1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.33 (SLES 10),
5 2.6.18-128.1.1.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: 1.40.11-sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
14 * File join has been disabled in this release, refer to Bugzilla 16929.
16 Severity : enhancement
18 Description: Add state history info file, enhance import info file
19 Details : Track import connection state changes in a new osc/mdc proc file;
20 add overview-type data to the osc/mdc import proc file.
22 Severity : enhancement
24 Description: MDS create should not wait for statfs RPC while holding DLM lock.
26 Severity : enhancement
28 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
33 Descriptoin: OST grant too much space to client even there are not enough space.
34 Details : Client will shrink its grant cache to OST if there are no write
35 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
36 this grant cache if there are already not enough avaible space
37 (left_space < total_clients * 32M).
40 Frequency : start MDS on uncleanly shutdowned MDS device
42 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
43 Details : stay in waiting mds<>ost recovery finished produce random bugs
44 due race between two ll_sync thread for one lov target. send
45 ACTIVATE event only if connect realy finished and import have
49 Frequency : rare, connect and disconnect target at same time
51 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
52 Details : don't call obd_disconnect under lov_lock. this long time
53 operation and can block ptlrpcd which answer to connect request.
58 Descriptoin: don't lose wakeup for imp_recovery_waitq
59 Details : recover_import_no_retry or invalidate_import and import_close can
60 both sleep on imp_recovery_waitq, but we was send only one wakeup
64 Frequency : always with long access acl
66 Descriptoin: mds can't pack reply with long acl.
67 Details : mds don't control size of acl but they limited by reint/getattr
70 Severity : enhancement
72 Description: Update to SLES10 kernel-2.6.16.60-0.33.
74 Severity : enhancement
76 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
79 Frequency : start MDS on uncleanly shutdowned MDS device
81 Descriptoin: aborting recovery hang on MDS
82 Details : don't throttle destroy RPCs for the MDT.
85 Frequency : on remount
87 Description: external journal device not working after the remount
88 Details : clear dev_rdonly flag for external journal devices in
94 Description: shutdown vs evict race
95 Details : client_disconnect_export vs connect request race.
96 if client will evicted at this time - we start invalidate
97 thread without referece to import and import can be freed
101 Frequency : rare, need acl's on inode.
103 Description: client can't handle ost additional correctly
104 Details : if ost was added after client connected to mds client can have
105 hit lnet_try_match_md ... to big messages to wide striped files.
106 in this case need teach client to handle config events about add
107 lov target and update client max ea size at that event.
109 Severity : enhancement
111 Description: Changelogs
112 Details : Changelogs are a lightweight mechanism to track filesystem
113 metadata and namespace changes. The changelog is recorded
114 permanently on the MDTs, and is periodically "consumed" / purged
115 when records are no longer needed.
117 Severity : enhancement
119 Description: compact fld format with extents
120 Details : Store range of seq rather than every seq in FLD. Seq
121 controller update FLD rather than clients. In Case of CMD, mdt0
122 has FLD, all other metadata server act as non persistent proxy
123 for FLD queries and cache fld entries in fld cache.
128 Description: don't skip ost target if they assigned to file
129 Details : Drop slow OSCs if we can, but not for requested start idx.
130 This means "if OSC is slow and it is not the requested
131 start OST, then it can be skipped, otherwise skip it only
132 if it is inactive/recovering/out-of-space.
136 Description: more cleanup in mds_lov
137 Details : not send LOV EA under replay, we can't know about they size at this
138 time. Don't allow client connect to mds before any ost connected,
139 for avoid problems with LOV EA size and returning EIO to client.
141 Severity : enhancement
143 Description: Interoperability at server side (Disk interoperability)
145 Severity : enhancement
147 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
149 Severity : enhancement
151 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
153 Severity : enhancement
155 Description: New client IO stack (CLIO).
157 Severity : enhancement
159 Description: Commit on sharing. Eliminate inter-client dependencies between
160 uncommitted transactions by doing transaction commits.
161 Thereby clients may recovery independently.
164 Frequency : Create a symlink file with a very long name
166 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
167 Details : If there is no extra space in the request for early cancels,
168 ldlm_req_handles_avail() returns 0 instead of a negative value.
170 Severity : enhancement
172 Description: Add /proc entry for import status
173 Details : The mdc, osc, and mgc import directories now have
174 an import directory that contains useful import data for debugging
177 Severity : enhancement
179 Description: Re-disable certain /proc logging
180 Details : Enable and disable client's offset_stats, extents_stats and
181 extents_stats_per_process stats logging on the fly.
184 Frequency : Only on FC kernels 2.6.22+
186 Description: oops in statahead
187 Details : Do not drop reference count for the dentry from VFS when lookup,
188 VFS will do that by itself.
190 Severity : enhancement
192 Description: Generic /proc file permissions
193 Details : Set /Proc file permissions in a more generic way to enable non-
194 root users operate on some /proc files.
198 Description: Hitting mdc_commit_close() ASSERTION
199 Details : Properly handle request reference release in
200 ll_release_openhandle().
204 Description: quota recovery deadlock during mds failover
205 Details : This patch includes att18982, att18236, att18237 in bz14840.
207 1. fix osts hang when mds does failover with quotaon
208 2. prevent watchdog storm when osts threads wait for the
213 Frequency : only patchless client
214 Description: add workaround for race between add/remove dentry from hash
216 Severity : enhancement
218 Description: Allow OST glimpses to return PW locks
222 Description: LBUG when llog conf file is full
223 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
227 Description: Prevent import from entering FULL state when server in recovery
231 Description: service mount cannot take device name with ":"
232 Details : Only when device name contains ":/" will mount treat it as
238 Description: replace ptlrpcd with the statahead thread to interpret the async
239 statahead RPC callback
243 Frequency : on recovery
244 Description: I/O failures after umount during fail back
245 Details : if client reconnected to restarted server we need join to recovery
246 instead of find server handler is changed and process self
247 eviction with cancel all locks.
249 Severity : enhancement
251 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
253 Severity : enhancement
255 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
257 Severity : enhancement
259 Description: Upcall on Lustre log has been dumped
260 Details : Allow for a user mode script to be called once a Lustre log has
261 been dumped. It passes the filename of the dumped log to the
262 script, the location of the script can be specified via
263 /proc/sys/lnet/debug_log_upcall.
268 Description: avoid idr_remove called for id which is not allocated.
269 Details : Move assigment s_dev for clustered nfs to end of initialization,
270 to avoid problem with error handling.
275 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
276 Details : When connection is reused this not moved from CONN_UNUSED_HASH
277 into CONN_USED_HASH and this prodice warning when put connection
278 again in unused hash.
281 Severity : enhancement
283 Description: File striping can now be set to use an arbitrary pool of OSTs.
285 Severity : enhancement
287 Description: Export bytes_read/bytes_write count on OSC/OST.
291 Description: Early reply size mismatch, MGC loses connection
292 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
293 the connect flags are properly negotiated.
298 Description: quotas are not honored with O_DIRECT
299 Details : all writes with the flag O_DIRECT will use grants which leads to
300 this problem. Now using OBD_BRW_SYNC to guard this.
304 Description: add quota statistics
305 Details : 1. sort out quota proc entries and proc code.
306 2. add quota statistics
308 Severity : enhancement
310 Description: enable quota support for HEAD.
314 Description: Properly propagate oinfo flags from lov to osc for statfs
315 Details : restore missing copy oi_flags to lov requests.
317 Severity : enhancement
319 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
320 : l_getgroups(8), lst(8), routerstat(8)
321 Details : included man pages for llobdstat(8), llstat(8),
322 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
324 Severity : enhancement
326 Description: configure's --enable-quota should check the
327 : kernel .config for CONFIG_QUOTA
328 Details : configure is terminated if --enable-quota is passed but
329 : no quota support is in kernel
333 Description: Remove portals compatibility
334 Details : Remove portals compatibility, not interoperable with releases
339 Description: Resolve device initialization race
340 Details : Prevent proc handler from accessing devices added to the
341 obd_devs array but yet be intialized.
343 Severity : enhancement
345 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
347 Severity : enhancement
349 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
354 Description: Using wrong pointer in osc_brw_prep_request
355 Details : Access to array[-1] can produce panic if kernel compiled with
356 CONFIG_PAGE_ALLOC enabled
360 Description: Client runs out of low memory
361 Details : Consider only lowmem when counting initial number of llap pages
365 Description: *optional* service tags registration
366 Details : if the "service tags" package is installed on a Lustre node
367 When the filesystem is mounted, a local-node service tag will
368 be created. See http://inventory.sun.com/ for more information
369 about the Service Tags asset management system.
373 Description: Kernel BUG tries to release flock
374 Details : Lustre does not destroy flock lock before last reference goes
375 away. So always drop flock locks when client is evicted and
376 perform unlock regardless of successfulness of speaking to MDS.
380 Description: add refcount protection for osc callbacks, avoid panic on shutdown
384 Description: sanity test 65a fails if stripecount of -1 is set
385 Details : handle -1 striping on filesystem in ll_dirstripe_verify
390 Description: ASSERTION(CheckWriteback(page,cmd)) failed
391 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
394 Severity : enhancement
396 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
400 Description: do not process already freed flock
401 Details : flock can possibly be freed by another thread before it reaches
402 to ldlm_flock_completion_ast.
406 Description: LBUG during stress test
407 Details : Need properly lock accesses the flock deadlock detection list.
411 Description: oops in page fault handler
412 Details : kernel page fault handler can return two special 'pages' in error
413 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
417 Description: timeout with invalidate import.
418 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
419 should be handled by ptlrpcd. This produce long age waiting and
420 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
422 Severity : enhancement
424 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
428 Description: enable MGS and MDT services start separately
429 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
430 a co-located MGS without starting the MGS, which is a complement
431 to 'nosvc' mount option.
435 Frequency : after recovery
436 Description: precreate to many object's after del orphan.
437 Details : del orphan st in oscc last_id == next_id and this triger growing
438 count of precreated objects. Set flag LOW to skip increase count
439 of precreated objects.
443 Frequency : rare, on clear nid stats
444 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
445 Details : when clean nid stats sometimes try destroy live entry,
446 and this produce panic in free.
450 Description: Stack overflow during MDS log replay
451 ease stack pressure by using a thread dealing llog_process.
455 Description: wait until IO finished before start new when do lock cancel.
456 Details : VM protocol want old IO finished before start new, in this case
457 need wait until PG_writeback is cleared until check dirty flag and
458 call writepages in lock cancel callback.
460 Severity : enhancement
462 Description: using special macro for print time and cleanup in includes.
466 Description: mds_mfd_close() ASSERTION(rc == 0)
467 Details : In mds_mfd_close(), we need protect inode's writecount change
468 within its orphan write semaphore to prevent possible races.
472 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
476 Frequency : rare, on shutdown ost
477 Description: don't hit live lock with umount ost.
478 Details : shrink_dcache_parent can be in long loop with destroy dentries,
479 use shrink_dcache_sb instead.
483 Description: don't panic with use echo client
484 Details : echo client pass NULL as client nid pointer and this produce null
489 Description: fix build on ppc32
490 Details : compile code with -m64 flag produce wrong object file for ppc32.
494 Description: add message levels for liblustreapi
498 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
499 Details : Move the 'good_osts' check before the 'total_bavail' check. This
500 will result in an -EAGAIN and in the exit call path we call
501 alloc_rr() which will with increasing aggressiveness attempt to
502 aquire precreated objects on the minimum number of required OSCs.
506 Description: Use old size assignment to avoid deadlock
507 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
508 the scheduling while holding a spinlock. We do not need locking
509 for size in ll_update_inode() because size is only updated from
510 the MDS for directories or files without objects, so there is no
511 other place to do the update, and concurrent access to such inodes
512 are protected by the inode lock.
516 Description: resolve "_IOWR redefined" build error on SLES10
520 Description: dump the memory debugging after all modules are unloaded to
521 suppress false negative in conf_sanity test 39
523 Severity : enhancement
525 Description: build kernel-ib packages for OFED 1.3 in our release cycle
530 Description: fix SLES kernel versioning
531 Details : the kernel version for our SLES 10 kernel did not include a "-"
532 before the "smp" at the end. while this was not a problem in
533 general, it did mean that software trying to use the kernel
534 version to try to detect a vendor specific kernel would fail.
535 this was most evident by the OFED build scripts.
539 Description: Don't update lov_desc members until making sure they are valid
540 Details : When updating lov_desc members via proc fs, need fix their
541 validities before doing the real update.
545 Description: don't put request into delay list while invalidate in flight.
546 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
547 import in flight. this produce timeout for invalidate and sometimes
548 can cause stale data.
552 Frequency : on ppc only
553 Description: not convert ost objects for directory because it's not exist.
554 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
558 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
560 Severity : enhancement
562 Description: Update to support 2.6.22.14 vanilla kernel.
566 Frequency : rare, on recovery
567 Description: read procfs can produce deadlock in some situation
568 Details : Holding lprocfs lock which send rpc can produce block for destroy
569 obd objects and this also block reconnect with -EALREADY. This
570 isn't fix all lprocfs bugs - but make it rare.
572 Severity : enhancement
574 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
577 Frequency : frequent on X2 node
579 Description: mdc_set_open_replay_data LBUG
580 Details : Set replay data for requests that are eligible for replay.
584 Description: lustre_mgs: operation 101 on unconnected MGS
585 Details : When MGC is disconnected from MGS long enough, MGS will evict the
586 MGC, and late on MGC cannot successfully connect to MGS and a lot
587 of the error messages complaining that MGS is not connected.
591 Frequency : on network error
592 Description: panic with double free request if network error
593 Details : mdc_finish_enqueue is finish request if any network error ocuring,
594 but it's true only for synchronus enqueue, for async enqueue
595 (via ptlrpcd) this incorrect and ptlrpcd want finish request
598 Severity : enhancement
600 Description: client-side metadata stat-ahead during readdir(directory readahead)
601 Details : perform client-side metadata stat-ahead when the client detects
602 readdir and sequential stat of dir entries therein
605 Frequency : on start mds
607 Description: Implement get_info(last_id) in obdfilter.
610 Frequency : occasional
612 Description: Correctly check stale fid, not start epoch if ost not support SOM
613 Details : open with flag O_CREATE need set old fid in op_fid3 because
614 op_fid2 was overwritten with new generated fid, but mds can answer
615 with one of these two fids and both is not stale. Setattr
616 incorrectly started an epoch and assume will be called
617 done_writing, but without SOM done_writing ever being called.
620 Frequency : rare, depends on device drivers and load
622 Description: MDS or OSS nodes crash due to stack overflow
623 Details : Code changes in 1.8.0 increased the stack usage of some functions.
624 In some cases, in conjunction with device drivers that use a lot
625 of stack the MDS (or possibly OSS) service threads could overflow
626 the stack. One change which was identified to consume additional
627 stack has been reworked to avoid the extra stack usage.
630 Frequency : occasional
632 Description: Do not fail import if osc_interpret_create gets -EAGAIN
633 Details : If osc_interpret_create got -EAGAIN it immediately exits and
634 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
635 oscc_has_objects and see OSC has no objests and call
636 oscc_internal_create to resend create request.
638 Severity : enhancement
640 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
642 Severity : enhancement
644 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
647 Frequency : very rare
649 Description: Fix replay if there is an un-replied request and open
650 Details : In some cases, older replay request will revert the
651 mcd->mcd_last_xid on MDS which is used to record the client's
654 Severity : enhancement
656 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
658 Severity : enhancement
660 Description: Add rhel5 support to HEAD.
662 Serverity : enhancement
664 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
669 Description: Don't allow skipping OSTs if index has been specified.
670 Details : Don't allow skipping OSTs if index has been specified, make
671 locking in internal create lots better.
675 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
676 Details : access to bitfield in structure is always rounded to long
677 and this produce problem with not atomic change any bit.
681 Description: Lustre make rpms failed.
682 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
683 builds Lustre rpms from tarball.
686 Frequency : rare on shutdown ost
688 Description: If llog cancel was not send before clean_exports phase, this can
689 produce deadlock in llog code.
690 Details : If llog thread has last reference to obd and call class_import_put
691 this produce deadlock because llog_cleanup_commit_master wait when
692 last llog_commit_thread exited, but this never success because was
693 called from llog_commit_thread.
697 Description: allow userland application know is lost one of stripes.
698 Details : fill lvb_blocks with error code on ost and return it to
699 application if error flag found.
703 Description: NULL lov_tgts causing MDS oops
704 Details : more safe checks for NULL lov_tgts for avoid oops.
706 Severity : enhancement
708 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
712 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
713 Details : If an OST is down the MDS will hang indefinitely in
714 obd_statfs_rqset() waiting for the statfs data. While for
715 MDS QOS usage of statfs, it should not stuck in waiting.
717 Severity : enhancement
719 Description: remote_acl support
720 Details : Support ACL-based permission check for remote user.
721 Support setfacl/getfacl for remote user with the utils
722 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
723 the system "{s,g}etfacl" utils.
725 Severity : enhancement
727 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
729 Severity : enhancement
731 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
735 Description: make mgs_setparam() handle fsname containing dash
736 Details : fsname containing a dash does not work with lctl conf_param
738 Severity : enhancement
740 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
742 Severity : enhancement
744 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
748 Frequency : rare, at shutdown
749 Description: access already free / zero obd_namespace.
750 Details : if client_disconnect_export was called without force flag set,
751 and exist connect request in flight, this can produce access to
752 NULL pointer (or already free pointer) when connect_interpret
753 store ocd flags in obd_namespace.
757 Frequency : only at startup
758 Description: not alloc memory with spinlock held.
759 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
760 if any spinlock held.
762 Severity : enhancement
764 Description: make lustre randomly fail allocating memory
765 Details : Make lustre randomly failed allocating memory for testing purpose.
767 Severity : enhancement
769 Description: lost problems with lov objid file
770 Details : Fixes some scability and access to not inited memory problems
771 in work with lov objdid file.
776 Description: lfs find does not continue on file error
777 Details : Continue other files processing when a file/dir is absent.
781 Description: Inconsistent usage of lustre_pack_reply()
782 Details : Standardize the usage of lustre_pack_reply() such that it
783 always generate a CERROR on failure.
786 Frequency : occasional
788 Description: MDS hang or stay in waiting lock
789 Details : If client receive lock with CBPENDING flag ldlm need send lock
790 cancel as separate rpc, to avoid situation when cancel request
791 can't processed due all i/o threads stay in wait lock.
795 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
799 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
803 Description: Update to SLES9 kernel-2.6.5-7.287.3.
807 Description: Update to RHEL4 latest kernel.
809 Severity : enhancement
811 Description: Build SLES10 patchless client fails
812 Details : The configure was broken by run ./configure with
813 --with-linux-obj=.... argument for patchless client. When the
814 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
815 can't search header adequately. Use absolute path such as
816 -I($LINUX)/include instead.
820 Description: interrupt oig_wait produce painc on resend.
821 Details : brw_redo_request can be used for resend requests from ptlrpcd and
822 private set, and this produce situation when rq_ptlrpcd_data not
823 copyed to new allocated request and triggered LBUG on assert
824 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
825 ptlrpcd set if request is changed and can be safety changed to use
830 Description: LASSERT_{REQ,REP}SWAB macros are buggy
831 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
832 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
833 these macros and replace them with their difinitions instead.
837 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
838 Details : Update which_patch & target file for SLES10 latest kernel.
841 Description: Remove client patches from SLES 10 kernel.
842 Details : This causes SLES 10 clients to behave as patchless clients
843 even on a Lustre-patched (server) kernel.
845 Severity : enhancement
847 Description: self-adjustable client's lru lists
848 Details : use adaptive algorithm for managing client cached locks lru
849 lists according to current server load, other client's work
850 pattern, memory activities, etc. Both, server and client
851 side namespaces provide number of proc tunables for controlling
854 Severity : enhancement
856 Description: light-weight GSS support
857 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
858 and reduce performance overhead.
860 Severity : enhancement
862 Description: Linux keyring support
863 Details : Support using service of Linux keyring for Lustre GSS internal
864 context refresh/cache mechanism.
868 Description: Fix errors in lfs documentation
869 Details : Fixes man pages
873 Description: don't use GFP_* in generic Lustre code.
874 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
878 Description: obdclass is limited by single OBD_ALLOC(idarray)
879 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
883 Description: warnings with build patchless client with vanila 2.6.19 and up
884 Details : change old ctl_table style and replace ctl_table/ctl_table_header
885 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
889 Frequency : Only for SLES
890 Description: sanity_quota fail test_1
891 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
892 file, which makes TSTID[2] inunique.
896 Frequency : Only for SLES9
897 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
898 Details : Update target/ChangeLog/which_patch .
902 Frequency : Only for test_10 in sanity-quota.sh
903 Description: a bug in quota test code
904 Details : checking a test flag in a wrong place
908 Frequency : Only for test_13 in sanity-quota.sh
909 Description: a bug in quota test script
910 Details : shouldn't check overfull allocated quota
914 Description: jbd statistics
915 Details : Port older jbd statistics patch for sles10
917 Severity : enhancement
919 Description: Add printing inode info into message about error in writepage.
924 Description: reply_lock_interpret crash due to race with it and lock cancel.
925 Details : Do not replay locks that are being cancelled. Do not reference
926 locks by their address during replay, just by their handle.
930 Frequency : When flocks are used.
931 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
932 Details : Flock locks might destroy just granted lock if it could be merged
933 with another existing flock, this is done in completion handler,
934 so teach ldlm_cli_enquque_fini that this is a valid case for
940 Description: Oops in read and write path when failing to allocate lock.
941 Details : Check if lock allocation failed and return error back.
945 Description: lstripe command fails for valid OST index
946 Details : The stripe offset is compared to lov->desc.ld_tgt_count
947 instead of lov->desc.ld_active_tgt_count.
951 Description: sanity.sh failed test 103
952 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
957 Description: df doesn't work properly if diskfs blocksize != 4K
958 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
962 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
963 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
965 Severity : enhancement
967 Description: Adaptive timeouts
968 Details : RPC timeouts adapt to changing server load and network
969 conditions to reduce resend attempts and improve recovery time.
973 Description: llapi_file_create() does not allow some changes
974 Details : add llapi_file_open() that allows specifying the mode and
975 open flags, and also returns an open file handle.
979 Description: merge and cleanup kernel patches.
980 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
984 Description: Add journal checksum support.(Kernel part)
985 Details : The journal checksum feature adds two new flags i.e
986 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
987 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
988 indicates that the commit block contains the checksum for
989 the blocks described by the descriptor blocks. Now commit
990 record can be sent to disk without waiting for descriptor
991 blocks to be written to disk. This behavior is controlled
992 using JBD2_FEATURE_ASYNC_COMMIT flag.
996 Description: OSS needs mutliple precreate threads
997 Details : Add ability to start more than one create thread per OSS.
1001 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1002 Details : execute lfs setstripe on client
1006 Description: mds_obd_create error creating tmp object
1007 Details : When the user sets quota on root, llog will be affected and can't
1008 create files and write files.
1011 Frequency : Always on ia64 patchless client, and possibly others.
1013 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1014 Details : This allows the patchless client to be loaded on architectures
1015 without this export.
1019 Description: RedHat Update kernel for RHEL5
1020 Details : Add the kernel config file for RHEL5.
1024 Description: RedHat Update kernel for RHEL5
1025 Details : Modify the kernel config file more closer RHEL5.
1029 Description: Build failure against Centos5 (RHEL5)
1030 Details : Define PAGE_SIZE when it isn't present.
1034 Description: test_7 on sanity-quota.sh can't be used on separate machines
1035 Details : add facet to handle it
1039 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1040 Details : with NFS, the anon dentry's parent was set to itself in
1041 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1042 corresponding dentry other than use rec->ur_name.
1044 Severity : enhancement
1046 Description: lfs setstripe enhancement
1047 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1051 Description: enable data checksumming by default
1052 Details : enable checksum by default, allow --disable-checksum
1053 configure option and "-o nochecksum" mount option. Checksums
1054 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1058 Description: lustre support for RHEL5
1059 Details : Add support for RHEL5.
1063 Description: Client eviction due to failover config
1064 Details : after a connection loss, the lustre client should attempt to
1065 reconnect to the last active server first before trying the
1066 other potential connections.
1070 Description: when mds and osts use different quota unit(32bit and 64bit),
1071 quota will be released repeatly.
1072 Details : void sending multiple quota reqs to mds, which will keep the
1073 status between the reqs.
1077 Description: osts not allocated evenly to files
1078 Details : change the condition to increase offset_idx
1081 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1083 Description: Data corruption for OSTs that are formatted larger than 2TB
1085 Details : When generating the bio request for lustre file writes the
1086 sector number would overflow a temporary variable before being
1087 used for the IO. The data reads correctly from Lustre (which
1088 will overflow in a similar manner) but other file data or
1089 filesystem metadata may be corrupted in some cases.
1093 Description: Tune the kernel for good SCSI performance.
1094 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1095 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1100 Description: rewrite ext2-derived code in obdclass/uuid.c
1101 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1102 from scratch preserving functionality.
1106 Description: "lfs find -obd UUID" prints directories
1107 Details : "lfs find -obd UUID" will return all directory names instead
1108 of just file names. It is incorrect because the directories
1109 do not reside on the OSTs.
1113 Description: use i_size_read and i_size_write in 2.6 port
1114 Details : replace inode->i_size access with i_size_read/write()
1117 Frequency : when removing large files
1119 Description: scheduling issue during removal of large Lustre files
1120 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1121 It causes scheduling issues when removing large files (17TB in the
1125 Frequency : only with liblustre clients on XT3
1127 Description: evictions taking too long
1128 Details : allow llrd to evict clients directly on OSTs
1131 Frequency : only on ppc
1133 Description: /proc/fs/lustre/devices broken on ppc
1134 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1135 We should make sure the type of 'index' is loff_t and then cast
1136 explicitly as needed below. Do not assign an explicitly cast
1140 Frequency : only for rhel5
1142 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1143 Details : Modify the target file & which_kernel.
1145 Severity : enhancement
1147 Description: omit set fsid for export NFS
1148 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1149 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1153 Frequency : Always, for kernels after 2.6.16
1154 Description: Fix warning idr_remove called for id=.. which is not allocated.
1155 Details : Last kernels save old s_dev before kill super and not allow
1156 to restore from callback - restore it before call kill_anon_super.
1160 Description: buffer overruns could theoretically occur
1161 Details : llapi_semantic_traverse() modifies the "path" argument by
1162 appending values to the end of the origin string, and a
1163 overrun may occur. Adding buffer overrun check in liblustreapi.
1167 Description: Fix error on 'ls .' at the top of the Lustre mount.
1168 Details : Don't revalidate dentry if it is a root dentry.
1172 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1173 Details : Modify vm-tunables-rhel4.patch.
1177 Description: Kernel config for 2.6.18-vanilla.
1178 Details : Modify targets/2.6-vanilla.target.in.
1179 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1180 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1181 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1182 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1186 Description: improve handling recoverable errors
1187 Details : if request processig with error which can be recoverable on server
1188 request should be resend, otherwise page released from cache and
1193 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1194 Details : Modify target file & which_patch.
1195 A flaw was found in the IA32 system call emulation provided
1196 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1197 value could be stored in the %RAX register, which could trigger an
1198 out-of-bounds system call table access. An untrusted local user
1199 could exploit this flaw to run code in the kernel
1200 (ie a root privilege escalation). (CVE-2007-4573).
1204 Description: O_DIRECT bypasses client statistics.
1205 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1206 read_ahead_stats were updated. copied the stats section from
1207 osc_send_oap_rpc() into async_internal().
1211 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1215 Description: change order of libsysio includes
1216 Details : '#include sysio.h' should always come before '#include xtio.h'
1220 Description: handle "serious error: objid * already exists" more gracefully
1221 Details : If LAST_ID value on disk is smaller than the objects existing in
1222 the O/0/d* directories, it indicates disk corruption and causes an
1223 LBUG(). If the object is 0-length, then we should use the existing
1224 object. This will help to avoid a full fsck in most cases.
1226 Severity : enhancement
1228 Description: adapt the lustre_config script to support the upgrade case
1229 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1230 server targets to 1.6 in parallel.
1234 Description: To avoid grant space > avaible space when the disk is almost
1235 full. Without this patch you might see the error "grant XXXX >
1236 available" or some LBUG about grant, when the disk is almost
1238 Details : In filter_check_grant, for non_grant cache write, we should
1239 check the left space by if (*left > ungranted + bytes), instead
1240 of (*left > ungranted), because only we are sure the left space
1241 is enough for another "bytes", then the ungrant space should be
1242 increase. In client, we should update cl_avail_grant only there
1243 is OBD_MD_FLGRANT in the reply.
1247 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1248 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1249 described in CVE-2007-4573. This problem could allow untrusted
1250 local users to gain root access.
1253 Frequency : when using O_DIRECT and quotas
1255 Description: Incorrect file ownership on O_DIRECT output files
1256 Details : block usage reported by 'lfs quota' does not take into account
1257 files that have been written with O_DIRECT.
1262 Description: touch file failed when fs is not full
1263 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1264 otherwise we can get ENOSP while fs is not full.
1268 Description: parallel lock callbacks
1269 Details : Instead of sending blocking and completion callbacks as separated
1270 requests, adding them to a set and sending in parallel.
1273 Frequency : only for Cray XT3
1274 Bugzilla : 12829/13455
1275 Description: Changing primary group doesn't change the group lustre assigns to
1277 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1278 overridden with the primary group provided by the group upcall,
1279 whereas the supplied fsgid can be trusted if it is in the list of
1280 supplementary groups returned by the group upcall.
1282 Severity : enhancement
1284 Description: Allow masking D_WARNING, D_ERROR messages from console
1285 Details : Console messages can now be disabled via lnet.printk.
1289 Description: User code with malformed file open parameter crashes client node
1290 Details : Before packing join_file req, all the related reference should be
1291 checked carefully in case some malformed flags cause fake
1292 join_file req on client.
1296 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1297 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1298 mode is equal to the granted mode. However, it should also check
1299 that LDLM_FL_CP_REQD is not set.
1303 Description: Heavy nfs access might result in deadlocks
1304 Details : After ELC code landed, it is now improper to enqueue any mds
1305 locks under och_sem, because enqueue might want to decide to
1306 cancel open locks for same inode we are holding och_sem for.
1310 Description: Client eviction while running blogbench
1311 Details : A lot of unlink operations with concurrent I/O can lead to a
1312 deadlock causing evictions. To address the problem, the number of
1313 oustanding OST_DESTROY requests is now throttled to
1314 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1315 callbacks are processed in priority.
1319 Description: enable ACLs on MDS by default
1320 Details : ACLs must be enabled on MDS by default.
1323 Frequency : PPC/PPC64 only
1325 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1326 Details : fix duplicated definitions between asm-ppc64/types.h and
1327 lustre_types.h on PPC.
1330 Frequency : PPC/PPC64 only
1332 Description: asm-ppc/segment.h does not exist
1333 Details : fix compile issue on PPC.
1337 Description: better handle error messages in extents code
1340 Frequency : RHEL4 only
1342 Description: mkfs is very slow on IA64/RHEL4
1343 Details : A performance regression has been discovered in the MPT Fusion
1344 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1345 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1346 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1348 Severity : enhancement
1350 Description: SNMP support enhancement
1351 Details : Adding total number of sampled request for an MDS node in snmp
1354 Severity : enhancement
1356 Description: Optimize ldlm waiting list processing for PR extent locks
1357 Details : When processing waiting list for read extent lock and meeting read
1358 lock that is same or wider to it that is not contended, skip
1359 processing rest of the list and immediatelly return current
1360 status of conflictness, since we are guaranteed there are no
1361 conflicting locks in the rest of the list.
1365 Description: Time out and refuse to reconnect
1366 Details : When the failover node is the primary node, it is possible
1367 to have two identical connections in imp_conn_list. We must
1368 compare not conn's pointers but NIDs, otherwise we can defeat
1369 connection throttling.
1373 Description: port llog fixes from b1_6 into HEAD
1374 Details : Port llog reference couting and some llog cleanups from b1_6
1375 (bug 10800) into HEAD, for protect from panic and access to already
1376 free llog structures.
1380 Description: Detect stride IO mode in read-ahead
1381 Details : When a client does stride read, read-ahead should detect that and
1382 read-ahead pages according to the detected stride pattern.
1386 Description: data checksumming impacts single node performance
1387 Details : add support for several checksum algorithm. Currently, only CRC32
1388 and Adler-32 are supported. The checksum type can be changed on
1389 the fly via /proc/fs/lustre/osc/*/checksum_type.
1393 Description: use adler32 for page checksums
1394 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1399 Description: build for x2 fails
1400 Details : fix compile issue on Cray systems.
1404 Description: Properly match for duplicate locks
1405 Details : Due to different lock order from skiplists code, we need to
1406 traverse entire list for now
1409 Frequency : only on PPC/SLES10
1411 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1412 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1413 be defined. Add a hack in mkfs_lustre.c to work around this compile
1418 Description: LASSERT on MDS when client holding flock lock dies
1419 Details : ldlm pool logic depends on number of granted locks equal to
1420 number of released locks which is not true for flock locks, so
1421 just exclude such locks from consideration.
1425 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1426 Details : Use fsfilt_sync() for both the whole filesystem sync and
1427 individual file sync to eliminate dangerous inode locking
1428 with I_LOCK that can lead to a deadlock.
1432 Description: performance in 1.6.3
1433 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1434 sure that 1MB requests are not fragmented by the block layer.
1436 Severity : enhancement
1438 Description: organize the server-side client stats on per-nid basis
1439 Details : Change the structure of stats under obdfilter and mds to
1447 The "uuid"s file would list the uuids of _active_ exports.
1448 And the clear entry is to clear all stats and stale nids.
1450 Severity : enhancement
1452 Description: eliminate client locks in face of contention
1453 Details : file contention detection and lockless i/o implementation
1454 for contended files.
1458 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1461 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1463 Description: MDS may lose file striping (and hence file data) in some cases
1464 Details : If there are additional extended attributes stored on the MDS,
1465 in particular ACLs, SELinux, or user attributes (if user_xattr
1466 is specified for the client mount options) then there is a risk
1467 of attribute loss. Additionally, the Lustre file striping
1468 needs to be larger than default (e.g. striped over all OSTs),
1469 and an additional attribute must be stored initially in the
1470 inode and then increase in size enough to be moved to the
1471 external attribute block (e.g. ACL growing in size) for file
1476 Description: skiplist implementation simplification
1477 Details : skiplists are used to group compatible locks on granted list
1478 that was implemented as tracking first and last lock of each
1479 lock group the patch changes that to using doubly linked lists
1483 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1484 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1485 because LASSERT on both the data supplied by a client, and the
1486 data on disk is dangerous and incorrect.
1488 Severity : enhancement
1490 Description: Slow truncate/writes to huge files at high offsets.
1491 Details : Directly associate cached pages to lock that protect those pages,
1492 this allows us to quickly find what pages to write and remove
1493 once lock callback is received.
1497 Description: more ldlm soft lockups
1498 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1499 starve other threads from the resource lock for a long time in
1500 case of long waiting queue, so change the debug level from
1501 D_OTHER to the less frequently used D_INFO.
1503 Severity : enhancement
1505 Description: add -gid, -group, -uid, -user options to lfs find
1509 Description: Hung threads in invalidate_inode_pages2_range
1510 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1511 once one is completed. As a result, some RPCs are stuck in the
1512 queue and are never sent.
1516 Description: filter threads hungs on waiting journal commit
1517 Details : Cleanup filter group llog code, then only filter group llog will
1518 be only created in the MDS/OST syncing process.
1522 Description: Procfs and llog threads access destoryed import sometimes.
1523 Details : Sync the import destoryed process with procfs and llog threads by
1524 the import refcount and semaphore.
1526 Severity : enhancement
1528 Description: openlock cache of b1_6 port to HEAD
1533 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1534 Details : If insertion of an extent fails, then discard the inode
1535 preallocation and free data blocks else it can lead to duplicate
1540 Description: don't always update ctime in ext3_xattr_set_handle()
1541 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1542 In some cases the ctime should not be updated, for example for
1543 2.0->1.8 compatibility it is necessary to delete an xattr and it
1544 should not update the ctime.
1548 Bugzilla : 15713/16362
1549 Description: Assertion in iopen_connect_dentry in 1.6.3
1550 Details : looking up an inode via iopen with the wrong generation number can
1551 populate the dcache with a disconneced dentry while the inode
1552 number is in the process of being reallocated. This causes an
1553 assertion failure in iopen since the inode's dentry list contains
1554 both a connected and disconnected dentry.
1558 Description: assertion failure in ldlm_handle2lock()
1559 Details : fix a race between class_handle_unhash() and class_handle2object()
1560 introduced in lustre 1.6.5 by bug 13622.
1565 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1566 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1571 Description: oops in sd_iostats_seq_show()
1572 Details : unloading/reloading the scsi low level driver triggers a kernel
1573 bug when trying to access the sd iostat file.
1578 Description: Kernel panics during QLogic driver reload
1579 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1580 patch, causing memory corruption.
1585 Description: journal_dev option does not work in b1_6
1586 Details : pass mount option during pre-mount.
1588 Severity : enhancement
1590 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1591 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1592 extent information of a file. It can be used to map logical blocks
1593 in a file to physical blocks in the block device.
1597 Description: LDLM soft lockups - improvement
1598 Details : It is be possible to send the lock handle along with each read
1599 or write request because the client is already doing a lock match
1600 itself so there isn't any reason the OST should have to re-do that
1604 Frequency : only on Cray X2
1606 Description: X2 build failures
1607 Details : fix build failures on Cray X2.
1611 Description: xid & resent requests
1612 Details : Initialize RPC XID from clock at startup (randomly if clock is
1615 Severity : enhancement
1617 Description: Add lustre_start utility to start or stop multiple Lustre servers
1622 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1623 Details : In case of memory pressure, list_del() can be called twice on
1624 req->rq_history_list, causing a kernel oops.
1628 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1629 Details : fix stack overflow in the distributed lock manager by defering
1630 export eviction after a failed AST to the elt thread instead of
1631 handling it in the dlm interpret routine.
1635 Description: Convert some comments to new format.
1636 Details : Update documenting comments to match doxygen conventions.
1640 Description: Grammar fixes.
1641 Details : A couple of trivial spelling fixes.
1645 Description: OSD_COUNTERS-mandatory
1646 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1647 to simplify the code.
1651 Description: simplify cmm_device freeing logic.
1652 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1656 Description: Add lockdep support to dt_object_operations locking interface.
1657 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1658 parameter indicating lock ordering. Update mdd code to use new
1663 Description: Introduce failloc constants for lockless IO tests.
1664 Details : Add two new failloc constants to test lockless IO. Only one of
1665 them in implemented---another is checked in yet to be landed
1670 Description: Add lockdep support for inode mutex.
1671 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1672 in the situations where Lustre has to lock more than one inode
1677 Description: Add optional invariants checking support.
1678 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1679 configure switch. This macro is to be used for consistency and
1680 sanity checks that are too expensive to be left in `production'
1685 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1686 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1691 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1692 Details : Use spin_lock_nested() in (the only) situation where more than
1693 one ldlm_lock is locked simultaneously. Also, fix possible
1694 dead-lock in ldlm_lock_change_resource() by enforcing particular
1699 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1700 Details : Instead of specifying each ldlm_lock call-back through separate
1701 parameter, wrap them into struct ldlm_callback_suite.
1705 Description: Kill join_lru obd method and its callers.
1706 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1707 in memory---a requirement implemented through ->o_join_lru() obd
1708 method in HEAD. Remove this method and its users.
1712 Description: Add asynchronous ldlm ENQUEUE completion handler.
1713 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1714 case that stock ldlm_completion_ast() cannot handle as it waits
1715 until lock is granted. Introduce new ldlm_completion_ast_async()
1716 for this. Also comment ldlm_completion_ast().
1720 Description: ldlm_error <-> errno conversion.
1721 Details : Add functions to map (rather arbitrary) between LDLM error codes
1722 and standard errno values. CLIO needs this to prevent LDLM specific
1723 constants from escaping ldlm and osc.
1727 Description: Kill unused ldlm_handle2lock_ns() function.
1728 Details : Kill unused ldlm_handle2lock_ns() function.
1732 Description: Add lu_ref support to ldlm_lock
1733 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1734 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1735 and ldlm_resource. LDLM interface has to be changed, because code that
1736 releases a reference on a lock, has to "know" what reference this is.
1737 In the most frequent case
1739 lock = ldlm_handle2lock(handle);
1741 LDLM_LOCK_PUT(lock);
1743 no changes are required. When any other reference (received _not_ from
1744 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1745 instead of LDLM_LOCK_PUT().
1747 Arguably, changes are pervasive, and interface requires some discipline
1748 for proper use. On the other hand, it was very instrumental in finding
1749 a few leaked lock references.
1753 Description: Add ldlm_lock_addref_try().
1754 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1755 attempts to addref a lock that might be being canceled
1760 Description: Add ldlm_weigh_callback().
1761 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1762 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1763 hard-coded `number of pages' logic.
1767 Description: Add lockdep annotations to llog code.
1768 Details : Use appropriately tagged _nested() locking calls in the places
1769 where llog takes more than one ->lgh_lock lock.
1773 Description: Add loi_kms_set().
1774 Details : Wrap kms updates into a helper function.
1778 Description: Constify instances of struct lsm_operations.
1779 Details : Constify instances of struct lsm_operations.
1783 Description: lu_conf support.
1784 Details : On a server, a file system object is uniquely identified
1785 by a fid, which is sufficient to locate and load all object
1786 state (inode). On a client, on the other hand, more data are
1787 necessary instantiate an object. Change lu_object_find() and
1788 friends to take additional `lu_conf' argument describing object.
1789 Typically this includes layout information.
1793 Description: lu_context fixes.
1794 Details : Introduce new lu_context functions that are needed on the client
1795 side, where some system threads (ptlrpcd) are shared by multiple
1796 modules, and so cannot be stopped during module shutdown.
1800 Description: Add start and stop methods to lu_device_type_operations.
1801 Details : Introduce two new methods in lu_device_type_operations, that are
1802 invoked when first instance of a given type is created and last one
1803 is destroyed respectively. This is need by CLIO.
1807 Description: Add lu_ref support to struct lu_device.
1808 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1809 track leaked references.
1813 Description: Introduce lu_kmem_descr.
1814 Details : lu_kmem_descr and its companion interface allow to create
1815 and destroy a number of kmem caches at once.
1819 Description: Fix lu_object finalization race.
1820 Details : Fix a race between lu_object_find() finding an object and its
1821 concurrent finalization. This race is (most likely) not possible
1822 on the server, but might happen on the client.
1826 Description: Introduce lu_ref interface.
1827 Details : lu_ref is a debugging module allowing to track references to
1828 a given object. It is quite cpu expensive, and has to be
1829 explicitly enabled with --enable-lu_ref. See usage description
1834 Description: Factor lu_site procfs stats into a separate function.
1835 Details : Separate lu_site stats printing code into a separate function
1836 that can be reused on a client.
1840 Description: Constify instances of struct {lu,dt,md}_device_operations.
1841 Details : Constify instances of struct {lu,dt,md}_device_operations.
1845 Description: Introduce struct md_site and move meta-data specific parts of
1846 struct lu_site here.
1847 Details : Move md-specific fields out of struct lu_site into special struct
1848 md_site, so that lu_site can be used on a client.
1852 Description: Kill mdd_lov_destroy().
1853 Details : Remove unused mdd code.
1857 Description: Add st_block checking to multistat.c.
1858 Details : Add st_block checking to multistat.c.
1862 Description: Add lu_ref support to struct obd_device.
1863 Details : Add lu_ref tracking to obd_device.
1867 Description: Kill obd_set_fail_loc().
1868 Details : Remove unused code.
1872 Description: Add special type for ptlrpc_request interpret functions.
1873 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1874 there. Actual usage will be in CLIO.
1878 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1879 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1880 doesn't work with lockdep.
1884 Description: Add rwv.c test program.
1885 Details : New testing program exercising readv(2) and writev(2) (Qian).
1889 Description: Add sendfile.c test program.
1890 Details : New testing program exercising sendfile(2) (Jay).
1894 Description: Ratelimit a message that can be very frequent.
1895 Details : Ratelimit a memory allocation failure message that can
1900 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1901 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1905 Description: Liblustre build fixes.
1906 Details : Liblustre build fixes.
1910 Description: libcfs: add cfs_{need,cond}_resched() interface.
1911 Details : libcfs: add cfs_{need,cond}_resched() definition and
1912 implementations for Linux, NT, and liblustre.
1914 Severity : enhancement
1916 Description: More exported tunables for mballoc
1917 Details : Add support for tunable preallocation window and new tunables for
1918 large/small requests
1922 Description: Detect corruption of block bitmap and checking for preallocations
1923 Details : Checks validity of on-disk block bitmap. Also it does better
1924 checking of number of applied preallocations. When corruption is
1925 found, it turns filesystem readonly to prevent further corruptions.
1929 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1930 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1931 window to 0. If it does hit cache miss, and read-pattern is still
1932 stride-io mode, does not reset the stride window, but also does not
1933 increase the stride window length in this case.
1937 Frequency : only for big-endian servers
1938 Description: Check if system is big-endian while mounting fs with extents feature
1939 Details : Mounting a filesystem with extents feature will fail on big-endian
1940 systems since ext3-based ldiskfs is not supported on big-endian
1941 systems. This can be over-riden with "bigendian_extents" mount option.
1943 Severity : enhancement
1945 Description: The root squash functionality
1946 Details : A security feature, which is to prevent users from being able
1947 to mount lustre on their desktop, run as root, and delete
1948 all of the files in the filesystem. The goal is accomplished by
1949 remapping user id (UID) and group id (GID) of the root user to
1950 a UID and GID specified by the system administartor via Lustre
1951 configuration management server (MGS). The functionality also
1952 allows to specify sets of clients for which the remapping does
1957 Description: Excessive recovery window
1958 Details : With AT enabled, the recovery window can be excessively long (6000+
1959 seconds). To address this problem, we no longer use
1960 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1961 timeout no longer depends on the service time, it is set to
1962 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1963 time via pb_service_time.
1967 Description: Watchdog triggered on MDS failover
1968 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
1969 the OSTs know that the MDS "UUID" can be reused for the same export
1970 from a different NID, so we do not need to wait for the export to be
1974 Frequency : rare, only if using MMP with Linux RAID
1976 Description: MMP doesn't work with Linux RAID
1977 Details : While using HA for Lustre servers with Linux RAID, it is possible
1978 that MMP will not detect multiple mounts. To make this work we
1979 need to unplug the device queue in RAID when the MMP block is being
1980 written. Also while reading the MMP block, we should read it from
1981 disk and not the cached one.
1983 Severity : enhancement
1985 Description: open file using fid
1986 Details : A file can be opened using just its fid, like
1987 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
1989 --------------------------------------------------------------------------------
1991 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1993 * Support for kernels:
1994 2.6.5-7.283 (SLES 9),
1995 2.6.9-55.EL (RHEL 4),
1996 2.6.16.46-0.14 (SLES 10),
1997 2.6.18.8 vanilla (kernel.org)
1998 * Client support for unpatched kernels:
1999 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2000 2.6.16 - 2.6.22 vanilla (kernel.org)
2001 * Due to recently discovered recovery problems, we do not recommend
2002 using patchless RHEL 4 clients with this or any earlier release.
2003 * Recommended e2fsprogs version: 1.39.cfs8
2004 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2005 * Starting with this release, the ldiskfs backing filesystem required
2006 by Lustre is now in its own package, lustre-ldiskfs. This package
2007 should be installed. It is versioned separately from Lustre and
2008 may be released separately in future.
2012 Description: block reactivating mgc import until all deactivates complete
2013 Details : Fix race when failing back MDT/MGS to itself (testing)
2016 Frequency : at statup only
2018 Description: mds_lov_synchronize race leads to various problems
2019 Details : simultaneous MDT->OST connections at startup can cause the
2020 sync to abort, leaving the OSC in a bad state.
2022 Severity : enhancement
2024 Description: add optional extra BUILD_VERSION info
2025 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2026 to override the lustre version.
2029 Frequency : 2.6.18 servers only
2031 Description: ll_kern_mount() doesn't release the module reference
2032 Details : The ldiskfs module reference count never drops down to 0
2033 because ll_kern_mount() doesn't release the module reference.
2038 Description: server LBUG when using old ost_num_threads parameter
2039 Details : Accept the old ost_num_threads parameter but warn that it
2040 is deprecated, and fix an off-by-one error that caused an LBUG.
2045 Description: Transient SCSI error results in persistent IO issue
2046 Details : iobuf->dr_error is not reinitialized to 0 between two
2050 Frequency : sometimes when underlying device returns I/O errors
2052 Description: OSTs not going read-only during write failures
2053 Details : OSTs are not remounted read-only when the journal commit threads
2054 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2055 instead of the ext3 wrappers.
2059 Description: poor connect scaling with increasing client count
2060 Details : Don't run filter_grant_sanity_check for more than 100 exports
2061 to improve scaling for large numbers of clients.
2064 Frequency : SLES10 only
2066 Description: sanity-quota.sh quotacheck failed: rc = -22
2067 Details : Quotas cannot be enabled on SLES10.
2070 Frequency : liblustre clients only
2072 Description: getdirentries does not give error when run on compute nodes
2073 Details : getdirentries does not fail when the size specified as an argument
2074 is too small to contain at least one entry
2076 Severity : enhancement
2078 Description: Add LNET router traceability for debug purposes
2079 Details : If a checksum failure occurs with a router as part of the
2080 IO path, the NID of the last router that forwarded the bulk data
2081 is printed so it can be identified.
2086 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2087 Details : Due to a race condition, liblustre clients were occasionally
2088 evicted incorrectly.
2090 Severity : enhancement
2092 Description: lfs setstripe use optional parameters instead of postional
2095 Severity : enhancement
2097 Description: Nanosecond timestamp support for ldiskfs
2098 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2099 resolution timestamps. There is not yet support for this at
2100 the Lustre filesystem level.
2103 Frequency : during server recovery
2105 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2106 Details : request with rq_no_resend flag not awake l_wait_event if they get
2110 Frequency : nfs export on patchless client
2112 Description: connectathon hang when test nfs export over patchless client
2113 Details : Disconnected dentry cannot be found with lookup, so we do not need
2114 to unhash it or make it invalid
2117 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2118 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2119 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2124 Description: open req refcounting wrong on reconnect
2125 Details : If reconnect happened between getting open reply from server and
2126 call to mdc_set_replay_data in ll_file_open, we will schedule
2127 replay for unreferenced request that we are about to free.
2128 Subsequent close will crash in variety of ways.
2129 Check that request is still eligible for replay in
2130 mdc_set_replay_data().
2135 Description: disable writes to filesystem when reading health_check file
2136 Details : the default for reading the health_check proc file has changed
2137 to NOT do a journal transaction and write to disk, because this
2138 can cause reads of the /proc file to hang and block HA state
2139 checking on a healthy but otherwise heavily loaded system. It
2140 is possible to return to the previous behaviour during configure
2141 with --enable-health-write.
2143 Severity : enhancement
2145 Description: 64-bit inode version
2146 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2147 to the inode. This will be required for version-based recovery.
2152 Description: MDS fails to start if a duplicate client export is detected
2153 Details : in some rare cases it was possible for a client to connect to
2154 an MDS multiple times. Upon recovery the MDS would detect this
2155 and fail during startup. Handle this more gracefully.
2157 Severity : enhancement
2159 Description: Add -o localflock option to simulate old noflock
2161 Details : This will achieve local-only flock/fcntl locks
2167 Description: log_commit_thread vs filter_destroy race leads to crash
2168 Details : Take import reference before releasing llog record semaphore
2173 Description: Wrong request locking in request set processing
2174 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2175 imp_delayed_list, in this place should be used imp_lock.
2178 Frequency : when reconnection
2180 Description: Grant Leak when osc reconnect to OST
2181 Details : When osc reconnect ost, OST(filter) should check whether it
2182 should grant more space to client by comparing fed_grant and
2183 cl_avail_grant, and return the granted space to client instead
2184 of "new granted" space, because client will call osc_init_grant
2185 to update the client grant space info.
2188 Frequency : when client reconnect to OST
2190 Description: Grant Leak when osc do resend and replay bulk write
2191 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2192 bulk write request, because the grant info while be sync between
2193 OSC and OST when reconnect, and we should ignore the grant info
2194 these of resend/replay write req.
2199 Description: Grant space more than avaiable left space sometimes.
2200 Details : When then OST is about to be full, if two bulk writing from
2201 different clients came to OST. Accord the avaliable space of the
2202 OST, the first req should be permitted, and the second one
2203 should be denied by ENOSPC. But if the seconde arrived before
2204 the first one is commited. The OST might wrongly permit second
2205 writing, which will cause grant space > avaiable space.
2208 Frequency : when client is evicted
2210 Description: Grant might be wrongly erased when osc is evicted by OST
2211 Details : when the import is evicted by server, it will fork another
2212 thread ptlrpc_invalidate_import_thread to invalidate the
2213 import, where the grant will be set to 0. While the original
2214 thread will update the grant it got when connecting. So if
2215 the former happened latter, the grant will be wrongly errased
2216 because of this race.
2221 Description: Checking Stale with correct fid
2222 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2223 to check whether it is stale, because sometimes, we want the
2224 enqueue happened anyway, and op_data.fid2 will not be initialized.
2226 Severity : enhancement
2228 Description: update patchless client
2229 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2232 Frequency : only with 2.4 kernel
2234 Description: random memory corruption
2235 Details : size of struct ll_inode_info is to big for union inode.u and this
2236 can be cause of random memory corruption.
2241 Description: Memory leak in recovery
2242 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2243 It should also check obd_fail before fsfilt_start, otherwise if
2244 fsfilt_start return -EROFS,(failover mds during mds recovery).
2245 then the req will return with repmsg->transno = 0 and rc = EROFS.
2246 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2247 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2248 be freed no matter whether fsfilt_commit success or not.
2251 Frequency : only with huge count clients
2253 Description: Prevents from taking the superblock lock in llap_from_page for
2255 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2256 for llap_from_page call in ll_removepage prevents from taking the
2257 superblock lock for a soon died page.
2262 Description: Not check open intent error before release open handle
2263 Details : in some rare cases, the open intent error is not checked before
2264 release open handle, which may cause
2265 ASSERTION(open_req->rq_transno != 0), because it tries to release
2266 the failed open handle.
2271 Description: Set cat log bitmap only after create log success.
2272 Details : in some rare cases, the cat log bitmap is set too early. and it
2273 should be set only after create log success.
2277 Description: Accessing a block bevice can re-enable I/O when Lustre is
2278 tearing down a device.
2279 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2284 Description: service threads may hog cpus when there are a lot of requests
2285 Details : Insert cond_resched to give other threads a chance to use some CPU
2290 Description: the cat log was not initialized in recovery
2291 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2292 unlink log on mds will not be initialized until mds post
2293 recovery. And also in mds post recovery, the unlink log will
2294 initialization will be done asynchronausly, so there will be race
2295 between add unlink log and unlink log initialization.
2299 Description: brw_stats were being printed incorrectly
2300 Details : brw_stats were being printed as log2 but all of them were not
2301 recorded as log2. Also remove some code duplication arising from
2302 filter_tally_{read,write}.
2306 Frequency : rare, only in recovery.
2307 Description: ASSERTION(req->rq_type != LI_POISON) failed
2308 Details : imp_lock should be held while iterating over imp_sending_list for
2309 prevent destroy request after get timeout in ptlrpc_queue_wait.
2313 Description: replay-single.sh test 52 fails
2314 Details : A lock's skiplist need to be cleanup when it being unlinked
2315 from its resource list.
2319 Description: Short directio read returns full requested size rather than
2321 Details : Direct I/O operations should return actual amount of bytes
2322 transferred rather than requested size.
2324 Severity : enhancement
2326 Description: metadata RPC reduction (e.g. for rm performance)
2327 Details : decrease the amount of synchronous RPC between clients and servers
2328 by canceling conflicing lock before the operation on the client
2329 and packing thier handles into the main operation RPC to server.
2331 Severity : enhancement
2333 Description: Async OSC create to avoid the blocking unnecessarily.
2334 Details : If a OST has no remain object, system will block on the creating
2335 when need to create a new object on this OST. Now, ways use
2336 pre-created objects when available, instead of blocking on an
2337 empty osc while others are not empty. If we must block, we block
2338 for the shortest possible period of time.
2342 Description: improve handling recoverable errors
2343 Details : if request processig with error which can be recoverable on server
2344 request should be resend, otherwise page released from cache and
2347 Severity : enhancement
2349 Description: refine locking for avoid write wrong info into lov_objid file
2350 Details : fix possible races with add new target and write/update data in
2353 --------------------------------------------------------------------------------
2355 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2360 Frequency : on some architectures
2362 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2363 Details : Uninitialized flags sometimes cause configuration commands to
2367 Frequency : patchless clients only
2369 Description: missing __iget() symbol export
2370 Details : The __iget() symbol export is missing. To avoid the need for
2371 this on patchless clients the deathrow inode reaper is turned
2372 off, and we depend on the VM to clean up old inodes. This
2373 dependency was during via the fix for bug 12181.
2375 --------------------------------------------------------------------------------
2377 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2379 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2380 INTEROPERATE with older versions automatically. In many cases a
2381 special upgrade step is needed. Please read the
2382 user documentation before upgrading any part of a live system.
2383 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2384 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2385 * WARNING: Lustre configuration and startup changes are required with
2386 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2390 Severity : enhancement
2392 Description: MountConf
2393 Details : Lustre configuration is now managed via mkfs and mount
2394 commands instead of lmc and lconf. New obd types (MGS, MGC)
2395 are added for dynamic configuration management. See
2396 https://mail.clusterfs.com/wikis/lustre/MountConf for
2399 Severity : enhancement
2401 Description: dynamic OST addition
2402 Details : OSTs can now be added to a live filesystem
2404 Severity : enhancement
2406 Description: startup order invariance
2407 Details : MDTs and OSTs can be started in any order. Clients only
2408 require the MDT to complete startup.
2410 Severity : enhancement
2412 Description: parallel, asynchronous orphan cleanup
2413 Details : orphan cleanup is now performed in separate threads for each
2414 OST, allowing parallel non-blocking operation.
2416 Severity : enhancement
2418 Description: optimized stripe assignment
2419 Details : stripe assignments are now made based on ost space available,
2420 ost previous usage, and OSS previous usage, in order to try
2421 to optimize storage space and networking resources.
2423 Severity : enhancement
2425 Description: Permanently set tunables
2426 Details : All writable /proc/fs/lustre tunables can now be permanently
2427 set on a per-server basis, at mkfs time or on a live system.
2429 Severity : enhancement
2431 Description: Lustre message v2
2432 Details : Add lustre message format v2.
2434 Severity : enhancement
2436 Description: client OST exclusion list
2437 Details : Clients can be started with a list of OSTs that should be
2438 declared "inactive" for known non-responsive OSTs.
2442 Description: SPEC SFS validation failure on NFS v2 over lustre.
2443 Details : Changes the blocksize for regular files to be 2x RPC size,
2444 and not depend on stripe size.
2446 Severity : enhancement
2448 Description: Multiple MD RPCs in flight.
2449 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2450 To avoid overly-overloading MDS, introduce a limit on number of
2451 MDS RPCs in flight for a single client and add /proc controls
2452 to adjust this limit.
2454 Severity : enhancement
2456 Description: client read/write statistics
2457 Details : Add client read/write call usage stats for performance
2458 analysis of user processes.
2459 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2460 file access. extents_stats shows chunk size distribution.
2461 extents_stats_per_process show chunk size distribution per
2464 Severity : enhancement
2466 Description: mds statistics
2467 Details : Add detailed mds operations statistics in
2468 /proc/fs/lustre/mds/*/stats.
2472 Description: Failure of copying files with lustre special EAs.
2473 Details : Client side always return success for setxattr call for lustre
2474 special xattr (currently only "trusted.lov").
2479 Description: Refcount LNET uuids
2480 Details : The global LNET uuid list grew linearly with every startup;
2481 refcount repeated list entries instead of always adding to
2484 Severity : enhancement
2486 Description: Dynamic service threads
2487 Details : Within a small range, start extra service threads
2488 automatically when the request queue builds up.
2491 Frequency : mixed-endian client/server environments
2493 Description: mixed-endian crashes
2494 Details : The new msg_v2 system had some failures in mixed-endian
2497 Severity : enhancement
2499 Description: Easy OST removal
2500 Details : OSTs can be permanently deactivated with e.g. 'lctl
2501 conf_param lustre-OST0001.osc.active=0'
2503 Severity : enhancement
2505 Description: MGS proc entries
2506 Details : Added basic proc entries for the MGS showing what filesystems
2509 Severity : enhancement
2511 Description: provide MGS failover
2512 Details : Added config lock reacquisition after MGS server failover.
2514 Severity : enhancement
2516 Description: add Linux 2.4 support
2517 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2521 Description: a large application tries to do I/O to the same resource and dies
2522 in the middle of it.
2523 Details : Check the req->rq_arrival time after the call to
2524 ost_brw_lock_get(), but before we do anything about
2525 processing it & sending the BULK transfer request. This
2526 should help move old stale pending locks off the queue as
2527 quickly as obd_timeout.
2530 Frequency : when an incorrect nid is specified during startup
2532 Description: ptlrpc connect to non-existant node causes kernel crash
2533 Details : LNET can't be re-entered from an event callback, which
2534 happened when we expire a message after the export has been
2535 cleaned up. Instead, hand the zombie cleanup off to another
2538 Severity : enhancement
2540 Description: plain/inodebits lock performance improvement
2541 Details : Grouping plain/inodebits in granted list by their request modes
2542 and bits policy, thus improving the performance of search through
2546 Frequency : only if OST filesystem is corrupted
2548 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2549 Details : for a short time RPCs with bulk IO are in the replay list,
2550 but replay of bulk IOs is unimplemented. If the OST filesystem
2551 is corrupted due to disk cache incoherency and then replay is
2552 started it is possible to trip an assertion. Avoid putting
2553 committed RPCs into the replay list at all to avoid this issue.
2556 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2558 Description: System hang on startup
2559 Details : This bug allowed the liblustre (e.g. catamount) client to
2560 return to the app before handling all startup RPCs. This
2561 could leave the node unresponsive to lustre network traffic
2562 and manifested as a server ptllnd timeout.
2564 Severity : enhancement
2566 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2567 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2568 boolean to control whether to print peer debug info when a
2569 client's RPC times out.
2572 Frequency : only for kernels with patches from Lustre below 1.4.3
2574 Description: Remove old rdonly API
2575 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2578 Frequency : only for devices with external journals
2580 Description: Set external device read-only also
2581 Details : During a commanded failover stop, we set the disk device
2582 read-only while the server shuts down. We now also set any
2583 external journal device read-only at the same time.
2586 Frequency : when upgrading from 1.4 while trying to change parameters
2588 Description: The wrong (new) MDC name was used when setting parameters for
2589 upgraded MDT's. Also allows changing of OSC (and MDC)
2590 parameters if --writeconf is specified at tunefs upgrade time.
2593 Frequency : when setting specific ost indicies
2595 Description: QOS code breaks on skipped indicies
2596 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2597 created with --index need not be sequential.
2599 Severity : enhancement
2601 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2602 Details : The uninit_groups feature works in conjunction with the kernel
2603 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2604 up the pass1 processing of e2fsck. This is a read-only feature
2605 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2606 mount filesystems that have had this feature enabled.
2608 Severity : enhancement
2610 Description: Improve multi-block allocation algorithm to avoid fragmentation
2611 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2612 allocation locality and avoid filesystem fragmentation.
2614 ------------------------------------------------------------------------------
2616 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2618 * Support for kernels:
2619 2.4.21-47.0.1.EL (RHEL 3)
2620 2.6.5-7.283 (SLES 9)
2621 2.6.9-42.0.10.EL (RHEL 4)
2622 2.6.12.6 vanilla (kernel.org)
2623 2.6.16.27-0.9 (SLES 10)
2624 * Recommended e2fsprogs version: 1.39.cfs5
2626 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2630 Frequency : occasional, depends on client load and configuration
2631 Bugzilla : 12181, 12203
2632 Description: data loss for recently-modified files
2634 Details : In some cases it is possible that recently written or created
2635 files may not be written to disk in a timely manner (this should
2636 normally be within 30s unless client IO load is very high).
2637 The problem appears as zero-length files or files that are a
2638 multiple of 1MB in size after a client crash or client eviction
2639 that are missing data at the end of the file.
2641 This problem is more likely to be hit on clients where files are
2642 repeatedly created and unlinked in the same directory, clients
2643 have a large amount of RAM, have many CPUs, the filesystem has
2644 many OSTs, the clients are rebooted frequently, and/or the files
2645 are not accessed by other nodes after being written.
2647 The presence of the problem can be detected by looking at
2648 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2649 smaller than the second (nr_unused) then dirty files will not
2650 be flushed automatically to disk. "sync; sleep 10" should be
2651 run several times on the node before unmounting it to update
2652 Lustre (this is also safe to run on nodes without this problem).
2654 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2655 that can cause this same problem, so customers using that kernel
2656 also need to update the kernel in addition to Lustre. In order
2657 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2659 It is normal that files written just before a client crash (less
2660 than 30s) may not yet have been flushed to disk, even for local
2664 Frequency : frequent on thin XT3 nodes
2666 Description: UUID collision on thin XT3 Linux nodes
2667 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2668 randomly, since we relied on an insufficiently-seeded PRNG.
2673 Description: OSS hangs after "All ost request buffers busy"
2674 Details : A deadlock between quota and journal operations caused OSS
2675 hangs after printing "All ost request buffers busy."
2678 Frequency : always on liblustre builds
2680 Description: Cleanup compiler warnings on liblustre
2683 Frequency : always on liblustre builds on XT3
2685 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2689 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2690 > 32000 subdirectories
2691 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2692 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2693 subdirectory count crosses 32000. This will aid e2fsck to
2694 correctly handle more than 32000 subdirectories.
2697 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2699 Description: System hang on startup
2700 Details : This bug allowed the liblustre (e.g. catamount) client to
2701 return to the app before handling all startup RPCs. This
2702 could leave the node unresponsive to lustre network traffic
2703 and manifested as a server ptllnd timeout.
2705 Severity : enhancement
2707 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2708 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2709 boolean to control whether to print peer debug info when a
2710 client's RPC times out.
2715 Description: make O_SYNC working on 2.6 kernels
2716 Details : 2.6 kernels use different method for mark pages for write,
2717 so need add a code to lustre for O_SYNC work.
2722 Description: Failure to close file and release space on NFS
2723 Details : Put inode details into lock acquired in ll_intent_file_open.
2724 Use mdc_intent_lock in ll_intent_open to properly
2725 detect all kind of errors unhandled by mdc_enqueue.
2730 Description: proc file read during shutdown sometimes raced obd removal,
2732 Details : Add lock to prevent obd access after proc file removal.
2735 Frequency : Only for files larger than 4GB on 32-bit clients.
2737 Description: improperly doing page alignment of locks
2738 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2739 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2744 Description: RPCs being resent when they shouldn't be
2745 Details : Some RPCs that should not be resent are being resent. This
2746 can cause inconsistencies in the RPC state machine. Do not
2747 resend such requests.
2750 Frequency : rare, only with NFS export
2752 Description: Crash on NFS re-export node
2753 Details : under very unusual load conditions an assertion is hit in
2754 ll_intent_file_open()
2757 Frequency : only if OST filesystem is corrupted
2759 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2760 Details : for a short time RPCs with bulk IO are in the replay list,
2761 but replay of bulk IOs is unimplemented. If the OST filesystem
2762 is corrupted due to disk cache incoherency and then replay is
2763 started it is possible to trip an assertion. Avoid putting
2764 committed RPCs into the replay list at all to avoid this issue.
2769 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2770 Details : Large single O_DIRECT read and write calls can fail to allocate
2771 a sufficiently large buffer to process the request. In case of
2772 allocation failure the allocation is retried with a smaller
2773 buffer and broken into smaller requests.
2775 Severity : enhancement
2777 Description: Add -o localflock option to simulate old noflock behaviour.
2778 Details : This will achieve local-only flock/fcntl locks coherentness.
2783 Description: versioning check is incomplete
2784 Details : Checking the version difference of client vs. server, report
2785 error if the gap is too big.
2790 Description: add support PG_writeback bit
2791 Details : add support for PG_writeback bit for Lustre, for more carefull
2792 work with page cache in 2.6 kernel. This also fix some deadlocks
2793 and remove hack for work O_SYNC with 2.6 kernel.
2795 Severity : enhancement
2797 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2798 Details : The uninit_groups feature works in conjunction with the kernel
2799 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2800 up the pass1 processing of e2fsck. This is a read-only feature
2801 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2802 mount filesystems that have had this feature enabled.
2804 Severity : enhancement
2806 Description: Improve multi-block allocation algorithm to avoid fragmentation
2807 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2808 allocation locality and avoid filesystem fragmentation.
2810 ------------------------------------------------------------------------------
2812 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2814 * Support for kernels:
2815 2.6.9-42.0.3.EL (RHEL 4)
2816 2.6.5-7.276 (SLES 9)
2817 2.4.21-47.0.1.EL (RHEL 3)
2818 2.6.12.6 vanilla (kernel.org)
2819 2.6.16.21-0.8 (SLES10)
2820 * Recommended e2fsprogs version: 1.39.cfs2-0
2822 * The backwards-compatible /proc/sys/portals symlink has been removed
2823 in this release. Before upgrading, please ensure that you change
2824 any configuration scripts or /etc/sysctl.conf files that access
2825 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2826 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2827 in advance of the upgrade on any system running Lustre 1.4.6 or
2828 newer, since /proc/sys/lnet was added in that version.
2829 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2833 Frequency : only when quota is used
2835 Description: avoid scanning export list for quota master
2836 Details : Change the algorithms to avoid scanning export list in order
2837 to improve the efficiency.
2840 Frequency : MDS failover only, very rarely
2842 Description: "went back in time" messages on mds failover
2843 Details : The greatest transno may be lost when the current operation
2844 finishes with an error (transno==0) and the client's last_rcvd
2845 record is over-written. Save the greatest transno in the
2846 mds_last_transno for this case.
2849 Frequency : always for specific kernels and striping counts
2851 Description: client may get "Matching packet too big" without ACL support
2852 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2853 when trying to access files in certain configurations. The
2854 clients should in fact be denied when mounting because they do
2855 not understand ACLs.
2858 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2860 Description: many clients connecting with IO in progress causes connect timeouts
2861 Details : Avoid synchronous journal commits to avoid delays caused by many
2862 clients connecting/disconnecting when bulk IO is in progress.
2863 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2864 OST_IO_PORTAL to avoid delays behind potentially many pending
2868 Frequency : occasionally with multiple writers to a single file
2870 Description: shared writes to file may result in wrong size reported by stat()
2871 Details : Allow growing of kms when extent lock is cancelled
2874 Frequency : always with random mmap IO to multi-striped file
2876 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2877 Details : Only if the hole is at the end of OST object so that kms is too
2878 small. Fix is to increase kms accordingly in ll_nopage.
2881 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2883 Description: writes to a missing object would leak memory on the OST
2884 Details : If there is an inconsistency between the MDS and OST filesystems,
2885 such that the MDS references an object that doesn't exist, writes
2886 to that object will leak memory due to incorrect cleanup in the
2887 error handling path, eventually running out of memory on the OST.
2892 Description: Creating too long symlink causes lustre errors
2893 Details : Check symlink and name lengths before sending requests to MDS.
2896 Frequency : only if flock is enabled (not on by default)
2898 Description: posix locks not released on fd closure on 2.6.9+
2899 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2900 kernels, this caused such locks not to be released on fd close and
2901 then assertions on fs unmount about still used locks.
2904 Frequency : MDS failover only, very rarely
2906 Description: clients may get ASSERTION(granted_lock != NULL)
2907 Details : When request was taking a long time, and a client was resending
2908 a getattr by name lock request. The were multiple lock requests
2909 with the same client lock handle and
2910 mds_getattr_name->fixup_handle_for_resent_request found one of the
2911 lock handles but later failed with ASSERTION(granted_lock != NULL).
2916 Description: handle->h_buffer_credits > 0, assertion failure
2917 Details : h_buffer_credits is zero after truncate, causing assertion
2918 failure. This patch extends the transaction or creates a new
2922 Frequency : NFS re-export or patchless client
2923 Bugzilla : 11179, 10796
2924 Description: Crash on NFS re-export node (__d_move)
2925 Details : We do not want to hash the dentry if we don't have a lock.
2926 But if this dentry is later used in d_move, we'd hit uninitialised
2927 list head d_hash, so we just do this to init d_hash field but
2928 leave dentry unhashed.
2931 Frequency : NFS re-export or patchless client
2933 Description: NFS exports has problem with symbolic link
2934 Details : lustre client didn't properly install dentry when re-exported
2935 to NFS or running patchless client.
2938 Frequency : NFS re-export or patchless client
2940 Description: Various nfs/patchless fixes.
2941 Details : fixes reuse disconected alias for lookup process - this fixes
2942 warning "find_exported_dentry: npd != pd",
2943 fix permission error with open files at nfs.
2944 fix apply umask when do revalidate.
2947 Frequency : occasional
2949 Description: Crash on NFS re-export node
2950 Details : calling clear_page() on the wrong pointer triggered oops in
2951 generic_mapping_read().
2954 Frequency : rarely, using O_DIRECT IO
2956 Description: unaligned directio crashes client with LASSERT
2957 Details : check for unaligned buffers before trying any requests.
2960 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2962 Description: stale data returned from RAID cache
2963 Details : If only a small amount of IO is done to the RAID device before
2964 reading it again it is possible to get stale data from the RAID
2965 cache instead of reading it from disk.
2968 Frequency : always for sles10 kernel
2970 Description: sles10 support
2971 Details : ll_follow_link: compile fixes and using of nd_set_link
2972 under newer kernels.
2975 Frequency : depends on arch, kernel and compiler version, always on sles10
2978 Description: recursive or deep enough symlinks cause stack overflow
2979 Details : getting rid of large stack-allocated variable in
2983 Frequency : depends on hardware
2985 Description: lustre write performance loss in the SLES10 kernel
2986 Details : the performance loss is caused by using of write barriers in the
2987 ext3 code. The SLES10 kernel turns barrier support on by
2988 default. The fix is to undo that change for ldiskfs.
2990 ------------------------------------------------------------------------------
2992 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2994 * Support for kernels:
2995 2.6.9-42.0.3EL (RHEL 4)
2996 2.6.5-7.276 (SLES 9)
2997 2.4.21-47.0.1.EL (RHEL 3)
2998 2.6.12.6 vanilla (kernel.org)
3002 Frequency : quota enabled and large files being deleted
3004 Description: releasing more than 4GB of quota at once hangs OST
3005 Details : If a user deletes more than 4GB of files on a single OST it
3006 will cause the OST to spin in an infinite loop. Release
3007 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3012 Description: statfs data retrieved from /proc may be stale or zero
3013 Details : When reading per-device statfs data from /proc, in the
3014 {kbytes,files}_{total,free,avail} files, it may appear
3015 as zero or be out of date.
3018 Frequency : systems with MD RAID1 external journal devices
3020 Description: lconf's call to blkid is confused by RAID1 journal devices
3021 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3022 of returning all block devices that match the journal UUID.
3025 Frequency : always, for aggregate stripe size over 4GB
3027 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3028 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3029 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3032 Frequency : always if "lfs find" used on a local file/directory
3034 Description: "lfs find" segfaults if used on a local file/directory
3035 Details : The case where a directory component was not specified wasn't
3036 handled correctly. Handle this properly.
3039 Frequency : always on ppc64
3041 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3042 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3043 perform endianness conversion.
3046 Frequency : rarely (truncate to non-zero file size after write under load)
3047 Bugzilla : 10730, 10687
3048 Description: Files padded with zeros to next 4K multiple
3049 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3050 it is possible that files that are truncated to a non-zero size
3051 immediately after being written are filled with zero bytes beyond
3052 the truncated size. No file data is lost.
3054 Severity : enhancement
3056 Description: Allow recovery/failover for liblustre clients.
3057 Details : liblustre clients were unaware of failover configurations until
3060 Severity : enhancement
3062 Description: user file locks should fail when not mounting with flock option
3063 Details : Set up an error-returning stub in ll_file_operations.lock field
3064 to prevent incorrect behaviour when client is mounted without
3065 flock option. Also, set up properly f_op->flock field for
3069 Frequency : always on ia64
3071 Description: "lfs df" loops on printing out MDS statfs information
3072 Details : The obd_ioctl_data was not initialized and in some systems
3073 this caused a failure during the ioctl that did not return
3074 an error. Initialize the struct and return an error on failure.
3077 Frequency : SLES 9 only
3079 Description: Error of copying files with lustre special EAs as root
3080 Details : Client side always return success for setxattr call for lustre
3081 special xattr (currently only "trusted.lov").
3084 Frequency : rarely on clusters with both ia64+i386 clients
3086 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3087 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3088 clients are doing concurrent writes to the same file it is
3089 possible that the ia64 clients may LASSERT because the OST
3090 extent locks are not PAGE_SIZE aligned. Ensure that grown
3091 locks are always aligned on the request boundary.
3094 Frequency : specific use, occasional
3096 Description: Overwriting in use executable truncates on-disk binary image
3097 Details : If one node attempts to overwrite an executable in use by
3098 another node, we now correctly return ETXTBSY instead of
3099 truncating the file.
3101 Severity : enhancement
3103 Description: Async OSC create to avoid the blocking unnecessarily.
3104 Details : If a OST has no remain object, system will block on the creating
3105 when need to create a new object on this OST. Now, ways use
3106 pre-created objects when available, instead of blocking on an
3107 empty osc while others are not empty. If we must block, we block
3108 for the shortest possible period of time.
3113 Description: chmod on Lustre root is propagated to other clients
3114 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3115 invalid by the follow_mount time.
3118 Frequency : liblustre clients only
3120 Description: Race in 'instant cancel' lock handling could lead to such locks
3121 never to be granted in case of SMP MDS
3122 Details : Do not destroy not yet granted but cbpending locks in
3126 Frequency : replay/resend of open
3128 Description: non null lock assetion failure in mds_intent_policy
3129 Details : Trying to replay/resend lockless open requests resulted in
3130 mds_open() returning 0 with no lock. Now it sets a flag if
3131 a lock is going to be returned.
3133 Severity : enhancement
3135 Description: Checksum enhancements
3136 Details : New checksum enhancements allow for resending RPCs that failed
3139 Severity : enhancement
3141 Description: Tunables on number of dirty pages in cacche
3142 Details : Allow to set limit on number of dirty pages cached.
3147 Description: client crash on unmount - lock still has references
3148 Details : In some error handling cases it was possible to leak a lock
3149 reference on a client while accessing a file. This was not
3150 harmful to the client during operation, but would cause the
3151 client to crash when the filesystem is unmounted.
3154 Frequency : specific case, rare
3156 Description: ETXTBSY on mds though file not in use by client
3157 Details : ETXTBSY is no longer incorrectly returned when attempting to
3158 chmod or chown a directory that the user previously tried to
3159 execute or a currently-executing binary.
3162 Frequency : extremely rare except on liblustre-based clients
3164 Description: Lustre space not freed when files are deleted
3165 Details : Clean up open-unlinked files after client eviction. Previously
3166 the unlink was skipped and the files remained as orphans.
3171 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3172 Details : In some cases it was possible to send a blocking callback to
3173 a client doing a glimpse, even though that client didn't get
3174 a lock granted. When the glimpse lock is cancelled on the OST
3175 the freed lock is left on the waiting list and corrupted the list.
3178 Frequency : all core dumps
3180 Description: Broke core dumps to lustre
3181 Details : Negative dentry may be unhashed if parent does not have UPDATE
3182 lock, but some callers, e.g. do_coredump, expect dentry to be
3183 hashed after successful create, hash it in ll_create_it.
3185 ------------------------------------------------------------------------------
3187 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3189 * Support for kernels:
3190 2.6.9-42.0.2.EL (RHEL 4)
3191 2.6.5-7.276 (SLES 9)
3192 2.4.21-40.EL (RHEL 3)
3193 2.6.12.6 vanilla (kernel.org)
3197 Frequency : always on RHEL 3
3199 Description: Number of open files grows over time
3200 Details : The number of open files grows over time, whether or not
3201 Lustre is started. This was due to a filp leak introduced
3202 by one of our kernel patches.
3204 ------------------------------------------------------------------------------
3206 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3208 * Support for kernels:
3209 2.6.9-42.EL (RHEL 4)
3210 2.6.5-7.276 (SLES 9)
3211 2.4.21-40.EL (RHEL 3)
3212 2.6.12.6 vanilla (kernel.org)
3217 Bugzilla : 5719, 9635, 9792, 9684
3218 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3219 Details : If a server is under heavy load and cannot reply to new
3220 connection requests before the client resends the (re)connect,
3221 the connection handling code can behave badly if two service
3222 threads are concurrently handing separate (re)connections from
3223 the same client. Add better locking to the connection handling
3224 code, and ensure that only a single connection will be processed
3225 for a given client UUID, even if the lock is dropped.
3227 Severity : enhancement
3229 Description: add TCP zero-copy support to kernel
3230 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3231 sends if the hardware supports scatter-gather and checksumming.
3232 This allows socklnd to do client-write and server-read more
3233 efficiently and reduce CPU utilization from skbuf copying.
3236 Frequency : only if NFS exporting from client
3238 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3239 Details : If setting attributes on a file created under NFS that had
3240 never been opened it would be possible to oops the client
3241 if the file had no objects.
3244 Frequency : always for liblustre
3246 Description: liblustre client does MDS+OSTs setattr RPC for each write
3247 Details : When doing a write from a liblustre client, the client
3248 incorrectly issued an RPC to the MDS and each OST the file was
3249 striped over in order to update the timestamps. When writing
3250 with small chunks and many clients this could overwhelm the MDS
3251 with RPCs. In all cases it would slow down the write because
3252 these RPCs are unnecessary.
3254 Severity : enhancement
3256 Description: allow number of MDS service threads to be changed at module load
3257 Details : It is now possible to change the number of MDS service threads
3258 running. Adding "options mds mds_num_threads={N}" to the MDS's
3259 /etc/modprobe.conf will set the number of threads for the next
3260 time Lustre is restarted (assuming the "mds" module is also
3261 reloaded at that time). The default number of threads will
3262 stay the same, 32 for most systems.
3267 Description: OST crash if filesystem is unformatted or corrupt
3268 Details : If an OST is started on a device that has never been formatted
3269 or if the filesystem is corrupt and cannot even mount then the
3270 error handling cleanup routines would dereference a NULL pointer.
3275 Description: NULL pointer deref in llap_from_page.
3276 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3277 incorrect) mapping if the page was truncated/reclaimed while it was
3278 searched for. Check for this condition and skip such pages when
3279 doing readahead. Introduce extra check to llap_from_page() to
3280 verify page->mapping->host is non-NULL (so page is not anonymous).
3283 Frequency : Sometimes when using sys_sendfile
3285 Description: "page not covered by a lock" warnings from ll_readpage
3286 Details : sendfile called ll_readpage without right page locks present.
3287 Now we introduced ll_file_sendfile that does necessary locking
3288 around call to generic_file_sendfile() much like we do in
3292 Frequency : with certain MDS communication failures at client mount time
3294 Description: NULL pointer deref after failed client mount
3295 Details : a client connection request may delayed by the network layer
3296 and not be sent until after the PTLRPC layer has timed out the
3297 request. If the client fails the mount immediately it will try
3298 to clean up before the network times out the request. Add a
3299 reference from the request import to the obd device and delay
3300 the cleanup until the network drops the request.
3303 Frequency : occasionally during client (re)connect
3305 Description: assertion failure during client (re)connect
3306 Details : processing a client connection request may be delayed by the
3307 client or server longer than the client connect timeout. This
3308 causes the client to resend the connection request. If the
3309 original connection request is replied in this interval, the
3310 client may trip an assertion failure in ptlrpc_connect_interpret()
3311 which thought it would be the only running connect process.
3314 Frequency : only with obd_echo servers and clients that are rebooted
3316 Description: kernel BUG accessing uninitialized data structure
3317 Details : When running an obd_echo server it did not start the ping_evictor
3318 thread, and when a client was evicted an uninitialized data
3319 structure was accessed. Start the ping_evictor in the RPC
3320 service startup instead of the OBD startup.
3322 Severity : enhancement
3323 Bugzilla : 10193 (patchless)
3324 Description: Remove dependency on various unexported kernel interfaces.
3325 Details : No longer need reparent_to_init, exit_mm, exit_files,
3326 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3329 Frequency : rare (only users of deprecated and unsupported LDAP config)
3331 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3332 Details : LDAP apparently contains 'lustreName' attributes instead of
3333 'name'. A simple remapping of the name is sufficient.
3336 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3338 Description: waiting_locks_callback trips kernel BUG if client is evicted
3339 Details : Running with the dump_on_timeout debug flag turned on makes
3340 it possible that the waiting_locks_callback() can try to dump
3341 the Lustre kernel debug logs from an interrupt handler. Defer
3342 this log dumping to the expired_lock_main() thread.
3344 Severity : enhancement
3346 Description: Support NFS exporting on 2.6 kernels.
3347 Details : Implement non-rawops metadata methods for NFS server to use without
3348 changing NFS server code.
3351 Frequency : very rare (synthetic metadata workload only)
3353 Description: two racing renames might cause an MDS thread to deadlock
3354 Details : Running the "racer" program may cause one MDS thread to rename
3355 a file from being the source of a rename to being the target of
3356 a rename at exactly the same time that another thread is doing
3357 so, and the second thread has already enqueued these locks after
3358 doing a lookup of the target and is trying to relock them in
3359 order. Ensure that we don't try to re-lock the same resource.
3362 Frequency : only very large systems with liblustre clients
3364 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3365 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3366 all OSTs in parallel. This allows the eviction of stale liblustre
3367 clients to proceed much faster than if they were done in series,
3368 and also offers similar improvements for other set_info RPCs.
3373 Description: excessive CPU usage during initial read phase on client
3374 Details : During the initial read phase on a client, it would agressively
3375 retry readahead on the file, consuming too much CPU and impacting
3376 performance (since 1.4.5.8). Improve the readahead algorithm
3377 to avoid this, and also improve some other common cases (read
3378 of small files in particular, where "small" is files smaller than
3379 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3384 Description: MDS crash when receiving packet with unknown intent.
3385 Details : Do not LBUG in unknown intent case, just return -EFAULT
3387 Severity : enhancement
3388 Bugzilla : 9293, 9385
3389 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3390 Details : Do not serialize getattr (non-intent version) and statfs.
3393 Frequency : occasional, when OST network is overloaded/intermittent
3395 Description: client evicted by OST after bulk IO timeout
3396 Details : If a client sends a bulk IO request (read or write) the OST
3397 may evict the client if it is unresposive to its data GET/PUT
3398 request. This is incorrect if the network is overloaded (takes
3399 too long to transfer the RPC data) or dropped the OST GET/PUT
3400 request. There is no need to evict the client at all, since
3401 the pinger and/or lock callbacks will handle this, and the
3402 client can restart the bulk request.
3405 Frequency : Always when mmapping file with no objects
3407 Description: client crashes when mmapping file with no objects
3408 Details : Check that we actually have objects in a file before doing any
3409 operations on objects in ll_vm_open, ll_vm_close and
3415 Description: Request leak when working with deleted CWD
3416 Details : Introduce advanced request refcount tracking for requests
3417 referenced from lustre intent.
3419 Severity : Enhancement
3421 Description: Cache open file handles on client.
3422 Details : MDS now will return special lock along with openhandle, if
3423 requested and client is allowed to hold openhandle, even if unused,
3424 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3425 closing files for every read/write openration.
3427 Severity : Enhancement
3429 Description: Cache open negative dentries on client when possible.
3430 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3431 negative dentries on lock revocation.
3436 Description: Remounting a client read-only wasn't possible with a zconf mount
3437 Details : It wasn't possible to remount a client read-only with llmount.
3439 Severity : enhancement
3440 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3441 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3442 MPICH is included to add Lustre-specific ADIO interfaces.
3443 This is based closely on the UFS ADIO layer and only differs
3444 in file creation, in order to allow the OST striping to be set.
3445 This is user-contributed code and not supported by CFS.
3450 Description: extended inode attributes (immutable, append-only) work improperly
3451 when 2.4 and 2.6 kernels are used on client/server or vice versa
3452 Details : Introduce kernel-independent values for these flags.
3454 Severity : enhancement
3457 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3458 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3459 granularity. Now, Fractional MB granularity is be supported,
3460 this is very useful for low memory system.
3462 Severity : enhancement
3464 Description: Getattr by fid
3465 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3466 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3467 have locks and avoids getattr rpc requests for it on every stat.
3470 Frequency : Always, for filesystems larger than 2TB
3472 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3473 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3474 filesystems larger than 2TB to be mounted with mballoc enabled.
3477 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3479 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3480 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3481 filesystems larger than 2TB is mounted then the kernel will
3482 silently corrupt the start of the filesystem. CONFIG_LBD is
3483 enabled for all CFS-supported kernels, but the possibility of
3484 this happening with a modified kernel config exists.
3486 Severity : enhancement
3488 Description: add client O_DIRECT support for 2.6 kernels
3489 Details : It is now possible to do O_DIRECT reads and writes to files
3490 in the Lustre client mountpoint on 2.6 kernel clients.
3492 Severity : enhancement
3494 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3495 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3496 parallel, not waiting for response from every OST before sending
3497 a rpc to the next OST.
3502 Description: setattr vs write race when updating file timestamps
3503 Details : Client processes that update a file timestamp into the past
3504 right after writing to the file (e.g. tar) it is possible that
3505 the updated file modification time can be reset to the current
3506 time due to a race between processing the setattr and write RPC.
3508 Severity : enhancement
3510 Description: Bring 'lfs find' closer in line with regular Linux find.
3511 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3512 -print0 options and obtains all the needed info through the lustre
3515 Severity : enhancement
3517 Description: support up to 1024 configured devices on one node
3518 Details : change obd_dev array from statically allocated to dynamically
3519 allocated structs as they are first used to reduce memory usage
3524 Description: Flush dirty partially truncated pages during truncate
3525 Details : Immediatelly flush partially truncated pages in filter_setattr,
3526 this way we completely avoid having any pages in page cache on OST
3527 and can retire ugly workarounds during writes to flush such pages.
3532 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3533 Details : protect inode from truncation within vfs_unlink() context
3534 just take a reference before calling vfs_unlink() and release it
3535 when parent's i_sem is free.
3540 Description: last_id value checked outside lock on OST caused LASSERT failure
3541 Details : If there were multiple MDS->OST object precreate requests in
3542 flight, it was possible that the OST's last object id was checked
3543 outside a lock and incorrectly tripped an assertion. Move checks
3544 inside locks, and discard old precreate requests.
3547 Frequency : always, if extents are used on OSTs
3549 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3550 Details : OSTs using the extents format would not zero the high 16 bits of
3551 the index physical block number. This is not a problem for any
3552 OST filesystems smaller than 16TB, and no kernels support ext3
3553 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3554 new/modified files) and can be fixed for existing filesystems
3555 with e2fsprogs-1.39-cfs1.
3560 Description: import connection selection may be incorrect if timer wraps
3561 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3562 import connections to be ignored if the 32-bit jiffies counter
3563 wraps. Use a 64-bit jiffies counter.
3566 Frequency : very large clusters immediately after boot
3568 Description: LNET request buffers exhausted under heavy short-term load
3569 Details : If a large number of client requests are generated on a service
3570 that has previously never seen so many requests it is possible
3571 that the request buffer growth cannot keep up with the spike in
3572 demand. Instead of dropping incoming requests, they are held in
3573 the LND until the RPC service can accept more requests.
3576 Frequency : Sometimes during replay
3578 Description: Assertion failure in ll_local_open after replay.
3579 Details : If replay happened on an open request reply before we were able
3580 to set replay handler, reply will become not swabbed tripping the
3581 assertion in ll_local_open. Now we set the handler right after
3582 recognising of open request
3585 Frequency : very rare
3587 Description: kernel reports "badness in vsnprintf"
3588 Details : Reading from the "recovery_status" /proc file in small chunks
3589 may cause a negative length in lprocfs_obd_rd_recovery_status()
3590 call to vsnprintf() (which is otherwise harmless). Exit early
3591 if there is no more space in the output buffer.
3593 Severity : enhancement
3595 Description: clear OBD RPC statistics by writing to them
3596 Details : It is now possible to clear the OBD RPC statistics by writing
3597 to the "stats" file.
3602 Description: Client mtime is not the same on different clients after utimes
3603 Details : In some cases, the client was using the utimes() syscall on
3604 a file cached on another node. The clients now validate the
3605 ctime from the MDS + OSTs to determine which one is right.
3610 Description: Inability to activate failout mode
3611 Details : lconf script incorrectly assumed that in pythong string's numeric
3612 value is used in comparisons.
3615 Frequency : always with multiple stripes per file
3617 Description: Inefficient object allocation for mutli-stripe files
3618 Details : When selecting which OSTs to stripe files over, for files with
3619 a stripe count that divides evenly into the number of OSTs,
3620 the MDS is always picking the same starting OST for each file.
3621 Return the OST selection heuristic to the original design.
3626 Description: mount failures may take full timeout to return an error
3627 Details : Under some heavy load conditions it is possible that a
3628 failed mount can wait for the full obd_timeout interval,
3629 possibly several minutes, before reporting an error.
3630 Instead return an error as soon as the status is known.
3632 Frequency : quota enabled and large files being deleted
3634 Description: releasing more than 4GB of quota at once hangs OST
3635 Details : If a user deletes more than 4GB of files on a single OST it
3636 will cause the OST to spin in an infinite loop. Release
3637 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3642 Description: statfs data retrieved from /proc may be stale or zero
3643 Details : When reading per-device statfs data from /proc, in the
3644 {kbytes,files}_{total,free,avail} files, it may appear
3645 as zero or be out of date.
3648 Frequency : systems with MD RAID1 external journal devices
3650 Description: lconf's call to blkid is confused by RAID1 journal devices
3651 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3652 of returning all block devices that match the journal UUID.
3655 Frequency : always, for aggregate stripe size over 4GB
3657 Description: assertion fails when trying to use 4GB stripe size
3658 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3659 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3662 Frequency : always on ppc64
3664 Description: the first write on an ext3 filesystem with mballoc got stuck
3665 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3666 perform endianness conversion.
3668 ------------------------------------------------------------------------------
3670 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3672 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3673 INTEROPERATE with older versions automatically. Please read the
3674 user documentation before upgrading any part of a live system.
3675 * WARNING: Lustre networking configuration changes are required with
3676 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3679 * Support for kernels:
3680 2.6.9-22.0.2.EL (RHEL 4)
3681 2.6.5-7.244 (SLES 9)
3682 2.6.12.6 vanilla (kernel.org)
3685 Severity : enhancement
3686 Bugzilla : 7981/8208
3687 Description: Introduced Lustre Networking (LNET)
3688 Details : LNET is new networking infrastructure for Lustre, it includes
3689 a reorganized network configuration mode (see the user
3690 documentation for full details) as well as support for routing
3691 between different network fabrics. Lustre Networking Devices
3692 (LNDs) for the supported network fabrics have also been
3693 created for this new infrastructure.
3695 Severity : enhancement
3696 Description: Introduced Access control lists
3697 Details : clients can set ACLs on files and directories in order to have
3698 more fine-grained permissions than the standard Unix UGO+RWX.
3699 The MDS must be started with the "-o acl" mount option.
3701 Severity : enhancement
3702 Description: Introduced filesystem quotas
3703 Details : Administrators may now establish per-user quotas on the
3706 Severity : enhancement
3708 Description: Configuration change for the XT3
3709 The PTLLND is now used to run Lustre over Portals on the XT3
3710 The configure option(s) --with-cray-portals are no longer used.
3711 Rather --with-portals=<path-to-portals-includes> is used to
3712 enable building on the XT3. In addition to enable XT3 specific
3713 features the option --enable-cray-xt3 must be used.
3718 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3719 Details : A race between allocating a new llcd and re-getting the llcd_lock
3720 allowed another thread to grab newly-allocated llcd.
3722 Severity : enhancement
3724 Description: 2.6 OST async journal commit and locking fix to improve performance
3725 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3726 2.6 kernels are now async as they already were in 2.4 kernels so
3727 that they can commit concurrently with the network bulk transfer.
3728 For block-allocated files the filter allocation semaphore is held
3729 to avoid filesystem fragmentation during allocation. BKL lock
3730 removed for 2.6 xattr operations where it is no longer needed.
3735 Description: lconf incorrectly determined whether two IP networks could talk
3736 Details : In some more complicated routing and multiple-network
3737 configurations, lconf will avoid trying to make a network
3738 connection to a disjoint part of the IP space. It was doing the
3739 math incorrectly for one set of cases.
3744 Description: Fix for potential infinite loop processing records in an llog.
3745 Details : If an llog record is corrupted/zeroed, it is possible to loop
3746 forever in llog_process(). Validate the llog record length
3747 and skip the remainder of the block on error.
3750 Frequency : occasional (liblustre only)
3752 Description: liblustre could not open files whose last component is a symlink
3753 Details : sysio_path_walk() would incorrectly pass the open intent to
3754 intermediate path components.
3757 Frequency : rare (liblustre only with non-standard tuning)
3758 Bugzilla : 7201 (7350)
3759 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3760 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3761 after it has been released, only do this for referenced locks
3763 Severity : enhancement
3765 Description: specifying an (invalid) directory default stripe_size of -1
3766 would reset the directory default striping
3767 Details : stripe_size -1 was used internally to signal directory stripe
3768 removal, now use "all default" to signal dir stripe removal
3769 as a directory striping of "all default" is not useful
3772 Frequency : common for large clusters running liblustre clients
3774 Description: doing an ls when liblustre clients are running is slow
3775 Details : sending a glimpse AST to a liblustre client waits for every AST
3776 to time out, as liblustre clients will not respond. Since they
3777 cannot cache data we refresh the OST lock LVB from disk instead.
3779 Severity : enhancement
3781 Description: doing an ls at the same time as file IO can be slow
3782 Details : enqueue and other "small" requests can be blocked behind many
3783 large IO requests. Create a new OST IO portal for non-IO
3784 requests so they can be processed faster.
3787 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3789 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3790 created file with mode 000
3791 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3792 allow an ftruncate() as a non-root user to a file with mode 000.
3793 Lustre now respects this flag to disable mode checks when
3794 truncating a file owned by the user
3797 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3800 Description: Revoking locks from clients that went dead or catatonic might take
3802 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3803 cancellation of such locks instant on servers without waiting for
3804 any reply from clients. Clients drops these locks when cancel
3805 notification from server is received without replying.
3808 Frequency : liblustre-only, when liblustre client dies or becomes busy
3810 Description: Doing ls on Linux clients can take a long time with active
3812 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3813 granting such locks to it in the first place if possible. Locks
3814 are taken by proxy on the OST during the read or write and
3815 dropped immediately afterward. Add connect flags handling, do
3816 not grant locks to liblustre clients for glimpse ASTs.
3818 Severity : enhancement
3820 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3821 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3822 sequential reads and behaves badly with applications doing random
3823 reads. Improve it by reading ahead at least read region, and
3824 avoiding excessive large RPC for small reads.
3826 Severity : enhancement
3828 Description: Creating more than 1000 files for a single job may cause a load
3829 imbalance on the OSTs if there are also a large number of OSTs.
3830 Details : qos_prep_create() uses an OST index reseed value that is an
3831 even multiple of the number of available OSTs so that if the
3832 reseed happens in the middle of the object allocation it will
3833 still utilize the OSTs as uniformly as possible.
3838 Description: OST or MDS may oops in ping_evictor_main()
3839 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3840 but doesn't restart at beginning of obd_exports_timed list
3843 Severity : enhancement
3845 Description: improve by-nid export eviction on the MDS and OST
3846 Details : allow multiple exports with the same NID to be evicted at one
3847 time without re-searching the exports list.
3850 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3852 Description: MDS may oops in groups_free()
3853 Details : in rare race conditions a newly allocated group_info struct is
3854 freed again, and this can be NULL. The 2.4 compatibility code
3855 for groups_free() checked for a NULL pointer, but 2.6 did not.
3858 Frequency : common for liblustre clients doing little filesystem IO
3859 Bugzilla : 9352, 7313
3860 Description: server may evict liblustre clients accessing contended locks
3861 Details : if a client is granted a lock or receives a completion AST
3862 with a blocking AST already set it would not reply to the AST
3863 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3866 Frequency : lfs setstripe, only systems with more than 160 OSTs
3868 Description: unable to set striping with a starting offset beyond OST 160
3869 Details : llapi_create_file() incorrectly limited the starting stripe
3870 index to the maximum single-file stripe count.
3873 Frequency : LDAP users only
3875 Description: lconf did not handle in-kernel recovery with LDAP properly
3876 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3878 Severity : enhancement
3880 Description: bind OST threads to NUMA nodes to improve performance
3881 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3882 node and do their allocations there to localize memory access
3884 Severity : enhancement
3886 Description: llmount can determine client NID directly from Myrinet (GM)
3887 Details : the client NID code from gmnalnid was moved directly into
3888 llmount, removing the need to use this or specifying the
3889 client NID explicitly when mounting GM clients with zeroconf
3892 Frequency : if client is started with down MDS
3894 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3895 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3896 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3897 ptlrpc_import_delayed_req() because we didn't check if the
3898 request was interrupted, and we also didn't break out of the
3899 event loop if there was no timeout
3904 Description: data loss during non-page-aligned writes to a single file from
3905 both multiple nodes and multiple threads on one node at same time
3906 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3907 inconsistency led to false short-reads, that were cached and later
3908 used by ->prepare_write() to fill in partially written page,
3909 leading to data loss.
3912 Frequency : always, if lconf --abort_recovery used
3914 Description: lconf --abort_recovery fails with 'Operation not supported'
3915 Details : lconf was attempting to abort recovery on the MDT device and not
3918 Severity : enhancement
3920 Description: remove cleanup logs
3921 Details : replace lconf-generated cleanup logs with lustre internal
3922 cleanup routines. Eliminates the need for client-cleanup and
3925 Severity : enhancement
3927 Description: add support for EAs (user and system) on lustre filesystems
3928 Details : it is now possible to store extended attributes in the Lustre
3929 client filesystem, and with the user_xattr mount option it
3930 is possible to allow users to store EAs on their files also
3932 Severity : enhancement
3934 Description: Add possibility (config option) to show minimal available OST free
3936 Details : When compiled with --enable-mindf configure option, statfs(2)
3937 (and so, df) will return least minimal free space available from
3938 all OSTs as amount of free space on FS, instead of summary of
3939 free spaces of all OSTs.
3941 Severity : enhancement
3943 Description: do not expand extent locks acquired on OST-side
3944 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3945 by server: they are not cached anyway.
3948 Frequency : when mmap is used/binaries executed from Lustre
3950 Description: Unmmap pages before throwing them away from read cache.
3951 Details : llap_shrink cache now attempts to unmap pages before discarding
3952 them (if unmapping failed - do not discard). SLES9 kernel has
3953 extra checks that trigger if this unmapping is not done first.
3958 Description: lconf didn't resolve symlinks before checking to see whether a
3959 given mountpoint was already in use
3962 Frequency : when migrating failover services
3963 Bugzilla : 6395, 9514
3964 Description: When migrating a subset of services from a node (e.g. failback
3965 from a failover service node) the remaining services would
3966 time out and evict clients.
3967 Details : lconf --force (implied by --failover) sets the global obd_timeout
3968 to 5 seconds in order to quickly disconnect, but this caused
3969 other RPCs to time out too quickly. Do not change the global
3970 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3972 Severity : enhancement
3973 Frequency : if MDS is started with down OST
3974 Bugzilla : 9439,5706
3975 Description: Allow startup/shutdown of an MDS without depending on the
3976 availability of the OSTs.
3977 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3978 Add appropriate locking and lov-osc refcounts for safe
3979 cleaning. Add osc abort_inflight calls in case the
3980 synchronize never started.
3983 Frequency : occasional (Cray XT3 only)
3985 Description: root not authorized to access files in CRAY_PORTALS environment
3986 Details : The client process capabilities were not honoured on the MDS in
3987 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3988 been accessed by an authorized user then root was able to access
3989 the file on the local client also. The root user capabilities
3990 are now allowed on the MDS, as this environment has secure UID.
3993 Frequency : occasional
3995 Description: ldiskfs "too long searching" message happens too often
3996 Details : A debugging message (otherwise harmless) prints too often on
3997 the OST console. This has been reduced to only happen when
3998 there are fragmentation problems on the filesystem.
4003 Description: Division by zero in statfs when all OSCs are inactive
4004 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4005 lov_statfs divides by value returned by lov_get_stripecnt().
4009 Bugzilla : 9489, 3273
4010 Description: First write from each client to each OST was only 4kB in size,
4011 to initialize client writeback cache, which caused sub-optimal
4012 RPCs and poor layout on disk for the first writen file.
4013 Details : Clients now request an initial cache grant at (re)connect time
4014 and so that they can start streaming writes to the cache right
4015 away and always do full-sized RPCs if there is enough data.
4016 If the OST is rebooted the client also re-establishes its grant
4017 so that client cached writes will be honoured under the grant.
4022 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4023 Details : Now I/O RPCs go to different portal number and (presumably) fast
4024 lock requests (and glimses) and other RPCs get their own service
4025 threads pool that should be able to service those RPCs
4028 Severity : enhancement
4030 Description: Ability to exchange lustre version between client and servers and
4031 issue warnings at client side if client is too old. Also for
4032 liblustre clients there is ability to refuse connection of too old
4034 Details : New 'version' field is added to connect data structure that is
4035 filled with version info. That info is later checked by server and
4039 Frequency : rare, liblustre only.
4040 Bugzilla : 9296, 9581
4041 Description: Two simultaneous writes from liblustre at offset within same page
4042 might proceed at the same time overwriting eachother with stale
4044 Details : I/O lock withing llu_file_prwv was released too early, before data
4045 actually was hitting the wire. Extended lock-holding time until
4046 server acknowledges receiving data.
4049 Frequency : extremely rare. Never observed in practice.
4051 Description: avoid generating lustre_handle cookie of 0.
4052 Details : class_handle_hash() generates handle cookies by incrementing
4053 global counter, and can hit 0 occasionaly (this is unlikely, but
4054 not impossible, because initial value of cookie counter is
4055 selected randonly). Value of 0 is used as a sentinel meaning
4056 "unassigned handle" --- avoid it. Also coalesce two critical
4057 sections in this function into one.
4059 Severity : enhancement
4061 Description: allow liblustre clients to delegate truncate locking to OST
4062 Details : To avoid overhead of locking, liblustre client instructs OST to
4063 take extent lock in ost_punch() on client's behalf. New connection
4064 flag is added to handle backward compatibility.
4066 Severity : enhancement
4067 Bugzilla : 4928, 7341, 9758
4068 Description: allow number of OST service threads to be specified
4069 Details : a module parameter allows the number of OST service threads
4070 to be specified via "options ost ost_num_threads={N}" in the
4071 OSS's /etc/modules.conf or /etc/modprobe.conf.
4075 Bugzilla : 6146, 9635, 9895
4076 Description: servers crash with bad pointer in target_handle_connect()
4077 Details : In rare cases when a client is reconnecting it was possible that
4078 the connection request was the last reference for that export.
4079 We would temporarily drop the export reference and get a new
4080 one, but this may have been the last reference and the export
4081 was just destroyed. Get new reference before dropping old one.
4083 Severity : enhancement
4084 Frequency : if client is started with failover MDS
4086 Description: Allow multiple MDS hostnames in the mount command
4087 Details : Try to read the configuration from all specified MDS
4088 hostnames during a client mount in case the "primary"
4091 Severity : enhancement
4093 Description: Stop sending data to evicted clients as soon as possible.
4094 Details : Check if the client we are about to send or are sending data to
4095 was evicted already. (Check is done every second of waiting,
4096 for which l_wait_event interface was extended to allow checking
4097 of exit condition at specified intervals).
4100 Frequency : rare, normally only when NFS exporting is done from client
4102 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4104 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4107 Frequency : always, if srand() is called before liblustre initialization
4109 Description: Liblustre uses system PRNG disturbing its usage by user application
4110 Details : Introduce internal to lustre fast and high-quality PRNG for
4111 lustre usage and make liblustre and some other places in generic
4112 lustre code to use it.
4114 Severity : enhancement
4115 Bugzilla : 9477, 9557, 9870
4116 Description: Verify that the MDS configuration logs are updated when xml is
4117 Details : Check if the .xml configuration logs are newer than the config
4118 logs stored on the MDS and report an error if this is the case.
4119 Request --write-conf, or allow starting with --old_conf.
4121 Severity : enhancement
4123 Description: Handle symlinks in the path when checking if Lustre is mounted.
4124 Details : Resolve intermediate symlinks when checking if a client has
4125 mounted a filesystem to avoid duplicate client mounts.
4130 Description: lconf can hit an error exception but still return success.
4131 Details : The lconf command catches the Command error exception at the top
4132 level script context and will exit with the associated exit
4133 status, but doesn't ensure that this exit status is non-zero.
4138 Description: failure of ptlrpc thread startup can cause oops
4139 Details : Starting a ptlrpc service thread can fail if there are a large
4140 number of threads or the server memory is very fragmented.
4141 Handle this without oopsing.
4144 Frequency : always, only if liblustre and non-default acceptor port was used
4146 Description: liblustre cannot connect to servers with non-default acceptor port
4147 Details : tcpnal_set_default_params() was not called and was therefore
4148 ignoring the environment varaible TCPNAL_PORT, as well as other
4149 TCPNAL_ environment variables
4154 Description: two objects could be created on the same OST for a single file
4155 Details : If an OST is down, in some cases it was possible to create two
4156 objects on a single OST for a single file. No problems other
4157 than potential performance impact and spurious error messages.
4161 Bugzilla : 5681, 9562
4162 Description: Client may oops in ll_unhash_aliases
4163 Details : Client dcache may become inconsistent in race condition.
4164 In some cases "getcwd" can fail if the current directory is
4170 Description: Inode refcounting problems in NFS export code
4171 Details : link_raw functions used to call d_instantiate without obtaining
4172 extra inode reference first.
4176 Bugzilla : 9942, 9903
4177 Description: Referencing freed requests leading to crash, memleaks with NFS.
4178 Details : We used to require that call to ll_revalidate_it was always
4179 followed by ll_lookup_it. Also with revalidate_special() it is
4180 possible to call ll_revalidate_it() twice for the same dentry
4181 even if first occurence returned success. This fix changes semantic
4182 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4183 reference on a request referred from the intent.
4184 ll_intent_release() then releases such a request.
4187 Frequency : rare, normally benchmark loads only
4189 Description: unlinked inodes were kept in memory on the client
4190 Details : If a client is repeatedly creating and unlinking files it
4191 can accumulate a lot of stale inodes in the inode slab cache.
4192 If there is no other client load running this can cause the
4193 client node to run out of memory. Instead flush old inodes
4194 from client cache that have the same inode number as a new inode.
4197 Frequency : SLES9 2.6.5 kernel and long filenames only
4198 Bugzilla : 9969, 10379
4199 Description: utime reports stale NFS file handle
4200 Details : SLES9 uses out-of-dentry names in some cases, which confused
4201 the lustre dentry revalidation. Change it to always use the
4205 Frequency : rare, unless heavy write-truncate concurrency is continuous
4206 Bugzilla : 4180, 6984, 7171, 9963, 9331
4207 Description: OST becomes very slow and/or deadlocked during object unlink
4208 Details : filter_destroy() was holding onto the parent directory lock
4209 while truncating+unlinking objects. For very large objects this
4210 may block other threads for a long time and slow overall OST
4211 responsiveness. It may also be possible to get a lock ordering
4212 deadlock in this case, or run out of journal credits because of
4213 the combined truncate+unlink. Solution is to do object truncate
4214 first in one transaction without parent lock, and then do the
4215 final unlink in a new transaction with the parent lock. This
4216 reduces the lock hold time dramatically.
4219 Frequency : rare, 2.4 kernels only
4221 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4222 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4223 before unmounting their filesystem in order to avoid deadlock.
4224 The kernel_locked() function in 2.4 kernels only checks whether
4225 the kernel lock is held, not whether it is this process that is
4226 holding it as 2.6 kernels do.
4231 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4232 Details : The client ptlrpc code may be trying to reconnect to a down
4233 server before a previous connection attempt has timed out.
4234 Increase the reconnect interval to be longer than the connection
4235 timeout interval to avoid sending duplicate connections to
4239 Frequency : echo_client brw_test command
4241 Description: fix echo_client to work with OST preallocated code
4242 Details : OST preallocation code (5137) didn't take echo_client IO path
4243 into account: echo_client calls filter methods outside of any
4244 OST thread and, hence, there is no per-thread preallocated
4245 pages and buffers to use. Solution: hijack pga pages for IO. As
4246 a byproduct, this avoids unnecessary data copying.
4250 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4251 Description: Client can oops in mdc_commit_close() after open replay
4252 Details : It was possible for the MDS to return an open request with no
4253 transaction number in mds_finish_transno() if the client was
4254 evicted, but without actually returning an error. Clients
4255 would later try to replay that open and may trip an assertion
4256 Simplify the client close codepath, and always return an error
4257 from the MDS in case the open is not successful.
4260 Frequency : rare, 2.6 OSTs only
4262 Description: OST may deadlock under high load on fragmented files
4263 Details : If there was a heavy load and highly-fragmented OST filesystems
4264 it was possible to have all the OST threads deadlock waiting on
4265 allocation of biovecs, because the biovecs were not released
4266 until the entire RPC IO was completed. Instead, release biovecs
4267 as soon as they are complete to ensure forward IO progress.
4269 Severity : enhancement
4271 Description: Support for specifying external journal device at mount
4272 Details : If an OST or MDS device is formatted with an external journal
4273 device, this device major/minor is stored in the ext3 superblock
4274 and may not be valid for failover. Allow detecting and
4275 specifying the external journal at mount time.
4280 Description: Mounting an MDS with pending unlinked files may cause oops
4281 Details : target_finish_recovery() calls mds_postrecov() which returned
4282 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4283 considers this an error and immediately begins cleaning up the
4284 lov, just after starting the mds_lov process
4286 Severity : enhancement
4288 Description: Implement 'lfs df' to report actual free space on per-OST basis
4289 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4290 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4291 usage of objects; '-h' to report in human readable format.
4293 ------------------------------------------------------------------------------
4295 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4302 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4303 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4304 at startup during orphan unlinking. Instead, skip buddy update
4305 before setup, as it will be regenerated after recovery is complete.
4310 Description: If an OST is inactive, its locks might reference stale inodes.
4311 Details : lov_change_cbdata() must iterate over all namespaces, even if
4312 they are inactive to clear inode references from the lock.
4314 Severity : enhancement
4315 Frequency : occasional, if non-standard max_dirty_mb used
4317 Description: Client will block write RPCs if not enough grant
4318 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4319 then the client will block writes while waiting for another RPC
4320 to complete instead of consuming its dirty limit. With change
4321 we get improved performance when max_dirty_mb is small.
4323 Severity : enhancement
4324 Bugzilla : 3389, 6253
4325 Description: Add support for supplementary groups on the MDS.
4326 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4327 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4328 lookups for user supplementary groups into a cache.
4332 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4333 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4334 for liblustre O_EXCL is enforced by MDS.
4338 Description: readdir never returns NULL in liblustre.
4339 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4340 dentry in struct dirent.
4344 Frequency : liblustre only, depends on application IO pattern
4345 Description: liblustre clients evicted if not contacting servers
4346 Details : Don't put liblustre clients into the ping_evictor list, so
4347 they will not be evicted by the pinger ever.
4349 Severity : enhancement
4351 Description: Add ability to evict clients by NID from MDS.
4352 Details : By echoing "nid:$NID" string into
4353 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4354 $NID would be instantly evicted from this MDS and from all active
4355 OSTs connected to it.
4359 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4360 Details : lookup_it_finish() used to query file size from OSTs that was not
4365 Description: service threads change working directory to that of init
4366 Details : Starting lustre service threads may pin the working directory
4367 of the parent thread, making that filesystem busy. Threads
4368 now change to the working directory of init to avoid this.
4372 Frequency : during shutdown only
4373 Description: shutdown with a failed MDS or OST can cause unmount to hang
4374 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4379 Frequency : 2.6 only
4380 Description: chmod/chown may include an extra supplementary group
4381 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4382 supplementary group and if none is specified this is used.
4385 Bugzilla : 5479 (6816)
4387 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4388 Details : If lookup is in progress on a file that is unlinked we might try
4389 to revalidate the inode and fail in revalidate after lookup is
4390 complete and ll_file_open() enqueues the open again but
4391 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4394 Frequency : always, if lconf --abort_recovery used
4396 Description: lconf --abort_recovery fails with 'Operation not supported'
4397 Details : lconf was attempting to abort recovery on the MDT device and not
4400 ------------------------------------------------------------------------------
4402 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4407 Frequency : rare (only unsupported configurations with a node running as an
4409 Bugzilla : 6514, 5137
4410 Description: Mounting a Lustre file system on a node running as an OST could
4412 Details : OSTs now preallocates memory needed to write out data at
4413 startup, instead of when needed, to avoid having to
4414 allocate memory in possibly low memory situations.
4415 Specifically, if the file system is mounted on on OST,
4416 memory pressure could force it to try to write out data,
4417 which it needed to allocate memory to do. Due to the low
4418 memory, it would be unable to do so and the node would
4419 become unresponsive.
4421 Severity : enhancement
4423 Description: Addition of lconf --service command line option
4424 Details : lconf now accepts a '--service <arg>' option, which is
4425 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4427 Severity : enhancement
4429 Description: Failover mode is now the default for OSTs.
4430 Details : By default, OSTs will now run in failover mode. To return to
4431 the old behaviour, add '--failout' to the lmc line for OSTs.
4433 Severity : enhancement
4435 Description: Health checks are now provided for MDS and OSTs
4436 Details : Additional detailed health check information on MSD and OSTs
4437 is now provided through the procfs health_check value.
4440 Frequency : occasional, depends on IO load
4442 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4443 after numerous create/delete cycles
4444 Details : The ext3 inode allocation policy would not allocate new inodes
4445 very well on the OSTs because there are no new directories
4446 being created. Instead we look for groups with free space if
4447 the parent directories are nearly full.
4452 Description: Network or server problems during mount may cause partially
4453 mounted clients instead of returning an error.
4454 Details : The config llog parsing code may overwrite the error return
4455 code during mount error handling, returning success instead
4461 Description: MDS can fail to allocate large reply buffers
4462 Details : After long uptimes the MDS can fail to allocate large reply
4463 buffers (e.g. zconf client mount config records) due to memory
4464 fragmentation or consumption by the buffer cache. Preallocate
4465 some large reply buffers so that these replies can be sent even
4466 under memory pressure.
4470 Frequency : rare (liblustre)
4471 Description: fsx running with liblustre complained that using truncate() to
4472 extend the file doesn't work. This patch corrects that issue.
4473 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4474 ATTR_SIZE and lsm use in llu_setattr_raw.
4478 Frequency : rare, only 2.6 kernels
4479 Description: Unusual file access patterns on the MDS may result in inode
4480 data being lost in very rare circumstances.
4481 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4482 code in the 2.6 kernel caused the inode and/or EA data not to
4483 be read from disk, causing single-file corruption.
4487 Frequency : rare, only 2.6 filesystems using extents
4488 Description: Heavy concurrent write and delete load may cause data corruption.
4489 Details : It was possible under high-load situations to have an extent
4490 metadata block in the block device cache from a just-unlinked
4491 file overwrite a newly-allocated data block. We now unmap any
4492 metadata buffers that alias just-allocated data blocks.
4496 Frequency : filesystems with default stripe_count larger than 77
4497 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4498 Details : lconf specifies an inode size of 4096 bytes when the default
4499 stripe_count is larger than 77. This conflicts with the default
4500 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4501 this case to avoid pinning too much memory for large EAs.
4503 ------------------------------------------------------------------------------
4505 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4510 Frequency : rare (extremely heavy IO load with hundreds of clients)
4512 Description: Client is evicted, gets IO error writing to file
4513 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4514 caused clients to be evicted for AST timeouts. The fixes in
4515 bug 5192 mean we no longer need to have such short AST timeouts
4516 so ldlm_timeout has been increased.
4519 Frequency : occasional during --force or --failover shutdown under load
4520 Bugzilla : 5949, 4834
4521 Description: Server oops/LBUG if stopped with --force or --failover under load
4522 Details : a collection of import/export refcount and cleanup ordering
4523 issues fixed for safer force cleanup
4526 Frequency : only filesystems larger than 120 OSTs
4527 Bugzilla : 5990, 6223
4528 Description: lfs getstripe would oops on a very large filesystem
4529 Details : lov_getconfig used kfree on vmalloc'd memory
4532 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4533 Bugzilla : 6242, 6243
4534 Description: reading from files that had been truncated to a non-zero size
4535 but never opened returned no data
4536 Details : ll_file_read() reads zeros from no-object files to EOF
4541 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4542 Details : There's little harm in aborting MDS/OSS recovery and letting it
4543 try again, so I removed the LASSERT and return an error instead.
4545 Severity : enhancement
4547 Description: New debugging infrastructure for tracking down data corruption
4548 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4549 (b) cover more of the client-side code path, and (c) try to narrow
4550 down where problems occurred
4554 Bugzilla : 3819, 4364, 4397, 6313
4555 Description: Racing close and eviction MDS could cause assertion in mds_close
4556 Details : It was possible to get multiple mfd references during close and
4557 client eviction, leading to one thread referencing a freed mfd.
4559 Severity: : enhancement
4560 Bugzilla : 3262, 6359
4561 Description: Attempts to reconnect to servers are now more aggressive.
4562 Details : This builds on the enhanced upcall-less recovery that was added
4563 in 1.4.2. When trying to reconnect to servers, clients will
4564 now try each server in the failover group every 10 seconds. By
4565 default, clients would previously try one server every 25 seconds.
4570 Description: After recovery, certain operations trigger a failed
4571 assertion on a client.
4572 Details : Failing over an mds, using lconf -d --failover, while a
4573 client was doing a readdir() call would cause the client to
4574 LBUG after recovery completed and the readdir() was resent.
4576 Severity : enhancement
4578 Description: Default groups are now added by lconf
4579 Details : You can now run lconf --group <servicename> without having to
4580 manually add groups with lmc.
4583 Frequency : occasional
4585 Description: Nodes with an elan id of 0 trigger a failed assertion
4588 Frequency : always when accessing e.g. tty/console device nodes
4590 Description: tty and some other devices nodes cannot be used on lustre
4591 Details : file's private_data field is used by device data and lustre
4592 values in there got lost. New field was added to struct file to
4593 store fs-specific private data.
4596 Frequency : when exporting Lustre via NFS
4598 Description: NFSD failed occasionally when looking up a path component
4599 Details : NFSD is looking up ".." which was broken in ext3 directories
4600 that had grown large enough to become hashed.
4603 Frequency : Clusters with multiple interfaces not on the same subnet
4605 Description: Nodes will repeatedly try to reconnect to an interface which it
4606 cannot reach and report an error to the log.
4607 Details : Extra peer list entries will be created by lconf with some peers
4608 unreachable. lconf now validates the peer before adding it.
4611 Frequency : Only if a default stripe is set on the filesystem root.
4613 Description: Setting a default stripe on the filesystem root prevented the
4614 filesystem from being remounted.
4615 Details : The client was sending extra request flags in the root getattr
4616 request and did not allocate a reply buffer for the dir EA.
4619 Frequency : occasional, higher if lots of files are accessed by one client
4620 Bugzilla : 6159, 6097
4621 Description: Client trips assertion regarding lsm mismatch/magic
4622 Details : While revalidating inodes the VFS looks up inodes with ifind()
4623 and in rare cases can find an inode that is being freed.
4624 The ll_test_inode() code will free the lsm during ifind()
4625 when it finds an existing inode and then the VFS later attaches
4626 this free lsm to a new inode.
4630 Bugzilla : 6422, 7030
4631 Description: MDS deadlock between mkdir and client eviction
4632 Details : Creating a new file via mkdir or mknod (starting a transaction
4633 and getting the ns lock) can deadlock with client eviction
4634 (gets ns lock and trying to finish a synchronous transaction).
4637 Frequency : occasional
4638 Description: While starting a server, the fsfilt_ext3 module could not be
4640 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4641 kernels. Previously, lconf would still use the ext3 name
4642 when trying to load modules. Now, it will correctly use
4643 ext3 on 2.4 and ldiskfs on 2.6.
4645 Severity : enhancement
4646 Description: The default stripe count has been changed to 1
4647 Details : The interpretation of the default stripe count (0, to lfs
4648 or lmc) has been changed to mean striping across a single
4649 OST, rather than all available. For general usage we have
4650 found a stripe count of 1 or 2 works best.
4652 Severity : enhancement
4653 Description: Add support for compiling against Cray portals.
4654 Details : Conditional compiling for some areas that are different
4658 Frequency : occasional
4659 Bugzilla : 6409, 6834
4660 Description: Creating files with an explicit stripe count may lead to
4661 a failed assertion on the MDS
4662 Details : If some OSTs are full or unavailable, creating files may
4663 trigger a failed assertion on the MDS. Now, Lustre will
4664 try to use other servers or return an error to the
4668 Frequency : occasional
4670 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4671 caused lock timeout during readahead (since 1.4.2).
4672 Details : Processes doing readahead might match a lock that hasn't been
4673 granted yet if there are overlapping and conflicting lock
4674 requests. The readahead process waits on ungranted lock
4675 (original lock is CBPENDING), while OST waits for that process
4676 to cancel CBPENDING read lock and eventually evicts client.
4678 Severity : enhancement
4680 Description: Initial enabling of flock support for clients
4681 Details : Implements fcntl advisory locking and file status functions.
4682 This feature is provided as an optional mount flag (default
4683 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4684 locking are implemented yet, and those that are are not guaranteed
4685 to be completely correct in production environments.
4686 mount -t lustre -o [flock|noflock] ...
4689 Frequency : occasional
4691 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4692 assertion in the ext3 JBD (journaling) layer.
4693 Details : The b_committed_data struct is protected by the big kernel lock
4694 in 2.4 kernels, serializing journal_commit_transaction() and
4695 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4696 access to this struct. In 2.6 kernels there is finer grained
4697 locking to improve SMP performance of the JBD layer.
4701 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4706 Description: merge and cleanup kernel patches.
4709 -----------------------------------------------------------------------------
4711 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4713 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4714 versions of Lustre. Please update all servers and clients to
4715 version 1.4.2 or later at the same time. You must also run
4716 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4717 to update the configuration logs.
4719 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4720 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4721 - swab llog records of type '0' so we get proper header size/idx (5861)
4722 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4723 - fix rename of one directory over another leaking an inode (5953)
4724 - avoid SetPageDirty on 2.6 (5981)
4725 - don't re-add just-being-destroyed locks to the waiting list (5653)
4726 - when creating new directories, inherit the parent's custom
4727 striping settings if present parent (3048)
4728 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4729 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4730 - don't hold client locks on temporary worklist from l_lru (5666)
4731 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4732 - automatically evict dead clients (5921)
4733 - Update file size properly in create+truncate+fstat case (6196)
4734 - Do not unhash mountpoint dentries, do not allow removal of
4736 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4737 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4738 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4739 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4740 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4741 - Correctly handle reads of files with no objects (6243)
4742 - lctl recover will also mark a device active if deactivate used (5933)
4744 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4745 - allow --write-conf on an MDS with different nettype than client (5619)
4746 - don't write config llogs to MDS for mounts not from that MDS (5617)
4747 - lconf should create multiple TCP connections from a client (5201)
4748 - init scripts are now turned off by default; run chkconfig --on
4749 lustre and chkconfig --on lustrefs to use them
4750 - upcalls are no longer needed for clients to recover to failover
4752 - add --abort-recovery option to lconf to abort recovery on device
4754 - add support for an arbitrary number of OSTs (3026)
4755 - Quota support protocol changes.
4756 - forward compatibility changes to wire structs (6007)
4757 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4758 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4759 - improved "lustre status" script
4760 - initialize blocksize for non-regular files (6062)
4761 - added --disable-server and --disable-client configure options (5782)
4762 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4763 - Vanilla 2.4.29 support
4764 - increase maximum number of obd devices to 520 (6242)
4765 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4766 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4767 SLES 9 2.6.5 kernels
4768 - limit stripes per file to 160 (the maximum EA size) (6093)
4770 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4773 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4774 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4775 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4776 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4777 - small ext3 extents cleanups and fixes (5733)
4778 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4779 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4780 - avoid needless client->OST connect, fix handle mismatch (5317)
4781 - fix DLM error path that led to out-of-sync client, long delays (5779)
4782 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4783 - fix several locking issues related to i_size (5492,5624,5654,5672)
4784 - don't move pending lock onto export if it is already evicted (5683)
4785 - fix kernel oops when creating .foo in unlinked directory (5548)
4786 - fix deadlock in obdfilter statistics vs. object create (5811)
4787 - use time_{before,after} to avoid timer jiffies wrap (5882)
4788 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4789 - Do not leak request if server was not able to process it (5154)
4790 - If mds_open unable to find parent dir, make that negative lookup(5154)
4791 - don't create new directories with extent-mapping (5909, 5936)
4793 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4794 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4795 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4796 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4797 - support for RHEL4 kernel on i686 (5773)
4798 - provide error messages when incompatible logs are encountered (5898)
4800 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4801 * version 1.4.0.10 (1.4.1 release candidate 1)
4803 - don't keep a lock reference when lock is not granted (4238)
4804 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4805 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4806 - drop import inflight refcount on signal_completed_replay error (5255)
4807 - unlock page after async write error during send (3677)
4808 - handle missing objects in filter_preprw_read properly (5265)
4809 - no transno return for symlink open, don't save no-trasno open (3440)
4810 - don't try to complete elan receive that already failed (4012)
4811 - free RPC server reply state on error (5406)
4812 - clean up thread from ptlrpc_start_thread() on error (5160)
4813 - readahead could read extra page into cache that wasn't ejected (5388)
4814 - prevent races in class_attach/setup/cleanup/detach (5260)
4815 - don't dereference de->d_inode after l_dput of de (5458)
4816 - use "int" for stripe value returned from lock_to_stripe (5544)
4817 - mballoc allocation and error-checking fixes in 2.6 (5504)
4818 - block device patches to fix I/O request sizes in 2.6 (5482)
4819 - look up hostnames for IB nals (5602)
4820 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4821 - don't start multiple acceptors for the same port (5277)
4822 - fix incorrect LASSERT in mds_getattr_name (5635)
4823 - export a proc file for general "ping" checking (5628)
4824 - fix "lfs check" to not block when the MDS is down (5628)
4826 - service request history (4965)
4827 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4828 - create an "evict_client" /proc entry on OSTs, like the MDS has
4829 - fix mount usage message, return errors per mount(8) (5168)
4830 - change grep [] to grep "[]" in tests so they work in more UMLs
4831 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4832 - remove ancient LOV_MAGIC_V0 EA support (5047)
4833 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4834 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4835 - allow readahead allocations to fail when low on memory (5383)
4836 - mmap locking landed again, after considerable improvement (2828)
4837 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4839 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4842 - send OST transaction number in read/write reply to free req (4966)
4843 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4844 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4845 - the watchdog thread now runs as interruptible (5246)
4846 - flock/lockf fixes (but it's still disabled, pending 5135)
4847 - don't use EXT3 constants in llite code (5094)
4848 - memory shortage at startup could cause assertion (5176)
4850 - reorganization of lov code
4851 - single portals codebase
4853 - add extents/mballoc support (5025)
4854 - direct I/O reads in the obdfilter (4048)
4855 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4857 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4860 - send OST transaction number in read/write reply to free req (4966)
4861 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4862 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4863 - the watchdog thread now runs as interruptible (5246)
4864 - handle missing objects in filter_preprw_read properly (5265)
4865 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4866 - drop import inflight refcount on signal_completed_replay error (5255)
4867 - unlock page after async write error during send (3677)
4868 - return original error code on reconstructed replies (3761)
4869 - no transno return for symlink open, don't save no-trasno open (3440)
4871 - add pid to ldlm debugging output (4922)
4872 - bump the watchdog timeouts -- we can't handle 30sec yet
4873 - extra debugging for orphan dentry/inode bug (5259)
4875 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4878 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4879 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4880 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4881 - fix NULL dereference in /proc/sys/portals/routes (4827)
4882 - allow failed mdc_close() operations to be interrupted (4561)
4883 - stop precreate on OST before MDS would time out on it (4778)
4884 - don't send partial-page writes before EOF from client (4410)
4885 - discard client grant for sub-page writes on large-page clients (4520)
4886 - don't free dentries not owned by NFS code, check generation (4806)
4887 - fix lsm leak if mds_create_objects() fails (4801)
4888 - limit debug_daemon file size, always print CERROR messages (4789)
4889 - use transno after validating reply (3892)
4890 - process timed out requests if import state changes (3754)
4891 - update mtime on OST during writes, return in glimpse (4829)
4892 - add mkfsoptions to LDAP (4679)
4893 - use ->max_readahead method instead of zapping global ra (5039)
4894 - don't interrupt __l_wait_event() during strace
4896 - add software watchdogs to catch hung threads quickly (4941)
4897 - make lustrefs init script start after nfs is mounted
4898 - fix CWARN/ERROR duplication (4930)
4899 - return async write errors to application if possible (2248)
4900 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4901 - print NAL number in %x format (4645)
4902 - update barely-supported suse-2.4.21-171 series (4842)
4903 - support for sles 9 %post scripts
4904 - support for building 2.6 kernel-source packages
4905 - support for sles km_* packages
4907 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4910 - ignore -ENOENT errors in osc_destroy (3639)
4911 - notify osc create thread that OSC is being cleaned up (4600)
4912 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4913 - reconstruct ost_handle() like mds_handle() (4657)
4914 - create a new thread to do import eviction to avoid deadlock (3969)
4915 - let lconf resolve symlinked-to devices (4629)
4916 - don't unlink "objects" from directory with default EA (4554)
4917 - hold socknal file ref over connect in case target is down (4394)
4918 - allow more than 32000 subdirectories in a single directory (3244)
4919 - fix blocks count for O_DIRECT writes (3751)
4920 - OST returns ENOSPC from object create when no space left (4539)
4921 - don't send truncate RPC if file size isn't changing (4410)
4922 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4923 - bind to privileged port in socknal and tcpnal (3689)
4925 - rate limit CERROR/CWARN console message to avoid overload (4519)
4926 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4927 - basic mmap support (3918)
4928 - kernel patch series update from b1_4 (4711)
4930 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4933 - avoid crash during MDS cleanup with OST shut down (2775)
4934 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4935 - don't use bad inodes on the MDS (3744)
4936 - dynamic object preallocation to improve recovery speed (4236)
4937 - don't hold spinlock over lock dumping or change debug flags (4401)
4938 - don't zero obd_dev when it is force cleaned (3651)
4939 - print grants to console if they go negative (4431)
4940 - "lctl deactivate" will stop automatic recovery attempts (3406)
4941 - look for existing locks in ldlm_handle_enqueue() (3764)
4942 - don't resolve lock handle twice in recovery avoiding race (4401)
4943 - revalidate should check working dir is a directory (4134)
4945 - don't always mark "slow" obdfilter messages as errors (4418)
4947 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4950 - don't close LustreDB during write_conf until it is done (3860)
4951 - fix typo in lconf for_each_profile (3821)
4952 - allow dumping logs from multiple threads at one time (3820)
4953 - don't allow multiple threads in OSC recovery (3812)
4954 - fix debug_size parameters (3864)
4955 - fix mds_postrecov to initialize import for llog ctxt (3121)
4956 - replace config semaphore with spinlock (3306)
4957 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4958 - don't allow enqueue to complete on a destroyed export (3822)
4959 - down write_lock before checking llog header bitmap (3825)
4960 - recover from lock replay timeout (3764)
4961 - up llog sem before sending rpc (3652)
4962 - reduce ns lock hold times when setting kms (3267)
4963 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4964 - fix NULL deref and obd_dev leak on setup error (3312)
4965 - replace some LBUG about llog ops with error handling (3841)
4966 - don't match INVALID dentries from d_lookup and spin (3784)
4967 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4968 - fix invalid assertion in ptlrpc_set_wait (3880)
4970 - add libwrap support for the TCP acceptor (3996)
4971 - add /proc/sys/portals/routes for non-root route listing (3994)
4972 - allow setting MDS UUID in .xml (2580)
4973 - print the stack of a process that LBUGs (4228)
4975 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4978 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4979 - make sure to unset replay flag from failed open requests (3440)
4980 - if default stripe count is 0, use OST count for inode size (3636)
4981 - update parent mtime/ctime on client for create/unlink (2611)
4982 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4983 - free recovery state on server during a forced cleanup (3571)
4984 - unregister_reply for resent reqs (3063)
4985 - loop back devices mounting and status check on 2.6 (3563)
4986 - fix resource-creation race that can provoke i_size == 0 (3513)
4987 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4988 - more debugging for page-accounting assertion (3746)
4989 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4990 - avoid deadlock after precreation failure (3758)
4991 - fix race and lock order deadlock in orphan handling (3450, 3750)
4992 - add validity checks when grabbing inodes from l_ast_data (3599)
4994 - add /proc/.../recovery_status to obdfilter (3428)
4995 - lightweight CDEBUG infrastructure, debug daemon (3668)
4996 - change default OSC RPC parameters to be better on small clusters
4997 - turn off OST read cache for files smaller than 32MB
4998 - install man pages and include them in rpms (3100)
4999 - add new init script for (un)mounting lustre filesystems (2593)
5000 - run chkconfig in %post for init scripts (3701)
5001 - drop scimac NAL (unmaintained)
5003 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5006 - clean kiobufs before and after use (3485)
5007 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5008 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5009 - change init script to fail more clearly if not run as root (1528)
5010 - allow clients to reconnect during replay (1742)
5011 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5012 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5013 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5014 - fix cond_resched (3554)
5015 - only evict unfinished clients after recovery (3515)
5016 - allow bulk resend, prevent data loss (3570)
5017 - dynamic ptlrpc request buffer allocation (2102)
5018 - don't allow unlinking open directory if it isn't empty (2904)
5019 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5020 - remove extraneous obd dereference causing LASSERT failure (3334)
5021 - don't use get_cycles() when creating temp. files on the mds (3156)
5022 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5023 - handle EEXIST for set-stripe, set proper directory name (3336)
5025 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5026 - fix ksocknal_fmb_callback() error messages (2918)
5028 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5031 - don't copy lvb into (possibly NULL) reply on error (2983)
5032 - don't deref dentry after dput, don't free lvb on error (2922)
5033 - use the kms to determine writeback rpc length (2947)
5034 - increment oti_logcookies when osc is inactive (2948)
5035 - update client's i_blocks count via lvb messages (2543)
5036 - handle intent open/close of special files properly (1557)
5037 - mount MDS with errors=remount-ro, like obdfilter (2009)
5038 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5039 - don't use cancelling-locks' kms values (2947)
5040 - use highest lock extent for kms, not last one (2925)
5041 - don't dereference ERR_PTR() dentry in error handling path (3107)
5042 - fix thread race in portals_debug_dumplog() (3122)
5043 - create lprocfs device entries at setup instead of at attach (1519)
5044 - common AST error handler, don't evict client on completion race (3145)
5045 - zero nameidata in detach_mnt in 2.6 (3118)
5046 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5047 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5048 - initialize RPC timeout timer earlier for 2.6 (3219)
5049 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5050 - print nal/nid for unknown nid (3258)
5051 - additional checks for oscc recovery before doing precreate (3284)
5052 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5053 - don't crash in mdc_close for bad permissions on open (3285)
5054 - zero i_rdev for non-device files (3147)
5055 - clear page->private before handing to FS, better assertion (3119)
5056 - tune the read pipeline (3236)
5057 - fix incorrect decref of invalidated dentry (2350)
5058 - provide read-ahead stats and refine rpc in flight stats (3328)
5059 - don't hold journal transaction open across create RPC (3313)
5060 - update atime on MDS at close time (3265)
5061 - close LDAP connection when recovering to avoid server load (3315)
5062 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5063 - don't leak open file on MDS after open resend (3325)
5064 - serialize filter_precreate and filter_destroy_precreated (3329)
5065 - loop device shouldn't call sync_dev() for nul device (3092)
5066 - clear page cache after eviction (2766)
5067 - resynchronize MDS->OST in background (2824)
5068 - refuse to mount the same filesystem twice on same mountpoint (3394)
5069 - allow llmount to create routes for mounting behind routers (3320)
5070 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5071 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5072 - fix rare problem with rename on htree directories (3417)
5074 - allow default OST striping configuration per directory (1414)
5075 - fix compilation for qswnal for 2.6 kernels (3125)
5076 - increase maximum number of MDS request buffers for large systems
5077 - change liblustreapi to be useful for external progs like lfsck (3098)
5078 - increase local configuration timeout for slow disks (3353)
5079 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5081 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5084 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5085 - don't overwrite extent policy data in reply if lock was blocked (2901)
5086 - drop filter export grants atomically with removal from device (2663)
5087 - del obd_self_export from work_list in class_disconnect_exports (2908)
5088 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5089 - swab reply message in mdc_close, other PPC fixes (2464)
5090 - fix destroying of named logs (2325)
5091 - overwrite old logs when running lconf --write_conf (2264)
5092 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5093 - fix race in target_handle_connect (2898)
5094 - mds_reint_create() should take same inode create lock (2926)
5095 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5096 - don't close files for self_export to avoid uninitialized obd (2936)
5097 - allow MDS with the same name as client node (2939)
5098 - hold dentry reference for closed log files for unlink (2325)
5099 - reserve space for all logs during transactions (2059)
5100 - don't evict page beyond end of stripe extent (2925)
5101 - don't oops on a deleted current working directory (2399)
5102 - handle hard links to targets without a parent properly (2517)
5103 - don't dereference NULL lock when racing during eviction (2867)
5104 - don't grow lock extents when lots of conflicting locks (2919)
5106 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5109 - account for cache space usage on clients to avoid data loss (974)
5110 - lfsck support in lustre kernel code (2349)
5111 - reduce journal credits needed for BRW writes (2370)
5112 - orphan handling to avoid losing space on client/server crashes
5113 - ptlrpcd can be blocked, stopping ALL progress (2477)
5114 - use lock value blocks to assist in proper KMS, faster stat (1021)
5115 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5116 - recovery for initial connections (2355)
5117 - fixes for mds_cleanup_orphans (1934)
5118 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5119 - block all file creations until orphan recovery completes (1901)
5120 - client remove rq_connection from request struct (2423)
5121 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5122 - recovery timer race (2670)
5123 - mdc_close recovey bug (2532)
5124 - ptlrpc cleanup bug (2710)
5125 - mds timeout on local locks (2588)
5126 - namespace lock held during RPCs (2431)
5127 - handle interrupted sync write properly (2503)
5128 - don't try to handle a message that hasn't been replied to (2699)
5129 - client assert failure during cleanup after abort recovery (2701)
5130 - leak mdc device after failed mount (2712)
5131 - ptlrpc_check_set allows timedout requests to complete (2714)
5132 - wait for inflight reqs when ptlrpcd finishes (2710)
5133 - make sure unregistered services are removed from the srv_list
5134 - reset bulk XID's when resending them (caught by 1138 test)
5135 - unregister_bulk after timeout
5136 - fix lconf error (2694)
5137 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5138 - readahead locks pages, leaves pending causing memory pressure (2673)
5139 - increase OST request buffers to 4096 on large machines (2729)
5140 - fix up permission of existing directories in simple_mkdir (2661)
5141 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5142 - don't assign transno to errored transactions (2742)
5143 - don't delete objects on OST if given a bogus objid from MDS (2751)
5144 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5145 - if rq_no_resend, then timeout request after recovery (2432)
5146 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5147 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5148 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5149 - MDS recovery shouldn't delete objects that it has given out (2730)
5150 - if enqueue arrives after completion, don't clobber LVB (2819)
5151 - don't unlock pages twice when trigger_group_io returns error (2814)
5152 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5153 - don't write pages to disk if there was an error (1450)
5154 - don't ping imports that have recovery disabled (2676)
5155 - take buffered bytes into account when balancing socknal conn (2817)
5156 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5157 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5158 - remove little-endian swabbing of llog records (1987)
5159 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5160 - retry reposting request buffers if they fail (1191)
5161 - grow extent at grant time to avoid granting a revoked lock (2809)
5162 - lock revoke doesn't evict page if covered by a second lock (2765)
5163 - disable VM readahead to avoid reading outside lock extents (2805)
5165 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5166 - updated kernel patches for hp-2.4.20 kernel (2681)
5168 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5171 - fix truncated write corruption (2366)
5172 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5174 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5175 - ensure len(uuid) < 37 in lmc (1171)
5176 - fix ia64 OOPS in llog_test (2255)
5177 - zero end of page at obdfilter for partial page writes (2648)
5178 - don't leave stale dentries around after renames (bug 2428)
5179 - fix timeouts when evicting a client with a single lock held (2642)
5180 - set deadline for the initial HELLO message to drain (2634)
5181 - print out dotted-quad IP addresses in the socknal (2302)
5183 - additional debugging for MDS client eviction problem (2443)
5184 - fix mkfsoptions support for osts (2603, 2604)
5186 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5189 - add series for the vanilla 2.6.0 kernel
5190 - add series for the vanilla 2.4.24 kernel
5191 - add series for a cray x86/64 UL kernel drop
5192 - fix xattr patches for the vanilla 2.4.19 series
5194 - generate true UUIDs in lmc (1171)
5195 - have portals stack dumping break in UML (2466)
5196 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5197 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5198 - wake callback waiting threads on client eviction (2460)
5199 - Add --ptldebug and --subsystem to lmc (1719)
5200 - update assertion to allow safe interrupt allocation
5201 - set rq_no_resend for cancel requests (2432)
5202 - recalculate ptlrpcd timeout after resend (2494)
5203 - call vfs_rmdir when removing pending directories (2368)
5204 - fix renaming a file to itself (2429)
5205 - lmc creates a default one-stripe lov (2454)
5206 - expand procfs space to handle large clusters (2326)
5207 - increase UML stack to avoid overflow
5208 - update lconf's list of debug and subsystem masks
5209 - fix lfs find --obd (2510)
5210 - /proc tunable for disabling filter read caching (2591)
5211 - stop rpm packages from altering slapd.conf (2301)
5212 - disable nagle in the socknal under 0conf (2578)
5213 - choose mds inode size based on stripe count (2572)
5214 - fix kernel-source rpm problems (2516)
5216 - add --disable-doc to avoid pdf generation (2421)
5217 - update documentation, tests, type-os, comments
5218 - avoid format warnings on ia64
5219 - remove the TOE NAL
5220 - tiny code cleanups by removing unused fields
5222 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5225 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5226 - remove the most blatant lies from BUILDING file (2371)
5227 - change default debug level to reasonable production setting
5228 - reduce client side cache size to reduce cache flush time
5229 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5230 - make TCP zerocopy and pinger support enabled by default (2476)
5231 - sync writes completed after process exits caused crashes (2319)
5232 - maintain correct mount count on the MDS (2356)
5233 - backout 1557, because 2316 wasn't really fixed
5234 - better file I/O statistics gathering in /proc
5235 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5236 - another kernel patch to fix zero-copy TCP function export
5237 - don't take duplicate lock when processing re-sent getattr (2420)
5238 - lctl uses obd_self_export instead of creating new conn (2353)
5239 - MDS/OST recovery case which requires object creation asserted (2425)
5240 - move lfs from /usr/sbin to /usr/bin in packages
5241 - fix race between mds_client_add and mds_client_free (2417)
5242 - use kmalloc instead of slabs in portals (2430)
5243 - don't create duplicate records when a failover MDS is present (2442)
5244 - remove unnecessary mount age check (2332)
5245 - don't remove directory inodes from locks prematurely (2451)
5246 - don't break if MDS service name is the same as hostname (2103)
5247 - fix races in client write RPC generation when cache full (2482)
5249 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5252 - remove now-unused request->rq_obd (278)
5253 - if an allocation fails, print out how much memory we've used (1933)
5254 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5255 - add the "configurable stack size" patch to most series files (1256)
5256 - ability to write large log records, for 100+ OST configs (2306)
5257 - fix NULL deref when filter_prep fails (2314)
5258 - fix operator precedence error in filter_sync
5259 - dynamic allocation of socknal TX descriptors (2315)
5260 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5261 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5262 - ensure that CWARN messages in Portals always get to the syslog
5263 - __init/__exit are not for prototype decls (ldlm_init/exit)
5264 - x86-64 compile warning fixes
5265 - fix gateway LMC keyword conflict (2318)
5266 - fix MDS lock inversions in getattr/reint paths (1844)
5267 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5268 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5269 - DLM race condition prevented some lock evictions (2328)
5270 - ENOMEM detection and retry on socknal sends (2230)
5271 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5272 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5273 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5274 - better error messages when a client is rejected during recovery (1505)
5275 - avoid cancelling locks which were never granted, after failure (2330)
5276 - fix i_sem/journal inversion in mds_client_add (2333)
5277 - fix truncate/getattr lock cycle deadlock (2334)
5278 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5279 - fix two rare exit paths which could leak an l_lock() ref (2321)
5280 - fencepost error in MDS/OST orphan recovery (2226)
5281 - make log record alignment 8 bytes (1988)
5282 - lstripe now fails when requested offset > ost_count (2237)
5283 - ensure that all kernel series have a complete list.h (1607)
5284 - fix crashes in special-file operations (2316)
5285 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5287 - allow configurable automake binary, for testing new versions
5288 - small update to the lfs documentation
5290 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5292 * fix negative export reference count in fsfilt_sync (2312)
5294 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5295 * release candidate 0.9.1
5297 - orphans are moved into the PENDING directory for possible recovery
5298 - replayed opens now open by fid for orphan/rename safety (1042)
5299 - last close of an orphan inode generates a transno (683)
5300 - chdir() and mount() now pin the directory entry (1020)
5301 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5302 - discard very old requests without processing them (1502)
5303 - remove some common, well-understood CERRORs (1505)
5304 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5305 - clear "grant" flags in OST replies until OST grant code lands (1644)
5306 - fix read performance by not clobbering i_blksize on client (1598)
5307 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5308 - make LRU size a /proc tunable, clears locks when reduced (707)
5309 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5310 - server locks take references on exports now (1558)
5311 - build fixes for 2.4.20-rh trees (1663)
5312 - return an error from lov_create if all OSCs are inactive (1751)
5313 - fix import levels when a reconnect happens without a timeout (1597)
5314 - exit early from mds_open if we get a lookup error (1749)
5315 - partial page read at EOF wouldn't wait for disk before sending (1642)
5316 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5317 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5318 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5319 - write/truncate lock inversion (1639)
5320 - Don't auto-load obdclass, portals modules during cleanup (1495)
5321 - fix timestamps from jumping to "now" (1763)
5322 - extra journal assertions (1648)
5323 - add an extra multiunlink test (1771)
5324 - fix read_record/write_record API (1776)
5325 - fix leak of offset_extent, possible incorrect i_size later (1772)
5326 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5327 - Debugging for the kqswnal_get_idle_tx problems (1820)
5328 - Allow recovery to be attempted multiple times (1536)
5329 - Write out MDS last_rcvd file after it is first created (1600)
5330 - Fix tx_descriptor leak in failed transmit situations (1827)
5331 - ext3 journaling fixes for assertion failure after IO error (1871)
5332 - class_export_put() on freed export after completion AST error (1896)
5333 - Fix revalidate looping in VFS (1322)
5334 - Don't access a freed export during MDS_REINT timeout (1521)
5335 - Add open-unlink recovery support on the MDS (1673,1764)
5336 - Return an error if no MDS data was read from last_rcvd (1946)
5337 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5338 - Don't setup a disk device that doesn't match exported UUID (317)
5339 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5340 - avoid committing NULL handle in force close
5341 - local.sh is now a one-stripe LOV configuration
5342 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5343 - fix ext3 htree duplicate directory entry corruption (1516)
5344 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5345 - update to new LOV EA format (2097)
5346 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5348 2003-06-15 Phil Schwan <phil@clusterfs.com>
5351 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5352 - per-import/export recovery handling (958, 931, 959)
5353 - multiple last-rcvd slots, for serving multiple FSes (949)
5354 - connections are again shared between multiple imp/exports (963, 964)
5355 - "umount -f" would hang if any requests needed to be sent (393, 978)
5356 - avoid pinning large req buffer by copying for queued messages (989)
5357 - add "uuid" to "lctl device" command to help upcalls (991)
5358 - "open" RPCs with transnos would confuse recovery counters (1037)
5359 - do proper endian conversion of all wire messages (288, 340, 891)
5360 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5361 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5362 - fix ptlrpc_connection leak in target_handle_connect (1174)
5363 - fix import refcounting bug in OST and MDS cleanup (1134)
5364 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5365 - fix very unlikely obd_types race condition (501)
5366 - remove osc_open hack for echo_client (1187)
5367 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5368 - a failure in read_inode2 leads to deadlock (1139)
5369 - cancel ack-locks as soon as transaction is committed (1072)
5370 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5371 - make sure to commitrw after any preprw to avoid deadlock (1162)
5372 - failing to execute a file in a lustre FS would lock inode (1203)
5373 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5374 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5375 - obd_uuid2tgt would walk off the end of the list (1255)
5376 - on IA64 the osc would give portals incorrect bulk size (1258)
5377 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5378 - fix lock inversion caused by new llite matching code (1282)
5379 - limit the number of dirty pages on a client to 10MB (1286)
5380 - timed out locks were not being corrected cancelled (1289)
5381 - fix O_DIRECT above 4GB on IA-32 (1292)
5382 * major user-visible changes
5383 - fail out/fail over policy now controlled by the upcall (993)
5385 - add OBD_PING to check server availability and failure (954)
5386 - lustre messages are now sent in sending host order (288, 340, 891)
5387 - add eadatalen to MDS getattr reply (340)
5388 - OST read replies may contain second buffer, with per-page status (593)
5390 2003-03-11 Phil Schwan <phil@clusterfs.com>
5393 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5394 - failed open()s could cause deadlock; fixed (867, 869)
5395 - stop cancelling OST locks when files are closed (481)
5396 - overlapping XID spaces caused network corruption (851, 853)
5397 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5398 - setattr_raw added, to do single-RPC, server-side setattrs
5399 - lmc/lconf syntax change for OST UUIDs
5400 - fix crashy race condition between ptlrpc_free_req and osc_close
5401 - don't use request in mdc_enqueue if we hit a timeout (889)
5402 - don't set the inode i_size for regular files from the MDS (896)
5403 - handle out of order completion AST (842)
5404 - don't LBUG if a lock request times out after receiving AST (913)
5405 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5406 - if a bad lock AST arrives, send an error instead of dropping entirely
5407 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5408 - fix leak in bulk IO when only partially completed (899, 900, 926)
5409 - fix O_DIRECT for ia64 (55)
5410 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5411 - C-z after timeout could hang a process forever; fixed (977)
5413 - client-side I/O cache (678, 924, 929, 941, 970)
5415 - READPAGE and SETATTRs which don't take server-side locks get
5418 2003-02-11 Phil Schwan <phil@clusterfs.com>
5421 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5422 - Fix fsfilt_extN_readpage() to read a full page of directory
5423 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5424 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5425 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5426 - Most intent operations were removed, in favour of a new RPC mode
5427 that does a single RPC to the server and bypasses most of the VFS
5428 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5429 - Aggressively cancel local locks on DLM servers
5430 - mds_reint_unlink sends EA to the client if it's the last nlink.
5431 client uses that EA to unlink OST objects.
5432 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5433 - recursive symlinks were fixed (439)
5434 - fixed NULL deref in DEBUG_REQ
5435 - filter_update_lastobjid no longer calls sync, which annoyed extN
5436 - fixed multi-client small-writes to a single file problem (445)
5437 - fixed mtime updates during file writes (607)
5438 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5439 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5440 - fixed Linux OST size reporting problem (444, 656)
5441 - OST now updates object mtime with writes or setattr (607, 619)
5442 - client verifies file size before zeroing page past EOF (445)
5443 - OST now writes last allocated objid to disk with allocation (108)
5444 - LOV on echo now works (409)
5446 - mds_reint_unlink sends a new buffer, with the EA included. this
5447 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5448 set if a regular file was being unlinked, and it was the last link
5449 - use PtlGet from the target for bulk writes (315)
5450 - OST now updates object mtime with writes or setattr (607, 619)
5451 - LDLM now has a grant-time callback to revalidate locked items, if
5453 - Many MDS operations were reorganized to combat race conditions
5455 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5456 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5458 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5461 - Fully reactivate OST imports after reconnection (512, others)
5462 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5463 - More graceful error handling for truncating on dead OST (515)
5464 - Don't error out unless we're actually accessing dead stripes (474)
5465 - Fix garbage sizes when stripes are missing (410)
5466 - LRU counters were broken, causing constant lock purge (433, 432)
5467 - garbage on read from stripes with failed OSTs (441)
5468 - mark OSCs as active before reconnecting during recovery (438)
5469 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5470 - lfind did not preserve OST order in output (443)
5471 - symlinks cause hung clients, incorrect data (439)
5472 - stop dereferencing request after dropping refcount (457)
5473 - don't LASSERT(spin_is_locked) on non-SMP (455)
5474 - fixes for many rename() bugs
5475 - fstat didn't correctly synchronize attributes (399)
5476 - server must handle lock cancellation during blocking AST prep (487)
5477 - bulk descriptors were free()d too soon (511)
5478 - fix paths in lconf, which would load incorrect modules (451, 507)
5479 - fix confusing lconf 'host not found' error message (386)
5480 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5481 - fix race condition in mdc_blocking_ast() for inode access (526)
5482 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5483 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5484 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5485 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5486 - fix lstripe in several ways (kernel side) (527)
5487 - fix request leak in ldlm_cli_enqueue (262)
5488 - incorrect OSC was marked inactive after OST failure
5489 - call mds_fs_cleanup before unmounting filesystem (524)
5490 - fix races between taking ns_lock and ldlm_lock_change_resource
5491 - fix races updating LOV export open file list
5492 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5493 - fix recovery NULL deref in ldlm_cli_cancel_unused
5494 - fix some DLM races by using new hash table for lock handles (419)
5495 - permit the client to specify desired inodes, at replay
5496 - duplicate requests when we queue them for replay reintegration
5497 - fix last_rcvd offset calculation
5498 - sync after each recovered transaction, so we always make progress
5499 - never, not always, ERESTART requests without transnos
5500 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5502 - skip replay if the MDS says that the client is already connected
5503 - don't check for a recovery-enabled export to match lctl's UUID
5504 - don't INC_USE_COUNT for phantom exports
5505 - don't crash when cleaning up phantom exports (567)
5506 - don't double-finish or set replay data for errored mdc_open requests
5507 - abort requests when they time out, so we don't get old replies
5508 - send/receive replies for AST messages again
5509 - if the client says that it doesn't have the lock, cancel it on the
5511 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5512 mark it as destroyed, it will all work out when decref is called
5513 - fix module use counts (22, 581)
5515 - ASTs now expect a reply (server cancels lock on error reply)
5517 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5520 - fix many simultaneous client startup (392)
5521 - fix dentry->d_it clobbering
5522 - credentials weren't being shipped for readdir/getattr operations
5523 - remove invalid assertions triggered during some concurrent MD
5525 - proper Lustre versions added (336, 389)
5526 - fix memory leak for create error case (398)
5527 - fix LOV locking bug that would get cli/srv out of sync
5528 - fix echo client over LOV (409)
5529 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5530 - fix double-O_EXCL intent crash (424)
5531 - avoid sending multiple lock CANCELs (352)
5533 - MDS can do multi-client recovery (modulo bugs in new code)
5535 - many updates, edits, cleanups
5537 2002-11-18 Phil Schwan <phil@clusterfs.com>
5540 - fix null d_it dereference (346)
5541 - fix full OST/dbench hang (333)
5542 - fix permission problem with file removal (286)
5543 - fix removal of OSCs from LOV when they fail
5544 - fix NULL deref during bulk timeout (214)
5545 - fix problems related to multiple filesystems on one MDS (241)
5546 - fixed serious subtle metadata locking bugs
5547 - free locks on clients when inodes are removed due to memory
5549 - fix inode pointer in lock data (285)
5550 - partial support for multiple MDS on a single host (241)
5551 - data locks weren't cancelled at clear_inode time (290, 311)
5552 - intent locks could lead to unbounded lock growth (205)
5553 - added a maximum lock count, an LRU list, and a flusher
5554 - fix multiple rename (365)
5555 - properly abstracted the echo client
5556 - OSC locked 1 byte too many; fixed
5557 - rewrote brw callback code:
5558 - fixed recovery bugs related to LOVs (306)
5559 - fixed too-many-pages-in-one-write crash (191)
5560 - fixed (again) crash in sync_io_timeout (214)
5561 - probably fixed callback-related race (385)
5563 - Add capability to MDS protocol
5564 - LDLM cancellations and callbacks on different portals
5566 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5569 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5570 - defer open object destruction to close time (601981, 138)
5571 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5572 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5573 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5574 - don't call obd_setattr in ll_file_release for destroyed objects
5575 * protocol change to lustre_msg: move |version| and add |flags|
5576 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5577 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5578 * added replay of create, unlink, link and rename operations during
5579 MDS failover; recovery should be much more robust now
5580 * remove failed OSCs from LOVs (only lov_create uses this so far)
5581 * the lustre-HOWTO was brought (more) up to date (582544)
5583 2002-10-23 Phil Schwan <phil@clusterfs.com>
5586 - in-use dentries weren't being reused properly (617851)
5587 - prevent multiple LDLM setup (599178)
5588 - fix LOV size calculations for truncate (617853)
5589 - fix client handling of MDS intent errors (POSIX)
5590 - fix permission bug in lovstripe.c test (624321)
5591 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5592 - truncate past end of file could corrupt data
5593 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5594 - a race in recovery could return ETIMEDOUT to apps (623947)
5595 - building outside the source directory was fixed
5596 * the lustre-HOWTO was brought (more) up to date (582544)
5597 * major progress was made on recovery functionality
5599 2002-10-10 Phil Schwan <phil@clusterfs.com>
5602 - recovery deadlock fix
5603 - rm -rf causes LBUG fix (617817)
5604 - file open by multiple tasks fix (618962)
5605 - directory permissions bugs (602707 and 620007)
5606 - journal_stop fixed with locking (611313)
5607 - O_APPEND failures resolved (618273, perhaps 614459)
5608 - lconf PATH fix (619770)
5609 - IA64 build fix (621450)
5610 - RPC buffer sizes scale with amount of memory
5612 2002-10-01 Phil Schwan <phil@clusterfs.com>
5615 - locks would be cancelled without throwing away data pages,
5616 resulting in inconsistent data (605627)
5617 - inode attributes were not always being refreshed (605627, 612449)
5618 - lconf now continues to cleanup after lctl reports an error
5619 - MDS now enforces user permissions (602707)
5620 - lprocfs cleanup fixed, but not yet enabled (614157)
5621 - fixed infinite server hang, should a client not respond to an AST
5622 - avoid going into recovery if user calls readlink() with a buffer
5623 that's too small (613941)
5624 - AST RPCs no longer require replies (614867) -- this may be changed
5625 - don't crash server if client sends an IOV that's too big (611336)
5626 - fixed lock conversion deadlock (611892)
5627 - fixed the following of symlinks (614622)
5628 * recovery: the server can remove locks from a client that dies, other
5629 clients can make progress
5630 * more extN patch fixes
5631 * compile-time configurable ptlrpc buffer allocations
5633 - collaborative read cache document
5634 - Lustre Lite Performance CDR document-in-progress
5636 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5639 - fix typo in patch-2.4.18
5641 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5644 - clear ptlrpc request each time in handle_incoming_request()
5645 - unlink of files now destroys the object on the OST
5647 2002-09-19 Peter Braam <braam@clusterfs.com>
5649 * add hard link support
5650 * change obdfile creation method
5651 * kernel patch changed
5653 2002-09-19 Peter Braam <braam@clusterfs.com>
5656 - stack overflow bug in extN fixed
5658 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5660 * documentation updates
5661 - add man pages for config tools
5662 - update tests/README to describe testing with new config tools
5663 - finish metadata API descriptions
5664 * bug fixes and cleanups
5665 - statfs workaround for 16TB limit
5666 - LOV stripe allocation improved, can stripe on subset of OSTs
5667 - LOV file size/IO offset was wrong for files > 4GB in size
5668 - object EA data was being dropped, caused files to be unreadable
5669 - memory overflow with non-LOV OST caused memory corruption
5670 - fixed regression tests to work with new config tools, obdfilter
5671 - fixed bug when directory size became larger than 1 block
5672 - fixed bug (for single client case) when PWD was deleted
5673 - invalidate local directory pages when doing intent-based ops
5674 - avoid LDLM oops when lock callback contained bad data
5676 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5678 * documentation updates
5679 * bug fixes and cleanups
5680 - configuration tools
5683 - 64-bit compile warnings
5684 - 64-bit internal statfs data
5686 * test_brw on persistent OST devices
5688 * lprocfs (disabled)
5690 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5692 * documentation updates
5693 * bug fixes and cleanups
5694 * configuration tools
5696 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5699 * many small fixes to 0_5_4
5700 * io/network handling
5701 * thinkos in MDS operations
5703 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5706 * crucial basic fixes to 0.5.3
5707 * IOR, Iozone work over Elan
5710 2002-08-07 Phil Schwan <phil@clusterfs.com>
5711 * version 0_5_3, our first alpha
5712 * we use the new Portals iovs
5713 * documentation updates
5714 * bug fixes and cleanups
5715 * small changes in the DLM wire protocol
5717 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5718 * version 0_5_1 with some initial stability,
5719 * locking on MD and file I/O.
5720 * documentation updates
5721 * several bug fixes since 0.5.0
5722 * small changes in wire protocol
5724 2002-07-18 Phil Schwan <phil@clusterfs.com>
5726 * delivered as Lustre Light Alpha
5727 * fixed a crash after handling invalid MDS requests
5728 * fixed directory pages for architectures with non-4k pages sizes
5730 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5731 * release version v0_4_4
5732 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5734 * Fixes a number of interruption problems with OST operations.
5735 * Update documentation for portals header changes
5736 * Move all wire protocol structs/defines to lustre_idl.h
5737 * Fixes symlink length bug.
5738 * Add tcpdump to repository.
5740 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5741 * release version v0_4_3
5742 * Fixes statfs for inodes on extN.
5743 * Fixes bug in runtests which would delete /etc/hosts.
5744 * Use 64-bit object IDs wherever possible (not into VFS though)
5745 Remove ost_get_info, which is unused by lustre, and out of date.
5747 2002-07-03 Peter Braam <braam@clusterfs.com>
5748 * release version v0_4_2 Fixes a lookup error (type not passed)
5749 * move forward to head of Portals
5750 * move forward to latest Lustre kernel
5752 2002-06-25 Peter Braam <braam@clusterfs.com>
5753 * release version v0_4_1. Hopefully stable on single node use.