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: Improved cleanup procedure for conf-sanity test_46a.
22 Description: Do not start bulk transfer, if client tried to reconnect.
23 Details : Do not start bulk transfer, if client tried to reconnect,
24 but failed due to active rpcs. Check for exp_abort_active_req
29 Description: Prevent inconsistences between linux and lustre mount structures.
30 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
31 Make the sleep interruptible.
33 Severity : enhancement
35 Description: provide server to client comms path
36 Details : server to client communications path via new LDLM_SET_INFO rpc
40 Description: truncate starts GFP_FS allocation under transaction causing deadlock
41 Details : ldiskfs_truncate calls grab_cache_page which may start page
42 allocation under an open transaction. This may lead to
43 calling prune_icache with consequent lustre reentrance.
47 Description: send correctly lsm on open replay
48 Details : MDS is trust to LSM size on replay open, but client can set wrong size
53 Description: enable client interface failover
54 Details : When a child reconnects from another NID, properly update export
55 nid hash position and ldlm reverse import.
57 Severity : enhancement
59 Description: Communicate OST degraded/readonly state via statfs to MDS
60 Details : Flags in the statfs returned from OSTs indicate whether the
61 OST is in a degraded RAID state, or if the filesystem has
62 turned read-only after a filesystem error is detected.
67 Description: don't panic if EPROTO was hit when reading symlink
68 Details : correctly handling request reference in error cases.
70 Severity : enhancement
72 Description: Add LustreNetLink, a kernel-userspace communcation path.
74 Severity : enhancement
76 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
81 Description: access to llog context before init.
82 Details : move handling CATALOGS file at osc layer and forbid access to llog
86 Frequency : rare, only when client is writing to OST on local node
88 Description: Avoid deadlock for local client writes
89 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
90 memory freeing context. This allows OST threads to set the
91 PF_MEMALLOC flag on task structures in order to allocate memory
92 from reserved pools and complete IO.
93 Use GFP_HIGHUSER for OST allocations for non-local client writes,
94 so that the OST threads generate memory pressure and allow
95 inactive pages to be reclaimed.
97 Severity : enhancement
99 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
104 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
105 Details : move ldlm namespace creation in setup phase to avoid grab
106 _lprocfs_lock with cli_sem held.
110 Description: Temporarily disable grant shrink.
111 Details : Disable the feature for debugging.
114 Frequency : only when formatting test filesystems
116 Description: Unable to run several mkfs.lustre on loop devices at the same time
117 Details : mkfs.lustre returns error 256 on the concurrent loop devices
118 formatting. The solution is to proper handle the error.
120 Severity : enhancement
122 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
124 Severity : enhancement
126 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
131 Description: resolve race between obd_disconnect and class_disconnect_exports
132 Details : if obd_disconnect will be called to already disconnected export he
133 forget release one reference and osc module can't unloaded.
135 Severity : enhancement
137 Description: Allow tuning service thread via /proc
138 Details : For each service a new
139 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
140 created that can be used to set min/max thread counts, and get the
141 current number of running threads.
146 Descriptoin: don't return error if have partially created objects for file.
147 Details : lov_update_create_set uses set->set_success as index for created
148 objects, so if some requests will be failed, they will have hole
149 at end of array and qos_shrink_lsm for allocate correct lsm.
151 Severity : enhancement
153 Description: Update OFED support to 1.4.1
158 Description: Reduce small size read RPC
159 Details : Set read-ahead limit for every file and only do read-ahead when
160 available read-ahead pages are bigger than 1M to avoid small size
163 Severity : enhancement
165 Description: move AT tunable parameters for more consistent usage
166 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
168 Severity : enhancement
170 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
171 Details : allow skip disconnected ost for send statfs request and hide error
177 Description: fix racy locking of mballoc block bitmaps causing BUG
178 Details : The locking of the mballoc buddy bitmap and the in-memory
179 block bitmap was using two different spin locks in some
180 cases. This made it possible to incorrectly access the
181 mballoc bitmap while another process was modifying it,
182 causing a sanity assertion to fail. While no on-disk corruption
183 was reported, there was some risk of this happening.
186 Frequency : rare, on llog test 6
188 Descriptoin: don't allow connect to already connected import
189 Details : allowing connect to already connected import is hide connecting problem.
194 Descriptoin: allow kill process which wait statahead result
195 Details : for some reasons 'ls' can stick in waiting result from statahead,
196 in this case need way for kill this process.
198 Severity : enhancement
200 Description: Add state history info file, enhance import info file
201 Details : Track import connection state changes in a new osc/mdc proc file;
202 add overview-type data to the osc/mdc import proc file.
204 Severity : enhancement
206 Description: MDS create should not wait for statfs RPC while holding DLM lock.
208 Severity : enhancement
210 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
215 Descriptoin: OST grant too much space to client even there are not enough space.
216 Details : Client will shrink its grant cache to OST if there are no write
217 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
218 this grant cache if there are already not enough avaible space
219 (left_space < total_clients * 32M).
222 Frequency : start MDS on uncleanly shutdowned MDS device
224 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
225 Details : stay in waiting mds<>ost recovery finished produce random bugs
226 due race between two ll_sync thread for one lov target. send
227 ACTIVATE event only if connect realy finished and import have
231 Frequency : rare, connect and disconnect target at same time
233 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
234 Details : don't call obd_disconnect under lov_lock. this long time
235 operation and can block ptlrpcd which answer to connect request.
240 Descriptoin: don't lose wakeup for imp_recovery_waitq
241 Details : recover_import_no_retry or invalidate_import and import_close can
242 both sleep on imp_recovery_waitq, but we was send only one wakeup
246 Frequency : always with long access acl
248 Descriptoin: mds can't pack reply with long acl.
249 Details : mds don't control size of acl but they limited by reint/getattr
252 Severity : enhancement
254 Description: Update to SLES10 kernel-2.6.16.60-0.33.
256 Severity : enhancement
258 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
261 Frequency : start MDS on uncleanly shutdowned MDS device
263 Descriptoin: aborting recovery hang on MDS
264 Details : don't throttle destroy RPCs for the MDT.
267 Frequency : on remount
269 Description: external journal device not working after the remount
270 Details : clear dev_rdonly flag for external journal devices in
276 Description: shutdown vs evict race
277 Details : client_disconnect_export vs connect request race.
278 if client will evicted at this time - we start invalidate
279 thread without referece to import and import can be freed
283 Frequency : rare, need acl's on inode.
285 Description: client can't handle ost additional correctly
286 Details : if ost was added after client connected to mds client can have
287 hit lnet_try_match_md ... to big messages to wide striped files.
288 in this case need teach client to handle config events about add
289 lov target and update client max ea size at that event.
291 Severity : enhancement
293 Description: Changelogs
294 Details : Changelogs are a lightweight mechanism to track filesystem
295 metadata and namespace changes. The changelog is recorded
296 permanently on the MDTs, and is periodically "consumed" / purged
297 when records are no longer needed.
299 Severity : enhancement
301 Description: compact fld format with extents
302 Details : Store range of seq rather than every seq in FLD. Seq
303 controller update FLD rather than clients. In Case of CMD, mdt0
304 has FLD, all other metadata server act as non persistent proxy
305 for FLD queries and cache fld entries in fld cache.
310 Description: don't skip ost target if they assigned to file
311 Details : Drop slow OSCs if we can, but not for requested start idx.
312 This means "if OSC is slow and it is not the requested
313 start OST, then it can be skipped, otherwise skip it only
314 if it is inactive/recovering/out-of-space.
318 Description: more cleanup in mds_lov
319 Details : not send LOV EA under replay, we can't know about they size at this
320 time. Don't allow client connect to mds before any ost connected,
321 for avoid problems with LOV EA size and returning EIO to client.
323 Severity : enhancement
325 Description: Interoperability at server side (Disk interoperability)
327 Severity : enhancement
329 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
331 Severity : enhancement
333 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
335 Severity : enhancement
337 Description: New client IO stack (CLIO).
339 Severity : enhancement
341 Description: Commit on sharing. Eliminate inter-client dependencies between
342 uncommitted transactions by doing transaction commits.
343 Thereby clients may recovery independently.
346 Frequency : Create a symlink file with a very long name
348 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
349 Details : If there is no extra space in the request for early cancels,
350 ldlm_req_handles_avail() returns 0 instead of a negative value.
352 Severity : enhancement
354 Description: Add /proc entry for import status
355 Details : The mdc, osc, and mgc import directories now have
356 an import directory that contains useful import data for debugging
359 Severity : enhancement
361 Description: Re-disable certain /proc logging
362 Details : Enable and disable client's offset_stats, extents_stats and
363 extents_stats_per_process stats logging on the fly.
366 Frequency : Only on FC kernels 2.6.22+
368 Description: oops in statahead
369 Details : Do not drop reference count for the dentry from VFS when lookup,
370 VFS will do that by itself.
372 Severity : enhancement
374 Description: Generic /proc file permissions
375 Details : Set /Proc file permissions in a more generic way to enable non-
376 root users operate on some /proc files.
380 Description: Hitting mdc_commit_close() ASSERTION
381 Details : Properly handle request reference release in
382 ll_release_openhandle().
386 Description: quota recovery deadlock during mds failover
387 Details : This patch includes att18982, att18236, att18237 in bz14840.
389 1. fix osts hang when mds does failover with quotaon
390 2. prevent watchdog storm when osts threads wait for the
395 Frequency : only patchless client
396 Description: add workaround for race between add/remove dentry from hash
398 Severity : enhancement
400 Description: Allow OST glimpses to return PW locks
404 Description: LBUG when llog conf file is full
405 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
409 Description: Prevent import from entering FULL state when server in recovery
413 Description: service mount cannot take device name with ":"
414 Details : Only when device name contains ":/" will mount treat it as
420 Description: replace ptlrpcd with the statahead thread to interpret the async
421 statahead RPC callback
425 Frequency : on recovery
426 Description: I/O failures after umount during fail back
427 Details : if client reconnected to restarted server we need join to recovery
428 instead of find server handler is changed and process self
429 eviction with cancel all locks.
431 Severity : enhancement
433 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
435 Severity : enhancement
437 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
439 Severity : enhancement
441 Description: Upcall on Lustre log has been dumped
442 Details : Allow for a user mode script to be called once a Lustre log has
443 been dumped. It passes the filename of the dumped log to the
444 script, the location of the script can be specified via
445 /proc/sys/lnet/debug_log_upcall.
450 Description: avoid idr_remove called for id which is not allocated.
451 Details : Move assigment s_dev for clustered nfs to end of initialization,
452 to avoid problem with error handling.
457 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
458 Details : When connection is reused this not moved from CONN_UNUSED_HASH
459 into CONN_USED_HASH and this prodice warning when put connection
460 again in unused hash.
463 Severity : enhancement
465 Description: File striping can now be set to use an arbitrary pool of OSTs.
467 Severity : enhancement
469 Description: Export bytes_read/bytes_write count on OSC/OST.
473 Description: Early reply size mismatch, MGC loses connection
474 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
475 the connect flags are properly negotiated.
480 Description: quotas are not honored with O_DIRECT
481 Details : all writes with the flag O_DIRECT will use grants which leads to
482 this problem. Now using OBD_BRW_SYNC to guard this.
486 Description: add quota statistics
487 Details : 1. sort out quota proc entries and proc code.
488 2. add quota statistics
490 Severity : enhancement
492 Description: enable quota support for HEAD.
496 Description: Properly propagate oinfo flags from lov to osc for statfs
497 Details : restore missing copy oi_flags to lov requests.
499 Severity : enhancement
501 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
502 : l_getgroups(8), lst(8), routerstat(8)
503 Details : included man pages for llobdstat(8), llstat(8),
504 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
506 Severity : enhancement
508 Description: configure's --enable-quota should check the
509 : kernel .config for CONFIG_QUOTA
510 Details : configure is terminated if --enable-quota is passed but
511 : no quota support is in kernel
515 Description: Remove portals compatibility
516 Details : Remove portals compatibility, not interoperable with releases
521 Description: Resolve device initialization race
522 Details : Prevent proc handler from accessing devices added to the
523 obd_devs array but yet be intialized.
525 Severity : enhancement
527 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
529 Severity : enhancement
531 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
536 Description: Using wrong pointer in osc_brw_prep_request
537 Details : Access to array[-1] can produce panic if kernel compiled with
538 CONFIG_PAGE_ALLOC enabled
542 Description: Client runs out of low memory
543 Details : Consider only lowmem when counting initial number of llap pages
547 Description: *optional* service tags registration
548 Details : if the "service tags" package is installed on a Lustre node
549 When the filesystem is mounted, a local-node service tag will
550 be created. See http://inventory.sun.com/ for more information
551 about the Service Tags asset management system.
555 Description: Kernel BUG tries to release flock
556 Details : Lustre does not destroy flock lock before last reference goes
557 away. So always drop flock locks when client is evicted and
558 perform unlock regardless of successfulness of speaking to MDS.
562 Description: add refcount protection for osc callbacks, avoid panic on shutdown
566 Description: sanity test 65a fails if stripecount of -1 is set
567 Details : handle -1 striping on filesystem in ll_dirstripe_verify
572 Description: ASSERTION(CheckWriteback(page,cmd)) failed
573 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
576 Severity : enhancement
578 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
582 Description: do not process already freed flock
583 Details : flock can possibly be freed by another thread before it reaches
584 to ldlm_flock_completion_ast.
588 Description: LBUG during stress test
589 Details : Need properly lock accesses the flock deadlock detection list.
593 Description: oops in page fault handler
594 Details : kernel page fault handler can return two special 'pages' in error
595 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
599 Description: timeout with invalidate import.
600 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
601 should be handled by ptlrpcd. This produce long age waiting and
602 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
604 Severity : enhancement
606 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
610 Description: enable MGS and MDT services start separately
611 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
612 a co-located MGS without starting the MGS, which is a complement
613 to 'nosvc' mount option.
617 Frequency : after recovery
618 Description: precreate to many object's after del orphan.
619 Details : del orphan st in oscc last_id == next_id and this triger growing
620 count of precreated objects. Set flag LOW to skip increase count
621 of precreated objects.
625 Frequency : rare, on clear nid stats
626 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
627 Details : when clean nid stats sometimes try destroy live entry,
628 and this produce panic in free.
632 Description: Stack overflow during MDS log replay
633 ease stack pressure by using a thread dealing llog_process.
637 Description: wait until IO finished before start new when do lock cancel.
638 Details : VM protocol want old IO finished before start new, in this case
639 need wait until PG_writeback is cleared until check dirty flag and
640 call writepages in lock cancel callback.
642 Severity : enhancement
644 Description: using special macro for print time and cleanup in includes.
648 Description: mds_mfd_close() ASSERTION(rc == 0)
649 Details : In mds_mfd_close(), we need protect inode's writecount change
650 within its orphan write semaphore to prevent possible races.
654 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
658 Frequency : rare, on shutdown ost
659 Description: don't hit live lock with umount ost.
660 Details : shrink_dcache_parent can be in long loop with destroy dentries,
661 use shrink_dcache_sb instead.
665 Description: don't panic with use echo client
666 Details : echo client pass NULL as client nid pointer and this produce null
671 Description: fix build on ppc32
672 Details : compile code with -m64 flag produce wrong object file for ppc32.
676 Description: add message levels for liblustreapi
680 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
681 Details : Move the 'good_osts' check before the 'total_bavail' check. This
682 will result in an -EAGAIN and in the exit call path we call
683 alloc_rr() which will with increasing aggressiveness attempt to
684 aquire precreated objects on the minimum number of required OSCs.
688 Description: Use old size assignment to avoid deadlock
689 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
690 the scheduling while holding a spinlock. We do not need locking
691 for size in ll_update_inode() because size is only updated from
692 the MDS for directories or files without objects, so there is no
693 other place to do the update, and concurrent access to such inodes
694 are protected by the inode lock.
698 Description: resolve "_IOWR redefined" build error on SLES10
702 Description: dump the memory debugging after all modules are unloaded to
703 suppress false negative in conf_sanity test 39
705 Severity : enhancement
707 Description: build kernel-ib packages for OFED 1.3 in our release cycle
712 Description: fix SLES kernel versioning
713 Details : the kernel version for our SLES 10 kernel did not include a "-"
714 before the "smp" at the end. while this was not a problem in
715 general, it did mean that software trying to use the kernel
716 version to try to detect a vendor specific kernel would fail.
717 this was most evident by the OFED build scripts.
721 Description: Don't update lov_desc members until making sure they are valid
722 Details : When updating lov_desc members via proc fs, need fix their
723 validities before doing the real update.
727 Description: don't put request into delay list while invalidate in flight.
728 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
729 import in flight. this produce timeout for invalidate and sometimes
730 can cause stale data.
734 Frequency : on ppc only
735 Description: not convert ost objects for directory because it's not exist.
736 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
740 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
742 Severity : enhancement
744 Description: Update to support 2.6.22.14 vanilla kernel.
748 Frequency : rare, on recovery
749 Description: read procfs can produce deadlock in some situation
750 Details : Holding lprocfs lock which send rpc can produce block for destroy
751 obd objects and this also block reconnect with -EALREADY. This
752 isn't fix all lprocfs bugs - but make it rare.
754 Severity : enhancement
756 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
759 Frequency : frequent on X2 node
761 Description: mdc_set_open_replay_data LBUG
762 Details : Set replay data for requests that are eligible for replay.
766 Description: lustre_mgs: operation 101 on unconnected MGS
767 Details : When MGC is disconnected from MGS long enough, MGS will evict the
768 MGC, and late on MGC cannot successfully connect to MGS and a lot
769 of the error messages complaining that MGS is not connected.
773 Frequency : on network error
774 Description: panic with double free request if network error
775 Details : mdc_finish_enqueue is finish request if any network error ocuring,
776 but it's true only for synchronus enqueue, for async enqueue
777 (via ptlrpcd) this incorrect and ptlrpcd want finish request
780 Severity : enhancement
782 Description: client-side metadata stat-ahead during readdir(directory readahead)
783 Details : perform client-side metadata stat-ahead when the client detects
784 readdir and sequential stat of dir entries therein
787 Frequency : on start mds
789 Description: Implement get_info(last_id) in obdfilter.
792 Frequency : occasional
794 Description: Correctly check stale fid, not start epoch if ost not support SOM
795 Details : open with flag O_CREATE need set old fid in op_fid3 because
796 op_fid2 was overwritten with new generated fid, but mds can answer
797 with one of these two fids and both is not stale. Setattr
798 incorrectly started an epoch and assume will be called
799 done_writing, but without SOM done_writing ever being called.
802 Frequency : rare, depends on device drivers and load
804 Description: MDS or OSS nodes crash due to stack overflow
805 Details : Code changes in 1.8.0 increased the stack usage of some functions.
806 In some cases, in conjunction with device drivers that use a lot
807 of stack the MDS (or possibly OSS) service threads could overflow
808 the stack. One change which was identified to consume additional
809 stack has been reworked to avoid the extra stack usage.
812 Frequency : occasional
814 Description: Do not fail import if osc_interpret_create gets -EAGAIN
815 Details : If osc_interpret_create got -EAGAIN it immediately exits and
816 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
817 oscc_has_objects and see OSC has no objests and call
818 oscc_internal_create to resend create request.
820 Severity : enhancement
822 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
824 Severity : enhancement
826 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
829 Frequency : very rare
831 Description: Fix replay if there is an un-replied request and open
832 Details : In some cases, older replay request will revert the
833 mcd->mcd_last_xid on MDS which is used to record the client's
836 Severity : enhancement
838 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
840 Severity : enhancement
842 Description: Add rhel5 support to HEAD.
844 Serverity : enhancement
846 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
851 Description: Don't allow skipping OSTs if index has been specified.
852 Details : Don't allow skipping OSTs if index has been specified, make
853 locking in internal create lots better.
857 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
858 Details : access to bitfield in structure is always rounded to long
859 and this produce problem with not atomic change any bit.
863 Description: Lustre make rpms failed.
864 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
865 builds Lustre rpms from tarball.
868 Frequency : rare on shutdown ost
870 Description: If llog cancel was not send before clean_exports phase, this can
871 produce deadlock in llog code.
872 Details : If llog thread has last reference to obd and call class_import_put
873 this produce deadlock because llog_cleanup_commit_master wait when
874 last llog_commit_thread exited, but this never success because was
875 called from llog_commit_thread.
879 Description: allow userland application know is lost one of stripes.
880 Details : fill lvb_blocks with error code on ost and return it to
881 application if error flag found.
885 Description: NULL lov_tgts causing MDS oops
886 Details : more safe checks for NULL lov_tgts for avoid oops.
888 Severity : enhancement
890 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
894 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
895 Details : If an OST is down the MDS will hang indefinitely in
896 obd_statfs_rqset() waiting for the statfs data. While for
897 MDS QOS usage of statfs, it should not stuck in waiting.
899 Severity : enhancement
901 Description: remote_acl support
902 Details : Support ACL-based permission check for remote user.
903 Support setfacl/getfacl for remote user with the utils
904 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
905 the system "{s,g}etfacl" utils.
907 Severity : enhancement
909 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
911 Severity : enhancement
913 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
917 Description: make mgs_setparam() handle fsname containing dash
918 Details : fsname containing a dash does not work with lctl conf_param
920 Severity : enhancement
922 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
924 Severity : enhancement
926 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
930 Frequency : rare, at shutdown
931 Description: access already free / zero obd_namespace.
932 Details : if client_disconnect_export was called without force flag set,
933 and exist connect request in flight, this can produce access to
934 NULL pointer (or already free pointer) when connect_interpret
935 store ocd flags in obd_namespace.
939 Frequency : only at startup
940 Description: not alloc memory with spinlock held.
941 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
942 if any spinlock held.
944 Severity : enhancement
946 Description: make lustre randomly fail allocating memory
947 Details : Make lustre randomly failed allocating memory for testing purpose.
949 Severity : enhancement
951 Description: lost problems with lov objid file
952 Details : Fixes some scability and access to not inited memory problems
953 in work with lov objdid file.
958 Description: lfs find does not continue on file error
959 Details : Continue other files processing when a file/dir is absent.
963 Description: Inconsistent usage of lustre_pack_reply()
964 Details : Standardize the usage of lustre_pack_reply() such that it
965 always generate a CERROR on failure.
968 Frequency : occasional
970 Description: MDS hang or stay in waiting lock
971 Details : If client receive lock with CBPENDING flag ldlm need send lock
972 cancel as separate rpc, to avoid situation when cancel request
973 can't processed due all i/o threads stay in wait lock.
977 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
981 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
985 Description: Update to SLES9 kernel-2.6.5-7.287.3.
989 Description: Update to RHEL4 latest kernel.
991 Severity : enhancement
993 Description: Build SLES10 patchless client fails
994 Details : The configure was broken by run ./configure with
995 --with-linux-obj=.... argument for patchless client. When the
996 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
997 can't search header adequately. Use absolute path such as
998 -I($LINUX)/include instead.
1002 Description: interrupt oig_wait produce painc on resend.
1003 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1004 private set, and this produce situation when rq_ptlrpcd_data not
1005 copyed to new allocated request and triggered LBUG on assert
1006 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1007 ptlrpcd set if request is changed and can be safety changed to use
1012 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1013 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1014 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1015 these macros and replace them with their difinitions instead.
1019 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1020 Details : Update which_patch & target file for SLES10 latest kernel.
1023 Description: Remove client patches from SLES 10 kernel.
1024 Details : This causes SLES 10 clients to behave as patchless clients
1025 even on a Lustre-patched (server) kernel.
1027 Severity : enhancement
1029 Description: self-adjustable client's lru lists
1030 Details : use adaptive algorithm for managing client cached locks lru
1031 lists according to current server load, other client's work
1032 pattern, memory activities, etc. Both, server and client
1033 side namespaces provide number of proc tunables for controlling
1036 Severity : enhancement
1038 Description: light-weight GSS support
1039 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1040 and reduce performance overhead.
1042 Severity : enhancement
1044 Description: Linux keyring support
1045 Details : Support using service of Linux keyring for Lustre GSS internal
1046 context refresh/cache mechanism.
1050 Description: Fix errors in lfs documentation
1051 Details : Fixes man pages
1055 Description: don't use GFP_* in generic Lustre code.
1056 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1060 Description: obdclass is limited by single OBD_ALLOC(idarray)
1061 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1065 Description: warnings with build patchless client with vanila 2.6.19 and up
1066 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1067 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1071 Frequency : Only for SLES
1072 Description: sanity_quota fail test_1
1073 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1074 file, which makes TSTID[2] inunique.
1078 Frequency : Only for SLES9
1079 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1080 Details : Update target/ChangeLog/which_patch .
1084 Frequency : Only for test_10 in sanity-quota.sh
1085 Description: a bug in quota test code
1086 Details : checking a test flag in a wrong place
1090 Frequency : Only for test_13 in sanity-quota.sh
1091 Description: a bug in quota test script
1092 Details : shouldn't check overfull allocated quota
1096 Description: jbd statistics
1097 Details : Port older jbd statistics patch for sles10
1099 Severity : enhancement
1101 Description: Add printing inode info into message about error in writepage.
1106 Description: reply_lock_interpret crash due to race with it and lock cancel.
1107 Details : Do not replay locks that are being cancelled. Do not reference
1108 locks by their address during replay, just by their handle.
1112 Frequency : When flocks are used.
1113 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1114 Details : Flock locks might destroy just granted lock if it could be merged
1115 with another existing flock, this is done in completion handler,
1116 so teach ldlm_cli_enquque_fini that this is a valid case for
1122 Description: Oops in read and write path when failing to allocate lock.
1123 Details : Check if lock allocation failed and return error back.
1127 Description: lstripe command fails for valid OST index
1128 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1129 instead of lov->desc.ld_active_tgt_count.
1133 Description: sanity.sh failed test 103
1134 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1139 Description: df doesn't work properly if diskfs blocksize != 4K
1140 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1144 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1145 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1147 Severity : enhancement
1149 Description: Adaptive timeouts
1150 Details : RPC timeouts adapt to changing server load and network
1151 conditions to reduce resend attempts and improve recovery time.
1155 Description: llapi_file_create() does not allow some changes
1156 Details : add llapi_file_open() that allows specifying the mode and
1157 open flags, and also returns an open file handle.
1161 Description: merge and cleanup kernel patches.
1162 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1166 Description: Add journal checksum support.(Kernel part)
1167 Details : The journal checksum feature adds two new flags i.e
1168 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1169 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1170 indicates that the commit block contains the checksum for
1171 the blocks described by the descriptor blocks. Now commit
1172 record can be sent to disk without waiting for descriptor
1173 blocks to be written to disk. This behavior is controlled
1174 using JBD2_FEATURE_ASYNC_COMMIT flag.
1178 Description: OSS needs mutliple precreate threads
1179 Details : Add ability to start more than one create thread per OSS.
1183 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1184 Details : execute lfs setstripe on client
1188 Description: mds_obd_create error creating tmp object
1189 Details : When the user sets quota on root, llog will be affected and can't
1190 create files and write files.
1193 Frequency : Always on ia64 patchless client, and possibly others.
1195 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1196 Details : This allows the patchless client to be loaded on architectures
1197 without this export.
1201 Description: RedHat Update kernel for RHEL5
1202 Details : Add the kernel config file for RHEL5.
1206 Description: RedHat Update kernel for RHEL5
1207 Details : Modify the kernel config file more closer RHEL5.
1211 Description: Build failure against Centos5 (RHEL5)
1212 Details : Define PAGE_SIZE when it isn't present.
1216 Description: test_7 on sanity-quota.sh can't be used on separate machines
1217 Details : add facet to handle it
1221 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1222 Details : with NFS, the anon dentry's parent was set to itself in
1223 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1224 corresponding dentry other than use rec->ur_name.
1226 Severity : enhancement
1228 Description: lfs setstripe enhancement
1229 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1233 Description: enable data checksumming by default
1234 Details : enable checksum by default, allow --disable-checksum
1235 configure option and "-o nochecksum" mount option. Checksums
1236 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1240 Description: lustre support for RHEL5
1241 Details : Add support for RHEL5.
1245 Description: Client eviction due to failover config
1246 Details : after a connection loss, the lustre client should attempt to
1247 reconnect to the last active server first before trying the
1248 other potential connections.
1252 Description: when mds and osts use different quota unit(32bit and 64bit),
1253 quota will be released repeatly.
1254 Details : void sending multiple quota reqs to mds, which will keep the
1255 status between the reqs.
1259 Description: osts not allocated evenly to files
1260 Details : change the condition to increase offset_idx
1263 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1265 Description: Data corruption for OSTs that are formatted larger than 2TB
1267 Details : When generating the bio request for lustre file writes the
1268 sector number would overflow a temporary variable before being
1269 used for the IO. The data reads correctly from Lustre (which
1270 will overflow in a similar manner) but other file data or
1271 filesystem metadata may be corrupted in some cases.
1275 Description: Tune the kernel for good SCSI performance.
1276 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1277 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1282 Description: rewrite ext2-derived code in obdclass/uuid.c
1283 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1284 from scratch preserving functionality.
1288 Description: "lfs find -obd UUID" prints directories
1289 Details : "lfs find -obd UUID" will return all directory names instead
1290 of just file names. It is incorrect because the directories
1291 do not reside on the OSTs.
1295 Description: use i_size_read and i_size_write in 2.6 port
1296 Details : replace inode->i_size access with i_size_read/write()
1299 Frequency : when removing large files
1301 Description: scheduling issue during removal of large Lustre files
1302 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1303 It causes scheduling issues when removing large files (17TB in the
1307 Frequency : only with liblustre clients on XT3
1309 Description: evictions taking too long
1310 Details : allow llrd to evict clients directly on OSTs
1313 Frequency : only on ppc
1315 Description: /proc/fs/lustre/devices broken on ppc
1316 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1317 We should make sure the type of 'index' is loff_t and then cast
1318 explicitly as needed below. Do not assign an explicitly cast
1322 Frequency : only for rhel5
1324 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1325 Details : Modify the target file & which_kernel.
1327 Severity : enhancement
1329 Description: omit set fsid for export NFS
1330 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1331 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1335 Frequency : Always, for kernels after 2.6.16
1336 Description: Fix warning idr_remove called for id=.. which is not allocated.
1337 Details : Last kernels save old s_dev before kill super and not allow
1338 to restore from callback - restore it before call kill_anon_super.
1342 Description: buffer overruns could theoretically occur
1343 Details : llapi_semantic_traverse() modifies the "path" argument by
1344 appending values to the end of the origin string, and a
1345 overrun may occur. Adding buffer overrun check in liblustreapi.
1349 Description: Fix error on 'ls .' at the top of the Lustre mount.
1350 Details : Don't revalidate dentry if it is a root dentry.
1354 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1355 Details : Modify vm-tunables-rhel4.patch.
1359 Description: Kernel config for 2.6.18-vanilla.
1360 Details : Modify targets/2.6-vanilla.target.in.
1361 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1362 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1363 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1364 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1368 Description: improve handling recoverable errors
1369 Details : if request processig with error which can be recoverable on server
1370 request should be resend, otherwise page released from cache and
1375 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1376 Details : Modify target file & which_patch.
1377 A flaw was found in the IA32 system call emulation provided
1378 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1379 value could be stored in the %RAX register, which could trigger an
1380 out-of-bounds system call table access. An untrusted local user
1381 could exploit this flaw to run code in the kernel
1382 (ie a root privilege escalation). (CVE-2007-4573).
1386 Description: O_DIRECT bypasses client statistics.
1387 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1388 read_ahead_stats were updated. copied the stats section from
1389 osc_send_oap_rpc() into async_internal().
1393 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1397 Description: change order of libsysio includes
1398 Details : '#include sysio.h' should always come before '#include xtio.h'
1402 Description: handle "serious error: objid * already exists" more gracefully
1403 Details : If LAST_ID value on disk is smaller than the objects existing in
1404 the O/0/d* directories, it indicates disk corruption and causes an
1405 LBUG(). If the object is 0-length, then we should use the existing
1406 object. This will help to avoid a full fsck in most cases.
1408 Severity : enhancement
1410 Description: adapt the lustre_config script to support the upgrade case
1411 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1412 server targets to 1.6 in parallel.
1416 Description: To avoid grant space > avaible space when the disk is almost
1417 full. Without this patch you might see the error "grant XXXX >
1418 available" or some LBUG about grant, when the disk is almost
1420 Details : In filter_check_grant, for non_grant cache write, we should
1421 check the left space by if (*left > ungranted + bytes), instead
1422 of (*left > ungranted), because only we are sure the left space
1423 is enough for another "bytes", then the ungrant space should be
1424 increase. In client, we should update cl_avail_grant only there
1425 is OBD_MD_FLGRANT in the reply.
1429 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1430 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1431 described in CVE-2007-4573. This problem could allow untrusted
1432 local users to gain root access.
1435 Frequency : when using O_DIRECT and quotas
1437 Description: Incorrect file ownership on O_DIRECT output files
1438 Details : block usage reported by 'lfs quota' does not take into account
1439 files that have been written with O_DIRECT.
1444 Description: touch file failed when fs is not full
1445 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1446 otherwise we can get ENOSP while fs is not full.
1450 Description: parallel lock callbacks
1451 Details : Instead of sending blocking and completion callbacks as separated
1452 requests, adding them to a set and sending in parallel.
1455 Frequency : only for Cray XT3
1456 Bugzilla : 12829/13455
1457 Description: Changing primary group doesn't change the group lustre assigns to
1459 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1460 overridden with the primary group provided by the group upcall,
1461 whereas the supplied fsgid can be trusted if it is in the list of
1462 supplementary groups returned by the group upcall.
1464 Severity : enhancement
1466 Description: Allow masking D_WARNING, D_ERROR messages from console
1467 Details : Console messages can now be disabled via lnet.printk.
1471 Description: User code with malformed file open parameter crashes client node
1472 Details : Before packing join_file req, all the related reference should be
1473 checked carefully in case some malformed flags cause fake
1474 join_file req on client.
1478 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1479 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1480 mode is equal to the granted mode. However, it should also check
1481 that LDLM_FL_CP_REQD is not set.
1485 Description: Heavy nfs access might result in deadlocks
1486 Details : After ELC code landed, it is now improper to enqueue any mds
1487 locks under och_sem, because enqueue might want to decide to
1488 cancel open locks for same inode we are holding och_sem for.
1492 Description: Client eviction while running blogbench
1493 Details : A lot of unlink operations with concurrent I/O can lead to a
1494 deadlock causing evictions. To address the problem, the number of
1495 oustanding OST_DESTROY requests is now throttled to
1496 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1497 callbacks are processed in priority.
1501 Description: enable ACLs on MDS by default
1502 Details : ACLs must be enabled on MDS by default.
1505 Frequency : PPC/PPC64 only
1507 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1508 Details : fix duplicated definitions between asm-ppc64/types.h and
1509 lustre_types.h on PPC.
1512 Frequency : PPC/PPC64 only
1514 Description: asm-ppc/segment.h does not exist
1515 Details : fix compile issue on PPC.
1519 Description: better handle error messages in extents code
1522 Frequency : RHEL4 only
1524 Description: mkfs is very slow on IA64/RHEL4
1525 Details : A performance regression has been discovered in the MPT Fusion
1526 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1527 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1528 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1530 Severity : enhancement
1532 Description: SNMP support enhancement
1533 Details : Adding total number of sampled request for an MDS node in snmp
1536 Severity : enhancement
1538 Description: Optimize ldlm waiting list processing for PR extent locks
1539 Details : When processing waiting list for read extent lock and meeting read
1540 lock that is same or wider to it that is not contended, skip
1541 processing rest of the list and immediatelly return current
1542 status of conflictness, since we are guaranteed there are no
1543 conflicting locks in the rest of the list.
1547 Description: Time out and refuse to reconnect
1548 Details : When the failover node is the primary node, it is possible
1549 to have two identical connections in imp_conn_list. We must
1550 compare not conn's pointers but NIDs, otherwise we can defeat
1551 connection throttling.
1555 Description: port llog fixes from b1_6 into HEAD
1556 Details : Port llog reference couting and some llog cleanups from b1_6
1557 (bug 10800) into HEAD, for protect from panic and access to already
1558 free llog structures.
1562 Description: Detect stride IO mode in read-ahead
1563 Details : When a client does stride read, read-ahead should detect that and
1564 read-ahead pages according to the detected stride pattern.
1568 Description: data checksumming impacts single node performance
1569 Details : add support for several checksum algorithm. Currently, only CRC32
1570 and Adler-32 are supported. The checksum type can be changed on
1571 the fly via /proc/fs/lustre/osc/*/checksum_type.
1575 Description: use adler32 for page checksums
1576 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1581 Description: build for x2 fails
1582 Details : fix compile issue on Cray systems.
1586 Description: Properly match for duplicate locks
1587 Details : Due to different lock order from skiplists code, we need to
1588 traverse entire list for now
1591 Frequency : only on PPC/SLES10
1593 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1594 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1595 be defined. Add a hack in mkfs_lustre.c to work around this compile
1600 Description: LASSERT on MDS when client holding flock lock dies
1601 Details : ldlm pool logic depends on number of granted locks equal to
1602 number of released locks which is not true for flock locks, so
1603 just exclude such locks from consideration.
1607 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1608 Details : Use fsfilt_sync() for both the whole filesystem sync and
1609 individual file sync to eliminate dangerous inode locking
1610 with I_LOCK that can lead to a deadlock.
1614 Description: performance in 1.6.3
1615 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1616 sure that 1MB requests are not fragmented by the block layer.
1618 Severity : enhancement
1620 Description: organize the server-side client stats on per-nid basis
1621 Details : Change the structure of stats under obdfilter and mds to
1629 The "uuid"s file would list the uuids of _active_ exports.
1630 And the clear entry is to clear all stats and stale nids.
1632 Severity : enhancement
1634 Description: eliminate client locks in face of contention
1635 Details : file contention detection and lockless i/o implementation
1636 for contended files.
1640 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1643 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1645 Description: MDS may lose file striping (and hence file data) in some cases
1646 Details : If there are additional extended attributes stored on the MDS,
1647 in particular ACLs, SELinux, or user attributes (if user_xattr
1648 is specified for the client mount options) then there is a risk
1649 of attribute loss. Additionally, the Lustre file striping
1650 needs to be larger than default (e.g. striped over all OSTs),
1651 and an additional attribute must be stored initially in the
1652 inode and then increase in size enough to be moved to the
1653 external attribute block (e.g. ACL growing in size) for file
1658 Description: skiplist implementation simplification
1659 Details : skiplists are used to group compatible locks on granted list
1660 that was implemented as tracking first and last lock of each
1661 lock group the patch changes that to using doubly linked lists
1665 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1666 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1667 because LASSERT on both the data supplied by a client, and the
1668 data on disk is dangerous and incorrect.
1670 Severity : enhancement
1672 Description: Slow truncate/writes to huge files at high offsets.
1673 Details : Directly associate cached pages to lock that protect those pages,
1674 this allows us to quickly find what pages to write and remove
1675 once lock callback is received.
1679 Description: more ldlm soft lockups
1680 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1681 starve other threads from the resource lock for a long time in
1682 case of long waiting queue, so change the debug level from
1683 D_OTHER to the less frequently used D_INFO.
1685 Severity : enhancement
1687 Description: add -gid, -group, -uid, -user options to lfs find
1691 Description: Hung threads in invalidate_inode_pages2_range
1692 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1693 once one is completed. As a result, some RPCs are stuck in the
1694 queue and are never sent.
1698 Description: filter threads hungs on waiting journal commit
1699 Details : Cleanup filter group llog code, then only filter group llog will
1700 be only created in the MDS/OST syncing process.
1704 Description: Procfs and llog threads access destoryed import sometimes.
1705 Details : Sync the import destoryed process with procfs and llog threads by
1706 the import refcount and semaphore.
1708 Severity : enhancement
1710 Description: openlock cache of b1_6 port to HEAD
1715 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1716 Details : If insertion of an extent fails, then discard the inode
1717 preallocation and free data blocks else it can lead to duplicate
1722 Description: don't always update ctime in ext3_xattr_set_handle()
1723 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1724 In some cases the ctime should not be updated, for example for
1725 2.0->1.8 compatibility it is necessary to delete an xattr and it
1726 should not update the ctime.
1730 Bugzilla : 15713/16362
1731 Description: Assertion in iopen_connect_dentry in 1.6.3
1732 Details : looking up an inode via iopen with the wrong generation number can
1733 populate the dcache with a disconneced dentry while the inode
1734 number is in the process of being reallocated. This causes an
1735 assertion failure in iopen since the inode's dentry list contains
1736 both a connected and disconnected dentry.
1740 Description: assertion failure in ldlm_handle2lock()
1741 Details : fix a race between class_handle_unhash() and class_handle2object()
1742 introduced in lustre 1.6.5 by bug 13622.
1747 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1748 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1753 Description: oops in sd_iostats_seq_show()
1754 Details : unloading/reloading the scsi low level driver triggers a kernel
1755 bug when trying to access the sd iostat file.
1760 Description: Kernel panics during QLogic driver reload
1761 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1762 patch, causing memory corruption.
1767 Description: journal_dev option does not work in b1_6
1768 Details : pass mount option during pre-mount.
1770 Severity : enhancement
1772 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1773 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1774 extent information of a file. It can be used to map logical blocks
1775 in a file to physical blocks in the block device.
1779 Description: LDLM soft lockups - improvement
1780 Details : It is be possible to send the lock handle along with each read
1781 or write request because the client is already doing a lock match
1782 itself so there isn't any reason the OST should have to re-do that
1786 Frequency : only on Cray X2
1788 Description: X2 build failures
1789 Details : fix build failures on Cray X2.
1793 Description: xid & resent requests
1794 Details : Initialize RPC XID from clock at startup (randomly if clock is
1797 Severity : enhancement
1799 Description: Add lustre_start utility to start or stop multiple Lustre servers
1804 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1805 Details : In case of memory pressure, list_del() can be called twice on
1806 req->rq_history_list, causing a kernel oops.
1810 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1811 Details : fix stack overflow in the distributed lock manager by defering
1812 export eviction after a failed AST to the elt thread instead of
1813 handling it in the dlm interpret routine.
1817 Description: Convert some comments to new format.
1818 Details : Update documenting comments to match doxygen conventions.
1822 Description: Grammar fixes.
1823 Details : A couple of trivial spelling fixes.
1827 Description: OSD_COUNTERS-mandatory
1828 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1829 to simplify the code.
1833 Description: simplify cmm_device freeing logic.
1834 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1838 Description: Add lockdep support to dt_object_operations locking interface.
1839 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1840 parameter indicating lock ordering. Update mdd code to use new
1845 Description: Introduce failloc constants for lockless IO tests.
1846 Details : Add two new failloc constants to test lockless IO. Only one of
1847 them in implemented---another is checked in yet to be landed
1852 Description: Add lockdep support for inode mutex.
1853 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1854 in the situations where Lustre has to lock more than one inode
1859 Description: Add optional invariants checking support.
1860 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1861 configure switch. This macro is to be used for consistency and
1862 sanity checks that are too expensive to be left in `production'
1867 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1868 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1873 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1874 Details : Use spin_lock_nested() in (the only) situation where more than
1875 one ldlm_lock is locked simultaneously. Also, fix possible
1876 dead-lock in ldlm_lock_change_resource() by enforcing particular
1881 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1882 Details : Instead of specifying each ldlm_lock call-back through separate
1883 parameter, wrap them into struct ldlm_callback_suite.
1887 Description: Kill join_lru obd method and its callers.
1888 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1889 in memory---a requirement implemented through ->o_join_lru() obd
1890 method in HEAD. Remove this method and its users.
1894 Description: Add asynchronous ldlm ENQUEUE completion handler.
1895 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1896 case that stock ldlm_completion_ast() cannot handle as it waits
1897 until lock is granted. Introduce new ldlm_completion_ast_async()
1898 for this. Also comment ldlm_completion_ast().
1902 Description: ldlm_error <-> errno conversion.
1903 Details : Add functions to map (rather arbitrary) between LDLM error codes
1904 and standard errno values. CLIO needs this to prevent LDLM specific
1905 constants from escaping ldlm and osc.
1909 Description: Kill unused ldlm_handle2lock_ns() function.
1910 Details : Kill unused ldlm_handle2lock_ns() function.
1914 Description: Add lu_ref support to ldlm_lock
1915 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1916 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1917 and ldlm_resource. LDLM interface has to be changed, because code that
1918 releases a reference on a lock, has to "know" what reference this is.
1919 In the most frequent case
1921 lock = ldlm_handle2lock(handle);
1923 LDLM_LOCK_PUT(lock);
1925 no changes are required. When any other reference (received _not_ from
1926 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1927 instead of LDLM_LOCK_PUT().
1929 Arguably, changes are pervasive, and interface requires some discipline
1930 for proper use. On the other hand, it was very instrumental in finding
1931 a few leaked lock references.
1935 Description: Add ldlm_lock_addref_try().
1936 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1937 attempts to addref a lock that might be being canceled
1942 Description: Add ldlm_weigh_callback().
1943 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1944 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1945 hard-coded `number of pages' logic.
1949 Description: Add lockdep annotations to llog code.
1950 Details : Use appropriately tagged _nested() locking calls in the places
1951 where llog takes more than one ->lgh_lock lock.
1955 Description: Add loi_kms_set().
1956 Details : Wrap kms updates into a helper function.
1960 Description: Constify instances of struct lsm_operations.
1961 Details : Constify instances of struct lsm_operations.
1965 Description: lu_conf support.
1966 Details : On a server, a file system object is uniquely identified
1967 by a fid, which is sufficient to locate and load all object
1968 state (inode). On a client, on the other hand, more data are
1969 necessary instantiate an object. Change lu_object_find() and
1970 friends to take additional `lu_conf' argument describing object.
1971 Typically this includes layout information.
1975 Description: lu_context fixes.
1976 Details : Introduce new lu_context functions that are needed on the client
1977 side, where some system threads (ptlrpcd) are shared by multiple
1978 modules, and so cannot be stopped during module shutdown.
1982 Description: Add start and stop methods to lu_device_type_operations.
1983 Details : Introduce two new methods in lu_device_type_operations, that are
1984 invoked when first instance of a given type is created and last one
1985 is destroyed respectively. This is need by CLIO.
1989 Description: Add lu_ref support to struct lu_device.
1990 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1991 track leaked references.
1995 Description: Introduce lu_kmem_descr.
1996 Details : lu_kmem_descr and its companion interface allow to create
1997 and destroy a number of kmem caches at once.
2001 Description: Fix lu_object finalization race.
2002 Details : Fix a race between lu_object_find() finding an object and its
2003 concurrent finalization. This race is (most likely) not possible
2004 on the server, but might happen on the client.
2008 Description: Introduce lu_ref interface.
2009 Details : lu_ref is a debugging module allowing to track references to
2010 a given object. It is quite cpu expensive, and has to be
2011 explicitly enabled with --enable-lu_ref. See usage description
2016 Description: Factor lu_site procfs stats into a separate function.
2017 Details : Separate lu_site stats printing code into a separate function
2018 that can be reused on a client.
2022 Description: Constify instances of struct {lu,dt,md}_device_operations.
2023 Details : Constify instances of struct {lu,dt,md}_device_operations.
2027 Description: Introduce struct md_site and move meta-data specific parts of
2028 struct lu_site here.
2029 Details : Move md-specific fields out of struct lu_site into special struct
2030 md_site, so that lu_site can be used on a client.
2034 Description: Kill mdd_lov_destroy().
2035 Details : Remove unused mdd code.
2039 Description: Add st_block checking to multistat.c.
2040 Details : Add st_block checking to multistat.c.
2044 Description: Add lu_ref support to struct obd_device.
2045 Details : Add lu_ref tracking to obd_device.
2049 Description: Kill obd_set_fail_loc().
2050 Details : Remove unused code.
2054 Description: Add special type for ptlrpc_request interpret functions.
2055 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2056 there. Actual usage will be in CLIO.
2060 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2061 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2062 doesn't work with lockdep.
2066 Description: Add rwv.c test program.
2067 Details : New testing program exercising readv(2) and writev(2) (Qian).
2071 Description: Add sendfile.c test program.
2072 Details : New testing program exercising sendfile(2) (Jay).
2076 Description: Ratelimit a message that can be very frequent.
2077 Details : Ratelimit a memory allocation failure message that can
2082 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2083 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2087 Description: Liblustre build fixes.
2088 Details : Liblustre build fixes.
2092 Description: libcfs: add cfs_{need,cond}_resched() interface.
2093 Details : libcfs: add cfs_{need,cond}_resched() definition and
2094 implementations for Linux, NT, and liblustre.
2096 Severity : enhancement
2098 Description: More exported tunables for mballoc
2099 Details : Add support for tunable preallocation window and new tunables for
2100 large/small requests
2104 Description: Detect corruption of block bitmap and checking for preallocations
2105 Details : Checks validity of on-disk block bitmap. Also it does better
2106 checking of number of applied preallocations. When corruption is
2107 found, it turns filesystem readonly to prevent further corruptions.
2111 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2112 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2113 window to 0. If it does hit cache miss, and read-pattern is still
2114 stride-io mode, does not reset the stride window, but also does not
2115 increase the stride window length in this case.
2119 Frequency : only for big-endian servers
2120 Description: Check if system is big-endian while mounting fs with extents feature
2121 Details : Mounting a filesystem with extents feature will fail on big-endian
2122 systems since ext3-based ldiskfs is not supported on big-endian
2123 systems. This can be over-riden with "bigendian_extents" mount option.
2125 Severity : enhancement
2127 Description: The root squash functionality
2128 Details : A security feature, which is to prevent users from being able
2129 to mount lustre on their desktop, run as root, and delete
2130 all of the files in the filesystem. The goal is accomplished by
2131 remapping user id (UID) and group id (GID) of the root user to
2132 a UID and GID specified by the system administartor via Lustre
2133 configuration management server (MGS). The functionality also
2134 allows to specify sets of clients for which the remapping does
2139 Description: Excessive recovery window
2140 Details : With AT enabled, the recovery window can be excessively long (6000+
2141 seconds). To address this problem, we no longer use
2142 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2143 timeout no longer depends on the service time, it is set to
2144 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2145 time via pb_service_time.
2149 Description: Watchdog triggered on MDS failover
2150 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2151 the OSTs know that the MDS "UUID" can be reused for the same export
2152 from a different NID, so we do not need to wait for the export to be
2156 Frequency : rare, only if using MMP with Linux RAID
2158 Description: MMP doesn't work with Linux RAID
2159 Details : While using HA for Lustre servers with Linux RAID, it is possible
2160 that MMP will not detect multiple mounts. To make this work we
2161 need to unplug the device queue in RAID when the MMP block is being
2162 written. Also while reading the MMP block, we should read it from
2163 disk and not the cached one.
2165 Severity : enhancement
2167 Description: open file using fid
2168 Details : A file can be opened using just its fid, like
2169 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2172 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2175 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2176 Details : add proper locking for creation of jbd_4k slab cache
2180 Description: MMP check in ext3_remount() fails without displaying any error
2181 Details : When multiple mount protection fails during remount, proper error
2184 Severity : enhancement
2186 Description: Allow stripe size to be up to 4G-64k
2187 Details : Fix math logic to allow large stripe sizes.
2191 Description: add check for >8TB ldiskfs filesystems
2192 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2193 Don't allow >8TB ldiskfs filesystems to be mounted without
2194 force_over_8tb mount option
2199 Descriptoin: OSS complains "too many blocks" and crashes
2200 Details : In filter_commitrw_write():
2201 1. lquota_chkquota();
2202 2. obdo_from_inode();
2203 3. lquota_pending_commit();
2204 "2." has changed uid and gid in this case, then caused
2205 lquota_chkquota() and lquota_pending_commit() to use
2206 different uid and gid. That is the root of the bug.
2208 --------------------------------------------------------------------------------
2210 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2212 * Support for kernels:
2213 2.6.5-7.283 (SLES 9),
2214 2.6.9-55.EL (RHEL 4),
2215 2.6.16.46-0.14 (SLES 10),
2216 2.6.18.8 vanilla (kernel.org)
2217 * Client support for unpatched kernels:
2218 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2219 2.6.16 - 2.6.22 vanilla (kernel.org)
2220 * Due to recently discovered recovery problems, we do not recommend
2221 using patchless RHEL 4 clients with this or any earlier release.
2222 * Recommended e2fsprogs version: 1.39.cfs8
2223 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2224 * Starting with this release, the ldiskfs backing filesystem required
2225 by Lustre is now in its own package, lustre-ldiskfs. This package
2226 should be installed. It is versioned separately from Lustre and
2227 may be released separately in future.
2231 Description: block reactivating mgc import until all deactivates complete
2232 Details : Fix race when failing back MDT/MGS to itself (testing)
2235 Frequency : at statup only
2237 Description: mds_lov_synchronize race leads to various problems
2238 Details : simultaneous MDT->OST connections at startup can cause the
2239 sync to abort, leaving the OSC in a bad state.
2241 Severity : enhancement
2243 Description: add optional extra BUILD_VERSION info
2244 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2245 to override the lustre version.
2248 Frequency : 2.6.18 servers only
2250 Description: ll_kern_mount() doesn't release the module reference
2251 Details : The ldiskfs module reference count never drops down to 0
2252 because ll_kern_mount() doesn't release the module reference.
2257 Description: server LBUG when using old ost_num_threads parameter
2258 Details : Accept the old ost_num_threads parameter but warn that it
2259 is deprecated, and fix an off-by-one error that caused an LBUG.
2264 Description: Transient SCSI error results in persistent IO issue
2265 Details : iobuf->dr_error is not reinitialized to 0 between two
2269 Frequency : sometimes when underlying device returns I/O errors
2271 Description: OSTs not going read-only during write failures
2272 Details : OSTs are not remounted read-only when the journal commit threads
2273 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2274 instead of the ext3 wrappers.
2278 Description: poor connect scaling with increasing client count
2279 Details : Don't run filter_grant_sanity_check for more than 100 exports
2280 to improve scaling for large numbers of clients.
2283 Frequency : SLES10 only
2285 Description: sanity-quota.sh quotacheck failed: rc = -22
2286 Details : Quotas cannot be enabled on SLES10.
2289 Frequency : liblustre clients only
2291 Description: getdirentries does not give error when run on compute nodes
2292 Details : getdirentries does not fail when the size specified as an argument
2293 is too small to contain at least one entry
2295 Severity : enhancement
2297 Description: Add LNET router traceability for debug purposes
2298 Details : If a checksum failure occurs with a router as part of the
2299 IO path, the NID of the last router that forwarded the bulk data
2300 is printed so it can be identified.
2305 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2306 Details : Due to a race condition, liblustre clients were occasionally
2307 evicted incorrectly.
2309 Severity : enhancement
2311 Description: lfs setstripe use optional parameters instead of postional
2314 Severity : enhancement
2316 Description: Nanosecond timestamp support for ldiskfs
2317 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2318 resolution timestamps. There is not yet support for this at
2319 the Lustre filesystem level.
2322 Frequency : during server recovery
2324 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2325 Details : request with rq_no_resend flag not awake l_wait_event if they get
2329 Frequency : nfs export on patchless client
2331 Description: connectathon hang when test nfs export over patchless client
2332 Details : Disconnected dentry cannot be found with lookup, so we do not need
2333 to unhash it or make it invalid
2336 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2337 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2338 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2343 Description: open req refcounting wrong on reconnect
2344 Details : If reconnect happened between getting open reply from server and
2345 call to mdc_set_replay_data in ll_file_open, we will schedule
2346 replay for unreferenced request that we are about to free.
2347 Subsequent close will crash in variety of ways.
2348 Check that request is still eligible for replay in
2349 mdc_set_replay_data().
2354 Description: disable writes to filesystem when reading health_check file
2355 Details : the default for reading the health_check proc file has changed
2356 to NOT do a journal transaction and write to disk, because this
2357 can cause reads of the /proc file to hang and block HA state
2358 checking on a healthy but otherwise heavily loaded system. It
2359 is possible to return to the previous behaviour during configure
2360 with --enable-health-write.
2362 Severity : enhancement
2364 Description: 64-bit inode version
2365 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2366 to the inode. This will be required for version-based recovery.
2371 Description: MDS fails to start if a duplicate client export is detected
2372 Details : in some rare cases it was possible for a client to connect to
2373 an MDS multiple times. Upon recovery the MDS would detect this
2374 and fail during startup. Handle this more gracefully.
2376 Severity : enhancement
2378 Description: Add -o localflock option to simulate old noflock
2380 Details : This will achieve local-only flock/fcntl locks
2386 Description: log_commit_thread vs filter_destroy race leads to crash
2387 Details : Take import reference before releasing llog record semaphore
2392 Description: Wrong request locking in request set processing
2393 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2394 imp_delayed_list, in this place should be used imp_lock.
2397 Frequency : when reconnection
2399 Description: Grant Leak when osc reconnect to OST
2400 Details : When osc reconnect ost, OST(filter) should check whether it
2401 should grant more space to client by comparing fed_grant and
2402 cl_avail_grant, and return the granted space to client instead
2403 of "new granted" space, because client will call osc_init_grant
2404 to update the client grant space info.
2407 Frequency : when client reconnect to OST
2409 Description: Grant Leak when osc do resend and replay bulk write
2410 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2411 bulk write request, because the grant info while be sync between
2412 OSC and OST when reconnect, and we should ignore the grant info
2413 these of resend/replay write req.
2418 Description: Grant space more than avaiable left space sometimes.
2419 Details : When then OST is about to be full, if two bulk writing from
2420 different clients came to OST. Accord the avaliable space of the
2421 OST, the first req should be permitted, and the second one
2422 should be denied by ENOSPC. But if the seconde arrived before
2423 the first one is commited. The OST might wrongly permit second
2424 writing, which will cause grant space > avaiable space.
2427 Frequency : when client is evicted
2429 Description: Grant might be wrongly erased when osc is evicted by OST
2430 Details : when the import is evicted by server, it will fork another
2431 thread ptlrpc_invalidate_import_thread to invalidate the
2432 import, where the grant will be set to 0. While the original
2433 thread will update the grant it got when connecting. So if
2434 the former happened latter, the grant will be wrongly errased
2435 because of this race.
2440 Description: Checking Stale with correct fid
2441 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2442 to check whether it is stale, because sometimes, we want the
2443 enqueue happened anyway, and op_data.fid2 will not be initialized.
2445 Severity : enhancement
2447 Description: update patchless client
2448 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2451 Frequency : only with 2.4 kernel
2453 Description: random memory corruption
2454 Details : size of struct ll_inode_info is to big for union inode.u and this
2455 can be cause of random memory corruption.
2460 Description: Memory leak in recovery
2461 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2462 It should also check obd_fail before fsfilt_start, otherwise if
2463 fsfilt_start return -EROFS,(failover mds during mds recovery).
2464 then the req will return with repmsg->transno = 0 and rc = EROFS.
2465 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2466 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2467 be freed no matter whether fsfilt_commit success or not.
2470 Frequency : only with huge count clients
2472 Description: Prevents from taking the superblock lock in llap_from_page for
2474 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2475 for llap_from_page call in ll_removepage prevents from taking the
2476 superblock lock for a soon died page.
2481 Description: Not check open intent error before release open handle
2482 Details : in some rare cases, the open intent error is not checked before
2483 release open handle, which may cause
2484 ASSERTION(open_req->rq_transno != 0), because it tries to release
2485 the failed open handle.
2490 Description: Set cat log bitmap only after create log success.
2491 Details : in some rare cases, the cat log bitmap is set too early. and it
2492 should be set only after create log success.
2496 Description: Accessing a block bevice can re-enable I/O when Lustre is
2497 tearing down a device.
2498 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2503 Description: service threads may hog cpus when there are a lot of requests
2504 Details : Insert cond_resched to give other threads a chance to use some CPU
2509 Description: the cat log was not initialized in recovery
2510 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2511 unlink log on mds will not be initialized until mds post
2512 recovery. And also in mds post recovery, the unlink log will
2513 initialization will be done asynchronausly, so there will be race
2514 between add unlink log and unlink log initialization.
2518 Description: brw_stats were being printed incorrectly
2519 Details : brw_stats were being printed as log2 but all of them were not
2520 recorded as log2. Also remove some code duplication arising from
2521 filter_tally_{read,write}.
2525 Frequency : rare, only in recovery.
2526 Description: ASSERTION(req->rq_type != LI_POISON) failed
2527 Details : imp_lock should be held while iterating over imp_sending_list for
2528 prevent destroy request after get timeout in ptlrpc_queue_wait.
2532 Description: replay-single.sh test 52 fails
2533 Details : A lock's skiplist need to be cleanup when it being unlinked
2534 from its resource list.
2538 Description: Short directio read returns full requested size rather than
2540 Details : Direct I/O operations should return actual amount of bytes
2541 transferred rather than requested size.
2543 Severity : enhancement
2545 Description: metadata RPC reduction (e.g. for rm performance)
2546 Details : decrease the amount of synchronous RPC between clients and servers
2547 by canceling conflicing lock before the operation on the client
2548 and packing thier handles into the main operation RPC to server.
2550 Severity : enhancement
2552 Description: Async OSC create to avoid the blocking unnecessarily.
2553 Details : If a OST has no remain object, system will block on the creating
2554 when need to create a new object on this OST. Now, ways use
2555 pre-created objects when available, instead of blocking on an
2556 empty osc while others are not empty. If we must block, we block
2557 for the shortest possible period of time.
2561 Description: improve handling recoverable errors
2562 Details : if request processig with error which can be recoverable on server
2563 request should be resend, otherwise page released from cache and
2566 Severity : enhancement
2568 Description: refine locking for avoid write wrong info into lov_objid file
2569 Details : fix possible races with add new target and write/update data in
2572 --------------------------------------------------------------------------------
2574 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2579 Frequency : on some architectures
2581 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2582 Details : Uninitialized flags sometimes cause configuration commands to
2586 Frequency : patchless clients only
2588 Description: missing __iget() symbol export
2589 Details : The __iget() symbol export is missing. To avoid the need for
2590 this on patchless clients the deathrow inode reaper is turned
2591 off, and we depend on the VM to clean up old inodes. This
2592 dependency was during via the fix for bug 12181.
2594 --------------------------------------------------------------------------------
2596 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2598 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2599 INTEROPERATE with older versions automatically. In many cases a
2600 special upgrade step is needed. Please read the
2601 user documentation before upgrading any part of a live system.
2602 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2603 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2604 * WARNING: Lustre configuration and startup changes are required with
2605 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2609 Severity : enhancement
2611 Description: MountConf
2612 Details : Lustre configuration is now managed via mkfs and mount
2613 commands instead of lmc and lconf. New obd types (MGS, MGC)
2614 are added for dynamic configuration management. See
2615 https://mail.clusterfs.com/wikis/lustre/MountConf for
2618 Severity : enhancement
2620 Description: dynamic OST addition
2621 Details : OSTs can now be added to a live filesystem
2623 Severity : enhancement
2625 Description: startup order invariance
2626 Details : MDTs and OSTs can be started in any order. Clients only
2627 require the MDT to complete startup.
2629 Severity : enhancement
2631 Description: parallel, asynchronous orphan cleanup
2632 Details : orphan cleanup is now performed in separate threads for each
2633 OST, allowing parallel non-blocking operation.
2635 Severity : enhancement
2637 Description: optimized stripe assignment
2638 Details : stripe assignments are now made based on ost space available,
2639 ost previous usage, and OSS previous usage, in order to try
2640 to optimize storage space and networking resources.
2642 Severity : enhancement
2644 Description: Permanently set tunables
2645 Details : All writable /proc/fs/lustre tunables can now be permanently
2646 set on a per-server basis, at mkfs time or on a live system.
2648 Severity : enhancement
2650 Description: Lustre message v2
2651 Details : Add lustre message format v2.
2653 Severity : enhancement
2655 Description: client OST exclusion list
2656 Details : Clients can be started with a list of OSTs that should be
2657 declared "inactive" for known non-responsive OSTs.
2661 Description: SPEC SFS validation failure on NFS v2 over lustre.
2662 Details : Changes the blocksize for regular files to be 2x RPC size,
2663 and not depend on stripe size.
2665 Severity : enhancement
2667 Description: Multiple MD RPCs in flight.
2668 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2669 To avoid overly-overloading MDS, introduce a limit on number of
2670 MDS RPCs in flight for a single client and add /proc controls
2671 to adjust this limit.
2673 Severity : enhancement
2675 Description: client read/write statistics
2676 Details : Add client read/write call usage stats for performance
2677 analysis of user processes.
2678 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2679 file access. extents_stats shows chunk size distribution.
2680 extents_stats_per_process show chunk size distribution per
2683 Severity : enhancement
2685 Description: mds statistics
2686 Details : Add detailed mds operations statistics in
2687 /proc/fs/lustre/mds/*/stats.
2691 Description: Failure of copying files with lustre special EAs.
2692 Details : Client side always return success for setxattr call for lustre
2693 special xattr (currently only "trusted.lov").
2698 Description: Refcount LNET uuids
2699 Details : The global LNET uuid list grew linearly with every startup;
2700 refcount repeated list entries instead of always adding to
2703 Severity : enhancement
2705 Description: Dynamic service threads
2706 Details : Within a small range, start extra service threads
2707 automatically when the request queue builds up.
2710 Frequency : mixed-endian client/server environments
2712 Description: mixed-endian crashes
2713 Details : The new msg_v2 system had some failures in mixed-endian
2716 Severity : enhancement
2718 Description: Easy OST removal
2719 Details : OSTs can be permanently deactivated with e.g. 'lctl
2720 conf_param lustre-OST0001.osc.active=0'
2722 Severity : enhancement
2724 Description: MGS proc entries
2725 Details : Added basic proc entries for the MGS showing what filesystems
2728 Severity : enhancement
2730 Description: provide MGS failover
2731 Details : Added config lock reacquisition after MGS server failover.
2733 Severity : enhancement
2735 Description: add Linux 2.4 support
2736 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2740 Description: a large application tries to do I/O to the same resource and dies
2741 in the middle of it.
2742 Details : Check the req->rq_arrival time after the call to
2743 ost_brw_lock_get(), but before we do anything about
2744 processing it & sending the BULK transfer request. This
2745 should help move old stale pending locks off the queue as
2746 quickly as obd_timeout.
2749 Frequency : when an incorrect nid is specified during startup
2751 Description: ptlrpc connect to non-existant node causes kernel crash
2752 Details : LNET can't be re-entered from an event callback, which
2753 happened when we expire a message after the export has been
2754 cleaned up. Instead, hand the zombie cleanup off to another
2757 Severity : enhancement
2759 Description: plain/inodebits lock performance improvement
2760 Details : Grouping plain/inodebits in granted list by their request modes
2761 and bits policy, thus improving the performance of search through
2765 Frequency : only if OST filesystem is corrupted
2767 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2768 Details : for a short time RPCs with bulk IO are in the replay list,
2769 but replay of bulk IOs is unimplemented. If the OST filesystem
2770 is corrupted due to disk cache incoherency and then replay is
2771 started it is possible to trip an assertion. Avoid putting
2772 committed RPCs into the replay list at all to avoid this issue.
2775 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2777 Description: System hang on startup
2778 Details : This bug allowed the liblustre (e.g. catamount) client to
2779 return to the app before handling all startup RPCs. This
2780 could leave the node unresponsive to lustre network traffic
2781 and manifested as a server ptllnd timeout.
2783 Severity : enhancement
2785 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2786 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2787 boolean to control whether to print peer debug info when a
2788 client's RPC times out.
2791 Frequency : only for kernels with patches from Lustre below 1.4.3
2793 Description: Remove old rdonly API
2794 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2797 Frequency : only for devices with external journals
2799 Description: Set external device read-only also
2800 Details : During a commanded failover stop, we set the disk device
2801 read-only while the server shuts down. We now also set any
2802 external journal device read-only at the same time.
2805 Frequency : when upgrading from 1.4 while trying to change parameters
2807 Description: The wrong (new) MDC name was used when setting parameters for
2808 upgraded MDT's. Also allows changing of OSC (and MDC)
2809 parameters if --writeconf is specified at tunefs upgrade time.
2812 Frequency : when setting specific ost indicies
2814 Description: QOS code breaks on skipped indicies
2815 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2816 created with --index need not be sequential.
2818 Severity : enhancement
2820 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2821 Details : The uninit_groups feature works in conjunction with the kernel
2822 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2823 up the pass1 processing of e2fsck. This is a read-only feature
2824 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2825 mount filesystems that have had this feature enabled.
2827 Severity : enhancement
2829 Description: Improve multi-block allocation algorithm to avoid fragmentation
2830 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2831 allocation locality and avoid filesystem fragmentation.
2833 ------------------------------------------------------------------------------
2835 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2837 * Support for kernels:
2838 2.4.21-47.0.1.EL (RHEL 3)
2839 2.6.5-7.283 (SLES 9)
2840 2.6.9-42.0.10.EL (RHEL 4)
2841 2.6.12.6 vanilla (kernel.org)
2842 2.6.16.27-0.9 (SLES 10)
2843 * Recommended e2fsprogs version: 1.39.cfs5
2845 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2849 Frequency : occasional, depends on client load and configuration
2850 Bugzilla : 12181, 12203
2851 Description: data loss for recently-modified files
2853 Details : In some cases it is possible that recently written or created
2854 files may not be written to disk in a timely manner (this should
2855 normally be within 30s unless client IO load is very high).
2856 The problem appears as zero-length files or files that are a
2857 multiple of 1MB in size after a client crash or client eviction
2858 that are missing data at the end of the file.
2860 This problem is more likely to be hit on clients where files are
2861 repeatedly created and unlinked in the same directory, clients
2862 have a large amount of RAM, have many CPUs, the filesystem has
2863 many OSTs, the clients are rebooted frequently, and/or the files
2864 are not accessed by other nodes after being written.
2866 The presence of the problem can be detected by looking at
2867 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2868 smaller than the second (nr_unused) then dirty files will not
2869 be flushed automatically to disk. "sync; sleep 10" should be
2870 run several times on the node before unmounting it to update
2871 Lustre (this is also safe to run on nodes without this problem).
2873 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2874 that can cause this same problem, so customers using that kernel
2875 also need to update the kernel in addition to Lustre. In order
2876 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2878 It is normal that files written just before a client crash (less
2879 than 30s) may not yet have been flushed to disk, even for local
2883 Frequency : frequent on thin XT3 nodes
2885 Description: UUID collision on thin XT3 Linux nodes
2886 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2887 randomly, since we relied on an insufficiently-seeded PRNG.
2892 Description: OSS hangs after "All ost request buffers busy"
2893 Details : A deadlock between quota and journal operations caused OSS
2894 hangs after printing "All ost request buffers busy."
2897 Frequency : always on liblustre builds
2899 Description: Cleanup compiler warnings on liblustre
2902 Frequency : always on liblustre builds on XT3
2904 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2908 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2909 > 32000 subdirectories
2910 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2911 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2912 subdirectory count crosses 32000. This will aid e2fsck to
2913 correctly handle more than 32000 subdirectories.
2916 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2918 Description: System hang on startup
2919 Details : This bug allowed the liblustre (e.g. catamount) client to
2920 return to the app before handling all startup RPCs. This
2921 could leave the node unresponsive to lustre network traffic
2922 and manifested as a server ptllnd timeout.
2924 Severity : enhancement
2926 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2927 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2928 boolean to control whether to print peer debug info when a
2929 client's RPC times out.
2934 Description: make O_SYNC working on 2.6 kernels
2935 Details : 2.6 kernels use different method for mark pages for write,
2936 so need add a code to lustre for O_SYNC work.
2941 Description: Failure to close file and release space on NFS
2942 Details : Put inode details into lock acquired in ll_intent_file_open.
2943 Use mdc_intent_lock in ll_intent_open to properly
2944 detect all kind of errors unhandled by mdc_enqueue.
2949 Description: proc file read during shutdown sometimes raced obd removal,
2951 Details : Add lock to prevent obd access after proc file removal.
2954 Frequency : Only for files larger than 4GB on 32-bit clients.
2956 Description: improperly doing page alignment of locks
2957 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2958 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2963 Description: RPCs being resent when they shouldn't be
2964 Details : Some RPCs that should not be resent are being resent. This
2965 can cause inconsistencies in the RPC state machine. Do not
2966 resend such requests.
2969 Frequency : rare, only with NFS export
2971 Description: Crash on NFS re-export node
2972 Details : under very unusual load conditions an assertion is hit in
2973 ll_intent_file_open()
2976 Frequency : only if OST filesystem is corrupted
2978 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2979 Details : for a short time RPCs with bulk IO are in the replay list,
2980 but replay of bulk IOs is unimplemented. If the OST filesystem
2981 is corrupted due to disk cache incoherency and then replay is
2982 started it is possible to trip an assertion. Avoid putting
2983 committed RPCs into the replay list at all to avoid this issue.
2988 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2989 Details : Large single O_DIRECT read and write calls can fail to allocate
2990 a sufficiently large buffer to process the request. In case of
2991 allocation failure the allocation is retried with a smaller
2992 buffer and broken into smaller requests.
2994 Severity : enhancement
2996 Description: Add -o localflock option to simulate old noflock behaviour.
2997 Details : This will achieve local-only flock/fcntl locks coherentness.
3002 Description: versioning check is incomplete
3003 Details : Checking the version difference of client vs. server, report
3004 error if the gap is too big.
3009 Description: add support PG_writeback bit
3010 Details : add support for PG_writeback bit for Lustre, for more carefull
3011 work with page cache in 2.6 kernel. This also fix some deadlocks
3012 and remove hack for work O_SYNC with 2.6 kernel.
3014 Severity : enhancement
3016 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3017 Details : The uninit_groups feature works in conjunction with the kernel
3018 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3019 up the pass1 processing of e2fsck. This is a read-only feature
3020 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3021 mount filesystems that have had this feature enabled.
3023 Severity : enhancement
3025 Description: Improve multi-block allocation algorithm to avoid fragmentation
3026 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3027 allocation locality and avoid filesystem fragmentation.
3029 ------------------------------------------------------------------------------
3031 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3033 * Support for kernels:
3034 2.6.9-42.0.3.EL (RHEL 4)
3035 2.6.5-7.276 (SLES 9)
3036 2.4.21-47.0.1.EL (RHEL 3)
3037 2.6.12.6 vanilla (kernel.org)
3038 2.6.16.21-0.8 (SLES10)
3039 * Recommended e2fsprogs version: 1.39.cfs2-0
3041 * The backwards-compatible /proc/sys/portals symlink has been removed
3042 in this release. Before upgrading, please ensure that you change
3043 any configuration scripts or /etc/sysctl.conf files that access
3044 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3045 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3046 in advance of the upgrade on any system running Lustre 1.4.6 or
3047 newer, since /proc/sys/lnet was added in that version.
3048 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3052 Frequency : only when quota is used
3054 Description: avoid scanning export list for quota master
3055 Details : Change the algorithms to avoid scanning export list in order
3056 to improve the efficiency.
3059 Frequency : MDS failover only, very rarely
3061 Description: "went back in time" messages on mds failover
3062 Details : The greatest transno may be lost when the current operation
3063 finishes with an error (transno==0) and the client's last_rcvd
3064 record is over-written. Save the greatest transno in the
3065 mds_last_transno for this case.
3068 Frequency : always for specific kernels and striping counts
3070 Description: client may get "Matching packet too big" without ACL support
3071 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3072 when trying to access files in certain configurations. The
3073 clients should in fact be denied when mounting because they do
3074 not understand ACLs.
3077 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3079 Description: many clients connecting with IO in progress causes connect timeouts
3080 Details : Avoid synchronous journal commits to avoid delays caused by many
3081 clients connecting/disconnecting when bulk IO is in progress.
3082 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3083 OST_IO_PORTAL to avoid delays behind potentially many pending
3087 Frequency : occasionally with multiple writers to a single file
3089 Description: shared writes to file may result in wrong size reported by stat()
3090 Details : Allow growing of kms when extent lock is cancelled
3093 Frequency : always with random mmap IO to multi-striped file
3095 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3096 Details : Only if the hole is at the end of OST object so that kms is too
3097 small. Fix is to increase kms accordingly in ll_nopage.
3100 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3102 Description: writes to a missing object would leak memory on the OST
3103 Details : If there is an inconsistency between the MDS and OST filesystems,
3104 such that the MDS references an object that doesn't exist, writes
3105 to that object will leak memory due to incorrect cleanup in the
3106 error handling path, eventually running out of memory on the OST.
3111 Description: Creating too long symlink causes lustre errors
3112 Details : Check symlink and name lengths before sending requests to MDS.
3115 Frequency : only if flock is enabled (not on by default)
3117 Description: posix locks not released on fd closure on 2.6.9+
3118 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3119 kernels, this caused such locks not to be released on fd close and
3120 then assertions on fs unmount about still used locks.
3123 Frequency : MDS failover only, very rarely
3125 Description: clients may get ASSERTION(granted_lock != NULL)
3126 Details : When request was taking a long time, and a client was resending
3127 a getattr by name lock request. The were multiple lock requests
3128 with the same client lock handle and
3129 mds_getattr_name->fixup_handle_for_resent_request found one of the
3130 lock handles but later failed with ASSERTION(granted_lock != NULL).
3135 Description: handle->h_buffer_credits > 0, assertion failure
3136 Details : h_buffer_credits is zero after truncate, causing assertion
3137 failure. This patch extends the transaction or creates a new
3141 Frequency : NFS re-export or patchless client
3142 Bugzilla : 11179, 10796
3143 Description: Crash on NFS re-export node (__d_move)
3144 Details : We do not want to hash the dentry if we don't have a lock.
3145 But if this dentry is later used in d_move, we'd hit uninitialised
3146 list head d_hash, so we just do this to init d_hash field but
3147 leave dentry unhashed.
3150 Frequency : NFS re-export or patchless client
3152 Description: NFS exports has problem with symbolic link
3153 Details : lustre client didn't properly install dentry when re-exported
3154 to NFS or running patchless client.
3157 Frequency : NFS re-export or patchless client
3159 Description: Various nfs/patchless fixes.
3160 Details : fixes reuse disconected alias for lookup process - this fixes
3161 warning "find_exported_dentry: npd != pd",
3162 fix permission error with open files at nfs.
3163 fix apply umask when do revalidate.
3166 Frequency : occasional
3168 Description: Crash on NFS re-export node
3169 Details : calling clear_page() on the wrong pointer triggered oops in
3170 generic_mapping_read().
3173 Frequency : rarely, using O_DIRECT IO
3175 Description: unaligned directio crashes client with LASSERT
3176 Details : check for unaligned buffers before trying any requests.
3179 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3181 Description: stale data returned from RAID cache
3182 Details : If only a small amount of IO is done to the RAID device before
3183 reading it again it is possible to get stale data from the RAID
3184 cache instead of reading it from disk.
3187 Frequency : always for sles10 kernel
3189 Description: sles10 support
3190 Details : ll_follow_link: compile fixes and using of nd_set_link
3191 under newer kernels.
3194 Frequency : depends on arch, kernel and compiler version, always on sles10
3197 Description: recursive or deep enough symlinks cause stack overflow
3198 Details : getting rid of large stack-allocated variable in
3202 Frequency : depends on hardware
3204 Description: lustre write performance loss in the SLES10 kernel
3205 Details : the performance loss is caused by using of write barriers in the
3206 ext3 code. The SLES10 kernel turns barrier support on by
3207 default. The fix is to undo that change for ldiskfs.
3209 ------------------------------------------------------------------------------
3211 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3213 * Support for kernels:
3214 2.6.9-42.0.3EL (RHEL 4)
3215 2.6.5-7.276 (SLES 9)
3216 2.4.21-47.0.1.EL (RHEL 3)
3217 2.6.12.6 vanilla (kernel.org)
3221 Frequency : quota enabled and large files being deleted
3223 Description: releasing more than 4GB of quota at once hangs OST
3224 Details : If a user deletes more than 4GB of files on a single OST it
3225 will cause the OST to spin in an infinite loop. Release
3226 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3231 Description: statfs data retrieved from /proc may be stale or zero
3232 Details : When reading per-device statfs data from /proc, in the
3233 {kbytes,files}_{total,free,avail} files, it may appear
3234 as zero or be out of date.
3237 Frequency : systems with MD RAID1 external journal devices
3239 Description: lconf's call to blkid is confused by RAID1 journal devices
3240 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3241 of returning all block devices that match the journal UUID.
3244 Frequency : always, for aggregate stripe size over 4GB
3246 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3247 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3248 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3251 Frequency : always if "lfs find" used on a local file/directory
3253 Description: "lfs find" segfaults if used on a local file/directory
3254 Details : The case where a directory component was not specified wasn't
3255 handled correctly. Handle this properly.
3258 Frequency : always on ppc64
3260 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3261 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3262 perform endianness conversion.
3265 Frequency : rarely (truncate to non-zero file size after write under load)
3266 Bugzilla : 10730, 10687
3267 Description: Files padded with zeros to next 4K multiple
3268 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3269 it is possible that files that are truncated to a non-zero size
3270 immediately after being written are filled with zero bytes beyond
3271 the truncated size. No file data is lost.
3273 Severity : enhancement
3275 Description: Allow recovery/failover for liblustre clients.
3276 Details : liblustre clients were unaware of failover configurations until
3279 Severity : enhancement
3281 Description: user file locks should fail when not mounting with flock option
3282 Details : Set up an error-returning stub in ll_file_operations.lock field
3283 to prevent incorrect behaviour when client is mounted without
3284 flock option. Also, set up properly f_op->flock field for
3288 Frequency : always on ia64
3290 Description: "lfs df" loops on printing out MDS statfs information
3291 Details : The obd_ioctl_data was not initialized and in some systems
3292 this caused a failure during the ioctl that did not return
3293 an error. Initialize the struct and return an error on failure.
3296 Frequency : SLES 9 only
3298 Description: Error of copying files with lustre special EAs as root
3299 Details : Client side always return success for setxattr call for lustre
3300 special xattr (currently only "trusted.lov").
3303 Frequency : rarely on clusters with both ia64+i386 clients
3305 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3306 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3307 clients are doing concurrent writes to the same file it is
3308 possible that the ia64 clients may LASSERT because the OST
3309 extent locks are not PAGE_SIZE aligned. Ensure that grown
3310 locks are always aligned on the request boundary.
3313 Frequency : specific use, occasional
3315 Description: Overwriting in use executable truncates on-disk binary image
3316 Details : If one node attempts to overwrite an executable in use by
3317 another node, we now correctly return ETXTBSY instead of
3318 truncating the file.
3320 Severity : enhancement
3322 Description: Async OSC create to avoid the blocking unnecessarily.
3323 Details : If a OST has no remain object, system will block on the creating
3324 when need to create a new object on this OST. Now, ways use
3325 pre-created objects when available, instead of blocking on an
3326 empty osc while others are not empty. If we must block, we block
3327 for the shortest possible period of time.
3332 Description: chmod on Lustre root is propagated to other clients
3333 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3334 invalid by the follow_mount time.
3337 Frequency : liblustre clients only
3339 Description: Race in 'instant cancel' lock handling could lead to such locks
3340 never to be granted in case of SMP MDS
3341 Details : Do not destroy not yet granted but cbpending locks in
3345 Frequency : replay/resend of open
3347 Description: non null lock assetion failure in mds_intent_policy
3348 Details : Trying to replay/resend lockless open requests resulted in
3349 mds_open() returning 0 with no lock. Now it sets a flag if
3350 a lock is going to be returned.
3352 Severity : enhancement
3354 Description: Checksum enhancements
3355 Details : New checksum enhancements allow for resending RPCs that failed
3358 Severity : enhancement
3360 Description: Tunables on number of dirty pages in cacche
3361 Details : Allow to set limit on number of dirty pages cached.
3366 Description: client crash on unmount - lock still has references
3367 Details : In some error handling cases it was possible to leak a lock
3368 reference on a client while accessing a file. This was not
3369 harmful to the client during operation, but would cause the
3370 client to crash when the filesystem is unmounted.
3373 Frequency : specific case, rare
3375 Description: ETXTBSY on mds though file not in use by client
3376 Details : ETXTBSY is no longer incorrectly returned when attempting to
3377 chmod or chown a directory that the user previously tried to
3378 execute or a currently-executing binary.
3381 Frequency : extremely rare except on liblustre-based clients
3383 Description: Lustre space not freed when files are deleted
3384 Details : Clean up open-unlinked files after client eviction. Previously
3385 the unlink was skipped and the files remained as orphans.
3390 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3391 Details : In some cases it was possible to send a blocking callback to
3392 a client doing a glimpse, even though that client didn't get
3393 a lock granted. When the glimpse lock is cancelled on the OST
3394 the freed lock is left on the waiting list and corrupted the list.
3397 Frequency : all core dumps
3399 Description: Broke core dumps to lustre
3400 Details : Negative dentry may be unhashed if parent does not have UPDATE
3401 lock, but some callers, e.g. do_coredump, expect dentry to be
3402 hashed after successful create, hash it in ll_create_it.
3404 ------------------------------------------------------------------------------
3406 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3408 * Support for kernels:
3409 2.6.9-42.0.2.EL (RHEL 4)
3410 2.6.5-7.276 (SLES 9)
3411 2.4.21-40.EL (RHEL 3)
3412 2.6.12.6 vanilla (kernel.org)
3416 Frequency : always on RHEL 3
3418 Description: Number of open files grows over time
3419 Details : The number of open files grows over time, whether or not
3420 Lustre is started. This was due to a filp leak introduced
3421 by one of our kernel patches.
3423 ------------------------------------------------------------------------------
3425 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3427 * Support for kernels:
3428 2.6.9-42.EL (RHEL 4)
3429 2.6.5-7.276 (SLES 9)
3430 2.4.21-40.EL (RHEL 3)
3431 2.6.12.6 vanilla (kernel.org)
3436 Bugzilla : 5719, 9635, 9792, 9684
3437 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3438 Details : If a server is under heavy load and cannot reply to new
3439 connection requests before the client resends the (re)connect,
3440 the connection handling code can behave badly if two service
3441 threads are concurrently handing separate (re)connections from
3442 the same client. Add better locking to the connection handling
3443 code, and ensure that only a single connection will be processed
3444 for a given client UUID, even if the lock is dropped.
3446 Severity : enhancement
3448 Description: add TCP zero-copy support to kernel
3449 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3450 sends if the hardware supports scatter-gather and checksumming.
3451 This allows socklnd to do client-write and server-read more
3452 efficiently and reduce CPU utilization from skbuf copying.
3455 Frequency : only if NFS exporting from client
3457 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3458 Details : If setting attributes on a file created under NFS that had
3459 never been opened it would be possible to oops the client
3460 if the file had no objects.
3463 Frequency : always for liblustre
3465 Description: liblustre client does MDS+OSTs setattr RPC for each write
3466 Details : When doing a write from a liblustre client, the client
3467 incorrectly issued an RPC to the MDS and each OST the file was
3468 striped over in order to update the timestamps. When writing
3469 with small chunks and many clients this could overwhelm the MDS
3470 with RPCs. In all cases it would slow down the write because
3471 these RPCs are unnecessary.
3473 Severity : enhancement
3475 Description: allow number of MDS service threads to be changed at module load
3476 Details : It is now possible to change the number of MDS service threads
3477 running. Adding "options mds mds_num_threads={N}" to the MDS's
3478 /etc/modprobe.conf will set the number of threads for the next
3479 time Lustre is restarted (assuming the "mds" module is also
3480 reloaded at that time). The default number of threads will
3481 stay the same, 32 for most systems.
3486 Description: OST crash if filesystem is unformatted or corrupt
3487 Details : If an OST is started on a device that has never been formatted
3488 or if the filesystem is corrupt and cannot even mount then the
3489 error handling cleanup routines would dereference a NULL pointer.
3494 Description: NULL pointer deref in llap_from_page.
3495 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3496 incorrect) mapping if the page was truncated/reclaimed while it was
3497 searched for. Check for this condition and skip such pages when
3498 doing readahead. Introduce extra check to llap_from_page() to
3499 verify page->mapping->host is non-NULL (so page is not anonymous).
3502 Frequency : Sometimes when using sys_sendfile
3504 Description: "page not covered by a lock" warnings from ll_readpage
3505 Details : sendfile called ll_readpage without right page locks present.
3506 Now we introduced ll_file_sendfile that does necessary locking
3507 around call to generic_file_sendfile() much like we do in
3511 Frequency : with certain MDS communication failures at client mount time
3513 Description: NULL pointer deref after failed client mount
3514 Details : a client connection request may delayed by the network layer
3515 and not be sent until after the PTLRPC layer has timed out the
3516 request. If the client fails the mount immediately it will try
3517 to clean up before the network times out the request. Add a
3518 reference from the request import to the obd device and delay
3519 the cleanup until the network drops the request.
3522 Frequency : occasionally during client (re)connect
3524 Description: assertion failure during client (re)connect
3525 Details : processing a client connection request may be delayed by the
3526 client or server longer than the client connect timeout. This
3527 causes the client to resend the connection request. If the
3528 original connection request is replied in this interval, the
3529 client may trip an assertion failure in ptlrpc_connect_interpret()
3530 which thought it would be the only running connect process.
3533 Frequency : only with obd_echo servers and clients that are rebooted
3535 Description: kernel BUG accessing uninitialized data structure
3536 Details : When running an obd_echo server it did not start the ping_evictor
3537 thread, and when a client was evicted an uninitialized data
3538 structure was accessed. Start the ping_evictor in the RPC
3539 service startup instead of the OBD startup.
3541 Severity : enhancement
3542 Bugzilla : 10193 (patchless)
3543 Description: Remove dependency on various unexported kernel interfaces.
3544 Details : No longer need reparent_to_init, exit_mm, exit_files,
3545 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3548 Frequency : rare (only users of deprecated and unsupported LDAP config)
3550 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3551 Details : LDAP apparently contains 'lustreName' attributes instead of
3552 'name'. A simple remapping of the name is sufficient.
3555 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3557 Description: waiting_locks_callback trips kernel BUG if client is evicted
3558 Details : Running with the dump_on_timeout debug flag turned on makes
3559 it possible that the waiting_locks_callback() can try to dump
3560 the Lustre kernel debug logs from an interrupt handler. Defer
3561 this log dumping to the expired_lock_main() thread.
3563 Severity : enhancement
3565 Description: Support NFS exporting on 2.6 kernels.
3566 Details : Implement non-rawops metadata methods for NFS server to use without
3567 changing NFS server code.
3570 Frequency : very rare (synthetic metadata workload only)
3572 Description: two racing renames might cause an MDS thread to deadlock
3573 Details : Running the "racer" program may cause one MDS thread to rename
3574 a file from being the source of a rename to being the target of
3575 a rename at exactly the same time that another thread is doing
3576 so, and the second thread has already enqueued these locks after
3577 doing a lookup of the target and is trying to relock them in
3578 order. Ensure that we don't try to re-lock the same resource.
3581 Frequency : only very large systems with liblustre clients
3583 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3584 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3585 all OSTs in parallel. This allows the eviction of stale liblustre
3586 clients to proceed much faster than if they were done in series,
3587 and also offers similar improvements for other set_info RPCs.
3592 Description: excessive CPU usage during initial read phase on client
3593 Details : During the initial read phase on a client, it would agressively
3594 retry readahead on the file, consuming too much CPU and impacting
3595 performance (since 1.4.5.8). Improve the readahead algorithm
3596 to avoid this, and also improve some other common cases (read
3597 of small files in particular, where "small" is files smaller than
3598 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3603 Description: MDS crash when receiving packet with unknown intent.
3604 Details : Do not LBUG in unknown intent case, just return -EFAULT
3606 Severity : enhancement
3607 Bugzilla : 9293, 9385
3608 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3609 Details : Do not serialize getattr (non-intent version) and statfs.
3612 Frequency : occasional, when OST network is overloaded/intermittent
3614 Description: client evicted by OST after bulk IO timeout
3615 Details : If a client sends a bulk IO request (read or write) the OST
3616 may evict the client if it is unresposive to its data GET/PUT
3617 request. This is incorrect if the network is overloaded (takes
3618 too long to transfer the RPC data) or dropped the OST GET/PUT
3619 request. There is no need to evict the client at all, since
3620 the pinger and/or lock callbacks will handle this, and the
3621 client can restart the bulk request.
3624 Frequency : Always when mmapping file with no objects
3626 Description: client crashes when mmapping file with no objects
3627 Details : Check that we actually have objects in a file before doing any
3628 operations on objects in ll_vm_open, ll_vm_close and
3634 Description: Request leak when working with deleted CWD
3635 Details : Introduce advanced request refcount tracking for requests
3636 referenced from lustre intent.
3638 Severity : Enhancement
3640 Description: Cache open file handles on client.
3641 Details : MDS now will return special lock along with openhandle, if
3642 requested and client is allowed to hold openhandle, even if unused,
3643 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3644 closing files for every read/write openration.
3646 Severity : Enhancement
3648 Description: Cache open negative dentries on client when possible.
3649 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3650 negative dentries on lock revocation.
3655 Description: Remounting a client read-only wasn't possible with a zconf mount
3656 Details : It wasn't possible to remount a client read-only with llmount.
3658 Severity : enhancement
3659 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3660 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3661 MPICH is included to add Lustre-specific ADIO interfaces.
3662 This is based closely on the UFS ADIO layer and only differs
3663 in file creation, in order to allow the OST striping to be set.
3664 This is user-contributed code and not supported by CFS.
3669 Description: extended inode attributes (immutable, append-only) work improperly
3670 when 2.4 and 2.6 kernels are used on client/server or vice versa
3671 Details : Introduce kernel-independent values for these flags.
3673 Severity : enhancement
3676 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3677 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3678 granularity. Now, Fractional MB granularity is be supported,
3679 this is very useful for low memory system.
3681 Severity : enhancement
3683 Description: Getattr by fid
3684 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3685 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3686 have locks and avoids getattr rpc requests for it on every stat.
3689 Frequency : Always, for filesystems larger than 2TB
3691 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3692 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3693 filesystems larger than 2TB to be mounted with mballoc enabled.
3696 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3698 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3699 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3700 filesystems larger than 2TB is mounted then the kernel will
3701 silently corrupt the start of the filesystem. CONFIG_LBD is
3702 enabled for all CFS-supported kernels, but the possibility of
3703 this happening with a modified kernel config exists.
3705 Severity : enhancement
3707 Description: add client O_DIRECT support for 2.6 kernels
3708 Details : It is now possible to do O_DIRECT reads and writes to files
3709 in the Lustre client mountpoint on 2.6 kernel clients.
3711 Severity : enhancement
3713 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3714 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3715 parallel, not waiting for response from every OST before sending
3716 a rpc to the next OST.
3721 Description: setattr vs write race when updating file timestamps
3722 Details : Client processes that update a file timestamp into the past
3723 right after writing to the file (e.g. tar) it is possible that
3724 the updated file modification time can be reset to the current
3725 time due to a race between processing the setattr and write RPC.
3727 Severity : enhancement
3729 Description: Bring 'lfs find' closer in line with regular Linux find.
3730 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3731 -print0 options and obtains all the needed info through the lustre
3734 Severity : enhancement
3736 Description: support up to 1024 configured devices on one node
3737 Details : change obd_dev array from statically allocated to dynamically
3738 allocated structs as they are first used to reduce memory usage
3743 Description: Flush dirty partially truncated pages during truncate
3744 Details : Immediatelly flush partially truncated pages in filter_setattr,
3745 this way we completely avoid having any pages in page cache on OST
3746 and can retire ugly workarounds during writes to flush such pages.
3751 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3752 Details : protect inode from truncation within vfs_unlink() context
3753 just take a reference before calling vfs_unlink() and release it
3754 when parent's i_sem is free.
3759 Description: last_id value checked outside lock on OST caused LASSERT failure
3760 Details : If there were multiple MDS->OST object precreate requests in
3761 flight, it was possible that the OST's last object id was checked
3762 outside a lock and incorrectly tripped an assertion. Move checks
3763 inside locks, and discard old precreate requests.
3766 Frequency : always, if extents are used on OSTs
3768 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3769 Details : OSTs using the extents format would not zero the high 16 bits of
3770 the index physical block number. This is not a problem for any
3771 OST filesystems smaller than 16TB, and no kernels support ext3
3772 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3773 new/modified files) and can be fixed for existing filesystems
3774 with e2fsprogs-1.39-cfs1.
3779 Description: import connection selection may be incorrect if timer wraps
3780 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3781 import connections to be ignored if the 32-bit jiffies counter
3782 wraps. Use a 64-bit jiffies counter.
3785 Frequency : very large clusters immediately after boot
3787 Description: LNET request buffers exhausted under heavy short-term load
3788 Details : If a large number of client requests are generated on a service
3789 that has previously never seen so many requests it is possible
3790 that the request buffer growth cannot keep up with the spike in
3791 demand. Instead of dropping incoming requests, they are held in
3792 the LND until the RPC service can accept more requests.
3795 Frequency : Sometimes during replay
3797 Description: Assertion failure in ll_local_open after replay.
3798 Details : If replay happened on an open request reply before we were able
3799 to set replay handler, reply will become not swabbed tripping the
3800 assertion in ll_local_open. Now we set the handler right after
3801 recognising of open request
3804 Frequency : very rare
3806 Description: kernel reports "badness in vsnprintf"
3807 Details : Reading from the "recovery_status" /proc file in small chunks
3808 may cause a negative length in lprocfs_obd_rd_recovery_status()
3809 call to vsnprintf() (which is otherwise harmless). Exit early
3810 if there is no more space in the output buffer.
3812 Severity : enhancement
3814 Description: clear OBD RPC statistics by writing to them
3815 Details : It is now possible to clear the OBD RPC statistics by writing
3816 to the "stats" file.
3821 Description: Client mtime is not the same on different clients after utimes
3822 Details : In some cases, the client was using the utimes() syscall on
3823 a file cached on another node. The clients now validate the
3824 ctime from the MDS + OSTs to determine which one is right.
3829 Description: Inability to activate failout mode
3830 Details : lconf script incorrectly assumed that in pythong string's numeric
3831 value is used in comparisons.
3834 Frequency : always with multiple stripes per file
3836 Description: Inefficient object allocation for mutli-stripe files
3837 Details : When selecting which OSTs to stripe files over, for files with
3838 a stripe count that divides evenly into the number of OSTs,
3839 the MDS is always picking the same starting OST for each file.
3840 Return the OST selection heuristic to the original design.
3845 Description: mount failures may take full timeout to return an error
3846 Details : Under some heavy load conditions it is possible that a
3847 failed mount can wait for the full obd_timeout interval,
3848 possibly several minutes, before reporting an error.
3849 Instead return an error as soon as the status is known.
3851 Frequency : quota enabled and large files being deleted
3853 Description: releasing more than 4GB of quota at once hangs OST
3854 Details : If a user deletes more than 4GB of files on a single OST it
3855 will cause the OST to spin in an infinite loop. Release
3856 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3861 Description: statfs data retrieved from /proc may be stale or zero
3862 Details : When reading per-device statfs data from /proc, in the
3863 {kbytes,files}_{total,free,avail} files, it may appear
3864 as zero or be out of date.
3867 Frequency : systems with MD RAID1 external journal devices
3869 Description: lconf's call to blkid is confused by RAID1 journal devices
3870 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3871 of returning all block devices that match the journal UUID.
3874 Frequency : always, for aggregate stripe size over 4GB
3876 Description: assertion fails when trying to use 4GB stripe size
3877 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3878 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3881 Frequency : always on ppc64
3883 Description: the first write on an ext3 filesystem with mballoc got stuck
3884 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3885 perform endianness conversion.
3887 ------------------------------------------------------------------------------
3889 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3891 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3892 INTEROPERATE with older versions automatically. Please read the
3893 user documentation before upgrading any part of a live system.
3894 * WARNING: Lustre networking configuration changes are required with
3895 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3898 * Support for kernels:
3899 2.6.9-22.0.2.EL (RHEL 4)
3900 2.6.5-7.244 (SLES 9)
3901 2.6.12.6 vanilla (kernel.org)
3904 Severity : enhancement
3905 Bugzilla : 7981/8208
3906 Description: Introduced Lustre Networking (LNET)
3907 Details : LNET is new networking infrastructure for Lustre, it includes
3908 a reorganized network configuration mode (see the user
3909 documentation for full details) as well as support for routing
3910 between different network fabrics. Lustre Networking Devices
3911 (LNDs) for the supported network fabrics have also been
3912 created for this new infrastructure.
3914 Severity : enhancement
3915 Description: Introduced Access control lists
3916 Details : clients can set ACLs on files and directories in order to have
3917 more fine-grained permissions than the standard Unix UGO+RWX.
3918 The MDS must be started with the "-o acl" mount option.
3920 Severity : enhancement
3921 Description: Introduced filesystem quotas
3922 Details : Administrators may now establish per-user quotas on the
3925 Severity : enhancement
3927 Description: Configuration change for the XT3
3928 The PTLLND is now used to run Lustre over Portals on the XT3
3929 The configure option(s) --with-cray-portals are no longer used.
3930 Rather --with-portals=<path-to-portals-includes> is used to
3931 enable building on the XT3. In addition to enable XT3 specific
3932 features the option --enable-cray-xt3 must be used.
3937 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3938 Details : A race between allocating a new llcd and re-getting the llcd_lock
3939 allowed another thread to grab newly-allocated llcd.
3941 Severity : enhancement
3943 Description: 2.6 OST async journal commit and locking fix to improve performance
3944 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3945 2.6 kernels are now async as they already were in 2.4 kernels so
3946 that they can commit concurrently with the network bulk transfer.
3947 For block-allocated files the filter allocation semaphore is held
3948 to avoid filesystem fragmentation during allocation. BKL lock
3949 removed for 2.6 xattr operations where it is no longer needed.
3954 Description: lconf incorrectly determined whether two IP networks could talk
3955 Details : In some more complicated routing and multiple-network
3956 configurations, lconf will avoid trying to make a network
3957 connection to a disjoint part of the IP space. It was doing the
3958 math incorrectly for one set of cases.
3963 Description: Fix for potential infinite loop processing records in an llog.
3964 Details : If an llog record is corrupted/zeroed, it is possible to loop
3965 forever in llog_process(). Validate the llog record length
3966 and skip the remainder of the block on error.
3969 Frequency : occasional (liblustre only)
3971 Description: liblustre could not open files whose last component is a symlink
3972 Details : sysio_path_walk() would incorrectly pass the open intent to
3973 intermediate path components.
3976 Frequency : rare (liblustre only with non-standard tuning)
3977 Bugzilla : 7201 (7350)
3978 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3979 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3980 after it has been released, only do this for referenced locks
3982 Severity : enhancement
3984 Description: specifying an (invalid) directory default stripe_size of -1
3985 would reset the directory default striping
3986 Details : stripe_size -1 was used internally to signal directory stripe
3987 removal, now use "all default" to signal dir stripe removal
3988 as a directory striping of "all default" is not useful
3991 Frequency : common for large clusters running liblustre clients
3993 Description: doing an ls when liblustre clients are running is slow
3994 Details : sending a glimpse AST to a liblustre client waits for every AST
3995 to time out, as liblustre clients will not respond. Since they
3996 cannot cache data we refresh the OST lock LVB from disk instead.
3998 Severity : enhancement
4000 Description: doing an ls at the same time as file IO can be slow
4001 Details : enqueue and other "small" requests can be blocked behind many
4002 large IO requests. Create a new OST IO portal for non-IO
4003 requests so they can be processed faster.
4006 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4008 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4009 created file with mode 000
4010 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4011 allow an ftruncate() as a non-root user to a file with mode 000.
4012 Lustre now respects this flag to disable mode checks when
4013 truncating a file owned by the user
4016 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4019 Description: Revoking locks from clients that went dead or catatonic might take
4021 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4022 cancellation of such locks instant on servers without waiting for
4023 any reply from clients. Clients drops these locks when cancel
4024 notification from server is received without replying.
4027 Frequency : liblustre-only, when liblustre client dies or becomes busy
4029 Description: Doing ls on Linux clients can take a long time with active
4031 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4032 granting such locks to it in the first place if possible. Locks
4033 are taken by proxy on the OST during the read or write and
4034 dropped immediately afterward. Add connect flags handling, do
4035 not grant locks to liblustre clients for glimpse ASTs.
4037 Severity : enhancement
4039 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4040 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4041 sequential reads and behaves badly with applications doing random
4042 reads. Improve it by reading ahead at least read region, and
4043 avoiding excessive large RPC for small reads.
4045 Severity : enhancement
4047 Description: Creating more than 1000 files for a single job may cause a load
4048 imbalance on the OSTs if there are also a large number of OSTs.
4049 Details : qos_prep_create() uses an OST index reseed value that is an
4050 even multiple of the number of available OSTs so that if the
4051 reseed happens in the middle of the object allocation it will
4052 still utilize the OSTs as uniformly as possible.
4057 Description: OST or MDS may oops in ping_evictor_main()
4058 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4059 but doesn't restart at beginning of obd_exports_timed list
4062 Severity : enhancement
4064 Description: improve by-nid export eviction on the MDS and OST
4065 Details : allow multiple exports with the same NID to be evicted at one
4066 time without re-searching the exports list.
4069 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4071 Description: MDS may oops in groups_free()
4072 Details : in rare race conditions a newly allocated group_info struct is
4073 freed again, and this can be NULL. The 2.4 compatibility code
4074 for groups_free() checked for a NULL pointer, but 2.6 did not.
4077 Frequency : common for liblustre clients doing little filesystem IO
4078 Bugzilla : 9352, 7313
4079 Description: server may evict liblustre clients accessing contended locks
4080 Details : if a client is granted a lock or receives a completion AST
4081 with a blocking AST already set it would not reply to the AST
4082 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4085 Frequency : lfs setstripe, only systems with more than 160 OSTs
4087 Description: unable to set striping with a starting offset beyond OST 160
4088 Details : llapi_create_file() incorrectly limited the starting stripe
4089 index to the maximum single-file stripe count.
4092 Frequency : LDAP users only
4094 Description: lconf did not handle in-kernel recovery with LDAP properly
4095 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4097 Severity : enhancement
4099 Description: bind OST threads to NUMA nodes to improve performance
4100 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4101 node and do their allocations there to localize memory access
4103 Severity : enhancement
4105 Description: llmount can determine client NID directly from Myrinet (GM)
4106 Details : the client NID code from gmnalnid was moved directly into
4107 llmount, removing the need to use this or specifying the
4108 client NID explicitly when mounting GM clients with zeroconf
4111 Frequency : if client is started with down MDS
4113 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4114 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4115 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4116 ptlrpc_import_delayed_req() because we didn't check if the
4117 request was interrupted, and we also didn't break out of the
4118 event loop if there was no timeout
4123 Description: data loss during non-page-aligned writes to a single file from
4124 both multiple nodes and multiple threads on one node at same time
4125 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4126 inconsistency led to false short-reads, that were cached and later
4127 used by ->prepare_write() to fill in partially written page,
4128 leading to data loss.
4131 Frequency : always, if lconf --abort_recovery used
4133 Description: lconf --abort_recovery fails with 'Operation not supported'
4134 Details : lconf was attempting to abort recovery on the MDT device and not
4137 Severity : enhancement
4139 Description: remove cleanup logs
4140 Details : replace lconf-generated cleanup logs with lustre internal
4141 cleanup routines. Eliminates the need for client-cleanup and
4144 Severity : enhancement
4146 Description: add support for EAs (user and system) on lustre filesystems
4147 Details : it is now possible to store extended attributes in the Lustre
4148 client filesystem, and with the user_xattr mount option it
4149 is possible to allow users to store EAs on their files also
4151 Severity : enhancement
4153 Description: Add possibility (config option) to show minimal available OST free
4155 Details : When compiled with --enable-mindf configure option, statfs(2)
4156 (and so, df) will return least minimal free space available from
4157 all OSTs as amount of free space on FS, instead of summary of
4158 free spaces of all OSTs.
4160 Severity : enhancement
4162 Description: do not expand extent locks acquired on OST-side
4163 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4164 by server: they are not cached anyway.
4167 Frequency : when mmap is used/binaries executed from Lustre
4169 Description: Unmmap pages before throwing them away from read cache.
4170 Details : llap_shrink cache now attempts to unmap pages before discarding
4171 them (if unmapping failed - do not discard). SLES9 kernel has
4172 extra checks that trigger if this unmapping is not done first.
4177 Description: lconf didn't resolve symlinks before checking to see whether a
4178 given mountpoint was already in use
4181 Frequency : when migrating failover services
4182 Bugzilla : 6395, 9514
4183 Description: When migrating a subset of services from a node (e.g. failback
4184 from a failover service node) the remaining services would
4185 time out and evict clients.
4186 Details : lconf --force (implied by --failover) sets the global obd_timeout
4187 to 5 seconds in order to quickly disconnect, but this caused
4188 other RPCs to time out too quickly. Do not change the global
4189 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4191 Severity : enhancement
4192 Frequency : if MDS is started with down OST
4193 Bugzilla : 9439,5706
4194 Description: Allow startup/shutdown of an MDS without depending on the
4195 availability of the OSTs.
4196 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4197 Add appropriate locking and lov-osc refcounts for safe
4198 cleaning. Add osc abort_inflight calls in case the
4199 synchronize never started.
4202 Frequency : occasional (Cray XT3 only)
4204 Description: root not authorized to access files in CRAY_PORTALS environment
4205 Details : The client process capabilities were not honoured on the MDS in
4206 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4207 been accessed by an authorized user then root was able to access
4208 the file on the local client also. The root user capabilities
4209 are now allowed on the MDS, as this environment has secure UID.
4212 Frequency : occasional
4214 Description: ldiskfs "too long searching" message happens too often
4215 Details : A debugging message (otherwise harmless) prints too often on
4216 the OST console. This has been reduced to only happen when
4217 there are fragmentation problems on the filesystem.
4222 Description: Division by zero in statfs when all OSCs are inactive
4223 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4224 lov_statfs divides by value returned by lov_get_stripecnt().
4228 Bugzilla : 9489, 3273
4229 Description: First write from each client to each OST was only 4kB in size,
4230 to initialize client writeback cache, which caused sub-optimal
4231 RPCs and poor layout on disk for the first writen file.
4232 Details : Clients now request an initial cache grant at (re)connect time
4233 and so that they can start streaming writes to the cache right
4234 away and always do full-sized RPCs if there is enough data.
4235 If the OST is rebooted the client also re-establishes its grant
4236 so that client cached writes will be honoured under the grant.
4241 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4242 Details : Now I/O RPCs go to different portal number and (presumably) fast
4243 lock requests (and glimses) and other RPCs get their own service
4244 threads pool that should be able to service those RPCs
4247 Severity : enhancement
4249 Description: Ability to exchange lustre version between client and servers and
4250 issue warnings at client side if client is too old. Also for
4251 liblustre clients there is ability to refuse connection of too old
4253 Details : New 'version' field is added to connect data structure that is
4254 filled with version info. That info is later checked by server and
4258 Frequency : rare, liblustre only.
4259 Bugzilla : 9296, 9581
4260 Description: Two simultaneous writes from liblustre at offset within same page
4261 might proceed at the same time overwriting eachother with stale
4263 Details : I/O lock withing llu_file_prwv was released too early, before data
4264 actually was hitting the wire. Extended lock-holding time until
4265 server acknowledges receiving data.
4268 Frequency : extremely rare. Never observed in practice.
4270 Description: avoid generating lustre_handle cookie of 0.
4271 Details : class_handle_hash() generates handle cookies by incrementing
4272 global counter, and can hit 0 occasionaly (this is unlikely, but
4273 not impossible, because initial value of cookie counter is
4274 selected randonly). Value of 0 is used as a sentinel meaning
4275 "unassigned handle" --- avoid it. Also coalesce two critical
4276 sections in this function into one.
4278 Severity : enhancement
4280 Description: allow liblustre clients to delegate truncate locking to OST
4281 Details : To avoid overhead of locking, liblustre client instructs OST to
4282 take extent lock in ost_punch() on client's behalf. New connection
4283 flag is added to handle backward compatibility.
4285 Severity : enhancement
4286 Bugzilla : 4928, 7341, 9758
4287 Description: allow number of OST service threads to be specified
4288 Details : a module parameter allows the number of OST service threads
4289 to be specified via "options ost ost_num_threads={N}" in the
4290 OSS's /etc/modules.conf or /etc/modprobe.conf.
4294 Bugzilla : 6146, 9635, 9895
4295 Description: servers crash with bad pointer in target_handle_connect()
4296 Details : In rare cases when a client is reconnecting it was possible that
4297 the connection request was the last reference for that export.
4298 We would temporarily drop the export reference and get a new
4299 one, but this may have been the last reference and the export
4300 was just destroyed. Get new reference before dropping old one.
4302 Severity : enhancement
4303 Frequency : if client is started with failover MDS
4305 Description: Allow multiple MDS hostnames in the mount command
4306 Details : Try to read the configuration from all specified MDS
4307 hostnames during a client mount in case the "primary"
4310 Severity : enhancement
4312 Description: Stop sending data to evicted clients as soon as possible.
4313 Details : Check if the client we are about to send or are sending data to
4314 was evicted already. (Check is done every second of waiting,
4315 for which l_wait_event interface was extended to allow checking
4316 of exit condition at specified intervals).
4319 Frequency : rare, normally only when NFS exporting is done from client
4321 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4323 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4326 Frequency : always, if srand() is called before liblustre initialization
4328 Description: Liblustre uses system PRNG disturbing its usage by user application
4329 Details : Introduce internal to lustre fast and high-quality PRNG for
4330 lustre usage and make liblustre and some other places in generic
4331 lustre code to use it.
4333 Severity : enhancement
4334 Bugzilla : 9477, 9557, 9870
4335 Description: Verify that the MDS configuration logs are updated when xml is
4336 Details : Check if the .xml configuration logs are newer than the config
4337 logs stored on the MDS and report an error if this is the case.
4338 Request --write-conf, or allow starting with --old_conf.
4340 Severity : enhancement
4342 Description: Handle symlinks in the path when checking if Lustre is mounted.
4343 Details : Resolve intermediate symlinks when checking if a client has
4344 mounted a filesystem to avoid duplicate client mounts.
4349 Description: lconf can hit an error exception but still return success.
4350 Details : The lconf command catches the Command error exception at the top
4351 level script context and will exit with the associated exit
4352 status, but doesn't ensure that this exit status is non-zero.
4357 Description: failure of ptlrpc thread startup can cause oops
4358 Details : Starting a ptlrpc service thread can fail if there are a large
4359 number of threads or the server memory is very fragmented.
4360 Handle this without oopsing.
4363 Frequency : always, only if liblustre and non-default acceptor port was used
4365 Description: liblustre cannot connect to servers with non-default acceptor port
4366 Details : tcpnal_set_default_params() was not called and was therefore
4367 ignoring the environment varaible TCPNAL_PORT, as well as other
4368 TCPNAL_ environment variables
4373 Description: two objects could be created on the same OST for a single file
4374 Details : If an OST is down, in some cases it was possible to create two
4375 objects on a single OST for a single file. No problems other
4376 than potential performance impact and spurious error messages.
4380 Bugzilla : 5681, 9562
4381 Description: Client may oops in ll_unhash_aliases
4382 Details : Client dcache may become inconsistent in race condition.
4383 In some cases "getcwd" can fail if the current directory is
4389 Description: Inode refcounting problems in NFS export code
4390 Details : link_raw functions used to call d_instantiate without obtaining
4391 extra inode reference first.
4395 Bugzilla : 9942, 9903
4396 Description: Referencing freed requests leading to crash, memleaks with NFS.
4397 Details : We used to require that call to ll_revalidate_it was always
4398 followed by ll_lookup_it. Also with revalidate_special() it is
4399 possible to call ll_revalidate_it() twice for the same dentry
4400 even if first occurence returned success. This fix changes semantic
4401 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4402 reference on a request referred from the intent.
4403 ll_intent_release() then releases such a request.
4406 Frequency : rare, normally benchmark loads only
4408 Description: unlinked inodes were kept in memory on the client
4409 Details : If a client is repeatedly creating and unlinking files it
4410 can accumulate a lot of stale inodes in the inode slab cache.
4411 If there is no other client load running this can cause the
4412 client node to run out of memory. Instead flush old inodes
4413 from client cache that have the same inode number as a new inode.
4416 Frequency : SLES9 2.6.5 kernel and long filenames only
4417 Bugzilla : 9969, 10379
4418 Description: utime reports stale NFS file handle
4419 Details : SLES9 uses out-of-dentry names in some cases, which confused
4420 the lustre dentry revalidation. Change it to always use the
4424 Frequency : rare, unless heavy write-truncate concurrency is continuous
4425 Bugzilla : 4180, 6984, 7171, 9963, 9331
4426 Description: OST becomes very slow and/or deadlocked during object unlink
4427 Details : filter_destroy() was holding onto the parent directory lock
4428 while truncating+unlinking objects. For very large objects this
4429 may block other threads for a long time and slow overall OST
4430 responsiveness. It may also be possible to get a lock ordering
4431 deadlock in this case, or run out of journal credits because of
4432 the combined truncate+unlink. Solution is to do object truncate
4433 first in one transaction without parent lock, and then do the
4434 final unlink in a new transaction with the parent lock. This
4435 reduces the lock hold time dramatically.
4438 Frequency : rare, 2.4 kernels only
4440 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4441 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4442 before unmounting their filesystem in order to avoid deadlock.
4443 The kernel_locked() function in 2.4 kernels only checks whether
4444 the kernel lock is held, not whether it is this process that is
4445 holding it as 2.6 kernels do.
4450 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4451 Details : The client ptlrpc code may be trying to reconnect to a down
4452 server before a previous connection attempt has timed out.
4453 Increase the reconnect interval to be longer than the connection
4454 timeout interval to avoid sending duplicate connections to
4458 Frequency : echo_client brw_test command
4460 Description: fix echo_client to work with OST preallocated code
4461 Details : OST preallocation code (5137) didn't take echo_client IO path
4462 into account: echo_client calls filter methods outside of any
4463 OST thread and, hence, there is no per-thread preallocated
4464 pages and buffers to use. Solution: hijack pga pages for IO. As
4465 a byproduct, this avoids unnecessary data copying.
4469 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4470 Description: Client can oops in mdc_commit_close() after open replay
4471 Details : It was possible for the MDS to return an open request with no
4472 transaction number in mds_finish_transno() if the client was
4473 evicted, but without actually returning an error. Clients
4474 would later try to replay that open and may trip an assertion
4475 Simplify the client close codepath, and always return an error
4476 from the MDS in case the open is not successful.
4479 Frequency : rare, 2.6 OSTs only
4481 Description: OST may deadlock under high load on fragmented files
4482 Details : If there was a heavy load and highly-fragmented OST filesystems
4483 it was possible to have all the OST threads deadlock waiting on
4484 allocation of biovecs, because the biovecs were not released
4485 until the entire RPC IO was completed. Instead, release biovecs
4486 as soon as they are complete to ensure forward IO progress.
4488 Severity : enhancement
4490 Description: Support for specifying external journal device at mount
4491 Details : If an OST or MDS device is formatted with an external journal
4492 device, this device major/minor is stored in the ext3 superblock
4493 and may not be valid for failover. Allow detecting and
4494 specifying the external journal at mount time.
4499 Description: Mounting an MDS with pending unlinked files may cause oops
4500 Details : target_finish_recovery() calls mds_postrecov() which returned
4501 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4502 considers this an error and immediately begins cleaning up the
4503 lov, just after starting the mds_lov process
4505 Severity : enhancement
4507 Description: Implement 'lfs df' to report actual free space on per-OST basis
4508 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4509 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4510 usage of objects; '-h' to report in human readable format.
4512 ------------------------------------------------------------------------------
4514 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4521 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4522 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4523 at startup during orphan unlinking. Instead, skip buddy update
4524 before setup, as it will be regenerated after recovery is complete.
4529 Description: If an OST is inactive, its locks might reference stale inodes.
4530 Details : lov_change_cbdata() must iterate over all namespaces, even if
4531 they are inactive to clear inode references from the lock.
4533 Severity : enhancement
4534 Frequency : occasional, if non-standard max_dirty_mb used
4536 Description: Client will block write RPCs if not enough grant
4537 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4538 then the client will block writes while waiting for another RPC
4539 to complete instead of consuming its dirty limit. With change
4540 we get improved performance when max_dirty_mb is small.
4542 Severity : enhancement
4543 Bugzilla : 3389, 6253
4544 Description: Add support for supplementary groups on the MDS.
4545 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4546 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4547 lookups for user supplementary groups into a cache.
4551 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4552 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4553 for liblustre O_EXCL is enforced by MDS.
4557 Description: readdir never returns NULL in liblustre.
4558 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4559 dentry in struct dirent.
4563 Frequency : liblustre only, depends on application IO pattern
4564 Description: liblustre clients evicted if not contacting servers
4565 Details : Don't put liblustre clients into the ping_evictor list, so
4566 they will not be evicted by the pinger ever.
4568 Severity : enhancement
4570 Description: Add ability to evict clients by NID from MDS.
4571 Details : By echoing "nid:$NID" string into
4572 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4573 $NID would be instantly evicted from this MDS and from all active
4574 OSTs connected to it.
4578 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4579 Details : lookup_it_finish() used to query file size from OSTs that was not
4584 Description: service threads change working directory to that of init
4585 Details : Starting lustre service threads may pin the working directory
4586 of the parent thread, making that filesystem busy. Threads
4587 now change to the working directory of init to avoid this.
4591 Frequency : during shutdown only
4592 Description: shutdown with a failed MDS or OST can cause unmount to hang
4593 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4598 Frequency : 2.6 only
4599 Description: chmod/chown may include an extra supplementary group
4600 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4601 supplementary group and if none is specified this is used.
4604 Bugzilla : 5479 (6816)
4606 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4607 Details : If lookup is in progress on a file that is unlinked we might try
4608 to revalidate the inode and fail in revalidate after lookup is
4609 complete and ll_file_open() enqueues the open again but
4610 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4613 Frequency : always, if lconf --abort_recovery used
4615 Description: lconf --abort_recovery fails with 'Operation not supported'
4616 Details : lconf was attempting to abort recovery on the MDT device and not
4619 ------------------------------------------------------------------------------
4621 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4626 Frequency : rare (only unsupported configurations with a node running as an
4628 Bugzilla : 6514, 5137
4629 Description: Mounting a Lustre file system on a node running as an OST could
4631 Details : OSTs now preallocates memory needed to write out data at
4632 startup, instead of when needed, to avoid having to
4633 allocate memory in possibly low memory situations.
4634 Specifically, if the file system is mounted on on OST,
4635 memory pressure could force it to try to write out data,
4636 which it needed to allocate memory to do. Due to the low
4637 memory, it would be unable to do so and the node would
4638 become unresponsive.
4640 Severity : enhancement
4642 Description: Addition of lconf --service command line option
4643 Details : lconf now accepts a '--service <arg>' option, which is
4644 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4646 Severity : enhancement
4648 Description: Failover mode is now the default for OSTs.
4649 Details : By default, OSTs will now run in failover mode. To return to
4650 the old behaviour, add '--failout' to the lmc line for OSTs.
4652 Severity : enhancement
4654 Description: Health checks are now provided for MDS and OSTs
4655 Details : Additional detailed health check information on MSD and OSTs
4656 is now provided through the procfs health_check value.
4659 Frequency : occasional, depends on IO load
4661 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4662 after numerous create/delete cycles
4663 Details : The ext3 inode allocation policy would not allocate new inodes
4664 very well on the OSTs because there are no new directories
4665 being created. Instead we look for groups with free space if
4666 the parent directories are nearly full.
4671 Description: Network or server problems during mount may cause partially
4672 mounted clients instead of returning an error.
4673 Details : The config llog parsing code may overwrite the error return
4674 code during mount error handling, returning success instead
4680 Description: MDS can fail to allocate large reply buffers
4681 Details : After long uptimes the MDS can fail to allocate large reply
4682 buffers (e.g. zconf client mount config records) due to memory
4683 fragmentation or consumption by the buffer cache. Preallocate
4684 some large reply buffers so that these replies can be sent even
4685 under memory pressure.
4689 Frequency : rare (liblustre)
4690 Description: fsx running with liblustre complained that using truncate() to
4691 extend the file doesn't work. This patch corrects that issue.
4692 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4693 ATTR_SIZE and lsm use in llu_setattr_raw.
4697 Frequency : rare, only 2.6 kernels
4698 Description: Unusual file access patterns on the MDS may result in inode
4699 data being lost in very rare circumstances.
4700 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4701 code in the 2.6 kernel caused the inode and/or EA data not to
4702 be read from disk, causing single-file corruption.
4706 Frequency : rare, only 2.6 filesystems using extents
4707 Description: Heavy concurrent write and delete load may cause data corruption.
4708 Details : It was possible under high-load situations to have an extent
4709 metadata block in the block device cache from a just-unlinked
4710 file overwrite a newly-allocated data block. We now unmap any
4711 metadata buffers that alias just-allocated data blocks.
4715 Frequency : filesystems with default stripe_count larger than 77
4716 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4717 Details : lconf specifies an inode size of 4096 bytes when the default
4718 stripe_count is larger than 77. This conflicts with the default
4719 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4720 this case to avoid pinning too much memory for large EAs.
4722 ------------------------------------------------------------------------------
4724 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4729 Frequency : rare (extremely heavy IO load with hundreds of clients)
4731 Description: Client is evicted, gets IO error writing to file
4732 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4733 caused clients to be evicted for AST timeouts. The fixes in
4734 bug 5192 mean we no longer need to have such short AST timeouts
4735 so ldlm_timeout has been increased.
4738 Frequency : occasional during --force or --failover shutdown under load
4739 Bugzilla : 5949, 4834
4740 Description: Server oops/LBUG if stopped with --force or --failover under load
4741 Details : a collection of import/export refcount and cleanup ordering
4742 issues fixed for safer force cleanup
4745 Frequency : only filesystems larger than 120 OSTs
4746 Bugzilla : 5990, 6223
4747 Description: lfs getstripe would oops on a very large filesystem
4748 Details : lov_getconfig used kfree on vmalloc'd memory
4751 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4752 Bugzilla : 6242, 6243
4753 Description: reading from files that had been truncated to a non-zero size
4754 but never opened returned no data
4755 Details : ll_file_read() reads zeros from no-object files to EOF
4760 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4761 Details : There's little harm in aborting MDS/OSS recovery and letting it
4762 try again, so I removed the LASSERT and return an error instead.
4764 Severity : enhancement
4766 Description: New debugging infrastructure for tracking down data corruption
4767 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4768 (b) cover more of the client-side code path, and (c) try to narrow
4769 down where problems occurred
4773 Bugzilla : 3819, 4364, 4397, 6313
4774 Description: Racing close and eviction MDS could cause assertion in mds_close
4775 Details : It was possible to get multiple mfd references during close and
4776 client eviction, leading to one thread referencing a freed mfd.
4778 Severity: : enhancement
4779 Bugzilla : 3262, 6359
4780 Description: Attempts to reconnect to servers are now more aggressive.
4781 Details : This builds on the enhanced upcall-less recovery that was added
4782 in 1.4.2. When trying to reconnect to servers, clients will
4783 now try each server in the failover group every 10 seconds. By
4784 default, clients would previously try one server every 25 seconds.
4789 Description: After recovery, certain operations trigger a failed
4790 assertion on a client.
4791 Details : Failing over an mds, using lconf -d --failover, while a
4792 client was doing a readdir() call would cause the client to
4793 LBUG after recovery completed and the readdir() was resent.
4795 Severity : enhancement
4797 Description: Default groups are now added by lconf
4798 Details : You can now run lconf --group <servicename> without having to
4799 manually add groups with lmc.
4802 Frequency : occasional
4804 Description: Nodes with an elan id of 0 trigger a failed assertion
4807 Frequency : always when accessing e.g. tty/console device nodes
4809 Description: tty and some other devices nodes cannot be used on lustre
4810 Details : file's private_data field is used by device data and lustre
4811 values in there got lost. New field was added to struct file to
4812 store fs-specific private data.
4815 Frequency : when exporting Lustre via NFS
4817 Description: NFSD failed occasionally when looking up a path component
4818 Details : NFSD is looking up ".." which was broken in ext3 directories
4819 that had grown large enough to become hashed.
4822 Frequency : Clusters with multiple interfaces not on the same subnet
4824 Description: Nodes will repeatedly try to reconnect to an interface which it
4825 cannot reach and report an error to the log.
4826 Details : Extra peer list entries will be created by lconf with some peers
4827 unreachable. lconf now validates the peer before adding it.
4830 Frequency : Only if a default stripe is set on the filesystem root.
4832 Description: Setting a default stripe on the filesystem root prevented the
4833 filesystem from being remounted.
4834 Details : The client was sending extra request flags in the root getattr
4835 request and did not allocate a reply buffer for the dir EA.
4838 Frequency : occasional, higher if lots of files are accessed by one client
4839 Bugzilla : 6159, 6097
4840 Description: Client trips assertion regarding lsm mismatch/magic
4841 Details : While revalidating inodes the VFS looks up inodes with ifind()
4842 and in rare cases can find an inode that is being freed.
4843 The ll_test_inode() code will free the lsm during ifind()
4844 when it finds an existing inode and then the VFS later attaches
4845 this free lsm to a new inode.
4849 Bugzilla : 6422, 7030
4850 Description: MDS deadlock between mkdir and client eviction
4851 Details : Creating a new file via mkdir or mknod (starting a transaction
4852 and getting the ns lock) can deadlock with client eviction
4853 (gets ns lock and trying to finish a synchronous transaction).
4856 Frequency : occasional
4857 Description: While starting a server, the fsfilt_ext3 module could not be
4859 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4860 kernels. Previously, lconf would still use the ext3 name
4861 when trying to load modules. Now, it will correctly use
4862 ext3 on 2.4 and ldiskfs on 2.6.
4864 Severity : enhancement
4865 Description: The default stripe count has been changed to 1
4866 Details : The interpretation of the default stripe count (0, to lfs
4867 or lmc) has been changed to mean striping across a single
4868 OST, rather than all available. For general usage we have
4869 found a stripe count of 1 or 2 works best.
4871 Severity : enhancement
4872 Description: Add support for compiling against Cray portals.
4873 Details : Conditional compiling for some areas that are different
4877 Frequency : occasional
4878 Bugzilla : 6409, 6834
4879 Description: Creating files with an explicit stripe count may lead to
4880 a failed assertion on the MDS
4881 Details : If some OSTs are full or unavailable, creating files may
4882 trigger a failed assertion on the MDS. Now, Lustre will
4883 try to use other servers or return an error to the
4887 Frequency : occasional
4889 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4890 caused lock timeout during readahead (since 1.4.2).
4891 Details : Processes doing readahead might match a lock that hasn't been
4892 granted yet if there are overlapping and conflicting lock
4893 requests. The readahead process waits on ungranted lock
4894 (original lock is CBPENDING), while OST waits for that process
4895 to cancel CBPENDING read lock and eventually evicts client.
4897 Severity : enhancement
4899 Description: Initial enabling of flock support for clients
4900 Details : Implements fcntl advisory locking and file status functions.
4901 This feature is provided as an optional mount flag (default
4902 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4903 locking are implemented yet, and those that are are not guaranteed
4904 to be completely correct in production environments.
4905 mount -t lustre -o [flock|noflock] ...
4908 Frequency : occasional
4910 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4911 assertion in the ext3 JBD (journaling) layer.
4912 Details : The b_committed_data struct is protected by the big kernel lock
4913 in 2.4 kernels, serializing journal_commit_transaction() and
4914 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4915 access to this struct. In 2.6 kernels there is finer grained
4916 locking to improve SMP performance of the JBD layer.
4920 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4925 Description: merge and cleanup kernel patches.
4928 -----------------------------------------------------------------------------
4930 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4932 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4933 versions of Lustre. Please update all servers and clients to
4934 version 1.4.2 or later at the same time. You must also run
4935 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4936 to update the configuration logs.
4938 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4939 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4940 - swab llog records of type '0' so we get proper header size/idx (5861)
4941 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4942 - fix rename of one directory over another leaking an inode (5953)
4943 - avoid SetPageDirty on 2.6 (5981)
4944 - don't re-add just-being-destroyed locks to the waiting list (5653)
4945 - when creating new directories, inherit the parent's custom
4946 striping settings if present parent (3048)
4947 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4948 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4949 - don't hold client locks on temporary worklist from l_lru (5666)
4950 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4951 - automatically evict dead clients (5921)
4952 - Update file size properly in create+truncate+fstat case (6196)
4953 - Do not unhash mountpoint dentries, do not allow removal of
4955 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4956 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4957 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4958 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4959 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4960 - Correctly handle reads of files with no objects (6243)
4961 - lctl recover will also mark a device active if deactivate used (5933)
4963 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4964 - allow --write-conf on an MDS with different nettype than client (5619)
4965 - don't write config llogs to MDS for mounts not from that MDS (5617)
4966 - lconf should create multiple TCP connections from a client (5201)
4967 - init scripts are now turned off by default; run chkconfig --on
4968 lustre and chkconfig --on lustrefs to use them
4969 - upcalls are no longer needed for clients to recover to failover
4971 - add --abort-recovery option to lconf to abort recovery on device
4973 - add support for an arbitrary number of OSTs (3026)
4974 - Quota support protocol changes.
4975 - forward compatibility changes to wire structs (6007)
4976 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4977 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4978 - improved "lustre status" script
4979 - initialize blocksize for non-regular files (6062)
4980 - added --disable-server and --disable-client configure options (5782)
4981 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4982 - Vanilla 2.4.29 support
4983 - increase maximum number of obd devices to 520 (6242)
4984 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4985 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4986 SLES 9 2.6.5 kernels
4987 - limit stripes per file to 160 (the maximum EA size) (6093)
4989 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4992 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4993 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4994 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4995 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4996 - small ext3 extents cleanups and fixes (5733)
4997 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4998 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4999 - avoid needless client->OST connect, fix handle mismatch (5317)
5000 - fix DLM error path that led to out-of-sync client, long delays (5779)
5001 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5002 - fix several locking issues related to i_size (5492,5624,5654,5672)
5003 - don't move pending lock onto export if it is already evicted (5683)
5004 - fix kernel oops when creating .foo in unlinked directory (5548)
5005 - fix deadlock in obdfilter statistics vs. object create (5811)
5006 - use time_{before,after} to avoid timer jiffies wrap (5882)
5007 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5008 - Do not leak request if server was not able to process it (5154)
5009 - If mds_open unable to find parent dir, make that negative lookup(5154)
5010 - don't create new directories with extent-mapping (5909, 5936)
5012 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5013 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5014 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5015 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5016 - support for RHEL4 kernel on i686 (5773)
5017 - provide error messages when incompatible logs are encountered (5898)
5019 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5020 * version 1.4.0.10 (1.4.1 release candidate 1)
5022 - don't keep a lock reference when lock is not granted (4238)
5023 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5024 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5025 - drop import inflight refcount on signal_completed_replay error (5255)
5026 - unlock page after async write error during send (3677)
5027 - handle missing objects in filter_preprw_read properly (5265)
5028 - no transno return for symlink open, don't save no-trasno open (3440)
5029 - don't try to complete elan receive that already failed (4012)
5030 - free RPC server reply state on error (5406)
5031 - clean up thread from ptlrpc_start_thread() on error (5160)
5032 - readahead could read extra page into cache that wasn't ejected (5388)
5033 - prevent races in class_attach/setup/cleanup/detach (5260)
5034 - don't dereference de->d_inode after l_dput of de (5458)
5035 - use "int" for stripe value returned from lock_to_stripe (5544)
5036 - mballoc allocation and error-checking fixes in 2.6 (5504)
5037 - block device patches to fix I/O request sizes in 2.6 (5482)
5038 - look up hostnames for IB nals (5602)
5039 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5040 - don't start multiple acceptors for the same port (5277)
5041 - fix incorrect LASSERT in mds_getattr_name (5635)
5042 - export a proc file for general "ping" checking (5628)
5043 - fix "lfs check" to not block when the MDS is down (5628)
5045 - service request history (4965)
5046 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5047 - create an "evict_client" /proc entry on OSTs, like the MDS has
5048 - fix mount usage message, return errors per mount(8) (5168)
5049 - change grep [] to grep "[]" in tests so they work in more UMLs
5050 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5051 - remove ancient LOV_MAGIC_V0 EA support (5047)
5052 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5053 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5054 - allow readahead allocations to fail when low on memory (5383)
5055 - mmap locking landed again, after considerable improvement (2828)
5056 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5058 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5061 - send OST transaction number in read/write reply to free req (4966)
5062 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5063 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5064 - the watchdog thread now runs as interruptible (5246)
5065 - flock/lockf fixes (but it's still disabled, pending 5135)
5066 - don't use EXT3 constants in llite code (5094)
5067 - memory shortage at startup could cause assertion (5176)
5069 - reorganization of lov code
5070 - single portals codebase
5072 - add extents/mballoc support (5025)
5073 - direct I/O reads in the obdfilter (4048)
5074 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5076 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5079 - send OST transaction number in read/write reply to free req (4966)
5080 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5081 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5082 - the watchdog thread now runs as interruptible (5246)
5083 - handle missing objects in filter_preprw_read properly (5265)
5084 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5085 - drop import inflight refcount on signal_completed_replay error (5255)
5086 - unlock page after async write error during send (3677)
5087 - return original error code on reconstructed replies (3761)
5088 - no transno return for symlink open, don't save no-trasno open (3440)
5090 - add pid to ldlm debugging output (4922)
5091 - bump the watchdog timeouts -- we can't handle 30sec yet
5092 - extra debugging for orphan dentry/inode bug (5259)
5094 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5097 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5098 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5099 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5100 - fix NULL dereference in /proc/sys/portals/routes (4827)
5101 - allow failed mdc_close() operations to be interrupted (4561)
5102 - stop precreate on OST before MDS would time out on it (4778)
5103 - don't send partial-page writes before EOF from client (4410)
5104 - discard client grant for sub-page writes on large-page clients (4520)
5105 - don't free dentries not owned by NFS code, check generation (4806)
5106 - fix lsm leak if mds_create_objects() fails (4801)
5107 - limit debug_daemon file size, always print CERROR messages (4789)
5108 - use transno after validating reply (3892)
5109 - process timed out requests if import state changes (3754)
5110 - update mtime on OST during writes, return in glimpse (4829)
5111 - add mkfsoptions to LDAP (4679)
5112 - use ->max_readahead method instead of zapping global ra (5039)
5113 - don't interrupt __l_wait_event() during strace
5115 - add software watchdogs to catch hung threads quickly (4941)
5116 - make lustrefs init script start after nfs is mounted
5117 - fix CWARN/ERROR duplication (4930)
5118 - return async write errors to application if possible (2248)
5119 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5120 - print NAL number in %x format (4645)
5121 - update barely-supported suse-2.4.21-171 series (4842)
5122 - support for sles 9 %post scripts
5123 - support for building 2.6 kernel-source packages
5124 - support for sles km_* packages
5126 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5129 - ignore -ENOENT errors in osc_destroy (3639)
5130 - notify osc create thread that OSC is being cleaned up (4600)
5131 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5132 - reconstruct ost_handle() like mds_handle() (4657)
5133 - create a new thread to do import eviction to avoid deadlock (3969)
5134 - let lconf resolve symlinked-to devices (4629)
5135 - don't unlink "objects" from directory with default EA (4554)
5136 - hold socknal file ref over connect in case target is down (4394)
5137 - allow more than 32000 subdirectories in a single directory (3244)
5138 - fix blocks count for O_DIRECT writes (3751)
5139 - OST returns ENOSPC from object create when no space left (4539)
5140 - don't send truncate RPC if file size isn't changing (4410)
5141 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5142 - bind to privileged port in socknal and tcpnal (3689)
5144 - rate limit CERROR/CWARN console message to avoid overload (4519)
5145 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5146 - basic mmap support (3918)
5147 - kernel patch series update from b1_4 (4711)
5149 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5152 - avoid crash during MDS cleanup with OST shut down (2775)
5153 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5154 - don't use bad inodes on the MDS (3744)
5155 - dynamic object preallocation to improve recovery speed (4236)
5156 - don't hold spinlock over lock dumping or change debug flags (4401)
5157 - don't zero obd_dev when it is force cleaned (3651)
5158 - print grants to console if they go negative (4431)
5159 - "lctl deactivate" will stop automatic recovery attempts (3406)
5160 - look for existing locks in ldlm_handle_enqueue() (3764)
5161 - don't resolve lock handle twice in recovery avoiding race (4401)
5162 - revalidate should check working dir is a directory (4134)
5164 - don't always mark "slow" obdfilter messages as errors (4418)
5166 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5169 - don't close LustreDB during write_conf until it is done (3860)
5170 - fix typo in lconf for_each_profile (3821)
5171 - allow dumping logs from multiple threads at one time (3820)
5172 - don't allow multiple threads in OSC recovery (3812)
5173 - fix debug_size parameters (3864)
5174 - fix mds_postrecov to initialize import for llog ctxt (3121)
5175 - replace config semaphore with spinlock (3306)
5176 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5177 - don't allow enqueue to complete on a destroyed export (3822)
5178 - down write_lock before checking llog header bitmap (3825)
5179 - recover from lock replay timeout (3764)
5180 - up llog sem before sending rpc (3652)
5181 - reduce ns lock hold times when setting kms (3267)
5182 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5183 - fix NULL deref and obd_dev leak on setup error (3312)
5184 - replace some LBUG about llog ops with error handling (3841)
5185 - don't match INVALID dentries from d_lookup and spin (3784)
5186 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5187 - fix invalid assertion in ptlrpc_set_wait (3880)
5189 - add libwrap support for the TCP acceptor (3996)
5190 - add /proc/sys/portals/routes for non-root route listing (3994)
5191 - allow setting MDS UUID in .xml (2580)
5192 - print the stack of a process that LBUGs (4228)
5194 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5197 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5198 - make sure to unset replay flag from failed open requests (3440)
5199 - if default stripe count is 0, use OST count for inode size (3636)
5200 - update parent mtime/ctime on client for create/unlink (2611)
5201 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5202 - free recovery state on server during a forced cleanup (3571)
5203 - unregister_reply for resent reqs (3063)
5204 - loop back devices mounting and status check on 2.6 (3563)
5205 - fix resource-creation race that can provoke i_size == 0 (3513)
5206 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5207 - more debugging for page-accounting assertion (3746)
5208 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5209 - avoid deadlock after precreation failure (3758)
5210 - fix race and lock order deadlock in orphan handling (3450, 3750)
5211 - add validity checks when grabbing inodes from l_ast_data (3599)
5213 - add /proc/.../recovery_status to obdfilter (3428)
5214 - lightweight CDEBUG infrastructure, debug daemon (3668)
5215 - change default OSC RPC parameters to be better on small clusters
5216 - turn off OST read cache for files smaller than 32MB
5217 - install man pages and include them in rpms (3100)
5218 - add new init script for (un)mounting lustre filesystems (2593)
5219 - run chkconfig in %post for init scripts (3701)
5220 - drop scimac NAL (unmaintained)
5222 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5225 - clean kiobufs before and after use (3485)
5226 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5227 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5228 - change init script to fail more clearly if not run as root (1528)
5229 - allow clients to reconnect during replay (1742)
5230 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5231 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5232 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5233 - fix cond_resched (3554)
5234 - only evict unfinished clients after recovery (3515)
5235 - allow bulk resend, prevent data loss (3570)
5236 - dynamic ptlrpc request buffer allocation (2102)
5237 - don't allow unlinking open directory if it isn't empty (2904)
5238 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5239 - remove extraneous obd dereference causing LASSERT failure (3334)
5240 - don't use get_cycles() when creating temp. files on the mds (3156)
5241 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5242 - handle EEXIST for set-stripe, set proper directory name (3336)
5244 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5245 - fix ksocknal_fmb_callback() error messages (2918)
5247 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5250 - don't copy lvb into (possibly NULL) reply on error (2983)
5251 - don't deref dentry after dput, don't free lvb on error (2922)
5252 - use the kms to determine writeback rpc length (2947)
5253 - increment oti_logcookies when osc is inactive (2948)
5254 - update client's i_blocks count via lvb messages (2543)
5255 - handle intent open/close of special files properly (1557)
5256 - mount MDS with errors=remount-ro, like obdfilter (2009)
5257 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5258 - don't use cancelling-locks' kms values (2947)
5259 - use highest lock extent for kms, not last one (2925)
5260 - don't dereference ERR_PTR() dentry in error handling path (3107)
5261 - fix thread race in portals_debug_dumplog() (3122)
5262 - create lprocfs device entries at setup instead of at attach (1519)
5263 - common AST error handler, don't evict client on completion race (3145)
5264 - zero nameidata in detach_mnt in 2.6 (3118)
5265 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5266 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5267 - initialize RPC timeout timer earlier for 2.6 (3219)
5268 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5269 - print nal/nid for unknown nid (3258)
5270 - additional checks for oscc recovery before doing precreate (3284)
5271 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5272 - don't crash in mdc_close for bad permissions on open (3285)
5273 - zero i_rdev for non-device files (3147)
5274 - clear page->private before handing to FS, better assertion (3119)
5275 - tune the read pipeline (3236)
5276 - fix incorrect decref of invalidated dentry (2350)
5277 - provide read-ahead stats and refine rpc in flight stats (3328)
5278 - don't hold journal transaction open across create RPC (3313)
5279 - update atime on MDS at close time (3265)
5280 - close LDAP connection when recovering to avoid server load (3315)
5281 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5282 - don't leak open file on MDS after open resend (3325)
5283 - serialize filter_precreate and filter_destroy_precreated (3329)
5284 - loop device shouldn't call sync_dev() for nul device (3092)
5285 - clear page cache after eviction (2766)
5286 - resynchronize MDS->OST in background (2824)
5287 - refuse to mount the same filesystem twice on same mountpoint (3394)
5288 - allow llmount to create routes for mounting behind routers (3320)
5289 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5290 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5291 - fix rare problem with rename on htree directories (3417)
5293 - allow default OST striping configuration per directory (1414)
5294 - fix compilation for qswnal for 2.6 kernels (3125)
5295 - increase maximum number of MDS request buffers for large systems
5296 - change liblustreapi to be useful for external progs like lfsck (3098)
5297 - increase local configuration timeout for slow disks (3353)
5298 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5300 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5303 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5304 - don't overwrite extent policy data in reply if lock was blocked (2901)
5305 - drop filter export grants atomically with removal from device (2663)
5306 - del obd_self_export from work_list in class_disconnect_exports (2908)
5307 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5308 - swab reply message in mdc_close, other PPC fixes (2464)
5309 - fix destroying of named logs (2325)
5310 - overwrite old logs when running lconf --write_conf (2264)
5311 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5312 - fix race in target_handle_connect (2898)
5313 - mds_reint_create() should take same inode create lock (2926)
5314 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5315 - don't close files for self_export to avoid uninitialized obd (2936)
5316 - allow MDS with the same name as client node (2939)
5317 - hold dentry reference for closed log files for unlink (2325)
5318 - reserve space for all logs during transactions (2059)
5319 - don't evict page beyond end of stripe extent (2925)
5320 - don't oops on a deleted current working directory (2399)
5321 - handle hard links to targets without a parent properly (2517)
5322 - don't dereference NULL lock when racing during eviction (2867)
5323 - don't grow lock extents when lots of conflicting locks (2919)
5325 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5328 - account for cache space usage on clients to avoid data loss (974)
5329 - lfsck support in lustre kernel code (2349)
5330 - reduce journal credits needed for BRW writes (2370)
5331 - orphan handling to avoid losing space on client/server crashes
5332 - ptlrpcd can be blocked, stopping ALL progress (2477)
5333 - use lock value blocks to assist in proper KMS, faster stat (1021)
5334 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5335 - recovery for initial connections (2355)
5336 - fixes for mds_cleanup_orphans (1934)
5337 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5338 - block all file creations until orphan recovery completes (1901)
5339 - client remove rq_connection from request struct (2423)
5340 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5341 - recovery timer race (2670)
5342 - mdc_close recovey bug (2532)
5343 - ptlrpc cleanup bug (2710)
5344 - mds timeout on local locks (2588)
5345 - namespace lock held during RPCs (2431)
5346 - handle interrupted sync write properly (2503)
5347 - don't try to handle a message that hasn't been replied to (2699)
5348 - client assert failure during cleanup after abort recovery (2701)
5349 - leak mdc device after failed mount (2712)
5350 - ptlrpc_check_set allows timedout requests to complete (2714)
5351 - wait for inflight reqs when ptlrpcd finishes (2710)
5352 - make sure unregistered services are removed from the srv_list
5353 - reset bulk XID's when resending them (caught by 1138 test)
5354 - unregister_bulk after timeout
5355 - fix lconf error (2694)
5356 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5357 - readahead locks pages, leaves pending causing memory pressure (2673)
5358 - increase OST request buffers to 4096 on large machines (2729)
5359 - fix up permission of existing directories in simple_mkdir (2661)
5360 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5361 - don't assign transno to errored transactions (2742)
5362 - don't delete objects on OST if given a bogus objid from MDS (2751)
5363 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5364 - if rq_no_resend, then timeout request after recovery (2432)
5365 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5366 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5367 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5368 - MDS recovery shouldn't delete objects that it has given out (2730)
5369 - if enqueue arrives after completion, don't clobber LVB (2819)
5370 - don't unlock pages twice when trigger_group_io returns error (2814)
5371 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5372 - don't write pages to disk if there was an error (1450)
5373 - don't ping imports that have recovery disabled (2676)
5374 - take buffered bytes into account when balancing socknal conn (2817)
5375 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5376 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5377 - remove little-endian swabbing of llog records (1987)
5378 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5379 - retry reposting request buffers if they fail (1191)
5380 - grow extent at grant time to avoid granting a revoked lock (2809)
5381 - lock revoke doesn't evict page if covered by a second lock (2765)
5382 - disable VM readahead to avoid reading outside lock extents (2805)
5384 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5385 - updated kernel patches for hp-2.4.20 kernel (2681)
5387 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5390 - fix truncated write corruption (2366)
5391 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5393 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5394 - ensure len(uuid) < 37 in lmc (1171)
5395 - fix ia64 OOPS in llog_test (2255)
5396 - zero end of page at obdfilter for partial page writes (2648)
5397 - don't leave stale dentries around after renames (bug 2428)
5398 - fix timeouts when evicting a client with a single lock held (2642)
5399 - set deadline for the initial HELLO message to drain (2634)
5400 - print out dotted-quad IP addresses in the socknal (2302)
5402 - additional debugging for MDS client eviction problem (2443)
5403 - fix mkfsoptions support for osts (2603, 2604)
5405 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5408 - add series for the vanilla 2.6.0 kernel
5409 - add series for the vanilla 2.4.24 kernel
5410 - add series for a cray x86/64 UL kernel drop
5411 - fix xattr patches for the vanilla 2.4.19 series
5413 - generate true UUIDs in lmc (1171)
5414 - have portals stack dumping break in UML (2466)
5415 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5416 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5417 - wake callback waiting threads on client eviction (2460)
5418 - Add --ptldebug and --subsystem to lmc (1719)
5419 - update assertion to allow safe interrupt allocation
5420 - set rq_no_resend for cancel requests (2432)
5421 - recalculate ptlrpcd timeout after resend (2494)
5422 - call vfs_rmdir when removing pending directories (2368)
5423 - fix renaming a file to itself (2429)
5424 - lmc creates a default one-stripe lov (2454)
5425 - expand procfs space to handle large clusters (2326)
5426 - increase UML stack to avoid overflow
5427 - update lconf's list of debug and subsystem masks
5428 - fix lfs find --obd (2510)
5429 - /proc tunable for disabling filter read caching (2591)
5430 - stop rpm packages from altering slapd.conf (2301)
5431 - disable nagle in the socknal under 0conf (2578)
5432 - choose mds inode size based on stripe count (2572)
5433 - fix kernel-source rpm problems (2516)
5435 - add --disable-doc to avoid pdf generation (2421)
5436 - update documentation, tests, type-os, comments
5437 - avoid format warnings on ia64
5438 - remove the TOE NAL
5439 - tiny code cleanups by removing unused fields
5441 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5444 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5445 - remove the most blatant lies from BUILDING file (2371)
5446 - change default debug level to reasonable production setting
5447 - reduce client side cache size to reduce cache flush time
5448 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5449 - make TCP zerocopy and pinger support enabled by default (2476)
5450 - sync writes completed after process exits caused crashes (2319)
5451 - maintain correct mount count on the MDS (2356)
5452 - backout 1557, because 2316 wasn't really fixed
5453 - better file I/O statistics gathering in /proc
5454 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5455 - another kernel patch to fix zero-copy TCP function export
5456 - don't take duplicate lock when processing re-sent getattr (2420)
5457 - lctl uses obd_self_export instead of creating new conn (2353)
5458 - MDS/OST recovery case which requires object creation asserted (2425)
5459 - move lfs from /usr/sbin to /usr/bin in packages
5460 - fix race between mds_client_add and mds_client_free (2417)
5461 - use kmalloc instead of slabs in portals (2430)
5462 - don't create duplicate records when a failover MDS is present (2442)
5463 - remove unnecessary mount age check (2332)
5464 - don't remove directory inodes from locks prematurely (2451)
5465 - don't break if MDS service name is the same as hostname (2103)
5466 - fix races in client write RPC generation when cache full (2482)
5468 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5471 - remove now-unused request->rq_obd (278)
5472 - if an allocation fails, print out how much memory we've used (1933)
5473 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5474 - add the "configurable stack size" patch to most series files (1256)
5475 - ability to write large log records, for 100+ OST configs (2306)
5476 - fix NULL deref when filter_prep fails (2314)
5477 - fix operator precedence error in filter_sync
5478 - dynamic allocation of socknal TX descriptors (2315)
5479 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5480 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5481 - ensure that CWARN messages in Portals always get to the syslog
5482 - __init/__exit are not for prototype decls (ldlm_init/exit)
5483 - x86-64 compile warning fixes
5484 - fix gateway LMC keyword conflict (2318)
5485 - fix MDS lock inversions in getattr/reint paths (1844)
5486 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5487 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5488 - DLM race condition prevented some lock evictions (2328)
5489 - ENOMEM detection and retry on socknal sends (2230)
5490 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5491 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5492 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5493 - better error messages when a client is rejected during recovery (1505)
5494 - avoid cancelling locks which were never granted, after failure (2330)
5495 - fix i_sem/journal inversion in mds_client_add (2333)
5496 - fix truncate/getattr lock cycle deadlock (2334)
5497 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5498 - fix two rare exit paths which could leak an l_lock() ref (2321)
5499 - fencepost error in MDS/OST orphan recovery (2226)
5500 - make log record alignment 8 bytes (1988)
5501 - lstripe now fails when requested offset > ost_count (2237)
5502 - ensure that all kernel series have a complete list.h (1607)
5503 - fix crashes in special-file operations (2316)
5504 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5506 - allow configurable automake binary, for testing new versions
5507 - small update to the lfs documentation
5509 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5511 * fix negative export reference count in fsfilt_sync (2312)
5513 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5514 * release candidate 0.9.1
5516 - orphans are moved into the PENDING directory for possible recovery
5517 - replayed opens now open by fid for orphan/rename safety (1042)
5518 - last close of an orphan inode generates a transno (683)
5519 - chdir() and mount() now pin the directory entry (1020)
5520 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5521 - discard very old requests without processing them (1502)
5522 - remove some common, well-understood CERRORs (1505)
5523 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5524 - clear "grant" flags in OST replies until OST grant code lands (1644)
5525 - fix read performance by not clobbering i_blksize on client (1598)
5526 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5527 - make LRU size a /proc tunable, clears locks when reduced (707)
5528 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5529 - server locks take references on exports now (1558)
5530 - build fixes for 2.4.20-rh trees (1663)
5531 - return an error from lov_create if all OSCs are inactive (1751)
5532 - fix import levels when a reconnect happens without a timeout (1597)
5533 - exit early from mds_open if we get a lookup error (1749)
5534 - partial page read at EOF wouldn't wait for disk before sending (1642)
5535 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5536 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5537 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5538 - write/truncate lock inversion (1639)
5539 - Don't auto-load obdclass, portals modules during cleanup (1495)
5540 - fix timestamps from jumping to "now" (1763)
5541 - extra journal assertions (1648)
5542 - add an extra multiunlink test (1771)
5543 - fix read_record/write_record API (1776)
5544 - fix leak of offset_extent, possible incorrect i_size later (1772)
5545 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5546 - Debugging for the kqswnal_get_idle_tx problems (1820)
5547 - Allow recovery to be attempted multiple times (1536)
5548 - Write out MDS last_rcvd file after it is first created (1600)
5549 - Fix tx_descriptor leak in failed transmit situations (1827)
5550 - ext3 journaling fixes for assertion failure after IO error (1871)
5551 - class_export_put() on freed export after completion AST error (1896)
5552 - Fix revalidate looping in VFS (1322)
5553 - Don't access a freed export during MDS_REINT timeout (1521)
5554 - Add open-unlink recovery support on the MDS (1673,1764)
5555 - Return an error if no MDS data was read from last_rcvd (1946)
5556 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5557 - Don't setup a disk device that doesn't match exported UUID (317)
5558 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5559 - avoid committing NULL handle in force close
5560 - local.sh is now a one-stripe LOV configuration
5561 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5562 - fix ext3 htree duplicate directory entry corruption (1516)
5563 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5564 - update to new LOV EA format (2097)
5565 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5567 2003-06-15 Phil Schwan <phil@clusterfs.com>
5570 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5571 - per-import/export recovery handling (958, 931, 959)
5572 - multiple last-rcvd slots, for serving multiple FSes (949)
5573 - connections are again shared between multiple imp/exports (963, 964)
5574 - "umount -f" would hang if any requests needed to be sent (393, 978)
5575 - avoid pinning large req buffer by copying for queued messages (989)
5576 - add "uuid" to "lctl device" command to help upcalls (991)
5577 - "open" RPCs with transnos would confuse recovery counters (1037)
5578 - do proper endian conversion of all wire messages (288, 340, 891)
5579 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5580 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5581 - fix ptlrpc_connection leak in target_handle_connect (1174)
5582 - fix import refcounting bug in OST and MDS cleanup (1134)
5583 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5584 - fix very unlikely obd_types race condition (501)
5585 - remove osc_open hack for echo_client (1187)
5586 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5587 - a failure in read_inode2 leads to deadlock (1139)
5588 - cancel ack-locks as soon as transaction is committed (1072)
5589 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5590 - make sure to commitrw after any preprw to avoid deadlock (1162)
5591 - failing to execute a file in a lustre FS would lock inode (1203)
5592 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5593 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5594 - obd_uuid2tgt would walk off the end of the list (1255)
5595 - on IA64 the osc would give portals incorrect bulk size (1258)
5596 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5597 - fix lock inversion caused by new llite matching code (1282)
5598 - limit the number of dirty pages on a client to 10MB (1286)
5599 - timed out locks were not being corrected cancelled (1289)
5600 - fix O_DIRECT above 4GB on IA-32 (1292)
5601 * major user-visible changes
5602 - fail out/fail over policy now controlled by the upcall (993)
5604 - add OBD_PING to check server availability and failure (954)
5605 - lustre messages are now sent in sending host order (288, 340, 891)
5606 - add eadatalen to MDS getattr reply (340)
5607 - OST read replies may contain second buffer, with per-page status (593)
5609 2003-03-11 Phil Schwan <phil@clusterfs.com>
5612 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5613 - failed open()s could cause deadlock; fixed (867, 869)
5614 - stop cancelling OST locks when files are closed (481)
5615 - overlapping XID spaces caused network corruption (851, 853)
5616 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5617 - setattr_raw added, to do single-RPC, server-side setattrs
5618 - lmc/lconf syntax change for OST UUIDs
5619 - fix crashy race condition between ptlrpc_free_req and osc_close
5620 - don't use request in mdc_enqueue if we hit a timeout (889)
5621 - don't set the inode i_size for regular files from the MDS (896)
5622 - handle out of order completion AST (842)
5623 - don't LBUG if a lock request times out after receiving AST (913)
5624 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5625 - if a bad lock AST arrives, send an error instead of dropping entirely
5626 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5627 - fix leak in bulk IO when only partially completed (899, 900, 926)
5628 - fix O_DIRECT for ia64 (55)
5629 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5630 - C-z after timeout could hang a process forever; fixed (977)
5632 - client-side I/O cache (678, 924, 929, 941, 970)
5634 - READPAGE and SETATTRs which don't take server-side locks get
5637 2003-02-11 Phil Schwan <phil@clusterfs.com>
5640 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5641 - Fix fsfilt_extN_readpage() to read a full page of directory
5642 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5643 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5644 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5645 - Most intent operations were removed, in favour of a new RPC mode
5646 that does a single RPC to the server and bypasses most of the VFS
5647 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5648 - Aggressively cancel local locks on DLM servers
5649 - mds_reint_unlink sends EA to the client if it's the last nlink.
5650 client uses that EA to unlink OST objects.
5651 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5652 - recursive symlinks were fixed (439)
5653 - fixed NULL deref in DEBUG_REQ
5654 - filter_update_lastobjid no longer calls sync, which annoyed extN
5655 - fixed multi-client small-writes to a single file problem (445)
5656 - fixed mtime updates during file writes (607)
5657 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5658 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5659 - fixed Linux OST size reporting problem (444, 656)
5660 - OST now updates object mtime with writes or setattr (607, 619)
5661 - client verifies file size before zeroing page past EOF (445)
5662 - OST now writes last allocated objid to disk with allocation (108)
5663 - LOV on echo now works (409)
5665 - mds_reint_unlink sends a new buffer, with the EA included. this
5666 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5667 set if a regular file was being unlinked, and it was the last link
5668 - use PtlGet from the target for bulk writes (315)
5669 - OST now updates object mtime with writes or setattr (607, 619)
5670 - LDLM now has a grant-time callback to revalidate locked items, if
5672 - Many MDS operations were reorganized to combat race conditions
5674 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5675 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5677 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5680 - Fully reactivate OST imports after reconnection (512, others)
5681 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5682 - More graceful error handling for truncating on dead OST (515)
5683 - Don't error out unless we're actually accessing dead stripes (474)
5684 - Fix garbage sizes when stripes are missing (410)
5685 - LRU counters were broken, causing constant lock purge (433, 432)
5686 - garbage on read from stripes with failed OSTs (441)
5687 - mark OSCs as active before reconnecting during recovery (438)
5688 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5689 - lfind did not preserve OST order in output (443)
5690 - symlinks cause hung clients, incorrect data (439)
5691 - stop dereferencing request after dropping refcount (457)
5692 - don't LASSERT(spin_is_locked) on non-SMP (455)
5693 - fixes for many rename() bugs
5694 - fstat didn't correctly synchronize attributes (399)
5695 - server must handle lock cancellation during blocking AST prep (487)
5696 - bulk descriptors were free()d too soon (511)
5697 - fix paths in lconf, which would load incorrect modules (451, 507)
5698 - fix confusing lconf 'host not found' error message (386)
5699 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5700 - fix race condition in mdc_blocking_ast() for inode access (526)
5701 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5702 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5703 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5704 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5705 - fix lstripe in several ways (kernel side) (527)
5706 - fix request leak in ldlm_cli_enqueue (262)
5707 - incorrect OSC was marked inactive after OST failure
5708 - call mds_fs_cleanup before unmounting filesystem (524)
5709 - fix races between taking ns_lock and ldlm_lock_change_resource
5710 - fix races updating LOV export open file list
5711 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5712 - fix recovery NULL deref in ldlm_cli_cancel_unused
5713 - fix some DLM races by using new hash table for lock handles (419)
5714 - permit the client to specify desired inodes, at replay
5715 - duplicate requests when we queue them for replay reintegration
5716 - fix last_rcvd offset calculation
5717 - sync after each recovered transaction, so we always make progress
5718 - never, not always, ERESTART requests without transnos
5719 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5721 - skip replay if the MDS says that the client is already connected
5722 - don't check for a recovery-enabled export to match lctl's UUID
5723 - don't INC_USE_COUNT for phantom exports
5724 - don't crash when cleaning up phantom exports (567)
5725 - don't double-finish or set replay data for errored mdc_open requests
5726 - abort requests when they time out, so we don't get old replies
5727 - send/receive replies for AST messages again
5728 - if the client says that it doesn't have the lock, cancel it on the
5730 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5731 mark it as destroyed, it will all work out when decref is called
5732 - fix module use counts (22, 581)
5734 - ASTs now expect a reply (server cancels lock on error reply)
5736 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5739 - fix many simultaneous client startup (392)
5740 - fix dentry->d_it clobbering
5741 - credentials weren't being shipped for readdir/getattr operations
5742 - remove invalid assertions triggered during some concurrent MD
5744 - proper Lustre versions added (336, 389)
5745 - fix memory leak for create error case (398)
5746 - fix LOV locking bug that would get cli/srv out of sync
5747 - fix echo client over LOV (409)
5748 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5749 - fix double-O_EXCL intent crash (424)
5750 - avoid sending multiple lock CANCELs (352)
5752 - MDS can do multi-client recovery (modulo bugs in new code)
5754 - many updates, edits, cleanups
5756 2002-11-18 Phil Schwan <phil@clusterfs.com>
5759 - fix null d_it dereference (346)
5760 - fix full OST/dbench hang (333)
5761 - fix permission problem with file removal (286)
5762 - fix removal of OSCs from LOV when they fail
5763 - fix NULL deref during bulk timeout (214)
5764 - fix problems related to multiple filesystems on one MDS (241)
5765 - fixed serious subtle metadata locking bugs
5766 - free locks on clients when inodes are removed due to memory
5768 - fix inode pointer in lock data (285)
5769 - partial support for multiple MDS on a single host (241)
5770 - data locks weren't cancelled at clear_inode time (290, 311)
5771 - intent locks could lead to unbounded lock growth (205)
5772 - added a maximum lock count, an LRU list, and a flusher
5773 - fix multiple rename (365)
5774 - properly abstracted the echo client
5775 - OSC locked 1 byte too many; fixed
5776 - rewrote brw callback code:
5777 - fixed recovery bugs related to LOVs (306)
5778 - fixed too-many-pages-in-one-write crash (191)
5779 - fixed (again) crash in sync_io_timeout (214)
5780 - probably fixed callback-related race (385)
5782 - Add capability to MDS protocol
5783 - LDLM cancellations and callbacks on different portals
5785 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5788 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5789 - defer open object destruction to close time (601981, 138)
5790 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5791 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5792 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5793 - don't call obd_setattr in ll_file_release for destroyed objects
5794 * protocol change to lustre_msg: move |version| and add |flags|
5795 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5796 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5797 * added replay of create, unlink, link and rename operations during
5798 MDS failover; recovery should be much more robust now
5799 * remove failed OSCs from LOVs (only lov_create uses this so far)
5800 * the lustre-HOWTO was brought (more) up to date (582544)
5802 2002-10-23 Phil Schwan <phil@clusterfs.com>
5805 - in-use dentries weren't being reused properly (617851)
5806 - prevent multiple LDLM setup (599178)
5807 - fix LOV size calculations for truncate (617853)
5808 - fix client handling of MDS intent errors (POSIX)
5809 - fix permission bug in lovstripe.c test (624321)
5810 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5811 - truncate past end of file could corrupt data
5812 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5813 - a race in recovery could return ETIMEDOUT to apps (623947)
5814 - building outside the source directory was fixed
5815 * the lustre-HOWTO was brought (more) up to date (582544)
5816 * major progress was made on recovery functionality
5818 2002-10-10 Phil Schwan <phil@clusterfs.com>
5821 - recovery deadlock fix
5822 - rm -rf causes LBUG fix (617817)
5823 - file open by multiple tasks fix (618962)
5824 - directory permissions bugs (602707 and 620007)
5825 - journal_stop fixed with locking (611313)
5826 - O_APPEND failures resolved (618273, perhaps 614459)
5827 - lconf PATH fix (619770)
5828 - IA64 build fix (621450)
5829 - RPC buffer sizes scale with amount of memory
5831 2002-10-01 Phil Schwan <phil@clusterfs.com>
5834 - locks would be cancelled without throwing away data pages,
5835 resulting in inconsistent data (605627)
5836 - inode attributes were not always being refreshed (605627, 612449)
5837 - lconf now continues to cleanup after lctl reports an error
5838 - MDS now enforces user permissions (602707)
5839 - lprocfs cleanup fixed, but not yet enabled (614157)
5840 - fixed infinite server hang, should a client not respond to an AST
5841 - avoid going into recovery if user calls readlink() with a buffer
5842 that's too small (613941)
5843 - AST RPCs no longer require replies (614867) -- this may be changed
5844 - don't crash server if client sends an IOV that's too big (611336)
5845 - fixed lock conversion deadlock (611892)
5846 - fixed the following of symlinks (614622)
5847 * recovery: the server can remove locks from a client that dies, other
5848 clients can make progress
5849 * more extN patch fixes
5850 * compile-time configurable ptlrpc buffer allocations
5852 - collaborative read cache document
5853 - Lustre Lite Performance CDR document-in-progress
5855 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5858 - fix typo in patch-2.4.18
5860 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5863 - clear ptlrpc request each time in handle_incoming_request()
5864 - unlink of files now destroys the object on the OST
5866 2002-09-19 Peter Braam <braam@clusterfs.com>
5868 * add hard link support
5869 * change obdfile creation method
5870 * kernel patch changed
5872 2002-09-19 Peter Braam <braam@clusterfs.com>
5875 - stack overflow bug in extN fixed
5877 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5879 * documentation updates
5880 - add man pages for config tools
5881 - update tests/README to describe testing with new config tools
5882 - finish metadata API descriptions
5883 * bug fixes and cleanups
5884 - statfs workaround for 16TB limit
5885 - LOV stripe allocation improved, can stripe on subset of OSTs
5886 - LOV file size/IO offset was wrong for files > 4GB in size
5887 - object EA data was being dropped, caused files to be unreadable
5888 - memory overflow with non-LOV OST caused memory corruption
5889 - fixed regression tests to work with new config tools, obdfilter
5890 - fixed bug when directory size became larger than 1 block
5891 - fixed bug (for single client case) when PWD was deleted
5892 - invalidate local directory pages when doing intent-based ops
5893 - avoid LDLM oops when lock callback contained bad data
5895 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5897 * documentation updates
5898 * bug fixes and cleanups
5899 - configuration tools
5902 - 64-bit compile warnings
5903 - 64-bit internal statfs data
5905 * test_brw on persistent OST devices
5907 * lprocfs (disabled)
5909 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5911 * documentation updates
5912 * bug fixes and cleanups
5913 * configuration tools
5915 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5918 * many small fixes to 0_5_4
5919 * io/network handling
5920 * thinkos in MDS operations
5922 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5925 * crucial basic fixes to 0.5.3
5926 * IOR, Iozone work over Elan
5929 2002-08-07 Phil Schwan <phil@clusterfs.com>
5930 * version 0_5_3, our first alpha
5931 * we use the new Portals iovs
5932 * documentation updates
5933 * bug fixes and cleanups
5934 * small changes in the DLM wire protocol
5936 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5937 * version 0_5_1 with some initial stability,
5938 * locking on MD and file I/O.
5939 * documentation updates
5940 * several bug fixes since 0.5.0
5941 * small changes in wire protocol
5943 2002-07-18 Phil Schwan <phil@clusterfs.com>
5945 * delivered as Lustre Light Alpha
5946 * fixed a crash after handling invalid MDS requests
5947 * fixed directory pages for architectures with non-4k pages sizes
5949 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5950 * release version v0_4_4
5951 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5953 * Fixes a number of interruption problems with OST operations.
5954 * Update documentation for portals header changes
5955 * Move all wire protocol structs/defines to lustre_idl.h
5956 * Fixes symlink length bug.
5957 * Add tcpdump to repository.
5959 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5960 * release version v0_4_3
5961 * Fixes statfs for inodes on extN.
5962 * Fixes bug in runtests which would delete /etc/hosts.
5963 * Use 64-bit object IDs wherever possible (not into VFS though)
5964 Remove ost_get_info, which is unused by lustre, and out of date.
5966 2002-07-03 Peter Braam <braam@clusterfs.com>
5967 * release version v0_4_2 Fixes a lookup error (type not passed)
5968 * move forward to head of Portals
5969 * move forward to latest Lustre kernel
5971 2002-06-25 Peter Braam <braam@clusterfs.com>
5972 * release version v0_4_1. Hopefully stable on single node use.