1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.33 (SLES 10),
5 2.6.18-92.1.22.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.40.11-sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
14 * File join has been disabled in this release, refer to Bugzilla 16929.
19 Descriptoin: OST grant too much space to client even there are not enough space.
20 Details : Client will shrink its grant cache to OST if there are no write
21 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
22 this grant cache if there are already not enough avaible space
23 (left_space < total_clients * 32M).
26 Frequency : start MDS on uncleanly shutdowned MDS device
28 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
29 Details : stay in waiting mds<>ost recovery finished produce random bugs
30 due race between two ll_sync thread for one lov target. send
31 ACTIVATE event only if connect realy finished and import have
35 Frequency : rare, connect and disconnect target at same time
37 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
38 Details : don't call obd_disconnect under lov_lock. this long time
39 operation and can block ptlrpcd which answer to connect request.
44 Descriptoin: don't lose wakeup for imp_recovery_waitq
45 Details : recover_import_no_retry or invalidate_import and import_close can
46 both sleep on imp_recovery_waitq, but we was send only one wakeup
50 Frequency : always with long access acl
52 Descriptoin: mds can't pack reply with long acl.
53 Details : mds don't control size of acl but they limited by reint/getattr
56 Severity : enhancement
58 Description: Update to SLES10 kernel-2.6.16.60-0.33.
60 Severity : enhancement
62 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
65 Frequency : start MDS on uncleanly shutdowned MDS device
67 Descriptoin: aborting recovery hang on MDS
68 Details : don't throttle destroy RPCs for the MDT.
71 Frequency : on remount
73 Description: external journal device not working after the remount
74 Details : clear dev_rdonly flag for external journal devices in
80 Description: shutdown vs evict race
81 Details : client_disconnect_export vs connect request race.
82 if client will evicted at this time - we start invalidate
83 thread without referece to import and import can be freed
87 Frequency : rare, need acl's on inode.
89 Description: client can't handle ost additional correctly
90 Details : if ost was added after client connected to mds client can have
91 hit lnet_try_match_md ... to big messages to wide striped files.
92 in this case need teach client to handle config events about add
93 lov target and update client max ea size at that event.
95 Severity : enhancement
97 Description: Changelogs
98 Details : Changelogs are a lightweight mechanism to track filesystem
99 metadata and namespace changes. The changelog is recorded
100 permanently on the MDTs, and is periodically "consumed" / purged
101 when records are no longer needed.
103 Severity : enhancement
105 Description: compact fld format with extents
106 Details : Store range of seq rather than every seq in FLD. Seq
107 controller update FLD rather than clients. In Case of CMD, mdt0
108 has FLD, all other metadata server act as non persistent proxy
109 for FLD queries and cache fld entries in fld cache.
114 Description: don't skip ost target if they assigned to file
115 Details : Drop slow OSCs if we can, but not for requested start idx.
116 This means "if OSC is slow and it is not the requested
117 start OST, then it can be skipped, otherwise skip it only
118 if it is inactive/recovering/out-of-space.
122 Description: more cleanup in mds_lov
123 Details : not send LOV EA under replay, we can't know about they size at this
124 time. Don't allow client connect to mds before any ost connected,
125 for avoid problems with LOV EA size and returning EIO to client.
127 Severity : enhancement
129 Description: Interoperability at server side (Disk interoperability)
131 Severity : enhancement
133 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
135 Severity : enhancement
137 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
139 Severity : enhancement
141 Description: New client IO stack (CLIO).
143 Severity : enhancement
145 Description: Commit on sharing. Eliminate inter-client dependencies between
146 uncommitted transactions by doing transaction commits.
147 Thereby clients may recovery independently.
150 Frequency : Create a symlink file with a very long name
152 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
153 Details : If there is no extra space in the request for early cancels,
154 ldlm_req_handles_avail() returns 0 instead of a negative value.
156 Severity : enhancement
158 Description: Add /proc entry for import status
159 Details : The mdc, osc, and mgc import directories now have
160 an import directory that contains useful import data for debugging
163 Severity : enhancement
165 Description: Re-disable certain /proc logging
166 Details : Enable and disable client's offset_stats, extents_stats and
167 extents_stats_per_process stats logging on the fly.
170 Frequency : Only on FC kernels 2.6.22+
172 Description: oops in statahead
173 Details : Do not drop reference count for the dentry from VFS when lookup,
174 VFS will do that by itself.
176 Severity : enhancement
178 Description: Generic /proc file permissions
179 Details : Set /Proc file permissions in a more generic way to enable non-
180 root users operate on some /proc files.
184 Description: Hitting mdc_commit_close() ASSERTION
185 Details : Properly handle request reference release in
186 ll_release_openhandle().
190 Description: quota recovery deadlock during mds failover
191 Details : This patch includes att18982, att18236, att18237 in bz14840.
193 1. fix osts hang when mds does failover with quotaon
194 2. prevent watchdog storm when osts threads wait for the
199 Frequency : only patchless client
200 Description: add workaround for race between add/remove dentry from hash
202 Severity : enhancement
204 Description: Allow OST glimpses to return PW locks
208 Description: LBUG when llog conf file is full
209 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
213 Description: Prevent import from entering FULL state when server in recovery
217 Description: service mount cannot take device name with ":"
218 Details : Only when device name contains ":/" will mount treat it as
224 Description: replace ptlrpcd with the statahead thread to interpret the async
225 statahead RPC callback
229 Frequency : on recovery
230 Description: I/O failures after umount during fail back
231 Details : if client reconnected to restarted server we need join to recovery
232 instead of find server handler is changed and process self
233 eviction with cancel all locks.
235 Severity : enhancement
237 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
239 Severity : enhancement
241 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
243 Severity : enhancement
245 Description: Upcall on Lustre log has been dumped
246 Details : Allow for a user mode script to be called once a Lustre log has
247 been dumped. It passes the filename of the dumped log to the
248 script, the location of the script can be specified via
249 /proc/sys/lnet/debug_log_upcall.
254 Description: avoid idr_remove called for id which is not allocated.
255 Details : Move assigment s_dev for clustered nfs to end of initialization,
256 to avoid problem with error handling.
261 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
262 Details : When connection is reused this not moved from CONN_UNUSED_HASH
263 into CONN_USED_HASH and this prodice warning when put connection
264 again in unused hash.
267 Severity : enhancement
269 Description: File striping can now be set to use an arbitrary pool of OSTs.
271 Severity : enhancement
273 Description: Export bytes_read/bytes_write count on OSC/OST.
277 Description: Early reply size mismatch, MGC loses connection
278 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
279 the connect flags are properly negotiated.
284 Description: quotas are not honored with O_DIRECT
285 Details : all writes with the flag O_DIRECT will use grants which leads to
286 this problem. Now using OBD_BRW_SYNC to guard this.
290 Description: add quota statistics
291 Details : 1. sort out quota proc entries and proc code.
292 2. add quota statistics
294 Severity : enhancement
296 Description: enable quota support for HEAD.
300 Description: Properly propagate oinfo flags from lov to osc for statfs
301 Details : restore missing copy oi_flags to lov requests.
303 Severity : enhancement
305 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
306 : l_getgroups(8), lst(8), routerstat(8)
307 Details : included man pages for llobdstat(8), llstat(8),
308 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
310 Severity : enhancement
312 Description: configure's --enable-quota should check the
313 : kernel .config for CONFIG_QUOTA
314 Details : configure is terminated if --enable-quota is passed but
315 : no quota support is in kernel
319 Description: Remove portals compatibility
320 Details : Remove portals compatibility, not interoperable with releases
325 Description: Resolve device initialization race
326 Details : Prevent proc handler from accessing devices added to the
327 obd_devs array but yet be intialized.
329 Severity : enhancement
331 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
333 Severity : enhancement
335 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
340 Description: Using wrong pointer in osc_brw_prep_request
341 Details : Access to array[-1] can produce panic if kernel compiled with
342 CONFIG_PAGE_ALLOC enabled
346 Description: Client runs out of low memory
347 Details : Consider only lowmem when counting initial number of llap pages
351 Description: *optional* service tags registration
352 Details : if the "service tags" package is installed on a Lustre node
353 When the filesystem is mounted, a local-node service tag will
354 be created. See http://inventory.sun.com/ for more information
355 about the Service Tags asset management system.
359 Description: Kernel BUG tries to release flock
360 Details : Lustre does not destroy flock lock before last reference goes
361 away. So always drop flock locks when client is evicted and
362 perform unlock regardless of successfulness of speaking to MDS.
366 Description: add refcount protection for osc callbacks, avoid panic on shutdown
370 Description: sanity test 65a fails if stripecount of -1 is set
371 Details : handle -1 striping on filesystem in ll_dirstripe_verify
376 Description: ASSERTION(CheckWriteback(page,cmd)) failed
377 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
380 Severity : enhancement
382 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
386 Description: do not process already freed flock
387 Details : flock can possibly be freed by another thread before it reaches
388 to ldlm_flock_completion_ast.
392 Description: LBUG during stress test
393 Details : Need properly lock accesses the flock deadlock detection list.
397 Description: oops in page fault handler
398 Details : kernel page fault handler can return two special 'pages' in error
399 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
403 Description: timeout with invalidate import.
404 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
405 should be handled by ptlrpcd. This produce long age waiting and
406 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
408 Severity : enhancement
410 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
414 Description: enable MGS and MDT services start separately
415 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
416 a co-located MGS without starting the MGS, which is a complement
417 to 'nosvc' mount option.
421 Frequency : after recovery
422 Description: precreate to many object's after del orphan.
423 Details : del orphan st in oscc last_id == next_id and this triger growing
424 count of precreated objects. Set flag LOW to skip increase count
425 of precreated objects.
429 Frequency : rare, on clear nid stats
430 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
431 Details : when clean nid stats sometimes try destroy live entry,
432 and this produce panic in free.
436 Description: Stack overflow during MDS log replay
437 ease stack pressure by using a thread dealing llog_process.
441 Description: wait until IO finished before start new when do lock cancel.
442 Details : VM protocol want old IO finished before start new, in this case
443 need wait until PG_writeback is cleared until check dirty flag and
444 call writepages in lock cancel callback.
446 Severity : enhancement
448 Description: using special macro for print time and cleanup in includes.
452 Description: mds_mfd_close() ASSERTION(rc == 0)
453 Details : In mds_mfd_close(), we need protect inode's writecount change
454 within its orphan write semaphore to prevent possible races.
458 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
462 Frequency : rare, on shutdown ost
463 Description: don't hit live lock with umount ost.
464 Details : shrink_dcache_parent can be in long loop with destroy dentries,
465 use shrink_dcache_sb instead.
469 Description: don't panic with use echo client
470 Details : echo client pass NULL as client nid pointer and this produce null
475 Description: fix build on ppc32
476 Details : compile code with -m64 flag produce wrong object file for ppc32.
480 Description: add message levels for liblustreapi
484 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
485 Details : Move the 'good_osts' check before the 'total_bavail' check. This
486 will result in an -EAGAIN and in the exit call path we call
487 alloc_rr() which will with increasing aggressiveness attempt to
488 aquire precreated objects on the minimum number of required OSCs.
492 Description: Use old size assignment to avoid deadlock
493 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
494 the scheduling while holding a spinlock. We do not need locking
495 for size in ll_update_inode() because size is only updated from
496 the MDS for directories or files without objects, so there is no
497 other place to do the update, and concurrent access to such inodes
498 are protected by the inode lock.
502 Description: resolve "_IOWR redefined" build error on SLES10
506 Description: dump the memory debugging after all modules are unloaded to
507 suppress false negative in conf_sanity test 39
509 Severity : enhancement
511 Description: build kernel-ib packages for OFED 1.3 in our release cycle
516 Description: fix SLES kernel versioning
517 Details : the kernel version for our SLES 10 kernel did not include a "-"
518 before the "smp" at the end. while this was not a problem in
519 general, it did mean that software trying to use the kernel
520 version to try to detect a vendor specific kernel would fail.
521 this was most evident by the OFED build scripts.
525 Description: Don't update lov_desc members until making sure they are valid
526 Details : When updating lov_desc members via proc fs, need fix their
527 validities before doing the real update.
531 Description: don't put request into delay list while invalidate in flight.
532 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
533 import in flight. this produce timeout for invalidate and sometimes
534 can cause stale data.
538 Frequency : on ppc only
539 Description: not convert ost objects for directory because it's not exist.
540 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
544 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
546 Severity : enhancement
548 Description: Update to support 2.6.22.14 vanilla kernel.
552 Frequency : rare, on recovery
553 Description: read procfs can produce deadlock in some situation
554 Details : Holding lprocfs lock which send rpc can produce block for destroy
555 obd objects and this also block reconnect with -EALREADY. This
556 isn't fix all lprocfs bugs - but make it rare.
558 Severity : enhancement
560 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
563 Frequency : frequent on X2 node
565 Description: mdc_set_open_replay_data LBUG
566 Details : Set replay data for requests that are eligible for replay.
570 Description: lustre_mgs: operation 101 on unconnected MGS
571 Details : When MGC is disconnected from MGS long enough, MGS will evict the
572 MGC, and late on MGC cannot successfully connect to MGS and a lot
573 of the error messages complaining that MGS is not connected.
577 Frequency : on network error
578 Description: panic with double free request if network error
579 Details : mdc_finish_enqueue is finish request if any network error ocuring,
580 but it's true only for synchronus enqueue, for async enqueue
581 (via ptlrpcd) this incorrect and ptlrpcd want finish request
584 Severity : enhancement
586 Description: client-side metadata stat-ahead during readdir(directory readahead)
587 Details : perform client-side metadata stat-ahead when the client detects
588 readdir and sequential stat of dir entries therein
591 Frequency : on start mds
593 Description: Implement get_info(last_id) in obdfilter.
596 Frequency : occasional
598 Description: Correctly check stale fid, not start epoch if ost not support SOM
599 Details : open with flag O_CREATE need set old fid in op_fid3 because
600 op_fid2 was overwritten with new generated fid, but mds can answer
601 with one of these two fids and both is not stale. Setattr
602 incorrectly started an epoch and assume will be called
603 done_writing, but without SOM done_writing ever being called.
606 Frequency : rare, depends on device drivers and load
608 Description: MDS or OSS nodes crash due to stack overflow
609 Details : Code changes in 1.8.0 increased the stack usage of some functions.
610 In some cases, in conjunction with device drivers that use a lot
611 of stack the MDS (or possibly OSS) service threads could overflow
612 the stack. One change which was identified to consume additional
613 stack has been reworked to avoid the extra stack usage.
616 Frequency : occasional
618 Description: Do not fail import if osc_interpret_create gets -EAGAIN
619 Details : If osc_interpret_create got -EAGAIN it immediately exits and
620 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
621 oscc_has_objects and see OSC has no objests and call
622 oscc_internal_create to resend create request.
624 Severity : enhancement
626 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
628 Severity : enhancement
630 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
633 Frequency : very rare
635 Description: Fix replay if there is an un-replied request and open
636 Details : In some cases, older replay request will revert the
637 mcd->mcd_last_xid on MDS which is used to record the client's
640 Severity : enhancement
642 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
644 Severity : enhancement
646 Description: Add rhel5 support to HEAD.
648 Serverity : enhancement
650 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
655 Description: Don't allow skipping OSTs if index has been specified.
656 Details : Don't allow skipping OSTs if index has been specified, make
657 locking in internal create lots better.
661 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
662 Details : access to bitfield in structure is always rounded to long
663 and this produce problem with not atomic change any bit.
667 Description: Lustre make rpms failed.
668 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
669 builds Lustre rpms from tarball.
672 Frequency : rare on shutdown ost
674 Description: If llog cancel was not send before clean_exports phase, this can
675 produce deadlock in llog code.
676 Details : If llog thread has last reference to obd and call class_import_put
677 this produce deadlock because llog_cleanup_commit_master wait when
678 last llog_commit_thread exited, but this never success because was
679 called from llog_commit_thread.
683 Description: allow userland application know is lost one of stripes.
684 Details : fill lvb_blocks with error code on ost and return it to
685 application if error flag found.
689 Description: NULL lov_tgts causing MDS oops
690 Details : more safe checks for NULL lov_tgts for avoid oops.
692 Severity : enhancement
694 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
698 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
699 Details : If an OST is down the MDS will hang indefinitely in
700 obd_statfs_rqset() waiting for the statfs data. While for
701 MDS QOS usage of statfs, it should not stuck in waiting.
703 Severity : enhancement
705 Description: remote_acl support
706 Details : Support ACL-based permission check for remote user.
707 Support setfacl/getfacl for remote user with the utils
708 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
709 the system "{s,g}etfacl" utils.
711 Severity : enhancement
713 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
715 Severity : enhancement
717 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
721 Description: make mgs_setparam() handle fsname containing dash
722 Details : fsname containing a dash does not work with lctl conf_param
724 Severity : enhancement
726 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
728 Severity : enhancement
730 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
734 Frequency : rare, at shutdown
735 Description: access already free / zero obd_namespace.
736 Details : if client_disconnect_export was called without force flag set,
737 and exist connect request in flight, this can produce access to
738 NULL pointer (or already free pointer) when connect_interpret
739 store ocd flags in obd_namespace.
743 Frequency : only at startup
744 Description: not alloc memory with spinlock held.
745 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
746 if any spinlock held.
748 Severity : enhancement
750 Description: make lustre randomly fail allocating memory
751 Details : Make lustre randomly failed allocating memory for testing purpose.
753 Severity : enhancement
755 Description: lost problems with lov objid file
756 Details : Fixes some scability and access to not inited memory problems
757 in work with lov objdid file.
762 Description: lfs find does not continue on file error
763 Details : Continue other files processing when a file/dir is absent.
767 Description: Inconsistent usage of lustre_pack_reply()
768 Details : Standardize the usage of lustre_pack_reply() such that it
769 always generate a CERROR on failure.
772 Frequency : occasional
774 Description: MDS hang or stay in waiting lock
775 Details : If client receive lock with CBPENDING flag ldlm need send lock
776 cancel as separate rpc, to avoid situation when cancel request
777 can't processed due all i/o threads stay in wait lock.
781 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
785 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
789 Description: Update to SLES9 kernel-2.6.5-7.287.3.
793 Description: Update to RHEL4 latest kernel.
795 Severity : enhancement
797 Description: Build SLES10 patchless client fails
798 Details : The configure was broken by run ./configure with
799 --with-linux-obj=.... argument for patchless client. When the
800 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
801 can't search header adequately. Use absolute path such as
802 -I($LINUX)/include instead.
806 Description: interrupt oig_wait produce painc on resend.
807 Details : brw_redo_request can be used for resend requests from ptlrpcd and
808 private set, and this produce situation when rq_ptlrpcd_data not
809 copyed to new allocated request and triggered LBUG on assert
810 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
811 ptlrpcd set if request is changed and can be safety changed to use
816 Description: LASSERT_{REQ,REP}SWAB macros are buggy
817 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
818 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
819 these macros and replace them with their difinitions instead.
823 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
824 Details : Update which_patch & target file for SLES10 latest kernel.
827 Description: Remove client patches from SLES 10 kernel.
828 Details : This causes SLES 10 clients to behave as patchless clients
829 even on a Lustre-patched (server) kernel.
831 Severity : enhancement
833 Description: self-adjustable client's lru lists
834 Details : use adaptive algorithm for managing client cached locks lru
835 lists according to current server load, other client's work
836 pattern, memory activities, etc. Both, server and client
837 side namespaces provide number of proc tunables for controlling
840 Severity : enhancement
842 Description: light-weight GSS support
843 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
844 and reduce performance overhead.
846 Severity : enhancement
848 Description: Linux keyring support
849 Details : Support using service of Linux keyring for Lustre GSS internal
850 context refresh/cache mechanism.
854 Description: Fix errors in lfs documentation
855 Details : Fixes man pages
859 Description: don't use GFP_* in generic Lustre code.
860 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
864 Description: obdclass is limited by single OBD_ALLOC(idarray)
865 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
869 Description: warnings with build patchless client with vanila 2.6.19 and up
870 Details : change old ctl_table style and replace ctl_table/ctl_table_header
871 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
875 Frequency : Only for SLES
876 Description: sanity_quota fail test_1
877 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
878 file, which makes TSTID[2] inunique.
882 Frequency : Only for SLES9
883 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
884 Details : Update target/ChangeLog/which_patch .
888 Frequency : Only for test_10 in sanity-quota.sh
889 Description: a bug in quota test code
890 Details : checking a test flag in a wrong place
894 Frequency : Only for test_13 in sanity-quota.sh
895 Description: a bug in quota test script
896 Details : shouldn't check overfull allocated quota
900 Description: jbd statistics
901 Details : Port older jbd statistics patch for sles10
903 Severity : enhancement
905 Description: Add printing inode info into message about error in writepage.
910 Description: reply_lock_interpret crash due to race with it and lock cancel.
911 Details : Do not replay locks that are being cancelled. Do not reference
912 locks by their address during replay, just by their handle.
916 Frequency : When flocks are used.
917 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
918 Details : Flock locks might destroy just granted lock if it could be merged
919 with another existing flock, this is done in completion handler,
920 so teach ldlm_cli_enquque_fini that this is a valid case for
926 Description: Oops in read and write path when failing to allocate lock.
927 Details : Check if lock allocation failed and return error back.
931 Description: lstripe command fails for valid OST index
932 Details : The stripe offset is compared to lov->desc.ld_tgt_count
933 instead of lov->desc.ld_active_tgt_count.
937 Description: sanity.sh failed test 103
938 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
943 Description: df doesn't work properly if diskfs blocksize != 4K
944 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
948 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
949 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
951 Severity : enhancement
953 Description: Adaptive timeouts
954 Details : RPC timeouts adapt to changing server load and network
955 conditions to reduce resend attempts and improve recovery time.
959 Description: llapi_file_create() does not allow some changes
960 Details : add llapi_file_open() that allows specifying the mode and
961 open flags, and also returns an open file handle.
965 Description: merge and cleanup kernel patches.
966 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
970 Description: Add journal checksum support.(Kernel part)
971 Details : The journal checksum feature adds two new flags i.e
972 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
973 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
974 indicates that the commit block contains the checksum for
975 the blocks described by the descriptor blocks. Now commit
976 record can be sent to disk without waiting for descriptor
977 blocks to be written to disk. This behavior is controlled
978 using JBD2_FEATURE_ASYNC_COMMIT flag.
982 Description: OSS needs mutliple precreate threads
983 Details : Add ability to start more than one create thread per OSS.
987 Description: Sanity.sh test_65k failed: setstripe should have succeeded
988 Details : execute lfs setstripe on client
992 Description: mds_obd_create error creating tmp object
993 Details : When the user sets quota on root, llog will be affected and can't
994 create files and write files.
997 Frequency : Always on ia64 patchless client, and possibly others.
999 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1000 Details : This allows the patchless client to be loaded on architectures
1001 without this export.
1005 Description: RedHat Update kernel for RHEL5
1006 Details : Add the kernel config file for RHEL5.
1010 Description: RedHat Update kernel for RHEL5
1011 Details : Modify the kernel config file more closer RHEL5.
1015 Description: Build failure against Centos5 (RHEL5)
1016 Details : Define PAGE_SIZE when it isn't present.
1020 Description: test_7 on sanity-quota.sh can't be used on separate machines
1021 Details : add facet to handle it
1025 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1026 Details : with NFS, the anon dentry's parent was set to itself in
1027 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1028 corresponding dentry other than use rec->ur_name.
1030 Severity : enhancement
1032 Description: lfs setstripe enhancement
1033 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1037 Description: enable data checksumming by default
1038 Details : enable checksum by default, allow --disable-checksum
1039 configure option and "-o nochecksum" mount option. Checksums
1040 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1044 Description: lustre support for RHEL5
1045 Details : Add support for RHEL5.
1049 Description: Client eviction due to failover config
1050 Details : after a connection loss, the lustre client should attempt to
1051 reconnect to the last active server first before trying the
1052 other potential connections.
1056 Description: when mds and osts use different quota unit(32bit and 64bit),
1057 quota will be released repeatly.
1058 Details : void sending multiple quota reqs to mds, which will keep the
1059 status between the reqs.
1063 Description: osts not allocated evenly to files
1064 Details : change the condition to increase offset_idx
1067 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1069 Description: Data corruption for OSTs that are formatted larger than 2TB
1071 Details : When generating the bio request for lustre file writes the
1072 sector number would overflow a temporary variable before being
1073 used for the IO. The data reads correctly from Lustre (which
1074 will overflow in a similar manner) but other file data or
1075 filesystem metadata may be corrupted in some cases.
1079 Description: Tune the kernel for good SCSI performance.
1080 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1081 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1086 Description: rewrite ext2-derived code in obdclass/uuid.c
1087 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1088 from scratch preserving functionality.
1092 Description: "lfs find -obd UUID" prints directories
1093 Details : "lfs find -obd UUID" will return all directory names instead
1094 of just file names. It is incorrect because the directories
1095 do not reside on the OSTs.
1099 Description: use i_size_read and i_size_write in 2.6 port
1100 Details : replace inode->i_size access with i_size_read/write()
1103 Frequency : when removing large files
1105 Description: scheduling issue during removal of large Lustre files
1106 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1107 It causes scheduling issues when removing large files (17TB in the
1111 Frequency : only with liblustre clients on XT3
1113 Description: evictions taking too long
1114 Details : allow llrd to evict clients directly on OSTs
1117 Frequency : only on ppc
1119 Description: /proc/fs/lustre/devices broken on ppc
1120 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1121 We should make sure the type of 'index' is loff_t and then cast
1122 explicitly as needed below. Do not assign an explicitly cast
1126 Frequency : only for rhel5
1128 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1129 Details : Modify the target file & which_kernel.
1131 Severity : enhancement
1133 Description: omit set fsid for export NFS
1134 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1135 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1139 Frequency : Always, for kernels after 2.6.16
1140 Description: Fix warning idr_remove called for id=.. which is not allocated.
1141 Details : Last kernels save old s_dev before kill super and not allow
1142 to restore from callback - restore it before call kill_anon_super.
1146 Description: buffer overruns could theoretically occur
1147 Details : llapi_semantic_traverse() modifies the "path" argument by
1148 appending values to the end of the origin string, and a
1149 overrun may occur. Adding buffer overrun check in liblustreapi.
1153 Description: Fix error on 'ls .' at the top of the Lustre mount.
1154 Details : Don't revalidate dentry if it is a root dentry.
1158 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1159 Details : Modify vm-tunables-rhel4.patch.
1163 Description: Kernel config for 2.6.18-vanilla.
1164 Details : Modify targets/2.6-vanilla.target.in.
1165 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1166 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1167 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1168 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1172 Description: improve handling recoverable errors
1173 Details : if request processig with error which can be recoverable on server
1174 request should be resend, otherwise page released from cache and
1179 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1180 Details : Modify target file & which_patch.
1181 A flaw was found in the IA32 system call emulation provided
1182 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1183 value could be stored in the %RAX register, which could trigger an
1184 out-of-bounds system call table access. An untrusted local user
1185 could exploit this flaw to run code in the kernel
1186 (ie a root privilege escalation). (CVE-2007-4573).
1190 Description: O_DIRECT bypasses client statistics.
1191 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1192 read_ahead_stats were updated. copied the stats section from
1193 osc_send_oap_rpc() into async_internal().
1197 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1201 Description: change order of libsysio includes
1202 Details : '#include sysio.h' should always come before '#include xtio.h'
1206 Description: handle "serious error: objid * already exists" more gracefully
1207 Details : If LAST_ID value on disk is smaller than the objects existing in
1208 the O/0/d* directories, it indicates disk corruption and causes an
1209 LBUG(). If the object is 0-length, then we should use the existing
1210 object. This will help to avoid a full fsck in most cases.
1212 Severity : enhancement
1214 Description: adapt the lustre_config script to support the upgrade case
1215 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1216 server targets to 1.6 in parallel.
1220 Description: To avoid grant space > avaible space when the disk is almost
1221 full. Without this patch you might see the error "grant XXXX >
1222 available" or some LBUG about grant, when the disk is almost
1224 Details : In filter_check_grant, for non_grant cache write, we should
1225 check the left space by if (*left > ungranted + bytes), instead
1226 of (*left > ungranted), because only we are sure the left space
1227 is enough for another "bytes", then the ungrant space should be
1228 increase. In client, we should update cl_avail_grant only there
1229 is OBD_MD_FLGRANT in the reply.
1233 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1234 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1235 described in CVE-2007-4573. This problem could allow untrusted
1236 local users to gain root access.
1239 Frequency : when using O_DIRECT and quotas
1241 Description: Incorrect file ownership on O_DIRECT output files
1242 Details : block usage reported by 'lfs quota' does not take into account
1243 files that have been written with O_DIRECT.
1248 Description: touch file failed when fs is not full
1249 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1250 otherwise we can get ENOSP while fs is not full.
1254 Description: parallel lock callbacks
1255 Details : Instead of sending blocking and completion callbacks as separated
1256 requests, adding them to a set and sending in parallel.
1259 Frequency : only for Cray XT3
1260 Bugzilla : 12829/13455
1261 Description: Changing primary group doesn't change the group lustre assigns to
1263 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1264 overridden with the primary group provided by the group upcall,
1265 whereas the supplied fsgid can be trusted if it is in the list of
1266 supplementary groups returned by the group upcall.
1268 Severity : enhancement
1270 Description: Allow masking D_WARNING, D_ERROR messages from console
1271 Details : Console messages can now be disabled via lnet.printk.
1275 Description: User code with malformed file open parameter crashes client node
1276 Details : Before packing join_file req, all the related reference should be
1277 checked carefully in case some malformed flags cause fake
1278 join_file req on client.
1282 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1283 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1284 mode is equal to the granted mode. However, it should also check
1285 that LDLM_FL_CP_REQD is not set.
1289 Description: Heavy nfs access might result in deadlocks
1290 Details : After ELC code landed, it is now improper to enqueue any mds
1291 locks under och_sem, because enqueue might want to decide to
1292 cancel open locks for same inode we are holding och_sem for.
1296 Description: Client eviction while running blogbench
1297 Details : A lot of unlink operations with concurrent I/O can lead to a
1298 deadlock causing evictions. To address the problem, the number of
1299 oustanding OST_DESTROY requests is now throttled to
1300 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1301 callbacks are processed in priority.
1305 Description: enable ACLs on MDS by default
1306 Details : ACLs must be enabled on MDS by default.
1309 Frequency : PPC/PPC64 only
1311 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1312 Details : fix duplicated definitions between asm-ppc64/types.h and
1313 lustre_types.h on PPC.
1316 Frequency : PPC/PPC64 only
1318 Description: asm-ppc/segment.h does not exist
1319 Details : fix compile issue on PPC.
1323 Description: better handle error messages in extents code
1326 Frequency : RHEL4 only
1328 Description: mkfs is very slow on IA64/RHEL4
1329 Details : A performance regression has been discovered in the MPT Fusion
1330 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1331 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1332 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1334 Severity : enhancement
1336 Description: SNMP support enhancement
1337 Details : Adding total number of sampled request for an MDS node in snmp
1340 Severity : enhancement
1342 Description: Optimize ldlm waiting list processing for PR extent locks
1343 Details : When processing waiting list for read extent lock and meeting read
1344 lock that is same or wider to it that is not contended, skip
1345 processing rest of the list and immediatelly return current
1346 status of conflictness, since we are guaranteed there are no
1347 conflicting locks in the rest of the list.
1351 Description: Time out and refuse to reconnect
1352 Details : When the failover node is the primary node, it is possible
1353 to have two identical connections in imp_conn_list. We must
1354 compare not conn's pointers but NIDs, otherwise we can defeat
1355 connection throttling.
1359 Description: port llog fixes from b1_6 into HEAD
1360 Details : Port llog reference couting and some llog cleanups from b1_6
1361 (bug 10800) into HEAD, for protect from panic and access to already
1362 free llog structures.
1366 Description: Detect stride IO mode in read-ahead
1367 Details : When a client does stride read, read-ahead should detect that and
1368 read-ahead pages according to the detected stride pattern.
1372 Description: data checksumming impacts single node performance
1373 Details : add support for several checksum algorithm. Currently, only CRC32
1374 and Adler-32 are supported. The checksum type can be changed on
1375 the fly via /proc/fs/lustre/osc/*/checksum_type.
1379 Description: use adler32 for page checksums
1380 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1385 Description: build for x2 fails
1386 Details : fix compile issue on Cray systems.
1390 Description: Properly match for duplicate locks
1391 Details : Due to different lock order from skiplists code, we need to
1392 traverse entire list for now
1395 Frequency : only on PPC/SLES10
1397 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1398 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1399 be defined. Add a hack in mkfs_lustre.c to work around this compile
1404 Description: LASSERT on MDS when client holding flock lock dies
1405 Details : ldlm pool logic depends on number of granted locks equal to
1406 number of released locks which is not true for flock locks, so
1407 just exclude such locks from consideration.
1411 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1412 Details : Use fsfilt_sync() for both the whole filesystem sync and
1413 individual file sync to eliminate dangerous inode locking
1414 with I_LOCK that can lead to a deadlock.
1418 Description: performance in 1.6.3
1419 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1420 sure that 1MB requests are not fragmented by the block layer.
1422 Severity : enhancement
1424 Description: organize the server-side client stats on per-nid basis
1425 Details : Change the structure of stats under obdfilter and mds to
1433 The "uuid"s file would list the uuids of _active_ exports.
1434 And the clear entry is to clear all stats and stale nids.
1436 Severity : enhancement
1438 Description: eliminate client locks in face of contention
1439 Details : file contention detection and lockless i/o implementation
1440 for contended files.
1444 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1447 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1449 Description: MDS may lose file striping (and hence file data) in some cases
1450 Details : If there are additional extended attributes stored on the MDS,
1451 in particular ACLs, SELinux, or user attributes (if user_xattr
1452 is specified for the client mount options) then there is a risk
1453 of attribute loss. Additionally, the Lustre file striping
1454 needs to be larger than default (e.g. striped over all OSTs),
1455 and an additional attribute must be stored initially in the
1456 inode and then increase in size enough to be moved to the
1457 external attribute block (e.g. ACL growing in size) for file
1462 Description: skiplist implementation simplification
1463 Details : skiplists are used to group compatible locks on granted list
1464 that was implemented as tracking first and last lock of each
1465 lock group the patch changes that to using doubly linked lists
1469 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1470 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1471 because LASSERT on both the data supplied by a client, and the
1472 data on disk is dangerous and incorrect.
1474 Severity : enhancement
1476 Description: Slow truncate/writes to huge files at high offsets.
1477 Details : Directly associate cached pages to lock that protect those pages,
1478 this allows us to quickly find what pages to write and remove
1479 once lock callback is received.
1483 Description: more ldlm soft lockups
1484 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1485 starve other threads from the resource lock for a long time in
1486 case of long waiting queue, so change the debug level from
1487 D_OTHER to the less frequently used D_INFO.
1489 Severity : enhancement
1491 Description: add -gid, -group, -uid, -user options to lfs find
1495 Description: Hung threads in invalidate_inode_pages2_range
1496 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1497 once one is completed. As a result, some RPCs are stuck in the
1498 queue and are never sent.
1502 Description: filter threads hungs on waiting journal commit
1503 Details : Cleanup filter group llog code, then only filter group llog will
1504 be only created in the MDS/OST syncing process.
1508 Description: Procfs and llog threads access destoryed import sometimes.
1509 Details : Sync the import destoryed process with procfs and llog threads by
1510 the import refcount and semaphore.
1512 Severity : enhancement
1514 Description: openlock cache of b1_6 port to HEAD
1519 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1520 Details : If insertion of an extent fails, then discard the inode
1521 preallocation and free data blocks else it can lead to duplicate
1526 Description: don't always update ctime in ext3_xattr_set_handle()
1527 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1528 In some cases the ctime should not be updated, for example for
1529 2.0->1.8 compatibility it is necessary to delete an xattr and it
1530 should not update the ctime.
1534 Bugzilla : 15713/16362
1535 Description: Assertion in iopen_connect_dentry in 1.6.3
1536 Details : looking up an inode via iopen with the wrong generation number can
1537 populate the dcache with a disconneced dentry while the inode
1538 number is in the process of being reallocated. This causes an
1539 assertion failure in iopen since the inode's dentry list contains
1540 both a connected and disconnected dentry.
1544 Description: assertion failure in ldlm_handle2lock()
1545 Details : fix a race between class_handle_unhash() and class_handle2object()
1546 introduced in lustre 1.6.5 by bug 13622.
1551 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1552 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1557 Description: oops in sd_iostats_seq_show()
1558 Details : unloading/reloading the scsi low level driver triggers a kernel
1559 bug when trying to access the sd iostat file.
1564 Description: Kernel panics during QLogic driver reload
1565 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1566 patch, causing memory corruption.
1571 Description: journal_dev option does not work in b1_6
1572 Details : pass mount option during pre-mount.
1574 Severity : enhancement
1576 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1577 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1578 extent information of a file. It can be used to map logical blocks
1579 in a file to physical blocks in the block device.
1583 Description: LDLM soft lockups - improvement
1584 Details : It is be possible to send the lock handle along with each read
1585 or write request because the client is already doing a lock match
1586 itself so there isn't any reason the OST should have to re-do that
1590 Frequency : only on Cray X2
1592 Description: X2 build failures
1593 Details : fix build failures on Cray X2.
1597 Description: xid & resent requests
1598 Details : Initialize RPC XID from clock at startup (randomly if clock is
1601 Severity : enhancement
1603 Description: Add lustre_start utility to start or stop multiple Lustre servers
1608 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1609 Details : In case of memory pressure, list_del() can be called twice on
1610 req->rq_history_list, causing a kernel oops.
1614 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1615 Details : fix stack overflow in the distributed lock manager by defering
1616 export eviction after a failed AST to the elt thread instead of
1617 handling it in the dlm interpret routine.
1621 Description: Convert some comments to new format.
1622 Details : Update documenting comments to match doxygen conventions.
1626 Description: Grammar fixes.
1627 Details : A couple of trivial spelling fixes.
1631 Description: OSD_COUNTERS-mandatory
1632 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1633 to simplify the code.
1637 Description: simplify cmm_device freeing logic.
1638 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1642 Description: Add lockdep support to dt_object_operations locking interface.
1643 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1644 parameter indicating lock ordering. Update mdd code to use new
1649 Description: Introduce failloc constants for lockless IO tests.
1650 Details : Add two new failloc constants to test lockless IO. Only one of
1651 them in implemented---another is checked in yet to be landed
1656 Description: Add lockdep support for inode mutex.
1657 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1658 in the situations where Lustre has to lock more than one inode
1663 Description: Add optional invariants checking support.
1664 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1665 configure switch. This macro is to be used for consistency and
1666 sanity checks that are too expensive to be left in `production'
1671 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1672 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1677 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1678 Details : Use spin_lock_nested() in (the only) situation where more than
1679 one ldlm_lock is locked simultaneously. Also, fix possible
1680 dead-lock in ldlm_lock_change_resource() by enforcing particular
1685 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1686 Details : Instead of specifying each ldlm_lock call-back through separate
1687 parameter, wrap them into struct ldlm_callback_suite.
1691 Description: Kill join_lru obd method and its callers.
1692 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1693 in memory---a requirement implemented through ->o_join_lru() obd
1694 method in HEAD. Remove this method and its users.
1698 Description: Add asynchronous ldlm ENQUEUE completion handler.
1699 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1700 case that stock ldlm_completion_ast() cannot handle as it waits
1701 until lock is granted. Introduce new ldlm_completion_ast_async()
1702 for this. Also comment ldlm_completion_ast().
1706 Description: ldlm_error <-> errno conversion.
1707 Details : Add functions to map (rather arbitrary) between LDLM error codes
1708 and standard errno values. CLIO needs this to prevent LDLM specific
1709 constants from escaping ldlm and osc.
1713 Description: Kill unused ldlm_handle2lock_ns() function.
1714 Details : Kill unused ldlm_handle2lock_ns() function.
1718 Description: Add lu_ref support to ldlm_lock
1719 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1720 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1721 and ldlm_resource. LDLM interface has to be changed, because code that
1722 releases a reference on a lock, has to "know" what reference this is.
1723 In the most frequent case
1725 lock = ldlm_handle2lock(handle);
1727 LDLM_LOCK_PUT(lock);
1729 no changes are required. When any other reference (received _not_ from
1730 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1731 instead of LDLM_LOCK_PUT().
1733 Arguably, changes are pervasive, and interface requires some discipline
1734 for proper use. On the other hand, it was very instrumental in finding
1735 a few leaked lock references.
1739 Description: Add ldlm_lock_addref_try().
1740 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1741 attempts to addref a lock that might be being canceled
1746 Description: Add ldlm_weigh_callback().
1747 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1748 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1749 hard-coded `number of pages' logic.
1753 Description: Add lockdep annotations to llog code.
1754 Details : Use appropriately tagged _nested() locking calls in the places
1755 where llog takes more than one ->lgh_lock lock.
1759 Description: Add loi_kms_set().
1760 Details : Wrap kms updates into a helper function.
1764 Description: Constify instances of struct lsm_operations.
1765 Details : Constify instances of struct lsm_operations.
1769 Description: lu_conf support.
1770 Details : On a server, a file system object is uniquely identified
1771 by a fid, which is sufficient to locate and load all object
1772 state (inode). On a client, on the other hand, more data are
1773 necessary instantiate an object. Change lu_object_find() and
1774 friends to take additional `lu_conf' argument describing object.
1775 Typically this includes layout information.
1779 Description: lu_context fixes.
1780 Details : Introduce new lu_context functions that are needed on the client
1781 side, where some system threads (ptlrpcd) are shared by multiple
1782 modules, and so cannot be stopped during module shutdown.
1786 Description: Add start and stop methods to lu_device_type_operations.
1787 Details : Introduce two new methods in lu_device_type_operations, that are
1788 invoked when first instance of a given type is created and last one
1789 is destroyed respectively. This is need by CLIO.
1793 Description: Add lu_ref support to struct lu_device.
1794 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1795 track leaked references.
1799 Description: Introduce lu_kmem_descr.
1800 Details : lu_kmem_descr and its companion interface allow to create
1801 and destroy a number of kmem caches at once.
1805 Description: Fix lu_object finalization race.
1806 Details : Fix a race between lu_object_find() finding an object and its
1807 concurrent finalization. This race is (most likely) not possible
1808 on the server, but might happen on the client.
1812 Description: Introduce lu_ref interface.
1813 Details : lu_ref is a debugging module allowing to track references to
1814 a given object. It is quite cpu expensive, and has to be
1815 explicitly enabled with --enable-lu_ref. See usage description
1820 Description: Factor lu_site procfs stats into a separate function.
1821 Details : Separate lu_site stats printing code into a separate function
1822 that can be reused on a client.
1826 Description: Constify instances of struct {lu,dt,md}_device_operations.
1827 Details : Constify instances of struct {lu,dt,md}_device_operations.
1831 Description: Introduce struct md_site and move meta-data specific parts of
1832 struct lu_site here.
1833 Details : Move md-specific fields out of struct lu_site into special struct
1834 md_site, so that lu_site can be used on a client.
1838 Description: Kill mdd_lov_destroy().
1839 Details : Remove unused mdd code.
1843 Description: Add st_block checking to multistat.c.
1844 Details : Add st_block checking to multistat.c.
1848 Description: Add lu_ref support to struct obd_device.
1849 Details : Add lu_ref tracking to obd_device.
1853 Description: Kill obd_set_fail_loc().
1854 Details : Remove unused code.
1858 Description: Add special type for ptlrpc_request interpret functions.
1859 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1860 there. Actual usage will be in CLIO.
1864 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1865 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1866 doesn't work with lockdep.
1870 Description: Add rwv.c test program.
1871 Details : New testing program exercising readv(2) and writev(2) (Qian).
1875 Description: Add sendfile.c test program.
1876 Details : New testing program exercising sendfile(2) (Jay).
1880 Description: Ratelimit a message that can be very frequent.
1881 Details : Ratelimit a memory allocation failure message that can
1886 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1887 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1891 Description: Liblustre build fixes.
1892 Details : Liblustre build fixes.
1896 Description: libcfs: add cfs_{need,cond}_resched() interface.
1897 Details : libcfs: add cfs_{need,cond}_resched() definition and
1898 implementations for Linux, NT, and liblustre.
1900 Severity : enhancement
1902 Description: More exported tunables for mballoc
1903 Details : Add support for tunable preallocation window and new tunables for
1904 large/small requests
1908 Description: Detect corruption of block bitmap and checking for preallocations
1909 Details : Checks validity of on-disk block bitmap. Also it does better
1910 checking of number of applied preallocations. When corruption is
1911 found, it turns filesystem readonly to prevent further corruptions.
1915 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1916 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1917 window to 0. If it does hit cache miss, and read-pattern is still
1918 stride-io mode, does not reset the stride window, but also does not
1919 increase the stride window length in this case.
1923 Frequency : only for big-endian servers
1924 Description: Check if system is big-endian while mounting fs with extents feature
1925 Details : Mounting a filesystem with extents feature will fail on big-endian
1926 systems since ext3-based ldiskfs is not supported on big-endian
1927 systems. This can be over-riden with "bigendian_extents" mount option.
1929 Severity : enhancement
1931 Description: The root squash functionality
1932 Details : A security feature, which is to prevent users from being able
1933 to mount lustre on their desktop, run as root, and delete
1934 all of the files in the filesystem. The goal is accomplished by
1935 remapping user id (UID) and group id (GID) of the root user to
1936 a UID and GID specified by the system administartor via Lustre
1937 configuration management server (MGS). The functionality also
1938 allows to specify sets of clients for which the remapping does
1943 Description: Excessive recovery window
1944 Details : With AT enabled, the recovery window can be excessively long (6000+
1945 seconds). To address this problem, we no longer use
1946 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1947 timeout no longer depends on the service time, it is set to
1948 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1949 time via pb_service_time.
1953 Description: Watchdog triggered on MDS failover
1954 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
1955 the OSTs know that the MDS "UUID" can be reused for the same export
1956 from a different NID, so we do not need to wait for the export to be
1960 Frequency : rare, only if using MMP with Linux RAID
1962 Description: MMP doesn't work with Linux RAID
1963 Details : While using HA for Lustre servers with Linux RAID, it is possible
1964 that MMP will not detect multiple mounts. To make this work we
1965 need to unplug the device queue in RAID when the MMP block is being
1966 written. Also while reading the MMP block, we should read it from
1967 disk and not the cached one.
1969 Severity : enhancement
1971 Description: open file using fid
1972 Details : A file can be opened using just its fid, like
1973 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
1975 --------------------------------------------------------------------------------
1977 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1979 * Support for kernels:
1980 2.6.5-7.283 (SLES 9),
1981 2.6.9-55.EL (RHEL 4),
1982 2.6.16.46-0.14 (SLES 10),
1983 2.6.18.8 vanilla (kernel.org)
1984 * Client support for unpatched kernels:
1985 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1986 2.6.16 - 2.6.22 vanilla (kernel.org)
1987 * Due to recently discovered recovery problems, we do not recommend
1988 using patchless RHEL 4 clients with this or any earlier release.
1989 * Recommended e2fsprogs version: 1.39.cfs8
1990 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1991 * Starting with this release, the ldiskfs backing filesystem required
1992 by Lustre is now in its own package, lustre-ldiskfs. This package
1993 should be installed. It is versioned separately from Lustre and
1994 may be released separately in future.
1998 Description: block reactivating mgc import until all deactivates complete
1999 Details : Fix race when failing back MDT/MGS to itself (testing)
2002 Frequency : at statup only
2004 Description: mds_lov_synchronize race leads to various problems
2005 Details : simultaneous MDT->OST connections at startup can cause the
2006 sync to abort, leaving the OSC in a bad state.
2008 Severity : enhancement
2010 Description: add optional extra BUILD_VERSION info
2011 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2012 to override the lustre version.
2015 Frequency : 2.6.18 servers only
2017 Description: ll_kern_mount() doesn't release the module reference
2018 Details : The ldiskfs module reference count never drops down to 0
2019 because ll_kern_mount() doesn't release the module reference.
2024 Description: server LBUG when using old ost_num_threads parameter
2025 Details : Accept the old ost_num_threads parameter but warn that it
2026 is deprecated, and fix an off-by-one error that caused an LBUG.
2031 Description: Transient SCSI error results in persistent IO issue
2032 Details : iobuf->dr_error is not reinitialized to 0 between two
2036 Frequency : sometimes when underlying device returns I/O errors
2038 Description: OSTs not going read-only during write failures
2039 Details : OSTs are not remounted read-only when the journal commit threads
2040 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2041 instead of the ext3 wrappers.
2045 Description: poor connect scaling with increasing client count
2046 Details : Don't run filter_grant_sanity_check for more than 100 exports
2047 to improve scaling for large numbers of clients.
2050 Frequency : SLES10 only
2052 Description: sanity-quota.sh quotacheck failed: rc = -22
2053 Details : Quotas cannot be enabled on SLES10.
2056 Frequency : liblustre clients only
2058 Description: getdirentries does not give error when run on compute nodes
2059 Details : getdirentries does not fail when the size specified as an argument
2060 is too small to contain at least one entry
2062 Severity : enhancement
2064 Description: Add LNET router traceability for debug purposes
2065 Details : If a checksum failure occurs with a router as part of the
2066 IO path, the NID of the last router that forwarded the bulk data
2067 is printed so it can be identified.
2072 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2073 Details : Due to a race condition, liblustre clients were occasionally
2074 evicted incorrectly.
2076 Severity : enhancement
2078 Description: lfs setstripe use optional parameters instead of postional
2081 Severity : enhancement
2083 Description: Nanosecond timestamp support for ldiskfs
2084 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2085 resolution timestamps. There is not yet support for this at
2086 the Lustre filesystem level.
2089 Frequency : during server recovery
2091 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2092 Details : request with rq_no_resend flag not awake l_wait_event if they get
2096 Frequency : nfs export on patchless client
2098 Description: connectathon hang when test nfs export over patchless client
2099 Details : Disconnected dentry cannot be found with lookup, so we do not need
2100 to unhash it or make it invalid
2103 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2104 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2105 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2110 Description: open req refcounting wrong on reconnect
2111 Details : If reconnect happened between getting open reply from server and
2112 call to mdc_set_replay_data in ll_file_open, we will schedule
2113 replay for unreferenced request that we are about to free.
2114 Subsequent close will crash in variety of ways.
2115 Check that request is still eligible for replay in
2116 mdc_set_replay_data().
2121 Description: disable writes to filesystem when reading health_check file
2122 Details : the default for reading the health_check proc file has changed
2123 to NOT do a journal transaction and write to disk, because this
2124 can cause reads of the /proc file to hang and block HA state
2125 checking on a healthy but otherwise heavily loaded system. It
2126 is possible to return to the previous behaviour during configure
2127 with --enable-health-write.
2129 Severity : enhancement
2131 Description: 64-bit inode version
2132 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2133 to the inode. This will be required for version-based recovery.
2138 Description: MDS fails to start if a duplicate client export is detected
2139 Details : in some rare cases it was possible for a client to connect to
2140 an MDS multiple times. Upon recovery the MDS would detect this
2141 and fail during startup. Handle this more gracefully.
2143 Severity : enhancement
2145 Description: Add -o localflock option to simulate old noflock
2147 Details : This will achieve local-only flock/fcntl locks
2153 Description: log_commit_thread vs filter_destroy race leads to crash
2154 Details : Take import reference before releasing llog record semaphore
2159 Description: Wrong request locking in request set processing
2160 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2161 imp_delayed_list, in this place should be used imp_lock.
2164 Frequency : when reconnection
2166 Description: Grant Leak when osc reconnect to OST
2167 Details : When osc reconnect ost, OST(filter) should check whether it
2168 should grant more space to client by comparing fed_grant and
2169 cl_avail_grant, and return the granted space to client instead
2170 of "new granted" space, because client will call osc_init_grant
2171 to update the client grant space info.
2174 Frequency : when client reconnect to OST
2176 Description: Grant Leak when osc do resend and replay bulk write
2177 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2178 bulk write request, because the grant info while be sync between
2179 OSC and OST when reconnect, and we should ignore the grant info
2180 these of resend/replay write req.
2185 Description: Grant space more than avaiable left space sometimes.
2186 Details : When then OST is about to be full, if two bulk writing from
2187 different clients came to OST. Accord the avaliable space of the
2188 OST, the first req should be permitted, and the second one
2189 should be denied by ENOSPC. But if the seconde arrived before
2190 the first one is commited. The OST might wrongly permit second
2191 writing, which will cause grant space > avaiable space.
2194 Frequency : when client is evicted
2196 Description: Grant might be wrongly erased when osc is evicted by OST
2197 Details : when the import is evicted by server, it will fork another
2198 thread ptlrpc_invalidate_import_thread to invalidate the
2199 import, where the grant will be set to 0. While the original
2200 thread will update the grant it got when connecting. So if
2201 the former happened latter, the grant will be wrongly errased
2202 because of this race.
2207 Description: Checking Stale with correct fid
2208 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2209 to check whether it is stale, because sometimes, we want the
2210 enqueue happened anyway, and op_data.fid2 will not be initialized.
2212 Severity : enhancement
2214 Description: update patchless client
2215 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2218 Frequency : only with 2.4 kernel
2220 Description: random memory corruption
2221 Details : size of struct ll_inode_info is to big for union inode.u and this
2222 can be cause of random memory corruption.
2227 Description: Memory leak in recovery
2228 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2229 It should also check obd_fail before fsfilt_start, otherwise if
2230 fsfilt_start return -EROFS,(failover mds during mds recovery).
2231 then the req will return with repmsg->transno = 0 and rc = EROFS.
2232 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2233 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2234 be freed no matter whether fsfilt_commit success or not.
2237 Frequency : only with huge count clients
2239 Description: Prevents from taking the superblock lock in llap_from_page for
2241 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2242 for llap_from_page call in ll_removepage prevents from taking the
2243 superblock lock for a soon died page.
2248 Description: Not check open intent error before release open handle
2249 Details : in some rare cases, the open intent error is not checked before
2250 release open handle, which may cause
2251 ASSERTION(open_req->rq_transno != 0), because it tries to release
2252 the failed open handle.
2257 Description: Set cat log bitmap only after create log success.
2258 Details : in some rare cases, the cat log bitmap is set too early. and it
2259 should be set only after create log success.
2263 Description: Accessing a block bevice can re-enable I/O when Lustre is
2264 tearing down a device.
2265 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2270 Description: service threads may hog cpus when there are a lot of requests
2271 Details : Insert cond_resched to give other threads a chance to use some CPU
2276 Description: the cat log was not initialized in recovery
2277 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2278 unlink log on mds will not be initialized until mds post
2279 recovery. And also in mds post recovery, the unlink log will
2280 initialization will be done asynchronausly, so there will be race
2281 between add unlink log and unlink log initialization.
2285 Description: brw_stats were being printed incorrectly
2286 Details : brw_stats were being printed as log2 but all of them were not
2287 recorded as log2. Also remove some code duplication arising from
2288 filter_tally_{read,write}.
2292 Frequency : rare, only in recovery.
2293 Description: ASSERTION(req->rq_type != LI_POISON) failed
2294 Details : imp_lock should be held while iterating over imp_sending_list for
2295 prevent destroy request after get timeout in ptlrpc_queue_wait.
2299 Description: replay-single.sh test 52 fails
2300 Details : A lock's skiplist need to be cleanup when it being unlinked
2301 from its resource list.
2305 Description: Short directio read returns full requested size rather than
2307 Details : Direct I/O operations should return actual amount of bytes
2308 transferred rather than requested size.
2310 Severity : enhancement
2312 Description: metadata RPC reduction (e.g. for rm performance)
2313 Details : decrease the amount of synchronous RPC between clients and servers
2314 by canceling conflicing lock before the operation on the client
2315 and packing thier handles into the main operation RPC to server.
2317 Severity : enhancement
2319 Description: Async OSC create to avoid the blocking unnecessarily.
2320 Details : If a OST has no remain object, system will block on the creating
2321 when need to create a new object on this OST. Now, ways use
2322 pre-created objects when available, instead of blocking on an
2323 empty osc while others are not empty. If we must block, we block
2324 for the shortest possible period of time.
2328 Description: improve handling recoverable errors
2329 Details : if request processig with error which can be recoverable on server
2330 request should be resend, otherwise page released from cache and
2333 Severity : enhancement
2335 Description: refine locking for avoid write wrong info into lov_objid file
2336 Details : fix possible races with add new target and write/update data in
2339 --------------------------------------------------------------------------------
2341 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2346 Frequency : on some architectures
2348 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2349 Details : Uninitialized flags sometimes cause configuration commands to
2353 Frequency : patchless clients only
2355 Description: missing __iget() symbol export
2356 Details : The __iget() symbol export is missing. To avoid the need for
2357 this on patchless clients the deathrow inode reaper is turned
2358 off, and we depend on the VM to clean up old inodes. This
2359 dependency was during via the fix for bug 12181.
2361 --------------------------------------------------------------------------------
2363 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2365 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2366 INTEROPERATE with older versions automatically. In many cases a
2367 special upgrade step is needed. Please read the
2368 user documentation before upgrading any part of a live system.
2369 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2370 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2371 * WARNING: Lustre configuration and startup changes are required with
2372 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2376 Severity : enhancement
2378 Description: MountConf
2379 Details : Lustre configuration is now managed via mkfs and mount
2380 commands instead of lmc and lconf. New obd types (MGS, MGC)
2381 are added for dynamic configuration management. See
2382 https://mail.clusterfs.com/wikis/lustre/MountConf for
2385 Severity : enhancement
2387 Description: dynamic OST addition
2388 Details : OSTs can now be added to a live filesystem
2390 Severity : enhancement
2392 Description: startup order invariance
2393 Details : MDTs and OSTs can be started in any order. Clients only
2394 require the MDT to complete startup.
2396 Severity : enhancement
2398 Description: parallel, asynchronous orphan cleanup
2399 Details : orphan cleanup is now performed in separate threads for each
2400 OST, allowing parallel non-blocking operation.
2402 Severity : enhancement
2404 Description: optimized stripe assignment
2405 Details : stripe assignments are now made based on ost space available,
2406 ost previous usage, and OSS previous usage, in order to try
2407 to optimize storage space and networking resources.
2409 Severity : enhancement
2411 Description: Permanently set tunables
2412 Details : All writable /proc/fs/lustre tunables can now be permanently
2413 set on a per-server basis, at mkfs time or on a live system.
2415 Severity : enhancement
2417 Description: Lustre message v2
2418 Details : Add lustre message format v2.
2420 Severity : enhancement
2422 Description: client OST exclusion list
2423 Details : Clients can be started with a list of OSTs that should be
2424 declared "inactive" for known non-responsive OSTs.
2428 Description: SPEC SFS validation failure on NFS v2 over lustre.
2429 Details : Changes the blocksize for regular files to be 2x RPC size,
2430 and not depend on stripe size.
2432 Severity : enhancement
2434 Description: Multiple MD RPCs in flight.
2435 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2436 To avoid overly-overloading MDS, introduce a limit on number of
2437 MDS RPCs in flight for a single client and add /proc controls
2438 to adjust this limit.
2440 Severity : enhancement
2442 Description: client read/write statistics
2443 Details : Add client read/write call usage stats for performance
2444 analysis of user processes.
2445 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2446 file access. extents_stats shows chunk size distribution.
2447 extents_stats_per_process show chunk size distribution per
2450 Severity : enhancement
2452 Description: mds statistics
2453 Details : Add detailed mds operations statistics in
2454 /proc/fs/lustre/mds/*/stats.
2458 Description: Failure of copying files with lustre special EAs.
2459 Details : Client side always return success for setxattr call for lustre
2460 special xattr (currently only "trusted.lov").
2465 Description: Refcount LNET uuids
2466 Details : The global LNET uuid list grew linearly with every startup;
2467 refcount repeated list entries instead of always adding to
2470 Severity : enhancement
2472 Description: Dynamic service threads
2473 Details : Within a small range, start extra service threads
2474 automatically when the request queue builds up.
2477 Frequency : mixed-endian client/server environments
2479 Description: mixed-endian crashes
2480 Details : The new msg_v2 system had some failures in mixed-endian
2483 Severity : enhancement
2485 Description: Easy OST removal
2486 Details : OSTs can be permanently deactivated with e.g. 'lctl
2487 conf_param lustre-OST0001.osc.active=0'
2489 Severity : enhancement
2491 Description: MGS proc entries
2492 Details : Added basic proc entries for the MGS showing what filesystems
2495 Severity : enhancement
2497 Description: provide MGS failover
2498 Details : Added config lock reacquisition after MGS server failover.
2500 Severity : enhancement
2502 Description: add Linux 2.4 support
2503 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2507 Description: a large application tries to do I/O to the same resource and dies
2508 in the middle of it.
2509 Details : Check the req->rq_arrival time after the call to
2510 ost_brw_lock_get(), but before we do anything about
2511 processing it & sending the BULK transfer request. This
2512 should help move old stale pending locks off the queue as
2513 quickly as obd_timeout.
2516 Frequency : when an incorrect nid is specified during startup
2518 Description: ptlrpc connect to non-existant node causes kernel crash
2519 Details : LNET can't be re-entered from an event callback, which
2520 happened when we expire a message after the export has been
2521 cleaned up. Instead, hand the zombie cleanup off to another
2524 Severity : enhancement
2526 Description: plain/inodebits lock performance improvement
2527 Details : Grouping plain/inodebits in granted list by their request modes
2528 and bits policy, thus improving the performance of search through
2532 Frequency : only if OST filesystem is corrupted
2534 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2535 Details : for a short time RPCs with bulk IO are in the replay list,
2536 but replay of bulk IOs is unimplemented. If the OST filesystem
2537 is corrupted due to disk cache incoherency and then replay is
2538 started it is possible to trip an assertion. Avoid putting
2539 committed RPCs into the replay list at all to avoid this issue.
2542 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2544 Description: System hang on startup
2545 Details : This bug allowed the liblustre (e.g. catamount) client to
2546 return to the app before handling all startup RPCs. This
2547 could leave the node unresponsive to lustre network traffic
2548 and manifested as a server ptllnd timeout.
2550 Severity : enhancement
2552 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2553 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2554 boolean to control whether to print peer debug info when a
2555 client's RPC times out.
2558 Frequency : only for kernels with patches from Lustre below 1.4.3
2560 Description: Remove old rdonly API
2561 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2564 Frequency : only for devices with external journals
2566 Description: Set external device read-only also
2567 Details : During a commanded failover stop, we set the disk device
2568 read-only while the server shuts down. We now also set any
2569 external journal device read-only at the same time.
2572 Frequency : when upgrading from 1.4 while trying to change parameters
2574 Description: The wrong (new) MDC name was used when setting parameters for
2575 upgraded MDT's. Also allows changing of OSC (and MDC)
2576 parameters if --writeconf is specified at tunefs upgrade time.
2579 Frequency : when setting specific ost indicies
2581 Description: QOS code breaks on skipped indicies
2582 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2583 created with --index need not be sequential.
2585 Severity : enhancement
2587 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2588 Details : The uninit_groups feature works in conjunction with the kernel
2589 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2590 up the pass1 processing of e2fsck. This is a read-only feature
2591 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2592 mount filesystems that have had this feature enabled.
2594 Severity : enhancement
2596 Description: Improve multi-block allocation algorithm to avoid fragmentation
2597 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2598 allocation locality and avoid filesystem fragmentation.
2600 ------------------------------------------------------------------------------
2602 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2604 * Support for kernels:
2605 2.4.21-47.0.1.EL (RHEL 3)
2606 2.6.5-7.283 (SLES 9)
2607 2.6.9-42.0.10.EL (RHEL 4)
2608 2.6.12.6 vanilla (kernel.org)
2609 2.6.16.27-0.9 (SLES 10)
2610 * Recommended e2fsprogs version: 1.39.cfs5
2612 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2616 Frequency : occasional, depends on client load and configuration
2617 Bugzilla : 12181, 12203
2618 Description: data loss for recently-modified files
2620 Details : In some cases it is possible that recently written or created
2621 files may not be written to disk in a timely manner (this should
2622 normally be within 30s unless client IO load is very high).
2623 The problem appears as zero-length files or files that are a
2624 multiple of 1MB in size after a client crash or client eviction
2625 that are missing data at the end of the file.
2627 This problem is more likely to be hit on clients where files are
2628 repeatedly created and unlinked in the same directory, clients
2629 have a large amount of RAM, have many CPUs, the filesystem has
2630 many OSTs, the clients are rebooted frequently, and/or the files
2631 are not accessed by other nodes after being written.
2633 The presence of the problem can be detected by looking at
2634 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2635 smaller than the second (nr_unused) then dirty files will not
2636 be flushed automatically to disk. "sync; sleep 10" should be
2637 run several times on the node before unmounting it to update
2638 Lustre (this is also safe to run on nodes without this problem).
2640 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2641 that can cause this same problem, so customers using that kernel
2642 also need to update the kernel in addition to Lustre. In order
2643 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2645 It is normal that files written just before a client crash (less
2646 than 30s) may not yet have been flushed to disk, even for local
2650 Frequency : frequent on thin XT3 nodes
2652 Description: UUID collision on thin XT3 Linux nodes
2653 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2654 randomly, since we relied on an insufficiently-seeded PRNG.
2659 Description: OSS hangs after "All ost request buffers busy"
2660 Details : A deadlock between quota and journal operations caused OSS
2661 hangs after printing "All ost request buffers busy."
2664 Frequency : always on liblustre builds
2666 Description: Cleanup compiler warnings on liblustre
2669 Frequency : always on liblustre builds on XT3
2671 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2675 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2676 > 32000 subdirectories
2677 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2678 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2679 subdirectory count crosses 32000. This will aid e2fsck to
2680 correctly handle more than 32000 subdirectories.
2683 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2685 Description: System hang on startup
2686 Details : This bug allowed the liblustre (e.g. catamount) client to
2687 return to the app before handling all startup RPCs. This
2688 could leave the node unresponsive to lustre network traffic
2689 and manifested as a server ptllnd timeout.
2691 Severity : enhancement
2693 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2694 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2695 boolean to control whether to print peer debug info when a
2696 client's RPC times out.
2701 Description: make O_SYNC working on 2.6 kernels
2702 Details : 2.6 kernels use different method for mark pages for write,
2703 so need add a code to lustre for O_SYNC work.
2708 Description: Failure to close file and release space on NFS
2709 Details : Put inode details into lock acquired in ll_intent_file_open.
2710 Use mdc_intent_lock in ll_intent_open to properly
2711 detect all kind of errors unhandled by mdc_enqueue.
2716 Description: proc file read during shutdown sometimes raced obd removal,
2718 Details : Add lock to prevent obd access after proc file removal.
2721 Frequency : Only for files larger than 4GB on 32-bit clients.
2723 Description: improperly doing page alignment of locks
2724 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2725 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2730 Description: RPCs being resent when they shouldn't be
2731 Details : Some RPCs that should not be resent are being resent. This
2732 can cause inconsistencies in the RPC state machine. Do not
2733 resend such requests.
2736 Frequency : rare, only with NFS export
2738 Description: Crash on NFS re-export node
2739 Details : under very unusual load conditions an assertion is hit in
2740 ll_intent_file_open()
2743 Frequency : only if OST filesystem is corrupted
2745 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2746 Details : for a short time RPCs with bulk IO are in the replay list,
2747 but replay of bulk IOs is unimplemented. If the OST filesystem
2748 is corrupted due to disk cache incoherency and then replay is
2749 started it is possible to trip an assertion. Avoid putting
2750 committed RPCs into the replay list at all to avoid this issue.
2755 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2756 Details : Large single O_DIRECT read and write calls can fail to allocate
2757 a sufficiently large buffer to process the request. In case of
2758 allocation failure the allocation is retried with a smaller
2759 buffer and broken into smaller requests.
2761 Severity : enhancement
2763 Description: Add -o localflock option to simulate old noflock behaviour.
2764 Details : This will achieve local-only flock/fcntl locks coherentness.
2769 Description: versioning check is incomplete
2770 Details : Checking the version difference of client vs. server, report
2771 error if the gap is too big.
2776 Description: add support PG_writeback bit
2777 Details : add support for PG_writeback bit for Lustre, for more carefull
2778 work with page cache in 2.6 kernel. This also fix some deadlocks
2779 and remove hack for work O_SYNC with 2.6 kernel.
2781 Severity : enhancement
2783 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2784 Details : The uninit_groups feature works in conjunction with the kernel
2785 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2786 up the pass1 processing of e2fsck. This is a read-only feature
2787 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2788 mount filesystems that have had this feature enabled.
2790 Severity : enhancement
2792 Description: Improve multi-block allocation algorithm to avoid fragmentation
2793 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2794 allocation locality and avoid filesystem fragmentation.
2796 ------------------------------------------------------------------------------
2798 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2800 * Support for kernels:
2801 2.6.9-42.0.3.EL (RHEL 4)
2802 2.6.5-7.276 (SLES 9)
2803 2.4.21-47.0.1.EL (RHEL 3)
2804 2.6.12.6 vanilla (kernel.org)
2805 2.6.16.21-0.8 (SLES10)
2806 * Recommended e2fsprogs version: 1.39.cfs2-0
2808 * The backwards-compatible /proc/sys/portals symlink has been removed
2809 in this release. Before upgrading, please ensure that you change
2810 any configuration scripts or /etc/sysctl.conf files that access
2811 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2812 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2813 in advance of the upgrade on any system running Lustre 1.4.6 or
2814 newer, since /proc/sys/lnet was added in that version.
2815 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2819 Frequency : only when quota is used
2821 Description: avoid scanning export list for quota master
2822 Details : Change the algorithms to avoid scanning export list in order
2823 to improve the efficiency.
2826 Frequency : MDS failover only, very rarely
2828 Description: "went back in time" messages on mds failover
2829 Details : The greatest transno may be lost when the current operation
2830 finishes with an error (transno==0) and the client's last_rcvd
2831 record is over-written. Save the greatest transno in the
2832 mds_last_transno for this case.
2835 Frequency : always for specific kernels and striping counts
2837 Description: client may get "Matching packet too big" without ACL support
2838 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2839 when trying to access files in certain configurations. The
2840 clients should in fact be denied when mounting because they do
2841 not understand ACLs.
2844 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2846 Description: many clients connecting with IO in progress causes connect timeouts
2847 Details : Avoid synchronous journal commits to avoid delays caused by many
2848 clients connecting/disconnecting when bulk IO is in progress.
2849 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2850 OST_IO_PORTAL to avoid delays behind potentially many pending
2854 Frequency : occasionally with multiple writers to a single file
2856 Description: shared writes to file may result in wrong size reported by stat()
2857 Details : Allow growing of kms when extent lock is cancelled
2860 Frequency : always with random mmap IO to multi-striped file
2862 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2863 Details : Only if the hole is at the end of OST object so that kms is too
2864 small. Fix is to increase kms accordingly in ll_nopage.
2867 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2869 Description: writes to a missing object would leak memory on the OST
2870 Details : If there is an inconsistency between the MDS and OST filesystems,
2871 such that the MDS references an object that doesn't exist, writes
2872 to that object will leak memory due to incorrect cleanup in the
2873 error handling path, eventually running out of memory on the OST.
2878 Description: Creating too long symlink causes lustre errors
2879 Details : Check symlink and name lengths before sending requests to MDS.
2882 Frequency : only if flock is enabled (not on by default)
2884 Description: posix locks not released on fd closure on 2.6.9+
2885 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2886 kernels, this caused such locks not to be released on fd close and
2887 then assertions on fs unmount about still used locks.
2890 Frequency : MDS failover only, very rarely
2892 Description: clients may get ASSERTION(granted_lock != NULL)
2893 Details : When request was taking a long time, and a client was resending
2894 a getattr by name lock request. The were multiple lock requests
2895 with the same client lock handle and
2896 mds_getattr_name->fixup_handle_for_resent_request found one of the
2897 lock handles but later failed with ASSERTION(granted_lock != NULL).
2902 Description: handle->h_buffer_credits > 0, assertion failure
2903 Details : h_buffer_credits is zero after truncate, causing assertion
2904 failure. This patch extends the transaction or creates a new
2908 Frequency : NFS re-export or patchless client
2909 Bugzilla : 11179, 10796
2910 Description: Crash on NFS re-export node (__d_move)
2911 Details : We do not want to hash the dentry if we don't have a lock.
2912 But if this dentry is later used in d_move, we'd hit uninitialised
2913 list head d_hash, so we just do this to init d_hash field but
2914 leave dentry unhashed.
2917 Frequency : NFS re-export or patchless client
2919 Description: NFS exports has problem with symbolic link
2920 Details : lustre client didn't properly install dentry when re-exported
2921 to NFS or running patchless client.
2924 Frequency : NFS re-export or patchless client
2926 Description: Various nfs/patchless fixes.
2927 Details : fixes reuse disconected alias for lookup process - this fixes
2928 warning "find_exported_dentry: npd != pd",
2929 fix permission error with open files at nfs.
2930 fix apply umask when do revalidate.
2933 Frequency : occasional
2935 Description: Crash on NFS re-export node
2936 Details : calling clear_page() on the wrong pointer triggered oops in
2937 generic_mapping_read().
2940 Frequency : rarely, using O_DIRECT IO
2942 Description: unaligned directio crashes client with LASSERT
2943 Details : check for unaligned buffers before trying any requests.
2946 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2948 Description: stale data returned from RAID cache
2949 Details : If only a small amount of IO is done to the RAID device before
2950 reading it again it is possible to get stale data from the RAID
2951 cache instead of reading it from disk.
2954 Frequency : always for sles10 kernel
2956 Description: sles10 support
2957 Details : ll_follow_link: compile fixes and using of nd_set_link
2958 under newer kernels.
2961 Frequency : depends on arch, kernel and compiler version, always on sles10
2964 Description: recursive or deep enough symlinks cause stack overflow
2965 Details : getting rid of large stack-allocated variable in
2969 Frequency : depends on hardware
2971 Description: lustre write performance loss in the SLES10 kernel
2972 Details : the performance loss is caused by using of write barriers in the
2973 ext3 code. The SLES10 kernel turns barrier support on by
2974 default. The fix is to undo that change for ldiskfs.
2976 ------------------------------------------------------------------------------
2978 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2980 * Support for kernels:
2981 2.6.9-42.0.3EL (RHEL 4)
2982 2.6.5-7.276 (SLES 9)
2983 2.4.21-47.0.1.EL (RHEL 3)
2984 2.6.12.6 vanilla (kernel.org)
2988 Frequency : quota enabled and large files being deleted
2990 Description: releasing more than 4GB of quota at once hangs OST
2991 Details : If a user deletes more than 4GB of files on a single OST it
2992 will cause the OST to spin in an infinite loop. Release
2993 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2998 Description: statfs data retrieved from /proc may be stale or zero
2999 Details : When reading per-device statfs data from /proc, in the
3000 {kbytes,files}_{total,free,avail} files, it may appear
3001 as zero or be out of date.
3004 Frequency : systems with MD RAID1 external journal devices
3006 Description: lconf's call to blkid is confused by RAID1 journal devices
3007 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3008 of returning all block devices that match the journal UUID.
3011 Frequency : always, for aggregate stripe size over 4GB
3013 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3014 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3015 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3018 Frequency : always if "lfs find" used on a local file/directory
3020 Description: "lfs find" segfaults if used on a local file/directory
3021 Details : The case where a directory component was not specified wasn't
3022 handled correctly. Handle this properly.
3025 Frequency : always on ppc64
3027 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3028 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3029 perform endianness conversion.
3032 Frequency : rarely (truncate to non-zero file size after write under load)
3033 Bugzilla : 10730, 10687
3034 Description: Files padded with zeros to next 4K multiple
3035 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3036 it is possible that files that are truncated to a non-zero size
3037 immediately after being written are filled with zero bytes beyond
3038 the truncated size. No file data is lost.
3040 Severity : enhancement
3042 Description: Allow recovery/failover for liblustre clients.
3043 Details : liblustre clients were unaware of failover configurations until
3046 Severity : enhancement
3048 Description: user file locks should fail when not mounting with flock option
3049 Details : Set up an error-returning stub in ll_file_operations.lock field
3050 to prevent incorrect behaviour when client is mounted without
3051 flock option. Also, set up properly f_op->flock field for
3055 Frequency : always on ia64
3057 Description: "lfs df" loops on printing out MDS statfs information
3058 Details : The obd_ioctl_data was not initialized and in some systems
3059 this caused a failure during the ioctl that did not return
3060 an error. Initialize the struct and return an error on failure.
3063 Frequency : SLES 9 only
3065 Description: Error of copying files with lustre special EAs as root
3066 Details : Client side always return success for setxattr call for lustre
3067 special xattr (currently only "trusted.lov").
3070 Frequency : rarely on clusters with both ia64+i386 clients
3072 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3073 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3074 clients are doing concurrent writes to the same file it is
3075 possible that the ia64 clients may LASSERT because the OST
3076 extent locks are not PAGE_SIZE aligned. Ensure that grown
3077 locks are always aligned on the request boundary.
3080 Frequency : specific use, occasional
3082 Description: Overwriting in use executable truncates on-disk binary image
3083 Details : If one node attempts to overwrite an executable in use by
3084 another node, we now correctly return ETXTBSY instead of
3085 truncating the file.
3087 Severity : enhancement
3089 Description: Async OSC create to avoid the blocking unnecessarily.
3090 Details : If a OST has no remain object, system will block on the creating
3091 when need to create a new object on this OST. Now, ways use
3092 pre-created objects when available, instead of blocking on an
3093 empty osc while others are not empty. If we must block, we block
3094 for the shortest possible period of time.
3099 Description: chmod on Lustre root is propagated to other clients
3100 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3101 invalid by the follow_mount time.
3104 Frequency : liblustre clients only
3106 Description: Race in 'instant cancel' lock handling could lead to such locks
3107 never to be granted in case of SMP MDS
3108 Details : Do not destroy not yet granted but cbpending locks in
3112 Frequency : replay/resend of open
3114 Description: non null lock assetion failure in mds_intent_policy
3115 Details : Trying to replay/resend lockless open requests resulted in
3116 mds_open() returning 0 with no lock. Now it sets a flag if
3117 a lock is going to be returned.
3119 Severity : enhancement
3121 Description: Checksum enhancements
3122 Details : New checksum enhancements allow for resending RPCs that failed
3125 Severity : enhancement
3127 Description: Tunables on number of dirty pages in cacche
3128 Details : Allow to set limit on number of dirty pages cached.
3133 Description: client crash on unmount - lock still has references
3134 Details : In some error handling cases it was possible to leak a lock
3135 reference on a client while accessing a file. This was not
3136 harmful to the client during operation, but would cause the
3137 client to crash when the filesystem is unmounted.
3140 Frequency : specific case, rare
3142 Description: ETXTBSY on mds though file not in use by client
3143 Details : ETXTBSY is no longer incorrectly returned when attempting to
3144 chmod or chown a directory that the user previously tried to
3145 execute or a currently-executing binary.
3148 Frequency : extremely rare except on liblustre-based clients
3150 Description: Lustre space not freed when files are deleted
3151 Details : Clean up open-unlinked files after client eviction. Previously
3152 the unlink was skipped and the files remained as orphans.
3157 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3158 Details : In some cases it was possible to send a blocking callback to
3159 a client doing a glimpse, even though that client didn't get
3160 a lock granted. When the glimpse lock is cancelled on the OST
3161 the freed lock is left on the waiting list and corrupted the list.
3164 Frequency : all core dumps
3166 Description: Broke core dumps to lustre
3167 Details : Negative dentry may be unhashed if parent does not have UPDATE
3168 lock, but some callers, e.g. do_coredump, expect dentry to be
3169 hashed after successful create, hash it in ll_create_it.
3171 ------------------------------------------------------------------------------
3173 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3175 * Support for kernels:
3176 2.6.9-42.0.2.EL (RHEL 4)
3177 2.6.5-7.276 (SLES 9)
3178 2.4.21-40.EL (RHEL 3)
3179 2.6.12.6 vanilla (kernel.org)
3183 Frequency : always on RHEL 3
3185 Description: Number of open files grows over time
3186 Details : The number of open files grows over time, whether or not
3187 Lustre is started. This was due to a filp leak introduced
3188 by one of our kernel patches.
3190 ------------------------------------------------------------------------------
3192 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3194 * Support for kernels:
3195 2.6.9-42.EL (RHEL 4)
3196 2.6.5-7.276 (SLES 9)
3197 2.4.21-40.EL (RHEL 3)
3198 2.6.12.6 vanilla (kernel.org)
3203 Bugzilla : 5719, 9635, 9792, 9684
3204 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3205 Details : If a server is under heavy load and cannot reply to new
3206 connection requests before the client resends the (re)connect,
3207 the connection handling code can behave badly if two service
3208 threads are concurrently handing separate (re)connections from
3209 the same client. Add better locking to the connection handling
3210 code, and ensure that only a single connection will be processed
3211 for a given client UUID, even if the lock is dropped.
3213 Severity : enhancement
3215 Description: add TCP zero-copy support to kernel
3216 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3217 sends if the hardware supports scatter-gather and checksumming.
3218 This allows socklnd to do client-write and server-read more
3219 efficiently and reduce CPU utilization from skbuf copying.
3222 Frequency : only if NFS exporting from client
3224 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3225 Details : If setting attributes on a file created under NFS that had
3226 never been opened it would be possible to oops the client
3227 if the file had no objects.
3230 Frequency : always for liblustre
3232 Description: liblustre client does MDS+OSTs setattr RPC for each write
3233 Details : When doing a write from a liblustre client, the client
3234 incorrectly issued an RPC to the MDS and each OST the file was
3235 striped over in order to update the timestamps. When writing
3236 with small chunks and many clients this could overwhelm the MDS
3237 with RPCs. In all cases it would slow down the write because
3238 these RPCs are unnecessary.
3240 Severity : enhancement
3242 Description: allow number of MDS service threads to be changed at module load
3243 Details : It is now possible to change the number of MDS service threads
3244 running. Adding "options mds mds_num_threads={N}" to the MDS's
3245 /etc/modprobe.conf will set the number of threads for the next
3246 time Lustre is restarted (assuming the "mds" module is also
3247 reloaded at that time). The default number of threads will
3248 stay the same, 32 for most systems.
3253 Description: OST crash if filesystem is unformatted or corrupt
3254 Details : If an OST is started on a device that has never been formatted
3255 or if the filesystem is corrupt and cannot even mount then the
3256 error handling cleanup routines would dereference a NULL pointer.
3261 Description: NULL pointer deref in llap_from_page.
3262 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3263 incorrect) mapping if the page was truncated/reclaimed while it was
3264 searched for. Check for this condition and skip such pages when
3265 doing readahead. Introduce extra check to llap_from_page() to
3266 verify page->mapping->host is non-NULL (so page is not anonymous).
3269 Frequency : Sometimes when using sys_sendfile
3271 Description: "page not covered by a lock" warnings from ll_readpage
3272 Details : sendfile called ll_readpage without right page locks present.
3273 Now we introduced ll_file_sendfile that does necessary locking
3274 around call to generic_file_sendfile() much like we do in
3278 Frequency : with certain MDS communication failures at client mount time
3280 Description: NULL pointer deref after failed client mount
3281 Details : a client connection request may delayed by the network layer
3282 and not be sent until after the PTLRPC layer has timed out the
3283 request. If the client fails the mount immediately it will try
3284 to clean up before the network times out the request. Add a
3285 reference from the request import to the obd device and delay
3286 the cleanup until the network drops the request.
3289 Frequency : occasionally during client (re)connect
3291 Description: assertion failure during client (re)connect
3292 Details : processing a client connection request may be delayed by the
3293 client or server longer than the client connect timeout. This
3294 causes the client to resend the connection request. If the
3295 original connection request is replied in this interval, the
3296 client may trip an assertion failure in ptlrpc_connect_interpret()
3297 which thought it would be the only running connect process.
3300 Frequency : only with obd_echo servers and clients that are rebooted
3302 Description: kernel BUG accessing uninitialized data structure
3303 Details : When running an obd_echo server it did not start the ping_evictor
3304 thread, and when a client was evicted an uninitialized data
3305 structure was accessed. Start the ping_evictor in the RPC
3306 service startup instead of the OBD startup.
3308 Severity : enhancement
3309 Bugzilla : 10193 (patchless)
3310 Description: Remove dependency on various unexported kernel interfaces.
3311 Details : No longer need reparent_to_init, exit_mm, exit_files,
3312 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3315 Frequency : rare (only users of deprecated and unsupported LDAP config)
3317 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3318 Details : LDAP apparently contains 'lustreName' attributes instead of
3319 'name'. A simple remapping of the name is sufficient.
3322 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3324 Description: waiting_locks_callback trips kernel BUG if client is evicted
3325 Details : Running with the dump_on_timeout debug flag turned on makes
3326 it possible that the waiting_locks_callback() can try to dump
3327 the Lustre kernel debug logs from an interrupt handler. Defer
3328 this log dumping to the expired_lock_main() thread.
3330 Severity : enhancement
3332 Description: Support NFS exporting on 2.6 kernels.
3333 Details : Implement non-rawops metadata methods for NFS server to use without
3334 changing NFS server code.
3337 Frequency : very rare (synthetic metadata workload only)
3339 Description: two racing renames might cause an MDS thread to deadlock
3340 Details : Running the "racer" program may cause one MDS thread to rename
3341 a file from being the source of a rename to being the target of
3342 a rename at exactly the same time that another thread is doing
3343 so, and the second thread has already enqueued these locks after
3344 doing a lookup of the target and is trying to relock them in
3345 order. Ensure that we don't try to re-lock the same resource.
3348 Frequency : only very large systems with liblustre clients
3350 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3351 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3352 all OSTs in parallel. This allows the eviction of stale liblustre
3353 clients to proceed much faster than if they were done in series,
3354 and also offers similar improvements for other set_info RPCs.
3359 Description: excessive CPU usage during initial read phase on client
3360 Details : During the initial read phase on a client, it would agressively
3361 retry readahead on the file, consuming too much CPU and impacting
3362 performance (since 1.4.5.8). Improve the readahead algorithm
3363 to avoid this, and also improve some other common cases (read
3364 of small files in particular, where "small" is files smaller than
3365 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3370 Description: MDS crash when receiving packet with unknown intent.
3371 Details : Do not LBUG in unknown intent case, just return -EFAULT
3373 Severity : enhancement
3374 Bugzilla : 9293, 9385
3375 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3376 Details : Do not serialize getattr (non-intent version) and statfs.
3379 Frequency : occasional, when OST network is overloaded/intermittent
3381 Description: client evicted by OST after bulk IO timeout
3382 Details : If a client sends a bulk IO request (read or write) the OST
3383 may evict the client if it is unresposive to its data GET/PUT
3384 request. This is incorrect if the network is overloaded (takes
3385 too long to transfer the RPC data) or dropped the OST GET/PUT
3386 request. There is no need to evict the client at all, since
3387 the pinger and/or lock callbacks will handle this, and the
3388 client can restart the bulk request.
3391 Frequency : Always when mmapping file with no objects
3393 Description: client crashes when mmapping file with no objects
3394 Details : Check that we actually have objects in a file before doing any
3395 operations on objects in ll_vm_open, ll_vm_close and
3401 Description: Request leak when working with deleted CWD
3402 Details : Introduce advanced request refcount tracking for requests
3403 referenced from lustre intent.
3405 Severity : Enhancement
3407 Description: Cache open file handles on client.
3408 Details : MDS now will return special lock along with openhandle, if
3409 requested and client is allowed to hold openhandle, even if unused,
3410 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3411 closing files for every read/write openration.
3413 Severity : Enhancement
3415 Description: Cache open negative dentries on client when possible.
3416 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3417 negative dentries on lock revocation.
3422 Description: Remounting a client read-only wasn't possible with a zconf mount
3423 Details : It wasn't possible to remount a client read-only with llmount.
3425 Severity : enhancement
3426 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3427 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3428 MPICH is included to add Lustre-specific ADIO interfaces.
3429 This is based closely on the UFS ADIO layer and only differs
3430 in file creation, in order to allow the OST striping to be set.
3431 This is user-contributed code and not supported by CFS.
3436 Description: extended inode attributes (immutable, append-only) work improperly
3437 when 2.4 and 2.6 kernels are used on client/server or vice versa
3438 Details : Introduce kernel-independent values for these flags.
3440 Severity : enhancement
3443 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3444 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3445 granularity. Now, Fractional MB granularity is be supported,
3446 this is very useful for low memory system.
3448 Severity : enhancement
3450 Description: Getattr by fid
3451 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3452 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3453 have locks and avoids getattr rpc requests for it on every stat.
3456 Frequency : Always, for filesystems larger than 2TB
3458 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3459 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3460 filesystems larger than 2TB to be mounted with mballoc enabled.
3463 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3465 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3466 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3467 filesystems larger than 2TB is mounted then the kernel will
3468 silently corrupt the start of the filesystem. CONFIG_LBD is
3469 enabled for all CFS-supported kernels, but the possibility of
3470 this happening with a modified kernel config exists.
3472 Severity : enhancement
3474 Description: add client O_DIRECT support for 2.6 kernels
3475 Details : It is now possible to do O_DIRECT reads and writes to files
3476 in the Lustre client mountpoint on 2.6 kernel clients.
3478 Severity : enhancement
3480 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3481 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3482 parallel, not waiting for response from every OST before sending
3483 a rpc to the next OST.
3488 Description: setattr vs write race when updating file timestamps
3489 Details : Client processes that update a file timestamp into the past
3490 right after writing to the file (e.g. tar) it is possible that
3491 the updated file modification time can be reset to the current
3492 time due to a race between processing the setattr and write RPC.
3494 Severity : enhancement
3496 Description: Bring 'lfs find' closer in line with regular Linux find.
3497 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3498 -print0 options and obtains all the needed info through the lustre
3501 Severity : enhancement
3503 Description: support up to 1024 configured devices on one node
3504 Details : change obd_dev array from statically allocated to dynamically
3505 allocated structs as they are first used to reduce memory usage
3510 Description: Flush dirty partially truncated pages during truncate
3511 Details : Immediatelly flush partially truncated pages in filter_setattr,
3512 this way we completely avoid having any pages in page cache on OST
3513 and can retire ugly workarounds during writes to flush such pages.
3518 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3519 Details : protect inode from truncation within vfs_unlink() context
3520 just take a reference before calling vfs_unlink() and release it
3521 when parent's i_sem is free.
3526 Description: last_id value checked outside lock on OST caused LASSERT failure
3527 Details : If there were multiple MDS->OST object precreate requests in
3528 flight, it was possible that the OST's last object id was checked
3529 outside a lock and incorrectly tripped an assertion. Move checks
3530 inside locks, and discard old precreate requests.
3533 Frequency : always, if extents are used on OSTs
3535 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3536 Details : OSTs using the extents format would not zero the high 16 bits of
3537 the index physical block number. This is not a problem for any
3538 OST filesystems smaller than 16TB, and no kernels support ext3
3539 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3540 new/modified files) and can be fixed for existing filesystems
3541 with e2fsprogs-1.39-cfs1.
3546 Description: import connection selection may be incorrect if timer wraps
3547 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3548 import connections to be ignored if the 32-bit jiffies counter
3549 wraps. Use a 64-bit jiffies counter.
3552 Frequency : very large clusters immediately after boot
3554 Description: LNET request buffers exhausted under heavy short-term load
3555 Details : If a large number of client requests are generated on a service
3556 that has previously never seen so many requests it is possible
3557 that the request buffer growth cannot keep up with the spike in
3558 demand. Instead of dropping incoming requests, they are held in
3559 the LND until the RPC service can accept more requests.
3562 Frequency : Sometimes during replay
3564 Description: Assertion failure in ll_local_open after replay.
3565 Details : If replay happened on an open request reply before we were able
3566 to set replay handler, reply will become not swabbed tripping the
3567 assertion in ll_local_open. Now we set the handler right after
3568 recognising of open request
3571 Frequency : very rare
3573 Description: kernel reports "badness in vsnprintf"
3574 Details : Reading from the "recovery_status" /proc file in small chunks
3575 may cause a negative length in lprocfs_obd_rd_recovery_status()
3576 call to vsnprintf() (which is otherwise harmless). Exit early
3577 if there is no more space in the output buffer.
3579 Severity : enhancement
3581 Description: clear OBD RPC statistics by writing to them
3582 Details : It is now possible to clear the OBD RPC statistics by writing
3583 to the "stats" file.
3588 Description: Client mtime is not the same on different clients after utimes
3589 Details : In some cases, the client was using the utimes() syscall on
3590 a file cached on another node. The clients now validate the
3591 ctime from the MDS + OSTs to determine which one is right.
3596 Description: Inability to activate failout mode
3597 Details : lconf script incorrectly assumed that in pythong string's numeric
3598 value is used in comparisons.
3601 Frequency : always with multiple stripes per file
3603 Description: Inefficient object allocation for mutli-stripe files
3604 Details : When selecting which OSTs to stripe files over, for files with
3605 a stripe count that divides evenly into the number of OSTs,
3606 the MDS is always picking the same starting OST for each file.
3607 Return the OST selection heuristic to the original design.
3612 Description: mount failures may take full timeout to return an error
3613 Details : Under some heavy load conditions it is possible that a
3614 failed mount can wait for the full obd_timeout interval,
3615 possibly several minutes, before reporting an error.
3616 Instead return an error as soon as the status is known.
3618 Frequency : quota enabled and large files being deleted
3620 Description: releasing more than 4GB of quota at once hangs OST
3621 Details : If a user deletes more than 4GB of files on a single OST it
3622 will cause the OST to spin in an infinite loop. Release
3623 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3628 Description: statfs data retrieved from /proc may be stale or zero
3629 Details : When reading per-device statfs data from /proc, in the
3630 {kbytes,files}_{total,free,avail} files, it may appear
3631 as zero or be out of date.
3634 Frequency : systems with MD RAID1 external journal devices
3636 Description: lconf's call to blkid is confused by RAID1 journal devices
3637 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3638 of returning all block devices that match the journal UUID.
3641 Frequency : always, for aggregate stripe size over 4GB
3643 Description: assertion fails when trying to use 4GB stripe size
3644 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3645 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3648 Frequency : always on ppc64
3650 Description: the first write on an ext3 filesystem with mballoc got stuck
3651 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3652 perform endianness conversion.
3654 ------------------------------------------------------------------------------
3656 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3658 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3659 INTEROPERATE with older versions automatically. Please read the
3660 user documentation before upgrading any part of a live system.
3661 * WARNING: Lustre networking configuration changes are required with
3662 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3665 * Support for kernels:
3666 2.6.9-22.0.2.EL (RHEL 4)
3667 2.6.5-7.244 (SLES 9)
3668 2.6.12.6 vanilla (kernel.org)
3671 Severity : enhancement
3672 Bugzilla : 7981/8208
3673 Description: Introduced Lustre Networking (LNET)
3674 Details : LNET is new networking infrastructure for Lustre, it includes
3675 a reorganized network configuration mode (see the user
3676 documentation for full details) as well as support for routing
3677 between different network fabrics. Lustre Networking Devices
3678 (LNDs) for the supported network fabrics have also been
3679 created for this new infrastructure.
3681 Severity : enhancement
3682 Description: Introduced Access control lists
3683 Details : clients can set ACLs on files and directories in order to have
3684 more fine-grained permissions than the standard Unix UGO+RWX.
3685 The MDS must be started with the "-o acl" mount option.
3687 Severity : enhancement
3688 Description: Introduced filesystem quotas
3689 Details : Administrators may now establish per-user quotas on the
3692 Severity : enhancement
3694 Description: Configuration change for the XT3
3695 The PTLLND is now used to run Lustre over Portals on the XT3
3696 The configure option(s) --with-cray-portals are no longer used.
3697 Rather --with-portals=<path-to-portals-includes> is used to
3698 enable building on the XT3. In addition to enable XT3 specific
3699 features the option --enable-cray-xt3 must be used.
3704 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3705 Details : A race between allocating a new llcd and re-getting the llcd_lock
3706 allowed another thread to grab newly-allocated llcd.
3708 Severity : enhancement
3710 Description: 2.6 OST async journal commit and locking fix to improve performance
3711 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3712 2.6 kernels are now async as they already were in 2.4 kernels so
3713 that they can commit concurrently with the network bulk transfer.
3714 For block-allocated files the filter allocation semaphore is held
3715 to avoid filesystem fragmentation during allocation. BKL lock
3716 removed for 2.6 xattr operations where it is no longer needed.
3721 Description: lconf incorrectly determined whether two IP networks could talk
3722 Details : In some more complicated routing and multiple-network
3723 configurations, lconf will avoid trying to make a network
3724 connection to a disjoint part of the IP space. It was doing the
3725 math incorrectly for one set of cases.
3730 Description: Fix for potential infinite loop processing records in an llog.
3731 Details : If an llog record is corrupted/zeroed, it is possible to loop
3732 forever in llog_process(). Validate the llog record length
3733 and skip the remainder of the block on error.
3736 Frequency : occasional (liblustre only)
3738 Description: liblustre could not open files whose last component is a symlink
3739 Details : sysio_path_walk() would incorrectly pass the open intent to
3740 intermediate path components.
3743 Frequency : rare (liblustre only with non-standard tuning)
3744 Bugzilla : 7201 (7350)
3745 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3746 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3747 after it has been released, only do this for referenced locks
3749 Severity : enhancement
3751 Description: specifying an (invalid) directory default stripe_size of -1
3752 would reset the directory default striping
3753 Details : stripe_size -1 was used internally to signal directory stripe
3754 removal, now use "all default" to signal dir stripe removal
3755 as a directory striping of "all default" is not useful
3758 Frequency : common for large clusters running liblustre clients
3760 Description: doing an ls when liblustre clients are running is slow
3761 Details : sending a glimpse AST to a liblustre client waits for every AST
3762 to time out, as liblustre clients will not respond. Since they
3763 cannot cache data we refresh the OST lock LVB from disk instead.
3765 Severity : enhancement
3767 Description: doing an ls at the same time as file IO can be slow
3768 Details : enqueue and other "small" requests can be blocked behind many
3769 large IO requests. Create a new OST IO portal for non-IO
3770 requests so they can be processed faster.
3773 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3775 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3776 created file with mode 000
3777 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3778 allow an ftruncate() as a non-root user to a file with mode 000.
3779 Lustre now respects this flag to disable mode checks when
3780 truncating a file owned by the user
3783 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3786 Description: Revoking locks from clients that went dead or catatonic might take
3788 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3789 cancellation of such locks instant on servers without waiting for
3790 any reply from clients. Clients drops these locks when cancel
3791 notification from server is received without replying.
3794 Frequency : liblustre-only, when liblustre client dies or becomes busy
3796 Description: Doing ls on Linux clients can take a long time with active
3798 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3799 granting such locks to it in the first place if possible. Locks
3800 are taken by proxy on the OST during the read or write and
3801 dropped immediately afterward. Add connect flags handling, do
3802 not grant locks to liblustre clients for glimpse ASTs.
3804 Severity : enhancement
3806 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3807 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3808 sequential reads and behaves badly with applications doing random
3809 reads. Improve it by reading ahead at least read region, and
3810 avoiding excessive large RPC for small reads.
3812 Severity : enhancement
3814 Description: Creating more than 1000 files for a single job may cause a load
3815 imbalance on the OSTs if there are also a large number of OSTs.
3816 Details : qos_prep_create() uses an OST index reseed value that is an
3817 even multiple of the number of available OSTs so that if the
3818 reseed happens in the middle of the object allocation it will
3819 still utilize the OSTs as uniformly as possible.
3824 Description: OST or MDS may oops in ping_evictor_main()
3825 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3826 but doesn't restart at beginning of obd_exports_timed list
3829 Severity : enhancement
3831 Description: improve by-nid export eviction on the MDS and OST
3832 Details : allow multiple exports with the same NID to be evicted at one
3833 time without re-searching the exports list.
3836 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3838 Description: MDS may oops in groups_free()
3839 Details : in rare race conditions a newly allocated group_info struct is
3840 freed again, and this can be NULL. The 2.4 compatibility code
3841 for groups_free() checked for a NULL pointer, but 2.6 did not.
3844 Frequency : common for liblustre clients doing little filesystem IO
3845 Bugzilla : 9352, 7313
3846 Description: server may evict liblustre clients accessing contended locks
3847 Details : if a client is granted a lock or receives a completion AST
3848 with a blocking AST already set it would not reply to the AST
3849 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3852 Frequency : lfs setstripe, only systems with more than 160 OSTs
3854 Description: unable to set striping with a starting offset beyond OST 160
3855 Details : llapi_create_file() incorrectly limited the starting stripe
3856 index to the maximum single-file stripe count.
3859 Frequency : LDAP users only
3861 Description: lconf did not handle in-kernel recovery with LDAP properly
3862 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3864 Severity : enhancement
3866 Description: bind OST threads to NUMA nodes to improve performance
3867 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3868 node and do their allocations there to localize memory access
3870 Severity : enhancement
3872 Description: llmount can determine client NID directly from Myrinet (GM)
3873 Details : the client NID code from gmnalnid was moved directly into
3874 llmount, removing the need to use this or specifying the
3875 client NID explicitly when mounting GM clients with zeroconf
3878 Frequency : if client is started with down MDS
3880 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3881 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3882 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3883 ptlrpc_import_delayed_req() because we didn't check if the
3884 request was interrupted, and we also didn't break out of the
3885 event loop if there was no timeout
3890 Description: data loss during non-page-aligned writes to a single file from
3891 both multiple nodes and multiple threads on one node at same time
3892 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3893 inconsistency led to false short-reads, that were cached and later
3894 used by ->prepare_write() to fill in partially written page,
3895 leading to data loss.
3898 Frequency : always, if lconf --abort_recovery used
3900 Description: lconf --abort_recovery fails with 'Operation not supported'
3901 Details : lconf was attempting to abort recovery on the MDT device and not
3904 Severity : enhancement
3906 Description: remove cleanup logs
3907 Details : replace lconf-generated cleanup logs with lustre internal
3908 cleanup routines. Eliminates the need for client-cleanup and
3911 Severity : enhancement
3913 Description: add support for EAs (user and system) on lustre filesystems
3914 Details : it is now possible to store extended attributes in the Lustre
3915 client filesystem, and with the user_xattr mount option it
3916 is possible to allow users to store EAs on their files also
3918 Severity : enhancement
3920 Description: Add possibility (config option) to show minimal available OST free
3922 Details : When compiled with --enable-mindf configure option, statfs(2)
3923 (and so, df) will return least minimal free space available from
3924 all OSTs as amount of free space on FS, instead of summary of
3925 free spaces of all OSTs.
3927 Severity : enhancement
3929 Description: do not expand extent locks acquired on OST-side
3930 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3931 by server: they are not cached anyway.
3934 Frequency : when mmap is used/binaries executed from Lustre
3936 Description: Unmmap pages before throwing them away from read cache.
3937 Details : llap_shrink cache now attempts to unmap pages before discarding
3938 them (if unmapping failed - do not discard). SLES9 kernel has
3939 extra checks that trigger if this unmapping is not done first.
3944 Description: lconf didn't resolve symlinks before checking to see whether a
3945 given mountpoint was already in use
3948 Frequency : when migrating failover services
3949 Bugzilla : 6395, 9514
3950 Description: When migrating a subset of services from a node (e.g. failback
3951 from a failover service node) the remaining services would
3952 time out and evict clients.
3953 Details : lconf --force (implied by --failover) sets the global obd_timeout
3954 to 5 seconds in order to quickly disconnect, but this caused
3955 other RPCs to time out too quickly. Do not change the global
3956 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3958 Severity : enhancement
3959 Frequency : if MDS is started with down OST
3960 Bugzilla : 9439,5706
3961 Description: Allow startup/shutdown of an MDS without depending on the
3962 availability of the OSTs.
3963 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3964 Add appropriate locking and lov-osc refcounts for safe
3965 cleaning. Add osc abort_inflight calls in case the
3966 synchronize never started.
3969 Frequency : occasional (Cray XT3 only)
3971 Description: root not authorized to access files in CRAY_PORTALS environment
3972 Details : The client process capabilities were not honoured on the MDS in
3973 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3974 been accessed by an authorized user then root was able to access
3975 the file on the local client also. The root user capabilities
3976 are now allowed on the MDS, as this environment has secure UID.
3979 Frequency : occasional
3981 Description: ldiskfs "too long searching" message happens too often
3982 Details : A debugging message (otherwise harmless) prints too often on
3983 the OST console. This has been reduced to only happen when
3984 there are fragmentation problems on the filesystem.
3989 Description: Division by zero in statfs when all OSCs are inactive
3990 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3991 lov_statfs divides by value returned by lov_get_stripecnt().
3995 Bugzilla : 9489, 3273
3996 Description: First write from each client to each OST was only 4kB in size,
3997 to initialize client writeback cache, which caused sub-optimal
3998 RPCs and poor layout on disk for the first writen file.
3999 Details : Clients now request an initial cache grant at (re)connect time
4000 and so that they can start streaming writes to the cache right
4001 away and always do full-sized RPCs if there is enough data.
4002 If the OST is rebooted the client also re-establishes its grant
4003 so that client cached writes will be honoured under the grant.
4008 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4009 Details : Now I/O RPCs go to different portal number and (presumably) fast
4010 lock requests (and glimses) and other RPCs get their own service
4011 threads pool that should be able to service those RPCs
4014 Severity : enhancement
4016 Description: Ability to exchange lustre version between client and servers and
4017 issue warnings at client side if client is too old. Also for
4018 liblustre clients there is ability to refuse connection of too old
4020 Details : New 'version' field is added to connect data structure that is
4021 filled with version info. That info is later checked by server and
4025 Frequency : rare, liblustre only.
4026 Bugzilla : 9296, 9581
4027 Description: Two simultaneous writes from liblustre at offset within same page
4028 might proceed at the same time overwriting eachother with stale
4030 Details : I/O lock withing llu_file_prwv was released too early, before data
4031 actually was hitting the wire. Extended lock-holding time until
4032 server acknowledges receiving data.
4035 Frequency : extremely rare. Never observed in practice.
4037 Description: avoid generating lustre_handle cookie of 0.
4038 Details : class_handle_hash() generates handle cookies by incrementing
4039 global counter, and can hit 0 occasionaly (this is unlikely, but
4040 not impossible, because initial value of cookie counter is
4041 selected randonly). Value of 0 is used as a sentinel meaning
4042 "unassigned handle" --- avoid it. Also coalesce two critical
4043 sections in this function into one.
4045 Severity : enhancement
4047 Description: allow liblustre clients to delegate truncate locking to OST
4048 Details : To avoid overhead of locking, liblustre client instructs OST to
4049 take extent lock in ost_punch() on client's behalf. New connection
4050 flag is added to handle backward compatibility.
4052 Severity : enhancement
4053 Bugzilla : 4928, 7341, 9758
4054 Description: allow number of OST service threads to be specified
4055 Details : a module parameter allows the number of OST service threads
4056 to be specified via "options ost ost_num_threads={N}" in the
4057 OSS's /etc/modules.conf or /etc/modprobe.conf.
4061 Bugzilla : 6146, 9635, 9895
4062 Description: servers crash with bad pointer in target_handle_connect()
4063 Details : In rare cases when a client is reconnecting it was possible that
4064 the connection request was the last reference for that export.
4065 We would temporarily drop the export reference and get a new
4066 one, but this may have been the last reference and the export
4067 was just destroyed. Get new reference before dropping old one.
4069 Severity : enhancement
4070 Frequency : if client is started with failover MDS
4072 Description: Allow multiple MDS hostnames in the mount command
4073 Details : Try to read the configuration from all specified MDS
4074 hostnames during a client mount in case the "primary"
4077 Severity : enhancement
4079 Description: Stop sending data to evicted clients as soon as possible.
4080 Details : Check if the client we are about to send or are sending data to
4081 was evicted already. (Check is done every second of waiting,
4082 for which l_wait_event interface was extended to allow checking
4083 of exit condition at specified intervals).
4086 Frequency : rare, normally only when NFS exporting is done from client
4088 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4090 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4093 Frequency : always, if srand() is called before liblustre initialization
4095 Description: Liblustre uses system PRNG disturbing its usage by user application
4096 Details : Introduce internal to lustre fast and high-quality PRNG for
4097 lustre usage and make liblustre and some other places in generic
4098 lustre code to use it.
4100 Severity : enhancement
4101 Bugzilla : 9477, 9557, 9870
4102 Description: Verify that the MDS configuration logs are updated when xml is
4103 Details : Check if the .xml configuration logs are newer than the config
4104 logs stored on the MDS and report an error if this is the case.
4105 Request --write-conf, or allow starting with --old_conf.
4107 Severity : enhancement
4109 Description: Handle symlinks in the path when checking if Lustre is mounted.
4110 Details : Resolve intermediate symlinks when checking if a client has
4111 mounted a filesystem to avoid duplicate client mounts.
4116 Description: lconf can hit an error exception but still return success.
4117 Details : The lconf command catches the Command error exception at the top
4118 level script context and will exit with the associated exit
4119 status, but doesn't ensure that this exit status is non-zero.
4124 Description: failure of ptlrpc thread startup can cause oops
4125 Details : Starting a ptlrpc service thread can fail if there are a large
4126 number of threads or the server memory is very fragmented.
4127 Handle this without oopsing.
4130 Frequency : always, only if liblustre and non-default acceptor port was used
4132 Description: liblustre cannot connect to servers with non-default acceptor port
4133 Details : tcpnal_set_default_params() was not called and was therefore
4134 ignoring the environment varaible TCPNAL_PORT, as well as other
4135 TCPNAL_ environment variables
4140 Description: two objects could be created on the same OST for a single file
4141 Details : If an OST is down, in some cases it was possible to create two
4142 objects on a single OST for a single file. No problems other
4143 than potential performance impact and spurious error messages.
4147 Bugzilla : 5681, 9562
4148 Description: Client may oops in ll_unhash_aliases
4149 Details : Client dcache may become inconsistent in race condition.
4150 In some cases "getcwd" can fail if the current directory is
4156 Description: Inode refcounting problems in NFS export code
4157 Details : link_raw functions used to call d_instantiate without obtaining
4158 extra inode reference first.
4162 Bugzilla : 9942, 9903
4163 Description: Referencing freed requests leading to crash, memleaks with NFS.
4164 Details : We used to require that call to ll_revalidate_it was always
4165 followed by ll_lookup_it. Also with revalidate_special() it is
4166 possible to call ll_revalidate_it() twice for the same dentry
4167 even if first occurence returned success. This fix changes semantic
4168 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4169 reference on a request referred from the intent.
4170 ll_intent_release() then releases such a request.
4173 Frequency : rare, normally benchmark loads only
4175 Description: unlinked inodes were kept in memory on the client
4176 Details : If a client is repeatedly creating and unlinking files it
4177 can accumulate a lot of stale inodes in the inode slab cache.
4178 If there is no other client load running this can cause the
4179 client node to run out of memory. Instead flush old inodes
4180 from client cache that have the same inode number as a new inode.
4183 Frequency : SLES9 2.6.5 kernel and long filenames only
4184 Bugzilla : 9969, 10379
4185 Description: utime reports stale NFS file handle
4186 Details : SLES9 uses out-of-dentry names in some cases, which confused
4187 the lustre dentry revalidation. Change it to always use the
4191 Frequency : rare, unless heavy write-truncate concurrency is continuous
4192 Bugzilla : 4180, 6984, 7171, 9963, 9331
4193 Description: OST becomes very slow and/or deadlocked during object unlink
4194 Details : filter_destroy() was holding onto the parent directory lock
4195 while truncating+unlinking objects. For very large objects this
4196 may block other threads for a long time and slow overall OST
4197 responsiveness. It may also be possible to get a lock ordering
4198 deadlock in this case, or run out of journal credits because of
4199 the combined truncate+unlink. Solution is to do object truncate
4200 first in one transaction without parent lock, and then do the
4201 final unlink in a new transaction with the parent lock. This
4202 reduces the lock hold time dramatically.
4205 Frequency : rare, 2.4 kernels only
4207 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4208 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4209 before unmounting their filesystem in order to avoid deadlock.
4210 The kernel_locked() function in 2.4 kernels only checks whether
4211 the kernel lock is held, not whether it is this process that is
4212 holding it as 2.6 kernels do.
4217 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4218 Details : The client ptlrpc code may be trying to reconnect to a down
4219 server before a previous connection attempt has timed out.
4220 Increase the reconnect interval to be longer than the connection
4221 timeout interval to avoid sending duplicate connections to
4225 Frequency : echo_client brw_test command
4227 Description: fix echo_client to work with OST preallocated code
4228 Details : OST preallocation code (5137) didn't take echo_client IO path
4229 into account: echo_client calls filter methods outside of any
4230 OST thread and, hence, there is no per-thread preallocated
4231 pages and buffers to use. Solution: hijack pga pages for IO. As
4232 a byproduct, this avoids unnecessary data copying.
4236 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4237 Description: Client can oops in mdc_commit_close() after open replay
4238 Details : It was possible for the MDS to return an open request with no
4239 transaction number in mds_finish_transno() if the client was
4240 evicted, but without actually returning an error. Clients
4241 would later try to replay that open and may trip an assertion
4242 Simplify the client close codepath, and always return an error
4243 from the MDS in case the open is not successful.
4246 Frequency : rare, 2.6 OSTs only
4248 Description: OST may deadlock under high load on fragmented files
4249 Details : If there was a heavy load and highly-fragmented OST filesystems
4250 it was possible to have all the OST threads deadlock waiting on
4251 allocation of biovecs, because the biovecs were not released
4252 until the entire RPC IO was completed. Instead, release biovecs
4253 as soon as they are complete to ensure forward IO progress.
4255 Severity : enhancement
4257 Description: Support for specifying external journal device at mount
4258 Details : If an OST or MDS device is formatted with an external journal
4259 device, this device major/minor is stored in the ext3 superblock
4260 and may not be valid for failover. Allow detecting and
4261 specifying the external journal at mount time.
4266 Description: Mounting an MDS with pending unlinked files may cause oops
4267 Details : target_finish_recovery() calls mds_postrecov() which returned
4268 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4269 considers this an error and immediately begins cleaning up the
4270 lov, just after starting the mds_lov process
4272 Severity : enhancement
4274 Description: Implement 'lfs df' to report actual free space on per-OST basis
4275 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4276 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4277 usage of objects; '-h' to report in human readable format.
4279 ------------------------------------------------------------------------------
4281 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4288 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4289 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4290 at startup during orphan unlinking. Instead, skip buddy update
4291 before setup, as it will be regenerated after recovery is complete.
4296 Description: If an OST is inactive, its locks might reference stale inodes.
4297 Details : lov_change_cbdata() must iterate over all namespaces, even if
4298 they are inactive to clear inode references from the lock.
4300 Severity : enhancement
4301 Frequency : occasional, if non-standard max_dirty_mb used
4303 Description: Client will block write RPCs if not enough grant
4304 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4305 then the client will block writes while waiting for another RPC
4306 to complete instead of consuming its dirty limit. With change
4307 we get improved performance when max_dirty_mb is small.
4309 Severity : enhancement
4310 Bugzilla : 3389, 6253
4311 Description: Add support for supplementary groups on the MDS.
4312 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4313 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4314 lookups for user supplementary groups into a cache.
4318 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4319 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4320 for liblustre O_EXCL is enforced by MDS.
4324 Description: readdir never returns NULL in liblustre.
4325 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4326 dentry in struct dirent.
4330 Frequency : liblustre only, depends on application IO pattern
4331 Description: liblustre clients evicted if not contacting servers
4332 Details : Don't put liblustre clients into the ping_evictor list, so
4333 they will not be evicted by the pinger ever.
4335 Severity : enhancement
4337 Description: Add ability to evict clients by NID from MDS.
4338 Details : By echoing "nid:$NID" string into
4339 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4340 $NID would be instantly evicted from this MDS and from all active
4341 OSTs connected to it.
4345 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4346 Details : lookup_it_finish() used to query file size from OSTs that was not
4351 Description: service threads change working directory to that of init
4352 Details : Starting lustre service threads may pin the working directory
4353 of the parent thread, making that filesystem busy. Threads
4354 now change to the working directory of init to avoid this.
4358 Frequency : during shutdown only
4359 Description: shutdown with a failed MDS or OST can cause unmount to hang
4360 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4365 Frequency : 2.6 only
4366 Description: chmod/chown may include an extra supplementary group
4367 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4368 supplementary group and if none is specified this is used.
4371 Bugzilla : 5479 (6816)
4373 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4374 Details : If lookup is in progress on a file that is unlinked we might try
4375 to revalidate the inode and fail in revalidate after lookup is
4376 complete and ll_file_open() enqueues the open again but
4377 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4380 Frequency : always, if lconf --abort_recovery used
4382 Description: lconf --abort_recovery fails with 'Operation not supported'
4383 Details : lconf was attempting to abort recovery on the MDT device and not
4386 ------------------------------------------------------------------------------
4388 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4393 Frequency : rare (only unsupported configurations with a node running as an
4395 Bugzilla : 6514, 5137
4396 Description: Mounting a Lustre file system on a node running as an OST could
4398 Details : OSTs now preallocates memory needed to write out data at
4399 startup, instead of when needed, to avoid having to
4400 allocate memory in possibly low memory situations.
4401 Specifically, if the file system is mounted on on OST,
4402 memory pressure could force it to try to write out data,
4403 which it needed to allocate memory to do. Due to the low
4404 memory, it would be unable to do so and the node would
4405 become unresponsive.
4407 Severity : enhancement
4409 Description: Addition of lconf --service command line option
4410 Details : lconf now accepts a '--service <arg>' option, which is
4411 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4413 Severity : enhancement
4415 Description: Failover mode is now the default for OSTs.
4416 Details : By default, OSTs will now run in failover mode. To return to
4417 the old behaviour, add '--failout' to the lmc line for OSTs.
4419 Severity : enhancement
4421 Description: Health checks are now provided for MDS and OSTs
4422 Details : Additional detailed health check information on MSD and OSTs
4423 is now provided through the procfs health_check value.
4426 Frequency : occasional, depends on IO load
4428 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4429 after numerous create/delete cycles
4430 Details : The ext3 inode allocation policy would not allocate new inodes
4431 very well on the OSTs because there are no new directories
4432 being created. Instead we look for groups with free space if
4433 the parent directories are nearly full.
4438 Description: Network or server problems during mount may cause partially
4439 mounted clients instead of returning an error.
4440 Details : The config llog parsing code may overwrite the error return
4441 code during mount error handling, returning success instead
4447 Description: MDS can fail to allocate large reply buffers
4448 Details : After long uptimes the MDS can fail to allocate large reply
4449 buffers (e.g. zconf client mount config records) due to memory
4450 fragmentation or consumption by the buffer cache. Preallocate
4451 some large reply buffers so that these replies can be sent even
4452 under memory pressure.
4456 Frequency : rare (liblustre)
4457 Description: fsx running with liblustre complained that using truncate() to
4458 extend the file doesn't work. This patch corrects that issue.
4459 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4460 ATTR_SIZE and lsm use in llu_setattr_raw.
4464 Frequency : rare, only 2.6 kernels
4465 Description: Unusual file access patterns on the MDS may result in inode
4466 data being lost in very rare circumstances.
4467 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4468 code in the 2.6 kernel caused the inode and/or EA data not to
4469 be read from disk, causing single-file corruption.
4473 Frequency : rare, only 2.6 filesystems using extents
4474 Description: Heavy concurrent write and delete load may cause data corruption.
4475 Details : It was possible under high-load situations to have an extent
4476 metadata block in the block device cache from a just-unlinked
4477 file overwrite a newly-allocated data block. We now unmap any
4478 metadata buffers that alias just-allocated data blocks.
4482 Frequency : filesystems with default stripe_count larger than 77
4483 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4484 Details : lconf specifies an inode size of 4096 bytes when the default
4485 stripe_count is larger than 77. This conflicts with the default
4486 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4487 this case to avoid pinning too much memory for large EAs.
4489 ------------------------------------------------------------------------------
4491 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4496 Frequency : rare (extremely heavy IO load with hundreds of clients)
4498 Description: Client is evicted, gets IO error writing to file
4499 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4500 caused clients to be evicted for AST timeouts. The fixes in
4501 bug 5192 mean we no longer need to have such short AST timeouts
4502 so ldlm_timeout has been increased.
4505 Frequency : occasional during --force or --failover shutdown under load
4506 Bugzilla : 5949, 4834
4507 Description: Server oops/LBUG if stopped with --force or --failover under load
4508 Details : a collection of import/export refcount and cleanup ordering
4509 issues fixed for safer force cleanup
4512 Frequency : only filesystems larger than 120 OSTs
4513 Bugzilla : 5990, 6223
4514 Description: lfs getstripe would oops on a very large filesystem
4515 Details : lov_getconfig used kfree on vmalloc'd memory
4518 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4519 Bugzilla : 6242, 6243
4520 Description: reading from files that had been truncated to a non-zero size
4521 but never opened returned no data
4522 Details : ll_file_read() reads zeros from no-object files to EOF
4527 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4528 Details : There's little harm in aborting MDS/OSS recovery and letting it
4529 try again, so I removed the LASSERT and return an error instead.
4531 Severity : enhancement
4533 Description: New debugging infrastructure for tracking down data corruption
4534 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4535 (b) cover more of the client-side code path, and (c) try to narrow
4536 down where problems occurred
4540 Bugzilla : 3819, 4364, 4397, 6313
4541 Description: Racing close and eviction MDS could cause assertion in mds_close
4542 Details : It was possible to get multiple mfd references during close and
4543 client eviction, leading to one thread referencing a freed mfd.
4545 Severity: : enhancement
4546 Bugzilla : 3262, 6359
4547 Description: Attempts to reconnect to servers are now more aggressive.
4548 Details : This builds on the enhanced upcall-less recovery that was added
4549 in 1.4.2. When trying to reconnect to servers, clients will
4550 now try each server in the failover group every 10 seconds. By
4551 default, clients would previously try one server every 25 seconds.
4556 Description: After recovery, certain operations trigger a failed
4557 assertion on a client.
4558 Details : Failing over an mds, using lconf -d --failover, while a
4559 client was doing a readdir() call would cause the client to
4560 LBUG after recovery completed and the readdir() was resent.
4562 Severity : enhancement
4564 Description: Default groups are now added by lconf
4565 Details : You can now run lconf --group <servicename> without having to
4566 manually add groups with lmc.
4569 Frequency : occasional
4571 Description: Nodes with an elan id of 0 trigger a failed assertion
4574 Frequency : always when accessing e.g. tty/console device nodes
4576 Description: tty and some other devices nodes cannot be used on lustre
4577 Details : file's private_data field is used by device data and lustre
4578 values in there got lost. New field was added to struct file to
4579 store fs-specific private data.
4582 Frequency : when exporting Lustre via NFS
4584 Description: NFSD failed occasionally when looking up a path component
4585 Details : NFSD is looking up ".." which was broken in ext3 directories
4586 that had grown large enough to become hashed.
4589 Frequency : Clusters with multiple interfaces not on the same subnet
4591 Description: Nodes will repeatedly try to reconnect to an interface which it
4592 cannot reach and report an error to the log.
4593 Details : Extra peer list entries will be created by lconf with some peers
4594 unreachable. lconf now validates the peer before adding it.
4597 Frequency : Only if a default stripe is set on the filesystem root.
4599 Description: Setting a default stripe on the filesystem root prevented the
4600 filesystem from being remounted.
4601 Details : The client was sending extra request flags in the root getattr
4602 request and did not allocate a reply buffer for the dir EA.
4605 Frequency : occasional, higher if lots of files are accessed by one client
4606 Bugzilla : 6159, 6097
4607 Description: Client trips assertion regarding lsm mismatch/magic
4608 Details : While revalidating inodes the VFS looks up inodes with ifind()
4609 and in rare cases can find an inode that is being freed.
4610 The ll_test_inode() code will free the lsm during ifind()
4611 when it finds an existing inode and then the VFS later attaches
4612 this free lsm to a new inode.
4616 Bugzilla : 6422, 7030
4617 Description: MDS deadlock between mkdir and client eviction
4618 Details : Creating a new file via mkdir or mknod (starting a transaction
4619 and getting the ns lock) can deadlock with client eviction
4620 (gets ns lock and trying to finish a synchronous transaction).
4623 Frequency : occasional
4624 Description: While starting a server, the fsfilt_ext3 module could not be
4626 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4627 kernels. Previously, lconf would still use the ext3 name
4628 when trying to load modules. Now, it will correctly use
4629 ext3 on 2.4 and ldiskfs on 2.6.
4631 Severity : enhancement
4632 Description: The default stripe count has been changed to 1
4633 Details : The interpretation of the default stripe count (0, to lfs
4634 or lmc) has been changed to mean striping across a single
4635 OST, rather than all available. For general usage we have
4636 found a stripe count of 1 or 2 works best.
4638 Severity : enhancement
4639 Description: Add support for compiling against Cray portals.
4640 Details : Conditional compiling for some areas that are different
4644 Frequency : occasional
4645 Bugzilla : 6409, 6834
4646 Description: Creating files with an explicit stripe count may lead to
4647 a failed assertion on the MDS
4648 Details : If some OSTs are full or unavailable, creating files may
4649 trigger a failed assertion on the MDS. Now, Lustre will
4650 try to use other servers or return an error to the
4654 Frequency : occasional
4656 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4657 caused lock timeout during readahead (since 1.4.2).
4658 Details : Processes doing readahead might match a lock that hasn't been
4659 granted yet if there are overlapping and conflicting lock
4660 requests. The readahead process waits on ungranted lock
4661 (original lock is CBPENDING), while OST waits for that process
4662 to cancel CBPENDING read lock and eventually evicts client.
4664 Severity : enhancement
4666 Description: Initial enabling of flock support for clients
4667 Details : Implements fcntl advisory locking and file status functions.
4668 This feature is provided as an optional mount flag (default
4669 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4670 locking are implemented yet, and those that are are not guaranteed
4671 to be completely correct in production environments.
4672 mount -t lustre -o [flock|noflock] ...
4675 Frequency : occasional
4677 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4678 assertion in the ext3 JBD (journaling) layer.
4679 Details : The b_committed_data struct is protected by the big kernel lock
4680 in 2.4 kernels, serializing journal_commit_transaction() and
4681 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4682 access to this struct. In 2.6 kernels there is finer grained
4683 locking to improve SMP performance of the JBD layer.
4687 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4692 Description: merge and cleanup kernel patches.
4695 -----------------------------------------------------------------------------
4697 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4699 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4700 versions of Lustre. Please update all servers and clients to
4701 version 1.4.2 or later at the same time. You must also run
4702 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4703 to update the configuration logs.
4705 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4706 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4707 - swab llog records of type '0' so we get proper header size/idx (5861)
4708 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4709 - fix rename of one directory over another leaking an inode (5953)
4710 - avoid SetPageDirty on 2.6 (5981)
4711 - don't re-add just-being-destroyed locks to the waiting list (5653)
4712 - when creating new directories, inherit the parent's custom
4713 striping settings if present parent (3048)
4714 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4715 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4716 - don't hold client locks on temporary worklist from l_lru (5666)
4717 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4718 - automatically evict dead clients (5921)
4719 - Update file size properly in create+truncate+fstat case (6196)
4720 - Do not unhash mountpoint dentries, do not allow removal of
4722 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4723 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4724 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4725 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4726 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4727 - Correctly handle reads of files with no objects (6243)
4728 - lctl recover will also mark a device active if deactivate used (5933)
4730 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4731 - allow --write-conf on an MDS with different nettype than client (5619)
4732 - don't write config llogs to MDS for mounts not from that MDS (5617)
4733 - lconf should create multiple TCP connections from a client (5201)
4734 - init scripts are now turned off by default; run chkconfig --on
4735 lustre and chkconfig --on lustrefs to use them
4736 - upcalls are no longer needed for clients to recover to failover
4738 - add --abort-recovery option to lconf to abort recovery on device
4740 - add support for an arbitrary number of OSTs (3026)
4741 - Quota support protocol changes.
4742 - forward compatibility changes to wire structs (6007)
4743 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4744 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4745 - improved "lustre status" script
4746 - initialize blocksize for non-regular files (6062)
4747 - added --disable-server and --disable-client configure options (5782)
4748 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4749 - Vanilla 2.4.29 support
4750 - increase maximum number of obd devices to 520 (6242)
4751 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4752 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4753 SLES 9 2.6.5 kernels
4754 - limit stripes per file to 160 (the maximum EA size) (6093)
4756 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4759 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4760 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4761 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4762 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4763 - small ext3 extents cleanups and fixes (5733)
4764 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4765 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4766 - avoid needless client->OST connect, fix handle mismatch (5317)
4767 - fix DLM error path that led to out-of-sync client, long delays (5779)
4768 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4769 - fix several locking issues related to i_size (5492,5624,5654,5672)
4770 - don't move pending lock onto export if it is already evicted (5683)
4771 - fix kernel oops when creating .foo in unlinked directory (5548)
4772 - fix deadlock in obdfilter statistics vs. object create (5811)
4773 - use time_{before,after} to avoid timer jiffies wrap (5882)
4774 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4775 - Do not leak request if server was not able to process it (5154)
4776 - If mds_open unable to find parent dir, make that negative lookup(5154)
4777 - don't create new directories with extent-mapping (5909, 5936)
4779 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4780 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4781 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4782 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4783 - support for RHEL4 kernel on i686 (5773)
4784 - provide error messages when incompatible logs are encountered (5898)
4786 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4787 * version 1.4.0.10 (1.4.1 release candidate 1)
4789 - don't keep a lock reference when lock is not granted (4238)
4790 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4791 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4792 - drop import inflight refcount on signal_completed_replay error (5255)
4793 - unlock page after async write error during send (3677)
4794 - handle missing objects in filter_preprw_read properly (5265)
4795 - no transno return for symlink open, don't save no-trasno open (3440)
4796 - don't try to complete elan receive that already failed (4012)
4797 - free RPC server reply state on error (5406)
4798 - clean up thread from ptlrpc_start_thread() on error (5160)
4799 - readahead could read extra page into cache that wasn't ejected (5388)
4800 - prevent races in class_attach/setup/cleanup/detach (5260)
4801 - don't dereference de->d_inode after l_dput of de (5458)
4802 - use "int" for stripe value returned from lock_to_stripe (5544)
4803 - mballoc allocation and error-checking fixes in 2.6 (5504)
4804 - block device patches to fix I/O request sizes in 2.6 (5482)
4805 - look up hostnames for IB nals (5602)
4806 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4807 - don't start multiple acceptors for the same port (5277)
4808 - fix incorrect LASSERT in mds_getattr_name (5635)
4809 - export a proc file for general "ping" checking (5628)
4810 - fix "lfs check" to not block when the MDS is down (5628)
4812 - service request history (4965)
4813 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4814 - create an "evict_client" /proc entry on OSTs, like the MDS has
4815 - fix mount usage message, return errors per mount(8) (5168)
4816 - change grep [] to grep "[]" in tests so they work in more UMLs
4817 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4818 - remove ancient LOV_MAGIC_V0 EA support (5047)
4819 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4820 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4821 - allow readahead allocations to fail when low on memory (5383)
4822 - mmap locking landed again, after considerable improvement (2828)
4823 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4825 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4828 - send OST transaction number in read/write reply to free req (4966)
4829 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4830 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4831 - the watchdog thread now runs as interruptible (5246)
4832 - flock/lockf fixes (but it's still disabled, pending 5135)
4833 - don't use EXT3 constants in llite code (5094)
4834 - memory shortage at startup could cause assertion (5176)
4836 - reorganization of lov code
4837 - single portals codebase
4839 - add extents/mballoc support (5025)
4840 - direct I/O reads in the obdfilter (4048)
4841 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4843 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4846 - send OST transaction number in read/write reply to free req (4966)
4847 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4848 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4849 - the watchdog thread now runs as interruptible (5246)
4850 - handle missing objects in filter_preprw_read properly (5265)
4851 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4852 - drop import inflight refcount on signal_completed_replay error (5255)
4853 - unlock page after async write error during send (3677)
4854 - return original error code on reconstructed replies (3761)
4855 - no transno return for symlink open, don't save no-trasno open (3440)
4857 - add pid to ldlm debugging output (4922)
4858 - bump the watchdog timeouts -- we can't handle 30sec yet
4859 - extra debugging for orphan dentry/inode bug (5259)
4861 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4864 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4865 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4866 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4867 - fix NULL dereference in /proc/sys/portals/routes (4827)
4868 - allow failed mdc_close() operations to be interrupted (4561)
4869 - stop precreate on OST before MDS would time out on it (4778)
4870 - don't send partial-page writes before EOF from client (4410)
4871 - discard client grant for sub-page writes on large-page clients (4520)
4872 - don't free dentries not owned by NFS code, check generation (4806)
4873 - fix lsm leak if mds_create_objects() fails (4801)
4874 - limit debug_daemon file size, always print CERROR messages (4789)
4875 - use transno after validating reply (3892)
4876 - process timed out requests if import state changes (3754)
4877 - update mtime on OST during writes, return in glimpse (4829)
4878 - add mkfsoptions to LDAP (4679)
4879 - use ->max_readahead method instead of zapping global ra (5039)
4880 - don't interrupt __l_wait_event() during strace
4882 - add software watchdogs to catch hung threads quickly (4941)
4883 - make lustrefs init script start after nfs is mounted
4884 - fix CWARN/ERROR duplication (4930)
4885 - return async write errors to application if possible (2248)
4886 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4887 - print NAL number in %x format (4645)
4888 - update barely-supported suse-2.4.21-171 series (4842)
4889 - support for sles 9 %post scripts
4890 - support for building 2.6 kernel-source packages
4891 - support for sles km_* packages
4893 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4896 - ignore -ENOENT errors in osc_destroy (3639)
4897 - notify osc create thread that OSC is being cleaned up (4600)
4898 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4899 - reconstruct ost_handle() like mds_handle() (4657)
4900 - create a new thread to do import eviction to avoid deadlock (3969)
4901 - let lconf resolve symlinked-to devices (4629)
4902 - don't unlink "objects" from directory with default EA (4554)
4903 - hold socknal file ref over connect in case target is down (4394)
4904 - allow more than 32000 subdirectories in a single directory (3244)
4905 - fix blocks count for O_DIRECT writes (3751)
4906 - OST returns ENOSPC from object create when no space left (4539)
4907 - don't send truncate RPC if file size isn't changing (4410)
4908 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4909 - bind to privileged port in socknal and tcpnal (3689)
4911 - rate limit CERROR/CWARN console message to avoid overload (4519)
4912 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4913 - basic mmap support (3918)
4914 - kernel patch series update from b1_4 (4711)
4916 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4919 - avoid crash during MDS cleanup with OST shut down (2775)
4920 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4921 - don't use bad inodes on the MDS (3744)
4922 - dynamic object preallocation to improve recovery speed (4236)
4923 - don't hold spinlock over lock dumping or change debug flags (4401)
4924 - don't zero obd_dev when it is force cleaned (3651)
4925 - print grants to console if they go negative (4431)
4926 - "lctl deactivate" will stop automatic recovery attempts (3406)
4927 - look for existing locks in ldlm_handle_enqueue() (3764)
4928 - don't resolve lock handle twice in recovery avoiding race (4401)
4929 - revalidate should check working dir is a directory (4134)
4931 - don't always mark "slow" obdfilter messages as errors (4418)
4933 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4936 - don't close LustreDB during write_conf until it is done (3860)
4937 - fix typo in lconf for_each_profile (3821)
4938 - allow dumping logs from multiple threads at one time (3820)
4939 - don't allow multiple threads in OSC recovery (3812)
4940 - fix debug_size parameters (3864)
4941 - fix mds_postrecov to initialize import for llog ctxt (3121)
4942 - replace config semaphore with spinlock (3306)
4943 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4944 - don't allow enqueue to complete on a destroyed export (3822)
4945 - down write_lock before checking llog header bitmap (3825)
4946 - recover from lock replay timeout (3764)
4947 - up llog sem before sending rpc (3652)
4948 - reduce ns lock hold times when setting kms (3267)
4949 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4950 - fix NULL deref and obd_dev leak on setup error (3312)
4951 - replace some LBUG about llog ops with error handling (3841)
4952 - don't match INVALID dentries from d_lookup and spin (3784)
4953 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4954 - fix invalid assertion in ptlrpc_set_wait (3880)
4956 - add libwrap support for the TCP acceptor (3996)
4957 - add /proc/sys/portals/routes for non-root route listing (3994)
4958 - allow setting MDS UUID in .xml (2580)
4959 - print the stack of a process that LBUGs (4228)
4961 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4964 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4965 - make sure to unset replay flag from failed open requests (3440)
4966 - if default stripe count is 0, use OST count for inode size (3636)
4967 - update parent mtime/ctime on client for create/unlink (2611)
4968 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4969 - free recovery state on server during a forced cleanup (3571)
4970 - unregister_reply for resent reqs (3063)
4971 - loop back devices mounting and status check on 2.6 (3563)
4972 - fix resource-creation race that can provoke i_size == 0 (3513)
4973 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4974 - more debugging for page-accounting assertion (3746)
4975 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4976 - avoid deadlock after precreation failure (3758)
4977 - fix race and lock order deadlock in orphan handling (3450, 3750)
4978 - add validity checks when grabbing inodes from l_ast_data (3599)
4980 - add /proc/.../recovery_status to obdfilter (3428)
4981 - lightweight CDEBUG infrastructure, debug daemon (3668)
4982 - change default OSC RPC parameters to be better on small clusters
4983 - turn off OST read cache for files smaller than 32MB
4984 - install man pages and include them in rpms (3100)
4985 - add new init script for (un)mounting lustre filesystems (2593)
4986 - run chkconfig in %post for init scripts (3701)
4987 - drop scimac NAL (unmaintained)
4989 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4992 - clean kiobufs before and after use (3485)
4993 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4994 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4995 - change init script to fail more clearly if not run as root (1528)
4996 - allow clients to reconnect during replay (1742)
4997 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4998 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4999 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5000 - fix cond_resched (3554)
5001 - only evict unfinished clients after recovery (3515)
5002 - allow bulk resend, prevent data loss (3570)
5003 - dynamic ptlrpc request buffer allocation (2102)
5004 - don't allow unlinking open directory if it isn't empty (2904)
5005 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5006 - remove extraneous obd dereference causing LASSERT failure (3334)
5007 - don't use get_cycles() when creating temp. files on the mds (3156)
5008 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5009 - handle EEXIST for set-stripe, set proper directory name (3336)
5011 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5012 - fix ksocknal_fmb_callback() error messages (2918)
5014 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5017 - don't copy lvb into (possibly NULL) reply on error (2983)
5018 - don't deref dentry after dput, don't free lvb on error (2922)
5019 - use the kms to determine writeback rpc length (2947)
5020 - increment oti_logcookies when osc is inactive (2948)
5021 - update client's i_blocks count via lvb messages (2543)
5022 - handle intent open/close of special files properly (1557)
5023 - mount MDS with errors=remount-ro, like obdfilter (2009)
5024 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5025 - don't use cancelling-locks' kms values (2947)
5026 - use highest lock extent for kms, not last one (2925)
5027 - don't dereference ERR_PTR() dentry in error handling path (3107)
5028 - fix thread race in portals_debug_dumplog() (3122)
5029 - create lprocfs device entries at setup instead of at attach (1519)
5030 - common AST error handler, don't evict client on completion race (3145)
5031 - zero nameidata in detach_mnt in 2.6 (3118)
5032 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5033 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5034 - initialize RPC timeout timer earlier for 2.6 (3219)
5035 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5036 - print nal/nid for unknown nid (3258)
5037 - additional checks for oscc recovery before doing precreate (3284)
5038 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5039 - don't crash in mdc_close for bad permissions on open (3285)
5040 - zero i_rdev for non-device files (3147)
5041 - clear page->private before handing to FS, better assertion (3119)
5042 - tune the read pipeline (3236)
5043 - fix incorrect decref of invalidated dentry (2350)
5044 - provide read-ahead stats and refine rpc in flight stats (3328)
5045 - don't hold journal transaction open across create RPC (3313)
5046 - update atime on MDS at close time (3265)
5047 - close LDAP connection when recovering to avoid server load (3315)
5048 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5049 - don't leak open file on MDS after open resend (3325)
5050 - serialize filter_precreate and filter_destroy_precreated (3329)
5051 - loop device shouldn't call sync_dev() for nul device (3092)
5052 - clear page cache after eviction (2766)
5053 - resynchronize MDS->OST in background (2824)
5054 - refuse to mount the same filesystem twice on same mountpoint (3394)
5055 - allow llmount to create routes for mounting behind routers (3320)
5056 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5057 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5058 - fix rare problem with rename on htree directories (3417)
5060 - allow default OST striping configuration per directory (1414)
5061 - fix compilation for qswnal for 2.6 kernels (3125)
5062 - increase maximum number of MDS request buffers for large systems
5063 - change liblustreapi to be useful for external progs like lfsck (3098)
5064 - increase local configuration timeout for slow disks (3353)
5065 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5067 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5070 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5071 - don't overwrite extent policy data in reply if lock was blocked (2901)
5072 - drop filter export grants atomically with removal from device (2663)
5073 - del obd_self_export from work_list in class_disconnect_exports (2908)
5074 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5075 - swab reply message in mdc_close, other PPC fixes (2464)
5076 - fix destroying of named logs (2325)
5077 - overwrite old logs when running lconf --write_conf (2264)
5078 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5079 - fix race in target_handle_connect (2898)
5080 - mds_reint_create() should take same inode create lock (2926)
5081 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5082 - don't close files for self_export to avoid uninitialized obd (2936)
5083 - allow MDS with the same name as client node (2939)
5084 - hold dentry reference for closed log files for unlink (2325)
5085 - reserve space for all logs during transactions (2059)
5086 - don't evict page beyond end of stripe extent (2925)
5087 - don't oops on a deleted current working directory (2399)
5088 - handle hard links to targets without a parent properly (2517)
5089 - don't dereference NULL lock when racing during eviction (2867)
5090 - don't grow lock extents when lots of conflicting locks (2919)
5092 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5095 - account for cache space usage on clients to avoid data loss (974)
5096 - lfsck support in lustre kernel code (2349)
5097 - reduce journal credits needed for BRW writes (2370)
5098 - orphan handling to avoid losing space on client/server crashes
5099 - ptlrpcd can be blocked, stopping ALL progress (2477)
5100 - use lock value blocks to assist in proper KMS, faster stat (1021)
5101 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5102 - recovery for initial connections (2355)
5103 - fixes for mds_cleanup_orphans (1934)
5104 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5105 - block all file creations until orphan recovery completes (1901)
5106 - client remove rq_connection from request struct (2423)
5107 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5108 - recovery timer race (2670)
5109 - mdc_close recovey bug (2532)
5110 - ptlrpc cleanup bug (2710)
5111 - mds timeout on local locks (2588)
5112 - namespace lock held during RPCs (2431)
5113 - handle interrupted sync write properly (2503)
5114 - don't try to handle a message that hasn't been replied to (2699)
5115 - client assert failure during cleanup after abort recovery (2701)
5116 - leak mdc device after failed mount (2712)
5117 - ptlrpc_check_set allows timedout requests to complete (2714)
5118 - wait for inflight reqs when ptlrpcd finishes (2710)
5119 - make sure unregistered services are removed from the srv_list
5120 - reset bulk XID's when resending them (caught by 1138 test)
5121 - unregister_bulk after timeout
5122 - fix lconf error (2694)
5123 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5124 - readahead locks pages, leaves pending causing memory pressure (2673)
5125 - increase OST request buffers to 4096 on large machines (2729)
5126 - fix up permission of existing directories in simple_mkdir (2661)
5127 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5128 - don't assign transno to errored transactions (2742)
5129 - don't delete objects on OST if given a bogus objid from MDS (2751)
5130 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5131 - if rq_no_resend, then timeout request after recovery (2432)
5132 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5133 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5134 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5135 - MDS recovery shouldn't delete objects that it has given out (2730)
5136 - if enqueue arrives after completion, don't clobber LVB (2819)
5137 - don't unlock pages twice when trigger_group_io returns error (2814)
5138 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5139 - don't write pages to disk if there was an error (1450)
5140 - don't ping imports that have recovery disabled (2676)
5141 - take buffered bytes into account when balancing socknal conn (2817)
5142 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5143 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5144 - remove little-endian swabbing of llog records (1987)
5145 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5146 - retry reposting request buffers if they fail (1191)
5147 - grow extent at grant time to avoid granting a revoked lock (2809)
5148 - lock revoke doesn't evict page if covered by a second lock (2765)
5149 - disable VM readahead to avoid reading outside lock extents (2805)
5151 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5152 - updated kernel patches for hp-2.4.20 kernel (2681)
5154 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5157 - fix truncated write corruption (2366)
5158 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5160 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5161 - ensure len(uuid) < 37 in lmc (1171)
5162 - fix ia64 OOPS in llog_test (2255)
5163 - zero end of page at obdfilter for partial page writes (2648)
5164 - don't leave stale dentries around after renames (bug 2428)
5165 - fix timeouts when evicting a client with a single lock held (2642)
5166 - set deadline for the initial HELLO message to drain (2634)
5167 - print out dotted-quad IP addresses in the socknal (2302)
5169 - additional debugging for MDS client eviction problem (2443)
5170 - fix mkfsoptions support for osts (2603, 2604)
5172 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5175 - add series for the vanilla 2.6.0 kernel
5176 - add series for the vanilla 2.4.24 kernel
5177 - add series for a cray x86/64 UL kernel drop
5178 - fix xattr patches for the vanilla 2.4.19 series
5180 - generate true UUIDs in lmc (1171)
5181 - have portals stack dumping break in UML (2466)
5182 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5183 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5184 - wake callback waiting threads on client eviction (2460)
5185 - Add --ptldebug and --subsystem to lmc (1719)
5186 - update assertion to allow safe interrupt allocation
5187 - set rq_no_resend for cancel requests (2432)
5188 - recalculate ptlrpcd timeout after resend (2494)
5189 - call vfs_rmdir when removing pending directories (2368)
5190 - fix renaming a file to itself (2429)
5191 - lmc creates a default one-stripe lov (2454)
5192 - expand procfs space to handle large clusters (2326)
5193 - increase UML stack to avoid overflow
5194 - update lconf's list of debug and subsystem masks
5195 - fix lfs find --obd (2510)
5196 - /proc tunable for disabling filter read caching (2591)
5197 - stop rpm packages from altering slapd.conf (2301)
5198 - disable nagle in the socknal under 0conf (2578)
5199 - choose mds inode size based on stripe count (2572)
5200 - fix kernel-source rpm problems (2516)
5202 - add --disable-doc to avoid pdf generation (2421)
5203 - update documentation, tests, type-os, comments
5204 - avoid format warnings on ia64
5205 - remove the TOE NAL
5206 - tiny code cleanups by removing unused fields
5208 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5211 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5212 - remove the most blatant lies from BUILDING file (2371)
5213 - change default debug level to reasonable production setting
5214 - reduce client side cache size to reduce cache flush time
5215 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5216 - make TCP zerocopy and pinger support enabled by default (2476)
5217 - sync writes completed after process exits caused crashes (2319)
5218 - maintain correct mount count on the MDS (2356)
5219 - backout 1557, because 2316 wasn't really fixed
5220 - better file I/O statistics gathering in /proc
5221 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5222 - another kernel patch to fix zero-copy TCP function export
5223 - don't take duplicate lock when processing re-sent getattr (2420)
5224 - lctl uses obd_self_export instead of creating new conn (2353)
5225 - MDS/OST recovery case which requires object creation asserted (2425)
5226 - move lfs from /usr/sbin to /usr/bin in packages
5227 - fix race between mds_client_add and mds_client_free (2417)
5228 - use kmalloc instead of slabs in portals (2430)
5229 - don't create duplicate records when a failover MDS is present (2442)
5230 - remove unnecessary mount age check (2332)
5231 - don't remove directory inodes from locks prematurely (2451)
5232 - don't break if MDS service name is the same as hostname (2103)
5233 - fix races in client write RPC generation when cache full (2482)
5235 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5238 - remove now-unused request->rq_obd (278)
5239 - if an allocation fails, print out how much memory we've used (1933)
5240 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5241 - add the "configurable stack size" patch to most series files (1256)
5242 - ability to write large log records, for 100+ OST configs (2306)
5243 - fix NULL deref when filter_prep fails (2314)
5244 - fix operator precedence error in filter_sync
5245 - dynamic allocation of socknal TX descriptors (2315)
5246 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5247 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5248 - ensure that CWARN messages in Portals always get to the syslog
5249 - __init/__exit are not for prototype decls (ldlm_init/exit)
5250 - x86-64 compile warning fixes
5251 - fix gateway LMC keyword conflict (2318)
5252 - fix MDS lock inversions in getattr/reint paths (1844)
5253 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5254 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5255 - DLM race condition prevented some lock evictions (2328)
5256 - ENOMEM detection and retry on socknal sends (2230)
5257 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5258 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5259 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5260 - better error messages when a client is rejected during recovery (1505)
5261 - avoid cancelling locks which were never granted, after failure (2330)
5262 - fix i_sem/journal inversion in mds_client_add (2333)
5263 - fix truncate/getattr lock cycle deadlock (2334)
5264 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5265 - fix two rare exit paths which could leak an l_lock() ref (2321)
5266 - fencepost error in MDS/OST orphan recovery (2226)
5267 - make log record alignment 8 bytes (1988)
5268 - lstripe now fails when requested offset > ost_count (2237)
5269 - ensure that all kernel series have a complete list.h (1607)
5270 - fix crashes in special-file operations (2316)
5271 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5273 - allow configurable automake binary, for testing new versions
5274 - small update to the lfs documentation
5276 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5278 * fix negative export reference count in fsfilt_sync (2312)
5280 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5281 * release candidate 0.9.1
5283 - orphans are moved into the PENDING directory for possible recovery
5284 - replayed opens now open by fid for orphan/rename safety (1042)
5285 - last close of an orphan inode generates a transno (683)
5286 - chdir() and mount() now pin the directory entry (1020)
5287 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5288 - discard very old requests without processing them (1502)
5289 - remove some common, well-understood CERRORs (1505)
5290 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5291 - clear "grant" flags in OST replies until OST grant code lands (1644)
5292 - fix read performance by not clobbering i_blksize on client (1598)
5293 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5294 - make LRU size a /proc tunable, clears locks when reduced (707)
5295 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5296 - server locks take references on exports now (1558)
5297 - build fixes for 2.4.20-rh trees (1663)
5298 - return an error from lov_create if all OSCs are inactive (1751)
5299 - fix import levels when a reconnect happens without a timeout (1597)
5300 - exit early from mds_open if we get a lookup error (1749)
5301 - partial page read at EOF wouldn't wait for disk before sending (1642)
5302 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5303 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5304 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5305 - write/truncate lock inversion (1639)
5306 - Don't auto-load obdclass, portals modules during cleanup (1495)
5307 - fix timestamps from jumping to "now" (1763)
5308 - extra journal assertions (1648)
5309 - add an extra multiunlink test (1771)
5310 - fix read_record/write_record API (1776)
5311 - fix leak of offset_extent, possible incorrect i_size later (1772)
5312 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5313 - Debugging for the kqswnal_get_idle_tx problems (1820)
5314 - Allow recovery to be attempted multiple times (1536)
5315 - Write out MDS last_rcvd file after it is first created (1600)
5316 - Fix tx_descriptor leak in failed transmit situations (1827)
5317 - ext3 journaling fixes for assertion failure after IO error (1871)
5318 - class_export_put() on freed export after completion AST error (1896)
5319 - Fix revalidate looping in VFS (1322)
5320 - Don't access a freed export during MDS_REINT timeout (1521)
5321 - Add open-unlink recovery support on the MDS (1673,1764)
5322 - Return an error if no MDS data was read from last_rcvd (1946)
5323 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5324 - Don't setup a disk device that doesn't match exported UUID (317)
5325 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5326 - avoid committing NULL handle in force close
5327 - local.sh is now a one-stripe LOV configuration
5328 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5329 - fix ext3 htree duplicate directory entry corruption (1516)
5330 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5331 - update to new LOV EA format (2097)
5332 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5334 2003-06-15 Phil Schwan <phil@clusterfs.com>
5337 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5338 - per-import/export recovery handling (958, 931, 959)
5339 - multiple last-rcvd slots, for serving multiple FSes (949)
5340 - connections are again shared between multiple imp/exports (963, 964)
5341 - "umount -f" would hang if any requests needed to be sent (393, 978)
5342 - avoid pinning large req buffer by copying for queued messages (989)
5343 - add "uuid" to "lctl device" command to help upcalls (991)
5344 - "open" RPCs with transnos would confuse recovery counters (1037)
5345 - do proper endian conversion of all wire messages (288, 340, 891)
5346 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5347 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5348 - fix ptlrpc_connection leak in target_handle_connect (1174)
5349 - fix import refcounting bug in OST and MDS cleanup (1134)
5350 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5351 - fix very unlikely obd_types race condition (501)
5352 - remove osc_open hack for echo_client (1187)
5353 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5354 - a failure in read_inode2 leads to deadlock (1139)
5355 - cancel ack-locks as soon as transaction is committed (1072)
5356 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5357 - make sure to commitrw after any preprw to avoid deadlock (1162)
5358 - failing to execute a file in a lustre FS would lock inode (1203)
5359 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5360 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5361 - obd_uuid2tgt would walk off the end of the list (1255)
5362 - on IA64 the osc would give portals incorrect bulk size (1258)
5363 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5364 - fix lock inversion caused by new llite matching code (1282)
5365 - limit the number of dirty pages on a client to 10MB (1286)
5366 - timed out locks were not being corrected cancelled (1289)
5367 - fix O_DIRECT above 4GB on IA-32 (1292)
5368 * major user-visible changes
5369 - fail out/fail over policy now controlled by the upcall (993)
5371 - add OBD_PING to check server availability and failure (954)
5372 - lustre messages are now sent in sending host order (288, 340, 891)
5373 - add eadatalen to MDS getattr reply (340)
5374 - OST read replies may contain second buffer, with per-page status (593)
5376 2003-03-11 Phil Schwan <phil@clusterfs.com>
5379 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5380 - failed open()s could cause deadlock; fixed (867, 869)
5381 - stop cancelling OST locks when files are closed (481)
5382 - overlapping XID spaces caused network corruption (851, 853)
5383 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5384 - setattr_raw added, to do single-RPC, server-side setattrs
5385 - lmc/lconf syntax change for OST UUIDs
5386 - fix crashy race condition between ptlrpc_free_req and osc_close
5387 - don't use request in mdc_enqueue if we hit a timeout (889)
5388 - don't set the inode i_size for regular files from the MDS (896)
5389 - handle out of order completion AST (842)
5390 - don't LBUG if a lock request times out after receiving AST (913)
5391 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5392 - if a bad lock AST arrives, send an error instead of dropping entirely
5393 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5394 - fix leak in bulk IO when only partially completed (899, 900, 926)
5395 - fix O_DIRECT for ia64 (55)
5396 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5397 - C-z after timeout could hang a process forever; fixed (977)
5399 - client-side I/O cache (678, 924, 929, 941, 970)
5401 - READPAGE and SETATTRs which don't take server-side locks get
5404 2003-02-11 Phil Schwan <phil@clusterfs.com>
5407 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5408 - Fix fsfilt_extN_readpage() to read a full page of directory
5409 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5410 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5411 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5412 - Most intent operations were removed, in favour of a new RPC mode
5413 that does a single RPC to the server and bypasses most of the VFS
5414 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5415 - Aggressively cancel local locks on DLM servers
5416 - mds_reint_unlink sends EA to the client if it's the last nlink.
5417 client uses that EA to unlink OST objects.
5418 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5419 - recursive symlinks were fixed (439)
5420 - fixed NULL deref in DEBUG_REQ
5421 - filter_update_lastobjid no longer calls sync, which annoyed extN
5422 - fixed multi-client small-writes to a single file problem (445)
5423 - fixed mtime updates during file writes (607)
5424 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5425 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5426 - fixed Linux OST size reporting problem (444, 656)
5427 - OST now updates object mtime with writes or setattr (607, 619)
5428 - client verifies file size before zeroing page past EOF (445)
5429 - OST now writes last allocated objid to disk with allocation (108)
5430 - LOV on echo now works (409)
5432 - mds_reint_unlink sends a new buffer, with the EA included. this
5433 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5434 set if a regular file was being unlinked, and it was the last link
5435 - use PtlGet from the target for bulk writes (315)
5436 - OST now updates object mtime with writes or setattr (607, 619)
5437 - LDLM now has a grant-time callback to revalidate locked items, if
5439 - Many MDS operations were reorganized to combat race conditions
5441 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5442 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5444 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5447 - Fully reactivate OST imports after reconnection (512, others)
5448 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5449 - More graceful error handling for truncating on dead OST (515)
5450 - Don't error out unless we're actually accessing dead stripes (474)
5451 - Fix garbage sizes when stripes are missing (410)
5452 - LRU counters were broken, causing constant lock purge (433, 432)
5453 - garbage on read from stripes with failed OSTs (441)
5454 - mark OSCs as active before reconnecting during recovery (438)
5455 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5456 - lfind did not preserve OST order in output (443)
5457 - symlinks cause hung clients, incorrect data (439)
5458 - stop dereferencing request after dropping refcount (457)
5459 - don't LASSERT(spin_is_locked) on non-SMP (455)
5460 - fixes for many rename() bugs
5461 - fstat didn't correctly synchronize attributes (399)
5462 - server must handle lock cancellation during blocking AST prep (487)
5463 - bulk descriptors were free()d too soon (511)
5464 - fix paths in lconf, which would load incorrect modules (451, 507)
5465 - fix confusing lconf 'host not found' error message (386)
5466 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5467 - fix race condition in mdc_blocking_ast() for inode access (526)
5468 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5469 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5470 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5471 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5472 - fix lstripe in several ways (kernel side) (527)
5473 - fix request leak in ldlm_cli_enqueue (262)
5474 - incorrect OSC was marked inactive after OST failure
5475 - call mds_fs_cleanup before unmounting filesystem (524)
5476 - fix races between taking ns_lock and ldlm_lock_change_resource
5477 - fix races updating LOV export open file list
5478 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5479 - fix recovery NULL deref in ldlm_cli_cancel_unused
5480 - fix some DLM races by using new hash table for lock handles (419)
5481 - permit the client to specify desired inodes, at replay
5482 - duplicate requests when we queue them for replay reintegration
5483 - fix last_rcvd offset calculation
5484 - sync after each recovered transaction, so we always make progress
5485 - never, not always, ERESTART requests without transnos
5486 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5488 - skip replay if the MDS says that the client is already connected
5489 - don't check for a recovery-enabled export to match lctl's UUID
5490 - don't INC_USE_COUNT for phantom exports
5491 - don't crash when cleaning up phantom exports (567)
5492 - don't double-finish or set replay data for errored mdc_open requests
5493 - abort requests when they time out, so we don't get old replies
5494 - send/receive replies for AST messages again
5495 - if the client says that it doesn't have the lock, cancel it on the
5497 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5498 mark it as destroyed, it will all work out when decref is called
5499 - fix module use counts (22, 581)
5501 - ASTs now expect a reply (server cancels lock on error reply)
5503 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5506 - fix many simultaneous client startup (392)
5507 - fix dentry->d_it clobbering
5508 - credentials weren't being shipped for readdir/getattr operations
5509 - remove invalid assertions triggered during some concurrent MD
5511 - proper Lustre versions added (336, 389)
5512 - fix memory leak for create error case (398)
5513 - fix LOV locking bug that would get cli/srv out of sync
5514 - fix echo client over LOV (409)
5515 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5516 - fix double-O_EXCL intent crash (424)
5517 - avoid sending multiple lock CANCELs (352)
5519 - MDS can do multi-client recovery (modulo bugs in new code)
5521 - many updates, edits, cleanups
5523 2002-11-18 Phil Schwan <phil@clusterfs.com>
5526 - fix null d_it dereference (346)
5527 - fix full OST/dbench hang (333)
5528 - fix permission problem with file removal (286)
5529 - fix removal of OSCs from LOV when they fail
5530 - fix NULL deref during bulk timeout (214)
5531 - fix problems related to multiple filesystems on one MDS (241)
5532 - fixed serious subtle metadata locking bugs
5533 - free locks on clients when inodes are removed due to memory
5535 - fix inode pointer in lock data (285)
5536 - partial support for multiple MDS on a single host (241)
5537 - data locks weren't cancelled at clear_inode time (290, 311)
5538 - intent locks could lead to unbounded lock growth (205)
5539 - added a maximum lock count, an LRU list, and a flusher
5540 - fix multiple rename (365)
5541 - properly abstracted the echo client
5542 - OSC locked 1 byte too many; fixed
5543 - rewrote brw callback code:
5544 - fixed recovery bugs related to LOVs (306)
5545 - fixed too-many-pages-in-one-write crash (191)
5546 - fixed (again) crash in sync_io_timeout (214)
5547 - probably fixed callback-related race (385)
5549 - Add capability to MDS protocol
5550 - LDLM cancellations and callbacks on different portals
5552 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5555 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5556 - defer open object destruction to close time (601981, 138)
5557 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5558 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5559 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5560 - don't call obd_setattr in ll_file_release for destroyed objects
5561 * protocol change to lustre_msg: move |version| and add |flags|
5562 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5563 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5564 * added replay of create, unlink, link and rename operations during
5565 MDS failover; recovery should be much more robust now
5566 * remove failed OSCs from LOVs (only lov_create uses this so far)
5567 * the lustre-HOWTO was brought (more) up to date (582544)
5569 2002-10-23 Phil Schwan <phil@clusterfs.com>
5572 - in-use dentries weren't being reused properly (617851)
5573 - prevent multiple LDLM setup (599178)
5574 - fix LOV size calculations for truncate (617853)
5575 - fix client handling of MDS intent errors (POSIX)
5576 - fix permission bug in lovstripe.c test (624321)
5577 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5578 - truncate past end of file could corrupt data
5579 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5580 - a race in recovery could return ETIMEDOUT to apps (623947)
5581 - building outside the source directory was fixed
5582 * the lustre-HOWTO was brought (more) up to date (582544)
5583 * major progress was made on recovery functionality
5585 2002-10-10 Phil Schwan <phil@clusterfs.com>
5588 - recovery deadlock fix
5589 - rm -rf causes LBUG fix (617817)
5590 - file open by multiple tasks fix (618962)
5591 - directory permissions bugs (602707 and 620007)
5592 - journal_stop fixed with locking (611313)
5593 - O_APPEND failures resolved (618273, perhaps 614459)
5594 - lconf PATH fix (619770)
5595 - IA64 build fix (621450)
5596 - RPC buffer sizes scale with amount of memory
5598 2002-10-01 Phil Schwan <phil@clusterfs.com>
5601 - locks would be cancelled without throwing away data pages,
5602 resulting in inconsistent data (605627)
5603 - inode attributes were not always being refreshed (605627, 612449)
5604 - lconf now continues to cleanup after lctl reports an error
5605 - MDS now enforces user permissions (602707)
5606 - lprocfs cleanup fixed, but not yet enabled (614157)
5607 - fixed infinite server hang, should a client not respond to an AST
5608 - avoid going into recovery if user calls readlink() with a buffer
5609 that's too small (613941)
5610 - AST RPCs no longer require replies (614867) -- this may be changed
5611 - don't crash server if client sends an IOV that's too big (611336)
5612 - fixed lock conversion deadlock (611892)
5613 - fixed the following of symlinks (614622)
5614 * recovery: the server can remove locks from a client that dies, other
5615 clients can make progress
5616 * more extN patch fixes
5617 * compile-time configurable ptlrpc buffer allocations
5619 - collaborative read cache document
5620 - Lustre Lite Performance CDR document-in-progress
5622 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5625 - fix typo in patch-2.4.18
5627 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5630 - clear ptlrpc request each time in handle_incoming_request()
5631 - unlink of files now destroys the object on the OST
5633 2002-09-19 Peter Braam <braam@clusterfs.com>
5635 * add hard link support
5636 * change obdfile creation method
5637 * kernel patch changed
5639 2002-09-19 Peter Braam <braam@clusterfs.com>
5642 - stack overflow bug in extN fixed
5644 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5646 * documentation updates
5647 - add man pages for config tools
5648 - update tests/README to describe testing with new config tools
5649 - finish metadata API descriptions
5650 * bug fixes and cleanups
5651 - statfs workaround for 16TB limit
5652 - LOV stripe allocation improved, can stripe on subset of OSTs
5653 - LOV file size/IO offset was wrong for files > 4GB in size
5654 - object EA data was being dropped, caused files to be unreadable
5655 - memory overflow with non-LOV OST caused memory corruption
5656 - fixed regression tests to work with new config tools, obdfilter
5657 - fixed bug when directory size became larger than 1 block
5658 - fixed bug (for single client case) when PWD was deleted
5659 - invalidate local directory pages when doing intent-based ops
5660 - avoid LDLM oops when lock callback contained bad data
5662 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5664 * documentation updates
5665 * bug fixes and cleanups
5666 - configuration tools
5669 - 64-bit compile warnings
5670 - 64-bit internal statfs data
5672 * test_brw on persistent OST devices
5674 * lprocfs (disabled)
5676 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5678 * documentation updates
5679 * bug fixes and cleanups
5680 * configuration tools
5682 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5685 * many small fixes to 0_5_4
5686 * io/network handling
5687 * thinkos in MDS operations
5689 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5692 * crucial basic fixes to 0.5.3
5693 * IOR, Iozone work over Elan
5696 2002-08-07 Phil Schwan <phil@clusterfs.com>
5697 * version 0_5_3, our first alpha
5698 * we use the new Portals iovs
5699 * documentation updates
5700 * bug fixes and cleanups
5701 * small changes in the DLM wire protocol
5703 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5704 * version 0_5_1 with some initial stability,
5705 * locking on MD and file I/O.
5706 * documentation updates
5707 * several bug fixes since 0.5.0
5708 * small changes in wire protocol
5710 2002-07-18 Phil Schwan <phil@clusterfs.com>
5712 * delivered as Lustre Light Alpha
5713 * fixed a crash after handling invalid MDS requests
5714 * fixed directory pages for architectures with non-4k pages sizes
5716 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5717 * release version v0_4_4
5718 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5720 * Fixes a number of interruption problems with OST operations.
5721 * Update documentation for portals header changes
5722 * Move all wire protocol structs/defines to lustre_idl.h
5723 * Fixes symlink length bug.
5724 * Add tcpdump to repository.
5726 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5727 * release version v0_4_3
5728 * Fixes statfs for inodes on extN.
5729 * Fixes bug in runtests which would delete /etc/hosts.
5730 * Use 64-bit object IDs wherever possible (not into VFS though)
5731 Remove ost_get_info, which is unused by lustre, and out of date.
5733 2002-07-03 Peter Braam <braam@clusterfs.com>
5734 * release version v0_4_2 Fixes a lookup error (type not passed)
5735 * move forward to head of Portals
5736 * move forward to latest Lustre kernel
5738 2002-06-25 Peter Braam <braam@clusterfs.com>
5739 * release version v0_4_1. Hopefully stable on single node use.