1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.39.3 (SLES 10),
5 2.6.18-128.1.14.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.41.6.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.
18 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
22 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
26 Description: Error handling in osc_statfs_interpret() has been improved.
27 Details : Check in osc_statfs_interpret() for EBADR.
31 Description: Do not send statfs() requests to OSTs disabled by administrator.
32 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
34 Severity : enhancement
36 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
38 Severity : enhancement
40 Description: Enhancement for lfs(1) command to use numeric uid/gid.
44 Description: Improved cleanup procedure for conf-sanity test_46a.
48 Description: Do not start bulk transfer, if client tried to reconnect.
49 Details : Do not start bulk transfer, if client tried to reconnect,
50 but failed due to active rpcs. Check for exp_abort_active_req
55 Description: Prevent inconsistences between linux and lustre mount structures.
56 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
57 Make the sleep interruptible.
59 Severity : enhancement
61 Description: provide server to client comms path
62 Details : server to client communications path via new LDLM_SET_INFO rpc
66 Description: truncate starts GFP_FS allocation under transaction causing deadlock
67 Details : ldiskfs_truncate calls grab_cache_page which may start page
68 allocation under an open transaction. This may lead to
69 calling prune_icache with consequent lustre reentrance.
73 Description: send correctly lsm on open replay
74 Details : MDS is trust to LSM size on replay open, but client can set wrong size
79 Description: enable client interface failover
80 Details : When a child reconnects from another NID, properly update export
81 nid hash position and ldlm reverse import.
83 Severity : enhancement
85 Description: Communicate OST degraded/readonly state via statfs to MDS
86 Details : Flags in the statfs returned from OSTs indicate whether the
87 OST is in a degraded RAID state, or if the filesystem has
88 turned read-only after a filesystem error is detected.
93 Description: don't panic if EPROTO was hit when reading symlink
94 Details : correctly handling request reference in error cases.
96 Severity : enhancement
98 Description: Add LustreNetLink, a kernel-userspace communcation path.
100 Severity : enhancement
102 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
107 Description: access to llog context before init.
108 Details : move handling CATALOGS file at osc layer and forbid access to llog
112 Frequency : rare, only when client is writing to OST on local node
114 Description: Avoid deadlock for local client writes
115 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
116 memory freeing context. This allows OST threads to set the
117 PF_MEMALLOC flag on task structures in order to allocate memory
118 from reserved pools and complete IO.
119 Use GFP_HIGHUSER for OST allocations for non-local client writes,
120 so that the OST threads generate memory pressure and allow
121 inactive pages to be reclaimed.
123 Severity : enhancement
125 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
130 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
131 Details : move ldlm namespace creation in setup phase to avoid grab
132 _lprocfs_lock with cli_sem held.
136 Description: Temporarily disable grant shrink.
137 Details : Disable the feature for debugging.
140 Frequency : only when formatting test filesystems
142 Description: Unable to run several mkfs.lustre on loop devices at the same time
143 Details : mkfs.lustre returns error 256 on the concurrent loop devices
144 formatting. The solution is to proper handle the error.
146 Severity : enhancement
148 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
150 Severity : enhancement
152 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
157 Description: resolve race between obd_disconnect and class_disconnect_exports
158 Details : if obd_disconnect will be called to already disconnected export he
159 forget release one reference and osc module can't unloaded.
161 Severity : enhancement
163 Description: Allow tuning service thread via /proc
164 Details : For each service a new
165 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
166 created that can be used to set min/max thread counts, and get the
167 current number of running threads.
172 Descriptoin: don't return error if have partially created objects for file.
173 Details : lov_update_create_set uses set->set_success as index for created
174 objects, so if some requests will be failed, they will have hole
175 at end of array and qos_shrink_lsm for allocate correct lsm.
177 Severity : enhancement
179 Description: Update OFED support to 1.4.1
184 Description: Reduce small size read RPC
185 Details : Set read-ahead limit for every file and only do read-ahead when
186 available read-ahead pages are bigger than 1M to avoid small size
189 Severity : enhancement
191 Description: move AT tunable parameters for more consistent usage
192 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
194 Severity : enhancement
196 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
197 Details : allow skip disconnected ost for send statfs request and hide error
203 Description: fix racy locking of mballoc block bitmaps causing BUG
204 Details : The locking of the mballoc buddy bitmap and the in-memory
205 block bitmap was using two different spin locks in some
206 cases. This made it possible to incorrectly access the
207 mballoc bitmap while another process was modifying it,
208 causing a sanity assertion to fail. While no on-disk corruption
209 was reported, there was some risk of this happening.
212 Frequency : rare, on llog test 6
214 Descriptoin: don't allow connect to already connected import
215 Details : allowing connect to already connected import is hide connecting problem.
220 Descriptoin: allow kill process which wait statahead result
221 Details : for some reasons 'ls' can stick in waiting result from statahead,
222 in this case need way for kill this process.
224 Severity : enhancement
226 Description: Add state history info file, enhance import info file
227 Details : Track import connection state changes in a new osc/mdc proc file;
228 add overview-type data to the osc/mdc import proc file.
230 Severity : enhancement
232 Description: MDS create should not wait for statfs RPC while holding DLM lock.
234 Severity : enhancement
236 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
241 Descriptoin: OST grant too much space to client even there are not enough space.
242 Details : Client will shrink its grant cache to OST if there are no write
243 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
244 this grant cache if there are already not enough avaible space
245 (left_space < total_clients * 32M).
248 Frequency : start MDS on uncleanly shutdowned MDS device
250 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
251 Details : stay in waiting mds<>ost recovery finished produce random bugs
252 due race between two ll_sync thread for one lov target. send
253 ACTIVATE event only if connect realy finished and import have
257 Frequency : rare, connect and disconnect target at same time
259 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
260 Details : don't call obd_disconnect under lov_lock. this long time
261 operation and can block ptlrpcd which answer to connect request.
266 Descriptoin: don't lose wakeup for imp_recovery_waitq
267 Details : recover_import_no_retry or invalidate_import and import_close can
268 both sleep on imp_recovery_waitq, but we was send only one wakeup
272 Frequency : always with long access acl
274 Descriptoin: mds can't pack reply with long acl.
275 Details : mds don't control size of acl but they limited by reint/getattr
278 Severity : enhancement
280 Description: Update to SLES10 kernel-2.6.16.60-0.33.
282 Severity : enhancement
284 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
287 Frequency : start MDS on uncleanly shutdowned MDS device
289 Descriptoin: aborting recovery hang on MDS
290 Details : don't throttle destroy RPCs for the MDT.
293 Frequency : on remount
295 Description: external journal device not working after the remount
296 Details : clear dev_rdonly flag for external journal devices in
302 Description: shutdown vs evict race
303 Details : client_disconnect_export vs connect request race.
304 if client will evicted at this time - we start invalidate
305 thread without referece to import and import can be freed
309 Frequency : rare, need acl's on inode.
311 Description: client can't handle ost additional correctly
312 Details : if ost was added after client connected to mds client can have
313 hit lnet_try_match_md ... to big messages to wide striped files.
314 in this case need teach client to handle config events about add
315 lov target and update client max ea size at that event.
317 Severity : enhancement
319 Description: Changelogs
320 Details : Changelogs are a lightweight mechanism to track filesystem
321 metadata and namespace changes. The changelog is recorded
322 permanently on the MDTs, and is periodically "consumed" / purged
323 when records are no longer needed.
325 Severity : enhancement
327 Description: compact fld format with extents
328 Details : Store range of seq rather than every seq in FLD. Seq
329 controller update FLD rather than clients. In Case of CMD, mdt0
330 has FLD, all other metadata server act as non persistent proxy
331 for FLD queries and cache fld entries in fld cache.
336 Description: don't skip ost target if they assigned to file
337 Details : Drop slow OSCs if we can, but not for requested start idx.
338 This means "if OSC is slow and it is not the requested
339 start OST, then it can be skipped, otherwise skip it only
340 if it is inactive/recovering/out-of-space.
344 Description: more cleanup in mds_lov
345 Details : not send LOV EA under replay, we can't know about they size at this
346 time. Don't allow client connect to mds before any ost connected,
347 for avoid problems with LOV EA size and returning EIO to client.
349 Severity : enhancement
351 Description: Interoperability at server side (Disk interoperability)
353 Severity : enhancement
355 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
357 Severity : enhancement
359 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
361 Severity : enhancement
363 Description: New client IO stack (CLIO).
365 Severity : enhancement
367 Description: Commit on sharing. Eliminate inter-client dependencies between
368 uncommitted transactions by doing transaction commits.
369 Thereby clients may recovery independently.
372 Frequency : Create a symlink file with a very long name
374 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
375 Details : If there is no extra space in the request for early cancels,
376 ldlm_req_handles_avail() returns 0 instead of a negative value.
378 Severity : enhancement
380 Description: Add /proc entry for import status
381 Details : The mdc, osc, and mgc import directories now have
382 an import directory that contains useful import data for debugging
385 Severity : enhancement
387 Description: Re-disable certain /proc logging
388 Details : Enable and disable client's offset_stats, extents_stats and
389 extents_stats_per_process stats logging on the fly.
392 Frequency : Only on FC kernels 2.6.22+
394 Description: oops in statahead
395 Details : Do not drop reference count for the dentry from VFS when lookup,
396 VFS will do that by itself.
398 Severity : enhancement
400 Description: Generic /proc file permissions
401 Details : Set /Proc file permissions in a more generic way to enable non-
402 root users operate on some /proc files.
406 Description: Hitting mdc_commit_close() ASSERTION
407 Details : Properly handle request reference release in
408 ll_release_openhandle().
412 Description: quota recovery deadlock during mds failover
413 Details : This patch includes att18982, att18236, att18237 in bz14840.
415 1. fix osts hang when mds does failover with quotaon
416 2. prevent watchdog storm when osts threads wait for the
421 Frequency : only patchless client
422 Description: add workaround for race between add/remove dentry from hash
424 Severity : enhancement
426 Description: Allow OST glimpses to return PW locks
430 Description: LBUG when llog conf file is full
431 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
435 Description: Prevent import from entering FULL state when server in recovery
439 Description: service mount cannot take device name with ":"
440 Details : Only when device name contains ":/" will mount treat it as
446 Description: replace ptlrpcd with the statahead thread to interpret the async
447 statahead RPC callback
451 Frequency : on recovery
452 Description: I/O failures after umount during fail back
453 Details : if client reconnected to restarted server we need join to recovery
454 instead of find server handler is changed and process self
455 eviction with cancel all locks.
457 Severity : enhancement
459 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
461 Severity : enhancement
463 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
465 Severity : enhancement
467 Description: Upcall on Lustre log has been dumped
468 Details : Allow for a user mode script to be called once a Lustre log has
469 been dumped. It passes the filename of the dumped log to the
470 script, the location of the script can be specified via
471 /proc/sys/lnet/debug_log_upcall.
476 Description: avoid idr_remove called for id which is not allocated.
477 Details : Move assigment s_dev for clustered nfs to end of initialization,
478 to avoid problem with error handling.
483 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
484 Details : When connection is reused this not moved from CONN_UNUSED_HASH
485 into CONN_USED_HASH and this prodice warning when put connection
486 again in unused hash.
489 Severity : enhancement
491 Description: File striping can now be set to use an arbitrary pool of OSTs.
493 Severity : enhancement
495 Description: Export bytes_read/bytes_write count on OSC/OST.
499 Description: Early reply size mismatch, MGC loses connection
500 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
501 the connect flags are properly negotiated.
506 Description: quotas are not honored with O_DIRECT
507 Details : all writes with the flag O_DIRECT will use grants which leads to
508 this problem. Now using OBD_BRW_SYNC to guard this.
512 Description: add quota statistics
513 Details : 1. sort out quota proc entries and proc code.
514 2. add quota statistics
516 Severity : enhancement
518 Description: enable quota support for HEAD.
522 Description: Properly propagate oinfo flags from lov to osc for statfs
523 Details : restore missing copy oi_flags to lov requests.
525 Severity : enhancement
527 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
528 : l_getgroups(8), lst(8), routerstat(8)
529 Details : included man pages for llobdstat(8), llstat(8),
530 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
532 Severity : enhancement
534 Description: configure's --enable-quota should check the
535 : kernel .config for CONFIG_QUOTA
536 Details : configure is terminated if --enable-quota is passed but
537 : no quota support is in kernel
541 Description: Remove portals compatibility
542 Details : Remove portals compatibility, not interoperable with releases
547 Description: Resolve device initialization race
548 Details : Prevent proc handler from accessing devices added to the
549 obd_devs array but yet be intialized.
551 Severity : enhancement
553 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
555 Severity : enhancement
557 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
562 Description: Using wrong pointer in osc_brw_prep_request
563 Details : Access to array[-1] can produce panic if kernel compiled with
564 CONFIG_PAGE_ALLOC enabled
568 Description: Client runs out of low memory
569 Details : Consider only lowmem when counting initial number of llap pages
573 Description: *optional* service tags registration
574 Details : if the "service tags" package is installed on a Lustre node
575 When the filesystem is mounted, a local-node service tag will
576 be created. See http://inventory.sun.com/ for more information
577 about the Service Tags asset management system.
581 Description: Kernel BUG tries to release flock
582 Details : Lustre does not destroy flock lock before last reference goes
583 away. So always drop flock locks when client is evicted and
584 perform unlock regardless of successfulness of speaking to MDS.
588 Description: add refcount protection for osc callbacks, avoid panic on shutdown
592 Description: sanity test 65a fails if stripecount of -1 is set
593 Details : handle -1 striping on filesystem in ll_dirstripe_verify
598 Description: ASSERTION(CheckWriteback(page,cmd)) failed
599 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
602 Severity : enhancement
604 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
608 Description: do not process already freed flock
609 Details : flock can possibly be freed by another thread before it reaches
610 to ldlm_flock_completion_ast.
614 Description: LBUG during stress test
615 Details : Need properly lock accesses the flock deadlock detection list.
619 Description: oops in page fault handler
620 Details : kernel page fault handler can return two special 'pages' in error
621 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
625 Description: timeout with invalidate import.
626 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
627 should be handled by ptlrpcd. This produce long age waiting and
628 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
630 Severity : enhancement
632 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
636 Description: enable MGS and MDT services start separately
637 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
638 a co-located MGS without starting the MGS, which is a complement
639 to 'nosvc' mount option.
643 Frequency : after recovery
644 Description: precreate to many object's after del orphan.
645 Details : del orphan st in oscc last_id == next_id and this triger growing
646 count of precreated objects. Set flag LOW to skip increase count
647 of precreated objects.
651 Frequency : rare, on clear nid stats
652 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
653 Details : when clean nid stats sometimes try destroy live entry,
654 and this produce panic in free.
658 Description: Stack overflow during MDS log replay
659 ease stack pressure by using a thread dealing llog_process.
663 Description: wait until IO finished before start new when do lock cancel.
664 Details : VM protocol want old IO finished before start new, in this case
665 need wait until PG_writeback is cleared until check dirty flag and
666 call writepages in lock cancel callback.
668 Severity : enhancement
670 Description: using special macro for print time and cleanup in includes.
674 Description: mds_mfd_close() ASSERTION(rc == 0)
675 Details : In mds_mfd_close(), we need protect inode's writecount change
676 within its orphan write semaphore to prevent possible races.
680 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
684 Frequency : rare, on shutdown ost
685 Description: don't hit live lock with umount ost.
686 Details : shrink_dcache_parent can be in long loop with destroy dentries,
687 use shrink_dcache_sb instead.
691 Description: don't panic with use echo client
692 Details : echo client pass NULL as client nid pointer and this produce null
697 Description: fix build on ppc32
698 Details : compile code with -m64 flag produce wrong object file for ppc32.
702 Description: add message levels for liblustreapi
706 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
707 Details : Move the 'good_osts' check before the 'total_bavail' check. This
708 will result in an -EAGAIN and in the exit call path we call
709 alloc_rr() which will with increasing aggressiveness attempt to
710 aquire precreated objects on the minimum number of required OSCs.
714 Description: Use old size assignment to avoid deadlock
715 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
716 the scheduling while holding a spinlock. We do not need locking
717 for size in ll_update_inode() because size is only updated from
718 the MDS for directories or files without objects, so there is no
719 other place to do the update, and concurrent access to such inodes
720 are protected by the inode lock.
724 Description: resolve "_IOWR redefined" build error on SLES10
728 Description: dump the memory debugging after all modules are unloaded to
729 suppress false negative in conf_sanity test 39
731 Severity : enhancement
733 Description: build kernel-ib packages for OFED 1.3 in our release cycle
738 Description: fix SLES kernel versioning
739 Details : the kernel version for our SLES 10 kernel did not include a "-"
740 before the "smp" at the end. while this was not a problem in
741 general, it did mean that software trying to use the kernel
742 version to try to detect a vendor specific kernel would fail.
743 this was most evident by the OFED build scripts.
747 Description: Don't update lov_desc members until making sure they are valid
748 Details : When updating lov_desc members via proc fs, need fix their
749 validities before doing the real update.
753 Description: don't put request into delay list while invalidate in flight.
754 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
755 import in flight. this produce timeout for invalidate and sometimes
756 can cause stale data.
760 Frequency : on ppc only
761 Description: not convert ost objects for directory because it's not exist.
762 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
766 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
768 Severity : enhancement
770 Description: Update to support 2.6.22.14 vanilla kernel.
774 Frequency : rare, on recovery
775 Description: read procfs can produce deadlock in some situation
776 Details : Holding lprocfs lock which send rpc can produce block for destroy
777 obd objects and this also block reconnect with -EALREADY. This
778 isn't fix all lprocfs bugs - but make it rare.
780 Severity : enhancement
782 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
785 Frequency : frequent on X2 node
787 Description: mdc_set_open_replay_data LBUG
788 Details : Set replay data for requests that are eligible for replay.
792 Description: lustre_mgs: operation 101 on unconnected MGS
793 Details : When MGC is disconnected from MGS long enough, MGS will evict the
794 MGC, and late on MGC cannot successfully connect to MGS and a lot
795 of the error messages complaining that MGS is not connected.
799 Frequency : on network error
800 Description: panic with double free request if network error
801 Details : mdc_finish_enqueue is finish request if any network error ocuring,
802 but it's true only for synchronus enqueue, for async enqueue
803 (via ptlrpcd) this incorrect and ptlrpcd want finish request
806 Severity : enhancement
808 Description: client-side metadata stat-ahead during readdir(directory readahead)
809 Details : perform client-side metadata stat-ahead when the client detects
810 readdir and sequential stat of dir entries therein
813 Frequency : on start mds
815 Description: Implement get_info(last_id) in obdfilter.
818 Frequency : occasional
820 Description: Correctly check stale fid, not start epoch if ost not support SOM
821 Details : open with flag O_CREATE need set old fid in op_fid3 because
822 op_fid2 was overwritten with new generated fid, but mds can answer
823 with one of these two fids and both is not stale. Setattr
824 incorrectly started an epoch and assume will be called
825 done_writing, but without SOM done_writing ever being called.
828 Frequency : rare, depends on device drivers and load
830 Description: MDS or OSS nodes crash due to stack overflow
831 Details : Code changes in 1.8.0 increased the stack usage of some functions.
832 In some cases, in conjunction with device drivers that use a lot
833 of stack the MDS (or possibly OSS) service threads could overflow
834 the stack. One change which was identified to consume additional
835 stack has been reworked to avoid the extra stack usage.
838 Frequency : occasional
840 Description: Do not fail import if osc_interpret_create gets -EAGAIN
841 Details : If osc_interpret_create got -EAGAIN it immediately exits and
842 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
843 oscc_has_objects and see OSC has no objests and call
844 oscc_internal_create to resend create request.
846 Severity : enhancement
848 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
850 Severity : enhancement
852 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
855 Frequency : very rare
857 Description: Fix replay if there is an un-replied request and open
858 Details : In some cases, older replay request will revert the
859 mcd->mcd_last_xid on MDS which is used to record the client's
862 Severity : enhancement
864 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
866 Severity : enhancement
868 Description: Add rhel5 support to HEAD.
870 Serverity : enhancement
872 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
877 Description: Don't allow skipping OSTs if index has been specified.
878 Details : Don't allow skipping OSTs if index has been specified, make
879 locking in internal create lots better.
883 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
884 Details : access to bitfield in structure is always rounded to long
885 and this produce problem with not atomic change any bit.
889 Description: Lustre make rpms failed.
890 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
891 builds Lustre rpms from tarball.
894 Frequency : rare on shutdown ost
896 Description: If llog cancel was not send before clean_exports phase, this can
897 produce deadlock in llog code.
898 Details : If llog thread has last reference to obd and call class_import_put
899 this produce deadlock because llog_cleanup_commit_master wait when
900 last llog_commit_thread exited, but this never success because was
901 called from llog_commit_thread.
905 Description: allow userland application know is lost one of stripes.
906 Details : fill lvb_blocks with error code on ost and return it to
907 application if error flag found.
911 Description: NULL lov_tgts causing MDS oops
912 Details : more safe checks for NULL lov_tgts for avoid oops.
914 Severity : enhancement
916 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
920 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
921 Details : If an OST is down the MDS will hang indefinitely in
922 obd_statfs_rqset() waiting for the statfs data. While for
923 MDS QOS usage of statfs, it should not stuck in waiting.
925 Severity : enhancement
927 Description: remote_acl support
928 Details : Support ACL-based permission check for remote user.
929 Support setfacl/getfacl for remote user with the utils
930 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
931 the system "{s,g}etfacl" utils.
933 Severity : enhancement
935 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
937 Severity : enhancement
939 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
943 Description: make mgs_setparam() handle fsname containing dash
944 Details : fsname containing a dash does not work with lctl conf_param
946 Severity : enhancement
948 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
950 Severity : enhancement
952 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
956 Frequency : rare, at shutdown
957 Description: access already free / zero obd_namespace.
958 Details : if client_disconnect_export was called without force flag set,
959 and exist connect request in flight, this can produce access to
960 NULL pointer (or already free pointer) when connect_interpret
961 store ocd flags in obd_namespace.
965 Frequency : only at startup
966 Description: not alloc memory with spinlock held.
967 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
968 if any spinlock held.
970 Severity : enhancement
972 Description: make lustre randomly fail allocating memory
973 Details : Make lustre randomly failed allocating memory for testing purpose.
975 Severity : enhancement
977 Description: lost problems with lov objid file
978 Details : Fixes some scability and access to not inited memory problems
979 in work with lov objdid file.
984 Description: lfs find does not continue on file error
985 Details : Continue other files processing when a file/dir is absent.
989 Description: Inconsistent usage of lustre_pack_reply()
990 Details : Standardize the usage of lustre_pack_reply() such that it
991 always generate a CERROR on failure.
994 Frequency : occasional
996 Description: MDS hang or stay in waiting lock
997 Details : If client receive lock with CBPENDING flag ldlm need send lock
998 cancel as separate rpc, to avoid situation when cancel request
999 can't processed due all i/o threads stay in wait lock.
1003 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1007 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1011 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1015 Description: Update to RHEL4 latest kernel.
1017 Severity : enhancement
1019 Description: Build SLES10 patchless client fails
1020 Details : The configure was broken by run ./configure with
1021 --with-linux-obj=.... argument for patchless client. When the
1022 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1023 can't search header adequately. Use absolute path such as
1024 -I($LINUX)/include instead.
1028 Description: interrupt oig_wait produce painc on resend.
1029 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1030 private set, and this produce situation when rq_ptlrpcd_data not
1031 copyed to new allocated request and triggered LBUG on assert
1032 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1033 ptlrpcd set if request is changed and can be safety changed to use
1038 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1039 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1040 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1041 these macros and replace them with their difinitions instead.
1045 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1046 Details : Update which_patch & target file for SLES10 latest kernel.
1049 Description: Remove client patches from SLES 10 kernel.
1050 Details : This causes SLES 10 clients to behave as patchless clients
1051 even on a Lustre-patched (server) kernel.
1053 Severity : enhancement
1055 Description: self-adjustable client's lru lists
1056 Details : use adaptive algorithm for managing client cached locks lru
1057 lists according to current server load, other client's work
1058 pattern, memory activities, etc. Both, server and client
1059 side namespaces provide number of proc tunables for controlling
1062 Severity : enhancement
1064 Description: light-weight GSS support
1065 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1066 and reduce performance overhead.
1068 Severity : enhancement
1070 Description: Linux keyring support
1071 Details : Support using service of Linux keyring for Lustre GSS internal
1072 context refresh/cache mechanism.
1076 Description: Fix errors in lfs documentation
1077 Details : Fixes man pages
1081 Description: don't use GFP_* in generic Lustre code.
1082 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1086 Description: obdclass is limited by single OBD_ALLOC(idarray)
1087 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1091 Description: warnings with build patchless client with vanila 2.6.19 and up
1092 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1093 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1097 Frequency : Only for SLES
1098 Description: sanity_quota fail test_1
1099 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1100 file, which makes TSTID[2] inunique.
1104 Frequency : Only for SLES9
1105 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1106 Details : Update target/ChangeLog/which_patch .
1110 Frequency : Only for test_10 in sanity-quota.sh
1111 Description: a bug in quota test code
1112 Details : checking a test flag in a wrong place
1116 Frequency : Only for test_13 in sanity-quota.sh
1117 Description: a bug in quota test script
1118 Details : shouldn't check overfull allocated quota
1122 Description: jbd statistics
1123 Details : Port older jbd statistics patch for sles10
1125 Severity : enhancement
1127 Description: Add printing inode info into message about error in writepage.
1132 Description: reply_lock_interpret crash due to race with it and lock cancel.
1133 Details : Do not replay locks that are being cancelled. Do not reference
1134 locks by their address during replay, just by their handle.
1138 Frequency : When flocks are used.
1139 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1140 Details : Flock locks might destroy just granted lock if it could be merged
1141 with another existing flock, this is done in completion handler,
1142 so teach ldlm_cli_enquque_fini that this is a valid case for
1148 Description: Oops in read and write path when failing to allocate lock.
1149 Details : Check if lock allocation failed and return error back.
1153 Description: lstripe command fails for valid OST index
1154 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1155 instead of lov->desc.ld_active_tgt_count.
1159 Description: sanity.sh failed test 103
1160 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1165 Description: df doesn't work properly if diskfs blocksize != 4K
1166 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1170 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1171 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1173 Severity : enhancement
1175 Description: Adaptive timeouts
1176 Details : RPC timeouts adapt to changing server load and network
1177 conditions to reduce resend attempts and improve recovery time.
1181 Description: llapi_file_create() does not allow some changes
1182 Details : add llapi_file_open() that allows specifying the mode and
1183 open flags, and also returns an open file handle.
1187 Description: merge and cleanup kernel patches.
1188 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1192 Description: Add journal checksum support.(Kernel part)
1193 Details : The journal checksum feature adds two new flags i.e
1194 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1195 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1196 indicates that the commit block contains the checksum for
1197 the blocks described by the descriptor blocks. Now commit
1198 record can be sent to disk without waiting for descriptor
1199 blocks to be written to disk. This behavior is controlled
1200 using JBD2_FEATURE_ASYNC_COMMIT flag.
1204 Description: OSS needs mutliple precreate threads
1205 Details : Add ability to start more than one create thread per OSS.
1209 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1210 Details : execute lfs setstripe on client
1214 Description: mds_obd_create error creating tmp object
1215 Details : When the user sets quota on root, llog will be affected and can't
1216 create files and write files.
1219 Frequency : Always on ia64 patchless client, and possibly others.
1221 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1222 Details : This allows the patchless client to be loaded on architectures
1223 without this export.
1227 Description: RedHat Update kernel for RHEL5
1228 Details : Add the kernel config file for RHEL5.
1232 Description: RedHat Update kernel for RHEL5
1233 Details : Modify the kernel config file more closer RHEL5.
1237 Description: Build failure against Centos5 (RHEL5)
1238 Details : Define PAGE_SIZE when it isn't present.
1242 Description: test_7 on sanity-quota.sh can't be used on separate machines
1243 Details : add facet to handle it
1247 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1248 Details : with NFS, the anon dentry's parent was set to itself in
1249 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1250 corresponding dentry other than use rec->ur_name.
1252 Severity : enhancement
1254 Description: lfs setstripe enhancement
1255 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1259 Description: enable data checksumming by default
1260 Details : enable checksum by default, allow --disable-checksum
1261 configure option and "-o nochecksum" mount option. Checksums
1262 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1266 Description: lustre support for RHEL5
1267 Details : Add support for RHEL5.
1271 Description: Client eviction due to failover config
1272 Details : after a connection loss, the lustre client should attempt to
1273 reconnect to the last active server first before trying the
1274 other potential connections.
1278 Description: when mds and osts use different quota unit(32bit and 64bit),
1279 quota will be released repeatly.
1280 Details : void sending multiple quota reqs to mds, which will keep the
1281 status between the reqs.
1285 Description: osts not allocated evenly to files
1286 Details : change the condition to increase offset_idx
1289 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1291 Description: Data corruption for OSTs that are formatted larger than 2TB
1293 Details : When generating the bio request for lustre file writes the
1294 sector number would overflow a temporary variable before being
1295 used for the IO. The data reads correctly from Lustre (which
1296 will overflow in a similar manner) but other file data or
1297 filesystem metadata may be corrupted in some cases.
1301 Description: Tune the kernel for good SCSI performance.
1302 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1303 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1308 Description: rewrite ext2-derived code in obdclass/uuid.c
1309 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1310 from scratch preserving functionality.
1314 Description: "lfs find -obd UUID" prints directories
1315 Details : "lfs find -obd UUID" will return all directory names instead
1316 of just file names. It is incorrect because the directories
1317 do not reside on the OSTs.
1321 Description: use i_size_read and i_size_write in 2.6 port
1322 Details : replace inode->i_size access with i_size_read/write()
1325 Frequency : when removing large files
1327 Description: scheduling issue during removal of large Lustre files
1328 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1329 It causes scheduling issues when removing large files (17TB in the
1333 Frequency : only with liblustre clients on XT3
1335 Description: evictions taking too long
1336 Details : allow llrd to evict clients directly on OSTs
1339 Frequency : only on ppc
1341 Description: /proc/fs/lustre/devices broken on ppc
1342 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1343 We should make sure the type of 'index' is loff_t and then cast
1344 explicitly as needed below. Do not assign an explicitly cast
1348 Frequency : only for rhel5
1350 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1351 Details : Modify the target file & which_kernel.
1353 Severity : enhancement
1355 Description: omit set fsid for export NFS
1356 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1357 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1361 Frequency : Always, for kernels after 2.6.16
1362 Description: Fix warning idr_remove called for id=.. which is not allocated.
1363 Details : Last kernels save old s_dev before kill super and not allow
1364 to restore from callback - restore it before call kill_anon_super.
1368 Description: buffer overruns could theoretically occur
1369 Details : llapi_semantic_traverse() modifies the "path" argument by
1370 appending values to the end of the origin string, and a
1371 overrun may occur. Adding buffer overrun check in liblustreapi.
1375 Description: Fix error on 'ls .' at the top of the Lustre mount.
1376 Details : Don't revalidate dentry if it is a root dentry.
1380 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1381 Details : Modify vm-tunables-rhel4.patch.
1385 Description: Kernel config for 2.6.18-vanilla.
1386 Details : Modify targets/2.6-vanilla.target.in.
1387 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1388 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1389 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1390 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1394 Description: improve handling recoverable errors
1395 Details : if request processig with error which can be recoverable on server
1396 request should be resend, otherwise page released from cache and
1401 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1402 Details : Modify target file & which_patch.
1403 A flaw was found in the IA32 system call emulation provided
1404 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1405 value could be stored in the %RAX register, which could trigger an
1406 out-of-bounds system call table access. An untrusted local user
1407 could exploit this flaw to run code in the kernel
1408 (ie a root privilege escalation). (CVE-2007-4573).
1412 Description: O_DIRECT bypasses client statistics.
1413 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1414 read_ahead_stats were updated. copied the stats section from
1415 osc_send_oap_rpc() into async_internal().
1419 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1423 Description: change order of libsysio includes
1424 Details : '#include sysio.h' should always come before '#include xtio.h'
1428 Description: handle "serious error: objid * already exists" more gracefully
1429 Details : If LAST_ID value on disk is smaller than the objects existing in
1430 the O/0/d* directories, it indicates disk corruption and causes an
1431 LBUG(). If the object is 0-length, then we should use the existing
1432 object. This will help to avoid a full fsck in most cases.
1434 Severity : enhancement
1436 Description: adapt the lustre_config script to support the upgrade case
1437 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1438 server targets to 1.6 in parallel.
1442 Description: To avoid grant space > avaible space when the disk is almost
1443 full. Without this patch you might see the error "grant XXXX >
1444 available" or some LBUG about grant, when the disk is almost
1446 Details : In filter_check_grant, for non_grant cache write, we should
1447 check the left space by if (*left > ungranted + bytes), instead
1448 of (*left > ungranted), because only we are sure the left space
1449 is enough for another "bytes", then the ungrant space should be
1450 increase. In client, we should update cl_avail_grant only there
1451 is OBD_MD_FLGRANT in the reply.
1455 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1456 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1457 described in CVE-2007-4573. This problem could allow untrusted
1458 local users to gain root access.
1461 Frequency : when using O_DIRECT and quotas
1463 Description: Incorrect file ownership on O_DIRECT output files
1464 Details : block usage reported by 'lfs quota' does not take into account
1465 files that have been written with O_DIRECT.
1470 Description: touch file failed when fs is not full
1471 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1472 otherwise we can get ENOSP while fs is not full.
1476 Description: parallel lock callbacks
1477 Details : Instead of sending blocking and completion callbacks as separated
1478 requests, adding them to a set and sending in parallel.
1481 Frequency : only for Cray XT3
1482 Bugzilla : 12829/13455
1483 Description: Changing primary group doesn't change the group lustre assigns to
1485 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1486 overridden with the primary group provided by the group upcall,
1487 whereas the supplied fsgid can be trusted if it is in the list of
1488 supplementary groups returned by the group upcall.
1490 Severity : enhancement
1492 Description: Allow masking D_WARNING, D_ERROR messages from console
1493 Details : Console messages can now be disabled via lnet.printk.
1497 Description: User code with malformed file open parameter crashes client node
1498 Details : Before packing join_file req, all the related reference should be
1499 checked carefully in case some malformed flags cause fake
1500 join_file req on client.
1504 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1505 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1506 mode is equal to the granted mode. However, it should also check
1507 that LDLM_FL_CP_REQD is not set.
1511 Description: Heavy nfs access might result in deadlocks
1512 Details : After ELC code landed, it is now improper to enqueue any mds
1513 locks under och_sem, because enqueue might want to decide to
1514 cancel open locks for same inode we are holding och_sem for.
1518 Description: Client eviction while running blogbench
1519 Details : A lot of unlink operations with concurrent I/O can lead to a
1520 deadlock causing evictions. To address the problem, the number of
1521 oustanding OST_DESTROY requests is now throttled to
1522 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1523 callbacks are processed in priority.
1527 Description: enable ACLs on MDS by default
1528 Details : ACLs must be enabled on MDS by default.
1531 Frequency : PPC/PPC64 only
1533 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1534 Details : fix duplicated definitions between asm-ppc64/types.h and
1535 lustre_types.h on PPC.
1538 Frequency : PPC/PPC64 only
1540 Description: asm-ppc/segment.h does not exist
1541 Details : fix compile issue on PPC.
1545 Description: better handle error messages in extents code
1548 Frequency : RHEL4 only
1550 Description: mkfs is very slow on IA64/RHEL4
1551 Details : A performance regression has been discovered in the MPT Fusion
1552 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1553 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1554 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1556 Severity : enhancement
1558 Description: SNMP support enhancement
1559 Details : Adding total number of sampled request for an MDS node in snmp
1562 Severity : enhancement
1564 Description: Optimize ldlm waiting list processing for PR extent locks
1565 Details : When processing waiting list for read extent lock and meeting read
1566 lock that is same or wider to it that is not contended, skip
1567 processing rest of the list and immediatelly return current
1568 status of conflictness, since we are guaranteed there are no
1569 conflicting locks in the rest of the list.
1573 Description: Time out and refuse to reconnect
1574 Details : When the failover node is the primary node, it is possible
1575 to have two identical connections in imp_conn_list. We must
1576 compare not conn's pointers but NIDs, otherwise we can defeat
1577 connection throttling.
1581 Description: port llog fixes from b1_6 into HEAD
1582 Details : Port llog reference couting and some llog cleanups from b1_6
1583 (bug 10800) into HEAD, for protect from panic and access to already
1584 free llog structures.
1588 Description: Detect stride IO mode in read-ahead
1589 Details : When a client does stride read, read-ahead should detect that and
1590 read-ahead pages according to the detected stride pattern.
1594 Description: data checksumming impacts single node performance
1595 Details : add support for several checksum algorithm. Currently, only CRC32
1596 and Adler-32 are supported. The checksum type can be changed on
1597 the fly via /proc/fs/lustre/osc/*/checksum_type.
1601 Description: use adler32 for page checksums
1602 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1607 Description: build for x2 fails
1608 Details : fix compile issue on Cray systems.
1612 Description: Properly match for duplicate locks
1613 Details : Due to different lock order from skiplists code, we need to
1614 traverse entire list for now
1617 Frequency : only on PPC/SLES10
1619 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1620 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1621 be defined. Add a hack in mkfs_lustre.c to work around this compile
1626 Description: LASSERT on MDS when client holding flock lock dies
1627 Details : ldlm pool logic depends on number of granted locks equal to
1628 number of released locks which is not true for flock locks, so
1629 just exclude such locks from consideration.
1633 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1634 Details : Use fsfilt_sync() for both the whole filesystem sync and
1635 individual file sync to eliminate dangerous inode locking
1636 with I_LOCK that can lead to a deadlock.
1640 Description: performance in 1.6.3
1641 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1642 sure that 1MB requests are not fragmented by the block layer.
1644 Severity : enhancement
1646 Description: organize the server-side client stats on per-nid basis
1647 Details : Change the structure of stats under obdfilter and mds to
1655 The "uuid"s file would list the uuids of _active_ exports.
1656 And the clear entry is to clear all stats and stale nids.
1658 Severity : enhancement
1660 Description: eliminate client locks in face of contention
1661 Details : file contention detection and lockless i/o implementation
1662 for contended files.
1666 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1669 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1671 Description: MDS may lose file striping (and hence file data) in some cases
1672 Details : If there are additional extended attributes stored on the MDS,
1673 in particular ACLs, SELinux, or user attributes (if user_xattr
1674 is specified for the client mount options) then there is a risk
1675 of attribute loss. Additionally, the Lustre file striping
1676 needs to be larger than default (e.g. striped over all OSTs),
1677 and an additional attribute must be stored initially in the
1678 inode and then increase in size enough to be moved to the
1679 external attribute block (e.g. ACL growing in size) for file
1684 Description: skiplist implementation simplification
1685 Details : skiplists are used to group compatible locks on granted list
1686 that was implemented as tracking first and last lock of each
1687 lock group the patch changes that to using doubly linked lists
1691 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1692 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1693 because LASSERT on both the data supplied by a client, and the
1694 data on disk is dangerous and incorrect.
1696 Severity : enhancement
1698 Description: Slow truncate/writes to huge files at high offsets.
1699 Details : Directly associate cached pages to lock that protect those pages,
1700 this allows us to quickly find what pages to write and remove
1701 once lock callback is received.
1705 Description: more ldlm soft lockups
1706 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1707 starve other threads from the resource lock for a long time in
1708 case of long waiting queue, so change the debug level from
1709 D_OTHER to the less frequently used D_INFO.
1711 Severity : enhancement
1713 Description: add -gid, -group, -uid, -user options to lfs find
1717 Description: Hung threads in invalidate_inode_pages2_range
1718 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1719 once one is completed. As a result, some RPCs are stuck in the
1720 queue and are never sent.
1724 Description: filter threads hungs on waiting journal commit
1725 Details : Cleanup filter group llog code, then only filter group llog will
1726 be only created in the MDS/OST syncing process.
1730 Description: Procfs and llog threads access destoryed import sometimes.
1731 Details : Sync the import destoryed process with procfs and llog threads by
1732 the import refcount and semaphore.
1734 Severity : enhancement
1736 Description: openlock cache of b1_6 port to HEAD
1741 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1742 Details : If insertion of an extent fails, then discard the inode
1743 preallocation and free data blocks else it can lead to duplicate
1748 Description: don't always update ctime in ext3_xattr_set_handle()
1749 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1750 In some cases the ctime should not be updated, for example for
1751 2.0->1.8 compatibility it is necessary to delete an xattr and it
1752 should not update the ctime.
1756 Bugzilla : 15713/16362
1757 Description: Assertion in iopen_connect_dentry in 1.6.3
1758 Details : looking up an inode via iopen with the wrong generation number can
1759 populate the dcache with a disconneced dentry while the inode
1760 number is in the process of being reallocated. This causes an
1761 assertion failure in iopen since the inode's dentry list contains
1762 both a connected and disconnected dentry.
1766 Description: assertion failure in ldlm_handle2lock()
1767 Details : fix a race between class_handle_unhash() and class_handle2object()
1768 introduced in lustre 1.6.5 by bug 13622.
1773 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1774 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1779 Description: oops in sd_iostats_seq_show()
1780 Details : unloading/reloading the scsi low level driver triggers a kernel
1781 bug when trying to access the sd iostat file.
1786 Description: Kernel panics during QLogic driver reload
1787 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1788 patch, causing memory corruption.
1793 Description: journal_dev option does not work in b1_6
1794 Details : pass mount option during pre-mount.
1796 Severity : enhancement
1798 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1799 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1800 extent information of a file. It can be used to map logical blocks
1801 in a file to physical blocks in the block device.
1805 Description: LDLM soft lockups - improvement
1806 Details : It is be possible to send the lock handle along with each read
1807 or write request because the client is already doing a lock match
1808 itself so there isn't any reason the OST should have to re-do that
1812 Frequency : only on Cray X2
1814 Description: X2 build failures
1815 Details : fix build failures on Cray X2.
1819 Description: xid & resent requests
1820 Details : Initialize RPC XID from clock at startup (randomly if clock is
1823 Severity : enhancement
1825 Description: Add lustre_start utility to start or stop multiple Lustre servers
1830 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1831 Details : In case of memory pressure, list_del() can be called twice on
1832 req->rq_history_list, causing a kernel oops.
1836 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1837 Details : fix stack overflow in the distributed lock manager by defering
1838 export eviction after a failed AST to the elt thread instead of
1839 handling it in the dlm interpret routine.
1843 Description: Convert some comments to new format.
1844 Details : Update documenting comments to match doxygen conventions.
1848 Description: Grammar fixes.
1849 Details : A couple of trivial spelling fixes.
1853 Description: OSD_COUNTERS-mandatory
1854 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1855 to simplify the code.
1859 Description: simplify cmm_device freeing logic.
1860 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1864 Description: Add lockdep support to dt_object_operations locking interface.
1865 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1866 parameter indicating lock ordering. Update mdd code to use new
1871 Description: Introduce failloc constants for lockless IO tests.
1872 Details : Add two new failloc constants to test lockless IO. Only one of
1873 them in implemented---another is checked in yet to be landed
1878 Description: Add lockdep support for inode mutex.
1879 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1880 in the situations where Lustre has to lock more than one inode
1885 Description: Add optional invariants checking support.
1886 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1887 configure switch. This macro is to be used for consistency and
1888 sanity checks that are too expensive to be left in `production'
1893 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1894 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1899 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1900 Details : Use spin_lock_nested() in (the only) situation where more than
1901 one ldlm_lock is locked simultaneously. Also, fix possible
1902 dead-lock in ldlm_lock_change_resource() by enforcing particular
1907 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1908 Details : Instead of specifying each ldlm_lock call-back through separate
1909 parameter, wrap them into struct ldlm_callback_suite.
1913 Description: Kill join_lru obd method and its callers.
1914 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1915 in memory---a requirement implemented through ->o_join_lru() obd
1916 method in HEAD. Remove this method and its users.
1920 Description: Add asynchronous ldlm ENQUEUE completion handler.
1921 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1922 case that stock ldlm_completion_ast() cannot handle as it waits
1923 until lock is granted. Introduce new ldlm_completion_ast_async()
1924 for this. Also comment ldlm_completion_ast().
1928 Description: ldlm_error <-> errno conversion.
1929 Details : Add functions to map (rather arbitrary) between LDLM error codes
1930 and standard errno values. CLIO needs this to prevent LDLM specific
1931 constants from escaping ldlm and osc.
1935 Description: Kill unused ldlm_handle2lock_ns() function.
1936 Details : Kill unused ldlm_handle2lock_ns() function.
1940 Description: Add lu_ref support to ldlm_lock
1941 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1942 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1943 and ldlm_resource. LDLM interface has to be changed, because code that
1944 releases a reference on a lock, has to "know" what reference this is.
1945 In the most frequent case
1947 lock = ldlm_handle2lock(handle);
1949 LDLM_LOCK_PUT(lock);
1951 no changes are required. When any other reference (received _not_ from
1952 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1953 instead of LDLM_LOCK_PUT().
1955 Arguably, changes are pervasive, and interface requires some discipline
1956 for proper use. On the other hand, it was very instrumental in finding
1957 a few leaked lock references.
1961 Description: Add ldlm_lock_addref_try().
1962 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1963 attempts to addref a lock that might be being canceled
1968 Description: Add ldlm_weigh_callback().
1969 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1970 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1971 hard-coded `number of pages' logic.
1975 Description: Add lockdep annotations to llog code.
1976 Details : Use appropriately tagged _nested() locking calls in the places
1977 where llog takes more than one ->lgh_lock lock.
1981 Description: Add loi_kms_set().
1982 Details : Wrap kms updates into a helper function.
1986 Description: Constify instances of struct lsm_operations.
1987 Details : Constify instances of struct lsm_operations.
1991 Description: lu_conf support.
1992 Details : On a server, a file system object is uniquely identified
1993 by a fid, which is sufficient to locate and load all object
1994 state (inode). On a client, on the other hand, more data are
1995 necessary instantiate an object. Change lu_object_find() and
1996 friends to take additional `lu_conf' argument describing object.
1997 Typically this includes layout information.
2001 Description: lu_context fixes.
2002 Details : Introduce new lu_context functions that are needed on the client
2003 side, where some system threads (ptlrpcd) are shared by multiple
2004 modules, and so cannot be stopped during module shutdown.
2008 Description: Add start and stop methods to lu_device_type_operations.
2009 Details : Introduce two new methods in lu_device_type_operations, that are
2010 invoked when first instance of a given type is created and last one
2011 is destroyed respectively. This is need by CLIO.
2015 Description: Add lu_ref support to struct lu_device.
2016 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2017 track leaked references.
2021 Description: Introduce lu_kmem_descr.
2022 Details : lu_kmem_descr and its companion interface allow to create
2023 and destroy a number of kmem caches at once.
2027 Description: Fix lu_object finalization race.
2028 Details : Fix a race between lu_object_find() finding an object and its
2029 concurrent finalization. This race is (most likely) not possible
2030 on the server, but might happen on the client.
2034 Description: Introduce lu_ref interface.
2035 Details : lu_ref is a debugging module allowing to track references to
2036 a given object. It is quite cpu expensive, and has to be
2037 explicitly enabled with --enable-lu_ref. See usage description
2042 Description: Factor lu_site procfs stats into a separate function.
2043 Details : Separate lu_site stats printing code into a separate function
2044 that can be reused on a client.
2048 Description: Constify instances of struct {lu,dt,md}_device_operations.
2049 Details : Constify instances of struct {lu,dt,md}_device_operations.
2053 Description: Introduce struct md_site and move meta-data specific parts of
2054 struct lu_site here.
2055 Details : Move md-specific fields out of struct lu_site into special struct
2056 md_site, so that lu_site can be used on a client.
2060 Description: Kill mdd_lov_destroy().
2061 Details : Remove unused mdd code.
2065 Description: Add st_block checking to multistat.c.
2066 Details : Add st_block checking to multistat.c.
2070 Description: Add lu_ref support to struct obd_device.
2071 Details : Add lu_ref tracking to obd_device.
2075 Description: Kill obd_set_fail_loc().
2076 Details : Remove unused code.
2080 Description: Add special type for ptlrpc_request interpret functions.
2081 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2082 there. Actual usage will be in CLIO.
2086 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2087 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2088 doesn't work with lockdep.
2092 Description: Add rwv.c test program.
2093 Details : New testing program exercising readv(2) and writev(2) (Qian).
2097 Description: Add sendfile.c test program.
2098 Details : New testing program exercising sendfile(2) (Jay).
2102 Description: Ratelimit a message that can be very frequent.
2103 Details : Ratelimit a memory allocation failure message that can
2108 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2109 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2113 Description: Liblustre build fixes.
2114 Details : Liblustre build fixes.
2118 Description: libcfs: add cfs_{need,cond}_resched() interface.
2119 Details : libcfs: add cfs_{need,cond}_resched() definition and
2120 implementations for Linux, NT, and liblustre.
2122 Severity : enhancement
2124 Description: More exported tunables for mballoc
2125 Details : Add support for tunable preallocation window and new tunables for
2126 large/small requests
2130 Description: Detect corruption of block bitmap and checking for preallocations
2131 Details : Checks validity of on-disk block bitmap. Also it does better
2132 checking of number of applied preallocations. When corruption is
2133 found, it turns filesystem readonly to prevent further corruptions.
2137 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2138 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2139 window to 0. If it does hit cache miss, and read-pattern is still
2140 stride-io mode, does not reset the stride window, but also does not
2141 increase the stride window length in this case.
2145 Frequency : only for big-endian servers
2146 Description: Check if system is big-endian while mounting fs with extents feature
2147 Details : Mounting a filesystem with extents feature will fail on big-endian
2148 systems since ext3-based ldiskfs is not supported on big-endian
2149 systems. This can be over-riden with "bigendian_extents" mount option.
2151 Severity : enhancement
2153 Description: The root squash functionality
2154 Details : A security feature, which is to prevent users from being able
2155 to mount lustre on their desktop, run as root, and delete
2156 all of the files in the filesystem. The goal is accomplished by
2157 remapping user id (UID) and group id (GID) of the root user to
2158 a UID and GID specified by the system administartor via Lustre
2159 configuration management server (MGS). The functionality also
2160 allows to specify sets of clients for which the remapping does
2165 Description: Excessive recovery window
2166 Details : With AT enabled, the recovery window can be excessively long (6000+
2167 seconds). To address this problem, we no longer use
2168 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2169 timeout no longer depends on the service time, it is set to
2170 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2171 time via pb_service_time.
2175 Description: Watchdog triggered on MDS failover
2176 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2177 the OSTs know that the MDS "UUID" can be reused for the same export
2178 from a different NID, so we do not need to wait for the export to be
2182 Frequency : rare, only if using MMP with Linux RAID
2184 Description: MMP doesn't work with Linux RAID
2185 Details : While using HA for Lustre servers with Linux RAID, it is possible
2186 that MMP will not detect multiple mounts. To make this work we
2187 need to unplug the device queue in RAID when the MMP block is being
2188 written. Also while reading the MMP block, we should read it from
2189 disk and not the cached one.
2191 Severity : enhancement
2193 Description: open file using fid
2194 Details : A file can be opened using just its fid, like
2195 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2198 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2201 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2202 Details : add proper locking for creation of jbd_4k slab cache
2206 Description: MMP check in ext3_remount() fails without displaying any error
2207 Details : When multiple mount protection fails during remount, proper error
2210 Severity : enhancement
2212 Description: Allow stripe size to be up to 4G-64k
2213 Details : Fix math logic to allow large stripe sizes.
2217 Description: add check for >8TB ldiskfs filesystems
2218 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2219 Don't allow >8TB ldiskfs filesystems to be mounted without
2220 force_over_8tb mount option
2225 Descriptoin: OSS complains "too many blocks" and crashes
2226 Details : In filter_commitrw_write():
2227 1. lquota_chkquota();
2228 2. obdo_from_inode();
2229 3. lquota_pending_commit();
2230 "2." has changed uid and gid in this case, then caused
2231 lquota_chkquota() and lquota_pending_commit() to use
2232 different uid and gid. That is the root of the bug.
2236 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2237 Details : Detect and print out the 'stripe count = -1' case properly.
2239 --------------------------------------------------------------------------------
2241 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2243 * Support for kernels:
2244 2.6.5-7.283 (SLES 9),
2245 2.6.9-55.EL (RHEL 4),
2246 2.6.16.46-0.14 (SLES 10),
2247 2.6.18.8 vanilla (kernel.org)
2248 * Client support for unpatched kernels:
2249 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2250 2.6.16 - 2.6.22 vanilla (kernel.org)
2251 * Due to recently discovered recovery problems, we do not recommend
2252 using patchless RHEL 4 clients with this or any earlier release.
2253 * Recommended e2fsprogs version: 1.39.cfs8
2254 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2255 * Starting with this release, the ldiskfs backing filesystem required
2256 by Lustre is now in its own package, lustre-ldiskfs. This package
2257 should be installed. It is versioned separately from Lustre and
2258 may be released separately in future.
2262 Description: block reactivating mgc import until all deactivates complete
2263 Details : Fix race when failing back MDT/MGS to itself (testing)
2266 Frequency : at statup only
2268 Description: mds_lov_synchronize race leads to various problems
2269 Details : simultaneous MDT->OST connections at startup can cause the
2270 sync to abort, leaving the OSC in a bad state.
2272 Severity : enhancement
2274 Description: add optional extra BUILD_VERSION info
2275 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2276 to override the lustre version.
2279 Frequency : 2.6.18 servers only
2281 Description: ll_kern_mount() doesn't release the module reference
2282 Details : The ldiskfs module reference count never drops down to 0
2283 because ll_kern_mount() doesn't release the module reference.
2288 Description: server LBUG when using old ost_num_threads parameter
2289 Details : Accept the old ost_num_threads parameter but warn that it
2290 is deprecated, and fix an off-by-one error that caused an LBUG.
2295 Description: Transient SCSI error results in persistent IO issue
2296 Details : iobuf->dr_error is not reinitialized to 0 between two
2300 Frequency : sometimes when underlying device returns I/O errors
2302 Description: OSTs not going read-only during write failures
2303 Details : OSTs are not remounted read-only when the journal commit threads
2304 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2305 instead of the ext3 wrappers.
2309 Description: poor connect scaling with increasing client count
2310 Details : Don't run filter_grant_sanity_check for more than 100 exports
2311 to improve scaling for large numbers of clients.
2314 Frequency : SLES10 only
2316 Description: sanity-quota.sh quotacheck failed: rc = -22
2317 Details : Quotas cannot be enabled on SLES10.
2320 Frequency : liblustre clients only
2322 Description: getdirentries does not give error when run on compute nodes
2323 Details : getdirentries does not fail when the size specified as an argument
2324 is too small to contain at least one entry
2326 Severity : enhancement
2328 Description: Add LNET router traceability for debug purposes
2329 Details : If a checksum failure occurs with a router as part of the
2330 IO path, the NID of the last router that forwarded the bulk data
2331 is printed so it can be identified.
2336 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2337 Details : Due to a race condition, liblustre clients were occasionally
2338 evicted incorrectly.
2340 Severity : enhancement
2342 Description: lfs setstripe use optional parameters instead of postional
2345 Severity : enhancement
2347 Description: Nanosecond timestamp support for ldiskfs
2348 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2349 resolution timestamps. There is not yet support for this at
2350 the Lustre filesystem level.
2353 Frequency : during server recovery
2355 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2356 Details : request with rq_no_resend flag not awake l_wait_event if they get
2360 Frequency : nfs export on patchless client
2362 Description: connectathon hang when test nfs export over patchless client
2363 Details : Disconnected dentry cannot be found with lookup, so we do not need
2364 to unhash it or make it invalid
2367 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2368 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2369 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2374 Description: open req refcounting wrong on reconnect
2375 Details : If reconnect happened between getting open reply from server and
2376 call to mdc_set_replay_data in ll_file_open, we will schedule
2377 replay for unreferenced request that we are about to free.
2378 Subsequent close will crash in variety of ways.
2379 Check that request is still eligible for replay in
2380 mdc_set_replay_data().
2385 Description: disable writes to filesystem when reading health_check file
2386 Details : the default for reading the health_check proc file has changed
2387 to NOT do a journal transaction and write to disk, because this
2388 can cause reads of the /proc file to hang and block HA state
2389 checking on a healthy but otherwise heavily loaded system. It
2390 is possible to return to the previous behaviour during configure
2391 with --enable-health-write.
2393 Severity : enhancement
2395 Description: 64-bit inode version
2396 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2397 to the inode. This will be required for version-based recovery.
2402 Description: MDS fails to start if a duplicate client export is detected
2403 Details : in some rare cases it was possible for a client to connect to
2404 an MDS multiple times. Upon recovery the MDS would detect this
2405 and fail during startup. Handle this more gracefully.
2407 Severity : enhancement
2409 Description: Add -o localflock option to simulate old noflock
2411 Details : This will achieve local-only flock/fcntl locks
2417 Description: log_commit_thread vs filter_destroy race leads to crash
2418 Details : Take import reference before releasing llog record semaphore
2423 Description: Wrong request locking in request set processing
2424 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2425 imp_delayed_list, in this place should be used imp_lock.
2428 Frequency : when reconnection
2430 Description: Grant Leak when osc reconnect to OST
2431 Details : When osc reconnect ost, OST(filter) should check whether it
2432 should grant more space to client by comparing fed_grant and
2433 cl_avail_grant, and return the granted space to client instead
2434 of "new granted" space, because client will call osc_init_grant
2435 to update the client grant space info.
2438 Frequency : when client reconnect to OST
2440 Description: Grant Leak when osc do resend and replay bulk write
2441 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2442 bulk write request, because the grant info while be sync between
2443 OSC and OST when reconnect, and we should ignore the grant info
2444 these of resend/replay write req.
2449 Description: Grant space more than avaiable left space sometimes.
2450 Details : When then OST is about to be full, if two bulk writing from
2451 different clients came to OST. Accord the avaliable space of the
2452 OST, the first req should be permitted, and the second one
2453 should be denied by ENOSPC. But if the seconde arrived before
2454 the first one is commited. The OST might wrongly permit second
2455 writing, which will cause grant space > avaiable space.
2458 Frequency : when client is evicted
2460 Description: Grant might be wrongly erased when osc is evicted by OST
2461 Details : when the import is evicted by server, it will fork another
2462 thread ptlrpc_invalidate_import_thread to invalidate the
2463 import, where the grant will be set to 0. While the original
2464 thread will update the grant it got when connecting. So if
2465 the former happened latter, the grant will be wrongly errased
2466 because of this race.
2471 Description: Checking Stale with correct fid
2472 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2473 to check whether it is stale, because sometimes, we want the
2474 enqueue happened anyway, and op_data.fid2 will not be initialized.
2476 Severity : enhancement
2478 Description: update patchless client
2479 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2482 Frequency : only with 2.4 kernel
2484 Description: random memory corruption
2485 Details : size of struct ll_inode_info is to big for union inode.u and this
2486 can be cause of random memory corruption.
2491 Description: Memory leak in recovery
2492 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2493 It should also check obd_fail before fsfilt_start, otherwise if
2494 fsfilt_start return -EROFS,(failover mds during mds recovery).
2495 then the req will return with repmsg->transno = 0 and rc = EROFS.
2496 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2497 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2498 be freed no matter whether fsfilt_commit success or not.
2501 Frequency : only with huge count clients
2503 Description: Prevents from taking the superblock lock in llap_from_page for
2505 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2506 for llap_from_page call in ll_removepage prevents from taking the
2507 superblock lock for a soon died page.
2512 Description: Not check open intent error before release open handle
2513 Details : in some rare cases, the open intent error is not checked before
2514 release open handle, which may cause
2515 ASSERTION(open_req->rq_transno != 0), because it tries to release
2516 the failed open handle.
2521 Description: Set cat log bitmap only after create log success.
2522 Details : in some rare cases, the cat log bitmap is set too early. and it
2523 should be set only after create log success.
2527 Description: Accessing a block bevice can re-enable I/O when Lustre is
2528 tearing down a device.
2529 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2534 Description: service threads may hog cpus when there are a lot of requests
2535 Details : Insert cond_resched to give other threads a chance to use some CPU
2540 Description: the cat log was not initialized in recovery
2541 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2542 unlink log on mds will not be initialized until mds post
2543 recovery. And also in mds post recovery, the unlink log will
2544 initialization will be done asynchronausly, so there will be race
2545 between add unlink log and unlink log initialization.
2549 Description: brw_stats were being printed incorrectly
2550 Details : brw_stats were being printed as log2 but all of them were not
2551 recorded as log2. Also remove some code duplication arising from
2552 filter_tally_{read,write}.
2556 Frequency : rare, only in recovery.
2557 Description: ASSERTION(req->rq_type != LI_POISON) failed
2558 Details : imp_lock should be held while iterating over imp_sending_list for
2559 prevent destroy request after get timeout in ptlrpc_queue_wait.
2563 Description: replay-single.sh test 52 fails
2564 Details : A lock's skiplist need to be cleanup when it being unlinked
2565 from its resource list.
2569 Description: Short directio read returns full requested size rather than
2571 Details : Direct I/O operations should return actual amount of bytes
2572 transferred rather than requested size.
2574 Severity : enhancement
2576 Description: metadata RPC reduction (e.g. for rm performance)
2577 Details : decrease the amount of synchronous RPC between clients and servers
2578 by canceling conflicing lock before the operation on the client
2579 and packing thier handles into the main operation RPC to server.
2581 Severity : enhancement
2583 Description: Async OSC create to avoid the blocking unnecessarily.
2584 Details : If a OST has no remain object, system will block on the creating
2585 when need to create a new object on this OST. Now, ways use
2586 pre-created objects when available, instead of blocking on an
2587 empty osc while others are not empty. If we must block, we block
2588 for the shortest possible period of time.
2592 Description: improve handling recoverable errors
2593 Details : if request processig with error which can be recoverable on server
2594 request should be resend, otherwise page released from cache and
2597 Severity : enhancement
2599 Description: refine locking for avoid write wrong info into lov_objid file
2600 Details : fix possible races with add new target and write/update data in
2603 --------------------------------------------------------------------------------
2605 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2610 Frequency : on some architectures
2612 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2613 Details : Uninitialized flags sometimes cause configuration commands to
2617 Frequency : patchless clients only
2619 Description: missing __iget() symbol export
2620 Details : The __iget() symbol export is missing. To avoid the need for
2621 this on patchless clients the deathrow inode reaper is turned
2622 off, and we depend on the VM to clean up old inodes. This
2623 dependency was during via the fix for bug 12181.
2625 --------------------------------------------------------------------------------
2627 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2629 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2630 INTEROPERATE with older versions automatically. In many cases a
2631 special upgrade step is needed. Please read the
2632 user documentation before upgrading any part of a live system.
2633 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2634 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2635 * WARNING: Lustre configuration and startup changes are required with
2636 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2640 Severity : enhancement
2642 Description: MountConf
2643 Details : Lustre configuration is now managed via mkfs and mount
2644 commands instead of lmc and lconf. New obd types (MGS, MGC)
2645 are added for dynamic configuration management. See
2646 https://mail.clusterfs.com/wikis/lustre/MountConf for
2649 Severity : enhancement
2651 Description: dynamic OST addition
2652 Details : OSTs can now be added to a live filesystem
2654 Severity : enhancement
2656 Description: startup order invariance
2657 Details : MDTs and OSTs can be started in any order. Clients only
2658 require the MDT to complete startup.
2660 Severity : enhancement
2662 Description: parallel, asynchronous orphan cleanup
2663 Details : orphan cleanup is now performed in separate threads for each
2664 OST, allowing parallel non-blocking operation.
2666 Severity : enhancement
2668 Description: optimized stripe assignment
2669 Details : stripe assignments are now made based on ost space available,
2670 ost previous usage, and OSS previous usage, in order to try
2671 to optimize storage space and networking resources.
2673 Severity : enhancement
2675 Description: Permanently set tunables
2676 Details : All writable /proc/fs/lustre tunables can now be permanently
2677 set on a per-server basis, at mkfs time or on a live system.
2679 Severity : enhancement
2681 Description: Lustre message v2
2682 Details : Add lustre message format v2.
2684 Severity : enhancement
2686 Description: client OST exclusion list
2687 Details : Clients can be started with a list of OSTs that should be
2688 declared "inactive" for known non-responsive OSTs.
2692 Description: SPEC SFS validation failure on NFS v2 over lustre.
2693 Details : Changes the blocksize for regular files to be 2x RPC size,
2694 and not depend on stripe size.
2696 Severity : enhancement
2698 Description: Multiple MD RPCs in flight.
2699 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2700 To avoid overly-overloading MDS, introduce a limit on number of
2701 MDS RPCs in flight for a single client and add /proc controls
2702 to adjust this limit.
2704 Severity : enhancement
2706 Description: client read/write statistics
2707 Details : Add client read/write call usage stats for performance
2708 analysis of user processes.
2709 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2710 file access. extents_stats shows chunk size distribution.
2711 extents_stats_per_process show chunk size distribution per
2714 Severity : enhancement
2716 Description: mds statistics
2717 Details : Add detailed mds operations statistics in
2718 /proc/fs/lustre/mds/*/stats.
2722 Description: Failure of copying files with lustre special EAs.
2723 Details : Client side always return success for setxattr call for lustre
2724 special xattr (currently only "trusted.lov").
2729 Description: Refcount LNET uuids
2730 Details : The global LNET uuid list grew linearly with every startup;
2731 refcount repeated list entries instead of always adding to
2734 Severity : enhancement
2736 Description: Dynamic service threads
2737 Details : Within a small range, start extra service threads
2738 automatically when the request queue builds up.
2741 Frequency : mixed-endian client/server environments
2743 Description: mixed-endian crashes
2744 Details : The new msg_v2 system had some failures in mixed-endian
2747 Severity : enhancement
2749 Description: Easy OST removal
2750 Details : OSTs can be permanently deactivated with e.g. 'lctl
2751 conf_param lustre-OST0001.osc.active=0'
2753 Severity : enhancement
2755 Description: MGS proc entries
2756 Details : Added basic proc entries for the MGS showing what filesystems
2759 Severity : enhancement
2761 Description: provide MGS failover
2762 Details : Added config lock reacquisition after MGS server failover.
2764 Severity : enhancement
2766 Description: add Linux 2.4 support
2767 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2771 Description: a large application tries to do I/O to the same resource and dies
2772 in the middle of it.
2773 Details : Check the req->rq_arrival time after the call to
2774 ost_brw_lock_get(), but before we do anything about
2775 processing it & sending the BULK transfer request. This
2776 should help move old stale pending locks off the queue as
2777 quickly as obd_timeout.
2780 Frequency : when an incorrect nid is specified during startup
2782 Description: ptlrpc connect to non-existant node causes kernel crash
2783 Details : LNET can't be re-entered from an event callback, which
2784 happened when we expire a message after the export has been
2785 cleaned up. Instead, hand the zombie cleanup off to another
2788 Severity : enhancement
2790 Description: plain/inodebits lock performance improvement
2791 Details : Grouping plain/inodebits in granted list by their request modes
2792 and bits policy, thus improving the performance of search through
2796 Frequency : only if OST filesystem is corrupted
2798 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2799 Details : for a short time RPCs with bulk IO are in the replay list,
2800 but replay of bulk IOs is unimplemented. If the OST filesystem
2801 is corrupted due to disk cache incoherency and then replay is
2802 started it is possible to trip an assertion. Avoid putting
2803 committed RPCs into the replay list at all to avoid this issue.
2806 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2808 Description: System hang on startup
2809 Details : This bug allowed the liblustre (e.g. catamount) client to
2810 return to the app before handling all startup RPCs. This
2811 could leave the node unresponsive to lustre network traffic
2812 and manifested as a server ptllnd timeout.
2814 Severity : enhancement
2816 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2817 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2818 boolean to control whether to print peer debug info when a
2819 client's RPC times out.
2822 Frequency : only for kernels with patches from Lustre below 1.4.3
2824 Description: Remove old rdonly API
2825 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2828 Frequency : only for devices with external journals
2830 Description: Set external device read-only also
2831 Details : During a commanded failover stop, we set the disk device
2832 read-only while the server shuts down. We now also set any
2833 external journal device read-only at the same time.
2836 Frequency : when upgrading from 1.4 while trying to change parameters
2838 Description: The wrong (new) MDC name was used when setting parameters for
2839 upgraded MDT's. Also allows changing of OSC (and MDC)
2840 parameters if --writeconf is specified at tunefs upgrade time.
2843 Frequency : when setting specific ost indicies
2845 Description: QOS code breaks on skipped indicies
2846 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2847 created with --index need not be sequential.
2849 Severity : enhancement
2851 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2852 Details : The uninit_groups feature works in conjunction with the kernel
2853 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2854 up the pass1 processing of e2fsck. This is a read-only feature
2855 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2856 mount filesystems that have had this feature enabled.
2858 Severity : enhancement
2860 Description: Improve multi-block allocation algorithm to avoid fragmentation
2861 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2862 allocation locality and avoid filesystem fragmentation.
2864 ------------------------------------------------------------------------------
2866 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2868 * Support for kernels:
2869 2.4.21-47.0.1.EL (RHEL 3)
2870 2.6.5-7.283 (SLES 9)
2871 2.6.9-42.0.10.EL (RHEL 4)
2872 2.6.12.6 vanilla (kernel.org)
2873 2.6.16.27-0.9 (SLES 10)
2874 * Recommended e2fsprogs version: 1.39.cfs5
2876 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2880 Frequency : occasional, depends on client load and configuration
2881 Bugzilla : 12181, 12203
2882 Description: data loss for recently-modified files
2884 Details : In some cases it is possible that recently written or created
2885 files may not be written to disk in a timely manner (this should
2886 normally be within 30s unless client IO load is very high).
2887 The problem appears as zero-length files or files that are a
2888 multiple of 1MB in size after a client crash or client eviction
2889 that are missing data at the end of the file.
2891 This problem is more likely to be hit on clients where files are
2892 repeatedly created and unlinked in the same directory, clients
2893 have a large amount of RAM, have many CPUs, the filesystem has
2894 many OSTs, the clients are rebooted frequently, and/or the files
2895 are not accessed by other nodes after being written.
2897 The presence of the problem can be detected by looking at
2898 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2899 smaller than the second (nr_unused) then dirty files will not
2900 be flushed automatically to disk. "sync; sleep 10" should be
2901 run several times on the node before unmounting it to update
2902 Lustre (this is also safe to run on nodes without this problem).
2904 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2905 that can cause this same problem, so customers using that kernel
2906 also need to update the kernel in addition to Lustre. In order
2907 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2909 It is normal that files written just before a client crash (less
2910 than 30s) may not yet have been flushed to disk, even for local
2914 Frequency : frequent on thin XT3 nodes
2916 Description: UUID collision on thin XT3 Linux nodes
2917 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2918 randomly, since we relied on an insufficiently-seeded PRNG.
2923 Description: OSS hangs after "All ost request buffers busy"
2924 Details : A deadlock between quota and journal operations caused OSS
2925 hangs after printing "All ost request buffers busy."
2928 Frequency : always on liblustre builds
2930 Description: Cleanup compiler warnings on liblustre
2933 Frequency : always on liblustre builds on XT3
2935 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2939 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2940 > 32000 subdirectories
2941 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2942 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2943 subdirectory count crosses 32000. This will aid e2fsck to
2944 correctly handle more than 32000 subdirectories.
2947 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2949 Description: System hang on startup
2950 Details : This bug allowed the liblustre (e.g. catamount) client to
2951 return to the app before handling all startup RPCs. This
2952 could leave the node unresponsive to lustre network traffic
2953 and manifested as a server ptllnd timeout.
2955 Severity : enhancement
2957 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2958 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2959 boolean to control whether to print peer debug info when a
2960 client's RPC times out.
2965 Description: make O_SYNC working on 2.6 kernels
2966 Details : 2.6 kernels use different method for mark pages for write,
2967 so need add a code to lustre for O_SYNC work.
2972 Description: Failure to close file and release space on NFS
2973 Details : Put inode details into lock acquired in ll_intent_file_open.
2974 Use mdc_intent_lock in ll_intent_open to properly
2975 detect all kind of errors unhandled by mdc_enqueue.
2980 Description: proc file read during shutdown sometimes raced obd removal,
2982 Details : Add lock to prevent obd access after proc file removal.
2985 Frequency : Only for files larger than 4GB on 32-bit clients.
2987 Description: improperly doing page alignment of locks
2988 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2989 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2994 Description: RPCs being resent when they shouldn't be
2995 Details : Some RPCs that should not be resent are being resent. This
2996 can cause inconsistencies in the RPC state machine. Do not
2997 resend such requests.
3000 Frequency : rare, only with NFS export
3002 Description: Crash on NFS re-export node
3003 Details : under very unusual load conditions an assertion is hit in
3004 ll_intent_file_open()
3007 Frequency : only if OST filesystem is corrupted
3009 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3010 Details : for a short time RPCs with bulk IO are in the replay list,
3011 but replay of bulk IOs is unimplemented. If the OST filesystem
3012 is corrupted due to disk cache incoherency and then replay is
3013 started it is possible to trip an assertion. Avoid putting
3014 committed RPCs into the replay list at all to avoid this issue.
3019 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3020 Details : Large single O_DIRECT read and write calls can fail to allocate
3021 a sufficiently large buffer to process the request. In case of
3022 allocation failure the allocation is retried with a smaller
3023 buffer and broken into smaller requests.
3025 Severity : enhancement
3027 Description: Add -o localflock option to simulate old noflock behaviour.
3028 Details : This will achieve local-only flock/fcntl locks coherentness.
3033 Description: versioning check is incomplete
3034 Details : Checking the version difference of client vs. server, report
3035 error if the gap is too big.
3040 Description: add support PG_writeback bit
3041 Details : add support for PG_writeback bit for Lustre, for more carefull
3042 work with page cache in 2.6 kernel. This also fix some deadlocks
3043 and remove hack for work O_SYNC with 2.6 kernel.
3045 Severity : enhancement
3047 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3048 Details : The uninit_groups feature works in conjunction with the kernel
3049 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3050 up the pass1 processing of e2fsck. This is a read-only feature
3051 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3052 mount filesystems that have had this feature enabled.
3054 Severity : enhancement
3056 Description: Improve multi-block allocation algorithm to avoid fragmentation
3057 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3058 allocation locality and avoid filesystem fragmentation.
3060 ------------------------------------------------------------------------------
3062 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3064 * Support for kernels:
3065 2.6.9-42.0.3.EL (RHEL 4)
3066 2.6.5-7.276 (SLES 9)
3067 2.4.21-47.0.1.EL (RHEL 3)
3068 2.6.12.6 vanilla (kernel.org)
3069 2.6.16.21-0.8 (SLES10)
3070 * Recommended e2fsprogs version: 1.39.cfs2-0
3072 * The backwards-compatible /proc/sys/portals symlink has been removed
3073 in this release. Before upgrading, please ensure that you change
3074 any configuration scripts or /etc/sysctl.conf files that access
3075 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3076 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3077 in advance of the upgrade on any system running Lustre 1.4.6 or
3078 newer, since /proc/sys/lnet was added in that version.
3079 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3083 Frequency : only when quota is used
3085 Description: avoid scanning export list for quota master
3086 Details : Change the algorithms to avoid scanning export list in order
3087 to improve the efficiency.
3090 Frequency : MDS failover only, very rarely
3092 Description: "went back in time" messages on mds failover
3093 Details : The greatest transno may be lost when the current operation
3094 finishes with an error (transno==0) and the client's last_rcvd
3095 record is over-written. Save the greatest transno in the
3096 mds_last_transno for this case.
3099 Frequency : always for specific kernels and striping counts
3101 Description: client may get "Matching packet too big" without ACL support
3102 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3103 when trying to access files in certain configurations. The
3104 clients should in fact be denied when mounting because they do
3105 not understand ACLs.
3108 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3110 Description: many clients connecting with IO in progress causes connect timeouts
3111 Details : Avoid synchronous journal commits to avoid delays caused by many
3112 clients connecting/disconnecting when bulk IO is in progress.
3113 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3114 OST_IO_PORTAL to avoid delays behind potentially many pending
3118 Frequency : occasionally with multiple writers to a single file
3120 Description: shared writes to file may result in wrong size reported by stat()
3121 Details : Allow growing of kms when extent lock is cancelled
3124 Frequency : always with random mmap IO to multi-striped file
3126 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3127 Details : Only if the hole is at the end of OST object so that kms is too
3128 small. Fix is to increase kms accordingly in ll_nopage.
3131 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3133 Description: writes to a missing object would leak memory on the OST
3134 Details : If there is an inconsistency between the MDS and OST filesystems,
3135 such that the MDS references an object that doesn't exist, writes
3136 to that object will leak memory due to incorrect cleanup in the
3137 error handling path, eventually running out of memory on the OST.
3142 Description: Creating too long symlink causes lustre errors
3143 Details : Check symlink and name lengths before sending requests to MDS.
3146 Frequency : only if flock is enabled (not on by default)
3148 Description: posix locks not released on fd closure on 2.6.9+
3149 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3150 kernels, this caused such locks not to be released on fd close and
3151 then assertions on fs unmount about still used locks.
3154 Frequency : MDS failover only, very rarely
3156 Description: clients may get ASSERTION(granted_lock != NULL)
3157 Details : When request was taking a long time, and a client was resending
3158 a getattr by name lock request. The were multiple lock requests
3159 with the same client lock handle and
3160 mds_getattr_name->fixup_handle_for_resent_request found one of the
3161 lock handles but later failed with ASSERTION(granted_lock != NULL).
3166 Description: handle->h_buffer_credits > 0, assertion failure
3167 Details : h_buffer_credits is zero after truncate, causing assertion
3168 failure. This patch extends the transaction or creates a new
3172 Frequency : NFS re-export or patchless client
3173 Bugzilla : 11179, 10796
3174 Description: Crash on NFS re-export node (__d_move)
3175 Details : We do not want to hash the dentry if we don't have a lock.
3176 But if this dentry is later used in d_move, we'd hit uninitialised
3177 list head d_hash, so we just do this to init d_hash field but
3178 leave dentry unhashed.
3181 Frequency : NFS re-export or patchless client
3183 Description: NFS exports has problem with symbolic link
3184 Details : lustre client didn't properly install dentry when re-exported
3185 to NFS or running patchless client.
3188 Frequency : NFS re-export or patchless client
3190 Description: Various nfs/patchless fixes.
3191 Details : fixes reuse disconected alias for lookup process - this fixes
3192 warning "find_exported_dentry: npd != pd",
3193 fix permission error with open files at nfs.
3194 fix apply umask when do revalidate.
3197 Frequency : occasional
3199 Description: Crash on NFS re-export node
3200 Details : calling clear_page() on the wrong pointer triggered oops in
3201 generic_mapping_read().
3204 Frequency : rarely, using O_DIRECT IO
3206 Description: unaligned directio crashes client with LASSERT
3207 Details : check for unaligned buffers before trying any requests.
3210 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3212 Description: stale data returned from RAID cache
3213 Details : If only a small amount of IO is done to the RAID device before
3214 reading it again it is possible to get stale data from the RAID
3215 cache instead of reading it from disk.
3218 Frequency : always for sles10 kernel
3220 Description: sles10 support
3221 Details : ll_follow_link: compile fixes and using of nd_set_link
3222 under newer kernels.
3225 Frequency : depends on arch, kernel and compiler version, always on sles10
3228 Description: recursive or deep enough symlinks cause stack overflow
3229 Details : getting rid of large stack-allocated variable in
3233 Frequency : depends on hardware
3235 Description: lustre write performance loss in the SLES10 kernel
3236 Details : the performance loss is caused by using of write barriers in the
3237 ext3 code. The SLES10 kernel turns barrier support on by
3238 default. The fix is to undo that change for ldiskfs.
3240 ------------------------------------------------------------------------------
3242 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3244 * Support for kernels:
3245 2.6.9-42.0.3EL (RHEL 4)
3246 2.6.5-7.276 (SLES 9)
3247 2.4.21-47.0.1.EL (RHEL 3)
3248 2.6.12.6 vanilla (kernel.org)
3252 Frequency : quota enabled and large files being deleted
3254 Description: releasing more than 4GB of quota at once hangs OST
3255 Details : If a user deletes more than 4GB of files on a single OST it
3256 will cause the OST to spin in an infinite loop. Release
3257 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3262 Description: statfs data retrieved from /proc may be stale or zero
3263 Details : When reading per-device statfs data from /proc, in the
3264 {kbytes,files}_{total,free,avail} files, it may appear
3265 as zero or be out of date.
3268 Frequency : systems with MD RAID1 external journal devices
3270 Description: lconf's call to blkid is confused by RAID1 journal devices
3271 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3272 of returning all block devices that match the journal UUID.
3275 Frequency : always, for aggregate stripe size over 4GB
3277 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3278 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3279 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3282 Frequency : always if "lfs find" used on a local file/directory
3284 Description: "lfs find" segfaults if used on a local file/directory
3285 Details : The case where a directory component was not specified wasn't
3286 handled correctly. Handle this properly.
3289 Frequency : always on ppc64
3291 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3292 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3293 perform endianness conversion.
3296 Frequency : rarely (truncate to non-zero file size after write under load)
3297 Bugzilla : 10730, 10687
3298 Description: Files padded with zeros to next 4K multiple
3299 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3300 it is possible that files that are truncated to a non-zero size
3301 immediately after being written are filled with zero bytes beyond
3302 the truncated size. No file data is lost.
3304 Severity : enhancement
3306 Description: Allow recovery/failover for liblustre clients.
3307 Details : liblustre clients were unaware of failover configurations until
3310 Severity : enhancement
3312 Description: user file locks should fail when not mounting with flock option
3313 Details : Set up an error-returning stub in ll_file_operations.lock field
3314 to prevent incorrect behaviour when client is mounted without
3315 flock option. Also, set up properly f_op->flock field for
3319 Frequency : always on ia64
3321 Description: "lfs df" loops on printing out MDS statfs information
3322 Details : The obd_ioctl_data was not initialized and in some systems
3323 this caused a failure during the ioctl that did not return
3324 an error. Initialize the struct and return an error on failure.
3327 Frequency : SLES 9 only
3329 Description: Error of copying files with lustre special EAs as root
3330 Details : Client side always return success for setxattr call for lustre
3331 special xattr (currently only "trusted.lov").
3334 Frequency : rarely on clusters with both ia64+i386 clients
3336 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3337 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3338 clients are doing concurrent writes to the same file it is
3339 possible that the ia64 clients may LASSERT because the OST
3340 extent locks are not PAGE_SIZE aligned. Ensure that grown
3341 locks are always aligned on the request boundary.
3344 Frequency : specific use, occasional
3346 Description: Overwriting in use executable truncates on-disk binary image
3347 Details : If one node attempts to overwrite an executable in use by
3348 another node, we now correctly return ETXTBSY instead of
3349 truncating the file.
3351 Severity : enhancement
3353 Description: Async OSC create to avoid the blocking unnecessarily.
3354 Details : If a OST has no remain object, system will block on the creating
3355 when need to create a new object on this OST. Now, ways use
3356 pre-created objects when available, instead of blocking on an
3357 empty osc while others are not empty. If we must block, we block
3358 for the shortest possible period of time.
3363 Description: chmod on Lustre root is propagated to other clients
3364 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3365 invalid by the follow_mount time.
3368 Frequency : liblustre clients only
3370 Description: Race in 'instant cancel' lock handling could lead to such locks
3371 never to be granted in case of SMP MDS
3372 Details : Do not destroy not yet granted but cbpending locks in
3376 Frequency : replay/resend of open
3378 Description: non null lock assetion failure in mds_intent_policy
3379 Details : Trying to replay/resend lockless open requests resulted in
3380 mds_open() returning 0 with no lock. Now it sets a flag if
3381 a lock is going to be returned.
3383 Severity : enhancement
3385 Description: Checksum enhancements
3386 Details : New checksum enhancements allow for resending RPCs that failed
3389 Severity : enhancement
3391 Description: Tunables on number of dirty pages in cacche
3392 Details : Allow to set limit on number of dirty pages cached.
3397 Description: client crash on unmount - lock still has references
3398 Details : In some error handling cases it was possible to leak a lock
3399 reference on a client while accessing a file. This was not
3400 harmful to the client during operation, but would cause the
3401 client to crash when the filesystem is unmounted.
3404 Frequency : specific case, rare
3406 Description: ETXTBSY on mds though file not in use by client
3407 Details : ETXTBSY is no longer incorrectly returned when attempting to
3408 chmod or chown a directory that the user previously tried to
3409 execute or a currently-executing binary.
3412 Frequency : extremely rare except on liblustre-based clients
3414 Description: Lustre space not freed when files are deleted
3415 Details : Clean up open-unlinked files after client eviction. Previously
3416 the unlink was skipped and the files remained as orphans.
3421 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3422 Details : In some cases it was possible to send a blocking callback to
3423 a client doing a glimpse, even though that client didn't get
3424 a lock granted. When the glimpse lock is cancelled on the OST
3425 the freed lock is left on the waiting list and corrupted the list.
3428 Frequency : all core dumps
3430 Description: Broke core dumps to lustre
3431 Details : Negative dentry may be unhashed if parent does not have UPDATE
3432 lock, but some callers, e.g. do_coredump, expect dentry to be
3433 hashed after successful create, hash it in ll_create_it.
3435 ------------------------------------------------------------------------------
3437 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3439 * Support for kernels:
3440 2.6.9-42.0.2.EL (RHEL 4)
3441 2.6.5-7.276 (SLES 9)
3442 2.4.21-40.EL (RHEL 3)
3443 2.6.12.6 vanilla (kernel.org)
3447 Frequency : always on RHEL 3
3449 Description: Number of open files grows over time
3450 Details : The number of open files grows over time, whether or not
3451 Lustre is started. This was due to a filp leak introduced
3452 by one of our kernel patches.
3454 ------------------------------------------------------------------------------
3456 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3458 * Support for kernels:
3459 2.6.9-42.EL (RHEL 4)
3460 2.6.5-7.276 (SLES 9)
3461 2.4.21-40.EL (RHEL 3)
3462 2.6.12.6 vanilla (kernel.org)
3467 Bugzilla : 5719, 9635, 9792, 9684
3468 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3469 Details : If a server is under heavy load and cannot reply to new
3470 connection requests before the client resends the (re)connect,
3471 the connection handling code can behave badly if two service
3472 threads are concurrently handing separate (re)connections from
3473 the same client. Add better locking to the connection handling
3474 code, and ensure that only a single connection will be processed
3475 for a given client UUID, even if the lock is dropped.
3477 Severity : enhancement
3479 Description: add TCP zero-copy support to kernel
3480 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3481 sends if the hardware supports scatter-gather and checksumming.
3482 This allows socklnd to do client-write and server-read more
3483 efficiently and reduce CPU utilization from skbuf copying.
3486 Frequency : only if NFS exporting from client
3488 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3489 Details : If setting attributes on a file created under NFS that had
3490 never been opened it would be possible to oops the client
3491 if the file had no objects.
3494 Frequency : always for liblustre
3496 Description: liblustre client does MDS+OSTs setattr RPC for each write
3497 Details : When doing a write from a liblustre client, the client
3498 incorrectly issued an RPC to the MDS and each OST the file was
3499 striped over in order to update the timestamps. When writing
3500 with small chunks and many clients this could overwhelm the MDS
3501 with RPCs. In all cases it would slow down the write because
3502 these RPCs are unnecessary.
3504 Severity : enhancement
3506 Description: allow number of MDS service threads to be changed at module load
3507 Details : It is now possible to change the number of MDS service threads
3508 running. Adding "options mds mds_num_threads={N}" to the MDS's
3509 /etc/modprobe.conf will set the number of threads for the next
3510 time Lustre is restarted (assuming the "mds" module is also
3511 reloaded at that time). The default number of threads will
3512 stay the same, 32 for most systems.
3517 Description: OST crash if filesystem is unformatted or corrupt
3518 Details : If an OST is started on a device that has never been formatted
3519 or if the filesystem is corrupt and cannot even mount then the
3520 error handling cleanup routines would dereference a NULL pointer.
3525 Description: NULL pointer deref in llap_from_page.
3526 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3527 incorrect) mapping if the page was truncated/reclaimed while it was
3528 searched for. Check for this condition and skip such pages when
3529 doing readahead. Introduce extra check to llap_from_page() to
3530 verify page->mapping->host is non-NULL (so page is not anonymous).
3533 Frequency : Sometimes when using sys_sendfile
3535 Description: "page not covered by a lock" warnings from ll_readpage
3536 Details : sendfile called ll_readpage without right page locks present.
3537 Now we introduced ll_file_sendfile that does necessary locking
3538 around call to generic_file_sendfile() much like we do in
3542 Frequency : with certain MDS communication failures at client mount time
3544 Description: NULL pointer deref after failed client mount
3545 Details : a client connection request may delayed by the network layer
3546 and not be sent until after the PTLRPC layer has timed out the
3547 request. If the client fails the mount immediately it will try
3548 to clean up before the network times out the request. Add a
3549 reference from the request import to the obd device and delay
3550 the cleanup until the network drops the request.
3553 Frequency : occasionally during client (re)connect
3555 Description: assertion failure during client (re)connect
3556 Details : processing a client connection request may be delayed by the
3557 client or server longer than the client connect timeout. This
3558 causes the client to resend the connection request. If the
3559 original connection request is replied in this interval, the
3560 client may trip an assertion failure in ptlrpc_connect_interpret()
3561 which thought it would be the only running connect process.
3564 Frequency : only with obd_echo servers and clients that are rebooted
3566 Description: kernel BUG accessing uninitialized data structure
3567 Details : When running an obd_echo server it did not start the ping_evictor
3568 thread, and when a client was evicted an uninitialized data
3569 structure was accessed. Start the ping_evictor in the RPC
3570 service startup instead of the OBD startup.
3572 Severity : enhancement
3573 Bugzilla : 10193 (patchless)
3574 Description: Remove dependency on various unexported kernel interfaces.
3575 Details : No longer need reparent_to_init, exit_mm, exit_files,
3576 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3579 Frequency : rare (only users of deprecated and unsupported LDAP config)
3581 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3582 Details : LDAP apparently contains 'lustreName' attributes instead of
3583 'name'. A simple remapping of the name is sufficient.
3586 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3588 Description: waiting_locks_callback trips kernel BUG if client is evicted
3589 Details : Running with the dump_on_timeout debug flag turned on makes
3590 it possible that the waiting_locks_callback() can try to dump
3591 the Lustre kernel debug logs from an interrupt handler. Defer
3592 this log dumping to the expired_lock_main() thread.
3594 Severity : enhancement
3596 Description: Support NFS exporting on 2.6 kernels.
3597 Details : Implement non-rawops metadata methods for NFS server to use without
3598 changing NFS server code.
3601 Frequency : very rare (synthetic metadata workload only)
3603 Description: two racing renames might cause an MDS thread to deadlock
3604 Details : Running the "racer" program may cause one MDS thread to rename
3605 a file from being the source of a rename to being the target of
3606 a rename at exactly the same time that another thread is doing
3607 so, and the second thread has already enqueued these locks after
3608 doing a lookup of the target and is trying to relock them in
3609 order. Ensure that we don't try to re-lock the same resource.
3612 Frequency : only very large systems with liblustre clients
3614 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3615 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3616 all OSTs in parallel. This allows the eviction of stale liblustre
3617 clients to proceed much faster than if they were done in series,
3618 and also offers similar improvements for other set_info RPCs.
3623 Description: excessive CPU usage during initial read phase on client
3624 Details : During the initial read phase on a client, it would agressively
3625 retry readahead on the file, consuming too much CPU and impacting
3626 performance (since 1.4.5.8). Improve the readahead algorithm
3627 to avoid this, and also improve some other common cases (read
3628 of small files in particular, where "small" is files smaller than
3629 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3634 Description: MDS crash when receiving packet with unknown intent.
3635 Details : Do not LBUG in unknown intent case, just return -EFAULT
3637 Severity : enhancement
3638 Bugzilla : 9293, 9385
3639 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3640 Details : Do not serialize getattr (non-intent version) and statfs.
3643 Frequency : occasional, when OST network is overloaded/intermittent
3645 Description: client evicted by OST after bulk IO timeout
3646 Details : If a client sends a bulk IO request (read or write) the OST
3647 may evict the client if it is unresposive to its data GET/PUT
3648 request. This is incorrect if the network is overloaded (takes
3649 too long to transfer the RPC data) or dropped the OST GET/PUT
3650 request. There is no need to evict the client at all, since
3651 the pinger and/or lock callbacks will handle this, and the
3652 client can restart the bulk request.
3655 Frequency : Always when mmapping file with no objects
3657 Description: client crashes when mmapping file with no objects
3658 Details : Check that we actually have objects in a file before doing any
3659 operations on objects in ll_vm_open, ll_vm_close and
3665 Description: Request leak when working with deleted CWD
3666 Details : Introduce advanced request refcount tracking for requests
3667 referenced from lustre intent.
3669 Severity : Enhancement
3671 Description: Cache open file handles on client.
3672 Details : MDS now will return special lock along with openhandle, if
3673 requested and client is allowed to hold openhandle, even if unused,
3674 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3675 closing files for every read/write openration.
3677 Severity : Enhancement
3679 Description: Cache open negative dentries on client when possible.
3680 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3681 negative dentries on lock revocation.
3686 Description: Remounting a client read-only wasn't possible with a zconf mount
3687 Details : It wasn't possible to remount a client read-only with llmount.
3689 Severity : enhancement
3690 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3691 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3692 MPICH is included to add Lustre-specific ADIO interfaces.
3693 This is based closely on the UFS ADIO layer and only differs
3694 in file creation, in order to allow the OST striping to be set.
3695 This is user-contributed code and not supported by CFS.
3700 Description: extended inode attributes (immutable, append-only) work improperly
3701 when 2.4 and 2.6 kernels are used on client/server or vice versa
3702 Details : Introduce kernel-independent values for these flags.
3704 Severity : enhancement
3707 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3708 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3709 granularity. Now, Fractional MB granularity is be supported,
3710 this is very useful for low memory system.
3712 Severity : enhancement
3714 Description: Getattr by fid
3715 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3716 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3717 have locks and avoids getattr rpc requests for it on every stat.
3720 Frequency : Always, for filesystems larger than 2TB
3722 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3723 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3724 filesystems larger than 2TB to be mounted with mballoc enabled.
3727 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3729 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3730 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3731 filesystems larger than 2TB is mounted then the kernel will
3732 silently corrupt the start of the filesystem. CONFIG_LBD is
3733 enabled for all CFS-supported kernels, but the possibility of
3734 this happening with a modified kernel config exists.
3736 Severity : enhancement
3738 Description: add client O_DIRECT support for 2.6 kernels
3739 Details : It is now possible to do O_DIRECT reads and writes to files
3740 in the Lustre client mountpoint on 2.6 kernel clients.
3742 Severity : enhancement
3744 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3745 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3746 parallel, not waiting for response from every OST before sending
3747 a rpc to the next OST.
3752 Description: setattr vs write race when updating file timestamps
3753 Details : Client processes that update a file timestamp into the past
3754 right after writing to the file (e.g. tar) it is possible that
3755 the updated file modification time can be reset to the current
3756 time due to a race between processing the setattr and write RPC.
3758 Severity : enhancement
3760 Description: Bring 'lfs find' closer in line with regular Linux find.
3761 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3762 -print0 options and obtains all the needed info through the lustre
3765 Severity : enhancement
3767 Description: support up to 1024 configured devices on one node
3768 Details : change obd_dev array from statically allocated to dynamically
3769 allocated structs as they are first used to reduce memory usage
3774 Description: Flush dirty partially truncated pages during truncate
3775 Details : Immediatelly flush partially truncated pages in filter_setattr,
3776 this way we completely avoid having any pages in page cache on OST
3777 and can retire ugly workarounds during writes to flush such pages.
3782 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3783 Details : protect inode from truncation within vfs_unlink() context
3784 just take a reference before calling vfs_unlink() and release it
3785 when parent's i_sem is free.
3790 Description: last_id value checked outside lock on OST caused LASSERT failure
3791 Details : If there were multiple MDS->OST object precreate requests in
3792 flight, it was possible that the OST's last object id was checked
3793 outside a lock and incorrectly tripped an assertion. Move checks
3794 inside locks, and discard old precreate requests.
3797 Frequency : always, if extents are used on OSTs
3799 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3800 Details : OSTs using the extents format would not zero the high 16 bits of
3801 the index physical block number. This is not a problem for any
3802 OST filesystems smaller than 16TB, and no kernels support ext3
3803 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3804 new/modified files) and can be fixed for existing filesystems
3805 with e2fsprogs-1.39-cfs1.
3810 Description: import connection selection may be incorrect if timer wraps
3811 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3812 import connections to be ignored if the 32-bit jiffies counter
3813 wraps. Use a 64-bit jiffies counter.
3816 Frequency : very large clusters immediately after boot
3818 Description: LNET request buffers exhausted under heavy short-term load
3819 Details : If a large number of client requests are generated on a service
3820 that has previously never seen so many requests it is possible
3821 that the request buffer growth cannot keep up with the spike in
3822 demand. Instead of dropping incoming requests, they are held in
3823 the LND until the RPC service can accept more requests.
3826 Frequency : Sometimes during replay
3828 Description: Assertion failure in ll_local_open after replay.
3829 Details : If replay happened on an open request reply before we were able
3830 to set replay handler, reply will become not swabbed tripping the
3831 assertion in ll_local_open. Now we set the handler right after
3832 recognising of open request
3835 Frequency : very rare
3837 Description: kernel reports "badness in vsnprintf"
3838 Details : Reading from the "recovery_status" /proc file in small chunks
3839 may cause a negative length in lprocfs_obd_rd_recovery_status()
3840 call to vsnprintf() (which is otherwise harmless). Exit early
3841 if there is no more space in the output buffer.
3843 Severity : enhancement
3845 Description: clear OBD RPC statistics by writing to them
3846 Details : It is now possible to clear the OBD RPC statistics by writing
3847 to the "stats" file.
3852 Description: Client mtime is not the same on different clients after utimes
3853 Details : In some cases, the client was using the utimes() syscall on
3854 a file cached on another node. The clients now validate the
3855 ctime from the MDS + OSTs to determine which one is right.
3860 Description: Inability to activate failout mode
3861 Details : lconf script incorrectly assumed that in pythong string's numeric
3862 value is used in comparisons.
3865 Frequency : always with multiple stripes per file
3867 Description: Inefficient object allocation for mutli-stripe files
3868 Details : When selecting which OSTs to stripe files over, for files with
3869 a stripe count that divides evenly into the number of OSTs,
3870 the MDS is always picking the same starting OST for each file.
3871 Return the OST selection heuristic to the original design.
3876 Description: mount failures may take full timeout to return an error
3877 Details : Under some heavy load conditions it is possible that a
3878 failed mount can wait for the full obd_timeout interval,
3879 possibly several minutes, before reporting an error.
3880 Instead return an error as soon as the status is known.
3882 Frequency : quota enabled and large files being deleted
3884 Description: releasing more than 4GB of quota at once hangs OST
3885 Details : If a user deletes more than 4GB of files on a single OST it
3886 will cause the OST to spin in an infinite loop. Release
3887 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3892 Description: statfs data retrieved from /proc may be stale or zero
3893 Details : When reading per-device statfs data from /proc, in the
3894 {kbytes,files}_{total,free,avail} files, it may appear
3895 as zero or be out of date.
3898 Frequency : systems with MD RAID1 external journal devices
3900 Description: lconf's call to blkid is confused by RAID1 journal devices
3901 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3902 of returning all block devices that match the journal UUID.
3905 Frequency : always, for aggregate stripe size over 4GB
3907 Description: assertion fails when trying to use 4GB stripe size
3908 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3909 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3912 Frequency : always on ppc64
3914 Description: the first write on an ext3 filesystem with mballoc got stuck
3915 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3916 perform endianness conversion.
3918 ------------------------------------------------------------------------------
3920 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3922 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3923 INTEROPERATE with older versions automatically. Please read the
3924 user documentation before upgrading any part of a live system.
3925 * WARNING: Lustre networking configuration changes are required with
3926 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3929 * Support for kernels:
3930 2.6.9-22.0.2.EL (RHEL 4)
3931 2.6.5-7.244 (SLES 9)
3932 2.6.12.6 vanilla (kernel.org)
3935 Severity : enhancement
3936 Bugzilla : 7981/8208
3937 Description: Introduced Lustre Networking (LNET)
3938 Details : LNET is new networking infrastructure for Lustre, it includes
3939 a reorganized network configuration mode (see the user
3940 documentation for full details) as well as support for routing
3941 between different network fabrics. Lustre Networking Devices
3942 (LNDs) for the supported network fabrics have also been
3943 created for this new infrastructure.
3945 Severity : enhancement
3946 Description: Introduced Access control lists
3947 Details : clients can set ACLs on files and directories in order to have
3948 more fine-grained permissions than the standard Unix UGO+RWX.
3949 The MDS must be started with the "-o acl" mount option.
3951 Severity : enhancement
3952 Description: Introduced filesystem quotas
3953 Details : Administrators may now establish per-user quotas on the
3956 Severity : enhancement
3958 Description: Configuration change for the XT3
3959 The PTLLND is now used to run Lustre over Portals on the XT3
3960 The configure option(s) --with-cray-portals are no longer used.
3961 Rather --with-portals=<path-to-portals-includes> is used to
3962 enable building on the XT3. In addition to enable XT3 specific
3963 features the option --enable-cray-xt3 must be used.
3968 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3969 Details : A race between allocating a new llcd and re-getting the llcd_lock
3970 allowed another thread to grab newly-allocated llcd.
3972 Severity : enhancement
3974 Description: 2.6 OST async journal commit and locking fix to improve performance
3975 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3976 2.6 kernels are now async as they already were in 2.4 kernels so
3977 that they can commit concurrently with the network bulk transfer.
3978 For block-allocated files the filter allocation semaphore is held
3979 to avoid filesystem fragmentation during allocation. BKL lock
3980 removed for 2.6 xattr operations where it is no longer needed.
3985 Description: lconf incorrectly determined whether two IP networks could talk
3986 Details : In some more complicated routing and multiple-network
3987 configurations, lconf will avoid trying to make a network
3988 connection to a disjoint part of the IP space. It was doing the
3989 math incorrectly for one set of cases.
3994 Description: Fix for potential infinite loop processing records in an llog.
3995 Details : If an llog record is corrupted/zeroed, it is possible to loop
3996 forever in llog_process(). Validate the llog record length
3997 and skip the remainder of the block on error.
4000 Frequency : occasional (liblustre only)
4002 Description: liblustre could not open files whose last component is a symlink
4003 Details : sysio_path_walk() would incorrectly pass the open intent to
4004 intermediate path components.
4007 Frequency : rare (liblustre only with non-standard tuning)
4008 Bugzilla : 7201 (7350)
4009 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4010 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4011 after it has been released, only do this for referenced locks
4013 Severity : enhancement
4015 Description: specifying an (invalid) directory default stripe_size of -1
4016 would reset the directory default striping
4017 Details : stripe_size -1 was used internally to signal directory stripe
4018 removal, now use "all default" to signal dir stripe removal
4019 as a directory striping of "all default" is not useful
4022 Frequency : common for large clusters running liblustre clients
4024 Description: doing an ls when liblustre clients are running is slow
4025 Details : sending a glimpse AST to a liblustre client waits for every AST
4026 to time out, as liblustre clients will not respond. Since they
4027 cannot cache data we refresh the OST lock LVB from disk instead.
4029 Severity : enhancement
4031 Description: doing an ls at the same time as file IO can be slow
4032 Details : enqueue and other "small" requests can be blocked behind many
4033 large IO requests. Create a new OST IO portal for non-IO
4034 requests so they can be processed faster.
4037 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4039 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4040 created file with mode 000
4041 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4042 allow an ftruncate() as a non-root user to a file with mode 000.
4043 Lustre now respects this flag to disable mode checks when
4044 truncating a file owned by the user
4047 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4050 Description: Revoking locks from clients that went dead or catatonic might take
4052 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4053 cancellation of such locks instant on servers without waiting for
4054 any reply from clients. Clients drops these locks when cancel
4055 notification from server is received without replying.
4058 Frequency : liblustre-only, when liblustre client dies or becomes busy
4060 Description: Doing ls on Linux clients can take a long time with active
4062 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4063 granting such locks to it in the first place if possible. Locks
4064 are taken by proxy on the OST during the read or write and
4065 dropped immediately afterward. Add connect flags handling, do
4066 not grant locks to liblustre clients for glimpse ASTs.
4068 Severity : enhancement
4070 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4071 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4072 sequential reads and behaves badly with applications doing random
4073 reads. Improve it by reading ahead at least read region, and
4074 avoiding excessive large RPC for small reads.
4076 Severity : enhancement
4078 Description: Creating more than 1000 files for a single job may cause a load
4079 imbalance on the OSTs if there are also a large number of OSTs.
4080 Details : qos_prep_create() uses an OST index reseed value that is an
4081 even multiple of the number of available OSTs so that if the
4082 reseed happens in the middle of the object allocation it will
4083 still utilize the OSTs as uniformly as possible.
4088 Description: OST or MDS may oops in ping_evictor_main()
4089 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4090 but doesn't restart at beginning of obd_exports_timed list
4093 Severity : enhancement
4095 Description: improve by-nid export eviction on the MDS and OST
4096 Details : allow multiple exports with the same NID to be evicted at one
4097 time without re-searching the exports list.
4100 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4102 Description: MDS may oops in groups_free()
4103 Details : in rare race conditions a newly allocated group_info struct is
4104 freed again, and this can be NULL. The 2.4 compatibility code
4105 for groups_free() checked for a NULL pointer, but 2.6 did not.
4108 Frequency : common for liblustre clients doing little filesystem IO
4109 Bugzilla : 9352, 7313
4110 Description: server may evict liblustre clients accessing contended locks
4111 Details : if a client is granted a lock or receives a completion AST
4112 with a blocking AST already set it would not reply to the AST
4113 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4116 Frequency : lfs setstripe, only systems with more than 160 OSTs
4118 Description: unable to set striping with a starting offset beyond OST 160
4119 Details : llapi_create_file() incorrectly limited the starting stripe
4120 index to the maximum single-file stripe count.
4123 Frequency : LDAP users only
4125 Description: lconf did not handle in-kernel recovery with LDAP properly
4126 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4128 Severity : enhancement
4130 Description: bind OST threads to NUMA nodes to improve performance
4131 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4132 node and do their allocations there to localize memory access
4134 Severity : enhancement
4136 Description: llmount can determine client NID directly from Myrinet (GM)
4137 Details : the client NID code from gmnalnid was moved directly into
4138 llmount, removing the need to use this or specifying the
4139 client NID explicitly when mounting GM clients with zeroconf
4142 Frequency : if client is started with down MDS
4144 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4145 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4146 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4147 ptlrpc_import_delayed_req() because we didn't check if the
4148 request was interrupted, and we also didn't break out of the
4149 event loop if there was no timeout
4154 Description: data loss during non-page-aligned writes to a single file from
4155 both multiple nodes and multiple threads on one node at same time
4156 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4157 inconsistency led to false short-reads, that were cached and later
4158 used by ->prepare_write() to fill in partially written page,
4159 leading to data loss.
4162 Frequency : always, if lconf --abort_recovery used
4164 Description: lconf --abort_recovery fails with 'Operation not supported'
4165 Details : lconf was attempting to abort recovery on the MDT device and not
4168 Severity : enhancement
4170 Description: remove cleanup logs
4171 Details : replace lconf-generated cleanup logs with lustre internal
4172 cleanup routines. Eliminates the need for client-cleanup and
4175 Severity : enhancement
4177 Description: add support for EAs (user and system) on lustre filesystems
4178 Details : it is now possible to store extended attributes in the Lustre
4179 client filesystem, and with the user_xattr mount option it
4180 is possible to allow users to store EAs on their files also
4182 Severity : enhancement
4184 Description: Add possibility (config option) to show minimal available OST free
4186 Details : When compiled with --enable-mindf configure option, statfs(2)
4187 (and so, df) will return least minimal free space available from
4188 all OSTs as amount of free space on FS, instead of summary of
4189 free spaces of all OSTs.
4191 Severity : enhancement
4193 Description: do not expand extent locks acquired on OST-side
4194 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4195 by server: they are not cached anyway.
4198 Frequency : when mmap is used/binaries executed from Lustre
4200 Description: Unmmap pages before throwing them away from read cache.
4201 Details : llap_shrink cache now attempts to unmap pages before discarding
4202 them (if unmapping failed - do not discard). SLES9 kernel has
4203 extra checks that trigger if this unmapping is not done first.
4208 Description: lconf didn't resolve symlinks before checking to see whether a
4209 given mountpoint was already in use
4212 Frequency : when migrating failover services
4213 Bugzilla : 6395, 9514
4214 Description: When migrating a subset of services from a node (e.g. failback
4215 from a failover service node) the remaining services would
4216 time out and evict clients.
4217 Details : lconf --force (implied by --failover) sets the global obd_timeout
4218 to 5 seconds in order to quickly disconnect, but this caused
4219 other RPCs to time out too quickly. Do not change the global
4220 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4222 Severity : enhancement
4223 Frequency : if MDS is started with down OST
4224 Bugzilla : 9439,5706
4225 Description: Allow startup/shutdown of an MDS without depending on the
4226 availability of the OSTs.
4227 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4228 Add appropriate locking and lov-osc refcounts for safe
4229 cleaning. Add osc abort_inflight calls in case the
4230 synchronize never started.
4233 Frequency : occasional (Cray XT3 only)
4235 Description: root not authorized to access files in CRAY_PORTALS environment
4236 Details : The client process capabilities were not honoured on the MDS in
4237 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4238 been accessed by an authorized user then root was able to access
4239 the file on the local client also. The root user capabilities
4240 are now allowed on the MDS, as this environment has secure UID.
4243 Frequency : occasional
4245 Description: ldiskfs "too long searching" message happens too often
4246 Details : A debugging message (otherwise harmless) prints too often on
4247 the OST console. This has been reduced to only happen when
4248 there are fragmentation problems on the filesystem.
4253 Description: Division by zero in statfs when all OSCs are inactive
4254 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4255 lov_statfs divides by value returned by lov_get_stripecnt().
4259 Bugzilla : 9489, 3273
4260 Description: First write from each client to each OST was only 4kB in size,
4261 to initialize client writeback cache, which caused sub-optimal
4262 RPCs and poor layout on disk for the first writen file.
4263 Details : Clients now request an initial cache grant at (re)connect time
4264 and so that they can start streaming writes to the cache right
4265 away and always do full-sized RPCs if there is enough data.
4266 If the OST is rebooted the client also re-establishes its grant
4267 so that client cached writes will be honoured under the grant.
4272 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4273 Details : Now I/O RPCs go to different portal number and (presumably) fast
4274 lock requests (and glimses) and other RPCs get their own service
4275 threads pool that should be able to service those RPCs
4278 Severity : enhancement
4280 Description: Ability to exchange lustre version between client and servers and
4281 issue warnings at client side if client is too old. Also for
4282 liblustre clients there is ability to refuse connection of too old
4284 Details : New 'version' field is added to connect data structure that is
4285 filled with version info. That info is later checked by server and
4289 Frequency : rare, liblustre only.
4290 Bugzilla : 9296, 9581
4291 Description: Two simultaneous writes from liblustre at offset within same page
4292 might proceed at the same time overwriting eachother with stale
4294 Details : I/O lock withing llu_file_prwv was released too early, before data
4295 actually was hitting the wire. Extended lock-holding time until
4296 server acknowledges receiving data.
4299 Frequency : extremely rare. Never observed in practice.
4301 Description: avoid generating lustre_handle cookie of 0.
4302 Details : class_handle_hash() generates handle cookies by incrementing
4303 global counter, and can hit 0 occasionaly (this is unlikely, but
4304 not impossible, because initial value of cookie counter is
4305 selected randonly). Value of 0 is used as a sentinel meaning
4306 "unassigned handle" --- avoid it. Also coalesce two critical
4307 sections in this function into one.
4309 Severity : enhancement
4311 Description: allow liblustre clients to delegate truncate locking to OST
4312 Details : To avoid overhead of locking, liblustre client instructs OST to
4313 take extent lock in ost_punch() on client's behalf. New connection
4314 flag is added to handle backward compatibility.
4316 Severity : enhancement
4317 Bugzilla : 4928, 7341, 9758
4318 Description: allow number of OST service threads to be specified
4319 Details : a module parameter allows the number of OST service threads
4320 to be specified via "options ost ost_num_threads={N}" in the
4321 OSS's /etc/modules.conf or /etc/modprobe.conf.
4325 Bugzilla : 6146, 9635, 9895
4326 Description: servers crash with bad pointer in target_handle_connect()
4327 Details : In rare cases when a client is reconnecting it was possible that
4328 the connection request was the last reference for that export.
4329 We would temporarily drop the export reference and get a new
4330 one, but this may have been the last reference and the export
4331 was just destroyed. Get new reference before dropping old one.
4333 Severity : enhancement
4334 Frequency : if client is started with failover MDS
4336 Description: Allow multiple MDS hostnames in the mount command
4337 Details : Try to read the configuration from all specified MDS
4338 hostnames during a client mount in case the "primary"
4341 Severity : enhancement
4343 Description: Stop sending data to evicted clients as soon as possible.
4344 Details : Check if the client we are about to send or are sending data to
4345 was evicted already. (Check is done every second of waiting,
4346 for which l_wait_event interface was extended to allow checking
4347 of exit condition at specified intervals).
4350 Frequency : rare, normally only when NFS exporting is done from client
4352 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4354 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4357 Frequency : always, if srand() is called before liblustre initialization
4359 Description: Liblustre uses system PRNG disturbing its usage by user application
4360 Details : Introduce internal to lustre fast and high-quality PRNG for
4361 lustre usage and make liblustre and some other places in generic
4362 lustre code to use it.
4364 Severity : enhancement
4365 Bugzilla : 9477, 9557, 9870
4366 Description: Verify that the MDS configuration logs are updated when xml is
4367 Details : Check if the .xml configuration logs are newer than the config
4368 logs stored on the MDS and report an error if this is the case.
4369 Request --write-conf, or allow starting with --old_conf.
4371 Severity : enhancement
4373 Description: Handle symlinks in the path when checking if Lustre is mounted.
4374 Details : Resolve intermediate symlinks when checking if a client has
4375 mounted a filesystem to avoid duplicate client mounts.
4380 Description: lconf can hit an error exception but still return success.
4381 Details : The lconf command catches the Command error exception at the top
4382 level script context and will exit with the associated exit
4383 status, but doesn't ensure that this exit status is non-zero.
4388 Description: failure of ptlrpc thread startup can cause oops
4389 Details : Starting a ptlrpc service thread can fail if there are a large
4390 number of threads or the server memory is very fragmented.
4391 Handle this without oopsing.
4394 Frequency : always, only if liblustre and non-default acceptor port was used
4396 Description: liblustre cannot connect to servers with non-default acceptor port
4397 Details : tcpnal_set_default_params() was not called and was therefore
4398 ignoring the environment varaible TCPNAL_PORT, as well as other
4399 TCPNAL_ environment variables
4404 Description: two objects could be created on the same OST for a single file
4405 Details : If an OST is down, in some cases it was possible to create two
4406 objects on a single OST for a single file. No problems other
4407 than potential performance impact and spurious error messages.
4411 Bugzilla : 5681, 9562
4412 Description: Client may oops in ll_unhash_aliases
4413 Details : Client dcache may become inconsistent in race condition.
4414 In some cases "getcwd" can fail if the current directory is
4420 Description: Inode refcounting problems in NFS export code
4421 Details : link_raw functions used to call d_instantiate without obtaining
4422 extra inode reference first.
4426 Bugzilla : 9942, 9903
4427 Description: Referencing freed requests leading to crash, memleaks with NFS.
4428 Details : We used to require that call to ll_revalidate_it was always
4429 followed by ll_lookup_it. Also with revalidate_special() it is
4430 possible to call ll_revalidate_it() twice for the same dentry
4431 even if first occurence returned success. This fix changes semantic
4432 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4433 reference on a request referred from the intent.
4434 ll_intent_release() then releases such a request.
4437 Frequency : rare, normally benchmark loads only
4439 Description: unlinked inodes were kept in memory on the client
4440 Details : If a client is repeatedly creating and unlinking files it
4441 can accumulate a lot of stale inodes in the inode slab cache.
4442 If there is no other client load running this can cause the
4443 client node to run out of memory. Instead flush old inodes
4444 from client cache that have the same inode number as a new inode.
4447 Frequency : SLES9 2.6.5 kernel and long filenames only
4448 Bugzilla : 9969, 10379
4449 Description: utime reports stale NFS file handle
4450 Details : SLES9 uses out-of-dentry names in some cases, which confused
4451 the lustre dentry revalidation. Change it to always use the
4455 Frequency : rare, unless heavy write-truncate concurrency is continuous
4456 Bugzilla : 4180, 6984, 7171, 9963, 9331
4457 Description: OST becomes very slow and/or deadlocked during object unlink
4458 Details : filter_destroy() was holding onto the parent directory lock
4459 while truncating+unlinking objects. For very large objects this
4460 may block other threads for a long time and slow overall OST
4461 responsiveness. It may also be possible to get a lock ordering
4462 deadlock in this case, or run out of journal credits because of
4463 the combined truncate+unlink. Solution is to do object truncate
4464 first in one transaction without parent lock, and then do the
4465 final unlink in a new transaction with the parent lock. This
4466 reduces the lock hold time dramatically.
4469 Frequency : rare, 2.4 kernels only
4471 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4472 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4473 before unmounting their filesystem in order to avoid deadlock.
4474 The kernel_locked() function in 2.4 kernels only checks whether
4475 the kernel lock is held, not whether it is this process that is
4476 holding it as 2.6 kernels do.
4481 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4482 Details : The client ptlrpc code may be trying to reconnect to a down
4483 server before a previous connection attempt has timed out.
4484 Increase the reconnect interval to be longer than the connection
4485 timeout interval to avoid sending duplicate connections to
4489 Frequency : echo_client brw_test command
4491 Description: fix echo_client to work with OST preallocated code
4492 Details : OST preallocation code (5137) didn't take echo_client IO path
4493 into account: echo_client calls filter methods outside of any
4494 OST thread and, hence, there is no per-thread preallocated
4495 pages and buffers to use. Solution: hijack pga pages for IO. As
4496 a byproduct, this avoids unnecessary data copying.
4500 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4501 Description: Client can oops in mdc_commit_close() after open replay
4502 Details : It was possible for the MDS to return an open request with no
4503 transaction number in mds_finish_transno() if the client was
4504 evicted, but without actually returning an error. Clients
4505 would later try to replay that open and may trip an assertion
4506 Simplify the client close codepath, and always return an error
4507 from the MDS in case the open is not successful.
4510 Frequency : rare, 2.6 OSTs only
4512 Description: OST may deadlock under high load on fragmented files
4513 Details : If there was a heavy load and highly-fragmented OST filesystems
4514 it was possible to have all the OST threads deadlock waiting on
4515 allocation of biovecs, because the biovecs were not released
4516 until the entire RPC IO was completed. Instead, release biovecs
4517 as soon as they are complete to ensure forward IO progress.
4519 Severity : enhancement
4521 Description: Support for specifying external journal device at mount
4522 Details : If an OST or MDS device is formatted with an external journal
4523 device, this device major/minor is stored in the ext3 superblock
4524 and may not be valid for failover. Allow detecting and
4525 specifying the external journal at mount time.
4530 Description: Mounting an MDS with pending unlinked files may cause oops
4531 Details : target_finish_recovery() calls mds_postrecov() which returned
4532 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4533 considers this an error and immediately begins cleaning up the
4534 lov, just after starting the mds_lov process
4536 Severity : enhancement
4538 Description: Implement 'lfs df' to report actual free space on per-OST basis
4539 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4540 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4541 usage of objects; '-h' to report in human readable format.
4543 ------------------------------------------------------------------------------
4545 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4552 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4553 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4554 at startup during orphan unlinking. Instead, skip buddy update
4555 before setup, as it will be regenerated after recovery is complete.
4560 Description: If an OST is inactive, its locks might reference stale inodes.
4561 Details : lov_change_cbdata() must iterate over all namespaces, even if
4562 they are inactive to clear inode references from the lock.
4564 Severity : enhancement
4565 Frequency : occasional, if non-standard max_dirty_mb used
4567 Description: Client will block write RPCs if not enough grant
4568 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4569 then the client will block writes while waiting for another RPC
4570 to complete instead of consuming its dirty limit. With change
4571 we get improved performance when max_dirty_mb is small.
4573 Severity : enhancement
4574 Bugzilla : 3389, 6253
4575 Description: Add support for supplementary groups on the MDS.
4576 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4577 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4578 lookups for user supplementary groups into a cache.
4582 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4583 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4584 for liblustre O_EXCL is enforced by MDS.
4588 Description: readdir never returns NULL in liblustre.
4589 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4590 dentry in struct dirent.
4594 Frequency : liblustre only, depends on application IO pattern
4595 Description: liblustre clients evicted if not contacting servers
4596 Details : Don't put liblustre clients into the ping_evictor list, so
4597 they will not be evicted by the pinger ever.
4599 Severity : enhancement
4601 Description: Add ability to evict clients by NID from MDS.
4602 Details : By echoing "nid:$NID" string into
4603 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4604 $NID would be instantly evicted from this MDS and from all active
4605 OSTs connected to it.
4609 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4610 Details : lookup_it_finish() used to query file size from OSTs that was not
4615 Description: service threads change working directory to that of init
4616 Details : Starting lustre service threads may pin the working directory
4617 of the parent thread, making that filesystem busy. Threads
4618 now change to the working directory of init to avoid this.
4622 Frequency : during shutdown only
4623 Description: shutdown with a failed MDS or OST can cause unmount to hang
4624 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4629 Frequency : 2.6 only
4630 Description: chmod/chown may include an extra supplementary group
4631 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4632 supplementary group and if none is specified this is used.
4635 Bugzilla : 5479 (6816)
4637 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4638 Details : If lookup is in progress on a file that is unlinked we might try
4639 to revalidate the inode and fail in revalidate after lookup is
4640 complete and ll_file_open() enqueues the open again but
4641 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4644 Frequency : always, if lconf --abort_recovery used
4646 Description: lconf --abort_recovery fails with 'Operation not supported'
4647 Details : lconf was attempting to abort recovery on the MDT device and not
4650 ------------------------------------------------------------------------------
4652 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4657 Frequency : rare (only unsupported configurations with a node running as an
4659 Bugzilla : 6514, 5137
4660 Description: Mounting a Lustre file system on a node running as an OST could
4662 Details : OSTs now preallocates memory needed to write out data at
4663 startup, instead of when needed, to avoid having to
4664 allocate memory in possibly low memory situations.
4665 Specifically, if the file system is mounted on on OST,
4666 memory pressure could force it to try to write out data,
4667 which it needed to allocate memory to do. Due to the low
4668 memory, it would be unable to do so and the node would
4669 become unresponsive.
4671 Severity : enhancement
4673 Description: Addition of lconf --service command line option
4674 Details : lconf now accepts a '--service <arg>' option, which is
4675 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4677 Severity : enhancement
4679 Description: Failover mode is now the default for OSTs.
4680 Details : By default, OSTs will now run in failover mode. To return to
4681 the old behaviour, add '--failout' to the lmc line for OSTs.
4683 Severity : enhancement
4685 Description: Health checks are now provided for MDS and OSTs
4686 Details : Additional detailed health check information on MSD and OSTs
4687 is now provided through the procfs health_check value.
4690 Frequency : occasional, depends on IO load
4692 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4693 after numerous create/delete cycles
4694 Details : The ext3 inode allocation policy would not allocate new inodes
4695 very well on the OSTs because there are no new directories
4696 being created. Instead we look for groups with free space if
4697 the parent directories are nearly full.
4702 Description: Network or server problems during mount may cause partially
4703 mounted clients instead of returning an error.
4704 Details : The config llog parsing code may overwrite the error return
4705 code during mount error handling, returning success instead
4711 Description: MDS can fail to allocate large reply buffers
4712 Details : After long uptimes the MDS can fail to allocate large reply
4713 buffers (e.g. zconf client mount config records) due to memory
4714 fragmentation or consumption by the buffer cache. Preallocate
4715 some large reply buffers so that these replies can be sent even
4716 under memory pressure.
4720 Frequency : rare (liblustre)
4721 Description: fsx running with liblustre complained that using truncate() to
4722 extend the file doesn't work. This patch corrects that issue.
4723 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4724 ATTR_SIZE and lsm use in llu_setattr_raw.
4728 Frequency : rare, only 2.6 kernels
4729 Description: Unusual file access patterns on the MDS may result in inode
4730 data being lost in very rare circumstances.
4731 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4732 code in the 2.6 kernel caused the inode and/or EA data not to
4733 be read from disk, causing single-file corruption.
4737 Frequency : rare, only 2.6 filesystems using extents
4738 Description: Heavy concurrent write and delete load may cause data corruption.
4739 Details : It was possible under high-load situations to have an extent
4740 metadata block in the block device cache from a just-unlinked
4741 file overwrite a newly-allocated data block. We now unmap any
4742 metadata buffers that alias just-allocated data blocks.
4746 Frequency : filesystems with default stripe_count larger than 77
4747 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4748 Details : lconf specifies an inode size of 4096 bytes when the default
4749 stripe_count is larger than 77. This conflicts with the default
4750 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4751 this case to avoid pinning too much memory for large EAs.
4753 ------------------------------------------------------------------------------
4755 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4760 Frequency : rare (extremely heavy IO load with hundreds of clients)
4762 Description: Client is evicted, gets IO error writing to file
4763 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4764 caused clients to be evicted for AST timeouts. The fixes in
4765 bug 5192 mean we no longer need to have such short AST timeouts
4766 so ldlm_timeout has been increased.
4769 Frequency : occasional during --force or --failover shutdown under load
4770 Bugzilla : 5949, 4834
4771 Description: Server oops/LBUG if stopped with --force or --failover under load
4772 Details : a collection of import/export refcount and cleanup ordering
4773 issues fixed for safer force cleanup
4776 Frequency : only filesystems larger than 120 OSTs
4777 Bugzilla : 5990, 6223
4778 Description: lfs getstripe would oops on a very large filesystem
4779 Details : lov_getconfig used kfree on vmalloc'd memory
4782 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4783 Bugzilla : 6242, 6243
4784 Description: reading from files that had been truncated to a non-zero size
4785 but never opened returned no data
4786 Details : ll_file_read() reads zeros from no-object files to EOF
4791 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4792 Details : There's little harm in aborting MDS/OSS recovery and letting it
4793 try again, so I removed the LASSERT and return an error instead.
4795 Severity : enhancement
4797 Description: New debugging infrastructure for tracking down data corruption
4798 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4799 (b) cover more of the client-side code path, and (c) try to narrow
4800 down where problems occurred
4804 Bugzilla : 3819, 4364, 4397, 6313
4805 Description: Racing close and eviction MDS could cause assertion in mds_close
4806 Details : It was possible to get multiple mfd references during close and
4807 client eviction, leading to one thread referencing a freed mfd.
4809 Severity: : enhancement
4810 Bugzilla : 3262, 6359
4811 Description: Attempts to reconnect to servers are now more aggressive.
4812 Details : This builds on the enhanced upcall-less recovery that was added
4813 in 1.4.2. When trying to reconnect to servers, clients will
4814 now try each server in the failover group every 10 seconds. By
4815 default, clients would previously try one server every 25 seconds.
4820 Description: After recovery, certain operations trigger a failed
4821 assertion on a client.
4822 Details : Failing over an mds, using lconf -d --failover, while a
4823 client was doing a readdir() call would cause the client to
4824 LBUG after recovery completed and the readdir() was resent.
4826 Severity : enhancement
4828 Description: Default groups are now added by lconf
4829 Details : You can now run lconf --group <servicename> without having to
4830 manually add groups with lmc.
4833 Frequency : occasional
4835 Description: Nodes with an elan id of 0 trigger a failed assertion
4838 Frequency : always when accessing e.g. tty/console device nodes
4840 Description: tty and some other devices nodes cannot be used on lustre
4841 Details : file's private_data field is used by device data and lustre
4842 values in there got lost. New field was added to struct file to
4843 store fs-specific private data.
4846 Frequency : when exporting Lustre via NFS
4848 Description: NFSD failed occasionally when looking up a path component
4849 Details : NFSD is looking up ".." which was broken in ext3 directories
4850 that had grown large enough to become hashed.
4853 Frequency : Clusters with multiple interfaces not on the same subnet
4855 Description: Nodes will repeatedly try to reconnect to an interface which it
4856 cannot reach and report an error to the log.
4857 Details : Extra peer list entries will be created by lconf with some peers
4858 unreachable. lconf now validates the peer before adding it.
4861 Frequency : Only if a default stripe is set on the filesystem root.
4863 Description: Setting a default stripe on the filesystem root prevented the
4864 filesystem from being remounted.
4865 Details : The client was sending extra request flags in the root getattr
4866 request and did not allocate a reply buffer for the dir EA.
4869 Frequency : occasional, higher if lots of files are accessed by one client
4870 Bugzilla : 6159, 6097
4871 Description: Client trips assertion regarding lsm mismatch/magic
4872 Details : While revalidating inodes the VFS looks up inodes with ifind()
4873 and in rare cases can find an inode that is being freed.
4874 The ll_test_inode() code will free the lsm during ifind()
4875 when it finds an existing inode and then the VFS later attaches
4876 this free lsm to a new inode.
4880 Bugzilla : 6422, 7030
4881 Description: MDS deadlock between mkdir and client eviction
4882 Details : Creating a new file via mkdir or mknod (starting a transaction
4883 and getting the ns lock) can deadlock with client eviction
4884 (gets ns lock and trying to finish a synchronous transaction).
4887 Frequency : occasional
4888 Description: While starting a server, the fsfilt_ext3 module could not be
4890 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4891 kernels. Previously, lconf would still use the ext3 name
4892 when trying to load modules. Now, it will correctly use
4893 ext3 on 2.4 and ldiskfs on 2.6.
4895 Severity : enhancement
4896 Description: The default stripe count has been changed to 1
4897 Details : The interpretation of the default stripe count (0, to lfs
4898 or lmc) has been changed to mean striping across a single
4899 OST, rather than all available. For general usage we have
4900 found a stripe count of 1 or 2 works best.
4902 Severity : enhancement
4903 Description: Add support for compiling against Cray portals.
4904 Details : Conditional compiling for some areas that are different
4908 Frequency : occasional
4909 Bugzilla : 6409, 6834
4910 Description: Creating files with an explicit stripe count may lead to
4911 a failed assertion on the MDS
4912 Details : If some OSTs are full or unavailable, creating files may
4913 trigger a failed assertion on the MDS. Now, Lustre will
4914 try to use other servers or return an error to the
4918 Frequency : occasional
4920 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4921 caused lock timeout during readahead (since 1.4.2).
4922 Details : Processes doing readahead might match a lock that hasn't been
4923 granted yet if there are overlapping and conflicting lock
4924 requests. The readahead process waits on ungranted lock
4925 (original lock is CBPENDING), while OST waits for that process
4926 to cancel CBPENDING read lock and eventually evicts client.
4928 Severity : enhancement
4930 Description: Initial enabling of flock support for clients
4931 Details : Implements fcntl advisory locking and file status functions.
4932 This feature is provided as an optional mount flag (default
4933 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4934 locking are implemented yet, and those that are are not guaranteed
4935 to be completely correct in production environments.
4936 mount -t lustre -o [flock|noflock] ...
4939 Frequency : occasional
4941 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4942 assertion in the ext3 JBD (journaling) layer.
4943 Details : The b_committed_data struct is protected by the big kernel lock
4944 in 2.4 kernels, serializing journal_commit_transaction() and
4945 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4946 access to this struct. In 2.6 kernels there is finer grained
4947 locking to improve SMP performance of the JBD layer.
4951 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4956 Description: merge and cleanup kernel patches.
4959 -----------------------------------------------------------------------------
4961 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4963 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4964 versions of Lustre. Please update all servers and clients to
4965 version 1.4.2 or later at the same time. You must also run
4966 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4967 to update the configuration logs.
4969 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4970 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4971 - swab llog records of type '0' so we get proper header size/idx (5861)
4972 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4973 - fix rename of one directory over another leaking an inode (5953)
4974 - avoid SetPageDirty on 2.6 (5981)
4975 - don't re-add just-being-destroyed locks to the waiting list (5653)
4976 - when creating new directories, inherit the parent's custom
4977 striping settings if present parent (3048)
4978 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4979 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4980 - don't hold client locks on temporary worklist from l_lru (5666)
4981 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4982 - automatically evict dead clients (5921)
4983 - Update file size properly in create+truncate+fstat case (6196)
4984 - Do not unhash mountpoint dentries, do not allow removal of
4986 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4987 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4988 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4989 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4990 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4991 - Correctly handle reads of files with no objects (6243)
4992 - lctl recover will also mark a device active if deactivate used (5933)
4994 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4995 - allow --write-conf on an MDS with different nettype than client (5619)
4996 - don't write config llogs to MDS for mounts not from that MDS (5617)
4997 - lconf should create multiple TCP connections from a client (5201)
4998 - init scripts are now turned off by default; run chkconfig --on
4999 lustre and chkconfig --on lustrefs to use them
5000 - upcalls are no longer needed for clients to recover to failover
5002 - add --abort-recovery option to lconf to abort recovery on device
5004 - add support for an arbitrary number of OSTs (3026)
5005 - Quota support protocol changes.
5006 - forward compatibility changes to wire structs (6007)
5007 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5008 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5009 - improved "lustre status" script
5010 - initialize blocksize for non-regular files (6062)
5011 - added --disable-server and --disable-client configure options (5782)
5012 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5013 - Vanilla 2.4.29 support
5014 - increase maximum number of obd devices to 520 (6242)
5015 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5016 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5017 SLES 9 2.6.5 kernels
5018 - limit stripes per file to 160 (the maximum EA size) (6093)
5020 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5023 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5024 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5025 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5026 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5027 - small ext3 extents cleanups and fixes (5733)
5028 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5029 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5030 - avoid needless client->OST connect, fix handle mismatch (5317)
5031 - fix DLM error path that led to out-of-sync client, long delays (5779)
5032 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5033 - fix several locking issues related to i_size (5492,5624,5654,5672)
5034 - don't move pending lock onto export if it is already evicted (5683)
5035 - fix kernel oops when creating .foo in unlinked directory (5548)
5036 - fix deadlock in obdfilter statistics vs. object create (5811)
5037 - use time_{before,after} to avoid timer jiffies wrap (5882)
5038 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5039 - Do not leak request if server was not able to process it (5154)
5040 - If mds_open unable to find parent dir, make that negative lookup(5154)
5041 - don't create new directories with extent-mapping (5909, 5936)
5043 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5044 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5045 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5046 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5047 - support for RHEL4 kernel on i686 (5773)
5048 - provide error messages when incompatible logs are encountered (5898)
5050 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5051 * version 1.4.0.10 (1.4.1 release candidate 1)
5053 - don't keep a lock reference when lock is not granted (4238)
5054 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5055 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5056 - drop import inflight refcount on signal_completed_replay error (5255)
5057 - unlock page after async write error during send (3677)
5058 - handle missing objects in filter_preprw_read properly (5265)
5059 - no transno return for symlink open, don't save no-trasno open (3440)
5060 - don't try to complete elan receive that already failed (4012)
5061 - free RPC server reply state on error (5406)
5062 - clean up thread from ptlrpc_start_thread() on error (5160)
5063 - readahead could read extra page into cache that wasn't ejected (5388)
5064 - prevent races in class_attach/setup/cleanup/detach (5260)
5065 - don't dereference de->d_inode after l_dput of de (5458)
5066 - use "int" for stripe value returned from lock_to_stripe (5544)
5067 - mballoc allocation and error-checking fixes in 2.6 (5504)
5068 - block device patches to fix I/O request sizes in 2.6 (5482)
5069 - look up hostnames for IB nals (5602)
5070 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5071 - don't start multiple acceptors for the same port (5277)
5072 - fix incorrect LASSERT in mds_getattr_name (5635)
5073 - export a proc file for general "ping" checking (5628)
5074 - fix "lfs check" to not block when the MDS is down (5628)
5076 - service request history (4965)
5077 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5078 - create an "evict_client" /proc entry on OSTs, like the MDS has
5079 - fix mount usage message, return errors per mount(8) (5168)
5080 - change grep [] to grep "[]" in tests so they work in more UMLs
5081 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5082 - remove ancient LOV_MAGIC_V0 EA support (5047)
5083 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5084 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5085 - allow readahead allocations to fail when low on memory (5383)
5086 - mmap locking landed again, after considerable improvement (2828)
5087 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5089 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5092 - send OST transaction number in read/write reply to free req (4966)
5093 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5094 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5095 - the watchdog thread now runs as interruptible (5246)
5096 - flock/lockf fixes (but it's still disabled, pending 5135)
5097 - don't use EXT3 constants in llite code (5094)
5098 - memory shortage at startup could cause assertion (5176)
5100 - reorganization of lov code
5101 - single portals codebase
5103 - add extents/mballoc support (5025)
5104 - direct I/O reads in the obdfilter (4048)
5105 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5107 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5110 - send OST transaction number in read/write reply to free req (4966)
5111 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5112 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5113 - the watchdog thread now runs as interruptible (5246)
5114 - handle missing objects in filter_preprw_read properly (5265)
5115 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5116 - drop import inflight refcount on signal_completed_replay error (5255)
5117 - unlock page after async write error during send (3677)
5118 - return original error code on reconstructed replies (3761)
5119 - no transno return for symlink open, don't save no-trasno open (3440)
5121 - add pid to ldlm debugging output (4922)
5122 - bump the watchdog timeouts -- we can't handle 30sec yet
5123 - extra debugging for orphan dentry/inode bug (5259)
5125 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5128 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5129 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5130 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5131 - fix NULL dereference in /proc/sys/portals/routes (4827)
5132 - allow failed mdc_close() operations to be interrupted (4561)
5133 - stop precreate on OST before MDS would time out on it (4778)
5134 - don't send partial-page writes before EOF from client (4410)
5135 - discard client grant for sub-page writes on large-page clients (4520)
5136 - don't free dentries not owned by NFS code, check generation (4806)
5137 - fix lsm leak if mds_create_objects() fails (4801)
5138 - limit debug_daemon file size, always print CERROR messages (4789)
5139 - use transno after validating reply (3892)
5140 - process timed out requests if import state changes (3754)
5141 - update mtime on OST during writes, return in glimpse (4829)
5142 - add mkfsoptions to LDAP (4679)
5143 - use ->max_readahead method instead of zapping global ra (5039)
5144 - don't interrupt __l_wait_event() during strace
5146 - add software watchdogs to catch hung threads quickly (4941)
5147 - make lustrefs init script start after nfs is mounted
5148 - fix CWARN/ERROR duplication (4930)
5149 - return async write errors to application if possible (2248)
5150 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5151 - print NAL number in %x format (4645)
5152 - update barely-supported suse-2.4.21-171 series (4842)
5153 - support for sles 9 %post scripts
5154 - support for building 2.6 kernel-source packages
5155 - support for sles km_* packages
5157 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5160 - ignore -ENOENT errors in osc_destroy (3639)
5161 - notify osc create thread that OSC is being cleaned up (4600)
5162 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5163 - reconstruct ost_handle() like mds_handle() (4657)
5164 - create a new thread to do import eviction to avoid deadlock (3969)
5165 - let lconf resolve symlinked-to devices (4629)
5166 - don't unlink "objects" from directory with default EA (4554)
5167 - hold socknal file ref over connect in case target is down (4394)
5168 - allow more than 32000 subdirectories in a single directory (3244)
5169 - fix blocks count for O_DIRECT writes (3751)
5170 - OST returns ENOSPC from object create when no space left (4539)
5171 - don't send truncate RPC if file size isn't changing (4410)
5172 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5173 - bind to privileged port in socknal and tcpnal (3689)
5175 - rate limit CERROR/CWARN console message to avoid overload (4519)
5176 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5177 - basic mmap support (3918)
5178 - kernel patch series update from b1_4 (4711)
5180 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5183 - avoid crash during MDS cleanup with OST shut down (2775)
5184 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5185 - don't use bad inodes on the MDS (3744)
5186 - dynamic object preallocation to improve recovery speed (4236)
5187 - don't hold spinlock over lock dumping or change debug flags (4401)
5188 - don't zero obd_dev when it is force cleaned (3651)
5189 - print grants to console if they go negative (4431)
5190 - "lctl deactivate" will stop automatic recovery attempts (3406)
5191 - look for existing locks in ldlm_handle_enqueue() (3764)
5192 - don't resolve lock handle twice in recovery avoiding race (4401)
5193 - revalidate should check working dir is a directory (4134)
5195 - don't always mark "slow" obdfilter messages as errors (4418)
5197 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5200 - don't close LustreDB during write_conf until it is done (3860)
5201 - fix typo in lconf for_each_profile (3821)
5202 - allow dumping logs from multiple threads at one time (3820)
5203 - don't allow multiple threads in OSC recovery (3812)
5204 - fix debug_size parameters (3864)
5205 - fix mds_postrecov to initialize import for llog ctxt (3121)
5206 - replace config semaphore with spinlock (3306)
5207 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5208 - don't allow enqueue to complete on a destroyed export (3822)
5209 - down write_lock before checking llog header bitmap (3825)
5210 - recover from lock replay timeout (3764)
5211 - up llog sem before sending rpc (3652)
5212 - reduce ns lock hold times when setting kms (3267)
5213 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5214 - fix NULL deref and obd_dev leak on setup error (3312)
5215 - replace some LBUG about llog ops with error handling (3841)
5216 - don't match INVALID dentries from d_lookup and spin (3784)
5217 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5218 - fix invalid assertion in ptlrpc_set_wait (3880)
5220 - add libwrap support for the TCP acceptor (3996)
5221 - add /proc/sys/portals/routes for non-root route listing (3994)
5222 - allow setting MDS UUID in .xml (2580)
5223 - print the stack of a process that LBUGs (4228)
5225 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5228 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5229 - make sure to unset replay flag from failed open requests (3440)
5230 - if default stripe count is 0, use OST count for inode size (3636)
5231 - update parent mtime/ctime on client for create/unlink (2611)
5232 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5233 - free recovery state on server during a forced cleanup (3571)
5234 - unregister_reply for resent reqs (3063)
5235 - loop back devices mounting and status check on 2.6 (3563)
5236 - fix resource-creation race that can provoke i_size == 0 (3513)
5237 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5238 - more debugging for page-accounting assertion (3746)
5239 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5240 - avoid deadlock after precreation failure (3758)
5241 - fix race and lock order deadlock in orphan handling (3450, 3750)
5242 - add validity checks when grabbing inodes from l_ast_data (3599)
5244 - add /proc/.../recovery_status to obdfilter (3428)
5245 - lightweight CDEBUG infrastructure, debug daemon (3668)
5246 - change default OSC RPC parameters to be better on small clusters
5247 - turn off OST read cache for files smaller than 32MB
5248 - install man pages and include them in rpms (3100)
5249 - add new init script for (un)mounting lustre filesystems (2593)
5250 - run chkconfig in %post for init scripts (3701)
5251 - drop scimac NAL (unmaintained)
5253 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5256 - clean kiobufs before and after use (3485)
5257 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5258 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5259 - change init script to fail more clearly if not run as root (1528)
5260 - allow clients to reconnect during replay (1742)
5261 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5262 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5263 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5264 - fix cond_resched (3554)
5265 - only evict unfinished clients after recovery (3515)
5266 - allow bulk resend, prevent data loss (3570)
5267 - dynamic ptlrpc request buffer allocation (2102)
5268 - don't allow unlinking open directory if it isn't empty (2904)
5269 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5270 - remove extraneous obd dereference causing LASSERT failure (3334)
5271 - don't use get_cycles() when creating temp. files on the mds (3156)
5272 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5273 - handle EEXIST for set-stripe, set proper directory name (3336)
5275 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5276 - fix ksocknal_fmb_callback() error messages (2918)
5278 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5281 - don't copy lvb into (possibly NULL) reply on error (2983)
5282 - don't deref dentry after dput, don't free lvb on error (2922)
5283 - use the kms to determine writeback rpc length (2947)
5284 - increment oti_logcookies when osc is inactive (2948)
5285 - update client's i_blocks count via lvb messages (2543)
5286 - handle intent open/close of special files properly (1557)
5287 - mount MDS with errors=remount-ro, like obdfilter (2009)
5288 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5289 - don't use cancelling-locks' kms values (2947)
5290 - use highest lock extent for kms, not last one (2925)
5291 - don't dereference ERR_PTR() dentry in error handling path (3107)
5292 - fix thread race in portals_debug_dumplog() (3122)
5293 - create lprocfs device entries at setup instead of at attach (1519)
5294 - common AST error handler, don't evict client on completion race (3145)
5295 - zero nameidata in detach_mnt in 2.6 (3118)
5296 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5297 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5298 - initialize RPC timeout timer earlier for 2.6 (3219)
5299 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5300 - print nal/nid for unknown nid (3258)
5301 - additional checks for oscc recovery before doing precreate (3284)
5302 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5303 - don't crash in mdc_close for bad permissions on open (3285)
5304 - zero i_rdev for non-device files (3147)
5305 - clear page->private before handing to FS, better assertion (3119)
5306 - tune the read pipeline (3236)
5307 - fix incorrect decref of invalidated dentry (2350)
5308 - provide read-ahead stats and refine rpc in flight stats (3328)
5309 - don't hold journal transaction open across create RPC (3313)
5310 - update atime on MDS at close time (3265)
5311 - close LDAP connection when recovering to avoid server load (3315)
5312 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5313 - don't leak open file on MDS after open resend (3325)
5314 - serialize filter_precreate and filter_destroy_precreated (3329)
5315 - loop device shouldn't call sync_dev() for nul device (3092)
5316 - clear page cache after eviction (2766)
5317 - resynchronize MDS->OST in background (2824)
5318 - refuse to mount the same filesystem twice on same mountpoint (3394)
5319 - allow llmount to create routes for mounting behind routers (3320)
5320 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5321 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5322 - fix rare problem with rename on htree directories (3417)
5324 - allow default OST striping configuration per directory (1414)
5325 - fix compilation for qswnal for 2.6 kernels (3125)
5326 - increase maximum number of MDS request buffers for large systems
5327 - change liblustreapi to be useful for external progs like lfsck (3098)
5328 - increase local configuration timeout for slow disks (3353)
5329 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5331 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5334 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5335 - don't overwrite extent policy data in reply if lock was blocked (2901)
5336 - drop filter export grants atomically with removal from device (2663)
5337 - del obd_self_export from work_list in class_disconnect_exports (2908)
5338 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5339 - swab reply message in mdc_close, other PPC fixes (2464)
5340 - fix destroying of named logs (2325)
5341 - overwrite old logs when running lconf --write_conf (2264)
5342 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5343 - fix race in target_handle_connect (2898)
5344 - mds_reint_create() should take same inode create lock (2926)
5345 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5346 - don't close files for self_export to avoid uninitialized obd (2936)
5347 - allow MDS with the same name as client node (2939)
5348 - hold dentry reference for closed log files for unlink (2325)
5349 - reserve space for all logs during transactions (2059)
5350 - don't evict page beyond end of stripe extent (2925)
5351 - don't oops on a deleted current working directory (2399)
5352 - handle hard links to targets without a parent properly (2517)
5353 - don't dereference NULL lock when racing during eviction (2867)
5354 - don't grow lock extents when lots of conflicting locks (2919)
5356 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5359 - account for cache space usage on clients to avoid data loss (974)
5360 - lfsck support in lustre kernel code (2349)
5361 - reduce journal credits needed for BRW writes (2370)
5362 - orphan handling to avoid losing space on client/server crashes
5363 - ptlrpcd can be blocked, stopping ALL progress (2477)
5364 - use lock value blocks to assist in proper KMS, faster stat (1021)
5365 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5366 - recovery for initial connections (2355)
5367 - fixes for mds_cleanup_orphans (1934)
5368 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5369 - block all file creations until orphan recovery completes (1901)
5370 - client remove rq_connection from request struct (2423)
5371 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5372 - recovery timer race (2670)
5373 - mdc_close recovey bug (2532)
5374 - ptlrpc cleanup bug (2710)
5375 - mds timeout on local locks (2588)
5376 - namespace lock held during RPCs (2431)
5377 - handle interrupted sync write properly (2503)
5378 - don't try to handle a message that hasn't been replied to (2699)
5379 - client assert failure during cleanup after abort recovery (2701)
5380 - leak mdc device after failed mount (2712)
5381 - ptlrpc_check_set allows timedout requests to complete (2714)
5382 - wait for inflight reqs when ptlrpcd finishes (2710)
5383 - make sure unregistered services are removed from the srv_list
5384 - reset bulk XID's when resending them (caught by 1138 test)
5385 - unregister_bulk after timeout
5386 - fix lconf error (2694)
5387 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5388 - readahead locks pages, leaves pending causing memory pressure (2673)
5389 - increase OST request buffers to 4096 on large machines (2729)
5390 - fix up permission of existing directories in simple_mkdir (2661)
5391 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5392 - don't assign transno to errored transactions (2742)
5393 - don't delete objects on OST if given a bogus objid from MDS (2751)
5394 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5395 - if rq_no_resend, then timeout request after recovery (2432)
5396 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5397 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5398 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5399 - MDS recovery shouldn't delete objects that it has given out (2730)
5400 - if enqueue arrives after completion, don't clobber LVB (2819)
5401 - don't unlock pages twice when trigger_group_io returns error (2814)
5402 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5403 - don't write pages to disk if there was an error (1450)
5404 - don't ping imports that have recovery disabled (2676)
5405 - take buffered bytes into account when balancing socknal conn (2817)
5406 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5407 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5408 - remove little-endian swabbing of llog records (1987)
5409 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5410 - retry reposting request buffers if they fail (1191)
5411 - grow extent at grant time to avoid granting a revoked lock (2809)
5412 - lock revoke doesn't evict page if covered by a second lock (2765)
5413 - disable VM readahead to avoid reading outside lock extents (2805)
5415 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5416 - updated kernel patches for hp-2.4.20 kernel (2681)
5418 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5421 - fix truncated write corruption (2366)
5422 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5424 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5425 - ensure len(uuid) < 37 in lmc (1171)
5426 - fix ia64 OOPS in llog_test (2255)
5427 - zero end of page at obdfilter for partial page writes (2648)
5428 - don't leave stale dentries around after renames (bug 2428)
5429 - fix timeouts when evicting a client with a single lock held (2642)
5430 - set deadline for the initial HELLO message to drain (2634)
5431 - print out dotted-quad IP addresses in the socknal (2302)
5433 - additional debugging for MDS client eviction problem (2443)
5434 - fix mkfsoptions support for osts (2603, 2604)
5436 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5439 - add series for the vanilla 2.6.0 kernel
5440 - add series for the vanilla 2.4.24 kernel
5441 - add series for a cray x86/64 UL kernel drop
5442 - fix xattr patches for the vanilla 2.4.19 series
5444 - generate true UUIDs in lmc (1171)
5445 - have portals stack dumping break in UML (2466)
5446 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5447 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5448 - wake callback waiting threads on client eviction (2460)
5449 - Add --ptldebug and --subsystem to lmc (1719)
5450 - update assertion to allow safe interrupt allocation
5451 - set rq_no_resend for cancel requests (2432)
5452 - recalculate ptlrpcd timeout after resend (2494)
5453 - call vfs_rmdir when removing pending directories (2368)
5454 - fix renaming a file to itself (2429)
5455 - lmc creates a default one-stripe lov (2454)
5456 - expand procfs space to handle large clusters (2326)
5457 - increase UML stack to avoid overflow
5458 - update lconf's list of debug and subsystem masks
5459 - fix lfs find --obd (2510)
5460 - /proc tunable for disabling filter read caching (2591)
5461 - stop rpm packages from altering slapd.conf (2301)
5462 - disable nagle in the socknal under 0conf (2578)
5463 - choose mds inode size based on stripe count (2572)
5464 - fix kernel-source rpm problems (2516)
5466 - add --disable-doc to avoid pdf generation (2421)
5467 - update documentation, tests, type-os, comments
5468 - avoid format warnings on ia64
5469 - remove the TOE NAL
5470 - tiny code cleanups by removing unused fields
5472 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5475 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5476 - remove the most blatant lies from BUILDING file (2371)
5477 - change default debug level to reasonable production setting
5478 - reduce client side cache size to reduce cache flush time
5479 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5480 - make TCP zerocopy and pinger support enabled by default (2476)
5481 - sync writes completed after process exits caused crashes (2319)
5482 - maintain correct mount count on the MDS (2356)
5483 - backout 1557, because 2316 wasn't really fixed
5484 - better file I/O statistics gathering in /proc
5485 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5486 - another kernel patch to fix zero-copy TCP function export
5487 - don't take duplicate lock when processing re-sent getattr (2420)
5488 - lctl uses obd_self_export instead of creating new conn (2353)
5489 - MDS/OST recovery case which requires object creation asserted (2425)
5490 - move lfs from /usr/sbin to /usr/bin in packages
5491 - fix race between mds_client_add and mds_client_free (2417)
5492 - use kmalloc instead of slabs in portals (2430)
5493 - don't create duplicate records when a failover MDS is present (2442)
5494 - remove unnecessary mount age check (2332)
5495 - don't remove directory inodes from locks prematurely (2451)
5496 - don't break if MDS service name is the same as hostname (2103)
5497 - fix races in client write RPC generation when cache full (2482)
5499 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5502 - remove now-unused request->rq_obd (278)
5503 - if an allocation fails, print out how much memory we've used (1933)
5504 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5505 - add the "configurable stack size" patch to most series files (1256)
5506 - ability to write large log records, for 100+ OST configs (2306)
5507 - fix NULL deref when filter_prep fails (2314)
5508 - fix operator precedence error in filter_sync
5509 - dynamic allocation of socknal TX descriptors (2315)
5510 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5511 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5512 - ensure that CWARN messages in Portals always get to the syslog
5513 - __init/__exit are not for prototype decls (ldlm_init/exit)
5514 - x86-64 compile warning fixes
5515 - fix gateway LMC keyword conflict (2318)
5516 - fix MDS lock inversions in getattr/reint paths (1844)
5517 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5518 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5519 - DLM race condition prevented some lock evictions (2328)
5520 - ENOMEM detection and retry on socknal sends (2230)
5521 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5522 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5523 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5524 - better error messages when a client is rejected during recovery (1505)
5525 - avoid cancelling locks which were never granted, after failure (2330)
5526 - fix i_sem/journal inversion in mds_client_add (2333)
5527 - fix truncate/getattr lock cycle deadlock (2334)
5528 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5529 - fix two rare exit paths which could leak an l_lock() ref (2321)
5530 - fencepost error in MDS/OST orphan recovery (2226)
5531 - make log record alignment 8 bytes (1988)
5532 - lstripe now fails when requested offset > ost_count (2237)
5533 - ensure that all kernel series have a complete list.h (1607)
5534 - fix crashes in special-file operations (2316)
5535 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5537 - allow configurable automake binary, for testing new versions
5538 - small update to the lfs documentation
5540 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5542 * fix negative export reference count in fsfilt_sync (2312)
5544 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5545 * release candidate 0.9.1
5547 - orphans are moved into the PENDING directory for possible recovery
5548 - replayed opens now open by fid for orphan/rename safety (1042)
5549 - last close of an orphan inode generates a transno (683)
5550 - chdir() and mount() now pin the directory entry (1020)
5551 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5552 - discard very old requests without processing them (1502)
5553 - remove some common, well-understood CERRORs (1505)
5554 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5555 - clear "grant" flags in OST replies until OST grant code lands (1644)
5556 - fix read performance by not clobbering i_blksize on client (1598)
5557 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5558 - make LRU size a /proc tunable, clears locks when reduced (707)
5559 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5560 - server locks take references on exports now (1558)
5561 - build fixes for 2.4.20-rh trees (1663)
5562 - return an error from lov_create if all OSCs are inactive (1751)
5563 - fix import levels when a reconnect happens without a timeout (1597)
5564 - exit early from mds_open if we get a lookup error (1749)
5565 - partial page read at EOF wouldn't wait for disk before sending (1642)
5566 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5567 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5568 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5569 - write/truncate lock inversion (1639)
5570 - Don't auto-load obdclass, portals modules during cleanup (1495)
5571 - fix timestamps from jumping to "now" (1763)
5572 - extra journal assertions (1648)
5573 - add an extra multiunlink test (1771)
5574 - fix read_record/write_record API (1776)
5575 - fix leak of offset_extent, possible incorrect i_size later (1772)
5576 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5577 - Debugging for the kqswnal_get_idle_tx problems (1820)
5578 - Allow recovery to be attempted multiple times (1536)
5579 - Write out MDS last_rcvd file after it is first created (1600)
5580 - Fix tx_descriptor leak in failed transmit situations (1827)
5581 - ext3 journaling fixes for assertion failure after IO error (1871)
5582 - class_export_put() on freed export after completion AST error (1896)
5583 - Fix revalidate looping in VFS (1322)
5584 - Don't access a freed export during MDS_REINT timeout (1521)
5585 - Add open-unlink recovery support on the MDS (1673,1764)
5586 - Return an error if no MDS data was read from last_rcvd (1946)
5587 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5588 - Don't setup a disk device that doesn't match exported UUID (317)
5589 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5590 - avoid committing NULL handle in force close
5591 - local.sh is now a one-stripe LOV configuration
5592 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5593 - fix ext3 htree duplicate directory entry corruption (1516)
5594 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5595 - update to new LOV EA format (2097)
5596 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5598 2003-06-15 Phil Schwan <phil@clusterfs.com>
5601 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5602 - per-import/export recovery handling (958, 931, 959)
5603 - multiple last-rcvd slots, for serving multiple FSes (949)
5604 - connections are again shared between multiple imp/exports (963, 964)
5605 - "umount -f" would hang if any requests needed to be sent (393, 978)
5606 - avoid pinning large req buffer by copying for queued messages (989)
5607 - add "uuid" to "lctl device" command to help upcalls (991)
5608 - "open" RPCs with transnos would confuse recovery counters (1037)
5609 - do proper endian conversion of all wire messages (288, 340, 891)
5610 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5611 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5612 - fix ptlrpc_connection leak in target_handle_connect (1174)
5613 - fix import refcounting bug in OST and MDS cleanup (1134)
5614 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5615 - fix very unlikely obd_types race condition (501)
5616 - remove osc_open hack for echo_client (1187)
5617 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5618 - a failure in read_inode2 leads to deadlock (1139)
5619 - cancel ack-locks as soon as transaction is committed (1072)
5620 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5621 - make sure to commitrw after any preprw to avoid deadlock (1162)
5622 - failing to execute a file in a lustre FS would lock inode (1203)
5623 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5624 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5625 - obd_uuid2tgt would walk off the end of the list (1255)
5626 - on IA64 the osc would give portals incorrect bulk size (1258)
5627 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5628 - fix lock inversion caused by new llite matching code (1282)
5629 - limit the number of dirty pages on a client to 10MB (1286)
5630 - timed out locks were not being corrected cancelled (1289)
5631 - fix O_DIRECT above 4GB on IA-32 (1292)
5632 * major user-visible changes
5633 - fail out/fail over policy now controlled by the upcall (993)
5635 - add OBD_PING to check server availability and failure (954)
5636 - lustre messages are now sent in sending host order (288, 340, 891)
5637 - add eadatalen to MDS getattr reply (340)
5638 - OST read replies may contain second buffer, with per-page status (593)
5640 2003-03-11 Phil Schwan <phil@clusterfs.com>
5643 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5644 - failed open()s could cause deadlock; fixed (867, 869)
5645 - stop cancelling OST locks when files are closed (481)
5646 - overlapping XID spaces caused network corruption (851, 853)
5647 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5648 - setattr_raw added, to do single-RPC, server-side setattrs
5649 - lmc/lconf syntax change for OST UUIDs
5650 - fix crashy race condition between ptlrpc_free_req and osc_close
5651 - don't use request in mdc_enqueue if we hit a timeout (889)
5652 - don't set the inode i_size for regular files from the MDS (896)
5653 - handle out of order completion AST (842)
5654 - don't LBUG if a lock request times out after receiving AST (913)
5655 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5656 - if a bad lock AST arrives, send an error instead of dropping entirely
5657 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5658 - fix leak in bulk IO when only partially completed (899, 900, 926)
5659 - fix O_DIRECT for ia64 (55)
5660 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5661 - C-z after timeout could hang a process forever; fixed (977)
5663 - client-side I/O cache (678, 924, 929, 941, 970)
5665 - READPAGE and SETATTRs which don't take server-side locks get
5668 2003-02-11 Phil Schwan <phil@clusterfs.com>
5671 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5672 - Fix fsfilt_extN_readpage() to read a full page of directory
5673 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5674 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5675 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5676 - Most intent operations were removed, in favour of a new RPC mode
5677 that does a single RPC to the server and bypasses most of the VFS
5678 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5679 - Aggressively cancel local locks on DLM servers
5680 - mds_reint_unlink sends EA to the client if it's the last nlink.
5681 client uses that EA to unlink OST objects.
5682 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5683 - recursive symlinks were fixed (439)
5684 - fixed NULL deref in DEBUG_REQ
5685 - filter_update_lastobjid no longer calls sync, which annoyed extN
5686 - fixed multi-client small-writes to a single file problem (445)
5687 - fixed mtime updates during file writes (607)
5688 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5689 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5690 - fixed Linux OST size reporting problem (444, 656)
5691 - OST now updates object mtime with writes or setattr (607, 619)
5692 - client verifies file size before zeroing page past EOF (445)
5693 - OST now writes last allocated objid to disk with allocation (108)
5694 - LOV on echo now works (409)
5696 - mds_reint_unlink sends a new buffer, with the EA included. this
5697 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5698 set if a regular file was being unlinked, and it was the last link
5699 - use PtlGet from the target for bulk writes (315)
5700 - OST now updates object mtime with writes or setattr (607, 619)
5701 - LDLM now has a grant-time callback to revalidate locked items, if
5703 - Many MDS operations were reorganized to combat race conditions
5705 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5706 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5708 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5711 - Fully reactivate OST imports after reconnection (512, others)
5712 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5713 - More graceful error handling for truncating on dead OST (515)
5714 - Don't error out unless we're actually accessing dead stripes (474)
5715 - Fix garbage sizes when stripes are missing (410)
5716 - LRU counters were broken, causing constant lock purge (433, 432)
5717 - garbage on read from stripes with failed OSTs (441)
5718 - mark OSCs as active before reconnecting during recovery (438)
5719 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5720 - lfind did not preserve OST order in output (443)
5721 - symlinks cause hung clients, incorrect data (439)
5722 - stop dereferencing request after dropping refcount (457)
5723 - don't LASSERT(spin_is_locked) on non-SMP (455)
5724 - fixes for many rename() bugs
5725 - fstat didn't correctly synchronize attributes (399)
5726 - server must handle lock cancellation during blocking AST prep (487)
5727 - bulk descriptors were free()d too soon (511)
5728 - fix paths in lconf, which would load incorrect modules (451, 507)
5729 - fix confusing lconf 'host not found' error message (386)
5730 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5731 - fix race condition in mdc_blocking_ast() for inode access (526)
5732 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5733 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5734 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5735 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5736 - fix lstripe in several ways (kernel side) (527)
5737 - fix request leak in ldlm_cli_enqueue (262)
5738 - incorrect OSC was marked inactive after OST failure
5739 - call mds_fs_cleanup before unmounting filesystem (524)
5740 - fix races between taking ns_lock and ldlm_lock_change_resource
5741 - fix races updating LOV export open file list
5742 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5743 - fix recovery NULL deref in ldlm_cli_cancel_unused
5744 - fix some DLM races by using new hash table for lock handles (419)
5745 - permit the client to specify desired inodes, at replay
5746 - duplicate requests when we queue them for replay reintegration
5747 - fix last_rcvd offset calculation
5748 - sync after each recovered transaction, so we always make progress
5749 - never, not always, ERESTART requests without transnos
5750 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5752 - skip replay if the MDS says that the client is already connected
5753 - don't check for a recovery-enabled export to match lctl's UUID
5754 - don't INC_USE_COUNT for phantom exports
5755 - don't crash when cleaning up phantom exports (567)
5756 - don't double-finish or set replay data for errored mdc_open requests
5757 - abort requests when they time out, so we don't get old replies
5758 - send/receive replies for AST messages again
5759 - if the client says that it doesn't have the lock, cancel it on the
5761 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5762 mark it as destroyed, it will all work out when decref is called
5763 - fix module use counts (22, 581)
5765 - ASTs now expect a reply (server cancels lock on error reply)
5767 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5770 - fix many simultaneous client startup (392)
5771 - fix dentry->d_it clobbering
5772 - credentials weren't being shipped for readdir/getattr operations
5773 - remove invalid assertions triggered during some concurrent MD
5775 - proper Lustre versions added (336, 389)
5776 - fix memory leak for create error case (398)
5777 - fix LOV locking bug that would get cli/srv out of sync
5778 - fix echo client over LOV (409)
5779 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5780 - fix double-O_EXCL intent crash (424)
5781 - avoid sending multiple lock CANCELs (352)
5783 - MDS can do multi-client recovery (modulo bugs in new code)
5785 - many updates, edits, cleanups
5787 2002-11-18 Phil Schwan <phil@clusterfs.com>
5790 - fix null d_it dereference (346)
5791 - fix full OST/dbench hang (333)
5792 - fix permission problem with file removal (286)
5793 - fix removal of OSCs from LOV when they fail
5794 - fix NULL deref during bulk timeout (214)
5795 - fix problems related to multiple filesystems on one MDS (241)
5796 - fixed serious subtle metadata locking bugs
5797 - free locks on clients when inodes are removed due to memory
5799 - fix inode pointer in lock data (285)
5800 - partial support for multiple MDS on a single host (241)
5801 - data locks weren't cancelled at clear_inode time (290, 311)
5802 - intent locks could lead to unbounded lock growth (205)
5803 - added a maximum lock count, an LRU list, and a flusher
5804 - fix multiple rename (365)
5805 - properly abstracted the echo client
5806 - OSC locked 1 byte too many; fixed
5807 - rewrote brw callback code:
5808 - fixed recovery bugs related to LOVs (306)
5809 - fixed too-many-pages-in-one-write crash (191)
5810 - fixed (again) crash in sync_io_timeout (214)
5811 - probably fixed callback-related race (385)
5813 - Add capability to MDS protocol
5814 - LDLM cancellations and callbacks on different portals
5816 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5819 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5820 - defer open object destruction to close time (601981, 138)
5821 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5822 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5823 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5824 - don't call obd_setattr in ll_file_release for destroyed objects
5825 * protocol change to lustre_msg: move |version| and add |flags|
5826 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5827 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5828 * added replay of create, unlink, link and rename operations during
5829 MDS failover; recovery should be much more robust now
5830 * remove failed OSCs from LOVs (only lov_create uses this so far)
5831 * the lustre-HOWTO was brought (more) up to date (582544)
5833 2002-10-23 Phil Schwan <phil@clusterfs.com>
5836 - in-use dentries weren't being reused properly (617851)
5837 - prevent multiple LDLM setup (599178)
5838 - fix LOV size calculations for truncate (617853)
5839 - fix client handling of MDS intent errors (POSIX)
5840 - fix permission bug in lovstripe.c test (624321)
5841 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5842 - truncate past end of file could corrupt data
5843 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5844 - a race in recovery could return ETIMEDOUT to apps (623947)
5845 - building outside the source directory was fixed
5846 * the lustre-HOWTO was brought (more) up to date (582544)
5847 * major progress was made on recovery functionality
5849 2002-10-10 Phil Schwan <phil@clusterfs.com>
5852 - recovery deadlock fix
5853 - rm -rf causes LBUG fix (617817)
5854 - file open by multiple tasks fix (618962)
5855 - directory permissions bugs (602707 and 620007)
5856 - journal_stop fixed with locking (611313)
5857 - O_APPEND failures resolved (618273, perhaps 614459)
5858 - lconf PATH fix (619770)
5859 - IA64 build fix (621450)
5860 - RPC buffer sizes scale with amount of memory
5862 2002-10-01 Phil Schwan <phil@clusterfs.com>
5865 - locks would be cancelled without throwing away data pages,
5866 resulting in inconsistent data (605627)
5867 - inode attributes were not always being refreshed (605627, 612449)
5868 - lconf now continues to cleanup after lctl reports an error
5869 - MDS now enforces user permissions (602707)
5870 - lprocfs cleanup fixed, but not yet enabled (614157)
5871 - fixed infinite server hang, should a client not respond to an AST
5872 - avoid going into recovery if user calls readlink() with a buffer
5873 that's too small (613941)
5874 - AST RPCs no longer require replies (614867) -- this may be changed
5875 - don't crash server if client sends an IOV that's too big (611336)
5876 - fixed lock conversion deadlock (611892)
5877 - fixed the following of symlinks (614622)
5878 * recovery: the server can remove locks from a client that dies, other
5879 clients can make progress
5880 * more extN patch fixes
5881 * compile-time configurable ptlrpc buffer allocations
5883 - collaborative read cache document
5884 - Lustre Lite Performance CDR document-in-progress
5886 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5889 - fix typo in patch-2.4.18
5891 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5894 - clear ptlrpc request each time in handle_incoming_request()
5895 - unlink of files now destroys the object on the OST
5897 2002-09-19 Peter Braam <braam@clusterfs.com>
5899 * add hard link support
5900 * change obdfile creation method
5901 * kernel patch changed
5903 2002-09-19 Peter Braam <braam@clusterfs.com>
5906 - stack overflow bug in extN fixed
5908 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5910 * documentation updates
5911 - add man pages for config tools
5912 - update tests/README to describe testing with new config tools
5913 - finish metadata API descriptions
5914 * bug fixes and cleanups
5915 - statfs workaround for 16TB limit
5916 - LOV stripe allocation improved, can stripe on subset of OSTs
5917 - LOV file size/IO offset was wrong for files > 4GB in size
5918 - object EA data was being dropped, caused files to be unreadable
5919 - memory overflow with non-LOV OST caused memory corruption
5920 - fixed regression tests to work with new config tools, obdfilter
5921 - fixed bug when directory size became larger than 1 block
5922 - fixed bug (for single client case) when PWD was deleted
5923 - invalidate local directory pages when doing intent-based ops
5924 - avoid LDLM oops when lock callback contained bad data
5926 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5928 * documentation updates
5929 * bug fixes and cleanups
5930 - configuration tools
5933 - 64-bit compile warnings
5934 - 64-bit internal statfs data
5936 * test_brw on persistent OST devices
5938 * lprocfs (disabled)
5940 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5942 * documentation updates
5943 * bug fixes and cleanups
5944 * configuration tools
5946 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5949 * many small fixes to 0_5_4
5950 * io/network handling
5951 * thinkos in MDS operations
5953 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5956 * crucial basic fixes to 0.5.3
5957 * IOR, Iozone work over Elan
5960 2002-08-07 Phil Schwan <phil@clusterfs.com>
5961 * version 0_5_3, our first alpha
5962 * we use the new Portals iovs
5963 * documentation updates
5964 * bug fixes and cleanups
5965 * small changes in the DLM wire protocol
5967 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5968 * version 0_5_1 with some initial stability,
5969 * locking on MD and file I/O.
5970 * documentation updates
5971 * several bug fixes since 0.5.0
5972 * small changes in wire protocol
5974 2002-07-18 Phil Schwan <phil@clusterfs.com>
5976 * delivered as Lustre Light Alpha
5977 * fixed a crash after handling invalid MDS requests
5978 * fixed directory pages for architectures with non-4k pages sizes
5980 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5981 * release version v0_4_4
5982 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5984 * Fixes a number of interruption problems with OST operations.
5985 * Update documentation for portals header changes
5986 * Move all wire protocol structs/defines to lustre_idl.h
5987 * Fixes symlink length bug.
5988 * Add tcpdump to repository.
5990 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5991 * release version v0_4_3
5992 * Fixes statfs for inodes on extN.
5993 * Fixes bug in runtests which would delete /etc/hosts.
5994 * Use 64-bit object IDs wherever possible (not into VFS though)
5995 Remove ost_get_info, which is unused by lustre, and out of date.
5997 2002-07-03 Peter Braam <braam@clusterfs.com>
5998 * release version v0_4_2 Fixes a lookup error (type not passed)
5999 * move forward to head of Portals
6000 * move forward to latest Lustre kernel
6002 2002-06-25 Peter Braam <braam@clusterfs.com>
6003 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob