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: 1.41.5.sun2
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
14 * File join has been disabled in this release, refer to Bugzilla 16929.
18 Description: Reduce small size read RPC
19 Details : Set read-ahead limite for every file and only do read-ahead when
20 available read-ahead pages are bigger than 1M to avoid small size
23 Severity : enhancement
25 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
26 Details : allow skip disconnected ost for send statfs request and hide error
32 Description: fix racy locking of mballoc block bitmaps causing BUG
33 Details : The locking of the mballoc buddy bitmap and the in-memory
34 block bitmap was using two different spin locks in some
35 cases. This made it possible to incorrectly access the
36 mballoc bitmap while another process was modifying it,
37 causing a sanity assertion to fail. While no on-disk corruption
38 was reported, there was some risk of this happening.
41 Frequency : rare, on llog test 6
43 Descriptoin: don't allow connect to already connected import
44 Details : allowing connect to already connected import is hide connecting problem.
49 Descriptoin: allow kill process which wait statahead result
50 Details : for some reasons 'ls' can stick in waiting result from statahead,
51 in this case need way for kill this process.
53 Severity : enhancement
55 Description: Add state history info file, enhance import info file
56 Details : Track import connection state changes in a new osc/mdc proc file;
57 add overview-type data to the osc/mdc import proc file.
59 Severity : enhancement
61 Description: MDS create should not wait for statfs RPC while holding DLM lock.
63 Severity : enhancement
65 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
70 Descriptoin: OST grant too much space to client even there are not enough space.
71 Details : Client will shrink its grant cache to OST if there are no write
72 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
73 this grant cache if there are already not enough avaible space
74 (left_space < total_clients * 32M).
77 Frequency : start MDS on uncleanly shutdowned MDS device
79 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
80 Details : stay in waiting mds<>ost recovery finished produce random bugs
81 due race between two ll_sync thread for one lov target. send
82 ACTIVATE event only if connect realy finished and import have
86 Frequency : rare, connect and disconnect target at same time
88 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
89 Details : don't call obd_disconnect under lov_lock. this long time
90 operation and can block ptlrpcd which answer to connect request.
95 Descriptoin: don't lose wakeup for imp_recovery_waitq
96 Details : recover_import_no_retry or invalidate_import and import_close can
97 both sleep on imp_recovery_waitq, but we was send only one wakeup
101 Frequency : always with long access acl
103 Descriptoin: mds can't pack reply with long acl.
104 Details : mds don't control size of acl but they limited by reint/getattr
107 Severity : enhancement
109 Description: Update to SLES10 kernel-2.6.16.60-0.33.
111 Severity : enhancement
113 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
116 Frequency : start MDS on uncleanly shutdowned MDS device
118 Descriptoin: aborting recovery hang on MDS
119 Details : don't throttle destroy RPCs for the MDT.
122 Frequency : on remount
124 Description: external journal device not working after the remount
125 Details : clear dev_rdonly flag for external journal devices in
131 Description: shutdown vs evict race
132 Details : client_disconnect_export vs connect request race.
133 if client will evicted at this time - we start invalidate
134 thread without referece to import and import can be freed
138 Frequency : rare, need acl's on inode.
140 Description: client can't handle ost additional correctly
141 Details : if ost was added after client connected to mds client can have
142 hit lnet_try_match_md ... to big messages to wide striped files.
143 in this case need teach client to handle config events about add
144 lov target and update client max ea size at that event.
146 Severity : enhancement
148 Description: Changelogs
149 Details : Changelogs are a lightweight mechanism to track filesystem
150 metadata and namespace changes. The changelog is recorded
151 permanently on the MDTs, and is periodically "consumed" / purged
152 when records are no longer needed.
154 Severity : enhancement
156 Description: compact fld format with extents
157 Details : Store range of seq rather than every seq in FLD. Seq
158 controller update FLD rather than clients. In Case of CMD, mdt0
159 has FLD, all other metadata server act as non persistent proxy
160 for FLD queries and cache fld entries in fld cache.
165 Description: don't skip ost target if they assigned to file
166 Details : Drop slow OSCs if we can, but not for requested start idx.
167 This means "if OSC is slow and it is not the requested
168 start OST, then it can be skipped, otherwise skip it only
169 if it is inactive/recovering/out-of-space.
173 Description: more cleanup in mds_lov
174 Details : not send LOV EA under replay, we can't know about they size at this
175 time. Don't allow client connect to mds before any ost connected,
176 for avoid problems with LOV EA size and returning EIO to client.
178 Severity : enhancement
180 Description: Interoperability at server side (Disk interoperability)
182 Severity : enhancement
184 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
186 Severity : enhancement
188 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
190 Severity : enhancement
192 Description: New client IO stack (CLIO).
194 Severity : enhancement
196 Description: Commit on sharing. Eliminate inter-client dependencies between
197 uncommitted transactions by doing transaction commits.
198 Thereby clients may recovery independently.
201 Frequency : Create a symlink file with a very long name
203 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
204 Details : If there is no extra space in the request for early cancels,
205 ldlm_req_handles_avail() returns 0 instead of a negative value.
207 Severity : enhancement
209 Description: Add /proc entry for import status
210 Details : The mdc, osc, and mgc import directories now have
211 an import directory that contains useful import data for debugging
214 Severity : enhancement
216 Description: Re-disable certain /proc logging
217 Details : Enable and disable client's offset_stats, extents_stats and
218 extents_stats_per_process stats logging on the fly.
221 Frequency : Only on FC kernels 2.6.22+
223 Description: oops in statahead
224 Details : Do not drop reference count for the dentry from VFS when lookup,
225 VFS will do that by itself.
227 Severity : enhancement
229 Description: Generic /proc file permissions
230 Details : Set /Proc file permissions in a more generic way to enable non-
231 root users operate on some /proc files.
235 Description: Hitting mdc_commit_close() ASSERTION
236 Details : Properly handle request reference release in
237 ll_release_openhandle().
241 Description: quota recovery deadlock during mds failover
242 Details : This patch includes att18982, att18236, att18237 in bz14840.
244 1. fix osts hang when mds does failover with quotaon
245 2. prevent watchdog storm when osts threads wait for the
250 Frequency : only patchless client
251 Description: add workaround for race between add/remove dentry from hash
253 Severity : enhancement
255 Description: Allow OST glimpses to return PW locks
259 Description: LBUG when llog conf file is full
260 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
264 Description: Prevent import from entering FULL state when server in recovery
268 Description: service mount cannot take device name with ":"
269 Details : Only when device name contains ":/" will mount treat it as
275 Description: replace ptlrpcd with the statahead thread to interpret the async
276 statahead RPC callback
280 Frequency : on recovery
281 Description: I/O failures after umount during fail back
282 Details : if client reconnected to restarted server we need join to recovery
283 instead of find server handler is changed and process self
284 eviction with cancel all locks.
286 Severity : enhancement
288 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
290 Severity : enhancement
292 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
294 Severity : enhancement
296 Description: Upcall on Lustre log has been dumped
297 Details : Allow for a user mode script to be called once a Lustre log has
298 been dumped. It passes the filename of the dumped log to the
299 script, the location of the script can be specified via
300 /proc/sys/lnet/debug_log_upcall.
305 Description: avoid idr_remove called for id which is not allocated.
306 Details : Move assigment s_dev for clustered nfs to end of initialization,
307 to avoid problem with error handling.
312 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
313 Details : When connection is reused this not moved from CONN_UNUSED_HASH
314 into CONN_USED_HASH and this prodice warning when put connection
315 again in unused hash.
318 Severity : enhancement
320 Description: File striping can now be set to use an arbitrary pool of OSTs.
322 Severity : enhancement
324 Description: Export bytes_read/bytes_write count on OSC/OST.
328 Description: Early reply size mismatch, MGC loses connection
329 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
330 the connect flags are properly negotiated.
335 Description: quotas are not honored with O_DIRECT
336 Details : all writes with the flag O_DIRECT will use grants which leads to
337 this problem. Now using OBD_BRW_SYNC to guard this.
341 Description: add quota statistics
342 Details : 1. sort out quota proc entries and proc code.
343 2. add quota statistics
345 Severity : enhancement
347 Description: enable quota support for HEAD.
351 Description: Properly propagate oinfo flags from lov to osc for statfs
352 Details : restore missing copy oi_flags to lov requests.
354 Severity : enhancement
356 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
357 : l_getgroups(8), lst(8), routerstat(8)
358 Details : included man pages for llobdstat(8), llstat(8),
359 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
361 Severity : enhancement
363 Description: configure's --enable-quota should check the
364 : kernel .config for CONFIG_QUOTA
365 Details : configure is terminated if --enable-quota is passed but
366 : no quota support is in kernel
370 Description: Remove portals compatibility
371 Details : Remove portals compatibility, not interoperable with releases
376 Description: Resolve device initialization race
377 Details : Prevent proc handler from accessing devices added to the
378 obd_devs array but yet be intialized.
380 Severity : enhancement
382 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
384 Severity : enhancement
386 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
391 Description: Using wrong pointer in osc_brw_prep_request
392 Details : Access to array[-1] can produce panic if kernel compiled with
393 CONFIG_PAGE_ALLOC enabled
397 Description: Client runs out of low memory
398 Details : Consider only lowmem when counting initial number of llap pages
402 Description: *optional* service tags registration
403 Details : if the "service tags" package is installed on a Lustre node
404 When the filesystem is mounted, a local-node service tag will
405 be created. See http://inventory.sun.com/ for more information
406 about the Service Tags asset management system.
410 Description: Kernel BUG tries to release flock
411 Details : Lustre does not destroy flock lock before last reference goes
412 away. So always drop flock locks when client is evicted and
413 perform unlock regardless of successfulness of speaking to MDS.
417 Description: add refcount protection for osc callbacks, avoid panic on shutdown
421 Description: sanity test 65a fails if stripecount of -1 is set
422 Details : handle -1 striping on filesystem in ll_dirstripe_verify
427 Description: ASSERTION(CheckWriteback(page,cmd)) failed
428 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
431 Severity : enhancement
433 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
437 Description: do not process already freed flock
438 Details : flock can possibly be freed by another thread before it reaches
439 to ldlm_flock_completion_ast.
443 Description: LBUG during stress test
444 Details : Need properly lock accesses the flock deadlock detection list.
448 Description: oops in page fault handler
449 Details : kernel page fault handler can return two special 'pages' in error
450 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
454 Description: timeout with invalidate import.
455 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
456 should be handled by ptlrpcd. This produce long age waiting and
457 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
459 Severity : enhancement
461 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
465 Description: enable MGS and MDT services start separately
466 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
467 a co-located MGS without starting the MGS, which is a complement
468 to 'nosvc' mount option.
472 Frequency : after recovery
473 Description: precreate to many object's after del orphan.
474 Details : del orphan st in oscc last_id == next_id and this triger growing
475 count of precreated objects. Set flag LOW to skip increase count
476 of precreated objects.
480 Frequency : rare, on clear nid stats
481 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
482 Details : when clean nid stats sometimes try destroy live entry,
483 and this produce panic in free.
487 Description: Stack overflow during MDS log replay
488 ease stack pressure by using a thread dealing llog_process.
492 Description: wait until IO finished before start new when do lock cancel.
493 Details : VM protocol want old IO finished before start new, in this case
494 need wait until PG_writeback is cleared until check dirty flag and
495 call writepages in lock cancel callback.
497 Severity : enhancement
499 Description: using special macro for print time and cleanup in includes.
503 Description: mds_mfd_close() ASSERTION(rc == 0)
504 Details : In mds_mfd_close(), we need protect inode's writecount change
505 within its orphan write semaphore to prevent possible races.
509 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
513 Frequency : rare, on shutdown ost
514 Description: don't hit live lock with umount ost.
515 Details : shrink_dcache_parent can be in long loop with destroy dentries,
516 use shrink_dcache_sb instead.
520 Description: don't panic with use echo client
521 Details : echo client pass NULL as client nid pointer and this produce null
526 Description: fix build on ppc32
527 Details : compile code with -m64 flag produce wrong object file for ppc32.
531 Description: add message levels for liblustreapi
535 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
536 Details : Move the 'good_osts' check before the 'total_bavail' check. This
537 will result in an -EAGAIN and in the exit call path we call
538 alloc_rr() which will with increasing aggressiveness attempt to
539 aquire precreated objects on the minimum number of required OSCs.
543 Description: Use old size assignment to avoid deadlock
544 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
545 the scheduling while holding a spinlock. We do not need locking
546 for size in ll_update_inode() because size is only updated from
547 the MDS for directories or files without objects, so there is no
548 other place to do the update, and concurrent access to such inodes
549 are protected by the inode lock.
553 Description: resolve "_IOWR redefined" build error on SLES10
557 Description: dump the memory debugging after all modules are unloaded to
558 suppress false negative in conf_sanity test 39
560 Severity : enhancement
562 Description: build kernel-ib packages for OFED 1.3 in our release cycle
567 Description: fix SLES kernel versioning
568 Details : the kernel version for our SLES 10 kernel did not include a "-"
569 before the "smp" at the end. while this was not a problem in
570 general, it did mean that software trying to use the kernel
571 version to try to detect a vendor specific kernel would fail.
572 this was most evident by the OFED build scripts.
576 Description: Don't update lov_desc members until making sure they are valid
577 Details : When updating lov_desc members via proc fs, need fix their
578 validities before doing the real update.
582 Description: don't put request into delay list while invalidate in flight.
583 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
584 import in flight. this produce timeout for invalidate and sometimes
585 can cause stale data.
589 Frequency : on ppc only
590 Description: not convert ost objects for directory because it's not exist.
591 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
595 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
597 Severity : enhancement
599 Description: Update to support 2.6.22.14 vanilla kernel.
603 Frequency : rare, on recovery
604 Description: read procfs can produce deadlock in some situation
605 Details : Holding lprocfs lock which send rpc can produce block for destroy
606 obd objects and this also block reconnect with -EALREADY. This
607 isn't fix all lprocfs bugs - but make it rare.
609 Severity : enhancement
611 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
614 Frequency : frequent on X2 node
616 Description: mdc_set_open_replay_data LBUG
617 Details : Set replay data for requests that are eligible for replay.
621 Description: lustre_mgs: operation 101 on unconnected MGS
622 Details : When MGC is disconnected from MGS long enough, MGS will evict the
623 MGC, and late on MGC cannot successfully connect to MGS and a lot
624 of the error messages complaining that MGS is not connected.
628 Frequency : on network error
629 Description: panic with double free request if network error
630 Details : mdc_finish_enqueue is finish request if any network error ocuring,
631 but it's true only for synchronus enqueue, for async enqueue
632 (via ptlrpcd) this incorrect and ptlrpcd want finish request
635 Severity : enhancement
637 Description: client-side metadata stat-ahead during readdir(directory readahead)
638 Details : perform client-side metadata stat-ahead when the client detects
639 readdir and sequential stat of dir entries therein
642 Frequency : on start mds
644 Description: Implement get_info(last_id) in obdfilter.
647 Frequency : occasional
649 Description: Correctly check stale fid, not start epoch if ost not support SOM
650 Details : open with flag O_CREATE need set old fid in op_fid3 because
651 op_fid2 was overwritten with new generated fid, but mds can answer
652 with one of these two fids and both is not stale. Setattr
653 incorrectly started an epoch and assume will be called
654 done_writing, but without SOM done_writing ever being called.
657 Frequency : rare, depends on device drivers and load
659 Description: MDS or OSS nodes crash due to stack overflow
660 Details : Code changes in 1.8.0 increased the stack usage of some functions.
661 In some cases, in conjunction with device drivers that use a lot
662 of stack the MDS (or possibly OSS) service threads could overflow
663 the stack. One change which was identified to consume additional
664 stack has been reworked to avoid the extra stack usage.
667 Frequency : occasional
669 Description: Do not fail import if osc_interpret_create gets -EAGAIN
670 Details : If osc_interpret_create got -EAGAIN it immediately exits and
671 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
672 oscc_has_objects and see OSC has no objests and call
673 oscc_internal_create to resend create request.
675 Severity : enhancement
677 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
679 Severity : enhancement
681 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
684 Frequency : very rare
686 Description: Fix replay if there is an un-replied request and open
687 Details : In some cases, older replay request will revert the
688 mcd->mcd_last_xid on MDS which is used to record the client's
691 Severity : enhancement
693 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
695 Severity : enhancement
697 Description: Add rhel5 support to HEAD.
699 Serverity : enhancement
701 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
706 Description: Don't allow skipping OSTs if index has been specified.
707 Details : Don't allow skipping OSTs if index has been specified, make
708 locking in internal create lots better.
712 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
713 Details : access to bitfield in structure is always rounded to long
714 and this produce problem with not atomic change any bit.
718 Description: Lustre make rpms failed.
719 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
720 builds Lustre rpms from tarball.
723 Frequency : rare on shutdown ost
725 Description: If llog cancel was not send before clean_exports phase, this can
726 produce deadlock in llog code.
727 Details : If llog thread has last reference to obd and call class_import_put
728 this produce deadlock because llog_cleanup_commit_master wait when
729 last llog_commit_thread exited, but this never success because was
730 called from llog_commit_thread.
734 Description: allow userland application know is lost one of stripes.
735 Details : fill lvb_blocks with error code on ost and return it to
736 application if error flag found.
740 Description: NULL lov_tgts causing MDS oops
741 Details : more safe checks for NULL lov_tgts for avoid oops.
743 Severity : enhancement
745 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
749 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
750 Details : If an OST is down the MDS will hang indefinitely in
751 obd_statfs_rqset() waiting for the statfs data. While for
752 MDS QOS usage of statfs, it should not stuck in waiting.
754 Severity : enhancement
756 Description: remote_acl support
757 Details : Support ACL-based permission check for remote user.
758 Support setfacl/getfacl for remote user with the utils
759 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
760 the system "{s,g}etfacl" utils.
762 Severity : enhancement
764 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
766 Severity : enhancement
768 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
772 Description: make mgs_setparam() handle fsname containing dash
773 Details : fsname containing a dash does not work with lctl conf_param
775 Severity : enhancement
777 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
779 Severity : enhancement
781 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
785 Frequency : rare, at shutdown
786 Description: access already free / zero obd_namespace.
787 Details : if client_disconnect_export was called without force flag set,
788 and exist connect request in flight, this can produce access to
789 NULL pointer (or already free pointer) when connect_interpret
790 store ocd flags in obd_namespace.
794 Frequency : only at startup
795 Description: not alloc memory with spinlock held.
796 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
797 if any spinlock held.
799 Severity : enhancement
801 Description: make lustre randomly fail allocating memory
802 Details : Make lustre randomly failed allocating memory for testing purpose.
804 Severity : enhancement
806 Description: lost problems with lov objid file
807 Details : Fixes some scability and access to not inited memory problems
808 in work with lov objdid file.
813 Description: lfs find does not continue on file error
814 Details : Continue other files processing when a file/dir is absent.
818 Description: Inconsistent usage of lustre_pack_reply()
819 Details : Standardize the usage of lustre_pack_reply() such that it
820 always generate a CERROR on failure.
823 Frequency : occasional
825 Description: MDS hang or stay in waiting lock
826 Details : If client receive lock with CBPENDING flag ldlm need send lock
827 cancel as separate rpc, to avoid situation when cancel request
828 can't processed due all i/o threads stay in wait lock.
832 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
836 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
840 Description: Update to SLES9 kernel-2.6.5-7.287.3.
844 Description: Update to RHEL4 latest kernel.
846 Severity : enhancement
848 Description: Build SLES10 patchless client fails
849 Details : The configure was broken by run ./configure with
850 --with-linux-obj=.... argument for patchless client. When the
851 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
852 can't search header adequately. Use absolute path such as
853 -I($LINUX)/include instead.
857 Description: interrupt oig_wait produce painc on resend.
858 Details : brw_redo_request can be used for resend requests from ptlrpcd and
859 private set, and this produce situation when rq_ptlrpcd_data not
860 copyed to new allocated request and triggered LBUG on assert
861 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
862 ptlrpcd set if request is changed and can be safety changed to use
867 Description: LASSERT_{REQ,REP}SWAB macros are buggy
868 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
869 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
870 these macros and replace them with their difinitions instead.
874 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
875 Details : Update which_patch & target file for SLES10 latest kernel.
878 Description: Remove client patches from SLES 10 kernel.
879 Details : This causes SLES 10 clients to behave as patchless clients
880 even on a Lustre-patched (server) kernel.
882 Severity : enhancement
884 Description: self-adjustable client's lru lists
885 Details : use adaptive algorithm for managing client cached locks lru
886 lists according to current server load, other client's work
887 pattern, memory activities, etc. Both, server and client
888 side namespaces provide number of proc tunables for controlling
891 Severity : enhancement
893 Description: light-weight GSS support
894 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
895 and reduce performance overhead.
897 Severity : enhancement
899 Description: Linux keyring support
900 Details : Support using service of Linux keyring for Lustre GSS internal
901 context refresh/cache mechanism.
905 Description: Fix errors in lfs documentation
906 Details : Fixes man pages
910 Description: don't use GFP_* in generic Lustre code.
911 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
915 Description: obdclass is limited by single OBD_ALLOC(idarray)
916 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
920 Description: warnings with build patchless client with vanila 2.6.19 and up
921 Details : change old ctl_table style and replace ctl_table/ctl_table_header
922 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
926 Frequency : Only for SLES
927 Description: sanity_quota fail test_1
928 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
929 file, which makes TSTID[2] inunique.
933 Frequency : Only for SLES9
934 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
935 Details : Update target/ChangeLog/which_patch .
939 Frequency : Only for test_10 in sanity-quota.sh
940 Description: a bug in quota test code
941 Details : checking a test flag in a wrong place
945 Frequency : Only for test_13 in sanity-quota.sh
946 Description: a bug in quota test script
947 Details : shouldn't check overfull allocated quota
951 Description: jbd statistics
952 Details : Port older jbd statistics patch for sles10
954 Severity : enhancement
956 Description: Add printing inode info into message about error in writepage.
961 Description: reply_lock_interpret crash due to race with it and lock cancel.
962 Details : Do not replay locks that are being cancelled. Do not reference
963 locks by their address during replay, just by their handle.
967 Frequency : When flocks are used.
968 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
969 Details : Flock locks might destroy just granted lock if it could be merged
970 with another existing flock, this is done in completion handler,
971 so teach ldlm_cli_enquque_fini that this is a valid case for
977 Description: Oops in read and write path when failing to allocate lock.
978 Details : Check if lock allocation failed and return error back.
982 Description: lstripe command fails for valid OST index
983 Details : The stripe offset is compared to lov->desc.ld_tgt_count
984 instead of lov->desc.ld_active_tgt_count.
988 Description: sanity.sh failed test 103
989 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
994 Description: df doesn't work properly if diskfs blocksize != 4K
995 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
999 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1000 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1002 Severity : enhancement
1004 Description: Adaptive timeouts
1005 Details : RPC timeouts adapt to changing server load and network
1006 conditions to reduce resend attempts and improve recovery time.
1010 Description: llapi_file_create() does not allow some changes
1011 Details : add llapi_file_open() that allows specifying the mode and
1012 open flags, and also returns an open file handle.
1016 Description: merge and cleanup kernel patches.
1017 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1021 Description: Add journal checksum support.(Kernel part)
1022 Details : The journal checksum feature adds two new flags i.e
1023 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1024 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1025 indicates that the commit block contains the checksum for
1026 the blocks described by the descriptor blocks. Now commit
1027 record can be sent to disk without waiting for descriptor
1028 blocks to be written to disk. This behavior is controlled
1029 using JBD2_FEATURE_ASYNC_COMMIT flag.
1033 Description: OSS needs mutliple precreate threads
1034 Details : Add ability to start more than one create thread per OSS.
1038 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1039 Details : execute lfs setstripe on client
1043 Description: mds_obd_create error creating tmp object
1044 Details : When the user sets quota on root, llog will be affected and can't
1045 create files and write files.
1048 Frequency : Always on ia64 patchless client, and possibly others.
1050 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1051 Details : This allows the patchless client to be loaded on architectures
1052 without this export.
1056 Description: RedHat Update kernel for RHEL5
1057 Details : Add the kernel config file for RHEL5.
1061 Description: RedHat Update kernel for RHEL5
1062 Details : Modify the kernel config file more closer RHEL5.
1066 Description: Build failure against Centos5 (RHEL5)
1067 Details : Define PAGE_SIZE when it isn't present.
1071 Description: test_7 on sanity-quota.sh can't be used on separate machines
1072 Details : add facet to handle it
1076 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1077 Details : with NFS, the anon dentry's parent was set to itself in
1078 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1079 corresponding dentry other than use rec->ur_name.
1081 Severity : enhancement
1083 Description: lfs setstripe enhancement
1084 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1088 Description: enable data checksumming by default
1089 Details : enable checksum by default, allow --disable-checksum
1090 configure option and "-o nochecksum" mount option. Checksums
1091 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1095 Description: lustre support for RHEL5
1096 Details : Add support for RHEL5.
1100 Description: Client eviction due to failover config
1101 Details : after a connection loss, the lustre client should attempt to
1102 reconnect to the last active server first before trying the
1103 other potential connections.
1107 Description: when mds and osts use different quota unit(32bit and 64bit),
1108 quota will be released repeatly.
1109 Details : void sending multiple quota reqs to mds, which will keep the
1110 status between the reqs.
1114 Description: osts not allocated evenly to files
1115 Details : change the condition to increase offset_idx
1118 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1120 Description: Data corruption for OSTs that are formatted larger than 2TB
1122 Details : When generating the bio request for lustre file writes the
1123 sector number would overflow a temporary variable before being
1124 used for the IO. The data reads correctly from Lustre (which
1125 will overflow in a similar manner) but other file data or
1126 filesystem metadata may be corrupted in some cases.
1130 Description: Tune the kernel for good SCSI performance.
1131 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1132 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1137 Description: rewrite ext2-derived code in obdclass/uuid.c
1138 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1139 from scratch preserving functionality.
1143 Description: "lfs find -obd UUID" prints directories
1144 Details : "lfs find -obd UUID" will return all directory names instead
1145 of just file names. It is incorrect because the directories
1146 do not reside on the OSTs.
1150 Description: use i_size_read and i_size_write in 2.6 port
1151 Details : replace inode->i_size access with i_size_read/write()
1154 Frequency : when removing large files
1156 Description: scheduling issue during removal of large Lustre files
1157 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1158 It causes scheduling issues when removing large files (17TB in the
1162 Frequency : only with liblustre clients on XT3
1164 Description: evictions taking too long
1165 Details : allow llrd to evict clients directly on OSTs
1168 Frequency : only on ppc
1170 Description: /proc/fs/lustre/devices broken on ppc
1171 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1172 We should make sure the type of 'index' is loff_t and then cast
1173 explicitly as needed below. Do not assign an explicitly cast
1177 Frequency : only for rhel5
1179 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1180 Details : Modify the target file & which_kernel.
1182 Severity : enhancement
1184 Description: omit set fsid for export NFS
1185 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1186 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1190 Frequency : Always, for kernels after 2.6.16
1191 Description: Fix warning idr_remove called for id=.. which is not allocated.
1192 Details : Last kernels save old s_dev before kill super and not allow
1193 to restore from callback - restore it before call kill_anon_super.
1197 Description: buffer overruns could theoretically occur
1198 Details : llapi_semantic_traverse() modifies the "path" argument by
1199 appending values to the end of the origin string, and a
1200 overrun may occur. Adding buffer overrun check in liblustreapi.
1204 Description: Fix error on 'ls .' at the top of the Lustre mount.
1205 Details : Don't revalidate dentry if it is a root dentry.
1209 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1210 Details : Modify vm-tunables-rhel4.patch.
1214 Description: Kernel config for 2.6.18-vanilla.
1215 Details : Modify targets/2.6-vanilla.target.in.
1216 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1217 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1218 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1219 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1223 Description: improve handling recoverable errors
1224 Details : if request processig with error which can be recoverable on server
1225 request should be resend, otherwise page released from cache and
1230 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1231 Details : Modify target file & which_patch.
1232 A flaw was found in the IA32 system call emulation provided
1233 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1234 value could be stored in the %RAX register, which could trigger an
1235 out-of-bounds system call table access. An untrusted local user
1236 could exploit this flaw to run code in the kernel
1237 (ie a root privilege escalation). (CVE-2007-4573).
1241 Description: O_DIRECT bypasses client statistics.
1242 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1243 read_ahead_stats were updated. copied the stats section from
1244 osc_send_oap_rpc() into async_internal().
1248 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1252 Description: change order of libsysio includes
1253 Details : '#include sysio.h' should always come before '#include xtio.h'
1257 Description: handle "serious error: objid * already exists" more gracefully
1258 Details : If LAST_ID value on disk is smaller than the objects existing in
1259 the O/0/d* directories, it indicates disk corruption and causes an
1260 LBUG(). If the object is 0-length, then we should use the existing
1261 object. This will help to avoid a full fsck in most cases.
1263 Severity : enhancement
1265 Description: adapt the lustre_config script to support the upgrade case
1266 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1267 server targets to 1.6 in parallel.
1271 Description: To avoid grant space > avaible space when the disk is almost
1272 full. Without this patch you might see the error "grant XXXX >
1273 available" or some LBUG about grant, when the disk is almost
1275 Details : In filter_check_grant, for non_grant cache write, we should
1276 check the left space by if (*left > ungranted + bytes), instead
1277 of (*left > ungranted), because only we are sure the left space
1278 is enough for another "bytes", then the ungrant space should be
1279 increase. In client, we should update cl_avail_grant only there
1280 is OBD_MD_FLGRANT in the reply.
1284 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1285 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1286 described in CVE-2007-4573. This problem could allow untrusted
1287 local users to gain root access.
1290 Frequency : when using O_DIRECT and quotas
1292 Description: Incorrect file ownership on O_DIRECT output files
1293 Details : block usage reported by 'lfs quota' does not take into account
1294 files that have been written with O_DIRECT.
1299 Description: touch file failed when fs is not full
1300 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1301 otherwise we can get ENOSP while fs is not full.
1305 Description: parallel lock callbacks
1306 Details : Instead of sending blocking and completion callbacks as separated
1307 requests, adding them to a set and sending in parallel.
1310 Frequency : only for Cray XT3
1311 Bugzilla : 12829/13455
1312 Description: Changing primary group doesn't change the group lustre assigns to
1314 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1315 overridden with the primary group provided by the group upcall,
1316 whereas the supplied fsgid can be trusted if it is in the list of
1317 supplementary groups returned by the group upcall.
1319 Severity : enhancement
1321 Description: Allow masking D_WARNING, D_ERROR messages from console
1322 Details : Console messages can now be disabled via lnet.printk.
1326 Description: User code with malformed file open parameter crashes client node
1327 Details : Before packing join_file req, all the related reference should be
1328 checked carefully in case some malformed flags cause fake
1329 join_file req on client.
1333 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1334 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1335 mode is equal to the granted mode. However, it should also check
1336 that LDLM_FL_CP_REQD is not set.
1340 Description: Heavy nfs access might result in deadlocks
1341 Details : After ELC code landed, it is now improper to enqueue any mds
1342 locks under och_sem, because enqueue might want to decide to
1343 cancel open locks for same inode we are holding och_sem for.
1347 Description: Client eviction while running blogbench
1348 Details : A lot of unlink operations with concurrent I/O can lead to a
1349 deadlock causing evictions. To address the problem, the number of
1350 oustanding OST_DESTROY requests is now throttled to
1351 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1352 callbacks are processed in priority.
1356 Description: enable ACLs on MDS by default
1357 Details : ACLs must be enabled on MDS by default.
1360 Frequency : PPC/PPC64 only
1362 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1363 Details : fix duplicated definitions between asm-ppc64/types.h and
1364 lustre_types.h on PPC.
1367 Frequency : PPC/PPC64 only
1369 Description: asm-ppc/segment.h does not exist
1370 Details : fix compile issue on PPC.
1374 Description: better handle error messages in extents code
1377 Frequency : RHEL4 only
1379 Description: mkfs is very slow on IA64/RHEL4
1380 Details : A performance regression has been discovered in the MPT Fusion
1381 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1382 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1383 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1385 Severity : enhancement
1387 Description: SNMP support enhancement
1388 Details : Adding total number of sampled request for an MDS node in snmp
1391 Severity : enhancement
1393 Description: Optimize ldlm waiting list processing for PR extent locks
1394 Details : When processing waiting list for read extent lock and meeting read
1395 lock that is same or wider to it that is not contended, skip
1396 processing rest of the list and immediatelly return current
1397 status of conflictness, since we are guaranteed there are no
1398 conflicting locks in the rest of the list.
1402 Description: Time out and refuse to reconnect
1403 Details : When the failover node is the primary node, it is possible
1404 to have two identical connections in imp_conn_list. We must
1405 compare not conn's pointers but NIDs, otherwise we can defeat
1406 connection throttling.
1410 Description: port llog fixes from b1_6 into HEAD
1411 Details : Port llog reference couting and some llog cleanups from b1_6
1412 (bug 10800) into HEAD, for protect from panic and access to already
1413 free llog structures.
1417 Description: Detect stride IO mode in read-ahead
1418 Details : When a client does stride read, read-ahead should detect that and
1419 read-ahead pages according to the detected stride pattern.
1423 Description: data checksumming impacts single node performance
1424 Details : add support for several checksum algorithm. Currently, only CRC32
1425 and Adler-32 are supported. The checksum type can be changed on
1426 the fly via /proc/fs/lustre/osc/*/checksum_type.
1430 Description: use adler32 for page checksums
1431 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1436 Description: build for x2 fails
1437 Details : fix compile issue on Cray systems.
1441 Description: Properly match for duplicate locks
1442 Details : Due to different lock order from skiplists code, we need to
1443 traverse entire list for now
1446 Frequency : only on PPC/SLES10
1448 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1449 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1450 be defined. Add a hack in mkfs_lustre.c to work around this compile
1455 Description: LASSERT on MDS when client holding flock lock dies
1456 Details : ldlm pool logic depends on number of granted locks equal to
1457 number of released locks which is not true for flock locks, so
1458 just exclude such locks from consideration.
1462 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1463 Details : Use fsfilt_sync() for both the whole filesystem sync and
1464 individual file sync to eliminate dangerous inode locking
1465 with I_LOCK that can lead to a deadlock.
1469 Description: performance in 1.6.3
1470 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1471 sure that 1MB requests are not fragmented by the block layer.
1473 Severity : enhancement
1475 Description: organize the server-side client stats on per-nid basis
1476 Details : Change the structure of stats under obdfilter and mds to
1484 The "uuid"s file would list the uuids of _active_ exports.
1485 And the clear entry is to clear all stats and stale nids.
1487 Severity : enhancement
1489 Description: eliminate client locks in face of contention
1490 Details : file contention detection and lockless i/o implementation
1491 for contended files.
1495 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1498 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1500 Description: MDS may lose file striping (and hence file data) in some cases
1501 Details : If there are additional extended attributes stored on the MDS,
1502 in particular ACLs, SELinux, or user attributes (if user_xattr
1503 is specified for the client mount options) then there is a risk
1504 of attribute loss. Additionally, the Lustre file striping
1505 needs to be larger than default (e.g. striped over all OSTs),
1506 and an additional attribute must be stored initially in the
1507 inode and then increase in size enough to be moved to the
1508 external attribute block (e.g. ACL growing in size) for file
1513 Description: skiplist implementation simplification
1514 Details : skiplists are used to group compatible locks on granted list
1515 that was implemented as tracking first and last lock of each
1516 lock group the patch changes that to using doubly linked lists
1520 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1521 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1522 because LASSERT on both the data supplied by a client, and the
1523 data on disk is dangerous and incorrect.
1525 Severity : enhancement
1527 Description: Slow truncate/writes to huge files at high offsets.
1528 Details : Directly associate cached pages to lock that protect those pages,
1529 this allows us to quickly find what pages to write and remove
1530 once lock callback is received.
1534 Description: more ldlm soft lockups
1535 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1536 starve other threads from the resource lock for a long time in
1537 case of long waiting queue, so change the debug level from
1538 D_OTHER to the less frequently used D_INFO.
1540 Severity : enhancement
1542 Description: add -gid, -group, -uid, -user options to lfs find
1546 Description: Hung threads in invalidate_inode_pages2_range
1547 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1548 once one is completed. As a result, some RPCs are stuck in the
1549 queue and are never sent.
1553 Description: filter threads hungs on waiting journal commit
1554 Details : Cleanup filter group llog code, then only filter group llog will
1555 be only created in the MDS/OST syncing process.
1559 Description: Procfs and llog threads access destoryed import sometimes.
1560 Details : Sync the import destoryed process with procfs and llog threads by
1561 the import refcount and semaphore.
1563 Severity : enhancement
1565 Description: openlock cache of b1_6 port to HEAD
1570 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1571 Details : If insertion of an extent fails, then discard the inode
1572 preallocation and free data blocks else it can lead to duplicate
1577 Description: don't always update ctime in ext3_xattr_set_handle()
1578 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1579 In some cases the ctime should not be updated, for example for
1580 2.0->1.8 compatibility it is necessary to delete an xattr and it
1581 should not update the ctime.
1585 Bugzilla : 15713/16362
1586 Description: Assertion in iopen_connect_dentry in 1.6.3
1587 Details : looking up an inode via iopen with the wrong generation number can
1588 populate the dcache with a disconneced dentry while the inode
1589 number is in the process of being reallocated. This causes an
1590 assertion failure in iopen since the inode's dentry list contains
1591 both a connected and disconnected dentry.
1595 Description: assertion failure in ldlm_handle2lock()
1596 Details : fix a race between class_handle_unhash() and class_handle2object()
1597 introduced in lustre 1.6.5 by bug 13622.
1602 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1603 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1608 Description: oops in sd_iostats_seq_show()
1609 Details : unloading/reloading the scsi low level driver triggers a kernel
1610 bug when trying to access the sd iostat file.
1615 Description: Kernel panics during QLogic driver reload
1616 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1617 patch, causing memory corruption.
1622 Description: journal_dev option does not work in b1_6
1623 Details : pass mount option during pre-mount.
1625 Severity : enhancement
1627 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1628 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1629 extent information of a file. It can be used to map logical blocks
1630 in a file to physical blocks in the block device.
1634 Description: LDLM soft lockups - improvement
1635 Details : It is be possible to send the lock handle along with each read
1636 or write request because the client is already doing a lock match
1637 itself so there isn't any reason the OST should have to re-do that
1641 Frequency : only on Cray X2
1643 Description: X2 build failures
1644 Details : fix build failures on Cray X2.
1648 Description: xid & resent requests
1649 Details : Initialize RPC XID from clock at startup (randomly if clock is
1652 Severity : enhancement
1654 Description: Add lustre_start utility to start or stop multiple Lustre servers
1659 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1660 Details : In case of memory pressure, list_del() can be called twice on
1661 req->rq_history_list, causing a kernel oops.
1665 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1666 Details : fix stack overflow in the distributed lock manager by defering
1667 export eviction after a failed AST to the elt thread instead of
1668 handling it in the dlm interpret routine.
1672 Description: Convert some comments to new format.
1673 Details : Update documenting comments to match doxygen conventions.
1677 Description: Grammar fixes.
1678 Details : A couple of trivial spelling fixes.
1682 Description: OSD_COUNTERS-mandatory
1683 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1684 to simplify the code.
1688 Description: simplify cmm_device freeing logic.
1689 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1693 Description: Add lockdep support to dt_object_operations locking interface.
1694 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1695 parameter indicating lock ordering. Update mdd code to use new
1700 Description: Introduce failloc constants for lockless IO tests.
1701 Details : Add two new failloc constants to test lockless IO. Only one of
1702 them in implemented---another is checked in yet to be landed
1707 Description: Add lockdep support for inode mutex.
1708 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1709 in the situations where Lustre has to lock more than one inode
1714 Description: Add optional invariants checking support.
1715 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1716 configure switch. This macro is to be used for consistency and
1717 sanity checks that are too expensive to be left in `production'
1722 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1723 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1728 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1729 Details : Use spin_lock_nested() in (the only) situation where more than
1730 one ldlm_lock is locked simultaneously. Also, fix possible
1731 dead-lock in ldlm_lock_change_resource() by enforcing particular
1736 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1737 Details : Instead of specifying each ldlm_lock call-back through separate
1738 parameter, wrap them into struct ldlm_callback_suite.
1742 Description: Kill join_lru obd method and its callers.
1743 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1744 in memory---a requirement implemented through ->o_join_lru() obd
1745 method in HEAD. Remove this method and its users.
1749 Description: Add asynchronous ldlm ENQUEUE completion handler.
1750 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1751 case that stock ldlm_completion_ast() cannot handle as it waits
1752 until lock is granted. Introduce new ldlm_completion_ast_async()
1753 for this. Also comment ldlm_completion_ast().
1757 Description: ldlm_error <-> errno conversion.
1758 Details : Add functions to map (rather arbitrary) between LDLM error codes
1759 and standard errno values. CLIO needs this to prevent LDLM specific
1760 constants from escaping ldlm and osc.
1764 Description: Kill unused ldlm_handle2lock_ns() function.
1765 Details : Kill unused ldlm_handle2lock_ns() function.
1769 Description: Add lu_ref support to ldlm_lock
1770 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1771 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1772 and ldlm_resource. LDLM interface has to be changed, because code that
1773 releases a reference on a lock, has to "know" what reference this is.
1774 In the most frequent case
1776 lock = ldlm_handle2lock(handle);
1778 LDLM_LOCK_PUT(lock);
1780 no changes are required. When any other reference (received _not_ from
1781 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1782 instead of LDLM_LOCK_PUT().
1784 Arguably, changes are pervasive, and interface requires some discipline
1785 for proper use. On the other hand, it was very instrumental in finding
1786 a few leaked lock references.
1790 Description: Add ldlm_lock_addref_try().
1791 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1792 attempts to addref a lock that might be being canceled
1797 Description: Add ldlm_weigh_callback().
1798 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1799 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1800 hard-coded `number of pages' logic.
1804 Description: Add lockdep annotations to llog code.
1805 Details : Use appropriately tagged _nested() locking calls in the places
1806 where llog takes more than one ->lgh_lock lock.
1810 Description: Add loi_kms_set().
1811 Details : Wrap kms updates into a helper function.
1815 Description: Constify instances of struct lsm_operations.
1816 Details : Constify instances of struct lsm_operations.
1820 Description: lu_conf support.
1821 Details : On a server, a file system object is uniquely identified
1822 by a fid, which is sufficient to locate and load all object
1823 state (inode). On a client, on the other hand, more data are
1824 necessary instantiate an object. Change lu_object_find() and
1825 friends to take additional `lu_conf' argument describing object.
1826 Typically this includes layout information.
1830 Description: lu_context fixes.
1831 Details : Introduce new lu_context functions that are needed on the client
1832 side, where some system threads (ptlrpcd) are shared by multiple
1833 modules, and so cannot be stopped during module shutdown.
1837 Description: Add start and stop methods to lu_device_type_operations.
1838 Details : Introduce two new methods in lu_device_type_operations, that are
1839 invoked when first instance of a given type is created and last one
1840 is destroyed respectively. This is need by CLIO.
1844 Description: Add lu_ref support to struct lu_device.
1845 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1846 track leaked references.
1850 Description: Introduce lu_kmem_descr.
1851 Details : lu_kmem_descr and its companion interface allow to create
1852 and destroy a number of kmem caches at once.
1856 Description: Fix lu_object finalization race.
1857 Details : Fix a race between lu_object_find() finding an object and its
1858 concurrent finalization. This race is (most likely) not possible
1859 on the server, but might happen on the client.
1863 Description: Introduce lu_ref interface.
1864 Details : lu_ref is a debugging module allowing to track references to
1865 a given object. It is quite cpu expensive, and has to be
1866 explicitly enabled with --enable-lu_ref. See usage description
1871 Description: Factor lu_site procfs stats into a separate function.
1872 Details : Separate lu_site stats printing code into a separate function
1873 that can be reused on a client.
1877 Description: Constify instances of struct {lu,dt,md}_device_operations.
1878 Details : Constify instances of struct {lu,dt,md}_device_operations.
1882 Description: Introduce struct md_site and move meta-data specific parts of
1883 struct lu_site here.
1884 Details : Move md-specific fields out of struct lu_site into special struct
1885 md_site, so that lu_site can be used on a client.
1889 Description: Kill mdd_lov_destroy().
1890 Details : Remove unused mdd code.
1894 Description: Add st_block checking to multistat.c.
1895 Details : Add st_block checking to multistat.c.
1899 Description: Add lu_ref support to struct obd_device.
1900 Details : Add lu_ref tracking to obd_device.
1904 Description: Kill obd_set_fail_loc().
1905 Details : Remove unused code.
1909 Description: Add special type for ptlrpc_request interpret functions.
1910 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1911 there. Actual usage will be in CLIO.
1915 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1916 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1917 doesn't work with lockdep.
1921 Description: Add rwv.c test program.
1922 Details : New testing program exercising readv(2) and writev(2) (Qian).
1926 Description: Add sendfile.c test program.
1927 Details : New testing program exercising sendfile(2) (Jay).
1931 Description: Ratelimit a message that can be very frequent.
1932 Details : Ratelimit a memory allocation failure message that can
1937 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1938 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1942 Description: Liblustre build fixes.
1943 Details : Liblustre build fixes.
1947 Description: libcfs: add cfs_{need,cond}_resched() interface.
1948 Details : libcfs: add cfs_{need,cond}_resched() definition and
1949 implementations for Linux, NT, and liblustre.
1951 Severity : enhancement
1953 Description: More exported tunables for mballoc
1954 Details : Add support for tunable preallocation window and new tunables for
1955 large/small requests
1959 Description: Detect corruption of block bitmap and checking for preallocations
1960 Details : Checks validity of on-disk block bitmap. Also it does better
1961 checking of number of applied preallocations. When corruption is
1962 found, it turns filesystem readonly to prevent further corruptions.
1966 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1967 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1968 window to 0. If it does hit cache miss, and read-pattern is still
1969 stride-io mode, does not reset the stride window, but also does not
1970 increase the stride window length in this case.
1974 Frequency : only for big-endian servers
1975 Description: Check if system is big-endian while mounting fs with extents feature
1976 Details : Mounting a filesystem with extents feature will fail on big-endian
1977 systems since ext3-based ldiskfs is not supported on big-endian
1978 systems. This can be over-riden with "bigendian_extents" mount option.
1980 Severity : enhancement
1982 Description: The root squash functionality
1983 Details : A security feature, which is to prevent users from being able
1984 to mount lustre on their desktop, run as root, and delete
1985 all of the files in the filesystem. The goal is accomplished by
1986 remapping user id (UID) and group id (GID) of the root user to
1987 a UID and GID specified by the system administartor via Lustre
1988 configuration management server (MGS). The functionality also
1989 allows to specify sets of clients for which the remapping does
1994 Description: Excessive recovery window
1995 Details : With AT enabled, the recovery window can be excessively long (6000+
1996 seconds). To address this problem, we no longer use
1997 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1998 timeout no longer depends on the service time, it is set to
1999 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2000 time via pb_service_time.
2004 Description: Watchdog triggered on MDS failover
2005 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2006 the OSTs know that the MDS "UUID" can be reused for the same export
2007 from a different NID, so we do not need to wait for the export to be
2011 Frequency : rare, only if using MMP with Linux RAID
2013 Description: MMP doesn't work with Linux RAID
2014 Details : While using HA for Lustre servers with Linux RAID, it is possible
2015 that MMP will not detect multiple mounts. To make this work we
2016 need to unplug the device queue in RAID when the MMP block is being
2017 written. Also while reading the MMP block, we should read it from
2018 disk and not the cached one.
2020 Severity : enhancement
2022 Description: open file using fid
2023 Details : A file can be opened using just its fid, like
2024 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2027 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2030 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2031 Details : add proper locking for creation of jbd_4k slab cache
2035 Description: MMP check in ext3_remount() fails without displaying any error
2036 Details : When multiple mount protection fails during remount, proper error
2039 --------------------------------------------------------------------------------
2041 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2043 * Support for kernels:
2044 2.6.5-7.283 (SLES 9),
2045 2.6.9-55.EL (RHEL 4),
2046 2.6.16.46-0.14 (SLES 10),
2047 2.6.18.8 vanilla (kernel.org)
2048 * Client support for unpatched kernels:
2049 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2050 2.6.16 - 2.6.22 vanilla (kernel.org)
2051 * Due to recently discovered recovery problems, we do not recommend
2052 using patchless RHEL 4 clients with this or any earlier release.
2053 * Recommended e2fsprogs version: 1.39.cfs8
2054 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2055 * Starting with this release, the ldiskfs backing filesystem required
2056 by Lustre is now in its own package, lustre-ldiskfs. This package
2057 should be installed. It is versioned separately from Lustre and
2058 may be released separately in future.
2062 Description: block reactivating mgc import until all deactivates complete
2063 Details : Fix race when failing back MDT/MGS to itself (testing)
2066 Frequency : at statup only
2068 Description: mds_lov_synchronize race leads to various problems
2069 Details : simultaneous MDT->OST connections at startup can cause the
2070 sync to abort, leaving the OSC in a bad state.
2072 Severity : enhancement
2074 Description: add optional extra BUILD_VERSION info
2075 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2076 to override the lustre version.
2079 Frequency : 2.6.18 servers only
2081 Description: ll_kern_mount() doesn't release the module reference
2082 Details : The ldiskfs module reference count never drops down to 0
2083 because ll_kern_mount() doesn't release the module reference.
2088 Description: server LBUG when using old ost_num_threads parameter
2089 Details : Accept the old ost_num_threads parameter but warn that it
2090 is deprecated, and fix an off-by-one error that caused an LBUG.
2095 Description: Transient SCSI error results in persistent IO issue
2096 Details : iobuf->dr_error is not reinitialized to 0 between two
2100 Frequency : sometimes when underlying device returns I/O errors
2102 Description: OSTs not going read-only during write failures
2103 Details : OSTs are not remounted read-only when the journal commit threads
2104 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2105 instead of the ext3 wrappers.
2109 Description: poor connect scaling with increasing client count
2110 Details : Don't run filter_grant_sanity_check for more than 100 exports
2111 to improve scaling for large numbers of clients.
2114 Frequency : SLES10 only
2116 Description: sanity-quota.sh quotacheck failed: rc = -22
2117 Details : Quotas cannot be enabled on SLES10.
2120 Frequency : liblustre clients only
2122 Description: getdirentries does not give error when run on compute nodes
2123 Details : getdirentries does not fail when the size specified as an argument
2124 is too small to contain at least one entry
2126 Severity : enhancement
2128 Description: Add LNET router traceability for debug purposes
2129 Details : If a checksum failure occurs with a router as part of the
2130 IO path, the NID of the last router that forwarded the bulk data
2131 is printed so it can be identified.
2136 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2137 Details : Due to a race condition, liblustre clients were occasionally
2138 evicted incorrectly.
2140 Severity : enhancement
2142 Description: lfs setstripe use optional parameters instead of postional
2145 Severity : enhancement
2147 Description: Nanosecond timestamp support for ldiskfs
2148 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2149 resolution timestamps. There is not yet support for this at
2150 the Lustre filesystem level.
2153 Frequency : during server recovery
2155 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2156 Details : request with rq_no_resend flag not awake l_wait_event if they get
2160 Frequency : nfs export on patchless client
2162 Description: connectathon hang when test nfs export over patchless client
2163 Details : Disconnected dentry cannot be found with lookup, so we do not need
2164 to unhash it or make it invalid
2167 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2168 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2169 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2174 Description: open req refcounting wrong on reconnect
2175 Details : If reconnect happened between getting open reply from server and
2176 call to mdc_set_replay_data in ll_file_open, we will schedule
2177 replay for unreferenced request that we are about to free.
2178 Subsequent close will crash in variety of ways.
2179 Check that request is still eligible for replay in
2180 mdc_set_replay_data().
2185 Description: disable writes to filesystem when reading health_check file
2186 Details : the default for reading the health_check proc file has changed
2187 to NOT do a journal transaction and write to disk, because this
2188 can cause reads of the /proc file to hang and block HA state
2189 checking on a healthy but otherwise heavily loaded system. It
2190 is possible to return to the previous behaviour during configure
2191 with --enable-health-write.
2193 Severity : enhancement
2195 Description: 64-bit inode version
2196 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2197 to the inode. This will be required for version-based recovery.
2202 Description: MDS fails to start if a duplicate client export is detected
2203 Details : in some rare cases it was possible for a client to connect to
2204 an MDS multiple times. Upon recovery the MDS would detect this
2205 and fail during startup. Handle this more gracefully.
2207 Severity : enhancement
2209 Description: Add -o localflock option to simulate old noflock
2211 Details : This will achieve local-only flock/fcntl locks
2217 Description: log_commit_thread vs filter_destroy race leads to crash
2218 Details : Take import reference before releasing llog record semaphore
2223 Description: Wrong request locking in request set processing
2224 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2225 imp_delayed_list, in this place should be used imp_lock.
2228 Frequency : when reconnection
2230 Description: Grant Leak when osc reconnect to OST
2231 Details : When osc reconnect ost, OST(filter) should check whether it
2232 should grant more space to client by comparing fed_grant and
2233 cl_avail_grant, and return the granted space to client instead
2234 of "new granted" space, because client will call osc_init_grant
2235 to update the client grant space info.
2238 Frequency : when client reconnect to OST
2240 Description: Grant Leak when osc do resend and replay bulk write
2241 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2242 bulk write request, because the grant info while be sync between
2243 OSC and OST when reconnect, and we should ignore the grant info
2244 these of resend/replay write req.
2249 Description: Grant space more than avaiable left space sometimes.
2250 Details : When then OST is about to be full, if two bulk writing from
2251 different clients came to OST. Accord the avaliable space of the
2252 OST, the first req should be permitted, and the second one
2253 should be denied by ENOSPC. But if the seconde arrived before
2254 the first one is commited. The OST might wrongly permit second
2255 writing, which will cause grant space > avaiable space.
2258 Frequency : when client is evicted
2260 Description: Grant might be wrongly erased when osc is evicted by OST
2261 Details : when the import is evicted by server, it will fork another
2262 thread ptlrpc_invalidate_import_thread to invalidate the
2263 import, where the grant will be set to 0. While the original
2264 thread will update the grant it got when connecting. So if
2265 the former happened latter, the grant will be wrongly errased
2266 because of this race.
2271 Description: Checking Stale with correct fid
2272 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2273 to check whether it is stale, because sometimes, we want the
2274 enqueue happened anyway, and op_data.fid2 will not be initialized.
2276 Severity : enhancement
2278 Description: update patchless client
2279 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2282 Frequency : only with 2.4 kernel
2284 Description: random memory corruption
2285 Details : size of struct ll_inode_info is to big for union inode.u and this
2286 can be cause of random memory corruption.
2291 Description: Memory leak in recovery
2292 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2293 It should also check obd_fail before fsfilt_start, otherwise if
2294 fsfilt_start return -EROFS,(failover mds during mds recovery).
2295 then the req will return with repmsg->transno = 0 and rc = EROFS.
2296 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2297 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2298 be freed no matter whether fsfilt_commit success or not.
2301 Frequency : only with huge count clients
2303 Description: Prevents from taking the superblock lock in llap_from_page for
2305 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2306 for llap_from_page call in ll_removepage prevents from taking the
2307 superblock lock for a soon died page.
2312 Description: Not check open intent error before release open handle
2313 Details : in some rare cases, the open intent error is not checked before
2314 release open handle, which may cause
2315 ASSERTION(open_req->rq_transno != 0), because it tries to release
2316 the failed open handle.
2321 Description: Set cat log bitmap only after create log success.
2322 Details : in some rare cases, the cat log bitmap is set too early. and it
2323 should be set only after create log success.
2327 Description: Accessing a block bevice can re-enable I/O when Lustre is
2328 tearing down a device.
2329 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2334 Description: service threads may hog cpus when there are a lot of requests
2335 Details : Insert cond_resched to give other threads a chance to use some CPU
2340 Description: the cat log was not initialized in recovery
2341 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2342 unlink log on mds will not be initialized until mds post
2343 recovery. And also in mds post recovery, the unlink log will
2344 initialization will be done asynchronausly, so there will be race
2345 between add unlink log and unlink log initialization.
2349 Description: brw_stats were being printed incorrectly
2350 Details : brw_stats were being printed as log2 but all of them were not
2351 recorded as log2. Also remove some code duplication arising from
2352 filter_tally_{read,write}.
2356 Frequency : rare, only in recovery.
2357 Description: ASSERTION(req->rq_type != LI_POISON) failed
2358 Details : imp_lock should be held while iterating over imp_sending_list for
2359 prevent destroy request after get timeout in ptlrpc_queue_wait.
2363 Description: replay-single.sh test 52 fails
2364 Details : A lock's skiplist need to be cleanup when it being unlinked
2365 from its resource list.
2369 Description: Short directio read returns full requested size rather than
2371 Details : Direct I/O operations should return actual amount of bytes
2372 transferred rather than requested size.
2374 Severity : enhancement
2376 Description: metadata RPC reduction (e.g. for rm performance)
2377 Details : decrease the amount of synchronous RPC between clients and servers
2378 by canceling conflicing lock before the operation on the client
2379 and packing thier handles into the main operation RPC to server.
2381 Severity : enhancement
2383 Description: Async OSC create to avoid the blocking unnecessarily.
2384 Details : If a OST has no remain object, system will block on the creating
2385 when need to create a new object on this OST. Now, ways use
2386 pre-created objects when available, instead of blocking on an
2387 empty osc while others are not empty. If we must block, we block
2388 for the shortest possible period of time.
2392 Description: improve handling recoverable errors
2393 Details : if request processig with error which can be recoverable on server
2394 request should be resend, otherwise page released from cache and
2397 Severity : enhancement
2399 Description: refine locking for avoid write wrong info into lov_objid file
2400 Details : fix possible races with add new target and write/update data in
2403 --------------------------------------------------------------------------------
2405 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2410 Frequency : on some architectures
2412 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2413 Details : Uninitialized flags sometimes cause configuration commands to
2417 Frequency : patchless clients only
2419 Description: missing __iget() symbol export
2420 Details : The __iget() symbol export is missing. To avoid the need for
2421 this on patchless clients the deathrow inode reaper is turned
2422 off, and we depend on the VM to clean up old inodes. This
2423 dependency was during via the fix for bug 12181.
2425 --------------------------------------------------------------------------------
2427 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2429 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2430 INTEROPERATE with older versions automatically. In many cases a
2431 special upgrade step is needed. Please read the
2432 user documentation before upgrading any part of a live system.
2433 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2434 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2435 * WARNING: Lustre configuration and startup changes are required with
2436 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2440 Severity : enhancement
2442 Description: MountConf
2443 Details : Lustre configuration is now managed via mkfs and mount
2444 commands instead of lmc and lconf. New obd types (MGS, MGC)
2445 are added for dynamic configuration management. See
2446 https://mail.clusterfs.com/wikis/lustre/MountConf for
2449 Severity : enhancement
2451 Description: dynamic OST addition
2452 Details : OSTs can now be added to a live filesystem
2454 Severity : enhancement
2456 Description: startup order invariance
2457 Details : MDTs and OSTs can be started in any order. Clients only
2458 require the MDT to complete startup.
2460 Severity : enhancement
2462 Description: parallel, asynchronous orphan cleanup
2463 Details : orphan cleanup is now performed in separate threads for each
2464 OST, allowing parallel non-blocking operation.
2466 Severity : enhancement
2468 Description: optimized stripe assignment
2469 Details : stripe assignments are now made based on ost space available,
2470 ost previous usage, and OSS previous usage, in order to try
2471 to optimize storage space and networking resources.
2473 Severity : enhancement
2475 Description: Permanently set tunables
2476 Details : All writable /proc/fs/lustre tunables can now be permanently
2477 set on a per-server basis, at mkfs time or on a live system.
2479 Severity : enhancement
2481 Description: Lustre message v2
2482 Details : Add lustre message format v2.
2484 Severity : enhancement
2486 Description: client OST exclusion list
2487 Details : Clients can be started with a list of OSTs that should be
2488 declared "inactive" for known non-responsive OSTs.
2492 Description: SPEC SFS validation failure on NFS v2 over lustre.
2493 Details : Changes the blocksize for regular files to be 2x RPC size,
2494 and not depend on stripe size.
2496 Severity : enhancement
2498 Description: Multiple MD RPCs in flight.
2499 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2500 To avoid overly-overloading MDS, introduce a limit on number of
2501 MDS RPCs in flight for a single client and add /proc controls
2502 to adjust this limit.
2504 Severity : enhancement
2506 Description: client read/write statistics
2507 Details : Add client read/write call usage stats for performance
2508 analysis of user processes.
2509 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2510 file access. extents_stats shows chunk size distribution.
2511 extents_stats_per_process show chunk size distribution per
2514 Severity : enhancement
2516 Description: mds statistics
2517 Details : Add detailed mds operations statistics in
2518 /proc/fs/lustre/mds/*/stats.
2522 Description: Failure of copying files with lustre special EAs.
2523 Details : Client side always return success for setxattr call for lustre
2524 special xattr (currently only "trusted.lov").
2529 Description: Refcount LNET uuids
2530 Details : The global LNET uuid list grew linearly with every startup;
2531 refcount repeated list entries instead of always adding to
2534 Severity : enhancement
2536 Description: Dynamic service threads
2537 Details : Within a small range, start extra service threads
2538 automatically when the request queue builds up.
2541 Frequency : mixed-endian client/server environments
2543 Description: mixed-endian crashes
2544 Details : The new msg_v2 system had some failures in mixed-endian
2547 Severity : enhancement
2549 Description: Easy OST removal
2550 Details : OSTs can be permanently deactivated with e.g. 'lctl
2551 conf_param lustre-OST0001.osc.active=0'
2553 Severity : enhancement
2555 Description: MGS proc entries
2556 Details : Added basic proc entries for the MGS showing what filesystems
2559 Severity : enhancement
2561 Description: provide MGS failover
2562 Details : Added config lock reacquisition after MGS server failover.
2564 Severity : enhancement
2566 Description: add Linux 2.4 support
2567 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2571 Description: a large application tries to do I/O to the same resource and dies
2572 in the middle of it.
2573 Details : Check the req->rq_arrival time after the call to
2574 ost_brw_lock_get(), but before we do anything about
2575 processing it & sending the BULK transfer request. This
2576 should help move old stale pending locks off the queue as
2577 quickly as obd_timeout.
2580 Frequency : when an incorrect nid is specified during startup
2582 Description: ptlrpc connect to non-existant node causes kernel crash
2583 Details : LNET can't be re-entered from an event callback, which
2584 happened when we expire a message after the export has been
2585 cleaned up. Instead, hand the zombie cleanup off to another
2588 Severity : enhancement
2590 Description: plain/inodebits lock performance improvement
2591 Details : Grouping plain/inodebits in granted list by their request modes
2592 and bits policy, thus improving the performance of search through
2596 Frequency : only if OST filesystem is corrupted
2598 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2599 Details : for a short time RPCs with bulk IO are in the replay list,
2600 but replay of bulk IOs is unimplemented. If the OST filesystem
2601 is corrupted due to disk cache incoherency and then replay is
2602 started it is possible to trip an assertion. Avoid putting
2603 committed RPCs into the replay list at all to avoid this issue.
2606 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2608 Description: System hang on startup
2609 Details : This bug allowed the liblustre (e.g. catamount) client to
2610 return to the app before handling all startup RPCs. This
2611 could leave the node unresponsive to lustre network traffic
2612 and manifested as a server ptllnd timeout.
2614 Severity : enhancement
2616 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2617 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2618 boolean to control whether to print peer debug info when a
2619 client's RPC times out.
2622 Frequency : only for kernels with patches from Lustre below 1.4.3
2624 Description: Remove old rdonly API
2625 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2628 Frequency : only for devices with external journals
2630 Description: Set external device read-only also
2631 Details : During a commanded failover stop, we set the disk device
2632 read-only while the server shuts down. We now also set any
2633 external journal device read-only at the same time.
2636 Frequency : when upgrading from 1.4 while trying to change parameters
2638 Description: The wrong (new) MDC name was used when setting parameters for
2639 upgraded MDT's. Also allows changing of OSC (and MDC)
2640 parameters if --writeconf is specified at tunefs upgrade time.
2643 Frequency : when setting specific ost indicies
2645 Description: QOS code breaks on skipped indicies
2646 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2647 created with --index need not be sequential.
2649 Severity : enhancement
2651 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2652 Details : The uninit_groups feature works in conjunction with the kernel
2653 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2654 up the pass1 processing of e2fsck. This is a read-only feature
2655 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2656 mount filesystems that have had this feature enabled.
2658 Severity : enhancement
2660 Description: Improve multi-block allocation algorithm to avoid fragmentation
2661 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2662 allocation locality and avoid filesystem fragmentation.
2664 ------------------------------------------------------------------------------
2666 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2668 * Support for kernels:
2669 2.4.21-47.0.1.EL (RHEL 3)
2670 2.6.5-7.283 (SLES 9)
2671 2.6.9-42.0.10.EL (RHEL 4)
2672 2.6.12.6 vanilla (kernel.org)
2673 2.6.16.27-0.9 (SLES 10)
2674 * Recommended e2fsprogs version: 1.39.cfs5
2676 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2680 Frequency : occasional, depends on client load and configuration
2681 Bugzilla : 12181, 12203
2682 Description: data loss for recently-modified files
2684 Details : In some cases it is possible that recently written or created
2685 files may not be written to disk in a timely manner (this should
2686 normally be within 30s unless client IO load is very high).
2687 The problem appears as zero-length files or files that are a
2688 multiple of 1MB in size after a client crash or client eviction
2689 that are missing data at the end of the file.
2691 This problem is more likely to be hit on clients where files are
2692 repeatedly created and unlinked in the same directory, clients
2693 have a large amount of RAM, have many CPUs, the filesystem has
2694 many OSTs, the clients are rebooted frequently, and/or the files
2695 are not accessed by other nodes after being written.
2697 The presence of the problem can be detected by looking at
2698 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2699 smaller than the second (nr_unused) then dirty files will not
2700 be flushed automatically to disk. "sync; sleep 10" should be
2701 run several times on the node before unmounting it to update
2702 Lustre (this is also safe to run on nodes without this problem).
2704 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2705 that can cause this same problem, so customers using that kernel
2706 also need to update the kernel in addition to Lustre. In order
2707 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2709 It is normal that files written just before a client crash (less
2710 than 30s) may not yet have been flushed to disk, even for local
2714 Frequency : frequent on thin XT3 nodes
2716 Description: UUID collision on thin XT3 Linux nodes
2717 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2718 randomly, since we relied on an insufficiently-seeded PRNG.
2723 Description: OSS hangs after "All ost request buffers busy"
2724 Details : A deadlock between quota and journal operations caused OSS
2725 hangs after printing "All ost request buffers busy."
2728 Frequency : always on liblustre builds
2730 Description: Cleanup compiler warnings on liblustre
2733 Frequency : always on liblustre builds on XT3
2735 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2739 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2740 > 32000 subdirectories
2741 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2742 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2743 subdirectory count crosses 32000. This will aid e2fsck to
2744 correctly handle more than 32000 subdirectories.
2747 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2749 Description: System hang on startup
2750 Details : This bug allowed the liblustre (e.g. catamount) client to
2751 return to the app before handling all startup RPCs. This
2752 could leave the node unresponsive to lustre network traffic
2753 and manifested as a server ptllnd timeout.
2755 Severity : enhancement
2757 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2758 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2759 boolean to control whether to print peer debug info when a
2760 client's RPC times out.
2765 Description: make O_SYNC working on 2.6 kernels
2766 Details : 2.6 kernels use different method for mark pages for write,
2767 so need add a code to lustre for O_SYNC work.
2772 Description: Failure to close file and release space on NFS
2773 Details : Put inode details into lock acquired in ll_intent_file_open.
2774 Use mdc_intent_lock in ll_intent_open to properly
2775 detect all kind of errors unhandled by mdc_enqueue.
2780 Description: proc file read during shutdown sometimes raced obd removal,
2782 Details : Add lock to prevent obd access after proc file removal.
2785 Frequency : Only for files larger than 4GB on 32-bit clients.
2787 Description: improperly doing page alignment of locks
2788 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2789 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2794 Description: RPCs being resent when they shouldn't be
2795 Details : Some RPCs that should not be resent are being resent. This
2796 can cause inconsistencies in the RPC state machine. Do not
2797 resend such requests.
2800 Frequency : rare, only with NFS export
2802 Description: Crash on NFS re-export node
2803 Details : under very unusual load conditions an assertion is hit in
2804 ll_intent_file_open()
2807 Frequency : only if OST filesystem is corrupted
2809 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2810 Details : for a short time RPCs with bulk IO are in the replay list,
2811 but replay of bulk IOs is unimplemented. If the OST filesystem
2812 is corrupted due to disk cache incoherency and then replay is
2813 started it is possible to trip an assertion. Avoid putting
2814 committed RPCs into the replay list at all to avoid this issue.
2819 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2820 Details : Large single O_DIRECT read and write calls can fail to allocate
2821 a sufficiently large buffer to process the request. In case of
2822 allocation failure the allocation is retried with a smaller
2823 buffer and broken into smaller requests.
2825 Severity : enhancement
2827 Description: Add -o localflock option to simulate old noflock behaviour.
2828 Details : This will achieve local-only flock/fcntl locks coherentness.
2833 Description: versioning check is incomplete
2834 Details : Checking the version difference of client vs. server, report
2835 error if the gap is too big.
2840 Description: add support PG_writeback bit
2841 Details : add support for PG_writeback bit for Lustre, for more carefull
2842 work with page cache in 2.6 kernel. This also fix some deadlocks
2843 and remove hack for work O_SYNC with 2.6 kernel.
2845 Severity : enhancement
2847 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2848 Details : The uninit_groups feature works in conjunction with the kernel
2849 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2850 up the pass1 processing of e2fsck. This is a read-only feature
2851 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2852 mount filesystems that have had this feature enabled.
2854 Severity : enhancement
2856 Description: Improve multi-block allocation algorithm to avoid fragmentation
2857 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2858 allocation locality and avoid filesystem fragmentation.
2860 ------------------------------------------------------------------------------
2862 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2864 * Support for kernels:
2865 2.6.9-42.0.3.EL (RHEL 4)
2866 2.6.5-7.276 (SLES 9)
2867 2.4.21-47.0.1.EL (RHEL 3)
2868 2.6.12.6 vanilla (kernel.org)
2869 2.6.16.21-0.8 (SLES10)
2870 * Recommended e2fsprogs version: 1.39.cfs2-0
2872 * The backwards-compatible /proc/sys/portals symlink has been removed
2873 in this release. Before upgrading, please ensure that you change
2874 any configuration scripts or /etc/sysctl.conf files that access
2875 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2876 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2877 in advance of the upgrade on any system running Lustre 1.4.6 or
2878 newer, since /proc/sys/lnet was added in that version.
2879 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2883 Frequency : only when quota is used
2885 Description: avoid scanning export list for quota master
2886 Details : Change the algorithms to avoid scanning export list in order
2887 to improve the efficiency.
2890 Frequency : MDS failover only, very rarely
2892 Description: "went back in time" messages on mds failover
2893 Details : The greatest transno may be lost when the current operation
2894 finishes with an error (transno==0) and the client's last_rcvd
2895 record is over-written. Save the greatest transno in the
2896 mds_last_transno for this case.
2899 Frequency : always for specific kernels and striping counts
2901 Description: client may get "Matching packet too big" without ACL support
2902 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2903 when trying to access files in certain configurations. The
2904 clients should in fact be denied when mounting because they do
2905 not understand ACLs.
2908 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2910 Description: many clients connecting with IO in progress causes connect timeouts
2911 Details : Avoid synchronous journal commits to avoid delays caused by many
2912 clients connecting/disconnecting when bulk IO is in progress.
2913 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2914 OST_IO_PORTAL to avoid delays behind potentially many pending
2918 Frequency : occasionally with multiple writers to a single file
2920 Description: shared writes to file may result in wrong size reported by stat()
2921 Details : Allow growing of kms when extent lock is cancelled
2924 Frequency : always with random mmap IO to multi-striped file
2926 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2927 Details : Only if the hole is at the end of OST object so that kms is too
2928 small. Fix is to increase kms accordingly in ll_nopage.
2931 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2933 Description: writes to a missing object would leak memory on the OST
2934 Details : If there is an inconsistency between the MDS and OST filesystems,
2935 such that the MDS references an object that doesn't exist, writes
2936 to that object will leak memory due to incorrect cleanup in the
2937 error handling path, eventually running out of memory on the OST.
2942 Description: Creating too long symlink causes lustre errors
2943 Details : Check symlink and name lengths before sending requests to MDS.
2946 Frequency : only if flock is enabled (not on by default)
2948 Description: posix locks not released on fd closure on 2.6.9+
2949 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2950 kernels, this caused such locks not to be released on fd close and
2951 then assertions on fs unmount about still used locks.
2954 Frequency : MDS failover only, very rarely
2956 Description: clients may get ASSERTION(granted_lock != NULL)
2957 Details : When request was taking a long time, and a client was resending
2958 a getattr by name lock request. The were multiple lock requests
2959 with the same client lock handle and
2960 mds_getattr_name->fixup_handle_for_resent_request found one of the
2961 lock handles but later failed with ASSERTION(granted_lock != NULL).
2966 Description: handle->h_buffer_credits > 0, assertion failure
2967 Details : h_buffer_credits is zero after truncate, causing assertion
2968 failure. This patch extends the transaction or creates a new
2972 Frequency : NFS re-export or patchless client
2973 Bugzilla : 11179, 10796
2974 Description: Crash on NFS re-export node (__d_move)
2975 Details : We do not want to hash the dentry if we don't have a lock.
2976 But if this dentry is later used in d_move, we'd hit uninitialised
2977 list head d_hash, so we just do this to init d_hash field but
2978 leave dentry unhashed.
2981 Frequency : NFS re-export or patchless client
2983 Description: NFS exports has problem with symbolic link
2984 Details : lustre client didn't properly install dentry when re-exported
2985 to NFS or running patchless client.
2988 Frequency : NFS re-export or patchless client
2990 Description: Various nfs/patchless fixes.
2991 Details : fixes reuse disconected alias for lookup process - this fixes
2992 warning "find_exported_dentry: npd != pd",
2993 fix permission error with open files at nfs.
2994 fix apply umask when do revalidate.
2997 Frequency : occasional
2999 Description: Crash on NFS re-export node
3000 Details : calling clear_page() on the wrong pointer triggered oops in
3001 generic_mapping_read().
3004 Frequency : rarely, using O_DIRECT IO
3006 Description: unaligned directio crashes client with LASSERT
3007 Details : check for unaligned buffers before trying any requests.
3010 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3012 Description: stale data returned from RAID cache
3013 Details : If only a small amount of IO is done to the RAID device before
3014 reading it again it is possible to get stale data from the RAID
3015 cache instead of reading it from disk.
3018 Frequency : always for sles10 kernel
3020 Description: sles10 support
3021 Details : ll_follow_link: compile fixes and using of nd_set_link
3022 under newer kernels.
3025 Frequency : depends on arch, kernel and compiler version, always on sles10
3028 Description: recursive or deep enough symlinks cause stack overflow
3029 Details : getting rid of large stack-allocated variable in
3033 Frequency : depends on hardware
3035 Description: lustre write performance loss in the SLES10 kernel
3036 Details : the performance loss is caused by using of write barriers in the
3037 ext3 code. The SLES10 kernel turns barrier support on by
3038 default. The fix is to undo that change for ldiskfs.
3040 ------------------------------------------------------------------------------
3042 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3044 * Support for kernels:
3045 2.6.9-42.0.3EL (RHEL 4)
3046 2.6.5-7.276 (SLES 9)
3047 2.4.21-47.0.1.EL (RHEL 3)
3048 2.6.12.6 vanilla (kernel.org)
3052 Frequency : quota enabled and large files being deleted
3054 Description: releasing more than 4GB of quota at once hangs OST
3055 Details : If a user deletes more than 4GB of files on a single OST it
3056 will cause the OST to spin in an infinite loop. Release
3057 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3062 Description: statfs data retrieved from /proc may be stale or zero
3063 Details : When reading per-device statfs data from /proc, in the
3064 {kbytes,files}_{total,free,avail} files, it may appear
3065 as zero or be out of date.
3068 Frequency : systems with MD RAID1 external journal devices
3070 Description: lconf's call to blkid is confused by RAID1 journal devices
3071 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3072 of returning all block devices that match the journal UUID.
3075 Frequency : always, for aggregate stripe size over 4GB
3077 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3078 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3079 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3082 Frequency : always if "lfs find" used on a local file/directory
3084 Description: "lfs find" segfaults if used on a local file/directory
3085 Details : The case where a directory component was not specified wasn't
3086 handled correctly. Handle this properly.
3089 Frequency : always on ppc64
3091 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3092 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3093 perform endianness conversion.
3096 Frequency : rarely (truncate to non-zero file size after write under load)
3097 Bugzilla : 10730, 10687
3098 Description: Files padded with zeros to next 4K multiple
3099 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3100 it is possible that files that are truncated to a non-zero size
3101 immediately after being written are filled with zero bytes beyond
3102 the truncated size. No file data is lost.
3104 Severity : enhancement
3106 Description: Allow recovery/failover for liblustre clients.
3107 Details : liblustre clients were unaware of failover configurations until
3110 Severity : enhancement
3112 Description: user file locks should fail when not mounting with flock option
3113 Details : Set up an error-returning stub in ll_file_operations.lock field
3114 to prevent incorrect behaviour when client is mounted without
3115 flock option. Also, set up properly f_op->flock field for
3119 Frequency : always on ia64
3121 Description: "lfs df" loops on printing out MDS statfs information
3122 Details : The obd_ioctl_data was not initialized and in some systems
3123 this caused a failure during the ioctl that did not return
3124 an error. Initialize the struct and return an error on failure.
3127 Frequency : SLES 9 only
3129 Description: Error of copying files with lustre special EAs as root
3130 Details : Client side always return success for setxattr call for lustre
3131 special xattr (currently only "trusted.lov").
3134 Frequency : rarely on clusters with both ia64+i386 clients
3136 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3137 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3138 clients are doing concurrent writes to the same file it is
3139 possible that the ia64 clients may LASSERT because the OST
3140 extent locks are not PAGE_SIZE aligned. Ensure that grown
3141 locks are always aligned on the request boundary.
3144 Frequency : specific use, occasional
3146 Description: Overwriting in use executable truncates on-disk binary image
3147 Details : If one node attempts to overwrite an executable in use by
3148 another node, we now correctly return ETXTBSY instead of
3149 truncating the file.
3151 Severity : enhancement
3153 Description: Async OSC create to avoid the blocking unnecessarily.
3154 Details : If a OST has no remain object, system will block on the creating
3155 when need to create a new object on this OST. Now, ways use
3156 pre-created objects when available, instead of blocking on an
3157 empty osc while others are not empty. If we must block, we block
3158 for the shortest possible period of time.
3163 Description: chmod on Lustre root is propagated to other clients
3164 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3165 invalid by the follow_mount time.
3168 Frequency : liblustre clients only
3170 Description: Race in 'instant cancel' lock handling could lead to such locks
3171 never to be granted in case of SMP MDS
3172 Details : Do not destroy not yet granted but cbpending locks in
3176 Frequency : replay/resend of open
3178 Description: non null lock assetion failure in mds_intent_policy
3179 Details : Trying to replay/resend lockless open requests resulted in
3180 mds_open() returning 0 with no lock. Now it sets a flag if
3181 a lock is going to be returned.
3183 Severity : enhancement
3185 Description: Checksum enhancements
3186 Details : New checksum enhancements allow for resending RPCs that failed
3189 Severity : enhancement
3191 Description: Tunables on number of dirty pages in cacche
3192 Details : Allow to set limit on number of dirty pages cached.
3197 Description: client crash on unmount - lock still has references
3198 Details : In some error handling cases it was possible to leak a lock
3199 reference on a client while accessing a file. This was not
3200 harmful to the client during operation, but would cause the
3201 client to crash when the filesystem is unmounted.
3204 Frequency : specific case, rare
3206 Description: ETXTBSY on mds though file not in use by client
3207 Details : ETXTBSY is no longer incorrectly returned when attempting to
3208 chmod or chown a directory that the user previously tried to
3209 execute or a currently-executing binary.
3212 Frequency : extremely rare except on liblustre-based clients
3214 Description: Lustre space not freed when files are deleted
3215 Details : Clean up open-unlinked files after client eviction. Previously
3216 the unlink was skipped and the files remained as orphans.
3221 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3222 Details : In some cases it was possible to send a blocking callback to
3223 a client doing a glimpse, even though that client didn't get
3224 a lock granted. When the glimpse lock is cancelled on the OST
3225 the freed lock is left on the waiting list and corrupted the list.
3228 Frequency : all core dumps
3230 Description: Broke core dumps to lustre
3231 Details : Negative dentry may be unhashed if parent does not have UPDATE
3232 lock, but some callers, e.g. do_coredump, expect dentry to be
3233 hashed after successful create, hash it in ll_create_it.
3235 ------------------------------------------------------------------------------
3237 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3239 * Support for kernels:
3240 2.6.9-42.0.2.EL (RHEL 4)
3241 2.6.5-7.276 (SLES 9)
3242 2.4.21-40.EL (RHEL 3)
3243 2.6.12.6 vanilla (kernel.org)
3247 Frequency : always on RHEL 3
3249 Description: Number of open files grows over time
3250 Details : The number of open files grows over time, whether or not
3251 Lustre is started. This was due to a filp leak introduced
3252 by one of our kernel patches.
3254 ------------------------------------------------------------------------------
3256 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3258 * Support for kernels:
3259 2.6.9-42.EL (RHEL 4)
3260 2.6.5-7.276 (SLES 9)
3261 2.4.21-40.EL (RHEL 3)
3262 2.6.12.6 vanilla (kernel.org)
3267 Bugzilla : 5719, 9635, 9792, 9684
3268 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3269 Details : If a server is under heavy load and cannot reply to new
3270 connection requests before the client resends the (re)connect,
3271 the connection handling code can behave badly if two service
3272 threads are concurrently handing separate (re)connections from
3273 the same client. Add better locking to the connection handling
3274 code, and ensure that only a single connection will be processed
3275 for a given client UUID, even if the lock is dropped.
3277 Severity : enhancement
3279 Description: add TCP zero-copy support to kernel
3280 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3281 sends if the hardware supports scatter-gather and checksumming.
3282 This allows socklnd to do client-write and server-read more
3283 efficiently and reduce CPU utilization from skbuf copying.
3286 Frequency : only if NFS exporting from client
3288 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3289 Details : If setting attributes on a file created under NFS that had
3290 never been opened it would be possible to oops the client
3291 if the file had no objects.
3294 Frequency : always for liblustre
3296 Description: liblustre client does MDS+OSTs setattr RPC for each write
3297 Details : When doing a write from a liblustre client, the client
3298 incorrectly issued an RPC to the MDS and each OST the file was
3299 striped over in order to update the timestamps. When writing
3300 with small chunks and many clients this could overwhelm the MDS
3301 with RPCs. In all cases it would slow down the write because
3302 these RPCs are unnecessary.
3304 Severity : enhancement
3306 Description: allow number of MDS service threads to be changed at module load
3307 Details : It is now possible to change the number of MDS service threads
3308 running. Adding "options mds mds_num_threads={N}" to the MDS's
3309 /etc/modprobe.conf will set the number of threads for the next
3310 time Lustre is restarted (assuming the "mds" module is also
3311 reloaded at that time). The default number of threads will
3312 stay the same, 32 for most systems.
3317 Description: OST crash if filesystem is unformatted or corrupt
3318 Details : If an OST is started on a device that has never been formatted
3319 or if the filesystem is corrupt and cannot even mount then the
3320 error handling cleanup routines would dereference a NULL pointer.
3325 Description: NULL pointer deref in llap_from_page.
3326 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3327 incorrect) mapping if the page was truncated/reclaimed while it was
3328 searched for. Check for this condition and skip such pages when
3329 doing readahead. Introduce extra check to llap_from_page() to
3330 verify page->mapping->host is non-NULL (so page is not anonymous).
3333 Frequency : Sometimes when using sys_sendfile
3335 Description: "page not covered by a lock" warnings from ll_readpage
3336 Details : sendfile called ll_readpage without right page locks present.
3337 Now we introduced ll_file_sendfile that does necessary locking
3338 around call to generic_file_sendfile() much like we do in
3342 Frequency : with certain MDS communication failures at client mount time
3344 Description: NULL pointer deref after failed client mount
3345 Details : a client connection request may delayed by the network layer
3346 and not be sent until after the PTLRPC layer has timed out the
3347 request. If the client fails the mount immediately it will try
3348 to clean up before the network times out the request. Add a
3349 reference from the request import to the obd device and delay
3350 the cleanup until the network drops the request.
3353 Frequency : occasionally during client (re)connect
3355 Description: assertion failure during client (re)connect
3356 Details : processing a client connection request may be delayed by the
3357 client or server longer than the client connect timeout. This
3358 causes the client to resend the connection request. If the
3359 original connection request is replied in this interval, the
3360 client may trip an assertion failure in ptlrpc_connect_interpret()
3361 which thought it would be the only running connect process.
3364 Frequency : only with obd_echo servers and clients that are rebooted
3366 Description: kernel BUG accessing uninitialized data structure
3367 Details : When running an obd_echo server it did not start the ping_evictor
3368 thread, and when a client was evicted an uninitialized data
3369 structure was accessed. Start the ping_evictor in the RPC
3370 service startup instead of the OBD startup.
3372 Severity : enhancement
3373 Bugzilla : 10193 (patchless)
3374 Description: Remove dependency on various unexported kernel interfaces.
3375 Details : No longer need reparent_to_init, exit_mm, exit_files,
3376 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3379 Frequency : rare (only users of deprecated and unsupported LDAP config)
3381 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3382 Details : LDAP apparently contains 'lustreName' attributes instead of
3383 'name'. A simple remapping of the name is sufficient.
3386 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3388 Description: waiting_locks_callback trips kernel BUG if client is evicted
3389 Details : Running with the dump_on_timeout debug flag turned on makes
3390 it possible that the waiting_locks_callback() can try to dump
3391 the Lustre kernel debug logs from an interrupt handler. Defer
3392 this log dumping to the expired_lock_main() thread.
3394 Severity : enhancement
3396 Description: Support NFS exporting on 2.6 kernels.
3397 Details : Implement non-rawops metadata methods for NFS server to use without
3398 changing NFS server code.
3401 Frequency : very rare (synthetic metadata workload only)
3403 Description: two racing renames might cause an MDS thread to deadlock
3404 Details : Running the "racer" program may cause one MDS thread to rename
3405 a file from being the source of a rename to being the target of
3406 a rename at exactly the same time that another thread is doing
3407 so, and the second thread has already enqueued these locks after
3408 doing a lookup of the target and is trying to relock them in
3409 order. Ensure that we don't try to re-lock the same resource.
3412 Frequency : only very large systems with liblustre clients
3414 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3415 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3416 all OSTs in parallel. This allows the eviction of stale liblustre
3417 clients to proceed much faster than if they were done in series,
3418 and also offers similar improvements for other set_info RPCs.
3423 Description: excessive CPU usage during initial read phase on client
3424 Details : During the initial read phase on a client, it would agressively
3425 retry readahead on the file, consuming too much CPU and impacting
3426 performance (since 1.4.5.8). Improve the readahead algorithm
3427 to avoid this, and also improve some other common cases (read
3428 of small files in particular, where "small" is files smaller than
3429 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3434 Description: MDS crash when receiving packet with unknown intent.
3435 Details : Do not LBUG in unknown intent case, just return -EFAULT
3437 Severity : enhancement
3438 Bugzilla : 9293, 9385
3439 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3440 Details : Do not serialize getattr (non-intent version) and statfs.
3443 Frequency : occasional, when OST network is overloaded/intermittent
3445 Description: client evicted by OST after bulk IO timeout
3446 Details : If a client sends a bulk IO request (read or write) the OST
3447 may evict the client if it is unresposive to its data GET/PUT
3448 request. This is incorrect if the network is overloaded (takes
3449 too long to transfer the RPC data) or dropped the OST GET/PUT
3450 request. There is no need to evict the client at all, since
3451 the pinger and/or lock callbacks will handle this, and the
3452 client can restart the bulk request.
3455 Frequency : Always when mmapping file with no objects
3457 Description: client crashes when mmapping file with no objects
3458 Details : Check that we actually have objects in a file before doing any
3459 operations on objects in ll_vm_open, ll_vm_close and
3465 Description: Request leak when working with deleted CWD
3466 Details : Introduce advanced request refcount tracking for requests
3467 referenced from lustre intent.
3469 Severity : Enhancement
3471 Description: Cache open file handles on client.
3472 Details : MDS now will return special lock along with openhandle, if
3473 requested and client is allowed to hold openhandle, even if unused,
3474 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3475 closing files for every read/write openration.
3477 Severity : Enhancement
3479 Description: Cache open negative dentries on client when possible.
3480 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3481 negative dentries on lock revocation.
3486 Description: Remounting a client read-only wasn't possible with a zconf mount
3487 Details : It wasn't possible to remount a client read-only with llmount.
3489 Severity : enhancement
3490 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3491 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3492 MPICH is included to add Lustre-specific ADIO interfaces.
3493 This is based closely on the UFS ADIO layer and only differs
3494 in file creation, in order to allow the OST striping to be set.
3495 This is user-contributed code and not supported by CFS.
3500 Description: extended inode attributes (immutable, append-only) work improperly
3501 when 2.4 and 2.6 kernels are used on client/server or vice versa
3502 Details : Introduce kernel-independent values for these flags.
3504 Severity : enhancement
3507 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3508 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3509 granularity. Now, Fractional MB granularity is be supported,
3510 this is very useful for low memory system.
3512 Severity : enhancement
3514 Description: Getattr by fid
3515 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3516 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3517 have locks and avoids getattr rpc requests for it on every stat.
3520 Frequency : Always, for filesystems larger than 2TB
3522 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3523 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3524 filesystems larger than 2TB to be mounted with mballoc enabled.
3527 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3529 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3530 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3531 filesystems larger than 2TB is mounted then the kernel will
3532 silently corrupt the start of the filesystem. CONFIG_LBD is
3533 enabled for all CFS-supported kernels, but the possibility of
3534 this happening with a modified kernel config exists.
3536 Severity : enhancement
3538 Description: add client O_DIRECT support for 2.6 kernels
3539 Details : It is now possible to do O_DIRECT reads and writes to files
3540 in the Lustre client mountpoint on 2.6 kernel clients.
3542 Severity : enhancement
3544 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3545 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3546 parallel, not waiting for response from every OST before sending
3547 a rpc to the next OST.
3552 Description: setattr vs write race when updating file timestamps
3553 Details : Client processes that update a file timestamp into the past
3554 right after writing to the file (e.g. tar) it is possible that
3555 the updated file modification time can be reset to the current
3556 time due to a race between processing the setattr and write RPC.
3558 Severity : enhancement
3560 Description: Bring 'lfs find' closer in line with regular Linux find.
3561 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3562 -print0 options and obtains all the needed info through the lustre
3565 Severity : enhancement
3567 Description: support up to 1024 configured devices on one node
3568 Details : change obd_dev array from statically allocated to dynamically
3569 allocated structs as they are first used to reduce memory usage
3574 Description: Flush dirty partially truncated pages during truncate
3575 Details : Immediatelly flush partially truncated pages in filter_setattr,
3576 this way we completely avoid having any pages in page cache on OST
3577 and can retire ugly workarounds during writes to flush such pages.
3582 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3583 Details : protect inode from truncation within vfs_unlink() context
3584 just take a reference before calling vfs_unlink() and release it
3585 when parent's i_sem is free.
3590 Description: last_id value checked outside lock on OST caused LASSERT failure
3591 Details : If there were multiple MDS->OST object precreate requests in
3592 flight, it was possible that the OST's last object id was checked
3593 outside a lock and incorrectly tripped an assertion. Move checks
3594 inside locks, and discard old precreate requests.
3597 Frequency : always, if extents are used on OSTs
3599 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3600 Details : OSTs using the extents format would not zero the high 16 bits of
3601 the index physical block number. This is not a problem for any
3602 OST filesystems smaller than 16TB, and no kernels support ext3
3603 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3604 new/modified files) and can be fixed for existing filesystems
3605 with e2fsprogs-1.39-cfs1.
3610 Description: import connection selection may be incorrect if timer wraps
3611 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3612 import connections to be ignored if the 32-bit jiffies counter
3613 wraps. Use a 64-bit jiffies counter.
3616 Frequency : very large clusters immediately after boot
3618 Description: LNET request buffers exhausted under heavy short-term load
3619 Details : If a large number of client requests are generated on a service
3620 that has previously never seen so many requests it is possible
3621 that the request buffer growth cannot keep up with the spike in
3622 demand. Instead of dropping incoming requests, they are held in
3623 the LND until the RPC service can accept more requests.
3626 Frequency : Sometimes during replay
3628 Description: Assertion failure in ll_local_open after replay.
3629 Details : If replay happened on an open request reply before we were able
3630 to set replay handler, reply will become not swabbed tripping the
3631 assertion in ll_local_open. Now we set the handler right after
3632 recognising of open request
3635 Frequency : very rare
3637 Description: kernel reports "badness in vsnprintf"
3638 Details : Reading from the "recovery_status" /proc file in small chunks
3639 may cause a negative length in lprocfs_obd_rd_recovery_status()
3640 call to vsnprintf() (which is otherwise harmless). Exit early
3641 if there is no more space in the output buffer.
3643 Severity : enhancement
3645 Description: clear OBD RPC statistics by writing to them
3646 Details : It is now possible to clear the OBD RPC statistics by writing
3647 to the "stats" file.
3652 Description: Client mtime is not the same on different clients after utimes
3653 Details : In some cases, the client was using the utimes() syscall on
3654 a file cached on another node. The clients now validate the
3655 ctime from the MDS + OSTs to determine which one is right.
3660 Description: Inability to activate failout mode
3661 Details : lconf script incorrectly assumed that in pythong string's numeric
3662 value is used in comparisons.
3665 Frequency : always with multiple stripes per file
3667 Description: Inefficient object allocation for mutli-stripe files
3668 Details : When selecting which OSTs to stripe files over, for files with
3669 a stripe count that divides evenly into the number of OSTs,
3670 the MDS is always picking the same starting OST for each file.
3671 Return the OST selection heuristic to the original design.
3676 Description: mount failures may take full timeout to return an error
3677 Details : Under some heavy load conditions it is possible that a
3678 failed mount can wait for the full obd_timeout interval,
3679 possibly several minutes, before reporting an error.
3680 Instead return an error as soon as the status is known.
3682 Frequency : quota enabled and large files being deleted
3684 Description: releasing more than 4GB of quota at once hangs OST
3685 Details : If a user deletes more than 4GB of files on a single OST it
3686 will cause the OST to spin in an infinite loop. Release
3687 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3692 Description: statfs data retrieved from /proc may be stale or zero
3693 Details : When reading per-device statfs data from /proc, in the
3694 {kbytes,files}_{total,free,avail} files, it may appear
3695 as zero or be out of date.
3698 Frequency : systems with MD RAID1 external journal devices
3700 Description: lconf's call to blkid is confused by RAID1 journal devices
3701 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3702 of returning all block devices that match the journal UUID.
3705 Frequency : always, for aggregate stripe size over 4GB
3707 Description: assertion fails when trying to use 4GB stripe size
3708 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3709 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3712 Frequency : always on ppc64
3714 Description: the first write on an ext3 filesystem with mballoc got stuck
3715 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3716 perform endianness conversion.
3718 ------------------------------------------------------------------------------
3720 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3722 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3723 INTEROPERATE with older versions automatically. Please read the
3724 user documentation before upgrading any part of a live system.
3725 * WARNING: Lustre networking configuration changes are required with
3726 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3729 * Support for kernels:
3730 2.6.9-22.0.2.EL (RHEL 4)
3731 2.6.5-7.244 (SLES 9)
3732 2.6.12.6 vanilla (kernel.org)
3735 Severity : enhancement
3736 Bugzilla : 7981/8208
3737 Description: Introduced Lustre Networking (LNET)
3738 Details : LNET is new networking infrastructure for Lustre, it includes
3739 a reorganized network configuration mode (see the user
3740 documentation for full details) as well as support for routing
3741 between different network fabrics. Lustre Networking Devices
3742 (LNDs) for the supported network fabrics have also been
3743 created for this new infrastructure.
3745 Severity : enhancement
3746 Description: Introduced Access control lists
3747 Details : clients can set ACLs on files and directories in order to have
3748 more fine-grained permissions than the standard Unix UGO+RWX.
3749 The MDS must be started with the "-o acl" mount option.
3751 Severity : enhancement
3752 Description: Introduced filesystem quotas
3753 Details : Administrators may now establish per-user quotas on the
3756 Severity : enhancement
3758 Description: Configuration change for the XT3
3759 The PTLLND is now used to run Lustre over Portals on the XT3
3760 The configure option(s) --with-cray-portals are no longer used.
3761 Rather --with-portals=<path-to-portals-includes> is used to
3762 enable building on the XT3. In addition to enable XT3 specific
3763 features the option --enable-cray-xt3 must be used.
3768 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3769 Details : A race between allocating a new llcd and re-getting the llcd_lock
3770 allowed another thread to grab newly-allocated llcd.
3772 Severity : enhancement
3774 Description: 2.6 OST async journal commit and locking fix to improve performance
3775 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3776 2.6 kernels are now async as they already were in 2.4 kernels so
3777 that they can commit concurrently with the network bulk transfer.
3778 For block-allocated files the filter allocation semaphore is held
3779 to avoid filesystem fragmentation during allocation. BKL lock
3780 removed for 2.6 xattr operations where it is no longer needed.
3785 Description: lconf incorrectly determined whether two IP networks could talk
3786 Details : In some more complicated routing and multiple-network
3787 configurations, lconf will avoid trying to make a network
3788 connection to a disjoint part of the IP space. It was doing the
3789 math incorrectly for one set of cases.
3794 Description: Fix for potential infinite loop processing records in an llog.
3795 Details : If an llog record is corrupted/zeroed, it is possible to loop
3796 forever in llog_process(). Validate the llog record length
3797 and skip the remainder of the block on error.
3800 Frequency : occasional (liblustre only)
3802 Description: liblustre could not open files whose last component is a symlink
3803 Details : sysio_path_walk() would incorrectly pass the open intent to
3804 intermediate path components.
3807 Frequency : rare (liblustre only with non-standard tuning)
3808 Bugzilla : 7201 (7350)
3809 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3810 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3811 after it has been released, only do this for referenced locks
3813 Severity : enhancement
3815 Description: specifying an (invalid) directory default stripe_size of -1
3816 would reset the directory default striping
3817 Details : stripe_size -1 was used internally to signal directory stripe
3818 removal, now use "all default" to signal dir stripe removal
3819 as a directory striping of "all default" is not useful
3822 Frequency : common for large clusters running liblustre clients
3824 Description: doing an ls when liblustre clients are running is slow
3825 Details : sending a glimpse AST to a liblustre client waits for every AST
3826 to time out, as liblustre clients will not respond. Since they
3827 cannot cache data we refresh the OST lock LVB from disk instead.
3829 Severity : enhancement
3831 Description: doing an ls at the same time as file IO can be slow
3832 Details : enqueue and other "small" requests can be blocked behind many
3833 large IO requests. Create a new OST IO portal for non-IO
3834 requests so they can be processed faster.
3837 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3839 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3840 created file with mode 000
3841 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3842 allow an ftruncate() as a non-root user to a file with mode 000.
3843 Lustre now respects this flag to disable mode checks when
3844 truncating a file owned by the user
3847 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3850 Description: Revoking locks from clients that went dead or catatonic might take
3852 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3853 cancellation of such locks instant on servers without waiting for
3854 any reply from clients. Clients drops these locks when cancel
3855 notification from server is received without replying.
3858 Frequency : liblustre-only, when liblustre client dies or becomes busy
3860 Description: Doing ls on Linux clients can take a long time with active
3862 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3863 granting such locks to it in the first place if possible. Locks
3864 are taken by proxy on the OST during the read or write and
3865 dropped immediately afterward. Add connect flags handling, do
3866 not grant locks to liblustre clients for glimpse ASTs.
3868 Severity : enhancement
3870 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3871 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3872 sequential reads and behaves badly with applications doing random
3873 reads. Improve it by reading ahead at least read region, and
3874 avoiding excessive large RPC for small reads.
3876 Severity : enhancement
3878 Description: Creating more than 1000 files for a single job may cause a load
3879 imbalance on the OSTs if there are also a large number of OSTs.
3880 Details : qos_prep_create() uses an OST index reseed value that is an
3881 even multiple of the number of available OSTs so that if the
3882 reseed happens in the middle of the object allocation it will
3883 still utilize the OSTs as uniformly as possible.
3888 Description: OST or MDS may oops in ping_evictor_main()
3889 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3890 but doesn't restart at beginning of obd_exports_timed list
3893 Severity : enhancement
3895 Description: improve by-nid export eviction on the MDS and OST
3896 Details : allow multiple exports with the same NID to be evicted at one
3897 time without re-searching the exports list.
3900 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3902 Description: MDS may oops in groups_free()
3903 Details : in rare race conditions a newly allocated group_info struct is
3904 freed again, and this can be NULL. The 2.4 compatibility code
3905 for groups_free() checked for a NULL pointer, but 2.6 did not.
3908 Frequency : common for liblustre clients doing little filesystem IO
3909 Bugzilla : 9352, 7313
3910 Description: server may evict liblustre clients accessing contended locks
3911 Details : if a client is granted a lock or receives a completion AST
3912 with a blocking AST already set it would not reply to the AST
3913 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3916 Frequency : lfs setstripe, only systems with more than 160 OSTs
3918 Description: unable to set striping with a starting offset beyond OST 160
3919 Details : llapi_create_file() incorrectly limited the starting stripe
3920 index to the maximum single-file stripe count.
3923 Frequency : LDAP users only
3925 Description: lconf did not handle in-kernel recovery with LDAP properly
3926 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3928 Severity : enhancement
3930 Description: bind OST threads to NUMA nodes to improve performance
3931 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3932 node and do their allocations there to localize memory access
3934 Severity : enhancement
3936 Description: llmount can determine client NID directly from Myrinet (GM)
3937 Details : the client NID code from gmnalnid was moved directly into
3938 llmount, removing the need to use this or specifying the
3939 client NID explicitly when mounting GM clients with zeroconf
3942 Frequency : if client is started with down MDS
3944 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3945 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3946 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3947 ptlrpc_import_delayed_req() because we didn't check if the
3948 request was interrupted, and we also didn't break out of the
3949 event loop if there was no timeout
3954 Description: data loss during non-page-aligned writes to a single file from
3955 both multiple nodes and multiple threads on one node at same time
3956 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3957 inconsistency led to false short-reads, that were cached and later
3958 used by ->prepare_write() to fill in partially written page,
3959 leading to data loss.
3962 Frequency : always, if lconf --abort_recovery used
3964 Description: lconf --abort_recovery fails with 'Operation not supported'
3965 Details : lconf was attempting to abort recovery on the MDT device and not
3968 Severity : enhancement
3970 Description: remove cleanup logs
3971 Details : replace lconf-generated cleanup logs with lustre internal
3972 cleanup routines. Eliminates the need for client-cleanup and
3975 Severity : enhancement
3977 Description: add support for EAs (user and system) on lustre filesystems
3978 Details : it is now possible to store extended attributes in the Lustre
3979 client filesystem, and with the user_xattr mount option it
3980 is possible to allow users to store EAs on their files also
3982 Severity : enhancement
3984 Description: Add possibility (config option) to show minimal available OST free
3986 Details : When compiled with --enable-mindf configure option, statfs(2)
3987 (and so, df) will return least minimal free space available from
3988 all OSTs as amount of free space on FS, instead of summary of
3989 free spaces of all OSTs.
3991 Severity : enhancement
3993 Description: do not expand extent locks acquired on OST-side
3994 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3995 by server: they are not cached anyway.
3998 Frequency : when mmap is used/binaries executed from Lustre
4000 Description: Unmmap pages before throwing them away from read cache.
4001 Details : llap_shrink cache now attempts to unmap pages before discarding
4002 them (if unmapping failed - do not discard). SLES9 kernel has
4003 extra checks that trigger if this unmapping is not done first.
4008 Description: lconf didn't resolve symlinks before checking to see whether a
4009 given mountpoint was already in use
4012 Frequency : when migrating failover services
4013 Bugzilla : 6395, 9514
4014 Description: When migrating a subset of services from a node (e.g. failback
4015 from a failover service node) the remaining services would
4016 time out and evict clients.
4017 Details : lconf --force (implied by --failover) sets the global obd_timeout
4018 to 5 seconds in order to quickly disconnect, but this caused
4019 other RPCs to time out too quickly. Do not change the global
4020 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4022 Severity : enhancement
4023 Frequency : if MDS is started with down OST
4024 Bugzilla : 9439,5706
4025 Description: Allow startup/shutdown of an MDS without depending on the
4026 availability of the OSTs.
4027 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4028 Add appropriate locking and lov-osc refcounts for safe
4029 cleaning. Add osc abort_inflight calls in case the
4030 synchronize never started.
4033 Frequency : occasional (Cray XT3 only)
4035 Description: root not authorized to access files in CRAY_PORTALS environment
4036 Details : The client process capabilities were not honoured on the MDS in
4037 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4038 been accessed by an authorized user then root was able to access
4039 the file on the local client also. The root user capabilities
4040 are now allowed on the MDS, as this environment has secure UID.
4043 Frequency : occasional
4045 Description: ldiskfs "too long searching" message happens too often
4046 Details : A debugging message (otherwise harmless) prints too often on
4047 the OST console. This has been reduced to only happen when
4048 there are fragmentation problems on the filesystem.
4053 Description: Division by zero in statfs when all OSCs are inactive
4054 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4055 lov_statfs divides by value returned by lov_get_stripecnt().
4059 Bugzilla : 9489, 3273
4060 Description: First write from each client to each OST was only 4kB in size,
4061 to initialize client writeback cache, which caused sub-optimal
4062 RPCs and poor layout on disk for the first writen file.
4063 Details : Clients now request an initial cache grant at (re)connect time
4064 and so that they can start streaming writes to the cache right
4065 away and always do full-sized RPCs if there is enough data.
4066 If the OST is rebooted the client also re-establishes its grant
4067 so that client cached writes will be honoured under the grant.
4072 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4073 Details : Now I/O RPCs go to different portal number and (presumably) fast
4074 lock requests (and glimses) and other RPCs get their own service
4075 threads pool that should be able to service those RPCs
4078 Severity : enhancement
4080 Description: Ability to exchange lustre version between client and servers and
4081 issue warnings at client side if client is too old. Also for
4082 liblustre clients there is ability to refuse connection of too old
4084 Details : New 'version' field is added to connect data structure that is
4085 filled with version info. That info is later checked by server and
4089 Frequency : rare, liblustre only.
4090 Bugzilla : 9296, 9581
4091 Description: Two simultaneous writes from liblustre at offset within same page
4092 might proceed at the same time overwriting eachother with stale
4094 Details : I/O lock withing llu_file_prwv was released too early, before data
4095 actually was hitting the wire. Extended lock-holding time until
4096 server acknowledges receiving data.
4099 Frequency : extremely rare. Never observed in practice.
4101 Description: avoid generating lustre_handle cookie of 0.
4102 Details : class_handle_hash() generates handle cookies by incrementing
4103 global counter, and can hit 0 occasionaly (this is unlikely, but
4104 not impossible, because initial value of cookie counter is
4105 selected randonly). Value of 0 is used as a sentinel meaning
4106 "unassigned handle" --- avoid it. Also coalesce two critical
4107 sections in this function into one.
4109 Severity : enhancement
4111 Description: allow liblustre clients to delegate truncate locking to OST
4112 Details : To avoid overhead of locking, liblustre client instructs OST to
4113 take extent lock in ost_punch() on client's behalf. New connection
4114 flag is added to handle backward compatibility.
4116 Severity : enhancement
4117 Bugzilla : 4928, 7341, 9758
4118 Description: allow number of OST service threads to be specified
4119 Details : a module parameter allows the number of OST service threads
4120 to be specified via "options ost ost_num_threads={N}" in the
4121 OSS's /etc/modules.conf or /etc/modprobe.conf.
4125 Bugzilla : 6146, 9635, 9895
4126 Description: servers crash with bad pointer in target_handle_connect()
4127 Details : In rare cases when a client is reconnecting it was possible that
4128 the connection request was the last reference for that export.
4129 We would temporarily drop the export reference and get a new
4130 one, but this may have been the last reference and the export
4131 was just destroyed. Get new reference before dropping old one.
4133 Severity : enhancement
4134 Frequency : if client is started with failover MDS
4136 Description: Allow multiple MDS hostnames in the mount command
4137 Details : Try to read the configuration from all specified MDS
4138 hostnames during a client mount in case the "primary"
4141 Severity : enhancement
4143 Description: Stop sending data to evicted clients as soon as possible.
4144 Details : Check if the client we are about to send or are sending data to
4145 was evicted already. (Check is done every second of waiting,
4146 for which l_wait_event interface was extended to allow checking
4147 of exit condition at specified intervals).
4150 Frequency : rare, normally only when NFS exporting is done from client
4152 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4154 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4157 Frequency : always, if srand() is called before liblustre initialization
4159 Description: Liblustre uses system PRNG disturbing its usage by user application
4160 Details : Introduce internal to lustre fast and high-quality PRNG for
4161 lustre usage and make liblustre and some other places in generic
4162 lustre code to use it.
4164 Severity : enhancement
4165 Bugzilla : 9477, 9557, 9870
4166 Description: Verify that the MDS configuration logs are updated when xml is
4167 Details : Check if the .xml configuration logs are newer than the config
4168 logs stored on the MDS and report an error if this is the case.
4169 Request --write-conf, or allow starting with --old_conf.
4171 Severity : enhancement
4173 Description: Handle symlinks in the path when checking if Lustre is mounted.
4174 Details : Resolve intermediate symlinks when checking if a client has
4175 mounted a filesystem to avoid duplicate client mounts.
4180 Description: lconf can hit an error exception but still return success.
4181 Details : The lconf command catches the Command error exception at the top
4182 level script context and will exit with the associated exit
4183 status, but doesn't ensure that this exit status is non-zero.
4188 Description: failure of ptlrpc thread startup can cause oops
4189 Details : Starting a ptlrpc service thread can fail if there are a large
4190 number of threads or the server memory is very fragmented.
4191 Handle this without oopsing.
4194 Frequency : always, only if liblustre and non-default acceptor port was used
4196 Description: liblustre cannot connect to servers with non-default acceptor port
4197 Details : tcpnal_set_default_params() was not called and was therefore
4198 ignoring the environment varaible TCPNAL_PORT, as well as other
4199 TCPNAL_ environment variables
4204 Description: two objects could be created on the same OST for a single file
4205 Details : If an OST is down, in some cases it was possible to create two
4206 objects on a single OST for a single file. No problems other
4207 than potential performance impact and spurious error messages.
4211 Bugzilla : 5681, 9562
4212 Description: Client may oops in ll_unhash_aliases
4213 Details : Client dcache may become inconsistent in race condition.
4214 In some cases "getcwd" can fail if the current directory is
4220 Description: Inode refcounting problems in NFS export code
4221 Details : link_raw functions used to call d_instantiate without obtaining
4222 extra inode reference first.
4226 Bugzilla : 9942, 9903
4227 Description: Referencing freed requests leading to crash, memleaks with NFS.
4228 Details : We used to require that call to ll_revalidate_it was always
4229 followed by ll_lookup_it. Also with revalidate_special() it is
4230 possible to call ll_revalidate_it() twice for the same dentry
4231 even if first occurence returned success. This fix changes semantic
4232 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4233 reference on a request referred from the intent.
4234 ll_intent_release() then releases such a request.
4237 Frequency : rare, normally benchmark loads only
4239 Description: unlinked inodes were kept in memory on the client
4240 Details : If a client is repeatedly creating and unlinking files it
4241 can accumulate a lot of stale inodes in the inode slab cache.
4242 If there is no other client load running this can cause the
4243 client node to run out of memory. Instead flush old inodes
4244 from client cache that have the same inode number as a new inode.
4247 Frequency : SLES9 2.6.5 kernel and long filenames only
4248 Bugzilla : 9969, 10379
4249 Description: utime reports stale NFS file handle
4250 Details : SLES9 uses out-of-dentry names in some cases, which confused
4251 the lustre dentry revalidation. Change it to always use the
4255 Frequency : rare, unless heavy write-truncate concurrency is continuous
4256 Bugzilla : 4180, 6984, 7171, 9963, 9331
4257 Description: OST becomes very slow and/or deadlocked during object unlink
4258 Details : filter_destroy() was holding onto the parent directory lock
4259 while truncating+unlinking objects. For very large objects this
4260 may block other threads for a long time and slow overall OST
4261 responsiveness. It may also be possible to get a lock ordering
4262 deadlock in this case, or run out of journal credits because of
4263 the combined truncate+unlink. Solution is to do object truncate
4264 first in one transaction without parent lock, and then do the
4265 final unlink in a new transaction with the parent lock. This
4266 reduces the lock hold time dramatically.
4269 Frequency : rare, 2.4 kernels only
4271 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4272 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4273 before unmounting their filesystem in order to avoid deadlock.
4274 The kernel_locked() function in 2.4 kernels only checks whether
4275 the kernel lock is held, not whether it is this process that is
4276 holding it as 2.6 kernels do.
4281 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4282 Details : The client ptlrpc code may be trying to reconnect to a down
4283 server before a previous connection attempt has timed out.
4284 Increase the reconnect interval to be longer than the connection
4285 timeout interval to avoid sending duplicate connections to
4289 Frequency : echo_client brw_test command
4291 Description: fix echo_client to work with OST preallocated code
4292 Details : OST preallocation code (5137) didn't take echo_client IO path
4293 into account: echo_client calls filter methods outside of any
4294 OST thread and, hence, there is no per-thread preallocated
4295 pages and buffers to use. Solution: hijack pga pages for IO. As
4296 a byproduct, this avoids unnecessary data copying.
4300 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4301 Description: Client can oops in mdc_commit_close() after open replay
4302 Details : It was possible for the MDS to return an open request with no
4303 transaction number in mds_finish_transno() if the client was
4304 evicted, but without actually returning an error. Clients
4305 would later try to replay that open and may trip an assertion
4306 Simplify the client close codepath, and always return an error
4307 from the MDS in case the open is not successful.
4310 Frequency : rare, 2.6 OSTs only
4312 Description: OST may deadlock under high load on fragmented files
4313 Details : If there was a heavy load and highly-fragmented OST filesystems
4314 it was possible to have all the OST threads deadlock waiting on
4315 allocation of biovecs, because the biovecs were not released
4316 until the entire RPC IO was completed. Instead, release biovecs
4317 as soon as they are complete to ensure forward IO progress.
4319 Severity : enhancement
4321 Description: Support for specifying external journal device at mount
4322 Details : If an OST or MDS device is formatted with an external journal
4323 device, this device major/minor is stored in the ext3 superblock
4324 and may not be valid for failover. Allow detecting and
4325 specifying the external journal at mount time.
4330 Description: Mounting an MDS with pending unlinked files may cause oops
4331 Details : target_finish_recovery() calls mds_postrecov() which returned
4332 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4333 considers this an error and immediately begins cleaning up the
4334 lov, just after starting the mds_lov process
4336 Severity : enhancement
4338 Description: Implement 'lfs df' to report actual free space on per-OST basis
4339 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4340 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4341 usage of objects; '-h' to report in human readable format.
4343 ------------------------------------------------------------------------------
4345 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4352 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4353 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4354 at startup during orphan unlinking. Instead, skip buddy update
4355 before setup, as it will be regenerated after recovery is complete.
4360 Description: If an OST is inactive, its locks might reference stale inodes.
4361 Details : lov_change_cbdata() must iterate over all namespaces, even if
4362 they are inactive to clear inode references from the lock.
4364 Severity : enhancement
4365 Frequency : occasional, if non-standard max_dirty_mb used
4367 Description: Client will block write RPCs if not enough grant
4368 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4369 then the client will block writes while waiting for another RPC
4370 to complete instead of consuming its dirty limit. With change
4371 we get improved performance when max_dirty_mb is small.
4373 Severity : enhancement
4374 Bugzilla : 3389, 6253
4375 Description: Add support for supplementary groups on the MDS.
4376 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4377 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4378 lookups for user supplementary groups into a cache.
4382 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4383 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4384 for liblustre O_EXCL is enforced by MDS.
4388 Description: readdir never returns NULL in liblustre.
4389 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4390 dentry in struct dirent.
4394 Frequency : liblustre only, depends on application IO pattern
4395 Description: liblustre clients evicted if not contacting servers
4396 Details : Don't put liblustre clients into the ping_evictor list, so
4397 they will not be evicted by the pinger ever.
4399 Severity : enhancement
4401 Description: Add ability to evict clients by NID from MDS.
4402 Details : By echoing "nid:$NID" string into
4403 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4404 $NID would be instantly evicted from this MDS and from all active
4405 OSTs connected to it.
4409 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4410 Details : lookup_it_finish() used to query file size from OSTs that was not
4415 Description: service threads change working directory to that of init
4416 Details : Starting lustre service threads may pin the working directory
4417 of the parent thread, making that filesystem busy. Threads
4418 now change to the working directory of init to avoid this.
4422 Frequency : during shutdown only
4423 Description: shutdown with a failed MDS or OST can cause unmount to hang
4424 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4429 Frequency : 2.6 only
4430 Description: chmod/chown may include an extra supplementary group
4431 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4432 supplementary group and if none is specified this is used.
4435 Bugzilla : 5479 (6816)
4437 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4438 Details : If lookup is in progress on a file that is unlinked we might try
4439 to revalidate the inode and fail in revalidate after lookup is
4440 complete and ll_file_open() enqueues the open again but
4441 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4444 Frequency : always, if lconf --abort_recovery used
4446 Description: lconf --abort_recovery fails with 'Operation not supported'
4447 Details : lconf was attempting to abort recovery on the MDT device and not
4450 ------------------------------------------------------------------------------
4452 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4457 Frequency : rare (only unsupported configurations with a node running as an
4459 Bugzilla : 6514, 5137
4460 Description: Mounting a Lustre file system on a node running as an OST could
4462 Details : OSTs now preallocates memory needed to write out data at
4463 startup, instead of when needed, to avoid having to
4464 allocate memory in possibly low memory situations.
4465 Specifically, if the file system is mounted on on OST,
4466 memory pressure could force it to try to write out data,
4467 which it needed to allocate memory to do. Due to the low
4468 memory, it would be unable to do so and the node would
4469 become unresponsive.
4471 Severity : enhancement
4473 Description: Addition of lconf --service command line option
4474 Details : lconf now accepts a '--service <arg>' option, which is
4475 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4477 Severity : enhancement
4479 Description: Failover mode is now the default for OSTs.
4480 Details : By default, OSTs will now run in failover mode. To return to
4481 the old behaviour, add '--failout' to the lmc line for OSTs.
4483 Severity : enhancement
4485 Description: Health checks are now provided for MDS and OSTs
4486 Details : Additional detailed health check information on MSD and OSTs
4487 is now provided through the procfs health_check value.
4490 Frequency : occasional, depends on IO load
4492 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4493 after numerous create/delete cycles
4494 Details : The ext3 inode allocation policy would not allocate new inodes
4495 very well on the OSTs because there are no new directories
4496 being created. Instead we look for groups with free space if
4497 the parent directories are nearly full.
4502 Description: Network or server problems during mount may cause partially
4503 mounted clients instead of returning an error.
4504 Details : The config llog parsing code may overwrite the error return
4505 code during mount error handling, returning success instead
4511 Description: MDS can fail to allocate large reply buffers
4512 Details : After long uptimes the MDS can fail to allocate large reply
4513 buffers (e.g. zconf client mount config records) due to memory
4514 fragmentation or consumption by the buffer cache. Preallocate
4515 some large reply buffers so that these replies can be sent even
4516 under memory pressure.
4520 Frequency : rare (liblustre)
4521 Description: fsx running with liblustre complained that using truncate() to
4522 extend the file doesn't work. This patch corrects that issue.
4523 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4524 ATTR_SIZE and lsm use in llu_setattr_raw.
4528 Frequency : rare, only 2.6 kernels
4529 Description: Unusual file access patterns on the MDS may result in inode
4530 data being lost in very rare circumstances.
4531 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4532 code in the 2.6 kernel caused the inode and/or EA data not to
4533 be read from disk, causing single-file corruption.
4537 Frequency : rare, only 2.6 filesystems using extents
4538 Description: Heavy concurrent write and delete load may cause data corruption.
4539 Details : It was possible under high-load situations to have an extent
4540 metadata block in the block device cache from a just-unlinked
4541 file overwrite a newly-allocated data block. We now unmap any
4542 metadata buffers that alias just-allocated data blocks.
4546 Frequency : filesystems with default stripe_count larger than 77
4547 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4548 Details : lconf specifies an inode size of 4096 bytes when the default
4549 stripe_count is larger than 77. This conflicts with the default
4550 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4551 this case to avoid pinning too much memory for large EAs.
4553 ------------------------------------------------------------------------------
4555 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4560 Frequency : rare (extremely heavy IO load with hundreds of clients)
4562 Description: Client is evicted, gets IO error writing to file
4563 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4564 caused clients to be evicted for AST timeouts. The fixes in
4565 bug 5192 mean we no longer need to have such short AST timeouts
4566 so ldlm_timeout has been increased.
4569 Frequency : occasional during --force or --failover shutdown under load
4570 Bugzilla : 5949, 4834
4571 Description: Server oops/LBUG if stopped with --force or --failover under load
4572 Details : a collection of import/export refcount and cleanup ordering
4573 issues fixed for safer force cleanup
4576 Frequency : only filesystems larger than 120 OSTs
4577 Bugzilla : 5990, 6223
4578 Description: lfs getstripe would oops on a very large filesystem
4579 Details : lov_getconfig used kfree on vmalloc'd memory
4582 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4583 Bugzilla : 6242, 6243
4584 Description: reading from files that had been truncated to a non-zero size
4585 but never opened returned no data
4586 Details : ll_file_read() reads zeros from no-object files to EOF
4591 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4592 Details : There's little harm in aborting MDS/OSS recovery and letting it
4593 try again, so I removed the LASSERT and return an error instead.
4595 Severity : enhancement
4597 Description: New debugging infrastructure for tracking down data corruption
4598 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4599 (b) cover more of the client-side code path, and (c) try to narrow
4600 down where problems occurred
4604 Bugzilla : 3819, 4364, 4397, 6313
4605 Description: Racing close and eviction MDS could cause assertion in mds_close
4606 Details : It was possible to get multiple mfd references during close and
4607 client eviction, leading to one thread referencing a freed mfd.
4609 Severity: : enhancement
4610 Bugzilla : 3262, 6359
4611 Description: Attempts to reconnect to servers are now more aggressive.
4612 Details : This builds on the enhanced upcall-less recovery that was added
4613 in 1.4.2. When trying to reconnect to servers, clients will
4614 now try each server in the failover group every 10 seconds. By
4615 default, clients would previously try one server every 25 seconds.
4620 Description: After recovery, certain operations trigger a failed
4621 assertion on a client.
4622 Details : Failing over an mds, using lconf -d --failover, while a
4623 client was doing a readdir() call would cause the client to
4624 LBUG after recovery completed and the readdir() was resent.
4626 Severity : enhancement
4628 Description: Default groups are now added by lconf
4629 Details : You can now run lconf --group <servicename> without having to
4630 manually add groups with lmc.
4633 Frequency : occasional
4635 Description: Nodes with an elan id of 0 trigger a failed assertion
4638 Frequency : always when accessing e.g. tty/console device nodes
4640 Description: tty and some other devices nodes cannot be used on lustre
4641 Details : file's private_data field is used by device data and lustre
4642 values in there got lost. New field was added to struct file to
4643 store fs-specific private data.
4646 Frequency : when exporting Lustre via NFS
4648 Description: NFSD failed occasionally when looking up a path component
4649 Details : NFSD is looking up ".." which was broken in ext3 directories
4650 that had grown large enough to become hashed.
4653 Frequency : Clusters with multiple interfaces not on the same subnet
4655 Description: Nodes will repeatedly try to reconnect to an interface which it
4656 cannot reach and report an error to the log.
4657 Details : Extra peer list entries will be created by lconf with some peers
4658 unreachable. lconf now validates the peer before adding it.
4661 Frequency : Only if a default stripe is set on the filesystem root.
4663 Description: Setting a default stripe on the filesystem root prevented the
4664 filesystem from being remounted.
4665 Details : The client was sending extra request flags in the root getattr
4666 request and did not allocate a reply buffer for the dir EA.
4669 Frequency : occasional, higher if lots of files are accessed by one client
4670 Bugzilla : 6159, 6097
4671 Description: Client trips assertion regarding lsm mismatch/magic
4672 Details : While revalidating inodes the VFS looks up inodes with ifind()
4673 and in rare cases can find an inode that is being freed.
4674 The ll_test_inode() code will free the lsm during ifind()
4675 when it finds an existing inode and then the VFS later attaches
4676 this free lsm to a new inode.
4680 Bugzilla : 6422, 7030
4681 Description: MDS deadlock between mkdir and client eviction
4682 Details : Creating a new file via mkdir or mknod (starting a transaction
4683 and getting the ns lock) can deadlock with client eviction
4684 (gets ns lock and trying to finish a synchronous transaction).
4687 Frequency : occasional
4688 Description: While starting a server, the fsfilt_ext3 module could not be
4690 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4691 kernels. Previously, lconf would still use the ext3 name
4692 when trying to load modules. Now, it will correctly use
4693 ext3 on 2.4 and ldiskfs on 2.6.
4695 Severity : enhancement
4696 Description: The default stripe count has been changed to 1
4697 Details : The interpretation of the default stripe count (0, to lfs
4698 or lmc) has been changed to mean striping across a single
4699 OST, rather than all available. For general usage we have
4700 found a stripe count of 1 or 2 works best.
4702 Severity : enhancement
4703 Description: Add support for compiling against Cray portals.
4704 Details : Conditional compiling for some areas that are different
4708 Frequency : occasional
4709 Bugzilla : 6409, 6834
4710 Description: Creating files with an explicit stripe count may lead to
4711 a failed assertion on the MDS
4712 Details : If some OSTs are full or unavailable, creating files may
4713 trigger a failed assertion on the MDS. Now, Lustre will
4714 try to use other servers or return an error to the
4718 Frequency : occasional
4720 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4721 caused lock timeout during readahead (since 1.4.2).
4722 Details : Processes doing readahead might match a lock that hasn't been
4723 granted yet if there are overlapping and conflicting lock
4724 requests. The readahead process waits on ungranted lock
4725 (original lock is CBPENDING), while OST waits for that process
4726 to cancel CBPENDING read lock and eventually evicts client.
4728 Severity : enhancement
4730 Description: Initial enabling of flock support for clients
4731 Details : Implements fcntl advisory locking and file status functions.
4732 This feature is provided as an optional mount flag (default
4733 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4734 locking are implemented yet, and those that are are not guaranteed
4735 to be completely correct in production environments.
4736 mount -t lustre -o [flock|noflock] ...
4739 Frequency : occasional
4741 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4742 assertion in the ext3 JBD (journaling) layer.
4743 Details : The b_committed_data struct is protected by the big kernel lock
4744 in 2.4 kernels, serializing journal_commit_transaction() and
4745 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4746 access to this struct. In 2.6 kernels there is finer grained
4747 locking to improve SMP performance of the JBD layer.
4751 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4756 Description: merge and cleanup kernel patches.
4759 -----------------------------------------------------------------------------
4761 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4763 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4764 versions of Lustre. Please update all servers and clients to
4765 version 1.4.2 or later at the same time. You must also run
4766 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4767 to update the configuration logs.
4769 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4770 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4771 - swab llog records of type '0' so we get proper header size/idx (5861)
4772 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4773 - fix rename of one directory over another leaking an inode (5953)
4774 - avoid SetPageDirty on 2.6 (5981)
4775 - don't re-add just-being-destroyed locks to the waiting list (5653)
4776 - when creating new directories, inherit the parent's custom
4777 striping settings if present parent (3048)
4778 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4779 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4780 - don't hold client locks on temporary worklist from l_lru (5666)
4781 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4782 - automatically evict dead clients (5921)
4783 - Update file size properly in create+truncate+fstat case (6196)
4784 - Do not unhash mountpoint dentries, do not allow removal of
4786 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4787 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4788 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4789 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4790 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4791 - Correctly handle reads of files with no objects (6243)
4792 - lctl recover will also mark a device active if deactivate used (5933)
4794 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4795 - allow --write-conf on an MDS with different nettype than client (5619)
4796 - don't write config llogs to MDS for mounts not from that MDS (5617)
4797 - lconf should create multiple TCP connections from a client (5201)
4798 - init scripts are now turned off by default; run chkconfig --on
4799 lustre and chkconfig --on lustrefs to use them
4800 - upcalls are no longer needed for clients to recover to failover
4802 - add --abort-recovery option to lconf to abort recovery on device
4804 - add support for an arbitrary number of OSTs (3026)
4805 - Quota support protocol changes.
4806 - forward compatibility changes to wire structs (6007)
4807 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4808 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4809 - improved "lustre status" script
4810 - initialize blocksize for non-regular files (6062)
4811 - added --disable-server and --disable-client configure options (5782)
4812 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4813 - Vanilla 2.4.29 support
4814 - increase maximum number of obd devices to 520 (6242)
4815 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4816 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4817 SLES 9 2.6.5 kernels
4818 - limit stripes per file to 160 (the maximum EA size) (6093)
4820 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4823 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4824 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4825 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4826 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4827 - small ext3 extents cleanups and fixes (5733)
4828 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4829 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4830 - avoid needless client->OST connect, fix handle mismatch (5317)
4831 - fix DLM error path that led to out-of-sync client, long delays (5779)
4832 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4833 - fix several locking issues related to i_size (5492,5624,5654,5672)
4834 - don't move pending lock onto export if it is already evicted (5683)
4835 - fix kernel oops when creating .foo in unlinked directory (5548)
4836 - fix deadlock in obdfilter statistics vs. object create (5811)
4837 - use time_{before,after} to avoid timer jiffies wrap (5882)
4838 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4839 - Do not leak request if server was not able to process it (5154)
4840 - If mds_open unable to find parent dir, make that negative lookup(5154)
4841 - don't create new directories with extent-mapping (5909, 5936)
4843 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4844 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4845 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4846 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4847 - support for RHEL4 kernel on i686 (5773)
4848 - provide error messages when incompatible logs are encountered (5898)
4850 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4851 * version 1.4.0.10 (1.4.1 release candidate 1)
4853 - don't keep a lock reference when lock is not granted (4238)
4854 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4855 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4856 - drop import inflight refcount on signal_completed_replay error (5255)
4857 - unlock page after async write error during send (3677)
4858 - handle missing objects in filter_preprw_read properly (5265)
4859 - no transno return for symlink open, don't save no-trasno open (3440)
4860 - don't try to complete elan receive that already failed (4012)
4861 - free RPC server reply state on error (5406)
4862 - clean up thread from ptlrpc_start_thread() on error (5160)
4863 - readahead could read extra page into cache that wasn't ejected (5388)
4864 - prevent races in class_attach/setup/cleanup/detach (5260)
4865 - don't dereference de->d_inode after l_dput of de (5458)
4866 - use "int" for stripe value returned from lock_to_stripe (5544)
4867 - mballoc allocation and error-checking fixes in 2.6 (5504)
4868 - block device patches to fix I/O request sizes in 2.6 (5482)
4869 - look up hostnames for IB nals (5602)
4870 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4871 - don't start multiple acceptors for the same port (5277)
4872 - fix incorrect LASSERT in mds_getattr_name (5635)
4873 - export a proc file for general "ping" checking (5628)
4874 - fix "lfs check" to not block when the MDS is down (5628)
4876 - service request history (4965)
4877 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4878 - create an "evict_client" /proc entry on OSTs, like the MDS has
4879 - fix mount usage message, return errors per mount(8) (5168)
4880 - change grep [] to grep "[]" in tests so they work in more UMLs
4881 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4882 - remove ancient LOV_MAGIC_V0 EA support (5047)
4883 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4884 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4885 - allow readahead allocations to fail when low on memory (5383)
4886 - mmap locking landed again, after considerable improvement (2828)
4887 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4889 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4892 - send OST transaction number in read/write reply to free req (4966)
4893 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4894 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4895 - the watchdog thread now runs as interruptible (5246)
4896 - flock/lockf fixes (but it's still disabled, pending 5135)
4897 - don't use EXT3 constants in llite code (5094)
4898 - memory shortage at startup could cause assertion (5176)
4900 - reorganization of lov code
4901 - single portals codebase
4903 - add extents/mballoc support (5025)
4904 - direct I/O reads in the obdfilter (4048)
4905 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4907 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4910 - send OST transaction number in read/write reply to free req (4966)
4911 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4912 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4913 - the watchdog thread now runs as interruptible (5246)
4914 - handle missing objects in filter_preprw_read properly (5265)
4915 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4916 - drop import inflight refcount on signal_completed_replay error (5255)
4917 - unlock page after async write error during send (3677)
4918 - return original error code on reconstructed replies (3761)
4919 - no transno return for symlink open, don't save no-trasno open (3440)
4921 - add pid to ldlm debugging output (4922)
4922 - bump the watchdog timeouts -- we can't handle 30sec yet
4923 - extra debugging for orphan dentry/inode bug (5259)
4925 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4928 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4929 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4930 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4931 - fix NULL dereference in /proc/sys/portals/routes (4827)
4932 - allow failed mdc_close() operations to be interrupted (4561)
4933 - stop precreate on OST before MDS would time out on it (4778)
4934 - don't send partial-page writes before EOF from client (4410)
4935 - discard client grant for sub-page writes on large-page clients (4520)
4936 - don't free dentries not owned by NFS code, check generation (4806)
4937 - fix lsm leak if mds_create_objects() fails (4801)
4938 - limit debug_daemon file size, always print CERROR messages (4789)
4939 - use transno after validating reply (3892)
4940 - process timed out requests if import state changes (3754)
4941 - update mtime on OST during writes, return in glimpse (4829)
4942 - add mkfsoptions to LDAP (4679)
4943 - use ->max_readahead method instead of zapping global ra (5039)
4944 - don't interrupt __l_wait_event() during strace
4946 - add software watchdogs to catch hung threads quickly (4941)
4947 - make lustrefs init script start after nfs is mounted
4948 - fix CWARN/ERROR duplication (4930)
4949 - return async write errors to application if possible (2248)
4950 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4951 - print NAL number in %x format (4645)
4952 - update barely-supported suse-2.4.21-171 series (4842)
4953 - support for sles 9 %post scripts
4954 - support for building 2.6 kernel-source packages
4955 - support for sles km_* packages
4957 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4960 - ignore -ENOENT errors in osc_destroy (3639)
4961 - notify osc create thread that OSC is being cleaned up (4600)
4962 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4963 - reconstruct ost_handle() like mds_handle() (4657)
4964 - create a new thread to do import eviction to avoid deadlock (3969)
4965 - let lconf resolve symlinked-to devices (4629)
4966 - don't unlink "objects" from directory with default EA (4554)
4967 - hold socknal file ref over connect in case target is down (4394)
4968 - allow more than 32000 subdirectories in a single directory (3244)
4969 - fix blocks count for O_DIRECT writes (3751)
4970 - OST returns ENOSPC from object create when no space left (4539)
4971 - don't send truncate RPC if file size isn't changing (4410)
4972 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4973 - bind to privileged port in socknal and tcpnal (3689)
4975 - rate limit CERROR/CWARN console message to avoid overload (4519)
4976 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4977 - basic mmap support (3918)
4978 - kernel patch series update from b1_4 (4711)
4980 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4983 - avoid crash during MDS cleanup with OST shut down (2775)
4984 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4985 - don't use bad inodes on the MDS (3744)
4986 - dynamic object preallocation to improve recovery speed (4236)
4987 - don't hold spinlock over lock dumping or change debug flags (4401)
4988 - don't zero obd_dev when it is force cleaned (3651)
4989 - print grants to console if they go negative (4431)
4990 - "lctl deactivate" will stop automatic recovery attempts (3406)
4991 - look for existing locks in ldlm_handle_enqueue() (3764)
4992 - don't resolve lock handle twice in recovery avoiding race (4401)
4993 - revalidate should check working dir is a directory (4134)
4995 - don't always mark "slow" obdfilter messages as errors (4418)
4997 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5000 - don't close LustreDB during write_conf until it is done (3860)
5001 - fix typo in lconf for_each_profile (3821)
5002 - allow dumping logs from multiple threads at one time (3820)
5003 - don't allow multiple threads in OSC recovery (3812)
5004 - fix debug_size parameters (3864)
5005 - fix mds_postrecov to initialize import for llog ctxt (3121)
5006 - replace config semaphore with spinlock (3306)
5007 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5008 - don't allow enqueue to complete on a destroyed export (3822)
5009 - down write_lock before checking llog header bitmap (3825)
5010 - recover from lock replay timeout (3764)
5011 - up llog sem before sending rpc (3652)
5012 - reduce ns lock hold times when setting kms (3267)
5013 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5014 - fix NULL deref and obd_dev leak on setup error (3312)
5015 - replace some LBUG about llog ops with error handling (3841)
5016 - don't match INVALID dentries from d_lookup and spin (3784)
5017 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5018 - fix invalid assertion in ptlrpc_set_wait (3880)
5020 - add libwrap support for the TCP acceptor (3996)
5021 - add /proc/sys/portals/routes for non-root route listing (3994)
5022 - allow setting MDS UUID in .xml (2580)
5023 - print the stack of a process that LBUGs (4228)
5025 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5028 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5029 - make sure to unset replay flag from failed open requests (3440)
5030 - if default stripe count is 0, use OST count for inode size (3636)
5031 - update parent mtime/ctime on client for create/unlink (2611)
5032 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5033 - free recovery state on server during a forced cleanup (3571)
5034 - unregister_reply for resent reqs (3063)
5035 - loop back devices mounting and status check on 2.6 (3563)
5036 - fix resource-creation race that can provoke i_size == 0 (3513)
5037 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5038 - more debugging for page-accounting assertion (3746)
5039 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5040 - avoid deadlock after precreation failure (3758)
5041 - fix race and lock order deadlock in orphan handling (3450, 3750)
5042 - add validity checks when grabbing inodes from l_ast_data (3599)
5044 - add /proc/.../recovery_status to obdfilter (3428)
5045 - lightweight CDEBUG infrastructure, debug daemon (3668)
5046 - change default OSC RPC parameters to be better on small clusters
5047 - turn off OST read cache for files smaller than 32MB
5048 - install man pages and include them in rpms (3100)
5049 - add new init script for (un)mounting lustre filesystems (2593)
5050 - run chkconfig in %post for init scripts (3701)
5051 - drop scimac NAL (unmaintained)
5053 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5056 - clean kiobufs before and after use (3485)
5057 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5058 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5059 - change init script to fail more clearly if not run as root (1528)
5060 - allow clients to reconnect during replay (1742)
5061 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5062 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5063 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5064 - fix cond_resched (3554)
5065 - only evict unfinished clients after recovery (3515)
5066 - allow bulk resend, prevent data loss (3570)
5067 - dynamic ptlrpc request buffer allocation (2102)
5068 - don't allow unlinking open directory if it isn't empty (2904)
5069 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5070 - remove extraneous obd dereference causing LASSERT failure (3334)
5071 - don't use get_cycles() when creating temp. files on the mds (3156)
5072 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5073 - handle EEXIST for set-stripe, set proper directory name (3336)
5075 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5076 - fix ksocknal_fmb_callback() error messages (2918)
5078 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5081 - don't copy lvb into (possibly NULL) reply on error (2983)
5082 - don't deref dentry after dput, don't free lvb on error (2922)
5083 - use the kms to determine writeback rpc length (2947)
5084 - increment oti_logcookies when osc is inactive (2948)
5085 - update client's i_blocks count via lvb messages (2543)
5086 - handle intent open/close of special files properly (1557)
5087 - mount MDS with errors=remount-ro, like obdfilter (2009)
5088 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5089 - don't use cancelling-locks' kms values (2947)
5090 - use highest lock extent for kms, not last one (2925)
5091 - don't dereference ERR_PTR() dentry in error handling path (3107)
5092 - fix thread race in portals_debug_dumplog() (3122)
5093 - create lprocfs device entries at setup instead of at attach (1519)
5094 - common AST error handler, don't evict client on completion race (3145)
5095 - zero nameidata in detach_mnt in 2.6 (3118)
5096 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5097 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5098 - initialize RPC timeout timer earlier for 2.6 (3219)
5099 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5100 - print nal/nid for unknown nid (3258)
5101 - additional checks for oscc recovery before doing precreate (3284)
5102 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5103 - don't crash in mdc_close for bad permissions on open (3285)
5104 - zero i_rdev for non-device files (3147)
5105 - clear page->private before handing to FS, better assertion (3119)
5106 - tune the read pipeline (3236)
5107 - fix incorrect decref of invalidated dentry (2350)
5108 - provide read-ahead stats and refine rpc in flight stats (3328)
5109 - don't hold journal transaction open across create RPC (3313)
5110 - update atime on MDS at close time (3265)
5111 - close LDAP connection when recovering to avoid server load (3315)
5112 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5113 - don't leak open file on MDS after open resend (3325)
5114 - serialize filter_precreate and filter_destroy_precreated (3329)
5115 - loop device shouldn't call sync_dev() for nul device (3092)
5116 - clear page cache after eviction (2766)
5117 - resynchronize MDS->OST in background (2824)
5118 - refuse to mount the same filesystem twice on same mountpoint (3394)
5119 - allow llmount to create routes for mounting behind routers (3320)
5120 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5121 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5122 - fix rare problem with rename on htree directories (3417)
5124 - allow default OST striping configuration per directory (1414)
5125 - fix compilation for qswnal for 2.6 kernels (3125)
5126 - increase maximum number of MDS request buffers for large systems
5127 - change liblustreapi to be useful for external progs like lfsck (3098)
5128 - increase local configuration timeout for slow disks (3353)
5129 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5131 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5134 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5135 - don't overwrite extent policy data in reply if lock was blocked (2901)
5136 - drop filter export grants atomically with removal from device (2663)
5137 - del obd_self_export from work_list in class_disconnect_exports (2908)
5138 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5139 - swab reply message in mdc_close, other PPC fixes (2464)
5140 - fix destroying of named logs (2325)
5141 - overwrite old logs when running lconf --write_conf (2264)
5142 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5143 - fix race in target_handle_connect (2898)
5144 - mds_reint_create() should take same inode create lock (2926)
5145 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5146 - don't close files for self_export to avoid uninitialized obd (2936)
5147 - allow MDS with the same name as client node (2939)
5148 - hold dentry reference for closed log files for unlink (2325)
5149 - reserve space for all logs during transactions (2059)
5150 - don't evict page beyond end of stripe extent (2925)
5151 - don't oops on a deleted current working directory (2399)
5152 - handle hard links to targets without a parent properly (2517)
5153 - don't dereference NULL lock when racing during eviction (2867)
5154 - don't grow lock extents when lots of conflicting locks (2919)
5156 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5159 - account for cache space usage on clients to avoid data loss (974)
5160 - lfsck support in lustre kernel code (2349)
5161 - reduce journal credits needed for BRW writes (2370)
5162 - orphan handling to avoid losing space on client/server crashes
5163 - ptlrpcd can be blocked, stopping ALL progress (2477)
5164 - use lock value blocks to assist in proper KMS, faster stat (1021)
5165 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5166 - recovery for initial connections (2355)
5167 - fixes for mds_cleanup_orphans (1934)
5168 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5169 - block all file creations until orphan recovery completes (1901)
5170 - client remove rq_connection from request struct (2423)
5171 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5172 - recovery timer race (2670)
5173 - mdc_close recovey bug (2532)
5174 - ptlrpc cleanup bug (2710)
5175 - mds timeout on local locks (2588)
5176 - namespace lock held during RPCs (2431)
5177 - handle interrupted sync write properly (2503)
5178 - don't try to handle a message that hasn't been replied to (2699)
5179 - client assert failure during cleanup after abort recovery (2701)
5180 - leak mdc device after failed mount (2712)
5181 - ptlrpc_check_set allows timedout requests to complete (2714)
5182 - wait for inflight reqs when ptlrpcd finishes (2710)
5183 - make sure unregistered services are removed from the srv_list
5184 - reset bulk XID's when resending them (caught by 1138 test)
5185 - unregister_bulk after timeout
5186 - fix lconf error (2694)
5187 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5188 - readahead locks pages, leaves pending causing memory pressure (2673)
5189 - increase OST request buffers to 4096 on large machines (2729)
5190 - fix up permission of existing directories in simple_mkdir (2661)
5191 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5192 - don't assign transno to errored transactions (2742)
5193 - don't delete objects on OST if given a bogus objid from MDS (2751)
5194 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5195 - if rq_no_resend, then timeout request after recovery (2432)
5196 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5197 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5198 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5199 - MDS recovery shouldn't delete objects that it has given out (2730)
5200 - if enqueue arrives after completion, don't clobber LVB (2819)
5201 - don't unlock pages twice when trigger_group_io returns error (2814)
5202 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5203 - don't write pages to disk if there was an error (1450)
5204 - don't ping imports that have recovery disabled (2676)
5205 - take buffered bytes into account when balancing socknal conn (2817)
5206 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5207 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5208 - remove little-endian swabbing of llog records (1987)
5209 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5210 - retry reposting request buffers if they fail (1191)
5211 - grow extent at grant time to avoid granting a revoked lock (2809)
5212 - lock revoke doesn't evict page if covered by a second lock (2765)
5213 - disable VM readahead to avoid reading outside lock extents (2805)
5215 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5216 - updated kernel patches for hp-2.4.20 kernel (2681)
5218 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5221 - fix truncated write corruption (2366)
5222 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5224 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5225 - ensure len(uuid) < 37 in lmc (1171)
5226 - fix ia64 OOPS in llog_test (2255)
5227 - zero end of page at obdfilter for partial page writes (2648)
5228 - don't leave stale dentries around after renames (bug 2428)
5229 - fix timeouts when evicting a client with a single lock held (2642)
5230 - set deadline for the initial HELLO message to drain (2634)
5231 - print out dotted-quad IP addresses in the socknal (2302)
5233 - additional debugging for MDS client eviction problem (2443)
5234 - fix mkfsoptions support for osts (2603, 2604)
5236 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5239 - add series for the vanilla 2.6.0 kernel
5240 - add series for the vanilla 2.4.24 kernel
5241 - add series for a cray x86/64 UL kernel drop
5242 - fix xattr patches for the vanilla 2.4.19 series
5244 - generate true UUIDs in lmc (1171)
5245 - have portals stack dumping break in UML (2466)
5246 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5247 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5248 - wake callback waiting threads on client eviction (2460)
5249 - Add --ptldebug and --subsystem to lmc (1719)
5250 - update assertion to allow safe interrupt allocation
5251 - set rq_no_resend for cancel requests (2432)
5252 - recalculate ptlrpcd timeout after resend (2494)
5253 - call vfs_rmdir when removing pending directories (2368)
5254 - fix renaming a file to itself (2429)
5255 - lmc creates a default one-stripe lov (2454)
5256 - expand procfs space to handle large clusters (2326)
5257 - increase UML stack to avoid overflow
5258 - update lconf's list of debug and subsystem masks
5259 - fix lfs find --obd (2510)
5260 - /proc tunable for disabling filter read caching (2591)
5261 - stop rpm packages from altering slapd.conf (2301)
5262 - disable nagle in the socknal under 0conf (2578)
5263 - choose mds inode size based on stripe count (2572)
5264 - fix kernel-source rpm problems (2516)
5266 - add --disable-doc to avoid pdf generation (2421)
5267 - update documentation, tests, type-os, comments
5268 - avoid format warnings on ia64
5269 - remove the TOE NAL
5270 - tiny code cleanups by removing unused fields
5272 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5275 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5276 - remove the most blatant lies from BUILDING file (2371)
5277 - change default debug level to reasonable production setting
5278 - reduce client side cache size to reduce cache flush time
5279 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5280 - make TCP zerocopy and pinger support enabled by default (2476)
5281 - sync writes completed after process exits caused crashes (2319)
5282 - maintain correct mount count on the MDS (2356)
5283 - backout 1557, because 2316 wasn't really fixed
5284 - better file I/O statistics gathering in /proc
5285 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5286 - another kernel patch to fix zero-copy TCP function export
5287 - don't take duplicate lock when processing re-sent getattr (2420)
5288 - lctl uses obd_self_export instead of creating new conn (2353)
5289 - MDS/OST recovery case which requires object creation asserted (2425)
5290 - move lfs from /usr/sbin to /usr/bin in packages
5291 - fix race between mds_client_add and mds_client_free (2417)
5292 - use kmalloc instead of slabs in portals (2430)
5293 - don't create duplicate records when a failover MDS is present (2442)
5294 - remove unnecessary mount age check (2332)
5295 - don't remove directory inodes from locks prematurely (2451)
5296 - don't break if MDS service name is the same as hostname (2103)
5297 - fix races in client write RPC generation when cache full (2482)
5299 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5302 - remove now-unused request->rq_obd (278)
5303 - if an allocation fails, print out how much memory we've used (1933)
5304 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5305 - add the "configurable stack size" patch to most series files (1256)
5306 - ability to write large log records, for 100+ OST configs (2306)
5307 - fix NULL deref when filter_prep fails (2314)
5308 - fix operator precedence error in filter_sync
5309 - dynamic allocation of socknal TX descriptors (2315)
5310 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5311 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5312 - ensure that CWARN messages in Portals always get to the syslog
5313 - __init/__exit are not for prototype decls (ldlm_init/exit)
5314 - x86-64 compile warning fixes
5315 - fix gateway LMC keyword conflict (2318)
5316 - fix MDS lock inversions in getattr/reint paths (1844)
5317 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5318 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5319 - DLM race condition prevented some lock evictions (2328)
5320 - ENOMEM detection and retry on socknal sends (2230)
5321 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5322 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5323 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5324 - better error messages when a client is rejected during recovery (1505)
5325 - avoid cancelling locks which were never granted, after failure (2330)
5326 - fix i_sem/journal inversion in mds_client_add (2333)
5327 - fix truncate/getattr lock cycle deadlock (2334)
5328 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5329 - fix two rare exit paths which could leak an l_lock() ref (2321)
5330 - fencepost error in MDS/OST orphan recovery (2226)
5331 - make log record alignment 8 bytes (1988)
5332 - lstripe now fails when requested offset > ost_count (2237)
5333 - ensure that all kernel series have a complete list.h (1607)
5334 - fix crashes in special-file operations (2316)
5335 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5337 - allow configurable automake binary, for testing new versions
5338 - small update to the lfs documentation
5340 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5342 * fix negative export reference count in fsfilt_sync (2312)
5344 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5345 * release candidate 0.9.1
5347 - orphans are moved into the PENDING directory for possible recovery
5348 - replayed opens now open by fid for orphan/rename safety (1042)
5349 - last close of an orphan inode generates a transno (683)
5350 - chdir() and mount() now pin the directory entry (1020)
5351 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5352 - discard very old requests without processing them (1502)
5353 - remove some common, well-understood CERRORs (1505)
5354 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5355 - clear "grant" flags in OST replies until OST grant code lands (1644)
5356 - fix read performance by not clobbering i_blksize on client (1598)
5357 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5358 - make LRU size a /proc tunable, clears locks when reduced (707)
5359 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5360 - server locks take references on exports now (1558)
5361 - build fixes for 2.4.20-rh trees (1663)
5362 - return an error from lov_create if all OSCs are inactive (1751)
5363 - fix import levels when a reconnect happens without a timeout (1597)
5364 - exit early from mds_open if we get a lookup error (1749)
5365 - partial page read at EOF wouldn't wait for disk before sending (1642)
5366 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5367 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5368 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5369 - write/truncate lock inversion (1639)
5370 - Don't auto-load obdclass, portals modules during cleanup (1495)
5371 - fix timestamps from jumping to "now" (1763)
5372 - extra journal assertions (1648)
5373 - add an extra multiunlink test (1771)
5374 - fix read_record/write_record API (1776)
5375 - fix leak of offset_extent, possible incorrect i_size later (1772)
5376 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5377 - Debugging for the kqswnal_get_idle_tx problems (1820)
5378 - Allow recovery to be attempted multiple times (1536)
5379 - Write out MDS last_rcvd file after it is first created (1600)
5380 - Fix tx_descriptor leak in failed transmit situations (1827)
5381 - ext3 journaling fixes for assertion failure after IO error (1871)
5382 - class_export_put() on freed export after completion AST error (1896)
5383 - Fix revalidate looping in VFS (1322)
5384 - Don't access a freed export during MDS_REINT timeout (1521)
5385 - Add open-unlink recovery support on the MDS (1673,1764)
5386 - Return an error if no MDS data was read from last_rcvd (1946)
5387 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5388 - Don't setup a disk device that doesn't match exported UUID (317)
5389 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5390 - avoid committing NULL handle in force close
5391 - local.sh is now a one-stripe LOV configuration
5392 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5393 - fix ext3 htree duplicate directory entry corruption (1516)
5394 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5395 - update to new LOV EA format (2097)
5396 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5398 2003-06-15 Phil Schwan <phil@clusterfs.com>
5401 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5402 - per-import/export recovery handling (958, 931, 959)
5403 - multiple last-rcvd slots, for serving multiple FSes (949)
5404 - connections are again shared between multiple imp/exports (963, 964)
5405 - "umount -f" would hang if any requests needed to be sent (393, 978)
5406 - avoid pinning large req buffer by copying for queued messages (989)
5407 - add "uuid" to "lctl device" command to help upcalls (991)
5408 - "open" RPCs with transnos would confuse recovery counters (1037)
5409 - do proper endian conversion of all wire messages (288, 340, 891)
5410 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5411 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5412 - fix ptlrpc_connection leak in target_handle_connect (1174)
5413 - fix import refcounting bug in OST and MDS cleanup (1134)
5414 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5415 - fix very unlikely obd_types race condition (501)
5416 - remove osc_open hack for echo_client (1187)
5417 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5418 - a failure in read_inode2 leads to deadlock (1139)
5419 - cancel ack-locks as soon as transaction is committed (1072)
5420 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5421 - make sure to commitrw after any preprw to avoid deadlock (1162)
5422 - failing to execute a file in a lustre FS would lock inode (1203)
5423 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5424 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5425 - obd_uuid2tgt would walk off the end of the list (1255)
5426 - on IA64 the osc would give portals incorrect bulk size (1258)
5427 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5428 - fix lock inversion caused by new llite matching code (1282)
5429 - limit the number of dirty pages on a client to 10MB (1286)
5430 - timed out locks were not being corrected cancelled (1289)
5431 - fix O_DIRECT above 4GB on IA-32 (1292)
5432 * major user-visible changes
5433 - fail out/fail over policy now controlled by the upcall (993)
5435 - add OBD_PING to check server availability and failure (954)
5436 - lustre messages are now sent in sending host order (288, 340, 891)
5437 - add eadatalen to MDS getattr reply (340)
5438 - OST read replies may contain second buffer, with per-page status (593)
5440 2003-03-11 Phil Schwan <phil@clusterfs.com>
5443 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5444 - failed open()s could cause deadlock; fixed (867, 869)
5445 - stop cancelling OST locks when files are closed (481)
5446 - overlapping XID spaces caused network corruption (851, 853)
5447 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5448 - setattr_raw added, to do single-RPC, server-side setattrs
5449 - lmc/lconf syntax change for OST UUIDs
5450 - fix crashy race condition between ptlrpc_free_req and osc_close
5451 - don't use request in mdc_enqueue if we hit a timeout (889)
5452 - don't set the inode i_size for regular files from the MDS (896)
5453 - handle out of order completion AST (842)
5454 - don't LBUG if a lock request times out after receiving AST (913)
5455 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5456 - if a bad lock AST arrives, send an error instead of dropping entirely
5457 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5458 - fix leak in bulk IO when only partially completed (899, 900, 926)
5459 - fix O_DIRECT for ia64 (55)
5460 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5461 - C-z after timeout could hang a process forever; fixed (977)
5463 - client-side I/O cache (678, 924, 929, 941, 970)
5465 - READPAGE and SETATTRs which don't take server-side locks get
5468 2003-02-11 Phil Schwan <phil@clusterfs.com>
5471 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5472 - Fix fsfilt_extN_readpage() to read a full page of directory
5473 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5474 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5475 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5476 - Most intent operations were removed, in favour of a new RPC mode
5477 that does a single RPC to the server and bypasses most of the VFS
5478 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5479 - Aggressively cancel local locks on DLM servers
5480 - mds_reint_unlink sends EA to the client if it's the last nlink.
5481 client uses that EA to unlink OST objects.
5482 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5483 - recursive symlinks were fixed (439)
5484 - fixed NULL deref in DEBUG_REQ
5485 - filter_update_lastobjid no longer calls sync, which annoyed extN
5486 - fixed multi-client small-writes to a single file problem (445)
5487 - fixed mtime updates during file writes (607)
5488 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5489 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5490 - fixed Linux OST size reporting problem (444, 656)
5491 - OST now updates object mtime with writes or setattr (607, 619)
5492 - client verifies file size before zeroing page past EOF (445)
5493 - OST now writes last allocated objid to disk with allocation (108)
5494 - LOV on echo now works (409)
5496 - mds_reint_unlink sends a new buffer, with the EA included. this
5497 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5498 set if a regular file was being unlinked, and it was the last link
5499 - use PtlGet from the target for bulk writes (315)
5500 - OST now updates object mtime with writes or setattr (607, 619)
5501 - LDLM now has a grant-time callback to revalidate locked items, if
5503 - Many MDS operations were reorganized to combat race conditions
5505 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5506 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5508 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5511 - Fully reactivate OST imports after reconnection (512, others)
5512 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5513 - More graceful error handling for truncating on dead OST (515)
5514 - Don't error out unless we're actually accessing dead stripes (474)
5515 - Fix garbage sizes when stripes are missing (410)
5516 - LRU counters were broken, causing constant lock purge (433, 432)
5517 - garbage on read from stripes with failed OSTs (441)
5518 - mark OSCs as active before reconnecting during recovery (438)
5519 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5520 - lfind did not preserve OST order in output (443)
5521 - symlinks cause hung clients, incorrect data (439)
5522 - stop dereferencing request after dropping refcount (457)
5523 - don't LASSERT(spin_is_locked) on non-SMP (455)
5524 - fixes for many rename() bugs
5525 - fstat didn't correctly synchronize attributes (399)
5526 - server must handle lock cancellation during blocking AST prep (487)
5527 - bulk descriptors were free()d too soon (511)
5528 - fix paths in lconf, which would load incorrect modules (451, 507)
5529 - fix confusing lconf 'host not found' error message (386)
5530 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5531 - fix race condition in mdc_blocking_ast() for inode access (526)
5532 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5533 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5534 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5535 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5536 - fix lstripe in several ways (kernel side) (527)
5537 - fix request leak in ldlm_cli_enqueue (262)
5538 - incorrect OSC was marked inactive after OST failure
5539 - call mds_fs_cleanup before unmounting filesystem (524)
5540 - fix races between taking ns_lock and ldlm_lock_change_resource
5541 - fix races updating LOV export open file list
5542 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5543 - fix recovery NULL deref in ldlm_cli_cancel_unused
5544 - fix some DLM races by using new hash table for lock handles (419)
5545 - permit the client to specify desired inodes, at replay
5546 - duplicate requests when we queue them for replay reintegration
5547 - fix last_rcvd offset calculation
5548 - sync after each recovered transaction, so we always make progress
5549 - never, not always, ERESTART requests without transnos
5550 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5552 - skip replay if the MDS says that the client is already connected
5553 - don't check for a recovery-enabled export to match lctl's UUID
5554 - don't INC_USE_COUNT for phantom exports
5555 - don't crash when cleaning up phantom exports (567)
5556 - don't double-finish or set replay data for errored mdc_open requests
5557 - abort requests when they time out, so we don't get old replies
5558 - send/receive replies for AST messages again
5559 - if the client says that it doesn't have the lock, cancel it on the
5561 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5562 mark it as destroyed, it will all work out when decref is called
5563 - fix module use counts (22, 581)
5565 - ASTs now expect a reply (server cancels lock on error reply)
5567 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5570 - fix many simultaneous client startup (392)
5571 - fix dentry->d_it clobbering
5572 - credentials weren't being shipped for readdir/getattr operations
5573 - remove invalid assertions triggered during some concurrent MD
5575 - proper Lustre versions added (336, 389)
5576 - fix memory leak for create error case (398)
5577 - fix LOV locking bug that would get cli/srv out of sync
5578 - fix echo client over LOV (409)
5579 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5580 - fix double-O_EXCL intent crash (424)
5581 - avoid sending multiple lock CANCELs (352)
5583 - MDS can do multi-client recovery (modulo bugs in new code)
5585 - many updates, edits, cleanups
5587 2002-11-18 Phil Schwan <phil@clusterfs.com>
5590 - fix null d_it dereference (346)
5591 - fix full OST/dbench hang (333)
5592 - fix permission problem with file removal (286)
5593 - fix removal of OSCs from LOV when they fail
5594 - fix NULL deref during bulk timeout (214)
5595 - fix problems related to multiple filesystems on one MDS (241)
5596 - fixed serious subtle metadata locking bugs
5597 - free locks on clients when inodes are removed due to memory
5599 - fix inode pointer in lock data (285)
5600 - partial support for multiple MDS on a single host (241)
5601 - data locks weren't cancelled at clear_inode time (290, 311)
5602 - intent locks could lead to unbounded lock growth (205)
5603 - added a maximum lock count, an LRU list, and a flusher
5604 - fix multiple rename (365)
5605 - properly abstracted the echo client
5606 - OSC locked 1 byte too many; fixed
5607 - rewrote brw callback code:
5608 - fixed recovery bugs related to LOVs (306)
5609 - fixed too-many-pages-in-one-write crash (191)
5610 - fixed (again) crash in sync_io_timeout (214)
5611 - probably fixed callback-related race (385)
5613 - Add capability to MDS protocol
5614 - LDLM cancellations and callbacks on different portals
5616 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5619 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5620 - defer open object destruction to close time (601981, 138)
5621 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5622 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5623 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5624 - don't call obd_setattr in ll_file_release for destroyed objects
5625 * protocol change to lustre_msg: move |version| and add |flags|
5626 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5627 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5628 * added replay of create, unlink, link and rename operations during
5629 MDS failover; recovery should be much more robust now
5630 * remove failed OSCs from LOVs (only lov_create uses this so far)
5631 * the lustre-HOWTO was brought (more) up to date (582544)
5633 2002-10-23 Phil Schwan <phil@clusterfs.com>
5636 - in-use dentries weren't being reused properly (617851)
5637 - prevent multiple LDLM setup (599178)
5638 - fix LOV size calculations for truncate (617853)
5639 - fix client handling of MDS intent errors (POSIX)
5640 - fix permission bug in lovstripe.c test (624321)
5641 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5642 - truncate past end of file could corrupt data
5643 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5644 - a race in recovery could return ETIMEDOUT to apps (623947)
5645 - building outside the source directory was fixed
5646 * the lustre-HOWTO was brought (more) up to date (582544)
5647 * major progress was made on recovery functionality
5649 2002-10-10 Phil Schwan <phil@clusterfs.com>
5652 - recovery deadlock fix
5653 - rm -rf causes LBUG fix (617817)
5654 - file open by multiple tasks fix (618962)
5655 - directory permissions bugs (602707 and 620007)
5656 - journal_stop fixed with locking (611313)
5657 - O_APPEND failures resolved (618273, perhaps 614459)
5658 - lconf PATH fix (619770)
5659 - IA64 build fix (621450)
5660 - RPC buffer sizes scale with amount of memory
5662 2002-10-01 Phil Schwan <phil@clusterfs.com>
5665 - locks would be cancelled without throwing away data pages,
5666 resulting in inconsistent data (605627)
5667 - inode attributes were not always being refreshed (605627, 612449)
5668 - lconf now continues to cleanup after lctl reports an error
5669 - MDS now enforces user permissions (602707)
5670 - lprocfs cleanup fixed, but not yet enabled (614157)
5671 - fixed infinite server hang, should a client not respond to an AST
5672 - avoid going into recovery if user calls readlink() with a buffer
5673 that's too small (613941)
5674 - AST RPCs no longer require replies (614867) -- this may be changed
5675 - don't crash server if client sends an IOV that's too big (611336)
5676 - fixed lock conversion deadlock (611892)
5677 - fixed the following of symlinks (614622)
5678 * recovery: the server can remove locks from a client that dies, other
5679 clients can make progress
5680 * more extN patch fixes
5681 * compile-time configurable ptlrpc buffer allocations
5683 - collaborative read cache document
5684 - Lustre Lite Performance CDR document-in-progress
5686 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5689 - fix typo in patch-2.4.18
5691 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5694 - clear ptlrpc request each time in handle_incoming_request()
5695 - unlink of files now destroys the object on the OST
5697 2002-09-19 Peter Braam <braam@clusterfs.com>
5699 * add hard link support
5700 * change obdfile creation method
5701 * kernel patch changed
5703 2002-09-19 Peter Braam <braam@clusterfs.com>
5706 - stack overflow bug in extN fixed
5708 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5710 * documentation updates
5711 - add man pages for config tools
5712 - update tests/README to describe testing with new config tools
5713 - finish metadata API descriptions
5714 * bug fixes and cleanups
5715 - statfs workaround for 16TB limit
5716 - LOV stripe allocation improved, can stripe on subset of OSTs
5717 - LOV file size/IO offset was wrong for files > 4GB in size
5718 - object EA data was being dropped, caused files to be unreadable
5719 - memory overflow with non-LOV OST caused memory corruption
5720 - fixed regression tests to work with new config tools, obdfilter
5721 - fixed bug when directory size became larger than 1 block
5722 - fixed bug (for single client case) when PWD was deleted
5723 - invalidate local directory pages when doing intent-based ops
5724 - avoid LDLM oops when lock callback contained bad data
5726 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5728 * documentation updates
5729 * bug fixes and cleanups
5730 - configuration tools
5733 - 64-bit compile warnings
5734 - 64-bit internal statfs data
5736 * test_brw on persistent OST devices
5738 * lprocfs (disabled)
5740 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5742 * documentation updates
5743 * bug fixes and cleanups
5744 * configuration tools
5746 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5749 * many small fixes to 0_5_4
5750 * io/network handling
5751 * thinkos in MDS operations
5753 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5756 * crucial basic fixes to 0.5.3
5757 * IOR, Iozone work over Elan
5760 2002-08-07 Phil Schwan <phil@clusterfs.com>
5761 * version 0_5_3, our first alpha
5762 * we use the new Portals iovs
5763 * documentation updates
5764 * bug fixes and cleanups
5765 * small changes in the DLM wire protocol
5767 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5768 * version 0_5_1 with some initial stability,
5769 * locking on MD and file I/O.
5770 * documentation updates
5771 * several bug fixes since 0.5.0
5772 * small changes in wire protocol
5774 2002-07-18 Phil Schwan <phil@clusterfs.com>
5776 * delivered as Lustre Light Alpha
5777 * fixed a crash after handling invalid MDS requests
5778 * fixed directory pages for architectures with non-4k pages sizes
5780 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5781 * release version v0_4_4
5782 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5784 * Fixes a number of interruption problems with OST operations.
5785 * Update documentation for portals header changes
5786 * Move all wire protocol structs/defines to lustre_idl.h
5787 * Fixes symlink length bug.
5788 * Add tcpdump to repository.
5790 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5791 * release version v0_4_3
5792 * Fixes statfs for inodes on extN.
5793 * Fixes bug in runtests which would delete /etc/hosts.
5794 * Use 64-bit object IDs wherever possible (not into VFS though)
5795 Remove ost_get_info, which is unused by lustre, and out of date.
5797 2002-07-03 Peter Braam <braam@clusterfs.com>
5798 * release version v0_4_2 Fixes a lookup error (type not passed)
5799 * move forward to head of Portals
5800 * move forward to latest Lustre kernel
5802 2002-06-25 Peter Braam <braam@clusterfs.com>
5803 * release version v0_4_1. Hopefully stable on single node use.