1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.37 (SLES 10),
5 2.6.18-128.1.6.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.41.6.sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
14 * File join has been disabled in this release, refer to Bugzilla 16929.
18 Description: Unable to run several mkfs.lustre on loop devices at the same
20 Details : mkfs.lustre returns error 256 on the concurrent loop devices
21 formatting. The solution is to proper handle the error.
23 Severity : enhancement
25 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
27 Severity : enhancement
29 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
33 Description: resolve race between obd_disconnect and class_disconnect_exports
34 Details : if obd_disconnect will be called to already disconnected export he
35 forget release one reference and osc module can't unloaded.
37 Severity : enhancement
39 Description: Allow tuning service thread via /proc
40 Details : For each service a new
41 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
42 created that can be used to set min/max thread counts, and get the
43 current number of running threads.
47 Descriptoin: don't return error if have particaly created objects for file.
48 Details : lov_update_create_set uses set->set_success as index for created objects,
49 so if some requests will be failed, they will have hole at end of
50 array and we can use qos_shrink_lsm for allocate correct lsm.
52 Severity : enhancement
54 Description: Update OFED support to 1.4.1
58 Description: Reduce small size read RPC
59 Details : Set read-ahead limite for every file and only do read-ahead when
60 available read-ahead pages are bigger than 1M to avoid small size
63 Severity : enhancement
65 Description: move AT tunable parameters for more consistent usage
66 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
68 Severity : enhancement
70 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
71 Details : allow skip disconnected ost for send statfs request and hide error
77 Description: fix racy locking of mballoc block bitmaps causing BUG
78 Details : The locking of the mballoc buddy bitmap and the in-memory
79 block bitmap was using two different spin locks in some
80 cases. This made it possible to incorrectly access the
81 mballoc bitmap while another process was modifying it,
82 causing a sanity assertion to fail. While no on-disk corruption
83 was reported, there was some risk of this happening.
86 Frequency : rare, on llog test 6
88 Descriptoin: don't allow connect to already connected import
89 Details : allowing connect to already connected import is hide connecting problem.
94 Descriptoin: allow kill process which wait statahead result
95 Details : for some reasons 'ls' can stick in waiting result from statahead,
96 in this case need way for kill this process.
98 Severity : enhancement
100 Description: Add state history info file, enhance import info file
101 Details : Track import connection state changes in a new osc/mdc proc file;
102 add overview-type data to the osc/mdc import proc file.
104 Severity : enhancement
106 Description: MDS create should not wait for statfs RPC while holding DLM lock.
108 Severity : enhancement
110 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
115 Descriptoin: OST grant too much space to client even there are not enough space.
116 Details : Client will shrink its grant cache to OST if there are no write
117 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
118 this grant cache if there are already not enough avaible space
119 (left_space < total_clients * 32M).
122 Frequency : start MDS on uncleanly shutdowned MDS device
124 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
125 Details : stay in waiting mds<>ost recovery finished produce random bugs
126 due race between two ll_sync thread for one lov target. send
127 ACTIVATE event only if connect realy finished and import have
131 Frequency : rare, connect and disconnect target at same time
133 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
134 Details : don't call obd_disconnect under lov_lock. this long time
135 operation and can block ptlrpcd which answer to connect request.
140 Descriptoin: don't lose wakeup for imp_recovery_waitq
141 Details : recover_import_no_retry or invalidate_import and import_close can
142 both sleep on imp_recovery_waitq, but we was send only one wakeup
146 Frequency : always with long access acl
148 Descriptoin: mds can't pack reply with long acl.
149 Details : mds don't control size of acl but they limited by reint/getattr
152 Severity : enhancement
154 Description: Update to SLES10 kernel-2.6.16.60-0.33.
156 Severity : enhancement
158 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
161 Frequency : start MDS on uncleanly shutdowned MDS device
163 Descriptoin: aborting recovery hang on MDS
164 Details : don't throttle destroy RPCs for the MDT.
167 Frequency : on remount
169 Description: external journal device not working after the remount
170 Details : clear dev_rdonly flag for external journal devices in
176 Description: shutdown vs evict race
177 Details : client_disconnect_export vs connect request race.
178 if client will evicted at this time - we start invalidate
179 thread without referece to import and import can be freed
183 Frequency : rare, need acl's on inode.
185 Description: client can't handle ost additional correctly
186 Details : if ost was added after client connected to mds client can have
187 hit lnet_try_match_md ... to big messages to wide striped files.
188 in this case need teach client to handle config events about add
189 lov target and update client max ea size at that event.
191 Severity : enhancement
193 Description: Changelogs
194 Details : Changelogs are a lightweight mechanism to track filesystem
195 metadata and namespace changes. The changelog is recorded
196 permanently on the MDTs, and is periodically "consumed" / purged
197 when records are no longer needed.
199 Severity : enhancement
201 Description: compact fld format with extents
202 Details : Store range of seq rather than every seq in FLD. Seq
203 controller update FLD rather than clients. In Case of CMD, mdt0
204 has FLD, all other metadata server act as non persistent proxy
205 for FLD queries and cache fld entries in fld cache.
210 Description: don't skip ost target if they assigned to file
211 Details : Drop slow OSCs if we can, but not for requested start idx.
212 This means "if OSC is slow and it is not the requested
213 start OST, then it can be skipped, otherwise skip it only
214 if it is inactive/recovering/out-of-space.
218 Description: more cleanup in mds_lov
219 Details : not send LOV EA under replay, we can't know about they size at this
220 time. Don't allow client connect to mds before any ost connected,
221 for avoid problems with LOV EA size and returning EIO to client.
223 Severity : enhancement
225 Description: Interoperability at server side (Disk interoperability)
227 Severity : enhancement
229 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
231 Severity : enhancement
233 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
235 Severity : enhancement
237 Description: New client IO stack (CLIO).
239 Severity : enhancement
241 Description: Commit on sharing. Eliminate inter-client dependencies between
242 uncommitted transactions by doing transaction commits.
243 Thereby clients may recovery independently.
246 Frequency : Create a symlink file with a very long name
248 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
249 Details : If there is no extra space in the request for early cancels,
250 ldlm_req_handles_avail() returns 0 instead of a negative value.
252 Severity : enhancement
254 Description: Add /proc entry for import status
255 Details : The mdc, osc, and mgc import directories now have
256 an import directory that contains useful import data for debugging
259 Severity : enhancement
261 Description: Re-disable certain /proc logging
262 Details : Enable and disable client's offset_stats, extents_stats and
263 extents_stats_per_process stats logging on the fly.
266 Frequency : Only on FC kernels 2.6.22+
268 Description: oops in statahead
269 Details : Do not drop reference count for the dentry from VFS when lookup,
270 VFS will do that by itself.
272 Severity : enhancement
274 Description: Generic /proc file permissions
275 Details : Set /Proc file permissions in a more generic way to enable non-
276 root users operate on some /proc files.
280 Description: Hitting mdc_commit_close() ASSERTION
281 Details : Properly handle request reference release in
282 ll_release_openhandle().
286 Description: quota recovery deadlock during mds failover
287 Details : This patch includes att18982, att18236, att18237 in bz14840.
289 1. fix osts hang when mds does failover with quotaon
290 2. prevent watchdog storm when osts threads wait for the
295 Frequency : only patchless client
296 Description: add workaround for race between add/remove dentry from hash
298 Severity : enhancement
300 Description: Allow OST glimpses to return PW locks
304 Description: LBUG when llog conf file is full
305 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
309 Description: Prevent import from entering FULL state when server in recovery
313 Description: service mount cannot take device name with ":"
314 Details : Only when device name contains ":/" will mount treat it as
320 Description: replace ptlrpcd with the statahead thread to interpret the async
321 statahead RPC callback
325 Frequency : on recovery
326 Description: I/O failures after umount during fail back
327 Details : if client reconnected to restarted server we need join to recovery
328 instead of find server handler is changed and process self
329 eviction with cancel all locks.
331 Severity : enhancement
333 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
335 Severity : enhancement
337 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
339 Severity : enhancement
341 Description: Upcall on Lustre log has been dumped
342 Details : Allow for a user mode script to be called once a Lustre log has
343 been dumped. It passes the filename of the dumped log to the
344 script, the location of the script can be specified via
345 /proc/sys/lnet/debug_log_upcall.
350 Description: avoid idr_remove called for id which is not allocated.
351 Details : Move assigment s_dev for clustered nfs to end of initialization,
352 to avoid problem with error handling.
357 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
358 Details : When connection is reused this not moved from CONN_UNUSED_HASH
359 into CONN_USED_HASH and this prodice warning when put connection
360 again in unused hash.
363 Severity : enhancement
365 Description: File striping can now be set to use an arbitrary pool of OSTs.
367 Severity : enhancement
369 Description: Export bytes_read/bytes_write count on OSC/OST.
373 Description: Early reply size mismatch, MGC loses connection
374 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
375 the connect flags are properly negotiated.
380 Description: quotas are not honored with O_DIRECT
381 Details : all writes with the flag O_DIRECT will use grants which leads to
382 this problem. Now using OBD_BRW_SYNC to guard this.
386 Description: add quota statistics
387 Details : 1. sort out quota proc entries and proc code.
388 2. add quota statistics
390 Severity : enhancement
392 Description: enable quota support for HEAD.
396 Description: Properly propagate oinfo flags from lov to osc for statfs
397 Details : restore missing copy oi_flags to lov requests.
399 Severity : enhancement
401 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
402 : l_getgroups(8), lst(8), routerstat(8)
403 Details : included man pages for llobdstat(8), llstat(8),
404 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
406 Severity : enhancement
408 Description: configure's --enable-quota should check the
409 : kernel .config for CONFIG_QUOTA
410 Details : configure is terminated if --enable-quota is passed but
411 : no quota support is in kernel
415 Description: Remove portals compatibility
416 Details : Remove portals compatibility, not interoperable with releases
421 Description: Resolve device initialization race
422 Details : Prevent proc handler from accessing devices added to the
423 obd_devs array but yet be intialized.
425 Severity : enhancement
427 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
429 Severity : enhancement
431 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
436 Description: Using wrong pointer in osc_brw_prep_request
437 Details : Access to array[-1] can produce panic if kernel compiled with
438 CONFIG_PAGE_ALLOC enabled
442 Description: Client runs out of low memory
443 Details : Consider only lowmem when counting initial number of llap pages
447 Description: *optional* service tags registration
448 Details : if the "service tags" package is installed on a Lustre node
449 When the filesystem is mounted, a local-node service tag will
450 be created. See http://inventory.sun.com/ for more information
451 about the Service Tags asset management system.
455 Description: Kernel BUG tries to release flock
456 Details : Lustre does not destroy flock lock before last reference goes
457 away. So always drop flock locks when client is evicted and
458 perform unlock regardless of successfulness of speaking to MDS.
462 Description: add refcount protection for osc callbacks, avoid panic on shutdown
466 Description: sanity test 65a fails if stripecount of -1 is set
467 Details : handle -1 striping on filesystem in ll_dirstripe_verify
472 Description: ASSERTION(CheckWriteback(page,cmd)) failed
473 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
476 Severity : enhancement
478 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
482 Description: do not process already freed flock
483 Details : flock can possibly be freed by another thread before it reaches
484 to ldlm_flock_completion_ast.
488 Description: LBUG during stress test
489 Details : Need properly lock accesses the flock deadlock detection list.
493 Description: oops in page fault handler
494 Details : kernel page fault handler can return two special 'pages' in error
495 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
499 Description: timeout with invalidate import.
500 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
501 should be handled by ptlrpcd. This produce long age waiting and
502 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
504 Severity : enhancement
506 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
510 Description: enable MGS and MDT services start separately
511 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
512 a co-located MGS without starting the MGS, which is a complement
513 to 'nosvc' mount option.
517 Frequency : after recovery
518 Description: precreate to many object's after del orphan.
519 Details : del orphan st in oscc last_id == next_id and this triger growing
520 count of precreated objects. Set flag LOW to skip increase count
521 of precreated objects.
525 Frequency : rare, on clear nid stats
526 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
527 Details : when clean nid stats sometimes try destroy live entry,
528 and this produce panic in free.
532 Description: Stack overflow during MDS log replay
533 ease stack pressure by using a thread dealing llog_process.
537 Description: wait until IO finished before start new when do lock cancel.
538 Details : VM protocol want old IO finished before start new, in this case
539 need wait until PG_writeback is cleared until check dirty flag and
540 call writepages in lock cancel callback.
542 Severity : enhancement
544 Description: using special macro for print time and cleanup in includes.
548 Description: mds_mfd_close() ASSERTION(rc == 0)
549 Details : In mds_mfd_close(), we need protect inode's writecount change
550 within its orphan write semaphore to prevent possible races.
554 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
558 Frequency : rare, on shutdown ost
559 Description: don't hit live lock with umount ost.
560 Details : shrink_dcache_parent can be in long loop with destroy dentries,
561 use shrink_dcache_sb instead.
565 Description: don't panic with use echo client
566 Details : echo client pass NULL as client nid pointer and this produce null
571 Description: fix build on ppc32
572 Details : compile code with -m64 flag produce wrong object file for ppc32.
576 Description: add message levels for liblustreapi
580 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
581 Details : Move the 'good_osts' check before the 'total_bavail' check. This
582 will result in an -EAGAIN and in the exit call path we call
583 alloc_rr() which will with increasing aggressiveness attempt to
584 aquire precreated objects on the minimum number of required OSCs.
588 Description: Use old size assignment to avoid deadlock
589 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
590 the scheduling while holding a spinlock. We do not need locking
591 for size in ll_update_inode() because size is only updated from
592 the MDS for directories or files without objects, so there is no
593 other place to do the update, and concurrent access to such inodes
594 are protected by the inode lock.
598 Description: resolve "_IOWR redefined" build error on SLES10
602 Description: dump the memory debugging after all modules are unloaded to
603 suppress false negative in conf_sanity test 39
605 Severity : enhancement
607 Description: build kernel-ib packages for OFED 1.3 in our release cycle
612 Description: fix SLES kernel versioning
613 Details : the kernel version for our SLES 10 kernel did not include a "-"
614 before the "smp" at the end. while this was not a problem in
615 general, it did mean that software trying to use the kernel
616 version to try to detect a vendor specific kernel would fail.
617 this was most evident by the OFED build scripts.
621 Description: Don't update lov_desc members until making sure they are valid
622 Details : When updating lov_desc members via proc fs, need fix their
623 validities before doing the real update.
627 Description: don't put request into delay list while invalidate in flight.
628 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
629 import in flight. this produce timeout for invalidate and sometimes
630 can cause stale data.
634 Frequency : on ppc only
635 Description: not convert ost objects for directory because it's not exist.
636 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
640 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
642 Severity : enhancement
644 Description: Update to support 2.6.22.14 vanilla kernel.
648 Frequency : rare, on recovery
649 Description: read procfs can produce deadlock in some situation
650 Details : Holding lprocfs lock which send rpc can produce block for destroy
651 obd objects and this also block reconnect with -EALREADY. This
652 isn't fix all lprocfs bugs - but make it rare.
654 Severity : enhancement
656 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
659 Frequency : frequent on X2 node
661 Description: mdc_set_open_replay_data LBUG
662 Details : Set replay data for requests that are eligible for replay.
666 Description: lustre_mgs: operation 101 on unconnected MGS
667 Details : When MGC is disconnected from MGS long enough, MGS will evict the
668 MGC, and late on MGC cannot successfully connect to MGS and a lot
669 of the error messages complaining that MGS is not connected.
673 Frequency : on network error
674 Description: panic with double free request if network error
675 Details : mdc_finish_enqueue is finish request if any network error ocuring,
676 but it's true only for synchronus enqueue, for async enqueue
677 (via ptlrpcd) this incorrect and ptlrpcd want finish request
680 Severity : enhancement
682 Description: client-side metadata stat-ahead during readdir(directory readahead)
683 Details : perform client-side metadata stat-ahead when the client detects
684 readdir and sequential stat of dir entries therein
687 Frequency : on start mds
689 Description: Implement get_info(last_id) in obdfilter.
692 Frequency : occasional
694 Description: Correctly check stale fid, not start epoch if ost not support SOM
695 Details : open with flag O_CREATE need set old fid in op_fid3 because
696 op_fid2 was overwritten with new generated fid, but mds can answer
697 with one of these two fids and both is not stale. Setattr
698 incorrectly started an epoch and assume will be called
699 done_writing, but without SOM done_writing ever being called.
702 Frequency : rare, depends on device drivers and load
704 Description: MDS or OSS nodes crash due to stack overflow
705 Details : Code changes in 1.8.0 increased the stack usage of some functions.
706 In some cases, in conjunction with device drivers that use a lot
707 of stack the MDS (or possibly OSS) service threads could overflow
708 the stack. One change which was identified to consume additional
709 stack has been reworked to avoid the extra stack usage.
712 Frequency : occasional
714 Description: Do not fail import if osc_interpret_create gets -EAGAIN
715 Details : If osc_interpret_create got -EAGAIN it immediately exits and
716 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
717 oscc_has_objects and see OSC has no objests and call
718 oscc_internal_create to resend create request.
720 Severity : enhancement
722 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
724 Severity : enhancement
726 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
729 Frequency : very rare
731 Description: Fix replay if there is an un-replied request and open
732 Details : In some cases, older replay request will revert the
733 mcd->mcd_last_xid on MDS which is used to record the client's
736 Severity : enhancement
738 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
740 Severity : enhancement
742 Description: Add rhel5 support to HEAD.
744 Serverity : enhancement
746 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
751 Description: Don't allow skipping OSTs if index has been specified.
752 Details : Don't allow skipping OSTs if index has been specified, make
753 locking in internal create lots better.
757 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
758 Details : access to bitfield in structure is always rounded to long
759 and this produce problem with not atomic change any bit.
763 Description: Lustre make rpms failed.
764 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
765 builds Lustre rpms from tarball.
768 Frequency : rare on shutdown ost
770 Description: If llog cancel was not send before clean_exports phase, this can
771 produce deadlock in llog code.
772 Details : If llog thread has last reference to obd and call class_import_put
773 this produce deadlock because llog_cleanup_commit_master wait when
774 last llog_commit_thread exited, but this never success because was
775 called from llog_commit_thread.
779 Description: allow userland application know is lost one of stripes.
780 Details : fill lvb_blocks with error code on ost and return it to
781 application if error flag found.
785 Description: NULL lov_tgts causing MDS oops
786 Details : more safe checks for NULL lov_tgts for avoid oops.
788 Severity : enhancement
790 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
794 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
795 Details : If an OST is down the MDS will hang indefinitely in
796 obd_statfs_rqset() waiting for the statfs data. While for
797 MDS QOS usage of statfs, it should not stuck in waiting.
799 Severity : enhancement
801 Description: remote_acl support
802 Details : Support ACL-based permission check for remote user.
803 Support setfacl/getfacl for remote user with the utils
804 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
805 the system "{s,g}etfacl" utils.
807 Severity : enhancement
809 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
811 Severity : enhancement
813 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
817 Description: make mgs_setparam() handle fsname containing dash
818 Details : fsname containing a dash does not work with lctl conf_param
820 Severity : enhancement
822 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
824 Severity : enhancement
826 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
830 Frequency : rare, at shutdown
831 Description: access already free / zero obd_namespace.
832 Details : if client_disconnect_export was called without force flag set,
833 and exist connect request in flight, this can produce access to
834 NULL pointer (or already free pointer) when connect_interpret
835 store ocd flags in obd_namespace.
839 Frequency : only at startup
840 Description: not alloc memory with spinlock held.
841 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
842 if any spinlock held.
844 Severity : enhancement
846 Description: make lustre randomly fail allocating memory
847 Details : Make lustre randomly failed allocating memory for testing purpose.
849 Severity : enhancement
851 Description: lost problems with lov objid file
852 Details : Fixes some scability and access to not inited memory problems
853 in work with lov objdid file.
858 Description: lfs find does not continue on file error
859 Details : Continue other files processing when a file/dir is absent.
863 Description: Inconsistent usage of lustre_pack_reply()
864 Details : Standardize the usage of lustre_pack_reply() such that it
865 always generate a CERROR on failure.
868 Frequency : occasional
870 Description: MDS hang or stay in waiting lock
871 Details : If client receive lock with CBPENDING flag ldlm need send lock
872 cancel as separate rpc, to avoid situation when cancel request
873 can't processed due all i/o threads stay in wait lock.
877 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
881 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
885 Description: Update to SLES9 kernel-2.6.5-7.287.3.
889 Description: Update to RHEL4 latest kernel.
891 Severity : enhancement
893 Description: Build SLES10 patchless client fails
894 Details : The configure was broken by run ./configure with
895 --with-linux-obj=.... argument for patchless client. When the
896 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
897 can't search header adequately. Use absolute path such as
898 -I($LINUX)/include instead.
902 Description: interrupt oig_wait produce painc on resend.
903 Details : brw_redo_request can be used for resend requests from ptlrpcd and
904 private set, and this produce situation when rq_ptlrpcd_data not
905 copyed to new allocated request and triggered LBUG on assert
906 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
907 ptlrpcd set if request is changed and can be safety changed to use
912 Description: LASSERT_{REQ,REP}SWAB macros are buggy
913 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
914 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
915 these macros and replace them with their difinitions instead.
919 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
920 Details : Update which_patch & target file for SLES10 latest kernel.
923 Description: Remove client patches from SLES 10 kernel.
924 Details : This causes SLES 10 clients to behave as patchless clients
925 even on a Lustre-patched (server) kernel.
927 Severity : enhancement
929 Description: self-adjustable client's lru lists
930 Details : use adaptive algorithm for managing client cached locks lru
931 lists according to current server load, other client's work
932 pattern, memory activities, etc. Both, server and client
933 side namespaces provide number of proc tunables for controlling
936 Severity : enhancement
938 Description: light-weight GSS support
939 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
940 and reduce performance overhead.
942 Severity : enhancement
944 Description: Linux keyring support
945 Details : Support using service of Linux keyring for Lustre GSS internal
946 context refresh/cache mechanism.
950 Description: Fix errors in lfs documentation
951 Details : Fixes man pages
955 Description: don't use GFP_* in generic Lustre code.
956 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
960 Description: obdclass is limited by single OBD_ALLOC(idarray)
961 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
965 Description: warnings with build patchless client with vanila 2.6.19 and up
966 Details : change old ctl_table style and replace ctl_table/ctl_table_header
967 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
971 Frequency : Only for SLES
972 Description: sanity_quota fail test_1
973 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
974 file, which makes TSTID[2] inunique.
978 Frequency : Only for SLES9
979 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
980 Details : Update target/ChangeLog/which_patch .
984 Frequency : Only for test_10 in sanity-quota.sh
985 Description: a bug in quota test code
986 Details : checking a test flag in a wrong place
990 Frequency : Only for test_13 in sanity-quota.sh
991 Description: a bug in quota test script
992 Details : shouldn't check overfull allocated quota
996 Description: jbd statistics
997 Details : Port older jbd statistics patch for sles10
999 Severity : enhancement
1001 Description: Add printing inode info into message about error in writepage.
1006 Description: reply_lock_interpret crash due to race with it and lock cancel.
1007 Details : Do not replay locks that are being cancelled. Do not reference
1008 locks by their address during replay, just by their handle.
1012 Frequency : When flocks are used.
1013 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1014 Details : Flock locks might destroy just granted lock if it could be merged
1015 with another existing flock, this is done in completion handler,
1016 so teach ldlm_cli_enquque_fini that this is a valid case for
1022 Description: Oops in read and write path when failing to allocate lock.
1023 Details : Check if lock allocation failed and return error back.
1027 Description: lstripe command fails for valid OST index
1028 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1029 instead of lov->desc.ld_active_tgt_count.
1033 Description: sanity.sh failed test 103
1034 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1039 Description: df doesn't work properly if diskfs blocksize != 4K
1040 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1044 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1045 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1047 Severity : enhancement
1049 Description: Adaptive timeouts
1050 Details : RPC timeouts adapt to changing server load and network
1051 conditions to reduce resend attempts and improve recovery time.
1055 Description: llapi_file_create() does not allow some changes
1056 Details : add llapi_file_open() that allows specifying the mode and
1057 open flags, and also returns an open file handle.
1061 Description: merge and cleanup kernel patches.
1062 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1066 Description: Add journal checksum support.(Kernel part)
1067 Details : The journal checksum feature adds two new flags i.e
1068 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1069 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1070 indicates that the commit block contains the checksum for
1071 the blocks described by the descriptor blocks. Now commit
1072 record can be sent to disk without waiting for descriptor
1073 blocks to be written to disk. This behavior is controlled
1074 using JBD2_FEATURE_ASYNC_COMMIT flag.
1078 Description: OSS needs mutliple precreate threads
1079 Details : Add ability to start more than one create thread per OSS.
1083 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1084 Details : execute lfs setstripe on client
1088 Description: mds_obd_create error creating tmp object
1089 Details : When the user sets quota on root, llog will be affected and can't
1090 create files and write files.
1093 Frequency : Always on ia64 patchless client, and possibly others.
1095 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1096 Details : This allows the patchless client to be loaded on architectures
1097 without this export.
1101 Description: RedHat Update kernel for RHEL5
1102 Details : Add the kernel config file for RHEL5.
1106 Description: RedHat Update kernel for RHEL5
1107 Details : Modify the kernel config file more closer RHEL5.
1111 Description: Build failure against Centos5 (RHEL5)
1112 Details : Define PAGE_SIZE when it isn't present.
1116 Description: test_7 on sanity-quota.sh can't be used on separate machines
1117 Details : add facet to handle it
1121 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1122 Details : with NFS, the anon dentry's parent was set to itself in
1123 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1124 corresponding dentry other than use rec->ur_name.
1126 Severity : enhancement
1128 Description: lfs setstripe enhancement
1129 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1133 Description: enable data checksumming by default
1134 Details : enable checksum by default, allow --disable-checksum
1135 configure option and "-o nochecksum" mount option. Checksums
1136 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1140 Description: lustre support for RHEL5
1141 Details : Add support for RHEL5.
1145 Description: Client eviction due to failover config
1146 Details : after a connection loss, the lustre client should attempt to
1147 reconnect to the last active server first before trying the
1148 other potential connections.
1152 Description: when mds and osts use different quota unit(32bit and 64bit),
1153 quota will be released repeatly.
1154 Details : void sending multiple quota reqs to mds, which will keep the
1155 status between the reqs.
1159 Description: osts not allocated evenly to files
1160 Details : change the condition to increase offset_idx
1163 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1165 Description: Data corruption for OSTs that are formatted larger than 2TB
1167 Details : When generating the bio request for lustre file writes the
1168 sector number would overflow a temporary variable before being
1169 used for the IO. The data reads correctly from Lustre (which
1170 will overflow in a similar manner) but other file data or
1171 filesystem metadata may be corrupted in some cases.
1175 Description: Tune the kernel for good SCSI performance.
1176 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1177 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1182 Description: rewrite ext2-derived code in obdclass/uuid.c
1183 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1184 from scratch preserving functionality.
1188 Description: "lfs find -obd UUID" prints directories
1189 Details : "lfs find -obd UUID" will return all directory names instead
1190 of just file names. It is incorrect because the directories
1191 do not reside on the OSTs.
1195 Description: use i_size_read and i_size_write in 2.6 port
1196 Details : replace inode->i_size access with i_size_read/write()
1199 Frequency : when removing large files
1201 Description: scheduling issue during removal of large Lustre files
1202 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1203 It causes scheduling issues when removing large files (17TB in the
1207 Frequency : only with liblustre clients on XT3
1209 Description: evictions taking too long
1210 Details : allow llrd to evict clients directly on OSTs
1213 Frequency : only on ppc
1215 Description: /proc/fs/lustre/devices broken on ppc
1216 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1217 We should make sure the type of 'index' is loff_t and then cast
1218 explicitly as needed below. Do not assign an explicitly cast
1222 Frequency : only for rhel5
1224 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1225 Details : Modify the target file & which_kernel.
1227 Severity : enhancement
1229 Description: omit set fsid for export NFS
1230 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1231 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1235 Frequency : Always, for kernels after 2.6.16
1236 Description: Fix warning idr_remove called for id=.. which is not allocated.
1237 Details : Last kernels save old s_dev before kill super and not allow
1238 to restore from callback - restore it before call kill_anon_super.
1242 Description: buffer overruns could theoretically occur
1243 Details : llapi_semantic_traverse() modifies the "path" argument by
1244 appending values to the end of the origin string, and a
1245 overrun may occur. Adding buffer overrun check in liblustreapi.
1249 Description: Fix error on 'ls .' at the top of the Lustre mount.
1250 Details : Don't revalidate dentry if it is a root dentry.
1254 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1255 Details : Modify vm-tunables-rhel4.patch.
1259 Description: Kernel config for 2.6.18-vanilla.
1260 Details : Modify targets/2.6-vanilla.target.in.
1261 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1262 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1263 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1264 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1268 Description: improve handling recoverable errors
1269 Details : if request processig with error which can be recoverable on server
1270 request should be resend, otherwise page released from cache and
1275 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1276 Details : Modify target file & which_patch.
1277 A flaw was found in the IA32 system call emulation provided
1278 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1279 value could be stored in the %RAX register, which could trigger an
1280 out-of-bounds system call table access. An untrusted local user
1281 could exploit this flaw to run code in the kernel
1282 (ie a root privilege escalation). (CVE-2007-4573).
1286 Description: O_DIRECT bypasses client statistics.
1287 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1288 read_ahead_stats were updated. copied the stats section from
1289 osc_send_oap_rpc() into async_internal().
1293 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1297 Description: change order of libsysio includes
1298 Details : '#include sysio.h' should always come before '#include xtio.h'
1302 Description: handle "serious error: objid * already exists" more gracefully
1303 Details : If LAST_ID value on disk is smaller than the objects existing in
1304 the O/0/d* directories, it indicates disk corruption and causes an
1305 LBUG(). If the object is 0-length, then we should use the existing
1306 object. This will help to avoid a full fsck in most cases.
1308 Severity : enhancement
1310 Description: adapt the lustre_config script to support the upgrade case
1311 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1312 server targets to 1.6 in parallel.
1316 Description: To avoid grant space > avaible space when the disk is almost
1317 full. Without this patch you might see the error "grant XXXX >
1318 available" or some LBUG about grant, when the disk is almost
1320 Details : In filter_check_grant, for non_grant cache write, we should
1321 check the left space by if (*left > ungranted + bytes), instead
1322 of (*left > ungranted), because only we are sure the left space
1323 is enough for another "bytes", then the ungrant space should be
1324 increase. In client, we should update cl_avail_grant only there
1325 is OBD_MD_FLGRANT in the reply.
1329 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1330 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1331 described in CVE-2007-4573. This problem could allow untrusted
1332 local users to gain root access.
1335 Frequency : when using O_DIRECT and quotas
1337 Description: Incorrect file ownership on O_DIRECT output files
1338 Details : block usage reported by 'lfs quota' does not take into account
1339 files that have been written with O_DIRECT.
1344 Description: touch file failed when fs is not full
1345 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1346 otherwise we can get ENOSP while fs is not full.
1350 Description: parallel lock callbacks
1351 Details : Instead of sending blocking and completion callbacks as separated
1352 requests, adding them to a set and sending in parallel.
1355 Frequency : only for Cray XT3
1356 Bugzilla : 12829/13455
1357 Description: Changing primary group doesn't change the group lustre assigns to
1359 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1360 overridden with the primary group provided by the group upcall,
1361 whereas the supplied fsgid can be trusted if it is in the list of
1362 supplementary groups returned by the group upcall.
1364 Severity : enhancement
1366 Description: Allow masking D_WARNING, D_ERROR messages from console
1367 Details : Console messages can now be disabled via lnet.printk.
1371 Description: User code with malformed file open parameter crashes client node
1372 Details : Before packing join_file req, all the related reference should be
1373 checked carefully in case some malformed flags cause fake
1374 join_file req on client.
1378 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1379 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1380 mode is equal to the granted mode. However, it should also check
1381 that LDLM_FL_CP_REQD is not set.
1385 Description: Heavy nfs access might result in deadlocks
1386 Details : After ELC code landed, it is now improper to enqueue any mds
1387 locks under och_sem, because enqueue might want to decide to
1388 cancel open locks for same inode we are holding och_sem for.
1392 Description: Client eviction while running blogbench
1393 Details : A lot of unlink operations with concurrent I/O can lead to a
1394 deadlock causing evictions. To address the problem, the number of
1395 oustanding OST_DESTROY requests is now throttled to
1396 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1397 callbacks are processed in priority.
1401 Description: enable ACLs on MDS by default
1402 Details : ACLs must be enabled on MDS by default.
1405 Frequency : PPC/PPC64 only
1407 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1408 Details : fix duplicated definitions between asm-ppc64/types.h and
1409 lustre_types.h on PPC.
1412 Frequency : PPC/PPC64 only
1414 Description: asm-ppc/segment.h does not exist
1415 Details : fix compile issue on PPC.
1419 Description: better handle error messages in extents code
1422 Frequency : RHEL4 only
1424 Description: mkfs is very slow on IA64/RHEL4
1425 Details : A performance regression has been discovered in the MPT Fusion
1426 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1427 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1428 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1430 Severity : enhancement
1432 Description: SNMP support enhancement
1433 Details : Adding total number of sampled request for an MDS node in snmp
1436 Severity : enhancement
1438 Description: Optimize ldlm waiting list processing for PR extent locks
1439 Details : When processing waiting list for read extent lock and meeting read
1440 lock that is same or wider to it that is not contended, skip
1441 processing rest of the list and immediatelly return current
1442 status of conflictness, since we are guaranteed there are no
1443 conflicting locks in the rest of the list.
1447 Description: Time out and refuse to reconnect
1448 Details : When the failover node is the primary node, it is possible
1449 to have two identical connections in imp_conn_list. We must
1450 compare not conn's pointers but NIDs, otherwise we can defeat
1451 connection throttling.
1455 Description: port llog fixes from b1_6 into HEAD
1456 Details : Port llog reference couting and some llog cleanups from b1_6
1457 (bug 10800) into HEAD, for protect from panic and access to already
1458 free llog structures.
1462 Description: Detect stride IO mode in read-ahead
1463 Details : When a client does stride read, read-ahead should detect that and
1464 read-ahead pages according to the detected stride pattern.
1468 Description: data checksumming impacts single node performance
1469 Details : add support for several checksum algorithm. Currently, only CRC32
1470 and Adler-32 are supported. The checksum type can be changed on
1471 the fly via /proc/fs/lustre/osc/*/checksum_type.
1475 Description: use adler32 for page checksums
1476 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1481 Description: build for x2 fails
1482 Details : fix compile issue on Cray systems.
1486 Description: Properly match for duplicate locks
1487 Details : Due to different lock order from skiplists code, we need to
1488 traverse entire list for now
1491 Frequency : only on PPC/SLES10
1493 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1494 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1495 be defined. Add a hack in mkfs_lustre.c to work around this compile
1500 Description: LASSERT on MDS when client holding flock lock dies
1501 Details : ldlm pool logic depends on number of granted locks equal to
1502 number of released locks which is not true for flock locks, so
1503 just exclude such locks from consideration.
1507 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1508 Details : Use fsfilt_sync() for both the whole filesystem sync and
1509 individual file sync to eliminate dangerous inode locking
1510 with I_LOCK that can lead to a deadlock.
1514 Description: performance in 1.6.3
1515 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1516 sure that 1MB requests are not fragmented by the block layer.
1518 Severity : enhancement
1520 Description: organize the server-side client stats on per-nid basis
1521 Details : Change the structure of stats under obdfilter and mds to
1529 The "uuid"s file would list the uuids of _active_ exports.
1530 And the clear entry is to clear all stats and stale nids.
1532 Severity : enhancement
1534 Description: eliminate client locks in face of contention
1535 Details : file contention detection and lockless i/o implementation
1536 for contended files.
1540 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1543 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1545 Description: MDS may lose file striping (and hence file data) in some cases
1546 Details : If there are additional extended attributes stored on the MDS,
1547 in particular ACLs, SELinux, or user attributes (if user_xattr
1548 is specified for the client mount options) then there is a risk
1549 of attribute loss. Additionally, the Lustre file striping
1550 needs to be larger than default (e.g. striped over all OSTs),
1551 and an additional attribute must be stored initially in the
1552 inode and then increase in size enough to be moved to the
1553 external attribute block (e.g. ACL growing in size) for file
1558 Description: skiplist implementation simplification
1559 Details : skiplists are used to group compatible locks on granted list
1560 that was implemented as tracking first and last lock of each
1561 lock group the patch changes that to using doubly linked lists
1565 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1566 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1567 because LASSERT on both the data supplied by a client, and the
1568 data on disk is dangerous and incorrect.
1570 Severity : enhancement
1572 Description: Slow truncate/writes to huge files at high offsets.
1573 Details : Directly associate cached pages to lock that protect those pages,
1574 this allows us to quickly find what pages to write and remove
1575 once lock callback is received.
1579 Description: more ldlm soft lockups
1580 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1581 starve other threads from the resource lock for a long time in
1582 case of long waiting queue, so change the debug level from
1583 D_OTHER to the less frequently used D_INFO.
1585 Severity : enhancement
1587 Description: add -gid, -group, -uid, -user options to lfs find
1591 Description: Hung threads in invalidate_inode_pages2_range
1592 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1593 once one is completed. As a result, some RPCs are stuck in the
1594 queue and are never sent.
1598 Description: filter threads hungs on waiting journal commit
1599 Details : Cleanup filter group llog code, then only filter group llog will
1600 be only created in the MDS/OST syncing process.
1604 Description: Procfs and llog threads access destoryed import sometimes.
1605 Details : Sync the import destoryed process with procfs and llog threads by
1606 the import refcount and semaphore.
1608 Severity : enhancement
1610 Description: openlock cache of b1_6 port to HEAD
1615 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1616 Details : If insertion of an extent fails, then discard the inode
1617 preallocation and free data blocks else it can lead to duplicate
1622 Description: don't always update ctime in ext3_xattr_set_handle()
1623 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1624 In some cases the ctime should not be updated, for example for
1625 2.0->1.8 compatibility it is necessary to delete an xattr and it
1626 should not update the ctime.
1630 Bugzilla : 15713/16362
1631 Description: Assertion in iopen_connect_dentry in 1.6.3
1632 Details : looking up an inode via iopen with the wrong generation number can
1633 populate the dcache with a disconneced dentry while the inode
1634 number is in the process of being reallocated. This causes an
1635 assertion failure in iopen since the inode's dentry list contains
1636 both a connected and disconnected dentry.
1640 Description: assertion failure in ldlm_handle2lock()
1641 Details : fix a race between class_handle_unhash() and class_handle2object()
1642 introduced in lustre 1.6.5 by bug 13622.
1647 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1648 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1653 Description: oops in sd_iostats_seq_show()
1654 Details : unloading/reloading the scsi low level driver triggers a kernel
1655 bug when trying to access the sd iostat file.
1660 Description: Kernel panics during QLogic driver reload
1661 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1662 patch, causing memory corruption.
1667 Description: journal_dev option does not work in b1_6
1668 Details : pass mount option during pre-mount.
1670 Severity : enhancement
1672 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1673 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1674 extent information of a file. It can be used to map logical blocks
1675 in a file to physical blocks in the block device.
1679 Description: LDLM soft lockups - improvement
1680 Details : It is be possible to send the lock handle along with each read
1681 or write request because the client is already doing a lock match
1682 itself so there isn't any reason the OST should have to re-do that
1686 Frequency : only on Cray X2
1688 Description: X2 build failures
1689 Details : fix build failures on Cray X2.
1693 Description: xid & resent requests
1694 Details : Initialize RPC XID from clock at startup (randomly if clock is
1697 Severity : enhancement
1699 Description: Add lustre_start utility to start or stop multiple Lustre servers
1704 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1705 Details : In case of memory pressure, list_del() can be called twice on
1706 req->rq_history_list, causing a kernel oops.
1710 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1711 Details : fix stack overflow in the distributed lock manager by defering
1712 export eviction after a failed AST to the elt thread instead of
1713 handling it in the dlm interpret routine.
1717 Description: Convert some comments to new format.
1718 Details : Update documenting comments to match doxygen conventions.
1722 Description: Grammar fixes.
1723 Details : A couple of trivial spelling fixes.
1727 Description: OSD_COUNTERS-mandatory
1728 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1729 to simplify the code.
1733 Description: simplify cmm_device freeing logic.
1734 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1738 Description: Add lockdep support to dt_object_operations locking interface.
1739 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1740 parameter indicating lock ordering. Update mdd code to use new
1745 Description: Introduce failloc constants for lockless IO tests.
1746 Details : Add two new failloc constants to test lockless IO. Only one of
1747 them in implemented---another is checked in yet to be landed
1752 Description: Add lockdep support for inode mutex.
1753 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1754 in the situations where Lustre has to lock more than one inode
1759 Description: Add optional invariants checking support.
1760 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1761 configure switch. This macro is to be used for consistency and
1762 sanity checks that are too expensive to be left in `production'
1767 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1768 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1773 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1774 Details : Use spin_lock_nested() in (the only) situation where more than
1775 one ldlm_lock is locked simultaneously. Also, fix possible
1776 dead-lock in ldlm_lock_change_resource() by enforcing particular
1781 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1782 Details : Instead of specifying each ldlm_lock call-back through separate
1783 parameter, wrap them into struct ldlm_callback_suite.
1787 Description: Kill join_lru obd method and its callers.
1788 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1789 in memory---a requirement implemented through ->o_join_lru() obd
1790 method in HEAD. Remove this method and its users.
1794 Description: Add asynchronous ldlm ENQUEUE completion handler.
1795 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1796 case that stock ldlm_completion_ast() cannot handle as it waits
1797 until lock is granted. Introduce new ldlm_completion_ast_async()
1798 for this. Also comment ldlm_completion_ast().
1802 Description: ldlm_error <-> errno conversion.
1803 Details : Add functions to map (rather arbitrary) between LDLM error codes
1804 and standard errno values. CLIO needs this to prevent LDLM specific
1805 constants from escaping ldlm and osc.
1809 Description: Kill unused ldlm_handle2lock_ns() function.
1810 Details : Kill unused ldlm_handle2lock_ns() function.
1814 Description: Add lu_ref support to ldlm_lock
1815 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1816 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1817 and ldlm_resource. LDLM interface has to be changed, because code that
1818 releases a reference on a lock, has to "know" what reference this is.
1819 In the most frequent case
1821 lock = ldlm_handle2lock(handle);
1823 LDLM_LOCK_PUT(lock);
1825 no changes are required. When any other reference (received _not_ from
1826 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1827 instead of LDLM_LOCK_PUT().
1829 Arguably, changes are pervasive, and interface requires some discipline
1830 for proper use. On the other hand, it was very instrumental in finding
1831 a few leaked lock references.
1835 Description: Add ldlm_lock_addref_try().
1836 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1837 attempts to addref a lock that might be being canceled
1842 Description: Add ldlm_weigh_callback().
1843 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1844 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1845 hard-coded `number of pages' logic.
1849 Description: Add lockdep annotations to llog code.
1850 Details : Use appropriately tagged _nested() locking calls in the places
1851 where llog takes more than one ->lgh_lock lock.
1855 Description: Add loi_kms_set().
1856 Details : Wrap kms updates into a helper function.
1860 Description: Constify instances of struct lsm_operations.
1861 Details : Constify instances of struct lsm_operations.
1865 Description: lu_conf support.
1866 Details : On a server, a file system object is uniquely identified
1867 by a fid, which is sufficient to locate and load all object
1868 state (inode). On a client, on the other hand, more data are
1869 necessary instantiate an object. Change lu_object_find() and
1870 friends to take additional `lu_conf' argument describing object.
1871 Typically this includes layout information.
1875 Description: lu_context fixes.
1876 Details : Introduce new lu_context functions that are needed on the client
1877 side, where some system threads (ptlrpcd) are shared by multiple
1878 modules, and so cannot be stopped during module shutdown.
1882 Description: Add start and stop methods to lu_device_type_operations.
1883 Details : Introduce two new methods in lu_device_type_operations, that are
1884 invoked when first instance of a given type is created and last one
1885 is destroyed respectively. This is need by CLIO.
1889 Description: Add lu_ref support to struct lu_device.
1890 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1891 track leaked references.
1895 Description: Introduce lu_kmem_descr.
1896 Details : lu_kmem_descr and its companion interface allow to create
1897 and destroy a number of kmem caches at once.
1901 Description: Fix lu_object finalization race.
1902 Details : Fix a race between lu_object_find() finding an object and its
1903 concurrent finalization. This race is (most likely) not possible
1904 on the server, but might happen on the client.
1908 Description: Introduce lu_ref interface.
1909 Details : lu_ref is a debugging module allowing to track references to
1910 a given object. It is quite cpu expensive, and has to be
1911 explicitly enabled with --enable-lu_ref. See usage description
1916 Description: Factor lu_site procfs stats into a separate function.
1917 Details : Separate lu_site stats printing code into a separate function
1918 that can be reused on a client.
1922 Description: Constify instances of struct {lu,dt,md}_device_operations.
1923 Details : Constify instances of struct {lu,dt,md}_device_operations.
1927 Description: Introduce struct md_site and move meta-data specific parts of
1928 struct lu_site here.
1929 Details : Move md-specific fields out of struct lu_site into special struct
1930 md_site, so that lu_site can be used on a client.
1934 Description: Kill mdd_lov_destroy().
1935 Details : Remove unused mdd code.
1939 Description: Add st_block checking to multistat.c.
1940 Details : Add st_block checking to multistat.c.
1944 Description: Add lu_ref support to struct obd_device.
1945 Details : Add lu_ref tracking to obd_device.
1949 Description: Kill obd_set_fail_loc().
1950 Details : Remove unused code.
1954 Description: Add special type for ptlrpc_request interpret functions.
1955 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1956 there. Actual usage will be in CLIO.
1960 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1961 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1962 doesn't work with lockdep.
1966 Description: Add rwv.c test program.
1967 Details : New testing program exercising readv(2) and writev(2) (Qian).
1971 Description: Add sendfile.c test program.
1972 Details : New testing program exercising sendfile(2) (Jay).
1976 Description: Ratelimit a message that can be very frequent.
1977 Details : Ratelimit a memory allocation failure message that can
1982 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1983 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1987 Description: Liblustre build fixes.
1988 Details : Liblustre build fixes.
1992 Description: libcfs: add cfs_{need,cond}_resched() interface.
1993 Details : libcfs: add cfs_{need,cond}_resched() definition and
1994 implementations for Linux, NT, and liblustre.
1996 Severity : enhancement
1998 Description: More exported tunables for mballoc
1999 Details : Add support for tunable preallocation window and new tunables for
2000 large/small requests
2004 Description: Detect corruption of block bitmap and checking for preallocations
2005 Details : Checks validity of on-disk block bitmap. Also it does better
2006 checking of number of applied preallocations. When corruption is
2007 found, it turns filesystem readonly to prevent further corruptions.
2011 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2012 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2013 window to 0. If it does hit cache miss, and read-pattern is still
2014 stride-io mode, does not reset the stride window, but also does not
2015 increase the stride window length in this case.
2019 Frequency : only for big-endian servers
2020 Description: Check if system is big-endian while mounting fs with extents feature
2021 Details : Mounting a filesystem with extents feature will fail on big-endian
2022 systems since ext3-based ldiskfs is not supported on big-endian
2023 systems. This can be over-riden with "bigendian_extents" mount option.
2025 Severity : enhancement
2027 Description: The root squash functionality
2028 Details : A security feature, which is to prevent users from being able
2029 to mount lustre on their desktop, run as root, and delete
2030 all of the files in the filesystem. The goal is accomplished by
2031 remapping user id (UID) and group id (GID) of the root user to
2032 a UID and GID specified by the system administartor via Lustre
2033 configuration management server (MGS). The functionality also
2034 allows to specify sets of clients for which the remapping does
2039 Description: Excessive recovery window
2040 Details : With AT enabled, the recovery window can be excessively long (6000+
2041 seconds). To address this problem, we no longer use
2042 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2043 timeout no longer depends on the service time, it is set to
2044 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2045 time via pb_service_time.
2049 Description: Watchdog triggered on MDS failover
2050 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2051 the OSTs know that the MDS "UUID" can be reused for the same export
2052 from a different NID, so we do not need to wait for the export to be
2056 Frequency : rare, only if using MMP with Linux RAID
2058 Description: MMP doesn't work with Linux RAID
2059 Details : While using HA for Lustre servers with Linux RAID, it is possible
2060 that MMP will not detect multiple mounts. To make this work we
2061 need to unplug the device queue in RAID when the MMP block is being
2062 written. Also while reading the MMP block, we should read it from
2063 disk and not the cached one.
2065 Severity : enhancement
2067 Description: open file using fid
2068 Details : A file can be opened using just its fid, like
2069 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2072 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2075 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2076 Details : add proper locking for creation of jbd_4k slab cache
2080 Description: MMP check in ext3_remount() fails without displaying any error
2081 Details : When multiple mount protection fails during remount, proper error
2084 Severity : enhancement
2086 Description: Allow stripe size to be up to 4G-64k
2087 Details : Fix math logic to allow large stripe sizes.
2091 Description: add check for >8TB ldiskfs filesystems
2092 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2093 Don't allow >8TB ldiskfs filesystems to be mounted without
2094 force_over_8tb mount option
2096 --------------------------------------------------------------------------------
2098 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2100 * Support for kernels:
2101 2.6.5-7.283 (SLES 9),
2102 2.6.9-55.EL (RHEL 4),
2103 2.6.16.46-0.14 (SLES 10),
2104 2.6.18.8 vanilla (kernel.org)
2105 * Client support for unpatched kernels:
2106 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2107 2.6.16 - 2.6.22 vanilla (kernel.org)
2108 * Due to recently discovered recovery problems, we do not recommend
2109 using patchless RHEL 4 clients with this or any earlier release.
2110 * Recommended e2fsprogs version: 1.39.cfs8
2111 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2112 * Starting with this release, the ldiskfs backing filesystem required
2113 by Lustre is now in its own package, lustre-ldiskfs. This package
2114 should be installed. It is versioned separately from Lustre and
2115 may be released separately in future.
2119 Description: block reactivating mgc import until all deactivates complete
2120 Details : Fix race when failing back MDT/MGS to itself (testing)
2123 Frequency : at statup only
2125 Description: mds_lov_synchronize race leads to various problems
2126 Details : simultaneous MDT->OST connections at startup can cause the
2127 sync to abort, leaving the OSC in a bad state.
2129 Severity : enhancement
2131 Description: add optional extra BUILD_VERSION info
2132 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2133 to override the lustre version.
2136 Frequency : 2.6.18 servers only
2138 Description: ll_kern_mount() doesn't release the module reference
2139 Details : The ldiskfs module reference count never drops down to 0
2140 because ll_kern_mount() doesn't release the module reference.
2145 Description: server LBUG when using old ost_num_threads parameter
2146 Details : Accept the old ost_num_threads parameter but warn that it
2147 is deprecated, and fix an off-by-one error that caused an LBUG.
2152 Description: Transient SCSI error results in persistent IO issue
2153 Details : iobuf->dr_error is not reinitialized to 0 between two
2157 Frequency : sometimes when underlying device returns I/O errors
2159 Description: OSTs not going read-only during write failures
2160 Details : OSTs are not remounted read-only when the journal commit threads
2161 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2162 instead of the ext3 wrappers.
2166 Description: poor connect scaling with increasing client count
2167 Details : Don't run filter_grant_sanity_check for more than 100 exports
2168 to improve scaling for large numbers of clients.
2171 Frequency : SLES10 only
2173 Description: sanity-quota.sh quotacheck failed: rc = -22
2174 Details : Quotas cannot be enabled on SLES10.
2177 Frequency : liblustre clients only
2179 Description: getdirentries does not give error when run on compute nodes
2180 Details : getdirentries does not fail when the size specified as an argument
2181 is too small to contain at least one entry
2183 Severity : enhancement
2185 Description: Add LNET router traceability for debug purposes
2186 Details : If a checksum failure occurs with a router as part of the
2187 IO path, the NID of the last router that forwarded the bulk data
2188 is printed so it can be identified.
2193 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2194 Details : Due to a race condition, liblustre clients were occasionally
2195 evicted incorrectly.
2197 Severity : enhancement
2199 Description: lfs setstripe use optional parameters instead of postional
2202 Severity : enhancement
2204 Description: Nanosecond timestamp support for ldiskfs
2205 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2206 resolution timestamps. There is not yet support for this at
2207 the Lustre filesystem level.
2210 Frequency : during server recovery
2212 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2213 Details : request with rq_no_resend flag not awake l_wait_event if they get
2217 Frequency : nfs export on patchless client
2219 Description: connectathon hang when test nfs export over patchless client
2220 Details : Disconnected dentry cannot be found with lookup, so we do not need
2221 to unhash it or make it invalid
2224 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2225 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2226 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2231 Description: open req refcounting wrong on reconnect
2232 Details : If reconnect happened between getting open reply from server and
2233 call to mdc_set_replay_data in ll_file_open, we will schedule
2234 replay for unreferenced request that we are about to free.
2235 Subsequent close will crash in variety of ways.
2236 Check that request is still eligible for replay in
2237 mdc_set_replay_data().
2242 Description: disable writes to filesystem when reading health_check file
2243 Details : the default for reading the health_check proc file has changed
2244 to NOT do a journal transaction and write to disk, because this
2245 can cause reads of the /proc file to hang and block HA state
2246 checking on a healthy but otherwise heavily loaded system. It
2247 is possible to return to the previous behaviour during configure
2248 with --enable-health-write.
2250 Severity : enhancement
2252 Description: 64-bit inode version
2253 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2254 to the inode. This will be required for version-based recovery.
2259 Description: MDS fails to start if a duplicate client export is detected
2260 Details : in some rare cases it was possible for a client to connect to
2261 an MDS multiple times. Upon recovery the MDS would detect this
2262 and fail during startup. Handle this more gracefully.
2264 Severity : enhancement
2266 Description: Add -o localflock option to simulate old noflock
2268 Details : This will achieve local-only flock/fcntl locks
2274 Description: log_commit_thread vs filter_destroy race leads to crash
2275 Details : Take import reference before releasing llog record semaphore
2280 Description: Wrong request locking in request set processing
2281 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2282 imp_delayed_list, in this place should be used imp_lock.
2285 Frequency : when reconnection
2287 Description: Grant Leak when osc reconnect to OST
2288 Details : When osc reconnect ost, OST(filter) should check whether it
2289 should grant more space to client by comparing fed_grant and
2290 cl_avail_grant, and return the granted space to client instead
2291 of "new granted" space, because client will call osc_init_grant
2292 to update the client grant space info.
2295 Frequency : when client reconnect to OST
2297 Description: Grant Leak when osc do resend and replay bulk write
2298 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2299 bulk write request, because the grant info while be sync between
2300 OSC and OST when reconnect, and we should ignore the grant info
2301 these of resend/replay write req.
2306 Description: Grant space more than avaiable left space sometimes.
2307 Details : When then OST is about to be full, if two bulk writing from
2308 different clients came to OST. Accord the avaliable space of the
2309 OST, the first req should be permitted, and the second one
2310 should be denied by ENOSPC. But if the seconde arrived before
2311 the first one is commited. The OST might wrongly permit second
2312 writing, which will cause grant space > avaiable space.
2315 Frequency : when client is evicted
2317 Description: Grant might be wrongly erased when osc is evicted by OST
2318 Details : when the import is evicted by server, it will fork another
2319 thread ptlrpc_invalidate_import_thread to invalidate the
2320 import, where the grant will be set to 0. While the original
2321 thread will update the grant it got when connecting. So if
2322 the former happened latter, the grant will be wrongly errased
2323 because of this race.
2328 Description: Checking Stale with correct fid
2329 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2330 to check whether it is stale, because sometimes, we want the
2331 enqueue happened anyway, and op_data.fid2 will not be initialized.
2333 Severity : enhancement
2335 Description: update patchless client
2336 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2339 Frequency : only with 2.4 kernel
2341 Description: random memory corruption
2342 Details : size of struct ll_inode_info is to big for union inode.u and this
2343 can be cause of random memory corruption.
2348 Description: Memory leak in recovery
2349 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2350 It should also check obd_fail before fsfilt_start, otherwise if
2351 fsfilt_start return -EROFS,(failover mds during mds recovery).
2352 then the req will return with repmsg->transno = 0 and rc = EROFS.
2353 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2354 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2355 be freed no matter whether fsfilt_commit success or not.
2358 Frequency : only with huge count clients
2360 Description: Prevents from taking the superblock lock in llap_from_page for
2362 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2363 for llap_from_page call in ll_removepage prevents from taking the
2364 superblock lock for a soon died page.
2369 Description: Not check open intent error before release open handle
2370 Details : in some rare cases, the open intent error is not checked before
2371 release open handle, which may cause
2372 ASSERTION(open_req->rq_transno != 0), because it tries to release
2373 the failed open handle.
2378 Description: Set cat log bitmap only after create log success.
2379 Details : in some rare cases, the cat log bitmap is set too early. and it
2380 should be set only after create log success.
2384 Description: Accessing a block bevice can re-enable I/O when Lustre is
2385 tearing down a device.
2386 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2391 Description: service threads may hog cpus when there are a lot of requests
2392 Details : Insert cond_resched to give other threads a chance to use some CPU
2397 Description: the cat log was not initialized in recovery
2398 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2399 unlink log on mds will not be initialized until mds post
2400 recovery. And also in mds post recovery, the unlink log will
2401 initialization will be done asynchronausly, so there will be race
2402 between add unlink log and unlink log initialization.
2406 Description: brw_stats were being printed incorrectly
2407 Details : brw_stats were being printed as log2 but all of them were not
2408 recorded as log2. Also remove some code duplication arising from
2409 filter_tally_{read,write}.
2413 Frequency : rare, only in recovery.
2414 Description: ASSERTION(req->rq_type != LI_POISON) failed
2415 Details : imp_lock should be held while iterating over imp_sending_list for
2416 prevent destroy request after get timeout in ptlrpc_queue_wait.
2420 Description: replay-single.sh test 52 fails
2421 Details : A lock's skiplist need to be cleanup when it being unlinked
2422 from its resource list.
2426 Description: Short directio read returns full requested size rather than
2428 Details : Direct I/O operations should return actual amount of bytes
2429 transferred rather than requested size.
2431 Severity : enhancement
2433 Description: metadata RPC reduction (e.g. for rm performance)
2434 Details : decrease the amount of synchronous RPC between clients and servers
2435 by canceling conflicing lock before the operation on the client
2436 and packing thier handles into the main operation RPC to server.
2438 Severity : enhancement
2440 Description: Async OSC create to avoid the blocking unnecessarily.
2441 Details : If a OST has no remain object, system will block on the creating
2442 when need to create a new object on this OST. Now, ways use
2443 pre-created objects when available, instead of blocking on an
2444 empty osc while others are not empty. If we must block, we block
2445 for the shortest possible period of time.
2449 Description: improve handling recoverable errors
2450 Details : if request processig with error which can be recoverable on server
2451 request should be resend, otherwise page released from cache and
2454 Severity : enhancement
2456 Description: refine locking for avoid write wrong info into lov_objid file
2457 Details : fix possible races with add new target and write/update data in
2460 --------------------------------------------------------------------------------
2462 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2467 Frequency : on some architectures
2469 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2470 Details : Uninitialized flags sometimes cause configuration commands to
2474 Frequency : patchless clients only
2476 Description: missing __iget() symbol export
2477 Details : The __iget() symbol export is missing. To avoid the need for
2478 this on patchless clients the deathrow inode reaper is turned
2479 off, and we depend on the VM to clean up old inodes. This
2480 dependency was during via the fix for bug 12181.
2482 --------------------------------------------------------------------------------
2484 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2486 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2487 INTEROPERATE with older versions automatically. In many cases a
2488 special upgrade step is needed. Please read the
2489 user documentation before upgrading any part of a live system.
2490 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2491 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2492 * WARNING: Lustre configuration and startup changes are required with
2493 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2497 Severity : enhancement
2499 Description: MountConf
2500 Details : Lustre configuration is now managed via mkfs and mount
2501 commands instead of lmc and lconf. New obd types (MGS, MGC)
2502 are added for dynamic configuration management. See
2503 https://mail.clusterfs.com/wikis/lustre/MountConf for
2506 Severity : enhancement
2508 Description: dynamic OST addition
2509 Details : OSTs can now be added to a live filesystem
2511 Severity : enhancement
2513 Description: startup order invariance
2514 Details : MDTs and OSTs can be started in any order. Clients only
2515 require the MDT to complete startup.
2517 Severity : enhancement
2519 Description: parallel, asynchronous orphan cleanup
2520 Details : orphan cleanup is now performed in separate threads for each
2521 OST, allowing parallel non-blocking operation.
2523 Severity : enhancement
2525 Description: optimized stripe assignment
2526 Details : stripe assignments are now made based on ost space available,
2527 ost previous usage, and OSS previous usage, in order to try
2528 to optimize storage space and networking resources.
2530 Severity : enhancement
2532 Description: Permanently set tunables
2533 Details : All writable /proc/fs/lustre tunables can now be permanently
2534 set on a per-server basis, at mkfs time or on a live system.
2536 Severity : enhancement
2538 Description: Lustre message v2
2539 Details : Add lustre message format v2.
2541 Severity : enhancement
2543 Description: client OST exclusion list
2544 Details : Clients can be started with a list of OSTs that should be
2545 declared "inactive" for known non-responsive OSTs.
2549 Description: SPEC SFS validation failure on NFS v2 over lustre.
2550 Details : Changes the blocksize for regular files to be 2x RPC size,
2551 and not depend on stripe size.
2553 Severity : enhancement
2555 Description: Multiple MD RPCs in flight.
2556 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2557 To avoid overly-overloading MDS, introduce a limit on number of
2558 MDS RPCs in flight for a single client and add /proc controls
2559 to adjust this limit.
2561 Severity : enhancement
2563 Description: client read/write statistics
2564 Details : Add client read/write call usage stats for performance
2565 analysis of user processes.
2566 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2567 file access. extents_stats shows chunk size distribution.
2568 extents_stats_per_process show chunk size distribution per
2571 Severity : enhancement
2573 Description: mds statistics
2574 Details : Add detailed mds operations statistics in
2575 /proc/fs/lustre/mds/*/stats.
2579 Description: Failure of copying files with lustre special EAs.
2580 Details : Client side always return success for setxattr call for lustre
2581 special xattr (currently only "trusted.lov").
2586 Description: Refcount LNET uuids
2587 Details : The global LNET uuid list grew linearly with every startup;
2588 refcount repeated list entries instead of always adding to
2591 Severity : enhancement
2593 Description: Dynamic service threads
2594 Details : Within a small range, start extra service threads
2595 automatically when the request queue builds up.
2598 Frequency : mixed-endian client/server environments
2600 Description: mixed-endian crashes
2601 Details : The new msg_v2 system had some failures in mixed-endian
2604 Severity : enhancement
2606 Description: Easy OST removal
2607 Details : OSTs can be permanently deactivated with e.g. 'lctl
2608 conf_param lustre-OST0001.osc.active=0'
2610 Severity : enhancement
2612 Description: MGS proc entries
2613 Details : Added basic proc entries for the MGS showing what filesystems
2616 Severity : enhancement
2618 Description: provide MGS failover
2619 Details : Added config lock reacquisition after MGS server failover.
2621 Severity : enhancement
2623 Description: add Linux 2.4 support
2624 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2628 Description: a large application tries to do I/O to the same resource and dies
2629 in the middle of it.
2630 Details : Check the req->rq_arrival time after the call to
2631 ost_brw_lock_get(), but before we do anything about
2632 processing it & sending the BULK transfer request. This
2633 should help move old stale pending locks off the queue as
2634 quickly as obd_timeout.
2637 Frequency : when an incorrect nid is specified during startup
2639 Description: ptlrpc connect to non-existant node causes kernel crash
2640 Details : LNET can't be re-entered from an event callback, which
2641 happened when we expire a message after the export has been
2642 cleaned up. Instead, hand the zombie cleanup off to another
2645 Severity : enhancement
2647 Description: plain/inodebits lock performance improvement
2648 Details : Grouping plain/inodebits in granted list by their request modes
2649 and bits policy, thus improving the performance of search through
2653 Frequency : only if OST filesystem is corrupted
2655 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2656 Details : for a short time RPCs with bulk IO are in the replay list,
2657 but replay of bulk IOs is unimplemented. If the OST filesystem
2658 is corrupted due to disk cache incoherency and then replay is
2659 started it is possible to trip an assertion. Avoid putting
2660 committed RPCs into the replay list at all to avoid this issue.
2663 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2665 Description: System hang on startup
2666 Details : This bug allowed the liblustre (e.g. catamount) client to
2667 return to the app before handling all startup RPCs. This
2668 could leave the node unresponsive to lustre network traffic
2669 and manifested as a server ptllnd timeout.
2671 Severity : enhancement
2673 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2674 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2675 boolean to control whether to print peer debug info when a
2676 client's RPC times out.
2679 Frequency : only for kernels with patches from Lustre below 1.4.3
2681 Description: Remove old rdonly API
2682 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2685 Frequency : only for devices with external journals
2687 Description: Set external device read-only also
2688 Details : During a commanded failover stop, we set the disk device
2689 read-only while the server shuts down. We now also set any
2690 external journal device read-only at the same time.
2693 Frequency : when upgrading from 1.4 while trying to change parameters
2695 Description: The wrong (new) MDC name was used when setting parameters for
2696 upgraded MDT's. Also allows changing of OSC (and MDC)
2697 parameters if --writeconf is specified at tunefs upgrade time.
2700 Frequency : when setting specific ost indicies
2702 Description: QOS code breaks on skipped indicies
2703 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2704 created with --index need not be sequential.
2706 Severity : enhancement
2708 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2709 Details : The uninit_groups feature works in conjunction with the kernel
2710 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2711 up the pass1 processing of e2fsck. This is a read-only feature
2712 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2713 mount filesystems that have had this feature enabled.
2715 Severity : enhancement
2717 Description: Improve multi-block allocation algorithm to avoid fragmentation
2718 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2719 allocation locality and avoid filesystem fragmentation.
2721 ------------------------------------------------------------------------------
2723 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2725 * Support for kernels:
2726 2.4.21-47.0.1.EL (RHEL 3)
2727 2.6.5-7.283 (SLES 9)
2728 2.6.9-42.0.10.EL (RHEL 4)
2729 2.6.12.6 vanilla (kernel.org)
2730 2.6.16.27-0.9 (SLES 10)
2731 * Recommended e2fsprogs version: 1.39.cfs5
2733 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2737 Frequency : occasional, depends on client load and configuration
2738 Bugzilla : 12181, 12203
2739 Description: data loss for recently-modified files
2741 Details : In some cases it is possible that recently written or created
2742 files may not be written to disk in a timely manner (this should
2743 normally be within 30s unless client IO load is very high).
2744 The problem appears as zero-length files or files that are a
2745 multiple of 1MB in size after a client crash or client eviction
2746 that are missing data at the end of the file.
2748 This problem is more likely to be hit on clients where files are
2749 repeatedly created and unlinked in the same directory, clients
2750 have a large amount of RAM, have many CPUs, the filesystem has
2751 many OSTs, the clients are rebooted frequently, and/or the files
2752 are not accessed by other nodes after being written.
2754 The presence of the problem can be detected by looking at
2755 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2756 smaller than the second (nr_unused) then dirty files will not
2757 be flushed automatically to disk. "sync; sleep 10" should be
2758 run several times on the node before unmounting it to update
2759 Lustre (this is also safe to run on nodes without this problem).
2761 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2762 that can cause this same problem, so customers using that kernel
2763 also need to update the kernel in addition to Lustre. In order
2764 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2766 It is normal that files written just before a client crash (less
2767 than 30s) may not yet have been flushed to disk, even for local
2771 Frequency : frequent on thin XT3 nodes
2773 Description: UUID collision on thin XT3 Linux nodes
2774 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2775 randomly, since we relied on an insufficiently-seeded PRNG.
2780 Description: OSS hangs after "All ost request buffers busy"
2781 Details : A deadlock between quota and journal operations caused OSS
2782 hangs after printing "All ost request buffers busy."
2785 Frequency : always on liblustre builds
2787 Description: Cleanup compiler warnings on liblustre
2790 Frequency : always on liblustre builds on XT3
2792 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2796 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2797 > 32000 subdirectories
2798 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2799 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2800 subdirectory count crosses 32000. This will aid e2fsck to
2801 correctly handle more than 32000 subdirectories.
2804 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2806 Description: System hang on startup
2807 Details : This bug allowed the liblustre (e.g. catamount) client to
2808 return to the app before handling all startup RPCs. This
2809 could leave the node unresponsive to lustre network traffic
2810 and manifested as a server ptllnd timeout.
2812 Severity : enhancement
2814 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2815 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2816 boolean to control whether to print peer debug info when a
2817 client's RPC times out.
2822 Description: make O_SYNC working on 2.6 kernels
2823 Details : 2.6 kernels use different method for mark pages for write,
2824 so need add a code to lustre for O_SYNC work.
2829 Description: Failure to close file and release space on NFS
2830 Details : Put inode details into lock acquired in ll_intent_file_open.
2831 Use mdc_intent_lock in ll_intent_open to properly
2832 detect all kind of errors unhandled by mdc_enqueue.
2837 Description: proc file read during shutdown sometimes raced obd removal,
2839 Details : Add lock to prevent obd access after proc file removal.
2842 Frequency : Only for files larger than 4GB on 32-bit clients.
2844 Description: improperly doing page alignment of locks
2845 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2846 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2851 Description: RPCs being resent when they shouldn't be
2852 Details : Some RPCs that should not be resent are being resent. This
2853 can cause inconsistencies in the RPC state machine. Do not
2854 resend such requests.
2857 Frequency : rare, only with NFS export
2859 Description: Crash on NFS re-export node
2860 Details : under very unusual load conditions an assertion is hit in
2861 ll_intent_file_open()
2864 Frequency : only if OST filesystem is corrupted
2866 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2867 Details : for a short time RPCs with bulk IO are in the replay list,
2868 but replay of bulk IOs is unimplemented. If the OST filesystem
2869 is corrupted due to disk cache incoherency and then replay is
2870 started it is possible to trip an assertion. Avoid putting
2871 committed RPCs into the replay list at all to avoid this issue.
2876 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2877 Details : Large single O_DIRECT read and write calls can fail to allocate
2878 a sufficiently large buffer to process the request. In case of
2879 allocation failure the allocation is retried with a smaller
2880 buffer and broken into smaller requests.
2882 Severity : enhancement
2884 Description: Add -o localflock option to simulate old noflock behaviour.
2885 Details : This will achieve local-only flock/fcntl locks coherentness.
2890 Description: versioning check is incomplete
2891 Details : Checking the version difference of client vs. server, report
2892 error if the gap is too big.
2897 Description: add support PG_writeback bit
2898 Details : add support for PG_writeback bit for Lustre, for more carefull
2899 work with page cache in 2.6 kernel. This also fix some deadlocks
2900 and remove hack for work O_SYNC with 2.6 kernel.
2902 Severity : enhancement
2904 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2905 Details : The uninit_groups feature works in conjunction with the kernel
2906 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2907 up the pass1 processing of e2fsck. This is a read-only feature
2908 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2909 mount filesystems that have had this feature enabled.
2911 Severity : enhancement
2913 Description: Improve multi-block allocation algorithm to avoid fragmentation
2914 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2915 allocation locality and avoid filesystem fragmentation.
2917 ------------------------------------------------------------------------------
2919 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2921 * Support for kernels:
2922 2.6.9-42.0.3.EL (RHEL 4)
2923 2.6.5-7.276 (SLES 9)
2924 2.4.21-47.0.1.EL (RHEL 3)
2925 2.6.12.6 vanilla (kernel.org)
2926 2.6.16.21-0.8 (SLES10)
2927 * Recommended e2fsprogs version: 1.39.cfs2-0
2929 * The backwards-compatible /proc/sys/portals symlink has been removed
2930 in this release. Before upgrading, please ensure that you change
2931 any configuration scripts or /etc/sysctl.conf files that access
2932 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2933 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2934 in advance of the upgrade on any system running Lustre 1.4.6 or
2935 newer, since /proc/sys/lnet was added in that version.
2936 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2940 Frequency : only when quota is used
2942 Description: avoid scanning export list for quota master
2943 Details : Change the algorithms to avoid scanning export list in order
2944 to improve the efficiency.
2947 Frequency : MDS failover only, very rarely
2949 Description: "went back in time" messages on mds failover
2950 Details : The greatest transno may be lost when the current operation
2951 finishes with an error (transno==0) and the client's last_rcvd
2952 record is over-written. Save the greatest transno in the
2953 mds_last_transno for this case.
2956 Frequency : always for specific kernels and striping counts
2958 Description: client may get "Matching packet too big" without ACL support
2959 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2960 when trying to access files in certain configurations. The
2961 clients should in fact be denied when mounting because they do
2962 not understand ACLs.
2965 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2967 Description: many clients connecting with IO in progress causes connect timeouts
2968 Details : Avoid synchronous journal commits to avoid delays caused by many
2969 clients connecting/disconnecting when bulk IO is in progress.
2970 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2971 OST_IO_PORTAL to avoid delays behind potentially many pending
2975 Frequency : occasionally with multiple writers to a single file
2977 Description: shared writes to file may result in wrong size reported by stat()
2978 Details : Allow growing of kms when extent lock is cancelled
2981 Frequency : always with random mmap IO to multi-striped file
2983 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2984 Details : Only if the hole is at the end of OST object so that kms is too
2985 small. Fix is to increase kms accordingly in ll_nopage.
2988 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2990 Description: writes to a missing object would leak memory on the OST
2991 Details : If there is an inconsistency between the MDS and OST filesystems,
2992 such that the MDS references an object that doesn't exist, writes
2993 to that object will leak memory due to incorrect cleanup in the
2994 error handling path, eventually running out of memory on the OST.
2999 Description: Creating too long symlink causes lustre errors
3000 Details : Check symlink and name lengths before sending requests to MDS.
3003 Frequency : only if flock is enabled (not on by default)
3005 Description: posix locks not released on fd closure on 2.6.9+
3006 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3007 kernels, this caused such locks not to be released on fd close and
3008 then assertions on fs unmount about still used locks.
3011 Frequency : MDS failover only, very rarely
3013 Description: clients may get ASSERTION(granted_lock != NULL)
3014 Details : When request was taking a long time, and a client was resending
3015 a getattr by name lock request. The were multiple lock requests
3016 with the same client lock handle and
3017 mds_getattr_name->fixup_handle_for_resent_request found one of the
3018 lock handles but later failed with ASSERTION(granted_lock != NULL).
3023 Description: handle->h_buffer_credits > 0, assertion failure
3024 Details : h_buffer_credits is zero after truncate, causing assertion
3025 failure. This patch extends the transaction or creates a new
3029 Frequency : NFS re-export or patchless client
3030 Bugzilla : 11179, 10796
3031 Description: Crash on NFS re-export node (__d_move)
3032 Details : We do not want to hash the dentry if we don't have a lock.
3033 But if this dentry is later used in d_move, we'd hit uninitialised
3034 list head d_hash, so we just do this to init d_hash field but
3035 leave dentry unhashed.
3038 Frequency : NFS re-export or patchless client
3040 Description: NFS exports has problem with symbolic link
3041 Details : lustre client didn't properly install dentry when re-exported
3042 to NFS or running patchless client.
3045 Frequency : NFS re-export or patchless client
3047 Description: Various nfs/patchless fixes.
3048 Details : fixes reuse disconected alias for lookup process - this fixes
3049 warning "find_exported_dentry: npd != pd",
3050 fix permission error with open files at nfs.
3051 fix apply umask when do revalidate.
3054 Frequency : occasional
3056 Description: Crash on NFS re-export node
3057 Details : calling clear_page() on the wrong pointer triggered oops in
3058 generic_mapping_read().
3061 Frequency : rarely, using O_DIRECT IO
3063 Description: unaligned directio crashes client with LASSERT
3064 Details : check for unaligned buffers before trying any requests.
3067 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3069 Description: stale data returned from RAID cache
3070 Details : If only a small amount of IO is done to the RAID device before
3071 reading it again it is possible to get stale data from the RAID
3072 cache instead of reading it from disk.
3075 Frequency : always for sles10 kernel
3077 Description: sles10 support
3078 Details : ll_follow_link: compile fixes and using of nd_set_link
3079 under newer kernels.
3082 Frequency : depends on arch, kernel and compiler version, always on sles10
3085 Description: recursive or deep enough symlinks cause stack overflow
3086 Details : getting rid of large stack-allocated variable in
3090 Frequency : depends on hardware
3092 Description: lustre write performance loss in the SLES10 kernel
3093 Details : the performance loss is caused by using of write barriers in the
3094 ext3 code. The SLES10 kernel turns barrier support on by
3095 default. The fix is to undo that change for ldiskfs.
3097 ------------------------------------------------------------------------------
3099 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3101 * Support for kernels:
3102 2.6.9-42.0.3EL (RHEL 4)
3103 2.6.5-7.276 (SLES 9)
3104 2.4.21-47.0.1.EL (RHEL 3)
3105 2.6.12.6 vanilla (kernel.org)
3109 Frequency : quota enabled and large files being deleted
3111 Description: releasing more than 4GB of quota at once hangs OST
3112 Details : If a user deletes more than 4GB of files on a single OST it
3113 will cause the OST to spin in an infinite loop. Release
3114 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3119 Description: statfs data retrieved from /proc may be stale or zero
3120 Details : When reading per-device statfs data from /proc, in the
3121 {kbytes,files}_{total,free,avail} files, it may appear
3122 as zero or be out of date.
3125 Frequency : systems with MD RAID1 external journal devices
3127 Description: lconf's call to blkid is confused by RAID1 journal devices
3128 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3129 of returning all block devices that match the journal UUID.
3132 Frequency : always, for aggregate stripe size over 4GB
3134 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3135 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3136 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3139 Frequency : always if "lfs find" used on a local file/directory
3141 Description: "lfs find" segfaults if used on a local file/directory
3142 Details : The case where a directory component was not specified wasn't
3143 handled correctly. Handle this properly.
3146 Frequency : always on ppc64
3148 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3149 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3150 perform endianness conversion.
3153 Frequency : rarely (truncate to non-zero file size after write under load)
3154 Bugzilla : 10730, 10687
3155 Description: Files padded with zeros to next 4K multiple
3156 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3157 it is possible that files that are truncated to a non-zero size
3158 immediately after being written are filled with zero bytes beyond
3159 the truncated size. No file data is lost.
3161 Severity : enhancement
3163 Description: Allow recovery/failover for liblustre clients.
3164 Details : liblustre clients were unaware of failover configurations until
3167 Severity : enhancement
3169 Description: user file locks should fail when not mounting with flock option
3170 Details : Set up an error-returning stub in ll_file_operations.lock field
3171 to prevent incorrect behaviour when client is mounted without
3172 flock option. Also, set up properly f_op->flock field for
3176 Frequency : always on ia64
3178 Description: "lfs df" loops on printing out MDS statfs information
3179 Details : The obd_ioctl_data was not initialized and in some systems
3180 this caused a failure during the ioctl that did not return
3181 an error. Initialize the struct and return an error on failure.
3184 Frequency : SLES 9 only
3186 Description: Error of copying files with lustre special EAs as root
3187 Details : Client side always return success for setxattr call for lustre
3188 special xattr (currently only "trusted.lov").
3191 Frequency : rarely on clusters with both ia64+i386 clients
3193 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3194 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3195 clients are doing concurrent writes to the same file it is
3196 possible that the ia64 clients may LASSERT because the OST
3197 extent locks are not PAGE_SIZE aligned. Ensure that grown
3198 locks are always aligned on the request boundary.
3201 Frequency : specific use, occasional
3203 Description: Overwriting in use executable truncates on-disk binary image
3204 Details : If one node attempts to overwrite an executable in use by
3205 another node, we now correctly return ETXTBSY instead of
3206 truncating the file.
3208 Severity : enhancement
3210 Description: Async OSC create to avoid the blocking unnecessarily.
3211 Details : If a OST has no remain object, system will block on the creating
3212 when need to create a new object on this OST. Now, ways use
3213 pre-created objects when available, instead of blocking on an
3214 empty osc while others are not empty. If we must block, we block
3215 for the shortest possible period of time.
3220 Description: chmod on Lustre root is propagated to other clients
3221 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3222 invalid by the follow_mount time.
3225 Frequency : liblustre clients only
3227 Description: Race in 'instant cancel' lock handling could lead to such locks
3228 never to be granted in case of SMP MDS
3229 Details : Do not destroy not yet granted but cbpending locks in
3233 Frequency : replay/resend of open
3235 Description: non null lock assetion failure in mds_intent_policy
3236 Details : Trying to replay/resend lockless open requests resulted in
3237 mds_open() returning 0 with no lock. Now it sets a flag if
3238 a lock is going to be returned.
3240 Severity : enhancement
3242 Description: Checksum enhancements
3243 Details : New checksum enhancements allow for resending RPCs that failed
3246 Severity : enhancement
3248 Description: Tunables on number of dirty pages in cacche
3249 Details : Allow to set limit on number of dirty pages cached.
3254 Description: client crash on unmount - lock still has references
3255 Details : In some error handling cases it was possible to leak a lock
3256 reference on a client while accessing a file. This was not
3257 harmful to the client during operation, but would cause the
3258 client to crash when the filesystem is unmounted.
3261 Frequency : specific case, rare
3263 Description: ETXTBSY on mds though file not in use by client
3264 Details : ETXTBSY is no longer incorrectly returned when attempting to
3265 chmod or chown a directory that the user previously tried to
3266 execute or a currently-executing binary.
3269 Frequency : extremely rare except on liblustre-based clients
3271 Description: Lustre space not freed when files are deleted
3272 Details : Clean up open-unlinked files after client eviction. Previously
3273 the unlink was skipped and the files remained as orphans.
3278 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3279 Details : In some cases it was possible to send a blocking callback to
3280 a client doing a glimpse, even though that client didn't get
3281 a lock granted. When the glimpse lock is cancelled on the OST
3282 the freed lock is left on the waiting list and corrupted the list.
3285 Frequency : all core dumps
3287 Description: Broke core dumps to lustre
3288 Details : Negative dentry may be unhashed if parent does not have UPDATE
3289 lock, but some callers, e.g. do_coredump, expect dentry to be
3290 hashed after successful create, hash it in ll_create_it.
3292 ------------------------------------------------------------------------------
3294 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3296 * Support for kernels:
3297 2.6.9-42.0.2.EL (RHEL 4)
3298 2.6.5-7.276 (SLES 9)
3299 2.4.21-40.EL (RHEL 3)
3300 2.6.12.6 vanilla (kernel.org)
3304 Frequency : always on RHEL 3
3306 Description: Number of open files grows over time
3307 Details : The number of open files grows over time, whether or not
3308 Lustre is started. This was due to a filp leak introduced
3309 by one of our kernel patches.
3311 ------------------------------------------------------------------------------
3313 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3315 * Support for kernels:
3316 2.6.9-42.EL (RHEL 4)
3317 2.6.5-7.276 (SLES 9)
3318 2.4.21-40.EL (RHEL 3)
3319 2.6.12.6 vanilla (kernel.org)
3324 Bugzilla : 5719, 9635, 9792, 9684
3325 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3326 Details : If a server is under heavy load and cannot reply to new
3327 connection requests before the client resends the (re)connect,
3328 the connection handling code can behave badly if two service
3329 threads are concurrently handing separate (re)connections from
3330 the same client. Add better locking to the connection handling
3331 code, and ensure that only a single connection will be processed
3332 for a given client UUID, even if the lock is dropped.
3334 Severity : enhancement
3336 Description: add TCP zero-copy support to kernel
3337 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3338 sends if the hardware supports scatter-gather and checksumming.
3339 This allows socklnd to do client-write and server-read more
3340 efficiently and reduce CPU utilization from skbuf copying.
3343 Frequency : only if NFS exporting from client
3345 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3346 Details : If setting attributes on a file created under NFS that had
3347 never been opened it would be possible to oops the client
3348 if the file had no objects.
3351 Frequency : always for liblustre
3353 Description: liblustre client does MDS+OSTs setattr RPC for each write
3354 Details : When doing a write from a liblustre client, the client
3355 incorrectly issued an RPC to the MDS and each OST the file was
3356 striped over in order to update the timestamps. When writing
3357 with small chunks and many clients this could overwhelm the MDS
3358 with RPCs. In all cases it would slow down the write because
3359 these RPCs are unnecessary.
3361 Severity : enhancement
3363 Description: allow number of MDS service threads to be changed at module load
3364 Details : It is now possible to change the number of MDS service threads
3365 running. Adding "options mds mds_num_threads={N}" to the MDS's
3366 /etc/modprobe.conf will set the number of threads for the next
3367 time Lustre is restarted (assuming the "mds" module is also
3368 reloaded at that time). The default number of threads will
3369 stay the same, 32 for most systems.
3374 Description: OST crash if filesystem is unformatted or corrupt
3375 Details : If an OST is started on a device that has never been formatted
3376 or if the filesystem is corrupt and cannot even mount then the
3377 error handling cleanup routines would dereference a NULL pointer.
3382 Description: NULL pointer deref in llap_from_page.
3383 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3384 incorrect) mapping if the page was truncated/reclaimed while it was
3385 searched for. Check for this condition and skip such pages when
3386 doing readahead. Introduce extra check to llap_from_page() to
3387 verify page->mapping->host is non-NULL (so page is not anonymous).
3390 Frequency : Sometimes when using sys_sendfile
3392 Description: "page not covered by a lock" warnings from ll_readpage
3393 Details : sendfile called ll_readpage without right page locks present.
3394 Now we introduced ll_file_sendfile that does necessary locking
3395 around call to generic_file_sendfile() much like we do in
3399 Frequency : with certain MDS communication failures at client mount time
3401 Description: NULL pointer deref after failed client mount
3402 Details : a client connection request may delayed by the network layer
3403 and not be sent until after the PTLRPC layer has timed out the
3404 request. If the client fails the mount immediately it will try
3405 to clean up before the network times out the request. Add a
3406 reference from the request import to the obd device and delay
3407 the cleanup until the network drops the request.
3410 Frequency : occasionally during client (re)connect
3412 Description: assertion failure during client (re)connect
3413 Details : processing a client connection request may be delayed by the
3414 client or server longer than the client connect timeout. This
3415 causes the client to resend the connection request. If the
3416 original connection request is replied in this interval, the
3417 client may trip an assertion failure in ptlrpc_connect_interpret()
3418 which thought it would be the only running connect process.
3421 Frequency : only with obd_echo servers and clients that are rebooted
3423 Description: kernel BUG accessing uninitialized data structure
3424 Details : When running an obd_echo server it did not start the ping_evictor
3425 thread, and when a client was evicted an uninitialized data
3426 structure was accessed. Start the ping_evictor in the RPC
3427 service startup instead of the OBD startup.
3429 Severity : enhancement
3430 Bugzilla : 10193 (patchless)
3431 Description: Remove dependency on various unexported kernel interfaces.
3432 Details : No longer need reparent_to_init, exit_mm, exit_files,
3433 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3436 Frequency : rare (only users of deprecated and unsupported LDAP config)
3438 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3439 Details : LDAP apparently contains 'lustreName' attributes instead of
3440 'name'. A simple remapping of the name is sufficient.
3443 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3445 Description: waiting_locks_callback trips kernel BUG if client is evicted
3446 Details : Running with the dump_on_timeout debug flag turned on makes
3447 it possible that the waiting_locks_callback() can try to dump
3448 the Lustre kernel debug logs from an interrupt handler. Defer
3449 this log dumping to the expired_lock_main() thread.
3451 Severity : enhancement
3453 Description: Support NFS exporting on 2.6 kernels.
3454 Details : Implement non-rawops metadata methods for NFS server to use without
3455 changing NFS server code.
3458 Frequency : very rare (synthetic metadata workload only)
3460 Description: two racing renames might cause an MDS thread to deadlock
3461 Details : Running the "racer" program may cause one MDS thread to rename
3462 a file from being the source of a rename to being the target of
3463 a rename at exactly the same time that another thread is doing
3464 so, and the second thread has already enqueued these locks after
3465 doing a lookup of the target and is trying to relock them in
3466 order. Ensure that we don't try to re-lock the same resource.
3469 Frequency : only very large systems with liblustre clients
3471 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3472 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3473 all OSTs in parallel. This allows the eviction of stale liblustre
3474 clients to proceed much faster than if they were done in series,
3475 and also offers similar improvements for other set_info RPCs.
3480 Description: excessive CPU usage during initial read phase on client
3481 Details : During the initial read phase on a client, it would agressively
3482 retry readahead on the file, consuming too much CPU and impacting
3483 performance (since 1.4.5.8). Improve the readahead algorithm
3484 to avoid this, and also improve some other common cases (read
3485 of small files in particular, where "small" is files smaller than
3486 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3491 Description: MDS crash when receiving packet with unknown intent.
3492 Details : Do not LBUG in unknown intent case, just return -EFAULT
3494 Severity : enhancement
3495 Bugzilla : 9293, 9385
3496 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3497 Details : Do not serialize getattr (non-intent version) and statfs.
3500 Frequency : occasional, when OST network is overloaded/intermittent
3502 Description: client evicted by OST after bulk IO timeout
3503 Details : If a client sends a bulk IO request (read or write) the OST
3504 may evict the client if it is unresposive to its data GET/PUT
3505 request. This is incorrect if the network is overloaded (takes
3506 too long to transfer the RPC data) or dropped the OST GET/PUT
3507 request. There is no need to evict the client at all, since
3508 the pinger and/or lock callbacks will handle this, and the
3509 client can restart the bulk request.
3512 Frequency : Always when mmapping file with no objects
3514 Description: client crashes when mmapping file with no objects
3515 Details : Check that we actually have objects in a file before doing any
3516 operations on objects in ll_vm_open, ll_vm_close and
3522 Description: Request leak when working with deleted CWD
3523 Details : Introduce advanced request refcount tracking for requests
3524 referenced from lustre intent.
3526 Severity : Enhancement
3528 Description: Cache open file handles on client.
3529 Details : MDS now will return special lock along with openhandle, if
3530 requested and client is allowed to hold openhandle, even if unused,
3531 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3532 closing files for every read/write openration.
3534 Severity : Enhancement
3536 Description: Cache open negative dentries on client when possible.
3537 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3538 negative dentries on lock revocation.
3543 Description: Remounting a client read-only wasn't possible with a zconf mount
3544 Details : It wasn't possible to remount a client read-only with llmount.
3546 Severity : enhancement
3547 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3548 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3549 MPICH is included to add Lustre-specific ADIO interfaces.
3550 This is based closely on the UFS ADIO layer and only differs
3551 in file creation, in order to allow the OST striping to be set.
3552 This is user-contributed code and not supported by CFS.
3557 Description: extended inode attributes (immutable, append-only) work improperly
3558 when 2.4 and 2.6 kernels are used on client/server or vice versa
3559 Details : Introduce kernel-independent values for these flags.
3561 Severity : enhancement
3564 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3565 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3566 granularity. Now, Fractional MB granularity is be supported,
3567 this is very useful for low memory system.
3569 Severity : enhancement
3571 Description: Getattr by fid
3572 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3573 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3574 have locks and avoids getattr rpc requests for it on every stat.
3577 Frequency : Always, for filesystems larger than 2TB
3579 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3580 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3581 filesystems larger than 2TB to be mounted with mballoc enabled.
3584 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3586 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3587 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3588 filesystems larger than 2TB is mounted then the kernel will
3589 silently corrupt the start of the filesystem. CONFIG_LBD is
3590 enabled for all CFS-supported kernels, but the possibility of
3591 this happening with a modified kernel config exists.
3593 Severity : enhancement
3595 Description: add client O_DIRECT support for 2.6 kernels
3596 Details : It is now possible to do O_DIRECT reads and writes to files
3597 in the Lustre client mountpoint on 2.6 kernel clients.
3599 Severity : enhancement
3601 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3602 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3603 parallel, not waiting for response from every OST before sending
3604 a rpc to the next OST.
3609 Description: setattr vs write race when updating file timestamps
3610 Details : Client processes that update a file timestamp into the past
3611 right after writing to the file (e.g. tar) it is possible that
3612 the updated file modification time can be reset to the current
3613 time due to a race between processing the setattr and write RPC.
3615 Severity : enhancement
3617 Description: Bring 'lfs find' closer in line with regular Linux find.
3618 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3619 -print0 options and obtains all the needed info through the lustre
3622 Severity : enhancement
3624 Description: support up to 1024 configured devices on one node
3625 Details : change obd_dev array from statically allocated to dynamically
3626 allocated structs as they are first used to reduce memory usage
3631 Description: Flush dirty partially truncated pages during truncate
3632 Details : Immediatelly flush partially truncated pages in filter_setattr,
3633 this way we completely avoid having any pages in page cache on OST
3634 and can retire ugly workarounds during writes to flush such pages.
3639 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3640 Details : protect inode from truncation within vfs_unlink() context
3641 just take a reference before calling vfs_unlink() and release it
3642 when parent's i_sem is free.
3647 Description: last_id value checked outside lock on OST caused LASSERT failure
3648 Details : If there were multiple MDS->OST object precreate requests in
3649 flight, it was possible that the OST's last object id was checked
3650 outside a lock and incorrectly tripped an assertion. Move checks
3651 inside locks, and discard old precreate requests.
3654 Frequency : always, if extents are used on OSTs
3656 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3657 Details : OSTs using the extents format would not zero the high 16 bits of
3658 the index physical block number. This is not a problem for any
3659 OST filesystems smaller than 16TB, and no kernels support ext3
3660 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3661 new/modified files) and can be fixed for existing filesystems
3662 with e2fsprogs-1.39-cfs1.
3667 Description: import connection selection may be incorrect if timer wraps
3668 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3669 import connections to be ignored if the 32-bit jiffies counter
3670 wraps. Use a 64-bit jiffies counter.
3673 Frequency : very large clusters immediately after boot
3675 Description: LNET request buffers exhausted under heavy short-term load
3676 Details : If a large number of client requests are generated on a service
3677 that has previously never seen so many requests it is possible
3678 that the request buffer growth cannot keep up with the spike in
3679 demand. Instead of dropping incoming requests, they are held in
3680 the LND until the RPC service can accept more requests.
3683 Frequency : Sometimes during replay
3685 Description: Assertion failure in ll_local_open after replay.
3686 Details : If replay happened on an open request reply before we were able
3687 to set replay handler, reply will become not swabbed tripping the
3688 assertion in ll_local_open. Now we set the handler right after
3689 recognising of open request
3692 Frequency : very rare
3694 Description: kernel reports "badness in vsnprintf"
3695 Details : Reading from the "recovery_status" /proc file in small chunks
3696 may cause a negative length in lprocfs_obd_rd_recovery_status()
3697 call to vsnprintf() (which is otherwise harmless). Exit early
3698 if there is no more space in the output buffer.
3700 Severity : enhancement
3702 Description: clear OBD RPC statistics by writing to them
3703 Details : It is now possible to clear the OBD RPC statistics by writing
3704 to the "stats" file.
3709 Description: Client mtime is not the same on different clients after utimes
3710 Details : In some cases, the client was using the utimes() syscall on
3711 a file cached on another node. The clients now validate the
3712 ctime from the MDS + OSTs to determine which one is right.
3717 Description: Inability to activate failout mode
3718 Details : lconf script incorrectly assumed that in pythong string's numeric
3719 value is used in comparisons.
3722 Frequency : always with multiple stripes per file
3724 Description: Inefficient object allocation for mutli-stripe files
3725 Details : When selecting which OSTs to stripe files over, for files with
3726 a stripe count that divides evenly into the number of OSTs,
3727 the MDS is always picking the same starting OST for each file.
3728 Return the OST selection heuristic to the original design.
3733 Description: mount failures may take full timeout to return an error
3734 Details : Under some heavy load conditions it is possible that a
3735 failed mount can wait for the full obd_timeout interval,
3736 possibly several minutes, before reporting an error.
3737 Instead return an error as soon as the status is known.
3739 Frequency : quota enabled and large files being deleted
3741 Description: releasing more than 4GB of quota at once hangs OST
3742 Details : If a user deletes more than 4GB of files on a single OST it
3743 will cause the OST to spin in an infinite loop. Release
3744 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3749 Description: statfs data retrieved from /proc may be stale or zero
3750 Details : When reading per-device statfs data from /proc, in the
3751 {kbytes,files}_{total,free,avail} files, it may appear
3752 as zero or be out of date.
3755 Frequency : systems with MD RAID1 external journal devices
3757 Description: lconf's call to blkid is confused by RAID1 journal devices
3758 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3759 of returning all block devices that match the journal UUID.
3762 Frequency : always, for aggregate stripe size over 4GB
3764 Description: assertion fails when trying to use 4GB stripe size
3765 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3766 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3769 Frequency : always on ppc64
3771 Description: the first write on an ext3 filesystem with mballoc got stuck
3772 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3773 perform endianness conversion.
3775 ------------------------------------------------------------------------------
3777 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3779 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3780 INTEROPERATE with older versions automatically. Please read the
3781 user documentation before upgrading any part of a live system.
3782 * WARNING: Lustre networking configuration changes are required with
3783 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3786 * Support for kernels:
3787 2.6.9-22.0.2.EL (RHEL 4)
3788 2.6.5-7.244 (SLES 9)
3789 2.6.12.6 vanilla (kernel.org)
3792 Severity : enhancement
3793 Bugzilla : 7981/8208
3794 Description: Introduced Lustre Networking (LNET)
3795 Details : LNET is new networking infrastructure for Lustre, it includes
3796 a reorganized network configuration mode (see the user
3797 documentation for full details) as well as support for routing
3798 between different network fabrics. Lustre Networking Devices
3799 (LNDs) for the supported network fabrics have also been
3800 created for this new infrastructure.
3802 Severity : enhancement
3803 Description: Introduced Access control lists
3804 Details : clients can set ACLs on files and directories in order to have
3805 more fine-grained permissions than the standard Unix UGO+RWX.
3806 The MDS must be started with the "-o acl" mount option.
3808 Severity : enhancement
3809 Description: Introduced filesystem quotas
3810 Details : Administrators may now establish per-user quotas on the
3813 Severity : enhancement
3815 Description: Configuration change for the XT3
3816 The PTLLND is now used to run Lustre over Portals on the XT3
3817 The configure option(s) --with-cray-portals are no longer used.
3818 Rather --with-portals=<path-to-portals-includes> is used to
3819 enable building on the XT3. In addition to enable XT3 specific
3820 features the option --enable-cray-xt3 must be used.
3825 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3826 Details : A race between allocating a new llcd and re-getting the llcd_lock
3827 allowed another thread to grab newly-allocated llcd.
3829 Severity : enhancement
3831 Description: 2.6 OST async journal commit and locking fix to improve performance
3832 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3833 2.6 kernels are now async as they already were in 2.4 kernels so
3834 that they can commit concurrently with the network bulk transfer.
3835 For block-allocated files the filter allocation semaphore is held
3836 to avoid filesystem fragmentation during allocation. BKL lock
3837 removed for 2.6 xattr operations where it is no longer needed.
3842 Description: lconf incorrectly determined whether two IP networks could talk
3843 Details : In some more complicated routing and multiple-network
3844 configurations, lconf will avoid trying to make a network
3845 connection to a disjoint part of the IP space. It was doing the
3846 math incorrectly for one set of cases.
3851 Description: Fix for potential infinite loop processing records in an llog.
3852 Details : If an llog record is corrupted/zeroed, it is possible to loop
3853 forever in llog_process(). Validate the llog record length
3854 and skip the remainder of the block on error.
3857 Frequency : occasional (liblustre only)
3859 Description: liblustre could not open files whose last component is a symlink
3860 Details : sysio_path_walk() would incorrectly pass the open intent to
3861 intermediate path components.
3864 Frequency : rare (liblustre only with non-standard tuning)
3865 Bugzilla : 7201 (7350)
3866 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3867 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3868 after it has been released, only do this for referenced locks
3870 Severity : enhancement
3872 Description: specifying an (invalid) directory default stripe_size of -1
3873 would reset the directory default striping
3874 Details : stripe_size -1 was used internally to signal directory stripe
3875 removal, now use "all default" to signal dir stripe removal
3876 as a directory striping of "all default" is not useful
3879 Frequency : common for large clusters running liblustre clients
3881 Description: doing an ls when liblustre clients are running is slow
3882 Details : sending a glimpse AST to a liblustre client waits for every AST
3883 to time out, as liblustre clients will not respond. Since they
3884 cannot cache data we refresh the OST lock LVB from disk instead.
3886 Severity : enhancement
3888 Description: doing an ls at the same time as file IO can be slow
3889 Details : enqueue and other "small" requests can be blocked behind many
3890 large IO requests. Create a new OST IO portal for non-IO
3891 requests so they can be processed faster.
3894 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3896 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3897 created file with mode 000
3898 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3899 allow an ftruncate() as a non-root user to a file with mode 000.
3900 Lustre now respects this flag to disable mode checks when
3901 truncating a file owned by the user
3904 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3907 Description: Revoking locks from clients that went dead or catatonic might take
3909 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3910 cancellation of such locks instant on servers without waiting for
3911 any reply from clients. Clients drops these locks when cancel
3912 notification from server is received without replying.
3915 Frequency : liblustre-only, when liblustre client dies or becomes busy
3917 Description: Doing ls on Linux clients can take a long time with active
3919 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3920 granting such locks to it in the first place if possible. Locks
3921 are taken by proxy on the OST during the read or write and
3922 dropped immediately afterward. Add connect flags handling, do
3923 not grant locks to liblustre clients for glimpse ASTs.
3925 Severity : enhancement
3927 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3928 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3929 sequential reads and behaves badly with applications doing random
3930 reads. Improve it by reading ahead at least read region, and
3931 avoiding excessive large RPC for small reads.
3933 Severity : enhancement
3935 Description: Creating more than 1000 files for a single job may cause a load
3936 imbalance on the OSTs if there are also a large number of OSTs.
3937 Details : qos_prep_create() uses an OST index reseed value that is an
3938 even multiple of the number of available OSTs so that if the
3939 reseed happens in the middle of the object allocation it will
3940 still utilize the OSTs as uniformly as possible.
3945 Description: OST or MDS may oops in ping_evictor_main()
3946 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3947 but doesn't restart at beginning of obd_exports_timed list
3950 Severity : enhancement
3952 Description: improve by-nid export eviction on the MDS and OST
3953 Details : allow multiple exports with the same NID to be evicted at one
3954 time without re-searching the exports list.
3957 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3959 Description: MDS may oops in groups_free()
3960 Details : in rare race conditions a newly allocated group_info struct is
3961 freed again, and this can be NULL. The 2.4 compatibility code
3962 for groups_free() checked for a NULL pointer, but 2.6 did not.
3965 Frequency : common for liblustre clients doing little filesystem IO
3966 Bugzilla : 9352, 7313
3967 Description: server may evict liblustre clients accessing contended locks
3968 Details : if a client is granted a lock or receives a completion AST
3969 with a blocking AST already set it would not reply to the AST
3970 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3973 Frequency : lfs setstripe, only systems with more than 160 OSTs
3975 Description: unable to set striping with a starting offset beyond OST 160
3976 Details : llapi_create_file() incorrectly limited the starting stripe
3977 index to the maximum single-file stripe count.
3980 Frequency : LDAP users only
3982 Description: lconf did not handle in-kernel recovery with LDAP properly
3983 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3985 Severity : enhancement
3987 Description: bind OST threads to NUMA nodes to improve performance
3988 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3989 node and do their allocations there to localize memory access
3991 Severity : enhancement
3993 Description: llmount can determine client NID directly from Myrinet (GM)
3994 Details : the client NID code from gmnalnid was moved directly into
3995 llmount, removing the need to use this or specifying the
3996 client NID explicitly when mounting GM clients with zeroconf
3999 Frequency : if client is started with down MDS
4001 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4002 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4003 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4004 ptlrpc_import_delayed_req() because we didn't check if the
4005 request was interrupted, and we also didn't break out of the
4006 event loop if there was no timeout
4011 Description: data loss during non-page-aligned writes to a single file from
4012 both multiple nodes and multiple threads on one node at same time
4013 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4014 inconsistency led to false short-reads, that were cached and later
4015 used by ->prepare_write() to fill in partially written page,
4016 leading to data loss.
4019 Frequency : always, if lconf --abort_recovery used
4021 Description: lconf --abort_recovery fails with 'Operation not supported'
4022 Details : lconf was attempting to abort recovery on the MDT device and not
4025 Severity : enhancement
4027 Description: remove cleanup logs
4028 Details : replace lconf-generated cleanup logs with lustre internal
4029 cleanup routines. Eliminates the need for client-cleanup and
4032 Severity : enhancement
4034 Description: add support for EAs (user and system) on lustre filesystems
4035 Details : it is now possible to store extended attributes in the Lustre
4036 client filesystem, and with the user_xattr mount option it
4037 is possible to allow users to store EAs on their files also
4039 Severity : enhancement
4041 Description: Add possibility (config option) to show minimal available OST free
4043 Details : When compiled with --enable-mindf configure option, statfs(2)
4044 (and so, df) will return least minimal free space available from
4045 all OSTs as amount of free space on FS, instead of summary of
4046 free spaces of all OSTs.
4048 Severity : enhancement
4050 Description: do not expand extent locks acquired on OST-side
4051 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4052 by server: they are not cached anyway.
4055 Frequency : when mmap is used/binaries executed from Lustre
4057 Description: Unmmap pages before throwing them away from read cache.
4058 Details : llap_shrink cache now attempts to unmap pages before discarding
4059 them (if unmapping failed - do not discard). SLES9 kernel has
4060 extra checks that trigger if this unmapping is not done first.
4065 Description: lconf didn't resolve symlinks before checking to see whether a
4066 given mountpoint was already in use
4069 Frequency : when migrating failover services
4070 Bugzilla : 6395, 9514
4071 Description: When migrating a subset of services from a node (e.g. failback
4072 from a failover service node) the remaining services would
4073 time out and evict clients.
4074 Details : lconf --force (implied by --failover) sets the global obd_timeout
4075 to 5 seconds in order to quickly disconnect, but this caused
4076 other RPCs to time out too quickly. Do not change the global
4077 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4079 Severity : enhancement
4080 Frequency : if MDS is started with down OST
4081 Bugzilla : 9439,5706
4082 Description: Allow startup/shutdown of an MDS without depending on the
4083 availability of the OSTs.
4084 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4085 Add appropriate locking and lov-osc refcounts for safe
4086 cleaning. Add osc abort_inflight calls in case the
4087 synchronize never started.
4090 Frequency : occasional (Cray XT3 only)
4092 Description: root not authorized to access files in CRAY_PORTALS environment
4093 Details : The client process capabilities were not honoured on the MDS in
4094 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4095 been accessed by an authorized user then root was able to access
4096 the file on the local client also. The root user capabilities
4097 are now allowed on the MDS, as this environment has secure UID.
4100 Frequency : occasional
4102 Description: ldiskfs "too long searching" message happens too often
4103 Details : A debugging message (otherwise harmless) prints too often on
4104 the OST console. This has been reduced to only happen when
4105 there are fragmentation problems on the filesystem.
4110 Description: Division by zero in statfs when all OSCs are inactive
4111 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4112 lov_statfs divides by value returned by lov_get_stripecnt().
4116 Bugzilla : 9489, 3273
4117 Description: First write from each client to each OST was only 4kB in size,
4118 to initialize client writeback cache, which caused sub-optimal
4119 RPCs and poor layout on disk for the first writen file.
4120 Details : Clients now request an initial cache grant at (re)connect time
4121 and so that they can start streaming writes to the cache right
4122 away and always do full-sized RPCs if there is enough data.
4123 If the OST is rebooted the client also re-establishes its grant
4124 so that client cached writes will be honoured under the grant.
4129 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4130 Details : Now I/O RPCs go to different portal number and (presumably) fast
4131 lock requests (and glimses) and other RPCs get their own service
4132 threads pool that should be able to service those RPCs
4135 Severity : enhancement
4137 Description: Ability to exchange lustre version between client and servers and
4138 issue warnings at client side if client is too old. Also for
4139 liblustre clients there is ability to refuse connection of too old
4141 Details : New 'version' field is added to connect data structure that is
4142 filled with version info. That info is later checked by server and
4146 Frequency : rare, liblustre only.
4147 Bugzilla : 9296, 9581
4148 Description: Two simultaneous writes from liblustre at offset within same page
4149 might proceed at the same time overwriting eachother with stale
4151 Details : I/O lock withing llu_file_prwv was released too early, before data
4152 actually was hitting the wire. Extended lock-holding time until
4153 server acknowledges receiving data.
4156 Frequency : extremely rare. Never observed in practice.
4158 Description: avoid generating lustre_handle cookie of 0.
4159 Details : class_handle_hash() generates handle cookies by incrementing
4160 global counter, and can hit 0 occasionaly (this is unlikely, but
4161 not impossible, because initial value of cookie counter is
4162 selected randonly). Value of 0 is used as a sentinel meaning
4163 "unassigned handle" --- avoid it. Also coalesce two critical
4164 sections in this function into one.
4166 Severity : enhancement
4168 Description: allow liblustre clients to delegate truncate locking to OST
4169 Details : To avoid overhead of locking, liblustre client instructs OST to
4170 take extent lock in ost_punch() on client's behalf. New connection
4171 flag is added to handle backward compatibility.
4173 Severity : enhancement
4174 Bugzilla : 4928, 7341, 9758
4175 Description: allow number of OST service threads to be specified
4176 Details : a module parameter allows the number of OST service threads
4177 to be specified via "options ost ost_num_threads={N}" in the
4178 OSS's /etc/modules.conf or /etc/modprobe.conf.
4182 Bugzilla : 6146, 9635, 9895
4183 Description: servers crash with bad pointer in target_handle_connect()
4184 Details : In rare cases when a client is reconnecting it was possible that
4185 the connection request was the last reference for that export.
4186 We would temporarily drop the export reference and get a new
4187 one, but this may have been the last reference and the export
4188 was just destroyed. Get new reference before dropping old one.
4190 Severity : enhancement
4191 Frequency : if client is started with failover MDS
4193 Description: Allow multiple MDS hostnames in the mount command
4194 Details : Try to read the configuration from all specified MDS
4195 hostnames during a client mount in case the "primary"
4198 Severity : enhancement
4200 Description: Stop sending data to evicted clients as soon as possible.
4201 Details : Check if the client we are about to send or are sending data to
4202 was evicted already. (Check is done every second of waiting,
4203 for which l_wait_event interface was extended to allow checking
4204 of exit condition at specified intervals).
4207 Frequency : rare, normally only when NFS exporting is done from client
4209 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4211 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4214 Frequency : always, if srand() is called before liblustre initialization
4216 Description: Liblustre uses system PRNG disturbing its usage by user application
4217 Details : Introduce internal to lustre fast and high-quality PRNG for
4218 lustre usage and make liblustre and some other places in generic
4219 lustre code to use it.
4221 Severity : enhancement
4222 Bugzilla : 9477, 9557, 9870
4223 Description: Verify that the MDS configuration logs are updated when xml is
4224 Details : Check if the .xml configuration logs are newer than the config
4225 logs stored on the MDS and report an error if this is the case.
4226 Request --write-conf, or allow starting with --old_conf.
4228 Severity : enhancement
4230 Description: Handle symlinks in the path when checking if Lustre is mounted.
4231 Details : Resolve intermediate symlinks when checking if a client has
4232 mounted a filesystem to avoid duplicate client mounts.
4237 Description: lconf can hit an error exception but still return success.
4238 Details : The lconf command catches the Command error exception at the top
4239 level script context and will exit with the associated exit
4240 status, but doesn't ensure that this exit status is non-zero.
4245 Description: failure of ptlrpc thread startup can cause oops
4246 Details : Starting a ptlrpc service thread can fail if there are a large
4247 number of threads or the server memory is very fragmented.
4248 Handle this without oopsing.
4251 Frequency : always, only if liblustre and non-default acceptor port was used
4253 Description: liblustre cannot connect to servers with non-default acceptor port
4254 Details : tcpnal_set_default_params() was not called and was therefore
4255 ignoring the environment varaible TCPNAL_PORT, as well as other
4256 TCPNAL_ environment variables
4261 Description: two objects could be created on the same OST for a single file
4262 Details : If an OST is down, in some cases it was possible to create two
4263 objects on a single OST for a single file. No problems other
4264 than potential performance impact and spurious error messages.
4268 Bugzilla : 5681, 9562
4269 Description: Client may oops in ll_unhash_aliases
4270 Details : Client dcache may become inconsistent in race condition.
4271 In some cases "getcwd" can fail if the current directory is
4277 Description: Inode refcounting problems in NFS export code
4278 Details : link_raw functions used to call d_instantiate without obtaining
4279 extra inode reference first.
4283 Bugzilla : 9942, 9903
4284 Description: Referencing freed requests leading to crash, memleaks with NFS.
4285 Details : We used to require that call to ll_revalidate_it was always
4286 followed by ll_lookup_it. Also with revalidate_special() it is
4287 possible to call ll_revalidate_it() twice for the same dentry
4288 even if first occurence returned success. This fix changes semantic
4289 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4290 reference on a request referred from the intent.
4291 ll_intent_release() then releases such a request.
4294 Frequency : rare, normally benchmark loads only
4296 Description: unlinked inodes were kept in memory on the client
4297 Details : If a client is repeatedly creating and unlinking files it
4298 can accumulate a lot of stale inodes in the inode slab cache.
4299 If there is no other client load running this can cause the
4300 client node to run out of memory. Instead flush old inodes
4301 from client cache that have the same inode number as a new inode.
4304 Frequency : SLES9 2.6.5 kernel and long filenames only
4305 Bugzilla : 9969, 10379
4306 Description: utime reports stale NFS file handle
4307 Details : SLES9 uses out-of-dentry names in some cases, which confused
4308 the lustre dentry revalidation. Change it to always use the
4312 Frequency : rare, unless heavy write-truncate concurrency is continuous
4313 Bugzilla : 4180, 6984, 7171, 9963, 9331
4314 Description: OST becomes very slow and/or deadlocked during object unlink
4315 Details : filter_destroy() was holding onto the parent directory lock
4316 while truncating+unlinking objects. For very large objects this
4317 may block other threads for a long time and slow overall OST
4318 responsiveness. It may also be possible to get a lock ordering
4319 deadlock in this case, or run out of journal credits because of
4320 the combined truncate+unlink. Solution is to do object truncate
4321 first in one transaction without parent lock, and then do the
4322 final unlink in a new transaction with the parent lock. This
4323 reduces the lock hold time dramatically.
4326 Frequency : rare, 2.4 kernels only
4328 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4329 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4330 before unmounting their filesystem in order to avoid deadlock.
4331 The kernel_locked() function in 2.4 kernels only checks whether
4332 the kernel lock is held, not whether it is this process that is
4333 holding it as 2.6 kernels do.
4338 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4339 Details : The client ptlrpc code may be trying to reconnect to a down
4340 server before a previous connection attempt has timed out.
4341 Increase the reconnect interval to be longer than the connection
4342 timeout interval to avoid sending duplicate connections to
4346 Frequency : echo_client brw_test command
4348 Description: fix echo_client to work with OST preallocated code
4349 Details : OST preallocation code (5137) didn't take echo_client IO path
4350 into account: echo_client calls filter methods outside of any
4351 OST thread and, hence, there is no per-thread preallocated
4352 pages and buffers to use. Solution: hijack pga pages for IO. As
4353 a byproduct, this avoids unnecessary data copying.
4357 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4358 Description: Client can oops in mdc_commit_close() after open replay
4359 Details : It was possible for the MDS to return an open request with no
4360 transaction number in mds_finish_transno() if the client was
4361 evicted, but without actually returning an error. Clients
4362 would later try to replay that open and may trip an assertion
4363 Simplify the client close codepath, and always return an error
4364 from the MDS in case the open is not successful.
4367 Frequency : rare, 2.6 OSTs only
4369 Description: OST may deadlock under high load on fragmented files
4370 Details : If there was a heavy load and highly-fragmented OST filesystems
4371 it was possible to have all the OST threads deadlock waiting on
4372 allocation of biovecs, because the biovecs were not released
4373 until the entire RPC IO was completed. Instead, release biovecs
4374 as soon as they are complete to ensure forward IO progress.
4376 Severity : enhancement
4378 Description: Support for specifying external journal device at mount
4379 Details : If an OST or MDS device is formatted with an external journal
4380 device, this device major/minor is stored in the ext3 superblock
4381 and may not be valid for failover. Allow detecting and
4382 specifying the external journal at mount time.
4387 Description: Mounting an MDS with pending unlinked files may cause oops
4388 Details : target_finish_recovery() calls mds_postrecov() which returned
4389 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4390 considers this an error and immediately begins cleaning up the
4391 lov, just after starting the mds_lov process
4393 Severity : enhancement
4395 Description: Implement 'lfs df' to report actual free space on per-OST basis
4396 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4397 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4398 usage of objects; '-h' to report in human readable format.
4400 ------------------------------------------------------------------------------
4402 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4409 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4410 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4411 at startup during orphan unlinking. Instead, skip buddy update
4412 before setup, as it will be regenerated after recovery is complete.
4417 Description: If an OST is inactive, its locks might reference stale inodes.
4418 Details : lov_change_cbdata() must iterate over all namespaces, even if
4419 they are inactive to clear inode references from the lock.
4421 Severity : enhancement
4422 Frequency : occasional, if non-standard max_dirty_mb used
4424 Description: Client will block write RPCs if not enough grant
4425 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4426 then the client will block writes while waiting for another RPC
4427 to complete instead of consuming its dirty limit. With change
4428 we get improved performance when max_dirty_mb is small.
4430 Severity : enhancement
4431 Bugzilla : 3389, 6253
4432 Description: Add support for supplementary groups on the MDS.
4433 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4434 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4435 lookups for user supplementary groups into a cache.
4439 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4440 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4441 for liblustre O_EXCL is enforced by MDS.
4445 Description: readdir never returns NULL in liblustre.
4446 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4447 dentry in struct dirent.
4451 Frequency : liblustre only, depends on application IO pattern
4452 Description: liblustre clients evicted if not contacting servers
4453 Details : Don't put liblustre clients into the ping_evictor list, so
4454 they will not be evicted by the pinger ever.
4456 Severity : enhancement
4458 Description: Add ability to evict clients by NID from MDS.
4459 Details : By echoing "nid:$NID" string into
4460 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4461 $NID would be instantly evicted from this MDS and from all active
4462 OSTs connected to it.
4466 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4467 Details : lookup_it_finish() used to query file size from OSTs that was not
4472 Description: service threads change working directory to that of init
4473 Details : Starting lustre service threads may pin the working directory
4474 of the parent thread, making that filesystem busy. Threads
4475 now change to the working directory of init to avoid this.
4479 Frequency : during shutdown only
4480 Description: shutdown with a failed MDS or OST can cause unmount to hang
4481 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4486 Frequency : 2.6 only
4487 Description: chmod/chown may include an extra supplementary group
4488 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4489 supplementary group and if none is specified this is used.
4492 Bugzilla : 5479 (6816)
4494 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4495 Details : If lookup is in progress on a file that is unlinked we might try
4496 to revalidate the inode and fail in revalidate after lookup is
4497 complete and ll_file_open() enqueues the open again but
4498 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4501 Frequency : always, if lconf --abort_recovery used
4503 Description: lconf --abort_recovery fails with 'Operation not supported'
4504 Details : lconf was attempting to abort recovery on the MDT device and not
4507 ------------------------------------------------------------------------------
4509 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4514 Frequency : rare (only unsupported configurations with a node running as an
4516 Bugzilla : 6514, 5137
4517 Description: Mounting a Lustre file system on a node running as an OST could
4519 Details : OSTs now preallocates memory needed to write out data at
4520 startup, instead of when needed, to avoid having to
4521 allocate memory in possibly low memory situations.
4522 Specifically, if the file system is mounted on on OST,
4523 memory pressure could force it to try to write out data,
4524 which it needed to allocate memory to do. Due to the low
4525 memory, it would be unable to do so and the node would
4526 become unresponsive.
4528 Severity : enhancement
4530 Description: Addition of lconf --service command line option
4531 Details : lconf now accepts a '--service <arg>' option, which is
4532 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4534 Severity : enhancement
4536 Description: Failover mode is now the default for OSTs.
4537 Details : By default, OSTs will now run in failover mode. To return to
4538 the old behaviour, add '--failout' to the lmc line for OSTs.
4540 Severity : enhancement
4542 Description: Health checks are now provided for MDS and OSTs
4543 Details : Additional detailed health check information on MSD and OSTs
4544 is now provided through the procfs health_check value.
4547 Frequency : occasional, depends on IO load
4549 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4550 after numerous create/delete cycles
4551 Details : The ext3 inode allocation policy would not allocate new inodes
4552 very well on the OSTs because there are no new directories
4553 being created. Instead we look for groups with free space if
4554 the parent directories are nearly full.
4559 Description: Network or server problems during mount may cause partially
4560 mounted clients instead of returning an error.
4561 Details : The config llog parsing code may overwrite the error return
4562 code during mount error handling, returning success instead
4568 Description: MDS can fail to allocate large reply buffers
4569 Details : After long uptimes the MDS can fail to allocate large reply
4570 buffers (e.g. zconf client mount config records) due to memory
4571 fragmentation or consumption by the buffer cache. Preallocate
4572 some large reply buffers so that these replies can be sent even
4573 under memory pressure.
4577 Frequency : rare (liblustre)
4578 Description: fsx running with liblustre complained that using truncate() to
4579 extend the file doesn't work. This patch corrects that issue.
4580 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4581 ATTR_SIZE and lsm use in llu_setattr_raw.
4585 Frequency : rare, only 2.6 kernels
4586 Description: Unusual file access patterns on the MDS may result in inode
4587 data being lost in very rare circumstances.
4588 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4589 code in the 2.6 kernel caused the inode and/or EA data not to
4590 be read from disk, causing single-file corruption.
4594 Frequency : rare, only 2.6 filesystems using extents
4595 Description: Heavy concurrent write and delete load may cause data corruption.
4596 Details : It was possible under high-load situations to have an extent
4597 metadata block in the block device cache from a just-unlinked
4598 file overwrite a newly-allocated data block. We now unmap any
4599 metadata buffers that alias just-allocated data blocks.
4603 Frequency : filesystems with default stripe_count larger than 77
4604 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4605 Details : lconf specifies an inode size of 4096 bytes when the default
4606 stripe_count is larger than 77. This conflicts with the default
4607 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4608 this case to avoid pinning too much memory for large EAs.
4610 ------------------------------------------------------------------------------
4612 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4617 Frequency : rare (extremely heavy IO load with hundreds of clients)
4619 Description: Client is evicted, gets IO error writing to file
4620 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4621 caused clients to be evicted for AST timeouts. The fixes in
4622 bug 5192 mean we no longer need to have such short AST timeouts
4623 so ldlm_timeout has been increased.
4626 Frequency : occasional during --force or --failover shutdown under load
4627 Bugzilla : 5949, 4834
4628 Description: Server oops/LBUG if stopped with --force or --failover under load
4629 Details : a collection of import/export refcount and cleanup ordering
4630 issues fixed for safer force cleanup
4633 Frequency : only filesystems larger than 120 OSTs
4634 Bugzilla : 5990, 6223
4635 Description: lfs getstripe would oops on a very large filesystem
4636 Details : lov_getconfig used kfree on vmalloc'd memory
4639 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4640 Bugzilla : 6242, 6243
4641 Description: reading from files that had been truncated to a non-zero size
4642 but never opened returned no data
4643 Details : ll_file_read() reads zeros from no-object files to EOF
4648 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4649 Details : There's little harm in aborting MDS/OSS recovery and letting it
4650 try again, so I removed the LASSERT and return an error instead.
4652 Severity : enhancement
4654 Description: New debugging infrastructure for tracking down data corruption
4655 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4656 (b) cover more of the client-side code path, and (c) try to narrow
4657 down where problems occurred
4661 Bugzilla : 3819, 4364, 4397, 6313
4662 Description: Racing close and eviction MDS could cause assertion in mds_close
4663 Details : It was possible to get multiple mfd references during close and
4664 client eviction, leading to one thread referencing a freed mfd.
4666 Severity: : enhancement
4667 Bugzilla : 3262, 6359
4668 Description: Attempts to reconnect to servers are now more aggressive.
4669 Details : This builds on the enhanced upcall-less recovery that was added
4670 in 1.4.2. When trying to reconnect to servers, clients will
4671 now try each server in the failover group every 10 seconds. By
4672 default, clients would previously try one server every 25 seconds.
4677 Description: After recovery, certain operations trigger a failed
4678 assertion on a client.
4679 Details : Failing over an mds, using lconf -d --failover, while a
4680 client was doing a readdir() call would cause the client to
4681 LBUG after recovery completed and the readdir() was resent.
4683 Severity : enhancement
4685 Description: Default groups are now added by lconf
4686 Details : You can now run lconf --group <servicename> without having to
4687 manually add groups with lmc.
4690 Frequency : occasional
4692 Description: Nodes with an elan id of 0 trigger a failed assertion
4695 Frequency : always when accessing e.g. tty/console device nodes
4697 Description: tty and some other devices nodes cannot be used on lustre
4698 Details : file's private_data field is used by device data and lustre
4699 values in there got lost. New field was added to struct file to
4700 store fs-specific private data.
4703 Frequency : when exporting Lustre via NFS
4705 Description: NFSD failed occasionally when looking up a path component
4706 Details : NFSD is looking up ".." which was broken in ext3 directories
4707 that had grown large enough to become hashed.
4710 Frequency : Clusters with multiple interfaces not on the same subnet
4712 Description: Nodes will repeatedly try to reconnect to an interface which it
4713 cannot reach and report an error to the log.
4714 Details : Extra peer list entries will be created by lconf with some peers
4715 unreachable. lconf now validates the peer before adding it.
4718 Frequency : Only if a default stripe is set on the filesystem root.
4720 Description: Setting a default stripe on the filesystem root prevented the
4721 filesystem from being remounted.
4722 Details : The client was sending extra request flags in the root getattr
4723 request and did not allocate a reply buffer for the dir EA.
4726 Frequency : occasional, higher if lots of files are accessed by one client
4727 Bugzilla : 6159, 6097
4728 Description: Client trips assertion regarding lsm mismatch/magic
4729 Details : While revalidating inodes the VFS looks up inodes with ifind()
4730 and in rare cases can find an inode that is being freed.
4731 The ll_test_inode() code will free the lsm during ifind()
4732 when it finds an existing inode and then the VFS later attaches
4733 this free lsm to a new inode.
4737 Bugzilla : 6422, 7030
4738 Description: MDS deadlock between mkdir and client eviction
4739 Details : Creating a new file via mkdir or mknod (starting a transaction
4740 and getting the ns lock) can deadlock with client eviction
4741 (gets ns lock and trying to finish a synchronous transaction).
4744 Frequency : occasional
4745 Description: While starting a server, the fsfilt_ext3 module could not be
4747 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4748 kernels. Previously, lconf would still use the ext3 name
4749 when trying to load modules. Now, it will correctly use
4750 ext3 on 2.4 and ldiskfs on 2.6.
4752 Severity : enhancement
4753 Description: The default stripe count has been changed to 1
4754 Details : The interpretation of the default stripe count (0, to lfs
4755 or lmc) has been changed to mean striping across a single
4756 OST, rather than all available. For general usage we have
4757 found a stripe count of 1 or 2 works best.
4759 Severity : enhancement
4760 Description: Add support for compiling against Cray portals.
4761 Details : Conditional compiling for some areas that are different
4765 Frequency : occasional
4766 Bugzilla : 6409, 6834
4767 Description: Creating files with an explicit stripe count may lead to
4768 a failed assertion on the MDS
4769 Details : If some OSTs are full or unavailable, creating files may
4770 trigger a failed assertion on the MDS. Now, Lustre will
4771 try to use other servers or return an error to the
4775 Frequency : occasional
4777 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4778 caused lock timeout during readahead (since 1.4.2).
4779 Details : Processes doing readahead might match a lock that hasn't been
4780 granted yet if there are overlapping and conflicting lock
4781 requests. The readahead process waits on ungranted lock
4782 (original lock is CBPENDING), while OST waits for that process
4783 to cancel CBPENDING read lock and eventually evicts client.
4785 Severity : enhancement
4787 Description: Initial enabling of flock support for clients
4788 Details : Implements fcntl advisory locking and file status functions.
4789 This feature is provided as an optional mount flag (default
4790 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4791 locking are implemented yet, and those that are are not guaranteed
4792 to be completely correct in production environments.
4793 mount -t lustre -o [flock|noflock] ...
4796 Frequency : occasional
4798 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4799 assertion in the ext3 JBD (journaling) layer.
4800 Details : The b_committed_data struct is protected by the big kernel lock
4801 in 2.4 kernels, serializing journal_commit_transaction() and
4802 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4803 access to this struct. In 2.6 kernels there is finer grained
4804 locking to improve SMP performance of the JBD layer.
4808 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4813 Description: merge and cleanup kernel patches.
4816 -----------------------------------------------------------------------------
4818 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4820 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4821 versions of Lustre. Please update all servers and clients to
4822 version 1.4.2 or later at the same time. You must also run
4823 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4824 to update the configuration logs.
4826 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4827 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4828 - swab llog records of type '0' so we get proper header size/idx (5861)
4829 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4830 - fix rename of one directory over another leaking an inode (5953)
4831 - avoid SetPageDirty on 2.6 (5981)
4832 - don't re-add just-being-destroyed locks to the waiting list (5653)
4833 - when creating new directories, inherit the parent's custom
4834 striping settings if present parent (3048)
4835 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4836 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4837 - don't hold client locks on temporary worklist from l_lru (5666)
4838 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4839 - automatically evict dead clients (5921)
4840 - Update file size properly in create+truncate+fstat case (6196)
4841 - Do not unhash mountpoint dentries, do not allow removal of
4843 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4844 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4845 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4846 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4847 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4848 - Correctly handle reads of files with no objects (6243)
4849 - lctl recover will also mark a device active if deactivate used (5933)
4851 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4852 - allow --write-conf on an MDS with different nettype than client (5619)
4853 - don't write config llogs to MDS for mounts not from that MDS (5617)
4854 - lconf should create multiple TCP connections from a client (5201)
4855 - init scripts are now turned off by default; run chkconfig --on
4856 lustre and chkconfig --on lustrefs to use them
4857 - upcalls are no longer needed for clients to recover to failover
4859 - add --abort-recovery option to lconf to abort recovery on device
4861 - add support for an arbitrary number of OSTs (3026)
4862 - Quota support protocol changes.
4863 - forward compatibility changes to wire structs (6007)
4864 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4865 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4866 - improved "lustre status" script
4867 - initialize blocksize for non-regular files (6062)
4868 - added --disable-server and --disable-client configure options (5782)
4869 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4870 - Vanilla 2.4.29 support
4871 - increase maximum number of obd devices to 520 (6242)
4872 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4873 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4874 SLES 9 2.6.5 kernels
4875 - limit stripes per file to 160 (the maximum EA size) (6093)
4877 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4880 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4881 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4882 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4883 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4884 - small ext3 extents cleanups and fixes (5733)
4885 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4886 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4887 - avoid needless client->OST connect, fix handle mismatch (5317)
4888 - fix DLM error path that led to out-of-sync client, long delays (5779)
4889 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4890 - fix several locking issues related to i_size (5492,5624,5654,5672)
4891 - don't move pending lock onto export if it is already evicted (5683)
4892 - fix kernel oops when creating .foo in unlinked directory (5548)
4893 - fix deadlock in obdfilter statistics vs. object create (5811)
4894 - use time_{before,after} to avoid timer jiffies wrap (5882)
4895 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4896 - Do not leak request if server was not able to process it (5154)
4897 - If mds_open unable to find parent dir, make that negative lookup(5154)
4898 - don't create new directories with extent-mapping (5909, 5936)
4900 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4901 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4902 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4903 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4904 - support for RHEL4 kernel on i686 (5773)
4905 - provide error messages when incompatible logs are encountered (5898)
4907 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4908 * version 1.4.0.10 (1.4.1 release candidate 1)
4910 - don't keep a lock reference when lock is not granted (4238)
4911 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4912 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4913 - drop import inflight refcount on signal_completed_replay error (5255)
4914 - unlock page after async write error during send (3677)
4915 - handle missing objects in filter_preprw_read properly (5265)
4916 - no transno return for symlink open, don't save no-trasno open (3440)
4917 - don't try to complete elan receive that already failed (4012)
4918 - free RPC server reply state on error (5406)
4919 - clean up thread from ptlrpc_start_thread() on error (5160)
4920 - readahead could read extra page into cache that wasn't ejected (5388)
4921 - prevent races in class_attach/setup/cleanup/detach (5260)
4922 - don't dereference de->d_inode after l_dput of de (5458)
4923 - use "int" for stripe value returned from lock_to_stripe (5544)
4924 - mballoc allocation and error-checking fixes in 2.6 (5504)
4925 - block device patches to fix I/O request sizes in 2.6 (5482)
4926 - look up hostnames for IB nals (5602)
4927 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4928 - don't start multiple acceptors for the same port (5277)
4929 - fix incorrect LASSERT in mds_getattr_name (5635)
4930 - export a proc file for general "ping" checking (5628)
4931 - fix "lfs check" to not block when the MDS is down (5628)
4933 - service request history (4965)
4934 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4935 - create an "evict_client" /proc entry on OSTs, like the MDS has
4936 - fix mount usage message, return errors per mount(8) (5168)
4937 - change grep [] to grep "[]" in tests so they work in more UMLs
4938 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4939 - remove ancient LOV_MAGIC_V0 EA support (5047)
4940 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4941 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4942 - allow readahead allocations to fail when low on memory (5383)
4943 - mmap locking landed again, after considerable improvement (2828)
4944 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4946 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4949 - send OST transaction number in read/write reply to free req (4966)
4950 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4951 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4952 - the watchdog thread now runs as interruptible (5246)
4953 - flock/lockf fixes (but it's still disabled, pending 5135)
4954 - don't use EXT3 constants in llite code (5094)
4955 - memory shortage at startup could cause assertion (5176)
4957 - reorganization of lov code
4958 - single portals codebase
4960 - add extents/mballoc support (5025)
4961 - direct I/O reads in the obdfilter (4048)
4962 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4964 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4967 - send OST transaction number in read/write reply to free req (4966)
4968 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4969 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4970 - the watchdog thread now runs as interruptible (5246)
4971 - handle missing objects in filter_preprw_read properly (5265)
4972 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4973 - drop import inflight refcount on signal_completed_replay error (5255)
4974 - unlock page after async write error during send (3677)
4975 - return original error code on reconstructed replies (3761)
4976 - no transno return for symlink open, don't save no-trasno open (3440)
4978 - add pid to ldlm debugging output (4922)
4979 - bump the watchdog timeouts -- we can't handle 30sec yet
4980 - extra debugging for orphan dentry/inode bug (5259)
4982 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4985 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4986 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4987 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4988 - fix NULL dereference in /proc/sys/portals/routes (4827)
4989 - allow failed mdc_close() operations to be interrupted (4561)
4990 - stop precreate on OST before MDS would time out on it (4778)
4991 - don't send partial-page writes before EOF from client (4410)
4992 - discard client grant for sub-page writes on large-page clients (4520)
4993 - don't free dentries not owned by NFS code, check generation (4806)
4994 - fix lsm leak if mds_create_objects() fails (4801)
4995 - limit debug_daemon file size, always print CERROR messages (4789)
4996 - use transno after validating reply (3892)
4997 - process timed out requests if import state changes (3754)
4998 - update mtime on OST during writes, return in glimpse (4829)
4999 - add mkfsoptions to LDAP (4679)
5000 - use ->max_readahead method instead of zapping global ra (5039)
5001 - don't interrupt __l_wait_event() during strace
5003 - add software watchdogs to catch hung threads quickly (4941)
5004 - make lustrefs init script start after nfs is mounted
5005 - fix CWARN/ERROR duplication (4930)
5006 - return async write errors to application if possible (2248)
5007 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5008 - print NAL number in %x format (4645)
5009 - update barely-supported suse-2.4.21-171 series (4842)
5010 - support for sles 9 %post scripts
5011 - support for building 2.6 kernel-source packages
5012 - support for sles km_* packages
5014 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5017 - ignore -ENOENT errors in osc_destroy (3639)
5018 - notify osc create thread that OSC is being cleaned up (4600)
5019 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5020 - reconstruct ost_handle() like mds_handle() (4657)
5021 - create a new thread to do import eviction to avoid deadlock (3969)
5022 - let lconf resolve symlinked-to devices (4629)
5023 - don't unlink "objects" from directory with default EA (4554)
5024 - hold socknal file ref over connect in case target is down (4394)
5025 - allow more than 32000 subdirectories in a single directory (3244)
5026 - fix blocks count for O_DIRECT writes (3751)
5027 - OST returns ENOSPC from object create when no space left (4539)
5028 - don't send truncate RPC if file size isn't changing (4410)
5029 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5030 - bind to privileged port in socknal and tcpnal (3689)
5032 - rate limit CERROR/CWARN console message to avoid overload (4519)
5033 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5034 - basic mmap support (3918)
5035 - kernel patch series update from b1_4 (4711)
5037 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5040 - avoid crash during MDS cleanup with OST shut down (2775)
5041 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5042 - don't use bad inodes on the MDS (3744)
5043 - dynamic object preallocation to improve recovery speed (4236)
5044 - don't hold spinlock over lock dumping or change debug flags (4401)
5045 - don't zero obd_dev when it is force cleaned (3651)
5046 - print grants to console if they go negative (4431)
5047 - "lctl deactivate" will stop automatic recovery attempts (3406)
5048 - look for existing locks in ldlm_handle_enqueue() (3764)
5049 - don't resolve lock handle twice in recovery avoiding race (4401)
5050 - revalidate should check working dir is a directory (4134)
5052 - don't always mark "slow" obdfilter messages as errors (4418)
5054 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5057 - don't close LustreDB during write_conf until it is done (3860)
5058 - fix typo in lconf for_each_profile (3821)
5059 - allow dumping logs from multiple threads at one time (3820)
5060 - don't allow multiple threads in OSC recovery (3812)
5061 - fix debug_size parameters (3864)
5062 - fix mds_postrecov to initialize import for llog ctxt (3121)
5063 - replace config semaphore with spinlock (3306)
5064 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5065 - don't allow enqueue to complete on a destroyed export (3822)
5066 - down write_lock before checking llog header bitmap (3825)
5067 - recover from lock replay timeout (3764)
5068 - up llog sem before sending rpc (3652)
5069 - reduce ns lock hold times when setting kms (3267)
5070 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5071 - fix NULL deref and obd_dev leak on setup error (3312)
5072 - replace some LBUG about llog ops with error handling (3841)
5073 - don't match INVALID dentries from d_lookup and spin (3784)
5074 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5075 - fix invalid assertion in ptlrpc_set_wait (3880)
5077 - add libwrap support for the TCP acceptor (3996)
5078 - add /proc/sys/portals/routes for non-root route listing (3994)
5079 - allow setting MDS UUID in .xml (2580)
5080 - print the stack of a process that LBUGs (4228)
5082 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5085 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5086 - make sure to unset replay flag from failed open requests (3440)
5087 - if default stripe count is 0, use OST count for inode size (3636)
5088 - update parent mtime/ctime on client for create/unlink (2611)
5089 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5090 - free recovery state on server during a forced cleanup (3571)
5091 - unregister_reply for resent reqs (3063)
5092 - loop back devices mounting and status check on 2.6 (3563)
5093 - fix resource-creation race that can provoke i_size == 0 (3513)
5094 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5095 - more debugging for page-accounting assertion (3746)
5096 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5097 - avoid deadlock after precreation failure (3758)
5098 - fix race and lock order deadlock in orphan handling (3450, 3750)
5099 - add validity checks when grabbing inodes from l_ast_data (3599)
5101 - add /proc/.../recovery_status to obdfilter (3428)
5102 - lightweight CDEBUG infrastructure, debug daemon (3668)
5103 - change default OSC RPC parameters to be better on small clusters
5104 - turn off OST read cache for files smaller than 32MB
5105 - install man pages and include them in rpms (3100)
5106 - add new init script for (un)mounting lustre filesystems (2593)
5107 - run chkconfig in %post for init scripts (3701)
5108 - drop scimac NAL (unmaintained)
5110 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5113 - clean kiobufs before and after use (3485)
5114 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5115 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5116 - change init script to fail more clearly if not run as root (1528)
5117 - allow clients to reconnect during replay (1742)
5118 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5119 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5120 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5121 - fix cond_resched (3554)
5122 - only evict unfinished clients after recovery (3515)
5123 - allow bulk resend, prevent data loss (3570)
5124 - dynamic ptlrpc request buffer allocation (2102)
5125 - don't allow unlinking open directory if it isn't empty (2904)
5126 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5127 - remove extraneous obd dereference causing LASSERT failure (3334)
5128 - don't use get_cycles() when creating temp. files on the mds (3156)
5129 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5130 - handle EEXIST for set-stripe, set proper directory name (3336)
5132 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5133 - fix ksocknal_fmb_callback() error messages (2918)
5135 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5138 - don't copy lvb into (possibly NULL) reply on error (2983)
5139 - don't deref dentry after dput, don't free lvb on error (2922)
5140 - use the kms to determine writeback rpc length (2947)
5141 - increment oti_logcookies when osc is inactive (2948)
5142 - update client's i_blocks count via lvb messages (2543)
5143 - handle intent open/close of special files properly (1557)
5144 - mount MDS with errors=remount-ro, like obdfilter (2009)
5145 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5146 - don't use cancelling-locks' kms values (2947)
5147 - use highest lock extent for kms, not last one (2925)
5148 - don't dereference ERR_PTR() dentry in error handling path (3107)
5149 - fix thread race in portals_debug_dumplog() (3122)
5150 - create lprocfs device entries at setup instead of at attach (1519)
5151 - common AST error handler, don't evict client on completion race (3145)
5152 - zero nameidata in detach_mnt in 2.6 (3118)
5153 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5154 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5155 - initialize RPC timeout timer earlier for 2.6 (3219)
5156 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5157 - print nal/nid for unknown nid (3258)
5158 - additional checks for oscc recovery before doing precreate (3284)
5159 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5160 - don't crash in mdc_close for bad permissions on open (3285)
5161 - zero i_rdev for non-device files (3147)
5162 - clear page->private before handing to FS, better assertion (3119)
5163 - tune the read pipeline (3236)
5164 - fix incorrect decref of invalidated dentry (2350)
5165 - provide read-ahead stats and refine rpc in flight stats (3328)
5166 - don't hold journal transaction open across create RPC (3313)
5167 - update atime on MDS at close time (3265)
5168 - close LDAP connection when recovering to avoid server load (3315)
5169 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5170 - don't leak open file on MDS after open resend (3325)
5171 - serialize filter_precreate and filter_destroy_precreated (3329)
5172 - loop device shouldn't call sync_dev() for nul device (3092)
5173 - clear page cache after eviction (2766)
5174 - resynchronize MDS->OST in background (2824)
5175 - refuse to mount the same filesystem twice on same mountpoint (3394)
5176 - allow llmount to create routes for mounting behind routers (3320)
5177 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5178 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5179 - fix rare problem with rename on htree directories (3417)
5181 - allow default OST striping configuration per directory (1414)
5182 - fix compilation for qswnal for 2.6 kernels (3125)
5183 - increase maximum number of MDS request buffers for large systems
5184 - change liblustreapi to be useful for external progs like lfsck (3098)
5185 - increase local configuration timeout for slow disks (3353)
5186 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5188 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5191 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5192 - don't overwrite extent policy data in reply if lock was blocked (2901)
5193 - drop filter export grants atomically with removal from device (2663)
5194 - del obd_self_export from work_list in class_disconnect_exports (2908)
5195 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5196 - swab reply message in mdc_close, other PPC fixes (2464)
5197 - fix destroying of named logs (2325)
5198 - overwrite old logs when running lconf --write_conf (2264)
5199 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5200 - fix race in target_handle_connect (2898)
5201 - mds_reint_create() should take same inode create lock (2926)
5202 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5203 - don't close files for self_export to avoid uninitialized obd (2936)
5204 - allow MDS with the same name as client node (2939)
5205 - hold dentry reference for closed log files for unlink (2325)
5206 - reserve space for all logs during transactions (2059)
5207 - don't evict page beyond end of stripe extent (2925)
5208 - don't oops on a deleted current working directory (2399)
5209 - handle hard links to targets without a parent properly (2517)
5210 - don't dereference NULL lock when racing during eviction (2867)
5211 - don't grow lock extents when lots of conflicting locks (2919)
5213 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5216 - account for cache space usage on clients to avoid data loss (974)
5217 - lfsck support in lustre kernel code (2349)
5218 - reduce journal credits needed for BRW writes (2370)
5219 - orphan handling to avoid losing space on client/server crashes
5220 - ptlrpcd can be blocked, stopping ALL progress (2477)
5221 - use lock value blocks to assist in proper KMS, faster stat (1021)
5222 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5223 - recovery for initial connections (2355)
5224 - fixes for mds_cleanup_orphans (1934)
5225 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5226 - block all file creations until orphan recovery completes (1901)
5227 - client remove rq_connection from request struct (2423)
5228 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5229 - recovery timer race (2670)
5230 - mdc_close recovey bug (2532)
5231 - ptlrpc cleanup bug (2710)
5232 - mds timeout on local locks (2588)
5233 - namespace lock held during RPCs (2431)
5234 - handle interrupted sync write properly (2503)
5235 - don't try to handle a message that hasn't been replied to (2699)
5236 - client assert failure during cleanup after abort recovery (2701)
5237 - leak mdc device after failed mount (2712)
5238 - ptlrpc_check_set allows timedout requests to complete (2714)
5239 - wait for inflight reqs when ptlrpcd finishes (2710)
5240 - make sure unregistered services are removed from the srv_list
5241 - reset bulk XID's when resending them (caught by 1138 test)
5242 - unregister_bulk after timeout
5243 - fix lconf error (2694)
5244 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5245 - readahead locks pages, leaves pending causing memory pressure (2673)
5246 - increase OST request buffers to 4096 on large machines (2729)
5247 - fix up permission of existing directories in simple_mkdir (2661)
5248 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5249 - don't assign transno to errored transactions (2742)
5250 - don't delete objects on OST if given a bogus objid from MDS (2751)
5251 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5252 - if rq_no_resend, then timeout request after recovery (2432)
5253 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5254 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5255 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5256 - MDS recovery shouldn't delete objects that it has given out (2730)
5257 - if enqueue arrives after completion, don't clobber LVB (2819)
5258 - don't unlock pages twice when trigger_group_io returns error (2814)
5259 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5260 - don't write pages to disk if there was an error (1450)
5261 - don't ping imports that have recovery disabled (2676)
5262 - take buffered bytes into account when balancing socknal conn (2817)
5263 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5264 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5265 - remove little-endian swabbing of llog records (1987)
5266 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5267 - retry reposting request buffers if they fail (1191)
5268 - grow extent at grant time to avoid granting a revoked lock (2809)
5269 - lock revoke doesn't evict page if covered by a second lock (2765)
5270 - disable VM readahead to avoid reading outside lock extents (2805)
5272 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5273 - updated kernel patches for hp-2.4.20 kernel (2681)
5275 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5278 - fix truncated write corruption (2366)
5279 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5281 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5282 - ensure len(uuid) < 37 in lmc (1171)
5283 - fix ia64 OOPS in llog_test (2255)
5284 - zero end of page at obdfilter for partial page writes (2648)
5285 - don't leave stale dentries around after renames (bug 2428)
5286 - fix timeouts when evicting a client with a single lock held (2642)
5287 - set deadline for the initial HELLO message to drain (2634)
5288 - print out dotted-quad IP addresses in the socknal (2302)
5290 - additional debugging for MDS client eviction problem (2443)
5291 - fix mkfsoptions support for osts (2603, 2604)
5293 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5296 - add series for the vanilla 2.6.0 kernel
5297 - add series for the vanilla 2.4.24 kernel
5298 - add series for a cray x86/64 UL kernel drop
5299 - fix xattr patches for the vanilla 2.4.19 series
5301 - generate true UUIDs in lmc (1171)
5302 - have portals stack dumping break in UML (2466)
5303 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5304 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5305 - wake callback waiting threads on client eviction (2460)
5306 - Add --ptldebug and --subsystem to lmc (1719)
5307 - update assertion to allow safe interrupt allocation
5308 - set rq_no_resend for cancel requests (2432)
5309 - recalculate ptlrpcd timeout after resend (2494)
5310 - call vfs_rmdir when removing pending directories (2368)
5311 - fix renaming a file to itself (2429)
5312 - lmc creates a default one-stripe lov (2454)
5313 - expand procfs space to handle large clusters (2326)
5314 - increase UML stack to avoid overflow
5315 - update lconf's list of debug and subsystem masks
5316 - fix lfs find --obd (2510)
5317 - /proc tunable for disabling filter read caching (2591)
5318 - stop rpm packages from altering slapd.conf (2301)
5319 - disable nagle in the socknal under 0conf (2578)
5320 - choose mds inode size based on stripe count (2572)
5321 - fix kernel-source rpm problems (2516)
5323 - add --disable-doc to avoid pdf generation (2421)
5324 - update documentation, tests, type-os, comments
5325 - avoid format warnings on ia64
5326 - remove the TOE NAL
5327 - tiny code cleanups by removing unused fields
5329 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5332 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5333 - remove the most blatant lies from BUILDING file (2371)
5334 - change default debug level to reasonable production setting
5335 - reduce client side cache size to reduce cache flush time
5336 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5337 - make TCP zerocopy and pinger support enabled by default (2476)
5338 - sync writes completed after process exits caused crashes (2319)
5339 - maintain correct mount count on the MDS (2356)
5340 - backout 1557, because 2316 wasn't really fixed
5341 - better file I/O statistics gathering in /proc
5342 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5343 - another kernel patch to fix zero-copy TCP function export
5344 - don't take duplicate lock when processing re-sent getattr (2420)
5345 - lctl uses obd_self_export instead of creating new conn (2353)
5346 - MDS/OST recovery case which requires object creation asserted (2425)
5347 - move lfs from /usr/sbin to /usr/bin in packages
5348 - fix race between mds_client_add and mds_client_free (2417)
5349 - use kmalloc instead of slabs in portals (2430)
5350 - don't create duplicate records when a failover MDS is present (2442)
5351 - remove unnecessary mount age check (2332)
5352 - don't remove directory inodes from locks prematurely (2451)
5353 - don't break if MDS service name is the same as hostname (2103)
5354 - fix races in client write RPC generation when cache full (2482)
5356 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5359 - remove now-unused request->rq_obd (278)
5360 - if an allocation fails, print out how much memory we've used (1933)
5361 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5362 - add the "configurable stack size" patch to most series files (1256)
5363 - ability to write large log records, for 100+ OST configs (2306)
5364 - fix NULL deref when filter_prep fails (2314)
5365 - fix operator precedence error in filter_sync
5366 - dynamic allocation of socknal TX descriptors (2315)
5367 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5368 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5369 - ensure that CWARN messages in Portals always get to the syslog
5370 - __init/__exit are not for prototype decls (ldlm_init/exit)
5371 - x86-64 compile warning fixes
5372 - fix gateway LMC keyword conflict (2318)
5373 - fix MDS lock inversions in getattr/reint paths (1844)
5374 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5375 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5376 - DLM race condition prevented some lock evictions (2328)
5377 - ENOMEM detection and retry on socknal sends (2230)
5378 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5379 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5380 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5381 - better error messages when a client is rejected during recovery (1505)
5382 - avoid cancelling locks which were never granted, after failure (2330)
5383 - fix i_sem/journal inversion in mds_client_add (2333)
5384 - fix truncate/getattr lock cycle deadlock (2334)
5385 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5386 - fix two rare exit paths which could leak an l_lock() ref (2321)
5387 - fencepost error in MDS/OST orphan recovery (2226)
5388 - make log record alignment 8 bytes (1988)
5389 - lstripe now fails when requested offset > ost_count (2237)
5390 - ensure that all kernel series have a complete list.h (1607)
5391 - fix crashes in special-file operations (2316)
5392 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5394 - allow configurable automake binary, for testing new versions
5395 - small update to the lfs documentation
5397 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5399 * fix negative export reference count in fsfilt_sync (2312)
5401 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5402 * release candidate 0.9.1
5404 - orphans are moved into the PENDING directory for possible recovery
5405 - replayed opens now open by fid for orphan/rename safety (1042)
5406 - last close of an orphan inode generates a transno (683)
5407 - chdir() and mount() now pin the directory entry (1020)
5408 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5409 - discard very old requests without processing them (1502)
5410 - remove some common, well-understood CERRORs (1505)
5411 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5412 - clear "grant" flags in OST replies until OST grant code lands (1644)
5413 - fix read performance by not clobbering i_blksize on client (1598)
5414 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5415 - make LRU size a /proc tunable, clears locks when reduced (707)
5416 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5417 - server locks take references on exports now (1558)
5418 - build fixes for 2.4.20-rh trees (1663)
5419 - return an error from lov_create if all OSCs are inactive (1751)
5420 - fix import levels when a reconnect happens without a timeout (1597)
5421 - exit early from mds_open if we get a lookup error (1749)
5422 - partial page read at EOF wouldn't wait for disk before sending (1642)
5423 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5424 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5425 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5426 - write/truncate lock inversion (1639)
5427 - Don't auto-load obdclass, portals modules during cleanup (1495)
5428 - fix timestamps from jumping to "now" (1763)
5429 - extra journal assertions (1648)
5430 - add an extra multiunlink test (1771)
5431 - fix read_record/write_record API (1776)
5432 - fix leak of offset_extent, possible incorrect i_size later (1772)
5433 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5434 - Debugging for the kqswnal_get_idle_tx problems (1820)
5435 - Allow recovery to be attempted multiple times (1536)
5436 - Write out MDS last_rcvd file after it is first created (1600)
5437 - Fix tx_descriptor leak in failed transmit situations (1827)
5438 - ext3 journaling fixes for assertion failure after IO error (1871)
5439 - class_export_put() on freed export after completion AST error (1896)
5440 - Fix revalidate looping in VFS (1322)
5441 - Don't access a freed export during MDS_REINT timeout (1521)
5442 - Add open-unlink recovery support on the MDS (1673,1764)
5443 - Return an error if no MDS data was read from last_rcvd (1946)
5444 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5445 - Don't setup a disk device that doesn't match exported UUID (317)
5446 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5447 - avoid committing NULL handle in force close
5448 - local.sh is now a one-stripe LOV configuration
5449 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5450 - fix ext3 htree duplicate directory entry corruption (1516)
5451 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5452 - update to new LOV EA format (2097)
5453 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5455 2003-06-15 Phil Schwan <phil@clusterfs.com>
5458 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5459 - per-import/export recovery handling (958, 931, 959)
5460 - multiple last-rcvd slots, for serving multiple FSes (949)
5461 - connections are again shared between multiple imp/exports (963, 964)
5462 - "umount -f" would hang if any requests needed to be sent (393, 978)
5463 - avoid pinning large req buffer by copying for queued messages (989)
5464 - add "uuid" to "lctl device" command to help upcalls (991)
5465 - "open" RPCs with transnos would confuse recovery counters (1037)
5466 - do proper endian conversion of all wire messages (288, 340, 891)
5467 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5468 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5469 - fix ptlrpc_connection leak in target_handle_connect (1174)
5470 - fix import refcounting bug in OST and MDS cleanup (1134)
5471 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5472 - fix very unlikely obd_types race condition (501)
5473 - remove osc_open hack for echo_client (1187)
5474 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5475 - a failure in read_inode2 leads to deadlock (1139)
5476 - cancel ack-locks as soon as transaction is committed (1072)
5477 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5478 - make sure to commitrw after any preprw to avoid deadlock (1162)
5479 - failing to execute a file in a lustre FS would lock inode (1203)
5480 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5481 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5482 - obd_uuid2tgt would walk off the end of the list (1255)
5483 - on IA64 the osc would give portals incorrect bulk size (1258)
5484 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5485 - fix lock inversion caused by new llite matching code (1282)
5486 - limit the number of dirty pages on a client to 10MB (1286)
5487 - timed out locks were not being corrected cancelled (1289)
5488 - fix O_DIRECT above 4GB on IA-32 (1292)
5489 * major user-visible changes
5490 - fail out/fail over policy now controlled by the upcall (993)
5492 - add OBD_PING to check server availability and failure (954)
5493 - lustre messages are now sent in sending host order (288, 340, 891)
5494 - add eadatalen to MDS getattr reply (340)
5495 - OST read replies may contain second buffer, with per-page status (593)
5497 2003-03-11 Phil Schwan <phil@clusterfs.com>
5500 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5501 - failed open()s could cause deadlock; fixed (867, 869)
5502 - stop cancelling OST locks when files are closed (481)
5503 - overlapping XID spaces caused network corruption (851, 853)
5504 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5505 - setattr_raw added, to do single-RPC, server-side setattrs
5506 - lmc/lconf syntax change for OST UUIDs
5507 - fix crashy race condition between ptlrpc_free_req and osc_close
5508 - don't use request in mdc_enqueue if we hit a timeout (889)
5509 - don't set the inode i_size for regular files from the MDS (896)
5510 - handle out of order completion AST (842)
5511 - don't LBUG if a lock request times out after receiving AST (913)
5512 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5513 - if a bad lock AST arrives, send an error instead of dropping entirely
5514 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5515 - fix leak in bulk IO when only partially completed (899, 900, 926)
5516 - fix O_DIRECT for ia64 (55)
5517 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5518 - C-z after timeout could hang a process forever; fixed (977)
5520 - client-side I/O cache (678, 924, 929, 941, 970)
5522 - READPAGE and SETATTRs which don't take server-side locks get
5525 2003-02-11 Phil Schwan <phil@clusterfs.com>
5528 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5529 - Fix fsfilt_extN_readpage() to read a full page of directory
5530 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5531 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5532 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5533 - Most intent operations were removed, in favour of a new RPC mode
5534 that does a single RPC to the server and bypasses most of the VFS
5535 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5536 - Aggressively cancel local locks on DLM servers
5537 - mds_reint_unlink sends EA to the client if it's the last nlink.
5538 client uses that EA to unlink OST objects.
5539 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5540 - recursive symlinks were fixed (439)
5541 - fixed NULL deref in DEBUG_REQ
5542 - filter_update_lastobjid no longer calls sync, which annoyed extN
5543 - fixed multi-client small-writes to a single file problem (445)
5544 - fixed mtime updates during file writes (607)
5545 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5546 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5547 - fixed Linux OST size reporting problem (444, 656)
5548 - OST now updates object mtime with writes or setattr (607, 619)
5549 - client verifies file size before zeroing page past EOF (445)
5550 - OST now writes last allocated objid to disk with allocation (108)
5551 - LOV on echo now works (409)
5553 - mds_reint_unlink sends a new buffer, with the EA included. this
5554 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5555 set if a regular file was being unlinked, and it was the last link
5556 - use PtlGet from the target for bulk writes (315)
5557 - OST now updates object mtime with writes or setattr (607, 619)
5558 - LDLM now has a grant-time callback to revalidate locked items, if
5560 - Many MDS operations were reorganized to combat race conditions
5562 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5563 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5565 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5568 - Fully reactivate OST imports after reconnection (512, others)
5569 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5570 - More graceful error handling for truncating on dead OST (515)
5571 - Don't error out unless we're actually accessing dead stripes (474)
5572 - Fix garbage sizes when stripes are missing (410)
5573 - LRU counters were broken, causing constant lock purge (433, 432)
5574 - garbage on read from stripes with failed OSTs (441)
5575 - mark OSCs as active before reconnecting during recovery (438)
5576 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5577 - lfind did not preserve OST order in output (443)
5578 - symlinks cause hung clients, incorrect data (439)
5579 - stop dereferencing request after dropping refcount (457)
5580 - don't LASSERT(spin_is_locked) on non-SMP (455)
5581 - fixes for many rename() bugs
5582 - fstat didn't correctly synchronize attributes (399)
5583 - server must handle lock cancellation during blocking AST prep (487)
5584 - bulk descriptors were free()d too soon (511)
5585 - fix paths in lconf, which would load incorrect modules (451, 507)
5586 - fix confusing lconf 'host not found' error message (386)
5587 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5588 - fix race condition in mdc_blocking_ast() for inode access (526)
5589 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5590 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5591 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5592 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5593 - fix lstripe in several ways (kernel side) (527)
5594 - fix request leak in ldlm_cli_enqueue (262)
5595 - incorrect OSC was marked inactive after OST failure
5596 - call mds_fs_cleanup before unmounting filesystem (524)
5597 - fix races between taking ns_lock and ldlm_lock_change_resource
5598 - fix races updating LOV export open file list
5599 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5600 - fix recovery NULL deref in ldlm_cli_cancel_unused
5601 - fix some DLM races by using new hash table for lock handles (419)
5602 - permit the client to specify desired inodes, at replay
5603 - duplicate requests when we queue them for replay reintegration
5604 - fix last_rcvd offset calculation
5605 - sync after each recovered transaction, so we always make progress
5606 - never, not always, ERESTART requests without transnos
5607 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5609 - skip replay if the MDS says that the client is already connected
5610 - don't check for a recovery-enabled export to match lctl's UUID
5611 - don't INC_USE_COUNT for phantom exports
5612 - don't crash when cleaning up phantom exports (567)
5613 - don't double-finish or set replay data for errored mdc_open requests
5614 - abort requests when they time out, so we don't get old replies
5615 - send/receive replies for AST messages again
5616 - if the client says that it doesn't have the lock, cancel it on the
5618 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5619 mark it as destroyed, it will all work out when decref is called
5620 - fix module use counts (22, 581)
5622 - ASTs now expect a reply (server cancels lock on error reply)
5624 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5627 - fix many simultaneous client startup (392)
5628 - fix dentry->d_it clobbering
5629 - credentials weren't being shipped for readdir/getattr operations
5630 - remove invalid assertions triggered during some concurrent MD
5632 - proper Lustre versions added (336, 389)
5633 - fix memory leak for create error case (398)
5634 - fix LOV locking bug that would get cli/srv out of sync
5635 - fix echo client over LOV (409)
5636 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5637 - fix double-O_EXCL intent crash (424)
5638 - avoid sending multiple lock CANCELs (352)
5640 - MDS can do multi-client recovery (modulo bugs in new code)
5642 - many updates, edits, cleanups
5644 2002-11-18 Phil Schwan <phil@clusterfs.com>
5647 - fix null d_it dereference (346)
5648 - fix full OST/dbench hang (333)
5649 - fix permission problem with file removal (286)
5650 - fix removal of OSCs from LOV when they fail
5651 - fix NULL deref during bulk timeout (214)
5652 - fix problems related to multiple filesystems on one MDS (241)
5653 - fixed serious subtle metadata locking bugs
5654 - free locks on clients when inodes are removed due to memory
5656 - fix inode pointer in lock data (285)
5657 - partial support for multiple MDS on a single host (241)
5658 - data locks weren't cancelled at clear_inode time (290, 311)
5659 - intent locks could lead to unbounded lock growth (205)
5660 - added a maximum lock count, an LRU list, and a flusher
5661 - fix multiple rename (365)
5662 - properly abstracted the echo client
5663 - OSC locked 1 byte too many; fixed
5664 - rewrote brw callback code:
5665 - fixed recovery bugs related to LOVs (306)
5666 - fixed too-many-pages-in-one-write crash (191)
5667 - fixed (again) crash in sync_io_timeout (214)
5668 - probably fixed callback-related race (385)
5670 - Add capability to MDS protocol
5671 - LDLM cancellations and callbacks on different portals
5673 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5676 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5677 - defer open object destruction to close time (601981, 138)
5678 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5679 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5680 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5681 - don't call obd_setattr in ll_file_release for destroyed objects
5682 * protocol change to lustre_msg: move |version| and add |flags|
5683 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5684 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5685 * added replay of create, unlink, link and rename operations during
5686 MDS failover; recovery should be much more robust now
5687 * remove failed OSCs from LOVs (only lov_create uses this so far)
5688 * the lustre-HOWTO was brought (more) up to date (582544)
5690 2002-10-23 Phil Schwan <phil@clusterfs.com>
5693 - in-use dentries weren't being reused properly (617851)
5694 - prevent multiple LDLM setup (599178)
5695 - fix LOV size calculations for truncate (617853)
5696 - fix client handling of MDS intent errors (POSIX)
5697 - fix permission bug in lovstripe.c test (624321)
5698 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5699 - truncate past end of file could corrupt data
5700 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5701 - a race in recovery could return ETIMEDOUT to apps (623947)
5702 - building outside the source directory was fixed
5703 * the lustre-HOWTO was brought (more) up to date (582544)
5704 * major progress was made on recovery functionality
5706 2002-10-10 Phil Schwan <phil@clusterfs.com>
5709 - recovery deadlock fix
5710 - rm -rf causes LBUG fix (617817)
5711 - file open by multiple tasks fix (618962)
5712 - directory permissions bugs (602707 and 620007)
5713 - journal_stop fixed with locking (611313)
5714 - O_APPEND failures resolved (618273, perhaps 614459)
5715 - lconf PATH fix (619770)
5716 - IA64 build fix (621450)
5717 - RPC buffer sizes scale with amount of memory
5719 2002-10-01 Phil Schwan <phil@clusterfs.com>
5722 - locks would be cancelled without throwing away data pages,
5723 resulting in inconsistent data (605627)
5724 - inode attributes were not always being refreshed (605627, 612449)
5725 - lconf now continues to cleanup after lctl reports an error
5726 - MDS now enforces user permissions (602707)
5727 - lprocfs cleanup fixed, but not yet enabled (614157)
5728 - fixed infinite server hang, should a client not respond to an AST
5729 - avoid going into recovery if user calls readlink() with a buffer
5730 that's too small (613941)
5731 - AST RPCs no longer require replies (614867) -- this may be changed
5732 - don't crash server if client sends an IOV that's too big (611336)
5733 - fixed lock conversion deadlock (611892)
5734 - fixed the following of symlinks (614622)
5735 * recovery: the server can remove locks from a client that dies, other
5736 clients can make progress
5737 * more extN patch fixes
5738 * compile-time configurable ptlrpc buffer allocations
5740 - collaborative read cache document
5741 - Lustre Lite Performance CDR document-in-progress
5743 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5746 - fix typo in patch-2.4.18
5748 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5751 - clear ptlrpc request each time in handle_incoming_request()
5752 - unlink of files now destroys the object on the OST
5754 2002-09-19 Peter Braam <braam@clusterfs.com>
5756 * add hard link support
5757 * change obdfile creation method
5758 * kernel patch changed
5760 2002-09-19 Peter Braam <braam@clusterfs.com>
5763 - stack overflow bug in extN fixed
5765 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5767 * documentation updates
5768 - add man pages for config tools
5769 - update tests/README to describe testing with new config tools
5770 - finish metadata API descriptions
5771 * bug fixes and cleanups
5772 - statfs workaround for 16TB limit
5773 - LOV stripe allocation improved, can stripe on subset of OSTs
5774 - LOV file size/IO offset was wrong for files > 4GB in size
5775 - object EA data was being dropped, caused files to be unreadable
5776 - memory overflow with non-LOV OST caused memory corruption
5777 - fixed regression tests to work with new config tools, obdfilter
5778 - fixed bug when directory size became larger than 1 block
5779 - fixed bug (for single client case) when PWD was deleted
5780 - invalidate local directory pages when doing intent-based ops
5781 - avoid LDLM oops when lock callback contained bad data
5783 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5785 * documentation updates
5786 * bug fixes and cleanups
5787 - configuration tools
5790 - 64-bit compile warnings
5791 - 64-bit internal statfs data
5793 * test_brw on persistent OST devices
5795 * lprocfs (disabled)
5797 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5799 * documentation updates
5800 * bug fixes and cleanups
5801 * configuration tools
5803 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5806 * many small fixes to 0_5_4
5807 * io/network handling
5808 * thinkos in MDS operations
5810 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5813 * crucial basic fixes to 0.5.3
5814 * IOR, Iozone work over Elan
5817 2002-08-07 Phil Schwan <phil@clusterfs.com>
5818 * version 0_5_3, our first alpha
5819 * we use the new Portals iovs
5820 * documentation updates
5821 * bug fixes and cleanups
5822 * small changes in the DLM wire protocol
5824 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5825 * version 0_5_1 with some initial stability,
5826 * locking on MD and file I/O.
5827 * documentation updates
5828 * several bug fixes since 0.5.0
5829 * small changes in wire protocol
5831 2002-07-18 Phil Schwan <phil@clusterfs.com>
5833 * delivered as Lustre Light Alpha
5834 * fixed a crash after handling invalid MDS requests
5835 * fixed directory pages for architectures with non-4k pages sizes
5837 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5838 * release version v0_4_4
5839 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5841 * Fixes a number of interruption problems with OST operations.
5842 * Update documentation for portals header changes
5843 * Move all wire protocol structs/defines to lustre_idl.h
5844 * Fixes symlink length bug.
5845 * Add tcpdump to repository.
5847 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5848 * release version v0_4_3
5849 * Fixes statfs for inodes on extN.
5850 * Fixes bug in runtests which would delete /etc/hosts.
5851 * Use 64-bit object IDs wherever possible (not into VFS though)
5852 Remove ost_get_info, which is unused by lustre, and out of date.
5854 2002-07-03 Peter Braam <braam@clusterfs.com>
5855 * release version v0_4_2 Fixes a lookup error (type not passed)
5856 * move forward to head of Portals
5857 * move forward to latest Lustre kernel
5859 2002-06-25 Peter Braam <braam@clusterfs.com>
5860 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob