1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.33 (SLES 10),
5 2.6.18-128.1.1.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.41.5.sun2
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
14 * File join has been disabled in this release, refer to Bugzilla 16929.
16 Severity : enhancement
18 Description: Update OFED support to 1.4.1
22 Description: Reduce small size read RPC
23 Details : Set read-ahead limite for every file and only do read-ahead when
24 available read-ahead pages are bigger than 1M to avoid small size
27 Severity : enhancement
29 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
30 Details : allow skip disconnected ost for send statfs request and hide error
36 Description: fix racy locking of mballoc block bitmaps causing BUG
37 Details : The locking of the mballoc buddy bitmap and the in-memory
38 block bitmap was using two different spin locks in some
39 cases. This made it possible to incorrectly access the
40 mballoc bitmap while another process was modifying it,
41 causing a sanity assertion to fail. While no on-disk corruption
42 was reported, there was some risk of this happening.
45 Frequency : rare, on llog test 6
47 Descriptoin: don't allow connect to already connected import
48 Details : allowing connect to already connected import is hide connecting problem.
53 Descriptoin: allow kill process which wait statahead result
54 Details : for some reasons 'ls' can stick in waiting result from statahead,
55 in this case need way for kill this process.
57 Severity : enhancement
59 Description: Add state history info file, enhance import info file
60 Details : Track import connection state changes in a new osc/mdc proc file;
61 add overview-type data to the osc/mdc import proc file.
63 Severity : enhancement
65 Description: MDS create should not wait for statfs RPC while holding DLM lock.
67 Severity : enhancement
69 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
74 Descriptoin: OST grant too much space to client even there are not enough space.
75 Details : Client will shrink its grant cache to OST if there are no write
76 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
77 this grant cache if there are already not enough avaible space
78 (left_space < total_clients * 32M).
81 Frequency : start MDS on uncleanly shutdowned MDS device
83 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
84 Details : stay in waiting mds<>ost recovery finished produce random bugs
85 due race between two ll_sync thread for one lov target. send
86 ACTIVATE event only if connect realy finished and import have
90 Frequency : rare, connect and disconnect target at same time
92 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
93 Details : don't call obd_disconnect under lov_lock. this long time
94 operation and can block ptlrpcd which answer to connect request.
99 Descriptoin: don't lose wakeup for imp_recovery_waitq
100 Details : recover_import_no_retry or invalidate_import and import_close can
101 both sleep on imp_recovery_waitq, but we was send only one wakeup
105 Frequency : always with long access acl
107 Descriptoin: mds can't pack reply with long acl.
108 Details : mds don't control size of acl but they limited by reint/getattr
111 Severity : enhancement
113 Description: Update to SLES10 kernel-2.6.16.60-0.33.
115 Severity : enhancement
117 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
120 Frequency : start MDS on uncleanly shutdowned MDS device
122 Descriptoin: aborting recovery hang on MDS
123 Details : don't throttle destroy RPCs for the MDT.
126 Frequency : on remount
128 Description: external journal device not working after the remount
129 Details : clear dev_rdonly flag for external journal devices in
135 Description: shutdown vs evict race
136 Details : client_disconnect_export vs connect request race.
137 if client will evicted at this time - we start invalidate
138 thread without referece to import and import can be freed
142 Frequency : rare, need acl's on inode.
144 Description: client can't handle ost additional correctly
145 Details : if ost was added after client connected to mds client can have
146 hit lnet_try_match_md ... to big messages to wide striped files.
147 in this case need teach client to handle config events about add
148 lov target and update client max ea size at that event.
150 Severity : enhancement
152 Description: Changelogs
153 Details : Changelogs are a lightweight mechanism to track filesystem
154 metadata and namespace changes. The changelog is recorded
155 permanently on the MDTs, and is periodically "consumed" / purged
156 when records are no longer needed.
158 Severity : enhancement
160 Description: compact fld format with extents
161 Details : Store range of seq rather than every seq in FLD. Seq
162 controller update FLD rather than clients. In Case of CMD, mdt0
163 has FLD, all other metadata server act as non persistent proxy
164 for FLD queries and cache fld entries in fld cache.
169 Description: don't skip ost target if they assigned to file
170 Details : Drop slow OSCs if we can, but not for requested start idx.
171 This means "if OSC is slow and it is not the requested
172 start OST, then it can be skipped, otherwise skip it only
173 if it is inactive/recovering/out-of-space.
177 Description: more cleanup in mds_lov
178 Details : not send LOV EA under replay, we can't know about they size at this
179 time. Don't allow client connect to mds before any ost connected,
180 for avoid problems with LOV EA size and returning EIO to client.
182 Severity : enhancement
184 Description: Interoperability at server side (Disk interoperability)
186 Severity : enhancement
188 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
190 Severity : enhancement
192 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
194 Severity : enhancement
196 Description: New client IO stack (CLIO).
198 Severity : enhancement
200 Description: Commit on sharing. Eliminate inter-client dependencies between
201 uncommitted transactions by doing transaction commits.
202 Thereby clients may recovery independently.
205 Frequency : Create a symlink file with a very long name
207 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
208 Details : If there is no extra space in the request for early cancels,
209 ldlm_req_handles_avail() returns 0 instead of a negative value.
211 Severity : enhancement
213 Description: Add /proc entry for import status
214 Details : The mdc, osc, and mgc import directories now have
215 an import directory that contains useful import data for debugging
218 Severity : enhancement
220 Description: Re-disable certain /proc logging
221 Details : Enable and disable client's offset_stats, extents_stats and
222 extents_stats_per_process stats logging on the fly.
225 Frequency : Only on FC kernels 2.6.22+
227 Description: oops in statahead
228 Details : Do not drop reference count for the dentry from VFS when lookup,
229 VFS will do that by itself.
231 Severity : enhancement
233 Description: Generic /proc file permissions
234 Details : Set /Proc file permissions in a more generic way to enable non-
235 root users operate on some /proc files.
239 Description: Hitting mdc_commit_close() ASSERTION
240 Details : Properly handle request reference release in
241 ll_release_openhandle().
245 Description: quota recovery deadlock during mds failover
246 Details : This patch includes att18982, att18236, att18237 in bz14840.
248 1. fix osts hang when mds does failover with quotaon
249 2. prevent watchdog storm when osts threads wait for the
254 Frequency : only patchless client
255 Description: add workaround for race between add/remove dentry from hash
257 Severity : enhancement
259 Description: Allow OST glimpses to return PW locks
263 Description: LBUG when llog conf file is full
264 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
268 Description: Prevent import from entering FULL state when server in recovery
272 Description: service mount cannot take device name with ":"
273 Details : Only when device name contains ":/" will mount treat it as
279 Description: replace ptlrpcd with the statahead thread to interpret the async
280 statahead RPC callback
284 Frequency : on recovery
285 Description: I/O failures after umount during fail back
286 Details : if client reconnected to restarted server we need join to recovery
287 instead of find server handler is changed and process self
288 eviction with cancel all locks.
290 Severity : enhancement
292 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
294 Severity : enhancement
296 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
298 Severity : enhancement
300 Description: Upcall on Lustre log has been dumped
301 Details : Allow for a user mode script to be called once a Lustre log has
302 been dumped. It passes the filename of the dumped log to the
303 script, the location of the script can be specified via
304 /proc/sys/lnet/debug_log_upcall.
309 Description: avoid idr_remove called for id which is not allocated.
310 Details : Move assigment s_dev for clustered nfs to end of initialization,
311 to avoid problem with error handling.
316 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
317 Details : When connection is reused this not moved from CONN_UNUSED_HASH
318 into CONN_USED_HASH and this prodice warning when put connection
319 again in unused hash.
322 Severity : enhancement
324 Description: File striping can now be set to use an arbitrary pool of OSTs.
326 Severity : enhancement
328 Description: Export bytes_read/bytes_write count on OSC/OST.
332 Description: Early reply size mismatch, MGC loses connection
333 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
334 the connect flags are properly negotiated.
339 Description: quotas are not honored with O_DIRECT
340 Details : all writes with the flag O_DIRECT will use grants which leads to
341 this problem. Now using OBD_BRW_SYNC to guard this.
345 Description: add quota statistics
346 Details : 1. sort out quota proc entries and proc code.
347 2. add quota statistics
349 Severity : enhancement
351 Description: enable quota support for HEAD.
355 Description: Properly propagate oinfo flags from lov to osc for statfs
356 Details : restore missing copy oi_flags to lov requests.
358 Severity : enhancement
360 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
361 : l_getgroups(8), lst(8), routerstat(8)
362 Details : included man pages for llobdstat(8), llstat(8),
363 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
365 Severity : enhancement
367 Description: configure's --enable-quota should check the
368 : kernel .config for CONFIG_QUOTA
369 Details : configure is terminated if --enable-quota is passed but
370 : no quota support is in kernel
374 Description: Remove portals compatibility
375 Details : Remove portals compatibility, not interoperable with releases
380 Description: Resolve device initialization race
381 Details : Prevent proc handler from accessing devices added to the
382 obd_devs array but yet be intialized.
384 Severity : enhancement
386 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
388 Severity : enhancement
390 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
395 Description: Using wrong pointer in osc_brw_prep_request
396 Details : Access to array[-1] can produce panic if kernel compiled with
397 CONFIG_PAGE_ALLOC enabled
401 Description: Client runs out of low memory
402 Details : Consider only lowmem when counting initial number of llap pages
406 Description: *optional* service tags registration
407 Details : if the "service tags" package is installed on a Lustre node
408 When the filesystem is mounted, a local-node service tag will
409 be created. See http://inventory.sun.com/ for more information
410 about the Service Tags asset management system.
414 Description: Kernel BUG tries to release flock
415 Details : Lustre does not destroy flock lock before last reference goes
416 away. So always drop flock locks when client is evicted and
417 perform unlock regardless of successfulness of speaking to MDS.
421 Description: add refcount protection for osc callbacks, avoid panic on shutdown
425 Description: sanity test 65a fails if stripecount of -1 is set
426 Details : handle -1 striping on filesystem in ll_dirstripe_verify
431 Description: ASSERTION(CheckWriteback(page,cmd)) failed
432 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
435 Severity : enhancement
437 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
441 Description: do not process already freed flock
442 Details : flock can possibly be freed by another thread before it reaches
443 to ldlm_flock_completion_ast.
447 Description: LBUG during stress test
448 Details : Need properly lock accesses the flock deadlock detection list.
452 Description: oops in page fault handler
453 Details : kernel page fault handler can return two special 'pages' in error
454 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
458 Description: timeout with invalidate import.
459 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
460 should be handled by ptlrpcd. This produce long age waiting and
461 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
463 Severity : enhancement
465 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
469 Description: enable MGS and MDT services start separately
470 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
471 a co-located MGS without starting the MGS, which is a complement
472 to 'nosvc' mount option.
476 Frequency : after recovery
477 Description: precreate to many object's after del orphan.
478 Details : del orphan st in oscc last_id == next_id and this triger growing
479 count of precreated objects. Set flag LOW to skip increase count
480 of precreated objects.
484 Frequency : rare, on clear nid stats
485 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
486 Details : when clean nid stats sometimes try destroy live entry,
487 and this produce panic in free.
491 Description: Stack overflow during MDS log replay
492 ease stack pressure by using a thread dealing llog_process.
496 Description: wait until IO finished before start new when do lock cancel.
497 Details : VM protocol want old IO finished before start new, in this case
498 need wait until PG_writeback is cleared until check dirty flag and
499 call writepages in lock cancel callback.
501 Severity : enhancement
503 Description: using special macro for print time and cleanup in includes.
507 Description: mds_mfd_close() ASSERTION(rc == 0)
508 Details : In mds_mfd_close(), we need protect inode's writecount change
509 within its orphan write semaphore to prevent possible races.
513 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
517 Frequency : rare, on shutdown ost
518 Description: don't hit live lock with umount ost.
519 Details : shrink_dcache_parent can be in long loop with destroy dentries,
520 use shrink_dcache_sb instead.
524 Description: don't panic with use echo client
525 Details : echo client pass NULL as client nid pointer and this produce null
530 Description: fix build on ppc32
531 Details : compile code with -m64 flag produce wrong object file for ppc32.
535 Description: add message levels for liblustreapi
539 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
540 Details : Move the 'good_osts' check before the 'total_bavail' check. This
541 will result in an -EAGAIN and in the exit call path we call
542 alloc_rr() which will with increasing aggressiveness attempt to
543 aquire precreated objects on the minimum number of required OSCs.
547 Description: Use old size assignment to avoid deadlock
548 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
549 the scheduling while holding a spinlock. We do not need locking
550 for size in ll_update_inode() because size is only updated from
551 the MDS for directories or files without objects, so there is no
552 other place to do the update, and concurrent access to such inodes
553 are protected by the inode lock.
557 Description: resolve "_IOWR redefined" build error on SLES10
561 Description: dump the memory debugging after all modules are unloaded to
562 suppress false negative in conf_sanity test 39
564 Severity : enhancement
566 Description: build kernel-ib packages for OFED 1.3 in our release cycle
571 Description: fix SLES kernel versioning
572 Details : the kernel version for our SLES 10 kernel did not include a "-"
573 before the "smp" at the end. while this was not a problem in
574 general, it did mean that software trying to use the kernel
575 version to try to detect a vendor specific kernel would fail.
576 this was most evident by the OFED build scripts.
580 Description: Don't update lov_desc members until making sure they are valid
581 Details : When updating lov_desc members via proc fs, need fix their
582 validities before doing the real update.
586 Description: don't put request into delay list while invalidate in flight.
587 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
588 import in flight. this produce timeout for invalidate and sometimes
589 can cause stale data.
593 Frequency : on ppc only
594 Description: not convert ost objects for directory because it's not exist.
595 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
599 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
601 Severity : enhancement
603 Description: Update to support 2.6.22.14 vanilla kernel.
607 Frequency : rare, on recovery
608 Description: read procfs can produce deadlock in some situation
609 Details : Holding lprocfs lock which send rpc can produce block for destroy
610 obd objects and this also block reconnect with -EALREADY. This
611 isn't fix all lprocfs bugs - but make it rare.
613 Severity : enhancement
615 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
618 Frequency : frequent on X2 node
620 Description: mdc_set_open_replay_data LBUG
621 Details : Set replay data for requests that are eligible for replay.
625 Description: lustre_mgs: operation 101 on unconnected MGS
626 Details : When MGC is disconnected from MGS long enough, MGS will evict the
627 MGC, and late on MGC cannot successfully connect to MGS and a lot
628 of the error messages complaining that MGS is not connected.
632 Frequency : on network error
633 Description: panic with double free request if network error
634 Details : mdc_finish_enqueue is finish request if any network error ocuring,
635 but it's true only for synchronus enqueue, for async enqueue
636 (via ptlrpcd) this incorrect and ptlrpcd want finish request
639 Severity : enhancement
641 Description: client-side metadata stat-ahead during readdir(directory readahead)
642 Details : perform client-side metadata stat-ahead when the client detects
643 readdir and sequential stat of dir entries therein
646 Frequency : on start mds
648 Description: Implement get_info(last_id) in obdfilter.
651 Frequency : occasional
653 Description: Correctly check stale fid, not start epoch if ost not support SOM
654 Details : open with flag O_CREATE need set old fid in op_fid3 because
655 op_fid2 was overwritten with new generated fid, but mds can answer
656 with one of these two fids and both is not stale. Setattr
657 incorrectly started an epoch and assume will be called
658 done_writing, but without SOM done_writing ever being called.
661 Frequency : rare, depends on device drivers and load
663 Description: MDS or OSS nodes crash due to stack overflow
664 Details : Code changes in 1.8.0 increased the stack usage of some functions.
665 In some cases, in conjunction with device drivers that use a lot
666 of stack the MDS (or possibly OSS) service threads could overflow
667 the stack. One change which was identified to consume additional
668 stack has been reworked to avoid the extra stack usage.
671 Frequency : occasional
673 Description: Do not fail import if osc_interpret_create gets -EAGAIN
674 Details : If osc_interpret_create got -EAGAIN it immediately exits and
675 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
676 oscc_has_objects and see OSC has no objests and call
677 oscc_internal_create to resend create request.
679 Severity : enhancement
681 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
683 Severity : enhancement
685 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
688 Frequency : very rare
690 Description: Fix replay if there is an un-replied request and open
691 Details : In some cases, older replay request will revert the
692 mcd->mcd_last_xid on MDS which is used to record the client's
695 Severity : enhancement
697 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
699 Severity : enhancement
701 Description: Add rhel5 support to HEAD.
703 Serverity : enhancement
705 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
710 Description: Don't allow skipping OSTs if index has been specified.
711 Details : Don't allow skipping OSTs if index has been specified, make
712 locking in internal create lots better.
716 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
717 Details : access to bitfield in structure is always rounded to long
718 and this produce problem with not atomic change any bit.
722 Description: Lustre make rpms failed.
723 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
724 builds Lustre rpms from tarball.
727 Frequency : rare on shutdown ost
729 Description: If llog cancel was not send before clean_exports phase, this can
730 produce deadlock in llog code.
731 Details : If llog thread has last reference to obd and call class_import_put
732 this produce deadlock because llog_cleanup_commit_master wait when
733 last llog_commit_thread exited, but this never success because was
734 called from llog_commit_thread.
738 Description: allow userland application know is lost one of stripes.
739 Details : fill lvb_blocks with error code on ost and return it to
740 application if error flag found.
744 Description: NULL lov_tgts causing MDS oops
745 Details : more safe checks for NULL lov_tgts for avoid oops.
747 Severity : enhancement
749 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
753 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
754 Details : If an OST is down the MDS will hang indefinitely in
755 obd_statfs_rqset() waiting for the statfs data. While for
756 MDS QOS usage of statfs, it should not stuck in waiting.
758 Severity : enhancement
760 Description: remote_acl support
761 Details : Support ACL-based permission check for remote user.
762 Support setfacl/getfacl for remote user with the utils
763 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
764 the system "{s,g}etfacl" utils.
766 Severity : enhancement
768 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
770 Severity : enhancement
772 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
776 Description: make mgs_setparam() handle fsname containing dash
777 Details : fsname containing a dash does not work with lctl conf_param
779 Severity : enhancement
781 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
783 Severity : enhancement
785 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
789 Frequency : rare, at shutdown
790 Description: access already free / zero obd_namespace.
791 Details : if client_disconnect_export was called without force flag set,
792 and exist connect request in flight, this can produce access to
793 NULL pointer (or already free pointer) when connect_interpret
794 store ocd flags in obd_namespace.
798 Frequency : only at startup
799 Description: not alloc memory with spinlock held.
800 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
801 if any spinlock held.
803 Severity : enhancement
805 Description: make lustre randomly fail allocating memory
806 Details : Make lustre randomly failed allocating memory for testing purpose.
808 Severity : enhancement
810 Description: lost problems with lov objid file
811 Details : Fixes some scability and access to not inited memory problems
812 in work with lov objdid file.
817 Description: lfs find does not continue on file error
818 Details : Continue other files processing when a file/dir is absent.
822 Description: Inconsistent usage of lustre_pack_reply()
823 Details : Standardize the usage of lustre_pack_reply() such that it
824 always generate a CERROR on failure.
827 Frequency : occasional
829 Description: MDS hang or stay in waiting lock
830 Details : If client receive lock with CBPENDING flag ldlm need send lock
831 cancel as separate rpc, to avoid situation when cancel request
832 can't processed due all i/o threads stay in wait lock.
836 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
840 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
844 Description: Update to SLES9 kernel-2.6.5-7.287.3.
848 Description: Update to RHEL4 latest kernel.
850 Severity : enhancement
852 Description: Build SLES10 patchless client fails
853 Details : The configure was broken by run ./configure with
854 --with-linux-obj=.... argument for patchless client. When the
855 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
856 can't search header adequately. Use absolute path such as
857 -I($LINUX)/include instead.
861 Description: interrupt oig_wait produce painc on resend.
862 Details : brw_redo_request can be used for resend requests from ptlrpcd and
863 private set, and this produce situation when rq_ptlrpcd_data not
864 copyed to new allocated request and triggered LBUG on assert
865 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
866 ptlrpcd set if request is changed and can be safety changed to use
871 Description: LASSERT_{REQ,REP}SWAB macros are buggy
872 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
873 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
874 these macros and replace them with their difinitions instead.
878 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
879 Details : Update which_patch & target file for SLES10 latest kernel.
882 Description: Remove client patches from SLES 10 kernel.
883 Details : This causes SLES 10 clients to behave as patchless clients
884 even on a Lustre-patched (server) kernel.
886 Severity : enhancement
888 Description: self-adjustable client's lru lists
889 Details : use adaptive algorithm for managing client cached locks lru
890 lists according to current server load, other client's work
891 pattern, memory activities, etc. Both, server and client
892 side namespaces provide number of proc tunables for controlling
895 Severity : enhancement
897 Description: light-weight GSS support
898 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
899 and reduce performance overhead.
901 Severity : enhancement
903 Description: Linux keyring support
904 Details : Support using service of Linux keyring for Lustre GSS internal
905 context refresh/cache mechanism.
909 Description: Fix errors in lfs documentation
910 Details : Fixes man pages
914 Description: don't use GFP_* in generic Lustre code.
915 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
919 Description: obdclass is limited by single OBD_ALLOC(idarray)
920 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
924 Description: warnings with build patchless client with vanila 2.6.19 and up
925 Details : change old ctl_table style and replace ctl_table/ctl_table_header
926 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
930 Frequency : Only for SLES
931 Description: sanity_quota fail test_1
932 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
933 file, which makes TSTID[2] inunique.
937 Frequency : Only for SLES9
938 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
939 Details : Update target/ChangeLog/which_patch .
943 Frequency : Only for test_10 in sanity-quota.sh
944 Description: a bug in quota test code
945 Details : checking a test flag in a wrong place
949 Frequency : Only for test_13 in sanity-quota.sh
950 Description: a bug in quota test script
951 Details : shouldn't check overfull allocated quota
955 Description: jbd statistics
956 Details : Port older jbd statistics patch for sles10
958 Severity : enhancement
960 Description: Add printing inode info into message about error in writepage.
965 Description: reply_lock_interpret crash due to race with it and lock cancel.
966 Details : Do not replay locks that are being cancelled. Do not reference
967 locks by their address during replay, just by their handle.
971 Frequency : When flocks are used.
972 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
973 Details : Flock locks might destroy just granted lock if it could be merged
974 with another existing flock, this is done in completion handler,
975 so teach ldlm_cli_enquque_fini that this is a valid case for
981 Description: Oops in read and write path when failing to allocate lock.
982 Details : Check if lock allocation failed and return error back.
986 Description: lstripe command fails for valid OST index
987 Details : The stripe offset is compared to lov->desc.ld_tgt_count
988 instead of lov->desc.ld_active_tgt_count.
992 Description: sanity.sh failed test 103
993 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
998 Description: df doesn't work properly if diskfs blocksize != 4K
999 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1003 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1004 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1006 Severity : enhancement
1008 Description: Adaptive timeouts
1009 Details : RPC timeouts adapt to changing server load and network
1010 conditions to reduce resend attempts and improve recovery time.
1014 Description: llapi_file_create() does not allow some changes
1015 Details : add llapi_file_open() that allows specifying the mode and
1016 open flags, and also returns an open file handle.
1020 Description: merge and cleanup kernel patches.
1021 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1025 Description: Add journal checksum support.(Kernel part)
1026 Details : The journal checksum feature adds two new flags i.e
1027 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1028 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1029 indicates that the commit block contains the checksum for
1030 the blocks described by the descriptor blocks. Now commit
1031 record can be sent to disk without waiting for descriptor
1032 blocks to be written to disk. This behavior is controlled
1033 using JBD2_FEATURE_ASYNC_COMMIT flag.
1037 Description: OSS needs mutliple precreate threads
1038 Details : Add ability to start more than one create thread per OSS.
1042 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1043 Details : execute lfs setstripe on client
1047 Description: mds_obd_create error creating tmp object
1048 Details : When the user sets quota on root, llog will be affected and can't
1049 create files and write files.
1052 Frequency : Always on ia64 patchless client, and possibly others.
1054 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1055 Details : This allows the patchless client to be loaded on architectures
1056 without this export.
1060 Description: RedHat Update kernel for RHEL5
1061 Details : Add the kernel config file for RHEL5.
1065 Description: RedHat Update kernel for RHEL5
1066 Details : Modify the kernel config file more closer RHEL5.
1070 Description: Build failure against Centos5 (RHEL5)
1071 Details : Define PAGE_SIZE when it isn't present.
1075 Description: test_7 on sanity-quota.sh can't be used on separate machines
1076 Details : add facet to handle it
1080 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1081 Details : with NFS, the anon dentry's parent was set to itself in
1082 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1083 corresponding dentry other than use rec->ur_name.
1085 Severity : enhancement
1087 Description: lfs setstripe enhancement
1088 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1092 Description: enable data checksumming by default
1093 Details : enable checksum by default, allow --disable-checksum
1094 configure option and "-o nochecksum" mount option. Checksums
1095 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1099 Description: lustre support for RHEL5
1100 Details : Add support for RHEL5.
1104 Description: Client eviction due to failover config
1105 Details : after a connection loss, the lustre client should attempt to
1106 reconnect to the last active server first before trying the
1107 other potential connections.
1111 Description: when mds and osts use different quota unit(32bit and 64bit),
1112 quota will be released repeatly.
1113 Details : void sending multiple quota reqs to mds, which will keep the
1114 status between the reqs.
1118 Description: osts not allocated evenly to files
1119 Details : change the condition to increase offset_idx
1122 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1124 Description: Data corruption for OSTs that are formatted larger than 2TB
1126 Details : When generating the bio request for lustre file writes the
1127 sector number would overflow a temporary variable before being
1128 used for the IO. The data reads correctly from Lustre (which
1129 will overflow in a similar manner) but other file data or
1130 filesystem metadata may be corrupted in some cases.
1134 Description: Tune the kernel for good SCSI performance.
1135 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1136 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1141 Description: rewrite ext2-derived code in obdclass/uuid.c
1142 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1143 from scratch preserving functionality.
1147 Description: "lfs find -obd UUID" prints directories
1148 Details : "lfs find -obd UUID" will return all directory names instead
1149 of just file names. It is incorrect because the directories
1150 do not reside on the OSTs.
1154 Description: use i_size_read and i_size_write in 2.6 port
1155 Details : replace inode->i_size access with i_size_read/write()
1158 Frequency : when removing large files
1160 Description: scheduling issue during removal of large Lustre files
1161 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1162 It causes scheduling issues when removing large files (17TB in the
1166 Frequency : only with liblustre clients on XT3
1168 Description: evictions taking too long
1169 Details : allow llrd to evict clients directly on OSTs
1172 Frequency : only on ppc
1174 Description: /proc/fs/lustre/devices broken on ppc
1175 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1176 We should make sure the type of 'index' is loff_t and then cast
1177 explicitly as needed below. Do not assign an explicitly cast
1181 Frequency : only for rhel5
1183 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1184 Details : Modify the target file & which_kernel.
1186 Severity : enhancement
1188 Description: omit set fsid for export NFS
1189 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1190 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1194 Frequency : Always, for kernels after 2.6.16
1195 Description: Fix warning idr_remove called for id=.. which is not allocated.
1196 Details : Last kernels save old s_dev before kill super and not allow
1197 to restore from callback - restore it before call kill_anon_super.
1201 Description: buffer overruns could theoretically occur
1202 Details : llapi_semantic_traverse() modifies the "path" argument by
1203 appending values to the end of the origin string, and a
1204 overrun may occur. Adding buffer overrun check in liblustreapi.
1208 Description: Fix error on 'ls .' at the top of the Lustre mount.
1209 Details : Don't revalidate dentry if it is a root dentry.
1213 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1214 Details : Modify vm-tunables-rhel4.patch.
1218 Description: Kernel config for 2.6.18-vanilla.
1219 Details : Modify targets/2.6-vanilla.target.in.
1220 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1221 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1222 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1223 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1227 Description: improve handling recoverable errors
1228 Details : if request processig with error which can be recoverable on server
1229 request should be resend, otherwise page released from cache and
1234 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1235 Details : Modify target file & which_patch.
1236 A flaw was found in the IA32 system call emulation provided
1237 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1238 value could be stored in the %RAX register, which could trigger an
1239 out-of-bounds system call table access. An untrusted local user
1240 could exploit this flaw to run code in the kernel
1241 (ie a root privilege escalation). (CVE-2007-4573).
1245 Description: O_DIRECT bypasses client statistics.
1246 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1247 read_ahead_stats were updated. copied the stats section from
1248 osc_send_oap_rpc() into async_internal().
1252 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1256 Description: change order of libsysio includes
1257 Details : '#include sysio.h' should always come before '#include xtio.h'
1261 Description: handle "serious error: objid * already exists" more gracefully
1262 Details : If LAST_ID value on disk is smaller than the objects existing in
1263 the O/0/d* directories, it indicates disk corruption and causes an
1264 LBUG(). If the object is 0-length, then we should use the existing
1265 object. This will help to avoid a full fsck in most cases.
1267 Severity : enhancement
1269 Description: adapt the lustre_config script to support the upgrade case
1270 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1271 server targets to 1.6 in parallel.
1275 Description: To avoid grant space > avaible space when the disk is almost
1276 full. Without this patch you might see the error "grant XXXX >
1277 available" or some LBUG about grant, when the disk is almost
1279 Details : In filter_check_grant, for non_grant cache write, we should
1280 check the left space by if (*left > ungranted + bytes), instead
1281 of (*left > ungranted), because only we are sure the left space
1282 is enough for another "bytes", then the ungrant space should be
1283 increase. In client, we should update cl_avail_grant only there
1284 is OBD_MD_FLGRANT in the reply.
1288 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1289 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1290 described in CVE-2007-4573. This problem could allow untrusted
1291 local users to gain root access.
1294 Frequency : when using O_DIRECT and quotas
1296 Description: Incorrect file ownership on O_DIRECT output files
1297 Details : block usage reported by 'lfs quota' does not take into account
1298 files that have been written with O_DIRECT.
1303 Description: touch file failed when fs is not full
1304 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1305 otherwise we can get ENOSP while fs is not full.
1309 Description: parallel lock callbacks
1310 Details : Instead of sending blocking and completion callbacks as separated
1311 requests, adding them to a set and sending in parallel.
1314 Frequency : only for Cray XT3
1315 Bugzilla : 12829/13455
1316 Description: Changing primary group doesn't change the group lustre assigns to
1318 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1319 overridden with the primary group provided by the group upcall,
1320 whereas the supplied fsgid can be trusted if it is in the list of
1321 supplementary groups returned by the group upcall.
1323 Severity : enhancement
1325 Description: Allow masking D_WARNING, D_ERROR messages from console
1326 Details : Console messages can now be disabled via lnet.printk.
1330 Description: User code with malformed file open parameter crashes client node
1331 Details : Before packing join_file req, all the related reference should be
1332 checked carefully in case some malformed flags cause fake
1333 join_file req on client.
1337 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1338 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1339 mode is equal to the granted mode. However, it should also check
1340 that LDLM_FL_CP_REQD is not set.
1344 Description: Heavy nfs access might result in deadlocks
1345 Details : After ELC code landed, it is now improper to enqueue any mds
1346 locks under och_sem, because enqueue might want to decide to
1347 cancel open locks for same inode we are holding och_sem for.
1351 Description: Client eviction while running blogbench
1352 Details : A lot of unlink operations with concurrent I/O can lead to a
1353 deadlock causing evictions. To address the problem, the number of
1354 oustanding OST_DESTROY requests is now throttled to
1355 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1356 callbacks are processed in priority.
1360 Description: enable ACLs on MDS by default
1361 Details : ACLs must be enabled on MDS by default.
1364 Frequency : PPC/PPC64 only
1366 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1367 Details : fix duplicated definitions between asm-ppc64/types.h and
1368 lustre_types.h on PPC.
1371 Frequency : PPC/PPC64 only
1373 Description: asm-ppc/segment.h does not exist
1374 Details : fix compile issue on PPC.
1378 Description: better handle error messages in extents code
1381 Frequency : RHEL4 only
1383 Description: mkfs is very slow on IA64/RHEL4
1384 Details : A performance regression has been discovered in the MPT Fusion
1385 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1386 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1387 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1389 Severity : enhancement
1391 Description: SNMP support enhancement
1392 Details : Adding total number of sampled request for an MDS node in snmp
1395 Severity : enhancement
1397 Description: Optimize ldlm waiting list processing for PR extent locks
1398 Details : When processing waiting list for read extent lock and meeting read
1399 lock that is same or wider to it that is not contended, skip
1400 processing rest of the list and immediatelly return current
1401 status of conflictness, since we are guaranteed there are no
1402 conflicting locks in the rest of the list.
1406 Description: Time out and refuse to reconnect
1407 Details : When the failover node is the primary node, it is possible
1408 to have two identical connections in imp_conn_list. We must
1409 compare not conn's pointers but NIDs, otherwise we can defeat
1410 connection throttling.
1414 Description: port llog fixes from b1_6 into HEAD
1415 Details : Port llog reference couting and some llog cleanups from b1_6
1416 (bug 10800) into HEAD, for protect from panic and access to already
1417 free llog structures.
1421 Description: Detect stride IO mode in read-ahead
1422 Details : When a client does stride read, read-ahead should detect that and
1423 read-ahead pages according to the detected stride pattern.
1427 Description: data checksumming impacts single node performance
1428 Details : add support for several checksum algorithm. Currently, only CRC32
1429 and Adler-32 are supported. The checksum type can be changed on
1430 the fly via /proc/fs/lustre/osc/*/checksum_type.
1434 Description: use adler32 for page checksums
1435 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1440 Description: build for x2 fails
1441 Details : fix compile issue on Cray systems.
1445 Description: Properly match for duplicate locks
1446 Details : Due to different lock order from skiplists code, we need to
1447 traverse entire list for now
1450 Frequency : only on PPC/SLES10
1452 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1453 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1454 be defined. Add a hack in mkfs_lustre.c to work around this compile
1459 Description: LASSERT on MDS when client holding flock lock dies
1460 Details : ldlm pool logic depends on number of granted locks equal to
1461 number of released locks which is not true for flock locks, so
1462 just exclude such locks from consideration.
1466 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1467 Details : Use fsfilt_sync() for both the whole filesystem sync and
1468 individual file sync to eliminate dangerous inode locking
1469 with I_LOCK that can lead to a deadlock.
1473 Description: performance in 1.6.3
1474 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1475 sure that 1MB requests are not fragmented by the block layer.
1477 Severity : enhancement
1479 Description: organize the server-side client stats on per-nid basis
1480 Details : Change the structure of stats under obdfilter and mds to
1488 The "uuid"s file would list the uuids of _active_ exports.
1489 And the clear entry is to clear all stats and stale nids.
1491 Severity : enhancement
1493 Description: eliminate client locks in face of contention
1494 Details : file contention detection and lockless i/o implementation
1495 for contended files.
1499 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1502 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1504 Description: MDS may lose file striping (and hence file data) in some cases
1505 Details : If there are additional extended attributes stored on the MDS,
1506 in particular ACLs, SELinux, or user attributes (if user_xattr
1507 is specified for the client mount options) then there is a risk
1508 of attribute loss. Additionally, the Lustre file striping
1509 needs to be larger than default (e.g. striped over all OSTs),
1510 and an additional attribute must be stored initially in the
1511 inode and then increase in size enough to be moved to the
1512 external attribute block (e.g. ACL growing in size) for file
1517 Description: skiplist implementation simplification
1518 Details : skiplists are used to group compatible locks on granted list
1519 that was implemented as tracking first and last lock of each
1520 lock group the patch changes that to using doubly linked lists
1524 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1525 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1526 because LASSERT on both the data supplied by a client, and the
1527 data on disk is dangerous and incorrect.
1529 Severity : enhancement
1531 Description: Slow truncate/writes to huge files at high offsets.
1532 Details : Directly associate cached pages to lock that protect those pages,
1533 this allows us to quickly find what pages to write and remove
1534 once lock callback is received.
1538 Description: more ldlm soft lockups
1539 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1540 starve other threads from the resource lock for a long time in
1541 case of long waiting queue, so change the debug level from
1542 D_OTHER to the less frequently used D_INFO.
1544 Severity : enhancement
1546 Description: add -gid, -group, -uid, -user options to lfs find
1550 Description: Hung threads in invalidate_inode_pages2_range
1551 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1552 once one is completed. As a result, some RPCs are stuck in the
1553 queue and are never sent.
1557 Description: filter threads hungs on waiting journal commit
1558 Details : Cleanup filter group llog code, then only filter group llog will
1559 be only created in the MDS/OST syncing process.
1563 Description: Procfs and llog threads access destoryed import sometimes.
1564 Details : Sync the import destoryed process with procfs and llog threads by
1565 the import refcount and semaphore.
1567 Severity : enhancement
1569 Description: openlock cache of b1_6 port to HEAD
1574 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1575 Details : If insertion of an extent fails, then discard the inode
1576 preallocation and free data blocks else it can lead to duplicate
1581 Description: don't always update ctime in ext3_xattr_set_handle()
1582 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1583 In some cases the ctime should not be updated, for example for
1584 2.0->1.8 compatibility it is necessary to delete an xattr and it
1585 should not update the ctime.
1589 Bugzilla : 15713/16362
1590 Description: Assertion in iopen_connect_dentry in 1.6.3
1591 Details : looking up an inode via iopen with the wrong generation number can
1592 populate the dcache with a disconneced dentry while the inode
1593 number is in the process of being reallocated. This causes an
1594 assertion failure in iopen since the inode's dentry list contains
1595 both a connected and disconnected dentry.
1599 Description: assertion failure in ldlm_handle2lock()
1600 Details : fix a race between class_handle_unhash() and class_handle2object()
1601 introduced in lustre 1.6.5 by bug 13622.
1606 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1607 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1612 Description: oops in sd_iostats_seq_show()
1613 Details : unloading/reloading the scsi low level driver triggers a kernel
1614 bug when trying to access the sd iostat file.
1619 Description: Kernel panics during QLogic driver reload
1620 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1621 patch, causing memory corruption.
1626 Description: journal_dev option does not work in b1_6
1627 Details : pass mount option during pre-mount.
1629 Severity : enhancement
1631 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1632 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1633 extent information of a file. It can be used to map logical blocks
1634 in a file to physical blocks in the block device.
1638 Description: LDLM soft lockups - improvement
1639 Details : It is be possible to send the lock handle along with each read
1640 or write request because the client is already doing a lock match
1641 itself so there isn't any reason the OST should have to re-do that
1645 Frequency : only on Cray X2
1647 Description: X2 build failures
1648 Details : fix build failures on Cray X2.
1652 Description: xid & resent requests
1653 Details : Initialize RPC XID from clock at startup (randomly if clock is
1656 Severity : enhancement
1658 Description: Add lustre_start utility to start or stop multiple Lustre servers
1663 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1664 Details : In case of memory pressure, list_del() can be called twice on
1665 req->rq_history_list, causing a kernel oops.
1669 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1670 Details : fix stack overflow in the distributed lock manager by defering
1671 export eviction after a failed AST to the elt thread instead of
1672 handling it in the dlm interpret routine.
1676 Description: Convert some comments to new format.
1677 Details : Update documenting comments to match doxygen conventions.
1681 Description: Grammar fixes.
1682 Details : A couple of trivial spelling fixes.
1686 Description: OSD_COUNTERS-mandatory
1687 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1688 to simplify the code.
1692 Description: simplify cmm_device freeing logic.
1693 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1697 Description: Add lockdep support to dt_object_operations locking interface.
1698 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1699 parameter indicating lock ordering. Update mdd code to use new
1704 Description: Introduce failloc constants for lockless IO tests.
1705 Details : Add two new failloc constants to test lockless IO. Only one of
1706 them in implemented---another is checked in yet to be landed
1711 Description: Add lockdep support for inode mutex.
1712 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1713 in the situations where Lustre has to lock more than one inode
1718 Description: Add optional invariants checking support.
1719 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1720 configure switch. This macro is to be used for consistency and
1721 sanity checks that are too expensive to be left in `production'
1726 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1727 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1732 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1733 Details : Use spin_lock_nested() in (the only) situation where more than
1734 one ldlm_lock is locked simultaneously. Also, fix possible
1735 dead-lock in ldlm_lock_change_resource() by enforcing particular
1740 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1741 Details : Instead of specifying each ldlm_lock call-back through separate
1742 parameter, wrap them into struct ldlm_callback_suite.
1746 Description: Kill join_lru obd method and its callers.
1747 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1748 in memory---a requirement implemented through ->o_join_lru() obd
1749 method in HEAD. Remove this method and its users.
1753 Description: Add asynchronous ldlm ENQUEUE completion handler.
1754 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1755 case that stock ldlm_completion_ast() cannot handle as it waits
1756 until lock is granted. Introduce new ldlm_completion_ast_async()
1757 for this. Also comment ldlm_completion_ast().
1761 Description: ldlm_error <-> errno conversion.
1762 Details : Add functions to map (rather arbitrary) between LDLM error codes
1763 and standard errno values. CLIO needs this to prevent LDLM specific
1764 constants from escaping ldlm and osc.
1768 Description: Kill unused ldlm_handle2lock_ns() function.
1769 Details : Kill unused ldlm_handle2lock_ns() function.
1773 Description: Add lu_ref support to ldlm_lock
1774 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1775 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1776 and ldlm_resource. LDLM interface has to be changed, because code that
1777 releases a reference on a lock, has to "know" what reference this is.
1778 In the most frequent case
1780 lock = ldlm_handle2lock(handle);
1782 LDLM_LOCK_PUT(lock);
1784 no changes are required. When any other reference (received _not_ from
1785 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1786 instead of LDLM_LOCK_PUT().
1788 Arguably, changes are pervasive, and interface requires some discipline
1789 for proper use. On the other hand, it was very instrumental in finding
1790 a few leaked lock references.
1794 Description: Add ldlm_lock_addref_try().
1795 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1796 attempts to addref a lock that might be being canceled
1801 Description: Add ldlm_weigh_callback().
1802 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1803 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1804 hard-coded `number of pages' logic.
1808 Description: Add lockdep annotations to llog code.
1809 Details : Use appropriately tagged _nested() locking calls in the places
1810 where llog takes more than one ->lgh_lock lock.
1814 Description: Add loi_kms_set().
1815 Details : Wrap kms updates into a helper function.
1819 Description: Constify instances of struct lsm_operations.
1820 Details : Constify instances of struct lsm_operations.
1824 Description: lu_conf support.
1825 Details : On a server, a file system object is uniquely identified
1826 by a fid, which is sufficient to locate and load all object
1827 state (inode). On a client, on the other hand, more data are
1828 necessary instantiate an object. Change lu_object_find() and
1829 friends to take additional `lu_conf' argument describing object.
1830 Typically this includes layout information.
1834 Description: lu_context fixes.
1835 Details : Introduce new lu_context functions that are needed on the client
1836 side, where some system threads (ptlrpcd) are shared by multiple
1837 modules, and so cannot be stopped during module shutdown.
1841 Description: Add start and stop methods to lu_device_type_operations.
1842 Details : Introduce two new methods in lu_device_type_operations, that are
1843 invoked when first instance of a given type is created and last one
1844 is destroyed respectively. This is need by CLIO.
1848 Description: Add lu_ref support to struct lu_device.
1849 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1850 track leaked references.
1854 Description: Introduce lu_kmem_descr.
1855 Details : lu_kmem_descr and its companion interface allow to create
1856 and destroy a number of kmem caches at once.
1860 Description: Fix lu_object finalization race.
1861 Details : Fix a race between lu_object_find() finding an object and its
1862 concurrent finalization. This race is (most likely) not possible
1863 on the server, but might happen on the client.
1867 Description: Introduce lu_ref interface.
1868 Details : lu_ref is a debugging module allowing to track references to
1869 a given object. It is quite cpu expensive, and has to be
1870 explicitly enabled with --enable-lu_ref. See usage description
1875 Description: Factor lu_site procfs stats into a separate function.
1876 Details : Separate lu_site stats printing code into a separate function
1877 that can be reused on a client.
1881 Description: Constify instances of struct {lu,dt,md}_device_operations.
1882 Details : Constify instances of struct {lu,dt,md}_device_operations.
1886 Description: Introduce struct md_site and move meta-data specific parts of
1887 struct lu_site here.
1888 Details : Move md-specific fields out of struct lu_site into special struct
1889 md_site, so that lu_site can be used on a client.
1893 Description: Kill mdd_lov_destroy().
1894 Details : Remove unused mdd code.
1898 Description: Add st_block checking to multistat.c.
1899 Details : Add st_block checking to multistat.c.
1903 Description: Add lu_ref support to struct obd_device.
1904 Details : Add lu_ref tracking to obd_device.
1908 Description: Kill obd_set_fail_loc().
1909 Details : Remove unused code.
1913 Description: Add special type for ptlrpc_request interpret functions.
1914 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1915 there. Actual usage will be in CLIO.
1919 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1920 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1921 doesn't work with lockdep.
1925 Description: Add rwv.c test program.
1926 Details : New testing program exercising readv(2) and writev(2) (Qian).
1930 Description: Add sendfile.c test program.
1931 Details : New testing program exercising sendfile(2) (Jay).
1935 Description: Ratelimit a message that can be very frequent.
1936 Details : Ratelimit a memory allocation failure message that can
1941 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1942 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1946 Description: Liblustre build fixes.
1947 Details : Liblustre build fixes.
1951 Description: libcfs: add cfs_{need,cond}_resched() interface.
1952 Details : libcfs: add cfs_{need,cond}_resched() definition and
1953 implementations for Linux, NT, and liblustre.
1955 Severity : enhancement
1957 Description: More exported tunables for mballoc
1958 Details : Add support for tunable preallocation window and new tunables for
1959 large/small requests
1963 Description: Detect corruption of block bitmap and checking for preallocations
1964 Details : Checks validity of on-disk block bitmap. Also it does better
1965 checking of number of applied preallocations. When corruption is
1966 found, it turns filesystem readonly to prevent further corruptions.
1970 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1971 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1972 window to 0. If it does hit cache miss, and read-pattern is still
1973 stride-io mode, does not reset the stride window, but also does not
1974 increase the stride window length in this case.
1978 Frequency : only for big-endian servers
1979 Description: Check if system is big-endian while mounting fs with extents feature
1980 Details : Mounting a filesystem with extents feature will fail on big-endian
1981 systems since ext3-based ldiskfs is not supported on big-endian
1982 systems. This can be over-riden with "bigendian_extents" mount option.
1984 Severity : enhancement
1986 Description: The root squash functionality
1987 Details : A security feature, which is to prevent users from being able
1988 to mount lustre on their desktop, run as root, and delete
1989 all of the files in the filesystem. The goal is accomplished by
1990 remapping user id (UID) and group id (GID) of the root user to
1991 a UID and GID specified by the system administartor via Lustre
1992 configuration management server (MGS). The functionality also
1993 allows to specify sets of clients for which the remapping does
1998 Description: Excessive recovery window
1999 Details : With AT enabled, the recovery window can be excessively long (6000+
2000 seconds). To address this problem, we no longer use
2001 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2002 timeout no longer depends on the service time, it is set to
2003 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2004 time via pb_service_time.
2008 Description: Watchdog triggered on MDS failover
2009 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2010 the OSTs know that the MDS "UUID" can be reused for the same export
2011 from a different NID, so we do not need to wait for the export to be
2015 Frequency : rare, only if using MMP with Linux RAID
2017 Description: MMP doesn't work with Linux RAID
2018 Details : While using HA for Lustre servers with Linux RAID, it is possible
2019 that MMP will not detect multiple mounts. To make this work we
2020 need to unplug the device queue in RAID when the MMP block is being
2021 written. Also while reading the MMP block, we should read it from
2022 disk and not the cached one.
2024 Severity : enhancement
2026 Description: open file using fid
2027 Details : A file can be opened using just its fid, like
2028 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2031 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2034 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2035 Details : add proper locking for creation of jbd_4k slab cache
2039 Description: MMP check in ext3_remount() fails without displaying any error
2040 Details : When multiple mount protection fails during remount, proper error
2043 --------------------------------------------------------------------------------
2045 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2047 * Support for kernels:
2048 2.6.5-7.283 (SLES 9),
2049 2.6.9-55.EL (RHEL 4),
2050 2.6.16.46-0.14 (SLES 10),
2051 2.6.18.8 vanilla (kernel.org)
2052 * Client support for unpatched kernels:
2053 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2054 2.6.16 - 2.6.22 vanilla (kernel.org)
2055 * Due to recently discovered recovery problems, we do not recommend
2056 using patchless RHEL 4 clients with this or any earlier release.
2057 * Recommended e2fsprogs version: 1.39.cfs8
2058 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2059 * Starting with this release, the ldiskfs backing filesystem required
2060 by Lustre is now in its own package, lustre-ldiskfs. This package
2061 should be installed. It is versioned separately from Lustre and
2062 may be released separately in future.
2066 Description: block reactivating mgc import until all deactivates complete
2067 Details : Fix race when failing back MDT/MGS to itself (testing)
2070 Frequency : at statup only
2072 Description: mds_lov_synchronize race leads to various problems
2073 Details : simultaneous MDT->OST connections at startup can cause the
2074 sync to abort, leaving the OSC in a bad state.
2076 Severity : enhancement
2078 Description: add optional extra BUILD_VERSION info
2079 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2080 to override the lustre version.
2083 Frequency : 2.6.18 servers only
2085 Description: ll_kern_mount() doesn't release the module reference
2086 Details : The ldiskfs module reference count never drops down to 0
2087 because ll_kern_mount() doesn't release the module reference.
2092 Description: server LBUG when using old ost_num_threads parameter
2093 Details : Accept the old ost_num_threads parameter but warn that it
2094 is deprecated, and fix an off-by-one error that caused an LBUG.
2099 Description: Transient SCSI error results in persistent IO issue
2100 Details : iobuf->dr_error is not reinitialized to 0 between two
2104 Frequency : sometimes when underlying device returns I/O errors
2106 Description: OSTs not going read-only during write failures
2107 Details : OSTs are not remounted read-only when the journal commit threads
2108 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2109 instead of the ext3 wrappers.
2113 Description: poor connect scaling with increasing client count
2114 Details : Don't run filter_grant_sanity_check for more than 100 exports
2115 to improve scaling for large numbers of clients.
2118 Frequency : SLES10 only
2120 Description: sanity-quota.sh quotacheck failed: rc = -22
2121 Details : Quotas cannot be enabled on SLES10.
2124 Frequency : liblustre clients only
2126 Description: getdirentries does not give error when run on compute nodes
2127 Details : getdirentries does not fail when the size specified as an argument
2128 is too small to contain at least one entry
2130 Severity : enhancement
2132 Description: Add LNET router traceability for debug purposes
2133 Details : If a checksum failure occurs with a router as part of the
2134 IO path, the NID of the last router that forwarded the bulk data
2135 is printed so it can be identified.
2140 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2141 Details : Due to a race condition, liblustre clients were occasionally
2142 evicted incorrectly.
2144 Severity : enhancement
2146 Description: lfs setstripe use optional parameters instead of postional
2149 Severity : enhancement
2151 Description: Nanosecond timestamp support for ldiskfs
2152 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2153 resolution timestamps. There is not yet support for this at
2154 the Lustre filesystem level.
2157 Frequency : during server recovery
2159 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2160 Details : request with rq_no_resend flag not awake l_wait_event if they get
2164 Frequency : nfs export on patchless client
2166 Description: connectathon hang when test nfs export over patchless client
2167 Details : Disconnected dentry cannot be found with lookup, so we do not need
2168 to unhash it or make it invalid
2171 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2172 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2173 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2178 Description: open req refcounting wrong on reconnect
2179 Details : If reconnect happened between getting open reply from server and
2180 call to mdc_set_replay_data in ll_file_open, we will schedule
2181 replay for unreferenced request that we are about to free.
2182 Subsequent close will crash in variety of ways.
2183 Check that request is still eligible for replay in
2184 mdc_set_replay_data().
2189 Description: disable writes to filesystem when reading health_check file
2190 Details : the default for reading the health_check proc file has changed
2191 to NOT do a journal transaction and write to disk, because this
2192 can cause reads of the /proc file to hang and block HA state
2193 checking on a healthy but otherwise heavily loaded system. It
2194 is possible to return to the previous behaviour during configure
2195 with --enable-health-write.
2197 Severity : enhancement
2199 Description: 64-bit inode version
2200 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2201 to the inode. This will be required for version-based recovery.
2206 Description: MDS fails to start if a duplicate client export is detected
2207 Details : in some rare cases it was possible for a client to connect to
2208 an MDS multiple times. Upon recovery the MDS would detect this
2209 and fail during startup. Handle this more gracefully.
2211 Severity : enhancement
2213 Description: Add -o localflock option to simulate old noflock
2215 Details : This will achieve local-only flock/fcntl locks
2221 Description: log_commit_thread vs filter_destroy race leads to crash
2222 Details : Take import reference before releasing llog record semaphore
2227 Description: Wrong request locking in request set processing
2228 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2229 imp_delayed_list, in this place should be used imp_lock.
2232 Frequency : when reconnection
2234 Description: Grant Leak when osc reconnect to OST
2235 Details : When osc reconnect ost, OST(filter) should check whether it
2236 should grant more space to client by comparing fed_grant and
2237 cl_avail_grant, and return the granted space to client instead
2238 of "new granted" space, because client will call osc_init_grant
2239 to update the client grant space info.
2242 Frequency : when client reconnect to OST
2244 Description: Grant Leak when osc do resend and replay bulk write
2245 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2246 bulk write request, because the grant info while be sync between
2247 OSC and OST when reconnect, and we should ignore the grant info
2248 these of resend/replay write req.
2253 Description: Grant space more than avaiable left space sometimes.
2254 Details : When then OST is about to be full, if two bulk writing from
2255 different clients came to OST. Accord the avaliable space of the
2256 OST, the first req should be permitted, and the second one
2257 should be denied by ENOSPC. But if the seconde arrived before
2258 the first one is commited. The OST might wrongly permit second
2259 writing, which will cause grant space > avaiable space.
2262 Frequency : when client is evicted
2264 Description: Grant might be wrongly erased when osc is evicted by OST
2265 Details : when the import is evicted by server, it will fork another
2266 thread ptlrpc_invalidate_import_thread to invalidate the
2267 import, where the grant will be set to 0. While the original
2268 thread will update the grant it got when connecting. So if
2269 the former happened latter, the grant will be wrongly errased
2270 because of this race.
2275 Description: Checking Stale with correct fid
2276 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2277 to check whether it is stale, because sometimes, we want the
2278 enqueue happened anyway, and op_data.fid2 will not be initialized.
2280 Severity : enhancement
2282 Description: update patchless client
2283 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2286 Frequency : only with 2.4 kernel
2288 Description: random memory corruption
2289 Details : size of struct ll_inode_info is to big for union inode.u and this
2290 can be cause of random memory corruption.
2295 Description: Memory leak in recovery
2296 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2297 It should also check obd_fail before fsfilt_start, otherwise if
2298 fsfilt_start return -EROFS,(failover mds during mds recovery).
2299 then the req will return with repmsg->transno = 0 and rc = EROFS.
2300 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2301 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2302 be freed no matter whether fsfilt_commit success or not.
2305 Frequency : only with huge count clients
2307 Description: Prevents from taking the superblock lock in llap_from_page for
2309 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2310 for llap_from_page call in ll_removepage prevents from taking the
2311 superblock lock for a soon died page.
2316 Description: Not check open intent error before release open handle
2317 Details : in some rare cases, the open intent error is not checked before
2318 release open handle, which may cause
2319 ASSERTION(open_req->rq_transno != 0), because it tries to release
2320 the failed open handle.
2325 Description: Set cat log bitmap only after create log success.
2326 Details : in some rare cases, the cat log bitmap is set too early. and it
2327 should be set only after create log success.
2331 Description: Accessing a block bevice can re-enable I/O when Lustre is
2332 tearing down a device.
2333 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2338 Description: service threads may hog cpus when there are a lot of requests
2339 Details : Insert cond_resched to give other threads a chance to use some CPU
2344 Description: the cat log was not initialized in recovery
2345 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2346 unlink log on mds will not be initialized until mds post
2347 recovery. And also in mds post recovery, the unlink log will
2348 initialization will be done asynchronausly, so there will be race
2349 between add unlink log and unlink log initialization.
2353 Description: brw_stats were being printed incorrectly
2354 Details : brw_stats were being printed as log2 but all of them were not
2355 recorded as log2. Also remove some code duplication arising from
2356 filter_tally_{read,write}.
2360 Frequency : rare, only in recovery.
2361 Description: ASSERTION(req->rq_type != LI_POISON) failed
2362 Details : imp_lock should be held while iterating over imp_sending_list for
2363 prevent destroy request after get timeout in ptlrpc_queue_wait.
2367 Description: replay-single.sh test 52 fails
2368 Details : A lock's skiplist need to be cleanup when it being unlinked
2369 from its resource list.
2373 Description: Short directio read returns full requested size rather than
2375 Details : Direct I/O operations should return actual amount of bytes
2376 transferred rather than requested size.
2378 Severity : enhancement
2380 Description: metadata RPC reduction (e.g. for rm performance)
2381 Details : decrease the amount of synchronous RPC between clients and servers
2382 by canceling conflicing lock before the operation on the client
2383 and packing thier handles into the main operation RPC to server.
2385 Severity : enhancement
2387 Description: Async OSC create to avoid the blocking unnecessarily.
2388 Details : If a OST has no remain object, system will block on the creating
2389 when need to create a new object on this OST. Now, ways use
2390 pre-created objects when available, instead of blocking on an
2391 empty osc while others are not empty. If we must block, we block
2392 for the shortest possible period of time.
2396 Description: improve handling recoverable errors
2397 Details : if request processig with error which can be recoverable on server
2398 request should be resend, otherwise page released from cache and
2401 Severity : enhancement
2403 Description: refine locking for avoid write wrong info into lov_objid file
2404 Details : fix possible races with add new target and write/update data in
2407 --------------------------------------------------------------------------------
2409 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2414 Frequency : on some architectures
2416 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2417 Details : Uninitialized flags sometimes cause configuration commands to
2421 Frequency : patchless clients only
2423 Description: missing __iget() symbol export
2424 Details : The __iget() symbol export is missing. To avoid the need for
2425 this on patchless clients the deathrow inode reaper is turned
2426 off, and we depend on the VM to clean up old inodes. This
2427 dependency was during via the fix for bug 12181.
2429 --------------------------------------------------------------------------------
2431 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2433 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2434 INTEROPERATE with older versions automatically. In many cases a
2435 special upgrade step is needed. Please read the
2436 user documentation before upgrading any part of a live system.
2437 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2438 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2439 * WARNING: Lustre configuration and startup changes are required with
2440 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2444 Severity : enhancement
2446 Description: MountConf
2447 Details : Lustre configuration is now managed via mkfs and mount
2448 commands instead of lmc and lconf. New obd types (MGS, MGC)
2449 are added for dynamic configuration management. See
2450 https://mail.clusterfs.com/wikis/lustre/MountConf for
2453 Severity : enhancement
2455 Description: dynamic OST addition
2456 Details : OSTs can now be added to a live filesystem
2458 Severity : enhancement
2460 Description: startup order invariance
2461 Details : MDTs and OSTs can be started in any order. Clients only
2462 require the MDT to complete startup.
2464 Severity : enhancement
2466 Description: parallel, asynchronous orphan cleanup
2467 Details : orphan cleanup is now performed in separate threads for each
2468 OST, allowing parallel non-blocking operation.
2470 Severity : enhancement
2472 Description: optimized stripe assignment
2473 Details : stripe assignments are now made based on ost space available,
2474 ost previous usage, and OSS previous usage, in order to try
2475 to optimize storage space and networking resources.
2477 Severity : enhancement
2479 Description: Permanently set tunables
2480 Details : All writable /proc/fs/lustre tunables can now be permanently
2481 set on a per-server basis, at mkfs time or on a live system.
2483 Severity : enhancement
2485 Description: Lustre message v2
2486 Details : Add lustre message format v2.
2488 Severity : enhancement
2490 Description: client OST exclusion list
2491 Details : Clients can be started with a list of OSTs that should be
2492 declared "inactive" for known non-responsive OSTs.
2496 Description: SPEC SFS validation failure on NFS v2 over lustre.
2497 Details : Changes the blocksize for regular files to be 2x RPC size,
2498 and not depend on stripe size.
2500 Severity : enhancement
2502 Description: Multiple MD RPCs in flight.
2503 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2504 To avoid overly-overloading MDS, introduce a limit on number of
2505 MDS RPCs in flight for a single client and add /proc controls
2506 to adjust this limit.
2508 Severity : enhancement
2510 Description: client read/write statistics
2511 Details : Add client read/write call usage stats for performance
2512 analysis of user processes.
2513 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2514 file access. extents_stats shows chunk size distribution.
2515 extents_stats_per_process show chunk size distribution per
2518 Severity : enhancement
2520 Description: mds statistics
2521 Details : Add detailed mds operations statistics in
2522 /proc/fs/lustre/mds/*/stats.
2526 Description: Failure of copying files with lustre special EAs.
2527 Details : Client side always return success for setxattr call for lustre
2528 special xattr (currently only "trusted.lov").
2533 Description: Refcount LNET uuids
2534 Details : The global LNET uuid list grew linearly with every startup;
2535 refcount repeated list entries instead of always adding to
2538 Severity : enhancement
2540 Description: Dynamic service threads
2541 Details : Within a small range, start extra service threads
2542 automatically when the request queue builds up.
2545 Frequency : mixed-endian client/server environments
2547 Description: mixed-endian crashes
2548 Details : The new msg_v2 system had some failures in mixed-endian
2551 Severity : enhancement
2553 Description: Easy OST removal
2554 Details : OSTs can be permanently deactivated with e.g. 'lctl
2555 conf_param lustre-OST0001.osc.active=0'
2557 Severity : enhancement
2559 Description: MGS proc entries
2560 Details : Added basic proc entries for the MGS showing what filesystems
2563 Severity : enhancement
2565 Description: provide MGS failover
2566 Details : Added config lock reacquisition after MGS server failover.
2568 Severity : enhancement
2570 Description: add Linux 2.4 support
2571 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2575 Description: a large application tries to do I/O to the same resource and dies
2576 in the middle of it.
2577 Details : Check the req->rq_arrival time after the call to
2578 ost_brw_lock_get(), but before we do anything about
2579 processing it & sending the BULK transfer request. This
2580 should help move old stale pending locks off the queue as
2581 quickly as obd_timeout.
2584 Frequency : when an incorrect nid is specified during startup
2586 Description: ptlrpc connect to non-existant node causes kernel crash
2587 Details : LNET can't be re-entered from an event callback, which
2588 happened when we expire a message after the export has been
2589 cleaned up. Instead, hand the zombie cleanup off to another
2592 Severity : enhancement
2594 Description: plain/inodebits lock performance improvement
2595 Details : Grouping plain/inodebits in granted list by their request modes
2596 and bits policy, thus improving the performance of search through
2600 Frequency : only if OST filesystem is corrupted
2602 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2603 Details : for a short time RPCs with bulk IO are in the replay list,
2604 but replay of bulk IOs is unimplemented. If the OST filesystem
2605 is corrupted due to disk cache incoherency and then replay is
2606 started it is possible to trip an assertion. Avoid putting
2607 committed RPCs into the replay list at all to avoid this issue.
2610 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2612 Description: System hang on startup
2613 Details : This bug allowed the liblustre (e.g. catamount) client to
2614 return to the app before handling all startup RPCs. This
2615 could leave the node unresponsive to lustre network traffic
2616 and manifested as a server ptllnd timeout.
2618 Severity : enhancement
2620 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2621 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2622 boolean to control whether to print peer debug info when a
2623 client's RPC times out.
2626 Frequency : only for kernels with patches from Lustre below 1.4.3
2628 Description: Remove old rdonly API
2629 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2632 Frequency : only for devices with external journals
2634 Description: Set external device read-only also
2635 Details : During a commanded failover stop, we set the disk device
2636 read-only while the server shuts down. We now also set any
2637 external journal device read-only at the same time.
2640 Frequency : when upgrading from 1.4 while trying to change parameters
2642 Description: The wrong (new) MDC name was used when setting parameters for
2643 upgraded MDT's. Also allows changing of OSC (and MDC)
2644 parameters if --writeconf is specified at tunefs upgrade time.
2647 Frequency : when setting specific ost indicies
2649 Description: QOS code breaks on skipped indicies
2650 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2651 created with --index need not be sequential.
2653 Severity : enhancement
2655 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2656 Details : The uninit_groups feature works in conjunction with the kernel
2657 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2658 up the pass1 processing of e2fsck. This is a read-only feature
2659 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2660 mount filesystems that have had this feature enabled.
2662 Severity : enhancement
2664 Description: Improve multi-block allocation algorithm to avoid fragmentation
2665 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2666 allocation locality and avoid filesystem fragmentation.
2668 ------------------------------------------------------------------------------
2670 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2672 * Support for kernels:
2673 2.4.21-47.0.1.EL (RHEL 3)
2674 2.6.5-7.283 (SLES 9)
2675 2.6.9-42.0.10.EL (RHEL 4)
2676 2.6.12.6 vanilla (kernel.org)
2677 2.6.16.27-0.9 (SLES 10)
2678 * Recommended e2fsprogs version: 1.39.cfs5
2680 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2684 Frequency : occasional, depends on client load and configuration
2685 Bugzilla : 12181, 12203
2686 Description: data loss for recently-modified files
2688 Details : In some cases it is possible that recently written or created
2689 files may not be written to disk in a timely manner (this should
2690 normally be within 30s unless client IO load is very high).
2691 The problem appears as zero-length files or files that are a
2692 multiple of 1MB in size after a client crash or client eviction
2693 that are missing data at the end of the file.
2695 This problem is more likely to be hit on clients where files are
2696 repeatedly created and unlinked in the same directory, clients
2697 have a large amount of RAM, have many CPUs, the filesystem has
2698 many OSTs, the clients are rebooted frequently, and/or the files
2699 are not accessed by other nodes after being written.
2701 The presence of the problem can be detected by looking at
2702 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2703 smaller than the second (nr_unused) then dirty files will not
2704 be flushed automatically to disk. "sync; sleep 10" should be
2705 run several times on the node before unmounting it to update
2706 Lustre (this is also safe to run on nodes without this problem).
2708 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2709 that can cause this same problem, so customers using that kernel
2710 also need to update the kernel in addition to Lustre. In order
2711 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2713 It is normal that files written just before a client crash (less
2714 than 30s) may not yet have been flushed to disk, even for local
2718 Frequency : frequent on thin XT3 nodes
2720 Description: UUID collision on thin XT3 Linux nodes
2721 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2722 randomly, since we relied on an insufficiently-seeded PRNG.
2727 Description: OSS hangs after "All ost request buffers busy"
2728 Details : A deadlock between quota and journal operations caused OSS
2729 hangs after printing "All ost request buffers busy."
2732 Frequency : always on liblustre builds
2734 Description: Cleanup compiler warnings on liblustre
2737 Frequency : always on liblustre builds on XT3
2739 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2743 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2744 > 32000 subdirectories
2745 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2746 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2747 subdirectory count crosses 32000. This will aid e2fsck to
2748 correctly handle more than 32000 subdirectories.
2751 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2753 Description: System hang on startup
2754 Details : This bug allowed the liblustre (e.g. catamount) client to
2755 return to the app before handling all startup RPCs. This
2756 could leave the node unresponsive to lustre network traffic
2757 and manifested as a server ptllnd timeout.
2759 Severity : enhancement
2761 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2762 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2763 boolean to control whether to print peer debug info when a
2764 client's RPC times out.
2769 Description: make O_SYNC working on 2.6 kernels
2770 Details : 2.6 kernels use different method for mark pages for write,
2771 so need add a code to lustre for O_SYNC work.
2776 Description: Failure to close file and release space on NFS
2777 Details : Put inode details into lock acquired in ll_intent_file_open.
2778 Use mdc_intent_lock in ll_intent_open to properly
2779 detect all kind of errors unhandled by mdc_enqueue.
2784 Description: proc file read during shutdown sometimes raced obd removal,
2786 Details : Add lock to prevent obd access after proc file removal.
2789 Frequency : Only for files larger than 4GB on 32-bit clients.
2791 Description: improperly doing page alignment of locks
2792 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2793 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2798 Description: RPCs being resent when they shouldn't be
2799 Details : Some RPCs that should not be resent are being resent. This
2800 can cause inconsistencies in the RPC state machine. Do not
2801 resend such requests.
2804 Frequency : rare, only with NFS export
2806 Description: Crash on NFS re-export node
2807 Details : under very unusual load conditions an assertion is hit in
2808 ll_intent_file_open()
2811 Frequency : only if OST filesystem is corrupted
2813 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2814 Details : for a short time RPCs with bulk IO are in the replay list,
2815 but replay of bulk IOs is unimplemented. If the OST filesystem
2816 is corrupted due to disk cache incoherency and then replay is
2817 started it is possible to trip an assertion. Avoid putting
2818 committed RPCs into the replay list at all to avoid this issue.
2823 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2824 Details : Large single O_DIRECT read and write calls can fail to allocate
2825 a sufficiently large buffer to process the request. In case of
2826 allocation failure the allocation is retried with a smaller
2827 buffer and broken into smaller requests.
2829 Severity : enhancement
2831 Description: Add -o localflock option to simulate old noflock behaviour.
2832 Details : This will achieve local-only flock/fcntl locks coherentness.
2837 Description: versioning check is incomplete
2838 Details : Checking the version difference of client vs. server, report
2839 error if the gap is too big.
2844 Description: add support PG_writeback bit
2845 Details : add support for PG_writeback bit for Lustre, for more carefull
2846 work with page cache in 2.6 kernel. This also fix some deadlocks
2847 and remove hack for work O_SYNC with 2.6 kernel.
2849 Severity : enhancement
2851 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2852 Details : The uninit_groups feature works in conjunction with the kernel
2853 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2854 up the pass1 processing of e2fsck. This is a read-only feature
2855 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2856 mount filesystems that have had this feature enabled.
2858 Severity : enhancement
2860 Description: Improve multi-block allocation algorithm to avoid fragmentation
2861 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2862 allocation locality and avoid filesystem fragmentation.
2864 ------------------------------------------------------------------------------
2866 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2868 * Support for kernels:
2869 2.6.9-42.0.3.EL (RHEL 4)
2870 2.6.5-7.276 (SLES 9)
2871 2.4.21-47.0.1.EL (RHEL 3)
2872 2.6.12.6 vanilla (kernel.org)
2873 2.6.16.21-0.8 (SLES10)
2874 * Recommended e2fsprogs version: 1.39.cfs2-0
2876 * The backwards-compatible /proc/sys/portals symlink has been removed
2877 in this release. Before upgrading, please ensure that you change
2878 any configuration scripts or /etc/sysctl.conf files that access
2879 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2880 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2881 in advance of the upgrade on any system running Lustre 1.4.6 or
2882 newer, since /proc/sys/lnet was added in that version.
2883 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2887 Frequency : only when quota is used
2889 Description: avoid scanning export list for quota master
2890 Details : Change the algorithms to avoid scanning export list in order
2891 to improve the efficiency.
2894 Frequency : MDS failover only, very rarely
2896 Description: "went back in time" messages on mds failover
2897 Details : The greatest transno may be lost when the current operation
2898 finishes with an error (transno==0) and the client's last_rcvd
2899 record is over-written. Save the greatest transno in the
2900 mds_last_transno for this case.
2903 Frequency : always for specific kernels and striping counts
2905 Description: client may get "Matching packet too big" without ACL support
2906 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2907 when trying to access files in certain configurations. The
2908 clients should in fact be denied when mounting because they do
2909 not understand ACLs.
2912 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2914 Description: many clients connecting with IO in progress causes connect timeouts
2915 Details : Avoid synchronous journal commits to avoid delays caused by many
2916 clients connecting/disconnecting when bulk IO is in progress.
2917 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2918 OST_IO_PORTAL to avoid delays behind potentially many pending
2922 Frequency : occasionally with multiple writers to a single file
2924 Description: shared writes to file may result in wrong size reported by stat()
2925 Details : Allow growing of kms when extent lock is cancelled
2928 Frequency : always with random mmap IO to multi-striped file
2930 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2931 Details : Only if the hole is at the end of OST object so that kms is too
2932 small. Fix is to increase kms accordingly in ll_nopage.
2935 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2937 Description: writes to a missing object would leak memory on the OST
2938 Details : If there is an inconsistency between the MDS and OST filesystems,
2939 such that the MDS references an object that doesn't exist, writes
2940 to that object will leak memory due to incorrect cleanup in the
2941 error handling path, eventually running out of memory on the OST.
2946 Description: Creating too long symlink causes lustre errors
2947 Details : Check symlink and name lengths before sending requests to MDS.
2950 Frequency : only if flock is enabled (not on by default)
2952 Description: posix locks not released on fd closure on 2.6.9+
2953 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2954 kernels, this caused such locks not to be released on fd close and
2955 then assertions on fs unmount about still used locks.
2958 Frequency : MDS failover only, very rarely
2960 Description: clients may get ASSERTION(granted_lock != NULL)
2961 Details : When request was taking a long time, and a client was resending
2962 a getattr by name lock request. The were multiple lock requests
2963 with the same client lock handle and
2964 mds_getattr_name->fixup_handle_for_resent_request found one of the
2965 lock handles but later failed with ASSERTION(granted_lock != NULL).
2970 Description: handle->h_buffer_credits > 0, assertion failure
2971 Details : h_buffer_credits is zero after truncate, causing assertion
2972 failure. This patch extends the transaction or creates a new
2976 Frequency : NFS re-export or patchless client
2977 Bugzilla : 11179, 10796
2978 Description: Crash on NFS re-export node (__d_move)
2979 Details : We do not want to hash the dentry if we don't have a lock.
2980 But if this dentry is later used in d_move, we'd hit uninitialised
2981 list head d_hash, so we just do this to init d_hash field but
2982 leave dentry unhashed.
2985 Frequency : NFS re-export or patchless client
2987 Description: NFS exports has problem with symbolic link
2988 Details : lustre client didn't properly install dentry when re-exported
2989 to NFS or running patchless client.
2992 Frequency : NFS re-export or patchless client
2994 Description: Various nfs/patchless fixes.
2995 Details : fixes reuse disconected alias for lookup process - this fixes
2996 warning "find_exported_dentry: npd != pd",
2997 fix permission error with open files at nfs.
2998 fix apply umask when do revalidate.
3001 Frequency : occasional
3003 Description: Crash on NFS re-export node
3004 Details : calling clear_page() on the wrong pointer triggered oops in
3005 generic_mapping_read().
3008 Frequency : rarely, using O_DIRECT IO
3010 Description: unaligned directio crashes client with LASSERT
3011 Details : check for unaligned buffers before trying any requests.
3014 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3016 Description: stale data returned from RAID cache
3017 Details : If only a small amount of IO is done to the RAID device before
3018 reading it again it is possible to get stale data from the RAID
3019 cache instead of reading it from disk.
3022 Frequency : always for sles10 kernel
3024 Description: sles10 support
3025 Details : ll_follow_link: compile fixes and using of nd_set_link
3026 under newer kernels.
3029 Frequency : depends on arch, kernel and compiler version, always on sles10
3032 Description: recursive or deep enough symlinks cause stack overflow
3033 Details : getting rid of large stack-allocated variable in
3037 Frequency : depends on hardware
3039 Description: lustre write performance loss in the SLES10 kernel
3040 Details : the performance loss is caused by using of write barriers in the
3041 ext3 code. The SLES10 kernel turns barrier support on by
3042 default. The fix is to undo that change for ldiskfs.
3044 ------------------------------------------------------------------------------
3046 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3048 * Support for kernels:
3049 2.6.9-42.0.3EL (RHEL 4)
3050 2.6.5-7.276 (SLES 9)
3051 2.4.21-47.0.1.EL (RHEL 3)
3052 2.6.12.6 vanilla (kernel.org)
3056 Frequency : quota enabled and large files being deleted
3058 Description: releasing more than 4GB of quota at once hangs OST
3059 Details : If a user deletes more than 4GB of files on a single OST it
3060 will cause the OST to spin in an infinite loop. Release
3061 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3066 Description: statfs data retrieved from /proc may be stale or zero
3067 Details : When reading per-device statfs data from /proc, in the
3068 {kbytes,files}_{total,free,avail} files, it may appear
3069 as zero or be out of date.
3072 Frequency : systems with MD RAID1 external journal devices
3074 Description: lconf's call to blkid is confused by RAID1 journal devices
3075 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3076 of returning all block devices that match the journal UUID.
3079 Frequency : always, for aggregate stripe size over 4GB
3081 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3082 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3083 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3086 Frequency : always if "lfs find" used on a local file/directory
3088 Description: "lfs find" segfaults if used on a local file/directory
3089 Details : The case where a directory component was not specified wasn't
3090 handled correctly. Handle this properly.
3093 Frequency : always on ppc64
3095 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3096 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3097 perform endianness conversion.
3100 Frequency : rarely (truncate to non-zero file size after write under load)
3101 Bugzilla : 10730, 10687
3102 Description: Files padded with zeros to next 4K multiple
3103 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3104 it is possible that files that are truncated to a non-zero size
3105 immediately after being written are filled with zero bytes beyond
3106 the truncated size. No file data is lost.
3108 Severity : enhancement
3110 Description: Allow recovery/failover for liblustre clients.
3111 Details : liblustre clients were unaware of failover configurations until
3114 Severity : enhancement
3116 Description: user file locks should fail when not mounting with flock option
3117 Details : Set up an error-returning stub in ll_file_operations.lock field
3118 to prevent incorrect behaviour when client is mounted without
3119 flock option. Also, set up properly f_op->flock field for
3123 Frequency : always on ia64
3125 Description: "lfs df" loops on printing out MDS statfs information
3126 Details : The obd_ioctl_data was not initialized and in some systems
3127 this caused a failure during the ioctl that did not return
3128 an error. Initialize the struct and return an error on failure.
3131 Frequency : SLES 9 only
3133 Description: Error of copying files with lustre special EAs as root
3134 Details : Client side always return success for setxattr call for lustre
3135 special xattr (currently only "trusted.lov").
3138 Frequency : rarely on clusters with both ia64+i386 clients
3140 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3141 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3142 clients are doing concurrent writes to the same file it is
3143 possible that the ia64 clients may LASSERT because the OST
3144 extent locks are not PAGE_SIZE aligned. Ensure that grown
3145 locks are always aligned on the request boundary.
3148 Frequency : specific use, occasional
3150 Description: Overwriting in use executable truncates on-disk binary image
3151 Details : If one node attempts to overwrite an executable in use by
3152 another node, we now correctly return ETXTBSY instead of
3153 truncating the file.
3155 Severity : enhancement
3157 Description: Async OSC create to avoid the blocking unnecessarily.
3158 Details : If a OST has no remain object, system will block on the creating
3159 when need to create a new object on this OST. Now, ways use
3160 pre-created objects when available, instead of blocking on an
3161 empty osc while others are not empty. If we must block, we block
3162 for the shortest possible period of time.
3167 Description: chmod on Lustre root is propagated to other clients
3168 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3169 invalid by the follow_mount time.
3172 Frequency : liblustre clients only
3174 Description: Race in 'instant cancel' lock handling could lead to such locks
3175 never to be granted in case of SMP MDS
3176 Details : Do not destroy not yet granted but cbpending locks in
3180 Frequency : replay/resend of open
3182 Description: non null lock assetion failure in mds_intent_policy
3183 Details : Trying to replay/resend lockless open requests resulted in
3184 mds_open() returning 0 with no lock. Now it sets a flag if
3185 a lock is going to be returned.
3187 Severity : enhancement
3189 Description: Checksum enhancements
3190 Details : New checksum enhancements allow for resending RPCs that failed
3193 Severity : enhancement
3195 Description: Tunables on number of dirty pages in cacche
3196 Details : Allow to set limit on number of dirty pages cached.
3201 Description: client crash on unmount - lock still has references
3202 Details : In some error handling cases it was possible to leak a lock
3203 reference on a client while accessing a file. This was not
3204 harmful to the client during operation, but would cause the
3205 client to crash when the filesystem is unmounted.
3208 Frequency : specific case, rare
3210 Description: ETXTBSY on mds though file not in use by client
3211 Details : ETXTBSY is no longer incorrectly returned when attempting to
3212 chmod or chown a directory that the user previously tried to
3213 execute or a currently-executing binary.
3216 Frequency : extremely rare except on liblustre-based clients
3218 Description: Lustre space not freed when files are deleted
3219 Details : Clean up open-unlinked files after client eviction. Previously
3220 the unlink was skipped and the files remained as orphans.
3225 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3226 Details : In some cases it was possible to send a blocking callback to
3227 a client doing a glimpse, even though that client didn't get
3228 a lock granted. When the glimpse lock is cancelled on the OST
3229 the freed lock is left on the waiting list and corrupted the list.
3232 Frequency : all core dumps
3234 Description: Broke core dumps to lustre
3235 Details : Negative dentry may be unhashed if parent does not have UPDATE
3236 lock, but some callers, e.g. do_coredump, expect dentry to be
3237 hashed after successful create, hash it in ll_create_it.
3239 ------------------------------------------------------------------------------
3241 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3243 * Support for kernels:
3244 2.6.9-42.0.2.EL (RHEL 4)
3245 2.6.5-7.276 (SLES 9)
3246 2.4.21-40.EL (RHEL 3)
3247 2.6.12.6 vanilla (kernel.org)
3251 Frequency : always on RHEL 3
3253 Description: Number of open files grows over time
3254 Details : The number of open files grows over time, whether or not
3255 Lustre is started. This was due to a filp leak introduced
3256 by one of our kernel patches.
3258 ------------------------------------------------------------------------------
3260 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3262 * Support for kernels:
3263 2.6.9-42.EL (RHEL 4)
3264 2.6.5-7.276 (SLES 9)
3265 2.4.21-40.EL (RHEL 3)
3266 2.6.12.6 vanilla (kernel.org)
3271 Bugzilla : 5719, 9635, 9792, 9684
3272 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3273 Details : If a server is under heavy load and cannot reply to new
3274 connection requests before the client resends the (re)connect,
3275 the connection handling code can behave badly if two service
3276 threads are concurrently handing separate (re)connections from
3277 the same client. Add better locking to the connection handling
3278 code, and ensure that only a single connection will be processed
3279 for a given client UUID, even if the lock is dropped.
3281 Severity : enhancement
3283 Description: add TCP zero-copy support to kernel
3284 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3285 sends if the hardware supports scatter-gather and checksumming.
3286 This allows socklnd to do client-write and server-read more
3287 efficiently and reduce CPU utilization from skbuf copying.
3290 Frequency : only if NFS exporting from client
3292 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3293 Details : If setting attributes on a file created under NFS that had
3294 never been opened it would be possible to oops the client
3295 if the file had no objects.
3298 Frequency : always for liblustre
3300 Description: liblustre client does MDS+OSTs setattr RPC for each write
3301 Details : When doing a write from a liblustre client, the client
3302 incorrectly issued an RPC to the MDS and each OST the file was
3303 striped over in order to update the timestamps. When writing
3304 with small chunks and many clients this could overwhelm the MDS
3305 with RPCs. In all cases it would slow down the write because
3306 these RPCs are unnecessary.
3308 Severity : enhancement
3310 Description: allow number of MDS service threads to be changed at module load
3311 Details : It is now possible to change the number of MDS service threads
3312 running. Adding "options mds mds_num_threads={N}" to the MDS's
3313 /etc/modprobe.conf will set the number of threads for the next
3314 time Lustre is restarted (assuming the "mds" module is also
3315 reloaded at that time). The default number of threads will
3316 stay the same, 32 for most systems.
3321 Description: OST crash if filesystem is unformatted or corrupt
3322 Details : If an OST is started on a device that has never been formatted
3323 or if the filesystem is corrupt and cannot even mount then the
3324 error handling cleanup routines would dereference a NULL pointer.
3329 Description: NULL pointer deref in llap_from_page.
3330 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3331 incorrect) mapping if the page was truncated/reclaimed while it was
3332 searched for. Check for this condition and skip such pages when
3333 doing readahead. Introduce extra check to llap_from_page() to
3334 verify page->mapping->host is non-NULL (so page is not anonymous).
3337 Frequency : Sometimes when using sys_sendfile
3339 Description: "page not covered by a lock" warnings from ll_readpage
3340 Details : sendfile called ll_readpage without right page locks present.
3341 Now we introduced ll_file_sendfile that does necessary locking
3342 around call to generic_file_sendfile() much like we do in
3346 Frequency : with certain MDS communication failures at client mount time
3348 Description: NULL pointer deref after failed client mount
3349 Details : a client connection request may delayed by the network layer
3350 and not be sent until after the PTLRPC layer has timed out the
3351 request. If the client fails the mount immediately it will try
3352 to clean up before the network times out the request. Add a
3353 reference from the request import to the obd device and delay
3354 the cleanup until the network drops the request.
3357 Frequency : occasionally during client (re)connect
3359 Description: assertion failure during client (re)connect
3360 Details : processing a client connection request may be delayed by the
3361 client or server longer than the client connect timeout. This
3362 causes the client to resend the connection request. If the
3363 original connection request is replied in this interval, the
3364 client may trip an assertion failure in ptlrpc_connect_interpret()
3365 which thought it would be the only running connect process.
3368 Frequency : only with obd_echo servers and clients that are rebooted
3370 Description: kernel BUG accessing uninitialized data structure
3371 Details : When running an obd_echo server it did not start the ping_evictor
3372 thread, and when a client was evicted an uninitialized data
3373 structure was accessed. Start the ping_evictor in the RPC
3374 service startup instead of the OBD startup.
3376 Severity : enhancement
3377 Bugzilla : 10193 (patchless)
3378 Description: Remove dependency on various unexported kernel interfaces.
3379 Details : No longer need reparent_to_init, exit_mm, exit_files,
3380 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3383 Frequency : rare (only users of deprecated and unsupported LDAP config)
3385 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3386 Details : LDAP apparently contains 'lustreName' attributes instead of
3387 'name'. A simple remapping of the name is sufficient.
3390 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3392 Description: waiting_locks_callback trips kernel BUG if client is evicted
3393 Details : Running with the dump_on_timeout debug flag turned on makes
3394 it possible that the waiting_locks_callback() can try to dump
3395 the Lustre kernel debug logs from an interrupt handler. Defer
3396 this log dumping to the expired_lock_main() thread.
3398 Severity : enhancement
3400 Description: Support NFS exporting on 2.6 kernels.
3401 Details : Implement non-rawops metadata methods for NFS server to use without
3402 changing NFS server code.
3405 Frequency : very rare (synthetic metadata workload only)
3407 Description: two racing renames might cause an MDS thread to deadlock
3408 Details : Running the "racer" program may cause one MDS thread to rename
3409 a file from being the source of a rename to being the target of
3410 a rename at exactly the same time that another thread is doing
3411 so, and the second thread has already enqueued these locks after
3412 doing a lookup of the target and is trying to relock them in
3413 order. Ensure that we don't try to re-lock the same resource.
3416 Frequency : only very large systems with liblustre clients
3418 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3419 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3420 all OSTs in parallel. This allows the eviction of stale liblustre
3421 clients to proceed much faster than if they were done in series,
3422 and also offers similar improvements for other set_info RPCs.
3427 Description: excessive CPU usage during initial read phase on client
3428 Details : During the initial read phase on a client, it would agressively
3429 retry readahead on the file, consuming too much CPU and impacting
3430 performance (since 1.4.5.8). Improve the readahead algorithm
3431 to avoid this, and also improve some other common cases (read
3432 of small files in particular, where "small" is files smaller than
3433 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3438 Description: MDS crash when receiving packet with unknown intent.
3439 Details : Do not LBUG in unknown intent case, just return -EFAULT
3441 Severity : enhancement
3442 Bugzilla : 9293, 9385
3443 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3444 Details : Do not serialize getattr (non-intent version) and statfs.
3447 Frequency : occasional, when OST network is overloaded/intermittent
3449 Description: client evicted by OST after bulk IO timeout
3450 Details : If a client sends a bulk IO request (read or write) the OST
3451 may evict the client if it is unresposive to its data GET/PUT
3452 request. This is incorrect if the network is overloaded (takes
3453 too long to transfer the RPC data) or dropped the OST GET/PUT
3454 request. There is no need to evict the client at all, since
3455 the pinger and/or lock callbacks will handle this, and the
3456 client can restart the bulk request.
3459 Frequency : Always when mmapping file with no objects
3461 Description: client crashes when mmapping file with no objects
3462 Details : Check that we actually have objects in a file before doing any
3463 operations on objects in ll_vm_open, ll_vm_close and
3469 Description: Request leak when working with deleted CWD
3470 Details : Introduce advanced request refcount tracking for requests
3471 referenced from lustre intent.
3473 Severity : Enhancement
3475 Description: Cache open file handles on client.
3476 Details : MDS now will return special lock along with openhandle, if
3477 requested and client is allowed to hold openhandle, even if unused,
3478 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3479 closing files for every read/write openration.
3481 Severity : Enhancement
3483 Description: Cache open negative dentries on client when possible.
3484 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3485 negative dentries on lock revocation.
3490 Description: Remounting a client read-only wasn't possible with a zconf mount
3491 Details : It wasn't possible to remount a client read-only with llmount.
3493 Severity : enhancement
3494 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3495 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3496 MPICH is included to add Lustre-specific ADIO interfaces.
3497 This is based closely on the UFS ADIO layer and only differs
3498 in file creation, in order to allow the OST striping to be set.
3499 This is user-contributed code and not supported by CFS.
3504 Description: extended inode attributes (immutable, append-only) work improperly
3505 when 2.4 and 2.6 kernels are used on client/server or vice versa
3506 Details : Introduce kernel-independent values for these flags.
3508 Severity : enhancement
3511 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3512 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3513 granularity. Now, Fractional MB granularity is be supported,
3514 this is very useful for low memory system.
3516 Severity : enhancement
3518 Description: Getattr by fid
3519 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3520 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3521 have locks and avoids getattr rpc requests for it on every stat.
3524 Frequency : Always, for filesystems larger than 2TB
3526 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3527 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3528 filesystems larger than 2TB to be mounted with mballoc enabled.
3531 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3533 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3534 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3535 filesystems larger than 2TB is mounted then the kernel will
3536 silently corrupt the start of the filesystem. CONFIG_LBD is
3537 enabled for all CFS-supported kernels, but the possibility of
3538 this happening with a modified kernel config exists.
3540 Severity : enhancement
3542 Description: add client O_DIRECT support for 2.6 kernels
3543 Details : It is now possible to do O_DIRECT reads and writes to files
3544 in the Lustre client mountpoint on 2.6 kernel clients.
3546 Severity : enhancement
3548 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3549 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3550 parallel, not waiting for response from every OST before sending
3551 a rpc to the next OST.
3556 Description: setattr vs write race when updating file timestamps
3557 Details : Client processes that update a file timestamp into the past
3558 right after writing to the file (e.g. tar) it is possible that
3559 the updated file modification time can be reset to the current
3560 time due to a race between processing the setattr and write RPC.
3562 Severity : enhancement
3564 Description: Bring 'lfs find' closer in line with regular Linux find.
3565 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3566 -print0 options and obtains all the needed info through the lustre
3569 Severity : enhancement
3571 Description: support up to 1024 configured devices on one node
3572 Details : change obd_dev array from statically allocated to dynamically
3573 allocated structs as they are first used to reduce memory usage
3578 Description: Flush dirty partially truncated pages during truncate
3579 Details : Immediatelly flush partially truncated pages in filter_setattr,
3580 this way we completely avoid having any pages in page cache on OST
3581 and can retire ugly workarounds during writes to flush such pages.
3586 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3587 Details : protect inode from truncation within vfs_unlink() context
3588 just take a reference before calling vfs_unlink() and release it
3589 when parent's i_sem is free.
3594 Description: last_id value checked outside lock on OST caused LASSERT failure
3595 Details : If there were multiple MDS->OST object precreate requests in
3596 flight, it was possible that the OST's last object id was checked
3597 outside a lock and incorrectly tripped an assertion. Move checks
3598 inside locks, and discard old precreate requests.
3601 Frequency : always, if extents are used on OSTs
3603 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3604 Details : OSTs using the extents format would not zero the high 16 bits of
3605 the index physical block number. This is not a problem for any
3606 OST filesystems smaller than 16TB, and no kernels support ext3
3607 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3608 new/modified files) and can be fixed for existing filesystems
3609 with e2fsprogs-1.39-cfs1.
3614 Description: import connection selection may be incorrect if timer wraps
3615 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3616 import connections to be ignored if the 32-bit jiffies counter
3617 wraps. Use a 64-bit jiffies counter.
3620 Frequency : very large clusters immediately after boot
3622 Description: LNET request buffers exhausted under heavy short-term load
3623 Details : If a large number of client requests are generated on a service
3624 that has previously never seen so many requests it is possible
3625 that the request buffer growth cannot keep up with the spike in
3626 demand. Instead of dropping incoming requests, they are held in
3627 the LND until the RPC service can accept more requests.
3630 Frequency : Sometimes during replay
3632 Description: Assertion failure in ll_local_open after replay.
3633 Details : If replay happened on an open request reply before we were able
3634 to set replay handler, reply will become not swabbed tripping the
3635 assertion in ll_local_open. Now we set the handler right after
3636 recognising of open request
3639 Frequency : very rare
3641 Description: kernel reports "badness in vsnprintf"
3642 Details : Reading from the "recovery_status" /proc file in small chunks
3643 may cause a negative length in lprocfs_obd_rd_recovery_status()
3644 call to vsnprintf() (which is otherwise harmless). Exit early
3645 if there is no more space in the output buffer.
3647 Severity : enhancement
3649 Description: clear OBD RPC statistics by writing to them
3650 Details : It is now possible to clear the OBD RPC statistics by writing
3651 to the "stats" file.
3656 Description: Client mtime is not the same on different clients after utimes
3657 Details : In some cases, the client was using the utimes() syscall on
3658 a file cached on another node. The clients now validate the
3659 ctime from the MDS + OSTs to determine which one is right.
3664 Description: Inability to activate failout mode
3665 Details : lconf script incorrectly assumed that in pythong string's numeric
3666 value is used in comparisons.
3669 Frequency : always with multiple stripes per file
3671 Description: Inefficient object allocation for mutli-stripe files
3672 Details : When selecting which OSTs to stripe files over, for files with
3673 a stripe count that divides evenly into the number of OSTs,
3674 the MDS is always picking the same starting OST for each file.
3675 Return the OST selection heuristic to the original design.
3680 Description: mount failures may take full timeout to return an error
3681 Details : Under some heavy load conditions it is possible that a
3682 failed mount can wait for the full obd_timeout interval,
3683 possibly several minutes, before reporting an error.
3684 Instead return an error as soon as the status is known.
3686 Frequency : quota enabled and large files being deleted
3688 Description: releasing more than 4GB of quota at once hangs OST
3689 Details : If a user deletes more than 4GB of files on a single OST it
3690 will cause the OST to spin in an infinite loop. Release
3691 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3696 Description: statfs data retrieved from /proc may be stale or zero
3697 Details : When reading per-device statfs data from /proc, in the
3698 {kbytes,files}_{total,free,avail} files, it may appear
3699 as zero or be out of date.
3702 Frequency : systems with MD RAID1 external journal devices
3704 Description: lconf's call to blkid is confused by RAID1 journal devices
3705 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3706 of returning all block devices that match the journal UUID.
3709 Frequency : always, for aggregate stripe size over 4GB
3711 Description: assertion fails when trying to use 4GB stripe size
3712 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3713 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3716 Frequency : always on ppc64
3718 Description: the first write on an ext3 filesystem with mballoc got stuck
3719 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3720 perform endianness conversion.
3722 ------------------------------------------------------------------------------
3724 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3726 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3727 INTEROPERATE with older versions automatically. Please read the
3728 user documentation before upgrading any part of a live system.
3729 * WARNING: Lustre networking configuration changes are required with
3730 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3733 * Support for kernels:
3734 2.6.9-22.0.2.EL (RHEL 4)
3735 2.6.5-7.244 (SLES 9)
3736 2.6.12.6 vanilla (kernel.org)
3739 Severity : enhancement
3740 Bugzilla : 7981/8208
3741 Description: Introduced Lustre Networking (LNET)
3742 Details : LNET is new networking infrastructure for Lustre, it includes
3743 a reorganized network configuration mode (see the user
3744 documentation for full details) as well as support for routing
3745 between different network fabrics. Lustre Networking Devices
3746 (LNDs) for the supported network fabrics have also been
3747 created for this new infrastructure.
3749 Severity : enhancement
3750 Description: Introduced Access control lists
3751 Details : clients can set ACLs on files and directories in order to have
3752 more fine-grained permissions than the standard Unix UGO+RWX.
3753 The MDS must be started with the "-o acl" mount option.
3755 Severity : enhancement
3756 Description: Introduced filesystem quotas
3757 Details : Administrators may now establish per-user quotas on the
3760 Severity : enhancement
3762 Description: Configuration change for the XT3
3763 The PTLLND is now used to run Lustre over Portals on the XT3
3764 The configure option(s) --with-cray-portals are no longer used.
3765 Rather --with-portals=<path-to-portals-includes> is used to
3766 enable building on the XT3. In addition to enable XT3 specific
3767 features the option --enable-cray-xt3 must be used.
3772 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3773 Details : A race between allocating a new llcd and re-getting the llcd_lock
3774 allowed another thread to grab newly-allocated llcd.
3776 Severity : enhancement
3778 Description: 2.6 OST async journal commit and locking fix to improve performance
3779 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3780 2.6 kernels are now async as they already were in 2.4 kernels so
3781 that they can commit concurrently with the network bulk transfer.
3782 For block-allocated files the filter allocation semaphore is held
3783 to avoid filesystem fragmentation during allocation. BKL lock
3784 removed for 2.6 xattr operations where it is no longer needed.
3789 Description: lconf incorrectly determined whether two IP networks could talk
3790 Details : In some more complicated routing and multiple-network
3791 configurations, lconf will avoid trying to make a network
3792 connection to a disjoint part of the IP space. It was doing the
3793 math incorrectly for one set of cases.
3798 Description: Fix for potential infinite loop processing records in an llog.
3799 Details : If an llog record is corrupted/zeroed, it is possible to loop
3800 forever in llog_process(). Validate the llog record length
3801 and skip the remainder of the block on error.
3804 Frequency : occasional (liblustre only)
3806 Description: liblustre could not open files whose last component is a symlink
3807 Details : sysio_path_walk() would incorrectly pass the open intent to
3808 intermediate path components.
3811 Frequency : rare (liblustre only with non-standard tuning)
3812 Bugzilla : 7201 (7350)
3813 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3814 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3815 after it has been released, only do this for referenced locks
3817 Severity : enhancement
3819 Description: specifying an (invalid) directory default stripe_size of -1
3820 would reset the directory default striping
3821 Details : stripe_size -1 was used internally to signal directory stripe
3822 removal, now use "all default" to signal dir stripe removal
3823 as a directory striping of "all default" is not useful
3826 Frequency : common for large clusters running liblustre clients
3828 Description: doing an ls when liblustre clients are running is slow
3829 Details : sending a glimpse AST to a liblustre client waits for every AST
3830 to time out, as liblustre clients will not respond. Since they
3831 cannot cache data we refresh the OST lock LVB from disk instead.
3833 Severity : enhancement
3835 Description: doing an ls at the same time as file IO can be slow
3836 Details : enqueue and other "small" requests can be blocked behind many
3837 large IO requests. Create a new OST IO portal for non-IO
3838 requests so they can be processed faster.
3841 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3843 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3844 created file with mode 000
3845 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3846 allow an ftruncate() as a non-root user to a file with mode 000.
3847 Lustre now respects this flag to disable mode checks when
3848 truncating a file owned by the user
3851 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3854 Description: Revoking locks from clients that went dead or catatonic might take
3856 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3857 cancellation of such locks instant on servers without waiting for
3858 any reply from clients. Clients drops these locks when cancel
3859 notification from server is received without replying.
3862 Frequency : liblustre-only, when liblustre client dies or becomes busy
3864 Description: Doing ls on Linux clients can take a long time with active
3866 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3867 granting such locks to it in the first place if possible. Locks
3868 are taken by proxy on the OST during the read or write and
3869 dropped immediately afterward. Add connect flags handling, do
3870 not grant locks to liblustre clients for glimpse ASTs.
3872 Severity : enhancement
3874 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3875 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3876 sequential reads and behaves badly with applications doing random
3877 reads. Improve it by reading ahead at least read region, and
3878 avoiding excessive large RPC for small reads.
3880 Severity : enhancement
3882 Description: Creating more than 1000 files for a single job may cause a load
3883 imbalance on the OSTs if there are also a large number of OSTs.
3884 Details : qos_prep_create() uses an OST index reseed value that is an
3885 even multiple of the number of available OSTs so that if the
3886 reseed happens in the middle of the object allocation it will
3887 still utilize the OSTs as uniformly as possible.
3892 Description: OST or MDS may oops in ping_evictor_main()
3893 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3894 but doesn't restart at beginning of obd_exports_timed list
3897 Severity : enhancement
3899 Description: improve by-nid export eviction on the MDS and OST
3900 Details : allow multiple exports with the same NID to be evicted at one
3901 time without re-searching the exports list.
3904 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3906 Description: MDS may oops in groups_free()
3907 Details : in rare race conditions a newly allocated group_info struct is
3908 freed again, and this can be NULL. The 2.4 compatibility code
3909 for groups_free() checked for a NULL pointer, but 2.6 did not.
3912 Frequency : common for liblustre clients doing little filesystem IO
3913 Bugzilla : 9352, 7313
3914 Description: server may evict liblustre clients accessing contended locks
3915 Details : if a client is granted a lock or receives a completion AST
3916 with a blocking AST already set it would not reply to the AST
3917 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3920 Frequency : lfs setstripe, only systems with more than 160 OSTs
3922 Description: unable to set striping with a starting offset beyond OST 160
3923 Details : llapi_create_file() incorrectly limited the starting stripe
3924 index to the maximum single-file stripe count.
3927 Frequency : LDAP users only
3929 Description: lconf did not handle in-kernel recovery with LDAP properly
3930 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3932 Severity : enhancement
3934 Description: bind OST threads to NUMA nodes to improve performance
3935 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3936 node and do their allocations there to localize memory access
3938 Severity : enhancement
3940 Description: llmount can determine client NID directly from Myrinet (GM)
3941 Details : the client NID code from gmnalnid was moved directly into
3942 llmount, removing the need to use this or specifying the
3943 client NID explicitly when mounting GM clients with zeroconf
3946 Frequency : if client is started with down MDS
3948 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3949 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3950 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3951 ptlrpc_import_delayed_req() because we didn't check if the
3952 request was interrupted, and we also didn't break out of the
3953 event loop if there was no timeout
3958 Description: data loss during non-page-aligned writes to a single file from
3959 both multiple nodes and multiple threads on one node at same time
3960 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3961 inconsistency led to false short-reads, that were cached and later
3962 used by ->prepare_write() to fill in partially written page,
3963 leading to data loss.
3966 Frequency : always, if lconf --abort_recovery used
3968 Description: lconf --abort_recovery fails with 'Operation not supported'
3969 Details : lconf was attempting to abort recovery on the MDT device and not
3972 Severity : enhancement
3974 Description: remove cleanup logs
3975 Details : replace lconf-generated cleanup logs with lustre internal
3976 cleanup routines. Eliminates the need for client-cleanup and
3979 Severity : enhancement
3981 Description: add support for EAs (user and system) on lustre filesystems
3982 Details : it is now possible to store extended attributes in the Lustre
3983 client filesystem, and with the user_xattr mount option it
3984 is possible to allow users to store EAs on their files also
3986 Severity : enhancement
3988 Description: Add possibility (config option) to show minimal available OST free
3990 Details : When compiled with --enable-mindf configure option, statfs(2)
3991 (and so, df) will return least minimal free space available from
3992 all OSTs as amount of free space on FS, instead of summary of
3993 free spaces of all OSTs.
3995 Severity : enhancement
3997 Description: do not expand extent locks acquired on OST-side
3998 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3999 by server: they are not cached anyway.
4002 Frequency : when mmap is used/binaries executed from Lustre
4004 Description: Unmmap pages before throwing them away from read cache.
4005 Details : llap_shrink cache now attempts to unmap pages before discarding
4006 them (if unmapping failed - do not discard). SLES9 kernel has
4007 extra checks that trigger if this unmapping is not done first.
4012 Description: lconf didn't resolve symlinks before checking to see whether a
4013 given mountpoint was already in use
4016 Frequency : when migrating failover services
4017 Bugzilla : 6395, 9514
4018 Description: When migrating a subset of services from a node (e.g. failback
4019 from a failover service node) the remaining services would
4020 time out and evict clients.
4021 Details : lconf --force (implied by --failover) sets the global obd_timeout
4022 to 5 seconds in order to quickly disconnect, but this caused
4023 other RPCs to time out too quickly. Do not change the global
4024 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4026 Severity : enhancement
4027 Frequency : if MDS is started with down OST
4028 Bugzilla : 9439,5706
4029 Description: Allow startup/shutdown of an MDS without depending on the
4030 availability of the OSTs.
4031 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4032 Add appropriate locking and lov-osc refcounts for safe
4033 cleaning. Add osc abort_inflight calls in case the
4034 synchronize never started.
4037 Frequency : occasional (Cray XT3 only)
4039 Description: root not authorized to access files in CRAY_PORTALS environment
4040 Details : The client process capabilities were not honoured on the MDS in
4041 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4042 been accessed by an authorized user then root was able to access
4043 the file on the local client also. The root user capabilities
4044 are now allowed on the MDS, as this environment has secure UID.
4047 Frequency : occasional
4049 Description: ldiskfs "too long searching" message happens too often
4050 Details : A debugging message (otherwise harmless) prints too often on
4051 the OST console. This has been reduced to only happen when
4052 there are fragmentation problems on the filesystem.
4057 Description: Division by zero in statfs when all OSCs are inactive
4058 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4059 lov_statfs divides by value returned by lov_get_stripecnt().
4063 Bugzilla : 9489, 3273
4064 Description: First write from each client to each OST was only 4kB in size,
4065 to initialize client writeback cache, which caused sub-optimal
4066 RPCs and poor layout on disk for the first writen file.
4067 Details : Clients now request an initial cache grant at (re)connect time
4068 and so that they can start streaming writes to the cache right
4069 away and always do full-sized RPCs if there is enough data.
4070 If the OST is rebooted the client also re-establishes its grant
4071 so that client cached writes will be honoured under the grant.
4076 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4077 Details : Now I/O RPCs go to different portal number and (presumably) fast
4078 lock requests (and glimses) and other RPCs get their own service
4079 threads pool that should be able to service those RPCs
4082 Severity : enhancement
4084 Description: Ability to exchange lustre version between client and servers and
4085 issue warnings at client side if client is too old. Also for
4086 liblustre clients there is ability to refuse connection of too old
4088 Details : New 'version' field is added to connect data structure that is
4089 filled with version info. That info is later checked by server and
4093 Frequency : rare, liblustre only.
4094 Bugzilla : 9296, 9581
4095 Description: Two simultaneous writes from liblustre at offset within same page
4096 might proceed at the same time overwriting eachother with stale
4098 Details : I/O lock withing llu_file_prwv was released too early, before data
4099 actually was hitting the wire. Extended lock-holding time until
4100 server acknowledges receiving data.
4103 Frequency : extremely rare. Never observed in practice.
4105 Description: avoid generating lustre_handle cookie of 0.
4106 Details : class_handle_hash() generates handle cookies by incrementing
4107 global counter, and can hit 0 occasionaly (this is unlikely, but
4108 not impossible, because initial value of cookie counter is
4109 selected randonly). Value of 0 is used as a sentinel meaning
4110 "unassigned handle" --- avoid it. Also coalesce two critical
4111 sections in this function into one.
4113 Severity : enhancement
4115 Description: allow liblustre clients to delegate truncate locking to OST
4116 Details : To avoid overhead of locking, liblustre client instructs OST to
4117 take extent lock in ost_punch() on client's behalf. New connection
4118 flag is added to handle backward compatibility.
4120 Severity : enhancement
4121 Bugzilla : 4928, 7341, 9758
4122 Description: allow number of OST service threads to be specified
4123 Details : a module parameter allows the number of OST service threads
4124 to be specified via "options ost ost_num_threads={N}" in the
4125 OSS's /etc/modules.conf or /etc/modprobe.conf.
4129 Bugzilla : 6146, 9635, 9895
4130 Description: servers crash with bad pointer in target_handle_connect()
4131 Details : In rare cases when a client is reconnecting it was possible that
4132 the connection request was the last reference for that export.
4133 We would temporarily drop the export reference and get a new
4134 one, but this may have been the last reference and the export
4135 was just destroyed. Get new reference before dropping old one.
4137 Severity : enhancement
4138 Frequency : if client is started with failover MDS
4140 Description: Allow multiple MDS hostnames in the mount command
4141 Details : Try to read the configuration from all specified MDS
4142 hostnames during a client mount in case the "primary"
4145 Severity : enhancement
4147 Description: Stop sending data to evicted clients as soon as possible.
4148 Details : Check if the client we are about to send or are sending data to
4149 was evicted already. (Check is done every second of waiting,
4150 for which l_wait_event interface was extended to allow checking
4151 of exit condition at specified intervals).
4154 Frequency : rare, normally only when NFS exporting is done from client
4156 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4158 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4161 Frequency : always, if srand() is called before liblustre initialization
4163 Description: Liblustre uses system PRNG disturbing its usage by user application
4164 Details : Introduce internal to lustre fast and high-quality PRNG for
4165 lustre usage and make liblustre and some other places in generic
4166 lustre code to use it.
4168 Severity : enhancement
4169 Bugzilla : 9477, 9557, 9870
4170 Description: Verify that the MDS configuration logs are updated when xml is
4171 Details : Check if the .xml configuration logs are newer than the config
4172 logs stored on the MDS and report an error if this is the case.
4173 Request --write-conf, or allow starting with --old_conf.
4175 Severity : enhancement
4177 Description: Handle symlinks in the path when checking if Lustre is mounted.
4178 Details : Resolve intermediate symlinks when checking if a client has
4179 mounted a filesystem to avoid duplicate client mounts.
4184 Description: lconf can hit an error exception but still return success.
4185 Details : The lconf command catches the Command error exception at the top
4186 level script context and will exit with the associated exit
4187 status, but doesn't ensure that this exit status is non-zero.
4192 Description: failure of ptlrpc thread startup can cause oops
4193 Details : Starting a ptlrpc service thread can fail if there are a large
4194 number of threads or the server memory is very fragmented.
4195 Handle this without oopsing.
4198 Frequency : always, only if liblustre and non-default acceptor port was used
4200 Description: liblustre cannot connect to servers with non-default acceptor port
4201 Details : tcpnal_set_default_params() was not called and was therefore
4202 ignoring the environment varaible TCPNAL_PORT, as well as other
4203 TCPNAL_ environment variables
4208 Description: two objects could be created on the same OST for a single file
4209 Details : If an OST is down, in some cases it was possible to create two
4210 objects on a single OST for a single file. No problems other
4211 than potential performance impact and spurious error messages.
4215 Bugzilla : 5681, 9562
4216 Description: Client may oops in ll_unhash_aliases
4217 Details : Client dcache may become inconsistent in race condition.
4218 In some cases "getcwd" can fail if the current directory is
4224 Description: Inode refcounting problems in NFS export code
4225 Details : link_raw functions used to call d_instantiate without obtaining
4226 extra inode reference first.
4230 Bugzilla : 9942, 9903
4231 Description: Referencing freed requests leading to crash, memleaks with NFS.
4232 Details : We used to require that call to ll_revalidate_it was always
4233 followed by ll_lookup_it. Also with revalidate_special() it is
4234 possible to call ll_revalidate_it() twice for the same dentry
4235 even if first occurence returned success. This fix changes semantic
4236 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4237 reference on a request referred from the intent.
4238 ll_intent_release() then releases such a request.
4241 Frequency : rare, normally benchmark loads only
4243 Description: unlinked inodes were kept in memory on the client
4244 Details : If a client is repeatedly creating and unlinking files it
4245 can accumulate a lot of stale inodes in the inode slab cache.
4246 If there is no other client load running this can cause the
4247 client node to run out of memory. Instead flush old inodes
4248 from client cache that have the same inode number as a new inode.
4251 Frequency : SLES9 2.6.5 kernel and long filenames only
4252 Bugzilla : 9969, 10379
4253 Description: utime reports stale NFS file handle
4254 Details : SLES9 uses out-of-dentry names in some cases, which confused
4255 the lustre dentry revalidation. Change it to always use the
4259 Frequency : rare, unless heavy write-truncate concurrency is continuous
4260 Bugzilla : 4180, 6984, 7171, 9963, 9331
4261 Description: OST becomes very slow and/or deadlocked during object unlink
4262 Details : filter_destroy() was holding onto the parent directory lock
4263 while truncating+unlinking objects. For very large objects this
4264 may block other threads for a long time and slow overall OST
4265 responsiveness. It may also be possible to get a lock ordering
4266 deadlock in this case, or run out of journal credits because of
4267 the combined truncate+unlink. Solution is to do object truncate
4268 first in one transaction without parent lock, and then do the
4269 final unlink in a new transaction with the parent lock. This
4270 reduces the lock hold time dramatically.
4273 Frequency : rare, 2.4 kernels only
4275 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4276 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4277 before unmounting their filesystem in order to avoid deadlock.
4278 The kernel_locked() function in 2.4 kernels only checks whether
4279 the kernel lock is held, not whether it is this process that is
4280 holding it as 2.6 kernels do.
4285 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4286 Details : The client ptlrpc code may be trying to reconnect to a down
4287 server before a previous connection attempt has timed out.
4288 Increase the reconnect interval to be longer than the connection
4289 timeout interval to avoid sending duplicate connections to
4293 Frequency : echo_client brw_test command
4295 Description: fix echo_client to work with OST preallocated code
4296 Details : OST preallocation code (5137) didn't take echo_client IO path
4297 into account: echo_client calls filter methods outside of any
4298 OST thread and, hence, there is no per-thread preallocated
4299 pages and buffers to use. Solution: hijack pga pages for IO. As
4300 a byproduct, this avoids unnecessary data copying.
4304 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4305 Description: Client can oops in mdc_commit_close() after open replay
4306 Details : It was possible for the MDS to return an open request with no
4307 transaction number in mds_finish_transno() if the client was
4308 evicted, but without actually returning an error. Clients
4309 would later try to replay that open and may trip an assertion
4310 Simplify the client close codepath, and always return an error
4311 from the MDS in case the open is not successful.
4314 Frequency : rare, 2.6 OSTs only
4316 Description: OST may deadlock under high load on fragmented files
4317 Details : If there was a heavy load and highly-fragmented OST filesystems
4318 it was possible to have all the OST threads deadlock waiting on
4319 allocation of biovecs, because the biovecs were not released
4320 until the entire RPC IO was completed. Instead, release biovecs
4321 as soon as they are complete to ensure forward IO progress.
4323 Severity : enhancement
4325 Description: Support for specifying external journal device at mount
4326 Details : If an OST or MDS device is formatted with an external journal
4327 device, this device major/minor is stored in the ext3 superblock
4328 and may not be valid for failover. Allow detecting and
4329 specifying the external journal at mount time.
4334 Description: Mounting an MDS with pending unlinked files may cause oops
4335 Details : target_finish_recovery() calls mds_postrecov() which returned
4336 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4337 considers this an error and immediately begins cleaning up the
4338 lov, just after starting the mds_lov process
4340 Severity : enhancement
4342 Description: Implement 'lfs df' to report actual free space on per-OST basis
4343 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4344 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4345 usage of objects; '-h' to report in human readable format.
4347 ------------------------------------------------------------------------------
4349 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4356 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4357 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4358 at startup during orphan unlinking. Instead, skip buddy update
4359 before setup, as it will be regenerated after recovery is complete.
4364 Description: If an OST is inactive, its locks might reference stale inodes.
4365 Details : lov_change_cbdata() must iterate over all namespaces, even if
4366 they are inactive to clear inode references from the lock.
4368 Severity : enhancement
4369 Frequency : occasional, if non-standard max_dirty_mb used
4371 Description: Client will block write RPCs if not enough grant
4372 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4373 then the client will block writes while waiting for another RPC
4374 to complete instead of consuming its dirty limit. With change
4375 we get improved performance when max_dirty_mb is small.
4377 Severity : enhancement
4378 Bugzilla : 3389, 6253
4379 Description: Add support for supplementary groups on the MDS.
4380 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4381 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4382 lookups for user supplementary groups into a cache.
4386 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4387 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4388 for liblustre O_EXCL is enforced by MDS.
4392 Description: readdir never returns NULL in liblustre.
4393 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4394 dentry in struct dirent.
4398 Frequency : liblustre only, depends on application IO pattern
4399 Description: liblustre clients evicted if not contacting servers
4400 Details : Don't put liblustre clients into the ping_evictor list, so
4401 they will not be evicted by the pinger ever.
4403 Severity : enhancement
4405 Description: Add ability to evict clients by NID from MDS.
4406 Details : By echoing "nid:$NID" string into
4407 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4408 $NID would be instantly evicted from this MDS and from all active
4409 OSTs connected to it.
4413 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4414 Details : lookup_it_finish() used to query file size from OSTs that was not
4419 Description: service threads change working directory to that of init
4420 Details : Starting lustre service threads may pin the working directory
4421 of the parent thread, making that filesystem busy. Threads
4422 now change to the working directory of init to avoid this.
4426 Frequency : during shutdown only
4427 Description: shutdown with a failed MDS or OST can cause unmount to hang
4428 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4433 Frequency : 2.6 only
4434 Description: chmod/chown may include an extra supplementary group
4435 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4436 supplementary group and if none is specified this is used.
4439 Bugzilla : 5479 (6816)
4441 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4442 Details : If lookup is in progress on a file that is unlinked we might try
4443 to revalidate the inode and fail in revalidate after lookup is
4444 complete and ll_file_open() enqueues the open again but
4445 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4448 Frequency : always, if lconf --abort_recovery used
4450 Description: lconf --abort_recovery fails with 'Operation not supported'
4451 Details : lconf was attempting to abort recovery on the MDT device and not
4454 ------------------------------------------------------------------------------
4456 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4461 Frequency : rare (only unsupported configurations with a node running as an
4463 Bugzilla : 6514, 5137
4464 Description: Mounting a Lustre file system on a node running as an OST could
4466 Details : OSTs now preallocates memory needed to write out data at
4467 startup, instead of when needed, to avoid having to
4468 allocate memory in possibly low memory situations.
4469 Specifically, if the file system is mounted on on OST,
4470 memory pressure could force it to try to write out data,
4471 which it needed to allocate memory to do. Due to the low
4472 memory, it would be unable to do so and the node would
4473 become unresponsive.
4475 Severity : enhancement
4477 Description: Addition of lconf --service command line option
4478 Details : lconf now accepts a '--service <arg>' option, which is
4479 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4481 Severity : enhancement
4483 Description: Failover mode is now the default for OSTs.
4484 Details : By default, OSTs will now run in failover mode. To return to
4485 the old behaviour, add '--failout' to the lmc line for OSTs.
4487 Severity : enhancement
4489 Description: Health checks are now provided for MDS and OSTs
4490 Details : Additional detailed health check information on MSD and OSTs
4491 is now provided through the procfs health_check value.
4494 Frequency : occasional, depends on IO load
4496 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4497 after numerous create/delete cycles
4498 Details : The ext3 inode allocation policy would not allocate new inodes
4499 very well on the OSTs because there are no new directories
4500 being created. Instead we look for groups with free space if
4501 the parent directories are nearly full.
4506 Description: Network or server problems during mount may cause partially
4507 mounted clients instead of returning an error.
4508 Details : The config llog parsing code may overwrite the error return
4509 code during mount error handling, returning success instead
4515 Description: MDS can fail to allocate large reply buffers
4516 Details : After long uptimes the MDS can fail to allocate large reply
4517 buffers (e.g. zconf client mount config records) due to memory
4518 fragmentation or consumption by the buffer cache. Preallocate
4519 some large reply buffers so that these replies can be sent even
4520 under memory pressure.
4524 Frequency : rare (liblustre)
4525 Description: fsx running with liblustre complained that using truncate() to
4526 extend the file doesn't work. This patch corrects that issue.
4527 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4528 ATTR_SIZE and lsm use in llu_setattr_raw.
4532 Frequency : rare, only 2.6 kernels
4533 Description: Unusual file access patterns on the MDS may result in inode
4534 data being lost in very rare circumstances.
4535 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4536 code in the 2.6 kernel caused the inode and/or EA data not to
4537 be read from disk, causing single-file corruption.
4541 Frequency : rare, only 2.6 filesystems using extents
4542 Description: Heavy concurrent write and delete load may cause data corruption.
4543 Details : It was possible under high-load situations to have an extent
4544 metadata block in the block device cache from a just-unlinked
4545 file overwrite a newly-allocated data block. We now unmap any
4546 metadata buffers that alias just-allocated data blocks.
4550 Frequency : filesystems with default stripe_count larger than 77
4551 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4552 Details : lconf specifies an inode size of 4096 bytes when the default
4553 stripe_count is larger than 77. This conflicts with the default
4554 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4555 this case to avoid pinning too much memory for large EAs.
4557 ------------------------------------------------------------------------------
4559 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4564 Frequency : rare (extremely heavy IO load with hundreds of clients)
4566 Description: Client is evicted, gets IO error writing to file
4567 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4568 caused clients to be evicted for AST timeouts. The fixes in
4569 bug 5192 mean we no longer need to have such short AST timeouts
4570 so ldlm_timeout has been increased.
4573 Frequency : occasional during --force or --failover shutdown under load
4574 Bugzilla : 5949, 4834
4575 Description: Server oops/LBUG if stopped with --force or --failover under load
4576 Details : a collection of import/export refcount and cleanup ordering
4577 issues fixed for safer force cleanup
4580 Frequency : only filesystems larger than 120 OSTs
4581 Bugzilla : 5990, 6223
4582 Description: lfs getstripe would oops on a very large filesystem
4583 Details : lov_getconfig used kfree on vmalloc'd memory
4586 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4587 Bugzilla : 6242, 6243
4588 Description: reading from files that had been truncated to a non-zero size
4589 but never opened returned no data
4590 Details : ll_file_read() reads zeros from no-object files to EOF
4595 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4596 Details : There's little harm in aborting MDS/OSS recovery and letting it
4597 try again, so I removed the LASSERT and return an error instead.
4599 Severity : enhancement
4601 Description: New debugging infrastructure for tracking down data corruption
4602 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4603 (b) cover more of the client-side code path, and (c) try to narrow
4604 down where problems occurred
4608 Bugzilla : 3819, 4364, 4397, 6313
4609 Description: Racing close and eviction MDS could cause assertion in mds_close
4610 Details : It was possible to get multiple mfd references during close and
4611 client eviction, leading to one thread referencing a freed mfd.
4613 Severity: : enhancement
4614 Bugzilla : 3262, 6359
4615 Description: Attempts to reconnect to servers are now more aggressive.
4616 Details : This builds on the enhanced upcall-less recovery that was added
4617 in 1.4.2. When trying to reconnect to servers, clients will
4618 now try each server in the failover group every 10 seconds. By
4619 default, clients would previously try one server every 25 seconds.
4624 Description: After recovery, certain operations trigger a failed
4625 assertion on a client.
4626 Details : Failing over an mds, using lconf -d --failover, while a
4627 client was doing a readdir() call would cause the client to
4628 LBUG after recovery completed and the readdir() was resent.
4630 Severity : enhancement
4632 Description: Default groups are now added by lconf
4633 Details : You can now run lconf --group <servicename> without having to
4634 manually add groups with lmc.
4637 Frequency : occasional
4639 Description: Nodes with an elan id of 0 trigger a failed assertion
4642 Frequency : always when accessing e.g. tty/console device nodes
4644 Description: tty and some other devices nodes cannot be used on lustre
4645 Details : file's private_data field is used by device data and lustre
4646 values in there got lost. New field was added to struct file to
4647 store fs-specific private data.
4650 Frequency : when exporting Lustre via NFS
4652 Description: NFSD failed occasionally when looking up a path component
4653 Details : NFSD is looking up ".." which was broken in ext3 directories
4654 that had grown large enough to become hashed.
4657 Frequency : Clusters with multiple interfaces not on the same subnet
4659 Description: Nodes will repeatedly try to reconnect to an interface which it
4660 cannot reach and report an error to the log.
4661 Details : Extra peer list entries will be created by lconf with some peers
4662 unreachable. lconf now validates the peer before adding it.
4665 Frequency : Only if a default stripe is set on the filesystem root.
4667 Description: Setting a default stripe on the filesystem root prevented the
4668 filesystem from being remounted.
4669 Details : The client was sending extra request flags in the root getattr
4670 request and did not allocate a reply buffer for the dir EA.
4673 Frequency : occasional, higher if lots of files are accessed by one client
4674 Bugzilla : 6159, 6097
4675 Description: Client trips assertion regarding lsm mismatch/magic
4676 Details : While revalidating inodes the VFS looks up inodes with ifind()
4677 and in rare cases can find an inode that is being freed.
4678 The ll_test_inode() code will free the lsm during ifind()
4679 when it finds an existing inode and then the VFS later attaches
4680 this free lsm to a new inode.
4684 Bugzilla : 6422, 7030
4685 Description: MDS deadlock between mkdir and client eviction
4686 Details : Creating a new file via mkdir or mknod (starting a transaction
4687 and getting the ns lock) can deadlock with client eviction
4688 (gets ns lock and trying to finish a synchronous transaction).
4691 Frequency : occasional
4692 Description: While starting a server, the fsfilt_ext3 module could not be
4694 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4695 kernels. Previously, lconf would still use the ext3 name
4696 when trying to load modules. Now, it will correctly use
4697 ext3 on 2.4 and ldiskfs on 2.6.
4699 Severity : enhancement
4700 Description: The default stripe count has been changed to 1
4701 Details : The interpretation of the default stripe count (0, to lfs
4702 or lmc) has been changed to mean striping across a single
4703 OST, rather than all available. For general usage we have
4704 found a stripe count of 1 or 2 works best.
4706 Severity : enhancement
4707 Description: Add support for compiling against Cray portals.
4708 Details : Conditional compiling for some areas that are different
4712 Frequency : occasional
4713 Bugzilla : 6409, 6834
4714 Description: Creating files with an explicit stripe count may lead to
4715 a failed assertion on the MDS
4716 Details : If some OSTs are full or unavailable, creating files may
4717 trigger a failed assertion on the MDS. Now, Lustre will
4718 try to use other servers or return an error to the
4722 Frequency : occasional
4724 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4725 caused lock timeout during readahead (since 1.4.2).
4726 Details : Processes doing readahead might match a lock that hasn't been
4727 granted yet if there are overlapping and conflicting lock
4728 requests. The readahead process waits on ungranted lock
4729 (original lock is CBPENDING), while OST waits for that process
4730 to cancel CBPENDING read lock and eventually evicts client.
4732 Severity : enhancement
4734 Description: Initial enabling of flock support for clients
4735 Details : Implements fcntl advisory locking and file status functions.
4736 This feature is provided as an optional mount flag (default
4737 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4738 locking are implemented yet, and those that are are not guaranteed
4739 to be completely correct in production environments.
4740 mount -t lustre -o [flock|noflock] ...
4743 Frequency : occasional
4745 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4746 assertion in the ext3 JBD (journaling) layer.
4747 Details : The b_committed_data struct is protected by the big kernel lock
4748 in 2.4 kernels, serializing journal_commit_transaction() and
4749 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4750 access to this struct. In 2.6 kernels there is finer grained
4751 locking to improve SMP performance of the JBD layer.
4755 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4760 Description: merge and cleanup kernel patches.
4763 -----------------------------------------------------------------------------
4765 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4767 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4768 versions of Lustre. Please update all servers and clients to
4769 version 1.4.2 or later at the same time. You must also run
4770 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4771 to update the configuration logs.
4773 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4774 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4775 - swab llog records of type '0' so we get proper header size/idx (5861)
4776 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4777 - fix rename of one directory over another leaking an inode (5953)
4778 - avoid SetPageDirty on 2.6 (5981)
4779 - don't re-add just-being-destroyed locks to the waiting list (5653)
4780 - when creating new directories, inherit the parent's custom
4781 striping settings if present parent (3048)
4782 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4783 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4784 - don't hold client locks on temporary worklist from l_lru (5666)
4785 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4786 - automatically evict dead clients (5921)
4787 - Update file size properly in create+truncate+fstat case (6196)
4788 - Do not unhash mountpoint dentries, do not allow removal of
4790 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4791 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4792 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4793 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4794 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4795 - Correctly handle reads of files with no objects (6243)
4796 - lctl recover will also mark a device active if deactivate used (5933)
4798 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4799 - allow --write-conf on an MDS with different nettype than client (5619)
4800 - don't write config llogs to MDS for mounts not from that MDS (5617)
4801 - lconf should create multiple TCP connections from a client (5201)
4802 - init scripts are now turned off by default; run chkconfig --on
4803 lustre and chkconfig --on lustrefs to use them
4804 - upcalls are no longer needed for clients to recover to failover
4806 - add --abort-recovery option to lconf to abort recovery on device
4808 - add support for an arbitrary number of OSTs (3026)
4809 - Quota support protocol changes.
4810 - forward compatibility changes to wire structs (6007)
4811 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4812 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4813 - improved "lustre status" script
4814 - initialize blocksize for non-regular files (6062)
4815 - added --disable-server and --disable-client configure options (5782)
4816 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4817 - Vanilla 2.4.29 support
4818 - increase maximum number of obd devices to 520 (6242)
4819 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4820 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4821 SLES 9 2.6.5 kernels
4822 - limit stripes per file to 160 (the maximum EA size) (6093)
4824 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4827 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4828 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4829 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4830 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4831 - small ext3 extents cleanups and fixes (5733)
4832 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4833 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4834 - avoid needless client->OST connect, fix handle mismatch (5317)
4835 - fix DLM error path that led to out-of-sync client, long delays (5779)
4836 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4837 - fix several locking issues related to i_size (5492,5624,5654,5672)
4838 - don't move pending lock onto export if it is already evicted (5683)
4839 - fix kernel oops when creating .foo in unlinked directory (5548)
4840 - fix deadlock in obdfilter statistics vs. object create (5811)
4841 - use time_{before,after} to avoid timer jiffies wrap (5882)
4842 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4843 - Do not leak request if server was not able to process it (5154)
4844 - If mds_open unable to find parent dir, make that negative lookup(5154)
4845 - don't create new directories with extent-mapping (5909, 5936)
4847 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4848 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4849 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4850 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4851 - support for RHEL4 kernel on i686 (5773)
4852 - provide error messages when incompatible logs are encountered (5898)
4854 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4855 * version 1.4.0.10 (1.4.1 release candidate 1)
4857 - don't keep a lock reference when lock is not granted (4238)
4858 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4859 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4860 - drop import inflight refcount on signal_completed_replay error (5255)
4861 - unlock page after async write error during send (3677)
4862 - handle missing objects in filter_preprw_read properly (5265)
4863 - no transno return for symlink open, don't save no-trasno open (3440)
4864 - don't try to complete elan receive that already failed (4012)
4865 - free RPC server reply state on error (5406)
4866 - clean up thread from ptlrpc_start_thread() on error (5160)
4867 - readahead could read extra page into cache that wasn't ejected (5388)
4868 - prevent races in class_attach/setup/cleanup/detach (5260)
4869 - don't dereference de->d_inode after l_dput of de (5458)
4870 - use "int" for stripe value returned from lock_to_stripe (5544)
4871 - mballoc allocation and error-checking fixes in 2.6 (5504)
4872 - block device patches to fix I/O request sizes in 2.6 (5482)
4873 - look up hostnames for IB nals (5602)
4874 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4875 - don't start multiple acceptors for the same port (5277)
4876 - fix incorrect LASSERT in mds_getattr_name (5635)
4877 - export a proc file for general "ping" checking (5628)
4878 - fix "lfs check" to not block when the MDS is down (5628)
4880 - service request history (4965)
4881 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4882 - create an "evict_client" /proc entry on OSTs, like the MDS has
4883 - fix mount usage message, return errors per mount(8) (5168)
4884 - change grep [] to grep "[]" in tests so they work in more UMLs
4885 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4886 - remove ancient LOV_MAGIC_V0 EA support (5047)
4887 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4888 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4889 - allow readahead allocations to fail when low on memory (5383)
4890 - mmap locking landed again, after considerable improvement (2828)
4891 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4893 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4896 - send OST transaction number in read/write reply to free req (4966)
4897 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4898 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4899 - the watchdog thread now runs as interruptible (5246)
4900 - flock/lockf fixes (but it's still disabled, pending 5135)
4901 - don't use EXT3 constants in llite code (5094)
4902 - memory shortage at startup could cause assertion (5176)
4904 - reorganization of lov code
4905 - single portals codebase
4907 - add extents/mballoc support (5025)
4908 - direct I/O reads in the obdfilter (4048)
4909 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4911 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4914 - send OST transaction number in read/write reply to free req (4966)
4915 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4916 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4917 - the watchdog thread now runs as interruptible (5246)
4918 - handle missing objects in filter_preprw_read properly (5265)
4919 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4920 - drop import inflight refcount on signal_completed_replay error (5255)
4921 - unlock page after async write error during send (3677)
4922 - return original error code on reconstructed replies (3761)
4923 - no transno return for symlink open, don't save no-trasno open (3440)
4925 - add pid to ldlm debugging output (4922)
4926 - bump the watchdog timeouts -- we can't handle 30sec yet
4927 - extra debugging for orphan dentry/inode bug (5259)
4929 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4932 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4933 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4934 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4935 - fix NULL dereference in /proc/sys/portals/routes (4827)
4936 - allow failed mdc_close() operations to be interrupted (4561)
4937 - stop precreate on OST before MDS would time out on it (4778)
4938 - don't send partial-page writes before EOF from client (4410)
4939 - discard client grant for sub-page writes on large-page clients (4520)
4940 - don't free dentries not owned by NFS code, check generation (4806)
4941 - fix lsm leak if mds_create_objects() fails (4801)
4942 - limit debug_daemon file size, always print CERROR messages (4789)
4943 - use transno after validating reply (3892)
4944 - process timed out requests if import state changes (3754)
4945 - update mtime on OST during writes, return in glimpse (4829)
4946 - add mkfsoptions to LDAP (4679)
4947 - use ->max_readahead method instead of zapping global ra (5039)
4948 - don't interrupt __l_wait_event() during strace
4950 - add software watchdogs to catch hung threads quickly (4941)
4951 - make lustrefs init script start after nfs is mounted
4952 - fix CWARN/ERROR duplication (4930)
4953 - return async write errors to application if possible (2248)
4954 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4955 - print NAL number in %x format (4645)
4956 - update barely-supported suse-2.4.21-171 series (4842)
4957 - support for sles 9 %post scripts
4958 - support for building 2.6 kernel-source packages
4959 - support for sles km_* packages
4961 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4964 - ignore -ENOENT errors in osc_destroy (3639)
4965 - notify osc create thread that OSC is being cleaned up (4600)
4966 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4967 - reconstruct ost_handle() like mds_handle() (4657)
4968 - create a new thread to do import eviction to avoid deadlock (3969)
4969 - let lconf resolve symlinked-to devices (4629)
4970 - don't unlink "objects" from directory with default EA (4554)
4971 - hold socknal file ref over connect in case target is down (4394)
4972 - allow more than 32000 subdirectories in a single directory (3244)
4973 - fix blocks count for O_DIRECT writes (3751)
4974 - OST returns ENOSPC from object create when no space left (4539)
4975 - don't send truncate RPC if file size isn't changing (4410)
4976 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4977 - bind to privileged port in socknal and tcpnal (3689)
4979 - rate limit CERROR/CWARN console message to avoid overload (4519)
4980 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4981 - basic mmap support (3918)
4982 - kernel patch series update from b1_4 (4711)
4984 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4987 - avoid crash during MDS cleanup with OST shut down (2775)
4988 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4989 - don't use bad inodes on the MDS (3744)
4990 - dynamic object preallocation to improve recovery speed (4236)
4991 - don't hold spinlock over lock dumping or change debug flags (4401)
4992 - don't zero obd_dev when it is force cleaned (3651)
4993 - print grants to console if they go negative (4431)
4994 - "lctl deactivate" will stop automatic recovery attempts (3406)
4995 - look for existing locks in ldlm_handle_enqueue() (3764)
4996 - don't resolve lock handle twice in recovery avoiding race (4401)
4997 - revalidate should check working dir is a directory (4134)
4999 - don't always mark "slow" obdfilter messages as errors (4418)
5001 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5004 - don't close LustreDB during write_conf until it is done (3860)
5005 - fix typo in lconf for_each_profile (3821)
5006 - allow dumping logs from multiple threads at one time (3820)
5007 - don't allow multiple threads in OSC recovery (3812)
5008 - fix debug_size parameters (3864)
5009 - fix mds_postrecov to initialize import for llog ctxt (3121)
5010 - replace config semaphore with spinlock (3306)
5011 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5012 - don't allow enqueue to complete on a destroyed export (3822)
5013 - down write_lock before checking llog header bitmap (3825)
5014 - recover from lock replay timeout (3764)
5015 - up llog sem before sending rpc (3652)
5016 - reduce ns lock hold times when setting kms (3267)
5017 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5018 - fix NULL deref and obd_dev leak on setup error (3312)
5019 - replace some LBUG about llog ops with error handling (3841)
5020 - don't match INVALID dentries from d_lookup and spin (3784)
5021 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5022 - fix invalid assertion in ptlrpc_set_wait (3880)
5024 - add libwrap support for the TCP acceptor (3996)
5025 - add /proc/sys/portals/routes for non-root route listing (3994)
5026 - allow setting MDS UUID in .xml (2580)
5027 - print the stack of a process that LBUGs (4228)
5029 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5032 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5033 - make sure to unset replay flag from failed open requests (3440)
5034 - if default stripe count is 0, use OST count for inode size (3636)
5035 - update parent mtime/ctime on client for create/unlink (2611)
5036 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5037 - free recovery state on server during a forced cleanup (3571)
5038 - unregister_reply for resent reqs (3063)
5039 - loop back devices mounting and status check on 2.6 (3563)
5040 - fix resource-creation race that can provoke i_size == 0 (3513)
5041 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5042 - more debugging for page-accounting assertion (3746)
5043 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5044 - avoid deadlock after precreation failure (3758)
5045 - fix race and lock order deadlock in orphan handling (3450, 3750)
5046 - add validity checks when grabbing inodes from l_ast_data (3599)
5048 - add /proc/.../recovery_status to obdfilter (3428)
5049 - lightweight CDEBUG infrastructure, debug daemon (3668)
5050 - change default OSC RPC parameters to be better on small clusters
5051 - turn off OST read cache for files smaller than 32MB
5052 - install man pages and include them in rpms (3100)
5053 - add new init script for (un)mounting lustre filesystems (2593)
5054 - run chkconfig in %post for init scripts (3701)
5055 - drop scimac NAL (unmaintained)
5057 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5060 - clean kiobufs before and after use (3485)
5061 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5062 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5063 - change init script to fail more clearly if not run as root (1528)
5064 - allow clients to reconnect during replay (1742)
5065 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5066 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5067 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5068 - fix cond_resched (3554)
5069 - only evict unfinished clients after recovery (3515)
5070 - allow bulk resend, prevent data loss (3570)
5071 - dynamic ptlrpc request buffer allocation (2102)
5072 - don't allow unlinking open directory if it isn't empty (2904)
5073 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5074 - remove extraneous obd dereference causing LASSERT failure (3334)
5075 - don't use get_cycles() when creating temp. files on the mds (3156)
5076 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5077 - handle EEXIST for set-stripe, set proper directory name (3336)
5079 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5080 - fix ksocknal_fmb_callback() error messages (2918)
5082 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5085 - don't copy lvb into (possibly NULL) reply on error (2983)
5086 - don't deref dentry after dput, don't free lvb on error (2922)
5087 - use the kms to determine writeback rpc length (2947)
5088 - increment oti_logcookies when osc is inactive (2948)
5089 - update client's i_blocks count via lvb messages (2543)
5090 - handle intent open/close of special files properly (1557)
5091 - mount MDS with errors=remount-ro, like obdfilter (2009)
5092 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5093 - don't use cancelling-locks' kms values (2947)
5094 - use highest lock extent for kms, not last one (2925)
5095 - don't dereference ERR_PTR() dentry in error handling path (3107)
5096 - fix thread race in portals_debug_dumplog() (3122)
5097 - create lprocfs device entries at setup instead of at attach (1519)
5098 - common AST error handler, don't evict client on completion race (3145)
5099 - zero nameidata in detach_mnt in 2.6 (3118)
5100 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5101 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5102 - initialize RPC timeout timer earlier for 2.6 (3219)
5103 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5104 - print nal/nid for unknown nid (3258)
5105 - additional checks for oscc recovery before doing precreate (3284)
5106 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5107 - don't crash in mdc_close for bad permissions on open (3285)
5108 - zero i_rdev for non-device files (3147)
5109 - clear page->private before handing to FS, better assertion (3119)
5110 - tune the read pipeline (3236)
5111 - fix incorrect decref of invalidated dentry (2350)
5112 - provide read-ahead stats and refine rpc in flight stats (3328)
5113 - don't hold journal transaction open across create RPC (3313)
5114 - update atime on MDS at close time (3265)
5115 - close LDAP connection when recovering to avoid server load (3315)
5116 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5117 - don't leak open file on MDS after open resend (3325)
5118 - serialize filter_precreate and filter_destroy_precreated (3329)
5119 - loop device shouldn't call sync_dev() for nul device (3092)
5120 - clear page cache after eviction (2766)
5121 - resynchronize MDS->OST in background (2824)
5122 - refuse to mount the same filesystem twice on same mountpoint (3394)
5123 - allow llmount to create routes for mounting behind routers (3320)
5124 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5125 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5126 - fix rare problem with rename on htree directories (3417)
5128 - allow default OST striping configuration per directory (1414)
5129 - fix compilation for qswnal for 2.6 kernels (3125)
5130 - increase maximum number of MDS request buffers for large systems
5131 - change liblustreapi to be useful for external progs like lfsck (3098)
5132 - increase local configuration timeout for slow disks (3353)
5133 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5135 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5138 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5139 - don't overwrite extent policy data in reply if lock was blocked (2901)
5140 - drop filter export grants atomically with removal from device (2663)
5141 - del obd_self_export from work_list in class_disconnect_exports (2908)
5142 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5143 - swab reply message in mdc_close, other PPC fixes (2464)
5144 - fix destroying of named logs (2325)
5145 - overwrite old logs when running lconf --write_conf (2264)
5146 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5147 - fix race in target_handle_connect (2898)
5148 - mds_reint_create() should take same inode create lock (2926)
5149 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5150 - don't close files for self_export to avoid uninitialized obd (2936)
5151 - allow MDS with the same name as client node (2939)
5152 - hold dentry reference for closed log files for unlink (2325)
5153 - reserve space for all logs during transactions (2059)
5154 - don't evict page beyond end of stripe extent (2925)
5155 - don't oops on a deleted current working directory (2399)
5156 - handle hard links to targets without a parent properly (2517)
5157 - don't dereference NULL lock when racing during eviction (2867)
5158 - don't grow lock extents when lots of conflicting locks (2919)
5160 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5163 - account for cache space usage on clients to avoid data loss (974)
5164 - lfsck support in lustre kernel code (2349)
5165 - reduce journal credits needed for BRW writes (2370)
5166 - orphan handling to avoid losing space on client/server crashes
5167 - ptlrpcd can be blocked, stopping ALL progress (2477)
5168 - use lock value blocks to assist in proper KMS, faster stat (1021)
5169 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5170 - recovery for initial connections (2355)
5171 - fixes for mds_cleanup_orphans (1934)
5172 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5173 - block all file creations until orphan recovery completes (1901)
5174 - client remove rq_connection from request struct (2423)
5175 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5176 - recovery timer race (2670)
5177 - mdc_close recovey bug (2532)
5178 - ptlrpc cleanup bug (2710)
5179 - mds timeout on local locks (2588)
5180 - namespace lock held during RPCs (2431)
5181 - handle interrupted sync write properly (2503)
5182 - don't try to handle a message that hasn't been replied to (2699)
5183 - client assert failure during cleanup after abort recovery (2701)
5184 - leak mdc device after failed mount (2712)
5185 - ptlrpc_check_set allows timedout requests to complete (2714)
5186 - wait for inflight reqs when ptlrpcd finishes (2710)
5187 - make sure unregistered services are removed from the srv_list
5188 - reset bulk XID's when resending them (caught by 1138 test)
5189 - unregister_bulk after timeout
5190 - fix lconf error (2694)
5191 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5192 - readahead locks pages, leaves pending causing memory pressure (2673)
5193 - increase OST request buffers to 4096 on large machines (2729)
5194 - fix up permission of existing directories in simple_mkdir (2661)
5195 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5196 - don't assign transno to errored transactions (2742)
5197 - don't delete objects on OST if given a bogus objid from MDS (2751)
5198 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5199 - if rq_no_resend, then timeout request after recovery (2432)
5200 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5201 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5202 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5203 - MDS recovery shouldn't delete objects that it has given out (2730)
5204 - if enqueue arrives after completion, don't clobber LVB (2819)
5205 - don't unlock pages twice when trigger_group_io returns error (2814)
5206 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5207 - don't write pages to disk if there was an error (1450)
5208 - don't ping imports that have recovery disabled (2676)
5209 - take buffered bytes into account when balancing socknal conn (2817)
5210 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5211 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5212 - remove little-endian swabbing of llog records (1987)
5213 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5214 - retry reposting request buffers if they fail (1191)
5215 - grow extent at grant time to avoid granting a revoked lock (2809)
5216 - lock revoke doesn't evict page if covered by a second lock (2765)
5217 - disable VM readahead to avoid reading outside lock extents (2805)
5219 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5220 - updated kernel patches for hp-2.4.20 kernel (2681)
5222 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5225 - fix truncated write corruption (2366)
5226 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5228 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5229 - ensure len(uuid) < 37 in lmc (1171)
5230 - fix ia64 OOPS in llog_test (2255)
5231 - zero end of page at obdfilter for partial page writes (2648)
5232 - don't leave stale dentries around after renames (bug 2428)
5233 - fix timeouts when evicting a client with a single lock held (2642)
5234 - set deadline for the initial HELLO message to drain (2634)
5235 - print out dotted-quad IP addresses in the socknal (2302)
5237 - additional debugging for MDS client eviction problem (2443)
5238 - fix mkfsoptions support for osts (2603, 2604)
5240 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5243 - add series for the vanilla 2.6.0 kernel
5244 - add series for the vanilla 2.4.24 kernel
5245 - add series for a cray x86/64 UL kernel drop
5246 - fix xattr patches for the vanilla 2.4.19 series
5248 - generate true UUIDs in lmc (1171)
5249 - have portals stack dumping break in UML (2466)
5250 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5251 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5252 - wake callback waiting threads on client eviction (2460)
5253 - Add --ptldebug and --subsystem to lmc (1719)
5254 - update assertion to allow safe interrupt allocation
5255 - set rq_no_resend for cancel requests (2432)
5256 - recalculate ptlrpcd timeout after resend (2494)
5257 - call vfs_rmdir when removing pending directories (2368)
5258 - fix renaming a file to itself (2429)
5259 - lmc creates a default one-stripe lov (2454)
5260 - expand procfs space to handle large clusters (2326)
5261 - increase UML stack to avoid overflow
5262 - update lconf's list of debug and subsystem masks
5263 - fix lfs find --obd (2510)
5264 - /proc tunable for disabling filter read caching (2591)
5265 - stop rpm packages from altering slapd.conf (2301)
5266 - disable nagle in the socknal under 0conf (2578)
5267 - choose mds inode size based on stripe count (2572)
5268 - fix kernel-source rpm problems (2516)
5270 - add --disable-doc to avoid pdf generation (2421)
5271 - update documentation, tests, type-os, comments
5272 - avoid format warnings on ia64
5273 - remove the TOE NAL
5274 - tiny code cleanups by removing unused fields
5276 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5279 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5280 - remove the most blatant lies from BUILDING file (2371)
5281 - change default debug level to reasonable production setting
5282 - reduce client side cache size to reduce cache flush time
5283 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5284 - make TCP zerocopy and pinger support enabled by default (2476)
5285 - sync writes completed after process exits caused crashes (2319)
5286 - maintain correct mount count on the MDS (2356)
5287 - backout 1557, because 2316 wasn't really fixed
5288 - better file I/O statistics gathering in /proc
5289 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5290 - another kernel patch to fix zero-copy TCP function export
5291 - don't take duplicate lock when processing re-sent getattr (2420)
5292 - lctl uses obd_self_export instead of creating new conn (2353)
5293 - MDS/OST recovery case which requires object creation asserted (2425)
5294 - move lfs from /usr/sbin to /usr/bin in packages
5295 - fix race between mds_client_add and mds_client_free (2417)
5296 - use kmalloc instead of slabs in portals (2430)
5297 - don't create duplicate records when a failover MDS is present (2442)
5298 - remove unnecessary mount age check (2332)
5299 - don't remove directory inodes from locks prematurely (2451)
5300 - don't break if MDS service name is the same as hostname (2103)
5301 - fix races in client write RPC generation when cache full (2482)
5303 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5306 - remove now-unused request->rq_obd (278)
5307 - if an allocation fails, print out how much memory we've used (1933)
5308 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5309 - add the "configurable stack size" patch to most series files (1256)
5310 - ability to write large log records, for 100+ OST configs (2306)
5311 - fix NULL deref when filter_prep fails (2314)
5312 - fix operator precedence error in filter_sync
5313 - dynamic allocation of socknal TX descriptors (2315)
5314 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5315 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5316 - ensure that CWARN messages in Portals always get to the syslog
5317 - __init/__exit are not for prototype decls (ldlm_init/exit)
5318 - x86-64 compile warning fixes
5319 - fix gateway LMC keyword conflict (2318)
5320 - fix MDS lock inversions in getattr/reint paths (1844)
5321 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5322 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5323 - DLM race condition prevented some lock evictions (2328)
5324 - ENOMEM detection and retry on socknal sends (2230)
5325 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5326 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5327 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5328 - better error messages when a client is rejected during recovery (1505)
5329 - avoid cancelling locks which were never granted, after failure (2330)
5330 - fix i_sem/journal inversion in mds_client_add (2333)
5331 - fix truncate/getattr lock cycle deadlock (2334)
5332 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5333 - fix two rare exit paths which could leak an l_lock() ref (2321)
5334 - fencepost error in MDS/OST orphan recovery (2226)
5335 - make log record alignment 8 bytes (1988)
5336 - lstripe now fails when requested offset > ost_count (2237)
5337 - ensure that all kernel series have a complete list.h (1607)
5338 - fix crashes in special-file operations (2316)
5339 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5341 - allow configurable automake binary, for testing new versions
5342 - small update to the lfs documentation
5344 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5346 * fix negative export reference count in fsfilt_sync (2312)
5348 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5349 * release candidate 0.9.1
5351 - orphans are moved into the PENDING directory for possible recovery
5352 - replayed opens now open by fid for orphan/rename safety (1042)
5353 - last close of an orphan inode generates a transno (683)
5354 - chdir() and mount() now pin the directory entry (1020)
5355 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5356 - discard very old requests without processing them (1502)
5357 - remove some common, well-understood CERRORs (1505)
5358 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5359 - clear "grant" flags in OST replies until OST grant code lands (1644)
5360 - fix read performance by not clobbering i_blksize on client (1598)
5361 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5362 - make LRU size a /proc tunable, clears locks when reduced (707)
5363 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5364 - server locks take references on exports now (1558)
5365 - build fixes for 2.4.20-rh trees (1663)
5366 - return an error from lov_create if all OSCs are inactive (1751)
5367 - fix import levels when a reconnect happens without a timeout (1597)
5368 - exit early from mds_open if we get a lookup error (1749)
5369 - partial page read at EOF wouldn't wait for disk before sending (1642)
5370 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5371 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5372 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5373 - write/truncate lock inversion (1639)
5374 - Don't auto-load obdclass, portals modules during cleanup (1495)
5375 - fix timestamps from jumping to "now" (1763)
5376 - extra journal assertions (1648)
5377 - add an extra multiunlink test (1771)
5378 - fix read_record/write_record API (1776)
5379 - fix leak of offset_extent, possible incorrect i_size later (1772)
5380 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5381 - Debugging for the kqswnal_get_idle_tx problems (1820)
5382 - Allow recovery to be attempted multiple times (1536)
5383 - Write out MDS last_rcvd file after it is first created (1600)
5384 - Fix tx_descriptor leak in failed transmit situations (1827)
5385 - ext3 journaling fixes for assertion failure after IO error (1871)
5386 - class_export_put() on freed export after completion AST error (1896)
5387 - Fix revalidate looping in VFS (1322)
5388 - Don't access a freed export during MDS_REINT timeout (1521)
5389 - Add open-unlink recovery support on the MDS (1673,1764)
5390 - Return an error if no MDS data was read from last_rcvd (1946)
5391 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5392 - Don't setup a disk device that doesn't match exported UUID (317)
5393 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5394 - avoid committing NULL handle in force close
5395 - local.sh is now a one-stripe LOV configuration
5396 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5397 - fix ext3 htree duplicate directory entry corruption (1516)
5398 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5399 - update to new LOV EA format (2097)
5400 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5402 2003-06-15 Phil Schwan <phil@clusterfs.com>
5405 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5406 - per-import/export recovery handling (958, 931, 959)
5407 - multiple last-rcvd slots, for serving multiple FSes (949)
5408 - connections are again shared between multiple imp/exports (963, 964)
5409 - "umount -f" would hang if any requests needed to be sent (393, 978)
5410 - avoid pinning large req buffer by copying for queued messages (989)
5411 - add "uuid" to "lctl device" command to help upcalls (991)
5412 - "open" RPCs with transnos would confuse recovery counters (1037)
5413 - do proper endian conversion of all wire messages (288, 340, 891)
5414 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5415 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5416 - fix ptlrpc_connection leak in target_handle_connect (1174)
5417 - fix import refcounting bug in OST and MDS cleanup (1134)
5418 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5419 - fix very unlikely obd_types race condition (501)
5420 - remove osc_open hack for echo_client (1187)
5421 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5422 - a failure in read_inode2 leads to deadlock (1139)
5423 - cancel ack-locks as soon as transaction is committed (1072)
5424 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5425 - make sure to commitrw after any preprw to avoid deadlock (1162)
5426 - failing to execute a file in a lustre FS would lock inode (1203)
5427 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5428 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5429 - obd_uuid2tgt would walk off the end of the list (1255)
5430 - on IA64 the osc would give portals incorrect bulk size (1258)
5431 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5432 - fix lock inversion caused by new llite matching code (1282)
5433 - limit the number of dirty pages on a client to 10MB (1286)
5434 - timed out locks were not being corrected cancelled (1289)
5435 - fix O_DIRECT above 4GB on IA-32 (1292)
5436 * major user-visible changes
5437 - fail out/fail over policy now controlled by the upcall (993)
5439 - add OBD_PING to check server availability and failure (954)
5440 - lustre messages are now sent in sending host order (288, 340, 891)
5441 - add eadatalen to MDS getattr reply (340)
5442 - OST read replies may contain second buffer, with per-page status (593)
5444 2003-03-11 Phil Schwan <phil@clusterfs.com>
5447 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5448 - failed open()s could cause deadlock; fixed (867, 869)
5449 - stop cancelling OST locks when files are closed (481)
5450 - overlapping XID spaces caused network corruption (851, 853)
5451 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5452 - setattr_raw added, to do single-RPC, server-side setattrs
5453 - lmc/lconf syntax change for OST UUIDs
5454 - fix crashy race condition between ptlrpc_free_req and osc_close
5455 - don't use request in mdc_enqueue if we hit a timeout (889)
5456 - don't set the inode i_size for regular files from the MDS (896)
5457 - handle out of order completion AST (842)
5458 - don't LBUG if a lock request times out after receiving AST (913)
5459 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5460 - if a bad lock AST arrives, send an error instead of dropping entirely
5461 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5462 - fix leak in bulk IO when only partially completed (899, 900, 926)
5463 - fix O_DIRECT for ia64 (55)
5464 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5465 - C-z after timeout could hang a process forever; fixed (977)
5467 - client-side I/O cache (678, 924, 929, 941, 970)
5469 - READPAGE and SETATTRs which don't take server-side locks get
5472 2003-02-11 Phil Schwan <phil@clusterfs.com>
5475 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5476 - Fix fsfilt_extN_readpage() to read a full page of directory
5477 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5478 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5479 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5480 - Most intent operations were removed, in favour of a new RPC mode
5481 that does a single RPC to the server and bypasses most of the VFS
5482 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5483 - Aggressively cancel local locks on DLM servers
5484 - mds_reint_unlink sends EA to the client if it's the last nlink.
5485 client uses that EA to unlink OST objects.
5486 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5487 - recursive symlinks were fixed (439)
5488 - fixed NULL deref in DEBUG_REQ
5489 - filter_update_lastobjid no longer calls sync, which annoyed extN
5490 - fixed multi-client small-writes to a single file problem (445)
5491 - fixed mtime updates during file writes (607)
5492 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5493 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5494 - fixed Linux OST size reporting problem (444, 656)
5495 - OST now updates object mtime with writes or setattr (607, 619)
5496 - client verifies file size before zeroing page past EOF (445)
5497 - OST now writes last allocated objid to disk with allocation (108)
5498 - LOV on echo now works (409)
5500 - mds_reint_unlink sends a new buffer, with the EA included. this
5501 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5502 set if a regular file was being unlinked, and it was the last link
5503 - use PtlGet from the target for bulk writes (315)
5504 - OST now updates object mtime with writes or setattr (607, 619)
5505 - LDLM now has a grant-time callback to revalidate locked items, if
5507 - Many MDS operations were reorganized to combat race conditions
5509 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5510 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5512 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5515 - Fully reactivate OST imports after reconnection (512, others)
5516 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5517 - More graceful error handling for truncating on dead OST (515)
5518 - Don't error out unless we're actually accessing dead stripes (474)
5519 - Fix garbage sizes when stripes are missing (410)
5520 - LRU counters were broken, causing constant lock purge (433, 432)
5521 - garbage on read from stripes with failed OSTs (441)
5522 - mark OSCs as active before reconnecting during recovery (438)
5523 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5524 - lfind did not preserve OST order in output (443)
5525 - symlinks cause hung clients, incorrect data (439)
5526 - stop dereferencing request after dropping refcount (457)
5527 - don't LASSERT(spin_is_locked) on non-SMP (455)
5528 - fixes for many rename() bugs
5529 - fstat didn't correctly synchronize attributes (399)
5530 - server must handle lock cancellation during blocking AST prep (487)
5531 - bulk descriptors were free()d too soon (511)
5532 - fix paths in lconf, which would load incorrect modules (451, 507)
5533 - fix confusing lconf 'host not found' error message (386)
5534 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5535 - fix race condition in mdc_blocking_ast() for inode access (526)
5536 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5537 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5538 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5539 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5540 - fix lstripe in several ways (kernel side) (527)
5541 - fix request leak in ldlm_cli_enqueue (262)
5542 - incorrect OSC was marked inactive after OST failure
5543 - call mds_fs_cleanup before unmounting filesystem (524)
5544 - fix races between taking ns_lock and ldlm_lock_change_resource
5545 - fix races updating LOV export open file list
5546 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5547 - fix recovery NULL deref in ldlm_cli_cancel_unused
5548 - fix some DLM races by using new hash table for lock handles (419)
5549 - permit the client to specify desired inodes, at replay
5550 - duplicate requests when we queue them for replay reintegration
5551 - fix last_rcvd offset calculation
5552 - sync after each recovered transaction, so we always make progress
5553 - never, not always, ERESTART requests without transnos
5554 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5556 - skip replay if the MDS says that the client is already connected
5557 - don't check for a recovery-enabled export to match lctl's UUID
5558 - don't INC_USE_COUNT for phantom exports
5559 - don't crash when cleaning up phantom exports (567)
5560 - don't double-finish or set replay data for errored mdc_open requests
5561 - abort requests when they time out, so we don't get old replies
5562 - send/receive replies for AST messages again
5563 - if the client says that it doesn't have the lock, cancel it on the
5565 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5566 mark it as destroyed, it will all work out when decref is called
5567 - fix module use counts (22, 581)
5569 - ASTs now expect a reply (server cancels lock on error reply)
5571 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5574 - fix many simultaneous client startup (392)
5575 - fix dentry->d_it clobbering
5576 - credentials weren't being shipped for readdir/getattr operations
5577 - remove invalid assertions triggered during some concurrent MD
5579 - proper Lustre versions added (336, 389)
5580 - fix memory leak for create error case (398)
5581 - fix LOV locking bug that would get cli/srv out of sync
5582 - fix echo client over LOV (409)
5583 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5584 - fix double-O_EXCL intent crash (424)
5585 - avoid sending multiple lock CANCELs (352)
5587 - MDS can do multi-client recovery (modulo bugs in new code)
5589 - many updates, edits, cleanups
5591 2002-11-18 Phil Schwan <phil@clusterfs.com>
5594 - fix null d_it dereference (346)
5595 - fix full OST/dbench hang (333)
5596 - fix permission problem with file removal (286)
5597 - fix removal of OSCs from LOV when they fail
5598 - fix NULL deref during bulk timeout (214)
5599 - fix problems related to multiple filesystems on one MDS (241)
5600 - fixed serious subtle metadata locking bugs
5601 - free locks on clients when inodes are removed due to memory
5603 - fix inode pointer in lock data (285)
5604 - partial support for multiple MDS on a single host (241)
5605 - data locks weren't cancelled at clear_inode time (290, 311)
5606 - intent locks could lead to unbounded lock growth (205)
5607 - added a maximum lock count, an LRU list, and a flusher
5608 - fix multiple rename (365)
5609 - properly abstracted the echo client
5610 - OSC locked 1 byte too many; fixed
5611 - rewrote brw callback code:
5612 - fixed recovery bugs related to LOVs (306)
5613 - fixed too-many-pages-in-one-write crash (191)
5614 - fixed (again) crash in sync_io_timeout (214)
5615 - probably fixed callback-related race (385)
5617 - Add capability to MDS protocol
5618 - LDLM cancellations and callbacks on different portals
5620 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5623 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5624 - defer open object destruction to close time (601981, 138)
5625 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5626 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5627 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5628 - don't call obd_setattr in ll_file_release for destroyed objects
5629 * protocol change to lustre_msg: move |version| and add |flags|
5630 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5631 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5632 * added replay of create, unlink, link and rename operations during
5633 MDS failover; recovery should be much more robust now
5634 * remove failed OSCs from LOVs (only lov_create uses this so far)
5635 * the lustre-HOWTO was brought (more) up to date (582544)
5637 2002-10-23 Phil Schwan <phil@clusterfs.com>
5640 - in-use dentries weren't being reused properly (617851)
5641 - prevent multiple LDLM setup (599178)
5642 - fix LOV size calculations for truncate (617853)
5643 - fix client handling of MDS intent errors (POSIX)
5644 - fix permission bug in lovstripe.c test (624321)
5645 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5646 - truncate past end of file could corrupt data
5647 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5648 - a race in recovery could return ETIMEDOUT to apps (623947)
5649 - building outside the source directory was fixed
5650 * the lustre-HOWTO was brought (more) up to date (582544)
5651 * major progress was made on recovery functionality
5653 2002-10-10 Phil Schwan <phil@clusterfs.com>
5656 - recovery deadlock fix
5657 - rm -rf causes LBUG fix (617817)
5658 - file open by multiple tasks fix (618962)
5659 - directory permissions bugs (602707 and 620007)
5660 - journal_stop fixed with locking (611313)
5661 - O_APPEND failures resolved (618273, perhaps 614459)
5662 - lconf PATH fix (619770)
5663 - IA64 build fix (621450)
5664 - RPC buffer sizes scale with amount of memory
5666 2002-10-01 Phil Schwan <phil@clusterfs.com>
5669 - locks would be cancelled without throwing away data pages,
5670 resulting in inconsistent data (605627)
5671 - inode attributes were not always being refreshed (605627, 612449)
5672 - lconf now continues to cleanup after lctl reports an error
5673 - MDS now enforces user permissions (602707)
5674 - lprocfs cleanup fixed, but not yet enabled (614157)
5675 - fixed infinite server hang, should a client not respond to an AST
5676 - avoid going into recovery if user calls readlink() with a buffer
5677 that's too small (613941)
5678 - AST RPCs no longer require replies (614867) -- this may be changed
5679 - don't crash server if client sends an IOV that's too big (611336)
5680 - fixed lock conversion deadlock (611892)
5681 - fixed the following of symlinks (614622)
5682 * recovery: the server can remove locks from a client that dies, other
5683 clients can make progress
5684 * more extN patch fixes
5685 * compile-time configurable ptlrpc buffer allocations
5687 - collaborative read cache document
5688 - Lustre Lite Performance CDR document-in-progress
5690 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5693 - fix typo in patch-2.4.18
5695 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5698 - clear ptlrpc request each time in handle_incoming_request()
5699 - unlink of files now destroys the object on the OST
5701 2002-09-19 Peter Braam <braam@clusterfs.com>
5703 * add hard link support
5704 * change obdfile creation method
5705 * kernel patch changed
5707 2002-09-19 Peter Braam <braam@clusterfs.com>
5710 - stack overflow bug in extN fixed
5712 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5714 * documentation updates
5715 - add man pages for config tools
5716 - update tests/README to describe testing with new config tools
5717 - finish metadata API descriptions
5718 * bug fixes and cleanups
5719 - statfs workaround for 16TB limit
5720 - LOV stripe allocation improved, can stripe on subset of OSTs
5721 - LOV file size/IO offset was wrong for files > 4GB in size
5722 - object EA data was being dropped, caused files to be unreadable
5723 - memory overflow with non-LOV OST caused memory corruption
5724 - fixed regression tests to work with new config tools, obdfilter
5725 - fixed bug when directory size became larger than 1 block
5726 - fixed bug (for single client case) when PWD was deleted
5727 - invalidate local directory pages when doing intent-based ops
5728 - avoid LDLM oops when lock callback contained bad data
5730 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5732 * documentation updates
5733 * bug fixes and cleanups
5734 - configuration tools
5737 - 64-bit compile warnings
5738 - 64-bit internal statfs data
5740 * test_brw on persistent OST devices
5742 * lprocfs (disabled)
5744 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5746 * documentation updates
5747 * bug fixes and cleanups
5748 * configuration tools
5750 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5753 * many small fixes to 0_5_4
5754 * io/network handling
5755 * thinkos in MDS operations
5757 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5760 * crucial basic fixes to 0.5.3
5761 * IOR, Iozone work over Elan
5764 2002-08-07 Phil Schwan <phil@clusterfs.com>
5765 * version 0_5_3, our first alpha
5766 * we use the new Portals iovs
5767 * documentation updates
5768 * bug fixes and cleanups
5769 * small changes in the DLM wire protocol
5771 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5772 * version 0_5_1 with some initial stability,
5773 * locking on MD and file I/O.
5774 * documentation updates
5775 * several bug fixes since 0.5.0
5776 * small changes in wire protocol
5778 2002-07-18 Phil Schwan <phil@clusterfs.com>
5780 * delivered as Lustre Light Alpha
5781 * fixed a crash after handling invalid MDS requests
5782 * fixed directory pages for architectures with non-4k pages sizes
5784 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5785 * release version v0_4_4
5786 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5788 * Fixes a number of interruption problems with OST operations.
5789 * Update documentation for portals header changes
5790 * Move all wire protocol structs/defines to lustre_idl.h
5791 * Fixes symlink length bug.
5792 * Add tcpdump to repository.
5794 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5795 * release version v0_4_3
5796 * Fixes statfs for inodes on extN.
5797 * Fixes bug in runtests which would delete /etc/hosts.
5798 * Use 64-bit object IDs wherever possible (not into VFS though)
5799 Remove ost_get_info, which is unused by lustre, and out of date.
5801 2002-07-03 Peter Braam <braam@clusterfs.com>
5802 * release version v0_4_2 Fixes a lookup error (type not passed)
5803 * move forward to head of Portals
5804 * move forward to latest Lustre kernel
5806 2002-06-25 Peter Braam <braam@clusterfs.com>
5807 * release version v0_4_1. Hopefully stable on single node use.