1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.33 (SLES 10),
5 2.6.18-128.1.1.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.40.11-sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
14 * File join has been disabled in this release, refer to Bugzilla 16929.
16 Severity : enhancement
18 Description: MDS create should not wait for statfs RPC while holding DLM lock.
20 Severity : enhancement
22 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
27 Descriptoin: OST grant too much space to client even there are not enough space.
28 Details : Client will shrink its grant cache to OST if there are no write
29 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
30 this grant cache if there are already not enough avaible space
31 (left_space < total_clients * 32M).
34 Frequency : start MDS on uncleanly shutdowned MDS device
36 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
37 Details : stay in waiting mds<>ost recovery finished produce random bugs
38 due race between two ll_sync thread for one lov target. send
39 ACTIVATE event only if connect realy finished and import have
43 Frequency : rare, connect and disconnect target at same time
45 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
46 Details : don't call obd_disconnect under lov_lock. this long time
47 operation and can block ptlrpcd which answer to connect request.
52 Descriptoin: don't lose wakeup for imp_recovery_waitq
53 Details : recover_import_no_retry or invalidate_import and import_close can
54 both sleep on imp_recovery_waitq, but we was send only one wakeup
58 Frequency : always with long access acl
60 Descriptoin: mds can't pack reply with long acl.
61 Details : mds don't control size of acl but they limited by reint/getattr
64 Severity : enhancement
66 Description: Update to SLES10 kernel-2.6.16.60-0.33.
68 Severity : enhancement
70 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
73 Frequency : start MDS on uncleanly shutdowned MDS device
75 Descriptoin: aborting recovery hang on MDS
76 Details : don't throttle destroy RPCs for the MDT.
79 Frequency : on remount
81 Description: external journal device not working after the remount
82 Details : clear dev_rdonly flag for external journal devices in
88 Description: shutdown vs evict race
89 Details : client_disconnect_export vs connect request race.
90 if client will evicted at this time - we start invalidate
91 thread without referece to import and import can be freed
95 Frequency : rare, need acl's on inode.
97 Description: client can't handle ost additional correctly
98 Details : if ost was added after client connected to mds client can have
99 hit lnet_try_match_md ... to big messages to wide striped files.
100 in this case need teach client to handle config events about add
101 lov target and update client max ea size at that event.
103 Severity : enhancement
105 Description: Changelogs
106 Details : Changelogs are a lightweight mechanism to track filesystem
107 metadata and namespace changes. The changelog is recorded
108 permanently on the MDTs, and is periodically "consumed" / purged
109 when records are no longer needed.
111 Severity : enhancement
113 Description: compact fld format with extents
114 Details : Store range of seq rather than every seq in FLD. Seq
115 controller update FLD rather than clients. In Case of CMD, mdt0
116 has FLD, all other metadata server act as non persistent proxy
117 for FLD queries and cache fld entries in fld cache.
122 Description: don't skip ost target if they assigned to file
123 Details : Drop slow OSCs if we can, but not for requested start idx.
124 This means "if OSC is slow and it is not the requested
125 start OST, then it can be skipped, otherwise skip it only
126 if it is inactive/recovering/out-of-space.
130 Description: more cleanup in mds_lov
131 Details : not send LOV EA under replay, we can't know about they size at this
132 time. Don't allow client connect to mds before any ost connected,
133 for avoid problems with LOV EA size and returning EIO to client.
135 Severity : enhancement
137 Description: Interoperability at server side (Disk interoperability)
139 Severity : enhancement
141 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
143 Severity : enhancement
145 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
147 Severity : enhancement
149 Description: New client IO stack (CLIO).
151 Severity : enhancement
153 Description: Commit on sharing. Eliminate inter-client dependencies between
154 uncommitted transactions by doing transaction commits.
155 Thereby clients may recovery independently.
158 Frequency : Create a symlink file with a very long name
160 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
161 Details : If there is no extra space in the request for early cancels,
162 ldlm_req_handles_avail() returns 0 instead of a negative value.
164 Severity : enhancement
166 Description: Add /proc entry for import status
167 Details : The mdc, osc, and mgc import directories now have
168 an import directory that contains useful import data for debugging
171 Severity : enhancement
173 Description: Re-disable certain /proc logging
174 Details : Enable and disable client's offset_stats, extents_stats and
175 extents_stats_per_process stats logging on the fly.
178 Frequency : Only on FC kernels 2.6.22+
180 Description: oops in statahead
181 Details : Do not drop reference count for the dentry from VFS when lookup,
182 VFS will do that by itself.
184 Severity : enhancement
186 Description: Generic /proc file permissions
187 Details : Set /Proc file permissions in a more generic way to enable non-
188 root users operate on some /proc files.
192 Description: Hitting mdc_commit_close() ASSERTION
193 Details : Properly handle request reference release in
194 ll_release_openhandle().
198 Description: quota recovery deadlock during mds failover
199 Details : This patch includes att18982, att18236, att18237 in bz14840.
201 1. fix osts hang when mds does failover with quotaon
202 2. prevent watchdog storm when osts threads wait for the
207 Frequency : only patchless client
208 Description: add workaround for race between add/remove dentry from hash
210 Severity : enhancement
212 Description: Allow OST glimpses to return PW locks
216 Description: LBUG when llog conf file is full
217 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
221 Description: Prevent import from entering FULL state when server in recovery
225 Description: service mount cannot take device name with ":"
226 Details : Only when device name contains ":/" will mount treat it as
232 Description: replace ptlrpcd with the statahead thread to interpret the async
233 statahead RPC callback
237 Frequency : on recovery
238 Description: I/O failures after umount during fail back
239 Details : if client reconnected to restarted server we need join to recovery
240 instead of find server handler is changed and process self
241 eviction with cancel all locks.
243 Severity : enhancement
245 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
247 Severity : enhancement
249 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
251 Severity : enhancement
253 Description: Upcall on Lustre log has been dumped
254 Details : Allow for a user mode script to be called once a Lustre log has
255 been dumped. It passes the filename of the dumped log to the
256 script, the location of the script can be specified via
257 /proc/sys/lnet/debug_log_upcall.
262 Description: avoid idr_remove called for id which is not allocated.
263 Details : Move assigment s_dev for clustered nfs to end of initialization,
264 to avoid problem with error handling.
269 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
270 Details : When connection is reused this not moved from CONN_UNUSED_HASH
271 into CONN_USED_HASH and this prodice warning when put connection
272 again in unused hash.
275 Severity : enhancement
277 Description: File striping can now be set to use an arbitrary pool of OSTs.
279 Severity : enhancement
281 Description: Export bytes_read/bytes_write count on OSC/OST.
285 Description: Early reply size mismatch, MGC loses connection
286 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
287 the connect flags are properly negotiated.
292 Description: quotas are not honored with O_DIRECT
293 Details : all writes with the flag O_DIRECT will use grants which leads to
294 this problem. Now using OBD_BRW_SYNC to guard this.
298 Description: add quota statistics
299 Details : 1. sort out quota proc entries and proc code.
300 2. add quota statistics
302 Severity : enhancement
304 Description: enable quota support for HEAD.
308 Description: Properly propagate oinfo flags from lov to osc for statfs
309 Details : restore missing copy oi_flags to lov requests.
311 Severity : enhancement
313 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
314 : l_getgroups(8), lst(8), routerstat(8)
315 Details : included man pages for llobdstat(8), llstat(8),
316 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
318 Severity : enhancement
320 Description: configure's --enable-quota should check the
321 : kernel .config for CONFIG_QUOTA
322 Details : configure is terminated if --enable-quota is passed but
323 : no quota support is in kernel
327 Description: Remove portals compatibility
328 Details : Remove portals compatibility, not interoperable with releases
333 Description: Resolve device initialization race
334 Details : Prevent proc handler from accessing devices added to the
335 obd_devs array but yet be intialized.
337 Severity : enhancement
339 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
341 Severity : enhancement
343 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
348 Description: Using wrong pointer in osc_brw_prep_request
349 Details : Access to array[-1] can produce panic if kernel compiled with
350 CONFIG_PAGE_ALLOC enabled
354 Description: Client runs out of low memory
355 Details : Consider only lowmem when counting initial number of llap pages
359 Description: *optional* service tags registration
360 Details : if the "service tags" package is installed on a Lustre node
361 When the filesystem is mounted, a local-node service tag will
362 be created. See http://inventory.sun.com/ for more information
363 about the Service Tags asset management system.
367 Description: Kernel BUG tries to release flock
368 Details : Lustre does not destroy flock lock before last reference goes
369 away. So always drop flock locks when client is evicted and
370 perform unlock regardless of successfulness of speaking to MDS.
374 Description: add refcount protection for osc callbacks, avoid panic on shutdown
378 Description: sanity test 65a fails if stripecount of -1 is set
379 Details : handle -1 striping on filesystem in ll_dirstripe_verify
384 Description: ASSERTION(CheckWriteback(page,cmd)) failed
385 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
388 Severity : enhancement
390 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
394 Description: do not process already freed flock
395 Details : flock can possibly be freed by another thread before it reaches
396 to ldlm_flock_completion_ast.
400 Description: LBUG during stress test
401 Details : Need properly lock accesses the flock deadlock detection list.
405 Description: oops in page fault handler
406 Details : kernel page fault handler can return two special 'pages' in error
407 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
411 Description: timeout with invalidate import.
412 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
413 should be handled by ptlrpcd. This produce long age waiting and
414 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
416 Severity : enhancement
418 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
422 Description: enable MGS and MDT services start separately
423 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
424 a co-located MGS without starting the MGS, which is a complement
425 to 'nosvc' mount option.
429 Frequency : after recovery
430 Description: precreate to many object's after del orphan.
431 Details : del orphan st in oscc last_id == next_id and this triger growing
432 count of precreated objects. Set flag LOW to skip increase count
433 of precreated objects.
437 Frequency : rare, on clear nid stats
438 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
439 Details : when clean nid stats sometimes try destroy live entry,
440 and this produce panic in free.
444 Description: Stack overflow during MDS log replay
445 ease stack pressure by using a thread dealing llog_process.
449 Description: wait until IO finished before start new when do lock cancel.
450 Details : VM protocol want old IO finished before start new, in this case
451 need wait until PG_writeback is cleared until check dirty flag and
452 call writepages in lock cancel callback.
454 Severity : enhancement
456 Description: using special macro for print time and cleanup in includes.
460 Description: mds_mfd_close() ASSERTION(rc == 0)
461 Details : In mds_mfd_close(), we need protect inode's writecount change
462 within its orphan write semaphore to prevent possible races.
466 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
470 Frequency : rare, on shutdown ost
471 Description: don't hit live lock with umount ost.
472 Details : shrink_dcache_parent can be in long loop with destroy dentries,
473 use shrink_dcache_sb instead.
477 Description: don't panic with use echo client
478 Details : echo client pass NULL as client nid pointer and this produce null
483 Description: fix build on ppc32
484 Details : compile code with -m64 flag produce wrong object file for ppc32.
488 Description: add message levels for liblustreapi
492 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
493 Details : Move the 'good_osts' check before the 'total_bavail' check. This
494 will result in an -EAGAIN and in the exit call path we call
495 alloc_rr() which will with increasing aggressiveness attempt to
496 aquire precreated objects on the minimum number of required OSCs.
500 Description: Use old size assignment to avoid deadlock
501 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
502 the scheduling while holding a spinlock. We do not need locking
503 for size in ll_update_inode() because size is only updated from
504 the MDS for directories or files without objects, so there is no
505 other place to do the update, and concurrent access to such inodes
506 are protected by the inode lock.
510 Description: resolve "_IOWR redefined" build error on SLES10
514 Description: dump the memory debugging after all modules are unloaded to
515 suppress false negative in conf_sanity test 39
517 Severity : enhancement
519 Description: build kernel-ib packages for OFED 1.3 in our release cycle
524 Description: fix SLES kernel versioning
525 Details : the kernel version for our SLES 10 kernel did not include a "-"
526 before the "smp" at the end. while this was not a problem in
527 general, it did mean that software trying to use the kernel
528 version to try to detect a vendor specific kernel would fail.
529 this was most evident by the OFED build scripts.
533 Description: Don't update lov_desc members until making sure they are valid
534 Details : When updating lov_desc members via proc fs, need fix their
535 validities before doing the real update.
539 Description: don't put request into delay list while invalidate in flight.
540 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
541 import in flight. this produce timeout for invalidate and sometimes
542 can cause stale data.
546 Frequency : on ppc only
547 Description: not convert ost objects for directory because it's not exist.
548 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
552 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
554 Severity : enhancement
556 Description: Update to support 2.6.22.14 vanilla kernel.
560 Frequency : rare, on recovery
561 Description: read procfs can produce deadlock in some situation
562 Details : Holding lprocfs lock which send rpc can produce block for destroy
563 obd objects and this also block reconnect with -EALREADY. This
564 isn't fix all lprocfs bugs - but make it rare.
566 Severity : enhancement
568 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
571 Frequency : frequent on X2 node
573 Description: mdc_set_open_replay_data LBUG
574 Details : Set replay data for requests that are eligible for replay.
578 Description: lustre_mgs: operation 101 on unconnected MGS
579 Details : When MGC is disconnected from MGS long enough, MGS will evict the
580 MGC, and late on MGC cannot successfully connect to MGS and a lot
581 of the error messages complaining that MGS is not connected.
585 Frequency : on network error
586 Description: panic with double free request if network error
587 Details : mdc_finish_enqueue is finish request if any network error ocuring,
588 but it's true only for synchronus enqueue, for async enqueue
589 (via ptlrpcd) this incorrect and ptlrpcd want finish request
592 Severity : enhancement
594 Description: client-side metadata stat-ahead during readdir(directory readahead)
595 Details : perform client-side metadata stat-ahead when the client detects
596 readdir and sequential stat of dir entries therein
599 Frequency : on start mds
601 Description: Implement get_info(last_id) in obdfilter.
604 Frequency : occasional
606 Description: Correctly check stale fid, not start epoch if ost not support SOM
607 Details : open with flag O_CREATE need set old fid in op_fid3 because
608 op_fid2 was overwritten with new generated fid, but mds can answer
609 with one of these two fids and both is not stale. Setattr
610 incorrectly started an epoch and assume will be called
611 done_writing, but without SOM done_writing ever being called.
614 Frequency : rare, depends on device drivers and load
616 Description: MDS or OSS nodes crash due to stack overflow
617 Details : Code changes in 1.8.0 increased the stack usage of some functions.
618 In some cases, in conjunction with device drivers that use a lot
619 of stack the MDS (or possibly OSS) service threads could overflow
620 the stack. One change which was identified to consume additional
621 stack has been reworked to avoid the extra stack usage.
624 Frequency : occasional
626 Description: Do not fail import if osc_interpret_create gets -EAGAIN
627 Details : If osc_interpret_create got -EAGAIN it immediately exits and
628 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
629 oscc_has_objects and see OSC has no objests and call
630 oscc_internal_create to resend create request.
632 Severity : enhancement
634 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
636 Severity : enhancement
638 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
641 Frequency : very rare
643 Description: Fix replay if there is an un-replied request and open
644 Details : In some cases, older replay request will revert the
645 mcd->mcd_last_xid on MDS which is used to record the client's
648 Severity : enhancement
650 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
652 Severity : enhancement
654 Description: Add rhel5 support to HEAD.
656 Serverity : enhancement
658 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
663 Description: Don't allow skipping OSTs if index has been specified.
664 Details : Don't allow skipping OSTs if index has been specified, make
665 locking in internal create lots better.
669 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
670 Details : access to bitfield in structure is always rounded to long
671 and this produce problem with not atomic change any bit.
675 Description: Lustre make rpms failed.
676 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
677 builds Lustre rpms from tarball.
680 Frequency : rare on shutdown ost
682 Description: If llog cancel was not send before clean_exports phase, this can
683 produce deadlock in llog code.
684 Details : If llog thread has last reference to obd and call class_import_put
685 this produce deadlock because llog_cleanup_commit_master wait when
686 last llog_commit_thread exited, but this never success because was
687 called from llog_commit_thread.
691 Description: allow userland application know is lost one of stripes.
692 Details : fill lvb_blocks with error code on ost and return it to
693 application if error flag found.
697 Description: NULL lov_tgts causing MDS oops
698 Details : more safe checks for NULL lov_tgts for avoid oops.
700 Severity : enhancement
702 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
706 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
707 Details : If an OST is down the MDS will hang indefinitely in
708 obd_statfs_rqset() waiting for the statfs data. While for
709 MDS QOS usage of statfs, it should not stuck in waiting.
711 Severity : enhancement
713 Description: remote_acl support
714 Details : Support ACL-based permission check for remote user.
715 Support setfacl/getfacl for remote user with the utils
716 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
717 the system "{s,g}etfacl" utils.
719 Severity : enhancement
721 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
723 Severity : enhancement
725 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
729 Description: make mgs_setparam() handle fsname containing dash
730 Details : fsname containing a dash does not work with lctl conf_param
732 Severity : enhancement
734 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
736 Severity : enhancement
738 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
742 Frequency : rare, at shutdown
743 Description: access already free / zero obd_namespace.
744 Details : if client_disconnect_export was called without force flag set,
745 and exist connect request in flight, this can produce access to
746 NULL pointer (or already free pointer) when connect_interpret
747 store ocd flags in obd_namespace.
751 Frequency : only at startup
752 Description: not alloc memory with spinlock held.
753 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
754 if any spinlock held.
756 Severity : enhancement
758 Description: make lustre randomly fail allocating memory
759 Details : Make lustre randomly failed allocating memory for testing purpose.
761 Severity : enhancement
763 Description: lost problems with lov objid file
764 Details : Fixes some scability and access to not inited memory problems
765 in work with lov objdid file.
770 Description: lfs find does not continue on file error
771 Details : Continue other files processing when a file/dir is absent.
775 Description: Inconsistent usage of lustre_pack_reply()
776 Details : Standardize the usage of lustre_pack_reply() such that it
777 always generate a CERROR on failure.
780 Frequency : occasional
782 Description: MDS hang or stay in waiting lock
783 Details : If client receive lock with CBPENDING flag ldlm need send lock
784 cancel as separate rpc, to avoid situation when cancel request
785 can't processed due all i/o threads stay in wait lock.
789 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
793 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
797 Description: Update to SLES9 kernel-2.6.5-7.287.3.
801 Description: Update to RHEL4 latest kernel.
803 Severity : enhancement
805 Description: Build SLES10 patchless client fails
806 Details : The configure was broken by run ./configure with
807 --with-linux-obj=.... argument for patchless client. When the
808 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
809 can't search header adequately. Use absolute path such as
810 -I($LINUX)/include instead.
814 Description: interrupt oig_wait produce painc on resend.
815 Details : brw_redo_request can be used for resend requests from ptlrpcd and
816 private set, and this produce situation when rq_ptlrpcd_data not
817 copyed to new allocated request and triggered LBUG on assert
818 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
819 ptlrpcd set if request is changed and can be safety changed to use
824 Description: LASSERT_{REQ,REP}SWAB macros are buggy
825 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
826 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
827 these macros and replace them with their difinitions instead.
831 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
832 Details : Update which_patch & target file for SLES10 latest kernel.
835 Description: Remove client patches from SLES 10 kernel.
836 Details : This causes SLES 10 clients to behave as patchless clients
837 even on a Lustre-patched (server) kernel.
839 Severity : enhancement
841 Description: self-adjustable client's lru lists
842 Details : use adaptive algorithm for managing client cached locks lru
843 lists according to current server load, other client's work
844 pattern, memory activities, etc. Both, server and client
845 side namespaces provide number of proc tunables for controlling
848 Severity : enhancement
850 Description: light-weight GSS support
851 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
852 and reduce performance overhead.
854 Severity : enhancement
856 Description: Linux keyring support
857 Details : Support using service of Linux keyring for Lustre GSS internal
858 context refresh/cache mechanism.
862 Description: Fix errors in lfs documentation
863 Details : Fixes man pages
867 Description: don't use GFP_* in generic Lustre code.
868 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
872 Description: obdclass is limited by single OBD_ALLOC(idarray)
873 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
877 Description: warnings with build patchless client with vanila 2.6.19 and up
878 Details : change old ctl_table style and replace ctl_table/ctl_table_header
879 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
883 Frequency : Only for SLES
884 Description: sanity_quota fail test_1
885 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
886 file, which makes TSTID[2] inunique.
890 Frequency : Only for SLES9
891 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
892 Details : Update target/ChangeLog/which_patch .
896 Frequency : Only for test_10 in sanity-quota.sh
897 Description: a bug in quota test code
898 Details : checking a test flag in a wrong place
902 Frequency : Only for test_13 in sanity-quota.sh
903 Description: a bug in quota test script
904 Details : shouldn't check overfull allocated quota
908 Description: jbd statistics
909 Details : Port older jbd statistics patch for sles10
911 Severity : enhancement
913 Description: Add printing inode info into message about error in writepage.
918 Description: reply_lock_interpret crash due to race with it and lock cancel.
919 Details : Do not replay locks that are being cancelled. Do not reference
920 locks by their address during replay, just by their handle.
924 Frequency : When flocks are used.
925 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
926 Details : Flock locks might destroy just granted lock if it could be merged
927 with another existing flock, this is done in completion handler,
928 so teach ldlm_cli_enquque_fini that this is a valid case for
934 Description: Oops in read and write path when failing to allocate lock.
935 Details : Check if lock allocation failed and return error back.
939 Description: lstripe command fails for valid OST index
940 Details : The stripe offset is compared to lov->desc.ld_tgt_count
941 instead of lov->desc.ld_active_tgt_count.
945 Description: sanity.sh failed test 103
946 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
951 Description: df doesn't work properly if diskfs blocksize != 4K
952 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
956 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
957 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
959 Severity : enhancement
961 Description: Adaptive timeouts
962 Details : RPC timeouts adapt to changing server load and network
963 conditions to reduce resend attempts and improve recovery time.
967 Description: llapi_file_create() does not allow some changes
968 Details : add llapi_file_open() that allows specifying the mode and
969 open flags, and also returns an open file handle.
973 Description: merge and cleanup kernel patches.
974 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
978 Description: Add journal checksum support.(Kernel part)
979 Details : The journal checksum feature adds two new flags i.e
980 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
981 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
982 indicates that the commit block contains the checksum for
983 the blocks described by the descriptor blocks. Now commit
984 record can be sent to disk without waiting for descriptor
985 blocks to be written to disk. This behavior is controlled
986 using JBD2_FEATURE_ASYNC_COMMIT flag.
990 Description: OSS needs mutliple precreate threads
991 Details : Add ability to start more than one create thread per OSS.
995 Description: Sanity.sh test_65k failed: setstripe should have succeeded
996 Details : execute lfs setstripe on client
1000 Description: mds_obd_create error creating tmp object
1001 Details : When the user sets quota on root, llog will be affected and can't
1002 create files and write files.
1005 Frequency : Always on ia64 patchless client, and possibly others.
1007 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1008 Details : This allows the patchless client to be loaded on architectures
1009 without this export.
1013 Description: RedHat Update kernel for RHEL5
1014 Details : Add the kernel config file for RHEL5.
1018 Description: RedHat Update kernel for RHEL5
1019 Details : Modify the kernel config file more closer RHEL5.
1023 Description: Build failure against Centos5 (RHEL5)
1024 Details : Define PAGE_SIZE when it isn't present.
1028 Description: test_7 on sanity-quota.sh can't be used on separate machines
1029 Details : add facet to handle it
1033 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1034 Details : with NFS, the anon dentry's parent was set to itself in
1035 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1036 corresponding dentry other than use rec->ur_name.
1038 Severity : enhancement
1040 Description: lfs setstripe enhancement
1041 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1045 Description: enable data checksumming by default
1046 Details : enable checksum by default, allow --disable-checksum
1047 configure option and "-o nochecksum" mount option. Checksums
1048 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1052 Description: lustre support for RHEL5
1053 Details : Add support for RHEL5.
1057 Description: Client eviction due to failover config
1058 Details : after a connection loss, the lustre client should attempt to
1059 reconnect to the last active server first before trying the
1060 other potential connections.
1064 Description: when mds and osts use different quota unit(32bit and 64bit),
1065 quota will be released repeatly.
1066 Details : void sending multiple quota reqs to mds, which will keep the
1067 status between the reqs.
1071 Description: osts not allocated evenly to files
1072 Details : change the condition to increase offset_idx
1075 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1077 Description: Data corruption for OSTs that are formatted larger than 2TB
1079 Details : When generating the bio request for lustre file writes the
1080 sector number would overflow a temporary variable before being
1081 used for the IO. The data reads correctly from Lustre (which
1082 will overflow in a similar manner) but other file data or
1083 filesystem metadata may be corrupted in some cases.
1087 Description: Tune the kernel for good SCSI performance.
1088 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1089 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1094 Description: rewrite ext2-derived code in obdclass/uuid.c
1095 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1096 from scratch preserving functionality.
1100 Description: "lfs find -obd UUID" prints directories
1101 Details : "lfs find -obd UUID" will return all directory names instead
1102 of just file names. It is incorrect because the directories
1103 do not reside on the OSTs.
1107 Description: use i_size_read and i_size_write in 2.6 port
1108 Details : replace inode->i_size access with i_size_read/write()
1111 Frequency : when removing large files
1113 Description: scheduling issue during removal of large Lustre files
1114 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1115 It causes scheduling issues when removing large files (17TB in the
1119 Frequency : only with liblustre clients on XT3
1121 Description: evictions taking too long
1122 Details : allow llrd to evict clients directly on OSTs
1125 Frequency : only on ppc
1127 Description: /proc/fs/lustre/devices broken on ppc
1128 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1129 We should make sure the type of 'index' is loff_t and then cast
1130 explicitly as needed below. Do not assign an explicitly cast
1134 Frequency : only for rhel5
1136 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1137 Details : Modify the target file & which_kernel.
1139 Severity : enhancement
1141 Description: omit set fsid for export NFS
1142 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1143 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1147 Frequency : Always, for kernels after 2.6.16
1148 Description: Fix warning idr_remove called for id=.. which is not allocated.
1149 Details : Last kernels save old s_dev before kill super and not allow
1150 to restore from callback - restore it before call kill_anon_super.
1154 Description: buffer overruns could theoretically occur
1155 Details : llapi_semantic_traverse() modifies the "path" argument by
1156 appending values to the end of the origin string, and a
1157 overrun may occur. Adding buffer overrun check in liblustreapi.
1161 Description: Fix error on 'ls .' at the top of the Lustre mount.
1162 Details : Don't revalidate dentry if it is a root dentry.
1166 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1167 Details : Modify vm-tunables-rhel4.patch.
1171 Description: Kernel config for 2.6.18-vanilla.
1172 Details : Modify targets/2.6-vanilla.target.in.
1173 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1174 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1175 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1176 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1180 Description: improve handling recoverable errors
1181 Details : if request processig with error which can be recoverable on server
1182 request should be resend, otherwise page released from cache and
1187 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1188 Details : Modify target file & which_patch.
1189 A flaw was found in the IA32 system call emulation provided
1190 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1191 value could be stored in the %RAX register, which could trigger an
1192 out-of-bounds system call table access. An untrusted local user
1193 could exploit this flaw to run code in the kernel
1194 (ie a root privilege escalation). (CVE-2007-4573).
1198 Description: O_DIRECT bypasses client statistics.
1199 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1200 read_ahead_stats were updated. copied the stats section from
1201 osc_send_oap_rpc() into async_internal().
1205 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1209 Description: change order of libsysio includes
1210 Details : '#include sysio.h' should always come before '#include xtio.h'
1214 Description: handle "serious error: objid * already exists" more gracefully
1215 Details : If LAST_ID value on disk is smaller than the objects existing in
1216 the O/0/d* directories, it indicates disk corruption and causes an
1217 LBUG(). If the object is 0-length, then we should use the existing
1218 object. This will help to avoid a full fsck in most cases.
1220 Severity : enhancement
1222 Description: adapt the lustre_config script to support the upgrade case
1223 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1224 server targets to 1.6 in parallel.
1228 Description: To avoid grant space > avaible space when the disk is almost
1229 full. Without this patch you might see the error "grant XXXX >
1230 available" or some LBUG about grant, when the disk is almost
1232 Details : In filter_check_grant, for non_grant cache write, we should
1233 check the left space by if (*left > ungranted + bytes), instead
1234 of (*left > ungranted), because only we are sure the left space
1235 is enough for another "bytes", then the ungrant space should be
1236 increase. In client, we should update cl_avail_grant only there
1237 is OBD_MD_FLGRANT in the reply.
1241 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1242 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1243 described in CVE-2007-4573. This problem could allow untrusted
1244 local users to gain root access.
1247 Frequency : when using O_DIRECT and quotas
1249 Description: Incorrect file ownership on O_DIRECT output files
1250 Details : block usage reported by 'lfs quota' does not take into account
1251 files that have been written with O_DIRECT.
1256 Description: touch file failed when fs is not full
1257 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1258 otherwise we can get ENOSP while fs is not full.
1262 Description: parallel lock callbacks
1263 Details : Instead of sending blocking and completion callbacks as separated
1264 requests, adding them to a set and sending in parallel.
1267 Frequency : only for Cray XT3
1268 Bugzilla : 12829/13455
1269 Description: Changing primary group doesn't change the group lustre assigns to
1271 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1272 overridden with the primary group provided by the group upcall,
1273 whereas the supplied fsgid can be trusted if it is in the list of
1274 supplementary groups returned by the group upcall.
1276 Severity : enhancement
1278 Description: Allow masking D_WARNING, D_ERROR messages from console
1279 Details : Console messages can now be disabled via lnet.printk.
1283 Description: User code with malformed file open parameter crashes client node
1284 Details : Before packing join_file req, all the related reference should be
1285 checked carefully in case some malformed flags cause fake
1286 join_file req on client.
1290 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1291 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1292 mode is equal to the granted mode. However, it should also check
1293 that LDLM_FL_CP_REQD is not set.
1297 Description: Heavy nfs access might result in deadlocks
1298 Details : After ELC code landed, it is now improper to enqueue any mds
1299 locks under och_sem, because enqueue might want to decide to
1300 cancel open locks for same inode we are holding och_sem for.
1304 Description: Client eviction while running blogbench
1305 Details : A lot of unlink operations with concurrent I/O can lead to a
1306 deadlock causing evictions. To address the problem, the number of
1307 oustanding OST_DESTROY requests is now throttled to
1308 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1309 callbacks are processed in priority.
1313 Description: enable ACLs on MDS by default
1314 Details : ACLs must be enabled on MDS by default.
1317 Frequency : PPC/PPC64 only
1319 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1320 Details : fix duplicated definitions between asm-ppc64/types.h and
1321 lustre_types.h on PPC.
1324 Frequency : PPC/PPC64 only
1326 Description: asm-ppc/segment.h does not exist
1327 Details : fix compile issue on PPC.
1331 Description: better handle error messages in extents code
1334 Frequency : RHEL4 only
1336 Description: mkfs is very slow on IA64/RHEL4
1337 Details : A performance regression has been discovered in the MPT Fusion
1338 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1339 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1340 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1342 Severity : enhancement
1344 Description: SNMP support enhancement
1345 Details : Adding total number of sampled request for an MDS node in snmp
1348 Severity : enhancement
1350 Description: Optimize ldlm waiting list processing for PR extent locks
1351 Details : When processing waiting list for read extent lock and meeting read
1352 lock that is same or wider to it that is not contended, skip
1353 processing rest of the list and immediatelly return current
1354 status of conflictness, since we are guaranteed there are no
1355 conflicting locks in the rest of the list.
1359 Description: Time out and refuse to reconnect
1360 Details : When the failover node is the primary node, it is possible
1361 to have two identical connections in imp_conn_list. We must
1362 compare not conn's pointers but NIDs, otherwise we can defeat
1363 connection throttling.
1367 Description: port llog fixes from b1_6 into HEAD
1368 Details : Port llog reference couting and some llog cleanups from b1_6
1369 (bug 10800) into HEAD, for protect from panic and access to already
1370 free llog structures.
1374 Description: Detect stride IO mode in read-ahead
1375 Details : When a client does stride read, read-ahead should detect that and
1376 read-ahead pages according to the detected stride pattern.
1380 Description: data checksumming impacts single node performance
1381 Details : add support for several checksum algorithm. Currently, only CRC32
1382 and Adler-32 are supported. The checksum type can be changed on
1383 the fly via /proc/fs/lustre/osc/*/checksum_type.
1387 Description: use adler32 for page checksums
1388 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1393 Description: build for x2 fails
1394 Details : fix compile issue on Cray systems.
1398 Description: Properly match for duplicate locks
1399 Details : Due to different lock order from skiplists code, we need to
1400 traverse entire list for now
1403 Frequency : only on PPC/SLES10
1405 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1406 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1407 be defined. Add a hack in mkfs_lustre.c to work around this compile
1412 Description: LASSERT on MDS when client holding flock lock dies
1413 Details : ldlm pool logic depends on number of granted locks equal to
1414 number of released locks which is not true for flock locks, so
1415 just exclude such locks from consideration.
1419 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1420 Details : Use fsfilt_sync() for both the whole filesystem sync and
1421 individual file sync to eliminate dangerous inode locking
1422 with I_LOCK that can lead to a deadlock.
1426 Description: performance in 1.6.3
1427 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1428 sure that 1MB requests are not fragmented by the block layer.
1430 Severity : enhancement
1432 Description: organize the server-side client stats on per-nid basis
1433 Details : Change the structure of stats under obdfilter and mds to
1441 The "uuid"s file would list the uuids of _active_ exports.
1442 And the clear entry is to clear all stats and stale nids.
1444 Severity : enhancement
1446 Description: eliminate client locks in face of contention
1447 Details : file contention detection and lockless i/o implementation
1448 for contended files.
1452 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1455 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1457 Description: MDS may lose file striping (and hence file data) in some cases
1458 Details : If there are additional extended attributes stored on the MDS,
1459 in particular ACLs, SELinux, or user attributes (if user_xattr
1460 is specified for the client mount options) then there is a risk
1461 of attribute loss. Additionally, the Lustre file striping
1462 needs to be larger than default (e.g. striped over all OSTs),
1463 and an additional attribute must be stored initially in the
1464 inode and then increase in size enough to be moved to the
1465 external attribute block (e.g. ACL growing in size) for file
1470 Description: skiplist implementation simplification
1471 Details : skiplists are used to group compatible locks on granted list
1472 that was implemented as tracking first and last lock of each
1473 lock group the patch changes that to using doubly linked lists
1477 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1478 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1479 because LASSERT on both the data supplied by a client, and the
1480 data on disk is dangerous and incorrect.
1482 Severity : enhancement
1484 Description: Slow truncate/writes to huge files at high offsets.
1485 Details : Directly associate cached pages to lock that protect those pages,
1486 this allows us to quickly find what pages to write and remove
1487 once lock callback is received.
1491 Description: more ldlm soft lockups
1492 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1493 starve other threads from the resource lock for a long time in
1494 case of long waiting queue, so change the debug level from
1495 D_OTHER to the less frequently used D_INFO.
1497 Severity : enhancement
1499 Description: add -gid, -group, -uid, -user options to lfs find
1503 Description: Hung threads in invalidate_inode_pages2_range
1504 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1505 once one is completed. As a result, some RPCs are stuck in the
1506 queue and are never sent.
1510 Description: filter threads hungs on waiting journal commit
1511 Details : Cleanup filter group llog code, then only filter group llog will
1512 be only created in the MDS/OST syncing process.
1516 Description: Procfs and llog threads access destoryed import sometimes.
1517 Details : Sync the import destoryed process with procfs and llog threads by
1518 the import refcount and semaphore.
1520 Severity : enhancement
1522 Description: openlock cache of b1_6 port to HEAD
1527 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1528 Details : If insertion of an extent fails, then discard the inode
1529 preallocation and free data blocks else it can lead to duplicate
1534 Description: don't always update ctime in ext3_xattr_set_handle()
1535 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1536 In some cases the ctime should not be updated, for example for
1537 2.0->1.8 compatibility it is necessary to delete an xattr and it
1538 should not update the ctime.
1542 Bugzilla : 15713/16362
1543 Description: Assertion in iopen_connect_dentry in 1.6.3
1544 Details : looking up an inode via iopen with the wrong generation number can
1545 populate the dcache with a disconneced dentry while the inode
1546 number is in the process of being reallocated. This causes an
1547 assertion failure in iopen since the inode's dentry list contains
1548 both a connected and disconnected dentry.
1552 Description: assertion failure in ldlm_handle2lock()
1553 Details : fix a race between class_handle_unhash() and class_handle2object()
1554 introduced in lustre 1.6.5 by bug 13622.
1559 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1560 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1565 Description: oops in sd_iostats_seq_show()
1566 Details : unloading/reloading the scsi low level driver triggers a kernel
1567 bug when trying to access the sd iostat file.
1572 Description: Kernel panics during QLogic driver reload
1573 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1574 patch, causing memory corruption.
1579 Description: journal_dev option does not work in b1_6
1580 Details : pass mount option during pre-mount.
1582 Severity : enhancement
1584 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1585 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1586 extent information of a file. It can be used to map logical blocks
1587 in a file to physical blocks in the block device.
1591 Description: LDLM soft lockups - improvement
1592 Details : It is be possible to send the lock handle along with each read
1593 or write request because the client is already doing a lock match
1594 itself so there isn't any reason the OST should have to re-do that
1598 Frequency : only on Cray X2
1600 Description: X2 build failures
1601 Details : fix build failures on Cray X2.
1605 Description: xid & resent requests
1606 Details : Initialize RPC XID from clock at startup (randomly if clock is
1609 Severity : enhancement
1611 Description: Add lustre_start utility to start or stop multiple Lustre servers
1616 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1617 Details : In case of memory pressure, list_del() can be called twice on
1618 req->rq_history_list, causing a kernel oops.
1622 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1623 Details : fix stack overflow in the distributed lock manager by defering
1624 export eviction after a failed AST to the elt thread instead of
1625 handling it in the dlm interpret routine.
1629 Description: Convert some comments to new format.
1630 Details : Update documenting comments to match doxygen conventions.
1634 Description: Grammar fixes.
1635 Details : A couple of trivial spelling fixes.
1639 Description: OSD_COUNTERS-mandatory
1640 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1641 to simplify the code.
1645 Description: simplify cmm_device freeing logic.
1646 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1650 Description: Add lockdep support to dt_object_operations locking interface.
1651 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1652 parameter indicating lock ordering. Update mdd code to use new
1657 Description: Introduce failloc constants for lockless IO tests.
1658 Details : Add two new failloc constants to test lockless IO. Only one of
1659 them in implemented---another is checked in yet to be landed
1664 Description: Add lockdep support for inode mutex.
1665 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1666 in the situations where Lustre has to lock more than one inode
1671 Description: Add optional invariants checking support.
1672 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1673 configure switch. This macro is to be used for consistency and
1674 sanity checks that are too expensive to be left in `production'
1679 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1680 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1685 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1686 Details : Use spin_lock_nested() in (the only) situation where more than
1687 one ldlm_lock is locked simultaneously. Also, fix possible
1688 dead-lock in ldlm_lock_change_resource() by enforcing particular
1693 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1694 Details : Instead of specifying each ldlm_lock call-back through separate
1695 parameter, wrap them into struct ldlm_callback_suite.
1699 Description: Kill join_lru obd method and its callers.
1700 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1701 in memory---a requirement implemented through ->o_join_lru() obd
1702 method in HEAD. Remove this method and its users.
1706 Description: Add asynchronous ldlm ENQUEUE completion handler.
1707 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1708 case that stock ldlm_completion_ast() cannot handle as it waits
1709 until lock is granted. Introduce new ldlm_completion_ast_async()
1710 for this. Also comment ldlm_completion_ast().
1714 Description: ldlm_error <-> errno conversion.
1715 Details : Add functions to map (rather arbitrary) between LDLM error codes
1716 and standard errno values. CLIO needs this to prevent LDLM specific
1717 constants from escaping ldlm and osc.
1721 Description: Kill unused ldlm_handle2lock_ns() function.
1722 Details : Kill unused ldlm_handle2lock_ns() function.
1726 Description: Add lu_ref support to ldlm_lock
1727 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1728 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1729 and ldlm_resource. LDLM interface has to be changed, because code that
1730 releases a reference on a lock, has to "know" what reference this is.
1731 In the most frequent case
1733 lock = ldlm_handle2lock(handle);
1735 LDLM_LOCK_PUT(lock);
1737 no changes are required. When any other reference (received _not_ from
1738 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1739 instead of LDLM_LOCK_PUT().
1741 Arguably, changes are pervasive, and interface requires some discipline
1742 for proper use. On the other hand, it was very instrumental in finding
1743 a few leaked lock references.
1747 Description: Add ldlm_lock_addref_try().
1748 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1749 attempts to addref a lock that might be being canceled
1754 Description: Add ldlm_weigh_callback().
1755 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1756 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1757 hard-coded `number of pages' logic.
1761 Description: Add lockdep annotations to llog code.
1762 Details : Use appropriately tagged _nested() locking calls in the places
1763 where llog takes more than one ->lgh_lock lock.
1767 Description: Add loi_kms_set().
1768 Details : Wrap kms updates into a helper function.
1772 Description: Constify instances of struct lsm_operations.
1773 Details : Constify instances of struct lsm_operations.
1777 Description: lu_conf support.
1778 Details : On a server, a file system object is uniquely identified
1779 by a fid, which is sufficient to locate and load all object
1780 state (inode). On a client, on the other hand, more data are
1781 necessary instantiate an object. Change lu_object_find() and
1782 friends to take additional `lu_conf' argument describing object.
1783 Typically this includes layout information.
1787 Description: lu_context fixes.
1788 Details : Introduce new lu_context functions that are needed on the client
1789 side, where some system threads (ptlrpcd) are shared by multiple
1790 modules, and so cannot be stopped during module shutdown.
1794 Description: Add start and stop methods to lu_device_type_operations.
1795 Details : Introduce two new methods in lu_device_type_operations, that are
1796 invoked when first instance of a given type is created and last one
1797 is destroyed respectively. This is need by CLIO.
1801 Description: Add lu_ref support to struct lu_device.
1802 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1803 track leaked references.
1807 Description: Introduce lu_kmem_descr.
1808 Details : lu_kmem_descr and its companion interface allow to create
1809 and destroy a number of kmem caches at once.
1813 Description: Fix lu_object finalization race.
1814 Details : Fix a race between lu_object_find() finding an object and its
1815 concurrent finalization. This race is (most likely) not possible
1816 on the server, but might happen on the client.
1820 Description: Introduce lu_ref interface.
1821 Details : lu_ref is a debugging module allowing to track references to
1822 a given object. It is quite cpu expensive, and has to be
1823 explicitly enabled with --enable-lu_ref. See usage description
1828 Description: Factor lu_site procfs stats into a separate function.
1829 Details : Separate lu_site stats printing code into a separate function
1830 that can be reused on a client.
1834 Description: Constify instances of struct {lu,dt,md}_device_operations.
1835 Details : Constify instances of struct {lu,dt,md}_device_operations.
1839 Description: Introduce struct md_site and move meta-data specific parts of
1840 struct lu_site here.
1841 Details : Move md-specific fields out of struct lu_site into special struct
1842 md_site, so that lu_site can be used on a client.
1846 Description: Kill mdd_lov_destroy().
1847 Details : Remove unused mdd code.
1851 Description: Add st_block checking to multistat.c.
1852 Details : Add st_block checking to multistat.c.
1856 Description: Add lu_ref support to struct obd_device.
1857 Details : Add lu_ref tracking to obd_device.
1861 Description: Kill obd_set_fail_loc().
1862 Details : Remove unused code.
1866 Description: Add special type for ptlrpc_request interpret functions.
1867 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1868 there. Actual usage will be in CLIO.
1872 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1873 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1874 doesn't work with lockdep.
1878 Description: Add rwv.c test program.
1879 Details : New testing program exercising readv(2) and writev(2) (Qian).
1883 Description: Add sendfile.c test program.
1884 Details : New testing program exercising sendfile(2) (Jay).
1888 Description: Ratelimit a message that can be very frequent.
1889 Details : Ratelimit a memory allocation failure message that can
1894 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1895 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1899 Description: Liblustre build fixes.
1900 Details : Liblustre build fixes.
1904 Description: libcfs: add cfs_{need,cond}_resched() interface.
1905 Details : libcfs: add cfs_{need,cond}_resched() definition and
1906 implementations for Linux, NT, and liblustre.
1908 Severity : enhancement
1910 Description: More exported tunables for mballoc
1911 Details : Add support for tunable preallocation window and new tunables for
1912 large/small requests
1916 Description: Detect corruption of block bitmap and checking for preallocations
1917 Details : Checks validity of on-disk block bitmap. Also it does better
1918 checking of number of applied preallocations. When corruption is
1919 found, it turns filesystem readonly to prevent further corruptions.
1923 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1924 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1925 window to 0. If it does hit cache miss, and read-pattern is still
1926 stride-io mode, does not reset the stride window, but also does not
1927 increase the stride window length in this case.
1931 Frequency : only for big-endian servers
1932 Description: Check if system is big-endian while mounting fs with extents feature
1933 Details : Mounting a filesystem with extents feature will fail on big-endian
1934 systems since ext3-based ldiskfs is not supported on big-endian
1935 systems. This can be over-riden with "bigendian_extents" mount option.
1937 Severity : enhancement
1939 Description: The root squash functionality
1940 Details : A security feature, which is to prevent users from being able
1941 to mount lustre on their desktop, run as root, and delete
1942 all of the files in the filesystem. The goal is accomplished by
1943 remapping user id (UID) and group id (GID) of the root user to
1944 a UID and GID specified by the system administartor via Lustre
1945 configuration management server (MGS). The functionality also
1946 allows to specify sets of clients for which the remapping does
1951 Description: Excessive recovery window
1952 Details : With AT enabled, the recovery window can be excessively long (6000+
1953 seconds). To address this problem, we no longer use
1954 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1955 timeout no longer depends on the service time, it is set to
1956 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1957 time via pb_service_time.
1961 Description: Watchdog triggered on MDS failover
1962 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
1963 the OSTs know that the MDS "UUID" can be reused for the same export
1964 from a different NID, so we do not need to wait for the export to be
1968 Frequency : rare, only if using MMP with Linux RAID
1970 Description: MMP doesn't work with Linux RAID
1971 Details : While using HA for Lustre servers with Linux RAID, it is possible
1972 that MMP will not detect multiple mounts. To make this work we
1973 need to unplug the device queue in RAID when the MMP block is being
1974 written. Also while reading the MMP block, we should read it from
1975 disk and not the cached one.
1977 Severity : enhancement
1979 Description: open file using fid
1980 Details : A file can be opened using just its fid, like
1981 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
1983 --------------------------------------------------------------------------------
1985 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1987 * Support for kernels:
1988 2.6.5-7.283 (SLES 9),
1989 2.6.9-55.EL (RHEL 4),
1990 2.6.16.46-0.14 (SLES 10),
1991 2.6.18.8 vanilla (kernel.org)
1992 * Client support for unpatched kernels:
1993 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1994 2.6.16 - 2.6.22 vanilla (kernel.org)
1995 * Due to recently discovered recovery problems, we do not recommend
1996 using patchless RHEL 4 clients with this or any earlier release.
1997 * Recommended e2fsprogs version: 1.39.cfs8
1998 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1999 * Starting with this release, the ldiskfs backing filesystem required
2000 by Lustre is now in its own package, lustre-ldiskfs. This package
2001 should be installed. It is versioned separately from Lustre and
2002 may be released separately in future.
2006 Description: block reactivating mgc import until all deactivates complete
2007 Details : Fix race when failing back MDT/MGS to itself (testing)
2010 Frequency : at statup only
2012 Description: mds_lov_synchronize race leads to various problems
2013 Details : simultaneous MDT->OST connections at startup can cause the
2014 sync to abort, leaving the OSC in a bad state.
2016 Severity : enhancement
2018 Description: add optional extra BUILD_VERSION info
2019 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2020 to override the lustre version.
2023 Frequency : 2.6.18 servers only
2025 Description: ll_kern_mount() doesn't release the module reference
2026 Details : The ldiskfs module reference count never drops down to 0
2027 because ll_kern_mount() doesn't release the module reference.
2032 Description: server LBUG when using old ost_num_threads parameter
2033 Details : Accept the old ost_num_threads parameter but warn that it
2034 is deprecated, and fix an off-by-one error that caused an LBUG.
2039 Description: Transient SCSI error results in persistent IO issue
2040 Details : iobuf->dr_error is not reinitialized to 0 between two
2044 Frequency : sometimes when underlying device returns I/O errors
2046 Description: OSTs not going read-only during write failures
2047 Details : OSTs are not remounted read-only when the journal commit threads
2048 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2049 instead of the ext3 wrappers.
2053 Description: poor connect scaling with increasing client count
2054 Details : Don't run filter_grant_sanity_check for more than 100 exports
2055 to improve scaling for large numbers of clients.
2058 Frequency : SLES10 only
2060 Description: sanity-quota.sh quotacheck failed: rc = -22
2061 Details : Quotas cannot be enabled on SLES10.
2064 Frequency : liblustre clients only
2066 Description: getdirentries does not give error when run on compute nodes
2067 Details : getdirentries does not fail when the size specified as an argument
2068 is too small to contain at least one entry
2070 Severity : enhancement
2072 Description: Add LNET router traceability for debug purposes
2073 Details : If a checksum failure occurs with a router as part of the
2074 IO path, the NID of the last router that forwarded the bulk data
2075 is printed so it can be identified.
2080 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2081 Details : Due to a race condition, liblustre clients were occasionally
2082 evicted incorrectly.
2084 Severity : enhancement
2086 Description: lfs setstripe use optional parameters instead of postional
2089 Severity : enhancement
2091 Description: Nanosecond timestamp support for ldiskfs
2092 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2093 resolution timestamps. There is not yet support for this at
2094 the Lustre filesystem level.
2097 Frequency : during server recovery
2099 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2100 Details : request with rq_no_resend flag not awake l_wait_event if they get
2104 Frequency : nfs export on patchless client
2106 Description: connectathon hang when test nfs export over patchless client
2107 Details : Disconnected dentry cannot be found with lookup, so we do not need
2108 to unhash it or make it invalid
2111 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2112 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2113 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2118 Description: open req refcounting wrong on reconnect
2119 Details : If reconnect happened between getting open reply from server and
2120 call to mdc_set_replay_data in ll_file_open, we will schedule
2121 replay for unreferenced request that we are about to free.
2122 Subsequent close will crash in variety of ways.
2123 Check that request is still eligible for replay in
2124 mdc_set_replay_data().
2129 Description: disable writes to filesystem when reading health_check file
2130 Details : the default for reading the health_check proc file has changed
2131 to NOT do a journal transaction and write to disk, because this
2132 can cause reads of the /proc file to hang and block HA state
2133 checking on a healthy but otherwise heavily loaded system. It
2134 is possible to return to the previous behaviour during configure
2135 with --enable-health-write.
2137 Severity : enhancement
2139 Description: 64-bit inode version
2140 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2141 to the inode. This will be required for version-based recovery.
2146 Description: MDS fails to start if a duplicate client export is detected
2147 Details : in some rare cases it was possible for a client to connect to
2148 an MDS multiple times. Upon recovery the MDS would detect this
2149 and fail during startup. Handle this more gracefully.
2151 Severity : enhancement
2153 Description: Add -o localflock option to simulate old noflock
2155 Details : This will achieve local-only flock/fcntl locks
2161 Description: log_commit_thread vs filter_destroy race leads to crash
2162 Details : Take import reference before releasing llog record semaphore
2167 Description: Wrong request locking in request set processing
2168 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2169 imp_delayed_list, in this place should be used imp_lock.
2172 Frequency : when reconnection
2174 Description: Grant Leak when osc reconnect to OST
2175 Details : When osc reconnect ost, OST(filter) should check whether it
2176 should grant more space to client by comparing fed_grant and
2177 cl_avail_grant, and return the granted space to client instead
2178 of "new granted" space, because client will call osc_init_grant
2179 to update the client grant space info.
2182 Frequency : when client reconnect to OST
2184 Description: Grant Leak when osc do resend and replay bulk write
2185 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2186 bulk write request, because the grant info while be sync between
2187 OSC and OST when reconnect, and we should ignore the grant info
2188 these of resend/replay write req.
2193 Description: Grant space more than avaiable left space sometimes.
2194 Details : When then OST is about to be full, if two bulk writing from
2195 different clients came to OST. Accord the avaliable space of the
2196 OST, the first req should be permitted, and the second one
2197 should be denied by ENOSPC. But if the seconde arrived before
2198 the first one is commited. The OST might wrongly permit second
2199 writing, which will cause grant space > avaiable space.
2202 Frequency : when client is evicted
2204 Description: Grant might be wrongly erased when osc is evicted by OST
2205 Details : when the import is evicted by server, it will fork another
2206 thread ptlrpc_invalidate_import_thread to invalidate the
2207 import, where the grant will be set to 0. While the original
2208 thread will update the grant it got when connecting. So if
2209 the former happened latter, the grant will be wrongly errased
2210 because of this race.
2215 Description: Checking Stale with correct fid
2216 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2217 to check whether it is stale, because sometimes, we want the
2218 enqueue happened anyway, and op_data.fid2 will not be initialized.
2220 Severity : enhancement
2222 Description: update patchless client
2223 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2226 Frequency : only with 2.4 kernel
2228 Description: random memory corruption
2229 Details : size of struct ll_inode_info is to big for union inode.u and this
2230 can be cause of random memory corruption.
2235 Description: Memory leak in recovery
2236 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2237 It should also check obd_fail before fsfilt_start, otherwise if
2238 fsfilt_start return -EROFS,(failover mds during mds recovery).
2239 then the req will return with repmsg->transno = 0 and rc = EROFS.
2240 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2241 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2242 be freed no matter whether fsfilt_commit success or not.
2245 Frequency : only with huge count clients
2247 Description: Prevents from taking the superblock lock in llap_from_page for
2249 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2250 for llap_from_page call in ll_removepage prevents from taking the
2251 superblock lock for a soon died page.
2256 Description: Not check open intent error before release open handle
2257 Details : in some rare cases, the open intent error is not checked before
2258 release open handle, which may cause
2259 ASSERTION(open_req->rq_transno != 0), because it tries to release
2260 the failed open handle.
2265 Description: Set cat log bitmap only after create log success.
2266 Details : in some rare cases, the cat log bitmap is set too early. and it
2267 should be set only after create log success.
2271 Description: Accessing a block bevice can re-enable I/O when Lustre is
2272 tearing down a device.
2273 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2278 Description: service threads may hog cpus when there are a lot of requests
2279 Details : Insert cond_resched to give other threads a chance to use some CPU
2284 Description: the cat log was not initialized in recovery
2285 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2286 unlink log on mds will not be initialized until mds post
2287 recovery. And also in mds post recovery, the unlink log will
2288 initialization will be done asynchronausly, so there will be race
2289 between add unlink log and unlink log initialization.
2293 Description: brw_stats were being printed incorrectly
2294 Details : brw_stats were being printed as log2 but all of them were not
2295 recorded as log2. Also remove some code duplication arising from
2296 filter_tally_{read,write}.
2300 Frequency : rare, only in recovery.
2301 Description: ASSERTION(req->rq_type != LI_POISON) failed
2302 Details : imp_lock should be held while iterating over imp_sending_list for
2303 prevent destroy request after get timeout in ptlrpc_queue_wait.
2307 Description: replay-single.sh test 52 fails
2308 Details : A lock's skiplist need to be cleanup when it being unlinked
2309 from its resource list.
2313 Description: Short directio read returns full requested size rather than
2315 Details : Direct I/O operations should return actual amount of bytes
2316 transferred rather than requested size.
2318 Severity : enhancement
2320 Description: metadata RPC reduction (e.g. for rm performance)
2321 Details : decrease the amount of synchronous RPC between clients and servers
2322 by canceling conflicing lock before the operation on the client
2323 and packing thier handles into the main operation RPC to server.
2325 Severity : enhancement
2327 Description: Async OSC create to avoid the blocking unnecessarily.
2328 Details : If a OST has no remain object, system will block on the creating
2329 when need to create a new object on this OST. Now, ways use
2330 pre-created objects when available, instead of blocking on an
2331 empty osc while others are not empty. If we must block, we block
2332 for the shortest possible period of time.
2336 Description: improve handling recoverable errors
2337 Details : if request processig with error which can be recoverable on server
2338 request should be resend, otherwise page released from cache and
2341 Severity : enhancement
2343 Description: refine locking for avoid write wrong info into lov_objid file
2344 Details : fix possible races with add new target and write/update data in
2347 --------------------------------------------------------------------------------
2349 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2354 Frequency : on some architectures
2356 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2357 Details : Uninitialized flags sometimes cause configuration commands to
2361 Frequency : patchless clients only
2363 Description: missing __iget() symbol export
2364 Details : The __iget() symbol export is missing. To avoid the need for
2365 this on patchless clients the deathrow inode reaper is turned
2366 off, and we depend on the VM to clean up old inodes. This
2367 dependency was during via the fix for bug 12181.
2369 --------------------------------------------------------------------------------
2371 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2373 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2374 INTEROPERATE with older versions automatically. In many cases a
2375 special upgrade step is needed. Please read the
2376 user documentation before upgrading any part of a live system.
2377 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2378 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2379 * WARNING: Lustre configuration and startup changes are required with
2380 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2384 Severity : enhancement
2386 Description: MountConf
2387 Details : Lustre configuration is now managed via mkfs and mount
2388 commands instead of lmc and lconf. New obd types (MGS, MGC)
2389 are added for dynamic configuration management. See
2390 https://mail.clusterfs.com/wikis/lustre/MountConf for
2393 Severity : enhancement
2395 Description: dynamic OST addition
2396 Details : OSTs can now be added to a live filesystem
2398 Severity : enhancement
2400 Description: startup order invariance
2401 Details : MDTs and OSTs can be started in any order. Clients only
2402 require the MDT to complete startup.
2404 Severity : enhancement
2406 Description: parallel, asynchronous orphan cleanup
2407 Details : orphan cleanup is now performed in separate threads for each
2408 OST, allowing parallel non-blocking operation.
2410 Severity : enhancement
2412 Description: optimized stripe assignment
2413 Details : stripe assignments are now made based on ost space available,
2414 ost previous usage, and OSS previous usage, in order to try
2415 to optimize storage space and networking resources.
2417 Severity : enhancement
2419 Description: Permanently set tunables
2420 Details : All writable /proc/fs/lustre tunables can now be permanently
2421 set on a per-server basis, at mkfs time or on a live system.
2423 Severity : enhancement
2425 Description: Lustre message v2
2426 Details : Add lustre message format v2.
2428 Severity : enhancement
2430 Description: client OST exclusion list
2431 Details : Clients can be started with a list of OSTs that should be
2432 declared "inactive" for known non-responsive OSTs.
2436 Description: SPEC SFS validation failure on NFS v2 over lustre.
2437 Details : Changes the blocksize for regular files to be 2x RPC size,
2438 and not depend on stripe size.
2440 Severity : enhancement
2442 Description: Multiple MD RPCs in flight.
2443 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2444 To avoid overly-overloading MDS, introduce a limit on number of
2445 MDS RPCs in flight for a single client and add /proc controls
2446 to adjust this limit.
2448 Severity : enhancement
2450 Description: client read/write statistics
2451 Details : Add client read/write call usage stats for performance
2452 analysis of user processes.
2453 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2454 file access. extents_stats shows chunk size distribution.
2455 extents_stats_per_process show chunk size distribution per
2458 Severity : enhancement
2460 Description: mds statistics
2461 Details : Add detailed mds operations statistics in
2462 /proc/fs/lustre/mds/*/stats.
2466 Description: Failure of copying files with lustre special EAs.
2467 Details : Client side always return success for setxattr call for lustre
2468 special xattr (currently only "trusted.lov").
2473 Description: Refcount LNET uuids
2474 Details : The global LNET uuid list grew linearly with every startup;
2475 refcount repeated list entries instead of always adding to
2478 Severity : enhancement
2480 Description: Dynamic service threads
2481 Details : Within a small range, start extra service threads
2482 automatically when the request queue builds up.
2485 Frequency : mixed-endian client/server environments
2487 Description: mixed-endian crashes
2488 Details : The new msg_v2 system had some failures in mixed-endian
2491 Severity : enhancement
2493 Description: Easy OST removal
2494 Details : OSTs can be permanently deactivated with e.g. 'lctl
2495 conf_param lustre-OST0001.osc.active=0'
2497 Severity : enhancement
2499 Description: MGS proc entries
2500 Details : Added basic proc entries for the MGS showing what filesystems
2503 Severity : enhancement
2505 Description: provide MGS failover
2506 Details : Added config lock reacquisition after MGS server failover.
2508 Severity : enhancement
2510 Description: add Linux 2.4 support
2511 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2515 Description: a large application tries to do I/O to the same resource and dies
2516 in the middle of it.
2517 Details : Check the req->rq_arrival time after the call to
2518 ost_brw_lock_get(), but before we do anything about
2519 processing it & sending the BULK transfer request. This
2520 should help move old stale pending locks off the queue as
2521 quickly as obd_timeout.
2524 Frequency : when an incorrect nid is specified during startup
2526 Description: ptlrpc connect to non-existant node causes kernel crash
2527 Details : LNET can't be re-entered from an event callback, which
2528 happened when we expire a message after the export has been
2529 cleaned up. Instead, hand the zombie cleanup off to another
2532 Severity : enhancement
2534 Description: plain/inodebits lock performance improvement
2535 Details : Grouping plain/inodebits in granted list by their request modes
2536 and bits policy, thus improving the performance of search through
2540 Frequency : only if OST filesystem is corrupted
2542 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2543 Details : for a short time RPCs with bulk IO are in the replay list,
2544 but replay of bulk IOs is unimplemented. If the OST filesystem
2545 is corrupted due to disk cache incoherency and then replay is
2546 started it is possible to trip an assertion. Avoid putting
2547 committed RPCs into the replay list at all to avoid this issue.
2550 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2552 Description: System hang on startup
2553 Details : This bug allowed the liblustre (e.g. catamount) client to
2554 return to the app before handling all startup RPCs. This
2555 could leave the node unresponsive to lustre network traffic
2556 and manifested as a server ptllnd timeout.
2558 Severity : enhancement
2560 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2561 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2562 boolean to control whether to print peer debug info when a
2563 client's RPC times out.
2566 Frequency : only for kernels with patches from Lustre below 1.4.3
2568 Description: Remove old rdonly API
2569 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2572 Frequency : only for devices with external journals
2574 Description: Set external device read-only also
2575 Details : During a commanded failover stop, we set the disk device
2576 read-only while the server shuts down. We now also set any
2577 external journal device read-only at the same time.
2580 Frequency : when upgrading from 1.4 while trying to change parameters
2582 Description: The wrong (new) MDC name was used when setting parameters for
2583 upgraded MDT's. Also allows changing of OSC (and MDC)
2584 parameters if --writeconf is specified at tunefs upgrade time.
2587 Frequency : when setting specific ost indicies
2589 Description: QOS code breaks on skipped indicies
2590 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2591 created with --index need not be sequential.
2593 Severity : enhancement
2595 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2596 Details : The uninit_groups feature works in conjunction with the kernel
2597 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2598 up the pass1 processing of e2fsck. This is a read-only feature
2599 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2600 mount filesystems that have had this feature enabled.
2602 Severity : enhancement
2604 Description: Improve multi-block allocation algorithm to avoid fragmentation
2605 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2606 allocation locality and avoid filesystem fragmentation.
2608 ------------------------------------------------------------------------------
2610 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2612 * Support for kernels:
2613 2.4.21-47.0.1.EL (RHEL 3)
2614 2.6.5-7.283 (SLES 9)
2615 2.6.9-42.0.10.EL (RHEL 4)
2616 2.6.12.6 vanilla (kernel.org)
2617 2.6.16.27-0.9 (SLES 10)
2618 * Recommended e2fsprogs version: 1.39.cfs5
2620 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2624 Frequency : occasional, depends on client load and configuration
2625 Bugzilla : 12181, 12203
2626 Description: data loss for recently-modified files
2628 Details : In some cases it is possible that recently written or created
2629 files may not be written to disk in a timely manner (this should
2630 normally be within 30s unless client IO load is very high).
2631 The problem appears as zero-length files or files that are a
2632 multiple of 1MB in size after a client crash or client eviction
2633 that are missing data at the end of the file.
2635 This problem is more likely to be hit on clients where files are
2636 repeatedly created and unlinked in the same directory, clients
2637 have a large amount of RAM, have many CPUs, the filesystem has
2638 many OSTs, the clients are rebooted frequently, and/or the files
2639 are not accessed by other nodes after being written.
2641 The presence of the problem can be detected by looking at
2642 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2643 smaller than the second (nr_unused) then dirty files will not
2644 be flushed automatically to disk. "sync; sleep 10" should be
2645 run several times on the node before unmounting it to update
2646 Lustre (this is also safe to run on nodes without this problem).
2648 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2649 that can cause this same problem, so customers using that kernel
2650 also need to update the kernel in addition to Lustre. In order
2651 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2653 It is normal that files written just before a client crash (less
2654 than 30s) may not yet have been flushed to disk, even for local
2658 Frequency : frequent on thin XT3 nodes
2660 Description: UUID collision on thin XT3 Linux nodes
2661 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2662 randomly, since we relied on an insufficiently-seeded PRNG.
2667 Description: OSS hangs after "All ost request buffers busy"
2668 Details : A deadlock between quota and journal operations caused OSS
2669 hangs after printing "All ost request buffers busy."
2672 Frequency : always on liblustre builds
2674 Description: Cleanup compiler warnings on liblustre
2677 Frequency : always on liblustre builds on XT3
2679 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2683 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2684 > 32000 subdirectories
2685 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2686 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2687 subdirectory count crosses 32000. This will aid e2fsck to
2688 correctly handle more than 32000 subdirectories.
2691 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2693 Description: System hang on startup
2694 Details : This bug allowed the liblustre (e.g. catamount) client to
2695 return to the app before handling all startup RPCs. This
2696 could leave the node unresponsive to lustre network traffic
2697 and manifested as a server ptllnd timeout.
2699 Severity : enhancement
2701 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2702 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2703 boolean to control whether to print peer debug info when a
2704 client's RPC times out.
2709 Description: make O_SYNC working on 2.6 kernels
2710 Details : 2.6 kernels use different method for mark pages for write,
2711 so need add a code to lustre for O_SYNC work.
2716 Description: Failure to close file and release space on NFS
2717 Details : Put inode details into lock acquired in ll_intent_file_open.
2718 Use mdc_intent_lock in ll_intent_open to properly
2719 detect all kind of errors unhandled by mdc_enqueue.
2724 Description: proc file read during shutdown sometimes raced obd removal,
2726 Details : Add lock to prevent obd access after proc file removal.
2729 Frequency : Only for files larger than 4GB on 32-bit clients.
2731 Description: improperly doing page alignment of locks
2732 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2733 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2738 Description: RPCs being resent when they shouldn't be
2739 Details : Some RPCs that should not be resent are being resent. This
2740 can cause inconsistencies in the RPC state machine. Do not
2741 resend such requests.
2744 Frequency : rare, only with NFS export
2746 Description: Crash on NFS re-export node
2747 Details : under very unusual load conditions an assertion is hit in
2748 ll_intent_file_open()
2751 Frequency : only if OST filesystem is corrupted
2753 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2754 Details : for a short time RPCs with bulk IO are in the replay list,
2755 but replay of bulk IOs is unimplemented. If the OST filesystem
2756 is corrupted due to disk cache incoherency and then replay is
2757 started it is possible to trip an assertion. Avoid putting
2758 committed RPCs into the replay list at all to avoid this issue.
2763 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2764 Details : Large single O_DIRECT read and write calls can fail to allocate
2765 a sufficiently large buffer to process the request. In case of
2766 allocation failure the allocation is retried with a smaller
2767 buffer and broken into smaller requests.
2769 Severity : enhancement
2771 Description: Add -o localflock option to simulate old noflock behaviour.
2772 Details : This will achieve local-only flock/fcntl locks coherentness.
2777 Description: versioning check is incomplete
2778 Details : Checking the version difference of client vs. server, report
2779 error if the gap is too big.
2784 Description: add support PG_writeback bit
2785 Details : add support for PG_writeback bit for Lustre, for more carefull
2786 work with page cache in 2.6 kernel. This also fix some deadlocks
2787 and remove hack for work O_SYNC with 2.6 kernel.
2789 Severity : enhancement
2791 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2792 Details : The uninit_groups feature works in conjunction with the kernel
2793 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2794 up the pass1 processing of e2fsck. This is a read-only feature
2795 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2796 mount filesystems that have had this feature enabled.
2798 Severity : enhancement
2800 Description: Improve multi-block allocation algorithm to avoid fragmentation
2801 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2802 allocation locality and avoid filesystem fragmentation.
2804 ------------------------------------------------------------------------------
2806 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2808 * Support for kernels:
2809 2.6.9-42.0.3.EL (RHEL 4)
2810 2.6.5-7.276 (SLES 9)
2811 2.4.21-47.0.1.EL (RHEL 3)
2812 2.6.12.6 vanilla (kernel.org)
2813 2.6.16.21-0.8 (SLES10)
2814 * Recommended e2fsprogs version: 1.39.cfs2-0
2816 * The backwards-compatible /proc/sys/portals symlink has been removed
2817 in this release. Before upgrading, please ensure that you change
2818 any configuration scripts or /etc/sysctl.conf files that access
2819 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2820 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2821 in advance of the upgrade on any system running Lustre 1.4.6 or
2822 newer, since /proc/sys/lnet was added in that version.
2823 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2827 Frequency : only when quota is used
2829 Description: avoid scanning export list for quota master
2830 Details : Change the algorithms to avoid scanning export list in order
2831 to improve the efficiency.
2834 Frequency : MDS failover only, very rarely
2836 Description: "went back in time" messages on mds failover
2837 Details : The greatest transno may be lost when the current operation
2838 finishes with an error (transno==0) and the client's last_rcvd
2839 record is over-written. Save the greatest transno in the
2840 mds_last_transno for this case.
2843 Frequency : always for specific kernels and striping counts
2845 Description: client may get "Matching packet too big" without ACL support
2846 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2847 when trying to access files in certain configurations. The
2848 clients should in fact be denied when mounting because they do
2849 not understand ACLs.
2852 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2854 Description: many clients connecting with IO in progress causes connect timeouts
2855 Details : Avoid synchronous journal commits to avoid delays caused by many
2856 clients connecting/disconnecting when bulk IO is in progress.
2857 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2858 OST_IO_PORTAL to avoid delays behind potentially many pending
2862 Frequency : occasionally with multiple writers to a single file
2864 Description: shared writes to file may result in wrong size reported by stat()
2865 Details : Allow growing of kms when extent lock is cancelled
2868 Frequency : always with random mmap IO to multi-striped file
2870 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2871 Details : Only if the hole is at the end of OST object so that kms is too
2872 small. Fix is to increase kms accordingly in ll_nopage.
2875 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2877 Description: writes to a missing object would leak memory on the OST
2878 Details : If there is an inconsistency between the MDS and OST filesystems,
2879 such that the MDS references an object that doesn't exist, writes
2880 to that object will leak memory due to incorrect cleanup in the
2881 error handling path, eventually running out of memory on the OST.
2886 Description: Creating too long symlink causes lustre errors
2887 Details : Check symlink and name lengths before sending requests to MDS.
2890 Frequency : only if flock is enabled (not on by default)
2892 Description: posix locks not released on fd closure on 2.6.9+
2893 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2894 kernels, this caused such locks not to be released on fd close and
2895 then assertions on fs unmount about still used locks.
2898 Frequency : MDS failover only, very rarely
2900 Description: clients may get ASSERTION(granted_lock != NULL)
2901 Details : When request was taking a long time, and a client was resending
2902 a getattr by name lock request. The were multiple lock requests
2903 with the same client lock handle and
2904 mds_getattr_name->fixup_handle_for_resent_request found one of the
2905 lock handles but later failed with ASSERTION(granted_lock != NULL).
2910 Description: handle->h_buffer_credits > 0, assertion failure
2911 Details : h_buffer_credits is zero after truncate, causing assertion
2912 failure. This patch extends the transaction or creates a new
2916 Frequency : NFS re-export or patchless client
2917 Bugzilla : 11179, 10796
2918 Description: Crash on NFS re-export node (__d_move)
2919 Details : We do not want to hash the dentry if we don't have a lock.
2920 But if this dentry is later used in d_move, we'd hit uninitialised
2921 list head d_hash, so we just do this to init d_hash field but
2922 leave dentry unhashed.
2925 Frequency : NFS re-export or patchless client
2927 Description: NFS exports has problem with symbolic link
2928 Details : lustre client didn't properly install dentry when re-exported
2929 to NFS or running patchless client.
2932 Frequency : NFS re-export or patchless client
2934 Description: Various nfs/patchless fixes.
2935 Details : fixes reuse disconected alias for lookup process - this fixes
2936 warning "find_exported_dentry: npd != pd",
2937 fix permission error with open files at nfs.
2938 fix apply umask when do revalidate.
2941 Frequency : occasional
2943 Description: Crash on NFS re-export node
2944 Details : calling clear_page() on the wrong pointer triggered oops in
2945 generic_mapping_read().
2948 Frequency : rarely, using O_DIRECT IO
2950 Description: unaligned directio crashes client with LASSERT
2951 Details : check for unaligned buffers before trying any requests.
2954 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2956 Description: stale data returned from RAID cache
2957 Details : If only a small amount of IO is done to the RAID device before
2958 reading it again it is possible to get stale data from the RAID
2959 cache instead of reading it from disk.
2962 Frequency : always for sles10 kernel
2964 Description: sles10 support
2965 Details : ll_follow_link: compile fixes and using of nd_set_link
2966 under newer kernels.
2969 Frequency : depends on arch, kernel and compiler version, always on sles10
2972 Description: recursive or deep enough symlinks cause stack overflow
2973 Details : getting rid of large stack-allocated variable in
2977 Frequency : depends on hardware
2979 Description: lustre write performance loss in the SLES10 kernel
2980 Details : the performance loss is caused by using of write barriers in the
2981 ext3 code. The SLES10 kernel turns barrier support on by
2982 default. The fix is to undo that change for ldiskfs.
2984 ------------------------------------------------------------------------------
2986 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2988 * Support for kernels:
2989 2.6.9-42.0.3EL (RHEL 4)
2990 2.6.5-7.276 (SLES 9)
2991 2.4.21-47.0.1.EL (RHEL 3)
2992 2.6.12.6 vanilla (kernel.org)
2996 Frequency : quota enabled and large files being deleted
2998 Description: releasing more than 4GB of quota at once hangs OST
2999 Details : If a user deletes more than 4GB of files on a single OST it
3000 will cause the OST to spin in an infinite loop. Release
3001 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3006 Description: statfs data retrieved from /proc may be stale or zero
3007 Details : When reading per-device statfs data from /proc, in the
3008 {kbytes,files}_{total,free,avail} files, it may appear
3009 as zero or be out of date.
3012 Frequency : systems with MD RAID1 external journal devices
3014 Description: lconf's call to blkid is confused by RAID1 journal devices
3015 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3016 of returning all block devices that match the journal UUID.
3019 Frequency : always, for aggregate stripe size over 4GB
3021 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3022 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3023 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3026 Frequency : always if "lfs find" used on a local file/directory
3028 Description: "lfs find" segfaults if used on a local file/directory
3029 Details : The case where a directory component was not specified wasn't
3030 handled correctly. Handle this properly.
3033 Frequency : always on ppc64
3035 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3036 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3037 perform endianness conversion.
3040 Frequency : rarely (truncate to non-zero file size after write under load)
3041 Bugzilla : 10730, 10687
3042 Description: Files padded with zeros to next 4K multiple
3043 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3044 it is possible that files that are truncated to a non-zero size
3045 immediately after being written are filled with zero bytes beyond
3046 the truncated size. No file data is lost.
3048 Severity : enhancement
3050 Description: Allow recovery/failover for liblustre clients.
3051 Details : liblustre clients were unaware of failover configurations until
3054 Severity : enhancement
3056 Description: user file locks should fail when not mounting with flock option
3057 Details : Set up an error-returning stub in ll_file_operations.lock field
3058 to prevent incorrect behaviour when client is mounted without
3059 flock option. Also, set up properly f_op->flock field for
3063 Frequency : always on ia64
3065 Description: "lfs df" loops on printing out MDS statfs information
3066 Details : The obd_ioctl_data was not initialized and in some systems
3067 this caused a failure during the ioctl that did not return
3068 an error. Initialize the struct and return an error on failure.
3071 Frequency : SLES 9 only
3073 Description: Error of copying files with lustre special EAs as root
3074 Details : Client side always return success for setxattr call for lustre
3075 special xattr (currently only "trusted.lov").
3078 Frequency : rarely on clusters with both ia64+i386 clients
3080 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3081 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3082 clients are doing concurrent writes to the same file it is
3083 possible that the ia64 clients may LASSERT because the OST
3084 extent locks are not PAGE_SIZE aligned. Ensure that grown
3085 locks are always aligned on the request boundary.
3088 Frequency : specific use, occasional
3090 Description: Overwriting in use executable truncates on-disk binary image
3091 Details : If one node attempts to overwrite an executable in use by
3092 another node, we now correctly return ETXTBSY instead of
3093 truncating the file.
3095 Severity : enhancement
3097 Description: Async OSC create to avoid the blocking unnecessarily.
3098 Details : If a OST has no remain object, system will block on the creating
3099 when need to create a new object on this OST. Now, ways use
3100 pre-created objects when available, instead of blocking on an
3101 empty osc while others are not empty. If we must block, we block
3102 for the shortest possible period of time.
3107 Description: chmod on Lustre root is propagated to other clients
3108 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3109 invalid by the follow_mount time.
3112 Frequency : liblustre clients only
3114 Description: Race in 'instant cancel' lock handling could lead to such locks
3115 never to be granted in case of SMP MDS
3116 Details : Do not destroy not yet granted but cbpending locks in
3120 Frequency : replay/resend of open
3122 Description: non null lock assetion failure in mds_intent_policy
3123 Details : Trying to replay/resend lockless open requests resulted in
3124 mds_open() returning 0 with no lock. Now it sets a flag if
3125 a lock is going to be returned.
3127 Severity : enhancement
3129 Description: Checksum enhancements
3130 Details : New checksum enhancements allow for resending RPCs that failed
3133 Severity : enhancement
3135 Description: Tunables on number of dirty pages in cacche
3136 Details : Allow to set limit on number of dirty pages cached.
3141 Description: client crash on unmount - lock still has references
3142 Details : In some error handling cases it was possible to leak a lock
3143 reference on a client while accessing a file. This was not
3144 harmful to the client during operation, but would cause the
3145 client to crash when the filesystem is unmounted.
3148 Frequency : specific case, rare
3150 Description: ETXTBSY on mds though file not in use by client
3151 Details : ETXTBSY is no longer incorrectly returned when attempting to
3152 chmod or chown a directory that the user previously tried to
3153 execute or a currently-executing binary.
3156 Frequency : extremely rare except on liblustre-based clients
3158 Description: Lustre space not freed when files are deleted
3159 Details : Clean up open-unlinked files after client eviction. Previously
3160 the unlink was skipped and the files remained as orphans.
3165 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3166 Details : In some cases it was possible to send a blocking callback to
3167 a client doing a glimpse, even though that client didn't get
3168 a lock granted. When the glimpse lock is cancelled on the OST
3169 the freed lock is left on the waiting list and corrupted the list.
3172 Frequency : all core dumps
3174 Description: Broke core dumps to lustre
3175 Details : Negative dentry may be unhashed if parent does not have UPDATE
3176 lock, but some callers, e.g. do_coredump, expect dentry to be
3177 hashed after successful create, hash it in ll_create_it.
3179 ------------------------------------------------------------------------------
3181 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3183 * Support for kernels:
3184 2.6.9-42.0.2.EL (RHEL 4)
3185 2.6.5-7.276 (SLES 9)
3186 2.4.21-40.EL (RHEL 3)
3187 2.6.12.6 vanilla (kernel.org)
3191 Frequency : always on RHEL 3
3193 Description: Number of open files grows over time
3194 Details : The number of open files grows over time, whether or not
3195 Lustre is started. This was due to a filp leak introduced
3196 by one of our kernel patches.
3198 ------------------------------------------------------------------------------
3200 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3202 * Support for kernels:
3203 2.6.9-42.EL (RHEL 4)
3204 2.6.5-7.276 (SLES 9)
3205 2.4.21-40.EL (RHEL 3)
3206 2.6.12.6 vanilla (kernel.org)
3211 Bugzilla : 5719, 9635, 9792, 9684
3212 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3213 Details : If a server is under heavy load and cannot reply to new
3214 connection requests before the client resends the (re)connect,
3215 the connection handling code can behave badly if two service
3216 threads are concurrently handing separate (re)connections from
3217 the same client. Add better locking to the connection handling
3218 code, and ensure that only a single connection will be processed
3219 for a given client UUID, even if the lock is dropped.
3221 Severity : enhancement
3223 Description: add TCP zero-copy support to kernel
3224 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3225 sends if the hardware supports scatter-gather and checksumming.
3226 This allows socklnd to do client-write and server-read more
3227 efficiently and reduce CPU utilization from skbuf copying.
3230 Frequency : only if NFS exporting from client
3232 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3233 Details : If setting attributes on a file created under NFS that had
3234 never been opened it would be possible to oops the client
3235 if the file had no objects.
3238 Frequency : always for liblustre
3240 Description: liblustre client does MDS+OSTs setattr RPC for each write
3241 Details : When doing a write from a liblustre client, the client
3242 incorrectly issued an RPC to the MDS and each OST the file was
3243 striped over in order to update the timestamps. When writing
3244 with small chunks and many clients this could overwhelm the MDS
3245 with RPCs. In all cases it would slow down the write because
3246 these RPCs are unnecessary.
3248 Severity : enhancement
3250 Description: allow number of MDS service threads to be changed at module load
3251 Details : It is now possible to change the number of MDS service threads
3252 running. Adding "options mds mds_num_threads={N}" to the MDS's
3253 /etc/modprobe.conf will set the number of threads for the next
3254 time Lustre is restarted (assuming the "mds" module is also
3255 reloaded at that time). The default number of threads will
3256 stay the same, 32 for most systems.
3261 Description: OST crash if filesystem is unformatted or corrupt
3262 Details : If an OST is started on a device that has never been formatted
3263 or if the filesystem is corrupt and cannot even mount then the
3264 error handling cleanup routines would dereference a NULL pointer.
3269 Description: NULL pointer deref in llap_from_page.
3270 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3271 incorrect) mapping if the page was truncated/reclaimed while it was
3272 searched for. Check for this condition and skip such pages when
3273 doing readahead. Introduce extra check to llap_from_page() to
3274 verify page->mapping->host is non-NULL (so page is not anonymous).
3277 Frequency : Sometimes when using sys_sendfile
3279 Description: "page not covered by a lock" warnings from ll_readpage
3280 Details : sendfile called ll_readpage without right page locks present.
3281 Now we introduced ll_file_sendfile that does necessary locking
3282 around call to generic_file_sendfile() much like we do in
3286 Frequency : with certain MDS communication failures at client mount time
3288 Description: NULL pointer deref after failed client mount
3289 Details : a client connection request may delayed by the network layer
3290 and not be sent until after the PTLRPC layer has timed out the
3291 request. If the client fails the mount immediately it will try
3292 to clean up before the network times out the request. Add a
3293 reference from the request import to the obd device and delay
3294 the cleanup until the network drops the request.
3297 Frequency : occasionally during client (re)connect
3299 Description: assertion failure during client (re)connect
3300 Details : processing a client connection request may be delayed by the
3301 client or server longer than the client connect timeout. This
3302 causes the client to resend the connection request. If the
3303 original connection request is replied in this interval, the
3304 client may trip an assertion failure in ptlrpc_connect_interpret()
3305 which thought it would be the only running connect process.
3308 Frequency : only with obd_echo servers and clients that are rebooted
3310 Description: kernel BUG accessing uninitialized data structure
3311 Details : When running an obd_echo server it did not start the ping_evictor
3312 thread, and when a client was evicted an uninitialized data
3313 structure was accessed. Start the ping_evictor in the RPC
3314 service startup instead of the OBD startup.
3316 Severity : enhancement
3317 Bugzilla : 10193 (patchless)
3318 Description: Remove dependency on various unexported kernel interfaces.
3319 Details : No longer need reparent_to_init, exit_mm, exit_files,
3320 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3323 Frequency : rare (only users of deprecated and unsupported LDAP config)
3325 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3326 Details : LDAP apparently contains 'lustreName' attributes instead of
3327 'name'. A simple remapping of the name is sufficient.
3330 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3332 Description: waiting_locks_callback trips kernel BUG if client is evicted
3333 Details : Running with the dump_on_timeout debug flag turned on makes
3334 it possible that the waiting_locks_callback() can try to dump
3335 the Lustre kernel debug logs from an interrupt handler. Defer
3336 this log dumping to the expired_lock_main() thread.
3338 Severity : enhancement
3340 Description: Support NFS exporting on 2.6 kernels.
3341 Details : Implement non-rawops metadata methods for NFS server to use without
3342 changing NFS server code.
3345 Frequency : very rare (synthetic metadata workload only)
3347 Description: two racing renames might cause an MDS thread to deadlock
3348 Details : Running the "racer" program may cause one MDS thread to rename
3349 a file from being the source of a rename to being the target of
3350 a rename at exactly the same time that another thread is doing
3351 so, and the second thread has already enqueued these locks after
3352 doing a lookup of the target and is trying to relock them in
3353 order. Ensure that we don't try to re-lock the same resource.
3356 Frequency : only very large systems with liblustre clients
3358 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3359 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3360 all OSTs in parallel. This allows the eviction of stale liblustre
3361 clients to proceed much faster than if they were done in series,
3362 and also offers similar improvements for other set_info RPCs.
3367 Description: excessive CPU usage during initial read phase on client
3368 Details : During the initial read phase on a client, it would agressively
3369 retry readahead on the file, consuming too much CPU and impacting
3370 performance (since 1.4.5.8). Improve the readahead algorithm
3371 to avoid this, and also improve some other common cases (read
3372 of small files in particular, where "small" is files smaller than
3373 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3378 Description: MDS crash when receiving packet with unknown intent.
3379 Details : Do not LBUG in unknown intent case, just return -EFAULT
3381 Severity : enhancement
3382 Bugzilla : 9293, 9385
3383 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3384 Details : Do not serialize getattr (non-intent version) and statfs.
3387 Frequency : occasional, when OST network is overloaded/intermittent
3389 Description: client evicted by OST after bulk IO timeout
3390 Details : If a client sends a bulk IO request (read or write) the OST
3391 may evict the client if it is unresposive to its data GET/PUT
3392 request. This is incorrect if the network is overloaded (takes
3393 too long to transfer the RPC data) or dropped the OST GET/PUT
3394 request. There is no need to evict the client at all, since
3395 the pinger and/or lock callbacks will handle this, and the
3396 client can restart the bulk request.
3399 Frequency : Always when mmapping file with no objects
3401 Description: client crashes when mmapping file with no objects
3402 Details : Check that we actually have objects in a file before doing any
3403 operations on objects in ll_vm_open, ll_vm_close and
3409 Description: Request leak when working with deleted CWD
3410 Details : Introduce advanced request refcount tracking for requests
3411 referenced from lustre intent.
3413 Severity : Enhancement
3415 Description: Cache open file handles on client.
3416 Details : MDS now will return special lock along with openhandle, if
3417 requested and client is allowed to hold openhandle, even if unused,
3418 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3419 closing files for every read/write openration.
3421 Severity : Enhancement
3423 Description: Cache open negative dentries on client when possible.
3424 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3425 negative dentries on lock revocation.
3430 Description: Remounting a client read-only wasn't possible with a zconf mount
3431 Details : It wasn't possible to remount a client read-only with llmount.
3433 Severity : enhancement
3434 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3435 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3436 MPICH is included to add Lustre-specific ADIO interfaces.
3437 This is based closely on the UFS ADIO layer and only differs
3438 in file creation, in order to allow the OST striping to be set.
3439 This is user-contributed code and not supported by CFS.
3444 Description: extended inode attributes (immutable, append-only) work improperly
3445 when 2.4 and 2.6 kernels are used on client/server or vice versa
3446 Details : Introduce kernel-independent values for these flags.
3448 Severity : enhancement
3451 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3452 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3453 granularity. Now, Fractional MB granularity is be supported,
3454 this is very useful for low memory system.
3456 Severity : enhancement
3458 Description: Getattr by fid
3459 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3460 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3461 have locks and avoids getattr rpc requests for it on every stat.
3464 Frequency : Always, for filesystems larger than 2TB
3466 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3467 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3468 filesystems larger than 2TB to be mounted with mballoc enabled.
3471 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3473 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3474 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3475 filesystems larger than 2TB is mounted then the kernel will
3476 silently corrupt the start of the filesystem. CONFIG_LBD is
3477 enabled for all CFS-supported kernels, but the possibility of
3478 this happening with a modified kernel config exists.
3480 Severity : enhancement
3482 Description: add client O_DIRECT support for 2.6 kernels
3483 Details : It is now possible to do O_DIRECT reads and writes to files
3484 in the Lustre client mountpoint on 2.6 kernel clients.
3486 Severity : enhancement
3488 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3489 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3490 parallel, not waiting for response from every OST before sending
3491 a rpc to the next OST.
3496 Description: setattr vs write race when updating file timestamps
3497 Details : Client processes that update a file timestamp into the past
3498 right after writing to the file (e.g. tar) it is possible that
3499 the updated file modification time can be reset to the current
3500 time due to a race between processing the setattr and write RPC.
3502 Severity : enhancement
3504 Description: Bring 'lfs find' closer in line with regular Linux find.
3505 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3506 -print0 options and obtains all the needed info through the lustre
3509 Severity : enhancement
3511 Description: support up to 1024 configured devices on one node
3512 Details : change obd_dev array from statically allocated to dynamically
3513 allocated structs as they are first used to reduce memory usage
3518 Description: Flush dirty partially truncated pages during truncate
3519 Details : Immediatelly flush partially truncated pages in filter_setattr,
3520 this way we completely avoid having any pages in page cache on OST
3521 and can retire ugly workarounds during writes to flush such pages.
3526 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3527 Details : protect inode from truncation within vfs_unlink() context
3528 just take a reference before calling vfs_unlink() and release it
3529 when parent's i_sem is free.
3534 Description: last_id value checked outside lock on OST caused LASSERT failure
3535 Details : If there were multiple MDS->OST object precreate requests in
3536 flight, it was possible that the OST's last object id was checked
3537 outside a lock and incorrectly tripped an assertion. Move checks
3538 inside locks, and discard old precreate requests.
3541 Frequency : always, if extents are used on OSTs
3543 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3544 Details : OSTs using the extents format would not zero the high 16 bits of
3545 the index physical block number. This is not a problem for any
3546 OST filesystems smaller than 16TB, and no kernels support ext3
3547 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3548 new/modified files) and can be fixed for existing filesystems
3549 with e2fsprogs-1.39-cfs1.
3554 Description: import connection selection may be incorrect if timer wraps
3555 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3556 import connections to be ignored if the 32-bit jiffies counter
3557 wraps. Use a 64-bit jiffies counter.
3560 Frequency : very large clusters immediately after boot
3562 Description: LNET request buffers exhausted under heavy short-term load
3563 Details : If a large number of client requests are generated on a service
3564 that has previously never seen so many requests it is possible
3565 that the request buffer growth cannot keep up with the spike in
3566 demand. Instead of dropping incoming requests, they are held in
3567 the LND until the RPC service can accept more requests.
3570 Frequency : Sometimes during replay
3572 Description: Assertion failure in ll_local_open after replay.
3573 Details : If replay happened on an open request reply before we were able
3574 to set replay handler, reply will become not swabbed tripping the
3575 assertion in ll_local_open. Now we set the handler right after
3576 recognising of open request
3579 Frequency : very rare
3581 Description: kernel reports "badness in vsnprintf"
3582 Details : Reading from the "recovery_status" /proc file in small chunks
3583 may cause a negative length in lprocfs_obd_rd_recovery_status()
3584 call to vsnprintf() (which is otherwise harmless). Exit early
3585 if there is no more space in the output buffer.
3587 Severity : enhancement
3589 Description: clear OBD RPC statistics by writing to them
3590 Details : It is now possible to clear the OBD RPC statistics by writing
3591 to the "stats" file.
3596 Description: Client mtime is not the same on different clients after utimes
3597 Details : In some cases, the client was using the utimes() syscall on
3598 a file cached on another node. The clients now validate the
3599 ctime from the MDS + OSTs to determine which one is right.
3604 Description: Inability to activate failout mode
3605 Details : lconf script incorrectly assumed that in pythong string's numeric
3606 value is used in comparisons.
3609 Frequency : always with multiple stripes per file
3611 Description: Inefficient object allocation for mutli-stripe files
3612 Details : When selecting which OSTs to stripe files over, for files with
3613 a stripe count that divides evenly into the number of OSTs,
3614 the MDS is always picking the same starting OST for each file.
3615 Return the OST selection heuristic to the original design.
3620 Description: mount failures may take full timeout to return an error
3621 Details : Under some heavy load conditions it is possible that a
3622 failed mount can wait for the full obd_timeout interval,
3623 possibly several minutes, before reporting an error.
3624 Instead return an error as soon as the status is known.
3626 Frequency : quota enabled and large files being deleted
3628 Description: releasing more than 4GB of quota at once hangs OST
3629 Details : If a user deletes more than 4GB of files on a single OST it
3630 will cause the OST to spin in an infinite loop. Release
3631 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3636 Description: statfs data retrieved from /proc may be stale or zero
3637 Details : When reading per-device statfs data from /proc, in the
3638 {kbytes,files}_{total,free,avail} files, it may appear
3639 as zero or be out of date.
3642 Frequency : systems with MD RAID1 external journal devices
3644 Description: lconf's call to blkid is confused by RAID1 journal devices
3645 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3646 of returning all block devices that match the journal UUID.
3649 Frequency : always, for aggregate stripe size over 4GB
3651 Description: assertion fails when trying to use 4GB stripe size
3652 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3653 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3656 Frequency : always on ppc64
3658 Description: the first write on an ext3 filesystem with mballoc got stuck
3659 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3660 perform endianness conversion.
3662 ------------------------------------------------------------------------------
3664 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3666 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3667 INTEROPERATE with older versions automatically. Please read the
3668 user documentation before upgrading any part of a live system.
3669 * WARNING: Lustre networking configuration changes are required with
3670 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3673 * Support for kernels:
3674 2.6.9-22.0.2.EL (RHEL 4)
3675 2.6.5-7.244 (SLES 9)
3676 2.6.12.6 vanilla (kernel.org)
3679 Severity : enhancement
3680 Bugzilla : 7981/8208
3681 Description: Introduced Lustre Networking (LNET)
3682 Details : LNET is new networking infrastructure for Lustre, it includes
3683 a reorganized network configuration mode (see the user
3684 documentation for full details) as well as support for routing
3685 between different network fabrics. Lustre Networking Devices
3686 (LNDs) for the supported network fabrics have also been
3687 created for this new infrastructure.
3689 Severity : enhancement
3690 Description: Introduced Access control lists
3691 Details : clients can set ACLs on files and directories in order to have
3692 more fine-grained permissions than the standard Unix UGO+RWX.
3693 The MDS must be started with the "-o acl" mount option.
3695 Severity : enhancement
3696 Description: Introduced filesystem quotas
3697 Details : Administrators may now establish per-user quotas on the
3700 Severity : enhancement
3702 Description: Configuration change for the XT3
3703 The PTLLND is now used to run Lustre over Portals on the XT3
3704 The configure option(s) --with-cray-portals are no longer used.
3705 Rather --with-portals=<path-to-portals-includes> is used to
3706 enable building on the XT3. In addition to enable XT3 specific
3707 features the option --enable-cray-xt3 must be used.
3712 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3713 Details : A race between allocating a new llcd and re-getting the llcd_lock
3714 allowed another thread to grab newly-allocated llcd.
3716 Severity : enhancement
3718 Description: 2.6 OST async journal commit and locking fix to improve performance
3719 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3720 2.6 kernels are now async as they already were in 2.4 kernels so
3721 that they can commit concurrently with the network bulk transfer.
3722 For block-allocated files the filter allocation semaphore is held
3723 to avoid filesystem fragmentation during allocation. BKL lock
3724 removed for 2.6 xattr operations where it is no longer needed.
3729 Description: lconf incorrectly determined whether two IP networks could talk
3730 Details : In some more complicated routing and multiple-network
3731 configurations, lconf will avoid trying to make a network
3732 connection to a disjoint part of the IP space. It was doing the
3733 math incorrectly for one set of cases.
3738 Description: Fix for potential infinite loop processing records in an llog.
3739 Details : If an llog record is corrupted/zeroed, it is possible to loop
3740 forever in llog_process(). Validate the llog record length
3741 and skip the remainder of the block on error.
3744 Frequency : occasional (liblustre only)
3746 Description: liblustre could not open files whose last component is a symlink
3747 Details : sysio_path_walk() would incorrectly pass the open intent to
3748 intermediate path components.
3751 Frequency : rare (liblustre only with non-standard tuning)
3752 Bugzilla : 7201 (7350)
3753 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3754 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3755 after it has been released, only do this for referenced locks
3757 Severity : enhancement
3759 Description: specifying an (invalid) directory default stripe_size of -1
3760 would reset the directory default striping
3761 Details : stripe_size -1 was used internally to signal directory stripe
3762 removal, now use "all default" to signal dir stripe removal
3763 as a directory striping of "all default" is not useful
3766 Frequency : common for large clusters running liblustre clients
3768 Description: doing an ls when liblustre clients are running is slow
3769 Details : sending a glimpse AST to a liblustre client waits for every AST
3770 to time out, as liblustre clients will not respond. Since they
3771 cannot cache data we refresh the OST lock LVB from disk instead.
3773 Severity : enhancement
3775 Description: doing an ls at the same time as file IO can be slow
3776 Details : enqueue and other "small" requests can be blocked behind many
3777 large IO requests. Create a new OST IO portal for non-IO
3778 requests so they can be processed faster.
3781 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3783 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3784 created file with mode 000
3785 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3786 allow an ftruncate() as a non-root user to a file with mode 000.
3787 Lustre now respects this flag to disable mode checks when
3788 truncating a file owned by the user
3791 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3794 Description: Revoking locks from clients that went dead or catatonic might take
3796 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3797 cancellation of such locks instant on servers without waiting for
3798 any reply from clients. Clients drops these locks when cancel
3799 notification from server is received without replying.
3802 Frequency : liblustre-only, when liblustre client dies or becomes busy
3804 Description: Doing ls on Linux clients can take a long time with active
3806 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3807 granting such locks to it in the first place if possible. Locks
3808 are taken by proxy on the OST during the read or write and
3809 dropped immediately afterward. Add connect flags handling, do
3810 not grant locks to liblustre clients for glimpse ASTs.
3812 Severity : enhancement
3814 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3815 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3816 sequential reads and behaves badly with applications doing random
3817 reads. Improve it by reading ahead at least read region, and
3818 avoiding excessive large RPC for small reads.
3820 Severity : enhancement
3822 Description: Creating more than 1000 files for a single job may cause a load
3823 imbalance on the OSTs if there are also a large number of OSTs.
3824 Details : qos_prep_create() uses an OST index reseed value that is an
3825 even multiple of the number of available OSTs so that if the
3826 reseed happens in the middle of the object allocation it will
3827 still utilize the OSTs as uniformly as possible.
3832 Description: OST or MDS may oops in ping_evictor_main()
3833 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3834 but doesn't restart at beginning of obd_exports_timed list
3837 Severity : enhancement
3839 Description: improve by-nid export eviction on the MDS and OST
3840 Details : allow multiple exports with the same NID to be evicted at one
3841 time without re-searching the exports list.
3844 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3846 Description: MDS may oops in groups_free()
3847 Details : in rare race conditions a newly allocated group_info struct is
3848 freed again, and this can be NULL. The 2.4 compatibility code
3849 for groups_free() checked for a NULL pointer, but 2.6 did not.
3852 Frequency : common for liblustre clients doing little filesystem IO
3853 Bugzilla : 9352, 7313
3854 Description: server may evict liblustre clients accessing contended locks
3855 Details : if a client is granted a lock or receives a completion AST
3856 with a blocking AST already set it would not reply to the AST
3857 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3860 Frequency : lfs setstripe, only systems with more than 160 OSTs
3862 Description: unable to set striping with a starting offset beyond OST 160
3863 Details : llapi_create_file() incorrectly limited the starting stripe
3864 index to the maximum single-file stripe count.
3867 Frequency : LDAP users only
3869 Description: lconf did not handle in-kernel recovery with LDAP properly
3870 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3872 Severity : enhancement
3874 Description: bind OST threads to NUMA nodes to improve performance
3875 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3876 node and do their allocations there to localize memory access
3878 Severity : enhancement
3880 Description: llmount can determine client NID directly from Myrinet (GM)
3881 Details : the client NID code from gmnalnid was moved directly into
3882 llmount, removing the need to use this or specifying the
3883 client NID explicitly when mounting GM clients with zeroconf
3886 Frequency : if client is started with down MDS
3888 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3889 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3890 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3891 ptlrpc_import_delayed_req() because we didn't check if the
3892 request was interrupted, and we also didn't break out of the
3893 event loop if there was no timeout
3898 Description: data loss during non-page-aligned writes to a single file from
3899 both multiple nodes and multiple threads on one node at same time
3900 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3901 inconsistency led to false short-reads, that were cached and later
3902 used by ->prepare_write() to fill in partially written page,
3903 leading to data loss.
3906 Frequency : always, if lconf --abort_recovery used
3908 Description: lconf --abort_recovery fails with 'Operation not supported'
3909 Details : lconf was attempting to abort recovery on the MDT device and not
3912 Severity : enhancement
3914 Description: remove cleanup logs
3915 Details : replace lconf-generated cleanup logs with lustre internal
3916 cleanup routines. Eliminates the need for client-cleanup and
3919 Severity : enhancement
3921 Description: add support for EAs (user and system) on lustre filesystems
3922 Details : it is now possible to store extended attributes in the Lustre
3923 client filesystem, and with the user_xattr mount option it
3924 is possible to allow users to store EAs on their files also
3926 Severity : enhancement
3928 Description: Add possibility (config option) to show minimal available OST free
3930 Details : When compiled with --enable-mindf configure option, statfs(2)
3931 (and so, df) will return least minimal free space available from
3932 all OSTs as amount of free space on FS, instead of summary of
3933 free spaces of all OSTs.
3935 Severity : enhancement
3937 Description: do not expand extent locks acquired on OST-side
3938 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3939 by server: they are not cached anyway.
3942 Frequency : when mmap is used/binaries executed from Lustre
3944 Description: Unmmap pages before throwing them away from read cache.
3945 Details : llap_shrink cache now attempts to unmap pages before discarding
3946 them (if unmapping failed - do not discard). SLES9 kernel has
3947 extra checks that trigger if this unmapping is not done first.
3952 Description: lconf didn't resolve symlinks before checking to see whether a
3953 given mountpoint was already in use
3956 Frequency : when migrating failover services
3957 Bugzilla : 6395, 9514
3958 Description: When migrating a subset of services from a node (e.g. failback
3959 from a failover service node) the remaining services would
3960 time out and evict clients.
3961 Details : lconf --force (implied by --failover) sets the global obd_timeout
3962 to 5 seconds in order to quickly disconnect, but this caused
3963 other RPCs to time out too quickly. Do not change the global
3964 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3966 Severity : enhancement
3967 Frequency : if MDS is started with down OST
3968 Bugzilla : 9439,5706
3969 Description: Allow startup/shutdown of an MDS without depending on the
3970 availability of the OSTs.
3971 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3972 Add appropriate locking and lov-osc refcounts for safe
3973 cleaning. Add osc abort_inflight calls in case the
3974 synchronize never started.
3977 Frequency : occasional (Cray XT3 only)
3979 Description: root not authorized to access files in CRAY_PORTALS environment
3980 Details : The client process capabilities were not honoured on the MDS in
3981 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3982 been accessed by an authorized user then root was able to access
3983 the file on the local client also. The root user capabilities
3984 are now allowed on the MDS, as this environment has secure UID.
3987 Frequency : occasional
3989 Description: ldiskfs "too long searching" message happens too often
3990 Details : A debugging message (otherwise harmless) prints too often on
3991 the OST console. This has been reduced to only happen when
3992 there are fragmentation problems on the filesystem.
3997 Description: Division by zero in statfs when all OSCs are inactive
3998 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3999 lov_statfs divides by value returned by lov_get_stripecnt().
4003 Bugzilla : 9489, 3273
4004 Description: First write from each client to each OST was only 4kB in size,
4005 to initialize client writeback cache, which caused sub-optimal
4006 RPCs and poor layout on disk for the first writen file.
4007 Details : Clients now request an initial cache grant at (re)connect time
4008 and so that they can start streaming writes to the cache right
4009 away and always do full-sized RPCs if there is enough data.
4010 If the OST is rebooted the client also re-establishes its grant
4011 so that client cached writes will be honoured under the grant.
4016 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4017 Details : Now I/O RPCs go to different portal number and (presumably) fast
4018 lock requests (and glimses) and other RPCs get their own service
4019 threads pool that should be able to service those RPCs
4022 Severity : enhancement
4024 Description: Ability to exchange lustre version between client and servers and
4025 issue warnings at client side if client is too old. Also for
4026 liblustre clients there is ability to refuse connection of too old
4028 Details : New 'version' field is added to connect data structure that is
4029 filled with version info. That info is later checked by server and
4033 Frequency : rare, liblustre only.
4034 Bugzilla : 9296, 9581
4035 Description: Two simultaneous writes from liblustre at offset within same page
4036 might proceed at the same time overwriting eachother with stale
4038 Details : I/O lock withing llu_file_prwv was released too early, before data
4039 actually was hitting the wire. Extended lock-holding time until
4040 server acknowledges receiving data.
4043 Frequency : extremely rare. Never observed in practice.
4045 Description: avoid generating lustre_handle cookie of 0.
4046 Details : class_handle_hash() generates handle cookies by incrementing
4047 global counter, and can hit 0 occasionaly (this is unlikely, but
4048 not impossible, because initial value of cookie counter is
4049 selected randonly). Value of 0 is used as a sentinel meaning
4050 "unassigned handle" --- avoid it. Also coalesce two critical
4051 sections in this function into one.
4053 Severity : enhancement
4055 Description: allow liblustre clients to delegate truncate locking to OST
4056 Details : To avoid overhead of locking, liblustre client instructs OST to
4057 take extent lock in ost_punch() on client's behalf. New connection
4058 flag is added to handle backward compatibility.
4060 Severity : enhancement
4061 Bugzilla : 4928, 7341, 9758
4062 Description: allow number of OST service threads to be specified
4063 Details : a module parameter allows the number of OST service threads
4064 to be specified via "options ost ost_num_threads={N}" in the
4065 OSS's /etc/modules.conf or /etc/modprobe.conf.
4069 Bugzilla : 6146, 9635, 9895
4070 Description: servers crash with bad pointer in target_handle_connect()
4071 Details : In rare cases when a client is reconnecting it was possible that
4072 the connection request was the last reference for that export.
4073 We would temporarily drop the export reference and get a new
4074 one, but this may have been the last reference and the export
4075 was just destroyed. Get new reference before dropping old one.
4077 Severity : enhancement
4078 Frequency : if client is started with failover MDS
4080 Description: Allow multiple MDS hostnames in the mount command
4081 Details : Try to read the configuration from all specified MDS
4082 hostnames during a client mount in case the "primary"
4085 Severity : enhancement
4087 Description: Stop sending data to evicted clients as soon as possible.
4088 Details : Check if the client we are about to send or are sending data to
4089 was evicted already. (Check is done every second of waiting,
4090 for which l_wait_event interface was extended to allow checking
4091 of exit condition at specified intervals).
4094 Frequency : rare, normally only when NFS exporting is done from client
4096 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4098 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4101 Frequency : always, if srand() is called before liblustre initialization
4103 Description: Liblustre uses system PRNG disturbing its usage by user application
4104 Details : Introduce internal to lustre fast and high-quality PRNG for
4105 lustre usage and make liblustre and some other places in generic
4106 lustre code to use it.
4108 Severity : enhancement
4109 Bugzilla : 9477, 9557, 9870
4110 Description: Verify that the MDS configuration logs are updated when xml is
4111 Details : Check if the .xml configuration logs are newer than the config
4112 logs stored on the MDS and report an error if this is the case.
4113 Request --write-conf, or allow starting with --old_conf.
4115 Severity : enhancement
4117 Description: Handle symlinks in the path when checking if Lustre is mounted.
4118 Details : Resolve intermediate symlinks when checking if a client has
4119 mounted a filesystem to avoid duplicate client mounts.
4124 Description: lconf can hit an error exception but still return success.
4125 Details : The lconf command catches the Command error exception at the top
4126 level script context and will exit with the associated exit
4127 status, but doesn't ensure that this exit status is non-zero.
4132 Description: failure of ptlrpc thread startup can cause oops
4133 Details : Starting a ptlrpc service thread can fail if there are a large
4134 number of threads or the server memory is very fragmented.
4135 Handle this without oopsing.
4138 Frequency : always, only if liblustre and non-default acceptor port was used
4140 Description: liblustre cannot connect to servers with non-default acceptor port
4141 Details : tcpnal_set_default_params() was not called and was therefore
4142 ignoring the environment varaible TCPNAL_PORT, as well as other
4143 TCPNAL_ environment variables
4148 Description: two objects could be created on the same OST for a single file
4149 Details : If an OST is down, in some cases it was possible to create two
4150 objects on a single OST for a single file. No problems other
4151 than potential performance impact and spurious error messages.
4155 Bugzilla : 5681, 9562
4156 Description: Client may oops in ll_unhash_aliases
4157 Details : Client dcache may become inconsistent in race condition.
4158 In some cases "getcwd" can fail if the current directory is
4164 Description: Inode refcounting problems in NFS export code
4165 Details : link_raw functions used to call d_instantiate without obtaining
4166 extra inode reference first.
4170 Bugzilla : 9942, 9903
4171 Description: Referencing freed requests leading to crash, memleaks with NFS.
4172 Details : We used to require that call to ll_revalidate_it was always
4173 followed by ll_lookup_it. Also with revalidate_special() it is
4174 possible to call ll_revalidate_it() twice for the same dentry
4175 even if first occurence returned success. This fix changes semantic
4176 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4177 reference on a request referred from the intent.
4178 ll_intent_release() then releases such a request.
4181 Frequency : rare, normally benchmark loads only
4183 Description: unlinked inodes were kept in memory on the client
4184 Details : If a client is repeatedly creating and unlinking files it
4185 can accumulate a lot of stale inodes in the inode slab cache.
4186 If there is no other client load running this can cause the
4187 client node to run out of memory. Instead flush old inodes
4188 from client cache that have the same inode number as a new inode.
4191 Frequency : SLES9 2.6.5 kernel and long filenames only
4192 Bugzilla : 9969, 10379
4193 Description: utime reports stale NFS file handle
4194 Details : SLES9 uses out-of-dentry names in some cases, which confused
4195 the lustre dentry revalidation. Change it to always use the
4199 Frequency : rare, unless heavy write-truncate concurrency is continuous
4200 Bugzilla : 4180, 6984, 7171, 9963, 9331
4201 Description: OST becomes very slow and/or deadlocked during object unlink
4202 Details : filter_destroy() was holding onto the parent directory lock
4203 while truncating+unlinking objects. For very large objects this
4204 may block other threads for a long time and slow overall OST
4205 responsiveness. It may also be possible to get a lock ordering
4206 deadlock in this case, or run out of journal credits because of
4207 the combined truncate+unlink. Solution is to do object truncate
4208 first in one transaction without parent lock, and then do the
4209 final unlink in a new transaction with the parent lock. This
4210 reduces the lock hold time dramatically.
4213 Frequency : rare, 2.4 kernels only
4215 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4216 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4217 before unmounting their filesystem in order to avoid deadlock.
4218 The kernel_locked() function in 2.4 kernels only checks whether
4219 the kernel lock is held, not whether it is this process that is
4220 holding it as 2.6 kernels do.
4225 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4226 Details : The client ptlrpc code may be trying to reconnect to a down
4227 server before a previous connection attempt has timed out.
4228 Increase the reconnect interval to be longer than the connection
4229 timeout interval to avoid sending duplicate connections to
4233 Frequency : echo_client brw_test command
4235 Description: fix echo_client to work with OST preallocated code
4236 Details : OST preallocation code (5137) didn't take echo_client IO path
4237 into account: echo_client calls filter methods outside of any
4238 OST thread and, hence, there is no per-thread preallocated
4239 pages and buffers to use. Solution: hijack pga pages for IO. As
4240 a byproduct, this avoids unnecessary data copying.
4244 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4245 Description: Client can oops in mdc_commit_close() after open replay
4246 Details : It was possible for the MDS to return an open request with no
4247 transaction number in mds_finish_transno() if the client was
4248 evicted, but without actually returning an error. Clients
4249 would later try to replay that open and may trip an assertion
4250 Simplify the client close codepath, and always return an error
4251 from the MDS in case the open is not successful.
4254 Frequency : rare, 2.6 OSTs only
4256 Description: OST may deadlock under high load on fragmented files
4257 Details : If there was a heavy load and highly-fragmented OST filesystems
4258 it was possible to have all the OST threads deadlock waiting on
4259 allocation of biovecs, because the biovecs were not released
4260 until the entire RPC IO was completed. Instead, release biovecs
4261 as soon as they are complete to ensure forward IO progress.
4263 Severity : enhancement
4265 Description: Support for specifying external journal device at mount
4266 Details : If an OST or MDS device is formatted with an external journal
4267 device, this device major/minor is stored in the ext3 superblock
4268 and may not be valid for failover. Allow detecting and
4269 specifying the external journal at mount time.
4274 Description: Mounting an MDS with pending unlinked files may cause oops
4275 Details : target_finish_recovery() calls mds_postrecov() which returned
4276 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4277 considers this an error and immediately begins cleaning up the
4278 lov, just after starting the mds_lov process
4280 Severity : enhancement
4282 Description: Implement 'lfs df' to report actual free space on per-OST basis
4283 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4284 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4285 usage of objects; '-h' to report in human readable format.
4287 ------------------------------------------------------------------------------
4289 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4296 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4297 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4298 at startup during orphan unlinking. Instead, skip buddy update
4299 before setup, as it will be regenerated after recovery is complete.
4304 Description: If an OST is inactive, its locks might reference stale inodes.
4305 Details : lov_change_cbdata() must iterate over all namespaces, even if
4306 they are inactive to clear inode references from the lock.
4308 Severity : enhancement
4309 Frequency : occasional, if non-standard max_dirty_mb used
4311 Description: Client will block write RPCs if not enough grant
4312 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4313 then the client will block writes while waiting for another RPC
4314 to complete instead of consuming its dirty limit. With change
4315 we get improved performance when max_dirty_mb is small.
4317 Severity : enhancement
4318 Bugzilla : 3389, 6253
4319 Description: Add support for supplementary groups on the MDS.
4320 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4321 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4322 lookups for user supplementary groups into a cache.
4326 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4327 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4328 for liblustre O_EXCL is enforced by MDS.
4332 Description: readdir never returns NULL in liblustre.
4333 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4334 dentry in struct dirent.
4338 Frequency : liblustre only, depends on application IO pattern
4339 Description: liblustre clients evicted if not contacting servers
4340 Details : Don't put liblustre clients into the ping_evictor list, so
4341 they will not be evicted by the pinger ever.
4343 Severity : enhancement
4345 Description: Add ability to evict clients by NID from MDS.
4346 Details : By echoing "nid:$NID" string into
4347 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4348 $NID would be instantly evicted from this MDS and from all active
4349 OSTs connected to it.
4353 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4354 Details : lookup_it_finish() used to query file size from OSTs that was not
4359 Description: service threads change working directory to that of init
4360 Details : Starting lustre service threads may pin the working directory
4361 of the parent thread, making that filesystem busy. Threads
4362 now change to the working directory of init to avoid this.
4366 Frequency : during shutdown only
4367 Description: shutdown with a failed MDS or OST can cause unmount to hang
4368 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4373 Frequency : 2.6 only
4374 Description: chmod/chown may include an extra supplementary group
4375 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4376 supplementary group and if none is specified this is used.
4379 Bugzilla : 5479 (6816)
4381 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4382 Details : If lookup is in progress on a file that is unlinked we might try
4383 to revalidate the inode and fail in revalidate after lookup is
4384 complete and ll_file_open() enqueues the open again but
4385 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4388 Frequency : always, if lconf --abort_recovery used
4390 Description: lconf --abort_recovery fails with 'Operation not supported'
4391 Details : lconf was attempting to abort recovery on the MDT device and not
4394 ------------------------------------------------------------------------------
4396 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4401 Frequency : rare (only unsupported configurations with a node running as an
4403 Bugzilla : 6514, 5137
4404 Description: Mounting a Lustre file system on a node running as an OST could
4406 Details : OSTs now preallocates memory needed to write out data at
4407 startup, instead of when needed, to avoid having to
4408 allocate memory in possibly low memory situations.
4409 Specifically, if the file system is mounted on on OST,
4410 memory pressure could force it to try to write out data,
4411 which it needed to allocate memory to do. Due to the low
4412 memory, it would be unable to do so and the node would
4413 become unresponsive.
4415 Severity : enhancement
4417 Description: Addition of lconf --service command line option
4418 Details : lconf now accepts a '--service <arg>' option, which is
4419 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4421 Severity : enhancement
4423 Description: Failover mode is now the default for OSTs.
4424 Details : By default, OSTs will now run in failover mode. To return to
4425 the old behaviour, add '--failout' to the lmc line for OSTs.
4427 Severity : enhancement
4429 Description: Health checks are now provided for MDS and OSTs
4430 Details : Additional detailed health check information on MSD and OSTs
4431 is now provided through the procfs health_check value.
4434 Frequency : occasional, depends on IO load
4436 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4437 after numerous create/delete cycles
4438 Details : The ext3 inode allocation policy would not allocate new inodes
4439 very well on the OSTs because there are no new directories
4440 being created. Instead we look for groups with free space if
4441 the parent directories are nearly full.
4446 Description: Network or server problems during mount may cause partially
4447 mounted clients instead of returning an error.
4448 Details : The config llog parsing code may overwrite the error return
4449 code during mount error handling, returning success instead
4455 Description: MDS can fail to allocate large reply buffers
4456 Details : After long uptimes the MDS can fail to allocate large reply
4457 buffers (e.g. zconf client mount config records) due to memory
4458 fragmentation or consumption by the buffer cache. Preallocate
4459 some large reply buffers so that these replies can be sent even
4460 under memory pressure.
4464 Frequency : rare (liblustre)
4465 Description: fsx running with liblustre complained that using truncate() to
4466 extend the file doesn't work. This patch corrects that issue.
4467 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4468 ATTR_SIZE and lsm use in llu_setattr_raw.
4472 Frequency : rare, only 2.6 kernels
4473 Description: Unusual file access patterns on the MDS may result in inode
4474 data being lost in very rare circumstances.
4475 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4476 code in the 2.6 kernel caused the inode and/or EA data not to
4477 be read from disk, causing single-file corruption.
4481 Frequency : rare, only 2.6 filesystems using extents
4482 Description: Heavy concurrent write and delete load may cause data corruption.
4483 Details : It was possible under high-load situations to have an extent
4484 metadata block in the block device cache from a just-unlinked
4485 file overwrite a newly-allocated data block. We now unmap any
4486 metadata buffers that alias just-allocated data blocks.
4490 Frequency : filesystems with default stripe_count larger than 77
4491 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4492 Details : lconf specifies an inode size of 4096 bytes when the default
4493 stripe_count is larger than 77. This conflicts with the default
4494 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4495 this case to avoid pinning too much memory for large EAs.
4497 ------------------------------------------------------------------------------
4499 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4504 Frequency : rare (extremely heavy IO load with hundreds of clients)
4506 Description: Client is evicted, gets IO error writing to file
4507 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4508 caused clients to be evicted for AST timeouts. The fixes in
4509 bug 5192 mean we no longer need to have such short AST timeouts
4510 so ldlm_timeout has been increased.
4513 Frequency : occasional during --force or --failover shutdown under load
4514 Bugzilla : 5949, 4834
4515 Description: Server oops/LBUG if stopped with --force or --failover under load
4516 Details : a collection of import/export refcount and cleanup ordering
4517 issues fixed for safer force cleanup
4520 Frequency : only filesystems larger than 120 OSTs
4521 Bugzilla : 5990, 6223
4522 Description: lfs getstripe would oops on a very large filesystem
4523 Details : lov_getconfig used kfree on vmalloc'd memory
4526 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4527 Bugzilla : 6242, 6243
4528 Description: reading from files that had been truncated to a non-zero size
4529 but never opened returned no data
4530 Details : ll_file_read() reads zeros from no-object files to EOF
4535 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4536 Details : There's little harm in aborting MDS/OSS recovery and letting it
4537 try again, so I removed the LASSERT and return an error instead.
4539 Severity : enhancement
4541 Description: New debugging infrastructure for tracking down data corruption
4542 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4543 (b) cover more of the client-side code path, and (c) try to narrow
4544 down where problems occurred
4548 Bugzilla : 3819, 4364, 4397, 6313
4549 Description: Racing close and eviction MDS could cause assertion in mds_close
4550 Details : It was possible to get multiple mfd references during close and
4551 client eviction, leading to one thread referencing a freed mfd.
4553 Severity: : enhancement
4554 Bugzilla : 3262, 6359
4555 Description: Attempts to reconnect to servers are now more aggressive.
4556 Details : This builds on the enhanced upcall-less recovery that was added
4557 in 1.4.2. When trying to reconnect to servers, clients will
4558 now try each server in the failover group every 10 seconds. By
4559 default, clients would previously try one server every 25 seconds.
4564 Description: After recovery, certain operations trigger a failed
4565 assertion on a client.
4566 Details : Failing over an mds, using lconf -d --failover, while a
4567 client was doing a readdir() call would cause the client to
4568 LBUG after recovery completed and the readdir() was resent.
4570 Severity : enhancement
4572 Description: Default groups are now added by lconf
4573 Details : You can now run lconf --group <servicename> without having to
4574 manually add groups with lmc.
4577 Frequency : occasional
4579 Description: Nodes with an elan id of 0 trigger a failed assertion
4582 Frequency : always when accessing e.g. tty/console device nodes
4584 Description: tty and some other devices nodes cannot be used on lustre
4585 Details : file's private_data field is used by device data and lustre
4586 values in there got lost. New field was added to struct file to
4587 store fs-specific private data.
4590 Frequency : when exporting Lustre via NFS
4592 Description: NFSD failed occasionally when looking up a path component
4593 Details : NFSD is looking up ".." which was broken in ext3 directories
4594 that had grown large enough to become hashed.
4597 Frequency : Clusters with multiple interfaces not on the same subnet
4599 Description: Nodes will repeatedly try to reconnect to an interface which it
4600 cannot reach and report an error to the log.
4601 Details : Extra peer list entries will be created by lconf with some peers
4602 unreachable. lconf now validates the peer before adding it.
4605 Frequency : Only if a default stripe is set on the filesystem root.
4607 Description: Setting a default stripe on the filesystem root prevented the
4608 filesystem from being remounted.
4609 Details : The client was sending extra request flags in the root getattr
4610 request and did not allocate a reply buffer for the dir EA.
4613 Frequency : occasional, higher if lots of files are accessed by one client
4614 Bugzilla : 6159, 6097
4615 Description: Client trips assertion regarding lsm mismatch/magic
4616 Details : While revalidating inodes the VFS looks up inodes with ifind()
4617 and in rare cases can find an inode that is being freed.
4618 The ll_test_inode() code will free the lsm during ifind()
4619 when it finds an existing inode and then the VFS later attaches
4620 this free lsm to a new inode.
4624 Bugzilla : 6422, 7030
4625 Description: MDS deadlock between mkdir and client eviction
4626 Details : Creating a new file via mkdir or mknod (starting a transaction
4627 and getting the ns lock) can deadlock with client eviction
4628 (gets ns lock and trying to finish a synchronous transaction).
4631 Frequency : occasional
4632 Description: While starting a server, the fsfilt_ext3 module could not be
4634 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4635 kernels. Previously, lconf would still use the ext3 name
4636 when trying to load modules. Now, it will correctly use
4637 ext3 on 2.4 and ldiskfs on 2.6.
4639 Severity : enhancement
4640 Description: The default stripe count has been changed to 1
4641 Details : The interpretation of the default stripe count (0, to lfs
4642 or lmc) has been changed to mean striping across a single
4643 OST, rather than all available. For general usage we have
4644 found a stripe count of 1 or 2 works best.
4646 Severity : enhancement
4647 Description: Add support for compiling against Cray portals.
4648 Details : Conditional compiling for some areas that are different
4652 Frequency : occasional
4653 Bugzilla : 6409, 6834
4654 Description: Creating files with an explicit stripe count may lead to
4655 a failed assertion on the MDS
4656 Details : If some OSTs are full or unavailable, creating files may
4657 trigger a failed assertion on the MDS. Now, Lustre will
4658 try to use other servers or return an error to the
4662 Frequency : occasional
4664 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4665 caused lock timeout during readahead (since 1.4.2).
4666 Details : Processes doing readahead might match a lock that hasn't been
4667 granted yet if there are overlapping and conflicting lock
4668 requests. The readahead process waits on ungranted lock
4669 (original lock is CBPENDING), while OST waits for that process
4670 to cancel CBPENDING read lock and eventually evicts client.
4672 Severity : enhancement
4674 Description: Initial enabling of flock support for clients
4675 Details : Implements fcntl advisory locking and file status functions.
4676 This feature is provided as an optional mount flag (default
4677 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4678 locking are implemented yet, and those that are are not guaranteed
4679 to be completely correct in production environments.
4680 mount -t lustre -o [flock|noflock] ...
4683 Frequency : occasional
4685 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4686 assertion in the ext3 JBD (journaling) layer.
4687 Details : The b_committed_data struct is protected by the big kernel lock
4688 in 2.4 kernels, serializing journal_commit_transaction() and
4689 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4690 access to this struct. In 2.6 kernels there is finer grained
4691 locking to improve SMP performance of the JBD layer.
4695 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4700 Description: merge and cleanup kernel patches.
4703 -----------------------------------------------------------------------------
4705 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4707 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4708 versions of Lustre. Please update all servers and clients to
4709 version 1.4.2 or later at the same time. You must also run
4710 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4711 to update the configuration logs.
4713 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4714 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4715 - swab llog records of type '0' so we get proper header size/idx (5861)
4716 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4717 - fix rename of one directory over another leaking an inode (5953)
4718 - avoid SetPageDirty on 2.6 (5981)
4719 - don't re-add just-being-destroyed locks to the waiting list (5653)
4720 - when creating new directories, inherit the parent's custom
4721 striping settings if present parent (3048)
4722 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4723 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4724 - don't hold client locks on temporary worklist from l_lru (5666)
4725 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4726 - automatically evict dead clients (5921)
4727 - Update file size properly in create+truncate+fstat case (6196)
4728 - Do not unhash mountpoint dentries, do not allow removal of
4730 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4731 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4732 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4733 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4734 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4735 - Correctly handle reads of files with no objects (6243)
4736 - lctl recover will also mark a device active if deactivate used (5933)
4738 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4739 - allow --write-conf on an MDS with different nettype than client (5619)
4740 - don't write config llogs to MDS for mounts not from that MDS (5617)
4741 - lconf should create multiple TCP connections from a client (5201)
4742 - init scripts are now turned off by default; run chkconfig --on
4743 lustre and chkconfig --on lustrefs to use them
4744 - upcalls are no longer needed for clients to recover to failover
4746 - add --abort-recovery option to lconf to abort recovery on device
4748 - add support for an arbitrary number of OSTs (3026)
4749 - Quota support protocol changes.
4750 - forward compatibility changes to wire structs (6007)
4751 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4752 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4753 - improved "lustre status" script
4754 - initialize blocksize for non-regular files (6062)
4755 - added --disable-server and --disable-client configure options (5782)
4756 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4757 - Vanilla 2.4.29 support
4758 - increase maximum number of obd devices to 520 (6242)
4759 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4760 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4761 SLES 9 2.6.5 kernels
4762 - limit stripes per file to 160 (the maximum EA size) (6093)
4764 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4767 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4768 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4769 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4770 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4771 - small ext3 extents cleanups and fixes (5733)
4772 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4773 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4774 - avoid needless client->OST connect, fix handle mismatch (5317)
4775 - fix DLM error path that led to out-of-sync client, long delays (5779)
4776 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4777 - fix several locking issues related to i_size (5492,5624,5654,5672)
4778 - don't move pending lock onto export if it is already evicted (5683)
4779 - fix kernel oops when creating .foo in unlinked directory (5548)
4780 - fix deadlock in obdfilter statistics vs. object create (5811)
4781 - use time_{before,after} to avoid timer jiffies wrap (5882)
4782 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4783 - Do not leak request if server was not able to process it (5154)
4784 - If mds_open unable to find parent dir, make that negative lookup(5154)
4785 - don't create new directories with extent-mapping (5909, 5936)
4787 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4788 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4789 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4790 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4791 - support for RHEL4 kernel on i686 (5773)
4792 - provide error messages when incompatible logs are encountered (5898)
4794 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4795 * version 1.4.0.10 (1.4.1 release candidate 1)
4797 - don't keep a lock reference when lock is not granted (4238)
4798 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4799 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4800 - drop import inflight refcount on signal_completed_replay error (5255)
4801 - unlock page after async write error during send (3677)
4802 - handle missing objects in filter_preprw_read properly (5265)
4803 - no transno return for symlink open, don't save no-trasno open (3440)
4804 - don't try to complete elan receive that already failed (4012)
4805 - free RPC server reply state on error (5406)
4806 - clean up thread from ptlrpc_start_thread() on error (5160)
4807 - readahead could read extra page into cache that wasn't ejected (5388)
4808 - prevent races in class_attach/setup/cleanup/detach (5260)
4809 - don't dereference de->d_inode after l_dput of de (5458)
4810 - use "int" for stripe value returned from lock_to_stripe (5544)
4811 - mballoc allocation and error-checking fixes in 2.6 (5504)
4812 - block device patches to fix I/O request sizes in 2.6 (5482)
4813 - look up hostnames for IB nals (5602)
4814 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4815 - don't start multiple acceptors for the same port (5277)
4816 - fix incorrect LASSERT in mds_getattr_name (5635)
4817 - export a proc file for general "ping" checking (5628)
4818 - fix "lfs check" to not block when the MDS is down (5628)
4820 - service request history (4965)
4821 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4822 - create an "evict_client" /proc entry on OSTs, like the MDS has
4823 - fix mount usage message, return errors per mount(8) (5168)
4824 - change grep [] to grep "[]" in tests so they work in more UMLs
4825 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4826 - remove ancient LOV_MAGIC_V0 EA support (5047)
4827 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4828 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4829 - allow readahead allocations to fail when low on memory (5383)
4830 - mmap locking landed again, after considerable improvement (2828)
4831 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4833 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4836 - send OST transaction number in read/write reply to free req (4966)
4837 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4838 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4839 - the watchdog thread now runs as interruptible (5246)
4840 - flock/lockf fixes (but it's still disabled, pending 5135)
4841 - don't use EXT3 constants in llite code (5094)
4842 - memory shortage at startup could cause assertion (5176)
4844 - reorganization of lov code
4845 - single portals codebase
4847 - add extents/mballoc support (5025)
4848 - direct I/O reads in the obdfilter (4048)
4849 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4851 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4854 - send OST transaction number in read/write reply to free req (4966)
4855 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4856 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4857 - the watchdog thread now runs as interruptible (5246)
4858 - handle missing objects in filter_preprw_read properly (5265)
4859 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4860 - drop import inflight refcount on signal_completed_replay error (5255)
4861 - unlock page after async write error during send (3677)
4862 - return original error code on reconstructed replies (3761)
4863 - no transno return for symlink open, don't save no-trasno open (3440)
4865 - add pid to ldlm debugging output (4922)
4866 - bump the watchdog timeouts -- we can't handle 30sec yet
4867 - extra debugging for orphan dentry/inode bug (5259)
4869 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4872 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4873 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4874 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4875 - fix NULL dereference in /proc/sys/portals/routes (4827)
4876 - allow failed mdc_close() operations to be interrupted (4561)
4877 - stop precreate on OST before MDS would time out on it (4778)
4878 - don't send partial-page writes before EOF from client (4410)
4879 - discard client grant for sub-page writes on large-page clients (4520)
4880 - don't free dentries not owned by NFS code, check generation (4806)
4881 - fix lsm leak if mds_create_objects() fails (4801)
4882 - limit debug_daemon file size, always print CERROR messages (4789)
4883 - use transno after validating reply (3892)
4884 - process timed out requests if import state changes (3754)
4885 - update mtime on OST during writes, return in glimpse (4829)
4886 - add mkfsoptions to LDAP (4679)
4887 - use ->max_readahead method instead of zapping global ra (5039)
4888 - don't interrupt __l_wait_event() during strace
4890 - add software watchdogs to catch hung threads quickly (4941)
4891 - make lustrefs init script start after nfs is mounted
4892 - fix CWARN/ERROR duplication (4930)
4893 - return async write errors to application if possible (2248)
4894 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4895 - print NAL number in %x format (4645)
4896 - update barely-supported suse-2.4.21-171 series (4842)
4897 - support for sles 9 %post scripts
4898 - support for building 2.6 kernel-source packages
4899 - support for sles km_* packages
4901 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4904 - ignore -ENOENT errors in osc_destroy (3639)
4905 - notify osc create thread that OSC is being cleaned up (4600)
4906 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4907 - reconstruct ost_handle() like mds_handle() (4657)
4908 - create a new thread to do import eviction to avoid deadlock (3969)
4909 - let lconf resolve symlinked-to devices (4629)
4910 - don't unlink "objects" from directory with default EA (4554)
4911 - hold socknal file ref over connect in case target is down (4394)
4912 - allow more than 32000 subdirectories in a single directory (3244)
4913 - fix blocks count for O_DIRECT writes (3751)
4914 - OST returns ENOSPC from object create when no space left (4539)
4915 - don't send truncate RPC if file size isn't changing (4410)
4916 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4917 - bind to privileged port in socknal and tcpnal (3689)
4919 - rate limit CERROR/CWARN console message to avoid overload (4519)
4920 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4921 - basic mmap support (3918)
4922 - kernel patch series update from b1_4 (4711)
4924 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4927 - avoid crash during MDS cleanup with OST shut down (2775)
4928 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4929 - don't use bad inodes on the MDS (3744)
4930 - dynamic object preallocation to improve recovery speed (4236)
4931 - don't hold spinlock over lock dumping or change debug flags (4401)
4932 - don't zero obd_dev when it is force cleaned (3651)
4933 - print grants to console if they go negative (4431)
4934 - "lctl deactivate" will stop automatic recovery attempts (3406)
4935 - look for existing locks in ldlm_handle_enqueue() (3764)
4936 - don't resolve lock handle twice in recovery avoiding race (4401)
4937 - revalidate should check working dir is a directory (4134)
4939 - don't always mark "slow" obdfilter messages as errors (4418)
4941 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4944 - don't close LustreDB during write_conf until it is done (3860)
4945 - fix typo in lconf for_each_profile (3821)
4946 - allow dumping logs from multiple threads at one time (3820)
4947 - don't allow multiple threads in OSC recovery (3812)
4948 - fix debug_size parameters (3864)
4949 - fix mds_postrecov to initialize import for llog ctxt (3121)
4950 - replace config semaphore with spinlock (3306)
4951 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4952 - don't allow enqueue to complete on a destroyed export (3822)
4953 - down write_lock before checking llog header bitmap (3825)
4954 - recover from lock replay timeout (3764)
4955 - up llog sem before sending rpc (3652)
4956 - reduce ns lock hold times when setting kms (3267)
4957 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4958 - fix NULL deref and obd_dev leak on setup error (3312)
4959 - replace some LBUG about llog ops with error handling (3841)
4960 - don't match INVALID dentries from d_lookup and spin (3784)
4961 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4962 - fix invalid assertion in ptlrpc_set_wait (3880)
4964 - add libwrap support for the TCP acceptor (3996)
4965 - add /proc/sys/portals/routes for non-root route listing (3994)
4966 - allow setting MDS UUID in .xml (2580)
4967 - print the stack of a process that LBUGs (4228)
4969 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4972 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4973 - make sure to unset replay flag from failed open requests (3440)
4974 - if default stripe count is 0, use OST count for inode size (3636)
4975 - update parent mtime/ctime on client for create/unlink (2611)
4976 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4977 - free recovery state on server during a forced cleanup (3571)
4978 - unregister_reply for resent reqs (3063)
4979 - loop back devices mounting and status check on 2.6 (3563)
4980 - fix resource-creation race that can provoke i_size == 0 (3513)
4981 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4982 - more debugging for page-accounting assertion (3746)
4983 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4984 - avoid deadlock after precreation failure (3758)
4985 - fix race and lock order deadlock in orphan handling (3450, 3750)
4986 - add validity checks when grabbing inodes from l_ast_data (3599)
4988 - add /proc/.../recovery_status to obdfilter (3428)
4989 - lightweight CDEBUG infrastructure, debug daemon (3668)
4990 - change default OSC RPC parameters to be better on small clusters
4991 - turn off OST read cache for files smaller than 32MB
4992 - install man pages and include them in rpms (3100)
4993 - add new init script for (un)mounting lustre filesystems (2593)
4994 - run chkconfig in %post for init scripts (3701)
4995 - drop scimac NAL (unmaintained)
4997 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5000 - clean kiobufs before and after use (3485)
5001 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5002 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5003 - change init script to fail more clearly if not run as root (1528)
5004 - allow clients to reconnect during replay (1742)
5005 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5006 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5007 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5008 - fix cond_resched (3554)
5009 - only evict unfinished clients after recovery (3515)
5010 - allow bulk resend, prevent data loss (3570)
5011 - dynamic ptlrpc request buffer allocation (2102)
5012 - don't allow unlinking open directory if it isn't empty (2904)
5013 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5014 - remove extraneous obd dereference causing LASSERT failure (3334)
5015 - don't use get_cycles() when creating temp. files on the mds (3156)
5016 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5017 - handle EEXIST for set-stripe, set proper directory name (3336)
5019 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5020 - fix ksocknal_fmb_callback() error messages (2918)
5022 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5025 - don't copy lvb into (possibly NULL) reply on error (2983)
5026 - don't deref dentry after dput, don't free lvb on error (2922)
5027 - use the kms to determine writeback rpc length (2947)
5028 - increment oti_logcookies when osc is inactive (2948)
5029 - update client's i_blocks count via lvb messages (2543)
5030 - handle intent open/close of special files properly (1557)
5031 - mount MDS with errors=remount-ro, like obdfilter (2009)
5032 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5033 - don't use cancelling-locks' kms values (2947)
5034 - use highest lock extent for kms, not last one (2925)
5035 - don't dereference ERR_PTR() dentry in error handling path (3107)
5036 - fix thread race in portals_debug_dumplog() (3122)
5037 - create lprocfs device entries at setup instead of at attach (1519)
5038 - common AST error handler, don't evict client on completion race (3145)
5039 - zero nameidata in detach_mnt in 2.6 (3118)
5040 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5041 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5042 - initialize RPC timeout timer earlier for 2.6 (3219)
5043 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5044 - print nal/nid for unknown nid (3258)
5045 - additional checks for oscc recovery before doing precreate (3284)
5046 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5047 - don't crash in mdc_close for bad permissions on open (3285)
5048 - zero i_rdev for non-device files (3147)
5049 - clear page->private before handing to FS, better assertion (3119)
5050 - tune the read pipeline (3236)
5051 - fix incorrect decref of invalidated dentry (2350)
5052 - provide read-ahead stats and refine rpc in flight stats (3328)
5053 - don't hold journal transaction open across create RPC (3313)
5054 - update atime on MDS at close time (3265)
5055 - close LDAP connection when recovering to avoid server load (3315)
5056 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5057 - don't leak open file on MDS after open resend (3325)
5058 - serialize filter_precreate and filter_destroy_precreated (3329)
5059 - loop device shouldn't call sync_dev() for nul device (3092)
5060 - clear page cache after eviction (2766)
5061 - resynchronize MDS->OST in background (2824)
5062 - refuse to mount the same filesystem twice on same mountpoint (3394)
5063 - allow llmount to create routes for mounting behind routers (3320)
5064 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5065 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5066 - fix rare problem with rename on htree directories (3417)
5068 - allow default OST striping configuration per directory (1414)
5069 - fix compilation for qswnal for 2.6 kernels (3125)
5070 - increase maximum number of MDS request buffers for large systems
5071 - change liblustreapi to be useful for external progs like lfsck (3098)
5072 - increase local configuration timeout for slow disks (3353)
5073 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5075 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5078 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5079 - don't overwrite extent policy data in reply if lock was blocked (2901)
5080 - drop filter export grants atomically with removal from device (2663)
5081 - del obd_self_export from work_list in class_disconnect_exports (2908)
5082 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5083 - swab reply message in mdc_close, other PPC fixes (2464)
5084 - fix destroying of named logs (2325)
5085 - overwrite old logs when running lconf --write_conf (2264)
5086 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5087 - fix race in target_handle_connect (2898)
5088 - mds_reint_create() should take same inode create lock (2926)
5089 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5090 - don't close files for self_export to avoid uninitialized obd (2936)
5091 - allow MDS with the same name as client node (2939)
5092 - hold dentry reference for closed log files for unlink (2325)
5093 - reserve space for all logs during transactions (2059)
5094 - don't evict page beyond end of stripe extent (2925)
5095 - don't oops on a deleted current working directory (2399)
5096 - handle hard links to targets without a parent properly (2517)
5097 - don't dereference NULL lock when racing during eviction (2867)
5098 - don't grow lock extents when lots of conflicting locks (2919)
5100 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5103 - account for cache space usage on clients to avoid data loss (974)
5104 - lfsck support in lustre kernel code (2349)
5105 - reduce journal credits needed for BRW writes (2370)
5106 - orphan handling to avoid losing space on client/server crashes
5107 - ptlrpcd can be blocked, stopping ALL progress (2477)
5108 - use lock value blocks to assist in proper KMS, faster stat (1021)
5109 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5110 - recovery for initial connections (2355)
5111 - fixes for mds_cleanup_orphans (1934)
5112 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5113 - block all file creations until orphan recovery completes (1901)
5114 - client remove rq_connection from request struct (2423)
5115 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5116 - recovery timer race (2670)
5117 - mdc_close recovey bug (2532)
5118 - ptlrpc cleanup bug (2710)
5119 - mds timeout on local locks (2588)
5120 - namespace lock held during RPCs (2431)
5121 - handle interrupted sync write properly (2503)
5122 - don't try to handle a message that hasn't been replied to (2699)
5123 - client assert failure during cleanup after abort recovery (2701)
5124 - leak mdc device after failed mount (2712)
5125 - ptlrpc_check_set allows timedout requests to complete (2714)
5126 - wait for inflight reqs when ptlrpcd finishes (2710)
5127 - make sure unregistered services are removed from the srv_list
5128 - reset bulk XID's when resending them (caught by 1138 test)
5129 - unregister_bulk after timeout
5130 - fix lconf error (2694)
5131 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5132 - readahead locks pages, leaves pending causing memory pressure (2673)
5133 - increase OST request buffers to 4096 on large machines (2729)
5134 - fix up permission of existing directories in simple_mkdir (2661)
5135 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5136 - don't assign transno to errored transactions (2742)
5137 - don't delete objects on OST if given a bogus objid from MDS (2751)
5138 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5139 - if rq_no_resend, then timeout request after recovery (2432)
5140 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5141 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5142 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5143 - MDS recovery shouldn't delete objects that it has given out (2730)
5144 - if enqueue arrives after completion, don't clobber LVB (2819)
5145 - don't unlock pages twice when trigger_group_io returns error (2814)
5146 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5147 - don't write pages to disk if there was an error (1450)
5148 - don't ping imports that have recovery disabled (2676)
5149 - take buffered bytes into account when balancing socknal conn (2817)
5150 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5151 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5152 - remove little-endian swabbing of llog records (1987)
5153 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5154 - retry reposting request buffers if they fail (1191)
5155 - grow extent at grant time to avoid granting a revoked lock (2809)
5156 - lock revoke doesn't evict page if covered by a second lock (2765)
5157 - disable VM readahead to avoid reading outside lock extents (2805)
5159 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5160 - updated kernel patches for hp-2.4.20 kernel (2681)
5162 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5165 - fix truncated write corruption (2366)
5166 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5168 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5169 - ensure len(uuid) < 37 in lmc (1171)
5170 - fix ia64 OOPS in llog_test (2255)
5171 - zero end of page at obdfilter for partial page writes (2648)
5172 - don't leave stale dentries around after renames (bug 2428)
5173 - fix timeouts when evicting a client with a single lock held (2642)
5174 - set deadline for the initial HELLO message to drain (2634)
5175 - print out dotted-quad IP addresses in the socknal (2302)
5177 - additional debugging for MDS client eviction problem (2443)
5178 - fix mkfsoptions support for osts (2603, 2604)
5180 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5183 - add series for the vanilla 2.6.0 kernel
5184 - add series for the vanilla 2.4.24 kernel
5185 - add series for a cray x86/64 UL kernel drop
5186 - fix xattr patches for the vanilla 2.4.19 series
5188 - generate true UUIDs in lmc (1171)
5189 - have portals stack dumping break in UML (2466)
5190 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5191 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5192 - wake callback waiting threads on client eviction (2460)
5193 - Add --ptldebug and --subsystem to lmc (1719)
5194 - update assertion to allow safe interrupt allocation
5195 - set rq_no_resend for cancel requests (2432)
5196 - recalculate ptlrpcd timeout after resend (2494)
5197 - call vfs_rmdir when removing pending directories (2368)
5198 - fix renaming a file to itself (2429)
5199 - lmc creates a default one-stripe lov (2454)
5200 - expand procfs space to handle large clusters (2326)
5201 - increase UML stack to avoid overflow
5202 - update lconf's list of debug and subsystem masks
5203 - fix lfs find --obd (2510)
5204 - /proc tunable for disabling filter read caching (2591)
5205 - stop rpm packages from altering slapd.conf (2301)
5206 - disable nagle in the socknal under 0conf (2578)
5207 - choose mds inode size based on stripe count (2572)
5208 - fix kernel-source rpm problems (2516)
5210 - add --disable-doc to avoid pdf generation (2421)
5211 - update documentation, tests, type-os, comments
5212 - avoid format warnings on ia64
5213 - remove the TOE NAL
5214 - tiny code cleanups by removing unused fields
5216 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5219 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5220 - remove the most blatant lies from BUILDING file (2371)
5221 - change default debug level to reasonable production setting
5222 - reduce client side cache size to reduce cache flush time
5223 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5224 - make TCP zerocopy and pinger support enabled by default (2476)
5225 - sync writes completed after process exits caused crashes (2319)
5226 - maintain correct mount count on the MDS (2356)
5227 - backout 1557, because 2316 wasn't really fixed
5228 - better file I/O statistics gathering in /proc
5229 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5230 - another kernel patch to fix zero-copy TCP function export
5231 - don't take duplicate lock when processing re-sent getattr (2420)
5232 - lctl uses obd_self_export instead of creating new conn (2353)
5233 - MDS/OST recovery case which requires object creation asserted (2425)
5234 - move lfs from /usr/sbin to /usr/bin in packages
5235 - fix race between mds_client_add and mds_client_free (2417)
5236 - use kmalloc instead of slabs in portals (2430)
5237 - don't create duplicate records when a failover MDS is present (2442)
5238 - remove unnecessary mount age check (2332)
5239 - don't remove directory inodes from locks prematurely (2451)
5240 - don't break if MDS service name is the same as hostname (2103)
5241 - fix races in client write RPC generation when cache full (2482)
5243 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5246 - remove now-unused request->rq_obd (278)
5247 - if an allocation fails, print out how much memory we've used (1933)
5248 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5249 - add the "configurable stack size" patch to most series files (1256)
5250 - ability to write large log records, for 100+ OST configs (2306)
5251 - fix NULL deref when filter_prep fails (2314)
5252 - fix operator precedence error in filter_sync
5253 - dynamic allocation of socknal TX descriptors (2315)
5254 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5255 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5256 - ensure that CWARN messages in Portals always get to the syslog
5257 - __init/__exit are not for prototype decls (ldlm_init/exit)
5258 - x86-64 compile warning fixes
5259 - fix gateway LMC keyword conflict (2318)
5260 - fix MDS lock inversions in getattr/reint paths (1844)
5261 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5262 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5263 - DLM race condition prevented some lock evictions (2328)
5264 - ENOMEM detection and retry on socknal sends (2230)
5265 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5266 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5267 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5268 - better error messages when a client is rejected during recovery (1505)
5269 - avoid cancelling locks which were never granted, after failure (2330)
5270 - fix i_sem/journal inversion in mds_client_add (2333)
5271 - fix truncate/getattr lock cycle deadlock (2334)
5272 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5273 - fix two rare exit paths which could leak an l_lock() ref (2321)
5274 - fencepost error in MDS/OST orphan recovery (2226)
5275 - make log record alignment 8 bytes (1988)
5276 - lstripe now fails when requested offset > ost_count (2237)
5277 - ensure that all kernel series have a complete list.h (1607)
5278 - fix crashes in special-file operations (2316)
5279 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5281 - allow configurable automake binary, for testing new versions
5282 - small update to the lfs documentation
5284 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5286 * fix negative export reference count in fsfilt_sync (2312)
5288 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5289 * release candidate 0.9.1
5291 - orphans are moved into the PENDING directory for possible recovery
5292 - replayed opens now open by fid for orphan/rename safety (1042)
5293 - last close of an orphan inode generates a transno (683)
5294 - chdir() and mount() now pin the directory entry (1020)
5295 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5296 - discard very old requests without processing them (1502)
5297 - remove some common, well-understood CERRORs (1505)
5298 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5299 - clear "grant" flags in OST replies until OST grant code lands (1644)
5300 - fix read performance by not clobbering i_blksize on client (1598)
5301 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5302 - make LRU size a /proc tunable, clears locks when reduced (707)
5303 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5304 - server locks take references on exports now (1558)
5305 - build fixes for 2.4.20-rh trees (1663)
5306 - return an error from lov_create if all OSCs are inactive (1751)
5307 - fix import levels when a reconnect happens without a timeout (1597)
5308 - exit early from mds_open if we get a lookup error (1749)
5309 - partial page read at EOF wouldn't wait for disk before sending (1642)
5310 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5311 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5312 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5313 - write/truncate lock inversion (1639)
5314 - Don't auto-load obdclass, portals modules during cleanup (1495)
5315 - fix timestamps from jumping to "now" (1763)
5316 - extra journal assertions (1648)
5317 - add an extra multiunlink test (1771)
5318 - fix read_record/write_record API (1776)
5319 - fix leak of offset_extent, possible incorrect i_size later (1772)
5320 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5321 - Debugging for the kqswnal_get_idle_tx problems (1820)
5322 - Allow recovery to be attempted multiple times (1536)
5323 - Write out MDS last_rcvd file after it is first created (1600)
5324 - Fix tx_descriptor leak in failed transmit situations (1827)
5325 - ext3 journaling fixes for assertion failure after IO error (1871)
5326 - class_export_put() on freed export after completion AST error (1896)
5327 - Fix revalidate looping in VFS (1322)
5328 - Don't access a freed export during MDS_REINT timeout (1521)
5329 - Add open-unlink recovery support on the MDS (1673,1764)
5330 - Return an error if no MDS data was read from last_rcvd (1946)
5331 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5332 - Don't setup a disk device that doesn't match exported UUID (317)
5333 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5334 - avoid committing NULL handle in force close
5335 - local.sh is now a one-stripe LOV configuration
5336 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5337 - fix ext3 htree duplicate directory entry corruption (1516)
5338 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5339 - update to new LOV EA format (2097)
5340 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5342 2003-06-15 Phil Schwan <phil@clusterfs.com>
5345 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5346 - per-import/export recovery handling (958, 931, 959)
5347 - multiple last-rcvd slots, for serving multiple FSes (949)
5348 - connections are again shared between multiple imp/exports (963, 964)
5349 - "umount -f" would hang if any requests needed to be sent (393, 978)
5350 - avoid pinning large req buffer by copying for queued messages (989)
5351 - add "uuid" to "lctl device" command to help upcalls (991)
5352 - "open" RPCs with transnos would confuse recovery counters (1037)
5353 - do proper endian conversion of all wire messages (288, 340, 891)
5354 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5355 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5356 - fix ptlrpc_connection leak in target_handle_connect (1174)
5357 - fix import refcounting bug in OST and MDS cleanup (1134)
5358 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5359 - fix very unlikely obd_types race condition (501)
5360 - remove osc_open hack for echo_client (1187)
5361 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5362 - a failure in read_inode2 leads to deadlock (1139)
5363 - cancel ack-locks as soon as transaction is committed (1072)
5364 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5365 - make sure to commitrw after any preprw to avoid deadlock (1162)
5366 - failing to execute a file in a lustre FS would lock inode (1203)
5367 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5368 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5369 - obd_uuid2tgt would walk off the end of the list (1255)
5370 - on IA64 the osc would give portals incorrect bulk size (1258)
5371 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5372 - fix lock inversion caused by new llite matching code (1282)
5373 - limit the number of dirty pages on a client to 10MB (1286)
5374 - timed out locks were not being corrected cancelled (1289)
5375 - fix O_DIRECT above 4GB on IA-32 (1292)
5376 * major user-visible changes
5377 - fail out/fail over policy now controlled by the upcall (993)
5379 - add OBD_PING to check server availability and failure (954)
5380 - lustre messages are now sent in sending host order (288, 340, 891)
5381 - add eadatalen to MDS getattr reply (340)
5382 - OST read replies may contain second buffer, with per-page status (593)
5384 2003-03-11 Phil Schwan <phil@clusterfs.com>
5387 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5388 - failed open()s could cause deadlock; fixed (867, 869)
5389 - stop cancelling OST locks when files are closed (481)
5390 - overlapping XID spaces caused network corruption (851, 853)
5391 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5392 - setattr_raw added, to do single-RPC, server-side setattrs
5393 - lmc/lconf syntax change for OST UUIDs
5394 - fix crashy race condition between ptlrpc_free_req and osc_close
5395 - don't use request in mdc_enqueue if we hit a timeout (889)
5396 - don't set the inode i_size for regular files from the MDS (896)
5397 - handle out of order completion AST (842)
5398 - don't LBUG if a lock request times out after receiving AST (913)
5399 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5400 - if a bad lock AST arrives, send an error instead of dropping entirely
5401 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5402 - fix leak in bulk IO when only partially completed (899, 900, 926)
5403 - fix O_DIRECT for ia64 (55)
5404 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5405 - C-z after timeout could hang a process forever; fixed (977)
5407 - client-side I/O cache (678, 924, 929, 941, 970)
5409 - READPAGE and SETATTRs which don't take server-side locks get
5412 2003-02-11 Phil Schwan <phil@clusterfs.com>
5415 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5416 - Fix fsfilt_extN_readpage() to read a full page of directory
5417 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5418 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5419 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5420 - Most intent operations were removed, in favour of a new RPC mode
5421 that does a single RPC to the server and bypasses most of the VFS
5422 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5423 - Aggressively cancel local locks on DLM servers
5424 - mds_reint_unlink sends EA to the client if it's the last nlink.
5425 client uses that EA to unlink OST objects.
5426 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5427 - recursive symlinks were fixed (439)
5428 - fixed NULL deref in DEBUG_REQ
5429 - filter_update_lastobjid no longer calls sync, which annoyed extN
5430 - fixed multi-client small-writes to a single file problem (445)
5431 - fixed mtime updates during file writes (607)
5432 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5433 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5434 - fixed Linux OST size reporting problem (444, 656)
5435 - OST now updates object mtime with writes or setattr (607, 619)
5436 - client verifies file size before zeroing page past EOF (445)
5437 - OST now writes last allocated objid to disk with allocation (108)
5438 - LOV on echo now works (409)
5440 - mds_reint_unlink sends a new buffer, with the EA included. this
5441 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5442 set if a regular file was being unlinked, and it was the last link
5443 - use PtlGet from the target for bulk writes (315)
5444 - OST now updates object mtime with writes or setattr (607, 619)
5445 - LDLM now has a grant-time callback to revalidate locked items, if
5447 - Many MDS operations were reorganized to combat race conditions
5449 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5450 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5452 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5455 - Fully reactivate OST imports after reconnection (512, others)
5456 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5457 - More graceful error handling for truncating on dead OST (515)
5458 - Don't error out unless we're actually accessing dead stripes (474)
5459 - Fix garbage sizes when stripes are missing (410)
5460 - LRU counters were broken, causing constant lock purge (433, 432)
5461 - garbage on read from stripes with failed OSTs (441)
5462 - mark OSCs as active before reconnecting during recovery (438)
5463 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5464 - lfind did not preserve OST order in output (443)
5465 - symlinks cause hung clients, incorrect data (439)
5466 - stop dereferencing request after dropping refcount (457)
5467 - don't LASSERT(spin_is_locked) on non-SMP (455)
5468 - fixes for many rename() bugs
5469 - fstat didn't correctly synchronize attributes (399)
5470 - server must handle lock cancellation during blocking AST prep (487)
5471 - bulk descriptors were free()d too soon (511)
5472 - fix paths in lconf, which would load incorrect modules (451, 507)
5473 - fix confusing lconf 'host not found' error message (386)
5474 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5475 - fix race condition in mdc_blocking_ast() for inode access (526)
5476 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5477 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5478 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5479 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5480 - fix lstripe in several ways (kernel side) (527)
5481 - fix request leak in ldlm_cli_enqueue (262)
5482 - incorrect OSC was marked inactive after OST failure
5483 - call mds_fs_cleanup before unmounting filesystem (524)
5484 - fix races between taking ns_lock and ldlm_lock_change_resource
5485 - fix races updating LOV export open file list
5486 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5487 - fix recovery NULL deref in ldlm_cli_cancel_unused
5488 - fix some DLM races by using new hash table for lock handles (419)
5489 - permit the client to specify desired inodes, at replay
5490 - duplicate requests when we queue them for replay reintegration
5491 - fix last_rcvd offset calculation
5492 - sync after each recovered transaction, so we always make progress
5493 - never, not always, ERESTART requests without transnos
5494 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5496 - skip replay if the MDS says that the client is already connected
5497 - don't check for a recovery-enabled export to match lctl's UUID
5498 - don't INC_USE_COUNT for phantom exports
5499 - don't crash when cleaning up phantom exports (567)
5500 - don't double-finish or set replay data for errored mdc_open requests
5501 - abort requests when they time out, so we don't get old replies
5502 - send/receive replies for AST messages again
5503 - if the client says that it doesn't have the lock, cancel it on the
5505 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5506 mark it as destroyed, it will all work out when decref is called
5507 - fix module use counts (22, 581)
5509 - ASTs now expect a reply (server cancels lock on error reply)
5511 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5514 - fix many simultaneous client startup (392)
5515 - fix dentry->d_it clobbering
5516 - credentials weren't being shipped for readdir/getattr operations
5517 - remove invalid assertions triggered during some concurrent MD
5519 - proper Lustre versions added (336, 389)
5520 - fix memory leak for create error case (398)
5521 - fix LOV locking bug that would get cli/srv out of sync
5522 - fix echo client over LOV (409)
5523 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5524 - fix double-O_EXCL intent crash (424)
5525 - avoid sending multiple lock CANCELs (352)
5527 - MDS can do multi-client recovery (modulo bugs in new code)
5529 - many updates, edits, cleanups
5531 2002-11-18 Phil Schwan <phil@clusterfs.com>
5534 - fix null d_it dereference (346)
5535 - fix full OST/dbench hang (333)
5536 - fix permission problem with file removal (286)
5537 - fix removal of OSCs from LOV when they fail
5538 - fix NULL deref during bulk timeout (214)
5539 - fix problems related to multiple filesystems on one MDS (241)
5540 - fixed serious subtle metadata locking bugs
5541 - free locks on clients when inodes are removed due to memory
5543 - fix inode pointer in lock data (285)
5544 - partial support for multiple MDS on a single host (241)
5545 - data locks weren't cancelled at clear_inode time (290, 311)
5546 - intent locks could lead to unbounded lock growth (205)
5547 - added a maximum lock count, an LRU list, and a flusher
5548 - fix multiple rename (365)
5549 - properly abstracted the echo client
5550 - OSC locked 1 byte too many; fixed
5551 - rewrote brw callback code:
5552 - fixed recovery bugs related to LOVs (306)
5553 - fixed too-many-pages-in-one-write crash (191)
5554 - fixed (again) crash in sync_io_timeout (214)
5555 - probably fixed callback-related race (385)
5557 - Add capability to MDS protocol
5558 - LDLM cancellations and callbacks on different portals
5560 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5563 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5564 - defer open object destruction to close time (601981, 138)
5565 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5566 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5567 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5568 - don't call obd_setattr in ll_file_release for destroyed objects
5569 * protocol change to lustre_msg: move |version| and add |flags|
5570 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5571 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5572 * added replay of create, unlink, link and rename operations during
5573 MDS failover; recovery should be much more robust now
5574 * remove failed OSCs from LOVs (only lov_create uses this so far)
5575 * the lustre-HOWTO was brought (more) up to date (582544)
5577 2002-10-23 Phil Schwan <phil@clusterfs.com>
5580 - in-use dentries weren't being reused properly (617851)
5581 - prevent multiple LDLM setup (599178)
5582 - fix LOV size calculations for truncate (617853)
5583 - fix client handling of MDS intent errors (POSIX)
5584 - fix permission bug in lovstripe.c test (624321)
5585 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5586 - truncate past end of file could corrupt data
5587 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5588 - a race in recovery could return ETIMEDOUT to apps (623947)
5589 - building outside the source directory was fixed
5590 * the lustre-HOWTO was brought (more) up to date (582544)
5591 * major progress was made on recovery functionality
5593 2002-10-10 Phil Schwan <phil@clusterfs.com>
5596 - recovery deadlock fix
5597 - rm -rf causes LBUG fix (617817)
5598 - file open by multiple tasks fix (618962)
5599 - directory permissions bugs (602707 and 620007)
5600 - journal_stop fixed with locking (611313)
5601 - O_APPEND failures resolved (618273, perhaps 614459)
5602 - lconf PATH fix (619770)
5603 - IA64 build fix (621450)
5604 - RPC buffer sizes scale with amount of memory
5606 2002-10-01 Phil Schwan <phil@clusterfs.com>
5609 - locks would be cancelled without throwing away data pages,
5610 resulting in inconsistent data (605627)
5611 - inode attributes were not always being refreshed (605627, 612449)
5612 - lconf now continues to cleanup after lctl reports an error
5613 - MDS now enforces user permissions (602707)
5614 - lprocfs cleanup fixed, but not yet enabled (614157)
5615 - fixed infinite server hang, should a client not respond to an AST
5616 - avoid going into recovery if user calls readlink() with a buffer
5617 that's too small (613941)
5618 - AST RPCs no longer require replies (614867) -- this may be changed
5619 - don't crash server if client sends an IOV that's too big (611336)
5620 - fixed lock conversion deadlock (611892)
5621 - fixed the following of symlinks (614622)
5622 * recovery: the server can remove locks from a client that dies, other
5623 clients can make progress
5624 * more extN patch fixes
5625 * compile-time configurable ptlrpc buffer allocations
5627 - collaborative read cache document
5628 - Lustre Lite Performance CDR document-in-progress
5630 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5633 - fix typo in patch-2.4.18
5635 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5638 - clear ptlrpc request each time in handle_incoming_request()
5639 - unlink of files now destroys the object on the OST
5641 2002-09-19 Peter Braam <braam@clusterfs.com>
5643 * add hard link support
5644 * change obdfile creation method
5645 * kernel patch changed
5647 2002-09-19 Peter Braam <braam@clusterfs.com>
5650 - stack overflow bug in extN fixed
5652 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5654 * documentation updates
5655 - add man pages for config tools
5656 - update tests/README to describe testing with new config tools
5657 - finish metadata API descriptions
5658 * bug fixes and cleanups
5659 - statfs workaround for 16TB limit
5660 - LOV stripe allocation improved, can stripe on subset of OSTs
5661 - LOV file size/IO offset was wrong for files > 4GB in size
5662 - object EA data was being dropped, caused files to be unreadable
5663 - memory overflow with non-LOV OST caused memory corruption
5664 - fixed regression tests to work with new config tools, obdfilter
5665 - fixed bug when directory size became larger than 1 block
5666 - fixed bug (for single client case) when PWD was deleted
5667 - invalidate local directory pages when doing intent-based ops
5668 - avoid LDLM oops when lock callback contained bad data
5670 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5672 * documentation updates
5673 * bug fixes and cleanups
5674 - configuration tools
5677 - 64-bit compile warnings
5678 - 64-bit internal statfs data
5680 * test_brw on persistent OST devices
5682 * lprocfs (disabled)
5684 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5686 * documentation updates
5687 * bug fixes and cleanups
5688 * configuration tools
5690 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5693 * many small fixes to 0_5_4
5694 * io/network handling
5695 * thinkos in MDS operations
5697 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5700 * crucial basic fixes to 0.5.3
5701 * IOR, Iozone work over Elan
5704 2002-08-07 Phil Schwan <phil@clusterfs.com>
5705 * version 0_5_3, our first alpha
5706 * we use the new Portals iovs
5707 * documentation updates
5708 * bug fixes and cleanups
5709 * small changes in the DLM wire protocol
5711 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5712 * version 0_5_1 with some initial stability,
5713 * locking on MD and file I/O.
5714 * documentation updates
5715 * several bug fixes since 0.5.0
5716 * small changes in wire protocol
5718 2002-07-18 Phil Schwan <phil@clusterfs.com>
5720 * delivered as Lustre Light Alpha
5721 * fixed a crash after handling invalid MDS requests
5722 * fixed directory pages for architectures with non-4k pages sizes
5724 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5725 * release version v0_4_4
5726 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5728 * Fixes a number of interruption problems with OST operations.
5729 * Update documentation for portals header changes
5730 * Move all wire protocol structs/defines to lustre_idl.h
5731 * Fixes symlink length bug.
5732 * Add tcpdump to repository.
5734 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5735 * release version v0_4_3
5736 * Fixes statfs for inodes on extN.
5737 * Fixes bug in runtests which would delete /etc/hosts.
5738 * Use 64-bit object IDs wherever possible (not into VFS though)
5739 Remove ost_get_info, which is unused by lustre, and out of date.
5741 2002-07-03 Peter Braam <braam@clusterfs.com>
5742 * release version v0_4_2 Fixes a lookup error (type not passed)
5743 * move forward to head of Portals
5744 * move forward to latest Lustre kernel
5746 2002-06-25 Peter Braam <braam@clusterfs.com>
5747 * release version v0_4_1. Hopefully stable on single node use.