1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.39.3 (SLES 10),
5 2.6.18-128.1.14.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.41.6.sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
14 * File join has been disabled in this release, refer to Bugzilla 16929.
18 Description: truncate starts GFP_FS allocation under transaction causing deadlock
19 Details : ldiskfs_truncate calls grab_cache_page which may start page
20 allocation under an open transaction. This may lead to
21 calling prune_icache with consequent lustre reentrance.
25 Description: enable client interface failover
26 Details : When a child reconnects from another NID, properly update export
27 nid hash position and ldlm reverse import.
29 Severity : enhancement
31 Description: Communicate OST degraded/readonly state via statfs to MDS
32 Details : Flags in the statfs returned from OSTs indicate whether the
33 OST is in a degraded RAID state, or if the filesystem has
34 turned read-only after a filesystem error is detected.
39 Description: don't panic if EPROTO was hit when reading symlink
40 Details : correctly handling request reference in error cases.
42 Severity : enhancement
44 Description: Add LustreNetLink, a kernel-userspace communcation path.
46 Severity : enhancement
48 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
53 Description: access to llog context before init.
54 Details : move handling CATALOGS file at osc layer and forbid access to llog
58 Frequency : rare, only when client is writing to OST on local node
60 Description: Avoid deadlock for local client writes
61 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
62 memory freeing context. This allows OST threads to set the
63 PF_MEMALLOC flag on task structures in order to allocate memory
64 from reserved pools and complete IO.
65 Use GFP_HIGHUSER for OST allocations for non-local client writes,
66 so that the OST threads generate memory pressure and allow
67 inactive pages to be reclaimed.
69 Severity : enhancement
71 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
76 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
77 Details : move ldlm namespace creation in setup phase to avoid grab
78 _lprocfs_lock with cli_sem held.
82 Description: Temporarily disable grant shrink.
83 Details : Disable the feature for debugging.
86 Frequency : only when formatting test filesystems
88 Description: Unable to run several mkfs.lustre on loop devices at the same time
89 Details : mkfs.lustre returns error 256 on the concurrent loop devices
90 formatting. The solution is to proper handle the error.
92 Severity : enhancement
94 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
96 Severity : enhancement
98 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
103 Description: resolve race between obd_disconnect and class_disconnect_exports
104 Details : if obd_disconnect will be called to already disconnected export he
105 forget release one reference and osc module can't unloaded.
107 Severity : enhancement
109 Description: Allow tuning service thread via /proc
110 Details : For each service a new
111 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
112 created that can be used to set min/max thread counts, and get the
113 current number of running threads.
118 Descriptoin: don't return error if have partially created objects for file.
119 Details : lov_update_create_set uses set->set_success as index for created
120 objects, so if some requests will be failed, they will have hole
121 at end of array and qos_shrink_lsm for allocate correct lsm.
123 Severity : enhancement
125 Description: Update OFED support to 1.4.1
130 Description: Reduce small size read RPC
131 Details : Set read-ahead limit for every file and only do read-ahead when
132 available read-ahead pages are bigger than 1M to avoid small size
135 Severity : enhancement
137 Description: move AT tunable parameters for more consistent usage
138 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
140 Severity : enhancement
142 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
143 Details : allow skip disconnected ost for send statfs request and hide error
149 Description: fix racy locking of mballoc block bitmaps causing BUG
150 Details : The locking of the mballoc buddy bitmap and the in-memory
151 block bitmap was using two different spin locks in some
152 cases. This made it possible to incorrectly access the
153 mballoc bitmap while another process was modifying it,
154 causing a sanity assertion to fail. While no on-disk corruption
155 was reported, there was some risk of this happening.
158 Frequency : rare, on llog test 6
160 Descriptoin: don't allow connect to already connected import
161 Details : allowing connect to already connected import is hide connecting problem.
166 Descriptoin: allow kill process which wait statahead result
167 Details : for some reasons 'ls' can stick in waiting result from statahead,
168 in this case need way for kill this process.
170 Severity : enhancement
172 Description: Add state history info file, enhance import info file
173 Details : Track import connection state changes in a new osc/mdc proc file;
174 add overview-type data to the osc/mdc import proc file.
176 Severity : enhancement
178 Description: MDS create should not wait for statfs RPC while holding DLM lock.
180 Severity : enhancement
182 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
187 Descriptoin: OST grant too much space to client even there are not enough space.
188 Details : Client will shrink its grant cache to OST if there are no write
189 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
190 this grant cache if there are already not enough avaible space
191 (left_space < total_clients * 32M).
194 Frequency : start MDS on uncleanly shutdowned MDS device
196 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
197 Details : stay in waiting mds<>ost recovery finished produce random bugs
198 due race between two ll_sync thread for one lov target. send
199 ACTIVATE event only if connect realy finished and import have
203 Frequency : rare, connect and disconnect target at same time
205 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
206 Details : don't call obd_disconnect under lov_lock. this long time
207 operation and can block ptlrpcd which answer to connect request.
212 Descriptoin: don't lose wakeup for imp_recovery_waitq
213 Details : recover_import_no_retry or invalidate_import and import_close can
214 both sleep on imp_recovery_waitq, but we was send only one wakeup
218 Frequency : always with long access acl
220 Descriptoin: mds can't pack reply with long acl.
221 Details : mds don't control size of acl but they limited by reint/getattr
224 Severity : enhancement
226 Description: Update to SLES10 kernel-2.6.16.60-0.33.
228 Severity : enhancement
230 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
233 Frequency : start MDS on uncleanly shutdowned MDS device
235 Descriptoin: aborting recovery hang on MDS
236 Details : don't throttle destroy RPCs for the MDT.
239 Frequency : on remount
241 Description: external journal device not working after the remount
242 Details : clear dev_rdonly flag for external journal devices in
248 Description: shutdown vs evict race
249 Details : client_disconnect_export vs connect request race.
250 if client will evicted at this time - we start invalidate
251 thread without referece to import and import can be freed
255 Frequency : rare, need acl's on inode.
257 Description: client can't handle ost additional correctly
258 Details : if ost was added after client connected to mds client can have
259 hit lnet_try_match_md ... to big messages to wide striped files.
260 in this case need teach client to handle config events about add
261 lov target and update client max ea size at that event.
263 Severity : enhancement
265 Description: Changelogs
266 Details : Changelogs are a lightweight mechanism to track filesystem
267 metadata and namespace changes. The changelog is recorded
268 permanently on the MDTs, and is periodically "consumed" / purged
269 when records are no longer needed.
271 Severity : enhancement
273 Description: compact fld format with extents
274 Details : Store range of seq rather than every seq in FLD. Seq
275 controller update FLD rather than clients. In Case of CMD, mdt0
276 has FLD, all other metadata server act as non persistent proxy
277 for FLD queries and cache fld entries in fld cache.
282 Description: don't skip ost target if they assigned to file
283 Details : Drop slow OSCs if we can, but not for requested start idx.
284 This means "if OSC is slow and it is not the requested
285 start OST, then it can be skipped, otherwise skip it only
286 if it is inactive/recovering/out-of-space.
290 Description: more cleanup in mds_lov
291 Details : not send LOV EA under replay, we can't know about they size at this
292 time. Don't allow client connect to mds before any ost connected,
293 for avoid problems with LOV EA size and returning EIO to client.
295 Severity : enhancement
297 Description: Interoperability at server side (Disk interoperability)
299 Severity : enhancement
301 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
303 Severity : enhancement
305 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
307 Severity : enhancement
309 Description: New client IO stack (CLIO).
311 Severity : enhancement
313 Description: Commit on sharing. Eliminate inter-client dependencies between
314 uncommitted transactions by doing transaction commits.
315 Thereby clients may recovery independently.
318 Frequency : Create a symlink file with a very long name
320 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
321 Details : If there is no extra space in the request for early cancels,
322 ldlm_req_handles_avail() returns 0 instead of a negative value.
324 Severity : enhancement
326 Description: Add /proc entry for import status
327 Details : The mdc, osc, and mgc import directories now have
328 an import directory that contains useful import data for debugging
331 Severity : enhancement
333 Description: Re-disable certain /proc logging
334 Details : Enable and disable client's offset_stats, extents_stats and
335 extents_stats_per_process stats logging on the fly.
338 Frequency : Only on FC kernels 2.6.22+
340 Description: oops in statahead
341 Details : Do not drop reference count for the dentry from VFS when lookup,
342 VFS will do that by itself.
344 Severity : enhancement
346 Description: Generic /proc file permissions
347 Details : Set /Proc file permissions in a more generic way to enable non-
348 root users operate on some /proc files.
352 Description: Hitting mdc_commit_close() ASSERTION
353 Details : Properly handle request reference release in
354 ll_release_openhandle().
358 Description: quota recovery deadlock during mds failover
359 Details : This patch includes att18982, att18236, att18237 in bz14840.
361 1. fix osts hang when mds does failover with quotaon
362 2. prevent watchdog storm when osts threads wait for the
367 Frequency : only patchless client
368 Description: add workaround for race between add/remove dentry from hash
370 Severity : enhancement
372 Description: Allow OST glimpses to return PW locks
376 Description: LBUG when llog conf file is full
377 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
381 Description: Prevent import from entering FULL state when server in recovery
385 Description: service mount cannot take device name with ":"
386 Details : Only when device name contains ":/" will mount treat it as
392 Description: replace ptlrpcd with the statahead thread to interpret the async
393 statahead RPC callback
397 Frequency : on recovery
398 Description: I/O failures after umount during fail back
399 Details : if client reconnected to restarted server we need join to recovery
400 instead of find server handler is changed and process self
401 eviction with cancel all locks.
403 Severity : enhancement
405 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
407 Severity : enhancement
409 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
411 Severity : enhancement
413 Description: Upcall on Lustre log has been dumped
414 Details : Allow for a user mode script to be called once a Lustre log has
415 been dumped. It passes the filename of the dumped log to the
416 script, the location of the script can be specified via
417 /proc/sys/lnet/debug_log_upcall.
422 Description: avoid idr_remove called for id which is not allocated.
423 Details : Move assigment s_dev for clustered nfs to end of initialization,
424 to avoid problem with error handling.
429 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
430 Details : When connection is reused this not moved from CONN_UNUSED_HASH
431 into CONN_USED_HASH and this prodice warning when put connection
432 again in unused hash.
435 Severity : enhancement
437 Description: File striping can now be set to use an arbitrary pool of OSTs.
439 Severity : enhancement
441 Description: Export bytes_read/bytes_write count on OSC/OST.
445 Description: Early reply size mismatch, MGC loses connection
446 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
447 the connect flags are properly negotiated.
452 Description: quotas are not honored with O_DIRECT
453 Details : all writes with the flag O_DIRECT will use grants which leads to
454 this problem. Now using OBD_BRW_SYNC to guard this.
458 Description: add quota statistics
459 Details : 1. sort out quota proc entries and proc code.
460 2. add quota statistics
462 Severity : enhancement
464 Description: enable quota support for HEAD.
468 Description: Properly propagate oinfo flags from lov to osc for statfs
469 Details : restore missing copy oi_flags to lov requests.
471 Severity : enhancement
473 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
474 : l_getgroups(8), lst(8), routerstat(8)
475 Details : included man pages for llobdstat(8), llstat(8),
476 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
478 Severity : enhancement
480 Description: configure's --enable-quota should check the
481 : kernel .config for CONFIG_QUOTA
482 Details : configure is terminated if --enable-quota is passed but
483 : no quota support is in kernel
487 Description: Remove portals compatibility
488 Details : Remove portals compatibility, not interoperable with releases
493 Description: Resolve device initialization race
494 Details : Prevent proc handler from accessing devices added to the
495 obd_devs array but yet be intialized.
497 Severity : enhancement
499 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
501 Severity : enhancement
503 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
508 Description: Using wrong pointer in osc_brw_prep_request
509 Details : Access to array[-1] can produce panic if kernel compiled with
510 CONFIG_PAGE_ALLOC enabled
514 Description: Client runs out of low memory
515 Details : Consider only lowmem when counting initial number of llap pages
519 Description: *optional* service tags registration
520 Details : if the "service tags" package is installed on a Lustre node
521 When the filesystem is mounted, a local-node service tag will
522 be created. See http://inventory.sun.com/ for more information
523 about the Service Tags asset management system.
527 Description: Kernel BUG tries to release flock
528 Details : Lustre does not destroy flock lock before last reference goes
529 away. So always drop flock locks when client is evicted and
530 perform unlock regardless of successfulness of speaking to MDS.
534 Description: add refcount protection for osc callbacks, avoid panic on shutdown
538 Description: sanity test 65a fails if stripecount of -1 is set
539 Details : handle -1 striping on filesystem in ll_dirstripe_verify
544 Description: ASSERTION(CheckWriteback(page,cmd)) failed
545 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
548 Severity : enhancement
550 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
554 Description: do not process already freed flock
555 Details : flock can possibly be freed by another thread before it reaches
556 to ldlm_flock_completion_ast.
560 Description: LBUG during stress test
561 Details : Need properly lock accesses the flock deadlock detection list.
565 Description: oops in page fault handler
566 Details : kernel page fault handler can return two special 'pages' in error
567 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
571 Description: timeout with invalidate import.
572 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
573 should be handled by ptlrpcd. This produce long age waiting and
574 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
576 Severity : enhancement
578 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
582 Description: enable MGS and MDT services start separately
583 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
584 a co-located MGS without starting the MGS, which is a complement
585 to 'nosvc' mount option.
589 Frequency : after recovery
590 Description: precreate to many object's after del orphan.
591 Details : del orphan st in oscc last_id == next_id and this triger growing
592 count of precreated objects. Set flag LOW to skip increase count
593 of precreated objects.
597 Frequency : rare, on clear nid stats
598 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
599 Details : when clean nid stats sometimes try destroy live entry,
600 and this produce panic in free.
604 Description: Stack overflow during MDS log replay
605 ease stack pressure by using a thread dealing llog_process.
609 Description: wait until IO finished before start new when do lock cancel.
610 Details : VM protocol want old IO finished before start new, in this case
611 need wait until PG_writeback is cleared until check dirty flag and
612 call writepages in lock cancel callback.
614 Severity : enhancement
616 Description: using special macro for print time and cleanup in includes.
620 Description: mds_mfd_close() ASSERTION(rc == 0)
621 Details : In mds_mfd_close(), we need protect inode's writecount change
622 within its orphan write semaphore to prevent possible races.
626 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
630 Frequency : rare, on shutdown ost
631 Description: don't hit live lock with umount ost.
632 Details : shrink_dcache_parent can be in long loop with destroy dentries,
633 use shrink_dcache_sb instead.
637 Description: don't panic with use echo client
638 Details : echo client pass NULL as client nid pointer and this produce null
643 Description: fix build on ppc32
644 Details : compile code with -m64 flag produce wrong object file for ppc32.
648 Description: add message levels for liblustreapi
652 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
653 Details : Move the 'good_osts' check before the 'total_bavail' check. This
654 will result in an -EAGAIN and in the exit call path we call
655 alloc_rr() which will with increasing aggressiveness attempt to
656 aquire precreated objects on the minimum number of required OSCs.
660 Description: Use old size assignment to avoid deadlock
661 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
662 the scheduling while holding a spinlock. We do not need locking
663 for size in ll_update_inode() because size is only updated from
664 the MDS for directories or files without objects, so there is no
665 other place to do the update, and concurrent access to such inodes
666 are protected by the inode lock.
670 Description: resolve "_IOWR redefined" build error on SLES10
674 Description: dump the memory debugging after all modules are unloaded to
675 suppress false negative in conf_sanity test 39
677 Severity : enhancement
679 Description: build kernel-ib packages for OFED 1.3 in our release cycle
684 Description: fix SLES kernel versioning
685 Details : the kernel version for our SLES 10 kernel did not include a "-"
686 before the "smp" at the end. while this was not a problem in
687 general, it did mean that software trying to use the kernel
688 version to try to detect a vendor specific kernel would fail.
689 this was most evident by the OFED build scripts.
693 Description: Don't update lov_desc members until making sure they are valid
694 Details : When updating lov_desc members via proc fs, need fix their
695 validities before doing the real update.
699 Description: don't put request into delay list while invalidate in flight.
700 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
701 import in flight. this produce timeout for invalidate and sometimes
702 can cause stale data.
706 Frequency : on ppc only
707 Description: not convert ost objects for directory because it's not exist.
708 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
712 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
714 Severity : enhancement
716 Description: Update to support 2.6.22.14 vanilla kernel.
720 Frequency : rare, on recovery
721 Description: read procfs can produce deadlock in some situation
722 Details : Holding lprocfs lock which send rpc can produce block for destroy
723 obd objects and this also block reconnect with -EALREADY. This
724 isn't fix all lprocfs bugs - but make it rare.
726 Severity : enhancement
728 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
731 Frequency : frequent on X2 node
733 Description: mdc_set_open_replay_data LBUG
734 Details : Set replay data for requests that are eligible for replay.
738 Description: lustre_mgs: operation 101 on unconnected MGS
739 Details : When MGC is disconnected from MGS long enough, MGS will evict the
740 MGC, and late on MGC cannot successfully connect to MGS and a lot
741 of the error messages complaining that MGS is not connected.
745 Frequency : on network error
746 Description: panic with double free request if network error
747 Details : mdc_finish_enqueue is finish request if any network error ocuring,
748 but it's true only for synchronus enqueue, for async enqueue
749 (via ptlrpcd) this incorrect and ptlrpcd want finish request
752 Severity : enhancement
754 Description: client-side metadata stat-ahead during readdir(directory readahead)
755 Details : perform client-side metadata stat-ahead when the client detects
756 readdir and sequential stat of dir entries therein
759 Frequency : on start mds
761 Description: Implement get_info(last_id) in obdfilter.
764 Frequency : occasional
766 Description: Correctly check stale fid, not start epoch if ost not support SOM
767 Details : open with flag O_CREATE need set old fid in op_fid3 because
768 op_fid2 was overwritten with new generated fid, but mds can answer
769 with one of these two fids and both is not stale. Setattr
770 incorrectly started an epoch and assume will be called
771 done_writing, but without SOM done_writing ever being called.
774 Frequency : rare, depends on device drivers and load
776 Description: MDS or OSS nodes crash due to stack overflow
777 Details : Code changes in 1.8.0 increased the stack usage of some functions.
778 In some cases, in conjunction with device drivers that use a lot
779 of stack the MDS (or possibly OSS) service threads could overflow
780 the stack. One change which was identified to consume additional
781 stack has been reworked to avoid the extra stack usage.
784 Frequency : occasional
786 Description: Do not fail import if osc_interpret_create gets -EAGAIN
787 Details : If osc_interpret_create got -EAGAIN it immediately exits and
788 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
789 oscc_has_objects and see OSC has no objests and call
790 oscc_internal_create to resend create request.
792 Severity : enhancement
794 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
796 Severity : enhancement
798 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
801 Frequency : very rare
803 Description: Fix replay if there is an un-replied request and open
804 Details : In some cases, older replay request will revert the
805 mcd->mcd_last_xid on MDS which is used to record the client's
808 Severity : enhancement
810 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
812 Severity : enhancement
814 Description: Add rhel5 support to HEAD.
816 Serverity : enhancement
818 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
823 Description: Don't allow skipping OSTs if index has been specified.
824 Details : Don't allow skipping OSTs if index has been specified, make
825 locking in internal create lots better.
829 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
830 Details : access to bitfield in structure is always rounded to long
831 and this produce problem with not atomic change any bit.
835 Description: Lustre make rpms failed.
836 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
837 builds Lustre rpms from tarball.
840 Frequency : rare on shutdown ost
842 Description: If llog cancel was not send before clean_exports phase, this can
843 produce deadlock in llog code.
844 Details : If llog thread has last reference to obd and call class_import_put
845 this produce deadlock because llog_cleanup_commit_master wait when
846 last llog_commit_thread exited, but this never success because was
847 called from llog_commit_thread.
851 Description: allow userland application know is lost one of stripes.
852 Details : fill lvb_blocks with error code on ost and return it to
853 application if error flag found.
857 Description: NULL lov_tgts causing MDS oops
858 Details : more safe checks for NULL lov_tgts for avoid oops.
860 Severity : enhancement
862 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
866 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
867 Details : If an OST is down the MDS will hang indefinitely in
868 obd_statfs_rqset() waiting for the statfs data. While for
869 MDS QOS usage of statfs, it should not stuck in waiting.
871 Severity : enhancement
873 Description: remote_acl support
874 Details : Support ACL-based permission check for remote user.
875 Support setfacl/getfacl for remote user with the utils
876 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
877 the system "{s,g}etfacl" utils.
879 Severity : enhancement
881 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
883 Severity : enhancement
885 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
889 Description: make mgs_setparam() handle fsname containing dash
890 Details : fsname containing a dash does not work with lctl conf_param
892 Severity : enhancement
894 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
896 Severity : enhancement
898 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
902 Frequency : rare, at shutdown
903 Description: access already free / zero obd_namespace.
904 Details : if client_disconnect_export was called without force flag set,
905 and exist connect request in flight, this can produce access to
906 NULL pointer (or already free pointer) when connect_interpret
907 store ocd flags in obd_namespace.
911 Frequency : only at startup
912 Description: not alloc memory with spinlock held.
913 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
914 if any spinlock held.
916 Severity : enhancement
918 Description: make lustre randomly fail allocating memory
919 Details : Make lustre randomly failed allocating memory for testing purpose.
921 Severity : enhancement
923 Description: lost problems with lov objid file
924 Details : Fixes some scability and access to not inited memory problems
925 in work with lov objdid file.
930 Description: lfs find does not continue on file error
931 Details : Continue other files processing when a file/dir is absent.
935 Description: Inconsistent usage of lustre_pack_reply()
936 Details : Standardize the usage of lustre_pack_reply() such that it
937 always generate a CERROR on failure.
940 Frequency : occasional
942 Description: MDS hang or stay in waiting lock
943 Details : If client receive lock with CBPENDING flag ldlm need send lock
944 cancel as separate rpc, to avoid situation when cancel request
945 can't processed due all i/o threads stay in wait lock.
949 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
953 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
957 Description: Update to SLES9 kernel-2.6.5-7.287.3.
961 Description: Update to RHEL4 latest kernel.
963 Severity : enhancement
965 Description: Build SLES10 patchless client fails
966 Details : The configure was broken by run ./configure with
967 --with-linux-obj=.... argument for patchless client. When the
968 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
969 can't search header adequately. Use absolute path such as
970 -I($LINUX)/include instead.
974 Description: interrupt oig_wait produce painc on resend.
975 Details : brw_redo_request can be used for resend requests from ptlrpcd and
976 private set, and this produce situation when rq_ptlrpcd_data not
977 copyed to new allocated request and triggered LBUG on assert
978 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
979 ptlrpcd set if request is changed and can be safety changed to use
984 Description: LASSERT_{REQ,REP}SWAB macros are buggy
985 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
986 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
987 these macros and replace them with their difinitions instead.
991 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
992 Details : Update which_patch & target file for SLES10 latest kernel.
995 Description: Remove client patches from SLES 10 kernel.
996 Details : This causes SLES 10 clients to behave as patchless clients
997 even on a Lustre-patched (server) kernel.
999 Severity : enhancement
1001 Description: self-adjustable client's lru lists
1002 Details : use adaptive algorithm for managing client cached locks lru
1003 lists according to current server load, other client's work
1004 pattern, memory activities, etc. Both, server and client
1005 side namespaces provide number of proc tunables for controlling
1008 Severity : enhancement
1010 Description: light-weight GSS support
1011 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1012 and reduce performance overhead.
1014 Severity : enhancement
1016 Description: Linux keyring support
1017 Details : Support using service of Linux keyring for Lustre GSS internal
1018 context refresh/cache mechanism.
1022 Description: Fix errors in lfs documentation
1023 Details : Fixes man pages
1027 Description: don't use GFP_* in generic Lustre code.
1028 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1032 Description: obdclass is limited by single OBD_ALLOC(idarray)
1033 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1037 Description: warnings with build patchless client with vanila 2.6.19 and up
1038 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1039 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1043 Frequency : Only for SLES
1044 Description: sanity_quota fail test_1
1045 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1046 file, which makes TSTID[2] inunique.
1050 Frequency : Only for SLES9
1051 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1052 Details : Update target/ChangeLog/which_patch .
1056 Frequency : Only for test_10 in sanity-quota.sh
1057 Description: a bug in quota test code
1058 Details : checking a test flag in a wrong place
1062 Frequency : Only for test_13 in sanity-quota.sh
1063 Description: a bug in quota test script
1064 Details : shouldn't check overfull allocated quota
1068 Description: jbd statistics
1069 Details : Port older jbd statistics patch for sles10
1071 Severity : enhancement
1073 Description: Add printing inode info into message about error in writepage.
1078 Description: reply_lock_interpret crash due to race with it and lock cancel.
1079 Details : Do not replay locks that are being cancelled. Do not reference
1080 locks by their address during replay, just by their handle.
1084 Frequency : When flocks are used.
1085 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1086 Details : Flock locks might destroy just granted lock if it could be merged
1087 with another existing flock, this is done in completion handler,
1088 so teach ldlm_cli_enquque_fini that this is a valid case for
1094 Description: Oops in read and write path when failing to allocate lock.
1095 Details : Check if lock allocation failed and return error back.
1099 Description: lstripe command fails for valid OST index
1100 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1101 instead of lov->desc.ld_active_tgt_count.
1105 Description: sanity.sh failed test 103
1106 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1111 Description: df doesn't work properly if diskfs blocksize != 4K
1112 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1116 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1117 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1119 Severity : enhancement
1121 Description: Adaptive timeouts
1122 Details : RPC timeouts adapt to changing server load and network
1123 conditions to reduce resend attempts and improve recovery time.
1127 Description: llapi_file_create() does not allow some changes
1128 Details : add llapi_file_open() that allows specifying the mode and
1129 open flags, and also returns an open file handle.
1133 Description: merge and cleanup kernel patches.
1134 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1138 Description: Add journal checksum support.(Kernel part)
1139 Details : The journal checksum feature adds two new flags i.e
1140 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1141 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1142 indicates that the commit block contains the checksum for
1143 the blocks described by the descriptor blocks. Now commit
1144 record can be sent to disk without waiting for descriptor
1145 blocks to be written to disk. This behavior is controlled
1146 using JBD2_FEATURE_ASYNC_COMMIT flag.
1150 Description: OSS needs mutliple precreate threads
1151 Details : Add ability to start more than one create thread per OSS.
1155 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1156 Details : execute lfs setstripe on client
1160 Description: mds_obd_create error creating tmp object
1161 Details : When the user sets quota on root, llog will be affected and can't
1162 create files and write files.
1165 Frequency : Always on ia64 patchless client, and possibly others.
1167 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1168 Details : This allows the patchless client to be loaded on architectures
1169 without this export.
1173 Description: RedHat Update kernel for RHEL5
1174 Details : Add the kernel config file for RHEL5.
1178 Description: RedHat Update kernel for RHEL5
1179 Details : Modify the kernel config file more closer RHEL5.
1183 Description: Build failure against Centos5 (RHEL5)
1184 Details : Define PAGE_SIZE when it isn't present.
1188 Description: test_7 on sanity-quota.sh can't be used on separate machines
1189 Details : add facet to handle it
1193 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1194 Details : with NFS, the anon dentry's parent was set to itself in
1195 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1196 corresponding dentry other than use rec->ur_name.
1198 Severity : enhancement
1200 Description: lfs setstripe enhancement
1201 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1205 Description: enable data checksumming by default
1206 Details : enable checksum by default, allow --disable-checksum
1207 configure option and "-o nochecksum" mount option. Checksums
1208 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1212 Description: lustre support for RHEL5
1213 Details : Add support for RHEL5.
1217 Description: Client eviction due to failover config
1218 Details : after a connection loss, the lustre client should attempt to
1219 reconnect to the last active server first before trying the
1220 other potential connections.
1224 Description: when mds and osts use different quota unit(32bit and 64bit),
1225 quota will be released repeatly.
1226 Details : void sending multiple quota reqs to mds, which will keep the
1227 status between the reqs.
1231 Description: osts not allocated evenly to files
1232 Details : change the condition to increase offset_idx
1235 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1237 Description: Data corruption for OSTs that are formatted larger than 2TB
1239 Details : When generating the bio request for lustre file writes the
1240 sector number would overflow a temporary variable before being
1241 used for the IO. The data reads correctly from Lustre (which
1242 will overflow in a similar manner) but other file data or
1243 filesystem metadata may be corrupted in some cases.
1247 Description: Tune the kernel for good SCSI performance.
1248 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1249 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1254 Description: rewrite ext2-derived code in obdclass/uuid.c
1255 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1256 from scratch preserving functionality.
1260 Description: "lfs find -obd UUID" prints directories
1261 Details : "lfs find -obd UUID" will return all directory names instead
1262 of just file names. It is incorrect because the directories
1263 do not reside on the OSTs.
1267 Description: use i_size_read and i_size_write in 2.6 port
1268 Details : replace inode->i_size access with i_size_read/write()
1271 Frequency : when removing large files
1273 Description: scheduling issue during removal of large Lustre files
1274 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1275 It causes scheduling issues when removing large files (17TB in the
1279 Frequency : only with liblustre clients on XT3
1281 Description: evictions taking too long
1282 Details : allow llrd to evict clients directly on OSTs
1285 Frequency : only on ppc
1287 Description: /proc/fs/lustre/devices broken on ppc
1288 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1289 We should make sure the type of 'index' is loff_t and then cast
1290 explicitly as needed below. Do not assign an explicitly cast
1294 Frequency : only for rhel5
1296 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1297 Details : Modify the target file & which_kernel.
1299 Severity : enhancement
1301 Description: omit set fsid for export NFS
1302 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1303 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1307 Frequency : Always, for kernels after 2.6.16
1308 Description: Fix warning idr_remove called for id=.. which is not allocated.
1309 Details : Last kernels save old s_dev before kill super and not allow
1310 to restore from callback - restore it before call kill_anon_super.
1314 Description: buffer overruns could theoretically occur
1315 Details : llapi_semantic_traverse() modifies the "path" argument by
1316 appending values to the end of the origin string, and a
1317 overrun may occur. Adding buffer overrun check in liblustreapi.
1321 Description: Fix error on 'ls .' at the top of the Lustre mount.
1322 Details : Don't revalidate dentry if it is a root dentry.
1326 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1327 Details : Modify vm-tunables-rhel4.patch.
1331 Description: Kernel config for 2.6.18-vanilla.
1332 Details : Modify targets/2.6-vanilla.target.in.
1333 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1334 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1335 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1336 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1340 Description: improve handling recoverable errors
1341 Details : if request processig with error which can be recoverable on server
1342 request should be resend, otherwise page released from cache and
1347 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1348 Details : Modify target file & which_patch.
1349 A flaw was found in the IA32 system call emulation provided
1350 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1351 value could be stored in the %RAX register, which could trigger an
1352 out-of-bounds system call table access. An untrusted local user
1353 could exploit this flaw to run code in the kernel
1354 (ie a root privilege escalation). (CVE-2007-4573).
1358 Description: O_DIRECT bypasses client statistics.
1359 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1360 read_ahead_stats were updated. copied the stats section from
1361 osc_send_oap_rpc() into async_internal().
1365 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1369 Description: change order of libsysio includes
1370 Details : '#include sysio.h' should always come before '#include xtio.h'
1374 Description: handle "serious error: objid * already exists" more gracefully
1375 Details : If LAST_ID value on disk is smaller than the objects existing in
1376 the O/0/d* directories, it indicates disk corruption and causes an
1377 LBUG(). If the object is 0-length, then we should use the existing
1378 object. This will help to avoid a full fsck in most cases.
1380 Severity : enhancement
1382 Description: adapt the lustre_config script to support the upgrade case
1383 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1384 server targets to 1.6 in parallel.
1388 Description: To avoid grant space > avaible space when the disk is almost
1389 full. Without this patch you might see the error "grant XXXX >
1390 available" or some LBUG about grant, when the disk is almost
1392 Details : In filter_check_grant, for non_grant cache write, we should
1393 check the left space by if (*left > ungranted + bytes), instead
1394 of (*left > ungranted), because only we are sure the left space
1395 is enough for another "bytes", then the ungrant space should be
1396 increase. In client, we should update cl_avail_grant only there
1397 is OBD_MD_FLGRANT in the reply.
1401 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1402 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1403 described in CVE-2007-4573. This problem could allow untrusted
1404 local users to gain root access.
1407 Frequency : when using O_DIRECT and quotas
1409 Description: Incorrect file ownership on O_DIRECT output files
1410 Details : block usage reported by 'lfs quota' does not take into account
1411 files that have been written with O_DIRECT.
1416 Description: touch file failed when fs is not full
1417 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1418 otherwise we can get ENOSP while fs is not full.
1422 Description: parallel lock callbacks
1423 Details : Instead of sending blocking and completion callbacks as separated
1424 requests, adding them to a set and sending in parallel.
1427 Frequency : only for Cray XT3
1428 Bugzilla : 12829/13455
1429 Description: Changing primary group doesn't change the group lustre assigns to
1431 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1432 overridden with the primary group provided by the group upcall,
1433 whereas the supplied fsgid can be trusted if it is in the list of
1434 supplementary groups returned by the group upcall.
1436 Severity : enhancement
1438 Description: Allow masking D_WARNING, D_ERROR messages from console
1439 Details : Console messages can now be disabled via lnet.printk.
1443 Description: User code with malformed file open parameter crashes client node
1444 Details : Before packing join_file req, all the related reference should be
1445 checked carefully in case some malformed flags cause fake
1446 join_file req on client.
1450 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1451 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1452 mode is equal to the granted mode. However, it should also check
1453 that LDLM_FL_CP_REQD is not set.
1457 Description: Heavy nfs access might result in deadlocks
1458 Details : After ELC code landed, it is now improper to enqueue any mds
1459 locks under och_sem, because enqueue might want to decide to
1460 cancel open locks for same inode we are holding och_sem for.
1464 Description: Client eviction while running blogbench
1465 Details : A lot of unlink operations with concurrent I/O can lead to a
1466 deadlock causing evictions. To address the problem, the number of
1467 oustanding OST_DESTROY requests is now throttled to
1468 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1469 callbacks are processed in priority.
1473 Description: enable ACLs on MDS by default
1474 Details : ACLs must be enabled on MDS by default.
1477 Frequency : PPC/PPC64 only
1479 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1480 Details : fix duplicated definitions between asm-ppc64/types.h and
1481 lustre_types.h on PPC.
1484 Frequency : PPC/PPC64 only
1486 Description: asm-ppc/segment.h does not exist
1487 Details : fix compile issue on PPC.
1491 Description: better handle error messages in extents code
1494 Frequency : RHEL4 only
1496 Description: mkfs is very slow on IA64/RHEL4
1497 Details : A performance regression has been discovered in the MPT Fusion
1498 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1499 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1500 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1502 Severity : enhancement
1504 Description: SNMP support enhancement
1505 Details : Adding total number of sampled request for an MDS node in snmp
1508 Severity : enhancement
1510 Description: Optimize ldlm waiting list processing for PR extent locks
1511 Details : When processing waiting list for read extent lock and meeting read
1512 lock that is same or wider to it that is not contended, skip
1513 processing rest of the list and immediatelly return current
1514 status of conflictness, since we are guaranteed there are no
1515 conflicting locks in the rest of the list.
1519 Description: Time out and refuse to reconnect
1520 Details : When the failover node is the primary node, it is possible
1521 to have two identical connections in imp_conn_list. We must
1522 compare not conn's pointers but NIDs, otherwise we can defeat
1523 connection throttling.
1527 Description: port llog fixes from b1_6 into HEAD
1528 Details : Port llog reference couting and some llog cleanups from b1_6
1529 (bug 10800) into HEAD, for protect from panic and access to already
1530 free llog structures.
1534 Description: Detect stride IO mode in read-ahead
1535 Details : When a client does stride read, read-ahead should detect that and
1536 read-ahead pages according to the detected stride pattern.
1540 Description: data checksumming impacts single node performance
1541 Details : add support for several checksum algorithm. Currently, only CRC32
1542 and Adler-32 are supported. The checksum type can be changed on
1543 the fly via /proc/fs/lustre/osc/*/checksum_type.
1547 Description: use adler32 for page checksums
1548 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1553 Description: build for x2 fails
1554 Details : fix compile issue on Cray systems.
1558 Description: Properly match for duplicate locks
1559 Details : Due to different lock order from skiplists code, we need to
1560 traverse entire list for now
1563 Frequency : only on PPC/SLES10
1565 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1566 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1567 be defined. Add a hack in mkfs_lustre.c to work around this compile
1572 Description: LASSERT on MDS when client holding flock lock dies
1573 Details : ldlm pool logic depends on number of granted locks equal to
1574 number of released locks which is not true for flock locks, so
1575 just exclude such locks from consideration.
1579 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1580 Details : Use fsfilt_sync() for both the whole filesystem sync and
1581 individual file sync to eliminate dangerous inode locking
1582 with I_LOCK that can lead to a deadlock.
1586 Description: performance in 1.6.3
1587 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1588 sure that 1MB requests are not fragmented by the block layer.
1590 Severity : enhancement
1592 Description: organize the server-side client stats on per-nid basis
1593 Details : Change the structure of stats under obdfilter and mds to
1601 The "uuid"s file would list the uuids of _active_ exports.
1602 And the clear entry is to clear all stats and stale nids.
1604 Severity : enhancement
1606 Description: eliminate client locks in face of contention
1607 Details : file contention detection and lockless i/o implementation
1608 for contended files.
1612 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1615 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1617 Description: MDS may lose file striping (and hence file data) in some cases
1618 Details : If there are additional extended attributes stored on the MDS,
1619 in particular ACLs, SELinux, or user attributes (if user_xattr
1620 is specified for the client mount options) then there is a risk
1621 of attribute loss. Additionally, the Lustre file striping
1622 needs to be larger than default (e.g. striped over all OSTs),
1623 and an additional attribute must be stored initially in the
1624 inode and then increase in size enough to be moved to the
1625 external attribute block (e.g. ACL growing in size) for file
1630 Description: skiplist implementation simplification
1631 Details : skiplists are used to group compatible locks on granted list
1632 that was implemented as tracking first and last lock of each
1633 lock group the patch changes that to using doubly linked lists
1637 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1638 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1639 because LASSERT on both the data supplied by a client, and the
1640 data on disk is dangerous and incorrect.
1642 Severity : enhancement
1644 Description: Slow truncate/writes to huge files at high offsets.
1645 Details : Directly associate cached pages to lock that protect those pages,
1646 this allows us to quickly find what pages to write and remove
1647 once lock callback is received.
1651 Description: more ldlm soft lockups
1652 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1653 starve other threads from the resource lock for a long time in
1654 case of long waiting queue, so change the debug level from
1655 D_OTHER to the less frequently used D_INFO.
1657 Severity : enhancement
1659 Description: add -gid, -group, -uid, -user options to lfs find
1663 Description: Hung threads in invalidate_inode_pages2_range
1664 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1665 once one is completed. As a result, some RPCs are stuck in the
1666 queue and are never sent.
1670 Description: filter threads hungs on waiting journal commit
1671 Details : Cleanup filter group llog code, then only filter group llog will
1672 be only created in the MDS/OST syncing process.
1676 Description: Procfs and llog threads access destoryed import sometimes.
1677 Details : Sync the import destoryed process with procfs and llog threads by
1678 the import refcount and semaphore.
1680 Severity : enhancement
1682 Description: openlock cache of b1_6 port to HEAD
1687 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1688 Details : If insertion of an extent fails, then discard the inode
1689 preallocation and free data blocks else it can lead to duplicate
1694 Description: don't always update ctime in ext3_xattr_set_handle()
1695 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1696 In some cases the ctime should not be updated, for example for
1697 2.0->1.8 compatibility it is necessary to delete an xattr and it
1698 should not update the ctime.
1702 Bugzilla : 15713/16362
1703 Description: Assertion in iopen_connect_dentry in 1.6.3
1704 Details : looking up an inode via iopen with the wrong generation number can
1705 populate the dcache with a disconneced dentry while the inode
1706 number is in the process of being reallocated. This causes an
1707 assertion failure in iopen since the inode's dentry list contains
1708 both a connected and disconnected dentry.
1712 Description: assertion failure in ldlm_handle2lock()
1713 Details : fix a race between class_handle_unhash() and class_handle2object()
1714 introduced in lustre 1.6.5 by bug 13622.
1719 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1720 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1725 Description: oops in sd_iostats_seq_show()
1726 Details : unloading/reloading the scsi low level driver triggers a kernel
1727 bug when trying to access the sd iostat file.
1732 Description: Kernel panics during QLogic driver reload
1733 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1734 patch, causing memory corruption.
1739 Description: journal_dev option does not work in b1_6
1740 Details : pass mount option during pre-mount.
1742 Severity : enhancement
1744 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1745 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1746 extent information of a file. It can be used to map logical blocks
1747 in a file to physical blocks in the block device.
1751 Description: LDLM soft lockups - improvement
1752 Details : It is be possible to send the lock handle along with each read
1753 or write request because the client is already doing a lock match
1754 itself so there isn't any reason the OST should have to re-do that
1758 Frequency : only on Cray X2
1760 Description: X2 build failures
1761 Details : fix build failures on Cray X2.
1765 Description: xid & resent requests
1766 Details : Initialize RPC XID from clock at startup (randomly if clock is
1769 Severity : enhancement
1771 Description: Add lustre_start utility to start or stop multiple Lustre servers
1776 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1777 Details : In case of memory pressure, list_del() can be called twice on
1778 req->rq_history_list, causing a kernel oops.
1782 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1783 Details : fix stack overflow in the distributed lock manager by defering
1784 export eviction after a failed AST to the elt thread instead of
1785 handling it in the dlm interpret routine.
1789 Description: Convert some comments to new format.
1790 Details : Update documenting comments to match doxygen conventions.
1794 Description: Grammar fixes.
1795 Details : A couple of trivial spelling fixes.
1799 Description: OSD_COUNTERS-mandatory
1800 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1801 to simplify the code.
1805 Description: simplify cmm_device freeing logic.
1806 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1810 Description: Add lockdep support to dt_object_operations locking interface.
1811 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1812 parameter indicating lock ordering. Update mdd code to use new
1817 Description: Introduce failloc constants for lockless IO tests.
1818 Details : Add two new failloc constants to test lockless IO. Only one of
1819 them in implemented---another is checked in yet to be landed
1824 Description: Add lockdep support for inode mutex.
1825 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1826 in the situations where Lustre has to lock more than one inode
1831 Description: Add optional invariants checking support.
1832 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1833 configure switch. This macro is to be used for consistency and
1834 sanity checks that are too expensive to be left in `production'
1839 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1840 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1845 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1846 Details : Use spin_lock_nested() in (the only) situation where more than
1847 one ldlm_lock is locked simultaneously. Also, fix possible
1848 dead-lock in ldlm_lock_change_resource() by enforcing particular
1853 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1854 Details : Instead of specifying each ldlm_lock call-back through separate
1855 parameter, wrap them into struct ldlm_callback_suite.
1859 Description: Kill join_lru obd method and its callers.
1860 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1861 in memory---a requirement implemented through ->o_join_lru() obd
1862 method in HEAD. Remove this method and its users.
1866 Description: Add asynchronous ldlm ENQUEUE completion handler.
1867 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1868 case that stock ldlm_completion_ast() cannot handle as it waits
1869 until lock is granted. Introduce new ldlm_completion_ast_async()
1870 for this. Also comment ldlm_completion_ast().
1874 Description: ldlm_error <-> errno conversion.
1875 Details : Add functions to map (rather arbitrary) between LDLM error codes
1876 and standard errno values. CLIO needs this to prevent LDLM specific
1877 constants from escaping ldlm and osc.
1881 Description: Kill unused ldlm_handle2lock_ns() function.
1882 Details : Kill unused ldlm_handle2lock_ns() function.
1886 Description: Add lu_ref support to ldlm_lock
1887 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1888 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1889 and ldlm_resource. LDLM interface has to be changed, because code that
1890 releases a reference on a lock, has to "know" what reference this is.
1891 In the most frequent case
1893 lock = ldlm_handle2lock(handle);
1895 LDLM_LOCK_PUT(lock);
1897 no changes are required. When any other reference (received _not_ from
1898 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1899 instead of LDLM_LOCK_PUT().
1901 Arguably, changes are pervasive, and interface requires some discipline
1902 for proper use. On the other hand, it was very instrumental in finding
1903 a few leaked lock references.
1907 Description: Add ldlm_lock_addref_try().
1908 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1909 attempts to addref a lock that might be being canceled
1914 Description: Add ldlm_weigh_callback().
1915 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1916 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1917 hard-coded `number of pages' logic.
1921 Description: Add lockdep annotations to llog code.
1922 Details : Use appropriately tagged _nested() locking calls in the places
1923 where llog takes more than one ->lgh_lock lock.
1927 Description: Add loi_kms_set().
1928 Details : Wrap kms updates into a helper function.
1932 Description: Constify instances of struct lsm_operations.
1933 Details : Constify instances of struct lsm_operations.
1937 Description: lu_conf support.
1938 Details : On a server, a file system object is uniquely identified
1939 by a fid, which is sufficient to locate and load all object
1940 state (inode). On a client, on the other hand, more data are
1941 necessary instantiate an object. Change lu_object_find() and
1942 friends to take additional `lu_conf' argument describing object.
1943 Typically this includes layout information.
1947 Description: lu_context fixes.
1948 Details : Introduce new lu_context functions that are needed on the client
1949 side, where some system threads (ptlrpcd) are shared by multiple
1950 modules, and so cannot be stopped during module shutdown.
1954 Description: Add start and stop methods to lu_device_type_operations.
1955 Details : Introduce two new methods in lu_device_type_operations, that are
1956 invoked when first instance of a given type is created and last one
1957 is destroyed respectively. This is need by CLIO.
1961 Description: Add lu_ref support to struct lu_device.
1962 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1963 track leaked references.
1967 Description: Introduce lu_kmem_descr.
1968 Details : lu_kmem_descr and its companion interface allow to create
1969 and destroy a number of kmem caches at once.
1973 Description: Fix lu_object finalization race.
1974 Details : Fix a race between lu_object_find() finding an object and its
1975 concurrent finalization. This race is (most likely) not possible
1976 on the server, but might happen on the client.
1980 Description: Introduce lu_ref interface.
1981 Details : lu_ref is a debugging module allowing to track references to
1982 a given object. It is quite cpu expensive, and has to be
1983 explicitly enabled with --enable-lu_ref. See usage description
1988 Description: Factor lu_site procfs stats into a separate function.
1989 Details : Separate lu_site stats printing code into a separate function
1990 that can be reused on a client.
1994 Description: Constify instances of struct {lu,dt,md}_device_operations.
1995 Details : Constify instances of struct {lu,dt,md}_device_operations.
1999 Description: Introduce struct md_site and move meta-data specific parts of
2000 struct lu_site here.
2001 Details : Move md-specific fields out of struct lu_site into special struct
2002 md_site, so that lu_site can be used on a client.
2006 Description: Kill mdd_lov_destroy().
2007 Details : Remove unused mdd code.
2011 Description: Add st_block checking to multistat.c.
2012 Details : Add st_block checking to multistat.c.
2016 Description: Add lu_ref support to struct obd_device.
2017 Details : Add lu_ref tracking to obd_device.
2021 Description: Kill obd_set_fail_loc().
2022 Details : Remove unused code.
2026 Description: Add special type for ptlrpc_request interpret functions.
2027 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2028 there. Actual usage will be in CLIO.
2032 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2033 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2034 doesn't work with lockdep.
2038 Description: Add rwv.c test program.
2039 Details : New testing program exercising readv(2) and writev(2) (Qian).
2043 Description: Add sendfile.c test program.
2044 Details : New testing program exercising sendfile(2) (Jay).
2048 Description: Ratelimit a message that can be very frequent.
2049 Details : Ratelimit a memory allocation failure message that can
2054 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2055 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2059 Description: Liblustre build fixes.
2060 Details : Liblustre build fixes.
2064 Description: libcfs: add cfs_{need,cond}_resched() interface.
2065 Details : libcfs: add cfs_{need,cond}_resched() definition and
2066 implementations for Linux, NT, and liblustre.
2068 Severity : enhancement
2070 Description: More exported tunables for mballoc
2071 Details : Add support for tunable preallocation window and new tunables for
2072 large/small requests
2076 Description: Detect corruption of block bitmap and checking for preallocations
2077 Details : Checks validity of on-disk block bitmap. Also it does better
2078 checking of number of applied preallocations. When corruption is
2079 found, it turns filesystem readonly to prevent further corruptions.
2083 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2084 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2085 window to 0. If it does hit cache miss, and read-pattern is still
2086 stride-io mode, does not reset the stride window, but also does not
2087 increase the stride window length in this case.
2091 Frequency : only for big-endian servers
2092 Description: Check if system is big-endian while mounting fs with extents feature
2093 Details : Mounting a filesystem with extents feature will fail on big-endian
2094 systems since ext3-based ldiskfs is not supported on big-endian
2095 systems. This can be over-riden with "bigendian_extents" mount option.
2097 Severity : enhancement
2099 Description: The root squash functionality
2100 Details : A security feature, which is to prevent users from being able
2101 to mount lustre on their desktop, run as root, and delete
2102 all of the files in the filesystem. The goal is accomplished by
2103 remapping user id (UID) and group id (GID) of the root user to
2104 a UID and GID specified by the system administartor via Lustre
2105 configuration management server (MGS). The functionality also
2106 allows to specify sets of clients for which the remapping does
2111 Description: Excessive recovery window
2112 Details : With AT enabled, the recovery window can be excessively long (6000+
2113 seconds). To address this problem, we no longer use
2114 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2115 timeout no longer depends on the service time, it is set to
2116 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2117 time via pb_service_time.
2121 Description: Watchdog triggered on MDS failover
2122 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2123 the OSTs know that the MDS "UUID" can be reused for the same export
2124 from a different NID, so we do not need to wait for the export to be
2128 Frequency : rare, only if using MMP with Linux RAID
2130 Description: MMP doesn't work with Linux RAID
2131 Details : While using HA for Lustre servers with Linux RAID, it is possible
2132 that MMP will not detect multiple mounts. To make this work we
2133 need to unplug the device queue in RAID when the MMP block is being
2134 written. Also while reading the MMP block, we should read it from
2135 disk and not the cached one.
2137 Severity : enhancement
2139 Description: open file using fid
2140 Details : A file can be opened using just its fid, like
2141 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2144 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2147 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2148 Details : add proper locking for creation of jbd_4k slab cache
2152 Description: MMP check in ext3_remount() fails without displaying any error
2153 Details : When multiple mount protection fails during remount, proper error
2156 Severity : enhancement
2158 Description: Allow stripe size to be up to 4G-64k
2159 Details : Fix math logic to allow large stripe sizes.
2163 Description: add check for >8TB ldiskfs filesystems
2164 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2165 Don't allow >8TB ldiskfs filesystems to be mounted without
2166 force_over_8tb mount option
2171 Descriptoin: OSS complains "too many blocks" and crashes
2172 Details : In filter_commitrw_write():
2173 1. lquota_chkquota();
2174 2. obdo_from_inode();
2175 3. lquota_pending_commit();
2176 "2." has changed uid and gid in this case, then caused
2177 lquota_chkquota() and lquota_pending_commit() to use
2178 different uid and gid. That is the root of the bug.
2180 --------------------------------------------------------------------------------
2182 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2184 * Support for kernels:
2185 2.6.5-7.283 (SLES 9),
2186 2.6.9-55.EL (RHEL 4),
2187 2.6.16.46-0.14 (SLES 10),
2188 2.6.18.8 vanilla (kernel.org)
2189 * Client support for unpatched kernels:
2190 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2191 2.6.16 - 2.6.22 vanilla (kernel.org)
2192 * Due to recently discovered recovery problems, we do not recommend
2193 using patchless RHEL 4 clients with this or any earlier release.
2194 * Recommended e2fsprogs version: 1.39.cfs8
2195 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2196 * Starting with this release, the ldiskfs backing filesystem required
2197 by Lustre is now in its own package, lustre-ldiskfs. This package
2198 should be installed. It is versioned separately from Lustre and
2199 may be released separately in future.
2203 Description: block reactivating mgc import until all deactivates complete
2204 Details : Fix race when failing back MDT/MGS to itself (testing)
2207 Frequency : at statup only
2209 Description: mds_lov_synchronize race leads to various problems
2210 Details : simultaneous MDT->OST connections at startup can cause the
2211 sync to abort, leaving the OSC in a bad state.
2213 Severity : enhancement
2215 Description: add optional extra BUILD_VERSION info
2216 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2217 to override the lustre version.
2220 Frequency : 2.6.18 servers only
2222 Description: ll_kern_mount() doesn't release the module reference
2223 Details : The ldiskfs module reference count never drops down to 0
2224 because ll_kern_mount() doesn't release the module reference.
2229 Description: server LBUG when using old ost_num_threads parameter
2230 Details : Accept the old ost_num_threads parameter but warn that it
2231 is deprecated, and fix an off-by-one error that caused an LBUG.
2236 Description: Transient SCSI error results in persistent IO issue
2237 Details : iobuf->dr_error is not reinitialized to 0 between two
2241 Frequency : sometimes when underlying device returns I/O errors
2243 Description: OSTs not going read-only during write failures
2244 Details : OSTs are not remounted read-only when the journal commit threads
2245 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2246 instead of the ext3 wrappers.
2250 Description: poor connect scaling with increasing client count
2251 Details : Don't run filter_grant_sanity_check for more than 100 exports
2252 to improve scaling for large numbers of clients.
2255 Frequency : SLES10 only
2257 Description: sanity-quota.sh quotacheck failed: rc = -22
2258 Details : Quotas cannot be enabled on SLES10.
2261 Frequency : liblustre clients only
2263 Description: getdirentries does not give error when run on compute nodes
2264 Details : getdirentries does not fail when the size specified as an argument
2265 is too small to contain at least one entry
2267 Severity : enhancement
2269 Description: Add LNET router traceability for debug purposes
2270 Details : If a checksum failure occurs with a router as part of the
2271 IO path, the NID of the last router that forwarded the bulk data
2272 is printed so it can be identified.
2277 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2278 Details : Due to a race condition, liblustre clients were occasionally
2279 evicted incorrectly.
2281 Severity : enhancement
2283 Description: lfs setstripe use optional parameters instead of postional
2286 Severity : enhancement
2288 Description: Nanosecond timestamp support for ldiskfs
2289 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2290 resolution timestamps. There is not yet support for this at
2291 the Lustre filesystem level.
2294 Frequency : during server recovery
2296 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2297 Details : request with rq_no_resend flag not awake l_wait_event if they get
2301 Frequency : nfs export on patchless client
2303 Description: connectathon hang when test nfs export over patchless client
2304 Details : Disconnected dentry cannot be found with lookup, so we do not need
2305 to unhash it or make it invalid
2308 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2309 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2310 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2315 Description: open req refcounting wrong on reconnect
2316 Details : If reconnect happened between getting open reply from server and
2317 call to mdc_set_replay_data in ll_file_open, we will schedule
2318 replay for unreferenced request that we are about to free.
2319 Subsequent close will crash in variety of ways.
2320 Check that request is still eligible for replay in
2321 mdc_set_replay_data().
2326 Description: disable writes to filesystem when reading health_check file
2327 Details : the default for reading the health_check proc file has changed
2328 to NOT do a journal transaction and write to disk, because this
2329 can cause reads of the /proc file to hang and block HA state
2330 checking on a healthy but otherwise heavily loaded system. It
2331 is possible to return to the previous behaviour during configure
2332 with --enable-health-write.
2334 Severity : enhancement
2336 Description: 64-bit inode version
2337 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2338 to the inode. This will be required for version-based recovery.
2343 Description: MDS fails to start if a duplicate client export is detected
2344 Details : in some rare cases it was possible for a client to connect to
2345 an MDS multiple times. Upon recovery the MDS would detect this
2346 and fail during startup. Handle this more gracefully.
2348 Severity : enhancement
2350 Description: Add -o localflock option to simulate old noflock
2352 Details : This will achieve local-only flock/fcntl locks
2358 Description: log_commit_thread vs filter_destroy race leads to crash
2359 Details : Take import reference before releasing llog record semaphore
2364 Description: Wrong request locking in request set processing
2365 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2366 imp_delayed_list, in this place should be used imp_lock.
2369 Frequency : when reconnection
2371 Description: Grant Leak when osc reconnect to OST
2372 Details : When osc reconnect ost, OST(filter) should check whether it
2373 should grant more space to client by comparing fed_grant and
2374 cl_avail_grant, and return the granted space to client instead
2375 of "new granted" space, because client will call osc_init_grant
2376 to update the client grant space info.
2379 Frequency : when client reconnect to OST
2381 Description: Grant Leak when osc do resend and replay bulk write
2382 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2383 bulk write request, because the grant info while be sync between
2384 OSC and OST when reconnect, and we should ignore the grant info
2385 these of resend/replay write req.
2390 Description: Grant space more than avaiable left space sometimes.
2391 Details : When then OST is about to be full, if two bulk writing from
2392 different clients came to OST. Accord the avaliable space of the
2393 OST, the first req should be permitted, and the second one
2394 should be denied by ENOSPC. But if the seconde arrived before
2395 the first one is commited. The OST might wrongly permit second
2396 writing, which will cause grant space > avaiable space.
2399 Frequency : when client is evicted
2401 Description: Grant might be wrongly erased when osc is evicted by OST
2402 Details : when the import is evicted by server, it will fork another
2403 thread ptlrpc_invalidate_import_thread to invalidate the
2404 import, where the grant will be set to 0. While the original
2405 thread will update the grant it got when connecting. So if
2406 the former happened latter, the grant will be wrongly errased
2407 because of this race.
2412 Description: Checking Stale with correct fid
2413 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2414 to check whether it is stale, because sometimes, we want the
2415 enqueue happened anyway, and op_data.fid2 will not be initialized.
2417 Severity : enhancement
2419 Description: update patchless client
2420 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2423 Frequency : only with 2.4 kernel
2425 Description: random memory corruption
2426 Details : size of struct ll_inode_info is to big for union inode.u and this
2427 can be cause of random memory corruption.
2432 Description: Memory leak in recovery
2433 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2434 It should also check obd_fail before fsfilt_start, otherwise if
2435 fsfilt_start return -EROFS,(failover mds during mds recovery).
2436 then the req will return with repmsg->transno = 0 and rc = EROFS.
2437 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2438 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2439 be freed no matter whether fsfilt_commit success or not.
2442 Frequency : only with huge count clients
2444 Description: Prevents from taking the superblock lock in llap_from_page for
2446 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2447 for llap_from_page call in ll_removepage prevents from taking the
2448 superblock lock for a soon died page.
2453 Description: Not check open intent error before release open handle
2454 Details : in some rare cases, the open intent error is not checked before
2455 release open handle, which may cause
2456 ASSERTION(open_req->rq_transno != 0), because it tries to release
2457 the failed open handle.
2462 Description: Set cat log bitmap only after create log success.
2463 Details : in some rare cases, the cat log bitmap is set too early. and it
2464 should be set only after create log success.
2468 Description: Accessing a block bevice can re-enable I/O when Lustre is
2469 tearing down a device.
2470 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2475 Description: service threads may hog cpus when there are a lot of requests
2476 Details : Insert cond_resched to give other threads a chance to use some CPU
2481 Description: the cat log was not initialized in recovery
2482 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2483 unlink log on mds will not be initialized until mds post
2484 recovery. And also in mds post recovery, the unlink log will
2485 initialization will be done asynchronausly, so there will be race
2486 between add unlink log and unlink log initialization.
2490 Description: brw_stats were being printed incorrectly
2491 Details : brw_stats were being printed as log2 but all of them were not
2492 recorded as log2. Also remove some code duplication arising from
2493 filter_tally_{read,write}.
2497 Frequency : rare, only in recovery.
2498 Description: ASSERTION(req->rq_type != LI_POISON) failed
2499 Details : imp_lock should be held while iterating over imp_sending_list for
2500 prevent destroy request after get timeout in ptlrpc_queue_wait.
2504 Description: replay-single.sh test 52 fails
2505 Details : A lock's skiplist need to be cleanup when it being unlinked
2506 from its resource list.
2510 Description: Short directio read returns full requested size rather than
2512 Details : Direct I/O operations should return actual amount of bytes
2513 transferred rather than requested size.
2515 Severity : enhancement
2517 Description: metadata RPC reduction (e.g. for rm performance)
2518 Details : decrease the amount of synchronous RPC between clients and servers
2519 by canceling conflicing lock before the operation on the client
2520 and packing thier handles into the main operation RPC to server.
2522 Severity : enhancement
2524 Description: Async OSC create to avoid the blocking unnecessarily.
2525 Details : If a OST has no remain object, system will block on the creating
2526 when need to create a new object on this OST. Now, ways use
2527 pre-created objects when available, instead of blocking on an
2528 empty osc while others are not empty. If we must block, we block
2529 for the shortest possible period of time.
2533 Description: improve handling recoverable errors
2534 Details : if request processig with error which can be recoverable on server
2535 request should be resend, otherwise page released from cache and
2538 Severity : enhancement
2540 Description: refine locking for avoid write wrong info into lov_objid file
2541 Details : fix possible races with add new target and write/update data in
2544 --------------------------------------------------------------------------------
2546 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2551 Frequency : on some architectures
2553 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2554 Details : Uninitialized flags sometimes cause configuration commands to
2558 Frequency : patchless clients only
2560 Description: missing __iget() symbol export
2561 Details : The __iget() symbol export is missing. To avoid the need for
2562 this on patchless clients the deathrow inode reaper is turned
2563 off, and we depend on the VM to clean up old inodes. This
2564 dependency was during via the fix for bug 12181.
2566 --------------------------------------------------------------------------------
2568 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2570 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2571 INTEROPERATE with older versions automatically. In many cases a
2572 special upgrade step is needed. Please read the
2573 user documentation before upgrading any part of a live system.
2574 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2575 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2576 * WARNING: Lustre configuration and startup changes are required with
2577 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2581 Severity : enhancement
2583 Description: MountConf
2584 Details : Lustre configuration is now managed via mkfs and mount
2585 commands instead of lmc and lconf. New obd types (MGS, MGC)
2586 are added for dynamic configuration management. See
2587 https://mail.clusterfs.com/wikis/lustre/MountConf for
2590 Severity : enhancement
2592 Description: dynamic OST addition
2593 Details : OSTs can now be added to a live filesystem
2595 Severity : enhancement
2597 Description: startup order invariance
2598 Details : MDTs and OSTs can be started in any order. Clients only
2599 require the MDT to complete startup.
2601 Severity : enhancement
2603 Description: parallel, asynchronous orphan cleanup
2604 Details : orphan cleanup is now performed in separate threads for each
2605 OST, allowing parallel non-blocking operation.
2607 Severity : enhancement
2609 Description: optimized stripe assignment
2610 Details : stripe assignments are now made based on ost space available,
2611 ost previous usage, and OSS previous usage, in order to try
2612 to optimize storage space and networking resources.
2614 Severity : enhancement
2616 Description: Permanently set tunables
2617 Details : All writable /proc/fs/lustre tunables can now be permanently
2618 set on a per-server basis, at mkfs time or on a live system.
2620 Severity : enhancement
2622 Description: Lustre message v2
2623 Details : Add lustre message format v2.
2625 Severity : enhancement
2627 Description: client OST exclusion list
2628 Details : Clients can be started with a list of OSTs that should be
2629 declared "inactive" for known non-responsive OSTs.
2633 Description: SPEC SFS validation failure on NFS v2 over lustre.
2634 Details : Changes the blocksize for regular files to be 2x RPC size,
2635 and not depend on stripe size.
2637 Severity : enhancement
2639 Description: Multiple MD RPCs in flight.
2640 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2641 To avoid overly-overloading MDS, introduce a limit on number of
2642 MDS RPCs in flight for a single client and add /proc controls
2643 to adjust this limit.
2645 Severity : enhancement
2647 Description: client read/write statistics
2648 Details : Add client read/write call usage stats for performance
2649 analysis of user processes.
2650 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2651 file access. extents_stats shows chunk size distribution.
2652 extents_stats_per_process show chunk size distribution per
2655 Severity : enhancement
2657 Description: mds statistics
2658 Details : Add detailed mds operations statistics in
2659 /proc/fs/lustre/mds/*/stats.
2663 Description: Failure of copying files with lustre special EAs.
2664 Details : Client side always return success for setxattr call for lustre
2665 special xattr (currently only "trusted.lov").
2670 Description: Refcount LNET uuids
2671 Details : The global LNET uuid list grew linearly with every startup;
2672 refcount repeated list entries instead of always adding to
2675 Severity : enhancement
2677 Description: Dynamic service threads
2678 Details : Within a small range, start extra service threads
2679 automatically when the request queue builds up.
2682 Frequency : mixed-endian client/server environments
2684 Description: mixed-endian crashes
2685 Details : The new msg_v2 system had some failures in mixed-endian
2688 Severity : enhancement
2690 Description: Easy OST removal
2691 Details : OSTs can be permanently deactivated with e.g. 'lctl
2692 conf_param lustre-OST0001.osc.active=0'
2694 Severity : enhancement
2696 Description: MGS proc entries
2697 Details : Added basic proc entries for the MGS showing what filesystems
2700 Severity : enhancement
2702 Description: provide MGS failover
2703 Details : Added config lock reacquisition after MGS server failover.
2705 Severity : enhancement
2707 Description: add Linux 2.4 support
2708 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2712 Description: a large application tries to do I/O to the same resource and dies
2713 in the middle of it.
2714 Details : Check the req->rq_arrival time after the call to
2715 ost_brw_lock_get(), but before we do anything about
2716 processing it & sending the BULK transfer request. This
2717 should help move old stale pending locks off the queue as
2718 quickly as obd_timeout.
2721 Frequency : when an incorrect nid is specified during startup
2723 Description: ptlrpc connect to non-existant node causes kernel crash
2724 Details : LNET can't be re-entered from an event callback, which
2725 happened when we expire a message after the export has been
2726 cleaned up. Instead, hand the zombie cleanup off to another
2729 Severity : enhancement
2731 Description: plain/inodebits lock performance improvement
2732 Details : Grouping plain/inodebits in granted list by their request modes
2733 and bits policy, thus improving the performance of search through
2737 Frequency : only if OST filesystem is corrupted
2739 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2740 Details : for a short time RPCs with bulk IO are in the replay list,
2741 but replay of bulk IOs is unimplemented. If the OST filesystem
2742 is corrupted due to disk cache incoherency and then replay is
2743 started it is possible to trip an assertion. Avoid putting
2744 committed RPCs into the replay list at all to avoid this issue.
2747 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2749 Description: System hang on startup
2750 Details : This bug allowed the liblustre (e.g. catamount) client to
2751 return to the app before handling all startup RPCs. This
2752 could leave the node unresponsive to lustre network traffic
2753 and manifested as a server ptllnd timeout.
2755 Severity : enhancement
2757 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2758 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2759 boolean to control whether to print peer debug info when a
2760 client's RPC times out.
2763 Frequency : only for kernels with patches from Lustre below 1.4.3
2765 Description: Remove old rdonly API
2766 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2769 Frequency : only for devices with external journals
2771 Description: Set external device read-only also
2772 Details : During a commanded failover stop, we set the disk device
2773 read-only while the server shuts down. We now also set any
2774 external journal device read-only at the same time.
2777 Frequency : when upgrading from 1.4 while trying to change parameters
2779 Description: The wrong (new) MDC name was used when setting parameters for
2780 upgraded MDT's. Also allows changing of OSC (and MDC)
2781 parameters if --writeconf is specified at tunefs upgrade time.
2784 Frequency : when setting specific ost indicies
2786 Description: QOS code breaks on skipped indicies
2787 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2788 created with --index need not be sequential.
2790 Severity : enhancement
2792 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2793 Details : The uninit_groups feature works in conjunction with the kernel
2794 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2795 up the pass1 processing of e2fsck. This is a read-only feature
2796 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2797 mount filesystems that have had this feature enabled.
2799 Severity : enhancement
2801 Description: Improve multi-block allocation algorithm to avoid fragmentation
2802 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2803 allocation locality and avoid filesystem fragmentation.
2805 ------------------------------------------------------------------------------
2807 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2809 * Support for kernels:
2810 2.4.21-47.0.1.EL (RHEL 3)
2811 2.6.5-7.283 (SLES 9)
2812 2.6.9-42.0.10.EL (RHEL 4)
2813 2.6.12.6 vanilla (kernel.org)
2814 2.6.16.27-0.9 (SLES 10)
2815 * Recommended e2fsprogs version: 1.39.cfs5
2817 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2821 Frequency : occasional, depends on client load and configuration
2822 Bugzilla : 12181, 12203
2823 Description: data loss for recently-modified files
2825 Details : In some cases it is possible that recently written or created
2826 files may not be written to disk in a timely manner (this should
2827 normally be within 30s unless client IO load is very high).
2828 The problem appears as zero-length files or files that are a
2829 multiple of 1MB in size after a client crash or client eviction
2830 that are missing data at the end of the file.
2832 This problem is more likely to be hit on clients where files are
2833 repeatedly created and unlinked in the same directory, clients
2834 have a large amount of RAM, have many CPUs, the filesystem has
2835 many OSTs, the clients are rebooted frequently, and/or the files
2836 are not accessed by other nodes after being written.
2838 The presence of the problem can be detected by looking at
2839 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2840 smaller than the second (nr_unused) then dirty files will not
2841 be flushed automatically to disk. "sync; sleep 10" should be
2842 run several times on the node before unmounting it to update
2843 Lustre (this is also safe to run on nodes without this problem).
2845 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2846 that can cause this same problem, so customers using that kernel
2847 also need to update the kernel in addition to Lustre. In order
2848 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2850 It is normal that files written just before a client crash (less
2851 than 30s) may not yet have been flushed to disk, even for local
2855 Frequency : frequent on thin XT3 nodes
2857 Description: UUID collision on thin XT3 Linux nodes
2858 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2859 randomly, since we relied on an insufficiently-seeded PRNG.
2864 Description: OSS hangs after "All ost request buffers busy"
2865 Details : A deadlock between quota and journal operations caused OSS
2866 hangs after printing "All ost request buffers busy."
2869 Frequency : always on liblustre builds
2871 Description: Cleanup compiler warnings on liblustre
2874 Frequency : always on liblustre builds on XT3
2876 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2880 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2881 > 32000 subdirectories
2882 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2883 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2884 subdirectory count crosses 32000. This will aid e2fsck to
2885 correctly handle more than 32000 subdirectories.
2888 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2890 Description: System hang on startup
2891 Details : This bug allowed the liblustre (e.g. catamount) client to
2892 return to the app before handling all startup RPCs. This
2893 could leave the node unresponsive to lustre network traffic
2894 and manifested as a server ptllnd timeout.
2896 Severity : enhancement
2898 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2899 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2900 boolean to control whether to print peer debug info when a
2901 client's RPC times out.
2906 Description: make O_SYNC working on 2.6 kernels
2907 Details : 2.6 kernels use different method for mark pages for write,
2908 so need add a code to lustre for O_SYNC work.
2913 Description: Failure to close file and release space on NFS
2914 Details : Put inode details into lock acquired in ll_intent_file_open.
2915 Use mdc_intent_lock in ll_intent_open to properly
2916 detect all kind of errors unhandled by mdc_enqueue.
2921 Description: proc file read during shutdown sometimes raced obd removal,
2923 Details : Add lock to prevent obd access after proc file removal.
2926 Frequency : Only for files larger than 4GB on 32-bit clients.
2928 Description: improperly doing page alignment of locks
2929 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2930 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2935 Description: RPCs being resent when they shouldn't be
2936 Details : Some RPCs that should not be resent are being resent. This
2937 can cause inconsistencies in the RPC state machine. Do not
2938 resend such requests.
2941 Frequency : rare, only with NFS export
2943 Description: Crash on NFS re-export node
2944 Details : under very unusual load conditions an assertion is hit in
2945 ll_intent_file_open()
2948 Frequency : only if OST filesystem is corrupted
2950 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2951 Details : for a short time RPCs with bulk IO are in the replay list,
2952 but replay of bulk IOs is unimplemented. If the OST filesystem
2953 is corrupted due to disk cache incoherency and then replay is
2954 started it is possible to trip an assertion. Avoid putting
2955 committed RPCs into the replay list at all to avoid this issue.
2960 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2961 Details : Large single O_DIRECT read and write calls can fail to allocate
2962 a sufficiently large buffer to process the request. In case of
2963 allocation failure the allocation is retried with a smaller
2964 buffer and broken into smaller requests.
2966 Severity : enhancement
2968 Description: Add -o localflock option to simulate old noflock behaviour.
2969 Details : This will achieve local-only flock/fcntl locks coherentness.
2974 Description: versioning check is incomplete
2975 Details : Checking the version difference of client vs. server, report
2976 error if the gap is too big.
2981 Description: add support PG_writeback bit
2982 Details : add support for PG_writeback bit for Lustre, for more carefull
2983 work with page cache in 2.6 kernel. This also fix some deadlocks
2984 and remove hack for work O_SYNC with 2.6 kernel.
2986 Severity : enhancement
2988 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2989 Details : The uninit_groups feature works in conjunction with the kernel
2990 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2991 up the pass1 processing of e2fsck. This is a read-only feature
2992 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2993 mount filesystems that have had this feature enabled.
2995 Severity : enhancement
2997 Description: Improve multi-block allocation algorithm to avoid fragmentation
2998 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2999 allocation locality and avoid filesystem fragmentation.
3001 ------------------------------------------------------------------------------
3003 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3005 * Support for kernels:
3006 2.6.9-42.0.3.EL (RHEL 4)
3007 2.6.5-7.276 (SLES 9)
3008 2.4.21-47.0.1.EL (RHEL 3)
3009 2.6.12.6 vanilla (kernel.org)
3010 2.6.16.21-0.8 (SLES10)
3011 * Recommended e2fsprogs version: 1.39.cfs2-0
3013 * The backwards-compatible /proc/sys/portals symlink has been removed
3014 in this release. Before upgrading, please ensure that you change
3015 any configuration scripts or /etc/sysctl.conf files that access
3016 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3017 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3018 in advance of the upgrade on any system running Lustre 1.4.6 or
3019 newer, since /proc/sys/lnet was added in that version.
3020 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3024 Frequency : only when quota is used
3026 Description: avoid scanning export list for quota master
3027 Details : Change the algorithms to avoid scanning export list in order
3028 to improve the efficiency.
3031 Frequency : MDS failover only, very rarely
3033 Description: "went back in time" messages on mds failover
3034 Details : The greatest transno may be lost when the current operation
3035 finishes with an error (transno==0) and the client's last_rcvd
3036 record is over-written. Save the greatest transno in the
3037 mds_last_transno for this case.
3040 Frequency : always for specific kernels and striping counts
3042 Description: client may get "Matching packet too big" without ACL support
3043 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3044 when trying to access files in certain configurations. The
3045 clients should in fact be denied when mounting because they do
3046 not understand ACLs.
3049 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3051 Description: many clients connecting with IO in progress causes connect timeouts
3052 Details : Avoid synchronous journal commits to avoid delays caused by many
3053 clients connecting/disconnecting when bulk IO is in progress.
3054 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3055 OST_IO_PORTAL to avoid delays behind potentially many pending
3059 Frequency : occasionally with multiple writers to a single file
3061 Description: shared writes to file may result in wrong size reported by stat()
3062 Details : Allow growing of kms when extent lock is cancelled
3065 Frequency : always with random mmap IO to multi-striped file
3067 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3068 Details : Only if the hole is at the end of OST object so that kms is too
3069 small. Fix is to increase kms accordingly in ll_nopage.
3072 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3074 Description: writes to a missing object would leak memory on the OST
3075 Details : If there is an inconsistency between the MDS and OST filesystems,
3076 such that the MDS references an object that doesn't exist, writes
3077 to that object will leak memory due to incorrect cleanup in the
3078 error handling path, eventually running out of memory on the OST.
3083 Description: Creating too long symlink causes lustre errors
3084 Details : Check symlink and name lengths before sending requests to MDS.
3087 Frequency : only if flock is enabled (not on by default)
3089 Description: posix locks not released on fd closure on 2.6.9+
3090 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3091 kernels, this caused such locks not to be released on fd close and
3092 then assertions on fs unmount about still used locks.
3095 Frequency : MDS failover only, very rarely
3097 Description: clients may get ASSERTION(granted_lock != NULL)
3098 Details : When request was taking a long time, and a client was resending
3099 a getattr by name lock request. The were multiple lock requests
3100 with the same client lock handle and
3101 mds_getattr_name->fixup_handle_for_resent_request found one of the
3102 lock handles but later failed with ASSERTION(granted_lock != NULL).
3107 Description: handle->h_buffer_credits > 0, assertion failure
3108 Details : h_buffer_credits is zero after truncate, causing assertion
3109 failure. This patch extends the transaction or creates a new
3113 Frequency : NFS re-export or patchless client
3114 Bugzilla : 11179, 10796
3115 Description: Crash on NFS re-export node (__d_move)
3116 Details : We do not want to hash the dentry if we don't have a lock.
3117 But if this dentry is later used in d_move, we'd hit uninitialised
3118 list head d_hash, so we just do this to init d_hash field but
3119 leave dentry unhashed.
3122 Frequency : NFS re-export or patchless client
3124 Description: NFS exports has problem with symbolic link
3125 Details : lustre client didn't properly install dentry when re-exported
3126 to NFS or running patchless client.
3129 Frequency : NFS re-export or patchless client
3131 Description: Various nfs/patchless fixes.
3132 Details : fixes reuse disconected alias for lookup process - this fixes
3133 warning "find_exported_dentry: npd != pd",
3134 fix permission error with open files at nfs.
3135 fix apply umask when do revalidate.
3138 Frequency : occasional
3140 Description: Crash on NFS re-export node
3141 Details : calling clear_page() on the wrong pointer triggered oops in
3142 generic_mapping_read().
3145 Frequency : rarely, using O_DIRECT IO
3147 Description: unaligned directio crashes client with LASSERT
3148 Details : check for unaligned buffers before trying any requests.
3151 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3153 Description: stale data returned from RAID cache
3154 Details : If only a small amount of IO is done to the RAID device before
3155 reading it again it is possible to get stale data from the RAID
3156 cache instead of reading it from disk.
3159 Frequency : always for sles10 kernel
3161 Description: sles10 support
3162 Details : ll_follow_link: compile fixes and using of nd_set_link
3163 under newer kernels.
3166 Frequency : depends on arch, kernel and compiler version, always on sles10
3169 Description: recursive or deep enough symlinks cause stack overflow
3170 Details : getting rid of large stack-allocated variable in
3174 Frequency : depends on hardware
3176 Description: lustre write performance loss in the SLES10 kernel
3177 Details : the performance loss is caused by using of write barriers in the
3178 ext3 code. The SLES10 kernel turns barrier support on by
3179 default. The fix is to undo that change for ldiskfs.
3181 ------------------------------------------------------------------------------
3183 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3185 * Support for kernels:
3186 2.6.9-42.0.3EL (RHEL 4)
3187 2.6.5-7.276 (SLES 9)
3188 2.4.21-47.0.1.EL (RHEL 3)
3189 2.6.12.6 vanilla (kernel.org)
3193 Frequency : quota enabled and large files being deleted
3195 Description: releasing more than 4GB of quota at once hangs OST
3196 Details : If a user deletes more than 4GB of files on a single OST it
3197 will cause the OST to spin in an infinite loop. Release
3198 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3203 Description: statfs data retrieved from /proc may be stale or zero
3204 Details : When reading per-device statfs data from /proc, in the
3205 {kbytes,files}_{total,free,avail} files, it may appear
3206 as zero or be out of date.
3209 Frequency : systems with MD RAID1 external journal devices
3211 Description: lconf's call to blkid is confused by RAID1 journal devices
3212 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3213 of returning all block devices that match the journal UUID.
3216 Frequency : always, for aggregate stripe size over 4GB
3218 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3219 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3220 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3223 Frequency : always if "lfs find" used on a local file/directory
3225 Description: "lfs find" segfaults if used on a local file/directory
3226 Details : The case where a directory component was not specified wasn't
3227 handled correctly. Handle this properly.
3230 Frequency : always on ppc64
3232 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3233 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3234 perform endianness conversion.
3237 Frequency : rarely (truncate to non-zero file size after write under load)
3238 Bugzilla : 10730, 10687
3239 Description: Files padded with zeros to next 4K multiple
3240 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3241 it is possible that files that are truncated to a non-zero size
3242 immediately after being written are filled with zero bytes beyond
3243 the truncated size. No file data is lost.
3245 Severity : enhancement
3247 Description: Allow recovery/failover for liblustre clients.
3248 Details : liblustre clients were unaware of failover configurations until
3251 Severity : enhancement
3253 Description: user file locks should fail when not mounting with flock option
3254 Details : Set up an error-returning stub in ll_file_operations.lock field
3255 to prevent incorrect behaviour when client is mounted without
3256 flock option. Also, set up properly f_op->flock field for
3260 Frequency : always on ia64
3262 Description: "lfs df" loops on printing out MDS statfs information
3263 Details : The obd_ioctl_data was not initialized and in some systems
3264 this caused a failure during the ioctl that did not return
3265 an error. Initialize the struct and return an error on failure.
3268 Frequency : SLES 9 only
3270 Description: Error of copying files with lustre special EAs as root
3271 Details : Client side always return success for setxattr call for lustre
3272 special xattr (currently only "trusted.lov").
3275 Frequency : rarely on clusters with both ia64+i386 clients
3277 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3278 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3279 clients are doing concurrent writes to the same file it is
3280 possible that the ia64 clients may LASSERT because the OST
3281 extent locks are not PAGE_SIZE aligned. Ensure that grown
3282 locks are always aligned on the request boundary.
3285 Frequency : specific use, occasional
3287 Description: Overwriting in use executable truncates on-disk binary image
3288 Details : If one node attempts to overwrite an executable in use by
3289 another node, we now correctly return ETXTBSY instead of
3290 truncating the file.
3292 Severity : enhancement
3294 Description: Async OSC create to avoid the blocking unnecessarily.
3295 Details : If a OST has no remain object, system will block on the creating
3296 when need to create a new object on this OST. Now, ways use
3297 pre-created objects when available, instead of blocking on an
3298 empty osc while others are not empty. If we must block, we block
3299 for the shortest possible period of time.
3304 Description: chmod on Lustre root is propagated to other clients
3305 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3306 invalid by the follow_mount time.
3309 Frequency : liblustre clients only
3311 Description: Race in 'instant cancel' lock handling could lead to such locks
3312 never to be granted in case of SMP MDS
3313 Details : Do not destroy not yet granted but cbpending locks in
3317 Frequency : replay/resend of open
3319 Description: non null lock assetion failure in mds_intent_policy
3320 Details : Trying to replay/resend lockless open requests resulted in
3321 mds_open() returning 0 with no lock. Now it sets a flag if
3322 a lock is going to be returned.
3324 Severity : enhancement
3326 Description: Checksum enhancements
3327 Details : New checksum enhancements allow for resending RPCs that failed
3330 Severity : enhancement
3332 Description: Tunables on number of dirty pages in cacche
3333 Details : Allow to set limit on number of dirty pages cached.
3338 Description: client crash on unmount - lock still has references
3339 Details : In some error handling cases it was possible to leak a lock
3340 reference on a client while accessing a file. This was not
3341 harmful to the client during operation, but would cause the
3342 client to crash when the filesystem is unmounted.
3345 Frequency : specific case, rare
3347 Description: ETXTBSY on mds though file not in use by client
3348 Details : ETXTBSY is no longer incorrectly returned when attempting to
3349 chmod or chown a directory that the user previously tried to
3350 execute or a currently-executing binary.
3353 Frequency : extremely rare except on liblustre-based clients
3355 Description: Lustre space not freed when files are deleted
3356 Details : Clean up open-unlinked files after client eviction. Previously
3357 the unlink was skipped and the files remained as orphans.
3362 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3363 Details : In some cases it was possible to send a blocking callback to
3364 a client doing a glimpse, even though that client didn't get
3365 a lock granted. When the glimpse lock is cancelled on the OST
3366 the freed lock is left on the waiting list and corrupted the list.
3369 Frequency : all core dumps
3371 Description: Broke core dumps to lustre
3372 Details : Negative dentry may be unhashed if parent does not have UPDATE
3373 lock, but some callers, e.g. do_coredump, expect dentry to be
3374 hashed after successful create, hash it in ll_create_it.
3376 ------------------------------------------------------------------------------
3378 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3380 * Support for kernels:
3381 2.6.9-42.0.2.EL (RHEL 4)
3382 2.6.5-7.276 (SLES 9)
3383 2.4.21-40.EL (RHEL 3)
3384 2.6.12.6 vanilla (kernel.org)
3388 Frequency : always on RHEL 3
3390 Description: Number of open files grows over time
3391 Details : The number of open files grows over time, whether or not
3392 Lustre is started. This was due to a filp leak introduced
3393 by one of our kernel patches.
3395 ------------------------------------------------------------------------------
3397 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3399 * Support for kernels:
3400 2.6.9-42.EL (RHEL 4)
3401 2.6.5-7.276 (SLES 9)
3402 2.4.21-40.EL (RHEL 3)
3403 2.6.12.6 vanilla (kernel.org)
3408 Bugzilla : 5719, 9635, 9792, 9684
3409 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3410 Details : If a server is under heavy load and cannot reply to new
3411 connection requests before the client resends the (re)connect,
3412 the connection handling code can behave badly if two service
3413 threads are concurrently handing separate (re)connections from
3414 the same client. Add better locking to the connection handling
3415 code, and ensure that only a single connection will be processed
3416 for a given client UUID, even if the lock is dropped.
3418 Severity : enhancement
3420 Description: add TCP zero-copy support to kernel
3421 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3422 sends if the hardware supports scatter-gather and checksumming.
3423 This allows socklnd to do client-write and server-read more
3424 efficiently and reduce CPU utilization from skbuf copying.
3427 Frequency : only if NFS exporting from client
3429 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3430 Details : If setting attributes on a file created under NFS that had
3431 never been opened it would be possible to oops the client
3432 if the file had no objects.
3435 Frequency : always for liblustre
3437 Description: liblustre client does MDS+OSTs setattr RPC for each write
3438 Details : When doing a write from a liblustre client, the client
3439 incorrectly issued an RPC to the MDS and each OST the file was
3440 striped over in order to update the timestamps. When writing
3441 with small chunks and many clients this could overwhelm the MDS
3442 with RPCs. In all cases it would slow down the write because
3443 these RPCs are unnecessary.
3445 Severity : enhancement
3447 Description: allow number of MDS service threads to be changed at module load
3448 Details : It is now possible to change the number of MDS service threads
3449 running. Adding "options mds mds_num_threads={N}" to the MDS's
3450 /etc/modprobe.conf will set the number of threads for the next
3451 time Lustre is restarted (assuming the "mds" module is also
3452 reloaded at that time). The default number of threads will
3453 stay the same, 32 for most systems.
3458 Description: OST crash if filesystem is unformatted or corrupt
3459 Details : If an OST is started on a device that has never been formatted
3460 or if the filesystem is corrupt and cannot even mount then the
3461 error handling cleanup routines would dereference a NULL pointer.
3466 Description: NULL pointer deref in llap_from_page.
3467 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3468 incorrect) mapping if the page was truncated/reclaimed while it was
3469 searched for. Check for this condition and skip such pages when
3470 doing readahead. Introduce extra check to llap_from_page() to
3471 verify page->mapping->host is non-NULL (so page is not anonymous).
3474 Frequency : Sometimes when using sys_sendfile
3476 Description: "page not covered by a lock" warnings from ll_readpage
3477 Details : sendfile called ll_readpage without right page locks present.
3478 Now we introduced ll_file_sendfile that does necessary locking
3479 around call to generic_file_sendfile() much like we do in
3483 Frequency : with certain MDS communication failures at client mount time
3485 Description: NULL pointer deref after failed client mount
3486 Details : a client connection request may delayed by the network layer
3487 and not be sent until after the PTLRPC layer has timed out the
3488 request. If the client fails the mount immediately it will try
3489 to clean up before the network times out the request. Add a
3490 reference from the request import to the obd device and delay
3491 the cleanup until the network drops the request.
3494 Frequency : occasionally during client (re)connect
3496 Description: assertion failure during client (re)connect
3497 Details : processing a client connection request may be delayed by the
3498 client or server longer than the client connect timeout. This
3499 causes the client to resend the connection request. If the
3500 original connection request is replied in this interval, the
3501 client may trip an assertion failure in ptlrpc_connect_interpret()
3502 which thought it would be the only running connect process.
3505 Frequency : only with obd_echo servers and clients that are rebooted
3507 Description: kernel BUG accessing uninitialized data structure
3508 Details : When running an obd_echo server it did not start the ping_evictor
3509 thread, and when a client was evicted an uninitialized data
3510 structure was accessed. Start the ping_evictor in the RPC
3511 service startup instead of the OBD startup.
3513 Severity : enhancement
3514 Bugzilla : 10193 (patchless)
3515 Description: Remove dependency on various unexported kernel interfaces.
3516 Details : No longer need reparent_to_init, exit_mm, exit_files,
3517 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3520 Frequency : rare (only users of deprecated and unsupported LDAP config)
3522 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3523 Details : LDAP apparently contains 'lustreName' attributes instead of
3524 'name'. A simple remapping of the name is sufficient.
3527 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3529 Description: waiting_locks_callback trips kernel BUG if client is evicted
3530 Details : Running with the dump_on_timeout debug flag turned on makes
3531 it possible that the waiting_locks_callback() can try to dump
3532 the Lustre kernel debug logs from an interrupt handler. Defer
3533 this log dumping to the expired_lock_main() thread.
3535 Severity : enhancement
3537 Description: Support NFS exporting on 2.6 kernels.
3538 Details : Implement non-rawops metadata methods for NFS server to use without
3539 changing NFS server code.
3542 Frequency : very rare (synthetic metadata workload only)
3544 Description: two racing renames might cause an MDS thread to deadlock
3545 Details : Running the "racer" program may cause one MDS thread to rename
3546 a file from being the source of a rename to being the target of
3547 a rename at exactly the same time that another thread is doing
3548 so, and the second thread has already enqueued these locks after
3549 doing a lookup of the target and is trying to relock them in
3550 order. Ensure that we don't try to re-lock the same resource.
3553 Frequency : only very large systems with liblustre clients
3555 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3556 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3557 all OSTs in parallel. This allows the eviction of stale liblustre
3558 clients to proceed much faster than if they were done in series,
3559 and also offers similar improvements for other set_info RPCs.
3564 Description: excessive CPU usage during initial read phase on client
3565 Details : During the initial read phase on a client, it would agressively
3566 retry readahead on the file, consuming too much CPU and impacting
3567 performance (since 1.4.5.8). Improve the readahead algorithm
3568 to avoid this, and also improve some other common cases (read
3569 of small files in particular, where "small" is files smaller than
3570 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3575 Description: MDS crash when receiving packet with unknown intent.
3576 Details : Do not LBUG in unknown intent case, just return -EFAULT
3578 Severity : enhancement
3579 Bugzilla : 9293, 9385
3580 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3581 Details : Do not serialize getattr (non-intent version) and statfs.
3584 Frequency : occasional, when OST network is overloaded/intermittent
3586 Description: client evicted by OST after bulk IO timeout
3587 Details : If a client sends a bulk IO request (read or write) the OST
3588 may evict the client if it is unresposive to its data GET/PUT
3589 request. This is incorrect if the network is overloaded (takes
3590 too long to transfer the RPC data) or dropped the OST GET/PUT
3591 request. There is no need to evict the client at all, since
3592 the pinger and/or lock callbacks will handle this, and the
3593 client can restart the bulk request.
3596 Frequency : Always when mmapping file with no objects
3598 Description: client crashes when mmapping file with no objects
3599 Details : Check that we actually have objects in a file before doing any
3600 operations on objects in ll_vm_open, ll_vm_close and
3606 Description: Request leak when working with deleted CWD
3607 Details : Introduce advanced request refcount tracking for requests
3608 referenced from lustre intent.
3610 Severity : Enhancement
3612 Description: Cache open file handles on client.
3613 Details : MDS now will return special lock along with openhandle, if
3614 requested and client is allowed to hold openhandle, even if unused,
3615 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3616 closing files for every read/write openration.
3618 Severity : Enhancement
3620 Description: Cache open negative dentries on client when possible.
3621 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3622 negative dentries on lock revocation.
3627 Description: Remounting a client read-only wasn't possible with a zconf mount
3628 Details : It wasn't possible to remount a client read-only with llmount.
3630 Severity : enhancement
3631 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3632 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3633 MPICH is included to add Lustre-specific ADIO interfaces.
3634 This is based closely on the UFS ADIO layer and only differs
3635 in file creation, in order to allow the OST striping to be set.
3636 This is user-contributed code and not supported by CFS.
3641 Description: extended inode attributes (immutable, append-only) work improperly
3642 when 2.4 and 2.6 kernels are used on client/server or vice versa
3643 Details : Introduce kernel-independent values for these flags.
3645 Severity : enhancement
3648 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3649 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3650 granularity. Now, Fractional MB granularity is be supported,
3651 this is very useful for low memory system.
3653 Severity : enhancement
3655 Description: Getattr by fid
3656 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3657 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3658 have locks and avoids getattr rpc requests for it on every stat.
3661 Frequency : Always, for filesystems larger than 2TB
3663 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3664 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3665 filesystems larger than 2TB to be mounted with mballoc enabled.
3668 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3670 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3671 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3672 filesystems larger than 2TB is mounted then the kernel will
3673 silently corrupt the start of the filesystem. CONFIG_LBD is
3674 enabled for all CFS-supported kernels, but the possibility of
3675 this happening with a modified kernel config exists.
3677 Severity : enhancement
3679 Description: add client O_DIRECT support for 2.6 kernels
3680 Details : It is now possible to do O_DIRECT reads and writes to files
3681 in the Lustre client mountpoint on 2.6 kernel clients.
3683 Severity : enhancement
3685 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3686 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3687 parallel, not waiting for response from every OST before sending
3688 a rpc to the next OST.
3693 Description: setattr vs write race when updating file timestamps
3694 Details : Client processes that update a file timestamp into the past
3695 right after writing to the file (e.g. tar) it is possible that
3696 the updated file modification time can be reset to the current
3697 time due to a race between processing the setattr and write RPC.
3699 Severity : enhancement
3701 Description: Bring 'lfs find' closer in line with regular Linux find.
3702 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3703 -print0 options and obtains all the needed info through the lustre
3706 Severity : enhancement
3708 Description: support up to 1024 configured devices on one node
3709 Details : change obd_dev array from statically allocated to dynamically
3710 allocated structs as they are first used to reduce memory usage
3715 Description: Flush dirty partially truncated pages during truncate
3716 Details : Immediatelly flush partially truncated pages in filter_setattr,
3717 this way we completely avoid having any pages in page cache on OST
3718 and can retire ugly workarounds during writes to flush such pages.
3723 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3724 Details : protect inode from truncation within vfs_unlink() context
3725 just take a reference before calling vfs_unlink() and release it
3726 when parent's i_sem is free.
3731 Description: last_id value checked outside lock on OST caused LASSERT failure
3732 Details : If there were multiple MDS->OST object precreate requests in
3733 flight, it was possible that the OST's last object id was checked
3734 outside a lock and incorrectly tripped an assertion. Move checks
3735 inside locks, and discard old precreate requests.
3738 Frequency : always, if extents are used on OSTs
3740 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3741 Details : OSTs using the extents format would not zero the high 16 bits of
3742 the index physical block number. This is not a problem for any
3743 OST filesystems smaller than 16TB, and no kernels support ext3
3744 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3745 new/modified files) and can be fixed for existing filesystems
3746 with e2fsprogs-1.39-cfs1.
3751 Description: import connection selection may be incorrect if timer wraps
3752 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3753 import connections to be ignored if the 32-bit jiffies counter
3754 wraps. Use a 64-bit jiffies counter.
3757 Frequency : very large clusters immediately after boot
3759 Description: LNET request buffers exhausted under heavy short-term load
3760 Details : If a large number of client requests are generated on a service
3761 that has previously never seen so many requests it is possible
3762 that the request buffer growth cannot keep up with the spike in
3763 demand. Instead of dropping incoming requests, they are held in
3764 the LND until the RPC service can accept more requests.
3767 Frequency : Sometimes during replay
3769 Description: Assertion failure in ll_local_open after replay.
3770 Details : If replay happened on an open request reply before we were able
3771 to set replay handler, reply will become not swabbed tripping the
3772 assertion in ll_local_open. Now we set the handler right after
3773 recognising of open request
3776 Frequency : very rare
3778 Description: kernel reports "badness in vsnprintf"
3779 Details : Reading from the "recovery_status" /proc file in small chunks
3780 may cause a negative length in lprocfs_obd_rd_recovery_status()
3781 call to vsnprintf() (which is otherwise harmless). Exit early
3782 if there is no more space in the output buffer.
3784 Severity : enhancement
3786 Description: clear OBD RPC statistics by writing to them
3787 Details : It is now possible to clear the OBD RPC statistics by writing
3788 to the "stats" file.
3793 Description: Client mtime is not the same on different clients after utimes
3794 Details : In some cases, the client was using the utimes() syscall on
3795 a file cached on another node. The clients now validate the
3796 ctime from the MDS + OSTs to determine which one is right.
3801 Description: Inability to activate failout mode
3802 Details : lconf script incorrectly assumed that in pythong string's numeric
3803 value is used in comparisons.
3806 Frequency : always with multiple stripes per file
3808 Description: Inefficient object allocation for mutli-stripe files
3809 Details : When selecting which OSTs to stripe files over, for files with
3810 a stripe count that divides evenly into the number of OSTs,
3811 the MDS is always picking the same starting OST for each file.
3812 Return the OST selection heuristic to the original design.
3817 Description: mount failures may take full timeout to return an error
3818 Details : Under some heavy load conditions it is possible that a
3819 failed mount can wait for the full obd_timeout interval,
3820 possibly several minutes, before reporting an error.
3821 Instead return an error as soon as the status is known.
3823 Frequency : quota enabled and large files being deleted
3825 Description: releasing more than 4GB of quota at once hangs OST
3826 Details : If a user deletes more than 4GB of files on a single OST it
3827 will cause the OST to spin in an infinite loop. Release
3828 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3833 Description: statfs data retrieved from /proc may be stale or zero
3834 Details : When reading per-device statfs data from /proc, in the
3835 {kbytes,files}_{total,free,avail} files, it may appear
3836 as zero or be out of date.
3839 Frequency : systems with MD RAID1 external journal devices
3841 Description: lconf's call to blkid is confused by RAID1 journal devices
3842 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3843 of returning all block devices that match the journal UUID.
3846 Frequency : always, for aggregate stripe size over 4GB
3848 Description: assertion fails when trying to use 4GB stripe size
3849 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3850 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3853 Frequency : always on ppc64
3855 Description: the first write on an ext3 filesystem with mballoc got stuck
3856 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3857 perform endianness conversion.
3859 ------------------------------------------------------------------------------
3861 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3863 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3864 INTEROPERATE with older versions automatically. Please read the
3865 user documentation before upgrading any part of a live system.
3866 * WARNING: Lustre networking configuration changes are required with
3867 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3870 * Support for kernels:
3871 2.6.9-22.0.2.EL (RHEL 4)
3872 2.6.5-7.244 (SLES 9)
3873 2.6.12.6 vanilla (kernel.org)
3876 Severity : enhancement
3877 Bugzilla : 7981/8208
3878 Description: Introduced Lustre Networking (LNET)
3879 Details : LNET is new networking infrastructure for Lustre, it includes
3880 a reorganized network configuration mode (see the user
3881 documentation for full details) as well as support for routing
3882 between different network fabrics. Lustre Networking Devices
3883 (LNDs) for the supported network fabrics have also been
3884 created for this new infrastructure.
3886 Severity : enhancement
3887 Description: Introduced Access control lists
3888 Details : clients can set ACLs on files and directories in order to have
3889 more fine-grained permissions than the standard Unix UGO+RWX.
3890 The MDS must be started with the "-o acl" mount option.
3892 Severity : enhancement
3893 Description: Introduced filesystem quotas
3894 Details : Administrators may now establish per-user quotas on the
3897 Severity : enhancement
3899 Description: Configuration change for the XT3
3900 The PTLLND is now used to run Lustre over Portals on the XT3
3901 The configure option(s) --with-cray-portals are no longer used.
3902 Rather --with-portals=<path-to-portals-includes> is used to
3903 enable building on the XT3. In addition to enable XT3 specific
3904 features the option --enable-cray-xt3 must be used.
3909 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3910 Details : A race between allocating a new llcd and re-getting the llcd_lock
3911 allowed another thread to grab newly-allocated llcd.
3913 Severity : enhancement
3915 Description: 2.6 OST async journal commit and locking fix to improve performance
3916 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3917 2.6 kernels are now async as they already were in 2.4 kernels so
3918 that they can commit concurrently with the network bulk transfer.
3919 For block-allocated files the filter allocation semaphore is held
3920 to avoid filesystem fragmentation during allocation. BKL lock
3921 removed for 2.6 xattr operations where it is no longer needed.
3926 Description: lconf incorrectly determined whether two IP networks could talk
3927 Details : In some more complicated routing and multiple-network
3928 configurations, lconf will avoid trying to make a network
3929 connection to a disjoint part of the IP space. It was doing the
3930 math incorrectly for one set of cases.
3935 Description: Fix for potential infinite loop processing records in an llog.
3936 Details : If an llog record is corrupted/zeroed, it is possible to loop
3937 forever in llog_process(). Validate the llog record length
3938 and skip the remainder of the block on error.
3941 Frequency : occasional (liblustre only)
3943 Description: liblustre could not open files whose last component is a symlink
3944 Details : sysio_path_walk() would incorrectly pass the open intent to
3945 intermediate path components.
3948 Frequency : rare (liblustre only with non-standard tuning)
3949 Bugzilla : 7201 (7350)
3950 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3951 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3952 after it has been released, only do this for referenced locks
3954 Severity : enhancement
3956 Description: specifying an (invalid) directory default stripe_size of -1
3957 would reset the directory default striping
3958 Details : stripe_size -1 was used internally to signal directory stripe
3959 removal, now use "all default" to signal dir stripe removal
3960 as a directory striping of "all default" is not useful
3963 Frequency : common for large clusters running liblustre clients
3965 Description: doing an ls when liblustre clients are running is slow
3966 Details : sending a glimpse AST to a liblustre client waits for every AST
3967 to time out, as liblustre clients will not respond. Since they
3968 cannot cache data we refresh the OST lock LVB from disk instead.
3970 Severity : enhancement
3972 Description: doing an ls at the same time as file IO can be slow
3973 Details : enqueue and other "small" requests can be blocked behind many
3974 large IO requests. Create a new OST IO portal for non-IO
3975 requests so they can be processed faster.
3978 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3980 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3981 created file with mode 000
3982 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3983 allow an ftruncate() as a non-root user to a file with mode 000.
3984 Lustre now respects this flag to disable mode checks when
3985 truncating a file owned by the user
3988 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3991 Description: Revoking locks from clients that went dead or catatonic might take
3993 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3994 cancellation of such locks instant on servers without waiting for
3995 any reply from clients. Clients drops these locks when cancel
3996 notification from server is received without replying.
3999 Frequency : liblustre-only, when liblustre client dies or becomes busy
4001 Description: Doing ls on Linux clients can take a long time with active
4003 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4004 granting such locks to it in the first place if possible. Locks
4005 are taken by proxy on the OST during the read or write and
4006 dropped immediately afterward. Add connect flags handling, do
4007 not grant locks to liblustre clients for glimpse ASTs.
4009 Severity : enhancement
4011 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4012 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4013 sequential reads and behaves badly with applications doing random
4014 reads. Improve it by reading ahead at least read region, and
4015 avoiding excessive large RPC for small reads.
4017 Severity : enhancement
4019 Description: Creating more than 1000 files for a single job may cause a load
4020 imbalance on the OSTs if there are also a large number of OSTs.
4021 Details : qos_prep_create() uses an OST index reseed value that is an
4022 even multiple of the number of available OSTs so that if the
4023 reseed happens in the middle of the object allocation it will
4024 still utilize the OSTs as uniformly as possible.
4029 Description: OST or MDS may oops in ping_evictor_main()
4030 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4031 but doesn't restart at beginning of obd_exports_timed list
4034 Severity : enhancement
4036 Description: improve by-nid export eviction on the MDS and OST
4037 Details : allow multiple exports with the same NID to be evicted at one
4038 time without re-searching the exports list.
4041 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4043 Description: MDS may oops in groups_free()
4044 Details : in rare race conditions a newly allocated group_info struct is
4045 freed again, and this can be NULL. The 2.4 compatibility code
4046 for groups_free() checked for a NULL pointer, but 2.6 did not.
4049 Frequency : common for liblustre clients doing little filesystem IO
4050 Bugzilla : 9352, 7313
4051 Description: server may evict liblustre clients accessing contended locks
4052 Details : if a client is granted a lock or receives a completion AST
4053 with a blocking AST already set it would not reply to the AST
4054 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4057 Frequency : lfs setstripe, only systems with more than 160 OSTs
4059 Description: unable to set striping with a starting offset beyond OST 160
4060 Details : llapi_create_file() incorrectly limited the starting stripe
4061 index to the maximum single-file stripe count.
4064 Frequency : LDAP users only
4066 Description: lconf did not handle in-kernel recovery with LDAP properly
4067 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4069 Severity : enhancement
4071 Description: bind OST threads to NUMA nodes to improve performance
4072 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4073 node and do their allocations there to localize memory access
4075 Severity : enhancement
4077 Description: llmount can determine client NID directly from Myrinet (GM)
4078 Details : the client NID code from gmnalnid was moved directly into
4079 llmount, removing the need to use this or specifying the
4080 client NID explicitly when mounting GM clients with zeroconf
4083 Frequency : if client is started with down MDS
4085 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4086 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4087 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4088 ptlrpc_import_delayed_req() because we didn't check if the
4089 request was interrupted, and we also didn't break out of the
4090 event loop if there was no timeout
4095 Description: data loss during non-page-aligned writes to a single file from
4096 both multiple nodes and multiple threads on one node at same time
4097 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4098 inconsistency led to false short-reads, that were cached and later
4099 used by ->prepare_write() to fill in partially written page,
4100 leading to data loss.
4103 Frequency : always, if lconf --abort_recovery used
4105 Description: lconf --abort_recovery fails with 'Operation not supported'
4106 Details : lconf was attempting to abort recovery on the MDT device and not
4109 Severity : enhancement
4111 Description: remove cleanup logs
4112 Details : replace lconf-generated cleanup logs with lustre internal
4113 cleanup routines. Eliminates the need for client-cleanup and
4116 Severity : enhancement
4118 Description: add support for EAs (user and system) on lustre filesystems
4119 Details : it is now possible to store extended attributes in the Lustre
4120 client filesystem, and with the user_xattr mount option it
4121 is possible to allow users to store EAs on their files also
4123 Severity : enhancement
4125 Description: Add possibility (config option) to show minimal available OST free
4127 Details : When compiled with --enable-mindf configure option, statfs(2)
4128 (and so, df) will return least minimal free space available from
4129 all OSTs as amount of free space on FS, instead of summary of
4130 free spaces of all OSTs.
4132 Severity : enhancement
4134 Description: do not expand extent locks acquired on OST-side
4135 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4136 by server: they are not cached anyway.
4139 Frequency : when mmap is used/binaries executed from Lustre
4141 Description: Unmmap pages before throwing them away from read cache.
4142 Details : llap_shrink cache now attempts to unmap pages before discarding
4143 them (if unmapping failed - do not discard). SLES9 kernel has
4144 extra checks that trigger if this unmapping is not done first.
4149 Description: lconf didn't resolve symlinks before checking to see whether a
4150 given mountpoint was already in use
4153 Frequency : when migrating failover services
4154 Bugzilla : 6395, 9514
4155 Description: When migrating a subset of services from a node (e.g. failback
4156 from a failover service node) the remaining services would
4157 time out and evict clients.
4158 Details : lconf --force (implied by --failover) sets the global obd_timeout
4159 to 5 seconds in order to quickly disconnect, but this caused
4160 other RPCs to time out too quickly. Do not change the global
4161 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4163 Severity : enhancement
4164 Frequency : if MDS is started with down OST
4165 Bugzilla : 9439,5706
4166 Description: Allow startup/shutdown of an MDS without depending on the
4167 availability of the OSTs.
4168 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4169 Add appropriate locking and lov-osc refcounts for safe
4170 cleaning. Add osc abort_inflight calls in case the
4171 synchronize never started.
4174 Frequency : occasional (Cray XT3 only)
4176 Description: root not authorized to access files in CRAY_PORTALS environment
4177 Details : The client process capabilities were not honoured on the MDS in
4178 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4179 been accessed by an authorized user then root was able to access
4180 the file on the local client also. The root user capabilities
4181 are now allowed on the MDS, as this environment has secure UID.
4184 Frequency : occasional
4186 Description: ldiskfs "too long searching" message happens too often
4187 Details : A debugging message (otherwise harmless) prints too often on
4188 the OST console. This has been reduced to only happen when
4189 there are fragmentation problems on the filesystem.
4194 Description: Division by zero in statfs when all OSCs are inactive
4195 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4196 lov_statfs divides by value returned by lov_get_stripecnt().
4200 Bugzilla : 9489, 3273
4201 Description: First write from each client to each OST was only 4kB in size,
4202 to initialize client writeback cache, which caused sub-optimal
4203 RPCs and poor layout on disk for the first writen file.
4204 Details : Clients now request an initial cache grant at (re)connect time
4205 and so that they can start streaming writes to the cache right
4206 away and always do full-sized RPCs if there is enough data.
4207 If the OST is rebooted the client also re-establishes its grant
4208 so that client cached writes will be honoured under the grant.
4213 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4214 Details : Now I/O RPCs go to different portal number and (presumably) fast
4215 lock requests (and glimses) and other RPCs get their own service
4216 threads pool that should be able to service those RPCs
4219 Severity : enhancement
4221 Description: Ability to exchange lustre version between client and servers and
4222 issue warnings at client side if client is too old. Also for
4223 liblustre clients there is ability to refuse connection of too old
4225 Details : New 'version' field is added to connect data structure that is
4226 filled with version info. That info is later checked by server and
4230 Frequency : rare, liblustre only.
4231 Bugzilla : 9296, 9581
4232 Description: Two simultaneous writes from liblustre at offset within same page
4233 might proceed at the same time overwriting eachother with stale
4235 Details : I/O lock withing llu_file_prwv was released too early, before data
4236 actually was hitting the wire. Extended lock-holding time until
4237 server acknowledges receiving data.
4240 Frequency : extremely rare. Never observed in practice.
4242 Description: avoid generating lustre_handle cookie of 0.
4243 Details : class_handle_hash() generates handle cookies by incrementing
4244 global counter, and can hit 0 occasionaly (this is unlikely, but
4245 not impossible, because initial value of cookie counter is
4246 selected randonly). Value of 0 is used as a sentinel meaning
4247 "unassigned handle" --- avoid it. Also coalesce two critical
4248 sections in this function into one.
4250 Severity : enhancement
4252 Description: allow liblustre clients to delegate truncate locking to OST
4253 Details : To avoid overhead of locking, liblustre client instructs OST to
4254 take extent lock in ost_punch() on client's behalf. New connection
4255 flag is added to handle backward compatibility.
4257 Severity : enhancement
4258 Bugzilla : 4928, 7341, 9758
4259 Description: allow number of OST service threads to be specified
4260 Details : a module parameter allows the number of OST service threads
4261 to be specified via "options ost ost_num_threads={N}" in the
4262 OSS's /etc/modules.conf or /etc/modprobe.conf.
4266 Bugzilla : 6146, 9635, 9895
4267 Description: servers crash with bad pointer in target_handle_connect()
4268 Details : In rare cases when a client is reconnecting it was possible that
4269 the connection request was the last reference for that export.
4270 We would temporarily drop the export reference and get a new
4271 one, but this may have been the last reference and the export
4272 was just destroyed. Get new reference before dropping old one.
4274 Severity : enhancement
4275 Frequency : if client is started with failover MDS
4277 Description: Allow multiple MDS hostnames in the mount command
4278 Details : Try to read the configuration from all specified MDS
4279 hostnames during a client mount in case the "primary"
4282 Severity : enhancement
4284 Description: Stop sending data to evicted clients as soon as possible.
4285 Details : Check if the client we are about to send or are sending data to
4286 was evicted already. (Check is done every second of waiting,
4287 for which l_wait_event interface was extended to allow checking
4288 of exit condition at specified intervals).
4291 Frequency : rare, normally only when NFS exporting is done from client
4293 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4295 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4298 Frequency : always, if srand() is called before liblustre initialization
4300 Description: Liblustre uses system PRNG disturbing its usage by user application
4301 Details : Introduce internal to lustre fast and high-quality PRNG for
4302 lustre usage and make liblustre and some other places in generic
4303 lustre code to use it.
4305 Severity : enhancement
4306 Bugzilla : 9477, 9557, 9870
4307 Description: Verify that the MDS configuration logs are updated when xml is
4308 Details : Check if the .xml configuration logs are newer than the config
4309 logs stored on the MDS and report an error if this is the case.
4310 Request --write-conf, or allow starting with --old_conf.
4312 Severity : enhancement
4314 Description: Handle symlinks in the path when checking if Lustre is mounted.
4315 Details : Resolve intermediate symlinks when checking if a client has
4316 mounted a filesystem to avoid duplicate client mounts.
4321 Description: lconf can hit an error exception but still return success.
4322 Details : The lconf command catches the Command error exception at the top
4323 level script context and will exit with the associated exit
4324 status, but doesn't ensure that this exit status is non-zero.
4329 Description: failure of ptlrpc thread startup can cause oops
4330 Details : Starting a ptlrpc service thread can fail if there are a large
4331 number of threads or the server memory is very fragmented.
4332 Handle this without oopsing.
4335 Frequency : always, only if liblustre and non-default acceptor port was used
4337 Description: liblustre cannot connect to servers with non-default acceptor port
4338 Details : tcpnal_set_default_params() was not called and was therefore
4339 ignoring the environment varaible TCPNAL_PORT, as well as other
4340 TCPNAL_ environment variables
4345 Description: two objects could be created on the same OST for a single file
4346 Details : If an OST is down, in some cases it was possible to create two
4347 objects on a single OST for a single file. No problems other
4348 than potential performance impact and spurious error messages.
4352 Bugzilla : 5681, 9562
4353 Description: Client may oops in ll_unhash_aliases
4354 Details : Client dcache may become inconsistent in race condition.
4355 In some cases "getcwd" can fail if the current directory is
4361 Description: Inode refcounting problems in NFS export code
4362 Details : link_raw functions used to call d_instantiate without obtaining
4363 extra inode reference first.
4367 Bugzilla : 9942, 9903
4368 Description: Referencing freed requests leading to crash, memleaks with NFS.
4369 Details : We used to require that call to ll_revalidate_it was always
4370 followed by ll_lookup_it. Also with revalidate_special() it is
4371 possible to call ll_revalidate_it() twice for the same dentry
4372 even if first occurence returned success. This fix changes semantic
4373 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4374 reference on a request referred from the intent.
4375 ll_intent_release() then releases such a request.
4378 Frequency : rare, normally benchmark loads only
4380 Description: unlinked inodes were kept in memory on the client
4381 Details : If a client is repeatedly creating and unlinking files it
4382 can accumulate a lot of stale inodes in the inode slab cache.
4383 If there is no other client load running this can cause the
4384 client node to run out of memory. Instead flush old inodes
4385 from client cache that have the same inode number as a new inode.
4388 Frequency : SLES9 2.6.5 kernel and long filenames only
4389 Bugzilla : 9969, 10379
4390 Description: utime reports stale NFS file handle
4391 Details : SLES9 uses out-of-dentry names in some cases, which confused
4392 the lustre dentry revalidation. Change it to always use the
4396 Frequency : rare, unless heavy write-truncate concurrency is continuous
4397 Bugzilla : 4180, 6984, 7171, 9963, 9331
4398 Description: OST becomes very slow and/or deadlocked during object unlink
4399 Details : filter_destroy() was holding onto the parent directory lock
4400 while truncating+unlinking objects. For very large objects this
4401 may block other threads for a long time and slow overall OST
4402 responsiveness. It may also be possible to get a lock ordering
4403 deadlock in this case, or run out of journal credits because of
4404 the combined truncate+unlink. Solution is to do object truncate
4405 first in one transaction without parent lock, and then do the
4406 final unlink in a new transaction with the parent lock. This
4407 reduces the lock hold time dramatically.
4410 Frequency : rare, 2.4 kernels only
4412 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4413 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4414 before unmounting their filesystem in order to avoid deadlock.
4415 The kernel_locked() function in 2.4 kernels only checks whether
4416 the kernel lock is held, not whether it is this process that is
4417 holding it as 2.6 kernels do.
4422 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4423 Details : The client ptlrpc code may be trying to reconnect to a down
4424 server before a previous connection attempt has timed out.
4425 Increase the reconnect interval to be longer than the connection
4426 timeout interval to avoid sending duplicate connections to
4430 Frequency : echo_client brw_test command
4432 Description: fix echo_client to work with OST preallocated code
4433 Details : OST preallocation code (5137) didn't take echo_client IO path
4434 into account: echo_client calls filter methods outside of any
4435 OST thread and, hence, there is no per-thread preallocated
4436 pages and buffers to use. Solution: hijack pga pages for IO. As
4437 a byproduct, this avoids unnecessary data copying.
4441 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4442 Description: Client can oops in mdc_commit_close() after open replay
4443 Details : It was possible for the MDS to return an open request with no
4444 transaction number in mds_finish_transno() if the client was
4445 evicted, but without actually returning an error. Clients
4446 would later try to replay that open and may trip an assertion
4447 Simplify the client close codepath, and always return an error
4448 from the MDS in case the open is not successful.
4451 Frequency : rare, 2.6 OSTs only
4453 Description: OST may deadlock under high load on fragmented files
4454 Details : If there was a heavy load and highly-fragmented OST filesystems
4455 it was possible to have all the OST threads deadlock waiting on
4456 allocation of biovecs, because the biovecs were not released
4457 until the entire RPC IO was completed. Instead, release biovecs
4458 as soon as they are complete to ensure forward IO progress.
4460 Severity : enhancement
4462 Description: Support for specifying external journal device at mount
4463 Details : If an OST or MDS device is formatted with an external journal
4464 device, this device major/minor is stored in the ext3 superblock
4465 and may not be valid for failover. Allow detecting and
4466 specifying the external journal at mount time.
4471 Description: Mounting an MDS with pending unlinked files may cause oops
4472 Details : target_finish_recovery() calls mds_postrecov() which returned
4473 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4474 considers this an error and immediately begins cleaning up the
4475 lov, just after starting the mds_lov process
4477 Severity : enhancement
4479 Description: Implement 'lfs df' to report actual free space on per-OST basis
4480 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4481 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4482 usage of objects; '-h' to report in human readable format.
4484 ------------------------------------------------------------------------------
4486 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4493 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4494 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4495 at startup during orphan unlinking. Instead, skip buddy update
4496 before setup, as it will be regenerated after recovery is complete.
4501 Description: If an OST is inactive, its locks might reference stale inodes.
4502 Details : lov_change_cbdata() must iterate over all namespaces, even if
4503 they are inactive to clear inode references from the lock.
4505 Severity : enhancement
4506 Frequency : occasional, if non-standard max_dirty_mb used
4508 Description: Client will block write RPCs if not enough grant
4509 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4510 then the client will block writes while waiting for another RPC
4511 to complete instead of consuming its dirty limit. With change
4512 we get improved performance when max_dirty_mb is small.
4514 Severity : enhancement
4515 Bugzilla : 3389, 6253
4516 Description: Add support for supplementary groups on the MDS.
4517 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4518 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4519 lookups for user supplementary groups into a cache.
4523 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4524 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4525 for liblustre O_EXCL is enforced by MDS.
4529 Description: readdir never returns NULL in liblustre.
4530 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4531 dentry in struct dirent.
4535 Frequency : liblustre only, depends on application IO pattern
4536 Description: liblustre clients evicted if not contacting servers
4537 Details : Don't put liblustre clients into the ping_evictor list, so
4538 they will not be evicted by the pinger ever.
4540 Severity : enhancement
4542 Description: Add ability to evict clients by NID from MDS.
4543 Details : By echoing "nid:$NID" string into
4544 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4545 $NID would be instantly evicted from this MDS and from all active
4546 OSTs connected to it.
4550 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4551 Details : lookup_it_finish() used to query file size from OSTs that was not
4556 Description: service threads change working directory to that of init
4557 Details : Starting lustre service threads may pin the working directory
4558 of the parent thread, making that filesystem busy. Threads
4559 now change to the working directory of init to avoid this.
4563 Frequency : during shutdown only
4564 Description: shutdown with a failed MDS or OST can cause unmount to hang
4565 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4570 Frequency : 2.6 only
4571 Description: chmod/chown may include an extra supplementary group
4572 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4573 supplementary group and if none is specified this is used.
4576 Bugzilla : 5479 (6816)
4578 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4579 Details : If lookup is in progress on a file that is unlinked we might try
4580 to revalidate the inode and fail in revalidate after lookup is
4581 complete and ll_file_open() enqueues the open again but
4582 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4585 Frequency : always, if lconf --abort_recovery used
4587 Description: lconf --abort_recovery fails with 'Operation not supported'
4588 Details : lconf was attempting to abort recovery on the MDT device and not
4591 ------------------------------------------------------------------------------
4593 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4598 Frequency : rare (only unsupported configurations with a node running as an
4600 Bugzilla : 6514, 5137
4601 Description: Mounting a Lustre file system on a node running as an OST could
4603 Details : OSTs now preallocates memory needed to write out data at
4604 startup, instead of when needed, to avoid having to
4605 allocate memory in possibly low memory situations.
4606 Specifically, if the file system is mounted on on OST,
4607 memory pressure could force it to try to write out data,
4608 which it needed to allocate memory to do. Due to the low
4609 memory, it would be unable to do so and the node would
4610 become unresponsive.
4612 Severity : enhancement
4614 Description: Addition of lconf --service command line option
4615 Details : lconf now accepts a '--service <arg>' option, which is
4616 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4618 Severity : enhancement
4620 Description: Failover mode is now the default for OSTs.
4621 Details : By default, OSTs will now run in failover mode. To return to
4622 the old behaviour, add '--failout' to the lmc line for OSTs.
4624 Severity : enhancement
4626 Description: Health checks are now provided for MDS and OSTs
4627 Details : Additional detailed health check information on MSD and OSTs
4628 is now provided through the procfs health_check value.
4631 Frequency : occasional, depends on IO load
4633 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4634 after numerous create/delete cycles
4635 Details : The ext3 inode allocation policy would not allocate new inodes
4636 very well on the OSTs because there are no new directories
4637 being created. Instead we look for groups with free space if
4638 the parent directories are nearly full.
4643 Description: Network or server problems during mount may cause partially
4644 mounted clients instead of returning an error.
4645 Details : The config llog parsing code may overwrite the error return
4646 code during mount error handling, returning success instead
4652 Description: MDS can fail to allocate large reply buffers
4653 Details : After long uptimes the MDS can fail to allocate large reply
4654 buffers (e.g. zconf client mount config records) due to memory
4655 fragmentation or consumption by the buffer cache. Preallocate
4656 some large reply buffers so that these replies can be sent even
4657 under memory pressure.
4661 Frequency : rare (liblustre)
4662 Description: fsx running with liblustre complained that using truncate() to
4663 extend the file doesn't work. This patch corrects that issue.
4664 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4665 ATTR_SIZE and lsm use in llu_setattr_raw.
4669 Frequency : rare, only 2.6 kernels
4670 Description: Unusual file access patterns on the MDS may result in inode
4671 data being lost in very rare circumstances.
4672 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4673 code in the 2.6 kernel caused the inode and/or EA data not to
4674 be read from disk, causing single-file corruption.
4678 Frequency : rare, only 2.6 filesystems using extents
4679 Description: Heavy concurrent write and delete load may cause data corruption.
4680 Details : It was possible under high-load situations to have an extent
4681 metadata block in the block device cache from a just-unlinked
4682 file overwrite a newly-allocated data block. We now unmap any
4683 metadata buffers that alias just-allocated data blocks.
4687 Frequency : filesystems with default stripe_count larger than 77
4688 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4689 Details : lconf specifies an inode size of 4096 bytes when the default
4690 stripe_count is larger than 77. This conflicts with the default
4691 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4692 this case to avoid pinning too much memory for large EAs.
4694 ------------------------------------------------------------------------------
4696 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4701 Frequency : rare (extremely heavy IO load with hundreds of clients)
4703 Description: Client is evicted, gets IO error writing to file
4704 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4705 caused clients to be evicted for AST timeouts. The fixes in
4706 bug 5192 mean we no longer need to have such short AST timeouts
4707 so ldlm_timeout has been increased.
4710 Frequency : occasional during --force or --failover shutdown under load
4711 Bugzilla : 5949, 4834
4712 Description: Server oops/LBUG if stopped with --force or --failover under load
4713 Details : a collection of import/export refcount and cleanup ordering
4714 issues fixed for safer force cleanup
4717 Frequency : only filesystems larger than 120 OSTs
4718 Bugzilla : 5990, 6223
4719 Description: lfs getstripe would oops on a very large filesystem
4720 Details : lov_getconfig used kfree on vmalloc'd memory
4723 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4724 Bugzilla : 6242, 6243
4725 Description: reading from files that had been truncated to a non-zero size
4726 but never opened returned no data
4727 Details : ll_file_read() reads zeros from no-object files to EOF
4732 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4733 Details : There's little harm in aborting MDS/OSS recovery and letting it
4734 try again, so I removed the LASSERT and return an error instead.
4736 Severity : enhancement
4738 Description: New debugging infrastructure for tracking down data corruption
4739 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4740 (b) cover more of the client-side code path, and (c) try to narrow
4741 down where problems occurred
4745 Bugzilla : 3819, 4364, 4397, 6313
4746 Description: Racing close and eviction MDS could cause assertion in mds_close
4747 Details : It was possible to get multiple mfd references during close and
4748 client eviction, leading to one thread referencing a freed mfd.
4750 Severity: : enhancement
4751 Bugzilla : 3262, 6359
4752 Description: Attempts to reconnect to servers are now more aggressive.
4753 Details : This builds on the enhanced upcall-less recovery that was added
4754 in 1.4.2. When trying to reconnect to servers, clients will
4755 now try each server in the failover group every 10 seconds. By
4756 default, clients would previously try one server every 25 seconds.
4761 Description: After recovery, certain operations trigger a failed
4762 assertion on a client.
4763 Details : Failing over an mds, using lconf -d --failover, while a
4764 client was doing a readdir() call would cause the client to
4765 LBUG after recovery completed and the readdir() was resent.
4767 Severity : enhancement
4769 Description: Default groups are now added by lconf
4770 Details : You can now run lconf --group <servicename> without having to
4771 manually add groups with lmc.
4774 Frequency : occasional
4776 Description: Nodes with an elan id of 0 trigger a failed assertion
4779 Frequency : always when accessing e.g. tty/console device nodes
4781 Description: tty and some other devices nodes cannot be used on lustre
4782 Details : file's private_data field is used by device data and lustre
4783 values in there got lost. New field was added to struct file to
4784 store fs-specific private data.
4787 Frequency : when exporting Lustre via NFS
4789 Description: NFSD failed occasionally when looking up a path component
4790 Details : NFSD is looking up ".." which was broken in ext3 directories
4791 that had grown large enough to become hashed.
4794 Frequency : Clusters with multiple interfaces not on the same subnet
4796 Description: Nodes will repeatedly try to reconnect to an interface which it
4797 cannot reach and report an error to the log.
4798 Details : Extra peer list entries will be created by lconf with some peers
4799 unreachable. lconf now validates the peer before adding it.
4802 Frequency : Only if a default stripe is set on the filesystem root.
4804 Description: Setting a default stripe on the filesystem root prevented the
4805 filesystem from being remounted.
4806 Details : The client was sending extra request flags in the root getattr
4807 request and did not allocate a reply buffer for the dir EA.
4810 Frequency : occasional, higher if lots of files are accessed by one client
4811 Bugzilla : 6159, 6097
4812 Description: Client trips assertion regarding lsm mismatch/magic
4813 Details : While revalidating inodes the VFS looks up inodes with ifind()
4814 and in rare cases can find an inode that is being freed.
4815 The ll_test_inode() code will free the lsm during ifind()
4816 when it finds an existing inode and then the VFS later attaches
4817 this free lsm to a new inode.
4821 Bugzilla : 6422, 7030
4822 Description: MDS deadlock between mkdir and client eviction
4823 Details : Creating a new file via mkdir or mknod (starting a transaction
4824 and getting the ns lock) can deadlock with client eviction
4825 (gets ns lock and trying to finish a synchronous transaction).
4828 Frequency : occasional
4829 Description: While starting a server, the fsfilt_ext3 module could not be
4831 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4832 kernels. Previously, lconf would still use the ext3 name
4833 when trying to load modules. Now, it will correctly use
4834 ext3 on 2.4 and ldiskfs on 2.6.
4836 Severity : enhancement
4837 Description: The default stripe count has been changed to 1
4838 Details : The interpretation of the default stripe count (0, to lfs
4839 or lmc) has been changed to mean striping across a single
4840 OST, rather than all available. For general usage we have
4841 found a stripe count of 1 or 2 works best.
4843 Severity : enhancement
4844 Description: Add support for compiling against Cray portals.
4845 Details : Conditional compiling for some areas that are different
4849 Frequency : occasional
4850 Bugzilla : 6409, 6834
4851 Description: Creating files with an explicit stripe count may lead to
4852 a failed assertion on the MDS
4853 Details : If some OSTs are full or unavailable, creating files may
4854 trigger a failed assertion on the MDS. Now, Lustre will
4855 try to use other servers or return an error to the
4859 Frequency : occasional
4861 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4862 caused lock timeout during readahead (since 1.4.2).
4863 Details : Processes doing readahead might match a lock that hasn't been
4864 granted yet if there are overlapping and conflicting lock
4865 requests. The readahead process waits on ungranted lock
4866 (original lock is CBPENDING), while OST waits for that process
4867 to cancel CBPENDING read lock and eventually evicts client.
4869 Severity : enhancement
4871 Description: Initial enabling of flock support for clients
4872 Details : Implements fcntl advisory locking and file status functions.
4873 This feature is provided as an optional mount flag (default
4874 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4875 locking are implemented yet, and those that are are not guaranteed
4876 to be completely correct in production environments.
4877 mount -t lustre -o [flock|noflock] ...
4880 Frequency : occasional
4882 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4883 assertion in the ext3 JBD (journaling) layer.
4884 Details : The b_committed_data struct is protected by the big kernel lock
4885 in 2.4 kernels, serializing journal_commit_transaction() and
4886 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4887 access to this struct. In 2.6 kernels there is finer grained
4888 locking to improve SMP performance of the JBD layer.
4892 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4897 Description: merge and cleanup kernel patches.
4900 -----------------------------------------------------------------------------
4902 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4904 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4905 versions of Lustre. Please update all servers and clients to
4906 version 1.4.2 or later at the same time. You must also run
4907 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4908 to update the configuration logs.
4910 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4911 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4912 - swab llog records of type '0' so we get proper header size/idx (5861)
4913 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4914 - fix rename of one directory over another leaking an inode (5953)
4915 - avoid SetPageDirty on 2.6 (5981)
4916 - don't re-add just-being-destroyed locks to the waiting list (5653)
4917 - when creating new directories, inherit the parent's custom
4918 striping settings if present parent (3048)
4919 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4920 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4921 - don't hold client locks on temporary worklist from l_lru (5666)
4922 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4923 - automatically evict dead clients (5921)
4924 - Update file size properly in create+truncate+fstat case (6196)
4925 - Do not unhash mountpoint dentries, do not allow removal of
4927 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4928 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4929 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4930 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4931 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4932 - Correctly handle reads of files with no objects (6243)
4933 - lctl recover will also mark a device active if deactivate used (5933)
4935 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4936 - allow --write-conf on an MDS with different nettype than client (5619)
4937 - don't write config llogs to MDS for mounts not from that MDS (5617)
4938 - lconf should create multiple TCP connections from a client (5201)
4939 - init scripts are now turned off by default; run chkconfig --on
4940 lustre and chkconfig --on lustrefs to use them
4941 - upcalls are no longer needed for clients to recover to failover
4943 - add --abort-recovery option to lconf to abort recovery on device
4945 - add support for an arbitrary number of OSTs (3026)
4946 - Quota support protocol changes.
4947 - forward compatibility changes to wire structs (6007)
4948 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4949 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4950 - improved "lustre status" script
4951 - initialize blocksize for non-regular files (6062)
4952 - added --disable-server and --disable-client configure options (5782)
4953 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4954 - Vanilla 2.4.29 support
4955 - increase maximum number of obd devices to 520 (6242)
4956 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4957 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4958 SLES 9 2.6.5 kernels
4959 - limit stripes per file to 160 (the maximum EA size) (6093)
4961 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4964 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4965 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4966 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4967 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4968 - small ext3 extents cleanups and fixes (5733)
4969 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4970 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4971 - avoid needless client->OST connect, fix handle mismatch (5317)
4972 - fix DLM error path that led to out-of-sync client, long delays (5779)
4973 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4974 - fix several locking issues related to i_size (5492,5624,5654,5672)
4975 - don't move pending lock onto export if it is already evicted (5683)
4976 - fix kernel oops when creating .foo in unlinked directory (5548)
4977 - fix deadlock in obdfilter statistics vs. object create (5811)
4978 - use time_{before,after} to avoid timer jiffies wrap (5882)
4979 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4980 - Do not leak request if server was not able to process it (5154)
4981 - If mds_open unable to find parent dir, make that negative lookup(5154)
4982 - don't create new directories with extent-mapping (5909, 5936)
4984 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4985 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4986 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4987 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4988 - support for RHEL4 kernel on i686 (5773)
4989 - provide error messages when incompatible logs are encountered (5898)
4991 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4992 * version 1.4.0.10 (1.4.1 release candidate 1)
4994 - don't keep a lock reference when lock is not granted (4238)
4995 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4996 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4997 - drop import inflight refcount on signal_completed_replay error (5255)
4998 - unlock page after async write error during send (3677)
4999 - handle missing objects in filter_preprw_read properly (5265)
5000 - no transno return for symlink open, don't save no-trasno open (3440)
5001 - don't try to complete elan receive that already failed (4012)
5002 - free RPC server reply state on error (5406)
5003 - clean up thread from ptlrpc_start_thread() on error (5160)
5004 - readahead could read extra page into cache that wasn't ejected (5388)
5005 - prevent races in class_attach/setup/cleanup/detach (5260)
5006 - don't dereference de->d_inode after l_dput of de (5458)
5007 - use "int" for stripe value returned from lock_to_stripe (5544)
5008 - mballoc allocation and error-checking fixes in 2.6 (5504)
5009 - block device patches to fix I/O request sizes in 2.6 (5482)
5010 - look up hostnames for IB nals (5602)
5011 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5012 - don't start multiple acceptors for the same port (5277)
5013 - fix incorrect LASSERT in mds_getattr_name (5635)
5014 - export a proc file for general "ping" checking (5628)
5015 - fix "lfs check" to not block when the MDS is down (5628)
5017 - service request history (4965)
5018 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5019 - create an "evict_client" /proc entry on OSTs, like the MDS has
5020 - fix mount usage message, return errors per mount(8) (5168)
5021 - change grep [] to grep "[]" in tests so they work in more UMLs
5022 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5023 - remove ancient LOV_MAGIC_V0 EA support (5047)
5024 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5025 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5026 - allow readahead allocations to fail when low on memory (5383)
5027 - mmap locking landed again, after considerable improvement (2828)
5028 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5030 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5033 - send OST transaction number in read/write reply to free req (4966)
5034 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5035 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5036 - the watchdog thread now runs as interruptible (5246)
5037 - flock/lockf fixes (but it's still disabled, pending 5135)
5038 - don't use EXT3 constants in llite code (5094)
5039 - memory shortage at startup could cause assertion (5176)
5041 - reorganization of lov code
5042 - single portals codebase
5044 - add extents/mballoc support (5025)
5045 - direct I/O reads in the obdfilter (4048)
5046 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5048 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5051 - send OST transaction number in read/write reply to free req (4966)
5052 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5053 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5054 - the watchdog thread now runs as interruptible (5246)
5055 - handle missing objects in filter_preprw_read properly (5265)
5056 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5057 - drop import inflight refcount on signal_completed_replay error (5255)
5058 - unlock page after async write error during send (3677)
5059 - return original error code on reconstructed replies (3761)
5060 - no transno return for symlink open, don't save no-trasno open (3440)
5062 - add pid to ldlm debugging output (4922)
5063 - bump the watchdog timeouts -- we can't handle 30sec yet
5064 - extra debugging for orphan dentry/inode bug (5259)
5066 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5069 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5070 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5071 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5072 - fix NULL dereference in /proc/sys/portals/routes (4827)
5073 - allow failed mdc_close() operations to be interrupted (4561)
5074 - stop precreate on OST before MDS would time out on it (4778)
5075 - don't send partial-page writes before EOF from client (4410)
5076 - discard client grant for sub-page writes on large-page clients (4520)
5077 - don't free dentries not owned by NFS code, check generation (4806)
5078 - fix lsm leak if mds_create_objects() fails (4801)
5079 - limit debug_daemon file size, always print CERROR messages (4789)
5080 - use transno after validating reply (3892)
5081 - process timed out requests if import state changes (3754)
5082 - update mtime on OST during writes, return in glimpse (4829)
5083 - add mkfsoptions to LDAP (4679)
5084 - use ->max_readahead method instead of zapping global ra (5039)
5085 - don't interrupt __l_wait_event() during strace
5087 - add software watchdogs to catch hung threads quickly (4941)
5088 - make lustrefs init script start after nfs is mounted
5089 - fix CWARN/ERROR duplication (4930)
5090 - return async write errors to application if possible (2248)
5091 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5092 - print NAL number in %x format (4645)
5093 - update barely-supported suse-2.4.21-171 series (4842)
5094 - support for sles 9 %post scripts
5095 - support for building 2.6 kernel-source packages
5096 - support for sles km_* packages
5098 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5101 - ignore -ENOENT errors in osc_destroy (3639)
5102 - notify osc create thread that OSC is being cleaned up (4600)
5103 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5104 - reconstruct ost_handle() like mds_handle() (4657)
5105 - create a new thread to do import eviction to avoid deadlock (3969)
5106 - let lconf resolve symlinked-to devices (4629)
5107 - don't unlink "objects" from directory with default EA (4554)
5108 - hold socknal file ref over connect in case target is down (4394)
5109 - allow more than 32000 subdirectories in a single directory (3244)
5110 - fix blocks count for O_DIRECT writes (3751)
5111 - OST returns ENOSPC from object create when no space left (4539)
5112 - don't send truncate RPC if file size isn't changing (4410)
5113 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5114 - bind to privileged port in socknal and tcpnal (3689)
5116 - rate limit CERROR/CWARN console message to avoid overload (4519)
5117 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5118 - basic mmap support (3918)
5119 - kernel patch series update from b1_4 (4711)
5121 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5124 - avoid crash during MDS cleanup with OST shut down (2775)
5125 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5126 - don't use bad inodes on the MDS (3744)
5127 - dynamic object preallocation to improve recovery speed (4236)
5128 - don't hold spinlock over lock dumping or change debug flags (4401)
5129 - don't zero obd_dev when it is force cleaned (3651)
5130 - print grants to console if they go negative (4431)
5131 - "lctl deactivate" will stop automatic recovery attempts (3406)
5132 - look for existing locks in ldlm_handle_enqueue() (3764)
5133 - don't resolve lock handle twice in recovery avoiding race (4401)
5134 - revalidate should check working dir is a directory (4134)
5136 - don't always mark "slow" obdfilter messages as errors (4418)
5138 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5141 - don't close LustreDB during write_conf until it is done (3860)
5142 - fix typo in lconf for_each_profile (3821)
5143 - allow dumping logs from multiple threads at one time (3820)
5144 - don't allow multiple threads in OSC recovery (3812)
5145 - fix debug_size parameters (3864)
5146 - fix mds_postrecov to initialize import for llog ctxt (3121)
5147 - replace config semaphore with spinlock (3306)
5148 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5149 - don't allow enqueue to complete on a destroyed export (3822)
5150 - down write_lock before checking llog header bitmap (3825)
5151 - recover from lock replay timeout (3764)
5152 - up llog sem before sending rpc (3652)
5153 - reduce ns lock hold times when setting kms (3267)
5154 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5155 - fix NULL deref and obd_dev leak on setup error (3312)
5156 - replace some LBUG about llog ops with error handling (3841)
5157 - don't match INVALID dentries from d_lookup and spin (3784)
5158 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5159 - fix invalid assertion in ptlrpc_set_wait (3880)
5161 - add libwrap support for the TCP acceptor (3996)
5162 - add /proc/sys/portals/routes for non-root route listing (3994)
5163 - allow setting MDS UUID in .xml (2580)
5164 - print the stack of a process that LBUGs (4228)
5166 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5169 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5170 - make sure to unset replay flag from failed open requests (3440)
5171 - if default stripe count is 0, use OST count for inode size (3636)
5172 - update parent mtime/ctime on client for create/unlink (2611)
5173 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5174 - free recovery state on server during a forced cleanup (3571)
5175 - unregister_reply for resent reqs (3063)
5176 - loop back devices mounting and status check on 2.6 (3563)
5177 - fix resource-creation race that can provoke i_size == 0 (3513)
5178 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5179 - more debugging for page-accounting assertion (3746)
5180 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5181 - avoid deadlock after precreation failure (3758)
5182 - fix race and lock order deadlock in orphan handling (3450, 3750)
5183 - add validity checks when grabbing inodes from l_ast_data (3599)
5185 - add /proc/.../recovery_status to obdfilter (3428)
5186 - lightweight CDEBUG infrastructure, debug daemon (3668)
5187 - change default OSC RPC parameters to be better on small clusters
5188 - turn off OST read cache for files smaller than 32MB
5189 - install man pages and include them in rpms (3100)
5190 - add new init script for (un)mounting lustre filesystems (2593)
5191 - run chkconfig in %post for init scripts (3701)
5192 - drop scimac NAL (unmaintained)
5194 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5197 - clean kiobufs before and after use (3485)
5198 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5199 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5200 - change init script to fail more clearly if not run as root (1528)
5201 - allow clients to reconnect during replay (1742)
5202 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5203 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5204 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5205 - fix cond_resched (3554)
5206 - only evict unfinished clients after recovery (3515)
5207 - allow bulk resend, prevent data loss (3570)
5208 - dynamic ptlrpc request buffer allocation (2102)
5209 - don't allow unlinking open directory if it isn't empty (2904)
5210 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5211 - remove extraneous obd dereference causing LASSERT failure (3334)
5212 - don't use get_cycles() when creating temp. files on the mds (3156)
5213 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5214 - handle EEXIST for set-stripe, set proper directory name (3336)
5216 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5217 - fix ksocknal_fmb_callback() error messages (2918)
5219 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5222 - don't copy lvb into (possibly NULL) reply on error (2983)
5223 - don't deref dentry after dput, don't free lvb on error (2922)
5224 - use the kms to determine writeback rpc length (2947)
5225 - increment oti_logcookies when osc is inactive (2948)
5226 - update client's i_blocks count via lvb messages (2543)
5227 - handle intent open/close of special files properly (1557)
5228 - mount MDS with errors=remount-ro, like obdfilter (2009)
5229 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5230 - don't use cancelling-locks' kms values (2947)
5231 - use highest lock extent for kms, not last one (2925)
5232 - don't dereference ERR_PTR() dentry in error handling path (3107)
5233 - fix thread race in portals_debug_dumplog() (3122)
5234 - create lprocfs device entries at setup instead of at attach (1519)
5235 - common AST error handler, don't evict client on completion race (3145)
5236 - zero nameidata in detach_mnt in 2.6 (3118)
5237 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5238 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5239 - initialize RPC timeout timer earlier for 2.6 (3219)
5240 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5241 - print nal/nid for unknown nid (3258)
5242 - additional checks for oscc recovery before doing precreate (3284)
5243 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5244 - don't crash in mdc_close for bad permissions on open (3285)
5245 - zero i_rdev for non-device files (3147)
5246 - clear page->private before handing to FS, better assertion (3119)
5247 - tune the read pipeline (3236)
5248 - fix incorrect decref of invalidated dentry (2350)
5249 - provide read-ahead stats and refine rpc in flight stats (3328)
5250 - don't hold journal transaction open across create RPC (3313)
5251 - update atime on MDS at close time (3265)
5252 - close LDAP connection when recovering to avoid server load (3315)
5253 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5254 - don't leak open file on MDS after open resend (3325)
5255 - serialize filter_precreate and filter_destroy_precreated (3329)
5256 - loop device shouldn't call sync_dev() for nul device (3092)
5257 - clear page cache after eviction (2766)
5258 - resynchronize MDS->OST in background (2824)
5259 - refuse to mount the same filesystem twice on same mountpoint (3394)
5260 - allow llmount to create routes for mounting behind routers (3320)
5261 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5262 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5263 - fix rare problem with rename on htree directories (3417)
5265 - allow default OST striping configuration per directory (1414)
5266 - fix compilation for qswnal for 2.6 kernels (3125)
5267 - increase maximum number of MDS request buffers for large systems
5268 - change liblustreapi to be useful for external progs like lfsck (3098)
5269 - increase local configuration timeout for slow disks (3353)
5270 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5272 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5275 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5276 - don't overwrite extent policy data in reply if lock was blocked (2901)
5277 - drop filter export grants atomically with removal from device (2663)
5278 - del obd_self_export from work_list in class_disconnect_exports (2908)
5279 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5280 - swab reply message in mdc_close, other PPC fixes (2464)
5281 - fix destroying of named logs (2325)
5282 - overwrite old logs when running lconf --write_conf (2264)
5283 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5284 - fix race in target_handle_connect (2898)
5285 - mds_reint_create() should take same inode create lock (2926)
5286 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5287 - don't close files for self_export to avoid uninitialized obd (2936)
5288 - allow MDS with the same name as client node (2939)
5289 - hold dentry reference for closed log files for unlink (2325)
5290 - reserve space for all logs during transactions (2059)
5291 - don't evict page beyond end of stripe extent (2925)
5292 - don't oops on a deleted current working directory (2399)
5293 - handle hard links to targets without a parent properly (2517)
5294 - don't dereference NULL lock when racing during eviction (2867)
5295 - don't grow lock extents when lots of conflicting locks (2919)
5297 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5300 - account for cache space usage on clients to avoid data loss (974)
5301 - lfsck support in lustre kernel code (2349)
5302 - reduce journal credits needed for BRW writes (2370)
5303 - orphan handling to avoid losing space on client/server crashes
5304 - ptlrpcd can be blocked, stopping ALL progress (2477)
5305 - use lock value blocks to assist in proper KMS, faster stat (1021)
5306 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5307 - recovery for initial connections (2355)
5308 - fixes for mds_cleanup_orphans (1934)
5309 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5310 - block all file creations until orphan recovery completes (1901)
5311 - client remove rq_connection from request struct (2423)
5312 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5313 - recovery timer race (2670)
5314 - mdc_close recovey bug (2532)
5315 - ptlrpc cleanup bug (2710)
5316 - mds timeout on local locks (2588)
5317 - namespace lock held during RPCs (2431)
5318 - handle interrupted sync write properly (2503)
5319 - don't try to handle a message that hasn't been replied to (2699)
5320 - client assert failure during cleanup after abort recovery (2701)
5321 - leak mdc device after failed mount (2712)
5322 - ptlrpc_check_set allows timedout requests to complete (2714)
5323 - wait for inflight reqs when ptlrpcd finishes (2710)
5324 - make sure unregistered services are removed from the srv_list
5325 - reset bulk XID's when resending them (caught by 1138 test)
5326 - unregister_bulk after timeout
5327 - fix lconf error (2694)
5328 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5329 - readahead locks pages, leaves pending causing memory pressure (2673)
5330 - increase OST request buffers to 4096 on large machines (2729)
5331 - fix up permission of existing directories in simple_mkdir (2661)
5332 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5333 - don't assign transno to errored transactions (2742)
5334 - don't delete objects on OST if given a bogus objid from MDS (2751)
5335 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5336 - if rq_no_resend, then timeout request after recovery (2432)
5337 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5338 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5339 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5340 - MDS recovery shouldn't delete objects that it has given out (2730)
5341 - if enqueue arrives after completion, don't clobber LVB (2819)
5342 - don't unlock pages twice when trigger_group_io returns error (2814)
5343 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5344 - don't write pages to disk if there was an error (1450)
5345 - don't ping imports that have recovery disabled (2676)
5346 - take buffered bytes into account when balancing socknal conn (2817)
5347 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5348 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5349 - remove little-endian swabbing of llog records (1987)
5350 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5351 - retry reposting request buffers if they fail (1191)
5352 - grow extent at grant time to avoid granting a revoked lock (2809)
5353 - lock revoke doesn't evict page if covered by a second lock (2765)
5354 - disable VM readahead to avoid reading outside lock extents (2805)
5356 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5357 - updated kernel patches for hp-2.4.20 kernel (2681)
5359 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5362 - fix truncated write corruption (2366)
5363 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5365 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5366 - ensure len(uuid) < 37 in lmc (1171)
5367 - fix ia64 OOPS in llog_test (2255)
5368 - zero end of page at obdfilter for partial page writes (2648)
5369 - don't leave stale dentries around after renames (bug 2428)
5370 - fix timeouts when evicting a client with a single lock held (2642)
5371 - set deadline for the initial HELLO message to drain (2634)
5372 - print out dotted-quad IP addresses in the socknal (2302)
5374 - additional debugging for MDS client eviction problem (2443)
5375 - fix mkfsoptions support for osts (2603, 2604)
5377 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5380 - add series for the vanilla 2.6.0 kernel
5381 - add series for the vanilla 2.4.24 kernel
5382 - add series for a cray x86/64 UL kernel drop
5383 - fix xattr patches for the vanilla 2.4.19 series
5385 - generate true UUIDs in lmc (1171)
5386 - have portals stack dumping break in UML (2466)
5387 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5388 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5389 - wake callback waiting threads on client eviction (2460)
5390 - Add --ptldebug and --subsystem to lmc (1719)
5391 - update assertion to allow safe interrupt allocation
5392 - set rq_no_resend for cancel requests (2432)
5393 - recalculate ptlrpcd timeout after resend (2494)
5394 - call vfs_rmdir when removing pending directories (2368)
5395 - fix renaming a file to itself (2429)
5396 - lmc creates a default one-stripe lov (2454)
5397 - expand procfs space to handle large clusters (2326)
5398 - increase UML stack to avoid overflow
5399 - update lconf's list of debug and subsystem masks
5400 - fix lfs find --obd (2510)
5401 - /proc tunable for disabling filter read caching (2591)
5402 - stop rpm packages from altering slapd.conf (2301)
5403 - disable nagle in the socknal under 0conf (2578)
5404 - choose mds inode size based on stripe count (2572)
5405 - fix kernel-source rpm problems (2516)
5407 - add --disable-doc to avoid pdf generation (2421)
5408 - update documentation, tests, type-os, comments
5409 - avoid format warnings on ia64
5410 - remove the TOE NAL
5411 - tiny code cleanups by removing unused fields
5413 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5416 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5417 - remove the most blatant lies from BUILDING file (2371)
5418 - change default debug level to reasonable production setting
5419 - reduce client side cache size to reduce cache flush time
5420 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5421 - make TCP zerocopy and pinger support enabled by default (2476)
5422 - sync writes completed after process exits caused crashes (2319)
5423 - maintain correct mount count on the MDS (2356)
5424 - backout 1557, because 2316 wasn't really fixed
5425 - better file I/O statistics gathering in /proc
5426 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5427 - another kernel patch to fix zero-copy TCP function export
5428 - don't take duplicate lock when processing re-sent getattr (2420)
5429 - lctl uses obd_self_export instead of creating new conn (2353)
5430 - MDS/OST recovery case which requires object creation asserted (2425)
5431 - move lfs from /usr/sbin to /usr/bin in packages
5432 - fix race between mds_client_add and mds_client_free (2417)
5433 - use kmalloc instead of slabs in portals (2430)
5434 - don't create duplicate records when a failover MDS is present (2442)
5435 - remove unnecessary mount age check (2332)
5436 - don't remove directory inodes from locks prematurely (2451)
5437 - don't break if MDS service name is the same as hostname (2103)
5438 - fix races in client write RPC generation when cache full (2482)
5440 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5443 - remove now-unused request->rq_obd (278)
5444 - if an allocation fails, print out how much memory we've used (1933)
5445 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5446 - add the "configurable stack size" patch to most series files (1256)
5447 - ability to write large log records, for 100+ OST configs (2306)
5448 - fix NULL deref when filter_prep fails (2314)
5449 - fix operator precedence error in filter_sync
5450 - dynamic allocation of socknal TX descriptors (2315)
5451 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5452 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5453 - ensure that CWARN messages in Portals always get to the syslog
5454 - __init/__exit are not for prototype decls (ldlm_init/exit)
5455 - x86-64 compile warning fixes
5456 - fix gateway LMC keyword conflict (2318)
5457 - fix MDS lock inversions in getattr/reint paths (1844)
5458 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5459 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5460 - DLM race condition prevented some lock evictions (2328)
5461 - ENOMEM detection and retry on socknal sends (2230)
5462 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5463 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5464 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5465 - better error messages when a client is rejected during recovery (1505)
5466 - avoid cancelling locks which were never granted, after failure (2330)
5467 - fix i_sem/journal inversion in mds_client_add (2333)
5468 - fix truncate/getattr lock cycle deadlock (2334)
5469 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5470 - fix two rare exit paths which could leak an l_lock() ref (2321)
5471 - fencepost error in MDS/OST orphan recovery (2226)
5472 - make log record alignment 8 bytes (1988)
5473 - lstripe now fails when requested offset > ost_count (2237)
5474 - ensure that all kernel series have a complete list.h (1607)
5475 - fix crashes in special-file operations (2316)
5476 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5478 - allow configurable automake binary, for testing new versions
5479 - small update to the lfs documentation
5481 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5483 * fix negative export reference count in fsfilt_sync (2312)
5485 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5486 * release candidate 0.9.1
5488 - orphans are moved into the PENDING directory for possible recovery
5489 - replayed opens now open by fid for orphan/rename safety (1042)
5490 - last close of an orphan inode generates a transno (683)
5491 - chdir() and mount() now pin the directory entry (1020)
5492 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5493 - discard very old requests without processing them (1502)
5494 - remove some common, well-understood CERRORs (1505)
5495 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5496 - clear "grant" flags in OST replies until OST grant code lands (1644)
5497 - fix read performance by not clobbering i_blksize on client (1598)
5498 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5499 - make LRU size a /proc tunable, clears locks when reduced (707)
5500 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5501 - server locks take references on exports now (1558)
5502 - build fixes for 2.4.20-rh trees (1663)
5503 - return an error from lov_create if all OSCs are inactive (1751)
5504 - fix import levels when a reconnect happens without a timeout (1597)
5505 - exit early from mds_open if we get a lookup error (1749)
5506 - partial page read at EOF wouldn't wait for disk before sending (1642)
5507 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5508 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5509 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5510 - write/truncate lock inversion (1639)
5511 - Don't auto-load obdclass, portals modules during cleanup (1495)
5512 - fix timestamps from jumping to "now" (1763)
5513 - extra journal assertions (1648)
5514 - add an extra multiunlink test (1771)
5515 - fix read_record/write_record API (1776)
5516 - fix leak of offset_extent, possible incorrect i_size later (1772)
5517 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5518 - Debugging for the kqswnal_get_idle_tx problems (1820)
5519 - Allow recovery to be attempted multiple times (1536)
5520 - Write out MDS last_rcvd file after it is first created (1600)
5521 - Fix tx_descriptor leak in failed transmit situations (1827)
5522 - ext3 journaling fixes for assertion failure after IO error (1871)
5523 - class_export_put() on freed export after completion AST error (1896)
5524 - Fix revalidate looping in VFS (1322)
5525 - Don't access a freed export during MDS_REINT timeout (1521)
5526 - Add open-unlink recovery support on the MDS (1673,1764)
5527 - Return an error if no MDS data was read from last_rcvd (1946)
5528 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5529 - Don't setup a disk device that doesn't match exported UUID (317)
5530 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5531 - avoid committing NULL handle in force close
5532 - local.sh is now a one-stripe LOV configuration
5533 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5534 - fix ext3 htree duplicate directory entry corruption (1516)
5535 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5536 - update to new LOV EA format (2097)
5537 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5539 2003-06-15 Phil Schwan <phil@clusterfs.com>
5542 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5543 - per-import/export recovery handling (958, 931, 959)
5544 - multiple last-rcvd slots, for serving multiple FSes (949)
5545 - connections are again shared between multiple imp/exports (963, 964)
5546 - "umount -f" would hang if any requests needed to be sent (393, 978)
5547 - avoid pinning large req buffer by copying for queued messages (989)
5548 - add "uuid" to "lctl device" command to help upcalls (991)
5549 - "open" RPCs with transnos would confuse recovery counters (1037)
5550 - do proper endian conversion of all wire messages (288, 340, 891)
5551 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5552 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5553 - fix ptlrpc_connection leak in target_handle_connect (1174)
5554 - fix import refcounting bug in OST and MDS cleanup (1134)
5555 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5556 - fix very unlikely obd_types race condition (501)
5557 - remove osc_open hack for echo_client (1187)
5558 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5559 - a failure in read_inode2 leads to deadlock (1139)
5560 - cancel ack-locks as soon as transaction is committed (1072)
5561 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5562 - make sure to commitrw after any preprw to avoid deadlock (1162)
5563 - failing to execute a file in a lustre FS would lock inode (1203)
5564 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5565 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5566 - obd_uuid2tgt would walk off the end of the list (1255)
5567 - on IA64 the osc would give portals incorrect bulk size (1258)
5568 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5569 - fix lock inversion caused by new llite matching code (1282)
5570 - limit the number of dirty pages on a client to 10MB (1286)
5571 - timed out locks were not being corrected cancelled (1289)
5572 - fix O_DIRECT above 4GB on IA-32 (1292)
5573 * major user-visible changes
5574 - fail out/fail over policy now controlled by the upcall (993)
5576 - add OBD_PING to check server availability and failure (954)
5577 - lustre messages are now sent in sending host order (288, 340, 891)
5578 - add eadatalen to MDS getattr reply (340)
5579 - OST read replies may contain second buffer, with per-page status (593)
5581 2003-03-11 Phil Schwan <phil@clusterfs.com>
5584 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5585 - failed open()s could cause deadlock; fixed (867, 869)
5586 - stop cancelling OST locks when files are closed (481)
5587 - overlapping XID spaces caused network corruption (851, 853)
5588 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5589 - setattr_raw added, to do single-RPC, server-side setattrs
5590 - lmc/lconf syntax change for OST UUIDs
5591 - fix crashy race condition between ptlrpc_free_req and osc_close
5592 - don't use request in mdc_enqueue if we hit a timeout (889)
5593 - don't set the inode i_size for regular files from the MDS (896)
5594 - handle out of order completion AST (842)
5595 - don't LBUG if a lock request times out after receiving AST (913)
5596 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5597 - if a bad lock AST arrives, send an error instead of dropping entirely
5598 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5599 - fix leak in bulk IO when only partially completed (899, 900, 926)
5600 - fix O_DIRECT for ia64 (55)
5601 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5602 - C-z after timeout could hang a process forever; fixed (977)
5604 - client-side I/O cache (678, 924, 929, 941, 970)
5606 - READPAGE and SETATTRs which don't take server-side locks get
5609 2003-02-11 Phil Schwan <phil@clusterfs.com>
5612 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5613 - Fix fsfilt_extN_readpage() to read a full page of directory
5614 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5615 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5616 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5617 - Most intent operations were removed, in favour of a new RPC mode
5618 that does a single RPC to the server and bypasses most of the VFS
5619 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5620 - Aggressively cancel local locks on DLM servers
5621 - mds_reint_unlink sends EA to the client if it's the last nlink.
5622 client uses that EA to unlink OST objects.
5623 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5624 - recursive symlinks were fixed (439)
5625 - fixed NULL deref in DEBUG_REQ
5626 - filter_update_lastobjid no longer calls sync, which annoyed extN
5627 - fixed multi-client small-writes to a single file problem (445)
5628 - fixed mtime updates during file writes (607)
5629 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5630 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5631 - fixed Linux OST size reporting problem (444, 656)
5632 - OST now updates object mtime with writes or setattr (607, 619)
5633 - client verifies file size before zeroing page past EOF (445)
5634 - OST now writes last allocated objid to disk with allocation (108)
5635 - LOV on echo now works (409)
5637 - mds_reint_unlink sends a new buffer, with the EA included. this
5638 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5639 set if a regular file was being unlinked, and it was the last link
5640 - use PtlGet from the target for bulk writes (315)
5641 - OST now updates object mtime with writes or setattr (607, 619)
5642 - LDLM now has a grant-time callback to revalidate locked items, if
5644 - Many MDS operations were reorganized to combat race conditions
5646 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5647 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5649 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5652 - Fully reactivate OST imports after reconnection (512, others)
5653 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5654 - More graceful error handling for truncating on dead OST (515)
5655 - Don't error out unless we're actually accessing dead stripes (474)
5656 - Fix garbage sizes when stripes are missing (410)
5657 - LRU counters were broken, causing constant lock purge (433, 432)
5658 - garbage on read from stripes with failed OSTs (441)
5659 - mark OSCs as active before reconnecting during recovery (438)
5660 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5661 - lfind did not preserve OST order in output (443)
5662 - symlinks cause hung clients, incorrect data (439)
5663 - stop dereferencing request after dropping refcount (457)
5664 - don't LASSERT(spin_is_locked) on non-SMP (455)
5665 - fixes for many rename() bugs
5666 - fstat didn't correctly synchronize attributes (399)
5667 - server must handle lock cancellation during blocking AST prep (487)
5668 - bulk descriptors were free()d too soon (511)
5669 - fix paths in lconf, which would load incorrect modules (451, 507)
5670 - fix confusing lconf 'host not found' error message (386)
5671 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5672 - fix race condition in mdc_blocking_ast() for inode access (526)
5673 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5674 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5675 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5676 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5677 - fix lstripe in several ways (kernel side) (527)
5678 - fix request leak in ldlm_cli_enqueue (262)
5679 - incorrect OSC was marked inactive after OST failure
5680 - call mds_fs_cleanup before unmounting filesystem (524)
5681 - fix races between taking ns_lock and ldlm_lock_change_resource
5682 - fix races updating LOV export open file list
5683 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5684 - fix recovery NULL deref in ldlm_cli_cancel_unused
5685 - fix some DLM races by using new hash table for lock handles (419)
5686 - permit the client to specify desired inodes, at replay
5687 - duplicate requests when we queue them for replay reintegration
5688 - fix last_rcvd offset calculation
5689 - sync after each recovered transaction, so we always make progress
5690 - never, not always, ERESTART requests without transnos
5691 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5693 - skip replay if the MDS says that the client is already connected
5694 - don't check for a recovery-enabled export to match lctl's UUID
5695 - don't INC_USE_COUNT for phantom exports
5696 - don't crash when cleaning up phantom exports (567)
5697 - don't double-finish or set replay data for errored mdc_open requests
5698 - abort requests when they time out, so we don't get old replies
5699 - send/receive replies for AST messages again
5700 - if the client says that it doesn't have the lock, cancel it on the
5702 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5703 mark it as destroyed, it will all work out when decref is called
5704 - fix module use counts (22, 581)
5706 - ASTs now expect a reply (server cancels lock on error reply)
5708 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5711 - fix many simultaneous client startup (392)
5712 - fix dentry->d_it clobbering
5713 - credentials weren't being shipped for readdir/getattr operations
5714 - remove invalid assertions triggered during some concurrent MD
5716 - proper Lustre versions added (336, 389)
5717 - fix memory leak for create error case (398)
5718 - fix LOV locking bug that would get cli/srv out of sync
5719 - fix echo client over LOV (409)
5720 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5721 - fix double-O_EXCL intent crash (424)
5722 - avoid sending multiple lock CANCELs (352)
5724 - MDS can do multi-client recovery (modulo bugs in new code)
5726 - many updates, edits, cleanups
5728 2002-11-18 Phil Schwan <phil@clusterfs.com>
5731 - fix null d_it dereference (346)
5732 - fix full OST/dbench hang (333)
5733 - fix permission problem with file removal (286)
5734 - fix removal of OSCs from LOV when they fail
5735 - fix NULL deref during bulk timeout (214)
5736 - fix problems related to multiple filesystems on one MDS (241)
5737 - fixed serious subtle metadata locking bugs
5738 - free locks on clients when inodes are removed due to memory
5740 - fix inode pointer in lock data (285)
5741 - partial support for multiple MDS on a single host (241)
5742 - data locks weren't cancelled at clear_inode time (290, 311)
5743 - intent locks could lead to unbounded lock growth (205)
5744 - added a maximum lock count, an LRU list, and a flusher
5745 - fix multiple rename (365)
5746 - properly abstracted the echo client
5747 - OSC locked 1 byte too many; fixed
5748 - rewrote brw callback code:
5749 - fixed recovery bugs related to LOVs (306)
5750 - fixed too-many-pages-in-one-write crash (191)
5751 - fixed (again) crash in sync_io_timeout (214)
5752 - probably fixed callback-related race (385)
5754 - Add capability to MDS protocol
5755 - LDLM cancellations and callbacks on different portals
5757 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5760 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5761 - defer open object destruction to close time (601981, 138)
5762 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5763 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5764 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5765 - don't call obd_setattr in ll_file_release for destroyed objects
5766 * protocol change to lustre_msg: move |version| and add |flags|
5767 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5768 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5769 * added replay of create, unlink, link and rename operations during
5770 MDS failover; recovery should be much more robust now
5771 * remove failed OSCs from LOVs (only lov_create uses this so far)
5772 * the lustre-HOWTO was brought (more) up to date (582544)
5774 2002-10-23 Phil Schwan <phil@clusterfs.com>
5777 - in-use dentries weren't being reused properly (617851)
5778 - prevent multiple LDLM setup (599178)
5779 - fix LOV size calculations for truncate (617853)
5780 - fix client handling of MDS intent errors (POSIX)
5781 - fix permission bug in lovstripe.c test (624321)
5782 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5783 - truncate past end of file could corrupt data
5784 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5785 - a race in recovery could return ETIMEDOUT to apps (623947)
5786 - building outside the source directory was fixed
5787 * the lustre-HOWTO was brought (more) up to date (582544)
5788 * major progress was made on recovery functionality
5790 2002-10-10 Phil Schwan <phil@clusterfs.com>
5793 - recovery deadlock fix
5794 - rm -rf causes LBUG fix (617817)
5795 - file open by multiple tasks fix (618962)
5796 - directory permissions bugs (602707 and 620007)
5797 - journal_stop fixed with locking (611313)
5798 - O_APPEND failures resolved (618273, perhaps 614459)
5799 - lconf PATH fix (619770)
5800 - IA64 build fix (621450)
5801 - RPC buffer sizes scale with amount of memory
5803 2002-10-01 Phil Schwan <phil@clusterfs.com>
5806 - locks would be cancelled without throwing away data pages,
5807 resulting in inconsistent data (605627)
5808 - inode attributes were not always being refreshed (605627, 612449)
5809 - lconf now continues to cleanup after lctl reports an error
5810 - MDS now enforces user permissions (602707)
5811 - lprocfs cleanup fixed, but not yet enabled (614157)
5812 - fixed infinite server hang, should a client not respond to an AST
5813 - avoid going into recovery if user calls readlink() with a buffer
5814 that's too small (613941)
5815 - AST RPCs no longer require replies (614867) -- this may be changed
5816 - don't crash server if client sends an IOV that's too big (611336)
5817 - fixed lock conversion deadlock (611892)
5818 - fixed the following of symlinks (614622)
5819 * recovery: the server can remove locks from a client that dies, other
5820 clients can make progress
5821 * more extN patch fixes
5822 * compile-time configurable ptlrpc buffer allocations
5824 - collaborative read cache document
5825 - Lustre Lite Performance CDR document-in-progress
5827 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5830 - fix typo in patch-2.4.18
5832 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5835 - clear ptlrpc request each time in handle_incoming_request()
5836 - unlink of files now destroys the object on the OST
5838 2002-09-19 Peter Braam <braam@clusterfs.com>
5840 * add hard link support
5841 * change obdfile creation method
5842 * kernel patch changed
5844 2002-09-19 Peter Braam <braam@clusterfs.com>
5847 - stack overflow bug in extN fixed
5849 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5851 * documentation updates
5852 - add man pages for config tools
5853 - update tests/README to describe testing with new config tools
5854 - finish metadata API descriptions
5855 * bug fixes and cleanups
5856 - statfs workaround for 16TB limit
5857 - LOV stripe allocation improved, can stripe on subset of OSTs
5858 - LOV file size/IO offset was wrong for files > 4GB in size
5859 - object EA data was being dropped, caused files to be unreadable
5860 - memory overflow with non-LOV OST caused memory corruption
5861 - fixed regression tests to work with new config tools, obdfilter
5862 - fixed bug when directory size became larger than 1 block
5863 - fixed bug (for single client case) when PWD was deleted
5864 - invalidate local directory pages when doing intent-based ops
5865 - avoid LDLM oops when lock callback contained bad data
5867 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5869 * documentation updates
5870 * bug fixes and cleanups
5871 - configuration tools
5874 - 64-bit compile warnings
5875 - 64-bit internal statfs data
5877 * test_brw on persistent OST devices
5879 * lprocfs (disabled)
5881 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5883 * documentation updates
5884 * bug fixes and cleanups
5885 * configuration tools
5887 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5890 * many small fixes to 0_5_4
5891 * io/network handling
5892 * thinkos in MDS operations
5894 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5897 * crucial basic fixes to 0.5.3
5898 * IOR, Iozone work over Elan
5901 2002-08-07 Phil Schwan <phil@clusterfs.com>
5902 * version 0_5_3, our first alpha
5903 * we use the new Portals iovs
5904 * documentation updates
5905 * bug fixes and cleanups
5906 * small changes in the DLM wire protocol
5908 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5909 * version 0_5_1 with some initial stability,
5910 * locking on MD and file I/O.
5911 * documentation updates
5912 * several bug fixes since 0.5.0
5913 * small changes in wire protocol
5915 2002-07-18 Phil Schwan <phil@clusterfs.com>
5917 * delivered as Lustre Light Alpha
5918 * fixed a crash after handling invalid MDS requests
5919 * fixed directory pages for architectures with non-4k pages sizes
5921 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5922 * release version v0_4_4
5923 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5925 * Fixes a number of interruption problems with OST operations.
5926 * Update documentation for portals header changes
5927 * Move all wire protocol structs/defines to lustre_idl.h
5928 * Fixes symlink length bug.
5929 * Add tcpdump to repository.
5931 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5932 * release version v0_4_3
5933 * Fixes statfs for inodes on extN.
5934 * Fixes bug in runtests which would delete /etc/hosts.
5935 * Use 64-bit object IDs wherever possible (not into VFS though)
5936 Remove ost_get_info, which is unused by lustre, and out of date.
5938 2002-07-03 Peter Braam <braam@clusterfs.com>
5939 * release version v0_4_2 Fixes a lookup error (type not passed)
5940 * move forward to head of Portals
5941 * move forward to latest Lustre kernel
5943 2002-06-25 Peter Braam <braam@clusterfs.com>
5944 * release version v0_4_1. Hopefully stable on single node use.