1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.33 (SLES 10),
5 2.6.18-128.1.1.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.40.11-sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
14 * File join has been disabled in this release, refer to Bugzilla 16929.
16 Severity : enhancement
18 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
23 Descriptoin: OST grant too much space to client even there are not enough space.
24 Details : Client will shrink its grant cache to OST if there are no write
25 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
26 this grant cache if there are already not enough avaible space
27 (left_space < total_clients * 32M).
30 Frequency : start MDS on uncleanly shutdowned MDS device
32 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
33 Details : stay in waiting mds<>ost recovery finished produce random bugs
34 due race between two ll_sync thread for one lov target. send
35 ACTIVATE event only if connect realy finished and import have
39 Frequency : rare, connect and disconnect target at same time
41 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
42 Details : don't call obd_disconnect under lov_lock. this long time
43 operation and can block ptlrpcd which answer to connect request.
48 Descriptoin: don't lose wakeup for imp_recovery_waitq
49 Details : recover_import_no_retry or invalidate_import and import_close can
50 both sleep on imp_recovery_waitq, but we was send only one wakeup
54 Frequency : always with long access acl
56 Descriptoin: mds can't pack reply with long acl.
57 Details : mds don't control size of acl but they limited by reint/getattr
60 Severity : enhancement
62 Description: Update to SLES10 kernel-2.6.16.60-0.33.
64 Severity : enhancement
66 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
69 Frequency : start MDS on uncleanly shutdowned MDS device
71 Descriptoin: aborting recovery hang on MDS
72 Details : don't throttle destroy RPCs for the MDT.
75 Frequency : on remount
77 Description: external journal device not working after the remount
78 Details : clear dev_rdonly flag for external journal devices in
84 Description: shutdown vs evict race
85 Details : client_disconnect_export vs connect request race.
86 if client will evicted at this time - we start invalidate
87 thread without referece to import and import can be freed
91 Frequency : rare, need acl's on inode.
93 Description: client can't handle ost additional correctly
94 Details : if ost was added after client connected to mds client can have
95 hit lnet_try_match_md ... to big messages to wide striped files.
96 in this case need teach client to handle config events about add
97 lov target and update client max ea size at that event.
99 Severity : enhancement
101 Description: Changelogs
102 Details : Changelogs are a lightweight mechanism to track filesystem
103 metadata and namespace changes. The changelog is recorded
104 permanently on the MDTs, and is periodically "consumed" / purged
105 when records are no longer needed.
107 Severity : enhancement
109 Description: compact fld format with extents
110 Details : Store range of seq rather than every seq in FLD. Seq
111 controller update FLD rather than clients. In Case of CMD, mdt0
112 has FLD, all other metadata server act as non persistent proxy
113 for FLD queries and cache fld entries in fld cache.
118 Description: don't skip ost target if they assigned to file
119 Details : Drop slow OSCs if we can, but not for requested start idx.
120 This means "if OSC is slow and it is not the requested
121 start OST, then it can be skipped, otherwise skip it only
122 if it is inactive/recovering/out-of-space.
126 Description: more cleanup in mds_lov
127 Details : not send LOV EA under replay, we can't know about they size at this
128 time. Don't allow client connect to mds before any ost connected,
129 for avoid problems with LOV EA size and returning EIO to client.
131 Severity : enhancement
133 Description: Interoperability at server side (Disk interoperability)
135 Severity : enhancement
137 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
139 Severity : enhancement
141 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
143 Severity : enhancement
145 Description: New client IO stack (CLIO).
147 Severity : enhancement
149 Description: Commit on sharing. Eliminate inter-client dependencies between
150 uncommitted transactions by doing transaction commits.
151 Thereby clients may recovery independently.
154 Frequency : Create a symlink file with a very long name
156 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
157 Details : If there is no extra space in the request for early cancels,
158 ldlm_req_handles_avail() returns 0 instead of a negative value.
160 Severity : enhancement
162 Description: Add /proc entry for import status
163 Details : The mdc, osc, and mgc import directories now have
164 an import directory that contains useful import data for debugging
167 Severity : enhancement
169 Description: Re-disable certain /proc logging
170 Details : Enable and disable client's offset_stats, extents_stats and
171 extents_stats_per_process stats logging on the fly.
174 Frequency : Only on FC kernels 2.6.22+
176 Description: oops in statahead
177 Details : Do not drop reference count for the dentry from VFS when lookup,
178 VFS will do that by itself.
180 Severity : enhancement
182 Description: Generic /proc file permissions
183 Details : Set /Proc file permissions in a more generic way to enable non-
184 root users operate on some /proc files.
188 Description: Hitting mdc_commit_close() ASSERTION
189 Details : Properly handle request reference release in
190 ll_release_openhandle().
194 Description: quota recovery deadlock during mds failover
195 Details : This patch includes att18982, att18236, att18237 in bz14840.
197 1. fix osts hang when mds does failover with quotaon
198 2. prevent watchdog storm when osts threads wait for the
203 Frequency : only patchless client
204 Description: add workaround for race between add/remove dentry from hash
206 Severity : enhancement
208 Description: Allow OST glimpses to return PW locks
212 Description: LBUG when llog conf file is full
213 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
217 Description: Prevent import from entering FULL state when server in recovery
221 Description: service mount cannot take device name with ":"
222 Details : Only when device name contains ":/" will mount treat it as
228 Description: replace ptlrpcd with the statahead thread to interpret the async
229 statahead RPC callback
233 Frequency : on recovery
234 Description: I/O failures after umount during fail back
235 Details : if client reconnected to restarted server we need join to recovery
236 instead of find server handler is changed and process self
237 eviction with cancel all locks.
239 Severity : enhancement
241 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
243 Severity : enhancement
245 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
247 Severity : enhancement
249 Description: Upcall on Lustre log has been dumped
250 Details : Allow for a user mode script to be called once a Lustre log has
251 been dumped. It passes the filename of the dumped log to the
252 script, the location of the script can be specified via
253 /proc/sys/lnet/debug_log_upcall.
258 Description: avoid idr_remove called for id which is not allocated.
259 Details : Move assigment s_dev for clustered nfs to end of initialization,
260 to avoid problem with error handling.
265 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
266 Details : When connection is reused this not moved from CONN_UNUSED_HASH
267 into CONN_USED_HASH and this prodice warning when put connection
268 again in unused hash.
271 Severity : enhancement
273 Description: File striping can now be set to use an arbitrary pool of OSTs.
275 Severity : enhancement
277 Description: Export bytes_read/bytes_write count on OSC/OST.
281 Description: Early reply size mismatch, MGC loses connection
282 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
283 the connect flags are properly negotiated.
288 Description: quotas are not honored with O_DIRECT
289 Details : all writes with the flag O_DIRECT will use grants which leads to
290 this problem. Now using OBD_BRW_SYNC to guard this.
294 Description: add quota statistics
295 Details : 1. sort out quota proc entries and proc code.
296 2. add quota statistics
298 Severity : enhancement
300 Description: enable quota support for HEAD.
304 Description: Properly propagate oinfo flags from lov to osc for statfs
305 Details : restore missing copy oi_flags to lov requests.
307 Severity : enhancement
309 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
310 : l_getgroups(8), lst(8), routerstat(8)
311 Details : included man pages for llobdstat(8), llstat(8),
312 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
314 Severity : enhancement
316 Description: configure's --enable-quota should check the
317 : kernel .config for CONFIG_QUOTA
318 Details : configure is terminated if --enable-quota is passed but
319 : no quota support is in kernel
323 Description: Remove portals compatibility
324 Details : Remove portals compatibility, not interoperable with releases
329 Description: Resolve device initialization race
330 Details : Prevent proc handler from accessing devices added to the
331 obd_devs array but yet be intialized.
333 Severity : enhancement
335 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
337 Severity : enhancement
339 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
344 Description: Using wrong pointer in osc_brw_prep_request
345 Details : Access to array[-1] can produce panic if kernel compiled with
346 CONFIG_PAGE_ALLOC enabled
350 Description: Client runs out of low memory
351 Details : Consider only lowmem when counting initial number of llap pages
355 Description: *optional* service tags registration
356 Details : if the "service tags" package is installed on a Lustre node
357 When the filesystem is mounted, a local-node service tag will
358 be created. See http://inventory.sun.com/ for more information
359 about the Service Tags asset management system.
363 Description: Kernel BUG tries to release flock
364 Details : Lustre does not destroy flock lock before last reference goes
365 away. So always drop flock locks when client is evicted and
366 perform unlock regardless of successfulness of speaking to MDS.
370 Description: add refcount protection for osc callbacks, avoid panic on shutdown
374 Description: sanity test 65a fails if stripecount of -1 is set
375 Details : handle -1 striping on filesystem in ll_dirstripe_verify
380 Description: ASSERTION(CheckWriteback(page,cmd)) failed
381 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
384 Severity : enhancement
386 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
390 Description: do not process already freed flock
391 Details : flock can possibly be freed by another thread before it reaches
392 to ldlm_flock_completion_ast.
396 Description: LBUG during stress test
397 Details : Need properly lock accesses the flock deadlock detection list.
401 Description: oops in page fault handler
402 Details : kernel page fault handler can return two special 'pages' in error
403 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
407 Description: timeout with invalidate import.
408 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
409 should be handled by ptlrpcd. This produce long age waiting and
410 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
412 Severity : enhancement
414 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
418 Description: enable MGS and MDT services start separately
419 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
420 a co-located MGS without starting the MGS, which is a complement
421 to 'nosvc' mount option.
425 Frequency : after recovery
426 Description: precreate to many object's after del orphan.
427 Details : del orphan st in oscc last_id == next_id and this triger growing
428 count of precreated objects. Set flag LOW to skip increase count
429 of precreated objects.
433 Frequency : rare, on clear nid stats
434 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
435 Details : when clean nid stats sometimes try destroy live entry,
436 and this produce panic in free.
440 Description: Stack overflow during MDS log replay
441 ease stack pressure by using a thread dealing llog_process.
445 Description: wait until IO finished before start new when do lock cancel.
446 Details : VM protocol want old IO finished before start new, in this case
447 need wait until PG_writeback is cleared until check dirty flag and
448 call writepages in lock cancel callback.
450 Severity : enhancement
452 Description: using special macro for print time and cleanup in includes.
456 Description: mds_mfd_close() ASSERTION(rc == 0)
457 Details : In mds_mfd_close(), we need protect inode's writecount change
458 within its orphan write semaphore to prevent possible races.
462 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
466 Frequency : rare, on shutdown ost
467 Description: don't hit live lock with umount ost.
468 Details : shrink_dcache_parent can be in long loop with destroy dentries,
469 use shrink_dcache_sb instead.
473 Description: don't panic with use echo client
474 Details : echo client pass NULL as client nid pointer and this produce null
479 Description: fix build on ppc32
480 Details : compile code with -m64 flag produce wrong object file for ppc32.
484 Description: add message levels for liblustreapi
488 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
489 Details : Move the 'good_osts' check before the 'total_bavail' check. This
490 will result in an -EAGAIN and in the exit call path we call
491 alloc_rr() which will with increasing aggressiveness attempt to
492 aquire precreated objects on the minimum number of required OSCs.
496 Description: Use old size assignment to avoid deadlock
497 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
498 the scheduling while holding a spinlock. We do not need locking
499 for size in ll_update_inode() because size is only updated from
500 the MDS for directories or files without objects, so there is no
501 other place to do the update, and concurrent access to such inodes
502 are protected by the inode lock.
506 Description: resolve "_IOWR redefined" build error on SLES10
510 Description: dump the memory debugging after all modules are unloaded to
511 suppress false negative in conf_sanity test 39
513 Severity : enhancement
515 Description: build kernel-ib packages for OFED 1.3 in our release cycle
520 Description: fix SLES kernel versioning
521 Details : the kernel version for our SLES 10 kernel did not include a "-"
522 before the "smp" at the end. while this was not a problem in
523 general, it did mean that software trying to use the kernel
524 version to try to detect a vendor specific kernel would fail.
525 this was most evident by the OFED build scripts.
529 Description: Don't update lov_desc members until making sure they are valid
530 Details : When updating lov_desc members via proc fs, need fix their
531 validities before doing the real update.
535 Description: don't put request into delay list while invalidate in flight.
536 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
537 import in flight. this produce timeout for invalidate and sometimes
538 can cause stale data.
542 Frequency : on ppc only
543 Description: not convert ost objects for directory because it's not exist.
544 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
548 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
550 Severity : enhancement
552 Description: Update to support 2.6.22.14 vanilla kernel.
556 Frequency : rare, on recovery
557 Description: read procfs can produce deadlock in some situation
558 Details : Holding lprocfs lock which send rpc can produce block for destroy
559 obd objects and this also block reconnect with -EALREADY. This
560 isn't fix all lprocfs bugs - but make it rare.
562 Severity : enhancement
564 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
567 Frequency : frequent on X2 node
569 Description: mdc_set_open_replay_data LBUG
570 Details : Set replay data for requests that are eligible for replay.
574 Description: lustre_mgs: operation 101 on unconnected MGS
575 Details : When MGC is disconnected from MGS long enough, MGS will evict the
576 MGC, and late on MGC cannot successfully connect to MGS and a lot
577 of the error messages complaining that MGS is not connected.
581 Frequency : on network error
582 Description: panic with double free request if network error
583 Details : mdc_finish_enqueue is finish request if any network error ocuring,
584 but it's true only for synchronus enqueue, for async enqueue
585 (via ptlrpcd) this incorrect and ptlrpcd want finish request
588 Severity : enhancement
590 Description: client-side metadata stat-ahead during readdir(directory readahead)
591 Details : perform client-side metadata stat-ahead when the client detects
592 readdir and sequential stat of dir entries therein
595 Frequency : on start mds
597 Description: Implement get_info(last_id) in obdfilter.
600 Frequency : occasional
602 Description: Correctly check stale fid, not start epoch if ost not support SOM
603 Details : open with flag O_CREATE need set old fid in op_fid3 because
604 op_fid2 was overwritten with new generated fid, but mds can answer
605 with one of these two fids and both is not stale. Setattr
606 incorrectly started an epoch and assume will be called
607 done_writing, but without SOM done_writing ever being called.
610 Frequency : rare, depends on device drivers and load
612 Description: MDS or OSS nodes crash due to stack overflow
613 Details : Code changes in 1.8.0 increased the stack usage of some functions.
614 In some cases, in conjunction with device drivers that use a lot
615 of stack the MDS (or possibly OSS) service threads could overflow
616 the stack. One change which was identified to consume additional
617 stack has been reworked to avoid the extra stack usage.
620 Frequency : occasional
622 Description: Do not fail import if osc_interpret_create gets -EAGAIN
623 Details : If osc_interpret_create got -EAGAIN it immediately exits and
624 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
625 oscc_has_objects and see OSC has no objests and call
626 oscc_internal_create to resend create request.
628 Severity : enhancement
630 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
632 Severity : enhancement
634 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
637 Frequency : very rare
639 Description: Fix replay if there is an un-replied request and open
640 Details : In some cases, older replay request will revert the
641 mcd->mcd_last_xid on MDS which is used to record the client's
644 Severity : enhancement
646 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
648 Severity : enhancement
650 Description: Add rhel5 support to HEAD.
652 Serverity : enhancement
654 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
659 Description: Don't allow skipping OSTs if index has been specified.
660 Details : Don't allow skipping OSTs if index has been specified, make
661 locking in internal create lots better.
665 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
666 Details : access to bitfield in structure is always rounded to long
667 and this produce problem with not atomic change any bit.
671 Description: Lustre make rpms failed.
672 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
673 builds Lustre rpms from tarball.
676 Frequency : rare on shutdown ost
678 Description: If llog cancel was not send before clean_exports phase, this can
679 produce deadlock in llog code.
680 Details : If llog thread has last reference to obd and call class_import_put
681 this produce deadlock because llog_cleanup_commit_master wait when
682 last llog_commit_thread exited, but this never success because was
683 called from llog_commit_thread.
687 Description: allow userland application know is lost one of stripes.
688 Details : fill lvb_blocks with error code on ost and return it to
689 application if error flag found.
693 Description: NULL lov_tgts causing MDS oops
694 Details : more safe checks for NULL lov_tgts for avoid oops.
696 Severity : enhancement
698 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
702 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
703 Details : If an OST is down the MDS will hang indefinitely in
704 obd_statfs_rqset() waiting for the statfs data. While for
705 MDS QOS usage of statfs, it should not stuck in waiting.
707 Severity : enhancement
709 Description: remote_acl support
710 Details : Support ACL-based permission check for remote user.
711 Support setfacl/getfacl for remote user with the utils
712 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
713 the system "{s,g}etfacl" utils.
715 Severity : enhancement
717 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
719 Severity : enhancement
721 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
725 Description: make mgs_setparam() handle fsname containing dash
726 Details : fsname containing a dash does not work with lctl conf_param
728 Severity : enhancement
730 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
732 Severity : enhancement
734 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
738 Frequency : rare, at shutdown
739 Description: access already free / zero obd_namespace.
740 Details : if client_disconnect_export was called without force flag set,
741 and exist connect request in flight, this can produce access to
742 NULL pointer (or already free pointer) when connect_interpret
743 store ocd flags in obd_namespace.
747 Frequency : only at startup
748 Description: not alloc memory with spinlock held.
749 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
750 if any spinlock held.
752 Severity : enhancement
754 Description: make lustre randomly fail allocating memory
755 Details : Make lustre randomly failed allocating memory for testing purpose.
757 Severity : enhancement
759 Description: lost problems with lov objid file
760 Details : Fixes some scability and access to not inited memory problems
761 in work with lov objdid file.
766 Description: lfs find does not continue on file error
767 Details : Continue other files processing when a file/dir is absent.
771 Description: Inconsistent usage of lustre_pack_reply()
772 Details : Standardize the usage of lustre_pack_reply() such that it
773 always generate a CERROR on failure.
776 Frequency : occasional
778 Description: MDS hang or stay in waiting lock
779 Details : If client receive lock with CBPENDING flag ldlm need send lock
780 cancel as separate rpc, to avoid situation when cancel request
781 can't processed due all i/o threads stay in wait lock.
785 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
789 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
793 Description: Update to SLES9 kernel-2.6.5-7.287.3.
797 Description: Update to RHEL4 latest kernel.
799 Severity : enhancement
801 Description: Build SLES10 patchless client fails
802 Details : The configure was broken by run ./configure with
803 --with-linux-obj=.... argument for patchless client. When the
804 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
805 can't search header adequately. Use absolute path such as
806 -I($LINUX)/include instead.
810 Description: interrupt oig_wait produce painc on resend.
811 Details : brw_redo_request can be used for resend requests from ptlrpcd and
812 private set, and this produce situation when rq_ptlrpcd_data not
813 copyed to new allocated request and triggered LBUG on assert
814 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
815 ptlrpcd set if request is changed and can be safety changed to use
820 Description: LASSERT_{REQ,REP}SWAB macros are buggy
821 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
822 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
823 these macros and replace them with their difinitions instead.
827 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
828 Details : Update which_patch & target file for SLES10 latest kernel.
831 Description: Remove client patches from SLES 10 kernel.
832 Details : This causes SLES 10 clients to behave as patchless clients
833 even on a Lustre-patched (server) kernel.
835 Severity : enhancement
837 Description: self-adjustable client's lru lists
838 Details : use adaptive algorithm for managing client cached locks lru
839 lists according to current server load, other client's work
840 pattern, memory activities, etc. Both, server and client
841 side namespaces provide number of proc tunables for controlling
844 Severity : enhancement
846 Description: light-weight GSS support
847 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
848 and reduce performance overhead.
850 Severity : enhancement
852 Description: Linux keyring support
853 Details : Support using service of Linux keyring for Lustre GSS internal
854 context refresh/cache mechanism.
858 Description: Fix errors in lfs documentation
859 Details : Fixes man pages
863 Description: don't use GFP_* in generic Lustre code.
864 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
868 Description: obdclass is limited by single OBD_ALLOC(idarray)
869 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
873 Description: warnings with build patchless client with vanila 2.6.19 and up
874 Details : change old ctl_table style and replace ctl_table/ctl_table_header
875 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
879 Frequency : Only for SLES
880 Description: sanity_quota fail test_1
881 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
882 file, which makes TSTID[2] inunique.
886 Frequency : Only for SLES9
887 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
888 Details : Update target/ChangeLog/which_patch .
892 Frequency : Only for test_10 in sanity-quota.sh
893 Description: a bug in quota test code
894 Details : checking a test flag in a wrong place
898 Frequency : Only for test_13 in sanity-quota.sh
899 Description: a bug in quota test script
900 Details : shouldn't check overfull allocated quota
904 Description: jbd statistics
905 Details : Port older jbd statistics patch for sles10
907 Severity : enhancement
909 Description: Add printing inode info into message about error in writepage.
914 Description: reply_lock_interpret crash due to race with it and lock cancel.
915 Details : Do not replay locks that are being cancelled. Do not reference
916 locks by their address during replay, just by their handle.
920 Frequency : When flocks are used.
921 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
922 Details : Flock locks might destroy just granted lock if it could be merged
923 with another existing flock, this is done in completion handler,
924 so teach ldlm_cli_enquque_fini that this is a valid case for
930 Description: Oops in read and write path when failing to allocate lock.
931 Details : Check if lock allocation failed and return error back.
935 Description: lstripe command fails for valid OST index
936 Details : The stripe offset is compared to lov->desc.ld_tgt_count
937 instead of lov->desc.ld_active_tgt_count.
941 Description: sanity.sh failed test 103
942 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
947 Description: df doesn't work properly if diskfs blocksize != 4K
948 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
952 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
953 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
955 Severity : enhancement
957 Description: Adaptive timeouts
958 Details : RPC timeouts adapt to changing server load and network
959 conditions to reduce resend attempts and improve recovery time.
963 Description: llapi_file_create() does not allow some changes
964 Details : add llapi_file_open() that allows specifying the mode and
965 open flags, and also returns an open file handle.
969 Description: merge and cleanup kernel patches.
970 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
974 Description: Add journal checksum support.(Kernel part)
975 Details : The journal checksum feature adds two new flags i.e
976 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
977 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
978 indicates that the commit block contains the checksum for
979 the blocks described by the descriptor blocks. Now commit
980 record can be sent to disk without waiting for descriptor
981 blocks to be written to disk. This behavior is controlled
982 using JBD2_FEATURE_ASYNC_COMMIT flag.
986 Description: OSS needs mutliple precreate threads
987 Details : Add ability to start more than one create thread per OSS.
991 Description: Sanity.sh test_65k failed: setstripe should have succeeded
992 Details : execute lfs setstripe on client
996 Description: mds_obd_create error creating tmp object
997 Details : When the user sets quota on root, llog will be affected and can't
998 create files and write files.
1001 Frequency : Always on ia64 patchless client, and possibly others.
1003 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1004 Details : This allows the patchless client to be loaded on architectures
1005 without this export.
1009 Description: RedHat Update kernel for RHEL5
1010 Details : Add the kernel config file for RHEL5.
1014 Description: RedHat Update kernel for RHEL5
1015 Details : Modify the kernel config file more closer RHEL5.
1019 Description: Build failure against Centos5 (RHEL5)
1020 Details : Define PAGE_SIZE when it isn't present.
1024 Description: test_7 on sanity-quota.sh can't be used on separate machines
1025 Details : add facet to handle it
1029 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1030 Details : with NFS, the anon dentry's parent was set to itself in
1031 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1032 corresponding dentry other than use rec->ur_name.
1034 Severity : enhancement
1036 Description: lfs setstripe enhancement
1037 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1041 Description: enable data checksumming by default
1042 Details : enable checksum by default, allow --disable-checksum
1043 configure option and "-o nochecksum" mount option. Checksums
1044 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1048 Description: lustre support for RHEL5
1049 Details : Add support for RHEL5.
1053 Description: Client eviction due to failover config
1054 Details : after a connection loss, the lustre client should attempt to
1055 reconnect to the last active server first before trying the
1056 other potential connections.
1060 Description: when mds and osts use different quota unit(32bit and 64bit),
1061 quota will be released repeatly.
1062 Details : void sending multiple quota reqs to mds, which will keep the
1063 status between the reqs.
1067 Description: osts not allocated evenly to files
1068 Details : change the condition to increase offset_idx
1071 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1073 Description: Data corruption for OSTs that are formatted larger than 2TB
1075 Details : When generating the bio request for lustre file writes the
1076 sector number would overflow a temporary variable before being
1077 used for the IO. The data reads correctly from Lustre (which
1078 will overflow in a similar manner) but other file data or
1079 filesystem metadata may be corrupted in some cases.
1083 Description: Tune the kernel for good SCSI performance.
1084 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1085 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1090 Description: rewrite ext2-derived code in obdclass/uuid.c
1091 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1092 from scratch preserving functionality.
1096 Description: "lfs find -obd UUID" prints directories
1097 Details : "lfs find -obd UUID" will return all directory names instead
1098 of just file names. It is incorrect because the directories
1099 do not reside on the OSTs.
1103 Description: use i_size_read and i_size_write in 2.6 port
1104 Details : replace inode->i_size access with i_size_read/write()
1107 Frequency : when removing large files
1109 Description: scheduling issue during removal of large Lustre files
1110 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1111 It causes scheduling issues when removing large files (17TB in the
1115 Frequency : only with liblustre clients on XT3
1117 Description: evictions taking too long
1118 Details : allow llrd to evict clients directly on OSTs
1121 Frequency : only on ppc
1123 Description: /proc/fs/lustre/devices broken on ppc
1124 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1125 We should make sure the type of 'index' is loff_t and then cast
1126 explicitly as needed below. Do not assign an explicitly cast
1130 Frequency : only for rhel5
1132 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1133 Details : Modify the target file & which_kernel.
1135 Severity : enhancement
1137 Description: omit set fsid for export NFS
1138 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1139 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1143 Frequency : Always, for kernels after 2.6.16
1144 Description: Fix warning idr_remove called for id=.. which is not allocated.
1145 Details : Last kernels save old s_dev before kill super and not allow
1146 to restore from callback - restore it before call kill_anon_super.
1150 Description: buffer overruns could theoretically occur
1151 Details : llapi_semantic_traverse() modifies the "path" argument by
1152 appending values to the end of the origin string, and a
1153 overrun may occur. Adding buffer overrun check in liblustreapi.
1157 Description: Fix error on 'ls .' at the top of the Lustre mount.
1158 Details : Don't revalidate dentry if it is a root dentry.
1162 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1163 Details : Modify vm-tunables-rhel4.patch.
1167 Description: Kernel config for 2.6.18-vanilla.
1168 Details : Modify targets/2.6-vanilla.target.in.
1169 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1170 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1171 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1172 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1176 Description: improve handling recoverable errors
1177 Details : if request processig with error which can be recoverable on server
1178 request should be resend, otherwise page released from cache and
1183 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1184 Details : Modify target file & which_patch.
1185 A flaw was found in the IA32 system call emulation provided
1186 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1187 value could be stored in the %RAX register, which could trigger an
1188 out-of-bounds system call table access. An untrusted local user
1189 could exploit this flaw to run code in the kernel
1190 (ie a root privilege escalation). (CVE-2007-4573).
1194 Description: O_DIRECT bypasses client statistics.
1195 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1196 read_ahead_stats were updated. copied the stats section from
1197 osc_send_oap_rpc() into async_internal().
1201 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1205 Description: change order of libsysio includes
1206 Details : '#include sysio.h' should always come before '#include xtio.h'
1210 Description: handle "serious error: objid * already exists" more gracefully
1211 Details : If LAST_ID value on disk is smaller than the objects existing in
1212 the O/0/d* directories, it indicates disk corruption and causes an
1213 LBUG(). If the object is 0-length, then we should use the existing
1214 object. This will help to avoid a full fsck in most cases.
1216 Severity : enhancement
1218 Description: adapt the lustre_config script to support the upgrade case
1219 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1220 server targets to 1.6 in parallel.
1224 Description: To avoid grant space > avaible space when the disk is almost
1225 full. Without this patch you might see the error "grant XXXX >
1226 available" or some LBUG about grant, when the disk is almost
1228 Details : In filter_check_grant, for non_grant cache write, we should
1229 check the left space by if (*left > ungranted + bytes), instead
1230 of (*left > ungranted), because only we are sure the left space
1231 is enough for another "bytes", then the ungrant space should be
1232 increase. In client, we should update cl_avail_grant only there
1233 is OBD_MD_FLGRANT in the reply.
1237 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1238 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1239 described in CVE-2007-4573. This problem could allow untrusted
1240 local users to gain root access.
1243 Frequency : when using O_DIRECT and quotas
1245 Description: Incorrect file ownership on O_DIRECT output files
1246 Details : block usage reported by 'lfs quota' does not take into account
1247 files that have been written with O_DIRECT.
1252 Description: touch file failed when fs is not full
1253 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1254 otherwise we can get ENOSP while fs is not full.
1258 Description: parallel lock callbacks
1259 Details : Instead of sending blocking and completion callbacks as separated
1260 requests, adding them to a set and sending in parallel.
1263 Frequency : only for Cray XT3
1264 Bugzilla : 12829/13455
1265 Description: Changing primary group doesn't change the group lustre assigns to
1267 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1268 overridden with the primary group provided by the group upcall,
1269 whereas the supplied fsgid can be trusted if it is in the list of
1270 supplementary groups returned by the group upcall.
1272 Severity : enhancement
1274 Description: Allow masking D_WARNING, D_ERROR messages from console
1275 Details : Console messages can now be disabled via lnet.printk.
1279 Description: User code with malformed file open parameter crashes client node
1280 Details : Before packing join_file req, all the related reference should be
1281 checked carefully in case some malformed flags cause fake
1282 join_file req on client.
1286 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1287 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1288 mode is equal to the granted mode. However, it should also check
1289 that LDLM_FL_CP_REQD is not set.
1293 Description: Heavy nfs access might result in deadlocks
1294 Details : After ELC code landed, it is now improper to enqueue any mds
1295 locks under och_sem, because enqueue might want to decide to
1296 cancel open locks for same inode we are holding och_sem for.
1300 Description: Client eviction while running blogbench
1301 Details : A lot of unlink operations with concurrent I/O can lead to a
1302 deadlock causing evictions. To address the problem, the number of
1303 oustanding OST_DESTROY requests is now throttled to
1304 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1305 callbacks are processed in priority.
1309 Description: enable ACLs on MDS by default
1310 Details : ACLs must be enabled on MDS by default.
1313 Frequency : PPC/PPC64 only
1315 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1316 Details : fix duplicated definitions between asm-ppc64/types.h and
1317 lustre_types.h on PPC.
1320 Frequency : PPC/PPC64 only
1322 Description: asm-ppc/segment.h does not exist
1323 Details : fix compile issue on PPC.
1327 Description: better handle error messages in extents code
1330 Frequency : RHEL4 only
1332 Description: mkfs is very slow on IA64/RHEL4
1333 Details : A performance regression has been discovered in the MPT Fusion
1334 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1335 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1336 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1338 Severity : enhancement
1340 Description: SNMP support enhancement
1341 Details : Adding total number of sampled request for an MDS node in snmp
1344 Severity : enhancement
1346 Description: Optimize ldlm waiting list processing for PR extent locks
1347 Details : When processing waiting list for read extent lock and meeting read
1348 lock that is same or wider to it that is not contended, skip
1349 processing rest of the list and immediatelly return current
1350 status of conflictness, since we are guaranteed there are no
1351 conflicting locks in the rest of the list.
1355 Description: Time out and refuse to reconnect
1356 Details : When the failover node is the primary node, it is possible
1357 to have two identical connections in imp_conn_list. We must
1358 compare not conn's pointers but NIDs, otherwise we can defeat
1359 connection throttling.
1363 Description: port llog fixes from b1_6 into HEAD
1364 Details : Port llog reference couting and some llog cleanups from b1_6
1365 (bug 10800) into HEAD, for protect from panic and access to already
1366 free llog structures.
1370 Description: Detect stride IO mode in read-ahead
1371 Details : When a client does stride read, read-ahead should detect that and
1372 read-ahead pages according to the detected stride pattern.
1376 Description: data checksumming impacts single node performance
1377 Details : add support for several checksum algorithm. Currently, only CRC32
1378 and Adler-32 are supported. The checksum type can be changed on
1379 the fly via /proc/fs/lustre/osc/*/checksum_type.
1383 Description: use adler32 for page checksums
1384 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1389 Description: build for x2 fails
1390 Details : fix compile issue on Cray systems.
1394 Description: Properly match for duplicate locks
1395 Details : Due to different lock order from skiplists code, we need to
1396 traverse entire list for now
1399 Frequency : only on PPC/SLES10
1401 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1402 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1403 be defined. Add a hack in mkfs_lustre.c to work around this compile
1408 Description: LASSERT on MDS when client holding flock lock dies
1409 Details : ldlm pool logic depends on number of granted locks equal to
1410 number of released locks which is not true for flock locks, so
1411 just exclude such locks from consideration.
1415 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1416 Details : Use fsfilt_sync() for both the whole filesystem sync and
1417 individual file sync to eliminate dangerous inode locking
1418 with I_LOCK that can lead to a deadlock.
1422 Description: performance in 1.6.3
1423 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1424 sure that 1MB requests are not fragmented by the block layer.
1426 Severity : enhancement
1428 Description: organize the server-side client stats on per-nid basis
1429 Details : Change the structure of stats under obdfilter and mds to
1437 The "uuid"s file would list the uuids of _active_ exports.
1438 And the clear entry is to clear all stats and stale nids.
1440 Severity : enhancement
1442 Description: eliminate client locks in face of contention
1443 Details : file contention detection and lockless i/o implementation
1444 for contended files.
1448 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1451 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1453 Description: MDS may lose file striping (and hence file data) in some cases
1454 Details : If there are additional extended attributes stored on the MDS,
1455 in particular ACLs, SELinux, or user attributes (if user_xattr
1456 is specified for the client mount options) then there is a risk
1457 of attribute loss. Additionally, the Lustre file striping
1458 needs to be larger than default (e.g. striped over all OSTs),
1459 and an additional attribute must be stored initially in the
1460 inode and then increase in size enough to be moved to the
1461 external attribute block (e.g. ACL growing in size) for file
1466 Description: skiplist implementation simplification
1467 Details : skiplists are used to group compatible locks on granted list
1468 that was implemented as tracking first and last lock of each
1469 lock group the patch changes that to using doubly linked lists
1473 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1474 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1475 because LASSERT on both the data supplied by a client, and the
1476 data on disk is dangerous and incorrect.
1478 Severity : enhancement
1480 Description: Slow truncate/writes to huge files at high offsets.
1481 Details : Directly associate cached pages to lock that protect those pages,
1482 this allows us to quickly find what pages to write and remove
1483 once lock callback is received.
1487 Description: more ldlm soft lockups
1488 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1489 starve other threads from the resource lock for a long time in
1490 case of long waiting queue, so change the debug level from
1491 D_OTHER to the less frequently used D_INFO.
1493 Severity : enhancement
1495 Description: add -gid, -group, -uid, -user options to lfs find
1499 Description: Hung threads in invalidate_inode_pages2_range
1500 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1501 once one is completed. As a result, some RPCs are stuck in the
1502 queue and are never sent.
1506 Description: filter threads hungs on waiting journal commit
1507 Details : Cleanup filter group llog code, then only filter group llog will
1508 be only created in the MDS/OST syncing process.
1512 Description: Procfs and llog threads access destoryed import sometimes.
1513 Details : Sync the import destoryed process with procfs and llog threads by
1514 the import refcount and semaphore.
1516 Severity : enhancement
1518 Description: openlock cache of b1_6 port to HEAD
1523 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1524 Details : If insertion of an extent fails, then discard the inode
1525 preallocation and free data blocks else it can lead to duplicate
1530 Description: don't always update ctime in ext3_xattr_set_handle()
1531 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1532 In some cases the ctime should not be updated, for example for
1533 2.0->1.8 compatibility it is necessary to delete an xattr and it
1534 should not update the ctime.
1538 Bugzilla : 15713/16362
1539 Description: Assertion in iopen_connect_dentry in 1.6.3
1540 Details : looking up an inode via iopen with the wrong generation number can
1541 populate the dcache with a disconneced dentry while the inode
1542 number is in the process of being reallocated. This causes an
1543 assertion failure in iopen since the inode's dentry list contains
1544 both a connected and disconnected dentry.
1548 Description: assertion failure in ldlm_handle2lock()
1549 Details : fix a race between class_handle_unhash() and class_handle2object()
1550 introduced in lustre 1.6.5 by bug 13622.
1555 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1556 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1561 Description: oops in sd_iostats_seq_show()
1562 Details : unloading/reloading the scsi low level driver triggers a kernel
1563 bug when trying to access the sd iostat file.
1568 Description: Kernel panics during QLogic driver reload
1569 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1570 patch, causing memory corruption.
1575 Description: journal_dev option does not work in b1_6
1576 Details : pass mount option during pre-mount.
1578 Severity : enhancement
1580 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1581 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1582 extent information of a file. It can be used to map logical blocks
1583 in a file to physical blocks in the block device.
1587 Description: LDLM soft lockups - improvement
1588 Details : It is be possible to send the lock handle along with each read
1589 or write request because the client is already doing a lock match
1590 itself so there isn't any reason the OST should have to re-do that
1594 Frequency : only on Cray X2
1596 Description: X2 build failures
1597 Details : fix build failures on Cray X2.
1601 Description: xid & resent requests
1602 Details : Initialize RPC XID from clock at startup (randomly if clock is
1605 Severity : enhancement
1607 Description: Add lustre_start utility to start or stop multiple Lustre servers
1612 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1613 Details : In case of memory pressure, list_del() can be called twice on
1614 req->rq_history_list, causing a kernel oops.
1618 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1619 Details : fix stack overflow in the distributed lock manager by defering
1620 export eviction after a failed AST to the elt thread instead of
1621 handling it in the dlm interpret routine.
1625 Description: Convert some comments to new format.
1626 Details : Update documenting comments to match doxygen conventions.
1630 Description: Grammar fixes.
1631 Details : A couple of trivial spelling fixes.
1635 Description: OSD_COUNTERS-mandatory
1636 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1637 to simplify the code.
1641 Description: simplify cmm_device freeing logic.
1642 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1646 Description: Add lockdep support to dt_object_operations locking interface.
1647 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1648 parameter indicating lock ordering. Update mdd code to use new
1653 Description: Introduce failloc constants for lockless IO tests.
1654 Details : Add two new failloc constants to test lockless IO. Only one of
1655 them in implemented---another is checked in yet to be landed
1660 Description: Add lockdep support for inode mutex.
1661 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1662 in the situations where Lustre has to lock more than one inode
1667 Description: Add optional invariants checking support.
1668 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1669 configure switch. This macro is to be used for consistency and
1670 sanity checks that are too expensive to be left in `production'
1675 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1676 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1681 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1682 Details : Use spin_lock_nested() in (the only) situation where more than
1683 one ldlm_lock is locked simultaneously. Also, fix possible
1684 dead-lock in ldlm_lock_change_resource() by enforcing particular
1689 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1690 Details : Instead of specifying each ldlm_lock call-back through separate
1691 parameter, wrap them into struct ldlm_callback_suite.
1695 Description: Kill join_lru obd method and its callers.
1696 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1697 in memory---a requirement implemented through ->o_join_lru() obd
1698 method in HEAD. Remove this method and its users.
1702 Description: Add asynchronous ldlm ENQUEUE completion handler.
1703 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1704 case that stock ldlm_completion_ast() cannot handle as it waits
1705 until lock is granted. Introduce new ldlm_completion_ast_async()
1706 for this. Also comment ldlm_completion_ast().
1710 Description: ldlm_error <-> errno conversion.
1711 Details : Add functions to map (rather arbitrary) between LDLM error codes
1712 and standard errno values. CLIO needs this to prevent LDLM specific
1713 constants from escaping ldlm and osc.
1717 Description: Kill unused ldlm_handle2lock_ns() function.
1718 Details : Kill unused ldlm_handle2lock_ns() function.
1722 Description: Add lu_ref support to ldlm_lock
1723 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1724 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1725 and ldlm_resource. LDLM interface has to be changed, because code that
1726 releases a reference on a lock, has to "know" what reference this is.
1727 In the most frequent case
1729 lock = ldlm_handle2lock(handle);
1731 LDLM_LOCK_PUT(lock);
1733 no changes are required. When any other reference (received _not_ from
1734 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1735 instead of LDLM_LOCK_PUT().
1737 Arguably, changes are pervasive, and interface requires some discipline
1738 for proper use. On the other hand, it was very instrumental in finding
1739 a few leaked lock references.
1743 Description: Add ldlm_lock_addref_try().
1744 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1745 attempts to addref a lock that might be being canceled
1750 Description: Add ldlm_weigh_callback().
1751 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1752 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1753 hard-coded `number of pages' logic.
1757 Description: Add lockdep annotations to llog code.
1758 Details : Use appropriately tagged _nested() locking calls in the places
1759 where llog takes more than one ->lgh_lock lock.
1763 Description: Add loi_kms_set().
1764 Details : Wrap kms updates into a helper function.
1768 Description: Constify instances of struct lsm_operations.
1769 Details : Constify instances of struct lsm_operations.
1773 Description: lu_conf support.
1774 Details : On a server, a file system object is uniquely identified
1775 by a fid, which is sufficient to locate and load all object
1776 state (inode). On a client, on the other hand, more data are
1777 necessary instantiate an object. Change lu_object_find() and
1778 friends to take additional `lu_conf' argument describing object.
1779 Typically this includes layout information.
1783 Description: lu_context fixes.
1784 Details : Introduce new lu_context functions that are needed on the client
1785 side, where some system threads (ptlrpcd) are shared by multiple
1786 modules, and so cannot be stopped during module shutdown.
1790 Description: Add start and stop methods to lu_device_type_operations.
1791 Details : Introduce two new methods in lu_device_type_operations, that are
1792 invoked when first instance of a given type is created and last one
1793 is destroyed respectively. This is need by CLIO.
1797 Description: Add lu_ref support to struct lu_device.
1798 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1799 track leaked references.
1803 Description: Introduce lu_kmem_descr.
1804 Details : lu_kmem_descr and its companion interface allow to create
1805 and destroy a number of kmem caches at once.
1809 Description: Fix lu_object finalization race.
1810 Details : Fix a race between lu_object_find() finding an object and its
1811 concurrent finalization. This race is (most likely) not possible
1812 on the server, but might happen on the client.
1816 Description: Introduce lu_ref interface.
1817 Details : lu_ref is a debugging module allowing to track references to
1818 a given object. It is quite cpu expensive, and has to be
1819 explicitly enabled with --enable-lu_ref. See usage description
1824 Description: Factor lu_site procfs stats into a separate function.
1825 Details : Separate lu_site stats printing code into a separate function
1826 that can be reused on a client.
1830 Description: Constify instances of struct {lu,dt,md}_device_operations.
1831 Details : Constify instances of struct {lu,dt,md}_device_operations.
1835 Description: Introduce struct md_site and move meta-data specific parts of
1836 struct lu_site here.
1837 Details : Move md-specific fields out of struct lu_site into special struct
1838 md_site, so that lu_site can be used on a client.
1842 Description: Kill mdd_lov_destroy().
1843 Details : Remove unused mdd code.
1847 Description: Add st_block checking to multistat.c.
1848 Details : Add st_block checking to multistat.c.
1852 Description: Add lu_ref support to struct obd_device.
1853 Details : Add lu_ref tracking to obd_device.
1857 Description: Kill obd_set_fail_loc().
1858 Details : Remove unused code.
1862 Description: Add special type for ptlrpc_request interpret functions.
1863 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1864 there. Actual usage will be in CLIO.
1868 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1869 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1870 doesn't work with lockdep.
1874 Description: Add rwv.c test program.
1875 Details : New testing program exercising readv(2) and writev(2) (Qian).
1879 Description: Add sendfile.c test program.
1880 Details : New testing program exercising sendfile(2) (Jay).
1884 Description: Ratelimit a message that can be very frequent.
1885 Details : Ratelimit a memory allocation failure message that can
1890 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1891 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1895 Description: Liblustre build fixes.
1896 Details : Liblustre build fixes.
1900 Description: libcfs: add cfs_{need,cond}_resched() interface.
1901 Details : libcfs: add cfs_{need,cond}_resched() definition and
1902 implementations for Linux, NT, and liblustre.
1904 Severity : enhancement
1906 Description: More exported tunables for mballoc
1907 Details : Add support for tunable preallocation window and new tunables for
1908 large/small requests
1912 Description: Detect corruption of block bitmap and checking for preallocations
1913 Details : Checks validity of on-disk block bitmap. Also it does better
1914 checking of number of applied preallocations. When corruption is
1915 found, it turns filesystem readonly to prevent further corruptions.
1919 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1920 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1921 window to 0. If it does hit cache miss, and read-pattern is still
1922 stride-io mode, does not reset the stride window, but also does not
1923 increase the stride window length in this case.
1927 Frequency : only for big-endian servers
1928 Description: Check if system is big-endian while mounting fs with extents feature
1929 Details : Mounting a filesystem with extents feature will fail on big-endian
1930 systems since ext3-based ldiskfs is not supported on big-endian
1931 systems. This can be over-riden with "bigendian_extents" mount option.
1933 Severity : enhancement
1935 Description: The root squash functionality
1936 Details : A security feature, which is to prevent users from being able
1937 to mount lustre on their desktop, run as root, and delete
1938 all of the files in the filesystem. The goal is accomplished by
1939 remapping user id (UID) and group id (GID) of the root user to
1940 a UID and GID specified by the system administartor via Lustre
1941 configuration management server (MGS). The functionality also
1942 allows to specify sets of clients for which the remapping does
1947 Description: Excessive recovery window
1948 Details : With AT enabled, the recovery window can be excessively long (6000+
1949 seconds). To address this problem, we no longer use
1950 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1951 timeout no longer depends on the service time, it is set to
1952 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1953 time via pb_service_time.
1957 Description: Watchdog triggered on MDS failover
1958 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
1959 the OSTs know that the MDS "UUID" can be reused for the same export
1960 from a different NID, so we do not need to wait for the export to be
1964 Frequency : rare, only if using MMP with Linux RAID
1966 Description: MMP doesn't work with Linux RAID
1967 Details : While using HA for Lustre servers with Linux RAID, it is possible
1968 that MMP will not detect multiple mounts. To make this work we
1969 need to unplug the device queue in RAID when the MMP block is being
1970 written. Also while reading the MMP block, we should read it from
1971 disk and not the cached one.
1973 Severity : enhancement
1975 Description: open file using fid
1976 Details : A file can be opened using just its fid, like
1977 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
1979 --------------------------------------------------------------------------------
1981 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1983 * Support for kernels:
1984 2.6.5-7.283 (SLES 9),
1985 2.6.9-55.EL (RHEL 4),
1986 2.6.16.46-0.14 (SLES 10),
1987 2.6.18.8 vanilla (kernel.org)
1988 * Client support for unpatched kernels:
1989 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1990 2.6.16 - 2.6.22 vanilla (kernel.org)
1991 * Due to recently discovered recovery problems, we do not recommend
1992 using patchless RHEL 4 clients with this or any earlier release.
1993 * Recommended e2fsprogs version: 1.39.cfs8
1994 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1995 * Starting with this release, the ldiskfs backing filesystem required
1996 by Lustre is now in its own package, lustre-ldiskfs. This package
1997 should be installed. It is versioned separately from Lustre and
1998 may be released separately in future.
2002 Description: block reactivating mgc import until all deactivates complete
2003 Details : Fix race when failing back MDT/MGS to itself (testing)
2006 Frequency : at statup only
2008 Description: mds_lov_synchronize race leads to various problems
2009 Details : simultaneous MDT->OST connections at startup can cause the
2010 sync to abort, leaving the OSC in a bad state.
2012 Severity : enhancement
2014 Description: add optional extra BUILD_VERSION info
2015 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2016 to override the lustre version.
2019 Frequency : 2.6.18 servers only
2021 Description: ll_kern_mount() doesn't release the module reference
2022 Details : The ldiskfs module reference count never drops down to 0
2023 because ll_kern_mount() doesn't release the module reference.
2028 Description: server LBUG when using old ost_num_threads parameter
2029 Details : Accept the old ost_num_threads parameter but warn that it
2030 is deprecated, and fix an off-by-one error that caused an LBUG.
2035 Description: Transient SCSI error results in persistent IO issue
2036 Details : iobuf->dr_error is not reinitialized to 0 between two
2040 Frequency : sometimes when underlying device returns I/O errors
2042 Description: OSTs not going read-only during write failures
2043 Details : OSTs are not remounted read-only when the journal commit threads
2044 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2045 instead of the ext3 wrappers.
2049 Description: poor connect scaling with increasing client count
2050 Details : Don't run filter_grant_sanity_check for more than 100 exports
2051 to improve scaling for large numbers of clients.
2054 Frequency : SLES10 only
2056 Description: sanity-quota.sh quotacheck failed: rc = -22
2057 Details : Quotas cannot be enabled on SLES10.
2060 Frequency : liblustre clients only
2062 Description: getdirentries does not give error when run on compute nodes
2063 Details : getdirentries does not fail when the size specified as an argument
2064 is too small to contain at least one entry
2066 Severity : enhancement
2068 Description: Add LNET router traceability for debug purposes
2069 Details : If a checksum failure occurs with a router as part of the
2070 IO path, the NID of the last router that forwarded the bulk data
2071 is printed so it can be identified.
2076 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2077 Details : Due to a race condition, liblustre clients were occasionally
2078 evicted incorrectly.
2080 Severity : enhancement
2082 Description: lfs setstripe use optional parameters instead of postional
2085 Severity : enhancement
2087 Description: Nanosecond timestamp support for ldiskfs
2088 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2089 resolution timestamps. There is not yet support for this at
2090 the Lustre filesystem level.
2093 Frequency : during server recovery
2095 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2096 Details : request with rq_no_resend flag not awake l_wait_event if they get
2100 Frequency : nfs export on patchless client
2102 Description: connectathon hang when test nfs export over patchless client
2103 Details : Disconnected dentry cannot be found with lookup, so we do not need
2104 to unhash it or make it invalid
2107 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2108 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2109 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2114 Description: open req refcounting wrong on reconnect
2115 Details : If reconnect happened between getting open reply from server and
2116 call to mdc_set_replay_data in ll_file_open, we will schedule
2117 replay for unreferenced request that we are about to free.
2118 Subsequent close will crash in variety of ways.
2119 Check that request is still eligible for replay in
2120 mdc_set_replay_data().
2125 Description: disable writes to filesystem when reading health_check file
2126 Details : the default for reading the health_check proc file has changed
2127 to NOT do a journal transaction and write to disk, because this
2128 can cause reads of the /proc file to hang and block HA state
2129 checking on a healthy but otherwise heavily loaded system. It
2130 is possible to return to the previous behaviour during configure
2131 with --enable-health-write.
2133 Severity : enhancement
2135 Description: 64-bit inode version
2136 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2137 to the inode. This will be required for version-based recovery.
2142 Description: MDS fails to start if a duplicate client export is detected
2143 Details : in some rare cases it was possible for a client to connect to
2144 an MDS multiple times. Upon recovery the MDS would detect this
2145 and fail during startup. Handle this more gracefully.
2147 Severity : enhancement
2149 Description: Add -o localflock option to simulate old noflock
2151 Details : This will achieve local-only flock/fcntl locks
2157 Description: log_commit_thread vs filter_destroy race leads to crash
2158 Details : Take import reference before releasing llog record semaphore
2163 Description: Wrong request locking in request set processing
2164 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2165 imp_delayed_list, in this place should be used imp_lock.
2168 Frequency : when reconnection
2170 Description: Grant Leak when osc reconnect to OST
2171 Details : When osc reconnect ost, OST(filter) should check whether it
2172 should grant more space to client by comparing fed_grant and
2173 cl_avail_grant, and return the granted space to client instead
2174 of "new granted" space, because client will call osc_init_grant
2175 to update the client grant space info.
2178 Frequency : when client reconnect to OST
2180 Description: Grant Leak when osc do resend and replay bulk write
2181 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2182 bulk write request, because the grant info while be sync between
2183 OSC and OST when reconnect, and we should ignore the grant info
2184 these of resend/replay write req.
2189 Description: Grant space more than avaiable left space sometimes.
2190 Details : When then OST is about to be full, if two bulk writing from
2191 different clients came to OST. Accord the avaliable space of the
2192 OST, the first req should be permitted, and the second one
2193 should be denied by ENOSPC. But if the seconde arrived before
2194 the first one is commited. The OST might wrongly permit second
2195 writing, which will cause grant space > avaiable space.
2198 Frequency : when client is evicted
2200 Description: Grant might be wrongly erased when osc is evicted by OST
2201 Details : when the import is evicted by server, it will fork another
2202 thread ptlrpc_invalidate_import_thread to invalidate the
2203 import, where the grant will be set to 0. While the original
2204 thread will update the grant it got when connecting. So if
2205 the former happened latter, the grant will be wrongly errased
2206 because of this race.
2211 Description: Checking Stale with correct fid
2212 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2213 to check whether it is stale, because sometimes, we want the
2214 enqueue happened anyway, and op_data.fid2 will not be initialized.
2216 Severity : enhancement
2218 Description: update patchless client
2219 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2222 Frequency : only with 2.4 kernel
2224 Description: random memory corruption
2225 Details : size of struct ll_inode_info is to big for union inode.u and this
2226 can be cause of random memory corruption.
2231 Description: Memory leak in recovery
2232 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2233 It should also check obd_fail before fsfilt_start, otherwise if
2234 fsfilt_start return -EROFS,(failover mds during mds recovery).
2235 then the req will return with repmsg->transno = 0 and rc = EROFS.
2236 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2237 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2238 be freed no matter whether fsfilt_commit success or not.
2241 Frequency : only with huge count clients
2243 Description: Prevents from taking the superblock lock in llap_from_page for
2245 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2246 for llap_from_page call in ll_removepage prevents from taking the
2247 superblock lock for a soon died page.
2252 Description: Not check open intent error before release open handle
2253 Details : in some rare cases, the open intent error is not checked before
2254 release open handle, which may cause
2255 ASSERTION(open_req->rq_transno != 0), because it tries to release
2256 the failed open handle.
2261 Description: Set cat log bitmap only after create log success.
2262 Details : in some rare cases, the cat log bitmap is set too early. and it
2263 should be set only after create log success.
2267 Description: Accessing a block bevice can re-enable I/O when Lustre is
2268 tearing down a device.
2269 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2274 Description: service threads may hog cpus when there are a lot of requests
2275 Details : Insert cond_resched to give other threads a chance to use some CPU
2280 Description: the cat log was not initialized in recovery
2281 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2282 unlink log on mds will not be initialized until mds post
2283 recovery. And also in mds post recovery, the unlink log will
2284 initialization will be done asynchronausly, so there will be race
2285 between add unlink log and unlink log initialization.
2289 Description: brw_stats were being printed incorrectly
2290 Details : brw_stats were being printed as log2 but all of them were not
2291 recorded as log2. Also remove some code duplication arising from
2292 filter_tally_{read,write}.
2296 Frequency : rare, only in recovery.
2297 Description: ASSERTION(req->rq_type != LI_POISON) failed
2298 Details : imp_lock should be held while iterating over imp_sending_list for
2299 prevent destroy request after get timeout in ptlrpc_queue_wait.
2303 Description: replay-single.sh test 52 fails
2304 Details : A lock's skiplist need to be cleanup when it being unlinked
2305 from its resource list.
2309 Description: Short directio read returns full requested size rather than
2311 Details : Direct I/O operations should return actual amount of bytes
2312 transferred rather than requested size.
2314 Severity : enhancement
2316 Description: metadata RPC reduction (e.g. for rm performance)
2317 Details : decrease the amount of synchronous RPC between clients and servers
2318 by canceling conflicing lock before the operation on the client
2319 and packing thier handles into the main operation RPC to server.
2321 Severity : enhancement
2323 Description: Async OSC create to avoid the blocking unnecessarily.
2324 Details : If a OST has no remain object, system will block on the creating
2325 when need to create a new object on this OST. Now, ways use
2326 pre-created objects when available, instead of blocking on an
2327 empty osc while others are not empty. If we must block, we block
2328 for the shortest possible period of time.
2332 Description: improve handling recoverable errors
2333 Details : if request processig with error which can be recoverable on server
2334 request should be resend, otherwise page released from cache and
2337 Severity : enhancement
2339 Description: refine locking for avoid write wrong info into lov_objid file
2340 Details : fix possible races with add new target and write/update data in
2343 --------------------------------------------------------------------------------
2345 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2350 Frequency : on some architectures
2352 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2353 Details : Uninitialized flags sometimes cause configuration commands to
2357 Frequency : patchless clients only
2359 Description: missing __iget() symbol export
2360 Details : The __iget() symbol export is missing. To avoid the need for
2361 this on patchless clients the deathrow inode reaper is turned
2362 off, and we depend on the VM to clean up old inodes. This
2363 dependency was during via the fix for bug 12181.
2365 --------------------------------------------------------------------------------
2367 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2369 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2370 INTEROPERATE with older versions automatically. In many cases a
2371 special upgrade step is needed. Please read the
2372 user documentation before upgrading any part of a live system.
2373 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2374 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2375 * WARNING: Lustre configuration and startup changes are required with
2376 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2380 Severity : enhancement
2382 Description: MountConf
2383 Details : Lustre configuration is now managed via mkfs and mount
2384 commands instead of lmc and lconf. New obd types (MGS, MGC)
2385 are added for dynamic configuration management. See
2386 https://mail.clusterfs.com/wikis/lustre/MountConf for
2389 Severity : enhancement
2391 Description: dynamic OST addition
2392 Details : OSTs can now be added to a live filesystem
2394 Severity : enhancement
2396 Description: startup order invariance
2397 Details : MDTs and OSTs can be started in any order. Clients only
2398 require the MDT to complete startup.
2400 Severity : enhancement
2402 Description: parallel, asynchronous orphan cleanup
2403 Details : orphan cleanup is now performed in separate threads for each
2404 OST, allowing parallel non-blocking operation.
2406 Severity : enhancement
2408 Description: optimized stripe assignment
2409 Details : stripe assignments are now made based on ost space available,
2410 ost previous usage, and OSS previous usage, in order to try
2411 to optimize storage space and networking resources.
2413 Severity : enhancement
2415 Description: Permanently set tunables
2416 Details : All writable /proc/fs/lustre tunables can now be permanently
2417 set on a per-server basis, at mkfs time or on a live system.
2419 Severity : enhancement
2421 Description: Lustre message v2
2422 Details : Add lustre message format v2.
2424 Severity : enhancement
2426 Description: client OST exclusion list
2427 Details : Clients can be started with a list of OSTs that should be
2428 declared "inactive" for known non-responsive OSTs.
2432 Description: SPEC SFS validation failure on NFS v2 over lustre.
2433 Details : Changes the blocksize for regular files to be 2x RPC size,
2434 and not depend on stripe size.
2436 Severity : enhancement
2438 Description: Multiple MD RPCs in flight.
2439 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2440 To avoid overly-overloading MDS, introduce a limit on number of
2441 MDS RPCs in flight for a single client and add /proc controls
2442 to adjust this limit.
2444 Severity : enhancement
2446 Description: client read/write statistics
2447 Details : Add client read/write call usage stats for performance
2448 analysis of user processes.
2449 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2450 file access. extents_stats shows chunk size distribution.
2451 extents_stats_per_process show chunk size distribution per
2454 Severity : enhancement
2456 Description: mds statistics
2457 Details : Add detailed mds operations statistics in
2458 /proc/fs/lustre/mds/*/stats.
2462 Description: Failure of copying files with lustre special EAs.
2463 Details : Client side always return success for setxattr call for lustre
2464 special xattr (currently only "trusted.lov").
2469 Description: Refcount LNET uuids
2470 Details : The global LNET uuid list grew linearly with every startup;
2471 refcount repeated list entries instead of always adding to
2474 Severity : enhancement
2476 Description: Dynamic service threads
2477 Details : Within a small range, start extra service threads
2478 automatically when the request queue builds up.
2481 Frequency : mixed-endian client/server environments
2483 Description: mixed-endian crashes
2484 Details : The new msg_v2 system had some failures in mixed-endian
2487 Severity : enhancement
2489 Description: Easy OST removal
2490 Details : OSTs can be permanently deactivated with e.g. 'lctl
2491 conf_param lustre-OST0001.osc.active=0'
2493 Severity : enhancement
2495 Description: MGS proc entries
2496 Details : Added basic proc entries for the MGS showing what filesystems
2499 Severity : enhancement
2501 Description: provide MGS failover
2502 Details : Added config lock reacquisition after MGS server failover.
2504 Severity : enhancement
2506 Description: add Linux 2.4 support
2507 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2511 Description: a large application tries to do I/O to the same resource and dies
2512 in the middle of it.
2513 Details : Check the req->rq_arrival time after the call to
2514 ost_brw_lock_get(), but before we do anything about
2515 processing it & sending the BULK transfer request. This
2516 should help move old stale pending locks off the queue as
2517 quickly as obd_timeout.
2520 Frequency : when an incorrect nid is specified during startup
2522 Description: ptlrpc connect to non-existant node causes kernel crash
2523 Details : LNET can't be re-entered from an event callback, which
2524 happened when we expire a message after the export has been
2525 cleaned up. Instead, hand the zombie cleanup off to another
2528 Severity : enhancement
2530 Description: plain/inodebits lock performance improvement
2531 Details : Grouping plain/inodebits in granted list by their request modes
2532 and bits policy, thus improving the performance of search through
2536 Frequency : only if OST filesystem is corrupted
2538 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2539 Details : for a short time RPCs with bulk IO are in the replay list,
2540 but replay of bulk IOs is unimplemented. If the OST filesystem
2541 is corrupted due to disk cache incoherency and then replay is
2542 started it is possible to trip an assertion. Avoid putting
2543 committed RPCs into the replay list at all to avoid this issue.
2546 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2548 Description: System hang on startup
2549 Details : This bug allowed the liblustre (e.g. catamount) client to
2550 return to the app before handling all startup RPCs. This
2551 could leave the node unresponsive to lustre network traffic
2552 and manifested as a server ptllnd timeout.
2554 Severity : enhancement
2556 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2557 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2558 boolean to control whether to print peer debug info when a
2559 client's RPC times out.
2562 Frequency : only for kernels with patches from Lustre below 1.4.3
2564 Description: Remove old rdonly API
2565 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2568 Frequency : only for devices with external journals
2570 Description: Set external device read-only also
2571 Details : During a commanded failover stop, we set the disk device
2572 read-only while the server shuts down. We now also set any
2573 external journal device read-only at the same time.
2576 Frequency : when upgrading from 1.4 while trying to change parameters
2578 Description: The wrong (new) MDC name was used when setting parameters for
2579 upgraded MDT's. Also allows changing of OSC (and MDC)
2580 parameters if --writeconf is specified at tunefs upgrade time.
2583 Frequency : when setting specific ost indicies
2585 Description: QOS code breaks on skipped indicies
2586 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2587 created with --index need not be sequential.
2589 Severity : enhancement
2591 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2592 Details : The uninit_groups feature works in conjunction with the kernel
2593 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2594 up the pass1 processing of e2fsck. This is a read-only feature
2595 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2596 mount filesystems that have had this feature enabled.
2598 Severity : enhancement
2600 Description: Improve multi-block allocation algorithm to avoid fragmentation
2601 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2602 allocation locality and avoid filesystem fragmentation.
2604 ------------------------------------------------------------------------------
2606 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2608 * Support for kernels:
2609 2.4.21-47.0.1.EL (RHEL 3)
2610 2.6.5-7.283 (SLES 9)
2611 2.6.9-42.0.10.EL (RHEL 4)
2612 2.6.12.6 vanilla (kernel.org)
2613 2.6.16.27-0.9 (SLES 10)
2614 * Recommended e2fsprogs version: 1.39.cfs5
2616 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2620 Frequency : occasional, depends on client load and configuration
2621 Bugzilla : 12181, 12203
2622 Description: data loss for recently-modified files
2624 Details : In some cases it is possible that recently written or created
2625 files may not be written to disk in a timely manner (this should
2626 normally be within 30s unless client IO load is very high).
2627 The problem appears as zero-length files or files that are a
2628 multiple of 1MB in size after a client crash or client eviction
2629 that are missing data at the end of the file.
2631 This problem is more likely to be hit on clients where files are
2632 repeatedly created and unlinked in the same directory, clients
2633 have a large amount of RAM, have many CPUs, the filesystem has
2634 many OSTs, the clients are rebooted frequently, and/or the files
2635 are not accessed by other nodes after being written.
2637 The presence of the problem can be detected by looking at
2638 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2639 smaller than the second (nr_unused) then dirty files will not
2640 be flushed automatically to disk. "sync; sleep 10" should be
2641 run several times on the node before unmounting it to update
2642 Lustre (this is also safe to run on nodes without this problem).
2644 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2645 that can cause this same problem, so customers using that kernel
2646 also need to update the kernel in addition to Lustre. In order
2647 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2649 It is normal that files written just before a client crash (less
2650 than 30s) may not yet have been flushed to disk, even for local
2654 Frequency : frequent on thin XT3 nodes
2656 Description: UUID collision on thin XT3 Linux nodes
2657 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2658 randomly, since we relied on an insufficiently-seeded PRNG.
2663 Description: OSS hangs after "All ost request buffers busy"
2664 Details : A deadlock between quota and journal operations caused OSS
2665 hangs after printing "All ost request buffers busy."
2668 Frequency : always on liblustre builds
2670 Description: Cleanup compiler warnings on liblustre
2673 Frequency : always on liblustre builds on XT3
2675 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2679 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2680 > 32000 subdirectories
2681 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2682 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2683 subdirectory count crosses 32000. This will aid e2fsck to
2684 correctly handle more than 32000 subdirectories.
2687 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2689 Description: System hang on startup
2690 Details : This bug allowed the liblustre (e.g. catamount) client to
2691 return to the app before handling all startup RPCs. This
2692 could leave the node unresponsive to lustre network traffic
2693 and manifested as a server ptllnd timeout.
2695 Severity : enhancement
2697 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2698 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2699 boolean to control whether to print peer debug info when a
2700 client's RPC times out.
2705 Description: make O_SYNC working on 2.6 kernels
2706 Details : 2.6 kernels use different method for mark pages for write,
2707 so need add a code to lustre for O_SYNC work.
2712 Description: Failure to close file and release space on NFS
2713 Details : Put inode details into lock acquired in ll_intent_file_open.
2714 Use mdc_intent_lock in ll_intent_open to properly
2715 detect all kind of errors unhandled by mdc_enqueue.
2720 Description: proc file read during shutdown sometimes raced obd removal,
2722 Details : Add lock to prevent obd access after proc file removal.
2725 Frequency : Only for files larger than 4GB on 32-bit clients.
2727 Description: improperly doing page alignment of locks
2728 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2729 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2734 Description: RPCs being resent when they shouldn't be
2735 Details : Some RPCs that should not be resent are being resent. This
2736 can cause inconsistencies in the RPC state machine. Do not
2737 resend such requests.
2740 Frequency : rare, only with NFS export
2742 Description: Crash on NFS re-export node
2743 Details : under very unusual load conditions an assertion is hit in
2744 ll_intent_file_open()
2747 Frequency : only if OST filesystem is corrupted
2749 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2750 Details : for a short time RPCs with bulk IO are in the replay list,
2751 but replay of bulk IOs is unimplemented. If the OST filesystem
2752 is corrupted due to disk cache incoherency and then replay is
2753 started it is possible to trip an assertion. Avoid putting
2754 committed RPCs into the replay list at all to avoid this issue.
2759 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2760 Details : Large single O_DIRECT read and write calls can fail to allocate
2761 a sufficiently large buffer to process the request. In case of
2762 allocation failure the allocation is retried with a smaller
2763 buffer and broken into smaller requests.
2765 Severity : enhancement
2767 Description: Add -o localflock option to simulate old noflock behaviour.
2768 Details : This will achieve local-only flock/fcntl locks coherentness.
2773 Description: versioning check is incomplete
2774 Details : Checking the version difference of client vs. server, report
2775 error if the gap is too big.
2780 Description: add support PG_writeback bit
2781 Details : add support for PG_writeback bit for Lustre, for more carefull
2782 work with page cache in 2.6 kernel. This also fix some deadlocks
2783 and remove hack for work O_SYNC with 2.6 kernel.
2785 Severity : enhancement
2787 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2788 Details : The uninit_groups feature works in conjunction with the kernel
2789 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2790 up the pass1 processing of e2fsck. This is a read-only feature
2791 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2792 mount filesystems that have had this feature enabled.
2794 Severity : enhancement
2796 Description: Improve multi-block allocation algorithm to avoid fragmentation
2797 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2798 allocation locality and avoid filesystem fragmentation.
2800 ------------------------------------------------------------------------------
2802 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2804 * Support for kernels:
2805 2.6.9-42.0.3.EL (RHEL 4)
2806 2.6.5-7.276 (SLES 9)
2807 2.4.21-47.0.1.EL (RHEL 3)
2808 2.6.12.6 vanilla (kernel.org)
2809 2.6.16.21-0.8 (SLES10)
2810 * Recommended e2fsprogs version: 1.39.cfs2-0
2812 * The backwards-compatible /proc/sys/portals symlink has been removed
2813 in this release. Before upgrading, please ensure that you change
2814 any configuration scripts or /etc/sysctl.conf files that access
2815 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2816 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2817 in advance of the upgrade on any system running Lustre 1.4.6 or
2818 newer, since /proc/sys/lnet was added in that version.
2819 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2823 Frequency : only when quota is used
2825 Description: avoid scanning export list for quota master
2826 Details : Change the algorithms to avoid scanning export list in order
2827 to improve the efficiency.
2830 Frequency : MDS failover only, very rarely
2832 Description: "went back in time" messages on mds failover
2833 Details : The greatest transno may be lost when the current operation
2834 finishes with an error (transno==0) and the client's last_rcvd
2835 record is over-written. Save the greatest transno in the
2836 mds_last_transno for this case.
2839 Frequency : always for specific kernels and striping counts
2841 Description: client may get "Matching packet too big" without ACL support
2842 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2843 when trying to access files in certain configurations. The
2844 clients should in fact be denied when mounting because they do
2845 not understand ACLs.
2848 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2850 Description: many clients connecting with IO in progress causes connect timeouts
2851 Details : Avoid synchronous journal commits to avoid delays caused by many
2852 clients connecting/disconnecting when bulk IO is in progress.
2853 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2854 OST_IO_PORTAL to avoid delays behind potentially many pending
2858 Frequency : occasionally with multiple writers to a single file
2860 Description: shared writes to file may result in wrong size reported by stat()
2861 Details : Allow growing of kms when extent lock is cancelled
2864 Frequency : always with random mmap IO to multi-striped file
2866 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2867 Details : Only if the hole is at the end of OST object so that kms is too
2868 small. Fix is to increase kms accordingly in ll_nopage.
2871 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2873 Description: writes to a missing object would leak memory on the OST
2874 Details : If there is an inconsistency between the MDS and OST filesystems,
2875 such that the MDS references an object that doesn't exist, writes
2876 to that object will leak memory due to incorrect cleanup in the
2877 error handling path, eventually running out of memory on the OST.
2882 Description: Creating too long symlink causes lustre errors
2883 Details : Check symlink and name lengths before sending requests to MDS.
2886 Frequency : only if flock is enabled (not on by default)
2888 Description: posix locks not released on fd closure on 2.6.9+
2889 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2890 kernels, this caused such locks not to be released on fd close and
2891 then assertions on fs unmount about still used locks.
2894 Frequency : MDS failover only, very rarely
2896 Description: clients may get ASSERTION(granted_lock != NULL)
2897 Details : When request was taking a long time, and a client was resending
2898 a getattr by name lock request. The were multiple lock requests
2899 with the same client lock handle and
2900 mds_getattr_name->fixup_handle_for_resent_request found one of the
2901 lock handles but later failed with ASSERTION(granted_lock != NULL).
2906 Description: handle->h_buffer_credits > 0, assertion failure
2907 Details : h_buffer_credits is zero after truncate, causing assertion
2908 failure. This patch extends the transaction or creates a new
2912 Frequency : NFS re-export or patchless client
2913 Bugzilla : 11179, 10796
2914 Description: Crash on NFS re-export node (__d_move)
2915 Details : We do not want to hash the dentry if we don't have a lock.
2916 But if this dentry is later used in d_move, we'd hit uninitialised
2917 list head d_hash, so we just do this to init d_hash field but
2918 leave dentry unhashed.
2921 Frequency : NFS re-export or patchless client
2923 Description: NFS exports has problem with symbolic link
2924 Details : lustre client didn't properly install dentry when re-exported
2925 to NFS or running patchless client.
2928 Frequency : NFS re-export or patchless client
2930 Description: Various nfs/patchless fixes.
2931 Details : fixes reuse disconected alias for lookup process - this fixes
2932 warning "find_exported_dentry: npd != pd",
2933 fix permission error with open files at nfs.
2934 fix apply umask when do revalidate.
2937 Frequency : occasional
2939 Description: Crash on NFS re-export node
2940 Details : calling clear_page() on the wrong pointer triggered oops in
2941 generic_mapping_read().
2944 Frequency : rarely, using O_DIRECT IO
2946 Description: unaligned directio crashes client with LASSERT
2947 Details : check for unaligned buffers before trying any requests.
2950 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2952 Description: stale data returned from RAID cache
2953 Details : If only a small amount of IO is done to the RAID device before
2954 reading it again it is possible to get stale data from the RAID
2955 cache instead of reading it from disk.
2958 Frequency : always for sles10 kernel
2960 Description: sles10 support
2961 Details : ll_follow_link: compile fixes and using of nd_set_link
2962 under newer kernels.
2965 Frequency : depends on arch, kernel and compiler version, always on sles10
2968 Description: recursive or deep enough symlinks cause stack overflow
2969 Details : getting rid of large stack-allocated variable in
2973 Frequency : depends on hardware
2975 Description: lustre write performance loss in the SLES10 kernel
2976 Details : the performance loss is caused by using of write barriers in the
2977 ext3 code. The SLES10 kernel turns barrier support on by
2978 default. The fix is to undo that change for ldiskfs.
2980 ------------------------------------------------------------------------------
2982 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2984 * Support for kernels:
2985 2.6.9-42.0.3EL (RHEL 4)
2986 2.6.5-7.276 (SLES 9)
2987 2.4.21-47.0.1.EL (RHEL 3)
2988 2.6.12.6 vanilla (kernel.org)
2992 Frequency : quota enabled and large files being deleted
2994 Description: releasing more than 4GB of quota at once hangs OST
2995 Details : If a user deletes more than 4GB of files on a single OST it
2996 will cause the OST to spin in an infinite loop. Release
2997 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3002 Description: statfs data retrieved from /proc may be stale or zero
3003 Details : When reading per-device statfs data from /proc, in the
3004 {kbytes,files}_{total,free,avail} files, it may appear
3005 as zero or be out of date.
3008 Frequency : systems with MD RAID1 external journal devices
3010 Description: lconf's call to blkid is confused by RAID1 journal devices
3011 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3012 of returning all block devices that match the journal UUID.
3015 Frequency : always, for aggregate stripe size over 4GB
3017 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3018 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3019 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3022 Frequency : always if "lfs find" used on a local file/directory
3024 Description: "lfs find" segfaults if used on a local file/directory
3025 Details : The case where a directory component was not specified wasn't
3026 handled correctly. Handle this properly.
3029 Frequency : always on ppc64
3031 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3032 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3033 perform endianness conversion.
3036 Frequency : rarely (truncate to non-zero file size after write under load)
3037 Bugzilla : 10730, 10687
3038 Description: Files padded with zeros to next 4K multiple
3039 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3040 it is possible that files that are truncated to a non-zero size
3041 immediately after being written are filled with zero bytes beyond
3042 the truncated size. No file data is lost.
3044 Severity : enhancement
3046 Description: Allow recovery/failover for liblustre clients.
3047 Details : liblustre clients were unaware of failover configurations until
3050 Severity : enhancement
3052 Description: user file locks should fail when not mounting with flock option
3053 Details : Set up an error-returning stub in ll_file_operations.lock field
3054 to prevent incorrect behaviour when client is mounted without
3055 flock option. Also, set up properly f_op->flock field for
3059 Frequency : always on ia64
3061 Description: "lfs df" loops on printing out MDS statfs information
3062 Details : The obd_ioctl_data was not initialized and in some systems
3063 this caused a failure during the ioctl that did not return
3064 an error. Initialize the struct and return an error on failure.
3067 Frequency : SLES 9 only
3069 Description: Error of copying files with lustre special EAs as root
3070 Details : Client side always return success for setxattr call for lustre
3071 special xattr (currently only "trusted.lov").
3074 Frequency : rarely on clusters with both ia64+i386 clients
3076 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3077 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3078 clients are doing concurrent writes to the same file it is
3079 possible that the ia64 clients may LASSERT because the OST
3080 extent locks are not PAGE_SIZE aligned. Ensure that grown
3081 locks are always aligned on the request boundary.
3084 Frequency : specific use, occasional
3086 Description: Overwriting in use executable truncates on-disk binary image
3087 Details : If one node attempts to overwrite an executable in use by
3088 another node, we now correctly return ETXTBSY instead of
3089 truncating the file.
3091 Severity : enhancement
3093 Description: Async OSC create to avoid the blocking unnecessarily.
3094 Details : If a OST has no remain object, system will block on the creating
3095 when need to create a new object on this OST. Now, ways use
3096 pre-created objects when available, instead of blocking on an
3097 empty osc while others are not empty. If we must block, we block
3098 for the shortest possible period of time.
3103 Description: chmod on Lustre root is propagated to other clients
3104 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3105 invalid by the follow_mount time.
3108 Frequency : liblustre clients only
3110 Description: Race in 'instant cancel' lock handling could lead to such locks
3111 never to be granted in case of SMP MDS
3112 Details : Do not destroy not yet granted but cbpending locks in
3116 Frequency : replay/resend of open
3118 Description: non null lock assetion failure in mds_intent_policy
3119 Details : Trying to replay/resend lockless open requests resulted in
3120 mds_open() returning 0 with no lock. Now it sets a flag if
3121 a lock is going to be returned.
3123 Severity : enhancement
3125 Description: Checksum enhancements
3126 Details : New checksum enhancements allow for resending RPCs that failed
3129 Severity : enhancement
3131 Description: Tunables on number of dirty pages in cacche
3132 Details : Allow to set limit on number of dirty pages cached.
3137 Description: client crash on unmount - lock still has references
3138 Details : In some error handling cases it was possible to leak a lock
3139 reference on a client while accessing a file. This was not
3140 harmful to the client during operation, but would cause the
3141 client to crash when the filesystem is unmounted.
3144 Frequency : specific case, rare
3146 Description: ETXTBSY on mds though file not in use by client
3147 Details : ETXTBSY is no longer incorrectly returned when attempting to
3148 chmod or chown a directory that the user previously tried to
3149 execute or a currently-executing binary.
3152 Frequency : extremely rare except on liblustre-based clients
3154 Description: Lustre space not freed when files are deleted
3155 Details : Clean up open-unlinked files after client eviction. Previously
3156 the unlink was skipped and the files remained as orphans.
3161 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3162 Details : In some cases it was possible to send a blocking callback to
3163 a client doing a glimpse, even though that client didn't get
3164 a lock granted. When the glimpse lock is cancelled on the OST
3165 the freed lock is left on the waiting list and corrupted the list.
3168 Frequency : all core dumps
3170 Description: Broke core dumps to lustre
3171 Details : Negative dentry may be unhashed if parent does not have UPDATE
3172 lock, but some callers, e.g. do_coredump, expect dentry to be
3173 hashed after successful create, hash it in ll_create_it.
3175 ------------------------------------------------------------------------------
3177 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3179 * Support for kernels:
3180 2.6.9-42.0.2.EL (RHEL 4)
3181 2.6.5-7.276 (SLES 9)
3182 2.4.21-40.EL (RHEL 3)
3183 2.6.12.6 vanilla (kernel.org)
3187 Frequency : always on RHEL 3
3189 Description: Number of open files grows over time
3190 Details : The number of open files grows over time, whether or not
3191 Lustre is started. This was due to a filp leak introduced
3192 by one of our kernel patches.
3194 ------------------------------------------------------------------------------
3196 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3198 * Support for kernels:
3199 2.6.9-42.EL (RHEL 4)
3200 2.6.5-7.276 (SLES 9)
3201 2.4.21-40.EL (RHEL 3)
3202 2.6.12.6 vanilla (kernel.org)
3207 Bugzilla : 5719, 9635, 9792, 9684
3208 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3209 Details : If a server is under heavy load and cannot reply to new
3210 connection requests before the client resends the (re)connect,
3211 the connection handling code can behave badly if two service
3212 threads are concurrently handing separate (re)connections from
3213 the same client. Add better locking to the connection handling
3214 code, and ensure that only a single connection will be processed
3215 for a given client UUID, even if the lock is dropped.
3217 Severity : enhancement
3219 Description: add TCP zero-copy support to kernel
3220 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3221 sends if the hardware supports scatter-gather and checksumming.
3222 This allows socklnd to do client-write and server-read more
3223 efficiently and reduce CPU utilization from skbuf copying.
3226 Frequency : only if NFS exporting from client
3228 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3229 Details : If setting attributes on a file created under NFS that had
3230 never been opened it would be possible to oops the client
3231 if the file had no objects.
3234 Frequency : always for liblustre
3236 Description: liblustre client does MDS+OSTs setattr RPC for each write
3237 Details : When doing a write from a liblustre client, the client
3238 incorrectly issued an RPC to the MDS and each OST the file was
3239 striped over in order to update the timestamps. When writing
3240 with small chunks and many clients this could overwhelm the MDS
3241 with RPCs. In all cases it would slow down the write because
3242 these RPCs are unnecessary.
3244 Severity : enhancement
3246 Description: allow number of MDS service threads to be changed at module load
3247 Details : It is now possible to change the number of MDS service threads
3248 running. Adding "options mds mds_num_threads={N}" to the MDS's
3249 /etc/modprobe.conf will set the number of threads for the next
3250 time Lustre is restarted (assuming the "mds" module is also
3251 reloaded at that time). The default number of threads will
3252 stay the same, 32 for most systems.
3257 Description: OST crash if filesystem is unformatted or corrupt
3258 Details : If an OST is started on a device that has never been formatted
3259 or if the filesystem is corrupt and cannot even mount then the
3260 error handling cleanup routines would dereference a NULL pointer.
3265 Description: NULL pointer deref in llap_from_page.
3266 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3267 incorrect) mapping if the page was truncated/reclaimed while it was
3268 searched for. Check for this condition and skip such pages when
3269 doing readahead. Introduce extra check to llap_from_page() to
3270 verify page->mapping->host is non-NULL (so page is not anonymous).
3273 Frequency : Sometimes when using sys_sendfile
3275 Description: "page not covered by a lock" warnings from ll_readpage
3276 Details : sendfile called ll_readpage without right page locks present.
3277 Now we introduced ll_file_sendfile that does necessary locking
3278 around call to generic_file_sendfile() much like we do in
3282 Frequency : with certain MDS communication failures at client mount time
3284 Description: NULL pointer deref after failed client mount
3285 Details : a client connection request may delayed by the network layer
3286 and not be sent until after the PTLRPC layer has timed out the
3287 request. If the client fails the mount immediately it will try
3288 to clean up before the network times out the request. Add a
3289 reference from the request import to the obd device and delay
3290 the cleanup until the network drops the request.
3293 Frequency : occasionally during client (re)connect
3295 Description: assertion failure during client (re)connect
3296 Details : processing a client connection request may be delayed by the
3297 client or server longer than the client connect timeout. This
3298 causes the client to resend the connection request. If the
3299 original connection request is replied in this interval, the
3300 client may trip an assertion failure in ptlrpc_connect_interpret()
3301 which thought it would be the only running connect process.
3304 Frequency : only with obd_echo servers and clients that are rebooted
3306 Description: kernel BUG accessing uninitialized data structure
3307 Details : When running an obd_echo server it did not start the ping_evictor
3308 thread, and when a client was evicted an uninitialized data
3309 structure was accessed. Start the ping_evictor in the RPC
3310 service startup instead of the OBD startup.
3312 Severity : enhancement
3313 Bugzilla : 10193 (patchless)
3314 Description: Remove dependency on various unexported kernel interfaces.
3315 Details : No longer need reparent_to_init, exit_mm, exit_files,
3316 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3319 Frequency : rare (only users of deprecated and unsupported LDAP config)
3321 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3322 Details : LDAP apparently contains 'lustreName' attributes instead of
3323 'name'. A simple remapping of the name is sufficient.
3326 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3328 Description: waiting_locks_callback trips kernel BUG if client is evicted
3329 Details : Running with the dump_on_timeout debug flag turned on makes
3330 it possible that the waiting_locks_callback() can try to dump
3331 the Lustre kernel debug logs from an interrupt handler. Defer
3332 this log dumping to the expired_lock_main() thread.
3334 Severity : enhancement
3336 Description: Support NFS exporting on 2.6 kernels.
3337 Details : Implement non-rawops metadata methods for NFS server to use without
3338 changing NFS server code.
3341 Frequency : very rare (synthetic metadata workload only)
3343 Description: two racing renames might cause an MDS thread to deadlock
3344 Details : Running the "racer" program may cause one MDS thread to rename
3345 a file from being the source of a rename to being the target of
3346 a rename at exactly the same time that another thread is doing
3347 so, and the second thread has already enqueued these locks after
3348 doing a lookup of the target and is trying to relock them in
3349 order. Ensure that we don't try to re-lock the same resource.
3352 Frequency : only very large systems with liblustre clients
3354 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3355 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3356 all OSTs in parallel. This allows the eviction of stale liblustre
3357 clients to proceed much faster than if they were done in series,
3358 and also offers similar improvements for other set_info RPCs.
3363 Description: excessive CPU usage during initial read phase on client
3364 Details : During the initial read phase on a client, it would agressively
3365 retry readahead on the file, consuming too much CPU and impacting
3366 performance (since 1.4.5.8). Improve the readahead algorithm
3367 to avoid this, and also improve some other common cases (read
3368 of small files in particular, where "small" is files smaller than
3369 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3374 Description: MDS crash when receiving packet with unknown intent.
3375 Details : Do not LBUG in unknown intent case, just return -EFAULT
3377 Severity : enhancement
3378 Bugzilla : 9293, 9385
3379 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3380 Details : Do not serialize getattr (non-intent version) and statfs.
3383 Frequency : occasional, when OST network is overloaded/intermittent
3385 Description: client evicted by OST after bulk IO timeout
3386 Details : If a client sends a bulk IO request (read or write) the OST
3387 may evict the client if it is unresposive to its data GET/PUT
3388 request. This is incorrect if the network is overloaded (takes
3389 too long to transfer the RPC data) or dropped the OST GET/PUT
3390 request. There is no need to evict the client at all, since
3391 the pinger and/or lock callbacks will handle this, and the
3392 client can restart the bulk request.
3395 Frequency : Always when mmapping file with no objects
3397 Description: client crashes when mmapping file with no objects
3398 Details : Check that we actually have objects in a file before doing any
3399 operations on objects in ll_vm_open, ll_vm_close and
3405 Description: Request leak when working with deleted CWD
3406 Details : Introduce advanced request refcount tracking for requests
3407 referenced from lustre intent.
3409 Severity : Enhancement
3411 Description: Cache open file handles on client.
3412 Details : MDS now will return special lock along with openhandle, if
3413 requested and client is allowed to hold openhandle, even if unused,
3414 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3415 closing files for every read/write openration.
3417 Severity : Enhancement
3419 Description: Cache open negative dentries on client when possible.
3420 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3421 negative dentries on lock revocation.
3426 Description: Remounting a client read-only wasn't possible with a zconf mount
3427 Details : It wasn't possible to remount a client read-only with llmount.
3429 Severity : enhancement
3430 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3431 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3432 MPICH is included to add Lustre-specific ADIO interfaces.
3433 This is based closely on the UFS ADIO layer and only differs
3434 in file creation, in order to allow the OST striping to be set.
3435 This is user-contributed code and not supported by CFS.
3440 Description: extended inode attributes (immutable, append-only) work improperly
3441 when 2.4 and 2.6 kernels are used on client/server or vice versa
3442 Details : Introduce kernel-independent values for these flags.
3444 Severity : enhancement
3447 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3448 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3449 granularity. Now, Fractional MB granularity is be supported,
3450 this is very useful for low memory system.
3452 Severity : enhancement
3454 Description: Getattr by fid
3455 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3456 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3457 have locks and avoids getattr rpc requests for it on every stat.
3460 Frequency : Always, for filesystems larger than 2TB
3462 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3463 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3464 filesystems larger than 2TB to be mounted with mballoc enabled.
3467 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3469 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3470 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3471 filesystems larger than 2TB is mounted then the kernel will
3472 silently corrupt the start of the filesystem. CONFIG_LBD is
3473 enabled for all CFS-supported kernels, but the possibility of
3474 this happening with a modified kernel config exists.
3476 Severity : enhancement
3478 Description: add client O_DIRECT support for 2.6 kernels
3479 Details : It is now possible to do O_DIRECT reads and writes to files
3480 in the Lustre client mountpoint on 2.6 kernel clients.
3482 Severity : enhancement
3484 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3485 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3486 parallel, not waiting for response from every OST before sending
3487 a rpc to the next OST.
3492 Description: setattr vs write race when updating file timestamps
3493 Details : Client processes that update a file timestamp into the past
3494 right after writing to the file (e.g. tar) it is possible that
3495 the updated file modification time can be reset to the current
3496 time due to a race between processing the setattr and write RPC.
3498 Severity : enhancement
3500 Description: Bring 'lfs find' closer in line with regular Linux find.
3501 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3502 -print0 options and obtains all the needed info through the lustre
3505 Severity : enhancement
3507 Description: support up to 1024 configured devices on one node
3508 Details : change obd_dev array from statically allocated to dynamically
3509 allocated structs as they are first used to reduce memory usage
3514 Description: Flush dirty partially truncated pages during truncate
3515 Details : Immediatelly flush partially truncated pages in filter_setattr,
3516 this way we completely avoid having any pages in page cache on OST
3517 and can retire ugly workarounds during writes to flush such pages.
3522 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3523 Details : protect inode from truncation within vfs_unlink() context
3524 just take a reference before calling vfs_unlink() and release it
3525 when parent's i_sem is free.
3530 Description: last_id value checked outside lock on OST caused LASSERT failure
3531 Details : If there were multiple MDS->OST object precreate requests in
3532 flight, it was possible that the OST's last object id was checked
3533 outside a lock and incorrectly tripped an assertion. Move checks
3534 inside locks, and discard old precreate requests.
3537 Frequency : always, if extents are used on OSTs
3539 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3540 Details : OSTs using the extents format would not zero the high 16 bits of
3541 the index physical block number. This is not a problem for any
3542 OST filesystems smaller than 16TB, and no kernels support ext3
3543 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3544 new/modified files) and can be fixed for existing filesystems
3545 with e2fsprogs-1.39-cfs1.
3550 Description: import connection selection may be incorrect if timer wraps
3551 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3552 import connections to be ignored if the 32-bit jiffies counter
3553 wraps. Use a 64-bit jiffies counter.
3556 Frequency : very large clusters immediately after boot
3558 Description: LNET request buffers exhausted under heavy short-term load
3559 Details : If a large number of client requests are generated on a service
3560 that has previously never seen so many requests it is possible
3561 that the request buffer growth cannot keep up with the spike in
3562 demand. Instead of dropping incoming requests, they are held in
3563 the LND until the RPC service can accept more requests.
3566 Frequency : Sometimes during replay
3568 Description: Assertion failure in ll_local_open after replay.
3569 Details : If replay happened on an open request reply before we were able
3570 to set replay handler, reply will become not swabbed tripping the
3571 assertion in ll_local_open. Now we set the handler right after
3572 recognising of open request
3575 Frequency : very rare
3577 Description: kernel reports "badness in vsnprintf"
3578 Details : Reading from the "recovery_status" /proc file in small chunks
3579 may cause a negative length in lprocfs_obd_rd_recovery_status()
3580 call to vsnprintf() (which is otherwise harmless). Exit early
3581 if there is no more space in the output buffer.
3583 Severity : enhancement
3585 Description: clear OBD RPC statistics by writing to them
3586 Details : It is now possible to clear the OBD RPC statistics by writing
3587 to the "stats" file.
3592 Description: Client mtime is not the same on different clients after utimes
3593 Details : In some cases, the client was using the utimes() syscall on
3594 a file cached on another node. The clients now validate the
3595 ctime from the MDS + OSTs to determine which one is right.
3600 Description: Inability to activate failout mode
3601 Details : lconf script incorrectly assumed that in pythong string's numeric
3602 value is used in comparisons.
3605 Frequency : always with multiple stripes per file
3607 Description: Inefficient object allocation for mutli-stripe files
3608 Details : When selecting which OSTs to stripe files over, for files with
3609 a stripe count that divides evenly into the number of OSTs,
3610 the MDS is always picking the same starting OST for each file.
3611 Return the OST selection heuristic to the original design.
3616 Description: mount failures may take full timeout to return an error
3617 Details : Under some heavy load conditions it is possible that a
3618 failed mount can wait for the full obd_timeout interval,
3619 possibly several minutes, before reporting an error.
3620 Instead return an error as soon as the status is known.
3622 Frequency : quota enabled and large files being deleted
3624 Description: releasing more than 4GB of quota at once hangs OST
3625 Details : If a user deletes more than 4GB of files on a single OST it
3626 will cause the OST to spin in an infinite loop. Release
3627 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3632 Description: statfs data retrieved from /proc may be stale or zero
3633 Details : When reading per-device statfs data from /proc, in the
3634 {kbytes,files}_{total,free,avail} files, it may appear
3635 as zero or be out of date.
3638 Frequency : systems with MD RAID1 external journal devices
3640 Description: lconf's call to blkid is confused by RAID1 journal devices
3641 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3642 of returning all block devices that match the journal UUID.
3645 Frequency : always, for aggregate stripe size over 4GB
3647 Description: assertion fails when trying to use 4GB stripe size
3648 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3649 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3652 Frequency : always on ppc64
3654 Description: the first write on an ext3 filesystem with mballoc got stuck
3655 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3656 perform endianness conversion.
3658 ------------------------------------------------------------------------------
3660 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3662 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3663 INTEROPERATE with older versions automatically. Please read the
3664 user documentation before upgrading any part of a live system.
3665 * WARNING: Lustre networking configuration changes are required with
3666 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3669 * Support for kernels:
3670 2.6.9-22.0.2.EL (RHEL 4)
3671 2.6.5-7.244 (SLES 9)
3672 2.6.12.6 vanilla (kernel.org)
3675 Severity : enhancement
3676 Bugzilla : 7981/8208
3677 Description: Introduced Lustre Networking (LNET)
3678 Details : LNET is new networking infrastructure for Lustre, it includes
3679 a reorganized network configuration mode (see the user
3680 documentation for full details) as well as support for routing
3681 between different network fabrics. Lustre Networking Devices
3682 (LNDs) for the supported network fabrics have also been
3683 created for this new infrastructure.
3685 Severity : enhancement
3686 Description: Introduced Access control lists
3687 Details : clients can set ACLs on files and directories in order to have
3688 more fine-grained permissions than the standard Unix UGO+RWX.
3689 The MDS must be started with the "-o acl" mount option.
3691 Severity : enhancement
3692 Description: Introduced filesystem quotas
3693 Details : Administrators may now establish per-user quotas on the
3696 Severity : enhancement
3698 Description: Configuration change for the XT3
3699 The PTLLND is now used to run Lustre over Portals on the XT3
3700 The configure option(s) --with-cray-portals are no longer used.
3701 Rather --with-portals=<path-to-portals-includes> is used to
3702 enable building on the XT3. In addition to enable XT3 specific
3703 features the option --enable-cray-xt3 must be used.
3708 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3709 Details : A race between allocating a new llcd and re-getting the llcd_lock
3710 allowed another thread to grab newly-allocated llcd.
3712 Severity : enhancement
3714 Description: 2.6 OST async journal commit and locking fix to improve performance
3715 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3716 2.6 kernels are now async as they already were in 2.4 kernels so
3717 that they can commit concurrently with the network bulk transfer.
3718 For block-allocated files the filter allocation semaphore is held
3719 to avoid filesystem fragmentation during allocation. BKL lock
3720 removed for 2.6 xattr operations where it is no longer needed.
3725 Description: lconf incorrectly determined whether two IP networks could talk
3726 Details : In some more complicated routing and multiple-network
3727 configurations, lconf will avoid trying to make a network
3728 connection to a disjoint part of the IP space. It was doing the
3729 math incorrectly for one set of cases.
3734 Description: Fix for potential infinite loop processing records in an llog.
3735 Details : If an llog record is corrupted/zeroed, it is possible to loop
3736 forever in llog_process(). Validate the llog record length
3737 and skip the remainder of the block on error.
3740 Frequency : occasional (liblustre only)
3742 Description: liblustre could not open files whose last component is a symlink
3743 Details : sysio_path_walk() would incorrectly pass the open intent to
3744 intermediate path components.
3747 Frequency : rare (liblustre only with non-standard tuning)
3748 Bugzilla : 7201 (7350)
3749 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3750 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3751 after it has been released, only do this for referenced locks
3753 Severity : enhancement
3755 Description: specifying an (invalid) directory default stripe_size of -1
3756 would reset the directory default striping
3757 Details : stripe_size -1 was used internally to signal directory stripe
3758 removal, now use "all default" to signal dir stripe removal
3759 as a directory striping of "all default" is not useful
3762 Frequency : common for large clusters running liblustre clients
3764 Description: doing an ls when liblustre clients are running is slow
3765 Details : sending a glimpse AST to a liblustre client waits for every AST
3766 to time out, as liblustre clients will not respond. Since they
3767 cannot cache data we refresh the OST lock LVB from disk instead.
3769 Severity : enhancement
3771 Description: doing an ls at the same time as file IO can be slow
3772 Details : enqueue and other "small" requests can be blocked behind many
3773 large IO requests. Create a new OST IO portal for non-IO
3774 requests so they can be processed faster.
3777 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3779 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3780 created file with mode 000
3781 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3782 allow an ftruncate() as a non-root user to a file with mode 000.
3783 Lustre now respects this flag to disable mode checks when
3784 truncating a file owned by the user
3787 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3790 Description: Revoking locks from clients that went dead or catatonic might take
3792 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3793 cancellation of such locks instant on servers without waiting for
3794 any reply from clients. Clients drops these locks when cancel
3795 notification from server is received without replying.
3798 Frequency : liblustre-only, when liblustre client dies or becomes busy
3800 Description: Doing ls on Linux clients can take a long time with active
3802 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3803 granting such locks to it in the first place if possible. Locks
3804 are taken by proxy on the OST during the read or write and
3805 dropped immediately afterward. Add connect flags handling, do
3806 not grant locks to liblustre clients for glimpse ASTs.
3808 Severity : enhancement
3810 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3811 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3812 sequential reads and behaves badly with applications doing random
3813 reads. Improve it by reading ahead at least read region, and
3814 avoiding excessive large RPC for small reads.
3816 Severity : enhancement
3818 Description: Creating more than 1000 files for a single job may cause a load
3819 imbalance on the OSTs if there are also a large number of OSTs.
3820 Details : qos_prep_create() uses an OST index reseed value that is an
3821 even multiple of the number of available OSTs so that if the
3822 reseed happens in the middle of the object allocation it will
3823 still utilize the OSTs as uniformly as possible.
3828 Description: OST or MDS may oops in ping_evictor_main()
3829 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3830 but doesn't restart at beginning of obd_exports_timed list
3833 Severity : enhancement
3835 Description: improve by-nid export eviction on the MDS and OST
3836 Details : allow multiple exports with the same NID to be evicted at one
3837 time without re-searching the exports list.
3840 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3842 Description: MDS may oops in groups_free()
3843 Details : in rare race conditions a newly allocated group_info struct is
3844 freed again, and this can be NULL. The 2.4 compatibility code
3845 for groups_free() checked for a NULL pointer, but 2.6 did not.
3848 Frequency : common for liblustre clients doing little filesystem IO
3849 Bugzilla : 9352, 7313
3850 Description: server may evict liblustre clients accessing contended locks
3851 Details : if a client is granted a lock or receives a completion AST
3852 with a blocking AST already set it would not reply to the AST
3853 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3856 Frequency : lfs setstripe, only systems with more than 160 OSTs
3858 Description: unable to set striping with a starting offset beyond OST 160
3859 Details : llapi_create_file() incorrectly limited the starting stripe
3860 index to the maximum single-file stripe count.
3863 Frequency : LDAP users only
3865 Description: lconf did not handle in-kernel recovery with LDAP properly
3866 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3868 Severity : enhancement
3870 Description: bind OST threads to NUMA nodes to improve performance
3871 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3872 node and do their allocations there to localize memory access
3874 Severity : enhancement
3876 Description: llmount can determine client NID directly from Myrinet (GM)
3877 Details : the client NID code from gmnalnid was moved directly into
3878 llmount, removing the need to use this or specifying the
3879 client NID explicitly when mounting GM clients with zeroconf
3882 Frequency : if client is started with down MDS
3884 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3885 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3886 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3887 ptlrpc_import_delayed_req() because we didn't check if the
3888 request was interrupted, and we also didn't break out of the
3889 event loop if there was no timeout
3894 Description: data loss during non-page-aligned writes to a single file from
3895 both multiple nodes and multiple threads on one node at same time
3896 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3897 inconsistency led to false short-reads, that were cached and later
3898 used by ->prepare_write() to fill in partially written page,
3899 leading to data loss.
3902 Frequency : always, if lconf --abort_recovery used
3904 Description: lconf --abort_recovery fails with 'Operation not supported'
3905 Details : lconf was attempting to abort recovery on the MDT device and not
3908 Severity : enhancement
3910 Description: remove cleanup logs
3911 Details : replace lconf-generated cleanup logs with lustre internal
3912 cleanup routines. Eliminates the need for client-cleanup and
3915 Severity : enhancement
3917 Description: add support for EAs (user and system) on lustre filesystems
3918 Details : it is now possible to store extended attributes in the Lustre
3919 client filesystem, and with the user_xattr mount option it
3920 is possible to allow users to store EAs on their files also
3922 Severity : enhancement
3924 Description: Add possibility (config option) to show minimal available OST free
3926 Details : When compiled with --enable-mindf configure option, statfs(2)
3927 (and so, df) will return least minimal free space available from
3928 all OSTs as amount of free space on FS, instead of summary of
3929 free spaces of all OSTs.
3931 Severity : enhancement
3933 Description: do not expand extent locks acquired on OST-side
3934 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3935 by server: they are not cached anyway.
3938 Frequency : when mmap is used/binaries executed from Lustre
3940 Description: Unmmap pages before throwing them away from read cache.
3941 Details : llap_shrink cache now attempts to unmap pages before discarding
3942 them (if unmapping failed - do not discard). SLES9 kernel has
3943 extra checks that trigger if this unmapping is not done first.
3948 Description: lconf didn't resolve symlinks before checking to see whether a
3949 given mountpoint was already in use
3952 Frequency : when migrating failover services
3953 Bugzilla : 6395, 9514
3954 Description: When migrating a subset of services from a node (e.g. failback
3955 from a failover service node) the remaining services would
3956 time out and evict clients.
3957 Details : lconf --force (implied by --failover) sets the global obd_timeout
3958 to 5 seconds in order to quickly disconnect, but this caused
3959 other RPCs to time out too quickly. Do not change the global
3960 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3962 Severity : enhancement
3963 Frequency : if MDS is started with down OST
3964 Bugzilla : 9439,5706
3965 Description: Allow startup/shutdown of an MDS without depending on the
3966 availability of the OSTs.
3967 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3968 Add appropriate locking and lov-osc refcounts for safe
3969 cleaning. Add osc abort_inflight calls in case the
3970 synchronize never started.
3973 Frequency : occasional (Cray XT3 only)
3975 Description: root not authorized to access files in CRAY_PORTALS environment
3976 Details : The client process capabilities were not honoured on the MDS in
3977 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3978 been accessed by an authorized user then root was able to access
3979 the file on the local client also. The root user capabilities
3980 are now allowed on the MDS, as this environment has secure UID.
3983 Frequency : occasional
3985 Description: ldiskfs "too long searching" message happens too often
3986 Details : A debugging message (otherwise harmless) prints too often on
3987 the OST console. This has been reduced to only happen when
3988 there are fragmentation problems on the filesystem.
3993 Description: Division by zero in statfs when all OSCs are inactive
3994 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3995 lov_statfs divides by value returned by lov_get_stripecnt().
3999 Bugzilla : 9489, 3273
4000 Description: First write from each client to each OST was only 4kB in size,
4001 to initialize client writeback cache, which caused sub-optimal
4002 RPCs and poor layout on disk for the first writen file.
4003 Details : Clients now request an initial cache grant at (re)connect time
4004 and so that they can start streaming writes to the cache right
4005 away and always do full-sized RPCs if there is enough data.
4006 If the OST is rebooted the client also re-establishes its grant
4007 so that client cached writes will be honoured under the grant.
4012 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4013 Details : Now I/O RPCs go to different portal number and (presumably) fast
4014 lock requests (and glimses) and other RPCs get their own service
4015 threads pool that should be able to service those RPCs
4018 Severity : enhancement
4020 Description: Ability to exchange lustre version between client and servers and
4021 issue warnings at client side if client is too old. Also for
4022 liblustre clients there is ability to refuse connection of too old
4024 Details : New 'version' field is added to connect data structure that is
4025 filled with version info. That info is later checked by server and
4029 Frequency : rare, liblustre only.
4030 Bugzilla : 9296, 9581
4031 Description: Two simultaneous writes from liblustre at offset within same page
4032 might proceed at the same time overwriting eachother with stale
4034 Details : I/O lock withing llu_file_prwv was released too early, before data
4035 actually was hitting the wire. Extended lock-holding time until
4036 server acknowledges receiving data.
4039 Frequency : extremely rare. Never observed in practice.
4041 Description: avoid generating lustre_handle cookie of 0.
4042 Details : class_handle_hash() generates handle cookies by incrementing
4043 global counter, and can hit 0 occasionaly (this is unlikely, but
4044 not impossible, because initial value of cookie counter is
4045 selected randonly). Value of 0 is used as a sentinel meaning
4046 "unassigned handle" --- avoid it. Also coalesce two critical
4047 sections in this function into one.
4049 Severity : enhancement
4051 Description: allow liblustre clients to delegate truncate locking to OST
4052 Details : To avoid overhead of locking, liblustre client instructs OST to
4053 take extent lock in ost_punch() on client's behalf. New connection
4054 flag is added to handle backward compatibility.
4056 Severity : enhancement
4057 Bugzilla : 4928, 7341, 9758
4058 Description: allow number of OST service threads to be specified
4059 Details : a module parameter allows the number of OST service threads
4060 to be specified via "options ost ost_num_threads={N}" in the
4061 OSS's /etc/modules.conf or /etc/modprobe.conf.
4065 Bugzilla : 6146, 9635, 9895
4066 Description: servers crash with bad pointer in target_handle_connect()
4067 Details : In rare cases when a client is reconnecting it was possible that
4068 the connection request was the last reference for that export.
4069 We would temporarily drop the export reference and get a new
4070 one, but this may have been the last reference and the export
4071 was just destroyed. Get new reference before dropping old one.
4073 Severity : enhancement
4074 Frequency : if client is started with failover MDS
4076 Description: Allow multiple MDS hostnames in the mount command
4077 Details : Try to read the configuration from all specified MDS
4078 hostnames during a client mount in case the "primary"
4081 Severity : enhancement
4083 Description: Stop sending data to evicted clients as soon as possible.
4084 Details : Check if the client we are about to send or are sending data to
4085 was evicted already. (Check is done every second of waiting,
4086 for which l_wait_event interface was extended to allow checking
4087 of exit condition at specified intervals).
4090 Frequency : rare, normally only when NFS exporting is done from client
4092 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4094 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4097 Frequency : always, if srand() is called before liblustre initialization
4099 Description: Liblustre uses system PRNG disturbing its usage by user application
4100 Details : Introduce internal to lustre fast and high-quality PRNG for
4101 lustre usage and make liblustre and some other places in generic
4102 lustre code to use it.
4104 Severity : enhancement
4105 Bugzilla : 9477, 9557, 9870
4106 Description: Verify that the MDS configuration logs are updated when xml is
4107 Details : Check if the .xml configuration logs are newer than the config
4108 logs stored on the MDS and report an error if this is the case.
4109 Request --write-conf, or allow starting with --old_conf.
4111 Severity : enhancement
4113 Description: Handle symlinks in the path when checking if Lustre is mounted.
4114 Details : Resolve intermediate symlinks when checking if a client has
4115 mounted a filesystem to avoid duplicate client mounts.
4120 Description: lconf can hit an error exception but still return success.
4121 Details : The lconf command catches the Command error exception at the top
4122 level script context and will exit with the associated exit
4123 status, but doesn't ensure that this exit status is non-zero.
4128 Description: failure of ptlrpc thread startup can cause oops
4129 Details : Starting a ptlrpc service thread can fail if there are a large
4130 number of threads or the server memory is very fragmented.
4131 Handle this without oopsing.
4134 Frequency : always, only if liblustre and non-default acceptor port was used
4136 Description: liblustre cannot connect to servers with non-default acceptor port
4137 Details : tcpnal_set_default_params() was not called and was therefore
4138 ignoring the environment varaible TCPNAL_PORT, as well as other
4139 TCPNAL_ environment variables
4144 Description: two objects could be created on the same OST for a single file
4145 Details : If an OST is down, in some cases it was possible to create two
4146 objects on a single OST for a single file. No problems other
4147 than potential performance impact and spurious error messages.
4151 Bugzilla : 5681, 9562
4152 Description: Client may oops in ll_unhash_aliases
4153 Details : Client dcache may become inconsistent in race condition.
4154 In some cases "getcwd" can fail if the current directory is
4160 Description: Inode refcounting problems in NFS export code
4161 Details : link_raw functions used to call d_instantiate without obtaining
4162 extra inode reference first.
4166 Bugzilla : 9942, 9903
4167 Description: Referencing freed requests leading to crash, memleaks with NFS.
4168 Details : We used to require that call to ll_revalidate_it was always
4169 followed by ll_lookup_it. Also with revalidate_special() it is
4170 possible to call ll_revalidate_it() twice for the same dentry
4171 even if first occurence returned success. This fix changes semantic
4172 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4173 reference on a request referred from the intent.
4174 ll_intent_release() then releases such a request.
4177 Frequency : rare, normally benchmark loads only
4179 Description: unlinked inodes were kept in memory on the client
4180 Details : If a client is repeatedly creating and unlinking files it
4181 can accumulate a lot of stale inodes in the inode slab cache.
4182 If there is no other client load running this can cause the
4183 client node to run out of memory. Instead flush old inodes
4184 from client cache that have the same inode number as a new inode.
4187 Frequency : SLES9 2.6.5 kernel and long filenames only
4188 Bugzilla : 9969, 10379
4189 Description: utime reports stale NFS file handle
4190 Details : SLES9 uses out-of-dentry names in some cases, which confused
4191 the lustre dentry revalidation. Change it to always use the
4195 Frequency : rare, unless heavy write-truncate concurrency is continuous
4196 Bugzilla : 4180, 6984, 7171, 9963, 9331
4197 Description: OST becomes very slow and/or deadlocked during object unlink
4198 Details : filter_destroy() was holding onto the parent directory lock
4199 while truncating+unlinking objects. For very large objects this
4200 may block other threads for a long time and slow overall OST
4201 responsiveness. It may also be possible to get a lock ordering
4202 deadlock in this case, or run out of journal credits because of
4203 the combined truncate+unlink. Solution is to do object truncate
4204 first in one transaction without parent lock, and then do the
4205 final unlink in a new transaction with the parent lock. This
4206 reduces the lock hold time dramatically.
4209 Frequency : rare, 2.4 kernels only
4211 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4212 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4213 before unmounting their filesystem in order to avoid deadlock.
4214 The kernel_locked() function in 2.4 kernels only checks whether
4215 the kernel lock is held, not whether it is this process that is
4216 holding it as 2.6 kernels do.
4221 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4222 Details : The client ptlrpc code may be trying to reconnect to a down
4223 server before a previous connection attempt has timed out.
4224 Increase the reconnect interval to be longer than the connection
4225 timeout interval to avoid sending duplicate connections to
4229 Frequency : echo_client brw_test command
4231 Description: fix echo_client to work with OST preallocated code
4232 Details : OST preallocation code (5137) didn't take echo_client IO path
4233 into account: echo_client calls filter methods outside of any
4234 OST thread and, hence, there is no per-thread preallocated
4235 pages and buffers to use. Solution: hijack pga pages for IO. As
4236 a byproduct, this avoids unnecessary data copying.
4240 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4241 Description: Client can oops in mdc_commit_close() after open replay
4242 Details : It was possible for the MDS to return an open request with no
4243 transaction number in mds_finish_transno() if the client was
4244 evicted, but without actually returning an error. Clients
4245 would later try to replay that open and may trip an assertion
4246 Simplify the client close codepath, and always return an error
4247 from the MDS in case the open is not successful.
4250 Frequency : rare, 2.6 OSTs only
4252 Description: OST may deadlock under high load on fragmented files
4253 Details : If there was a heavy load and highly-fragmented OST filesystems
4254 it was possible to have all the OST threads deadlock waiting on
4255 allocation of biovecs, because the biovecs were not released
4256 until the entire RPC IO was completed. Instead, release biovecs
4257 as soon as they are complete to ensure forward IO progress.
4259 Severity : enhancement
4261 Description: Support for specifying external journal device at mount
4262 Details : If an OST or MDS device is formatted with an external journal
4263 device, this device major/minor is stored in the ext3 superblock
4264 and may not be valid for failover. Allow detecting and
4265 specifying the external journal at mount time.
4270 Description: Mounting an MDS with pending unlinked files may cause oops
4271 Details : target_finish_recovery() calls mds_postrecov() which returned
4272 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4273 considers this an error and immediately begins cleaning up the
4274 lov, just after starting the mds_lov process
4276 Severity : enhancement
4278 Description: Implement 'lfs df' to report actual free space on per-OST basis
4279 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4280 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4281 usage of objects; '-h' to report in human readable format.
4283 ------------------------------------------------------------------------------
4285 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4292 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4293 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4294 at startup during orphan unlinking. Instead, skip buddy update
4295 before setup, as it will be regenerated after recovery is complete.
4300 Description: If an OST is inactive, its locks might reference stale inodes.
4301 Details : lov_change_cbdata() must iterate over all namespaces, even if
4302 they are inactive to clear inode references from the lock.
4304 Severity : enhancement
4305 Frequency : occasional, if non-standard max_dirty_mb used
4307 Description: Client will block write RPCs if not enough grant
4308 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4309 then the client will block writes while waiting for another RPC
4310 to complete instead of consuming its dirty limit. With change
4311 we get improved performance when max_dirty_mb is small.
4313 Severity : enhancement
4314 Bugzilla : 3389, 6253
4315 Description: Add support for supplementary groups on the MDS.
4316 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4317 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4318 lookups for user supplementary groups into a cache.
4322 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4323 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4324 for liblustre O_EXCL is enforced by MDS.
4328 Description: readdir never returns NULL in liblustre.
4329 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4330 dentry in struct dirent.
4334 Frequency : liblustre only, depends on application IO pattern
4335 Description: liblustre clients evicted if not contacting servers
4336 Details : Don't put liblustre clients into the ping_evictor list, so
4337 they will not be evicted by the pinger ever.
4339 Severity : enhancement
4341 Description: Add ability to evict clients by NID from MDS.
4342 Details : By echoing "nid:$NID" string into
4343 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4344 $NID would be instantly evicted from this MDS and from all active
4345 OSTs connected to it.
4349 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4350 Details : lookup_it_finish() used to query file size from OSTs that was not
4355 Description: service threads change working directory to that of init
4356 Details : Starting lustre service threads may pin the working directory
4357 of the parent thread, making that filesystem busy. Threads
4358 now change to the working directory of init to avoid this.
4362 Frequency : during shutdown only
4363 Description: shutdown with a failed MDS or OST can cause unmount to hang
4364 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4369 Frequency : 2.6 only
4370 Description: chmod/chown may include an extra supplementary group
4371 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4372 supplementary group and if none is specified this is used.
4375 Bugzilla : 5479 (6816)
4377 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4378 Details : If lookup is in progress on a file that is unlinked we might try
4379 to revalidate the inode and fail in revalidate after lookup is
4380 complete and ll_file_open() enqueues the open again but
4381 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4384 Frequency : always, if lconf --abort_recovery used
4386 Description: lconf --abort_recovery fails with 'Operation not supported'
4387 Details : lconf was attempting to abort recovery on the MDT device and not
4390 ------------------------------------------------------------------------------
4392 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4397 Frequency : rare (only unsupported configurations with a node running as an
4399 Bugzilla : 6514, 5137
4400 Description: Mounting a Lustre file system on a node running as an OST could
4402 Details : OSTs now preallocates memory needed to write out data at
4403 startup, instead of when needed, to avoid having to
4404 allocate memory in possibly low memory situations.
4405 Specifically, if the file system is mounted on on OST,
4406 memory pressure could force it to try to write out data,
4407 which it needed to allocate memory to do. Due to the low
4408 memory, it would be unable to do so and the node would
4409 become unresponsive.
4411 Severity : enhancement
4413 Description: Addition of lconf --service command line option
4414 Details : lconf now accepts a '--service <arg>' option, which is
4415 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4417 Severity : enhancement
4419 Description: Failover mode is now the default for OSTs.
4420 Details : By default, OSTs will now run in failover mode. To return to
4421 the old behaviour, add '--failout' to the lmc line for OSTs.
4423 Severity : enhancement
4425 Description: Health checks are now provided for MDS and OSTs
4426 Details : Additional detailed health check information on MSD and OSTs
4427 is now provided through the procfs health_check value.
4430 Frequency : occasional, depends on IO load
4432 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4433 after numerous create/delete cycles
4434 Details : The ext3 inode allocation policy would not allocate new inodes
4435 very well on the OSTs because there are no new directories
4436 being created. Instead we look for groups with free space if
4437 the parent directories are nearly full.
4442 Description: Network or server problems during mount may cause partially
4443 mounted clients instead of returning an error.
4444 Details : The config llog parsing code may overwrite the error return
4445 code during mount error handling, returning success instead
4451 Description: MDS can fail to allocate large reply buffers
4452 Details : After long uptimes the MDS can fail to allocate large reply
4453 buffers (e.g. zconf client mount config records) due to memory
4454 fragmentation or consumption by the buffer cache. Preallocate
4455 some large reply buffers so that these replies can be sent even
4456 under memory pressure.
4460 Frequency : rare (liblustre)
4461 Description: fsx running with liblustre complained that using truncate() to
4462 extend the file doesn't work. This patch corrects that issue.
4463 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4464 ATTR_SIZE and lsm use in llu_setattr_raw.
4468 Frequency : rare, only 2.6 kernels
4469 Description: Unusual file access patterns on the MDS may result in inode
4470 data being lost in very rare circumstances.
4471 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4472 code in the 2.6 kernel caused the inode and/or EA data not to
4473 be read from disk, causing single-file corruption.
4477 Frequency : rare, only 2.6 filesystems using extents
4478 Description: Heavy concurrent write and delete load may cause data corruption.
4479 Details : It was possible under high-load situations to have an extent
4480 metadata block in the block device cache from a just-unlinked
4481 file overwrite a newly-allocated data block. We now unmap any
4482 metadata buffers that alias just-allocated data blocks.
4486 Frequency : filesystems with default stripe_count larger than 77
4487 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4488 Details : lconf specifies an inode size of 4096 bytes when the default
4489 stripe_count is larger than 77. This conflicts with the default
4490 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4491 this case to avoid pinning too much memory for large EAs.
4493 ------------------------------------------------------------------------------
4495 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4500 Frequency : rare (extremely heavy IO load with hundreds of clients)
4502 Description: Client is evicted, gets IO error writing to file
4503 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4504 caused clients to be evicted for AST timeouts. The fixes in
4505 bug 5192 mean we no longer need to have such short AST timeouts
4506 so ldlm_timeout has been increased.
4509 Frequency : occasional during --force or --failover shutdown under load
4510 Bugzilla : 5949, 4834
4511 Description: Server oops/LBUG if stopped with --force or --failover under load
4512 Details : a collection of import/export refcount and cleanup ordering
4513 issues fixed for safer force cleanup
4516 Frequency : only filesystems larger than 120 OSTs
4517 Bugzilla : 5990, 6223
4518 Description: lfs getstripe would oops on a very large filesystem
4519 Details : lov_getconfig used kfree on vmalloc'd memory
4522 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4523 Bugzilla : 6242, 6243
4524 Description: reading from files that had been truncated to a non-zero size
4525 but never opened returned no data
4526 Details : ll_file_read() reads zeros from no-object files to EOF
4531 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4532 Details : There's little harm in aborting MDS/OSS recovery and letting it
4533 try again, so I removed the LASSERT and return an error instead.
4535 Severity : enhancement
4537 Description: New debugging infrastructure for tracking down data corruption
4538 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4539 (b) cover more of the client-side code path, and (c) try to narrow
4540 down where problems occurred
4544 Bugzilla : 3819, 4364, 4397, 6313
4545 Description: Racing close and eviction MDS could cause assertion in mds_close
4546 Details : It was possible to get multiple mfd references during close and
4547 client eviction, leading to one thread referencing a freed mfd.
4549 Severity: : enhancement
4550 Bugzilla : 3262, 6359
4551 Description: Attempts to reconnect to servers are now more aggressive.
4552 Details : This builds on the enhanced upcall-less recovery that was added
4553 in 1.4.2. When trying to reconnect to servers, clients will
4554 now try each server in the failover group every 10 seconds. By
4555 default, clients would previously try one server every 25 seconds.
4560 Description: After recovery, certain operations trigger a failed
4561 assertion on a client.
4562 Details : Failing over an mds, using lconf -d --failover, while a
4563 client was doing a readdir() call would cause the client to
4564 LBUG after recovery completed and the readdir() was resent.
4566 Severity : enhancement
4568 Description: Default groups are now added by lconf
4569 Details : You can now run lconf --group <servicename> without having to
4570 manually add groups with lmc.
4573 Frequency : occasional
4575 Description: Nodes with an elan id of 0 trigger a failed assertion
4578 Frequency : always when accessing e.g. tty/console device nodes
4580 Description: tty and some other devices nodes cannot be used on lustre
4581 Details : file's private_data field is used by device data and lustre
4582 values in there got lost. New field was added to struct file to
4583 store fs-specific private data.
4586 Frequency : when exporting Lustre via NFS
4588 Description: NFSD failed occasionally when looking up a path component
4589 Details : NFSD is looking up ".." which was broken in ext3 directories
4590 that had grown large enough to become hashed.
4593 Frequency : Clusters with multiple interfaces not on the same subnet
4595 Description: Nodes will repeatedly try to reconnect to an interface which it
4596 cannot reach and report an error to the log.
4597 Details : Extra peer list entries will be created by lconf with some peers
4598 unreachable. lconf now validates the peer before adding it.
4601 Frequency : Only if a default stripe is set on the filesystem root.
4603 Description: Setting a default stripe on the filesystem root prevented the
4604 filesystem from being remounted.
4605 Details : The client was sending extra request flags in the root getattr
4606 request and did not allocate a reply buffer for the dir EA.
4609 Frequency : occasional, higher if lots of files are accessed by one client
4610 Bugzilla : 6159, 6097
4611 Description: Client trips assertion regarding lsm mismatch/magic
4612 Details : While revalidating inodes the VFS looks up inodes with ifind()
4613 and in rare cases can find an inode that is being freed.
4614 The ll_test_inode() code will free the lsm during ifind()
4615 when it finds an existing inode and then the VFS later attaches
4616 this free lsm to a new inode.
4620 Bugzilla : 6422, 7030
4621 Description: MDS deadlock between mkdir and client eviction
4622 Details : Creating a new file via mkdir or mknod (starting a transaction
4623 and getting the ns lock) can deadlock with client eviction
4624 (gets ns lock and trying to finish a synchronous transaction).
4627 Frequency : occasional
4628 Description: While starting a server, the fsfilt_ext3 module could not be
4630 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4631 kernels. Previously, lconf would still use the ext3 name
4632 when trying to load modules. Now, it will correctly use
4633 ext3 on 2.4 and ldiskfs on 2.6.
4635 Severity : enhancement
4636 Description: The default stripe count has been changed to 1
4637 Details : The interpretation of the default stripe count (0, to lfs
4638 or lmc) has been changed to mean striping across a single
4639 OST, rather than all available. For general usage we have
4640 found a stripe count of 1 or 2 works best.
4642 Severity : enhancement
4643 Description: Add support for compiling against Cray portals.
4644 Details : Conditional compiling for some areas that are different
4648 Frequency : occasional
4649 Bugzilla : 6409, 6834
4650 Description: Creating files with an explicit stripe count may lead to
4651 a failed assertion on the MDS
4652 Details : If some OSTs are full or unavailable, creating files may
4653 trigger a failed assertion on the MDS. Now, Lustre will
4654 try to use other servers or return an error to the
4658 Frequency : occasional
4660 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4661 caused lock timeout during readahead (since 1.4.2).
4662 Details : Processes doing readahead might match a lock that hasn't been
4663 granted yet if there are overlapping and conflicting lock
4664 requests. The readahead process waits on ungranted lock
4665 (original lock is CBPENDING), while OST waits for that process
4666 to cancel CBPENDING read lock and eventually evicts client.
4668 Severity : enhancement
4670 Description: Initial enabling of flock support for clients
4671 Details : Implements fcntl advisory locking and file status functions.
4672 This feature is provided as an optional mount flag (default
4673 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4674 locking are implemented yet, and those that are are not guaranteed
4675 to be completely correct in production environments.
4676 mount -t lustre -o [flock|noflock] ...
4679 Frequency : occasional
4681 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4682 assertion in the ext3 JBD (journaling) layer.
4683 Details : The b_committed_data struct is protected by the big kernel lock
4684 in 2.4 kernels, serializing journal_commit_transaction() and
4685 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4686 access to this struct. In 2.6 kernels there is finer grained
4687 locking to improve SMP performance of the JBD layer.
4691 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4696 Description: merge and cleanup kernel patches.
4699 -----------------------------------------------------------------------------
4701 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4703 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4704 versions of Lustre. Please update all servers and clients to
4705 version 1.4.2 or later at the same time. You must also run
4706 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4707 to update the configuration logs.
4709 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4710 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4711 - swab llog records of type '0' so we get proper header size/idx (5861)
4712 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4713 - fix rename of one directory over another leaking an inode (5953)
4714 - avoid SetPageDirty on 2.6 (5981)
4715 - don't re-add just-being-destroyed locks to the waiting list (5653)
4716 - when creating new directories, inherit the parent's custom
4717 striping settings if present parent (3048)
4718 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4719 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4720 - don't hold client locks on temporary worklist from l_lru (5666)
4721 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4722 - automatically evict dead clients (5921)
4723 - Update file size properly in create+truncate+fstat case (6196)
4724 - Do not unhash mountpoint dentries, do not allow removal of
4726 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4727 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4728 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4729 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4730 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4731 - Correctly handle reads of files with no objects (6243)
4732 - lctl recover will also mark a device active if deactivate used (5933)
4734 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4735 - allow --write-conf on an MDS with different nettype than client (5619)
4736 - don't write config llogs to MDS for mounts not from that MDS (5617)
4737 - lconf should create multiple TCP connections from a client (5201)
4738 - init scripts are now turned off by default; run chkconfig --on
4739 lustre and chkconfig --on lustrefs to use them
4740 - upcalls are no longer needed for clients to recover to failover
4742 - add --abort-recovery option to lconf to abort recovery on device
4744 - add support for an arbitrary number of OSTs (3026)
4745 - Quota support protocol changes.
4746 - forward compatibility changes to wire structs (6007)
4747 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4748 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4749 - improved "lustre status" script
4750 - initialize blocksize for non-regular files (6062)
4751 - added --disable-server and --disable-client configure options (5782)
4752 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4753 - Vanilla 2.4.29 support
4754 - increase maximum number of obd devices to 520 (6242)
4755 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4756 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4757 SLES 9 2.6.5 kernels
4758 - limit stripes per file to 160 (the maximum EA size) (6093)
4760 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4763 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4764 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4765 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4766 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4767 - small ext3 extents cleanups and fixes (5733)
4768 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4769 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4770 - avoid needless client->OST connect, fix handle mismatch (5317)
4771 - fix DLM error path that led to out-of-sync client, long delays (5779)
4772 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4773 - fix several locking issues related to i_size (5492,5624,5654,5672)
4774 - don't move pending lock onto export if it is already evicted (5683)
4775 - fix kernel oops when creating .foo in unlinked directory (5548)
4776 - fix deadlock in obdfilter statistics vs. object create (5811)
4777 - use time_{before,after} to avoid timer jiffies wrap (5882)
4778 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4779 - Do not leak request if server was not able to process it (5154)
4780 - If mds_open unable to find parent dir, make that negative lookup(5154)
4781 - don't create new directories with extent-mapping (5909, 5936)
4783 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4784 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4785 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4786 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4787 - support for RHEL4 kernel on i686 (5773)
4788 - provide error messages when incompatible logs are encountered (5898)
4790 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4791 * version 1.4.0.10 (1.4.1 release candidate 1)
4793 - don't keep a lock reference when lock is not granted (4238)
4794 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4795 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4796 - drop import inflight refcount on signal_completed_replay error (5255)
4797 - unlock page after async write error during send (3677)
4798 - handle missing objects in filter_preprw_read properly (5265)
4799 - no transno return for symlink open, don't save no-trasno open (3440)
4800 - don't try to complete elan receive that already failed (4012)
4801 - free RPC server reply state on error (5406)
4802 - clean up thread from ptlrpc_start_thread() on error (5160)
4803 - readahead could read extra page into cache that wasn't ejected (5388)
4804 - prevent races in class_attach/setup/cleanup/detach (5260)
4805 - don't dereference de->d_inode after l_dput of de (5458)
4806 - use "int" for stripe value returned from lock_to_stripe (5544)
4807 - mballoc allocation and error-checking fixes in 2.6 (5504)
4808 - block device patches to fix I/O request sizes in 2.6 (5482)
4809 - look up hostnames for IB nals (5602)
4810 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4811 - don't start multiple acceptors for the same port (5277)
4812 - fix incorrect LASSERT in mds_getattr_name (5635)
4813 - export a proc file for general "ping" checking (5628)
4814 - fix "lfs check" to not block when the MDS is down (5628)
4816 - service request history (4965)
4817 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4818 - create an "evict_client" /proc entry on OSTs, like the MDS has
4819 - fix mount usage message, return errors per mount(8) (5168)
4820 - change grep [] to grep "[]" in tests so they work in more UMLs
4821 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4822 - remove ancient LOV_MAGIC_V0 EA support (5047)
4823 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4824 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4825 - allow readahead allocations to fail when low on memory (5383)
4826 - mmap locking landed again, after considerable improvement (2828)
4827 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4829 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4832 - send OST transaction number in read/write reply to free req (4966)
4833 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4834 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4835 - the watchdog thread now runs as interruptible (5246)
4836 - flock/lockf fixes (but it's still disabled, pending 5135)
4837 - don't use EXT3 constants in llite code (5094)
4838 - memory shortage at startup could cause assertion (5176)
4840 - reorganization of lov code
4841 - single portals codebase
4843 - add extents/mballoc support (5025)
4844 - direct I/O reads in the obdfilter (4048)
4845 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4847 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4850 - send OST transaction number in read/write reply to free req (4966)
4851 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4852 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4853 - the watchdog thread now runs as interruptible (5246)
4854 - handle missing objects in filter_preprw_read properly (5265)
4855 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4856 - drop import inflight refcount on signal_completed_replay error (5255)
4857 - unlock page after async write error during send (3677)
4858 - return original error code on reconstructed replies (3761)
4859 - no transno return for symlink open, don't save no-trasno open (3440)
4861 - add pid to ldlm debugging output (4922)
4862 - bump the watchdog timeouts -- we can't handle 30sec yet
4863 - extra debugging for orphan dentry/inode bug (5259)
4865 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4868 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4869 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4870 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4871 - fix NULL dereference in /proc/sys/portals/routes (4827)
4872 - allow failed mdc_close() operations to be interrupted (4561)
4873 - stop precreate on OST before MDS would time out on it (4778)
4874 - don't send partial-page writes before EOF from client (4410)
4875 - discard client grant for sub-page writes on large-page clients (4520)
4876 - don't free dentries not owned by NFS code, check generation (4806)
4877 - fix lsm leak if mds_create_objects() fails (4801)
4878 - limit debug_daemon file size, always print CERROR messages (4789)
4879 - use transno after validating reply (3892)
4880 - process timed out requests if import state changes (3754)
4881 - update mtime on OST during writes, return in glimpse (4829)
4882 - add mkfsoptions to LDAP (4679)
4883 - use ->max_readahead method instead of zapping global ra (5039)
4884 - don't interrupt __l_wait_event() during strace
4886 - add software watchdogs to catch hung threads quickly (4941)
4887 - make lustrefs init script start after nfs is mounted
4888 - fix CWARN/ERROR duplication (4930)
4889 - return async write errors to application if possible (2248)
4890 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4891 - print NAL number in %x format (4645)
4892 - update barely-supported suse-2.4.21-171 series (4842)
4893 - support for sles 9 %post scripts
4894 - support for building 2.6 kernel-source packages
4895 - support for sles km_* packages
4897 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4900 - ignore -ENOENT errors in osc_destroy (3639)
4901 - notify osc create thread that OSC is being cleaned up (4600)
4902 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4903 - reconstruct ost_handle() like mds_handle() (4657)
4904 - create a new thread to do import eviction to avoid deadlock (3969)
4905 - let lconf resolve symlinked-to devices (4629)
4906 - don't unlink "objects" from directory with default EA (4554)
4907 - hold socknal file ref over connect in case target is down (4394)
4908 - allow more than 32000 subdirectories in a single directory (3244)
4909 - fix blocks count for O_DIRECT writes (3751)
4910 - OST returns ENOSPC from object create when no space left (4539)
4911 - don't send truncate RPC if file size isn't changing (4410)
4912 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4913 - bind to privileged port in socknal and tcpnal (3689)
4915 - rate limit CERROR/CWARN console message to avoid overload (4519)
4916 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4917 - basic mmap support (3918)
4918 - kernel patch series update from b1_4 (4711)
4920 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4923 - avoid crash during MDS cleanup with OST shut down (2775)
4924 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4925 - don't use bad inodes on the MDS (3744)
4926 - dynamic object preallocation to improve recovery speed (4236)
4927 - don't hold spinlock over lock dumping or change debug flags (4401)
4928 - don't zero obd_dev when it is force cleaned (3651)
4929 - print grants to console if they go negative (4431)
4930 - "lctl deactivate" will stop automatic recovery attempts (3406)
4931 - look for existing locks in ldlm_handle_enqueue() (3764)
4932 - don't resolve lock handle twice in recovery avoiding race (4401)
4933 - revalidate should check working dir is a directory (4134)
4935 - don't always mark "slow" obdfilter messages as errors (4418)
4937 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4940 - don't close LustreDB during write_conf until it is done (3860)
4941 - fix typo in lconf for_each_profile (3821)
4942 - allow dumping logs from multiple threads at one time (3820)
4943 - don't allow multiple threads in OSC recovery (3812)
4944 - fix debug_size parameters (3864)
4945 - fix mds_postrecov to initialize import for llog ctxt (3121)
4946 - replace config semaphore with spinlock (3306)
4947 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4948 - don't allow enqueue to complete on a destroyed export (3822)
4949 - down write_lock before checking llog header bitmap (3825)
4950 - recover from lock replay timeout (3764)
4951 - up llog sem before sending rpc (3652)
4952 - reduce ns lock hold times when setting kms (3267)
4953 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4954 - fix NULL deref and obd_dev leak on setup error (3312)
4955 - replace some LBUG about llog ops with error handling (3841)
4956 - don't match INVALID dentries from d_lookup and spin (3784)
4957 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4958 - fix invalid assertion in ptlrpc_set_wait (3880)
4960 - add libwrap support for the TCP acceptor (3996)
4961 - add /proc/sys/portals/routes for non-root route listing (3994)
4962 - allow setting MDS UUID in .xml (2580)
4963 - print the stack of a process that LBUGs (4228)
4965 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4968 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4969 - make sure to unset replay flag from failed open requests (3440)
4970 - if default stripe count is 0, use OST count for inode size (3636)
4971 - update parent mtime/ctime on client for create/unlink (2611)
4972 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4973 - free recovery state on server during a forced cleanup (3571)
4974 - unregister_reply for resent reqs (3063)
4975 - loop back devices mounting and status check on 2.6 (3563)
4976 - fix resource-creation race that can provoke i_size == 0 (3513)
4977 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4978 - more debugging for page-accounting assertion (3746)
4979 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4980 - avoid deadlock after precreation failure (3758)
4981 - fix race and lock order deadlock in orphan handling (3450, 3750)
4982 - add validity checks when grabbing inodes from l_ast_data (3599)
4984 - add /proc/.../recovery_status to obdfilter (3428)
4985 - lightweight CDEBUG infrastructure, debug daemon (3668)
4986 - change default OSC RPC parameters to be better on small clusters
4987 - turn off OST read cache for files smaller than 32MB
4988 - install man pages and include them in rpms (3100)
4989 - add new init script for (un)mounting lustre filesystems (2593)
4990 - run chkconfig in %post for init scripts (3701)
4991 - drop scimac NAL (unmaintained)
4993 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4996 - clean kiobufs before and after use (3485)
4997 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4998 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4999 - change init script to fail more clearly if not run as root (1528)
5000 - allow clients to reconnect during replay (1742)
5001 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5002 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5003 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5004 - fix cond_resched (3554)
5005 - only evict unfinished clients after recovery (3515)
5006 - allow bulk resend, prevent data loss (3570)
5007 - dynamic ptlrpc request buffer allocation (2102)
5008 - don't allow unlinking open directory if it isn't empty (2904)
5009 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5010 - remove extraneous obd dereference causing LASSERT failure (3334)
5011 - don't use get_cycles() when creating temp. files on the mds (3156)
5012 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5013 - handle EEXIST for set-stripe, set proper directory name (3336)
5015 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5016 - fix ksocknal_fmb_callback() error messages (2918)
5018 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5021 - don't copy lvb into (possibly NULL) reply on error (2983)
5022 - don't deref dentry after dput, don't free lvb on error (2922)
5023 - use the kms to determine writeback rpc length (2947)
5024 - increment oti_logcookies when osc is inactive (2948)
5025 - update client's i_blocks count via lvb messages (2543)
5026 - handle intent open/close of special files properly (1557)
5027 - mount MDS with errors=remount-ro, like obdfilter (2009)
5028 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5029 - don't use cancelling-locks' kms values (2947)
5030 - use highest lock extent for kms, not last one (2925)
5031 - don't dereference ERR_PTR() dentry in error handling path (3107)
5032 - fix thread race in portals_debug_dumplog() (3122)
5033 - create lprocfs device entries at setup instead of at attach (1519)
5034 - common AST error handler, don't evict client on completion race (3145)
5035 - zero nameidata in detach_mnt in 2.6 (3118)
5036 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5037 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5038 - initialize RPC timeout timer earlier for 2.6 (3219)
5039 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5040 - print nal/nid for unknown nid (3258)
5041 - additional checks for oscc recovery before doing precreate (3284)
5042 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5043 - don't crash in mdc_close for bad permissions on open (3285)
5044 - zero i_rdev for non-device files (3147)
5045 - clear page->private before handing to FS, better assertion (3119)
5046 - tune the read pipeline (3236)
5047 - fix incorrect decref of invalidated dentry (2350)
5048 - provide read-ahead stats and refine rpc in flight stats (3328)
5049 - don't hold journal transaction open across create RPC (3313)
5050 - update atime on MDS at close time (3265)
5051 - close LDAP connection when recovering to avoid server load (3315)
5052 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5053 - don't leak open file on MDS after open resend (3325)
5054 - serialize filter_precreate and filter_destroy_precreated (3329)
5055 - loop device shouldn't call sync_dev() for nul device (3092)
5056 - clear page cache after eviction (2766)
5057 - resynchronize MDS->OST in background (2824)
5058 - refuse to mount the same filesystem twice on same mountpoint (3394)
5059 - allow llmount to create routes for mounting behind routers (3320)
5060 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5061 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5062 - fix rare problem with rename on htree directories (3417)
5064 - allow default OST striping configuration per directory (1414)
5065 - fix compilation for qswnal for 2.6 kernels (3125)
5066 - increase maximum number of MDS request buffers for large systems
5067 - change liblustreapi to be useful for external progs like lfsck (3098)
5068 - increase local configuration timeout for slow disks (3353)
5069 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5071 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5074 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5075 - don't overwrite extent policy data in reply if lock was blocked (2901)
5076 - drop filter export grants atomically with removal from device (2663)
5077 - del obd_self_export from work_list in class_disconnect_exports (2908)
5078 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5079 - swab reply message in mdc_close, other PPC fixes (2464)
5080 - fix destroying of named logs (2325)
5081 - overwrite old logs when running lconf --write_conf (2264)
5082 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5083 - fix race in target_handle_connect (2898)
5084 - mds_reint_create() should take same inode create lock (2926)
5085 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5086 - don't close files for self_export to avoid uninitialized obd (2936)
5087 - allow MDS with the same name as client node (2939)
5088 - hold dentry reference for closed log files for unlink (2325)
5089 - reserve space for all logs during transactions (2059)
5090 - don't evict page beyond end of stripe extent (2925)
5091 - don't oops on a deleted current working directory (2399)
5092 - handle hard links to targets without a parent properly (2517)
5093 - don't dereference NULL lock when racing during eviction (2867)
5094 - don't grow lock extents when lots of conflicting locks (2919)
5096 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5099 - account for cache space usage on clients to avoid data loss (974)
5100 - lfsck support in lustre kernel code (2349)
5101 - reduce journal credits needed for BRW writes (2370)
5102 - orphan handling to avoid losing space on client/server crashes
5103 - ptlrpcd can be blocked, stopping ALL progress (2477)
5104 - use lock value blocks to assist in proper KMS, faster stat (1021)
5105 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5106 - recovery for initial connections (2355)
5107 - fixes for mds_cleanup_orphans (1934)
5108 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5109 - block all file creations until orphan recovery completes (1901)
5110 - client remove rq_connection from request struct (2423)
5111 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5112 - recovery timer race (2670)
5113 - mdc_close recovey bug (2532)
5114 - ptlrpc cleanup bug (2710)
5115 - mds timeout on local locks (2588)
5116 - namespace lock held during RPCs (2431)
5117 - handle interrupted sync write properly (2503)
5118 - don't try to handle a message that hasn't been replied to (2699)
5119 - client assert failure during cleanup after abort recovery (2701)
5120 - leak mdc device after failed mount (2712)
5121 - ptlrpc_check_set allows timedout requests to complete (2714)
5122 - wait for inflight reqs when ptlrpcd finishes (2710)
5123 - make sure unregistered services are removed from the srv_list
5124 - reset bulk XID's when resending them (caught by 1138 test)
5125 - unregister_bulk after timeout
5126 - fix lconf error (2694)
5127 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5128 - readahead locks pages, leaves pending causing memory pressure (2673)
5129 - increase OST request buffers to 4096 on large machines (2729)
5130 - fix up permission of existing directories in simple_mkdir (2661)
5131 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5132 - don't assign transno to errored transactions (2742)
5133 - don't delete objects on OST if given a bogus objid from MDS (2751)
5134 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5135 - if rq_no_resend, then timeout request after recovery (2432)
5136 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5137 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5138 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5139 - MDS recovery shouldn't delete objects that it has given out (2730)
5140 - if enqueue arrives after completion, don't clobber LVB (2819)
5141 - don't unlock pages twice when trigger_group_io returns error (2814)
5142 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5143 - don't write pages to disk if there was an error (1450)
5144 - don't ping imports that have recovery disabled (2676)
5145 - take buffered bytes into account when balancing socknal conn (2817)
5146 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5147 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5148 - remove little-endian swabbing of llog records (1987)
5149 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5150 - retry reposting request buffers if they fail (1191)
5151 - grow extent at grant time to avoid granting a revoked lock (2809)
5152 - lock revoke doesn't evict page if covered by a second lock (2765)
5153 - disable VM readahead to avoid reading outside lock extents (2805)
5155 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5156 - updated kernel patches for hp-2.4.20 kernel (2681)
5158 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5161 - fix truncated write corruption (2366)
5162 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5164 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5165 - ensure len(uuid) < 37 in lmc (1171)
5166 - fix ia64 OOPS in llog_test (2255)
5167 - zero end of page at obdfilter for partial page writes (2648)
5168 - don't leave stale dentries around after renames (bug 2428)
5169 - fix timeouts when evicting a client with a single lock held (2642)
5170 - set deadline for the initial HELLO message to drain (2634)
5171 - print out dotted-quad IP addresses in the socknal (2302)
5173 - additional debugging for MDS client eviction problem (2443)
5174 - fix mkfsoptions support for osts (2603, 2604)
5176 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5179 - add series for the vanilla 2.6.0 kernel
5180 - add series for the vanilla 2.4.24 kernel
5181 - add series for a cray x86/64 UL kernel drop
5182 - fix xattr patches for the vanilla 2.4.19 series
5184 - generate true UUIDs in lmc (1171)
5185 - have portals stack dumping break in UML (2466)
5186 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5187 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5188 - wake callback waiting threads on client eviction (2460)
5189 - Add --ptldebug and --subsystem to lmc (1719)
5190 - update assertion to allow safe interrupt allocation
5191 - set rq_no_resend for cancel requests (2432)
5192 - recalculate ptlrpcd timeout after resend (2494)
5193 - call vfs_rmdir when removing pending directories (2368)
5194 - fix renaming a file to itself (2429)
5195 - lmc creates a default one-stripe lov (2454)
5196 - expand procfs space to handle large clusters (2326)
5197 - increase UML stack to avoid overflow
5198 - update lconf's list of debug and subsystem masks
5199 - fix lfs find --obd (2510)
5200 - /proc tunable for disabling filter read caching (2591)
5201 - stop rpm packages from altering slapd.conf (2301)
5202 - disable nagle in the socknal under 0conf (2578)
5203 - choose mds inode size based on stripe count (2572)
5204 - fix kernel-source rpm problems (2516)
5206 - add --disable-doc to avoid pdf generation (2421)
5207 - update documentation, tests, type-os, comments
5208 - avoid format warnings on ia64
5209 - remove the TOE NAL
5210 - tiny code cleanups by removing unused fields
5212 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5215 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5216 - remove the most blatant lies from BUILDING file (2371)
5217 - change default debug level to reasonable production setting
5218 - reduce client side cache size to reduce cache flush time
5219 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5220 - make TCP zerocopy and pinger support enabled by default (2476)
5221 - sync writes completed after process exits caused crashes (2319)
5222 - maintain correct mount count on the MDS (2356)
5223 - backout 1557, because 2316 wasn't really fixed
5224 - better file I/O statistics gathering in /proc
5225 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5226 - another kernel patch to fix zero-copy TCP function export
5227 - don't take duplicate lock when processing re-sent getattr (2420)
5228 - lctl uses obd_self_export instead of creating new conn (2353)
5229 - MDS/OST recovery case which requires object creation asserted (2425)
5230 - move lfs from /usr/sbin to /usr/bin in packages
5231 - fix race between mds_client_add and mds_client_free (2417)
5232 - use kmalloc instead of slabs in portals (2430)
5233 - don't create duplicate records when a failover MDS is present (2442)
5234 - remove unnecessary mount age check (2332)
5235 - don't remove directory inodes from locks prematurely (2451)
5236 - don't break if MDS service name is the same as hostname (2103)
5237 - fix races in client write RPC generation when cache full (2482)
5239 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5242 - remove now-unused request->rq_obd (278)
5243 - if an allocation fails, print out how much memory we've used (1933)
5244 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5245 - add the "configurable stack size" patch to most series files (1256)
5246 - ability to write large log records, for 100+ OST configs (2306)
5247 - fix NULL deref when filter_prep fails (2314)
5248 - fix operator precedence error in filter_sync
5249 - dynamic allocation of socknal TX descriptors (2315)
5250 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5251 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5252 - ensure that CWARN messages in Portals always get to the syslog
5253 - __init/__exit are not for prototype decls (ldlm_init/exit)
5254 - x86-64 compile warning fixes
5255 - fix gateway LMC keyword conflict (2318)
5256 - fix MDS lock inversions in getattr/reint paths (1844)
5257 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5258 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5259 - DLM race condition prevented some lock evictions (2328)
5260 - ENOMEM detection and retry on socknal sends (2230)
5261 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5262 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5263 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5264 - better error messages when a client is rejected during recovery (1505)
5265 - avoid cancelling locks which were never granted, after failure (2330)
5266 - fix i_sem/journal inversion in mds_client_add (2333)
5267 - fix truncate/getattr lock cycle deadlock (2334)
5268 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5269 - fix two rare exit paths which could leak an l_lock() ref (2321)
5270 - fencepost error in MDS/OST orphan recovery (2226)
5271 - make log record alignment 8 bytes (1988)
5272 - lstripe now fails when requested offset > ost_count (2237)
5273 - ensure that all kernel series have a complete list.h (1607)
5274 - fix crashes in special-file operations (2316)
5275 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5277 - allow configurable automake binary, for testing new versions
5278 - small update to the lfs documentation
5280 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5282 * fix negative export reference count in fsfilt_sync (2312)
5284 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5285 * release candidate 0.9.1
5287 - orphans are moved into the PENDING directory for possible recovery
5288 - replayed opens now open by fid for orphan/rename safety (1042)
5289 - last close of an orphan inode generates a transno (683)
5290 - chdir() and mount() now pin the directory entry (1020)
5291 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5292 - discard very old requests without processing them (1502)
5293 - remove some common, well-understood CERRORs (1505)
5294 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5295 - clear "grant" flags in OST replies until OST grant code lands (1644)
5296 - fix read performance by not clobbering i_blksize on client (1598)
5297 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5298 - make LRU size a /proc tunable, clears locks when reduced (707)
5299 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5300 - server locks take references on exports now (1558)
5301 - build fixes for 2.4.20-rh trees (1663)
5302 - return an error from lov_create if all OSCs are inactive (1751)
5303 - fix import levels when a reconnect happens without a timeout (1597)
5304 - exit early from mds_open if we get a lookup error (1749)
5305 - partial page read at EOF wouldn't wait for disk before sending (1642)
5306 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5307 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5308 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5309 - write/truncate lock inversion (1639)
5310 - Don't auto-load obdclass, portals modules during cleanup (1495)
5311 - fix timestamps from jumping to "now" (1763)
5312 - extra journal assertions (1648)
5313 - add an extra multiunlink test (1771)
5314 - fix read_record/write_record API (1776)
5315 - fix leak of offset_extent, possible incorrect i_size later (1772)
5316 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5317 - Debugging for the kqswnal_get_idle_tx problems (1820)
5318 - Allow recovery to be attempted multiple times (1536)
5319 - Write out MDS last_rcvd file after it is first created (1600)
5320 - Fix tx_descriptor leak in failed transmit situations (1827)
5321 - ext3 journaling fixes for assertion failure after IO error (1871)
5322 - class_export_put() on freed export after completion AST error (1896)
5323 - Fix revalidate looping in VFS (1322)
5324 - Don't access a freed export during MDS_REINT timeout (1521)
5325 - Add open-unlink recovery support on the MDS (1673,1764)
5326 - Return an error if no MDS data was read from last_rcvd (1946)
5327 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5328 - Don't setup a disk device that doesn't match exported UUID (317)
5329 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5330 - avoid committing NULL handle in force close
5331 - local.sh is now a one-stripe LOV configuration
5332 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5333 - fix ext3 htree duplicate directory entry corruption (1516)
5334 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5335 - update to new LOV EA format (2097)
5336 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5338 2003-06-15 Phil Schwan <phil@clusterfs.com>
5341 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5342 - per-import/export recovery handling (958, 931, 959)
5343 - multiple last-rcvd slots, for serving multiple FSes (949)
5344 - connections are again shared between multiple imp/exports (963, 964)
5345 - "umount -f" would hang if any requests needed to be sent (393, 978)
5346 - avoid pinning large req buffer by copying for queued messages (989)
5347 - add "uuid" to "lctl device" command to help upcalls (991)
5348 - "open" RPCs with transnos would confuse recovery counters (1037)
5349 - do proper endian conversion of all wire messages (288, 340, 891)
5350 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5351 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5352 - fix ptlrpc_connection leak in target_handle_connect (1174)
5353 - fix import refcounting bug in OST and MDS cleanup (1134)
5354 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5355 - fix very unlikely obd_types race condition (501)
5356 - remove osc_open hack for echo_client (1187)
5357 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5358 - a failure in read_inode2 leads to deadlock (1139)
5359 - cancel ack-locks as soon as transaction is committed (1072)
5360 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5361 - make sure to commitrw after any preprw to avoid deadlock (1162)
5362 - failing to execute a file in a lustre FS would lock inode (1203)
5363 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5364 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5365 - obd_uuid2tgt would walk off the end of the list (1255)
5366 - on IA64 the osc would give portals incorrect bulk size (1258)
5367 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5368 - fix lock inversion caused by new llite matching code (1282)
5369 - limit the number of dirty pages on a client to 10MB (1286)
5370 - timed out locks were not being corrected cancelled (1289)
5371 - fix O_DIRECT above 4GB on IA-32 (1292)
5372 * major user-visible changes
5373 - fail out/fail over policy now controlled by the upcall (993)
5375 - add OBD_PING to check server availability and failure (954)
5376 - lustre messages are now sent in sending host order (288, 340, 891)
5377 - add eadatalen to MDS getattr reply (340)
5378 - OST read replies may contain second buffer, with per-page status (593)
5380 2003-03-11 Phil Schwan <phil@clusterfs.com>
5383 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5384 - failed open()s could cause deadlock; fixed (867, 869)
5385 - stop cancelling OST locks when files are closed (481)
5386 - overlapping XID spaces caused network corruption (851, 853)
5387 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5388 - setattr_raw added, to do single-RPC, server-side setattrs
5389 - lmc/lconf syntax change for OST UUIDs
5390 - fix crashy race condition between ptlrpc_free_req and osc_close
5391 - don't use request in mdc_enqueue if we hit a timeout (889)
5392 - don't set the inode i_size for regular files from the MDS (896)
5393 - handle out of order completion AST (842)
5394 - don't LBUG if a lock request times out after receiving AST (913)
5395 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5396 - if a bad lock AST arrives, send an error instead of dropping entirely
5397 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5398 - fix leak in bulk IO when only partially completed (899, 900, 926)
5399 - fix O_DIRECT for ia64 (55)
5400 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5401 - C-z after timeout could hang a process forever; fixed (977)
5403 - client-side I/O cache (678, 924, 929, 941, 970)
5405 - READPAGE and SETATTRs which don't take server-side locks get
5408 2003-02-11 Phil Schwan <phil@clusterfs.com>
5411 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5412 - Fix fsfilt_extN_readpage() to read a full page of directory
5413 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5414 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5415 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5416 - Most intent operations were removed, in favour of a new RPC mode
5417 that does a single RPC to the server and bypasses most of the VFS
5418 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5419 - Aggressively cancel local locks on DLM servers
5420 - mds_reint_unlink sends EA to the client if it's the last nlink.
5421 client uses that EA to unlink OST objects.
5422 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5423 - recursive symlinks were fixed (439)
5424 - fixed NULL deref in DEBUG_REQ
5425 - filter_update_lastobjid no longer calls sync, which annoyed extN
5426 - fixed multi-client small-writes to a single file problem (445)
5427 - fixed mtime updates during file writes (607)
5428 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5429 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5430 - fixed Linux OST size reporting problem (444, 656)
5431 - OST now updates object mtime with writes or setattr (607, 619)
5432 - client verifies file size before zeroing page past EOF (445)
5433 - OST now writes last allocated objid to disk with allocation (108)
5434 - LOV on echo now works (409)
5436 - mds_reint_unlink sends a new buffer, with the EA included. this
5437 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5438 set if a regular file was being unlinked, and it was the last link
5439 - use PtlGet from the target for bulk writes (315)
5440 - OST now updates object mtime with writes or setattr (607, 619)
5441 - LDLM now has a grant-time callback to revalidate locked items, if
5443 - Many MDS operations were reorganized to combat race conditions
5445 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5446 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5448 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5451 - Fully reactivate OST imports after reconnection (512, others)
5452 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5453 - More graceful error handling for truncating on dead OST (515)
5454 - Don't error out unless we're actually accessing dead stripes (474)
5455 - Fix garbage sizes when stripes are missing (410)
5456 - LRU counters were broken, causing constant lock purge (433, 432)
5457 - garbage on read from stripes with failed OSTs (441)
5458 - mark OSCs as active before reconnecting during recovery (438)
5459 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5460 - lfind did not preserve OST order in output (443)
5461 - symlinks cause hung clients, incorrect data (439)
5462 - stop dereferencing request after dropping refcount (457)
5463 - don't LASSERT(spin_is_locked) on non-SMP (455)
5464 - fixes for many rename() bugs
5465 - fstat didn't correctly synchronize attributes (399)
5466 - server must handle lock cancellation during blocking AST prep (487)
5467 - bulk descriptors were free()d too soon (511)
5468 - fix paths in lconf, which would load incorrect modules (451, 507)
5469 - fix confusing lconf 'host not found' error message (386)
5470 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5471 - fix race condition in mdc_blocking_ast() for inode access (526)
5472 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5473 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5474 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5475 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5476 - fix lstripe in several ways (kernel side) (527)
5477 - fix request leak in ldlm_cli_enqueue (262)
5478 - incorrect OSC was marked inactive after OST failure
5479 - call mds_fs_cleanup before unmounting filesystem (524)
5480 - fix races between taking ns_lock and ldlm_lock_change_resource
5481 - fix races updating LOV export open file list
5482 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5483 - fix recovery NULL deref in ldlm_cli_cancel_unused
5484 - fix some DLM races by using new hash table for lock handles (419)
5485 - permit the client to specify desired inodes, at replay
5486 - duplicate requests when we queue them for replay reintegration
5487 - fix last_rcvd offset calculation
5488 - sync after each recovered transaction, so we always make progress
5489 - never, not always, ERESTART requests without transnos
5490 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5492 - skip replay if the MDS says that the client is already connected
5493 - don't check for a recovery-enabled export to match lctl's UUID
5494 - don't INC_USE_COUNT for phantom exports
5495 - don't crash when cleaning up phantom exports (567)
5496 - don't double-finish or set replay data for errored mdc_open requests
5497 - abort requests when they time out, so we don't get old replies
5498 - send/receive replies for AST messages again
5499 - if the client says that it doesn't have the lock, cancel it on the
5501 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5502 mark it as destroyed, it will all work out when decref is called
5503 - fix module use counts (22, 581)
5505 - ASTs now expect a reply (server cancels lock on error reply)
5507 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5510 - fix many simultaneous client startup (392)
5511 - fix dentry->d_it clobbering
5512 - credentials weren't being shipped for readdir/getattr operations
5513 - remove invalid assertions triggered during some concurrent MD
5515 - proper Lustre versions added (336, 389)
5516 - fix memory leak for create error case (398)
5517 - fix LOV locking bug that would get cli/srv out of sync
5518 - fix echo client over LOV (409)
5519 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5520 - fix double-O_EXCL intent crash (424)
5521 - avoid sending multiple lock CANCELs (352)
5523 - MDS can do multi-client recovery (modulo bugs in new code)
5525 - many updates, edits, cleanups
5527 2002-11-18 Phil Schwan <phil@clusterfs.com>
5530 - fix null d_it dereference (346)
5531 - fix full OST/dbench hang (333)
5532 - fix permission problem with file removal (286)
5533 - fix removal of OSCs from LOV when they fail
5534 - fix NULL deref during bulk timeout (214)
5535 - fix problems related to multiple filesystems on one MDS (241)
5536 - fixed serious subtle metadata locking bugs
5537 - free locks on clients when inodes are removed due to memory
5539 - fix inode pointer in lock data (285)
5540 - partial support for multiple MDS on a single host (241)
5541 - data locks weren't cancelled at clear_inode time (290, 311)
5542 - intent locks could lead to unbounded lock growth (205)
5543 - added a maximum lock count, an LRU list, and a flusher
5544 - fix multiple rename (365)
5545 - properly abstracted the echo client
5546 - OSC locked 1 byte too many; fixed
5547 - rewrote brw callback code:
5548 - fixed recovery bugs related to LOVs (306)
5549 - fixed too-many-pages-in-one-write crash (191)
5550 - fixed (again) crash in sync_io_timeout (214)
5551 - probably fixed callback-related race (385)
5553 - Add capability to MDS protocol
5554 - LDLM cancellations and callbacks on different portals
5556 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5559 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5560 - defer open object destruction to close time (601981, 138)
5561 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5562 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5563 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5564 - don't call obd_setattr in ll_file_release for destroyed objects
5565 * protocol change to lustre_msg: move |version| and add |flags|
5566 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5567 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5568 * added replay of create, unlink, link and rename operations during
5569 MDS failover; recovery should be much more robust now
5570 * remove failed OSCs from LOVs (only lov_create uses this so far)
5571 * the lustre-HOWTO was brought (more) up to date (582544)
5573 2002-10-23 Phil Schwan <phil@clusterfs.com>
5576 - in-use dentries weren't being reused properly (617851)
5577 - prevent multiple LDLM setup (599178)
5578 - fix LOV size calculations for truncate (617853)
5579 - fix client handling of MDS intent errors (POSIX)
5580 - fix permission bug in lovstripe.c test (624321)
5581 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5582 - truncate past end of file could corrupt data
5583 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5584 - a race in recovery could return ETIMEDOUT to apps (623947)
5585 - building outside the source directory was fixed
5586 * the lustre-HOWTO was brought (more) up to date (582544)
5587 * major progress was made on recovery functionality
5589 2002-10-10 Phil Schwan <phil@clusterfs.com>
5592 - recovery deadlock fix
5593 - rm -rf causes LBUG fix (617817)
5594 - file open by multiple tasks fix (618962)
5595 - directory permissions bugs (602707 and 620007)
5596 - journal_stop fixed with locking (611313)
5597 - O_APPEND failures resolved (618273, perhaps 614459)
5598 - lconf PATH fix (619770)
5599 - IA64 build fix (621450)
5600 - RPC buffer sizes scale with amount of memory
5602 2002-10-01 Phil Schwan <phil@clusterfs.com>
5605 - locks would be cancelled without throwing away data pages,
5606 resulting in inconsistent data (605627)
5607 - inode attributes were not always being refreshed (605627, 612449)
5608 - lconf now continues to cleanup after lctl reports an error
5609 - MDS now enforces user permissions (602707)
5610 - lprocfs cleanup fixed, but not yet enabled (614157)
5611 - fixed infinite server hang, should a client not respond to an AST
5612 - avoid going into recovery if user calls readlink() with a buffer
5613 that's too small (613941)
5614 - AST RPCs no longer require replies (614867) -- this may be changed
5615 - don't crash server if client sends an IOV that's too big (611336)
5616 - fixed lock conversion deadlock (611892)
5617 - fixed the following of symlinks (614622)
5618 * recovery: the server can remove locks from a client that dies, other
5619 clients can make progress
5620 * more extN patch fixes
5621 * compile-time configurable ptlrpc buffer allocations
5623 - collaborative read cache document
5624 - Lustre Lite Performance CDR document-in-progress
5626 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5629 - fix typo in patch-2.4.18
5631 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5634 - clear ptlrpc request each time in handle_incoming_request()
5635 - unlink of files now destroys the object on the OST
5637 2002-09-19 Peter Braam <braam@clusterfs.com>
5639 * add hard link support
5640 * change obdfile creation method
5641 * kernel patch changed
5643 2002-09-19 Peter Braam <braam@clusterfs.com>
5646 - stack overflow bug in extN fixed
5648 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5650 * documentation updates
5651 - add man pages for config tools
5652 - update tests/README to describe testing with new config tools
5653 - finish metadata API descriptions
5654 * bug fixes and cleanups
5655 - statfs workaround for 16TB limit
5656 - LOV stripe allocation improved, can stripe on subset of OSTs
5657 - LOV file size/IO offset was wrong for files > 4GB in size
5658 - object EA data was being dropped, caused files to be unreadable
5659 - memory overflow with non-LOV OST caused memory corruption
5660 - fixed regression tests to work with new config tools, obdfilter
5661 - fixed bug when directory size became larger than 1 block
5662 - fixed bug (for single client case) when PWD was deleted
5663 - invalidate local directory pages when doing intent-based ops
5664 - avoid LDLM oops when lock callback contained bad data
5666 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5668 * documentation updates
5669 * bug fixes and cleanups
5670 - configuration tools
5673 - 64-bit compile warnings
5674 - 64-bit internal statfs data
5676 * test_brw on persistent OST devices
5678 * lprocfs (disabled)
5680 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5682 * documentation updates
5683 * bug fixes and cleanups
5684 * configuration tools
5686 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5689 * many small fixes to 0_5_4
5690 * io/network handling
5691 * thinkos in MDS operations
5693 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5696 * crucial basic fixes to 0.5.3
5697 * IOR, Iozone work over Elan
5700 2002-08-07 Phil Schwan <phil@clusterfs.com>
5701 * version 0_5_3, our first alpha
5702 * we use the new Portals iovs
5703 * documentation updates
5704 * bug fixes and cleanups
5705 * small changes in the DLM wire protocol
5707 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5708 * version 0_5_1 with some initial stability,
5709 * locking on MD and file I/O.
5710 * documentation updates
5711 * several bug fixes since 0.5.0
5712 * small changes in wire protocol
5714 2002-07-18 Phil Schwan <phil@clusterfs.com>
5716 * delivered as Lustre Light Alpha
5717 * fixed a crash after handling invalid MDS requests
5718 * fixed directory pages for architectures with non-4k pages sizes
5720 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5721 * release version v0_4_4
5722 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5724 * Fixes a number of interruption problems with OST operations.
5725 * Update documentation for portals header changes
5726 * Move all wire protocol structs/defines to lustre_idl.h
5727 * Fixes symlink length bug.
5728 * Add tcpdump to repository.
5730 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5731 * release version v0_4_3
5732 * Fixes statfs for inodes on extN.
5733 * Fixes bug in runtests which would delete /etc/hosts.
5734 * Use 64-bit object IDs wherever possible (not into VFS though)
5735 Remove ost_get_info, which is unused by lustre, and out of date.
5737 2002-07-03 Peter Braam <braam@clusterfs.com>
5738 * release version v0_4_2 Fixes a lookup error (type not passed)
5739 * move forward to head of Portals
5740 * move forward to latest Lustre kernel
5742 2002-06-25 Peter Braam <braam@clusterfs.com>
5743 * release version v0_4_1. Hopefully stable on single node use.