconnect_exports
91 Details : if obd_disconnect will be called to already disconnected export he
92 forget release one reference and osc module can't unloaded.
94 Severity : enhancement
96 Description: Allow tuning service thread via /proc
97 Details : For each service a new
98 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
99 created that can be used to set min/max thread counts, and get the
100 current number of running threads.
105 Descriptoin: don't return error if have partially created objects for file.
106 Details : lov_update_create_set uses set->set_success as index for created
107 objects, so if some requests will be failed, they will have hole
108 at end of array and qos_shrink_lsm for allocate correct lsm.
110 Severity : enhancement
112 Description: Update OFED support to 1.4.1
117 Description: Reduce small size read RPC
118 Details : Set read-ahead limit for every file and only do read-ahead when
119 available read-ahead pages are bigger than 1M to avoid small size
122 Severity : enhancement
124 Description: move AT tunable parameters for more consistent usage
125 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
127 Severity : enhancement
129 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
130 Details : allow skip disconnected ost for send statfs request and hide error
136 Description: fix racy locking of mballoc block bitmaps causing BUG
137 Details : The locking of the mballoc buddy bitmap and the in-memory
138 block bitmap was using two different spin locks in some
139 cases. This made it possible to incorrectly access the
140 mballoc bitmap while another process was modifying it,
141 causing a sanity assertion to fail. While no on-disk corruption
142 was reported, there was some risk of this happening.
145 Frequency : rare, on llog test 6
147 Descriptoin: don't allow connect to already connected import
148 Details : allowing connect to already connected import is hide connecting problem.
153 Descriptoin: allow kill process which wait statahead result
154 Details : for some reasons 'ls' can stick in waiting result from statahead,
155 in this case need way for kill this process.
157 Severity : enhancement
159 Description: Add state history info file, enhance import info file
160 Details : Track import connection state changes in a new osc/mdc proc file;
161 add overview-type data to the osc/mdc import proc file.
163 Severity : enhancement
165 Description: MDS create should not wait for statfs RPC while holding DLM lock.
167 Severity : enhancement
169 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
174 Descriptoin: OST grant too much space to client even there are not enough space.
175 Details : Client will shrink its grant cache to OST if there are no write
176 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
177 this grant cache if there are already not enough avaible space
178 (left_space < total_clients * 32M).
181 Frequency : start MDS on uncleanly shutdowned MDS device
183 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
184 Details : stay in waiting mds<>ost recovery finished produce random bugs
185 due race between two ll_sync thread for one lov target. send
186 ACTIVATE event only if connect realy finished and import have
190 Frequency : rare, connect and disconnect target at same time
192 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
193 Details : don't call obd_disconnect under lov_lock. this long time
194 operation and can block ptlrpcd which answer to connect request.
199 Descriptoin: don't lose wakeup for imp_recovery_waitq
200 Details : recover_import_no_retry or invalidate_import and import_close can
201 both sleep on imp_recovery_waitq, but we was send only one wakeup
205 Frequency : always with long access acl
207 Descriptoin: mds can't pack reply with long acl.
208 Details : mds don't control size of acl but they limited by reint/getattr
211 Severity : enhancement
213 Description: Update to SLES10 kernel-2.6.16.60-0.33.
215 Severity : enhancement
217 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
220 Frequency : start MDS on uncleanly shutdowned MDS device
222 Descriptoin: aborting recovery hang on MDS
223 Details : don't throttle destroy RPCs for the MDT.
226 Frequency : on remount
228 Description: external journal device not working after the remount
229 Details : clear dev_rdonly flag for external journal devices in
235 Description: shutdown vs evict race
236 Details : client_disconnect_export vs connect request race.
237 if client will evicted at this time - we start invalidate
238 thread without referece to import and import can be freed
242 Frequency : rare, need acl's on inode.
244 Description: client can't handle ost additional correctly
245 Details : if ost was added after client connected to mds client can have
246 hit lnet_try_match_md ... to big messages to wide striped files.
247 in this case need teach client to handle config events about add
248 lov target and update client max ea size at that event.
250 Severity : enhancement
252 Description: Changelogs
253 Details : Changelogs are a lightweight mechanism to track filesystem
254 metadata and namespace changes. The changelog is recorded
255 permanently on the MDTs, and is periodically "consumed" / purged
256 when records are no longer needed.
258 Severity : enhancement
260 Description: compact fld format with extents
261 Details : Store range of seq rather than every seq in FLD. Seq
262 controller update FLD rather than clients. In Case of CMD, mdt0
263 has FLD, all other metadata server act as non persistent proxy
264 for FLD queries and cache fld entries in fld cache.
269 Description: don't skip ost target if they assigned to file
270 Details : Drop slow OSCs if we can, but not for requested start idx.
271 This means "if OSC is slow and it is not the requested
272 start OST, then it can be skipped, otherwise skip it only
273 if it is inactive/recovering/out-of-space.
277 Description: more cleanup in mds_lov
278 Details : not send LOV EA under replay, we can't know about they size at this
279 time. Don't allow client connect to mds before any ost connected,
280 for avoid problems with LOV EA size and returning EIO to client.
282 Severity : enhancement
284 Description: Interoperability at server side (Disk interoperability)
286 Severity : enhancement
288 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
290 Severity : enhancement
292 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
294 Severity : enhancement
296 Description: New client IO stack (CLIO).
298 Severity : enhancement
300 Description: Commit on sharing. Eliminate inter-client dependencies between
301 uncommitted transactions by doing transaction commits.
302 Thereby clients may recovery independently.
305 Frequency : Create a symlink file with a very long name
307 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
308 Details : If there is no extra space in the request for early cancels,
309 ldlm_req_handles_avail() returns 0 instead of a negative value.
311 Severity : enhancement
313 Description: Add /proc entry for import status
314 Details : The mdc, osc, and mgc import directories now have
315 an import directory that contains useful import data for debugging
318 Severity : enhancement
320 Description: Re-disable certain /proc logging
321 Details : Enable and disable client's offset_stats, extents_stats and
322 extents_stats_per_process stats logging on the fly.
325 Frequency : Only on FC kernels 2.6.22+
327 Description: oops in statahead
328 Details : Do not drop reference count for the dentry from VFS when lookup,
329 VFS will do that by itself.
331 Severity : enhancement
333 Description: Generic /proc file permissions
334 Details : Set /Proc file permissions in a more generic way to enable non-
335 root users operate on some /proc files.
339 Description: Hitting mdc_commit_close() ASSERTION
340 Details : Properly handle request reference release in
341 ll_release_openhandle().
345 Description: quota recovery deadlock during mds failover
346 Details : This patch includes att18982, att18236, att18237 in bz14840.
348 1. fix osts hang when mds does failover with quotaon
349 2. prevent watchdog storm when osts threads wait for the
354 Frequency : only patchless client
355 Description: add workaround for race between add/remove dentry from hash
357 Severity : enhancement
359 Description: Allow OST glimpses to return PW locks
363 Description: LBUG when llog conf file is full
364 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
368 Description: Prevent import from entering FULL state when server in recovery
372 Description: service mount cannot take device name with ":"
373 Details : Only when device name contains ":/" will mount treat it as
379 Description: replace ptlrpcd with the statahead thread to interpret the async
380 statahead RPC callback
384 Frequency : on recovery
385 Description: I/O failures after umount during fail back
386 Details : if client reconnected to restarted server we need join to recovery
387 instead of find server handler is changed and process self
388 eviction with cancel all locks.
390 Severity : enhancement
392 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
394 Severity : enhancement
396 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
398 Severity : enhancement
400 Description: Upcall on Lustre log has been dumped
401 Details : Allow for a user mode script to be called once a Lustre log has
402 been dumped. It passes the filename of the dumped log to the
403 script, the location of the script can be specified via
404 /proc/sys/lnet/debug_log_upcall.
409 Description: avoid idr_remove called for id which is not allocated.
410 Details : Move assigment s_dev for clustered nfs to end of initialization,
411 to avoid problem with error handling.
416 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
417 Details : When connection is reused this not moved from CONN_UNUSED_HASH
418 into CONN_USED_HASH and this prodice warning when put connection
419 again in unused hash.
422 Severity : enhancement
424 Description: File striping can now be set to use an arbitrary pool of OSTs.
426 Severity : enhancement
428 Description: Export bytes_read/bytes_write count on OSC/OST.
432 Description: Early reply size mismatch, MGC loses connection
433 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
434 the connect flags are properly negotiated.
439 Description: quotas are not honored with O_DIRECT
440 Details : all writes with the flag O_DIRECT will use grants which leads to
441 this problem. Now using OBD_BRW_SYNC to guard this.
445 Description: add quota statistics
446 Details : 1. sort out quota proc entries and proc code.
447 2. add quota statistics
449 Severity : enhancement
451 Description: enable quota support for HEAD.
455 Description: Properly propagate oinfo flags from lov to osc for statfs
456 Details : restore missing copy oi_flags to lov requests.
458 Severity : enhancement
460 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
461 : l_getgroups(8), lst(8), routerstat(8)
462 Details : included man pages for llobdstat(8), llstat(8),
463 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
465 Severity : enhancement
467 Description: configure's --enable-quota should check the
468 : kernel .config for CONFIG_QUOTA
469 Details : configure is terminated if --enable-quota is passed but
470 : no quota support is in kernel
474 Description: Remove portals compatibility
475 Details : Remove portals compatibility, not interoperable with releases
480 Description: Resolve device initialization race
481 Details : Prevent proc handler from accessing devices added to the
482 obd_devs array but yet be intialized.
484 Severity : enhancement
486 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
488 Severity : enhancement
490 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
495 Description: Using wrong pointer in osc_brw_prep_request
496 Details : Access to array[-1] can produce panic if kernel compiled with
497 CONFIG_PAGE_ALLOC enabled
501 Description: Client runs out of low memory
502 Details : Consider only lowmem when counting initial number of llap pages
506 Description: *optional* service tags registration
507 Details : if the "service tags" package is installed on a Lustre node
508 When the filesystem is mounted, a local-node service tag will
509 be created. See http://inventory.sun.com/ for more information
510 about the Service Tags asset management system.
514 Description: Kernel BUG tries to release flock
515 Details : Lustre does not destroy flock lock before last reference goes
516 away. So always drop flock locks when client is evicted and
517 perform unlock regardless of successfulness of speaking to MDS.
521 Description: add refcount protection for osc callbacks, avoid panic on shutdown
525 Description: sanity test 65a fails if stripecount of -1 is set
526 Details : handle -1 striping on filesystem in ll_dirstripe_verify
531 Description: ASSERTION(CheckWriteback(page,cmd)) failed
532 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
535 Severity : enhancement
537 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
541 Description: do not process already freed flock
542 Details : flock can possibly be freed by another thread before it reaches
543 to ldlm_flock_completion_ast.
547 Description: LBUG during stress test
548 Details : Need properly lock accesses the flock deadlock detection list.
552 Description: oops in page fault handler
553 Details : kernel page fault handler can return two special 'pages' in error
554 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
558 Description: timeout with invalidate import.
559 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
560 should be handled by ptlrpcd. This produce long age waiting and
561 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
563 Severity : enhancement
565 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
569 Description: enable MGS and MDT services start separately
570 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
571 a co-located MGS without starting the MGS, which is a complement
572 to 'nosvc' mount option.
576 Frequency : after recovery
577 Description: precreate to many object's after del orphan.
578 Details : del orphan st in oscc last_id == next_id and this triger growing
579 count of precreated objects. Set flag LOW to skip increase count
580 of precreated objects.
584 Frequency : rare, on clear nid stats
585 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
586 Details : when clean nid stats sometimes try destroy live entry,
587 and this produce panic in free.
591 Description: Stack overflow during MDS log replay
592 ease stack pressure by using a thread dealing llog_process.
596 Description: wait until IO finished before start new when do lock cancel.
597 Details : VM protocol want old IO finished before start new, in this case
598 need wait until PG_writeback is cleared until check dirty flag and
599 call writepages in lock cancel callback.
601 Severity : enhancement
603 Description: using special macro for print time and cleanup in includes.
607 Description: mds_mfd_close() ASSERTION(rc == 0)
608 Details : In mds_mfd_close(), we need protect inode's writecount change
609 within its orphan write semaphore to prevent possible races.
613 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
617 Frequency : rare, on shutdown ost
618 Description: don't hit live lock with umount ost.
619 Details : shrink_dcache_parent can be in long loop with destroy dentries,
620 use shrink_dcache_sb instead.
624 Description: don't panic with use echo client
625 Details : echo client pass NULL as client nid pointer and this produce null
630 Description: fix build on ppc32
631 Details : compile code with -m64 flag produce wrong object file for ppc32.
635 Description: add message levels for liblustreapi
639 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
640 Details : Move the 'good_osts' check before the 'total_bavail' check. This
641 will result in an -EAGAIN and in the exit call path we call
642 alloc_rr() which will with increasing aggressiveness attempt to
643 aquire precreated objects on the minimum number of required OSCs.
647 Description: Use old size assignment to avoid deadlock
648 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
649 the scheduling while holding a spinlock. We do not need locking
650 for size in ll_update_inode() because size is only updated from
651 the MDS for directories or files without objects, so there is no
652 other place to do the update, and concurrent access to such inodes
653 are protected by the inode lock.
657 Description: resolve "_IOWR redefined" build error on SLES10
661 Description: dump the memory debugging after all modules are unloaded to
662 suppress false negative in conf_sanity test 39
664 Severity : enhancement
666 Description: build kernel-ib packages for OFED 1.3 in our release cycle
671 Description: fix SLES kernel versioning
672 Details : the kernel version for our SLES 10 kernel did not include a "-"
673 before the "smp" at the end. while this was not a problem in
674 general, it did mean that software trying to use the kernel
675 version to try to detect a vendor specific kernel would fail.
676 this was most evident by the OFED build scripts.
680 Description: Don't update lov_desc members until making sure they are valid
681 Details : When updating lov_desc members via proc fs, need fix their
682 validities before doing the real update.
686 Description: don't put request into delay list while invalidate in flight.
687 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
688 import in flight. this produce timeout for invalidate and sometimes
689 can cause stale data.
693 Frequency : on ppc only
694 Description: not convert ost objects for directory because it's not exist.
695 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
699 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
701 Severity : enhancement
703 Description: Update to support 2.6.22.14 vanilla kernel.
707 Frequency : rare, on recovery
708 Description: read procfs can produce deadlock in some situation
709 Details : Holding lprocfs lock which send rpc can produce block for destroy
710 obd objects and this also block reconnect with -EALREADY. This
711 isn't fix all lprocfs bugs - but make it rare.
713 Severity : enhancement
715 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
718 Frequency : frequent on X2 node
720 Description: mdc_set_open_replay_data LBUG
721 Details : Set replay data for requests that are eligible for replay.
725 Description: lustre_mgs: operation 101 on unconnected MGS
726 Details : When MGC is disconnected from MGS long enough, MGS will evict the
727 MGC, and late on MGC cannot successfully connect to MGS and a lot
728 of the error messages complaining that MGS is not connected.
732 Frequency : on network error
733 Description: panic with double free request if network error
734 Details : mdc_finish_enqueue is finish request if any network error ocuring,
735 but it's true only for synchronus enqueue, for async enqueue
736 (via ptlrpcd) this incorrect and ptlrpcd want finish request
739 Severity : enhancement
741 Description: client-side metadata stat-ahead during readdir(directory readahead)
742 Details : perform client-side metadata stat-ahead when the client detects
743 readdir and sequential stat of dir entries therein
746 Frequency : on start mds
748 Description: Implement get_info(last_id) in obdfilter.
751 Frequency : occasional
753 Description: Correctly check stale fid, not start epoch if ost not support SOM
754 Details : open with flag O_CREATE need set old fid in op_fid3 because
755 op_fid2 was overwritten with new generated fid, but mds can answer
756 with one of these two fids and both is not stale. Setattr
757 incorrectly started an epoch and assume will be called
758 done_writing, but without SOM done_writing ever being called.
761 Frequency : rare, depends on device drivers and load
763 Description: MDS or OSS nodes crash due to stack overflow
764 Details : Code changes in 1.8.0 increased the stack usage of some functions.
765 In some cases, in conjunction with device drivers that use a lot
766 of stack the MDS (or possibly OSS) service threads could overflow
767 the stack. One change which was identified to consume additional
768 stack has been reworked to avoid the extra stack usage.
771 Frequency : occasional
773 Description: Do not fail import if osc_interpret_create gets -EAGAIN
774 Details : If osc_interpret_create got -EAGAIN it immediately exits and
775 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
776 oscc_has_objects and see OSC has no objests and call
777 oscc_internal_create to resend create request.
779 Severity : enhancement
781 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
783 Severity : enhancement
785 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
788 Frequency : very rare
790 Description: Fix replay if there is an un-replied request and open
791 Details : In some cases, older replay request will revert the
792 mcd->mcd_last_xid on MDS which is used to record the client's
795 Severity : enhancement
797 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
799 Severity : enhancement
801 Description: Add rhel5 support to HEAD.
803 Serverity : enhancement
805 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
810 Description: Don't allow skipping OSTs if index has been specified.
811 Details : Don't allow skipping OSTs if index has been specified, make
812 locking in internal create lots better.
816 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
817 Details : access to bitfield in structure is always rounded to long
818 and this produce problem with not atomic change any bit.
822 Description: Lustre make rpms failed.
823 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
824 builds Lustre rpms from tarball.
827 Frequency : rare on shutdown ost
829 Description: If llog cancel was not send before clean_exports phase, this can
830 produce deadlock in llog code.
831 Details : If llog thread has last reference to obd and call class_import_put
832 this produce deadlock because llog_cleanup_commit_master wait when
833 last llog_commit_thread exited, but this never success because was
834 called from llog_commit_thread.
838 Description: allow userland application know is lost one of stripes.
839 Details : fill lvb_blocks with error code on ost and return it to
840 application if error flag found.
844 Description: NULL lov_tgts causing MDS oops
845 Details : more safe checks for NULL lov_tgts for avoid oops.
847 Severity : enhancement
849 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
853 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
854 Details : If an OST is down the MDS will hang indefinitely in
855 obd_statfs_rqset() waiting for the statfs data. While for
856 MDS QOS usage of statfs, it should not stuck in waiting.
858 Severity : enhancement
860 Description: remote_acl support
861 Details : Support ACL-based permission check for remote user.
862 Support setfacl/getfacl for remote user with the utils
863 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
864 the system "{s,g}etfacl" utils.
866 Severity : enhancement
868 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
870 Severity : enhancement
872 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
876 Description: make mgs_setparam() handle fsname containing dash
877 Details : fsname containing a dash does not work with lctl conf_param
879 Severity : enhancement
881 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
883 Severity : enhancement
885 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
889 Frequency : rare, at shutdown
890 Description: access already free / zero obd_namespace.
891 Details : if client_disconnect_export was called without force flag set,
892 and exist connect request in flight, this can produce access to
893 NULL pointer (or already free pointer) when connect_interpret
894 store ocd flags in obd_namespace.
898 Frequency : only at startup
899 Description: not alloc memory with spinlock held.
900 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
901 if any spinlock held.
903 Severity : enhancement
905 Description: make lustre randomly fail allocating memory
906 Details : Make lustre randomly failed allocating memory for testing purpose.
908 Severity : enhancement
910 Description: lost problems with lov objid file
911 Details : Fixes some scability and access to not inited memory problems
912 in work with lov objdid file.
917 Description: lfs find does not continue on file error
918 Details : Continue other files processing when a file/dir is absent.
922 Description: Inconsistent usage of lustre_pack_reply()
923 Details : Standardize the usage of lustre_pack_reply() such that it
924 always generate a CERROR on failure.
927 Frequency : occasional
929 Description: MDS hang or stay in waiting lock
930 Details : If client receive lock with CBPENDING flag ldlm need send lock
931 cancel as separate rpc, to avoid situation when cancel request
932 can't processed due all i/o threads stay in wait lock.
936 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
940 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
944 Description: Update to SLES9 kernel-2.6.5-7.287.3.
948 Description: Update to RHEL4 latest kernel.
950 Severity : enhancement
952 Description: Build SLES10 patchless client fails
953 Details : The configure was broken by run ./configure with
954 --with-linux-obj=.... argument for patchless client. When the
955 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
956 can't search header adequately. Use absolute path such as
957 -I($LINUX)/include instead.
961 Description: interrupt oig_wait produce painc on resend.
962 Details : brw_redo_request can be used for resend requests from ptlrpcd and
963 private set, and this produce situation when rq_ptlrpcd_data not
964 copyed to new allocated request and triggered LBUG on assert
965 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
966 ptlrpcd set if request is changed and can be safety changed to use
971 Description: LASSERT_{REQ,REP}SWAB macros are buggy
972 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
973 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
974 these macros and replace them with their difinitions instead.
978 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
979 Details : Update which_patch & target file for SLES10 latest kernel.
982 Description: Remove client patches from SLES 10 kernel.
983 Details : This causes SLES 10 clients to behave as patchless clients
984 even on a Lustre-patched (server) kernel.
986 Severity : enhancement
988 Description: self-adjustable client's lru lists
989 Details : use adaptive algorithm for managing client cached locks lru
990 lists according to current server load, other client's work
991 pattern, memory activities, etc. Both, server and client
992 side namespaces provide number of proc tunables for controlling
995 Severity : enhancement
997 Description: light-weight GSS support
998 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
999 and reduce performance overhead.
1001 Severity : enhancement
1003 Description: Linux keyring support
1004 Details : Support using service of Linux keyring for Lustre GSS internal
1005 context refresh/cache mechanism.
1009 Description: Fix errors in lfs documentation
1010 Details : Fixes man pages
1014 Description: don't use GFP_* in generic Lustre code.
1015 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1019 Description: obdclass is limited by single OBD_ALLOC(idarray)
1020 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1024 Description: warnings with build patchless client with vanila 2.6.19 and up
1025 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1026 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1030 Frequency : Only for SLES
1031 Description: sanity_quota fail test_1
1032 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1033 file, which makes TSTID[2] inunique.
1037 Frequency : Only for SLES9
1038 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1039 Details : Update target/ChangeLog/which_patch .
1043 Frequency : Only for test_10 in sanity-quota.sh
1044 Description: a bug in quota test code
1045 Details : checking a test flag in a wrong place
1049 Frequency : Only for test_13 in sanity-quota.sh
1050 Description: a bug in quota test script
1051 Details : shouldn't check overfull allocated quota
1055 Description: jbd statistics
1056 Details : Port older jbd statistics patch for sles10
1058 Severity : enhancement
1060 Description: Add printing inode info into message about error in writepage.
1065 Description: reply_lock_interpret crash due to race with it and lock cancel.
1066 Details : Do not replay locks that are being cancelled. Do not reference
1067 locks by their address during replay, just by their handle.
1071 Frequency : When flocks are used.
1072 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1073 Details : Flock locks might destroy just granted lock if it could be merged
1074 with another existing flock, this is done in completion handler,
1075 so teach ldlm_cli_enquque_fini that this is a valid case for
1081 Description: Oops in read and write path when failing to allocate lock.
1082 Details : Check if lock allocation failed and return error back.
1086 Description: lstripe command fails for valid OST index
1087 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1088 instead of lov->desc.ld_active_tgt_count.
1092 Description: sanity.sh failed test 103
1093 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1098 Description: df doesn't work properly if diskfs blocksize != 4K
1099 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1103 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1104 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1106 Severity : enhancement
1108 Description: Adaptive timeouts
1109 Details : RPC timeouts adapt to changing server load and network
1110 conditions to reduce resend attempts and improve recovery time.
1114 Description: llapi_file_create() does not allow some changes
1115 Details : add llapi_file_open() that allows specifying the mode and
1116 open flags, and also returns an open file handle.
1120 Description: merge and cleanup kernel patches.
1121 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1125 Description: Add journal checksum support.(Kernel part)
1126 Details : The journal checksum feature adds two new flags i.e
1127 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1128 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1129 indicates that the commit block contains the checksum for
1130 the blocks described by the descriptor blocks. Now commit
1131 record can be sent to disk without waiting for descriptor
1132 blocks to be written to disk. This behavior is controlled
1133 using JBD2_FEATURE_ASYNC_COMMIT flag.
1137 Description: OSS needs mutliple precreate threads
1138 Details : Add ability to start more than one create thread per OSS.
1142 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1143 Details : execute lfs setstripe on client
1147 Description: mds_obd_create error creating tmp object
1148 Details : When the user sets quota on root, llog will be affected and can't
1149 create files and write files.
1152 Frequency : Always on ia64 patchless client, and possibly others.
1154 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1155 Details : This allows the patchless client to be loaded on architectures
1156 without this export.
1160 Description: RedHat Update kernel for RHEL5
1161 Details : Add the kernel config file for RHEL5.
1165 Description: RedHat Update kernel for RHEL5
1166 Details : Modify the kernel config file more closer RHEL5.
1170 Description: Build failure against Centos5 (RHEL5)
1171 Details : Define PAGE_SIZE when it isn't present.
1175 Description: test_7 on sanity-quota.sh can't be used on separate machines
1176 Details : add facet to handle it
1180 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1181 Details : with NFS, the anon dentry's parent was set to itself in
1182 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1183 corresponding dentry other than use rec->ur_name.
1185 Severity : enhancement
1187 Description: lfs setstripe enhancement
1188 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1192 Description: enable data checksumming by default
1193 Details : enable checksum by default, allow --disable-checksum
1194 configure option and "-o nochecksum" mount option. Checksums
1195 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1199 Description: lustre support for RHEL5
1200 Details : Add support for RHEL5.
1204 Description: Client eviction due to failover config
1205 Details : after a connection loss, the lustre client should attempt to
1206 reconnect to the last active server first before trying the
1207 other potential connections.
1211 Description: when mds and osts use different quota unit(32bit and 64bit),
1212 quota will be released repeatly.
1213 Details : void sending multiple quota reqs to mds, which will keep the
1214 status between the reqs.
1218 Description: osts not allocated evenly to files
1219 Details : change the condition to increase offset_idx
1222 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1224 Description: Data corruption for OSTs that are formatted larger than 2TB
1226 Details : When generating the bio request for lustre file writes the
1227 sector number would overflow a temporary variable before being
1228 used for the IO. The data reads correctly from Lustre (which
1229 will overflow in a similar manner) but other file data or
1230 filesystem metadata may be corrupted in some cases.
1234 Description: Tune the kernel for good SCSI performance.
1235 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1236 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1241 Description: rewrite ext2-derived code in obdclass/uuid.c
1242 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1243 from scratch preserving functionality.
1247 Description: "lfs find -obd UUID" prints directories
1248 Details : "lfs find -obd UUID" will return all directory names instead
1249 of just file names. It is incorrect because the directories
1250 do not reside on the OSTs.
1254 Description: use i_size_read and i_size_write in 2.6 port
1255 Details : replace inode->i_size access with i_size_read/write()
1258 Frequency : when removing large files
1260 Description: scheduling issue during removal of large Lustre files
1261 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1262 It causes scheduling issues when removing large files (17TB in the
1266 Frequency : only with liblustre clients on XT3
1268 Description: evictions taking too long
1269 Details : allow llrd to evict clients directly on OSTs
1272 Frequency : only on ppc
1274 Description: /proc/fs/lustre/devices broken on ppc
1275 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1276 We should make sure the type of 'index' is loff_t and then cast
1277 explicitly as needed below. Do not assign an explicitly cast
1281 Frequency : only for rhel5
1283 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1284 Details : Modify the target file & which_kernel.
1286 Severity : enhancement
1288 Description: omit set fsid for export NFS
1289 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1290 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1294 Frequency : Always, for kernels after 2.6.16
1295 Description: Fix warning idr_remove called for id=.. which is not allocated.
1296 Details : Last kernels save old s_dev before kill super and not allow
1297 to restore from callback - restore it before call kill_anon_super.
1301 Description: buffer overruns could theoretically occur
1302 Details : llapi_semantic_traverse() modifies the "path" argument by
1303 appending values to the end of the origin string, and a
1304 overrun may occur. Adding buffer overrun check in liblustreapi.
1308 Description: Fix error on 'ls .' at the top of the Lustre mount.
1309 Details : Don't revalidate dentry if it is a root dentry.
1313 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1314 Details : Modify vm-tunables-rhel4.patch.
1318 Description: Kernel config for 2.6.18-vanilla.
1319 Details : Modify targets/2.6-vanilla.target.in.
1320 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1321 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1322 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1323 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1327 Description: improve handling recoverable errors
1328 Details : if request processig with error which can be recoverable on server
1329 request should be resend, otherwise page released from cache and
1334 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1335 Details : Modify target file & which_patch.
1336 A flaw was found in the IA32 system call emulation provided
1337 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1338 value could be stored in the %RAX register, which could trigger an
1339 out-of-bounds system call table access. An untrusted local user
1340 could exploit this flaw to run code in the kernel
1341 (ie a root privilege escalation). (CVE-2007-4573).
1345 Description: O_DIRECT bypasses client statistics.
1346 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1347 read_ahead_stats were updated. copied the stats section from
1348 osc_send_oap_rpc() into async_internal().
1352 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1356 Description: change order of libsysio includes
1357 Details : '#include sysio.h' should always come before '#include xtio.h'
1361 Description: handle "serious error: objid * already exists" more gracefully
1362 Details : If LAST_ID value on disk is smaller than the objects existing in
1363 the O/0/d* directories, it indicates disk corruption and causes an
1364 LBUG(). If the object is 0-length, then we should use the existing
1365 object. This will help to avoid a full fsck in most cases.
1367 Severity : enhancement
1369 Description: adapt the lustre_config script to support the upgrade case
1370 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1371 server targets to 1.6 in parallel.
1375 Description: To avoid grant space > avaible space when the disk is almost
1376 full. Without this patch you might see the error "grant XXXX >
1377 available" or some LBUG about grant, when the disk is almost
1379 Details : In filter_check_grant, for non_grant cache write, we should
1380 check the left space by if (*left > ungranted + bytes), instead
1381 of (*left > ungranted), because only we are sure the left space
1382 is enough for another "bytes", then the ungrant space should be
1383 increase. In client, we should update cl_avail_grant only there
1384 is OBD_MD_FLGRANT in the reply.
1388 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1389 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1390 described in CVE-2007-4573. This problem could allow untrusted
1391 local users to gain root access.
1394 Frequency : when using O_DIRECT and quotas
1396 Description: Incorrect file ownership on O_DIRECT output files
1397 Details : block usage reported by 'lfs quota' does not take into account
1398 files that have been written with O_DIRECT.
1403 Description: touch file failed when fs is not full
1404 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1405 otherwise we can get ENOSP while fs is not full.
1409 Description: parallel lock callbacks
1410 Details : Instead of sending blocking and completion callbacks as separated
1411 requests, adding them to a set and sending in parallel.
1414 Frequency : only for Cray XT3
1415 Bugzilla : 12829/13455
1416 Description: Changing primary group doesn't change the group lustre assigns to
1418 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1419 overridden with the primary group provided by the group upcall,
1420 whereas the supplied fsgid can be trusted if it is in the list of
1421 supplementary groups returned by the group upcall.
1423 Severity : enhancement
1425 Description: Allow masking D_WARNING, D_ERROR messages from console
1426 Details : Console messages can now be disabled via lnet.printk.
1430 Description: User code with malformed file open parameter crashes client node
1431 Details : Before packing join_file req, all the related reference should be
1432 checked carefully in case some malformed flags cause fake
1433 join_file req on client.
1437 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1438 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1439 mode is equal to the granted mode. However, it should also check
1440 that LDLM_FL_CP_REQD is not set.
1444 Description: Heavy nfs access might result in deadlocks
1445 Details : After ELC code landed, it is now improper to enqueue any mds
1446 locks under och_sem, because enqueue might want to decide to
1447 cancel open locks for same inode we are holding och_sem for.
1451 Description: Client eviction while running blogbench
1452 Details : A lot of unlink operations with concurrent I/O can lead to a
1453 deadlock causing evictions. To address the problem, the number of
1454 oustanding OST_DESTROY requests is now throttled to
1455 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1456 callbacks are processed in priority.
1460 Description: enable ACLs on MDS by default
1461 Details : ACLs must be enabled on MDS by default.
1464 Frequency : PPC/PPC64 only
1466 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1467 Details : fix duplicated definitions between asm-ppc64/types.h and
1468 lustre_types.h on PPC.
1471 Frequency : PPC/PPC64 only
1473 Description: asm-ppc/segment.h does not exist
1474 Details : fix compile issue on PPC.
1478 Description: better handle error messages in extents code
1481 Frequency : RHEL4 only
1483 Description: mkfs is very slow on IA64/RHEL4
1484 Details : A performance regression has been discovered in the MPT Fusion
1485 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1486 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1487 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1489 Severity : enhancement
1491 Description: SNMP support enhancement
1492 Details : Adding total number of sampled request for an MDS node in snmp
1495 Severity : enhancement
1497 Description: Optimize ldlm waiting list processing for PR extent locks
1498 Details : When processing waiting list for read extent lock and meeting read
1499 lock that is same or wider to it that is not contended, skip
1500 processing rest of the list and immediatelly return current
1501 status of conflictness, since we are guaranteed there are no
1502 conflicting locks in the rest of the list.
1506 Description: Time out and refuse to reconnect
1507 Details : When the failover node is the primary node, it is possible
1508 to have two identical connections in imp_conn_list. We must
1509 compare not conn's pointers but NIDs, otherwise we can defeat
1510 connection throttling.
1514 Description: port llog fixes from b1_6 into HEAD
1515 Details : Port llog reference couting and some llog cleanups from b1_6
1516 (bug 10800) into HEAD, for protect from panic and access to already
1517 free llog structures.
1521 Description: Detect stride IO mode in read-ahead
1522 Details : When a client does stride read, read-ahead should detect that and
1523 read-ahead pages according to the detected stride pattern.
1527 Description: data checksumming impacts single node performance
1528 Details : add support for several checksum algorithm. Currently, only CRC32
1529 and Adler-32 are supported. The checksum type can be changed on
1530 the fly via /proc/fs/lustre/osc/*/checksum_type.
1534 Description: use adler32 for page checksums
1535 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1540 Description: build for x2 fails
1541 Details : fix compile issue on Cray systems.
1545 Description: Properly match for duplicate locks
1546 Details : Due to different lock order from skiplists code, we need to
1547 traverse entire list for now
1550 Frequency : only on PPC/SLES10
1552 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1553 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1554 be defined. Add a hack in mkfs_lustre.c to work around this compile
1559 Description: LASSERT on MDS when client holding flock lock dies
1560 Details : ldlm pool logic depends on number of granted locks equal to
1561 number of released locks which is not true for flock locks, so
1562 just exclude such locks from consideration.
1566 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1567 Details : Use fsfilt_sync() for both the whole filesystem sync and
1568 individual file sync to eliminate dangerous inode locking
1569 with I_LOCK that can lead to a deadlock.
1573 Description: performance in 1.6.3
1574 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1575 sure that 1MB requests are not fragmented by the block layer.
1577 Severity : enhancement
1579 Description: organize the server-side client stats on per-nid basis
1580 Details : Change the structure of stats under obdfilter and mds to
1588 The "uuid"s file would list the uuids of _active_ exports.
1589 And the clear entry is to clear all stats and stale nids.
1591 Severity : enhancement
1593 Description: eliminate client locks in face of contention
1594 Details : file contention detection and lockless i/o implementation
1595 for contended files.
1599 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1602 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1604 Description: MDS may lose file striping (and hence file data) in some cases
1605 Details : If there are additional extended attributes stored on the MDS,
1606 in particular ACLs, SELinux, or user attributes (if user_xattr
1607 is specified for the client mount options) then there is a risk
1608 of attribute loss. Additionally, the Lustre file striping
1609 needs to be larger than default (e.g. striped over all OSTs),
1610 and an additional attribute must be stored initially in the
1611 inode and then increase in size enough to be moved to the
1612 external attribute block (e.g. ACL growing in size) for file
1617 Description: skiplist implementation simplification
1618 Details : skiplists are used to group compatible locks on granted list
1619 that was implemented as tracking first and last lock of each
1620 lock group the patch changes that to using doubly linked lists
1624 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1625 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1626 because LASSERT on both the data supplied by a client, and the
1627 data on disk is dangerous and incorrect.
1629 Severity : enhancement
1631 Description: Slow truncate/writes to huge files at high offsets.
1632 Details : Directly associate cached pages to lock that protect those pages,
1633 this allows us to quickly find what pages to write and remove
1634 once lock callback is received.
1638 Description: more ldlm soft lockups
1639 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1640 starve other threads from the resource lock for a long time in
1641 case of long waiting queue, so change the debug level from
1642 D_OTHER to the less frequently used D_INFO.
1644 Severity : enhancement
1646 Description: add -gid, -group, -uid, -user options to lfs find
1650 Description: Hung threads in invalidate_inode_pages2_range
1651 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1652 once one is completed. As a result, some RPCs are stuck in the
1653 queue and are never sent.
1657 Description: filter threads hungs on waiting journal commit
1658 Details : Cleanup filter group llog code, then only filter group llog will
1659 be only created in the MDS/OST syncing process.
1663 Description: Procfs and llog threads access destoryed import sometimes.
1664 Details : Sync the import destoryed process with procfs and llog threads by
1665 the import refcount and semaphore.
1667 Severity : enhancement
1669 Description: openlock cache of b1_6 port to HEAD
1674 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1675 Details : If insertion of an extent fails, then discard the inode
1676 preallocation and free data blocks else it can lead to duplicate
1681 Description: don't always update ctime in ext3_xattr_set_handle()
1682 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1683 In some cases the ctime should not be updated, for example for
1684 2.0->1.8 compatibility it is necessary to delete an xattr and it
1685 should not update the ctime.
1689 Bugzilla : 15713/16362
1690 Description: Assertion in iopen_connect_dentry in 1.6.3
1691 Details : looking up an inode via iopen with the wrong generation number can
1692 populate the dcache with a disconneced dentry while the inode
1693 number is in the process of being reallocated. This causes an
1694 assertion failure in iopen since the inode's dentry list contains
1695 both a connected and disconnected dentry.
1699 Description: assertion failure in ldlm_handle2lock()
1700 Details : fix a race between class_handle_unhash() and class_handle2object()
1701 introduced in lustre 1.6.5 by bug 13622.
1706 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1707 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1712 Description: oops in sd_iostats_seq_show()
1713 Details : unloading/reloading the scsi low level driver triggers a kernel
1714 bug when trying to access the sd iostat file.
1719 Description: Kernel panics during QLogic driver reload
1720 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1721 patch, causing memory corruption.
1726 Description: journal_dev option does not work in b1_6
1727 Details : pass mount option during pre-mount.
1729 Severity : enhancement
1731 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1732 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1733 extent information of a file. It can be used to map logical blocks
1734 in a file to physical blocks in the block device.
1738 Description: LDLM soft lockups - improvement
1739 Details : It is be possible to send the lock handle along with each read
1740 or write request because the client is already doing a lock match
1741 itself so there isn't any reason the OST should have to re-do that
1745 Frequency : only on Cray X2
1747 Description: X2 build failures
1748 Details : fix build failures on Cray X2.
1752 Description: xid & resent requests
1753 Details : Initialize RPC XID from clock at startup (randomly if clock is
1756 Severity : enhancement
1758 Description: Add lustre_start utility to start or stop multiple Lustre servers
1763 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1764 Details : In case of memory pressure, list_del() can be called twice on
1765 req->rq_history_list, causing a kernel oops.
1769 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1770 Details : fix stack overflow in the distributed lock manager by defering
1771 export eviction after a failed AST to the elt thread instead of
1772 handling it in the dlm interpret routine.
1776 Description: Convert some comments to new format.
1777 Details : Update documenting comments to match doxygen conventions.
1781 Description: Grammar fixes.
1782 Details : A couple of trivial spelling fixes.
1786 Description: OSD_COUNTERS-mandatory
1787 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1788 to simplify the code.
1792 Description: simplify cmm_device freeing logic.
1793 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1797 Description: Add lockdep support to dt_object_operations locking interface.
1798 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1799 parameter indicating lock ordering. Update mdd code to use new
1804 Description: Introduce failloc constants for lockless IO tests.
1805 Details : Add two new failloc constants to test lockless IO. Only one of
1806 them in implemented---another is checked in yet to be landed
1811 Description: Add lockdep support for inode mutex.
1812 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1813 in the situations where Lustre has to lock more than one inode
1818 Description: Add optional invariants checking support.
1819 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1820 configure switch. This macro is to be used for consistency and
1821 sanity checks that are too expensive to be left in `production'
1826 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1827 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1832 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1833 Details : Use spin_lock_nested() in (the only) situation where more than
1834 one ldlm_lock is locked simultaneously. Also, fix possible
1835 dead-lock in ldlm_lock_change_resource() by enforcing particular
1840 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1841 Details : Instead of specifying each ldlm_lock call-back through separate
1842 parameter, wrap them into struct ldlm_callback_suite.
1846 Description: Kill join_lru obd method and its callers.
1847 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1848 in memory---a requirement implemented through ->o_join_lru() obd
1849 method in HEAD. Remove this method and its users.
1853 Description: Add asynchronous ldlm ENQUEUE completion handler.
1854 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1855 case that stock ldlm_completion_ast() cannot handle as it waits
1856 until lock is granted. Introduce new ldlm_completion_ast_async()
1857 for this. Also comment ldlm_completion_ast().
1861 Description: ldlm_error <-> errno conversion.
1862 Details : Add functions to map (rather arbitrary) between LDLM error codes
1863 and standard errno values. CLIO needs this to prevent LDLM specific
1864 constants from escaping ldlm and osc.
1868 Description: Kill unused ldlm_handle2lock_ns() function.
1869 Details : Kill unused ldlm_handle2lock_ns() function.
1873 Description: Add lu_ref support to ldlm_lock
1874 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1875 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1876 and ldlm_resource. LDLM interface has to be changed, because code that
1877 releases a reference on a lock, has to "know" what reference this is.
1878 In the most frequent case
1880 lock = ldlm_handle2lock(handle);
1882 LDLM_LOCK_PUT(lock);
1884 no changes are required. When any other reference (received _not_ from
1885 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1886 instead of LDLM_LOCK_PUT().
1888 Arguably, changes are pervasive, and interface requires some discipline
1889 for proper use. On the other hand, it was very instrumental in finding
1890 a few leaked lock references.
1894 Description: Add ldlm_lock_addref_try().
1895 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1896 attempts to addref a lock that might be being canceled
1901 Description: Add ldlm_weigh_callback().
1902 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1903 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1904 hard-coded `number of pages' logic.
1908 Description: Add lockdep annotations to llog code.
1909 Details : Use appropriately tagged _nested() locking calls in the places
1910 where llog takes more than one ->lgh_lock lock.
1914 Description: Add loi_kms_set().
1915 Details : Wrap kms updates into a helper function.
1919 Description: Constify instances of struct lsm_operations.
1920 Details : Constify instances of struct lsm_operations.
1924 Description: lu_conf support.
1925 Details : On a server, a file system object is uniquely identified
1926 by a fid, which is sufficient to locate and load all object
1927 state (inode). On a client, on the other hand, more data are
1928 necessary instantiate an object. Change lu_object_find() and
1929 friends to take additional `lu_conf' argument describing object.
1930 Typically this includes layout information.
1934 Description: lu_context fixes.
1935 Details : Introduce new lu_context functions that are needed on the client
1936 side, where some system threads (ptlrpcd) are shared by multiple
1937 modules, and so cannot be stopped during module shutdown.
1941 Description: Add start and stop methods to lu_device_type_operations.
1942 Details : Introduce two new methods in lu_device_type_operations, that are
1943 invoked when first instance of a given type is created and last one
1944 is destroyed respectively. This is need by CLIO.
1948 Description: Add lu_ref support to struct lu_device.
1949 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1950 track leaked references.
1954 Description: Introduce lu_kmem_descr.
1955 Details : lu_kmem_descr and its companion interface allow to create
1956 and destroy a number of kmem caches at once.
1960 Description: Fix lu_object finalization race.
1961 Details : Fix a race between lu_object_find() finding an object and its
1962 concurrent finalization. This race is (most likely) not possible
1963 on the server, but might happen on the client.
1967 Description: Introduce lu_ref interface.
1968 Details : lu_ref is a debugging module allowing to track references to
1969 a given object. It is quite cpu expensive, and has to be
1970 explicitly enabled with --enable-lu_ref. See usage description
1975 Description: Factor lu_site procfs stats into a separate function.
1976 Details : Separate lu_site stats printing code into a separate function
1977 that can be reused on a client.
1981 Description: Constify instances of struct {lu,dt,md}_device_operations.
1982 Details : Constify instances of struct {lu,dt,md}_device_operations.
1986 Description: Introduce struct md_site and move meta-data specific parts of
1987 struct lu_site here.
1988 Details : Move md-specific fields out of struct lu_site into special struct
1989 md_site, so that lu_site can be used on a client.
1993 Description: Kill mdd_lov_destroy().
1994 Details : Remove unused mdd code.
1998 Description: Add st_block checking to multistat.c.
1999 Details : Add st_block checking to multistat.c.
2003 Description: Add lu_ref support to struct obd_device.
2004 Details : Add lu_ref tracking to obd_device.
2008 Description: Kill obd_set_fail_loc().
2009 Details : Remove unused code.
2013 Description: Add special type for ptlrpc_request interpret functions.
2014 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2015 there. Actual usage will be in CLIO.
2019 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2020 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2021 doesn't work with lockdep.
2025 Description: Add rwv.c test program.
2026 Details : New testing program exercising readv(2) and writev(2) (Qian).
2030 Description: Add sendfile.c test program.
2031 Details : New testing program exercising sendfile(2) (Jay).
2035 Description: Ratelimit a message that can be very frequent.
2036 Details : Ratelimit a memory allocation failure message that can
2041 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2042 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2046 Description: Liblustre build fixes.
2047 Details : Liblustre build fixes.
2051 Description: libcfs: add cfs_{need,cond}_resched() interface.
2052 Details : libcfs: add cfs_{need,cond}_resched() definition and
2053 implementations for Linux, NT, and liblustre.
2055 Severity : enhancement
2057 Description: More exported tunables for mballoc
2058 Details : Add support for tunable preallocation window and new tunables for
2059 large/small requests
2063 Description: Detect corruption of block bitmap and checking for preallocations
2064 Details : Checks validity of on-disk block bitmap. Also it does better
2065 checking of number of applied preallocations. When corruption is
2066 found, it turns filesystem readonly to prevent further corruptions.
2070 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2071 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2072 window to 0. If it does hit cache miss, and read-pattern is still
2073 stride-io mode, does not reset the stride window, but also does not
2074 increase the stride window length in this case.
2078 Frequency : only for big-endian servers
2079 Description: Check if system is big-endian while mounting fs with extents feature
2080 Details : Mounting a filesystem with extents feature will fail on big-endian
2081 systems since ext3-based ldiskfs is not supported on big-endian
2082 systems. This can be over-riden with "bigendian_extents" mount option.
2084 Severity : enhancement
2086 Description: The root squash functionality
2087 Details : A security feature, which is to prevent users from being able
2088 to mount lustre on their desktop, run as root, and delete
2089 all of the files in the filesystem. The goal is accomplished by
2090 remapping user id (UID) and group id (GID) of the root user to
2091 a UID and GID specified by the system administartor via Lustre
2092 configuration management server (MGS). The functionality also
2093 allows to specify sets of clients for which the remapping does
2098 Description: Excessive recovery window
2099 Details : With AT enabled, the recovery window can be excessively long (6000+
2100 seconds). To address this problem, we no longer use
2101 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2102 timeout no longer depends on the service time, it is set to
2103 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2104 time via pb_service_time.
2108 Description: Watchdog triggered on MDS failover
2109 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2110 the OSTs know that the MDS "UUID" can be reused for the same export
2111 from a different NID, so we do not need to wait for the export to be
2115 Frequency : rare, only if using MMP with Linux RAID
2117 Description: MMP doesn't work with Linux RAID
2118 Details : While using HA for Lustre servers with Linux RAID, it is possible
2119 that MMP will not detect multiple mounts. To make this work we
2120 need to unplug the device queue in RAID when the MMP block is being
2121 written. Also while reading the MMP block, we should read it from
2122 disk and not the cached one.
2124 Severity : enhancement
2126 Description: open file using fid
2127 Details : A file can be opened using just its fid, like
2128 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2131 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2134 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2135 Details : add proper locking for creation of jbd_4k slab cache
2139 Description: MMP check in ext3_remount() fails without displaying any error
2140 Details : When multiple mount protection fails during remount, proper error
2143 Severity : enhancement
2145 Description: Allow stripe size to be up to 4G-64k
2146 Details : Fix math logic to allow large stripe sizes.
2150 Description: add check for >8TB ldiskfs filesystems
2151 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2152 Don't allow >8TB ldiskfs filesystems to be mounted without
2153 force_over_8tb mount option
2158 Descriptoin: OSS complains "too many blocks" and crashes
2159 Details : In filter_commitrw_write():
2160 1. lquota_chkquota();
2161 2. obdo_from_inode();
2162 3. lquota_pending_commit();
2163 "2." has changed uid and gid in this case, then caused
2164 lquota_chkquota() and lquota_pending_commit() to use
2165 different uid and gid. That is the root of the bug.
2167 --------------------------------------------------------------------------------
2169 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2171 * Support for kernels:
2172 2.6.5-7.283 (SLES 9),
2173 2.6.9-55.EL (RHEL 4),
2174 2.6.16.46-0.14 (SLES 10),
2175 2.6.18.8 vanilla (kernel.org)
2176 * Client support for unpatched kernels:
2177 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2178 2.6.16 - 2.6.22 vanilla (kernel.org)
2179 * Due to recently discovered recovery problems, we do not recommend
2180 using patchless RHEL 4 clients with this or any earlier release.
2181 * Recommended e2fsprogs version: 1.39.cfs8
2182 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2183 * Starting with this release, the ldiskfs backing filesystem required
2184 by Lustre is now in its own package, lustre-ldiskfs. This package
2185 should be installed. It is versioned separately from Lustre and
2186 may be released separately in future.
2190 Description: block reactivating mgc import until all deactivates complete
2191 Details : Fix race when failing back MDT/MGS to itself (testing)
2194 Frequency : at statup only
2196 Description: mds_lov_synchronize race leads to various problems
2197 Details : simultaneous MDT->OST connections at startup can cause the
2198 sync to abort, leaving the OSC in a bad state.
2200 Severity : enhancement
2202 Description: add optional extra BUILD_VERSION info
2203 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2204 to override the lustre version.
2207 Frequency : 2.6.18 servers only
2209 Description: ll_kern_mount() doesn't release the module reference
2210 Details : The ldiskfs module reference count never drops down to 0
2211 because ll_kern_mount() doesn't release the module reference.
2216 Description: server LBUG when using old ost_num_threads parameter
2217 Details : Accept the old ost_num_threads parameter but warn that it
2218 is deprecated, and fix an off-by-one error that caused an LBUG.
2223 Description: Transient SCSI error results in persistent IO issue
2224 Details : iobuf->dr_error is not reinitialized to 0 between two
2228 Frequency : sometimes when underlying device returns I/O errors
2230 Description: OSTs not going read-only during write failures
2231 Details : OSTs are not remounted read-only when the journal commit threads
2232 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2233 instead of the ext3 wrappers.
2237 Description: poor connect scaling with increasing client count
2238 Details : Don't run filter_grant_sanity_check for more than 100 exports
2239 to improve scaling for large numbers of clients.
2242 Frequency : SLES10 only
2244 Description: sanity-quota.sh quotacheck failed: rc = -22
2245 Details : Quotas cannot be enabled on SLES10.
2248 Frequency : liblustre clients only
2250 Description: getdirentries does not give error when run on compute nodes
2251 Details : getdirentries does not fail when the size specified as an argument
2252 is too small to contain at least one entry
2254 Severity : enhancement
2256 Description: Add LNET router traceability for debug purposes
2257 Details : If a checksum failure occurs with a router as part of the
2258 IO path, the NID of the last router that forwarded the bulk data
2259 is printed so it can be identified.
2264 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2265 Details : Due to a race condition, liblustre clients were occasionally
2266 evicted incorrectly.
2268 Severity : enhancement
2270 Description: lfs setstripe use optional parameters instead of postional
2273 Severity : enhancement
2275 Description: Nanosecond timestamp support for ldiskfs
2276 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2277 resolution timestamps. There is not yet support for this at
2278 the Lustre filesystem level.
2281 Frequency : during server recovery
2283 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2284 Details : request with rq_no_resend flag not awake l_wait_event if they get
2288 Frequency : nfs export on patchless client
2290 Description: connectathon hang when test nfs export over patchless client
2291 Details : Disconnected dentry cannot be found with lookup, so we do not need
2292 to unhash it or make it invalid
2295 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2296 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2297 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2302 Description: open req refcounting wrong on reconnect
2303 Details : If reconnect happened between getting open reply from server and
2304 call to mdc_set_replay_data in ll_file_open, we will schedule
2305 replay for unreferenced request that we are about to free.
2306 Subsequent close will crash in variety of ways.
2307 Check that request is still eligible for replay in
2308 mdc_set_replay_data().
2313 Description: disable writes to filesystem when reading health_check file
2314 Details : the default for reading the health_check proc file has changed
2315 to NOT do a journal transaction and write to disk, because this
2316 can cause reads of the /proc file to hang and block HA state
2317 checking on a healthy but otherwise heavily loaded system. It
2318 is possible to return to the previous behaviour during configure
2319 with --enable-health-write.
2321 Severity : enhancement
2323 Description: 64-bit inode version
2324 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2325 to the inode. This will be required for version-based recovery.
2330 Description: MDS fails to start if a duplicate client export is detected
2331 Details : in some rare cases it was possible for a client to connect to
2332 an MDS multiple times. Upon recovery the MDS would detect this
2333 and fail during startup. Handle this more gracefully.
2335 Severity : enhancement
2337 Description: Add -o localflock option to simulate old noflock
2339 Details : This will achieve local-only flock/fcntl locks
2345 Description: log_commit_thread vs filter_destroy race leads to crash
2346 Details : Take import reference before releasing llog record semaphore
2351 Description: Wrong request locking in request set processing
2352 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2353 imp_delayed_list, in this place should be used imp_lock.
2356 Frequency : when reconnection
2358 Description: Grant Leak when osc reconnect to OST
2359 Details : When osc reconnect ost, OST(filter) should check whether it
2360 should grant more space to client by comparing fed_grant and
2361 cl_avail_grant, and return the granted space to client instead
2362 of "new granted" space, because client will call osc_init_grant
2363 to update the client grant space info.
2366 Frequency : when client reconnect to OST
2368 Description: Grant Leak when osc do resend and replay bulk write
2369 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2370 bulk write request, because the grant info while be sync between
2371 OSC and OST when reconnect, and we should ignore the grant info
2372 these of resend/replay write req.
2377 Description: Grant space more than avaiable left space sometimes.
2378 Details : When then OST is about to be full, if two bulk writing from
2379 different clients came to OST. Accord the avaliable space of the
2380 OST, the first req should be permitted, and the second one
2381 should be denied by ENOSPC. But if the seconde arrived before
2382 the first one is commited. The OST might wrongly permit second
2383 writing, which will cause grant space > avaiable space.
2386 Frequency : when client is evicted
2388 Description: Grant might be wrongly erased when osc is evicted by OST
2389 Details : when the import is evicted by server, it will fork another
2390 thread ptlrpc_invalidate_import_thread to invalidate the
2391 import, where the grant will be set to 0. While the original
2392 thread will update the grant it got when connecting. So if
2393 the former happened latter, the grant will be wrongly errased
2394 because of this race.
2399 Description: Checking Stale with correct fid
2400 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2401 to check whether it is stale, because sometimes, we want the
2402 enqueue happened anyway, and op_data.fid2 will not be initialized.
2404 Severity : enhancement
2406 Description: update patchless client
2407 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2410 Frequency : only with 2.4 kernel
2412 Description: random memory corruption
2413 Details : size of struct ll_inode_info is to big for union inode.u and this
2414 can be cause of random memory corruption.
2419 Description: Memory leak in recovery
2420 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2421 It should also check obd_fail before fsfilt_start, otherwise if
2422 fsfilt_start return -EROFS,(failover mds during mds recovery).
2423 then the req will return with repmsg->transno = 0 and rc = EROFS.
2424 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2425 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2426 be freed no matter whether fsfilt_commit success or not.
2429 Frequency : only with huge count clients
2431 Description: Prevents from taking the superblock lock in llap_from_page for
2433 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2434 for llap_from_page call in ll_removepage prevents from taking the
2435 superblock lock for a soon died page.
2440 Description: Not check open intent error before release open handle
2441 Details : in some rare cases, the open intent error is not checked before
2442 release open handle, which may cause
2443 ASSERTION(open_req->rq_transno != 0), because it tries to release
2444 the failed open handle.
2449 Description: Set cat log bitmap only after create log success.
2450 Details : in some rare cases, the cat log bitmap is set too early. and it
2451 should be set only after create log success.
2455 Description: Accessing a block bevice can re-enable I/O when Lustre is
2456 tearing down a device.
2457 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2462 Description: service threads may hog cpus when there are a lot of requests
2463 Details : Insert cond_resched to give other threads a chance to use some CPU
2468 Description: the cat log was not initialized in recovery
2469 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2470 unlink log on mds will not be initialized until mds post
2471 recovery. And also in mds post recovery, the unlink log will
2472 initialization will be done asynchronausly, so there will be race
2473 between add unlink log and unlink log initialization.
2477 Description: brw_stats were being printed incorrectly
2478 Details : brw_stats were being printed as log2 but all of them were not
2479 recorded as log2. Also remove some code duplication arising from
2480 filter_tally_{read,write}.
2484 Frequency : rare, only in recovery.
2485 Description: ASSERTION(req->rq_type != LI_POISON) failed
2486 Details : imp_lock should be held while iterating over imp_sending_list for
2487 prevent destroy request after get timeout in ptlrpc_queue_wait.
2491 Description: replay-single.sh test 52 fails
2492 Details : A lock's skiplist need to be cleanup when it being unlinked
2493 from its resource list.
2497 Description: Short directio read returns full requested size rather than
2499 Details : Direct I/O operations should return actual amount of bytes
2500 transferred rather than requested size.
2502 Severity : enhancement
2504 Description: metadata RPC reduction (e.g. for rm performance)
2505 Details : decrease the amount of synchronous RPC between clients and servers
2506 by canceling conflicing lock before the operation on the client
2507 and packing thier handles into the main operation RPC to server.
2509 Severity : enhancement
2511 Description: Async OSC create to avoid the blocking unnecessarily.
2512 Details : If a OST has no remain object, system will block on the creating
2513 when need to create a new object on this OST. Now, ways use
2514 pre-created objects when available, instead of blocking on an
2515 empty osc while others are not empty. If we must block, we block
2516 for the shortest possible period of time.
2520 Description: improve handling recoverable errors
2521 Details : if request processig with error which can be recoverable on server
2522 request should be resend, otherwise page released from cache and
2525 Severity : enhancement
2527 Description: refine locking for avoid write wrong info into lov_objid file
2528 Details : fix possible races with add new target and write/update data in
2531 --------------------------------------------------------------------------------
2533 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2538 Frequency : on some architectures
2540 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2541 Details : Uninitialized flags sometimes cause configuration commands to
2545 Frequency : patchless clients only
2547 Description: missing __iget() symbol export
2548 Details : The __iget() symbol export is missing. To avoid the need for
2549 this on patchless clients the deathrow inode reaper is turned
2550 off, and we depend on the VM to clean up old inodes. This
2551 dependency was during via the fix for bug 12181.
2553 --------------------------------------------------------------------------------
2555 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2557 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2558 INTEROPERATE with older versions automatically. In many cases a
2559 special upgrade step is needed. Please read the
2560 user documentation before upgrading any part of a live system.
2561 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2562 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2563 * WARNING: Lustre configuration and startup changes are required with
2564 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2568 Severity : enhancement
2570 Description: MountConf
2571 Details : Lustre configuration is now managed via mkfs and mount
2572 commands instead of lmc and lconf. New obd types (MGS, MGC)
2573 are added for dynamic configuration management. See
2574 https://mail.clusterfs.com/wikis/lustre/MountConf for
2577 Severity : enhancement
2579 Description: dynamic OST addition
2580 Details : OSTs can now be added to a live filesystem
2582 Severity : enhancement
2584 Description: startup order invariance
2585 Details : MDTs and OSTs can be started in any order. Clients only
2586 require the MDT to complete startup.
2588 Severity : enhancement
2590 Description: parallel, asynchronous orphan cleanup
2591 Details : orphan cleanup is now performed in separate threads for each
2592 OST, allowing parallel non-blocking operation.
2594 Severity : enhancement
2596 Description: optimized stripe assignment
2597 Details : stripe assignments are now made based on ost space available,
2598 ost previous usage, and OSS previous usage, in order to try
2599 to optimize storage space and networking resources.
2601 Severity : enhancement
2603 Description: Permanently set tunables
2604 Details : All writable /proc/fs/lustre tunables can now be permanently
2605 set on a per-server basis, at mkfs time or on a live system.
2607 Severity : enhancement
2609 Description: Lustre message v2
2610 Details : Add lustre message format v2.
2612 Severity : enhancement
2614 Description: client OST exclusion list
2615 Details : Clients can be started with a list of OSTs that should be
2616 declared "inactive" for known non-responsive OSTs.
2620 Description: SPEC SFS validation failure on NFS v2 over lustre.
2621 Details : Changes the blocksize for regular files to be 2x RPC size,
2622 and not depend on stripe size.
2624 Severity : enhancement
2626 Description: Multiple MD RPCs in flight.
2627 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2628 To avoid overly-overloading MDS, introduce a limit on number of
2629 MDS RPCs in flight for a single client and add /proc controls
2630 to adjust this limit.
2632 Severity : enhancement
2634 Description: client read/write statistics
2635 Details : Add client read/write call usage stats for performance
2636 analysis of user processes.
2637 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2638 file access. extents_stats shows chunk size distribution.
2639 extents_stats_per_process show chunk size distribution per
2642 Severity : enhancement
2644 Description: mds statistics
2645 Details : Add detailed mds operations statistics in
2646 /proc/fs/lustre/mds/*/stats.
2650 Description: Failure of copying files with lustre special EAs.
2651 Details : Client side always return success for setxattr call for lustre
2652 special xattr (currently only "trusted.lov").
2657 Description: Refcount LNET uuids
2658 Details : The global LNET uuid list grew linearly with every startup;
2659 refcount repeated list entries instead of always adding to
2662 Severity : enhancement
2664 Description: Dynamic service threads
2665 Details : Within a small range, start extra service threads
2666 automatically when the request queue builds up.
2669 Frequency : mixed-endian client/server environments
2671 Description: mixed-endian crashes
2672 Details : The new msg_v2 system had some failures in mixed-endian
2675 Severity : enhancement
2677 Description: Easy OST removal
2678 Details : OSTs can be permanently deactivated with e.g. 'lctl
2679 conf_param lustre-OST0001.osc.active=0'
2681 Severity : enhancement
2683 Description: MGS proc entries
2684 Details : Added basic proc entries for the MGS showing what filesystems
2687 Severity : enhancement
2689 Description: provide MGS failover
2690 Details : Added config lock reacquisition after MGS server failover.
2692 Severity : enhancement
2694 Description: add Linux 2.4 support
2695 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2699 Description: a large application tries to do I/O to the same resource and dies
2700 in the middle of it.
2701 Details : Check the req->rq_arrival time after the call to
2702 ost_brw_lock_get(), but before we do anything about
2703 processing it & sending the BULK transfer request. This
2704 should help move old stale pending locks off the queue as
2705 quickly as obd_timeout.
2708 Frequency : when an incorrect nid is specified during startup
2710 Description: ptlrpc connect to non-existant node causes kernel crash
2711 Details : LNET can't be re-entered from an event callback, which
2712 happened when we expire a message after the export has been
2713 cleaned up. Instead, hand the zombie cleanup off to another
2716 Severity : enhancement
2718 Description: plain/inodebits lock performance improvement
2719 Details : Grouping plain/inodebits in granted list by their request modes
2720 and bits policy, thus improving the performance of search through
2724 Frequency : only if OST filesystem is corrupted
2726 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2727 Details : for a short time RPCs with bulk IO are in the replay list,
2728 but replay of bulk IOs is unimplemented. If the OST filesystem
2729 is corrupted due to disk cache incoherency and then replay is
2730 started it is possible to trip an assertion. Avoid putting
2731 committed RPCs into the replay list at all to avoid this issue.
2734 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2736 Description: System hang on startup
2737 Details : This bug allowed the liblustre (e.g. catamount) client to
2738 return to the app before handling all startup RPCs. This
2739 could leave the node unresponsive to lustre network traffic
2740 and manifested as a server ptllnd timeout.
2742 Severity : enhancement
2744 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2745 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2746 boolean to control whether to print peer debug info when a
2747 client's RPC times out.
2750 Frequency : only for kernels with patches from Lustre below 1.4.3
2752 Description: Remove old rdonly API
2753 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2756 Frequency : only for devices with external journals
2758 Description: Set external device read-only also
2759 Details : During a commanded failover stop, we set the disk device
2760 read-only while the server shuts down. We now also set any
2761 external journal device read-only at the same time.
2764 Frequency : when upgrading from 1.4 while trying to change parameters
2766 Description: The wrong (new) MDC name was used when setting parameters for
2767 upgraded MDT's. Also allows changing of OSC (and MDC)
2768 parameters if --writeconf is specified at tunefs upgrade time.
2771 Frequency : when setting specific ost indicies
2773 Description: QOS code breaks on skipped indicies
2774 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2775 created with --index need not be sequential.
2777 Severity : enhancement
2779 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2780 Details : The uninit_groups feature works in conjunction with the kernel
2781 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2782 up the pass1 processing of e2fsck. This is a read-only feature
2783 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2784 mount filesystems that have had this feature enabled.
2786 Severity : enhancement
2788 Description: Improve multi-block allocation algorithm to avoid fragmentation
2789 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2790 allocation locality and avoid filesystem fragmentation.
2792 ------------------------------------------------------------------------------
2794 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2796 * Support for kernels:
2797 2.4.21-47.0.1.EL (RHEL 3)
2798 2.6.5-7.283 (SLES 9)
2799 2.6.9-42.0.10.EL (RHEL 4)
2800 2.6.12.6 vanilla (kernel.org)
2801 2.6.16.27-0.9 (SLES 10)
2802 * Recommended e2fsprogs version: 1.39.cfs5
2804 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2808 Frequency : occasional, depends on client load and configuration
2809 Bugzilla : 12181, 12203
2810 Description: data loss for recently-modified files
2812 Details : In some cases it is possible that recently written or created
2813 files may not be written to disk in a timely manner (this should
2814 normally be within 30s unless client IO load is very high).
2815 The problem appears as zero-length files or files that are a
2816 multiple of 1MB in size after a client crash or client eviction
2817 that are missing data at the end of the file.
2819 This problem is more likely to be hit on clients where files are
2820 repeatedly created and unlinked in the same directory, clients
2821 have a large amount of RAM, have many CPUs, the filesystem has
2822 many OSTs, the clients are rebooted frequently, and/or the files
2823 are not accessed by other nodes after being written.
2825 The presence of the problem can be detected by looking at
2826 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2827 smaller than the second (nr_unused) then dirty files will not
2828 be flushed automatically to disk. "sync; sleep 10" should be
2829 run several times on the node before unmounting it to update
2830 Lustre (this is also safe to run on nodes without this problem).
2832 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2833 that can cause this same problem, so customers using that kernel
2834 also need to update the kernel in addition to Lustre. In order
2835 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2837 It is normal that files written just before a client crash (less
2838 than 30s) may not yet have been flushed to disk, even for local
2842 Frequency : frequent on thin XT3 nodes
2844 Description: UUID collision on thin XT3 Linux nodes
2845 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2846 randomly, since we relied on an insufficiently-seeded PRNG.
2851 Description: OSS hangs after "All ost request buffers busy"
2852 Details : A deadlock between quota and journal operations caused OSS
2853 hangs after printing "All ost request buffers busy."
2856 Frequency : always on liblustre builds
2858 Description: Cleanup compiler warnings on liblustre
2861 Frequency : always on liblustre builds on XT3
2863 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2867 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2868 > 32000 subdirectories
2869 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2870 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2871 subdirectory count crosses 32000. This will aid e2fsck to
2872 correctly handle more than 32000 subdirectories.
2875 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2877 Description: System hang on startup
2878 Details : This bug allowed the liblustre (e.g. catamount) client to
2879 return to the app before handling all startup RPCs. This
2880 could leave the node unresponsive to lustre network traffic
2881 and manifested as a server ptllnd timeout.
2883 Severity : enhancement
2885 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2886 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2887 boolean to control whether to print peer debug info when a
2888 client's RPC times out.
2893 Description: make O_SYNC working on 2.6 kernels
2894 Details : 2.6 kernels use different method for mark pages for write,
2895 so need add a code to lustre for O_SYNC work.
2900 Description: Failure to close file and release space on NFS
2901 Details : Put inode details into lock acquired in ll_intent_file_open.
2902 Use mdc_intent_lock in ll_intent_open to properly
2903 detect all kind of errors unhandled by mdc_enqueue.
2908 Description: proc file read during shutdown sometimes raced obd removal,
2910 Details : Add lock to prevent obd access after proc file removal.
2913 Frequency : Only for files larger than 4GB on 32-bit clients.
2915 Description: improperly doing page alignment of locks
2916 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2917 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2922 Description: RPCs being resent when they shouldn't be
2923 Details : Some RPCs that should not be resent are being resent. This
2924 can cause inconsistencies in the RPC state machine. Do not
2925 resend such requests.
2928 Frequency : rare, only with NFS export
2930 Description: Crash on NFS re-export node
2931 Details : under very unusual load conditions an assertion is hit in
2932 ll_intent_file_open()
2935 Frequency : only if OST filesystem is corrupted
2937 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2938 Details : for a short time RPCs with bulk IO are in the replay list,
2939 but replay of bulk IOs is unimplemented. If the OST filesystem
2940 is corrupted due to disk cache incoherency and then replay is
2941 started it is possible to trip an assertion. Avoid putting
2942 committed RPCs into the replay list at all to avoid this issue.
2947 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2948 Details : Large single O_DIRECT read and write calls can fail to allocate
2949 a sufficiently large buffer to process the request. In case of
2950 allocation failure the allocation is retried with a smaller
2951 buffer and broken into smaller requests.
2953 Severity : enhancement
2955 Description: Add -o localflock option to simulate old noflock behaviour.
2956 Details : This will achieve local-only flock/fcntl locks coherentness.
2961 Description: versioning check is incomplete
2962 Details : Checking the version difference of client vs. server, report
2963 error if the gap is too big.
2968 Description: add support PG_writeback bit
2969 Details : add support for PG_writeback bit for Lustre, for more carefull
2970 work with page cache in 2.6 kernel. This also fix some deadlocks
2971 and remove hack for work O_SYNC with 2.6 kernel.
2973 Severity : enhancement
2975 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2976 Details : The uninit_groups feature works in conjunction with the kernel
2977 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2978 up the pass1 processing of e2fsck. This is a read-only feature
2979 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2980 mount filesystems that have had this feature enabled.
2982 Severity : enhancement
2984 Description: Improve multi-block allocation algorithm to avoid fragmentation
2985 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2986 allocation locality and avoid filesystem fragmentation.
2988 ------------------------------------------------------------------------------
2990 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2992 * Support for kernels:
2993 2.6.9-42.0.3.EL (RHEL 4)
2994 2.6.5-7.276 (SLES 9)
2995 2.4.21-47.0.1.EL (RHEL 3)
2996 2.6.12.6 vanilla (kernel.org)
2997 2.6.16.21-0.8 (SLES10)
2998 * Recommended e2fsprogs version: 1.39.cfs2-0
3000 * The backwards-compatible /proc/sys/portals symlink has been removed
3001 in this release. Before upgrading, please ensure that you change
3002 any configuration scripts or /etc/sysctl.conf files that access
3003 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3004 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3005 in advance of the upgrade on any system running Lustre 1.4.6 or
3006 newer, since /proc/sys/lnet was added in that version.
3007 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3011 Frequency : only when quota is used
3013 Description: avoid scanning export list for quota master
3014 Details : Change the algorithms to avoid scanning export list in order
3015 to improve the efficiency.
3018 Frequency : MDS failover only, very rarely
3020 Description: "went back in time" messages on mds failover
3021 Details : The greatest transno may be lost when the current operation
3022 finishes with an error (transno==0) and the client's last_rcvd
3023 record is over-written. Save the greatest transno in the
3024 mds_last_transno for this case.
3027 Frequency : always for specific kernels and striping counts
3029 Description: client may get "Matching packet too big" without ACL support
3030 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3031 when trying to access files in certain configurations. The
3032 clients should in fact be denied when mounting because they do
3033 not understand ACLs.
3036 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3038 Description: many clients connecting with IO in progress causes connect timeouts
3039 Details : Avoid synchronous journal commits to avoid delays caused by many
3040 clients connecting/disconnecting when bulk IO is in progress.
3041 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3042 OST_IO_PORTAL to avoid delays behind potentially many pending
3046 Frequency : occasionally with multiple writers to a single file
3048 Description: shared writes to file may result in wrong size reported by stat()
3049 Details : Allow growing of kms when extent lock is cancelled
3052 Frequency : always with random mmap IO to multi-striped file
3054 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3055 Details : Only if the hole is at the end of OST object so that kms is too
3056 small. Fix is to increase kms accordingly in ll_nopage.
3059 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3061 Description: writes to a missing object would leak memory on the OST
3062 Details : If there is an inconsistency between the MDS and OST filesystems,
3063 such that the MDS references an object that doesn't exist, writes
3064 to that object will leak memory due to incorrect cleanup in the
3065 error handling path, eventually running out of memory on the OST.
3070 Description: Creating too long symlink causes lustre errors
3071 Details : Check symlink and name lengths before sending requests to MDS.
3074 Frequency : only if flock is enabled (not on by default)
3076 Description: posix locks not released on fd closure on 2.6.9+
3077 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3078 kernels, this caused such locks not to be released on fd close and
3079 then assertions on fs unmount about still used locks.
3082 Frequency : MDS failover only, very rarely
3084 Description: clients may get ASSERTION(granted_lock != NULL)
3085 Details : When request was taking a long time, and a client was resending
3086 a getattr by name lock request. The were multiple lock requests
3087 with the same client lock handle and
3088 mds_getattr_name->fixup_handle_for_resent_request found one of the
3089 lock handles but later failed with ASSERTION(granted_lock != NULL).
3094 Description: handle->h_buffer_credits > 0, assertion failure
3095 Details : h_buffer_credits is zero after truncate, causing assertion
3096 failure. This patch extends the transaction or creates a new
3100 Frequency : NFS re-export or patchless client
3101 Bugzilla : 11179, 10796
3102 Description: Crash on NFS re-export node (__d_move)
3103 Details : We do not want to hash the dentry if we don't have a lock.
3104 But if this dentry is later used in d_move, we'd hit uninitialised
3105 list head d_hash, so we just do this to init d_hash field but
3106 leave dentry unhashed.
3109 Frequency : NFS re-export or patchless client
3111 Description: NFS exports has problem with symbolic link
3112 Details : lustre client didn't properly install dentry when re-exported
3113 to NFS or running patchless client.
3116 Frequency : NFS re-export or patchless client
3118 Description: Various nfs/patchless fixes.
3119 Details : fixes reuse disconected alias for lookup process - this fixes
3120 warning "find_exported_dentry: npd != pd",
3121 fix permission error with open files at nfs.
3122 fix apply umask when do revalidate.
3125 Frequency : occasional
3127 Description: Crash on NFS re-export node
3128 Details : calling clear_page() on the wrong pointer triggered oops in
3129 generic_mapping_read().
3132 Frequency : rarely, using O_DIRECT IO
3134 Description: unaligned directio crashes client with LASSERT
3135 Details : check for unaligned buffers before trying any requests.
3138 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3140 Description: stale data returned from RAID cache
3141 Details : If only a small amount of IO is done to the RAID device before
3142 reading it again it is possible to get stale data from the RAID
3143 cache instead of reading it from disk.
3146 Frequency : always for sles10 kernel
3148 Description: sles10 support
3149 Details : ll_follow_link: compile fixes and using of nd_set_link
3150 under newer kernels.
3153 Frequency : depends on arch, kernel and compiler version, always on sles10
3156 Description: recursive or deep enough symlinks cause stack overflow
3157 Details : getting rid of large stack-allocated variable in
3161 Frequency : depends on hardware
3163 Description: lustre write performance loss in the SLES10 kernel
3164 Details : the performance loss is caused by using of write barriers in the
3165 ext3 code. The SLES10 kernel turns barrier support on by
3166 default. The fix is to undo that change for ldiskfs.
3168 ------------------------------------------------------------------------------
3170 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3172 * Support for kernels:
3173 2.6.9-42.0.3EL (RHEL 4)
3174 2.6.5-7.276 (SLES 9)
3175 2.4.21-47.0.1.EL (RHEL 3)
3176 2.6.12.6 vanilla (kernel.org)
3180 Frequency : quota enabled and large files being deleted
3182 Description: releasing more than 4GB of quota at once hangs OST
3183 Details : If a user deletes more than 4GB of files on a single OST it
3184 will cause the OST to spin in an infinite loop. Release
3185 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3190 Description: statfs data retrieved from /proc may be stale or zero
3191 Details : When reading per-device statfs data from /proc, in the
3192 {kbytes,files}_{total,free,avail} files, it may appear
3193 as zero or be out of date.
3196 Frequency : systems with MD RAID1 external journal devices
3198 Description: lconf's call to blkid is confused by RAID1 journal devices
3199 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3200 of returning all block devices that match the journal UUID.
3203 Frequency : always, for aggregate stripe size over 4GB
3205 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3206 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3207 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3210 Frequency : always if "lfs find" used on a local file/directory
3212 Description: "lfs find" segfaults if used on a local file/directory
3213 Details : The case where a directory component was not specified wasn't
3214 handled correctly. Handle this properly.
3217 Frequency : always on ppc64
3219 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3220 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3221 perform endianness conversion.
3224 Frequency : rarely (truncate to non-zero file size after write under load)
3225 Bugzilla : 10730, 10687
3226 Description: Files padded with zeros to next 4K multiple
3227 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3228 it is possible that files that are truncated to a non-zero size
3229 immediately after being written are filled with zero bytes beyond
3230 the truncated size. No file data is lost.
3232 Severity : enhancement
3234 Description: Allow recovery/failover for liblustre clients.
3235 Details : liblustre clients were unaware of failover configurations until
3238 Severity : enhancement
3240 Description: user file locks should fail when not mounting with flock option
3241 Details : Set up an error-returning stub in ll_file_operations.lock field
3242 to prevent incorrect behaviour when client is mounted without
3243 flock option. Also, set up properly f_op->flock field for
3247 Frequency : always on ia64
3249 Description: "lfs df" loops on printing out MDS statfs information
3250 Details : The obd_ioctl_data was not initialized and in some systems
3251 this caused a failure during the ioctl that did not return
3252 an error. Initialize the struct and return an error on failure.
3255 Frequency : SLES 9 only
3257 Description: Error of copying files with lustre special EAs as root
3258 Details : Client side always return success for setxattr call for lustre
3259 special xattr (currently only "trusted.lov").
3262 Frequency : rarely on clusters with both ia64+i386 clients
3264 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3265 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3266 clients are doing concurrent writes to the same file it is
3267 possible that the ia64 clients may LASSERT because the OST
3268 extent locks are not PAGE_SIZE aligned. Ensure that grown
3269 locks are always aligned on the request boundary.
3272 Frequency : specific use, occasional
3274 Description: Overwriting in use executable truncates on-disk binary image
3275 Details : If one node attempts to overwrite an executable in use by
3276 another node, we now correctly return ETXTBSY instead of
3277 truncating the file.
3279 Severity : enhancement
3281 Description: Async OSC create to avoid the blocking unnecessarily.
3282 Details : If a OST has no remain object, system will block on the creating
3283 when need to create a new object on this OST. Now, ways use
3284 pre-created objects when available, instead of blocking on an
3285 empty osc while others are not empty. If we must block, we block
3286 for the shortest possible period of time.
3291 Description: chmod on Lustre root is propagated to other clients
3292 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3293 invalid by the follow_mount time.
3296 Frequency : liblustre clients only
3298 Description: Race in 'instant cancel' lock handling could lead to such locks
3299 never to be granted in case of SMP MDS
3300 Details : Do not destroy not yet granted but cbpending locks in
3304 Frequency : replay/resend of open
3306 Description: non null lock assetion failure in mds_intent_policy
3307 Details : Trying to replay/resend lockless open requests resulted in
3308 mds_open() returning 0 with no lock. Now it sets a flag if
3309 a lock is going to be returned.
3311 Severity : enhancement
3313 Description: Checksum enhancements
3314 Details : New checksum enhancements allow for resending RPCs that failed
3317 Severity : enhancement
3319 Description: Tunables on number of dirty pages in cacche
3320 Details : Allow to set limit on number of dirty pages cached.
3325 Description: client crash on unmount - lock still has references
3326 Details : In some error handling cases it was possible to leak a lock
3327 reference on a client while accessing a file. This was not
3328 harmful to the client during operation, but would cause the
3329 client to crash when the filesystem is unmounted.
3332 Frequency : specific case, rare
3334 Description: ETXTBSY on mds though file not in use by client
3335 Details : ETXTBSY is no longer incorrectly returned when attempting to
3336 chmod or chown a directory that the user previously tried to
3337 execute or a currently-executing binary.
3340 Frequency : extremely rare except on liblustre-based clients
3342 Description: Lustre space not freed when files are deleted
3343 Details : Clean up open-unlinked files after client eviction. Previously
3344 the unlink was skipped and the files remained as orphans.
3349 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3350 Details : In some cases it was possible to send a blocking callback to
3351 a client doing a glimpse, even though that client didn't get
3352 a lock granted. When the glimpse lock is cancelled on the OST
3353 the freed lock is left on the waiting list and corrupted the list.
3356 Frequency : all core dumps
3358 Description: Broke core dumps to lustre
3359 Details : Negative dentry may be unhashed if parent does not have UPDATE
3360 lock, but some callers, e.g. do_coredump, expect dentry to be
3361 hashed after successful create, hash it in ll_create_it.
3363 ------------------------------------------------------------------------------
3365 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3367 * Support for kernels:
3368 2.6.9-42.0.2.EL (RHEL 4)
3369 2.6.5-7.276 (SLES 9)
3370 2.4.21-40.EL (RHEL 3)
3371 2.6.12.6 vanilla (kernel.org)
3375 Frequency : always on RHEL 3
3377 Description: Number of open files grows over time
3378 Details : The number of open files grows over time, whether or not
3379 Lustre is started. This was due to a filp leak introduced
3380 by one of our kernel patches.
3382 ------------------------------------------------------------------------------
3384 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3386 * Support for kernels:
3387 2.6.9-42.EL (RHEL 4)
3388 2.6.5-7.276 (SLES 9)
3389 2.4.21-40.EL (RHEL 3)
3390 2.6.12.6 vanilla (kernel.org)
3395 Bugzilla : 5719, 9635, 9792, 9684
3396 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3397 Details : If a server is under heavy load and cannot reply to new
3398 connection requests before the client resends the (re)connect,
3399 the connection handling code can behave badly if two service
3400 threads are concurrently handing separate (re)connections from
3401 the same client. Add better locking to the connection handling
3402 code, and ensure that only a single connection will be processed
3403 for a given client UUID, even if the lock is dropped.
3405 Severity : enhancement
3407 Description: add TCP zero-copy support to kernel
3408 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3409 sends if the hardware supports scatter-gather and checksumming.
3410 This allows socklnd to do client-write and server-read more
3411 efficiently and reduce CPU utilization from skbuf copying.
3414 Frequency : only if NFS exporting from client
3416 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3417 Details : If setting attributes on a file created under NFS that had
3418 never been opened it would be possible to oops the client
3419 if the file had no objects.
3422 Frequency : always for liblustre
3424 Description: liblustre client does MDS+OSTs setattr RPC for each write
3425 Details : When doing a write from a liblustre client, the client
3426 incorrectly issued an RPC to the MDS and each OST the file was
3427 striped over in order to update the timestamps. When writing
3428 with small chunks and many clients this could overwhelm the MDS
3429 with RPCs. In all cases it would slow down the write because
3430 these RPCs are unnecessary.
3432 Severity : enhancement
3434 Description: allow number of MDS service threads to be changed at module load
3435 Details : It is now possible to change the number of MDS service threads
3436 running. Adding "options mds mds_num_threads={N}" to the MDS's
3437 /etc/modprobe.conf will set the number of threads for the next
3438 time Lustre is restarted (assuming the "mds" module is also
3439 reloaded at that time). The default number of threads will
3440 stay the same, 32 for most systems.
3445 Description: OST crash if filesystem is unformatted or corrupt
3446 Details : If an OST is started on a device that has never been formatted
3447 or if the filesystem is corrupt and cannot even mount then the
3448 error handling cleanup routines would dereference a NULL pointer.
3453 Description: NULL pointer deref in llap_from_page.
3454 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3455 incorrect) mapping if the page was truncated/reclaimed while it was
3456 searched for. Check for this condition and skip such pages when
3457 doing readahead. Introduce extra check to llap_from_page() to
3458 verify page->mapping->host is non-NULL (so page is not anonymous).
3461 Frequency : Sometimes when using sys_sendfile
3463 Description: "page not covered by a lock" warnings from ll_readpage
3464 Details : sendfile called ll_readpage without right page locks present.
3465 Now we introduced ll_file_sendfile that does necessary locking
3466 around call to generic_file_sendfile() much like we do in
3470 Frequency : with certain MDS communication failures at client mount time
3472 Description: NULL pointer deref after failed client mount
3473 Details : a client connection request may delayed by the network layer
3474 and not be sent until after the PTLRPC layer has timed out the
3475 request. If the client fails the mount immediately it will try
3476 to clean up before the network times out the request. Add a
3477 reference from the request import to the obd device and delay
3478 the cleanup until the network drops the request.
3481 Frequency : occasionally during client (re)connect
3483 Description: assertion failure during client (re)connect
3484 Details : processing a client connection request may be delayed by the
3485 client or server longer than the client connect timeout. This
3486 causes the client to resend the connection request. If the
3487 original connection request is replied in this interval, the
3488 client may trip an assertion failure in ptlrpc_connect_interpret()
3489 which thought it would be the only running connect process.
3492 Frequency : only with obd_echo servers and clients that are rebooted
3494 Description: kernel BUG accessing uninitialized data structure
3495 Details : When running an obd_echo server it did not start the ping_evictor
3496 thread, and when a client was evicted an uninitialized data
3497 structure was accessed. Start the ping_evictor in the RPC
3498 service startup instead of the OBD startup.
3500 Severity : enhancement
3501 Bugzilla : 10193 (patchless)
3502 Description: Remove dependency on various unexported kernel interfaces.
3503 Details : No longer need reparent_to_init, exit_mm, exit_files,
3504 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3507 Frequency : rare (only users of deprecated and unsupported LDAP config)
3509 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3510 Details : LDAP apparently contains 'lustreName' attributes instead of
3511 'name'. A simple remapping of the name is sufficient.
3514 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3516 Description: waiting_locks_callback trips kernel BUG if client is evicted
3517 Details : Running with the dump_on_timeout debug flag turned on makes
3518 it possible that the waiting_locks_callback() can try to dump
3519 the Lustre kernel debug logs from an interrupt handler. Defer
3520 this log dumping to the expired_lock_main() thread.
3522 Severity : enhancement
3524 Description: Support NFS exporting on 2.6 kernels.
3525 Details : Implement non-rawops metadata methods for NFS server to use without
3526 changing NFS server code.
3529 Frequency : very rare (synthetic metadata workload only)
3531 Description: two racing renames might cause an MDS thread to deadlock
3532 Details : Running the "racer" program may cause one MDS thread to rename
3533 a file from being the source of a rename to being the target of
3534 a rename at exactly the same time that another thread is doing
3535 so, and the second thread has already enqueued these locks after
3536 doing a lookup of the target and is trying to relock them in
3537 order. Ensure that we don't try to re-lock the same resource.
3540 Frequency : only very large systems with liblustre clients
3542 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3543 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3544 all OSTs in parallel. This allows the eviction of stale liblustre
3545 clients to proceed much faster than if they were done in series,
3546 and also offers similar improvements for other set_info RPCs.
3551 Description: excessive CPU usage during initial read phase on client
3552 Details : During the initial read phase on a client, it would agressively
3553 retry readahead on the file, consuming too much CPU and impacting
3554 performance (since 1.4.5.8). Improve the readahead algorithm
3555 to avoid this, and also improve some other common cases (read
3556 of small files in particular, where "small" is files smaller than
3557 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3562 Description: MDS crash when receiving packet with unknown intent.
3563 Details : Do not LBUG in unknown intent case, just return -EFAULT
3565 Severity : enhancement
3566 Bugzilla : 9293, 9385
3567 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3568 Details : Do not serialize getattr (non-intent version) and statfs.
3571 Frequency : occasional, when OST network is overloaded/intermittent
3573 Description: client evicted by OST after bulk IO timeout
3574 Details : If a client sends a bulk IO request (read or write) the OST
3575 may evict the client if it is unresposive to its data GET/PUT
3576 request. This is incorrect if the network is overloaded (takes
3577 too long to transfer the RPC data) or dropped the OST GET/PUT
3578 request. There is no need to evict the client at all, since
3579 the pinger and/or lock callbacks will handle this, and the
3580 client can restart the bulk request.
3583 Frequency : Always when mmapping file with no objects
3585 Description: client crashes when mmapping file with no objects
3586 Details : Check that we actually have objects in a file before doing any
3587 operations on objects in ll_vm_open, ll_vm_close and
3593 Description: Request leak when working with deleted CWD
3594 Details : Introduce advanced request refcount tracking for requests
3595 referenced from lustre intent.
3597 Severity : Enhancement
3599 Description: Cache open file handles on client.
3600 Details : MDS now will return special lock along with openhandle, if
3601 requested and client is allowed to hold openhandle, even if unused,
3602 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3603 closing files for every read/write openration.
3605 Severity : Enhancement
3607 Description: Cache open negative dentries on client when possible.
3608 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3609 negative dentries on lock revocation.
3614 Description: Remounting a client read-only wasn't possible with a zconf mount
3615 Details : It wasn't possible to remount a client read-only with llmount.
3617 Severity : enhancement
3618 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3619 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3620 MPICH is included to add Lustre-specific ADIO interfaces.
3621 This is based closely on the UFS ADIO layer and only differs
3622 in file creation, in order to allow the OST striping to be set.
3623 This is user-contributed code and not supported by CFS.
3628 Description: extended inode attributes (immutable, append-only) work improperly
3629 when 2.4 and 2.6 kernels are used on client/server or vice versa
3630 Details : Introduce kernel-independent values for these flags.
3632 Severity : enhancement
3635 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3636 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3637 granularity. Now, Fractional MB granularity is be supported,
3638 this is very useful for low memory system.
3640 Severity : enhancement
3642 Description: Getattr by fid
3643 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3644 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3645 have locks and avoids getattr rpc requests for it on every stat.
3648 Frequency : Always, for filesystems larger than 2TB
3650 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3651 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3652 filesystems larger than 2TB to be mounted with mballoc enabled.
3655 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3657 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3658 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3659 filesystems larger than 2TB is mounted then the kernel will
3660 silently corrupt the start of the filesystem. CONFIG_LBD is
3661 enabled for all CFS-supported kernels, but the possibility of
3662 this happening with a modified kernel config exists.
3664 Severity : enhancement
3666 Description: add client O_DIRECT support for 2.6 kernels
3667 Details : It is now possible to do O_DIRECT reads and writes to files
3668 in the Lustre client mountpoint on 2.6 kernel clients.
3670 Severity : enhancement
3672 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3673 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3674 parallel, not waiting for response from every OST before sending
3675 a rpc to the next OST.
3680 Description: setattr vs write race when updating file timestamps
3681 Details : Client processes that update a file timestamp into the past
3682 right after writing to the file (e.g. tar) it is possible that
3683 the updated file modification time can be reset to the current
3684 time due to a race between processing the setattr and write RPC.
3686 Severity : enhancement
3688 Description: Bring 'lfs find' closer in line with regular Linux find.
3689 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3690 -print0 options and obtains all the needed info through the lustre
3693 Severity : enhancement
3695 Description: support up to 1024 configured devices on one node
3696 Details : change obd_dev array from statically allocated to dynamically
3697 allocated structs as they are first used to reduce memory usage
3702 Description: Flush dirty partially truncated pages during truncate
3703 Details : Immediatelly flush partially truncated pages in filter_setattr,
3704 this way we completely avoid having any pages in page cache on OST
3705 and can retire ugly workarounds during writes to flush such pages.
3710 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3711 Details : protect inode from truncation within vfs_unlink() context
3712 just take a reference before calling vfs_unlink() and release it
3713 when parent's i_sem is free.
3718 Description: last_id value checked outside lock on OST caused LASSERT failure
3719 Details : If there were multiple MDS->OST object precreate requests in
3720 flight, it was possible that the OST's last object id was checked
3721 outside a lock and incorrectly tripped an assertion. Move checks
3722 inside locks, and discard old precreate requests.
3725 Frequency : always, if extents are used on OSTs
3727 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3728 Details : OSTs using the extents format would not zero the high 16 bits of
3729 the index physical block number. This is not a problem for any
3730 OST filesystems smaller than 16TB, and no kernels support ext3
3731 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3732 new/modified files) and can be fixed for existing filesystems
3733 with e2fsprogs-1.39-cfs1.
3738 Description: import connection selection may be incorrect if timer wraps
3739 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3740 import connections to be ignored if the 32-bit jiffies counter
3741 wraps. Use a 64-bit jiffies counter.
3744 Frequency : very large clusters immediately after boot
3746 Description: LNET request buffers exhausted under heavy short-term load
3747 Details : If a large number of client requests are generated on a service
3748 that has previously never seen so many requests it is possible
3749 that the request buffer growth cannot keep up with the spike in
3750 demand. Instead of dropping incoming requests, they are held in
3751 the LND until the RPC service can accept more requests.
3754 Frequency : Sometimes during replay
3756 Description: Assertion failure in ll_local_open after replay.
3757 Details : If replay happened on an open request reply before we were able
3758 to set replay handler, reply will become not swabbed tripping the
3759 assertion in ll_local_open. Now we set the handler right after
3760 recognising of open request
3763 Frequency : very rare
3765 Description: kernel reports "badness in vsnprintf"
3766 Details : Reading from the "recovery_status" /proc file in small chunks
3767 may cause a negative length in lprocfs_obd_rd_recovery_status()
3768 call to vsnprintf() (which is otherwise harmless). Exit early
3769 if there is no more space in the output buffer.
3771 Severity : enhancement
3773 Description: clear OBD RPC statistics by writing to them
3774 Details : It is now possible to clear the OBD RPC statistics by writing
3775 to the "stats" file.
3780 Description: Client mtime is not the same on different clients after utimes
3781 Details : In some cases, the client was using the utimes() syscall on
3782 a file cached on another node. The clients now validate the
3783 ctime from the MDS + OSTs to determine which one is right.
3788 Description: Inability to activate failout mode
3789 Details : lconf script incorrectly assumed that in pythong string's numeric
3790 value is used in comparisons.
3793 Frequency : always with multiple stripes per file
3795 Description: Inefficient object allocation for mutli-stripe files
3796 Details : When selecting which OSTs to stripe files over, for files with
3797 a stripe count that divides evenly into the number of OSTs,
3798 the MDS is always picking the same starting OST for each file.
3799 Return the OST selection heuristic to the original design.
3804 Description: mount failures may take full timeout to return an error
3805 Details : Under some heavy load conditions it is possible that a
3806 failed mount can wait for the full obd_timeout interval,
3807 possibly several minutes, before reporting an error.
3808 Instead return an error as soon as the status is known.
3810 Frequency : quota enabled and large files being deleted
3812 Description: releasing more than 4GB of quota at once hangs OST
3813 Details : If a user deletes more than 4GB of files on a single OST it
3814 will cause the OST to spin in an infinite loop. Release
3815 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3820 Description: statfs data retrieved from /proc may be stale or zero
3821 Details : When reading per-device statfs data from /proc, in the
3822 {kbytes,files}_{total,free,avail} files, it may appear
3823 as zero or be out of date.
3826 Frequency : systems with MD RAID1 external journal devices
3828 Description: lconf's call to blkid is confused by RAID1 journal devices
3829 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3830 of returning all block devices that match the journal UUID.
3833 Frequency : always, for aggregate stripe size over 4GB
3835 Description: assertion fails when trying to use 4GB stripe size
3836 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3837 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3840 Frequency : always on ppc64
3842 Description: the first write on an ext3 filesystem with mballoc got stuck
3843 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3844 perform endianness conversion.
3846 ------------------------------------------------------------------------------
3848 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3850 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3851 INTEROPERATE with older versions automatically. Please read the
3852 user documentation before upgrading any part of a live system.
3853 * WARNING: Lustre networking configuration changes are required with
3854 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3857 * Support for kernels:
3858 2.6.9-22.0.2.EL (RHEL 4)
3859 2.6.5-7.244 (SLES 9)
3860 2.6.12.6 vanilla (kernel.org)
3863 Severity : enhancement
3864 Bugzilla : 7981/8208
3865 Description: Introduced Lustre Networking (LNET)
3866 Details : LNET is new networking infrastructure for Lustre, it includes
3867 a reorganized network configuration mode (see the user
3868 documentation for full details) as well as support for routing
3869 between different network fabrics. Lustre Networking Devices
3870 (LNDs) for the supported network fabrics have also been
3871 created for this new infrastructure.
3873 Severity : enhancement
3874 Description: Introduced Access control lists
3875 Details : clients can set ACLs on files and directories in order to have
3876 more fine-grained permissions than the standard Unix UGO+RWX.
3877 The MDS must be started with the "-o acl" mount option.
3879 Severity : enhancement
3880 Description: Introduced filesystem quotas
3881 Details : Administrators may now establish per-user quotas on the
3884 Severity : enhancement
3886 Description: Configuration change for the XT3
3887 The PTLLND is now used to run Lustre over Portals on the XT3
3888 The configure option(s) --with-cray-portals are no longer used.
3889 Rather --with-portals=<path-to-portals-includes> is used to
3890 enable building on the XT3. In addition to enable XT3 specific
3891 features the option --enable-cray-xt3 must be used.
3896 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3897 Details : A race between allocating a new llcd and re-getting the llcd_lock
3898 allowed another thread to grab newly-allocated llcd.
3900 Severity : enhancement
3902 Description: 2.6 OST async journal commit and locking fix to improve performance
3903 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3904 2.6 kernels are now async as they already were in 2.4 kernels so
3905 that they can commit concurrently with the network bulk transfer.
3906 For block-allocated files the filter allocation semaphore is held
3907 to avoid filesystem fragmentation during allocation. BKL lock
3908 removed for 2.6 xattr operations where it is no longer needed.
3913 Description: lconf incorrectly determined whether two IP networks could talk
3914 Details : In some more complicated routing and multiple-network
3915 configurations, lconf will avoid trying to make a network
3916 connection to a disjoint part of the IP space. It was doing the
3917 math incorrectly for one set of cases.
3922 Description: Fix for potential infinite loop processing records in an llog.
3923 Details : If an llog record is corrupted/zeroed, it is possible to loop
3924 forever in llog_process(). Validate the llog record length
3925 and skip the remainder of the block on error.
3928 Frequency : occasional (liblustre only)
3930 Description: liblustre could not open files whose last component is a symlink
3931 Details : sysio_path_walk() would incorrectly pass the open intent to
3932 intermediate path components.
3935 Frequency : rare (liblustre only with non-standard tuning)
3936 Bugzilla : 7201 (7350)
3937 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3938 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3939 after it has been released, only do this for referenced locks
3941 Severity : enhancement
3943 Description: specifying an (invalid) directory default stripe_size of -1
3944 would reset the directory default striping
3945 Details : stripe_size -1 was used internally to signal directory stripe
3946 removal, now use "all default" to signal dir stripe removal
3947 as a directory striping of "all default" is not useful
3950 Frequency : common for large clusters running liblustre clients
3952 Description: doing an ls when liblustre clients are running is slow
3953 Details : sending a glimpse AST to a liblustre client waits for every AST
3954 to time out, as liblustre clients will not respond. Since they
3955 cannot cache data we refresh the OST lock LVB from disk instead.
3957 Severity : enhancement
3959 Description: doing an ls at the same time as file IO can be slow
3960 Details : enqueue and other "small" requests can be blocked behind many
3961 large IO requests. Create a new OST IO portal for non-IO
3962 requests so they can be processed faster.
3965 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3967 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3968 created file with mode 000
3969 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3970 allow an ftruncate() as a non-root user to a file with mode 000.
3971 Lustre now respects this flag to disable mode checks when
3972 truncating a file owned by the user
3975 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3978 Description: Revoking locks from clients that went dead or catatonic might take
3980 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3981 cancellation of such locks instant on servers without waiting for
3982 any reply from clients. Clients drops these locks when cancel
3983 notification from server is received without replying.
3986 Frequency : liblustre-only, when liblustre client dies or becomes busy
3988 Description: Doing ls on Linux clients can take a long time with active
3990 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3991 granting such locks to it in the first place if possible. Locks
3992 are taken by proxy on the OST during the read or write and
3993 dropped immediately afterward. Add connect flags handling, do
3994 not grant locks to liblustre clients for glimpse ASTs.
3996 Severity : enhancement
3998 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3999 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4000 sequential reads and behaves badly with applications doing random
4001 reads. Improve it by reading ahead at least read region, and
4002 avoiding excessive large RPC for small reads.
4004 Severity : enhancement
4006 Description: Creating more than 1000 files for a single job may cause a load
4007 imbalance on the OSTs if there are also a large number of OSTs.
4008 Details : qos_prep_create() uses an OST index reseed value that is an
4009 even multiple of the number of available OSTs so that if the
4010 reseed happens in the middle of the object allocation it will
4011 still utilize the OSTs as uniformly as possible.
4016 Description: OST or MDS may oops in ping_evictor_main()
4017 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4018 but doesn't restart at beginning of obd_exports_timed list
4021 Severity : enhancement
4023 Description: improve by-nid export eviction on the MDS and OST
4024 Details : allow multiple exports with the same NID to be evicted at one
4025 time without re-searching the exports list.
4028 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4030 Description: MDS may oops in groups_free()
4031 Details : in rare race conditions a newly allocated group_info struct is
4032 freed again, and this can be NULL. The 2.4 compatibility code
4033 for groups_free() checked for a NULL pointer, but 2.6 did not.
4036 Frequency : common for liblustre clients doing little filesystem IO
4037 Bugzilla : 9352, 7313
4038 Description: server may evict liblustre clients accessing contended locks
4039 Details : if a client is granted a lock or receives a completion AST
4040 with a blocking AST already set it would not reply to the AST
4041 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4044 Frequency : lfs setstripe, only systems with more than 160 OSTs
4046 Description: unable to set striping with a starting offset beyond OST 160
4047 Details : llapi_create_file() incorrectly limited the starting stripe
4048 index to the maximum single-file stripe count.
4051 Frequency : LDAP users only
4053 Description: lconf did not handle in-kernel recovery with LDAP properly
4054 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4056 Severity : enhancement
4058 Description: bind OST threads to NUMA nodes to improve performance
4059 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4060 node and do their allocations there to localize memory access
4062 Severity : enhancement
4064 Description: llmount can determine client NID directly from Myrinet (GM)
4065 Details : the client NID code from gmnalnid was moved directly into
4066 llmount, removing the need to use this or specifying the
4067 client NID explicitly when mounting GM clients with zeroconf
4070 Frequency : if client is started with down MDS
4072 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4073 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4074 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4075 ptlrpc_import_delayed_req() because we didn't check if the
4076 request was interrupted, and we also didn't break out of the
4077 event loop if there was no timeout
4082 Description: data loss during non-page-aligned writes to a single file from
4083 both multiple nodes and multiple threads on one node at same time
4084 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4085 inconsistency led to false short-reads, that were cached and later
4086 used by ->prepare_write() to fill in partially written page,
4087 leading to data loss.
4090 Frequency : always, if lconf --abort_recovery used
4092 Description: lconf --abort_recovery fails with 'Operation not supported'
4093 Details : lconf was attempting to abort recovery on the MDT device and not
4096 Severity : enhancement
4098 Description: remove cleanup logs
4099 Details : replace lconf-generated cleanup logs with lustre internal
4100 cleanup routines. Eliminates the need for client-cleanup and
4103 Severity : enhancement
4105 Description: add support for EAs (user and system) on lustre filesystems
4106 Details : it is now possible to store extended attributes in the Lustre
4107 client filesystem, and with the user_xattr mount option it
4108 is possible to allow users to store EAs on their files also
4110 Severity : enhancement
4112 Description: Add possibility (config option) to show minimal available OST free
4114 Details : When compiled with --enable-mindf configure option, statfs(2)
4115 (and so, df) will return least minimal free space available from
4116 all OSTs as amount of free space on FS, instead of summary of
4117 free spaces of all OSTs.
4119 Severity : enhancement
4121 Description: do not expand extent locks acquired on OST-side
4122 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4123 by server: they are not cached anyway.
4126 Frequency : when mmap is used/binaries executed from Lustre
4128 Description: Unmmap pages before throwing them away from read cache.
4129 Details : llap_shrink cache now attempts to unmap pages before discarding
4130 them (if unmapping failed - do not discard). SLES9 kernel has
4131 extra checks that trigger if this unmapping is not done first.
4136 Description: lconf didn't resolve symlinks before checking to see whether a
4137 given mountpoint was already in use
4140 Frequency : when migrating failover services
4141 Bugzilla : 6395, 9514
4142 Description: When migrating a subset of services from a node (e.g. failback
4143 from a failover service node) the remaining services would
4144 time out and evict clients.
4145 Details : lconf --force (implied by --failover) sets the global obd_timeout
4146 to 5 seconds in order to quickly disconnect, but this caused
4147 other RPCs to time out too quickly. Do not change the global
4148 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4150 Severity : enhancement
4151 Frequency : if MDS is started with down OST
4152 Bugzilla : 9439,5706
4153 Description: Allow startup/shutdown of an MDS without depending on the
4154 availability of the OSTs.
4155 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4156 Add appropriate locking and lov-osc refcounts for safe
4157 cleaning. Add osc abort_inflight calls in case the
4158 synchronize never started.
4161 Frequency : occasional (Cray XT3 only)
4163 Description: root not authorized to access files in CRAY_PORTALS environment
4164 Details : The client process capabilities were not honoured on the MDS in
4165 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4166 been accessed by an authorized user then root was able to access
4167 the file on the local client also. The root user capabilities
4168 are now allowed on the MDS, as this environment has secure UID.
4171 Frequency : occasional
4173 Description: ldiskfs "too long searching" message happens too often
4174 Details : A debugging message (otherwise harmless) prints too often on
4175 the OST console. This has been reduced to only happen when
4176 there are fragmentation problems on the filesystem.
4181 Description: Division by zero in statfs when all OSCs are inactive
4182 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4183 lov_statfs divides by value returned by lov_get_stripecnt().
4187 Bugzilla : 9489, 3273
4188 Description: First write from each client to each OST was only 4kB in size,
4189 to initialize client writeback cache, which caused sub-optimal
4190 RPCs and poor layout on disk for the first writen file.
4191 Details : Clients now request an initial cache grant at (re)connect time
4192 and so that they can start streaming writes to the cache right
4193 away and always do full-sized RPCs if there is enough data.
4194 If the OST is rebooted the client also re-establishes its grant
4195 so that client cached writes will be honoured under the grant.
4200 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4201 Details : Now I/O RPCs go to different portal number and (presumably) fast
4202 lock requests (and glimses) and other RPCs get their own service
4203 threads pool that should be able to service those RPCs
4206 Severity : enhancement
4208 Description: Ability to exchange lustre version between client and servers and
4209 issue warnings at client side if client is too old. Also for
4210 liblustre clients there is ability to refuse connection of too old
4212 Details : New 'version' field is added to connect data structure that is
4213 filled with version info. That info is later checked by server and
4217 Frequency : rare, liblustre only.
4218 Bugzilla : 9296, 9581
4219 Description: Two simultaneous writes from liblustre at offset within same page
4220 might proceed at the same time overwriting eachother with stale
4222 Details : I/O lock withing llu_file_prwv was released too early, before data
4223 actually was hitting the wire. Extended lock-holding time until
4224 server acknowledges receiving data.
4227 Frequency : extremely rare. Never observed in practice.
4229 Description: avoid generating lustre_handle cookie of 0.
4230 Details : class_handle_hash() generates handle cookies by incrementing
4231 global counter, and can hit 0 occasionaly (this is unlikely, but
4232 not impossible, because initial value of cookie counter is
4233 selected randonly). Value of 0 is used as a sentinel meaning
4234 "unassigned handle" --- avoid it. Also coalesce two critical
4235 sections in this function into one.
4237 Severity : enhancement
4239 Description: allow liblustre clients to delegate truncate locking to OST
4240 Details : To avoid overhead of locking, liblustre client instructs OST to
4241 take extent lock in ost_punch() on client's behalf. New connection
4242 flag is added to handle backward compatibility.
4244 Severity : enhancement
4245 Bugzilla : 4928, 7341, 9758
4246 Description: allow number of OST service threads to be specified
4247 Details : a module parameter allows the number of OST service threads
4248 to be specified via "options ost ost_num_threads={N}" in the
4249 OSS's /etc/modules.conf or /etc/modprobe.conf.
4253 Bugzilla : 6146, 9635, 9895
4254 Description: servers crash with bad pointer in target_handle_connect()
4255 Details : In rare cases when a client is reconnecting it was possible that
4256 the connection request was the last reference for that export.
4257 We would temporarily drop the export reference and get a new
4258 one, but this may have been the last reference and the export
4259 was just destroyed. Get new reference before dropping old one.
4261 Severity : enhancement
4262 Frequency : if client is started with failover MDS
4264 Description: Allow multiple MDS hostnames in the mount command
4265 Details : Try to read the configuration from all specified MDS
4266 hostnames during a client mount in case the "primary"
4269 Severity : enhancement
4271 Description: Stop sending data to evicted clients as soon as possible.
4272 Details : Check if the client we are about to send or are sending data to
4273 was evicted already. (Check is done every second of waiting,
4274 for which l_wait_event interface was extended to allow checking
4275 of exit condition at specified intervals).
4278 Frequency : rare, normally only when NFS exporting is done from client
4280 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4282 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4285 Frequency : always, if srand() is called before liblustre initialization
4287 Description: Liblustre uses system PRNG disturbing its usage by user application
4288 Details : Introduce internal to lustre fast and high-quality PRNG for
4289 lustre usage and make liblustre and some other places in generic
4290 lustre code to use it.
4292 Severity : enhancement
4293 Bugzilla : 9477, 9557, 9870
4294 Description: Verify that the MDS configuration logs are updated when xml is
4295 Details : Check if the .xml configuration logs are newer than the config
4296 logs stored on the MDS and report an error if this is the case.
4297 Request --write-conf, or allow starting with --old_conf.
4299 Severity : enhancement
4301 Description: Handle symlinks in the path when checking if Lustre is mounted.
4302 Details : Resolve intermediate symlinks when checking if a client has
4303 mounted a filesystem to avoid duplicate client mounts.
4308 Description: lconf can hit an error exception but still return success.
4309 Details : The lconf command catches the Command error exception at the top
4310 level script context and will exit with the associated exit
4311 status, but doesn't ensure that this exit status is non-zero.
4316 Description: failure of ptlrpc thread startup can cause oops
4317 Details : Starting a ptlrpc service thread can fail if there are a large
4318 number of threads or the server memory is very fragmented.
4319 Handle this without oopsing.
4322 Frequency : always, only if liblustre and non-default acceptor port was used
4324 Description: liblustre cannot connect to servers with non-default acceptor port
4325 Details : tcpnal_set_default_params() was not called and was therefore
4326 ignoring the environment varaible TCPNAL_PORT, as well as other
4327 TCPNAL_ environment variables
4332 Description: two objects could be created on the same OST for a single file
4333 Details : If an OST is down, in some cases it was possible to create two
4334 objects on a single OST for a single file. No problems other
4335 than potential performance impact and spurious error messages.
4339 Bugzilla : 5681, 9562
4340 Description: Client may oops in ll_unhash_aliases
4341 Details : Client dcache may become inconsistent in race condition.
4342 In some cases "getcwd" can fail if the current directory is
4348 Description: Inode refcounting problems in NFS export code
4349 Details : link_raw functions used to call d_instantiate without obtaining
4350 extra inode reference first.
4354 Bugzilla : 9942, 9903
4355 Description: Referencing freed requests leading to crash, memleaks with NFS.
4356 Details : We used to require that call to ll_revalidate_it was always
4357 followed by ll_lookup_it. Also with revalidate_special() it is
4358 possible to call ll_revalidate_it() twice for the same dentry
4359 even if first occurence returned success. This fix changes semantic
4360 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4361 reference on a request referred from the intent.
4362 ll_intent_release() then releases such a request.
4365 Frequency : rare, normally benchmark loads only
4367 Description: unlinked inodes were kept in memory on the client
4368 Details : If a client is repeatedly creating and unlinking files it
4369 can accumulate a lot of stale inodes in the inode slab cache.
4370 If there is no other client load running this can cause the
4371 client node to run out of memory. Instead flush old inodes
4372 from client cache that have the same inode number as a new inode.
4375 Frequency : SLES9 2.6.5 kernel and long filenames only
4376 Bugzilla : 9969, 10379
4377 Description: utime reports stale NFS file handle
4378 Details : SLES9 uses out-of-dentry names in some cases, which confused
4379 the lustre dentry revalidation. Change it to always use the
4383 Frequency : rare, unless heavy write-truncate concurrency is continuous
4384 Bugzilla : 4180, 6984, 7171, 9963, 9331
4385 Description: OST becomes very slow and/or deadlocked during object unlink
4386 Details : filter_destroy() was holding onto the parent directory lock
4387 while truncating+unlinking objects. For very large objects this
4388 may block other threads for a long time and slow overall OST
4389 responsiveness. It may also be possible to get a lock ordering
4390 deadlock in this case, or run out of journal credits because of
4391 the combined truncate+unlink. Solution is to do object truncate
4392 first in one transaction without parent lock, and then do the
4393 final unlink in a new transaction with the parent lock. This
4394 reduces the lock hold time dramatically.
4397 Frequency : rare, 2.4 kernels only
4399 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4400 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4401 before unmounting their filesystem in order to avoid deadlock.
4402 The kernel_locked() function in 2.4 kernels only checks whether
4403 the kernel lock is held, not whether it is this process that is
4404 holding it as 2.6 kernels do.
4409 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4410 Details : The client ptlrpc code may be trying to reconnect to a down
4411 server before a previous connection attempt has timed out.
4412 Increase the reconnect interval to be longer than the connection
4413 timeout interval to avoid sending duplicate connections to
4417 Frequency : echo_client brw_test command
4419 Description: fix echo_client to work with OST preallocated code
4420 Details : OST preallocation code (5137) didn't take echo_client IO path
4421 into account: echo_client calls filter methods outside of any
4422 OST thread and, hence, there is no per-thread preallocated
4423 pages and buffers to use. Solution: hijack pga pages for IO. As
4424 a byproduct, this avoids unnecessary data copying.
4428 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4429 Description: Client can oops in mdc_commit_close() after open replay
4430 Details : It was possible for the MDS to return an open request with no
4431 transaction number in mds_finish_transno() if the client was
4432 evicted, but without actually returning an error. Clients
4433 would later try to replay that open and may trip an assertion
4434 Simplify the client close codepath, and always return an error
4435 from the MDS in case the open is not successful.
4438 Frequency : rare, 2.6 OSTs only
4440 Description: OST may deadlock under high load on fragmented files
4441 Details : If there was a heavy load and highly-fragmented OST filesystems
4442 it was possible to have all the OST threads deadlock waiting on
4443 allocation of biovecs, because the biovecs were not released
4444 until the entire RPC IO was completed. Instead, release biovecs
4445 as soon as they are complete to ensure forward IO progress.
4447 Severity : enhancement
4449 Description: Support for specifying external journal device at mount
4450 Details : If an OST or MDS device is formatted with an external journal
4451 device, this device major/minor is stored in the ext3 superblock
4452 and may not be valid for failover. Allow detecting and
4453 specifying the external journal at mount time.
4458 Description: Mounting an MDS with pending unlinked files may cause oops
4459 Details : target_finish_recovery() calls mds_postrecov() which returned
4460 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4461 considers this an error and immediately begins cleaning up the
4462 lov, just after starting the mds_lov process
4464 Severity : enhancement
4466 Description: Implement 'lfs df' to report actual free space on per-OST basis
4467 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4468 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4469 usage of objects; '-h' to report in human readable format.
4471 ------------------------------------------------------------------------------
4473 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4480 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4481 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4482 at startup during orphan unlinking. Instead, skip buddy update
4483 before setup, as it will be regenerated after recovery is complete.
4488 Description: If an OST is inactive, its locks might reference stale inodes.
4489 Details : lov_change_cbdata() must iterate over all namespaces, even if
4490 they are inactive to clear inode references from the lock.
4492 Severity : enhancement
4493 Frequency : occasional, if non-standard max_dirty_mb used
4495 Description: Client will block write RPCs if not enough grant
4496 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4497 then the client will block writes while waiting for another RPC
4498 to complete instead of consuming its dirty limit. With change
4499 we get improved performance when max_dirty_mb is small.
4501 Severity : enhancement
4502 Bugzilla : 3389, 6253
4503 Description: Add support for supplementary groups on the MDS.
4504 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4505 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4506 lookups for user supplementary groups into a cache.
4510 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4511 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4512 for liblustre O_EXCL is enforced by MDS.
4516 Description: readdir never returns NULL in liblustre.
4517 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4518 dentry in struct dirent.
4522 Frequency : liblustre only, depends on application IO pattern
4523 Description: liblustre clients evicted if not contacting servers
4524 Details : Don't put liblustre clients into the ping_evictor list, so
4525 they will not be evicted by the pinger ever.
4527 Severity : enhancement
4529 Description: Add ability to evict clients by NID from MDS.
4530 Details : By echoing "nid:$NID" string into
4531 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4532 $NID would be instantly evicted from this MDS and from all active
4533 OSTs connected to it.
4537 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4538 Details : lookup_it_finish() used to query file size from OSTs that was not
4543 Description: service threads change working directory to that of init
4544 Details : Starting lustre service threads may pin the working directory
4545 of the parent thread, making that filesystem busy. Threads
4546 now change to the working directory of init to avoid this.
4550 Frequency : during shutdown only
4551 Description: shutdown with a failed MDS or OST can cause unmount to hang
4552 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4557 Frequency : 2.6 only
4558 Description: chmod/chown may include an extra supplementary group
4559 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4560 supplementary group and if none is specified this is used.
4563 Bugzilla : 5479 (6816)
4565 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4566 Details : If lookup is in progress on a file that is unlinked we might try
4567 to revalidate the inode and fail in revalidate after lookup is
4568 complete and ll_file_open() enqueues the open again but
4569 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4572 Frequency : always, if lconf --abort_recovery used
4574 Description: lconf --abort_recovery fails with 'Operation not supported'
4575 Details : lconf was attempting to abort recovery on the MDT device and not
4578 ------------------------------------------------------------------------------
4580 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4585 Frequency : rare (only unsupported configurations with a node running as an
4587 Bugzilla : 6514, 5137
4588 Description: Mounting a Lustre file system on a node running as an OST could
4590 Details : OSTs now preallocates memory needed to write out data at
4591 startup, instead of when needed, to avoid having to
4592 allocate memory in possibly low memory situations.
4593 Specifically, if the file system is mounted on on OST,
4594 memory pressure could force it to try to write out data,
4595 which it needed to allocate memory to do. Due to the low
4596 memory, it would be unable to do so and the node would
4597 become unresponsive.
4599 Severity : enhancement
4601 Description: Addition of lconf --service command line option
4602 Details : lconf now accepts a '--service <arg>' option, which is
4603 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4605 Severity : enhancement
4607 Description: Failover mode is now the default for OSTs.
4608 Details : By default, OSTs will now run in failover mode. To return to
4609 the old behaviour, add '--failout' to the lmc line for OSTs.
4611 Severity : enhancement
4613 Description: Health checks are now provided for MDS and OSTs
4614 Details : Additional detailed health check information on MSD and OSTs
4615 is now provided through the procfs health_check value.
4618 Frequency : occasional, depends on IO load
4620 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4621 after numerous create/delete cycles
4622 Details : The ext3 inode allocation policy would not allocate new inodes
4623 very well on the OSTs because there are no new directories
4624 being created. Instead we look for groups with free space if
4625 the parent directories are nearly full.
4630 Description: Network or server problems during mount may cause partially
4631 mounted clients instead of returning an error.
4632 Details : The config llog parsing code may overwrite the error return
4633 code during mount error handling, returning success instead
4639 Description: MDS can fail to allocate large reply buffers
4640 Details : After long uptimes the MDS can fail to allocate large reply
4641 buffers (e.g. zconf client mount config records) due to memory
4642 fragmentation or consumption by the buffer cache. Preallocate
4643 some large reply buffers so that these replies can be sent even
4644 under memory pressure.
4648 Frequency : rare (liblustre)
4649 Description: fsx running with liblustre complained that using truncate() to
4650 extend the file doesn't work. This patch corrects that issue.
4651 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4652 ATTR_SIZE and lsm use in llu_setattr_raw.
4656 Frequency : rare, only 2.6 kernels
4657 Description: Unusual file access patterns on the MDS may result in inode
4658 data being lost in very rare circumstances.
4659 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4660 code in the 2.6 kernel caused the inode and/or EA data not to
4661 be read from disk, causing single-file corruption.
4665 Frequency : rare, only 2.6 filesystems using extents
4666 Description: Heavy concurrent write and delete load may cause data corruption.
4667 Details : It was possible under high-load situations to have an extent
4668 metadata block in the block device cache from a just-unlinked
4669 file overwrite a newly-allocated data block. We now unmap any
4670 metadata buffers that alias just-allocated data blocks.
4674 Frequency : filesystems with default stripe_count larger than 77
4675 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4676 Details : lconf specifies an inode size of 4096 bytes when the default
4677 stripe_count is larger than 77. This conflicts with the default
4678 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4679 this case to avoid pinning too much memory for large EAs.
4681 ------------------------------------------------------------------------------
4683 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4688 Frequency : rare (extremely heavy IO load with hundreds of clients)
4690 Description: Client is evicted, gets IO error writing to file
4691 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4692 caused clients to be evicted for AST timeouts. The fixes in
4693 bug 5192 mean we no longer need to have such short AST timeouts
4694 so ldlm_timeout has been increased.
4697 Frequency : occasional during --force or --failover shutdown under load
4698 Bugzilla : 5949, 4834
4699 Description: Server oops/LBUG if stopped with --force or --failover under load
4700 Details : a collection of import/export refcount and cleanup ordering
4701 issues fixed for safer force cleanup
4704 Frequency : only filesystems larger than 120 OSTs
4705 Bugzilla : 5990, 6223
4706 Description: lfs getstripe would oops on a very large filesystem
4707 Details : lov_getconfig used kfree on vmalloc'd memory
4710 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4711 Bugzilla : 6242, 6243
4712 Description: reading from files that had been truncated to a non-zero size
4713 but never opened returned no data
4714 Details : ll_file_read() reads zeros from no-object files to EOF
4719 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4720 Details : There's little harm in aborting MDS/OSS recovery and letting it
4721 try again, so I removed the LASSERT and return an error instead.
4723 Severity : enhancement
4725 Description: New debugging infrastructure for tracking down data corruption
4726 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4727 (b) cover more of the client-side code path, and (c) try to narrow
4728 down where problems occurred
4732 Bugzilla : 3819, 4364, 4397, 6313
4733 Description: Racing close and eviction MDS could cause assertion in mds_close
4734 Details : It was possible to get multiple mfd references during close and
4735 client eviction, leading to one thread referencing a freed mfd.
4737 Severity: : enhancement
4738 Bugzilla : 3262, 6359
4739 Description: Attempts to reconnect to servers are now more aggressive.
4740 Details : This builds on the enhanced upcall-less recovery that was added
4741 in 1.4.2. When trying to reconnect to servers, clients will
4742 now try each server in the failover group every 10 seconds. By
4743 default, clients would previously try one server every 25 seconds.
4748 Description: After recovery, certain operations trigger a failed
4749 assertion on a client.
4750 Details : Failing over an mds, using lconf -d --failover, while a
4751 client was doing a readdir() call would cause the client to
4752 LBUG after recovery completed and the readdir() was resent.
4754 Severity : enhancement
4756 Description: Default groups are now added by lconf
4757 Details : You can now run lconf --group <servicename> without having to
4758 manually add groups with lmc.
4761 Frequency : occasional
4763 Description: Nodes with an elan id of 0 trigger a failed assertion
4766 Frequency : always when accessing e.g. tty/console device nodes
4768 Description: tty and some other devices nodes cannot be used on lustre
4769 Details : file's private_data field is used by device data and lustre
4770 values in there got lost. New field was added to struct file to
4771 store fs-specific private data.
4774 Frequency : when exporting Lustre via NFS
4776 Description: NFSD failed occasionally when looking up a path component
4777 Details : NFSD is looking up ".." which was broken in ext3 directories
4778 that had grown large enough to become hashed.
4781 Frequency : Clusters with multiple interfaces not on the same subnet
4783 Description: Nodes will repeatedly try to reconnect to an interface which it
4784 cannot reach and report an error to the log.
4785 Details : Extra peer list entries will be created by lconf with some peers
4786 unreachable. lconf now validates the peer before adding it.
4789 Frequency : Only if a default stripe is set on the filesystem root.
4791 Description: Setting a default stripe on the filesystem root prevented the
4792 filesystem from being remounted.
4793 Details : The client was sending extra request flags in the root getattr
4794 request and did not allocate a reply buffer for the dir EA.
4797 Frequency : occasional, higher if lots of files are accessed by one client
4798 Bugzilla : 6159, 6097
4799 Description: Client trips assertion regarding lsm mismatch/magic
4800 Details : While revalidating inodes the VFS looks up inodes with ifind()
4801 and in rare cases can find an inode that is being freed.
4802 The ll_test_inode() code will free the lsm during ifind()
4803 when it finds an existing inode and then the VFS later attaches
4804 this free lsm to a new inode.
4808 Bugzilla : 6422, 7030
4809 Description: MDS deadlock between mkdir and client eviction
4810 Details : Creating a new file via mkdir or mknod (starting a transaction
4811 and getting the ns lock) can deadlock with client eviction
4812 (gets ns lock and trying to finish a synchronous transaction).
4815 Frequency : occasional
4816 Description: While starting a server, the fsfilt_ext3 module could not be
4818 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4819 kernels. Previously, lconf would still use the ext3 name
4820 when trying to load modules. Now, it will correctly use
4821 ext3 on 2.4 and ldiskfs on 2.6.
4823 Severity : enhancement
4824 Description: The default stripe count has been changed to 1
4825 Details : The interpretation of the default stripe count (0, to lfs
4826 or lmc) has been changed to mean striping across a single
4827 OST, rather than all available. For general usage we have
4828 found a stripe count of 1 or 2 works best.
4830 Severity : enhancement
4831 Description: Add support for compiling against Cray portals.
4832 Details : Conditional compiling for some areas that are different
4836 Frequency : occasional
4837 Bugzilla : 6409, 6834
4838 Description: Creating files with an explicit stripe count may lead to
4839 a failed assertion on the MDS
4840 Details : If some OSTs are full or unavailable, creating files may
4841 trigger a failed assertion on the MDS. Now, Lustre will
4842 try to use other servers or return an error to the
4846 Frequency : occasional
4848 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4849 caused lock timeout during readahead (since 1.4.2).
4850 Details : Processes doing readahead might match a lock that hasn't been
4851 granted yet if there are overlapping and conflicting lock
4852 requests. The readahead process waits on ungranted lock
4853 (original lock is CBPENDING), while OST waits for that process
4854 to cancel CBPENDING read lock and eventually evicts client.
4856 Severity : enhancement
4858 Description: Initial enabling of flock support for clients
4859 Details : Implements fcntl advisory locking and file status functions.
4860 This feature is provided as an optional mount flag (default
4861 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4862 locking are implemented yet, and those that are are not guaranteed
4863 to be completely correct in production environments.
4864 mount -t lustre -o [flock|noflock] ...
4867 Frequency : occasional
4869 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4870 assertion in the ext3 JBD (journaling) layer.
4871 Details : The b_committed_data struct is protected by the big kernel lock
4872 in 2.4 kernels, serializing journal_commit_transaction() and
4873 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4874 access to this struct. In 2.6 kernels there is finer grained
4875 locking to improve SMP performance of the JBD layer.
4879 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4884 Description: merge and cleanup kernel patches.
4887 -----------------------------------------------------------------------------
4889 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4891 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4892 versions of Lustre. Please update all servers and clients to
4893 version 1.4.2 or later at the same time. You must also run
4894 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4895 to update the configuration logs.
4897 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4898 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4899 - swab llog records of type '0' so we get proper header size/idx (5861)
4900 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4901 - fix rename of one directory over another leaking an inode (5953)
4902 - avoid SetPageDirty on 2.6 (5981)
4903 - don't re-add just-being-destroyed locks to the waiting list (5653)
4904 - when creating new directories, inherit the parent's custom
4905 striping settings if present parent (3048)
4906 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4907 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4908 - don't hold client locks on temporary worklist from l_lru (5666)
4909 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4910 - automatically evict dead clients (5921)
4911 - Update file size properly in create+truncate+fstat case (6196)
4912 - Do not unhash mountpoint dentries, do not allow removal of
4914 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4915 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4916 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4917 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4918 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4919 - Correctly handle reads of files with no objects (6243)
4920 - lctl recover will also mark a device active if deactivate used (5933)
4922 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4923 - allow --write-conf on an MDS with different nettype than client (5619)
4924 - don't write config llogs to MDS for mounts not from that MDS (5617)
4925 - lconf should create multiple TCP connections from a client (5201)
4926 - init scripts are now turned off by default; run chkconfig --on
4927 lustre and chkconfig --on lustrefs to use them
4928 - upcalls are no longer needed for clients to recover to failover
4930 - add --abort-recovery option to lconf to abort recovery on device
4932 - add support for an arbitrary number of OSTs (3026)
4933 - Quota support protocol changes.
4934 - forward compatibility changes to wire structs (6007)
4935 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4936 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4937 - improved "lustre status" script
4938 - initialize blocksize for non-regular files (6062)
4939 - added --disable-server and --disable-client configure options (5782)
4940 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4941 - Vanilla 2.4.29 support
4942 - increase maximum number of obd devices to 520 (6242)
4943 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4944 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4945 SLES 9 2.6.5 kernels
4946 - limit stripes per file to 160 (the maximum EA size) (6093)
4948 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4951 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4952 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4953 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4954 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4955 - small ext3 extents cleanups and fixes (5733)
4956 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4957 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4958 - avoid needless client->OST connect, fix handle mismatch (5317)
4959 - fix DLM error path that led to out-of-sync client, long delays (5779)
4960 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4961 - fix several locking issues related to i_size (5492,5624,5654,5672)
4962 - don't move pending lock onto export if it is already evicted (5683)
4963 - fix kernel oops when creating .foo in unlinked directory (5548)
4964 - fix deadlock in obdfilter statistics vs. object create (5811)
4965 - use time_{before,after} to avoid timer jiffies wrap (5882)
4966 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4967 - Do not leak request if server was not able to process it (5154)
4968 - If mds_open unable to find parent dir, make that negative lookup(5154)
4969 - don't create new directories with extent-mapping (5909, 5936)
4971 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4972 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4973 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4974 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4975 - support for RHEL4 kernel on i686 (5773)
4976 - provide error messages when incompatible logs are encountered (5898)
4978 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4979 * version 1.4.0.10 (1.4.1 release candidate 1)
4981 - don't keep a lock reference when lock is not granted (4238)
4982 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4983 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4984 - drop import inflight refcount on signal_completed_replay error (5255)
4985 - unlock page after async write error during send (3677)
4986 - handle missing objects in filter_preprw_read properly (5265)
4987 - no transno return for symlink open, don't save no-trasno open (3440)
4988 - don't try to complete elan receive that already failed (4012)
4989 - free RPC server reply state on error (5406)
4990 - clean up thread from ptlrpc_start_thread() on error (5160)
4991 - readahead could read extra page into cache that wasn't ejected (5388)
4992 - prevent races in class_attach/setup/cleanup/detach (5260)
4993 - don't dereference de->d_inode after l_dput of de (5458)
4994 - use "int" for stripe value returned from lock_to_stripe (5544)
4995 - mballoc allocation and error-checking fixes in 2.6 (5504)
4996 - block device patches to fix I/O request sizes in 2.6 (5482)
4997 - look up hostnames for IB nals (5602)
4998 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4999 - don't start multiple acceptors for the same port (5277)
5000 - fix incorrect LASSERT in mds_getattr_name (5635)
5001 - export a proc file for general "ping" checking (5628)
5002 - fix "lfs check" to not block when the MDS is down (5628)
5004 - service request history (4965)
5005 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5006 - create an "evict_client" /proc entry on OSTs, like the MDS has
5007 - fix mount usage message, return errors per mount(8) (5168)
5008 - change grep [] to grep "[]" in tests so they work in more UMLs
5009 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5010 - remove ancient LOV_MAGIC_V0 EA support (5047)
5011 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5012 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5013 - allow readahead allocations to fail when low on memory (5383)
5014 - mmap locking landed again, after considerable improvement (2828)
5015 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5017 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5020 - send OST transaction number in read/write reply to free req (4966)
5021 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5022 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5023 - the watchdog thread now runs as interruptible (5246)
5024 - flock/lockf fixes (but it's still disabled, pending 5135)
5025 - don't use EXT3 constants in llite code (5094)
5026 - memory shortage at startup could cause assertion (5176)
5028 - reorganization of lov code
5029 - single portals codebase
5031 - add extents/mballoc support (5025)
5032 - direct I/O reads in the obdfilter (4048)
5033 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5035 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5038 - send OST transaction number in read/write reply to free req (4966)
5039 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5040 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5041 - the watchdog thread now runs as interruptible (5246)
5042 - handle missing objects in filter_preprw_read properly (5265)
5043 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5044 - drop import inflight refcount on signal_completed_replay error (5255)
5045 - unlock page after async write error during send (3677)
5046 - return original error code on reconstructed replies (3761)
5047 - no transno return for symlink open, don't save no-trasno open (3440)
5049 - add pid to ldlm debugging output (4922)
5050 - bump the watchdog timeouts -- we can't handle 30sec yet
5051 - extra debugging for orphan dentry/inode bug (5259)
5053 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5056 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5057 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5058 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5059 - fix NULL dereference in /proc/sys/portals/routes (4827)
5060 - allow failed mdc_close() operations to be interrupted (4561)
5061 - stop precreate on OST before MDS would time out on it (4778)
5062 - don't send partial-page writes before EOF from client (4410)
5063 - discard client grant for sub-page writes on large-page clients (4520)
5064 - don't free dentries not owned by NFS code, check generation (4806)
5065 - fix lsm leak if mds_create_objects() fails (4801)
5066 - limit debug_daemon file size, always print CERROR messages (4789)
5067 - use transno after validating reply (3892)
5068 - process timed out requests if import state changes (3754)
5069 - update mtime on OST during writes, return in glimpse (4829)
5070 - add mkfsoptions to LDAP (4679)
5071 - use ->max_readahead method instead of zapping global ra (5039)
5072 - don't interrupt __l_wait_event() during strace
5074 - add software watchdogs to catch hung threads quickly (4941)
5075 - make lustrefs init script start after nfs is mounted
5076 - fix CWARN/ERROR duplication (4930)
5077 - return async write errors to application if possible (2248)
5078 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5079 - print NAL number in %x format (4645)
5080 - update barely-supported suse-2.4.21-171 series (4842)
5081 - support for sles 9 %post scripts
5082 - support for building 2.6 kernel-source packages
5083 - support for sles km_* packages
5085 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5088 - ignore -ENOENT errors in osc_destroy (3639)
5089 - notify osc create thread that OSC is being cleaned up (4600)
5090 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5091 - reconstruct ost_handle() like mds_handle() (4657)
5092 - create a new thread to do import eviction to avoid deadlock (3969)
5093 - let lconf resolve symlinked-to devices (4629)
5094 - don't unlink "objects" from directory with default EA (4554)
5095 - hold socknal file ref over connect in case target is down (4394)
5096 - allow more than 32000 subdirectories in a single directory (3244)
5097 - fix blocks count for O_DIRECT writes (3751)
5098 - OST returns ENOSPC from object create when no space left (4539)
5099 - don't send truncate RPC if file size isn't changing (4410)
5100 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5101 - bind to privileged port in socknal and tcpnal (3689)
5103 - rate limit CERROR/CWARN console message to avoid overload (4519)
5104 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5105 - basic mmap support (3918)
5106 - kernel patch series update from b1_4 (4711)
5108 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5111 - avoid crash during MDS cleanup with OST shut down (2775)
5112 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5113 - don't use bad inodes on the MDS (3744)
5114 - dynamic object preallocation to improve recovery speed (4236)
5115 - don't hold spinlock over lock dumping or change debug flags (4401)
5116 - don't zero obd_dev when it is force cleaned (3651)
5117 - print grants to console if they go negative (4431)
5118 - "lctl deactivate" will stop automatic recovery attempts (3406)
5119 - look for existing locks in ldlm_handle_enqueue() (3764)
5120 - don't resolve lock handle twice in recovery avoiding race (4401)
5121 - revalidate should check working dir is a directory (4134)
5123 - don't always mark "slow" obdfilter messages as errors (4418)
5125 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5128 - don't close LustreDB during write_conf until it is done (3860)
5129 - fix typo in lconf for_each_profile (3821)
5130 - allow dumping logs from multiple threads at one time (3820)
5131 - don't allow multiple threads in OSC recovery (3812)
5132 - fix debug_size parameters (3864)
5133 - fix mds_postrecov to initialize import for llog ctxt (3121)
5134 - replace config semaphore with spinlock (3306)
5135 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5136 - don't allow enqueue to complete on a destroyed export (3822)
5137 - down write_lock before checking llog header bitmap (3825)
5138 - recover from lock replay timeout (3764)
5139 - up llog sem before sending rpc (3652)
5140 - reduce ns lock hold times when setting kms (3267)
5141 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5142 - fix NULL deref and obd_dev leak on setup error (3312)
5143 - replace some LBUG about llog ops with error handling (3841)
5144 - don't match INVALID dentries from d_lookup and spin (3784)
5145 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5146 - fix invalid assertion in ptlrpc_set_wait (3880)
5148 - add libwrap support for the TCP acceptor (3996)
5149 - add /proc/sys/portals/routes for non-root route listing (3994)
5150 - allow setting MDS UUID in .xml (2580)
5151 - print the stack of a process that LBUGs (4228)
5153 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5156 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5157 - make sure to unset replay flag from failed open requests (3440)
5158 - if default stripe count is 0, use OST count for inode size (3636)
5159 - update parent mtime/ctime on client for create/unlink (2611)
5160 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5161 - free recovery state on server during a forced cleanup (3571)
5162 - unregister_reply for resent reqs (3063)
5163 - loop back devices mounting and status check on 2.6 (3563)
5164 - fix resource-creation race that can provoke i_size == 0 (3513)
5165 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5166 - more debugging for page-accounting assertion (3746)
5167 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5168 - avoid deadlock after precreation failure (3758)
5169 - fix race and lock order deadlock in orphan handling (3450, 3750)
5170 - add validity checks when grabbing inodes from l_ast_data (3599)
5172 - add /proc/.../recovery_status to obdfilter (3428)
5173 - lightweight CDEBUG infrastructure, debug daemon (3668)
5174 - change default OSC RPC parameters to be better on small clusters
5175 - turn off OST read cache for files smaller than 32MB
5176 - install man pages and include them in rpms (3100)
5177 - add new init script for (un)mounting lustre filesystems (2593)
5178 - run chkconfig in %post for init scripts (3701)
5179 - drop scimac NAL (unmaintained)
5181 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5184 - clean kiobufs before and after use (3485)
5185 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5186 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5187 - change init script to fail more clearly if not run as root (1528)
5188 - allow clients to reconnect during replay (1742)
5189 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5190 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5191 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5192 - fix cond_resched (3554)
5193 - only evict unfinished clients after recovery (3515)
5194 - allow bulk resend, prevent data loss (3570)
5195 - dynamic ptlrpc request buffer allocation (2102)
5196 - don't allow unlinking open directory if it isn't empty (2904)
5197 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5198 - remove extraneous obd dereference causing LASSERT failure (3334)
5199 - don't use get_cycles() when creating temp. files on the mds (3156)
5200 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5201 - handle EEXIST for set-stripe, set proper directory name (3336)
5203 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5204 - fix ksocknal_fmb_callback() error messages (2918)
5206 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5209 - don't copy lvb into (possibly NULL) reply on error (2983)
5210 - don't deref dentry after dput, don't free lvb on error (2922)
5211 - use the kms to determine writeback rpc length (2947)
5212 - increment oti_logcookies when osc is inactive (2948)
5213 - update client's i_blocks count via lvb messages (2543)
5214 - handle intent open/close of special files properly (1557)
5215 - mount MDS with errors=remount-ro, like obdfilter (2009)
5216 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5217 - don't use cancelling-locks' kms values (2947)
5218 - use highest lock extent for kms, not last one (2925)
5219 - don't dereference ERR_PTR() dentry in error handling path (3107)
5220 - fix thread race in portals_debug_dumplog() (3122)
5221 - create lprocfs device entries at setup instead of at attach (1519)
5222 - common AST error handler, don't evict client on completion race (3145)
5223 - zero nameidata in detach_mnt in 2.6 (3118)
5224 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5225 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5226 - initialize RPC timeout timer earlier for 2.6 (3219)
5227 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5228 - print nal/nid for unknown nid (3258)
5229 - additional checks for oscc recovery before doing precreate (3284)
5230 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5231 - don't crash in mdc_close for bad permissions on open (3285)
5232 - zero i_rdev for non-device files (3147)
5233 - clear page->private before handing to FS, better assertion (3119)
5234 - tune the read pipeline (3236)
5235 - fix incorrect decref of invalidated dentry (2350)
5236 - provide read-ahead stats and refine rpc in flight stats (3328)
5237 - don't hold journal transaction open across create RPC (3313)
5238 - update atime on MDS at close time (3265)
5239 - close LDAP connection when recovering to avoid server load (3315)
5240 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5241 - don't leak open file on MDS after open resend (3325)
5242 - serialize filter_precreate and filter_destroy_precreated (3329)
5243 - loop device shouldn't call sync_dev() for nul device (3092)
5244 - clear page cache after eviction (2766)
5245 - resynchronize MDS->OST in background (2824)
5246 - refuse to mount the same filesystem twice on same mountpoint (3394)
5247 - allow llmount to create routes for mounting behind routers (3320)
5248 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5249 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5250 - fix rare problem with rename on htree directories (3417)
5252 - allow default OST striping configuration per directory (1414)
5253 - fix compilation for qswnal for 2.6 kernels (3125)
5254 - increase maximum number of MDS request buffers for large systems
5255 - change liblustreapi to be useful for external progs like lfsck (3098)
5256 - increase local configuration timeout for slow disks (3353)
5257 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5259 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5262 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5263 - don't overwrite extent policy data in reply if lock was blocked (2901)
5264 - drop filter export grants atomically with removal from device (2663)
5265 - del obd_self_export from work_list in class_disconnect_exports (2908)
5266 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5267 - swab reply message in mdc_close, other PPC fixes (2464)
5268 - fix destroying of named logs (2325)
5269 - overwrite old logs when running lconf --write_conf (2264)
5270 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5271 - fix race in target_handle_connect (2898)
5272 - mds_reint_create() should take same inode create lock (2926)
5273 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5274 - don't close files for self_export to avoid uninitialized obd (2936)
5275 - allow MDS with the same name as client node (2939)
5276 - hold dentry reference for closed log files for unlink (2325)
5277 - reserve space for all logs during transactions (2059)
5278 - don't evict page beyond end of stripe extent (2925)
5279 - don't oops on a deleted current working directory (2399)
5280 - handle hard links to targets without a parent properly (2517)
5281 - don't dereference NULL lock when racing during eviction (2867)
5282 - don't grow lock extents when lots of conflicting locks (2919)
5284 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5287 - account for cache space usage on clients to avoid data loss (974)
5288 - lfsck support in lustre kernel code (2349)
5289 - reduce journal credits needed for BRW writes (2370)
5290 - orphan handling to avoid losing space on client/server crashes
5291 - ptlrpcd can be blocked, stopping ALL progress (2477)
5292 - use lock value blocks to assist in proper KMS, faster stat (1021)
5293 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5294 - recovery for initial connections (2355)
5295 - fixes for mds_cleanup_orphans (1934)
5296 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5297 - block all file creations until orphan recovery completes (1901)
5298 - client remove rq_connection from request struct (2423)
5299 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5300 - recovery timer race (2670)
5301 - mdc_close recovey bug (2532)
5302 - ptlrpc cleanup bug (2710)
5303 - mds timeout on local locks (2588)
5304 - namespace lock held during RPCs (2431)
5305 - handle interrupted sync write properly (2503)
5306 - don't try to handle a message that hasn't been replied to (2699)
5307 - client assert failure during cleanup after abort recovery (2701)
5308 - leak mdc device after failed mount (2712)
5309 - ptlrpc_check_set allows timedout requests to complete (2714)
5310 - wait for inflight reqs when ptlrpcd finishes (2710)
5311 - make sure unregistered services are removed from the srv_list
5312 - reset bulk XID's when resending them (caught by 1138 test)
5313 - unregister_bulk after timeout
5314 - fix lconf error (2694)
5315 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5316 - readahead locks pages, leaves pending causing memory pressure (2673)
5317 - increase OST request buffers to 4096 on large machines (2729)
5318 - fix up permission of existing directories in simple_mkdir (2661)
5319 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5320 - don't assign transno to errored transactions (2742)
5321 - don't delete objects on OST if given a bogus objid from MDS (2751)
5322 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5323 - if rq_no_resend, then timeout request after recovery (2432)
5324 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5325 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5326 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5327 - MDS recovery shouldn't delete objects that it has given out (2730)
5328 - if enqueue arrives after completion, don't clobber LVB (2819)
5329 - don't unlock pages twice when trigger_group_io returns error (2814)
5330 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5331 - don't write pages to disk if there was an error (1450)
5332 - don't ping imports that have recovery disabled (2676)
5333 - take buffered bytes into account when balancing socknal conn (2817)
5334 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5335 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5336 - remove little-endian swabbing of llog records (1987)
5337 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5338 - retry reposting request buffers if they fail (1191)
5339 - grow extent at grant time to avoid granting a revoked lock (2809)
5340 - lock revoke doesn't evict page if covered by a second lock (2765)
5341 - disable VM readahead to avoid reading outside lock extents (2805)
5343 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5344 - updated kernel patches for hp-2.4.20 kernel (2681)
5346 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5349 - fix truncated write corruption (2366)
5350 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5352 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5353 - ensure len(uuid) < 37 in lmc (1171)
5354 - fix ia64 OOPS in llog_test (2255)
5355 - zero end of page at obdfilter for partial page writes (2648)
5356 - don't leave stale dentries around after renames (bug 2428)
5357 - fix timeouts when evicting a client with a single lock held (2642)
5358 - set deadline for the initial HELLO message to drain (2634)
5359 - print out dotted-quad IP addresses in the socknal (2302)
5361 - additional debugging for MDS client eviction problem (2443)
5362 - fix mkfsoptions support for osts (2603, 2604)
5364 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5367 - add series for the vanilla 2.6.0 kernel
5368 - add series for the vanilla 2.4.24 kernel
5369 - add series for a cray x86/64 UL kernel drop
5370 - fix xattr patches for the vanilla 2.4.19 series
5372 - generate true UUIDs in lmc (1171)
5373 - have portals stack dumping break in UML (2466)
5374 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5375 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5376 - wake callback waiting threads on client eviction (2460)
5377 - Add --ptldebug and --subsystem to lmc (1719)
5378 - update assertion to allow safe interrupt allocation
5379 - set rq_no_resend for cancel requests (2432)
5380 - recalculate ptlrpcd timeout after resend (2494)
5381 - call vfs_rmdir when removing pending directories (2368)
5382 - fix renaming a file to itself (2429)
5383 - lmc creates a default one-stripe lov (2454)
5384 - expand procfs space to handle large clusters (2326)
5385 - increase UML stack to avoid overflow
5386 - update lconf's list of debug and subsystem masks
5387 - fix lfs find --obd (2510)
5388 - /proc tunable for disabling filter read caching (2591)
5389 - stop rpm packages from altering slapd.conf (2301)
5390 - disable nagle in the socknal under 0conf (2578)
5391 - choose mds inode size based on stripe count (2572)
5392 - fix kernel-source rpm problems (2516)
5394 - add --disable-doc to avoid pdf generation (2421)
5395 - update documentation, tests, type-os, comments
5396 - avoid format warnings on ia64
5397 - remove the TOE NAL
5398 - tiny code cleanups by removing unused fields
5400 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5403 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5404 - remove the most blatant lies from BUILDING file (2371)
5405 - change default debug level to reasonable production setting
5406 - reduce client side cache size to reduce cache flush time
5407 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5408 - make TCP zerocopy and pinger support enabled by default (2476)
5409 - sync writes completed after process exits caused crashes (2319)
5410 - maintain correct mount count on the MDS (2356)
5411 - backout 1557, because 2316 wasn't really fixed
5412 - better file I/O statistics gathering in /proc
5413 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5414 - another kernel patch to fix zero-copy TCP function export
5415 - don't take duplicate lock when processing re-sent getattr (2420)
5416 - lctl uses obd_self_export instead of creating new conn (2353)
5417 - MDS/OST recovery case which requires object creation asserted (2425)
5418 - move lfs from /usr/sbin to /usr/bin in packages
5419 - fix race between mds_client_add and mds_client_free (2417)
5420 - use kmalloc instead of slabs in portals (2430)
5421 - don't create duplicate records when a failover MDS is present (2442)
5422 - remove unnecessary mount age check (2332)
5423 - don't remove directory inodes from locks prematurely (2451)
5424 - don't break if MDS service name is the same as hostname (2103)
5425 - fix races in client write RPC generation when cache full (2482)
5427 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5430 - remove now-unused request->rq_obd (278)
5431 - if an allocation fails, print out how much memory we've used (1933)
5432 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5433 - add the "configurable stack size" patch to most series files (1256)
5434 - ability to write large log records, for 100+ OST configs (2306)
5435 - fix NULL deref when filter_prep fails (2314)
5436 - fix operator precedence error in filter_sync
5437 - dynamic allocation of socknal TX descriptors (2315)
5438 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5439 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5440 - ensure that CWARN messages in Portals always get to the syslog
5441 - __init/__exit are not for prototype decls (ldlm_init/exit)
5442 - x86-64 compile warning fixes
5443 - fix gateway LMC keyword conflict (2318)
5444 - fix MDS lock inversions in getattr/reint paths (1844)
5445 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5446 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5447 - DLM race condition prevented some lock evictions (2328)
5448 - ENOMEM detection and retry on socknal sends (2230)
5449 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5450 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5451 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5452 - better error messages when a client is rejected during recovery (1505)
5453 - avoid cancelling locks which were never granted, after failure (2330)
5454 - fix i_sem/journal inversion in mds_client_add (2333)
5455 - fix truncate/getattr lock cycle deadlock (2334)
5456 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5457 - fix two rare exit paths which could leak an l_lock() ref (2321)
5458 - fencepost error in MDS/OST orphan recovery (2226)
5459 - make log record alignment 8 bytes (1988)
5460 - lstripe now fails when requested offset > ost_count (2237)
5461 - ensure that all kernel series have a complete list.h (1607)
5462 - fix crashes in special-file operations (2316)
5463 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5465 - allow configurable automake binary, for testing new versions
5466 - small update to the lfs documentation
5468 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5470 * fix negative export reference count in fsfilt_sync (2312)
5472 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5473 * release candidate 0.9.1
5475 - orphans are moved into the PENDING directory for possible recovery
5476 - replayed opens now open by fid for orphan/rename safety (1042)
5477 - last close of an orphan inode generates a transno (683)
5478 - chdir() and mount() now pin the directory entry (1020)
5479 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5480 - discard very old requests without processing them (1502)
5481 - remove some common, well-understood CERRORs (1505)
5482 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5483 - clear "grant" flags in OST replies until OST grant code lands (1644)
5484 - fix read performance by not clobbering i_blksize on client (1598)
5485 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5486 - make LRU size a /proc tunable, clears locks when reduced (707)
5487 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5488 - server locks take references on exports now (1558)
5489 - build fixes for 2.4.20-rh trees (1663)
5490 - return an error from lov_create if all OSCs are inactive (1751)
5491 - fix import levels when a reconnect happens without a timeout (1597)
5492 - exit early from mds_open if we get a lookup error (1749)
5493 - partial page read at EOF wouldn't wait for disk before sending (1642)
5494 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5495 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5496 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5497 - write/truncate lock inversion (1639)
5498 - Don't auto-load obdclass, portals modules during cleanup (1495)
5499 - fix timestamps from jumping to "now" (1763)
5500 - extra journal assertions (1648)
5501 - add an extra multiunlink test (1771)
5502 - fix read_record/write_record API (1776)
5503 - fix leak of offset_extent, possible incorrect i_size later (1772)
5504 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5505 - Debugging for the kqswnal_get_idle_tx problems (1820)
5506 - Allow recovery to be attempted multiple times (1536)
5507 - Write out MDS last_rcvd file after it is first created (1600)
5508 - Fix tx_descriptor leak in failed transmit situations (1827)
5509 - ext3 journaling fixes for assertion failure after IO error (1871)
5510 - class_export_put() on freed export after completion AST error (1896)
5511 - Fix revalidate looping in VFS (1322)
5512 - Don't access a freed export during MDS_REINT timeout (1521)
5513 - Add open-unlink recovery support on the MDS (1673,1764)
5514 - Return an error if no MDS data was read from last_rcvd (1946)
5515 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5516 - Don't setup a disk device that doesn't match exported UUID (317)
5517 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5518 - avoid committing NULL handle in force close
5519 - local.sh is now a one-stripe LOV configuration
5520 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5521 - fix ext3 htree duplicate directory entry corruption (1516)
5522 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5523 - update to new LOV EA format (2097)
5524 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5526 2003-06-15 Phil Schwan <phil@clusterfs.com>
5529 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5530 - per-import/export recovery handling (958, 931, 959)
5531 - multiple last-rcvd slots, for serving multiple FSes (949)
5532 - connections are again shared between multiple imp/exports (963, 964)
5533 - "umount -f" would hang if any requests needed to be sent (393, 978)
5534 - avoid pinning large req buffer by copying for queued messages (989)
5535 - add "uuid" to "lctl device" command to help upcalls (991)
5536 - "open" RPCs with transnos would confuse recovery counters (1037)
5537 - do proper endian conversion of all wire messages (288, 340, 891)
5538 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5539 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5540 - fix ptlrpc_connection leak in target_handle_connect (1174)
5541 - fix import refcounting bug in OST and MDS cleanup (1134)
5542 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5543 - fix very unlikely obd_types race condition (501)
5544 - remove osc_open hack for echo_client (1187)
5545 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5546 - a failure in read_inode2 leads to deadlock (1139)
5547 - cancel ack-locks as soon as transaction is committed (1072)
5548 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5549 - make sure to commitrw after any preprw to avoid deadlock (1162)
5550 - failing to execute a file in a lustre FS would lock inode (1203)
5551 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5552 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5553 - obd_uuid2tgt would walk off the end of the list (1255)
5554 - on IA64 the osc would give portals incorrect bulk size (1258)
5555 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5556 - fix lock inversion caused by new llite matching code (1282)
5557 - limit the number of dirty pages on a client to 10MB (1286)
5558 - timed out locks were not being corrected cancelled (1289)
5559 - fix O_DIRECT above 4GB on IA-32 (1292)
5560 * major user-visible changes
5561 - fail out/fail over policy now controlled by the upcall (993)
5563 - add OBD_PING to check server availability and failure (954)
5564 - lustre messages are now sent in sending host order (288, 340, 891)
5565 - add eadatalen to MDS getattr reply (340)
5566 - OST read replies may contain second buffer, with per-page status (593)
5568 2003-03-11 Phil Schwan <phil@clusterfs.com>
5571 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5572 - failed open()s could cause deadlock; fixed (867, 869)
5573 - stop cancelling OST locks when files are closed (481)
5574 - overlapping XID spaces caused network corruption (851, 853)
5575 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5576 - setattr_raw added, to do single-RPC, server-side setattrs
5577 - lmc/lconf syntax change for OST UUIDs
5578 - fix crashy race condition between ptlrpc_free_req and osc_close
5579 - don't use request in mdc_enqueue if we hit a timeout (889)
5580 - don't set the inode i_size for regular files from the MDS (896)
5581 - handle out of order completion AST (842)
5582 - don't LBUG if a lock request times out after receiving AST (913)
5583 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5584 - if a bad lock AST arrives, send an error instead of dropping entirely
5585 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5586 - fix leak in bulk IO when only partially completed (899, 900, 926)
5587 - fix O_DIRECT for ia64 (55)
5588 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5589 - C-z after timeout could hang a process forever; fixed (977)
5591 - client-side I/O cache (678, 924, 929, 941, 970)
5593 - READPAGE and SETATTRs which don't take server-side locks get
5596 2003-02-11 Phil Schwan <phil@clusterfs.com>
5599 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5600 - Fix fsfilt_extN_readpage() to read a full page of directory
5601 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5602 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5603 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5604 - Most intent operations were removed, in favour of a new RPC mode
5605 that does a single RPC to the server and bypasses most of the VFS
5606 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5607 - Aggressively cancel local locks on DLM servers
5608 - mds_reint_unlink sends EA to the client if it's the last nlink.
5609 client uses that EA to unlink OST objects.
5610 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5611 - recursive symlinks were fixed (439)
5612 - fixed NULL deref in DEBUG_REQ
5613 - filter_update_lastobjid no longer calls sync, which annoyed extN
5614 - fixed multi-client small-writes to a single file problem (445)
5615 - fixed mtime updates during file writes (607)
5616 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5617 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5618 - fixed Linux OST size reporting problem (444, 656)
5619 - OST now updates object mtime with writes or setattr (607, 619)
5620 - client verifies file size before zeroing page past EOF (445)
5621 - OST now writes last allocated objid to disk with allocation (108)
5622 - LOV on echo now works (409)
5624 - mds_reint_unlink sends a new buffer, with the EA included. this
5625 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5626 set if a regular file was being unlinked, and it was the last link
5627 - use PtlGet from the target for bulk writes (315)
5628 - OST now updates object mtime with writes or setattr (607, 619)
5629 - LDLM now has a grant-time callback to revalidate locked items, if
5631 - Many MDS operations were reorganized to combat race conditions
5633 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5634 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5636 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5639 - Fully reactivate OST imports after reconnection (512, others)
5640 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5641 - More graceful error handling for truncating on dead OST (515)
5642 - Don't error out unless we're actually accessing dead stripes (474)
5643 - Fix garbage sizes when stripes are missing (410)
5644 - LRU counters were broken, causing constant lock purge (433, 432)
5645 - garbage on read from stripes with failed OSTs (441)
5646 - mark OSCs as active before reconnecting during recovery (438)
5647 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5648 - lfind did not preserve OST order in output (443)
5649 - symlinks cause hung clients, incorrect data (439)
5650 - stop dereferencing request after dropping refcount (457)
5651 - don't LASSERT(spin_is_locked) on non-SMP (455)
5652 - fixes for many rename() bugs
5653 - fstat didn't correctly synchronize attributes (399)
5654 - server must handle lock cancellation during blocking AST prep (487)
5655 - bulk descriptors were free()d too soon (511)
5656 - fix paths in lconf, which would load incorrect modules (451, 507)
5657 - fix confusing lconf 'host not found' error message (386)
5658 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5659 - fix race condition in mdc_blocking_ast() for inode access (526)
5660 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5661 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5662 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5663 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5664 - fix lstripe in several ways (kernel side) (527)
5665 - fix request leak in ldlm_cli_enqueue (262)
5666 - incorrect OSC was marked inactive after OST failure
5667 - call mds_fs_cleanup before unmounting filesystem (524)
5668 - fix races between taking ns_lock and ldlm_lock_change_resource
5669 - fix races updating LOV export open file list
5670 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5671 - fix recovery NULL deref in ldlm_cli_cancel_unused
5672 - fix some DLM races by using new hash table for lock handles (419)
5673 - permit the client to specify desired inodes, at replay
5674 - duplicate requests when we queue them for replay reintegration
5675 - fix last_rcvd offset calculation
5676 - sync after each recovered transaction, so we always make progress
5677 - never, not always, ERESTART requests without transnos
5678 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5680 - skip replay if the MDS says that the client is already connected
5681 - don't check for a recovery-enabled export to match lctl's UUID
5682 - don't INC_USE_COUNT for phantom exports
5683 - don't crash when cleaning up phantom exports (567)
5684 - don't double-finish or set replay data for errored mdc_open requests
5685 - abort requests when they time out, so we don't get old replies
5686 - send/receive replies for AST messages again
5687 - if the client says that it doesn't have the lock, cancel it on the
5689 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5690 mark it as destroyed, it will all work out when decref is called
5691 - fix module use counts (22, 581)
5693 - ASTs now expect a reply (server cancels lock on error reply)
5695 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5698 - fix many simultaneous client startup (392)
5699 - fix dentry->d_it clobbering
5700 - credentials weren't being shipped for readdir/getattr operations
5701 - remove invalid assertions triggered during some concurrent MD
5703 - proper Lustre versions added (336, 389)
5704 - fix memory leak for create error case (398)
5705 - fix LOV locking bug that would get cli/srv out of sync
5706 - fix echo client over LOV (409)
5707 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5708 - fix double-O_EXCL intent crash (424)
5709 - avoid sending multiple lock CANCELs (352)
5711 - MDS can do multi-client recovery (modulo bugs in new code)
5713 - many updates, edits, cleanups
5715 2002-11-18 Phil Schwan <phil@clusterfs.com>
5718 - fix null d_it dereference (346)
5719 - fix full OST/dbench hang (333)
5720 - fix permission problem with file removal (286)
5721 - fix removal of OSCs from LOV when they fail
5722 - fix NULL deref during bulk timeout (214)
5723 - fix problems related to multiple filesystems on one MDS (241)
5724 - fixed serious subtle metadata locking bugs
5725 - free locks on clients when inodes are removed due to memory
5727 - fix inode pointer in lock data (285)
5728 - partial support for multiple MDS on a single host (241)
5729 - data locks weren't cancelled at clear_inode time (290, 311)
5730 - intent locks could lead to unbounded lock growth (205)
5731 - added a maximum lock count, an LRU list, and a flusher
5732 - fix multiple rename (365)
5733 - properly abstracted the echo client
5734 - OSC locked 1 byte too many; fixed
5735 - rewrote brw callback code:
5736 - fixed recovery bugs related to LOVs (306)
5737 - fixed too-many-pages-in-one-write crash (191)
5738 - fixed (again) crash in sync_io_timeout (214)
5739 - probably fixed callback-related race (385)
5741 - Add capability to MDS protocol
5742 - LDLM cancellations and callbacks on different portals
5744 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5747 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5748 - defer open object destruction to close time (601981, 138)
5749 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5750 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5751 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5752 - don't call obd_setattr in ll_file_release for destroyed objects
5753 * protocol change to lustre_msg: move |version| and add |flags|
5754 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5755 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5756 * added replay of create, unlink, link and rename operations during
5757 MDS failover; recovery should be much more robust now
5758 * remove failed OSCs from LOVs (only lov_create uses this so far)
5759 * the lustre-HOWTO was brought (more) up to date (582544)
5761 2002-10-23 Phil Schwan <phil@clusterfs.com>
5764 - in-use dentries weren't being reused properly (617851)
5765 - prevent multiple LDLM setup (599178)
5766 - fix LOV size calculations for truncate (617853)
5767 - fix client handling of MDS intent errors (POSIX)
5768 - fix permission bug in lovstripe.c test (624321)
5769 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5770 - truncate past end of file could corrupt data
5771 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5772 - a race in recovery could return ETIMEDOUT to apps (623947)
5773 - building outside the source directory was fixed
5774 * the lustre-HOWTO was brought (more) up to date (582544)
5775 * major progress was made on recovery functionality
5777 2002-10-10 Phil Schwan <phil@clusterfs.com>
5780 - recovery deadlock fix
5781 - rm -rf causes LBUG fix (617817)
5782 - file open by multiple tasks fix (618962)
5783 - directory permissions bugs (602707 and 620007)
5784 - journal_stop fixed with locking (611313)
5785 - O_APPEND failures resolved (618273, perhaps 614459)
5786 - lconf PATH fix (619770)
5787 - IA64 build fix (621450)
5788 - RPC buffer sizes scale with amount of memory
5790 2002-10-01 Phil Schwan <phil@clusterfs.com>
5793 - locks would be cancelled without throwing away data pages,
5794 resulting in inconsistent data (605627)
5795 - inode attributes were not always being refreshed (605627, 612449)
5796 - lconf now continues to cleanup after lctl reports an error
5797 - MDS now enforces user permissions (602707)
5798 - lprocfs cleanup fixed, but not yet enabled (614157)
5799 - fixed infinite server hang, should a client not respond to an AST
5800 - avoid going into recovery if user calls readlink() with a buffer
5801 that's too small (613941)
5802 - AST RPCs no longer require replies (614867) -- this may be changed
5803 - don't crash server if client sends an IOV that's too big (611336)
5804 - fixed lock conversion deadlock (611892)
5805 - fixed the following of symlinks (614622)
5806 * recovery: the server can remove locks from a client that dies, other
5807 clients can make progress
5808 * more extN patch fixes
5809 * compile-time configurable ptlrpc buffer allocations
5811 - collaborative read cache document
5812 - Lustre Lite Performance CDR document-in-progress
5814 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5817 - fix typo in patch-2.4.18
5819 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5822 - clear ptlrpc request each time in handle_incoming_request()
5823 - unlink of files now destroys the object on the OST
5825 2002-09-19 Peter Braam <braam@clusterfs.com>
5827 * add hard link support
5828 * change obdfile creation method
5829 * kernel patch changed
5831 2002-09-19 Peter Braam <braam@clusterfs.com>
5834 - stack overflow bug in extN fixed
5836 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5838 * documentation updates
5839 - add man pages for config tools
5840 - update tests/README to describe testing with new config tools
5841 - finish metadata API descriptions
5842 * bug fixes and cleanups
5843 - statfs workaround for 16TB limit
5844 - LOV stripe allocation improved, can stripe on subset of OSTs
5845 - LOV file size/IO offset was wrong for files > 4GB in size
5846 - object EA data was being dropped, caused files to be unreadable
5847 - memory overflow with non-LOV OST caused memory corruption
5848 - fixed regression tests to work with new config tools, obdfilter
5849 - fixed bug when directory size became larger than 1 block
5850 - fixed bug (for single client case) when PWD was deleted
5851 - invalidate local directory pages when doing intent-based ops
5852 - avoid LDLM oops when lock callback contained bad data
5854 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5856 * documentation updates
5857 * bug fixes and cleanups
5858 - configuration tools
5861 - 64-bit compile warnings
5862 - 64-bit internal statfs data
5864 * test_brw on persistent OST devices
5866 * lprocfs (disabled)
5868 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5870 * documentation updates
5871 * bug fixes and cleanups
5872 * configuration tools
5874 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5877 * many small fixes to 0_5_4
5878 * io/network handling
5879 * thinkos in MDS operations
5881 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5884 * crucial basic fixes to 0.5.3
5885 * IOR, Iozone work over Elan
5888 2002-08-07 Phil Schwan <phil@clusterfs.com>
5889 * version 0_5_3, our first alpha
5890 * we use the new Portals iovs
5891 * documentation updates
5892 * bug fixes and cleanups
5893 * small changes in the DLM wire protocol
5895 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5896 * version 0_5_1 with some initial stability,
5897 * locking on MD and file I/O.
5898 * documentation updates
5899 * several bug fixes since 0.5.0
5900 * small changes in wire protocol
5902 2002-07-18 Phil Schwan <phil@clusterfs.com>
5904 * delivered as Lustre Light Alpha
5905 * fixed a crash after handling invalid MDS requests
5906 * fixed directory pages for architectures with non-4k pages sizes
5908 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5909 * release version v0_4_4
5910 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5912 * Fixes a number of interruption problems with OST operations.
5913 * Update documentation for portals header changes
5914 * Move all wire protocol structs/defines to lustre_idl.h
5915 * Fixes symlink length bug.
5916 * Add tcpdump to repository.
5918 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5919 * release version v0_4_3
5920 * Fixes statfs for inodes on extN.
5921 * Fixes bug in runtests which would delete /etc/hosts.
5922 * Use 64-bit object IDs wherever possible (not into VFS though)
5923 Remove ost_get_info, which is unused by lustre, and out of date.
5925 2002-07-03 Peter Braam <braam@clusterfs.com>
5926 * release version v0_4_2 Fixes a lookup error (type not passed)
5927 * move forward to head of Portals
5928 * move forward to latest Lustre kernel
5930 2002-06-25 Peter Braam <braam@clusterfs.com>
5931 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob