1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.33 (SLES 10),
5 2.6.18-92.1.22.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.
17 Frequency : start MDS on uncleanly shutdowned MDS device
19 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
20 Details : stay in waiting mds<>ost recovery finished produce random bugs
21 due race between two ll_sync thread for one lov target. send
22 ACTIVATE event only if connect realy finished and import have
26 Frequency : rare, connect and disconnect target at same time
28 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
29 Details : don't call obd_disconnect under lov_lock. this long time
30 operation and can block ptlrpcd which answer to connect request.
35 Descriptoin: don't lose wakeup for imp_recovery_waitq
36 Details : recover_import_no_retry or invalidate_import and import_close can
37 both sleep on imp_recovery_waitq, but we was send only one wakeup
41 Frequency : always with long access acl
43 Descriptoin: mds can't pack reply with long acl.
44 Details : mds don't control size of acl but they limited by reint/getattr
47 Severity : enhancement
49 Description: Update to SLES10 kernel-2.6.16.60-0.33.
51 Severity : enhancement
53 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
56 Frequency : start MDS on uncleanly shutdowned MDS device
58 Descriptoin: aborting recovery hang on MDS
59 Details : don't throttle destroy RPCs for the MDT.
62 Frequency : on remount
64 Description: external journal device not working after the remount
65 Details : clear dev_rdonly flag for external journal devices in
71 Description: shutdown vs evict race
72 Details : client_disconnect_export vs connect request race.
73 if client will evicted at this time - we start invalidate
74 thread without referece to import and import can be freed
78 Frequency : rare, need acl's on inode.
80 Description: client can't handle ost additional correctly
81 Details : if ost was added after client connected to mds client can have
82 hit lnet_try_match_md ... to big messages to wide striped files.
83 in this case need teach client to handle config events about add
84 lov target and update client max ea size at that event.
86 Severity : enhancement
88 Description: Changelogs
89 Details : Changelogs are a lightweight mechanism to track filesystem
90 metadata and namespace changes. The changelog is recorded
91 permanently on the MDTs, and is periodically "consumed" / purged
92 when records are no longer needed.
94 Severity : enhancement
96 Description: compact fld format with extents
97 Details : Store range of seq rather than every seq in FLD. Seq
98 controller update FLD rather than clients. In Case of CMD, mdt0
99 has FLD, all other metadata server act as non persistent proxy
100 for FLD queries and cache fld entries in fld cache.
105 Description: don't skip ost target if they assigned to file
106 Details : Drop slow OSCs if we can, but not for requested start idx.
107 This means "if OSC is slow and it is not the requested
108 start OST, then it can be skipped, otherwise skip it only
109 if it is inactive/recovering/out-of-space.
113 Description: more cleanup in mds_lov
114 Details : not send LOV EA under replay, we can't know about they size at this
115 time. Don't allow client connect to mds before any ost connected,
116 for avoid problems with LOV EA size and returning EIO to client.
118 Severity : enhancement
120 Description: Interoperability at server side (Disk interoperability)
122 Severity : enhancement
124 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
126 Severity : enhancement
128 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
130 Severity : enhancement
132 Description: New client IO stack (CLIO).
134 Severity : enhancement
136 Description: Commit on sharing. Eliminate inter-client dependencies between
137 uncommitted transactions by doing transaction commits.
138 Thereby clients may recovery independently.
141 Frequency : Create a symlink file with a very long name
143 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
144 Details : If there is no extra space in the request for early cancels,
145 ldlm_req_handles_avail() returns 0 instead of a negative value.
147 Severity : enhancement
149 Description: Add /proc entry for import status
150 Details : The mdc, osc, and mgc import directories now have
151 an import directory that contains useful import data for debugging
154 Severity : enhancement
156 Description: Re-disable certain /proc logging
157 Details : Enable and disable client's offset_stats, extents_stats and
158 extents_stats_per_process stats logging on the fly.
161 Frequency : Only on FC kernels 2.6.22+
163 Description: oops in statahead
164 Details : Do not drop reference count for the dentry from VFS when lookup,
165 VFS will do that by itself.
167 Severity : enhancement
169 Description: Generic /proc file permissions
170 Details : Set /Proc file permissions in a more generic way to enable non-
171 root users operate on some /proc files.
175 Description: Hitting mdc_commit_close() ASSERTION
176 Details : Properly handle request reference release in
177 ll_release_openhandle().
181 Description: quota recovery deadlock during mds failover
182 Details : This patch includes att18982, att18236, att18237 in bz14840.
184 1. fix osts hang when mds does failover with quotaon
185 2. prevent watchdog storm when osts threads wait for the
190 Frequency : only patchless client
191 Description: add workaround for race between add/remove dentry from hash
193 Severity : enhancement
195 Description: Allow OST glimpses to return PW locks
199 Description: LBUG when llog conf file is full
200 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
204 Description: Prevent import from entering FULL state when server in recovery
208 Description: service mount cannot take device name with ":"
209 Details : Only when device name contains ":/" will mount treat it as
215 Description: replace ptlrpcd with the statahead thread to interpret the async
216 statahead RPC callback
220 Frequency : on recovery
221 Description: I/O failures after umount during fail back
222 Details : if client reconnected to restarted server we need join to recovery
223 instead of find server handler is changed and process self
224 eviction with cancel all locks.
226 Severity : enhancement
228 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
230 Severity : enhancement
232 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
234 Severity : enhancement
236 Description: Upcall on Lustre log has been dumped
237 Details : Allow for a user mode script to be called once a Lustre log has
238 been dumped. It passes the filename of the dumped log to the
239 script, the location of the script can be specified via
240 /proc/sys/lnet/debug_log_upcall.
245 Description: avoid idr_remove called for id which is not allocated.
246 Details : Move assigment s_dev for clustered nfs to end of initialization,
247 to avoid problem with error handling.
252 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
253 Details : When connection is reused this not moved from CONN_UNUSED_HASH
254 into CONN_USED_HASH and this prodice warning when put connection
255 again in unused hash.
258 Severity : enhancement
260 Description: File striping can now be set to use an arbitrary pool of OSTs.
262 Severity : enhancement
264 Description: Export bytes_read/bytes_write count on OSC/OST.
268 Description: Early reply size mismatch, MGC loses connection
269 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
270 the connect flags are properly negotiated.
275 Description: quotas are not honored with O_DIRECT
276 Details : all writes with the flag O_DIRECT will use grants which leads to
277 this problem. Now using OBD_BRW_SYNC to guard this.
281 Description: add quota statistics
282 Details : 1. sort out quota proc entries and proc code.
283 2. add quota statistics
285 Severity : enhancement
287 Description: enable quota support for HEAD.
291 Description: Properly propagate oinfo flags from lov to osc for statfs
292 Details : restore missing copy oi_flags to lov requests.
294 Severity : enhancement
296 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
297 : l_getgroups(8), lst(8), routerstat(8)
298 Details : included man pages for llobdstat(8), llstat(8),
299 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
301 Severity : enhancement
303 Description: configure's --enable-quota should check the
304 : kernel .config for CONFIG_QUOTA
305 Details : configure is terminated if --enable-quota is passed but
306 : no quota support is in kernel
310 Description: Remove portals compatibility
311 Details : Remove portals compatibility, not interoperable with releases
316 Description: Resolve device initialization race
317 Details : Prevent proc handler from accessing devices added to the
318 obd_devs array but yet be intialized.
320 Severity : enhancement
322 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
324 Severity : enhancement
326 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
331 Description: Using wrong pointer in osc_brw_prep_request
332 Details : Access to array[-1] can produce panic if kernel compiled with
333 CONFIG_PAGE_ALLOC enabled
337 Description: Client runs out of low memory
338 Details : Consider only lowmem when counting initial number of llap pages
342 Description: *optional* service tags registration
343 Details : if the "service tags" package is installed on a Lustre node
344 When the filesystem is mounted, a local-node service tag will
345 be created. See http://inventory.sun.com/ for more information
346 about the Service Tags asset management system.
350 Description: Kernel BUG tries to release flock
351 Details : Lustre does not destroy flock lock before last reference goes
352 away. So always drop flock locks when client is evicted and
353 perform unlock regardless of successfulness of speaking to MDS.
357 Description: add refcount protection for osc callbacks, avoid panic on shutdown
361 Description: sanity test 65a fails if stripecount of -1 is set
362 Details : handle -1 striping on filesystem in ll_dirstripe_verify
367 Description: ASSERTION(CheckWriteback(page,cmd)) failed
368 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
371 Severity : enhancement
373 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
377 Description: do not process already freed flock
378 Details : flock can possibly be freed by another thread before it reaches
379 to ldlm_flock_completion_ast.
383 Description: LBUG during stress test
384 Details : Need properly lock accesses the flock deadlock detection list.
388 Description: oops in page fault handler
389 Details : kernel page fault handler can return two special 'pages' in error
390 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
394 Description: timeout with invalidate import.
395 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
396 should be handled by ptlrpcd. This produce long age waiting and
397 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
399 Severity : enhancement
401 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
405 Description: enable MGS and MDT services start separately
406 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
407 a co-located MGS without starting the MGS, which is a complement
408 to 'nosvc' mount option.
412 Frequency : after recovery
413 Description: precreate to many object's after del orphan.
414 Details : del orphan st in oscc last_id == next_id and this triger growing
415 count of precreated objects. Set flag LOW to skip increase count
416 of precreated objects.
420 Frequency : rare, on clear nid stats
421 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
422 Details : when clean nid stats sometimes try destroy live entry,
423 and this produce panic in free.
427 Description: Stack overflow during MDS log replay
428 ease stack pressure by using a thread dealing llog_process.
432 Description: wait until IO finished before start new when do lock cancel.
433 Details : VM protocol want old IO finished before start new, in this case
434 need wait until PG_writeback is cleared until check dirty flag and
435 call writepages in lock cancel callback.
437 Severity : enhancement
439 Description: using special macro for print time and cleanup in includes.
443 Description: mds_mfd_close() ASSERTION(rc == 0)
444 Details : In mds_mfd_close(), we need protect inode's writecount change
445 within its orphan write semaphore to prevent possible races.
449 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
453 Frequency : rare, on shutdown ost
454 Description: don't hit live lock with umount ost.
455 Details : shrink_dcache_parent can be in long loop with destroy dentries,
456 use shrink_dcache_sb instead.
460 Description: don't panic with use echo client
461 Details : echo client pass NULL as client nid pointer and this produce null
466 Description: fix build on ppc32
467 Details : compile code with -m64 flag produce wrong object file for ppc32.
471 Description: add message levels for liblustreapi
475 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
476 Details : Move the 'good_osts' check before the 'total_bavail' check. This
477 will result in an -EAGAIN and in the exit call path we call
478 alloc_rr() which will with increasing aggressiveness attempt to
479 aquire precreated objects on the minimum number of required OSCs.
483 Description: Use old size assignment to avoid deadlock
484 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
485 the scheduling while holding a spinlock. We do not need locking
486 for size in ll_update_inode() because size is only updated from
487 the MDS for directories or files without objects, so there is no
488 other place to do the update, and concurrent access to such inodes
489 are protected by the inode lock.
493 Description: resolve "_IOWR redefined" build error on SLES10
497 Description: dump the memory debugging after all modules are unloaded to
498 suppress false negative in conf_sanity test 39
500 Severity : enhancement
502 Description: build kernel-ib packages for OFED 1.3 in our release cycle
507 Description: fix SLES kernel versioning
508 Details : the kernel version for our SLES 10 kernel did not include a "-"
509 before the "smp" at the end. while this was not a problem in
510 general, it did mean that software trying to use the kernel
511 version to try to detect a vendor specific kernel would fail.
512 this was most evident by the OFED build scripts.
516 Description: Don't update lov_desc members until making sure they are valid
517 Details : When updating lov_desc members via proc fs, need fix their
518 validities before doing the real update.
522 Description: don't put request into delay list while invalidate in flight.
523 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
524 import in flight. this produce timeout for invalidate and sometimes
525 can cause stale data.
529 Frequency : on ppc only
530 Description: not convert ost objects for directory because it's not exist.
531 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
535 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
537 Severity : enhancement
539 Description: Update to support 2.6.22.14 vanilla kernel.
543 Frequency : rare, on recovery
544 Description: read procfs can produce deadlock in some situation
545 Details : Holding lprocfs lock which send rpc can produce block for destroy
546 obd objects and this also block reconnect with -EALREADY. This
547 isn't fix all lprocfs bugs - but make it rare.
549 Severity : enhancement
551 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
554 Frequency : frequent on X2 node
556 Description: mdc_set_open_replay_data LBUG
557 Details : Set replay data for requests that are eligible for replay.
561 Description: lustre_mgs: operation 101 on unconnected MGS
562 Details : When MGC is disconnected from MGS long enough, MGS will evict the
563 MGC, and late on MGC cannot successfully connect to MGS and a lot
564 of the error messages complaining that MGS is not connected.
568 Frequency : on network error
569 Description: panic with double free request if network error
570 Details : mdc_finish_enqueue is finish request if any network error ocuring,
571 but it's true only for synchronus enqueue, for async enqueue
572 (via ptlrpcd) this incorrect and ptlrpcd want finish request
575 Severity : enhancement
577 Description: client-side metadata stat-ahead during readdir(directory readahead)
578 Details : perform client-side metadata stat-ahead when the client detects
579 readdir and sequential stat of dir entries therein
582 Frequency : on start mds
584 Description: Implement get_info(last_id) in obdfilter.
587 Frequency : occasional
589 Description: Correctly check stale fid, not start epoch if ost not support SOM
590 Details : open with flag O_CREATE need set old fid in op_fid3 because
591 op_fid2 was overwritten with new generated fid, but mds can answer
592 with one of these two fids and both is not stale. Setattr
593 incorrectly started an epoch and assume will be called
594 done_writing, but without SOM done_writing ever being called.
597 Frequency : rare, depends on device drivers and load
599 Description: MDS or OSS nodes crash due to stack overflow
600 Details : Code changes in 1.8.0 increased the stack usage of some functions.
601 In some cases, in conjunction with device drivers that use a lot
602 of stack the MDS (or possibly OSS) service threads could overflow
603 the stack. One change which was identified to consume additional
604 stack has been reworked to avoid the extra stack usage.
607 Frequency : occasional
609 Description: Do not fail import if osc_interpret_create gets -EAGAIN
610 Details : If osc_interpret_create got -EAGAIN it immediately exits and
611 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
612 oscc_has_objects and see OSC has no objests and call
613 oscc_internal_create to resend create request.
615 Severity : enhancement
617 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
619 Severity : enhancement
621 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
624 Frequency : very rare
626 Description: Fix replay if there is an un-replied request and open
627 Details : In some cases, older replay request will revert the
628 mcd->mcd_last_xid on MDS which is used to record the client's
631 Severity : enhancement
633 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
635 Severity : enhancement
637 Description: Add rhel5 support to HEAD.
639 Serverity : enhancement
641 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
646 Description: Don't allow skipping OSTs if index has been specified.
647 Details : Don't allow skipping OSTs if index has been specified, make
648 locking in internal create lots better.
652 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
653 Details : access to bitfield in structure is always rounded to long
654 and this produce problem with not atomic change any bit.
658 Description: Lustre make rpms failed.
659 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
660 builds Lustre rpms from tarball.
663 Frequency : rare on shutdown ost
665 Description: If llog cancel was not send before clean_exports phase, this can
666 produce deadlock in llog code.
667 Details : If llog thread has last reference to obd and call class_import_put
668 this produce deadlock because llog_cleanup_commit_master wait when
669 last llog_commit_thread exited, but this never success because was
670 called from llog_commit_thread.
674 Description: allow userland application know is lost one of stripes.
675 Details : fill lvb_blocks with error code on ost and return it to
676 application if error flag found.
680 Description: NULL lov_tgts causing MDS oops
681 Details : more safe checks for NULL lov_tgts for avoid oops.
683 Severity : enhancement
685 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
689 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
690 Details : If an OST is down the MDS will hang indefinitely in
691 obd_statfs_rqset() waiting for the statfs data. While for
692 MDS QOS usage of statfs, it should not stuck in waiting.
694 Severity : enhancement
696 Description: remote_acl support
697 Details : Support ACL-based permission check for remote user.
698 Support setfacl/getfacl for remote user with the utils
699 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
700 the system "{s,g}etfacl" utils.
702 Severity : enhancement
704 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
706 Severity : enhancement
708 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
712 Description: make mgs_setparam() handle fsname containing dash
713 Details : fsname containing a dash does not work with lctl conf_param
715 Severity : enhancement
717 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
719 Severity : enhancement
721 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
725 Frequency : rare, at shutdown
726 Description: access already free / zero obd_namespace.
727 Details : if client_disconnect_export was called without force flag set,
728 and exist connect request in flight, this can produce access to
729 NULL pointer (or already free pointer) when connect_interpret
730 store ocd flags in obd_namespace.
734 Frequency : only at startup
735 Description: not alloc memory with spinlock held.
736 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
737 if any spinlock held.
739 Severity : enhancement
741 Description: make lustre randomly fail allocating memory
742 Details : Make lustre randomly failed allocating memory for testing purpose.
744 Severity : enhancement
746 Description: lost problems with lov objid file
747 Details : Fixes some scability and access to not inited memory problems
748 in work with lov objdid file.
753 Description: lfs find does not continue on file error
754 Details : Continue other files processing when a file/dir is absent.
758 Description: Inconsistent usage of lustre_pack_reply()
759 Details : Standardize the usage of lustre_pack_reply() such that it
760 always generate a CERROR on failure.
763 Frequency : occasional
765 Description: MDS hang or stay in waiting lock
766 Details : If client receive lock with CBPENDING flag ldlm need send lock
767 cancel as separate rpc, to avoid situation when cancel request
768 can't processed due all i/o threads stay in wait lock.
772 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
776 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
780 Description: Update to SLES9 kernel-2.6.5-7.287.3.
784 Description: Update to RHEL4 latest kernel.
786 Severity : enhancement
788 Description: Build SLES10 patchless client fails
789 Details : The configure was broken by run ./configure with
790 --with-linux-obj=.... argument for patchless client. When the
791 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
792 can't search header adequately. Use absolute path such as
793 -I($LINUX)/include instead.
797 Description: interrupt oig_wait produce painc on resend.
798 Details : brw_redo_request can be used for resend requests from ptlrpcd and
799 private set, and this produce situation when rq_ptlrpcd_data not
800 copyed to new allocated request and triggered LBUG on assert
801 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
802 ptlrpcd set if request is changed and can be safety changed to use
807 Description: LASSERT_{REQ,REP}SWAB macros are buggy
808 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
809 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
810 these macros and replace them with their difinitions instead.
814 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
815 Details : Update which_patch & target file for SLES10 latest kernel.
818 Description: Remove client patches from SLES 10 kernel.
819 Details : This causes SLES 10 clients to behave as patchless clients
820 even on a Lustre-patched (server) kernel.
822 Severity : enhancement
824 Description: self-adjustable client's lru lists
825 Details : use adaptive algorithm for managing client cached locks lru
826 lists according to current server load, other client's work
827 pattern, memory activities, etc. Both, server and client
828 side namespaces provide number of proc tunables for controlling
831 Severity : enhancement
833 Description: light-weight GSS support
834 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
835 and reduce performance overhead.
837 Severity : enhancement
839 Description: Linux keyring support
840 Details : Support using service of Linux keyring for Lustre GSS internal
841 context refresh/cache mechanism.
845 Description: Fix errors in lfs documentation
846 Details : Fixes man pages
850 Description: don't use GFP_* in generic Lustre code.
851 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
855 Description: obdclass is limited by single OBD_ALLOC(idarray)
856 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
860 Description: warnings with build patchless client with vanila 2.6.19 and up
861 Details : change old ctl_table style and replace ctl_table/ctl_table_header
862 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
866 Frequency : Only for SLES
867 Description: sanity_quota fail test_1
868 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
869 file, which makes TSTID[2] inunique.
873 Frequency : Only for SLES9
874 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
875 Details : Update target/ChangeLog/which_patch .
879 Frequency : Only for test_10 in sanity-quota.sh
880 Description: a bug in quota test code
881 Details : checking a test flag in a wrong place
885 Frequency : Only for test_13 in sanity-quota.sh
886 Description: a bug in quota test script
887 Details : shouldn't check overfull allocated quota
891 Description: jbd statistics
892 Details : Port older jbd statistics patch for sles10
894 Severity : enhancement
896 Description: Add printing inode info into message about error in writepage.
901 Description: reply_lock_interpret crash due to race with it and lock cancel.
902 Details : Do not replay locks that are being cancelled. Do not reference
903 locks by their address during replay, just by their handle.
907 Frequency : When flocks are used.
908 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
909 Details : Flock locks might destroy just granted lock if it could be merged
910 with another existing flock, this is done in completion handler,
911 so teach ldlm_cli_enquque_fini that this is a valid case for
917 Description: Oops in read and write path when failing to allocate lock.
918 Details : Check if lock allocation failed and return error back.
922 Description: lstripe command fails for valid OST index
923 Details : The stripe offset is compared to lov->desc.ld_tgt_count
924 instead of lov->desc.ld_active_tgt_count.
928 Description: sanity.sh failed test 103
929 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
934 Description: df doesn't work properly if diskfs blocksize != 4K
935 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
939 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
940 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
942 Severity : enhancement
944 Description: Adaptive timeouts
945 Details : RPC timeouts adapt to changing server load and network
946 conditions to reduce resend attempts and improve recovery time.
950 Description: llapi_file_create() does not allow some changes
951 Details : add llapi_file_open() that allows specifying the mode and
952 open flags, and also returns an open file handle.
956 Description: merge and cleanup kernel patches.
957 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
961 Description: Add journal checksum support.(Kernel part)
962 Details : The journal checksum feature adds two new flags i.e
963 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
964 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
965 indicates that the commit block contains the checksum for
966 the blocks described by the descriptor blocks. Now commit
967 record can be sent to disk without waiting for descriptor
968 blocks to be written to disk. This behavior is controlled
969 using JBD2_FEATURE_ASYNC_COMMIT flag.
973 Description: OSS needs mutliple precreate threads
974 Details : Add ability to start more than one create thread per OSS.
978 Description: Sanity.sh test_65k failed: setstripe should have succeeded
979 Details : execute lfs setstripe on client
983 Description: mds_obd_create error creating tmp object
984 Details : When the user sets quota on root, llog will be affected and can't
985 create files and write files.
988 Frequency : Always on ia64 patchless client, and possibly others.
990 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
991 Details : This allows the patchless client to be loaded on architectures
996 Description: RedHat Update kernel for RHEL5
997 Details : Add the kernel config file for RHEL5.
1001 Description: RedHat Update kernel for RHEL5
1002 Details : Modify the kernel config file more closer RHEL5.
1006 Description: Build failure against Centos5 (RHEL5)
1007 Details : Define PAGE_SIZE when it isn't present.
1011 Description: test_7 on sanity-quota.sh can't be used on separate machines
1012 Details : add facet to handle it
1016 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1017 Details : with NFS, the anon dentry's parent was set to itself in
1018 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1019 corresponding dentry other than use rec->ur_name.
1021 Severity : enhancement
1023 Description: lfs setstripe enhancement
1024 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1028 Description: enable data checksumming by default
1029 Details : enable checksum by default, allow --disable-checksum
1030 configure option and "-o nochecksum" mount option. Checksums
1031 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1035 Description: lustre support for RHEL5
1036 Details : Add support for RHEL5.
1040 Description: Client eviction due to failover config
1041 Details : after a connection loss, the lustre client should attempt to
1042 reconnect to the last active server first before trying the
1043 other potential connections.
1047 Description: when mds and osts use different quota unit(32bit and 64bit),
1048 quota will be released repeatly.
1049 Details : void sending multiple quota reqs to mds, which will keep the
1050 status between the reqs.
1054 Description: osts not allocated evenly to files
1055 Details : change the condition to increase offset_idx
1058 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1060 Description: Data corruption for OSTs that are formatted larger than 2TB
1062 Details : When generating the bio request for lustre file writes the
1063 sector number would overflow a temporary variable before being
1064 used for the IO. The data reads correctly from Lustre (which
1065 will overflow in a similar manner) but other file data or
1066 filesystem metadata may be corrupted in some cases.
1070 Description: Tune the kernel for good SCSI performance.
1071 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1072 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1077 Description: rewrite ext2-derived code in obdclass/uuid.c
1078 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1079 from scratch preserving functionality.
1083 Description: "lfs find -obd UUID" prints directories
1084 Details : "lfs find -obd UUID" will return all directory names instead
1085 of just file names. It is incorrect because the directories
1086 do not reside on the OSTs.
1090 Description: use i_size_read and i_size_write in 2.6 port
1091 Details : replace inode->i_size access with i_size_read/write()
1094 Frequency : when removing large files
1096 Description: scheduling issue during removal of large Lustre files
1097 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1098 It causes scheduling issues when removing large files (17TB in the
1102 Frequency : only with liblustre clients on XT3
1104 Description: evictions taking too long
1105 Details : allow llrd to evict clients directly on OSTs
1108 Frequency : only on ppc
1110 Description: /proc/fs/lustre/devices broken on ppc
1111 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1112 We should make sure the type of 'index' is loff_t and then cast
1113 explicitly as needed below. Do not assign an explicitly cast
1117 Frequency : only for rhel5
1119 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1120 Details : Modify the target file & which_kernel.
1122 Severity : enhancement
1124 Description: omit set fsid for export NFS
1125 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1126 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1130 Frequency : Always, for kernels after 2.6.16
1131 Description: Fix warning idr_remove called for id=.. which is not allocated.
1132 Details : Last kernels save old s_dev before kill super and not allow
1133 to restore from callback - restore it before call kill_anon_super.
1137 Description: buffer overruns could theoretically occur
1138 Details : llapi_semantic_traverse() modifies the "path" argument by
1139 appending values to the end of the origin string, and a
1140 overrun may occur. Adding buffer overrun check in liblustreapi.
1144 Description: Fix error on 'ls .' at the top of the Lustre mount.
1145 Details : Don't revalidate dentry if it is a root dentry.
1149 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1150 Details : Modify vm-tunables-rhel4.patch.
1154 Description: Kernel config for 2.6.18-vanilla.
1155 Details : Modify targets/2.6-vanilla.target.in.
1156 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1157 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1158 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1159 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1163 Description: improve handling recoverable errors
1164 Details : if request processig with error which can be recoverable on server
1165 request should be resend, otherwise page released from cache and
1170 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1171 Details : Modify target file & which_patch.
1172 A flaw was found in the IA32 system call emulation provided
1173 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1174 value could be stored in the %RAX register, which could trigger an
1175 out-of-bounds system call table access. An untrusted local user
1176 could exploit this flaw to run code in the kernel
1177 (ie a root privilege escalation). (CVE-2007-4573).
1181 Description: O_DIRECT bypasses client statistics.
1182 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1183 read_ahead_stats were updated. copied the stats section from
1184 osc_send_oap_rpc() into async_internal().
1188 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1192 Description: change order of libsysio includes
1193 Details : '#include sysio.h' should always come before '#include xtio.h'
1197 Description: handle "serious error: objid * already exists" more gracefully
1198 Details : If LAST_ID value on disk is smaller than the objects existing in
1199 the O/0/d* directories, it indicates disk corruption and causes an
1200 LBUG(). If the object is 0-length, then we should use the existing
1201 object. This will help to avoid a full fsck in most cases.
1203 Severity : enhancement
1205 Description: adapt the lustre_config script to support the upgrade case
1206 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1207 server targets to 1.6 in parallel.
1211 Description: To avoid grant space > avaible space when the disk is almost
1212 full. Without this patch you might see the error "grant XXXX >
1213 available" or some LBUG about grant, when the disk is almost
1215 Details : In filter_check_grant, for non_grant cache write, we should
1216 check the left space by if (*left > ungranted + bytes), instead
1217 of (*left > ungranted), because only we are sure the left space
1218 is enough for another "bytes", then the ungrant space should be
1219 increase. In client, we should update cl_avail_grant only there
1220 is OBD_MD_FLGRANT in the reply.
1224 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1225 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1226 described in CVE-2007-4573. This problem could allow untrusted
1227 local users to gain root access.
1230 Frequency : when using O_DIRECT and quotas
1232 Description: Incorrect file ownership on O_DIRECT output files
1233 Details : block usage reported by 'lfs quota' does not take into account
1234 files that have been written with O_DIRECT.
1239 Description: touch file failed when fs is not full
1240 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1241 otherwise we can get ENOSP while fs is not full.
1245 Description: parallel lock callbacks
1246 Details : Instead of sending blocking and completion callbacks as separated
1247 requests, adding them to a set and sending in parallel.
1250 Frequency : only for Cray XT3
1251 Bugzilla : 12829/13455
1252 Description: Changing primary group doesn't change the group lustre assigns to
1254 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1255 overridden with the primary group provided by the group upcall,
1256 whereas the supplied fsgid can be trusted if it is in the list of
1257 supplementary groups returned by the group upcall.
1259 Severity : enhancement
1261 Description: Allow masking D_WARNING, D_ERROR messages from console
1262 Details : Console messages can now be disabled via lnet.printk.
1266 Description: User code with malformed file open parameter crashes client node
1267 Details : Before packing join_file req, all the related reference should be
1268 checked carefully in case some malformed flags cause fake
1269 join_file req on client.
1273 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1274 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1275 mode is equal to the granted mode. However, it should also check
1276 that LDLM_FL_CP_REQD is not set.
1280 Description: Heavy nfs access might result in deadlocks
1281 Details : After ELC code landed, it is now improper to enqueue any mds
1282 locks under och_sem, because enqueue might want to decide to
1283 cancel open locks for same inode we are holding och_sem for.
1287 Description: Client eviction while running blogbench
1288 Details : A lot of unlink operations with concurrent I/O can lead to a
1289 deadlock causing evictions. To address the problem, the number of
1290 oustanding OST_DESTROY requests is now throttled to
1291 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1292 callbacks are processed in priority.
1296 Description: enable ACLs on MDS by default
1297 Details : ACLs must be enabled on MDS by default.
1300 Frequency : PPC/PPC64 only
1302 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1303 Details : fix duplicated definitions between asm-ppc64/types.h and
1304 lustre_types.h on PPC.
1307 Frequency : PPC/PPC64 only
1309 Description: asm-ppc/segment.h does not exist
1310 Details : fix compile issue on PPC.
1314 Description: better handle error messages in extents code
1317 Frequency : RHEL4 only
1319 Description: mkfs is very slow on IA64/RHEL4
1320 Details : A performance regression has been discovered in the MPT Fusion
1321 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1322 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1323 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1325 Severity : enhancement
1327 Description: SNMP support enhancement
1328 Details : Adding total number of sampled request for an MDS node in snmp
1331 Severity : enhancement
1333 Description: Optimize ldlm waiting list processing for PR extent locks
1334 Details : When processing waiting list for read extent lock and meeting read
1335 lock that is same or wider to it that is not contended, skip
1336 processing rest of the list and immediatelly return current
1337 status of conflictness, since we are guaranteed there are no
1338 conflicting locks in the rest of the list.
1342 Description: Time out and refuse to reconnect
1343 Details : When the failover node is the primary node, it is possible
1344 to have two identical connections in imp_conn_list. We must
1345 compare not conn's pointers but NIDs, otherwise we can defeat
1346 connection throttling.
1350 Description: port llog fixes from b1_6 into HEAD
1351 Details : Port llog reference couting and some llog cleanups from b1_6
1352 (bug 10800) into HEAD, for protect from panic and access to already
1353 free llog structures.
1357 Description: Detect stride IO mode in read-ahead
1358 Details : When a client does stride read, read-ahead should detect that and
1359 read-ahead pages according to the detected stride pattern.
1363 Description: data checksumming impacts single node performance
1364 Details : add support for several checksum algorithm. Currently, only CRC32
1365 and Adler-32 are supported. The checksum type can be changed on
1366 the fly via /proc/fs/lustre/osc/*/checksum_type.
1370 Description: use adler32 for page checksums
1371 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1376 Description: build for x2 fails
1377 Details : fix compile issue on Cray systems.
1381 Description: Properly match for duplicate locks
1382 Details : Due to different lock order from skiplists code, we need to
1383 traverse entire list for now
1386 Frequency : only on PPC/SLES10
1388 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1389 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1390 be defined. Add a hack in mkfs_lustre.c to work around this compile
1395 Description: LASSERT on MDS when client holding flock lock dies
1396 Details : ldlm pool logic depends on number of granted locks equal to
1397 number of released locks which is not true for flock locks, so
1398 just exclude such locks from consideration.
1402 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1403 Details : Use fsfilt_sync() for both the whole filesystem sync and
1404 individual file sync to eliminate dangerous inode locking
1405 with I_LOCK that can lead to a deadlock.
1409 Description: performance in 1.6.3
1410 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1411 sure that 1MB requests are not fragmented by the block layer.
1413 Severity : enhancement
1415 Description: organize the server-side client stats on per-nid basis
1416 Details : Change the structure of stats under obdfilter and mds to
1424 The "uuid"s file would list the uuids of _active_ exports.
1425 And the clear entry is to clear all stats and stale nids.
1427 Severity : enhancement
1429 Description: eliminate client locks in face of contention
1430 Details : file contention detection and lockless i/o implementation
1431 for contended files.
1435 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1438 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1440 Description: MDS may lose file striping (and hence file data) in some cases
1441 Details : If there are additional extended attributes stored on the MDS,
1442 in particular ACLs, SELinux, or user attributes (if user_xattr
1443 is specified for the client mount options) then there is a risk
1444 of attribute loss. Additionally, the Lustre file striping
1445 needs to be larger than default (e.g. striped over all OSTs),
1446 and an additional attribute must be stored initially in the
1447 inode and then increase in size enough to be moved to the
1448 external attribute block (e.g. ACL growing in size) for file
1453 Description: skiplist implementation simplification
1454 Details : skiplists are used to group compatible locks on granted list
1455 that was implemented as tracking first and last lock of each
1456 lock group the patch changes that to using doubly linked lists
1460 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1461 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1462 because LASSERT on both the data supplied by a client, and the
1463 data on disk is dangerous and incorrect.
1465 Severity : enhancement
1467 Description: Slow truncate/writes to huge files at high offsets.
1468 Details : Directly associate cached pages to lock that protect those pages,
1469 this allows us to quickly find what pages to write and remove
1470 once lock callback is received.
1474 Description: more ldlm soft lockups
1475 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1476 starve other threads from the resource lock for a long time in
1477 case of long waiting queue, so change the debug level from
1478 D_OTHER to the less frequently used D_INFO.
1480 Severity : enhancement
1482 Description: add -gid, -group, -uid, -user options to lfs find
1486 Description: Hung threads in invalidate_inode_pages2_range
1487 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1488 once one is completed. As a result, some RPCs are stuck in the
1489 queue and are never sent.
1493 Description: filter threads hungs on waiting journal commit
1494 Details : Cleanup filter group llog code, then only filter group llog will
1495 be only created in the MDS/OST syncing process.
1499 Description: Procfs and llog threads access destoryed import sometimes.
1500 Details : Sync the import destoryed process with procfs and llog threads by
1501 the import refcount and semaphore.
1503 Severity : enhancement
1505 Description: openlock cache of b1_6 port to HEAD
1510 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1511 Details : If insertion of an extent fails, then discard the inode
1512 preallocation and free data blocks else it can lead to duplicate
1517 Description: don't always update ctime in ext3_xattr_set_handle()
1518 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1519 In some cases the ctime should not be updated, for example for
1520 2.0->1.8 compatibility it is necessary to delete an xattr and it
1521 should not update the ctime.
1525 Bugzilla : 15713/16362
1526 Description: Assertion in iopen_connect_dentry in 1.6.3
1527 Details : looking up an inode via iopen with the wrong generation number can
1528 populate the dcache with a disconneced dentry while the inode
1529 number is in the process of being reallocated. This causes an
1530 assertion failure in iopen since the inode's dentry list contains
1531 both a connected and disconnected dentry.
1535 Description: assertion failure in ldlm_handle2lock()
1536 Details : fix a race between class_handle_unhash() and class_handle2object()
1537 introduced in lustre 1.6.5 by bug 13622.
1542 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1543 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1548 Description: oops in sd_iostats_seq_show()
1549 Details : unloading/reloading the scsi low level driver triggers a kernel
1550 bug when trying to access the sd iostat file.
1555 Description: Kernel panics during QLogic driver reload
1556 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1557 patch, causing memory corruption.
1562 Description: journal_dev option does not work in b1_6
1563 Details : pass mount option during pre-mount.
1565 Severity : enhancement
1567 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1568 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1569 extent information of a file. It can be used to map logical blocks
1570 in a file to physical blocks in the block device.
1574 Description: LDLM soft lockups - improvement
1575 Details : It is be possible to send the lock handle along with each read
1576 or write request because the client is already doing a lock match
1577 itself so there isn't any reason the OST should have to re-do that
1581 Frequency : only on Cray X2
1583 Description: X2 build failures
1584 Details : fix build failures on Cray X2.
1588 Description: xid & resent requests
1589 Details : Initialize RPC XID from clock at startup (randomly if clock is
1592 Severity : enhancement
1594 Description: Add lustre_start utility to start or stop multiple Lustre servers
1599 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1600 Details : In case of memory pressure, list_del() can be called twice on
1601 req->rq_history_list, causing a kernel oops.
1605 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1606 Details : fix stack overflow in the distributed lock manager by defering
1607 export eviction after a failed AST to the elt thread instead of
1608 handling it in the dlm interpret routine.
1612 Description: Convert some comments to new format.
1613 Details : Update documenting comments to match doxygen conventions.
1617 Description: Grammar fixes.
1618 Details : A couple of trivial spelling fixes.
1622 Description: OSD_COUNTERS-mandatory
1623 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1624 to simplify the code.
1628 Description: simplify cmm_device freeing logic.
1629 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1633 Description: Add lockdep support to dt_object_operations locking interface.
1634 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1635 parameter indicating lock ordering. Update mdd code to use new
1640 Description: Introduce failloc constants for lockless IO tests.
1641 Details : Add two new failloc constants to test lockless IO. Only one of
1642 them in implemented---another is checked in yet to be landed
1647 Description: Add lockdep support for inode mutex.
1648 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1649 in the situations where Lustre has to lock more than one inode
1654 Description: Add optional invariants checking support.
1655 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1656 configure switch. This macro is to be used for consistency and
1657 sanity checks that are too expensive to be left in `production'
1662 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1663 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1668 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1669 Details : Use spin_lock_nested() in (the only) situation where more than
1670 one ldlm_lock is locked simultaneously. Also, fix possible
1671 dead-lock in ldlm_lock_change_resource() by enforcing particular
1676 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1677 Details : Instead of specifying each ldlm_lock call-back through separate
1678 parameter, wrap them into struct ldlm_callback_suite.
1682 Description: Kill join_lru obd method and its callers.
1683 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1684 in memory---a requirement implemented through ->o_join_lru() obd
1685 method in HEAD. Remove this method and its users.
1689 Description: Add asynchronous ldlm ENQUEUE completion handler.
1690 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1691 case that stock ldlm_completion_ast() cannot handle as it waits
1692 until lock is granted. Introduce new ldlm_completion_ast_async()
1693 for this. Also comment ldlm_completion_ast().
1697 Description: ldlm_error <-> errno conversion.
1698 Details : Add functions to map (rather arbitrary) between LDLM error codes
1699 and standard errno values. CLIO needs this to prevent LDLM specific
1700 constants from escaping ldlm and osc.
1704 Description: Kill unused ldlm_handle2lock_ns() function.
1705 Details : Kill unused ldlm_handle2lock_ns() function.
1709 Description: Add lu_ref support to ldlm_lock
1710 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1711 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1712 and ldlm_resource. LDLM interface has to be changed, because code that
1713 releases a reference on a lock, has to "know" what reference this is.
1714 In the most frequent case
1716 lock = ldlm_handle2lock(handle);
1718 LDLM_LOCK_PUT(lock);
1720 no changes are required. When any other reference (received _not_ from
1721 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1722 instead of LDLM_LOCK_PUT().
1724 Arguably, changes are pervasive, and interface requires some discipline
1725 for proper use. On the other hand, it was very instrumental in finding
1726 a few leaked lock references.
1730 Description: Add ldlm_lock_addref_try().
1731 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1732 attempts to addref a lock that might be being canceled
1737 Description: Add ldlm_weigh_callback().
1738 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1739 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1740 hard-coded `number of pages' logic.
1744 Description: Add lockdep annotations to llog code.
1745 Details : Use appropriately tagged _nested() locking calls in the places
1746 where llog takes more than one ->lgh_lock lock.
1750 Description: Add loi_kms_set().
1751 Details : Wrap kms updates into a helper function.
1755 Description: Constify instances of struct lsm_operations.
1756 Details : Constify instances of struct lsm_operations.
1760 Description: lu_conf support.
1761 Details : On a server, a file system object is uniquely identified
1762 by a fid, which is sufficient to locate and load all object
1763 state (inode). On a client, on the other hand, more data are
1764 necessary instantiate an object. Change lu_object_find() and
1765 friends to take additional `lu_conf' argument describing object.
1766 Typically this includes layout information.
1770 Description: lu_context fixes.
1771 Details : Introduce new lu_context functions that are needed on the client
1772 side, where some system threads (ptlrpcd) are shared by multiple
1773 modules, and so cannot be stopped during module shutdown.
1777 Description: Add start and stop methods to lu_device_type_operations.
1778 Details : Introduce two new methods in lu_device_type_operations, that are
1779 invoked when first instance of a given type is created and last one
1780 is destroyed respectively. This is need by CLIO.
1784 Description: Add lu_ref support to struct lu_device.
1785 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1786 track leaked references.
1790 Description: Introduce lu_kmem_descr.
1791 Details : lu_kmem_descr and its companion interface allow to create
1792 and destroy a number of kmem caches at once.
1796 Description: Fix lu_object finalization race.
1797 Details : Fix a race between lu_object_find() finding an object and its
1798 concurrent finalization. This race is (most likely) not possible
1799 on the server, but might happen on the client.
1803 Description: Introduce lu_ref interface.
1804 Details : lu_ref is a debugging module allowing to track references to
1805 a given object. It is quite cpu expensive, and has to be
1806 explicitly enabled with --enable-lu_ref. See usage description
1811 Description: Factor lu_site procfs stats into a separate function.
1812 Details : Separate lu_site stats printing code into a separate function
1813 that can be reused on a client.
1817 Description: Constify instances of struct {lu,dt,md}_device_operations.
1818 Details : Constify instances of struct {lu,dt,md}_device_operations.
1822 Description: Introduce struct md_site and move meta-data specific parts of
1823 struct lu_site here.
1824 Details : Move md-specific fields out of struct lu_site into special struct
1825 md_site, so that lu_site can be used on a client.
1829 Description: Kill mdd_lov_destroy().
1830 Details : Remove unused mdd code.
1834 Description: Add st_block checking to multistat.c.
1835 Details : Add st_block checking to multistat.c.
1839 Description: Add lu_ref support to struct obd_device.
1840 Details : Add lu_ref tracking to obd_device.
1844 Description: Kill obd_set_fail_loc().
1845 Details : Remove unused code.
1849 Description: Add special type for ptlrpc_request interpret functions.
1850 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1851 there. Actual usage will be in CLIO.
1855 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1856 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1857 doesn't work with lockdep.
1861 Description: Add rwv.c test program.
1862 Details : New testing program exercising readv(2) and writev(2) (Qian).
1866 Description: Add sendfile.c test program.
1867 Details : New testing program exercising sendfile(2) (Jay).
1871 Description: Ratelimit a message that can be very frequent.
1872 Details : Ratelimit a memory allocation failure message that can
1877 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1878 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1882 Description: Liblustre build fixes.
1883 Details : Liblustre build fixes.
1887 Description: libcfs: add cfs_{need,cond}_resched() interface.
1888 Details : libcfs: add cfs_{need,cond}_resched() definition and
1889 implementations for Linux, NT, and liblustre.
1891 Severity : enhancement
1893 Description: More exported tunables for mballoc
1894 Details : Add support for tunable preallocation window and new tunables for
1895 large/small requests
1899 Description: Detect corruption of block bitmap and checking for preallocations
1900 Details : Checks validity of on-disk block bitmap. Also it does better
1901 checking of number of applied preallocations. When corruption is
1902 found, it turns filesystem readonly to prevent further corruptions.
1906 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1907 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1908 window to 0. If it does hit cache miss, and read-pattern is still
1909 stride-io mode, does not reset the stride window, but also does not
1910 increase the stride window length in this case.
1914 Frequency : only for big-endian servers
1915 Description: Check if system is big-endian while mounting fs with extents feature
1916 Details : Mounting a filesystem with extents feature will fail on big-endian
1917 systems since ext3-based ldiskfs is not supported on big-endian
1918 systems. This can be over-riden with "bigendian_extents" mount option.
1920 Severity : enhancement
1922 Description: The root squash functionality
1923 Details : A security feature, which is to prevent users from being able
1924 to mount lustre on their desktop, run as root, and delete
1925 all of the files in the filesystem. The goal is accomplished by
1926 remapping user id (UID) and group id (GID) of the root user to
1927 a UID and GID specified by the system administartor via Lustre
1928 configuration management server (MGS). The functionality also
1929 allows to specify sets of clients for which the remapping does
1934 Description: Excessive recovery window
1935 Details : With AT enabled, the recovery window can be excessively long (6000+
1936 seconds). To address this problem, we no longer use
1937 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1938 timeout no longer depends on the service time, it is set to
1939 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1940 time via pb_service_time.
1944 Description: Watchdog triggered on MDS failover
1945 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
1946 the OSTs know that the MDS "UUID" can be reused for the same export
1947 from a different NID, so we do not need to wait for the export to be
1951 Frequency : rare, only if using MMP with Linux RAID
1953 Description: MMP doesn't work with Linux RAID
1954 Details : While using HA for Lustre servers with Linux RAID, it is possible
1955 that MMP will not detect multiple mounts. To make this work we
1956 need to unplug the device queue in RAID when the MMP block is being
1957 written. Also while reading the MMP block, we should read it from
1958 disk and not the cached one.
1960 Severity : enhancement
1962 Description: open file using fid
1963 Details : A file can be opened using just its fid, like
1964 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
1966 --------------------------------------------------------------------------------
1968 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1970 * Support for kernels:
1971 2.6.5-7.283 (SLES 9),
1972 2.6.9-55.EL (RHEL 4),
1973 2.6.16.46-0.14 (SLES 10),
1974 2.6.18.8 vanilla (kernel.org)
1975 * Client support for unpatched kernels:
1976 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1977 2.6.16 - 2.6.22 vanilla (kernel.org)
1978 * Due to recently discovered recovery problems, we do not recommend
1979 using patchless RHEL 4 clients with this or any earlier release.
1980 * Recommended e2fsprogs version: 1.39.cfs8
1981 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1982 * Starting with this release, the ldiskfs backing filesystem required
1983 by Lustre is now in its own package, lustre-ldiskfs. This package
1984 should be installed. It is versioned separately from Lustre and
1985 may be released separately in future.
1989 Description: block reactivating mgc import until all deactivates complete
1990 Details : Fix race when failing back MDT/MGS to itself (testing)
1993 Frequency : at statup only
1995 Description: mds_lov_synchronize race leads to various problems
1996 Details : simultaneous MDT->OST connections at startup can cause the
1997 sync to abort, leaving the OSC in a bad state.
1999 Severity : enhancement
2001 Description: add optional extra BUILD_VERSION info
2002 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2003 to override the lustre version.
2006 Frequency : 2.6.18 servers only
2008 Description: ll_kern_mount() doesn't release the module reference
2009 Details : The ldiskfs module reference count never drops down to 0
2010 because ll_kern_mount() doesn't release the module reference.
2015 Description: server LBUG when using old ost_num_threads parameter
2016 Details : Accept the old ost_num_threads parameter but warn that it
2017 is deprecated, and fix an off-by-one error that caused an LBUG.
2022 Description: Transient SCSI error results in persistent IO issue
2023 Details : iobuf->dr_error is not reinitialized to 0 between two
2027 Frequency : sometimes when underlying device returns I/O errors
2029 Description: OSTs not going read-only during write failures
2030 Details : OSTs are not remounted read-only when the journal commit threads
2031 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2032 instead of the ext3 wrappers.
2036 Description: poor connect scaling with increasing client count
2037 Details : Don't run filter_grant_sanity_check for more than 100 exports
2038 to improve scaling for large numbers of clients.
2041 Frequency : SLES10 only
2043 Description: sanity-quota.sh quotacheck failed: rc = -22
2044 Details : Quotas cannot be enabled on SLES10.
2047 Frequency : liblustre clients only
2049 Description: getdirentries does not give error when run on compute nodes
2050 Details : getdirentries does not fail when the size specified as an argument
2051 is too small to contain at least one entry
2053 Severity : enhancement
2055 Description: Add LNET router traceability for debug purposes
2056 Details : If a checksum failure occurs with a router as part of the
2057 IO path, the NID of the last router that forwarded the bulk data
2058 is printed so it can be identified.
2063 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2064 Details : Due to a race condition, liblustre clients were occasionally
2065 evicted incorrectly.
2067 Severity : enhancement
2069 Description: lfs setstripe use optional parameters instead of postional
2072 Severity : enhancement
2074 Description: Nanosecond timestamp support for ldiskfs
2075 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2076 resolution timestamps. There is not yet support for this at
2077 the Lustre filesystem level.
2080 Frequency : during server recovery
2082 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2083 Details : request with rq_no_resend flag not awake l_wait_event if they get
2087 Frequency : nfs export on patchless client
2089 Description: connectathon hang when test nfs export over patchless client
2090 Details : Disconnected dentry cannot be found with lookup, so we do not need
2091 to unhash it or make it invalid
2094 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2095 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2096 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2101 Description: open req refcounting wrong on reconnect
2102 Details : If reconnect happened between getting open reply from server and
2103 call to mdc_set_replay_data in ll_file_open, we will schedule
2104 replay for unreferenced request that we are about to free.
2105 Subsequent close will crash in variety of ways.
2106 Check that request is still eligible for replay in
2107 mdc_set_replay_data().
2112 Description: disable writes to filesystem when reading health_check file
2113 Details : the default for reading the health_check proc file has changed
2114 to NOT do a journal transaction and write to disk, because this
2115 can cause reads of the /proc file to hang and block HA state
2116 checking on a healthy but otherwise heavily loaded system. It
2117 is possible to return to the previous behaviour during configure
2118 with --enable-health-write.
2120 Severity : enhancement
2122 Description: 64-bit inode version
2123 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2124 to the inode. This will be required for version-based recovery.
2129 Description: MDS fails to start if a duplicate client export is detected
2130 Details : in some rare cases it was possible for a client to connect to
2131 an MDS multiple times. Upon recovery the MDS would detect this
2132 and fail during startup. Handle this more gracefully.
2134 Severity : enhancement
2136 Description: Add -o localflock option to simulate old noflock
2138 Details : This will achieve local-only flock/fcntl locks
2144 Description: log_commit_thread vs filter_destroy race leads to crash
2145 Details : Take import reference before releasing llog record semaphore
2150 Description: Wrong request locking in request set processing
2151 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2152 imp_delayed_list, in this place should be used imp_lock.
2155 Frequency : when reconnection
2157 Description: Grant Leak when osc reconnect to OST
2158 Details : When osc reconnect ost, OST(filter) should check whether it
2159 should grant more space to client by comparing fed_grant and
2160 cl_avail_grant, and return the granted space to client instead
2161 of "new granted" space, because client will call osc_init_grant
2162 to update the client grant space info.
2165 Frequency : when client reconnect to OST
2167 Description: Grant Leak when osc do resend and replay bulk write
2168 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2169 bulk write request, because the grant info while be sync between
2170 OSC and OST when reconnect, and we should ignore the grant info
2171 these of resend/replay write req.
2176 Description: Grant space more than avaiable left space sometimes.
2177 Details : When then OST is about to be full, if two bulk writing from
2178 different clients came to OST. Accord the avaliable space of the
2179 OST, the first req should be permitted, and the second one
2180 should be denied by ENOSPC. But if the seconde arrived before
2181 the first one is commited. The OST might wrongly permit second
2182 writing, which will cause grant space > avaiable space.
2185 Frequency : when client is evicted
2187 Description: Grant might be wrongly erased when osc is evicted by OST
2188 Details : when the import is evicted by server, it will fork another
2189 thread ptlrpc_invalidate_import_thread to invalidate the
2190 import, where the grant will be set to 0. While the original
2191 thread will update the grant it got when connecting. So if
2192 the former happened latter, the grant will be wrongly errased
2193 because of this race.
2198 Description: Checking Stale with correct fid
2199 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2200 to check whether it is stale, because sometimes, we want the
2201 enqueue happened anyway, and op_data.fid2 will not be initialized.
2203 Severity : enhancement
2205 Description: update patchless client
2206 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2209 Frequency : only with 2.4 kernel
2211 Description: random memory corruption
2212 Details : size of struct ll_inode_info is to big for union inode.u and this
2213 can be cause of random memory corruption.
2218 Description: Memory leak in recovery
2219 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2220 It should also check obd_fail before fsfilt_start, otherwise if
2221 fsfilt_start return -EROFS,(failover mds during mds recovery).
2222 then the req will return with repmsg->transno = 0 and rc = EROFS.
2223 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2224 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2225 be freed no matter whether fsfilt_commit success or not.
2228 Frequency : only with huge count clients
2230 Description: Prevents from taking the superblock lock in llap_from_page for
2232 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2233 for llap_from_page call in ll_removepage prevents from taking the
2234 superblock lock for a soon died page.
2239 Description: Not check open intent error before release open handle
2240 Details : in some rare cases, the open intent error is not checked before
2241 release open handle, which may cause
2242 ASSERTION(open_req->rq_transno != 0), because it tries to release
2243 the failed open handle.
2248 Description: Set cat log bitmap only after create log success.
2249 Details : in some rare cases, the cat log bitmap is set too early. and it
2250 should be set only after create log success.
2254 Description: Accessing a block bevice can re-enable I/O when Lustre is
2255 tearing down a device.
2256 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2261 Description: service threads may hog cpus when there are a lot of requests
2262 Details : Insert cond_resched to give other threads a chance to use some CPU
2267 Description: the cat log was not initialized in recovery
2268 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2269 unlink log on mds will not be initialized until mds post
2270 recovery. And also in mds post recovery, the unlink log will
2271 initialization will be done asynchronausly, so there will be race
2272 between add unlink log and unlink log initialization.
2276 Description: brw_stats were being printed incorrectly
2277 Details : brw_stats were being printed as log2 but all of them were not
2278 recorded as log2. Also remove some code duplication arising from
2279 filter_tally_{read,write}.
2283 Frequency : rare, only in recovery.
2284 Description: ASSERTION(req->rq_type != LI_POISON) failed
2285 Details : imp_lock should be held while iterating over imp_sending_list for
2286 prevent destroy request after get timeout in ptlrpc_queue_wait.
2290 Description: replay-single.sh test 52 fails
2291 Details : A lock's skiplist need to be cleanup when it being unlinked
2292 from its resource list.
2296 Description: Short directio read returns full requested size rather than
2298 Details : Direct I/O operations should return actual amount of bytes
2299 transferred rather than requested size.
2301 Severity : enhancement
2303 Description: metadata RPC reduction (e.g. for rm performance)
2304 Details : decrease the amount of synchronous RPC between clients and servers
2305 by canceling conflicing lock before the operation on the client
2306 and packing thier handles into the main operation RPC to server.
2308 Severity : enhancement
2310 Description: Async OSC create to avoid the blocking unnecessarily.
2311 Details : If a OST has no remain object, system will block on the creating
2312 when need to create a new object on this OST. Now, ways use
2313 pre-created objects when available, instead of blocking on an
2314 empty osc while others are not empty. If we must block, we block
2315 for the shortest possible period of time.
2319 Description: improve handling recoverable errors
2320 Details : if request processig with error which can be recoverable on server
2321 request should be resend, otherwise page released from cache and
2324 Severity : enhancement
2326 Description: refine locking for avoid write wrong info into lov_objid file
2327 Details : fix possible races with add new target and write/update data in
2330 --------------------------------------------------------------------------------
2332 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2337 Frequency : on some architectures
2339 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2340 Details : Uninitialized flags sometimes cause configuration commands to
2344 Frequency : patchless clients only
2346 Description: missing __iget() symbol export
2347 Details : The __iget() symbol export is missing. To avoid the need for
2348 this on patchless clients the deathrow inode reaper is turned
2349 off, and we depend on the VM to clean up old inodes. This
2350 dependency was during via the fix for bug 12181.
2352 --------------------------------------------------------------------------------
2354 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2356 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2357 INTEROPERATE with older versions automatically. In many cases a
2358 special upgrade step is needed. Please read the
2359 user documentation before upgrading any part of a live system.
2360 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2361 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2362 * WARNING: Lustre configuration and startup changes are required with
2363 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2367 Severity : enhancement
2369 Description: MountConf
2370 Details : Lustre configuration is now managed via mkfs and mount
2371 commands instead of lmc and lconf. New obd types (MGS, MGC)
2372 are added for dynamic configuration management. See
2373 https://mail.clusterfs.com/wikis/lustre/MountConf for
2376 Severity : enhancement
2378 Description: dynamic OST addition
2379 Details : OSTs can now be added to a live filesystem
2381 Severity : enhancement
2383 Description: startup order invariance
2384 Details : MDTs and OSTs can be started in any order. Clients only
2385 require the MDT to complete startup.
2387 Severity : enhancement
2389 Description: parallel, asynchronous orphan cleanup
2390 Details : orphan cleanup is now performed in separate threads for each
2391 OST, allowing parallel non-blocking operation.
2393 Severity : enhancement
2395 Description: optimized stripe assignment
2396 Details : stripe assignments are now made based on ost space available,
2397 ost previous usage, and OSS previous usage, in order to try
2398 to optimize storage space and networking resources.
2400 Severity : enhancement
2402 Description: Permanently set tunables
2403 Details : All writable /proc/fs/lustre tunables can now be permanently
2404 set on a per-server basis, at mkfs time or on a live system.
2406 Severity : enhancement
2408 Description: Lustre message v2
2409 Details : Add lustre message format v2.
2411 Severity : enhancement
2413 Description: client OST exclusion list
2414 Details : Clients can be started with a list of OSTs that should be
2415 declared "inactive" for known non-responsive OSTs.
2419 Description: SPEC SFS validation failure on NFS v2 over lustre.
2420 Details : Changes the blocksize for regular files to be 2x RPC size,
2421 and not depend on stripe size.
2423 Severity : enhancement
2425 Description: Multiple MD RPCs in flight.
2426 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2427 To avoid overly-overloading MDS, introduce a limit on number of
2428 MDS RPCs in flight for a single client and add /proc controls
2429 to adjust this limit.
2431 Severity : enhancement
2433 Description: client read/write statistics
2434 Details : Add client read/write call usage stats for performance
2435 analysis of user processes.
2436 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2437 file access. extents_stats shows chunk size distribution.
2438 extents_stats_per_process show chunk size distribution per
2441 Severity : enhancement
2443 Description: mds statistics
2444 Details : Add detailed mds operations statistics in
2445 /proc/fs/lustre/mds/*/stats.
2449 Description: Failure of copying files with lustre special EAs.
2450 Details : Client side always return success for setxattr call for lustre
2451 special xattr (currently only "trusted.lov").
2456 Description: Refcount LNET uuids
2457 Details : The global LNET uuid list grew linearly with every startup;
2458 refcount repeated list entries instead of always adding to
2461 Severity : enhancement
2463 Description: Dynamic service threads
2464 Details : Within a small range, start extra service threads
2465 automatically when the request queue builds up.
2468 Frequency : mixed-endian client/server environments
2470 Description: mixed-endian crashes
2471 Details : The new msg_v2 system had some failures in mixed-endian
2474 Severity : enhancement
2476 Description: Easy OST removal
2477 Details : OSTs can be permanently deactivated with e.g. 'lctl
2478 conf_param lustre-OST0001.osc.active=0'
2480 Severity : enhancement
2482 Description: MGS proc entries
2483 Details : Added basic proc entries for the MGS showing what filesystems
2486 Severity : enhancement
2488 Description: provide MGS failover
2489 Details : Added config lock reacquisition after MGS server failover.
2491 Severity : enhancement
2493 Description: add Linux 2.4 support
2494 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2498 Description: a large application tries to do I/O to the same resource and dies
2499 in the middle of it.
2500 Details : Check the req->rq_arrival time after the call to
2501 ost_brw_lock_get(), but before we do anything about
2502 processing it & sending the BULK transfer request. This
2503 should help move old stale pending locks off the queue as
2504 quickly as obd_timeout.
2507 Frequency : when an incorrect nid is specified during startup
2509 Description: ptlrpc connect to non-existant node causes kernel crash
2510 Details : LNET can't be re-entered from an event callback, which
2511 happened when we expire a message after the export has been
2512 cleaned up. Instead, hand the zombie cleanup off to another
2515 Severity : enhancement
2517 Description: plain/inodebits lock performance improvement
2518 Details : Grouping plain/inodebits in granted list by their request modes
2519 and bits policy, thus improving the performance of search through
2523 Frequency : only if OST filesystem is corrupted
2525 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2526 Details : for a short time RPCs with bulk IO are in the replay list,
2527 but replay of bulk IOs is unimplemented. If the OST filesystem
2528 is corrupted due to disk cache incoherency and then replay is
2529 started it is possible to trip an assertion. Avoid putting
2530 committed RPCs into the replay list at all to avoid this issue.
2533 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2535 Description: System hang on startup
2536 Details : This bug allowed the liblustre (e.g. catamount) client to
2537 return to the app before handling all startup RPCs. This
2538 could leave the node unresponsive to lustre network traffic
2539 and manifested as a server ptllnd timeout.
2541 Severity : enhancement
2543 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2544 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2545 boolean to control whether to print peer debug info when a
2546 client's RPC times out.
2549 Frequency : only for kernels with patches from Lustre below 1.4.3
2551 Description: Remove old rdonly API
2552 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2555 Frequency : only for devices with external journals
2557 Description: Set external device read-only also
2558 Details : During a commanded failover stop, we set the disk device
2559 read-only while the server shuts down. We now also set any
2560 external journal device read-only at the same time.
2563 Frequency : when upgrading from 1.4 while trying to change parameters
2565 Description: The wrong (new) MDC name was used when setting parameters for
2566 upgraded MDT's. Also allows changing of OSC (and MDC)
2567 parameters if --writeconf is specified at tunefs upgrade time.
2570 Frequency : when setting specific ost indicies
2572 Description: QOS code breaks on skipped indicies
2573 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2574 created with --index need not be sequential.
2576 Severity : enhancement
2578 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2579 Details : The uninit_groups feature works in conjunction with the kernel
2580 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2581 up the pass1 processing of e2fsck. This is a read-only feature
2582 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2583 mount filesystems that have had this feature enabled.
2585 Severity : enhancement
2587 Description: Improve multi-block allocation algorithm to avoid fragmentation
2588 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2589 allocation locality and avoid filesystem fragmentation.
2591 ------------------------------------------------------------------------------
2593 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2595 * Support for kernels:
2596 2.4.21-47.0.1.EL (RHEL 3)
2597 2.6.5-7.283 (SLES 9)
2598 2.6.9-42.0.10.EL (RHEL 4)
2599 2.6.12.6 vanilla (kernel.org)
2600 2.6.16.27-0.9 (SLES 10)
2601 * Recommended e2fsprogs version: 1.39.cfs5
2603 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2607 Frequency : occasional, depends on client load and configuration
2608 Bugzilla : 12181, 12203
2609 Description: data loss for recently-modified files
2611 Details : In some cases it is possible that recently written or created
2612 files may not be written to disk in a timely manner (this should
2613 normally be within 30s unless client IO load is very high).
2614 The problem appears as zero-length files or files that are a
2615 multiple of 1MB in size after a client crash or client eviction
2616 that are missing data at the end of the file.
2618 This problem is more likely to be hit on clients where files are
2619 repeatedly created and unlinked in the same directory, clients
2620 have a large amount of RAM, have many CPUs, the filesystem has
2621 many OSTs, the clients are rebooted frequently, and/or the files
2622 are not accessed by other nodes after being written.
2624 The presence of the problem can be detected by looking at
2625 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2626 smaller than the second (nr_unused) then dirty files will not
2627 be flushed automatically to disk. "sync; sleep 10" should be
2628 run several times on the node before unmounting it to update
2629 Lustre (this is also safe to run on nodes without this problem).
2631 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2632 that can cause this same problem, so customers using that kernel
2633 also need to update the kernel in addition to Lustre. In order
2634 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2636 It is normal that files written just before a client crash (less
2637 than 30s) may not yet have been flushed to disk, even for local
2641 Frequency : frequent on thin XT3 nodes
2643 Description: UUID collision on thin XT3 Linux nodes
2644 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2645 randomly, since we relied on an insufficiently-seeded PRNG.
2650 Description: OSS hangs after "All ost request buffers busy"
2651 Details : A deadlock between quota and journal operations caused OSS
2652 hangs after printing "All ost request buffers busy."
2655 Frequency : always on liblustre builds
2657 Description: Cleanup compiler warnings on liblustre
2660 Frequency : always on liblustre builds on XT3
2662 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2666 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2667 > 32000 subdirectories
2668 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2669 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2670 subdirectory count crosses 32000. This will aid e2fsck to
2671 correctly handle more than 32000 subdirectories.
2674 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2676 Description: System hang on startup
2677 Details : This bug allowed the liblustre (e.g. catamount) client to
2678 return to the app before handling all startup RPCs. This
2679 could leave the node unresponsive to lustre network traffic
2680 and manifested as a server ptllnd timeout.
2682 Severity : enhancement
2684 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2685 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2686 boolean to control whether to print peer debug info when a
2687 client's RPC times out.
2692 Description: make O_SYNC working on 2.6 kernels
2693 Details : 2.6 kernels use different method for mark pages for write,
2694 so need add a code to lustre for O_SYNC work.
2699 Description: Failure to close file and release space on NFS
2700 Details : Put inode details into lock acquired in ll_intent_file_open.
2701 Use mdc_intent_lock in ll_intent_open to properly
2702 detect all kind of errors unhandled by mdc_enqueue.
2707 Description: proc file read during shutdown sometimes raced obd removal,
2709 Details : Add lock to prevent obd access after proc file removal.
2712 Frequency : Only for files larger than 4GB on 32-bit clients.
2714 Description: improperly doing page alignment of locks
2715 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2716 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2721 Description: RPCs being resent when they shouldn't be
2722 Details : Some RPCs that should not be resent are being resent. This
2723 can cause inconsistencies in the RPC state machine. Do not
2724 resend such requests.
2727 Frequency : rare, only with NFS export
2729 Description: Crash on NFS re-export node
2730 Details : under very unusual load conditions an assertion is hit in
2731 ll_intent_file_open()
2734 Frequency : only if OST filesystem is corrupted
2736 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2737 Details : for a short time RPCs with bulk IO are in the replay list,
2738 but replay of bulk IOs is unimplemented. If the OST filesystem
2739 is corrupted due to disk cache incoherency and then replay is
2740 started it is possible to trip an assertion. Avoid putting
2741 committed RPCs into the replay list at all to avoid this issue.
2746 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2747 Details : Large single O_DIRECT read and write calls can fail to allocate
2748 a sufficiently large buffer to process the request. In case of
2749 allocation failure the allocation is retried with a smaller
2750 buffer and broken into smaller requests.
2752 Severity : enhancement
2754 Description: Add -o localflock option to simulate old noflock behaviour.
2755 Details : This will achieve local-only flock/fcntl locks coherentness.
2760 Description: versioning check is incomplete
2761 Details : Checking the version difference of client vs. server, report
2762 error if the gap is too big.
2767 Description: add support PG_writeback bit
2768 Details : add support for PG_writeback bit for Lustre, for more carefull
2769 work with page cache in 2.6 kernel. This also fix some deadlocks
2770 and remove hack for work O_SYNC with 2.6 kernel.
2772 Severity : enhancement
2774 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2775 Details : The uninit_groups feature works in conjunction with the kernel
2776 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2777 up the pass1 processing of e2fsck. This is a read-only feature
2778 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2779 mount filesystems that have had this feature enabled.
2781 Severity : enhancement
2783 Description: Improve multi-block allocation algorithm to avoid fragmentation
2784 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2785 allocation locality and avoid filesystem fragmentation.
2787 ------------------------------------------------------------------------------
2789 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2791 * Support for kernels:
2792 2.6.9-42.0.3.EL (RHEL 4)
2793 2.6.5-7.276 (SLES 9)
2794 2.4.21-47.0.1.EL (RHEL 3)
2795 2.6.12.6 vanilla (kernel.org)
2796 2.6.16.21-0.8 (SLES10)
2797 * Recommended e2fsprogs version: 1.39.cfs2-0
2799 * The backwards-compatible /proc/sys/portals symlink has been removed
2800 in this release. Before upgrading, please ensure that you change
2801 any configuration scripts or /etc/sysctl.conf files that access
2802 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2803 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2804 in advance of the upgrade on any system running Lustre 1.4.6 or
2805 newer, since /proc/sys/lnet was added in that version.
2806 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2810 Frequency : only when quota is used
2812 Description: avoid scanning export list for quota master
2813 Details : Change the algorithms to avoid scanning export list in order
2814 to improve the efficiency.
2817 Frequency : MDS failover only, very rarely
2819 Description: "went back in time" messages on mds failover
2820 Details : The greatest transno may be lost when the current operation
2821 finishes with an error (transno==0) and the client's last_rcvd
2822 record is over-written. Save the greatest transno in the
2823 mds_last_transno for this case.
2826 Frequency : always for specific kernels and striping counts
2828 Description: client may get "Matching packet too big" without ACL support
2829 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2830 when trying to access files in certain configurations. The
2831 clients should in fact be denied when mounting because they do
2832 not understand ACLs.
2835 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2837 Description: many clients connecting with IO in progress causes connect timeouts
2838 Details : Avoid synchronous journal commits to avoid delays caused by many
2839 clients connecting/disconnecting when bulk IO is in progress.
2840 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2841 OST_IO_PORTAL to avoid delays behind potentially many pending
2845 Frequency : occasionally with multiple writers to a single file
2847 Description: shared writes to file may result in wrong size reported by stat()
2848 Details : Allow growing of kms when extent lock is cancelled
2851 Frequency : always with random mmap IO to multi-striped file
2853 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2854 Details : Only if the hole is at the end of OST object so that kms is too
2855 small. Fix is to increase kms accordingly in ll_nopage.
2858 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2860 Description: writes to a missing object would leak memory on the OST
2861 Details : If there is an inconsistency between the MDS and OST filesystems,
2862 such that the MDS references an object that doesn't exist, writes
2863 to that object will leak memory due to incorrect cleanup in the
2864 error handling path, eventually running out of memory on the OST.
2869 Description: Creating too long symlink causes lustre errors
2870 Details : Check symlink and name lengths before sending requests to MDS.
2873 Frequency : only if flock is enabled (not on by default)
2875 Description: posix locks not released on fd closure on 2.6.9+
2876 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2877 kernels, this caused such locks not to be released on fd close and
2878 then assertions on fs unmount about still used locks.
2881 Frequency : MDS failover only, very rarely
2883 Description: clients may get ASSERTION(granted_lock != NULL)
2884 Details : When request was taking a long time, and a client was resending
2885 a getattr by name lock request. The were multiple lock requests
2886 with the same client lock handle and
2887 mds_getattr_name->fixup_handle_for_resent_request found one of the
2888 lock handles but later failed with ASSERTION(granted_lock != NULL).
2893 Description: handle->h_buffer_credits > 0, assertion failure
2894 Details : h_buffer_credits is zero after truncate, causing assertion
2895 failure. This patch extends the transaction or creates a new
2899 Frequency : NFS re-export or patchless client
2900 Bugzilla : 11179, 10796
2901 Description: Crash on NFS re-export node (__d_move)
2902 Details : We do not want to hash the dentry if we don't have a lock.
2903 But if this dentry is later used in d_move, we'd hit uninitialised
2904 list head d_hash, so we just do this to init d_hash field but
2905 leave dentry unhashed.
2908 Frequency : NFS re-export or patchless client
2910 Description: NFS exports has problem with symbolic link
2911 Details : lustre client didn't properly install dentry when re-exported
2912 to NFS or running patchless client.
2915 Frequency : NFS re-export or patchless client
2917 Description: Various nfs/patchless fixes.
2918 Details : fixes reuse disconected alias for lookup process - this fixes
2919 warning "find_exported_dentry: npd != pd",
2920 fix permission error with open files at nfs.
2921 fix apply umask when do revalidate.
2924 Frequency : occasional
2926 Description: Crash on NFS re-export node
2927 Details : calling clear_page() on the wrong pointer triggered oops in
2928 generic_mapping_read().
2931 Frequency : rarely, using O_DIRECT IO
2933 Description: unaligned directio crashes client with LASSERT
2934 Details : check for unaligned buffers before trying any requests.
2937 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2939 Description: stale data returned from RAID cache
2940 Details : If only a small amount of IO is done to the RAID device before
2941 reading it again it is possible to get stale data from the RAID
2942 cache instead of reading it from disk.
2945 Frequency : always for sles10 kernel
2947 Description: sles10 support
2948 Details : ll_follow_link: compile fixes and using of nd_set_link
2949 under newer kernels.
2952 Frequency : depends on arch, kernel and compiler version, always on sles10
2955 Description: recursive or deep enough symlinks cause stack overflow
2956 Details : getting rid of large stack-allocated variable in
2960 Frequency : depends on hardware
2962 Description: lustre write performance loss in the SLES10 kernel
2963 Details : the performance loss is caused by using of write barriers in the
2964 ext3 code. The SLES10 kernel turns barrier support on by
2965 default. The fix is to undo that change for ldiskfs.
2967 ------------------------------------------------------------------------------
2969 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2971 * Support for kernels:
2972 2.6.9-42.0.3EL (RHEL 4)
2973 2.6.5-7.276 (SLES 9)
2974 2.4.21-47.0.1.EL (RHEL 3)
2975 2.6.12.6 vanilla (kernel.org)
2979 Frequency : quota enabled and large files being deleted
2981 Description: releasing more than 4GB of quota at once hangs OST
2982 Details : If a user deletes more than 4GB of files on a single OST it
2983 will cause the OST to spin in an infinite loop. Release
2984 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2989 Description: statfs data retrieved from /proc may be stale or zero
2990 Details : When reading per-device statfs data from /proc, in the
2991 {kbytes,files}_{total,free,avail} files, it may appear
2992 as zero or be out of date.
2995 Frequency : systems with MD RAID1 external journal devices
2997 Description: lconf's call to blkid is confused by RAID1 journal devices
2998 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2999 of returning all block devices that match the journal UUID.
3002 Frequency : always, for aggregate stripe size over 4GB
3004 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3005 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3006 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3009 Frequency : always if "lfs find" used on a local file/directory
3011 Description: "lfs find" segfaults if used on a local file/directory
3012 Details : The case where a directory component was not specified wasn't
3013 handled correctly. Handle this properly.
3016 Frequency : always on ppc64
3018 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3019 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3020 perform endianness conversion.
3023 Frequency : rarely (truncate to non-zero file size after write under load)
3024 Bugzilla : 10730, 10687
3025 Description: Files padded with zeros to next 4K multiple
3026 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3027 it is possible that files that are truncated to a non-zero size
3028 immediately after being written are filled with zero bytes beyond
3029 the truncated size. No file data is lost.
3031 Severity : enhancement
3033 Description: Allow recovery/failover for liblustre clients.
3034 Details : liblustre clients were unaware of failover configurations until
3037 Severity : enhancement
3039 Description: user file locks should fail when not mounting with flock option
3040 Details : Set up an error-returning stub in ll_file_operations.lock field
3041 to prevent incorrect behaviour when client is mounted without
3042 flock option. Also, set up properly f_op->flock field for
3046 Frequency : always on ia64
3048 Description: "lfs df" loops on printing out MDS statfs information
3049 Details : The obd_ioctl_data was not initialized and in some systems
3050 this caused a failure during the ioctl that did not return
3051 an error. Initialize the struct and return an error on failure.
3054 Frequency : SLES 9 only
3056 Description: Error of copying files with lustre special EAs as root
3057 Details : Client side always return success for setxattr call for lustre
3058 special xattr (currently only "trusted.lov").
3061 Frequency : rarely on clusters with both ia64+i386 clients
3063 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3064 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3065 clients are doing concurrent writes to the same file it is
3066 possible that the ia64 clients may LASSERT because the OST
3067 extent locks are not PAGE_SIZE aligned. Ensure that grown
3068 locks are always aligned on the request boundary.
3071 Frequency : specific use, occasional
3073 Description: Overwriting in use executable truncates on-disk binary image
3074 Details : If one node attempts to overwrite an executable in use by
3075 another node, we now correctly return ETXTBSY instead of
3076 truncating the file.
3078 Severity : enhancement
3080 Description: Async OSC create to avoid the blocking unnecessarily.
3081 Details : If a OST has no remain object, system will block on the creating
3082 when need to create a new object on this OST. Now, ways use
3083 pre-created objects when available, instead of blocking on an
3084 empty osc while others are not empty. If we must block, we block
3085 for the shortest possible period of time.
3090 Description: chmod on Lustre root is propagated to other clients
3091 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3092 invalid by the follow_mount time.
3095 Frequency : liblustre clients only
3097 Description: Race in 'instant cancel' lock handling could lead to such locks
3098 never to be granted in case of SMP MDS
3099 Details : Do not destroy not yet granted but cbpending locks in
3103 Frequency : replay/resend of open
3105 Description: non null lock assetion failure in mds_intent_policy
3106 Details : Trying to replay/resend lockless open requests resulted in
3107 mds_open() returning 0 with no lock. Now it sets a flag if
3108 a lock is going to be returned.
3110 Severity : enhancement
3112 Description: Checksum enhancements
3113 Details : New checksum enhancements allow for resending RPCs that failed
3116 Severity : enhancement
3118 Description: Tunables on number of dirty pages in cacche
3119 Details : Allow to set limit on number of dirty pages cached.
3124 Description: client crash on unmount - lock still has references
3125 Details : In some error handling cases it was possible to leak a lock
3126 reference on a client while accessing a file. This was not
3127 harmful to the client during operation, but would cause the
3128 client to crash when the filesystem is unmounted.
3131 Frequency : specific case, rare
3133 Description: ETXTBSY on mds though file not in use by client
3134 Details : ETXTBSY is no longer incorrectly returned when attempting to
3135 chmod or chown a directory that the user previously tried to
3136 execute or a currently-executing binary.
3139 Frequency : extremely rare except on liblustre-based clients
3141 Description: Lustre space not freed when files are deleted
3142 Details : Clean up open-unlinked files after client eviction. Previously
3143 the unlink was skipped and the files remained as orphans.
3148 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3149 Details : In some cases it was possible to send a blocking callback to
3150 a client doing a glimpse, even though that client didn't get
3151 a lock granted. When the glimpse lock is cancelled on the OST
3152 the freed lock is left on the waiting list and corrupted the list.
3155 Frequency : all core dumps
3157 Description: Broke core dumps to lustre
3158 Details : Negative dentry may be unhashed if parent does not have UPDATE
3159 lock, but some callers, e.g. do_coredump, expect dentry to be
3160 hashed after successful create, hash it in ll_create_it.
3162 ------------------------------------------------------------------------------
3164 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3166 * Support for kernels:
3167 2.6.9-42.0.2.EL (RHEL 4)
3168 2.6.5-7.276 (SLES 9)
3169 2.4.21-40.EL (RHEL 3)
3170 2.6.12.6 vanilla (kernel.org)
3174 Frequency : always on RHEL 3
3176 Description: Number of open files grows over time
3177 Details : The number of open files grows over time, whether or not
3178 Lustre is started. This was due to a filp leak introduced
3179 by one of our kernel patches.
3181 ------------------------------------------------------------------------------
3183 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3185 * Support for kernels:
3186 2.6.9-42.EL (RHEL 4)
3187 2.6.5-7.276 (SLES 9)
3188 2.4.21-40.EL (RHEL 3)
3189 2.6.12.6 vanilla (kernel.org)
3194 Bugzilla : 5719, 9635, 9792, 9684
3195 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3196 Details : If a server is under heavy load and cannot reply to new
3197 connection requests before the client resends the (re)connect,
3198 the connection handling code can behave badly if two service
3199 threads are concurrently handing separate (re)connections from
3200 the same client. Add better locking to the connection handling
3201 code, and ensure that only a single connection will be processed
3202 for a given client UUID, even if the lock is dropped.
3204 Severity : enhancement
3206 Description: add TCP zero-copy support to kernel
3207 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3208 sends if the hardware supports scatter-gather and checksumming.
3209 This allows socklnd to do client-write and server-read more
3210 efficiently and reduce CPU utilization from skbuf copying.
3213 Frequency : only if NFS exporting from client
3215 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3216 Details : If setting attributes on a file created under NFS that had
3217 never been opened it would be possible to oops the client
3218 if the file had no objects.
3221 Frequency : always for liblustre
3223 Description: liblustre client does MDS+OSTs setattr RPC for each write
3224 Details : When doing a write from a liblustre client, the client
3225 incorrectly issued an RPC to the MDS and each OST the file was
3226 striped over in order to update the timestamps. When writing
3227 with small chunks and many clients this could overwhelm the MDS
3228 with RPCs. In all cases it would slow down the write because
3229 these RPCs are unnecessary.
3231 Severity : enhancement
3233 Description: allow number of MDS service threads to be changed at module load
3234 Details : It is now possible to change the number of MDS service threads
3235 running. Adding "options mds mds_num_threads={N}" to the MDS's
3236 /etc/modprobe.conf will set the number of threads for the next
3237 time Lustre is restarted (assuming the "mds" module is also
3238 reloaded at that time). The default number of threads will
3239 stay the same, 32 for most systems.
3244 Description: OST crash if filesystem is unformatted or corrupt
3245 Details : If an OST is started on a device that has never been formatted
3246 or if the filesystem is corrupt and cannot even mount then the
3247 error handling cleanup routines would dereference a NULL pointer.
3252 Description: NULL pointer deref in llap_from_page.
3253 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3254 incorrect) mapping if the page was truncated/reclaimed while it was
3255 searched for. Check for this condition and skip such pages when
3256 doing readahead. Introduce extra check to llap_from_page() to
3257 verify page->mapping->host is non-NULL (so page is not anonymous).
3260 Frequency : Sometimes when using sys_sendfile
3262 Description: "page not covered by a lock" warnings from ll_readpage
3263 Details : sendfile called ll_readpage without right page locks present.
3264 Now we introduced ll_file_sendfile that does necessary locking
3265 around call to generic_file_sendfile() much like we do in
3269 Frequency : with certain MDS communication failures at client mount time
3271 Description: NULL pointer deref after failed client mount
3272 Details : a client connection request may delayed by the network layer
3273 and not be sent until after the PTLRPC layer has timed out the
3274 request. If the client fails the mount immediately it will try
3275 to clean up before the network times out the request. Add a
3276 reference from the request import to the obd device and delay
3277 the cleanup until the network drops the request.
3280 Frequency : occasionally during client (re)connect
3282 Description: assertion failure during client (re)connect
3283 Details : processing a client connection request may be delayed by the
3284 client or server longer than the client connect timeout. This
3285 causes the client to resend the connection request. If the
3286 original connection request is replied in this interval, the
3287 client may trip an assertion failure in ptlrpc_connect_interpret()
3288 which thought it would be the only running connect process.
3291 Frequency : only with obd_echo servers and clients that are rebooted
3293 Description: kernel BUG accessing uninitialized data structure
3294 Details : When running an obd_echo server it did not start the ping_evictor
3295 thread, and when a client was evicted an uninitialized data
3296 structure was accessed. Start the ping_evictor in the RPC
3297 service startup instead of the OBD startup.
3299 Severity : enhancement
3300 Bugzilla : 10193 (patchless)
3301 Description: Remove dependency on various unexported kernel interfaces.
3302 Details : No longer need reparent_to_init, exit_mm, exit_files,
3303 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3306 Frequency : rare (only users of deprecated and unsupported LDAP config)
3308 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3309 Details : LDAP apparently contains 'lustreName' attributes instead of
3310 'name'. A simple remapping of the name is sufficient.
3313 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3315 Description: waiting_locks_callback trips kernel BUG if client is evicted
3316 Details : Running with the dump_on_timeout debug flag turned on makes
3317 it possible that the waiting_locks_callback() can try to dump
3318 the Lustre kernel debug logs from an interrupt handler. Defer
3319 this log dumping to the expired_lock_main() thread.
3321 Severity : enhancement
3323 Description: Support NFS exporting on 2.6 kernels.
3324 Details : Implement non-rawops metadata methods for NFS server to use without
3325 changing NFS server code.
3328 Frequency : very rare (synthetic metadata workload only)
3330 Description: two racing renames might cause an MDS thread to deadlock
3331 Details : Running the "racer" program may cause one MDS thread to rename
3332 a file from being the source of a rename to being the target of
3333 a rename at exactly the same time that another thread is doing
3334 so, and the second thread has already enqueued these locks after
3335 doing a lookup of the target and is trying to relock them in
3336 order. Ensure that we don't try to re-lock the same resource.
3339 Frequency : only very large systems with liblustre clients
3341 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3342 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3343 all OSTs in parallel. This allows the eviction of stale liblustre
3344 clients to proceed much faster than if they were done in series,
3345 and also offers similar improvements for other set_info RPCs.
3350 Description: excessive CPU usage during initial read phase on client
3351 Details : During the initial read phase on a client, it would agressively
3352 retry readahead on the file, consuming too much CPU and impacting
3353 performance (since 1.4.5.8). Improve the readahead algorithm
3354 to avoid this, and also improve some other common cases (read
3355 of small files in particular, where "small" is files smaller than
3356 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3361 Description: MDS crash when receiving packet with unknown intent.
3362 Details : Do not LBUG in unknown intent case, just return -EFAULT
3364 Severity : enhancement
3365 Bugzilla : 9293, 9385
3366 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3367 Details : Do not serialize getattr (non-intent version) and statfs.
3370 Frequency : occasional, when OST network is overloaded/intermittent
3372 Description: client evicted by OST after bulk IO timeout
3373 Details : If a client sends a bulk IO request (read or write) the OST
3374 may evict the client if it is unresposive to its data GET/PUT
3375 request. This is incorrect if the network is overloaded (takes
3376 too long to transfer the RPC data) or dropped the OST GET/PUT
3377 request. There is no need to evict the client at all, since
3378 the pinger and/or lock callbacks will handle this, and the
3379 client can restart the bulk request.
3382 Frequency : Always when mmapping file with no objects
3384 Description: client crashes when mmapping file with no objects
3385 Details : Check that we actually have objects in a file before doing any
3386 operations on objects in ll_vm_open, ll_vm_close and
3392 Description: Request leak when working with deleted CWD
3393 Details : Introduce advanced request refcount tracking for requests
3394 referenced from lustre intent.
3396 Severity : Enhancement
3398 Description: Cache open file handles on client.
3399 Details : MDS now will return special lock along with openhandle, if
3400 requested and client is allowed to hold openhandle, even if unused,
3401 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3402 closing files for every read/write openration.
3404 Severity : Enhancement
3406 Description: Cache open negative dentries on client when possible.
3407 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3408 negative dentries on lock revocation.
3413 Description: Remounting a client read-only wasn't possible with a zconf mount
3414 Details : It wasn't possible to remount a client read-only with llmount.
3416 Severity : enhancement
3417 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3418 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3419 MPICH is included to add Lustre-specific ADIO interfaces.
3420 This is based closely on the UFS ADIO layer and only differs
3421 in file creation, in order to allow the OST striping to be set.
3422 This is user-contributed code and not supported by CFS.
3427 Description: extended inode attributes (immutable, append-only) work improperly
3428 when 2.4 and 2.6 kernels are used on client/server or vice versa
3429 Details : Introduce kernel-independent values for these flags.
3431 Severity : enhancement
3434 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3435 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3436 granularity. Now, Fractional MB granularity is be supported,
3437 this is very useful for low memory system.
3439 Severity : enhancement
3441 Description: Getattr by fid
3442 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3443 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3444 have locks and avoids getattr rpc requests for it on every stat.
3447 Frequency : Always, for filesystems larger than 2TB
3449 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3450 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3451 filesystems larger than 2TB to be mounted with mballoc enabled.
3454 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3456 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3457 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3458 filesystems larger than 2TB is mounted then the kernel will
3459 silently corrupt the start of the filesystem. CONFIG_LBD is
3460 enabled for all CFS-supported kernels, but the possibility of
3461 this happening with a modified kernel config exists.
3463 Severity : enhancement
3465 Description: add client O_DIRECT support for 2.6 kernels
3466 Details : It is now possible to do O_DIRECT reads and writes to files
3467 in the Lustre client mountpoint on 2.6 kernel clients.
3469 Severity : enhancement
3471 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3472 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3473 parallel, not waiting for response from every OST before sending
3474 a rpc to the next OST.
3479 Description: setattr vs write race when updating file timestamps
3480 Details : Client processes that update a file timestamp into the past
3481 right after writing to the file (e.g. tar) it is possible that
3482 the updated file modification time can be reset to the current
3483 time due to a race between processing the setattr and write RPC.
3485 Severity : enhancement
3487 Description: Bring 'lfs find' closer in line with regular Linux find.
3488 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3489 -print0 options and obtains all the needed info through the lustre
3492 Severity : enhancement
3494 Description: support up to 1024 configured devices on one node
3495 Details : change obd_dev array from statically allocated to dynamically
3496 allocated structs as they are first used to reduce memory usage
3501 Description: Flush dirty partially truncated pages during truncate
3502 Details : Immediatelly flush partially truncated pages in filter_setattr,
3503 this way we completely avoid having any pages in page cache on OST
3504 and can retire ugly workarounds during writes to flush such pages.
3509 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3510 Details : protect inode from truncation within vfs_unlink() context
3511 just take a reference before calling vfs_unlink() and release it
3512 when parent's i_sem is free.
3517 Description: last_id value checked outside lock on OST caused LASSERT failure
3518 Details : If there were multiple MDS->OST object precreate requests in
3519 flight, it was possible that the OST's last object id was checked
3520 outside a lock and incorrectly tripped an assertion. Move checks
3521 inside locks, and discard old precreate requests.
3524 Frequency : always, if extents are used on OSTs
3526 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3527 Details : OSTs using the extents format would not zero the high 16 bits of
3528 the index physical block number. This is not a problem for any
3529 OST filesystems smaller than 16TB, and no kernels support ext3
3530 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3531 new/modified files) and can be fixed for existing filesystems
3532 with e2fsprogs-1.39-cfs1.
3537 Description: import connection selection may be incorrect if timer wraps
3538 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3539 import connections to be ignored if the 32-bit jiffies counter
3540 wraps. Use a 64-bit jiffies counter.
3543 Frequency : very large clusters immediately after boot
3545 Description: LNET request buffers exhausted under heavy short-term load
3546 Details : If a large number of client requests are generated on a service
3547 that has previously never seen so many requests it is possible
3548 that the request buffer growth cannot keep up with the spike in
3549 demand. Instead of dropping incoming requests, they are held in
3550 the LND until the RPC service can accept more requests.
3553 Frequency : Sometimes during replay
3555 Description: Assertion failure in ll_local_open after replay.
3556 Details : If replay happened on an open request reply before we were able
3557 to set replay handler, reply will become not swabbed tripping the
3558 assertion in ll_local_open. Now we set the handler right after
3559 recognising of open request
3562 Frequency : very rare
3564 Description: kernel reports "badness in vsnprintf"
3565 Details : Reading from the "recovery_status" /proc file in small chunks
3566 may cause a negative length in lprocfs_obd_rd_recovery_status()
3567 call to vsnprintf() (which is otherwise harmless). Exit early
3568 if there is no more space in the output buffer.
3570 Severity : enhancement
3572 Description: clear OBD RPC statistics by writing to them
3573 Details : It is now possible to clear the OBD RPC statistics by writing
3574 to the "stats" file.
3579 Description: Client mtime is not the same on different clients after utimes
3580 Details : In some cases, the client was using the utimes() syscall on
3581 a file cached on another node. The clients now validate the
3582 ctime from the MDS + OSTs to determine which one is right.
3587 Description: Inability to activate failout mode
3588 Details : lconf script incorrectly assumed that in pythong string's numeric
3589 value is used in comparisons.
3592 Frequency : always with multiple stripes per file
3594 Description: Inefficient object allocation for mutli-stripe files
3595 Details : When selecting which OSTs to stripe files over, for files with
3596 a stripe count that divides evenly into the number of OSTs,
3597 the MDS is always picking the same starting OST for each file.
3598 Return the OST selection heuristic to the original design.
3603 Description: mount failures may take full timeout to return an error
3604 Details : Under some heavy load conditions it is possible that a
3605 failed mount can wait for the full obd_timeout interval,
3606 possibly several minutes, before reporting an error.
3607 Instead return an error as soon as the status is known.
3609 Frequency : quota enabled and large files being deleted
3611 Description: releasing more than 4GB of quota at once hangs OST
3612 Details : If a user deletes more than 4GB of files on a single OST it
3613 will cause the OST to spin in an infinite loop. Release
3614 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3619 Description: statfs data retrieved from /proc may be stale or zero
3620 Details : When reading per-device statfs data from /proc, in the
3621 {kbytes,files}_{total,free,avail} files, it may appear
3622 as zero or be out of date.
3625 Frequency : systems with MD RAID1 external journal devices
3627 Description: lconf's call to blkid is confused by RAID1 journal devices
3628 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3629 of returning all block devices that match the journal UUID.
3632 Frequency : always, for aggregate stripe size over 4GB
3634 Description: assertion fails when trying to use 4GB stripe size
3635 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3636 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3639 Frequency : always on ppc64
3641 Description: the first write on an ext3 filesystem with mballoc got stuck
3642 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3643 perform endianness conversion.
3645 ------------------------------------------------------------------------------
3647 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3649 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3650 INTEROPERATE with older versions automatically. Please read the
3651 user documentation before upgrading any part of a live system.
3652 * WARNING: Lustre networking configuration changes are required with
3653 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3656 * Support for kernels:
3657 2.6.9-22.0.2.EL (RHEL 4)
3658 2.6.5-7.244 (SLES 9)
3659 2.6.12.6 vanilla (kernel.org)
3662 Severity : enhancement
3663 Bugzilla : 7981/8208
3664 Description: Introduced Lustre Networking (LNET)
3665 Details : LNET is new networking infrastructure for Lustre, it includes
3666 a reorganized network configuration mode (see the user
3667 documentation for full details) as well as support for routing
3668 between different network fabrics. Lustre Networking Devices
3669 (LNDs) for the supported network fabrics have also been
3670 created for this new infrastructure.
3672 Severity : enhancement
3673 Description: Introduced Access control lists
3674 Details : clients can set ACLs on files and directories in order to have
3675 more fine-grained permissions than the standard Unix UGO+RWX.
3676 The MDS must be started with the "-o acl" mount option.
3678 Severity : enhancement
3679 Description: Introduced filesystem quotas
3680 Details : Administrators may now establish per-user quotas on the
3683 Severity : enhancement
3685 Description: Configuration change for the XT3
3686 The PTLLND is now used to run Lustre over Portals on the XT3
3687 The configure option(s) --with-cray-portals are no longer used.
3688 Rather --with-portals=<path-to-portals-includes> is used to
3689 enable building on the XT3. In addition to enable XT3 specific
3690 features the option --enable-cray-xt3 must be used.
3695 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3696 Details : A race between allocating a new llcd and re-getting the llcd_lock
3697 allowed another thread to grab newly-allocated llcd.
3699 Severity : enhancement
3701 Description: 2.6 OST async journal commit and locking fix to improve performance
3702 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3703 2.6 kernels are now async as they already were in 2.4 kernels so
3704 that they can commit concurrently with the network bulk transfer.
3705 For block-allocated files the filter allocation semaphore is held
3706 to avoid filesystem fragmentation during allocation. BKL lock
3707 removed for 2.6 xattr operations where it is no longer needed.
3712 Description: lconf incorrectly determined whether two IP networks could talk
3713 Details : In some more complicated routing and multiple-network
3714 configurations, lconf will avoid trying to make a network
3715 connection to a disjoint part of the IP space. It was doing the
3716 math incorrectly for one set of cases.
3721 Description: Fix for potential infinite loop processing records in an llog.
3722 Details : If an llog record is corrupted/zeroed, it is possible to loop
3723 forever in llog_process(). Validate the llog record length
3724 and skip the remainder of the block on error.
3727 Frequency : occasional (liblustre only)
3729 Description: liblustre could not open files whose last component is a symlink
3730 Details : sysio_path_walk() would incorrectly pass the open intent to
3731 intermediate path components.
3734 Frequency : rare (liblustre only with non-standard tuning)
3735 Bugzilla : 7201 (7350)
3736 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3737 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3738 after it has been released, only do this for referenced locks
3740 Severity : enhancement
3742 Description: specifying an (invalid) directory default stripe_size of -1
3743 would reset the directory default striping
3744 Details : stripe_size -1 was used internally to signal directory stripe
3745 removal, now use "all default" to signal dir stripe removal
3746 as a directory striping of "all default" is not useful
3749 Frequency : common for large clusters running liblustre clients
3751 Description: doing an ls when liblustre clients are running is slow
3752 Details : sending a glimpse AST to a liblustre client waits for every AST
3753 to time out, as liblustre clients will not respond. Since they
3754 cannot cache data we refresh the OST lock LVB from disk instead.
3756 Severity : enhancement
3758 Description: doing an ls at the same time as file IO can be slow
3759 Details : enqueue and other "small" requests can be blocked behind many
3760 large IO requests. Create a new OST IO portal for non-IO
3761 requests so they can be processed faster.
3764 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3766 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3767 created file with mode 000
3768 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3769 allow an ftruncate() as a non-root user to a file with mode 000.
3770 Lustre now respects this flag to disable mode checks when
3771 truncating a file owned by the user
3774 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3777 Description: Revoking locks from clients that went dead or catatonic might take
3779 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3780 cancellation of such locks instant on servers without waiting for
3781 any reply from clients. Clients drops these locks when cancel
3782 notification from server is received without replying.
3785 Frequency : liblustre-only, when liblustre client dies or becomes busy
3787 Description: Doing ls on Linux clients can take a long time with active
3789 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3790 granting such locks to it in the first place if possible. Locks
3791 are taken by proxy on the OST during the read or write and
3792 dropped immediately afterward. Add connect flags handling, do
3793 not grant locks to liblustre clients for glimpse ASTs.
3795 Severity : enhancement
3797 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3798 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3799 sequential reads and behaves badly with applications doing random
3800 reads. Improve it by reading ahead at least read region, and
3801 avoiding excessive large RPC for small reads.
3803 Severity : enhancement
3805 Description: Creating more than 1000 files for a single job may cause a load
3806 imbalance on the OSTs if there are also a large number of OSTs.
3807 Details : qos_prep_create() uses an OST index reseed value that is an
3808 even multiple of the number of available OSTs so that if the
3809 reseed happens in the middle of the object allocation it will
3810 still utilize the OSTs as uniformly as possible.
3815 Description: OST or MDS may oops in ping_evictor_main()
3816 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3817 but doesn't restart at beginning of obd_exports_timed list
3820 Severity : enhancement
3822 Description: improve by-nid export eviction on the MDS and OST
3823 Details : allow multiple exports with the same NID to be evicted at one
3824 time without re-searching the exports list.
3827 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3829 Description: MDS may oops in groups_free()
3830 Details : in rare race conditions a newly allocated group_info struct is
3831 freed again, and this can be NULL. The 2.4 compatibility code
3832 for groups_free() checked for a NULL pointer, but 2.6 did not.
3835 Frequency : common for liblustre clients doing little filesystem IO
3836 Bugzilla : 9352, 7313
3837 Description: server may evict liblustre clients accessing contended locks
3838 Details : if a client is granted a lock or receives a completion AST
3839 with a blocking AST already set it would not reply to the AST
3840 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3843 Frequency : lfs setstripe, only systems with more than 160 OSTs
3845 Description: unable to set striping with a starting offset beyond OST 160
3846 Details : llapi_create_file() incorrectly limited the starting stripe
3847 index to the maximum single-file stripe count.
3850 Frequency : LDAP users only
3852 Description: lconf did not handle in-kernel recovery with LDAP properly
3853 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3855 Severity : enhancement
3857 Description: bind OST threads to NUMA nodes to improve performance
3858 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3859 node and do their allocations there to localize memory access
3861 Severity : enhancement
3863 Description: llmount can determine client NID directly from Myrinet (GM)
3864 Details : the client NID code from gmnalnid was moved directly into
3865 llmount, removing the need to use this or specifying the
3866 client NID explicitly when mounting GM clients with zeroconf
3869 Frequency : if client is started with down MDS
3871 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3872 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3873 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3874 ptlrpc_import_delayed_req() because we didn't check if the
3875 request was interrupted, and we also didn't break out of the
3876 event loop if there was no timeout
3881 Description: data loss during non-page-aligned writes to a single file from
3882 both multiple nodes and multiple threads on one node at same time
3883 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3884 inconsistency led to false short-reads, that were cached and later
3885 used by ->prepare_write() to fill in partially written page,
3886 leading to data loss.
3889 Frequency : always, if lconf --abort_recovery used
3891 Description: lconf --abort_recovery fails with 'Operation not supported'
3892 Details : lconf was attempting to abort recovery on the MDT device and not
3895 Severity : enhancement
3897 Description: remove cleanup logs
3898 Details : replace lconf-generated cleanup logs with lustre internal
3899 cleanup routines. Eliminates the need for client-cleanup and
3902 Severity : enhancement
3904 Description: add support for EAs (user and system) on lustre filesystems
3905 Details : it is now possible to store extended attributes in the Lustre
3906 client filesystem, and with the user_xattr mount option it
3907 is possible to allow users to store EAs on their files also
3909 Severity : enhancement
3911 Description: Add possibility (config option) to show minimal available OST free
3913 Details : When compiled with --enable-mindf configure option, statfs(2)
3914 (and so, df) will return least minimal free space available from
3915 all OSTs as amount of free space on FS, instead of summary of
3916 free spaces of all OSTs.
3918 Severity : enhancement
3920 Description: do not expand extent locks acquired on OST-side
3921 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3922 by server: they are not cached anyway.
3925 Frequency : when mmap is used/binaries executed from Lustre
3927 Description: Unmmap pages before throwing them away from read cache.
3928 Details : llap_shrink cache now attempts to unmap pages before discarding
3929 them (if unmapping failed - do not discard). SLES9 kernel has
3930 extra checks that trigger if this unmapping is not done first.
3935 Description: lconf didn't resolve symlinks before checking to see whether a
3936 given mountpoint was already in use
3939 Frequency : when migrating failover services
3940 Bugzilla : 6395, 9514
3941 Description: When migrating a subset of services from a node (e.g. failback
3942 from a failover service node) the remaining services would
3943 time out and evict clients.
3944 Details : lconf --force (implied by --failover) sets the global obd_timeout
3945 to 5 seconds in order to quickly disconnect, but this caused
3946 other RPCs to time out too quickly. Do not change the global
3947 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3949 Severity : enhancement
3950 Frequency : if MDS is started with down OST
3951 Bugzilla : 9439,5706
3952 Description: Allow startup/shutdown of an MDS without depending on the
3953 availability of the OSTs.
3954 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3955 Add appropriate locking and lov-osc refcounts for safe
3956 cleaning. Add osc abort_inflight calls in case the
3957 synchronize never started.
3960 Frequency : occasional (Cray XT3 only)
3962 Description: root not authorized to access files in CRAY_PORTALS environment
3963 Details : The client process capabilities were not honoured on the MDS in
3964 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3965 been accessed by an authorized user then root was able to access
3966 the file on the local client also. The root user capabilities
3967 are now allowed on the MDS, as this environment has secure UID.
3970 Frequency : occasional
3972 Description: ldiskfs "too long searching" message happens too often
3973 Details : A debugging message (otherwise harmless) prints too often on
3974 the OST console. This has been reduced to only happen when
3975 there are fragmentation problems on the filesystem.
3980 Description: Division by zero in statfs when all OSCs are inactive
3981 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3982 lov_statfs divides by value returned by lov_get_stripecnt().
3986 Bugzilla : 9489, 3273
3987 Description: First write from each client to each OST was only 4kB in size,
3988 to initialize client writeback cache, which caused sub-optimal
3989 RPCs and poor layout on disk for the first writen file.
3990 Details : Clients now request an initial cache grant at (re)connect time
3991 and so that they can start streaming writes to the cache right
3992 away and always do full-sized RPCs if there is enough data.
3993 If the OST is rebooted the client also re-establishes its grant
3994 so that client cached writes will be honoured under the grant.
3999 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4000 Details : Now I/O RPCs go to different portal number and (presumably) fast
4001 lock requests (and glimses) and other RPCs get their own service
4002 threads pool that should be able to service those RPCs
4005 Severity : enhancement
4007 Description: Ability to exchange lustre version between client and servers and
4008 issue warnings at client side if client is too old. Also for
4009 liblustre clients there is ability to refuse connection of too old
4011 Details : New 'version' field is added to connect data structure that is
4012 filled with version info. That info is later checked by server and
4016 Frequency : rare, liblustre only.
4017 Bugzilla : 9296, 9581
4018 Description: Two simultaneous writes from liblustre at offset within same page
4019 might proceed at the same time overwriting eachother with stale
4021 Details : I/O lock withing llu_file_prwv was released too early, before data
4022 actually was hitting the wire. Extended lock-holding time until
4023 server acknowledges receiving data.
4026 Frequency : extremely rare. Never observed in practice.
4028 Description: avoid generating lustre_handle cookie of 0.
4029 Details : class_handle_hash() generates handle cookies by incrementing
4030 global counter, and can hit 0 occasionaly (this is unlikely, but
4031 not impossible, because initial value of cookie counter is
4032 selected randonly). Value of 0 is used as a sentinel meaning
4033 "unassigned handle" --- avoid it. Also coalesce two critical
4034 sections in this function into one.
4036 Severity : enhancement
4038 Description: allow liblustre clients to delegate truncate locking to OST
4039 Details : To avoid overhead of locking, liblustre client instructs OST to
4040 take extent lock in ost_punch() on client's behalf. New connection
4041 flag is added to handle backward compatibility.
4043 Severity : enhancement
4044 Bugzilla : 4928, 7341, 9758
4045 Description: allow number of OST service threads to be specified
4046 Details : a module parameter allows the number of OST service threads
4047 to be specified via "options ost ost_num_threads={N}" in the
4048 OSS's /etc/modules.conf or /etc/modprobe.conf.
4052 Bugzilla : 6146, 9635, 9895
4053 Description: servers crash with bad pointer in target_handle_connect()
4054 Details : In rare cases when a client is reconnecting it was possible that
4055 the connection request was the last reference for that export.
4056 We would temporarily drop the export reference and get a new
4057 one, but this may have been the last reference and the export
4058 was just destroyed. Get new reference before dropping old one.
4060 Severity : enhancement
4061 Frequency : if client is started with failover MDS
4063 Description: Allow multiple MDS hostnames in the mount command
4064 Details : Try to read the configuration from all specified MDS
4065 hostnames during a client mount in case the "primary"
4068 Severity : enhancement
4070 Description: Stop sending data to evicted clients as soon as possible.
4071 Details : Check if the client we are about to send or are sending data to
4072 was evicted already. (Check is done every second of waiting,
4073 for which l_wait_event interface was extended to allow checking
4074 of exit condition at specified intervals).
4077 Frequency : rare, normally only when NFS exporting is done from client
4079 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4081 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4084 Frequency : always, if srand() is called before liblustre initialization
4086 Description: Liblustre uses system PRNG disturbing its usage by user application
4087 Details : Introduce internal to lustre fast and high-quality PRNG for
4088 lustre usage and make liblustre and some other places in generic
4089 lustre code to use it.
4091 Severity : enhancement
4092 Bugzilla : 9477, 9557, 9870
4093 Description: Verify that the MDS configuration logs are updated when xml is
4094 Details : Check if the .xml configuration logs are newer than the config
4095 logs stored on the MDS and report an error if this is the case.
4096 Request --write-conf, or allow starting with --old_conf.
4098 Severity : enhancement
4100 Description: Handle symlinks in the path when checking if Lustre is mounted.
4101 Details : Resolve intermediate symlinks when checking if a client has
4102 mounted a filesystem to avoid duplicate client mounts.
4107 Description: lconf can hit an error exception but still return success.
4108 Details : The lconf command catches the Command error exception at the top
4109 level script context and will exit with the associated exit
4110 status, but doesn't ensure that this exit status is non-zero.
4115 Description: failure of ptlrpc thread startup can cause oops
4116 Details : Starting a ptlrpc service thread can fail if there are a large
4117 number of threads or the server memory is very fragmented.
4118 Handle this without oopsing.
4121 Frequency : always, only if liblustre and non-default acceptor port was used
4123 Description: liblustre cannot connect to servers with non-default acceptor port
4124 Details : tcpnal_set_default_params() was not called and was therefore
4125 ignoring the environment varaible TCPNAL_PORT, as well as other
4126 TCPNAL_ environment variables
4131 Description: two objects could be created on the same OST for a single file
4132 Details : If an OST is down, in some cases it was possible to create two
4133 objects on a single OST for a single file. No problems other
4134 than potential performance impact and spurious error messages.
4138 Bugzilla : 5681, 9562
4139 Description: Client may oops in ll_unhash_aliases
4140 Details : Client dcache may become inconsistent in race condition.
4141 In some cases "getcwd" can fail if the current directory is
4147 Description: Inode refcounting problems in NFS export code
4148 Details : link_raw functions used to call d_instantiate without obtaining
4149 extra inode reference first.
4153 Bugzilla : 9942, 9903
4154 Description: Referencing freed requests leading to crash, memleaks with NFS.
4155 Details : We used to require that call to ll_revalidate_it was always
4156 followed by ll_lookup_it. Also with revalidate_special() it is
4157 possible to call ll_revalidate_it() twice for the same dentry
4158 even if first occurence returned success. This fix changes semantic
4159 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4160 reference on a request referred from the intent.
4161 ll_intent_release() then releases such a request.
4164 Frequency : rare, normally benchmark loads only
4166 Description: unlinked inodes were kept in memory on the client
4167 Details : If a client is repeatedly creating and unlinking files it
4168 can accumulate a lot of stale inodes in the inode slab cache.
4169 If there is no other client load running this can cause the
4170 client node to run out of memory. Instead flush old inodes
4171 from client cache that have the same inode number as a new inode.
4174 Frequency : SLES9 2.6.5 kernel and long filenames only
4175 Bugzilla : 9969, 10379
4176 Description: utime reports stale NFS file handle
4177 Details : SLES9 uses out-of-dentry names in some cases, which confused
4178 the lustre dentry revalidation. Change it to always use the
4182 Frequency : rare, unless heavy write-truncate concurrency is continuous
4183 Bugzilla : 4180, 6984, 7171, 9963, 9331
4184 Description: OST becomes very slow and/or deadlocked during object unlink
4185 Details : filter_destroy() was holding onto the parent directory lock
4186 while truncating+unlinking objects. For very large objects this
4187 may block other threads for a long time and slow overall OST
4188 responsiveness. It may also be possible to get a lock ordering
4189 deadlock in this case, or run out of journal credits because of
4190 the combined truncate+unlink. Solution is to do object truncate
4191 first in one transaction without parent lock, and then do the
4192 final unlink in a new transaction with the parent lock. This
4193 reduces the lock hold time dramatically.
4196 Frequency : rare, 2.4 kernels only
4198 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4199 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4200 before unmounting their filesystem in order to avoid deadlock.
4201 The kernel_locked() function in 2.4 kernels only checks whether
4202 the kernel lock is held, not whether it is this process that is
4203 holding it as 2.6 kernels do.
4208 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4209 Details : The client ptlrpc code may be trying to reconnect to a down
4210 server before a previous connection attempt has timed out.
4211 Increase the reconnect interval to be longer than the connection
4212 timeout interval to avoid sending duplicate connections to
4216 Frequency : echo_client brw_test command
4218 Description: fix echo_client to work with OST preallocated code
4219 Details : OST preallocation code (5137) didn't take echo_client IO path
4220 into account: echo_client calls filter methods outside of any
4221 OST thread and, hence, there is no per-thread preallocated
4222 pages and buffers to use. Solution: hijack pga pages for IO. As
4223 a byproduct, this avoids unnecessary data copying.
4227 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4228 Description: Client can oops in mdc_commit_close() after open replay
4229 Details : It was possible for the MDS to return an open request with no
4230 transaction number in mds_finish_transno() if the client was
4231 evicted, but without actually returning an error. Clients
4232 would later try to replay that open and may trip an assertion
4233 Simplify the client close codepath, and always return an error
4234 from the MDS in case the open is not successful.
4237 Frequency : rare, 2.6 OSTs only
4239 Description: OST may deadlock under high load on fragmented files
4240 Details : If there was a heavy load and highly-fragmented OST filesystems
4241 it was possible to have all the OST threads deadlock waiting on
4242 allocation of biovecs, because the biovecs were not released
4243 until the entire RPC IO was completed. Instead, release biovecs
4244 as soon as they are complete to ensure forward IO progress.
4246 Severity : enhancement
4248 Description: Support for specifying external journal device at mount
4249 Details : If an OST or MDS device is formatted with an external journal
4250 device, this device major/minor is stored in the ext3 superblock
4251 and may not be valid for failover. Allow detecting and
4252 specifying the external journal at mount time.
4257 Description: Mounting an MDS with pending unlinked files may cause oops
4258 Details : target_finish_recovery() calls mds_postrecov() which returned
4259 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4260 considers this an error and immediately begins cleaning up the
4261 lov, just after starting the mds_lov process
4263 Severity : enhancement
4265 Description: Implement 'lfs df' to report actual free space on per-OST basis
4266 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4267 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4268 usage of objects; '-h' to report in human readable format.
4270 ------------------------------------------------------------------------------
4272 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4279 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4280 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4281 at startup during orphan unlinking. Instead, skip buddy update
4282 before setup, as it will be regenerated after recovery is complete.
4287 Description: If an OST is inactive, its locks might reference stale inodes.
4288 Details : lov_change_cbdata() must iterate over all namespaces, even if
4289 they are inactive to clear inode references from the lock.
4291 Severity : enhancement
4292 Frequency : occasional, if non-standard max_dirty_mb used
4294 Description: Client will block write RPCs if not enough grant
4295 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4296 then the client will block writes while waiting for another RPC
4297 to complete instead of consuming its dirty limit. With change
4298 we get improved performance when max_dirty_mb is small.
4300 Severity : enhancement
4301 Bugzilla : 3389, 6253
4302 Description: Add support for supplementary groups on the MDS.
4303 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4304 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4305 lookups for user supplementary groups into a cache.
4309 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4310 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4311 for liblustre O_EXCL is enforced by MDS.
4315 Description: readdir never returns NULL in liblustre.
4316 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4317 dentry in struct dirent.
4321 Frequency : liblustre only, depends on application IO pattern
4322 Description: liblustre clients evicted if not contacting servers
4323 Details : Don't put liblustre clients into the ping_evictor list, so
4324 they will not be evicted by the pinger ever.
4326 Severity : enhancement
4328 Description: Add ability to evict clients by NID from MDS.
4329 Details : By echoing "nid:$NID" string into
4330 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4331 $NID would be instantly evicted from this MDS and from all active
4332 OSTs connected to it.
4336 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4337 Details : lookup_it_finish() used to query file size from OSTs that was not
4342 Description: service threads change working directory to that of init
4343 Details : Starting lustre service threads may pin the working directory
4344 of the parent thread, making that filesystem busy. Threads
4345 now change to the working directory of init to avoid this.
4349 Frequency : during shutdown only
4350 Description: shutdown with a failed MDS or OST can cause unmount to hang
4351 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4356 Frequency : 2.6 only
4357 Description: chmod/chown may include an extra supplementary group
4358 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4359 supplementary group and if none is specified this is used.
4362 Bugzilla : 5479 (6816)
4364 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4365 Details : If lookup is in progress on a file that is unlinked we might try
4366 to revalidate the inode and fail in revalidate after lookup is
4367 complete and ll_file_open() enqueues the open again but
4368 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4371 Frequency : always, if lconf --abort_recovery used
4373 Description: lconf --abort_recovery fails with 'Operation not supported'
4374 Details : lconf was attempting to abort recovery on the MDT device and not
4377 ------------------------------------------------------------------------------
4379 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4384 Frequency : rare (only unsupported configurations with a node running as an
4386 Bugzilla : 6514, 5137
4387 Description: Mounting a Lustre file system on a node running as an OST could
4389 Details : OSTs now preallocates memory needed to write out data at
4390 startup, instead of when needed, to avoid having to
4391 allocate memory in possibly low memory situations.
4392 Specifically, if the file system is mounted on on OST,
4393 memory pressure could force it to try to write out data,
4394 which it needed to allocate memory to do. Due to the low
4395 memory, it would be unable to do so and the node would
4396 become unresponsive.
4398 Severity : enhancement
4400 Description: Addition of lconf --service command line option
4401 Details : lconf now accepts a '--service <arg>' option, which is
4402 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4404 Severity : enhancement
4406 Description: Failover mode is now the default for OSTs.
4407 Details : By default, OSTs will now run in failover mode. To return to
4408 the old behaviour, add '--failout' to the lmc line for OSTs.
4410 Severity : enhancement
4412 Description: Health checks are now provided for MDS and OSTs
4413 Details : Additional detailed health check information on MSD and OSTs
4414 is now provided through the procfs health_check value.
4417 Frequency : occasional, depends on IO load
4419 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4420 after numerous create/delete cycles
4421 Details : The ext3 inode allocation policy would not allocate new inodes
4422 very well on the OSTs because there are no new directories
4423 being created. Instead we look for groups with free space if
4424 the parent directories are nearly full.
4429 Description: Network or server problems during mount may cause partially
4430 mounted clients instead of returning an error.
4431 Details : The config llog parsing code may overwrite the error return
4432 code during mount error handling, returning success instead
4438 Description: MDS can fail to allocate large reply buffers
4439 Details : After long uptimes the MDS can fail to allocate large reply
4440 buffers (e.g. zconf client mount config records) due to memory
4441 fragmentation or consumption by the buffer cache. Preallocate
4442 some large reply buffers so that these replies can be sent even
4443 under memory pressure.
4447 Frequency : rare (liblustre)
4448 Description: fsx running with liblustre complained that using truncate() to
4449 extend the file doesn't work. This patch corrects that issue.
4450 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4451 ATTR_SIZE and lsm use in llu_setattr_raw.
4455 Frequency : rare, only 2.6 kernels
4456 Description: Unusual file access patterns on the MDS may result in inode
4457 data being lost in very rare circumstances.
4458 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4459 code in the 2.6 kernel caused the inode and/or EA data not to
4460 be read from disk, causing single-file corruption.
4464 Frequency : rare, only 2.6 filesystems using extents
4465 Description: Heavy concurrent write and delete load may cause data corruption.
4466 Details : It was possible under high-load situations to have an extent
4467 metadata block in the block device cache from a just-unlinked
4468 file overwrite a newly-allocated data block. We now unmap any
4469 metadata buffers that alias just-allocated data blocks.
4473 Frequency : filesystems with default stripe_count larger than 77
4474 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4475 Details : lconf specifies an inode size of 4096 bytes when the default
4476 stripe_count is larger than 77. This conflicts with the default
4477 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4478 this case to avoid pinning too much memory for large EAs.
4480 ------------------------------------------------------------------------------
4482 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4487 Frequency : rare (extremely heavy IO load with hundreds of clients)
4489 Description: Client is evicted, gets IO error writing to file
4490 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4491 caused clients to be evicted for AST timeouts. The fixes in
4492 bug 5192 mean we no longer need to have such short AST timeouts
4493 so ldlm_timeout has been increased.
4496 Frequency : occasional during --force or --failover shutdown under load
4497 Bugzilla : 5949, 4834
4498 Description: Server oops/LBUG if stopped with --force or --failover under load
4499 Details : a collection of import/export refcount and cleanup ordering
4500 issues fixed for safer force cleanup
4503 Frequency : only filesystems larger than 120 OSTs
4504 Bugzilla : 5990, 6223
4505 Description: lfs getstripe would oops on a very large filesystem
4506 Details : lov_getconfig used kfree on vmalloc'd memory
4509 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4510 Bugzilla : 6242, 6243
4511 Description: reading from files that had been truncated to a non-zero size
4512 but never opened returned no data
4513 Details : ll_file_read() reads zeros from no-object files to EOF
4518 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4519 Details : There's little harm in aborting MDS/OSS recovery and letting it
4520 try again, so I removed the LASSERT and return an error instead.
4522 Severity : enhancement
4524 Description: New debugging infrastructure for tracking down data corruption
4525 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4526 (b) cover more of the client-side code path, and (c) try to narrow
4527 down where problems occurred
4531 Bugzilla : 3819, 4364, 4397, 6313
4532 Description: Racing close and eviction MDS could cause assertion in mds_close
4533 Details : It was possible to get multiple mfd references during close and
4534 client eviction, leading to one thread referencing a freed mfd.
4536 Severity: : enhancement
4537 Bugzilla : 3262, 6359
4538 Description: Attempts to reconnect to servers are now more aggressive.
4539 Details : This builds on the enhanced upcall-less recovery that was added
4540 in 1.4.2. When trying to reconnect to servers, clients will
4541 now try each server in the failover group every 10 seconds. By
4542 default, clients would previously try one server every 25 seconds.
4547 Description: After recovery, certain operations trigger a failed
4548 assertion on a client.
4549 Details : Failing over an mds, using lconf -d --failover, while a
4550 client was doing a readdir() call would cause the client to
4551 LBUG after recovery completed and the readdir() was resent.
4553 Severity : enhancement
4555 Description: Default groups are now added by lconf
4556 Details : You can now run lconf --group <servicename> without having to
4557 manually add groups with lmc.
4560 Frequency : occasional
4562 Description: Nodes with an elan id of 0 trigger a failed assertion
4565 Frequency : always when accessing e.g. tty/console device nodes
4567 Description: tty and some other devices nodes cannot be used on lustre
4568 Details : file's private_data field is used by device data and lustre
4569 values in there got lost. New field was added to struct file to
4570 store fs-specific private data.
4573 Frequency : when exporting Lustre via NFS
4575 Description: NFSD failed occasionally when looking up a path component
4576 Details : NFSD is looking up ".." which was broken in ext3 directories
4577 that had grown large enough to become hashed.
4580 Frequency : Clusters with multiple interfaces not on the same subnet
4582 Description: Nodes will repeatedly try to reconnect to an interface which it
4583 cannot reach and report an error to the log.
4584 Details : Extra peer list entries will be created by lconf with some peers
4585 unreachable. lconf now validates the peer before adding it.
4588 Frequency : Only if a default stripe is set on the filesystem root.
4590 Description: Setting a default stripe on the filesystem root prevented the
4591 filesystem from being remounted.
4592 Details : The client was sending extra request flags in the root getattr
4593 request and did not allocate a reply buffer for the dir EA.
4596 Frequency : occasional, higher if lots of files are accessed by one client
4597 Bugzilla : 6159, 6097
4598 Description: Client trips assertion regarding lsm mismatch/magic
4599 Details : While revalidating inodes the VFS looks up inodes with ifind()
4600 and in rare cases can find an inode that is being freed.
4601 The ll_test_inode() code will free the lsm during ifind()
4602 when it finds an existing inode and then the VFS later attaches
4603 this free lsm to a new inode.
4607 Bugzilla : 6422, 7030
4608 Description: MDS deadlock between mkdir and client eviction
4609 Details : Creating a new file via mkdir or mknod (starting a transaction
4610 and getting the ns lock) can deadlock with client eviction
4611 (gets ns lock and trying to finish a synchronous transaction).
4614 Frequency : occasional
4615 Description: While starting a server, the fsfilt_ext3 module could not be
4617 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4618 kernels. Previously, lconf would still use the ext3 name
4619 when trying to load modules. Now, it will correctly use
4620 ext3 on 2.4 and ldiskfs on 2.6.
4622 Severity : enhancement
4623 Description: The default stripe count has been changed to 1
4624 Details : The interpretation of the default stripe count (0, to lfs
4625 or lmc) has been changed to mean striping across a single
4626 OST, rather than all available. For general usage we have
4627 found a stripe count of 1 or 2 works best.
4629 Severity : enhancement
4630 Description: Add support for compiling against Cray portals.
4631 Details : Conditional compiling for some areas that are different
4635 Frequency : occasional
4636 Bugzilla : 6409, 6834
4637 Description: Creating files with an explicit stripe count may lead to
4638 a failed assertion on the MDS
4639 Details : If some OSTs are full or unavailable, creating files may
4640 trigger a failed assertion on the MDS. Now, Lustre will
4641 try to use other servers or return an error to the
4645 Frequency : occasional
4647 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4648 caused lock timeout during readahead (since 1.4.2).
4649 Details : Processes doing readahead might match a lock that hasn't been
4650 granted yet if there are overlapping and conflicting lock
4651 requests. The readahead process waits on ungranted lock
4652 (original lock is CBPENDING), while OST waits for that process
4653 to cancel CBPENDING read lock and eventually evicts client.
4655 Severity : enhancement
4657 Description: Initial enabling of flock support for clients
4658 Details : Implements fcntl advisory locking and file status functions.
4659 This feature is provided as an optional mount flag (default
4660 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4661 locking are implemented yet, and those that are are not guaranteed
4662 to be completely correct in production environments.
4663 mount -t lustre -o [flock|noflock] ...
4666 Frequency : occasional
4668 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4669 assertion in the ext3 JBD (journaling) layer.
4670 Details : The b_committed_data struct is protected by the big kernel lock
4671 in 2.4 kernels, serializing journal_commit_transaction() and
4672 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4673 access to this struct. In 2.6 kernels there is finer grained
4674 locking to improve SMP performance of the JBD layer.
4678 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4683 Description: merge and cleanup kernel patches.
4686 -----------------------------------------------------------------------------
4688 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4690 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4691 versions of Lustre. Please update all servers and clients to
4692 version 1.4.2 or later at the same time. You must also run
4693 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4694 to update the configuration logs.
4696 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4697 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4698 - swab llog records of type '0' so we get proper header size/idx (5861)
4699 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4700 - fix rename of one directory over another leaking an inode (5953)
4701 - avoid SetPageDirty on 2.6 (5981)
4702 - don't re-add just-being-destroyed locks to the waiting list (5653)
4703 - when creating new directories, inherit the parent's custom
4704 striping settings if present parent (3048)
4705 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4706 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4707 - don't hold client locks on temporary worklist from l_lru (5666)
4708 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4709 - automatically evict dead clients (5921)
4710 - Update file size properly in create+truncate+fstat case (6196)
4711 - Do not unhash mountpoint dentries, do not allow removal of
4713 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4714 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4715 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4716 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4717 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4718 - Correctly handle reads of files with no objects (6243)
4719 - lctl recover will also mark a device active if deactivate used (5933)
4721 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4722 - allow --write-conf on an MDS with different nettype than client (5619)
4723 - don't write config llogs to MDS for mounts not from that MDS (5617)
4724 - lconf should create multiple TCP connections from a client (5201)
4725 - init scripts are now turned off by default; run chkconfig --on
4726 lustre and chkconfig --on lustrefs to use them
4727 - upcalls are no longer needed for clients to recover to failover
4729 - add --abort-recovery option to lconf to abort recovery on device
4731 - add support for an arbitrary number of OSTs (3026)
4732 - Quota support protocol changes.
4733 - forward compatibility changes to wire structs (6007)
4734 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4735 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4736 - improved "lustre status" script
4737 - initialize blocksize for non-regular files (6062)
4738 - added --disable-server and --disable-client configure options (5782)
4739 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4740 - Vanilla 2.4.29 support
4741 - increase maximum number of obd devices to 520 (6242)
4742 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4743 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4744 SLES 9 2.6.5 kernels
4745 - limit stripes per file to 160 (the maximum EA size) (6093)
4747 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4750 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4751 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4752 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4753 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4754 - small ext3 extents cleanups and fixes (5733)
4755 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4756 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4757 - avoid needless client->OST connect, fix handle mismatch (5317)
4758 - fix DLM error path that led to out-of-sync client, long delays (5779)
4759 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4760 - fix several locking issues related to i_size (5492,5624,5654,5672)
4761 - don't move pending lock onto export if it is already evicted (5683)
4762 - fix kernel oops when creating .foo in unlinked directory (5548)
4763 - fix deadlock in obdfilter statistics vs. object create (5811)
4764 - use time_{before,after} to avoid timer jiffies wrap (5882)
4765 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4766 - Do not leak request if server was not able to process it (5154)
4767 - If mds_open unable to find parent dir, make that negative lookup(5154)
4768 - don't create new directories with extent-mapping (5909, 5936)
4770 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4771 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4772 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4773 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4774 - support for RHEL4 kernel on i686 (5773)
4775 - provide error messages when incompatible logs are encountered (5898)
4777 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4778 * version 1.4.0.10 (1.4.1 release candidate 1)
4780 - don't keep a lock reference when lock is not granted (4238)
4781 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4782 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4783 - drop import inflight refcount on signal_completed_replay error (5255)
4784 - unlock page after async write error during send (3677)
4785 - handle missing objects in filter_preprw_read properly (5265)
4786 - no transno return for symlink open, don't save no-trasno open (3440)
4787 - don't try to complete elan receive that already failed (4012)
4788 - free RPC server reply state on error (5406)
4789 - clean up thread from ptlrpc_start_thread() on error (5160)
4790 - readahead could read extra page into cache that wasn't ejected (5388)
4791 - prevent races in class_attach/setup/cleanup/detach (5260)
4792 - don't dereference de->d_inode after l_dput of de (5458)
4793 - use "int" for stripe value returned from lock_to_stripe (5544)
4794 - mballoc allocation and error-checking fixes in 2.6 (5504)
4795 - block device patches to fix I/O request sizes in 2.6 (5482)
4796 - look up hostnames for IB nals (5602)
4797 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4798 - don't start multiple acceptors for the same port (5277)
4799 - fix incorrect LASSERT in mds_getattr_name (5635)
4800 - export a proc file for general "ping" checking (5628)
4801 - fix "lfs check" to not block when the MDS is down (5628)
4803 - service request history (4965)
4804 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4805 - create an "evict_client" /proc entry on OSTs, like the MDS has
4806 - fix mount usage message, return errors per mount(8) (5168)
4807 - change grep [] to grep "[]" in tests so they work in more UMLs
4808 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4809 - remove ancient LOV_MAGIC_V0 EA support (5047)
4810 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4811 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4812 - allow readahead allocations to fail when low on memory (5383)
4813 - mmap locking landed again, after considerable improvement (2828)
4814 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4816 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4819 - send OST transaction number in read/write reply to free req (4966)
4820 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4821 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4822 - the watchdog thread now runs as interruptible (5246)
4823 - flock/lockf fixes (but it's still disabled, pending 5135)
4824 - don't use EXT3 constants in llite code (5094)
4825 - memory shortage at startup could cause assertion (5176)
4827 - reorganization of lov code
4828 - single portals codebase
4830 - add extents/mballoc support (5025)
4831 - direct I/O reads in the obdfilter (4048)
4832 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4834 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4837 - send OST transaction number in read/write reply to free req (4966)
4838 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4839 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4840 - the watchdog thread now runs as interruptible (5246)
4841 - handle missing objects in filter_preprw_read properly (5265)
4842 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4843 - drop import inflight refcount on signal_completed_replay error (5255)
4844 - unlock page after async write error during send (3677)
4845 - return original error code on reconstructed replies (3761)
4846 - no transno return for symlink open, don't save no-trasno open (3440)
4848 - add pid to ldlm debugging output (4922)
4849 - bump the watchdog timeouts -- we can't handle 30sec yet
4850 - extra debugging for orphan dentry/inode bug (5259)
4852 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4855 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4856 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4857 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4858 - fix NULL dereference in /proc/sys/portals/routes (4827)
4859 - allow failed mdc_close() operations to be interrupted (4561)
4860 - stop precreate on OST before MDS would time out on it (4778)
4861 - don't send partial-page writes before EOF from client (4410)
4862 - discard client grant for sub-page writes on large-page clients (4520)
4863 - don't free dentries not owned by NFS code, check generation (4806)
4864 - fix lsm leak if mds_create_objects() fails (4801)
4865 - limit debug_daemon file size, always print CERROR messages (4789)
4866 - use transno after validating reply (3892)
4867 - process timed out requests if import state changes (3754)
4868 - update mtime on OST during writes, return in glimpse (4829)
4869 - add mkfsoptions to LDAP (4679)
4870 - use ->max_readahead method instead of zapping global ra (5039)
4871 - don't interrupt __l_wait_event() during strace
4873 - add software watchdogs to catch hung threads quickly (4941)
4874 - make lustrefs init script start after nfs is mounted
4875 - fix CWARN/ERROR duplication (4930)
4876 - return async write errors to application if possible (2248)
4877 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4878 - print NAL number in %x format (4645)
4879 - update barely-supported suse-2.4.21-171 series (4842)
4880 - support for sles 9 %post scripts
4881 - support for building 2.6 kernel-source packages
4882 - support for sles km_* packages
4884 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4887 - ignore -ENOENT errors in osc_destroy (3639)
4888 - notify osc create thread that OSC is being cleaned up (4600)
4889 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4890 - reconstruct ost_handle() like mds_handle() (4657)
4891 - create a new thread to do import eviction to avoid deadlock (3969)
4892 - let lconf resolve symlinked-to devices (4629)
4893 - don't unlink "objects" from directory with default EA (4554)
4894 - hold socknal file ref over connect in case target is down (4394)
4895 - allow more than 32000 subdirectories in a single directory (3244)
4896 - fix blocks count for O_DIRECT writes (3751)
4897 - OST returns ENOSPC from object create when no space left (4539)
4898 - don't send truncate RPC if file size isn't changing (4410)
4899 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4900 - bind to privileged port in socknal and tcpnal (3689)
4902 - rate limit CERROR/CWARN console message to avoid overload (4519)
4903 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4904 - basic mmap support (3918)
4905 - kernel patch series update from b1_4 (4711)
4907 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4910 - avoid crash during MDS cleanup with OST shut down (2775)
4911 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4912 - don't use bad inodes on the MDS (3744)
4913 - dynamic object preallocation to improve recovery speed (4236)
4914 - don't hold spinlock over lock dumping or change debug flags (4401)
4915 - don't zero obd_dev when it is force cleaned (3651)
4916 - print grants to console if they go negative (4431)
4917 - "lctl deactivate" will stop automatic recovery attempts (3406)
4918 - look for existing locks in ldlm_handle_enqueue() (3764)
4919 - don't resolve lock handle twice in recovery avoiding race (4401)
4920 - revalidate should check working dir is a directory (4134)
4922 - don't always mark "slow" obdfilter messages as errors (4418)
4924 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4927 - don't close LustreDB during write_conf until it is done (3860)
4928 - fix typo in lconf for_each_profile (3821)
4929 - allow dumping logs from multiple threads at one time (3820)
4930 - don't allow multiple threads in OSC recovery (3812)
4931 - fix debug_size parameters (3864)
4932 - fix mds_postrecov to initialize import for llog ctxt (3121)
4933 - replace config semaphore with spinlock (3306)
4934 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4935 - don't allow enqueue to complete on a destroyed export (3822)
4936 - down write_lock before checking llog header bitmap (3825)
4937 - recover from lock replay timeout (3764)
4938 - up llog sem before sending rpc (3652)
4939 - reduce ns lock hold times when setting kms (3267)
4940 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4941 - fix NULL deref and obd_dev leak on setup error (3312)
4942 - replace some LBUG about llog ops with error handling (3841)
4943 - don't match INVALID dentries from d_lookup and spin (3784)
4944 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4945 - fix invalid assertion in ptlrpc_set_wait (3880)
4947 - add libwrap support for the TCP acceptor (3996)
4948 - add /proc/sys/portals/routes for non-root route listing (3994)
4949 - allow setting MDS UUID in .xml (2580)
4950 - print the stack of a process that LBUGs (4228)
4952 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4955 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4956 - make sure to unset replay flag from failed open requests (3440)
4957 - if default stripe count is 0, use OST count for inode size (3636)
4958 - update parent mtime/ctime on client for create/unlink (2611)
4959 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4960 - free recovery state on server during a forced cleanup (3571)
4961 - unregister_reply for resent reqs (3063)
4962 - loop back devices mounting and status check on 2.6 (3563)
4963 - fix resource-creation race that can provoke i_size == 0 (3513)
4964 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4965 - more debugging for page-accounting assertion (3746)
4966 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4967 - avoid deadlock after precreation failure (3758)
4968 - fix race and lock order deadlock in orphan handling (3450, 3750)
4969 - add validity checks when grabbing inodes from l_ast_data (3599)
4971 - add /proc/.../recovery_status to obdfilter (3428)
4972 - lightweight CDEBUG infrastructure, debug daemon (3668)
4973 - change default OSC RPC parameters to be better on small clusters
4974 - turn off OST read cache for files smaller than 32MB
4975 - install man pages and include them in rpms (3100)
4976 - add new init script for (un)mounting lustre filesystems (2593)
4977 - run chkconfig in %post for init scripts (3701)
4978 - drop scimac NAL (unmaintained)
4980 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4983 - clean kiobufs before and after use (3485)
4984 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4985 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4986 - change init script to fail more clearly if not run as root (1528)
4987 - allow clients to reconnect during replay (1742)
4988 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4989 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4990 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4991 - fix cond_resched (3554)
4992 - only evict unfinished clients after recovery (3515)
4993 - allow bulk resend, prevent data loss (3570)
4994 - dynamic ptlrpc request buffer allocation (2102)
4995 - don't allow unlinking open directory if it isn't empty (2904)
4996 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4997 - remove extraneous obd dereference causing LASSERT failure (3334)
4998 - don't use get_cycles() when creating temp. files on the mds (3156)
4999 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5000 - handle EEXIST for set-stripe, set proper directory name (3336)
5002 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5003 - fix ksocknal_fmb_callback() error messages (2918)
5005 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5008 - don't copy lvb into (possibly NULL) reply on error (2983)
5009 - don't deref dentry after dput, don't free lvb on error (2922)
5010 - use the kms to determine writeback rpc length (2947)
5011 - increment oti_logcookies when osc is inactive (2948)
5012 - update client's i_blocks count via lvb messages (2543)
5013 - handle intent open/close of special files properly (1557)
5014 - mount MDS with errors=remount-ro, like obdfilter (2009)
5015 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5016 - don't use cancelling-locks' kms values (2947)
5017 - use highest lock extent for kms, not last one (2925)
5018 - don't dereference ERR_PTR() dentry in error handling path (3107)
5019 - fix thread race in portals_debug_dumplog() (3122)
5020 - create lprocfs device entries at setup instead of at attach (1519)
5021 - common AST error handler, don't evict client on completion race (3145)
5022 - zero nameidata in detach_mnt in 2.6 (3118)
5023 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5024 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5025 - initialize RPC timeout timer earlier for 2.6 (3219)
5026 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5027 - print nal/nid for unknown nid (3258)
5028 - additional checks for oscc recovery before doing precreate (3284)
5029 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5030 - don't crash in mdc_close for bad permissions on open (3285)
5031 - zero i_rdev for non-device files (3147)
5032 - clear page->private before handing to FS, better assertion (3119)
5033 - tune the read pipeline (3236)
5034 - fix incorrect decref of invalidated dentry (2350)
5035 - provide read-ahead stats and refine rpc in flight stats (3328)
5036 - don't hold journal transaction open across create RPC (3313)
5037 - update atime on MDS at close time (3265)
5038 - close LDAP connection when recovering to avoid server load (3315)
5039 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5040 - don't leak open file on MDS after open resend (3325)
5041 - serialize filter_precreate and filter_destroy_precreated (3329)
5042 - loop device shouldn't call sync_dev() for nul device (3092)
5043 - clear page cache after eviction (2766)
5044 - resynchronize MDS->OST in background (2824)
5045 - refuse to mount the same filesystem twice on same mountpoint (3394)
5046 - allow llmount to create routes for mounting behind routers (3320)
5047 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5048 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5049 - fix rare problem with rename on htree directories (3417)
5051 - allow default OST striping configuration per directory (1414)
5052 - fix compilation for qswnal for 2.6 kernels (3125)
5053 - increase maximum number of MDS request buffers for large systems
5054 - change liblustreapi to be useful for external progs like lfsck (3098)
5055 - increase local configuration timeout for slow disks (3353)
5056 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5058 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5061 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5062 - don't overwrite extent policy data in reply if lock was blocked (2901)
5063 - drop filter export grants atomically with removal from device (2663)
5064 - del obd_self_export from work_list in class_disconnect_exports (2908)
5065 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5066 - swab reply message in mdc_close, other PPC fixes (2464)
5067 - fix destroying of named logs (2325)
5068 - overwrite old logs when running lconf --write_conf (2264)
5069 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5070 - fix race in target_handle_connect (2898)
5071 - mds_reint_create() should take same inode create lock (2926)
5072 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5073 - don't close files for self_export to avoid uninitialized obd (2936)
5074 - allow MDS with the same name as client node (2939)
5075 - hold dentry reference for closed log files for unlink (2325)
5076 - reserve space for all logs during transactions (2059)
5077 - don't evict page beyond end of stripe extent (2925)
5078 - don't oops on a deleted current working directory (2399)
5079 - handle hard links to targets without a parent properly (2517)
5080 - don't dereference NULL lock when racing during eviction (2867)
5081 - don't grow lock extents when lots of conflicting locks (2919)
5083 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5086 - account for cache space usage on clients to avoid data loss (974)
5087 - lfsck support in lustre kernel code (2349)
5088 - reduce journal credits needed for BRW writes (2370)
5089 - orphan handling to avoid losing space on client/server crashes
5090 - ptlrpcd can be blocked, stopping ALL progress (2477)
5091 - use lock value blocks to assist in proper KMS, faster stat (1021)
5092 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5093 - recovery for initial connections (2355)
5094 - fixes for mds_cleanup_orphans (1934)
5095 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5096 - block all file creations until orphan recovery completes (1901)
5097 - client remove rq_connection from request struct (2423)
5098 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5099 - recovery timer race (2670)
5100 - mdc_close recovey bug (2532)
5101 - ptlrpc cleanup bug (2710)
5102 - mds timeout on local locks (2588)
5103 - namespace lock held during RPCs (2431)
5104 - handle interrupted sync write properly (2503)
5105 - don't try to handle a message that hasn't been replied to (2699)
5106 - client assert failure during cleanup after abort recovery (2701)
5107 - leak mdc device after failed mount (2712)
5108 - ptlrpc_check_set allows timedout requests to complete (2714)
5109 - wait for inflight reqs when ptlrpcd finishes (2710)
5110 - make sure unregistered services are removed from the srv_list
5111 - reset bulk XID's when resending them (caught by 1138 test)
5112 - unregister_bulk after timeout
5113 - fix lconf error (2694)
5114 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5115 - readahead locks pages, leaves pending causing memory pressure (2673)
5116 - increase OST request buffers to 4096 on large machines (2729)
5117 - fix up permission of existing directories in simple_mkdir (2661)
5118 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5119 - don't assign transno to errored transactions (2742)
5120 - don't delete objects on OST if given a bogus objid from MDS (2751)
5121 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5122 - if rq_no_resend, then timeout request after recovery (2432)
5123 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5124 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5125 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5126 - MDS recovery shouldn't delete objects that it has given out (2730)
5127 - if enqueue arrives after completion, don't clobber LVB (2819)
5128 - don't unlock pages twice when trigger_group_io returns error (2814)
5129 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5130 - don't write pages to disk if there was an error (1450)
5131 - don't ping imports that have recovery disabled (2676)
5132 - take buffered bytes into account when balancing socknal conn (2817)
5133 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5134 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5135 - remove little-endian swabbing of llog records (1987)
5136 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5137 - retry reposting request buffers if they fail (1191)
5138 - grow extent at grant time to avoid granting a revoked lock (2809)
5139 - lock revoke doesn't evict page if covered by a second lock (2765)
5140 - disable VM readahead to avoid reading outside lock extents (2805)
5142 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5143 - updated kernel patches for hp-2.4.20 kernel (2681)
5145 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5148 - fix truncated write corruption (2366)
5149 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5151 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5152 - ensure len(uuid) < 37 in lmc (1171)
5153 - fix ia64 OOPS in llog_test (2255)
5154 - zero end of page at obdfilter for partial page writes (2648)
5155 - don't leave stale dentries around after renames (bug 2428)
5156 - fix timeouts when evicting a client with a single lock held (2642)
5157 - set deadline for the initial HELLO message to drain (2634)
5158 - print out dotted-quad IP addresses in the socknal (2302)
5160 - additional debugging for MDS client eviction problem (2443)
5161 - fix mkfsoptions support for osts (2603, 2604)
5163 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5166 - add series for the vanilla 2.6.0 kernel
5167 - add series for the vanilla 2.4.24 kernel
5168 - add series for a cray x86/64 UL kernel drop
5169 - fix xattr patches for the vanilla 2.4.19 series
5171 - generate true UUIDs in lmc (1171)
5172 - have portals stack dumping break in UML (2466)
5173 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5174 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5175 - wake callback waiting threads on client eviction (2460)
5176 - Add --ptldebug and --subsystem to lmc (1719)
5177 - update assertion to allow safe interrupt allocation
5178 - set rq_no_resend for cancel requests (2432)
5179 - recalculate ptlrpcd timeout after resend (2494)
5180 - call vfs_rmdir when removing pending directories (2368)
5181 - fix renaming a file to itself (2429)
5182 - lmc creates a default one-stripe lov (2454)
5183 - expand procfs space to handle large clusters (2326)
5184 - increase UML stack to avoid overflow
5185 - update lconf's list of debug and subsystem masks
5186 - fix lfs find --obd (2510)
5187 - /proc tunable for disabling filter read caching (2591)
5188 - stop rpm packages from altering slapd.conf (2301)
5189 - disable nagle in the socknal under 0conf (2578)
5190 - choose mds inode size based on stripe count (2572)
5191 - fix kernel-source rpm problems (2516)
5193 - add --disable-doc to avoid pdf generation (2421)
5194 - update documentation, tests, type-os, comments
5195 - avoid format warnings on ia64
5196 - remove the TOE NAL
5197 - tiny code cleanups by removing unused fields
5199 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5202 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5203 - remove the most blatant lies from BUILDING file (2371)
5204 - change default debug level to reasonable production setting
5205 - reduce client side cache size to reduce cache flush time
5206 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5207 - make TCP zerocopy and pinger support enabled by default (2476)
5208 - sync writes completed after process exits caused crashes (2319)
5209 - maintain correct mount count on the MDS (2356)
5210 - backout 1557, because 2316 wasn't really fixed
5211 - better file I/O statistics gathering in /proc
5212 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5213 - another kernel patch to fix zero-copy TCP function export
5214 - don't take duplicate lock when processing re-sent getattr (2420)
5215 - lctl uses obd_self_export instead of creating new conn (2353)
5216 - MDS/OST recovery case which requires object creation asserted (2425)
5217 - move lfs from /usr/sbin to /usr/bin in packages
5218 - fix race between mds_client_add and mds_client_free (2417)
5219 - use kmalloc instead of slabs in portals (2430)
5220 - don't create duplicate records when a failover MDS is present (2442)
5221 - remove unnecessary mount age check (2332)
5222 - don't remove directory inodes from locks prematurely (2451)
5223 - don't break if MDS service name is the same as hostname (2103)
5224 - fix races in client write RPC generation when cache full (2482)
5226 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5229 - remove now-unused request->rq_obd (278)
5230 - if an allocation fails, print out how much memory we've used (1933)
5231 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5232 - add the "configurable stack size" patch to most series files (1256)
5233 - ability to write large log records, for 100+ OST configs (2306)
5234 - fix NULL deref when filter_prep fails (2314)
5235 - fix operator precedence error in filter_sync
5236 - dynamic allocation of socknal TX descriptors (2315)
5237 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5238 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5239 - ensure that CWARN messages in Portals always get to the syslog
5240 - __init/__exit are not for prototype decls (ldlm_init/exit)
5241 - x86-64 compile warning fixes
5242 - fix gateway LMC keyword conflict (2318)
5243 - fix MDS lock inversions in getattr/reint paths (1844)
5244 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5245 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5246 - DLM race condition prevented some lock evictions (2328)
5247 - ENOMEM detection and retry on socknal sends (2230)
5248 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5249 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5250 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5251 - better error messages when a client is rejected during recovery (1505)
5252 - avoid cancelling locks which were never granted, after failure (2330)
5253 - fix i_sem/journal inversion in mds_client_add (2333)
5254 - fix truncate/getattr lock cycle deadlock (2334)
5255 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5256 - fix two rare exit paths which could leak an l_lock() ref (2321)
5257 - fencepost error in MDS/OST orphan recovery (2226)
5258 - make log record alignment 8 bytes (1988)
5259 - lstripe now fails when requested offset > ost_count (2237)
5260 - ensure that all kernel series have a complete list.h (1607)
5261 - fix crashes in special-file operations (2316)
5262 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5264 - allow configurable automake binary, for testing new versions
5265 - small update to the lfs documentation
5267 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5269 * fix negative export reference count in fsfilt_sync (2312)
5271 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5272 * release candidate 0.9.1
5274 - orphans are moved into the PENDING directory for possible recovery
5275 - replayed opens now open by fid for orphan/rename safety (1042)
5276 - last close of an orphan inode generates a transno (683)
5277 - chdir() and mount() now pin the directory entry (1020)
5278 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5279 - discard very old requests without processing them (1502)
5280 - remove some common, well-understood CERRORs (1505)
5281 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5282 - clear "grant" flags in OST replies until OST grant code lands (1644)
5283 - fix read performance by not clobbering i_blksize on client (1598)
5284 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5285 - make LRU size a /proc tunable, clears locks when reduced (707)
5286 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5287 - server locks take references on exports now (1558)
5288 - build fixes for 2.4.20-rh trees (1663)
5289 - return an error from lov_create if all OSCs are inactive (1751)
5290 - fix import levels when a reconnect happens without a timeout (1597)
5291 - exit early from mds_open if we get a lookup error (1749)
5292 - partial page read at EOF wouldn't wait for disk before sending (1642)
5293 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5294 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5295 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5296 - write/truncate lock inversion (1639)
5297 - Don't auto-load obdclass, portals modules during cleanup (1495)
5298 - fix timestamps from jumping to "now" (1763)
5299 - extra journal assertions (1648)
5300 - add an extra multiunlink test (1771)
5301 - fix read_record/write_record API (1776)
5302 - fix leak of offset_extent, possible incorrect i_size later (1772)
5303 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5304 - Debugging for the kqswnal_get_idle_tx problems (1820)
5305 - Allow recovery to be attempted multiple times (1536)
5306 - Write out MDS last_rcvd file after it is first created (1600)
5307 - Fix tx_descriptor leak in failed transmit situations (1827)
5308 - ext3 journaling fixes for assertion failure after IO error (1871)
5309 - class_export_put() on freed export after completion AST error (1896)
5310 - Fix revalidate looping in VFS (1322)
5311 - Don't access a freed export during MDS_REINT timeout (1521)
5312 - Add open-unlink recovery support on the MDS (1673,1764)
5313 - Return an error if no MDS data was read from last_rcvd (1946)
5314 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5315 - Don't setup a disk device that doesn't match exported UUID (317)
5316 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5317 - avoid committing NULL handle in force close
5318 - local.sh is now a one-stripe LOV configuration
5319 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5320 - fix ext3 htree duplicate directory entry corruption (1516)
5321 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5322 - update to new LOV EA format (2097)
5323 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5325 2003-06-15 Phil Schwan <phil@clusterfs.com>
5328 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5329 - per-import/export recovery handling (958, 931, 959)
5330 - multiple last-rcvd slots, for serving multiple FSes (949)
5331 - connections are again shared between multiple imp/exports (963, 964)
5332 - "umount -f" would hang if any requests needed to be sent (393, 978)
5333 - avoid pinning large req buffer by copying for queued messages (989)
5334 - add "uuid" to "lctl device" command to help upcalls (991)
5335 - "open" RPCs with transnos would confuse recovery counters (1037)
5336 - do proper endian conversion of all wire messages (288, 340, 891)
5337 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5338 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5339 - fix ptlrpc_connection leak in target_handle_connect (1174)
5340 - fix import refcounting bug in OST and MDS cleanup (1134)
5341 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5342 - fix very unlikely obd_types race condition (501)
5343 - remove osc_open hack for echo_client (1187)
5344 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5345 - a failure in read_inode2 leads to deadlock (1139)
5346 - cancel ack-locks as soon as transaction is committed (1072)
5347 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5348 - make sure to commitrw after any preprw to avoid deadlock (1162)
5349 - failing to execute a file in a lustre FS would lock inode (1203)
5350 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5351 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5352 - obd_uuid2tgt would walk off the end of the list (1255)
5353 - on IA64 the osc would give portals incorrect bulk size (1258)
5354 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5355 - fix lock inversion caused by new llite matching code (1282)
5356 - limit the number of dirty pages on a client to 10MB (1286)
5357 - timed out locks were not being corrected cancelled (1289)
5358 - fix O_DIRECT above 4GB on IA-32 (1292)
5359 * major user-visible changes
5360 - fail out/fail over policy now controlled by the upcall (993)
5362 - add OBD_PING to check server availability and failure (954)
5363 - lustre messages are now sent in sending host order (288, 340, 891)
5364 - add eadatalen to MDS getattr reply (340)
5365 - OST read replies may contain second buffer, with per-page status (593)
5367 2003-03-11 Phil Schwan <phil@clusterfs.com>
5370 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5371 - failed open()s could cause deadlock; fixed (867, 869)
5372 - stop cancelling OST locks when files are closed (481)
5373 - overlapping XID spaces caused network corruption (851, 853)
5374 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5375 - setattr_raw added, to do single-RPC, server-side setattrs
5376 - lmc/lconf syntax change for OST UUIDs
5377 - fix crashy race condition between ptlrpc_free_req and osc_close
5378 - don't use request in mdc_enqueue if we hit a timeout (889)
5379 - don't set the inode i_size for regular files from the MDS (896)
5380 - handle out of order completion AST (842)
5381 - don't LBUG if a lock request times out after receiving AST (913)
5382 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5383 - if a bad lock AST arrives, send an error instead of dropping entirely
5384 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5385 - fix leak in bulk IO when only partially completed (899, 900, 926)
5386 - fix O_DIRECT for ia64 (55)
5387 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5388 - C-z after timeout could hang a process forever; fixed (977)
5390 - client-side I/O cache (678, 924, 929, 941, 970)
5392 - READPAGE and SETATTRs which don't take server-side locks get
5395 2003-02-11 Phil Schwan <phil@clusterfs.com>
5398 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5399 - Fix fsfilt_extN_readpage() to read a full page of directory
5400 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5401 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5402 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5403 - Most intent operations were removed, in favour of a new RPC mode
5404 that does a single RPC to the server and bypasses most of the VFS
5405 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5406 - Aggressively cancel local locks on DLM servers
5407 - mds_reint_unlink sends EA to the client if it's the last nlink.
5408 client uses that EA to unlink OST objects.
5409 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5410 - recursive symlinks were fixed (439)
5411 - fixed NULL deref in DEBUG_REQ
5412 - filter_update_lastobjid no longer calls sync, which annoyed extN
5413 - fixed multi-client small-writes to a single file problem (445)
5414 - fixed mtime updates during file writes (607)
5415 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5416 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5417 - fixed Linux OST size reporting problem (444, 656)
5418 - OST now updates object mtime with writes or setattr (607, 619)
5419 - client verifies file size before zeroing page past EOF (445)
5420 - OST now writes last allocated objid to disk with allocation (108)
5421 - LOV on echo now works (409)
5423 - mds_reint_unlink sends a new buffer, with the EA included. this
5424 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5425 set if a regular file was being unlinked, and it was the last link
5426 - use PtlGet from the target for bulk writes (315)
5427 - OST now updates object mtime with writes or setattr (607, 619)
5428 - LDLM now has a grant-time callback to revalidate locked items, if
5430 - Many MDS operations were reorganized to combat race conditions
5432 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5433 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5435 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5438 - Fully reactivate OST imports after reconnection (512, others)
5439 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5440 - More graceful error handling for truncating on dead OST (515)
5441 - Don't error out unless we're actually accessing dead stripes (474)
5442 - Fix garbage sizes when stripes are missing (410)
5443 - LRU counters were broken, causing constant lock purge (433, 432)
5444 - garbage on read from stripes with failed OSTs (441)
5445 - mark OSCs as active before reconnecting during recovery (438)
5446 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5447 - lfind did not preserve OST order in output (443)
5448 - symlinks cause hung clients, incorrect data (439)
5449 - stop dereferencing request after dropping refcount (457)
5450 - don't LASSERT(spin_is_locked) on non-SMP (455)
5451 - fixes for many rename() bugs
5452 - fstat didn't correctly synchronize attributes (399)
5453 - server must handle lock cancellation during blocking AST prep (487)
5454 - bulk descriptors were free()d too soon (511)
5455 - fix paths in lconf, which would load incorrect modules (451, 507)
5456 - fix confusing lconf 'host not found' error message (386)
5457 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5458 - fix race condition in mdc_blocking_ast() for inode access (526)
5459 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5460 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5461 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5462 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5463 - fix lstripe in several ways (kernel side) (527)
5464 - fix request leak in ldlm_cli_enqueue (262)
5465 - incorrect OSC was marked inactive after OST failure
5466 - call mds_fs_cleanup before unmounting filesystem (524)
5467 - fix races between taking ns_lock and ldlm_lock_change_resource
5468 - fix races updating LOV export open file list
5469 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5470 - fix recovery NULL deref in ldlm_cli_cancel_unused
5471 - fix some DLM races by using new hash table for lock handles (419)
5472 - permit the client to specify desired inodes, at replay
5473 - duplicate requests when we queue them for replay reintegration
5474 - fix last_rcvd offset calculation
5475 - sync after each recovered transaction, so we always make progress
5476 - never, not always, ERESTART requests without transnos
5477 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5479 - skip replay if the MDS says that the client is already connected
5480 - don't check for a recovery-enabled export to match lctl's UUID
5481 - don't INC_USE_COUNT for phantom exports
5482 - don't crash when cleaning up phantom exports (567)
5483 - don't double-finish or set replay data for errored mdc_open requests
5484 - abort requests when they time out, so we don't get old replies
5485 - send/receive replies for AST messages again
5486 - if the client says that it doesn't have the lock, cancel it on the
5488 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5489 mark it as destroyed, it will all work out when decref is called
5490 - fix module use counts (22, 581)
5492 - ASTs now expect a reply (server cancels lock on error reply)
5494 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5497 - fix many simultaneous client startup (392)
5498 - fix dentry->d_it clobbering
5499 - credentials weren't being shipped for readdir/getattr operations
5500 - remove invalid assertions triggered during some concurrent MD
5502 - proper Lustre versions added (336, 389)
5503 - fix memory leak for create error case (398)
5504 - fix LOV locking bug that would get cli/srv out of sync
5505 - fix echo client over LOV (409)
5506 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5507 - fix double-O_EXCL intent crash (424)
5508 - avoid sending multiple lock CANCELs (352)
5510 - MDS can do multi-client recovery (modulo bugs in new code)
5512 - many updates, edits, cleanups
5514 2002-11-18 Phil Schwan <phil@clusterfs.com>
5517 - fix null d_it dereference (346)
5518 - fix full OST/dbench hang (333)
5519 - fix permission problem with file removal (286)
5520 - fix removal of OSCs from LOV when they fail
5521 - fix NULL deref during bulk timeout (214)
5522 - fix problems related to multiple filesystems on one MDS (241)
5523 - fixed serious subtle metadata locking bugs
5524 - free locks on clients when inodes are removed due to memory
5526 - fix inode pointer in lock data (285)
5527 - partial support for multiple MDS on a single host (241)
5528 - data locks weren't cancelled at clear_inode time (290, 311)
5529 - intent locks could lead to unbounded lock growth (205)
5530 - added a maximum lock count, an LRU list, and a flusher
5531 - fix multiple rename (365)
5532 - properly abstracted the echo client
5533 - OSC locked 1 byte too many; fixed
5534 - rewrote brw callback code:
5535 - fixed recovery bugs related to LOVs (306)
5536 - fixed too-many-pages-in-one-write crash (191)
5537 - fixed (again) crash in sync_io_timeout (214)
5538 - probably fixed callback-related race (385)
5540 - Add capability to MDS protocol
5541 - LDLM cancellations and callbacks on different portals
5543 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5546 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5547 - defer open object destruction to close time (601981, 138)
5548 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5549 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5550 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5551 - don't call obd_setattr in ll_file_release for destroyed objects
5552 * protocol change to lustre_msg: move |version| and add |flags|
5553 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5554 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5555 * added replay of create, unlink, link and rename operations during
5556 MDS failover; recovery should be much more robust now
5557 * remove failed OSCs from LOVs (only lov_create uses this so far)
5558 * the lustre-HOWTO was brought (more) up to date (582544)
5560 2002-10-23 Phil Schwan <phil@clusterfs.com>
5563 - in-use dentries weren't being reused properly (617851)
5564 - prevent multiple LDLM setup (599178)
5565 - fix LOV size calculations for truncate (617853)
5566 - fix client handling of MDS intent errors (POSIX)
5567 - fix permission bug in lovstripe.c test (624321)
5568 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5569 - truncate past end of file could corrupt data
5570 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5571 - a race in recovery could return ETIMEDOUT to apps (623947)
5572 - building outside the source directory was fixed
5573 * the lustre-HOWTO was brought (more) up to date (582544)
5574 * major progress was made on recovery functionality
5576 2002-10-10 Phil Schwan <phil@clusterfs.com>
5579 - recovery deadlock fix
5580 - rm -rf causes LBUG fix (617817)
5581 - file open by multiple tasks fix (618962)
5582 - directory permissions bugs (602707 and 620007)
5583 - journal_stop fixed with locking (611313)
5584 - O_APPEND failures resolved (618273, perhaps 614459)
5585 - lconf PATH fix (619770)
5586 - IA64 build fix (621450)
5587 - RPC buffer sizes scale with amount of memory
5589 2002-10-01 Phil Schwan <phil@clusterfs.com>
5592 - locks would be cancelled without throwing away data pages,
5593 resulting in inconsistent data (605627)
5594 - inode attributes were not always being refreshed (605627, 612449)
5595 - lconf now continues to cleanup after lctl reports an error
5596 - MDS now enforces user permissions (602707)
5597 - lprocfs cleanup fixed, but not yet enabled (614157)
5598 - fixed infinite server hang, should a client not respond to an AST
5599 - avoid going into recovery if user calls readlink() with a buffer
5600 that's too small (613941)
5601 - AST RPCs no longer require replies (614867) -- this may be changed
5602 - don't crash server if client sends an IOV that's too big (611336)
5603 - fixed lock conversion deadlock (611892)
5604 - fixed the following of symlinks (614622)
5605 * recovery: the server can remove locks from a client that dies, other
5606 clients can make progress
5607 * more extN patch fixes
5608 * compile-time configurable ptlrpc buffer allocations
5610 - collaborative read cache document
5611 - Lustre Lite Performance CDR document-in-progress
5613 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5616 - fix typo in patch-2.4.18
5618 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5621 - clear ptlrpc request each time in handle_incoming_request()
5622 - unlink of files now destroys the object on the OST
5624 2002-09-19 Peter Braam <braam@clusterfs.com>
5626 * add hard link support
5627 * change obdfile creation method
5628 * kernel patch changed
5630 2002-09-19 Peter Braam <braam@clusterfs.com>
5633 - stack overflow bug in extN fixed
5635 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5637 * documentation updates
5638 - add man pages for config tools
5639 - update tests/README to describe testing with new config tools
5640 - finish metadata API descriptions
5641 * bug fixes and cleanups
5642 - statfs workaround for 16TB limit
5643 - LOV stripe allocation improved, can stripe on subset of OSTs
5644 - LOV file size/IO offset was wrong for files > 4GB in size
5645 - object EA data was being dropped, caused files to be unreadable
5646 - memory overflow with non-LOV OST caused memory corruption
5647 - fixed regression tests to work with new config tools, obdfilter
5648 - fixed bug when directory size became larger than 1 block
5649 - fixed bug (for single client case) when PWD was deleted
5650 - invalidate local directory pages when doing intent-based ops
5651 - avoid LDLM oops when lock callback contained bad data
5653 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5655 * documentation updates
5656 * bug fixes and cleanups
5657 - configuration tools
5660 - 64-bit compile warnings
5661 - 64-bit internal statfs data
5663 * test_brw on persistent OST devices
5665 * lprocfs (disabled)
5667 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5669 * documentation updates
5670 * bug fixes and cleanups
5671 * configuration tools
5673 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5676 * many small fixes to 0_5_4
5677 * io/network handling
5678 * thinkos in MDS operations
5680 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5683 * crucial basic fixes to 0.5.3
5684 * IOR, Iozone work over Elan
5687 2002-08-07 Phil Schwan <phil@clusterfs.com>
5688 * version 0_5_3, our first alpha
5689 * we use the new Portals iovs
5690 * documentation updates
5691 * bug fixes and cleanups
5692 * small changes in the DLM wire protocol
5694 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5695 * version 0_5_1 with some initial stability,
5696 * locking on MD and file I/O.
5697 * documentation updates
5698 * several bug fixes since 0.5.0
5699 * small changes in wire protocol
5701 2002-07-18 Phil Schwan <phil@clusterfs.com>
5703 * delivered as Lustre Light Alpha
5704 * fixed a crash after handling invalid MDS requests
5705 * fixed directory pages for architectures with non-4k pages sizes
5707 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5708 * release version v0_4_4
5709 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5711 * Fixes a number of interruption problems with OST operations.
5712 * Update documentation for portals header changes
5713 * Move all wire protocol structs/defines to lustre_idl.h
5714 * Fixes symlink length bug.
5715 * Add tcpdump to repository.
5717 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5718 * release version v0_4_3
5719 * Fixes statfs for inodes on extN.
5720 * Fixes bug in runtests which would delete /etc/hosts.
5721 * Use 64-bit object IDs wherever possible (not into VFS though)
5722 Remove ost_get_info, which is unused by lustre, and out of date.
5724 2002-07-03 Peter Braam <braam@clusterfs.com>
5725 * release version v0_4_2 Fixes a lookup error (type not passed)
5726 * move forward to head of Portals
5727 * move forward to latest Lustre kernel
5729 2002-06-25 Peter Braam <braam@clusterfs.com>
5730 * release version v0_4_1. Hopefully stable on single node use.