1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.31 (SLES 10),
5 2.6.18-92.1.17.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.40.11-sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
14 * File join has been disabled in this release, refer to Bugzilla 16929.
16 Severity : enhancement
18 Description: Changelogs
19 Details : Changelogs are a lightweight mechanism to track filesystem
20 metadata and namespace changes. The changelog is recorded
21 permanently on the MDTs, and is periodically "consumed" / purged
22 when records are no longer needed.
24 Severity : enhancement
26 Description: compact fld format with extents
27 Details : Store range of seq rather than every seq in FLD. Seq
28 controller update FLD rather than clients. In Case of CMD, mdt0
29 has FLD, all other metadata server act as non persistent proxy
30 for FLD queries and cache fld entries in fld cache.
35 Description: don't skip ost target if they assigned to file
36 Details : Drop slow OSCs if we can, but not for requested start idx.
37 This means "if OSC is slow and it is not the requested
38 start OST, then it can be skipped, otherwise skip it only
39 if it is inactive/recovering/out-of-space.
43 Description: more cleanup in mds_lov
44 Details : not send LOV EA under replay, we can't know about they size at this
45 time. Don't allow client connect to mds before any ost connected,
46 for avoid problems with LOV EA size and returning EIO to client.
48 Severity : enhancement
50 Description: Interoperability at server side (Disk interoperability)
52 Severity : enhancement
54 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
56 Severity : enhancement
58 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
60 Severity : enhancement
62 Description: New client IO stack (CLIO).
64 Severity : enhancement
66 Description: Commit on sharing. Eliminate inter-client dependencies between
67 uncommitted transactions by doing transaction commits.
68 Thereby clients may recovery independently.
71 Frequency : Create a symlink file with a very long name
73 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
74 Details : If there is no extra space in the request for early cancels,
75 ldlm_req_handles_avail() returns 0 instead of a negative value.
77 Severity : enhancement
79 Description: Add /proc entry for import status
80 Details : The mdc, osc, and mgc import directories now have
81 an import directory that contains useful import data for debugging
84 Severity : enhancement
86 Description: Re-disable certain /proc logging
87 Details : Enable and disable client's offset_stats, extents_stats and
88 extents_stats_per_process stats logging on the fly.
91 Frequency : Only on FC kernels 2.6.22+
93 Description: oops in statahead
94 Details : Do not drop reference count for the dentry from VFS when lookup,
95 VFS will do that by itself.
97 Severity : enhancement
99 Description: Generic /proc file permissions
100 Details : Set /Proc file permissions in a more generic way to enable non-
101 root users operate on some /proc files.
105 Description: Hitting mdc_commit_close() ASSERTION
106 Details : Properly handle request reference release in
107 ll_release_openhandle().
111 Description: quota recovery deadlock during mds failover
112 Details : This patch includes att18982, att18236, att18237 in bz14840.
114 1. fix osts hang when mds does failover with quotaon
115 2. prevent watchdog storm when osts threads wait for the
120 Frequency : only patchless client
121 Description: add workaround for race between add/remove dentry from hash
123 Severity : enhancement
125 Description: Allow OST glimpses to return PW locks
129 Description: LBUG when llog conf file is full
130 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
134 Description: Prevent import from entering FULL state when server in recovery
138 Description: service mount cannot take device name with ":"
139 Details : Only when device name contains ":/" will mount treat it as
145 Description: replace ptlrpcd with the statahead thread to interpret the async
146 statahead RPC callback
150 Frequency : on recovery
151 Description: I/O failures after umount during fail back
152 Details : if client reconnected to restarted server we need join to recovery
153 instead of find server handler is changed and process self
154 eviction with cancel all locks.
156 Severity : enhancement
158 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
160 Severity : enhancement
162 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
164 Severity : enhancement
166 Description: Upcall on Lustre log has been dumped
167 Details : Allow for a user mode script to be called once a Lustre log has
168 been dumped. It passes the filename of the dumped log to the
169 script, the location of the script can be specified via
170 /proc/sys/lnet/debug_log_upcall.
175 Description: avoid idr_remove called for id which is not allocated.
176 Details : Move assigment s_dev for clustered nfs to end of initialization,
177 to avoid problem with error handling.
182 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
183 Details : When connection is reused this not moved from CONN_UNUSED_HASH
184 into CONN_USED_HASH and this prodice warning when put connection
185 again in unused hash.
188 Severity : enhancement
190 Description: File striping can now be set to use an arbitrary pool of OSTs.
192 Severity : enhancement
194 Description: Export bytes_read/bytes_write count on OSC/OST.
198 Description: Early reply size mismatch, MGC loses connection
199 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
200 the connect flags are properly negotiated.
205 Description: quotas are not honored with O_DIRECT
206 Details : all writes with the flag O_DIRECT will use grants which leads to
207 this problem. Now using OBD_BRW_SYNC to guard this.
211 Description: add quota statistics
212 Details : 1. sort out quota proc entries and proc code.
213 2. add quota statistics
215 Severity : enhancement
217 Description: enable quota support for HEAD.
221 Description: Properly propagate oinfo flags from lov to osc for statfs
222 Details : restore missing copy oi_flags to lov requests.
224 Severity : enhancement
226 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
227 : l_getgroups(8), lst(8), routerstat(8)
228 Details : included man pages for llobdstat(8), llstat(8),
229 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
231 Severity : enhancement
233 Description: configure's --enable-quota should check the
234 : kernel .config for CONFIG_QUOTA
235 Details : configure is terminated if --enable-quota is passed but
236 : no quota support is in kernel
240 Description: Remove portals compatibility
241 Details : Remove portals compatibility, not interoperable with releases
246 Description: Resolve device initialization race
247 Details : Prevent proc handler from accessing devices added to the
248 obd_devs array but yet be intialized.
250 Severity : enhancement
252 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
254 Severity : enhancement
256 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
261 Description: Using wrong pointer in osc_brw_prep_request
262 Details : Access to array[-1] can produce panic if kernel compiled with
263 CONFIG_PAGE_ALLOC enabled
267 Description: Client runs out of low memory
268 Details : Consider only lowmem when counting initial number of llap pages
272 Description: *optional* service tags registration
273 Details : if the "service tags" package is installed on a Lustre node
274 When the filesystem is mounted, a local-node service tag will
275 be created. See http://inventory.sun.com/ for more information
276 about the Service Tags asset management system.
280 Description: Kernel BUG tries to release flock
281 Details : Lustre does not destroy flock lock before last reference goes
282 away. So always drop flock locks when client is evicted and
283 perform unlock regardless of successfulness of speaking to MDS.
287 Description: add refcount protection for osc callbacks, avoid panic on shutdown
291 Description: sanity test 65a fails if stripecount of -1 is set
292 Details : handle -1 striping on filesystem in ll_dirstripe_verify
297 Description: ASSERTION(CheckWriteback(page,cmd)) failed
298 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
301 Severity : enhancement
303 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
307 Description: do not process already freed flock
308 Details : flock can possibly be freed by another thread before it reaches
309 to ldlm_flock_completion_ast.
313 Description: LBUG during stress test
314 Details : Need properly lock accesses the flock deadlock detection list.
318 Description: oops in page fault handler
319 Details : kernel page fault handler can return two special 'pages' in error
320 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
324 Description: timeout with invalidate import.
325 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
326 should be handled by ptlrpcd. This produce long age waiting and
327 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
329 Severity : enhancement
331 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
335 Description: enable MGS and MDT services start separately
336 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
337 a co-located MGS without starting the MGS, which is a complement
338 to 'nosvc' mount option.
342 Frequency : after recovery
343 Description: precreate to many object's after del orphan.
344 Details : del orphan st in oscc last_id == next_id and this triger growing
345 count of precreated objects. Set flag LOW to skip increase count
346 of precreated objects.
350 Frequency : rare, on clear nid stats
351 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
352 Details : when clean nid stats sometimes try destroy live entry,
353 and this produce panic in free.
357 Description: Stack overflow during MDS log replay
358 ease stack pressure by using a thread dealing llog_process.
362 Description: wait until IO finished before start new when do lock cancel.
363 Details : VM protocol want old IO finished before start new, in this case
364 need wait until PG_writeback is cleared until check dirty flag and
365 call writepages in lock cancel callback.
367 Severity : enhancement
369 Description: using special macro for print time and cleanup in includes.
373 Description: mds_mfd_close() ASSERTION(rc == 0)
374 Details : In mds_mfd_close(), we need protect inode's writecount change
375 within its orphan write semaphore to prevent possible races.
379 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
383 Frequency : rare, on shutdown ost
384 Description: don't hit live lock with umount ost.
385 Details : shrink_dcache_parent can be in long loop with destroy dentries,
386 use shrink_dcache_sb instead.
390 Description: don't panic with use echo client
391 Details : echo client pass NULL as client nid pointer and this produce null
396 Description: fix build on ppc32
397 Details : compile code with -m64 flag produce wrong object file for ppc32.
401 Description: add message levels for liblustreapi
405 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
406 Details : Move the 'good_osts' check before the 'total_bavail' check. This
407 will result in an -EAGAIN and in the exit call path we call
408 alloc_rr() which will with increasing aggressiveness attempt to
409 aquire precreated objects on the minimum number of required OSCs.
413 Description: Use old size assignment to avoid deadlock
414 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
415 the scheduling while holding a spinlock. We do not need locking
416 for size in ll_update_inode() because size is only updated from
417 the MDS for directories or files without objects, so there is no
418 other place to do the update, and concurrent access to such inodes
419 are protected by the inode lock.
423 Description: resolve "_IOWR redefined" build error on SLES10
427 Description: dump the memory debugging after all modules are unloaded to
428 suppress false negative in conf_sanity test 39
430 Severity : enhancement
432 Description: build kernel-ib packages for OFED 1.3 in our release cycle
437 Description: fix SLES kernel versioning
438 Details : the kernel version for our SLES 10 kernel did not include a "-"
439 before the "smp" at the end. while this was not a problem in
440 general, it did mean that software trying to use the kernel
441 version to try to detect a vendor specific kernel would fail.
442 this was most evident by the OFED build scripts.
446 Description: Don't update lov_desc members until making sure they are valid
447 Details : When updating lov_desc members via proc fs, need fix their
448 validities before doing the real update.
452 Description: don't put request into delay list while invalidate in flight.
453 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
454 import in flight. this produce timeout for invalidate and sometimes
455 can cause stale data.
459 Frequency : on ppc only
460 Description: not convert ost objects for directory because it's not exist.
461 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
465 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
467 Severity : enhancement
469 Description: Update to support 2.6.22.14 vanilla kernel.
473 Frequency : rare, on recovery
474 Description: read procfs can produce deadlock in some situation
475 Details : Holding lprocfs lock which send rpc can produce block for destroy
476 obd objects and this also block reconnect with -EALREADY. This
477 isn't fix all lprocfs bugs - but make it rare.
479 Severity : enhancement
481 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
484 Frequency : frequent on X2 node
486 Description: mdc_set_open_replay_data LBUG
487 Details : Set replay data for requests that are eligible for replay.
491 Description: lustre_mgs: operation 101 on unconnected MGS
492 Details : When MGC is disconnected from MGS long enough, MGS will evict the
493 MGC, and late on MGC cannot successfully connect to MGS and a lot
494 of the error messages complaining that MGS is not connected.
498 Frequency : on network error
499 Description: panic with double free request if network error
500 Details : mdc_finish_enqueue is finish request if any network error ocuring,
501 but it's true only for synchronus enqueue, for async enqueue
502 (via ptlrpcd) this incorrect and ptlrpcd want finish request
505 Severity : enhancement
507 Description: client-side metadata stat-ahead during readdir(directory readahead)
508 Details : perform client-side metadata stat-ahead when the client detects
509 readdir and sequential stat of dir entries therein
512 Frequency : on start mds
514 Description: Implement get_info(last_id) in obdfilter.
517 Frequency : occasional
519 Description: Correctly check stale fid, not start epoch if ost not support SOM
520 Details : open with flag O_CREATE need set old fid in op_fid3 because
521 op_fid2 was overwritten with new generated fid, but mds can answer
522 with one of these two fids and both is not stale. Setattr
523 incorrectly started an epoch and assume will be called
524 done_writing, but without SOM done_writing ever being called.
527 Frequency : rare, depends on device drivers and load
529 Description: MDS or OSS nodes crash due to stack overflow
530 Details : Code changes in 1.8.0 increased the stack usage of some functions.
531 In some cases, in conjunction with device drivers that use a lot
532 of stack the MDS (or possibly OSS) service threads could overflow
533 the stack. One change which was identified to consume additional
534 stack has been reworked to avoid the extra stack usage.
537 Frequency : occasional
539 Description: Do not fail import if osc_interpret_create gets -EAGAIN
540 Details : If osc_interpret_create got -EAGAIN it immediately exits and
541 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
542 oscc_has_objects and see OSC has no objests and call
543 oscc_internal_create to resend create request.
545 Severity : enhancement
547 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
549 Severity : enhancement
551 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
554 Frequency : very rare
556 Description: Fix replay if there is an un-replied request and open
557 Details : In some cases, older replay request will revert the
558 mcd->mcd_last_xid on MDS which is used to record the client's
561 Severity : enhancement
563 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
565 Severity : enhancement
567 Description: Add rhel5 support to HEAD.
569 Serverity : enhancement
571 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
576 Description: Don't allow skipping OSTs if index has been specified.
577 Details : Don't allow skipping OSTs if index has been specified, make
578 locking in internal create lots better.
582 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
583 Details : access to bitfield in structure is always rounded to long
584 and this produce problem with not atomic change any bit.
588 Description: Lustre make rpms failed.
589 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
590 builds Lustre rpms from tarball.
593 Frequency : rare on shutdown ost
595 Description: If llog cancel was not send before clean_exports phase, this can
596 produce deadlock in llog code.
597 Details : If llog thread has last reference to obd and call class_import_put
598 this produce deadlock because llog_cleanup_commit_master wait when
599 last llog_commit_thread exited, but this never success because was
600 called from llog_commit_thread.
604 Description: allow userland application know is lost one of stripes.
605 Details : fill lvb_blocks with error code on ost and return it to
606 application if error flag found.
610 Description: NULL lov_tgts causing MDS oops
611 Details : more safe checks for NULL lov_tgts for avoid oops.
613 Severity : enhancement
615 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
619 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
620 Details : If an OST is down the MDS will hang indefinitely in
621 obd_statfs_rqset() waiting for the statfs data. While for
622 MDS QOS usage of statfs, it should not stuck in waiting.
624 Severity : enhancement
626 Description: remote_acl support
627 Details : Support ACL-based permission check for remote user.
628 Support setfacl/getfacl for remote user with the utils
629 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
630 the system "{s,g}etfacl" utils.
632 Severity : enhancement
634 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
636 Severity : enhancement
638 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
642 Description: make mgs_setparam() handle fsname containing dash
643 Details : fsname containing a dash does not work with lctl conf_param
645 Severity : enhancement
647 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
649 Severity : enhancement
651 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
655 Frequency : rare, at shutdown
656 Description: access already free / zero obd_namespace.
657 Details : if client_disconnect_export was called without force flag set,
658 and exist connect request in flight, this can produce access to
659 NULL pointer (or already free pointer) when connect_interpret
660 store ocd flags in obd_namespace.
664 Frequency : only at startup
665 Description: not alloc memory with spinlock held.
666 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
667 if any spinlock held.
669 Severity : enhancement
671 Description: make lustre randomly fail allocating memory
672 Details : Make lustre randomly failed allocating memory for testing purpose.
674 Severity : enhancement
676 Description: lost problems with lov objid file
677 Details : Fixes some scability and access to not inited memory problems
678 in work with lov objdid file.
683 Description: lfs find does not continue on file error
684 Details : Continue other files processing when a file/dir is absent.
688 Description: Inconsistent usage of lustre_pack_reply()
689 Details : Standardize the usage of lustre_pack_reply() such that it
690 always generate a CERROR on failure.
693 Frequency : occasional
695 Description: MDS hang or stay in waiting lock
696 Details : If client receive lock with CBPENDING flag ldlm need send lock
697 cancel as separate rpc, to avoid situation when cancel request
698 can't processed due all i/o threads stay in wait lock.
702 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
706 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
710 Description: Update to SLES9 kernel-2.6.5-7.287.3.
714 Description: Update to RHEL4 latest kernel.
716 Severity : enhancement
718 Description: Build SLES10 patchless client fails
719 Details : The configure was broken by run ./configure with
720 --with-linux-obj=.... argument for patchless client. When the
721 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
722 can't search header adequately. Use absolute path such as
723 -I($LINUX)/include instead.
727 Description: interrupt oig_wait produce painc on resend.
728 Details : brw_redo_request can be used for resend requests from ptlrpcd and
729 private set, and this produce situation when rq_ptlrpcd_data not
730 copyed to new allocated request and triggered LBUG on assert
731 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
732 ptlrpcd set if request is changed and can be safety changed to use
737 Description: LASSERT_{REQ,REP}SWAB macros are buggy
738 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
739 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
740 these macros and replace them with their difinitions instead.
744 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
745 Details : Update which_patch & target file for SLES10 latest kernel.
748 Description: Remove client patches from SLES 10 kernel.
749 Details : This causes SLES 10 clients to behave as patchless clients
750 even on a Lustre-patched (server) kernel.
752 Severity : enhancement
754 Description: self-adjustable client's lru lists
755 Details : use adaptive algorithm for managing client cached locks lru
756 lists according to current server load, other client's work
757 pattern, memory activities, etc. Both, server and client
758 side namespaces provide number of proc tunables for controlling
761 Severity : enhancement
763 Description: light-weight GSS support
764 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
765 and reduce performance overhead.
767 Severity : enhancement
769 Description: Linux keyring support
770 Details : Support using service of Linux keyring for Lustre GSS internal
771 context refresh/cache mechanism.
775 Description: Fix errors in lfs documentation
776 Details : Fixes man pages
780 Description: don't use GFP_* in generic Lustre code.
781 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
785 Description: obdclass is limited by single OBD_ALLOC(idarray)
786 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
790 Description: warnings with build patchless client with vanila 2.6.19 and up
791 Details : change old ctl_table style and replace ctl_table/ctl_table_header
792 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
796 Frequency : Only for SLES
797 Description: sanity_quota fail test_1
798 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
799 file, which makes TSTID[2] inunique.
803 Frequency : Only for SLES9
804 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
805 Details : Update target/ChangeLog/which_patch .
809 Frequency : Only for test_10 in sanity-quota.sh
810 Description: a bug in quota test code
811 Details : checking a test flag in a wrong place
815 Frequency : Only for test_13 in sanity-quota.sh
816 Description: a bug in quota test script
817 Details : shouldn't check overfull allocated quota
821 Description: jbd statistics
822 Details : Port older jbd statistics patch for sles10
824 Severity : enhancement
826 Description: Add printing inode info into message about error in writepage.
831 Description: reply_lock_interpret crash due to race with it and lock cancel.
832 Details : Do not replay locks that are being cancelled. Do not reference
833 locks by their address during replay, just by their handle.
837 Frequency : When flocks are used.
838 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
839 Details : Flock locks might destroy just granted lock if it could be merged
840 with another existing flock, this is done in completion handler,
841 so teach ldlm_cli_enquque_fini that this is a valid case for
847 Description: Oops in read and write path when failing to allocate lock.
848 Details : Check if lock allocation failed and return error back.
852 Description: lstripe command fails for valid OST index
853 Details : The stripe offset is compared to lov->desc.ld_tgt_count
854 instead of lov->desc.ld_active_tgt_count.
858 Description: sanity.sh failed test 103
859 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
864 Description: df doesn't work properly if diskfs blocksize != 4K
865 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
869 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
870 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
872 Severity : enhancement
874 Description: Adaptive timeouts
875 Details : RPC timeouts adapt to changing server load and network
876 conditions to reduce resend attempts and improve recovery time.
880 Description: llapi_file_create() does not allow some changes
881 Details : add llapi_file_open() that allows specifying the mode and
882 open flags, and also returns an open file handle.
886 Description: merge and cleanup kernel patches.
887 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
891 Description: Add journal checksum support.(Kernel part)
892 Details : The journal checksum feature adds two new flags i.e
893 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
894 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
895 indicates that the commit block contains the checksum for
896 the blocks described by the descriptor blocks. Now commit
897 record can be sent to disk without waiting for descriptor
898 blocks to be written to disk. This behavior is controlled
899 using JBD2_FEATURE_ASYNC_COMMIT flag.
903 Description: OSS needs mutliple precreate threads
904 Details : Add ability to start more than one create thread per OSS.
908 Description: Sanity.sh test_65k failed: setstripe should have succeeded
909 Details : execute lfs setstripe on client
913 Description: mds_obd_create error creating tmp object
914 Details : When the user sets quota on root, llog will be affected and can't
915 create files and write files.
918 Frequency : Always on ia64 patchless client, and possibly others.
920 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
921 Details : This allows the patchless client to be loaded on architectures
926 Description: RedHat Update kernel for RHEL5
927 Details : Add the kernel config file for RHEL5.
931 Description: RedHat Update kernel for RHEL5
932 Details : Modify the kernel config file more closer RHEL5.
936 Description: Build failure against Centos5 (RHEL5)
937 Details : Define PAGE_SIZE when it isn't present.
941 Description: test_7 on sanity-quota.sh can't be used on separate machines
942 Details : add facet to handle it
946 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
947 Details : with NFS, the anon dentry's parent was set to itself in
948 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
949 corresponding dentry other than use rec->ur_name.
951 Severity : enhancement
953 Description: lfs setstripe enhancement
954 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
958 Description: enable data checksumming by default
959 Details : enable checksum by default, allow --disable-checksum
960 configure option and "-o nochecksum" mount option. Checksums
961 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
965 Description: lustre support for RHEL5
966 Details : Add support for RHEL5.
970 Description: Client eviction due to failover config
971 Details : after a connection loss, the lustre client should attempt to
972 reconnect to the last active server first before trying the
973 other potential connections.
977 Description: when mds and osts use different quota unit(32bit and 64bit),
978 quota will be released repeatly.
979 Details : void sending multiple quota reqs to mds, which will keep the
980 status between the reqs.
984 Description: osts not allocated evenly to files
985 Details : change the condition to increase offset_idx
988 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
990 Description: Data corruption for OSTs that are formatted larger than 2TB
992 Details : When generating the bio request for lustre file writes the
993 sector number would overflow a temporary variable before being
994 used for the IO. The data reads correctly from Lustre (which
995 will overflow in a similar manner) but other file data or
996 filesystem metadata may be corrupted in some cases.
1000 Description: Tune the kernel for good SCSI performance.
1001 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1002 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1007 Description: rewrite ext2-derived code in obdclass/uuid.c
1008 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1009 from scratch preserving functionality.
1013 Description: "lfs find -obd UUID" prints directories
1014 Details : "lfs find -obd UUID" will return all directory names instead
1015 of just file names. It is incorrect because the directories
1016 do not reside on the OSTs.
1020 Description: use i_size_read and i_size_write in 2.6 port
1021 Details : replace inode->i_size access with i_size_read/write()
1024 Frequency : when removing large files
1026 Description: scheduling issue during removal of large Lustre files
1027 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1028 It causes scheduling issues when removing large files (17TB in the
1032 Frequency : only with liblustre clients on XT3
1034 Description: evictions taking too long
1035 Details : allow llrd to evict clients directly on OSTs
1038 Frequency : only on ppc
1040 Description: /proc/fs/lustre/devices broken on ppc
1041 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1042 We should make sure the type of 'index' is loff_t and then cast
1043 explicitly as needed below. Do not assign an explicitly cast
1047 Frequency : only for rhel5
1049 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1050 Details : Modify the target file & which_kernel.
1052 Severity : enhancement
1054 Description: omit set fsid for export NFS
1055 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1056 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1060 Frequency : Always, for kernels after 2.6.16
1061 Description: Fix warning idr_remove called for id=.. which is not allocated.
1062 Details : Last kernels save old s_dev before kill super and not allow
1063 to restore from callback - restore it before call kill_anon_super.
1067 Description: buffer overruns could theoretically occur
1068 Details : llapi_semantic_traverse() modifies the "path" argument by
1069 appending values to the end of the origin string, and a
1070 overrun may occur. Adding buffer overrun check in liblustreapi.
1074 Description: Fix error on 'ls .' at the top of the Lustre mount.
1075 Details : Don't revalidate dentry if it is a root dentry.
1079 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1080 Details : Modify vm-tunables-rhel4.patch.
1084 Description: Kernel config for 2.6.18-vanilla.
1085 Details : Modify targets/2.6-vanilla.target.in.
1086 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1087 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1088 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1089 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1093 Description: improve handling recoverable errors
1094 Details : if request processig with error which can be recoverable on server
1095 request should be resend, otherwise page released from cache and
1100 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1101 Details : Modify target file & which_patch.
1102 A flaw was found in the IA32 system call emulation provided
1103 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1104 value could be stored in the %RAX register, which could trigger an
1105 out-of-bounds system call table access. An untrusted local user
1106 could exploit this flaw to run code in the kernel
1107 (ie a root privilege escalation). (CVE-2007-4573).
1111 Description: O_DIRECT bypasses client statistics.
1112 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1113 read_ahead_stats were updated. copied the stats section from
1114 osc_send_oap_rpc() into async_internal().
1118 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1122 Description: change order of libsysio includes
1123 Details : '#include sysio.h' should always come before '#include xtio.h'
1127 Description: handle "serious error: objid * already exists" more gracefully
1128 Details : If LAST_ID value on disk is smaller than the objects existing in
1129 the O/0/d* directories, it indicates disk corruption and causes an
1130 LBUG(). If the object is 0-length, then we should use the existing
1131 object. This will help to avoid a full fsck in most cases.
1133 Severity : enhancement
1135 Description: adapt the lustre_config script to support the upgrade case
1136 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1137 server targets to 1.6 in parallel.
1141 Description: To avoid grant space > avaible space when the disk is almost
1142 full. Without this patch you might see the error "grant XXXX >
1143 available" or some LBUG about grant, when the disk is almost
1145 Details : In filter_check_grant, for non_grant cache write, we should
1146 check the left space by if (*left > ungranted + bytes), instead
1147 of (*left > ungranted), because only we are sure the left space
1148 is enough for another "bytes", then the ungrant space should be
1149 increase. In client, we should update cl_avail_grant only there
1150 is OBD_MD_FLGRANT in the reply.
1154 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1155 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1156 described in CVE-2007-4573. This problem could allow untrusted
1157 local users to gain root access.
1160 Frequency : when using O_DIRECT and quotas
1162 Description: Incorrect file ownership on O_DIRECT output files
1163 Details : block usage reported by 'lfs quota' does not take into account
1164 files that have been written with O_DIRECT.
1169 Description: touch file failed when fs is not full
1170 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1171 otherwise we can get ENOSP while fs is not full.
1175 Description: parallel lock callbacks
1176 Details : Instead of sending blocking and completion callbacks as separated
1177 requests, adding them to a set and sending in parallel.
1180 Frequency : only for Cray XT3
1181 Bugzilla : 12829/13455
1182 Description: Changing primary group doesn't change the group lustre assigns to
1184 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1185 overridden with the primary group provided by the group upcall,
1186 whereas the supplied fsgid can be trusted if it is in the list of
1187 supplementary groups returned by the group upcall.
1189 Severity : enhancement
1191 Description: Allow masking D_WARNING, D_ERROR messages from console
1192 Details : Console messages can now be disabled via lnet.printk.
1196 Description: User code with malformed file open parameter crashes client node
1197 Details : Before packing join_file req, all the related reference should be
1198 checked carefully in case some malformed flags cause fake
1199 join_file req on client.
1203 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1204 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1205 mode is equal to the granted mode. However, it should also check
1206 that LDLM_FL_CP_REQD is not set.
1210 Description: Heavy nfs access might result in deadlocks
1211 Details : After ELC code landed, it is now improper to enqueue any mds
1212 locks under och_sem, because enqueue might want to decide to
1213 cancel open locks for same inode we are holding och_sem for.
1217 Description: Client eviction while running blogbench
1218 Details : A lot of unlink operations with concurrent I/O can lead to a
1219 deadlock causing evictions. To address the problem, the number of
1220 oustanding OST_DESTROY requests is now throttled to
1221 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1222 callbacks are processed in priority.
1226 Description: enable ACLs on MDS by default
1227 Details : ACLs must be enabled on MDS by default.
1230 Frequency : PPC/PPC64 only
1232 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1233 Details : fix duplicated definitions between asm-ppc64/types.h and
1234 lustre_types.h on PPC.
1237 Frequency : PPC/PPC64 only
1239 Description: asm-ppc/segment.h does not exist
1240 Details : fix compile issue on PPC.
1244 Description: better handle error messages in extents code
1247 Frequency : RHEL4 only
1249 Description: mkfs is very slow on IA64/RHEL4
1250 Details : A performance regression has been discovered in the MPT Fusion
1251 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1252 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1253 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1255 Severity : enhancement
1257 Description: SNMP support enhancement
1258 Details : Adding total number of sampled request for an MDS node in snmp
1261 Severity : enhancement
1263 Description: Optimize ldlm waiting list processing for PR extent locks
1264 Details : When processing waiting list for read extent lock and meeting read
1265 lock that is same or wider to it that is not contended, skip
1266 processing rest of the list and immediatelly return current
1267 status of conflictness, since we are guaranteed there are no
1268 conflicting locks in the rest of the list.
1272 Description: Time out and refuse to reconnect
1273 Details : When the failover node is the primary node, it is possible
1274 to have two identical connections in imp_conn_list. We must
1275 compare not conn's pointers but NIDs, otherwise we can defeat
1276 connection throttling.
1280 Description: port llog fixes from b1_6 into HEAD
1281 Details : Port llog reference couting and some llog cleanups from b1_6
1282 (bug 10800) into HEAD, for protect from panic and access to already
1283 free llog structures.
1287 Description: Detect stride IO mode in read-ahead
1288 Details : When a client does stride read, read-ahead should detect that and
1289 read-ahead pages according to the detected stride pattern.
1293 Description: data checksumming impacts single node performance
1294 Details : add support for several checksum algorithm. Currently, only CRC32
1295 and Adler-32 are supported. The checksum type can be changed on
1296 the fly via /proc/fs/lustre/osc/*/checksum_type.
1300 Description: use adler32 for page checksums
1301 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1306 Description: build for x2 fails
1307 Details : fix compile issue on Cray systems.
1311 Description: Properly match for duplicate locks
1312 Details : Due to different lock order from skiplists code, we need to
1313 traverse entire list for now
1316 Frequency : only on PPC/SLES10
1318 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1319 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1320 be defined. Add a hack in mkfs_lustre.c to work around this compile
1325 Description: LASSERT on MDS when client holding flock lock dies
1326 Details : ldlm pool logic depends on number of granted locks equal to
1327 number of released locks which is not true for flock locks, so
1328 just exclude such locks from consideration.
1332 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1333 Details : Use fsfilt_sync() for both the whole filesystem sync and
1334 individual file sync to eliminate dangerous inode locking
1335 with I_LOCK that can lead to a deadlock.
1339 Description: performance in 1.6.3
1340 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1341 sure that 1MB requests are not fragmented by the block layer.
1343 Severity : enhancement
1345 Description: organize the server-side client stats on per-nid basis
1346 Details : Change the structure of stats under obdfilter and mds to
1354 The "uuid"s file would list the uuids of _active_ exports.
1355 And the clear entry is to clear all stats and stale nids.
1357 Severity : enhancement
1359 Description: eliminate client locks in face of contention
1360 Details : file contention detection and lockless i/o implementation
1361 for contended files.
1365 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1368 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1370 Description: MDS may lose file striping (and hence file data) in some cases
1371 Details : If there are additional extended attributes stored on the MDS,
1372 in particular ACLs, SELinux, or user attributes (if user_xattr
1373 is specified for the client mount options) then there is a risk
1374 of attribute loss. Additionally, the Lustre file striping
1375 needs to be larger than default (e.g. striped over all OSTs),
1376 and an additional attribute must be stored initially in the
1377 inode and then increase in size enough to be moved to the
1378 external attribute block (e.g. ACL growing in size) for file
1383 Description: skiplist implementation simplification
1384 Details : skiplists are used to group compatible locks on granted list
1385 that was implemented as tracking first and last lock of each
1386 lock group the patch changes that to using doubly linked lists
1390 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1391 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1392 because LASSERT on both the data supplied by a client, and the
1393 data on disk is dangerous and incorrect.
1395 Severity : enhancement
1397 Description: Slow truncate/writes to huge files at high offsets.
1398 Details : Directly associate cached pages to lock that protect those pages,
1399 this allows us to quickly find what pages to write and remove
1400 once lock callback is received.
1404 Description: more ldlm soft lockups
1405 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1406 starve other threads from the resource lock for a long time in
1407 case of long waiting queue, so change the debug level from
1408 D_OTHER to the less frequently used D_INFO.
1410 Severity : enhancement
1412 Description: add -gid, -group, -uid, -user options to lfs find
1416 Description: Hung threads in invalidate_inode_pages2_range
1417 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1418 once one is completed. As a result, some RPCs are stuck in the
1419 queue and are never sent.
1423 Description: filter threads hungs on waiting journal commit
1424 Details : Cleanup filter group llog code, then only filter group llog will
1425 be only created in the MDS/OST syncing process.
1429 Description: Procfs and llog threads access destoryed import sometimes.
1430 Details : Sync the import destoryed process with procfs and llog threads by
1431 the import refcount and semaphore.
1433 Severity : enhancement
1435 Description: openlock cache of b1_6 port to HEAD
1440 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1441 Details : If insertion of an extent fails, then discard the inode
1442 preallocation and free data blocks else it can lead to duplicate
1447 Description: don't always update ctime in ext3_xattr_set_handle()
1448 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1449 In some cases the ctime should not be updated, for example for
1450 2.0->1.8 compatibility it is necessary to delete an xattr and it
1451 should not update the ctime.
1455 Bugzilla : 15713/16362
1456 Description: Assertion in iopen_connect_dentry in 1.6.3
1457 Details : looking up an inode via iopen with the wrong generation number can
1458 populate the dcache with a disconneced dentry while the inode
1459 number is in the process of being reallocated. This causes an
1460 assertion failure in iopen since the inode's dentry list contains
1461 both a connected and disconnected dentry.
1465 Description: assertion failure in ldlm_handle2lock()
1466 Details : fix a race between class_handle_unhash() and class_handle2object()
1467 introduced in lustre 1.6.5 by bug 13622.
1472 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1473 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1478 Description: oops in sd_iostats_seq_show()
1479 Details : unloading/reloading the scsi low level driver triggers a kernel
1480 bug when trying to access the sd iostat file.
1485 Description: Kernel panics during QLogic driver reload
1486 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1487 patch, causing memory corruption.
1492 Description: journal_dev option does not work in b1_6
1493 Details : pass mount option during pre-mount.
1495 Severity : enhancement
1497 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1498 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1499 extent information of a file. It can be used to map logical blocks
1500 in a file to physical blocks in the block device.
1504 Description: LDLM soft lockups - improvement
1505 Details : It is be possible to send the lock handle along with each read
1506 or write request because the client is already doing a lock match
1507 itself so there isn't any reason the OST should have to re-do that
1511 Frequency : only on Cray X2
1513 Description: X2 build failures
1514 Details : fix build failures on Cray X2.
1518 Description: xid & resent requests
1519 Details : Initialize RPC XID from clock at startup (randomly if clock is
1522 Severity : enhancement
1524 Description: Add lustre_start utility to start or stop multiple Lustre servers
1529 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1530 Details : In case of memory pressure, list_del() can be called twice on
1531 req->rq_history_list, causing a kernel oops.
1535 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1536 Details : fix stack overflow in the distributed lock manager by defering
1537 export eviction after a failed AST to the elt thread instead of
1538 handling it in the dlm interpret routine.
1542 Description: Convert some comments to new format.
1543 Details : Update documenting comments to match doxygen conventions.
1547 Description: Grammar fixes.
1548 Details : A couple of trivial spelling fixes.
1552 Description: OSD_COUNTERS-mandatory
1553 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1554 to simplify the code.
1558 Description: simplify cmm_device freeing logic.
1559 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1563 Description: Add lockdep support to dt_object_operations locking interface.
1564 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1565 parameter indicating lock ordering. Update mdd code to use new
1570 Description: Introduce failloc constants for lockless IO tests.
1571 Details : Add two new failloc constants to test lockless IO. Only one of
1572 them in implemented---another is checked in yet to be landed
1577 Description: Add lockdep support for inode mutex.
1578 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1579 in the situations where Lustre has to lock more than one inode
1584 Description: Add optional invariants checking support.
1585 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1586 configure switch. This macro is to be used for consistency and
1587 sanity checks that are too expensive to be left in `production'
1592 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1593 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1598 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1599 Details : Use spin_lock_nested() in (the only) situation where more than
1600 one ldlm_lock is locked simultaneously. Also, fix possible
1601 dead-lock in ldlm_lock_change_resource() by enforcing particular
1606 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1607 Details : Instead of specifying each ldlm_lock call-back through separate
1608 parameter, wrap them into struct ldlm_callback_suite.
1612 Description: Kill join_lru obd method and its callers.
1613 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1614 in memory---a requirement implemented through ->o_join_lru() obd
1615 method in HEAD. Remove this method and its users.
1619 Description: Add asynchronous ldlm ENQUEUE completion handler.
1620 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1621 case that stock ldlm_completion_ast() cannot handle as it waits
1622 until lock is granted. Introduce new ldlm_completion_ast_async()
1623 for this. Also comment ldlm_completion_ast().
1627 Description: ldlm_error <-> errno conversion.
1628 Details : Add functions to map (rather arbitrary) between LDLM error codes
1629 and standard errno values. CLIO needs this to prevent LDLM specific
1630 constants from escaping ldlm and osc.
1634 Description: Kill unused ldlm_handle2lock_ns() function.
1635 Details : Kill unused ldlm_handle2lock_ns() function.
1639 Description: Add lu_ref support to ldlm_lock
1640 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1641 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1642 and ldlm_resource. LDLM interface has to be changed, because code that
1643 releases a reference on a lock, has to "know" what reference this is.
1644 In the most frequent case
1646 lock = ldlm_handle2lock(handle);
1648 LDLM_LOCK_PUT(lock);
1650 no changes are required. When any other reference (received _not_ from
1651 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1652 instead of LDLM_LOCK_PUT().
1654 Arguably, changes are pervasive, and interface requires some discipline
1655 for proper use. On the other hand, it was very instrumental in finding
1656 a few leaked lock references.
1660 Description: Add ldlm_lock_addref_try().
1661 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1662 attempts to addref a lock that might be being canceled
1667 Description: Add ldlm_weigh_callback().
1668 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1669 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1670 hard-coded `number of pages' logic.
1674 Description: Add lockdep annotations to llog code.
1675 Details : Use appropriately tagged _nested() locking calls in the places
1676 where llog takes more than one ->lgh_lock lock.
1680 Description: Add loi_kms_set().
1681 Details : Wrap kms updates into a helper function.
1685 Description: Constify instances of struct lsm_operations.
1686 Details : Constify instances of struct lsm_operations.
1690 Description: lu_conf support.
1691 Details : On a server, a file system object is uniquely identified
1692 by a fid, which is sufficient to locate and load all object
1693 state (inode). On a client, on the other hand, more data are
1694 necessary instantiate an object. Change lu_object_find() and
1695 friends to take additional `lu_conf' argument describing object.
1696 Typically this includes layout information.
1700 Description: lu_context fixes.
1701 Details : Introduce new lu_context functions that are needed on the client
1702 side, where some system threads (ptlrpcd) are shared by multiple
1703 modules, and so cannot be stopped during module shutdown.
1707 Description: Add start and stop methods to lu_device_type_operations.
1708 Details : Introduce two new methods in lu_device_type_operations, that are
1709 invoked when first instance of a given type is created and last one
1710 is destroyed respectively. This is need by CLIO.
1714 Description: Add lu_ref support to struct lu_device.
1715 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1716 track leaked references.
1720 Description: Introduce lu_kmem_descr.
1721 Details : lu_kmem_descr and its companion interface allow to create
1722 and destroy a number of kmem caches at once.
1726 Description: Fix lu_object finalization race.
1727 Details : Fix a race between lu_object_find() finding an object and its
1728 concurrent finalization. This race is (most likely) not possible
1729 on the server, but might happen on the client.
1733 Description: Introduce lu_ref interface.
1734 Details : lu_ref is a debugging module allowing to track references to
1735 a given object. It is quite cpu expensive, and has to be
1736 explicitly enabled with --enable-lu_ref. See usage description
1741 Description: Factor lu_site procfs stats into a separate function.
1742 Details : Separate lu_site stats printing code into a separate function
1743 that can be reused on a client.
1747 Description: Constify instances of struct {lu,dt,md}_device_operations.
1748 Details : Constify instances of struct {lu,dt,md}_device_operations.
1752 Description: Introduce struct md_site and move meta-data specific parts of
1753 struct lu_site here.
1754 Details : Move md-specific fields out of struct lu_site into special struct
1755 md_site, so that lu_site can be used on a client.
1759 Description: Kill mdd_lov_destroy().
1760 Details : Remove unused mdd code.
1764 Description: Add st_block checking to multistat.c.
1765 Details : Add st_block checking to multistat.c.
1769 Description: Add lu_ref support to struct obd_device.
1770 Details : Add lu_ref tracking to obd_device.
1774 Description: Kill obd_set_fail_loc().
1775 Details : Remove unused code.
1779 Description: Add special type for ptlrpc_request interpret functions.
1780 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1781 there. Actual usage will be in CLIO.
1785 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1786 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1787 doesn't work with lockdep.
1791 Description: Add rwv.c test program.
1792 Details : New testing program exercising readv(2) and writev(2) (Qian).
1796 Description: Add sendfile.c test program.
1797 Details : New testing program exercising sendfile(2) (Jay).
1801 Description: Ratelimit a message that can be very frequent.
1802 Details : Ratelimit a memory allocation failure message that can
1807 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1808 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1812 Description: Liblustre build fixes.
1813 Details : Liblustre build fixes.
1817 Description: libcfs: add cfs_{need,cond}_resched() interface.
1818 Details : libcfs: add cfs_{need,cond}_resched() definition and
1819 implementations for Linux, NT, and liblustre.
1821 Severity : enhancement
1823 Description: More exported tunables for mballoc
1824 Details : Add support for tunable preallocation window and new tunables for
1825 large/small requests
1829 Description: Detect corruption of block bitmap and checking for preallocations
1830 Details : Checks validity of on-disk block bitmap. Also it does better
1831 checking of number of applied preallocations. When corruption is
1832 found, it turns filesystem readonly to prevent further corruptions.
1836 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1837 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1838 window to 0. If it does hit cache miss, and read-pattern is still
1839 stride-io mode, does not reset the stride window, but also does not
1840 increase the stride window length in this case.
1844 Frequency : only for big-endian servers
1845 Description: Check if system is big-endian while mounting fs with extents feature
1846 Details : Mounting a filesystem with extents feature will fail on big-endian
1847 systems since ext3-based ldiskfs is not supported on big-endian
1848 systems. This can be over-riden with "bigendian_extents" mount option.
1850 Severity : enhancement
1852 Description: The root squash functionality
1853 Details : A security feature, which is to prevent users from being able
1854 to mount lustre on their desktop, run as root, and delete
1855 all of the files in the filesystem. The goal is accomplished by
1856 remapping user id (UID) and group id (GID) of the root user to
1857 a UID and GID specified by the system administartor via Lustre
1858 configuration management server (MGS). The functionality also
1859 allows to specify sets of clients for which the remapping does
1864 Description: Excessive recovery window
1865 Details : With AT enabled, the recovery window can be excessively long (6000+
1866 seconds). To address this problem, we no longer use
1867 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1868 timeout no longer depends on the service time, it is set to
1869 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1870 time via pb_service_time.
1872 --------------------------------------------------------------------------------
1874 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1876 * Support for kernels:
1877 2.6.5-7.283 (SLES 9),
1878 2.6.9-55.EL (RHEL 4),
1879 2.6.16.46-0.14 (SLES 10),
1880 2.6.18.8 vanilla (kernel.org)
1881 * Client support for unpatched kernels:
1882 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1883 2.6.16 - 2.6.22 vanilla (kernel.org)
1884 * Due to recently discovered recovery problems, we do not recommend
1885 using patchless RHEL 4 clients with this or any earlier release.
1886 * Recommended e2fsprogs version: 1.39.cfs8
1887 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1888 * Starting with this release, the ldiskfs backing filesystem required
1889 by Lustre is now in its own package, lustre-ldiskfs. This package
1890 should be installed. It is versioned separately from Lustre and
1891 may be released separately in future.
1895 Description: block reactivating mgc import until all deactivates complete
1896 Details : Fix race when failing back MDT/MGS to itself (testing)
1899 Frequency : at statup only
1901 Description: mds_lov_synchronize race leads to various problems
1902 Details : simultaneous MDT->OST connections at startup can cause the
1903 sync to abort, leaving the OSC in a bad state.
1905 Severity : enhancement
1907 Description: add optional extra BUILD_VERSION info
1908 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1909 to override the lustre version.
1912 Frequency : 2.6.18 servers only
1914 Description: ll_kern_mount() doesn't release the module reference
1915 Details : The ldiskfs module reference count never drops down to 0
1916 because ll_kern_mount() doesn't release the module reference.
1921 Description: server LBUG when using old ost_num_threads parameter
1922 Details : Accept the old ost_num_threads parameter but warn that it
1923 is deprecated, and fix an off-by-one error that caused an LBUG.
1928 Description: Transient SCSI error results in persistent IO issue
1929 Details : iobuf->dr_error is not reinitialized to 0 between two
1933 Frequency : sometimes when underlying device returns I/O errors
1935 Description: OSTs not going read-only during write failures
1936 Details : OSTs are not remounted read-only when the journal commit threads
1937 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1938 instead of the ext3 wrappers.
1942 Description: poor connect scaling with increasing client count
1943 Details : Don't run filter_grant_sanity_check for more than 100 exports
1944 to improve scaling for large numbers of clients.
1947 Frequency : SLES10 only
1949 Description: sanity-quota.sh quotacheck failed: rc = -22
1950 Details : Quotas cannot be enabled on SLES10.
1953 Frequency : liblustre clients only
1955 Description: getdirentries does not give error when run on compute nodes
1956 Details : getdirentries does not fail when the size specified as an argument
1957 is too small to contain at least one entry
1959 Severity : enhancement
1961 Description: Add LNET router traceability for debug purposes
1962 Details : If a checksum failure occurs with a router as part of the
1963 IO path, the NID of the last router that forwarded the bulk data
1964 is printed so it can be identified.
1969 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1970 Details : Due to a race condition, liblustre clients were occasionally
1971 evicted incorrectly.
1973 Severity : enhancement
1975 Description: lfs setstripe use optional parameters instead of postional
1978 Severity : enhancement
1980 Description: Nanosecond timestamp support for ldiskfs
1981 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1982 resolution timestamps. There is not yet support for this at
1983 the Lustre filesystem level.
1986 Frequency : during server recovery
1988 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1989 Details : request with rq_no_resend flag not awake l_wait_event if they get
1993 Frequency : nfs export on patchless client
1995 Description: connectathon hang when test nfs export over patchless client
1996 Details : Disconnected dentry cannot be found with lookup, so we do not need
1997 to unhash it or make it invalid
2000 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2001 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2002 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2007 Description: open req refcounting wrong on reconnect
2008 Details : If reconnect happened between getting open reply from server and
2009 call to mdc_set_replay_data in ll_file_open, we will schedule
2010 replay for unreferenced request that we are about to free.
2011 Subsequent close will crash in variety of ways.
2012 Check that request is still eligible for replay in
2013 mdc_set_replay_data().
2018 Description: disable writes to filesystem when reading health_check file
2019 Details : the default for reading the health_check proc file has changed
2020 to NOT do a journal transaction and write to disk, because this
2021 can cause reads of the /proc file to hang and block HA state
2022 checking on a healthy but otherwise heavily loaded system. It
2023 is possible to return to the previous behaviour during configure
2024 with --enable-health-write.
2026 Severity : enhancement
2028 Description: 64-bit inode version
2029 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2030 to the inode. This will be required for version-based recovery.
2035 Description: MDS fails to start if a duplicate client export is detected
2036 Details : in some rare cases it was possible for a client to connect to
2037 an MDS multiple times. Upon recovery the MDS would detect this
2038 and fail during startup. Handle this more gracefully.
2040 Severity : enhancement
2042 Description: Add -o localflock option to simulate old noflock
2044 Details : This will achieve local-only flock/fcntl locks
2050 Description: log_commit_thread vs filter_destroy race leads to crash
2051 Details : Take import reference before releasing llog record semaphore
2056 Description: Wrong request locking in request set processing
2057 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2058 imp_delayed_list, in this place should be used imp_lock.
2061 Frequency : when reconnection
2063 Description: Grant Leak when osc reconnect to OST
2064 Details : When osc reconnect ost, OST(filter) should check whether it
2065 should grant more space to client by comparing fed_grant and
2066 cl_avail_grant, and return the granted space to client instead
2067 of "new granted" space, because client will call osc_init_grant
2068 to update the client grant space info.
2071 Frequency : when client reconnect to OST
2073 Description: Grant Leak when osc do resend and replay bulk write
2074 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2075 bulk write request, because the grant info while be sync between
2076 OSC and OST when reconnect, and we should ignore the grant info
2077 these of resend/replay write req.
2082 Description: Grant space more than avaiable left space sometimes.
2083 Details : When then OST is about to be full, if two bulk writing from
2084 different clients came to OST. Accord the avaliable space of the
2085 OST, the first req should be permitted, and the second one
2086 should be denied by ENOSPC. But if the seconde arrived before
2087 the first one is commited. The OST might wrongly permit second
2088 writing, which will cause grant space > avaiable space.
2091 Frequency : when client is evicted
2093 Description: Grant might be wrongly erased when osc is evicted by OST
2094 Details : when the import is evicted by server, it will fork another
2095 thread ptlrpc_invalidate_import_thread to invalidate the
2096 import, where the grant will be set to 0. While the original
2097 thread will update the grant it got when connecting. So if
2098 the former happened latter, the grant will be wrongly errased
2099 because of this race.
2104 Description: Checking Stale with correct fid
2105 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2106 to check whether it is stale, because sometimes, we want the
2107 enqueue happened anyway, and op_data.fid2 will not be initialized.
2109 Severity : enhancement
2111 Description: update patchless client
2112 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2115 Frequency : only with 2.4 kernel
2117 Description: random memory corruption
2118 Details : size of struct ll_inode_info is to big for union inode.u and this
2119 can be cause of random memory corruption.
2124 Description: Memory leak in recovery
2125 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2126 It should also check obd_fail before fsfilt_start, otherwise if
2127 fsfilt_start return -EROFS,(failover mds during mds recovery).
2128 then the req will return with repmsg->transno = 0 and rc = EROFS.
2129 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2130 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2131 be freed no matter whether fsfilt_commit success or not.
2134 Frequency : only with huge count clients
2136 Description: Prevents from taking the superblock lock in llap_from_page for
2138 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2139 for llap_from_page call in ll_removepage prevents from taking the
2140 superblock lock for a soon died page.
2145 Description: Not check open intent error before release open handle
2146 Details : in some rare cases, the open intent error is not checked before
2147 release open handle, which may cause
2148 ASSERTION(open_req->rq_transno != 0), because it tries to release
2149 the failed open handle.
2154 Description: Set cat log bitmap only after create log success.
2155 Details : in some rare cases, the cat log bitmap is set too early. and it
2156 should be set only after create log success.
2160 Description: Accessing a block bevice can re-enable I/O when Lustre is
2161 tearing down a device.
2162 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2167 Description: service threads may hog cpus when there are a lot of requests
2168 Details : Insert cond_resched to give other threads a chance to use some CPU
2173 Description: the cat log was not initialized in recovery
2174 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2175 unlink log on mds will not be initialized until mds post
2176 recovery. And also in mds post recovery, the unlink log will
2177 initialization will be done asynchronausly, so there will be race
2178 between add unlink log and unlink log initialization.
2182 Description: brw_stats were being printed incorrectly
2183 Details : brw_stats were being printed as log2 but all of them were not
2184 recorded as log2. Also remove some code duplication arising from
2185 filter_tally_{read,write}.
2189 Frequency : rare, only in recovery.
2190 Description: ASSERTION(req->rq_type != LI_POISON) failed
2191 Details : imp_lock should be held while iterating over imp_sending_list for
2192 prevent destroy request after get timeout in ptlrpc_queue_wait.
2196 Description: replay-single.sh test 52 fails
2197 Details : A lock's skiplist need to be cleanup when it being unlinked
2198 from its resource list.
2202 Description: Short directio read returns full requested size rather than
2204 Details : Direct I/O operations should return actual amount of bytes
2205 transferred rather than requested size.
2207 Severity : enhancement
2209 Description: metadata RPC reduction (e.g. for rm performance)
2210 Details : decrease the amount of synchronous RPC between clients and servers
2211 by canceling conflicing lock before the operation on the client
2212 and packing thier handles into the main operation RPC to server.
2214 Severity : enhancement
2216 Description: Async OSC create to avoid the blocking unnecessarily.
2217 Details : If a OST has no remain object, system will block on the creating
2218 when need to create a new object on this OST. Now, ways use
2219 pre-created objects when available, instead of blocking on an
2220 empty osc while others are not empty. If we must block, we block
2221 for the shortest possible period of time.
2225 Description: improve handling recoverable errors
2226 Details : if request processig with error which can be recoverable on server
2227 request should be resend, otherwise page released from cache and
2230 Severity : enhancement
2232 Description: refine locking for avoid write wrong info into lov_objid file
2233 Details : fix possible races with add new target and write/update data in
2236 --------------------------------------------------------------------------------
2238 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2243 Frequency : on some architectures
2245 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2246 Details : Uninitialized flags sometimes cause configuration commands to
2250 Frequency : patchless clients only
2252 Description: missing __iget() symbol export
2253 Details : The __iget() symbol export is missing. To avoid the need for
2254 this on patchless clients the deathrow inode reaper is turned
2255 off, and we depend on the VM to clean up old inodes. This
2256 dependency was during via the fix for bug 12181.
2258 --------------------------------------------------------------------------------
2260 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2262 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2263 INTEROPERATE with older versions automatically. In many cases a
2264 special upgrade step is needed. Please read the
2265 user documentation before upgrading any part of a live system.
2266 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2267 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2268 * WARNING: Lustre configuration and startup changes are required with
2269 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2273 Severity : enhancement
2275 Description: MountConf
2276 Details : Lustre configuration is now managed via mkfs and mount
2277 commands instead of lmc and lconf. New obd types (MGS, MGC)
2278 are added for dynamic configuration management. See
2279 https://mail.clusterfs.com/wikis/lustre/MountConf for
2282 Severity : enhancement
2284 Description: dynamic OST addition
2285 Details : OSTs can now be added to a live filesystem
2287 Severity : enhancement
2289 Description: startup order invariance
2290 Details : MDTs and OSTs can be started in any order. Clients only
2291 require the MDT to complete startup.
2293 Severity : enhancement
2295 Description: parallel, asynchronous orphan cleanup
2296 Details : orphan cleanup is now performed in separate threads for each
2297 OST, allowing parallel non-blocking operation.
2299 Severity : enhancement
2301 Description: optimized stripe assignment
2302 Details : stripe assignments are now made based on ost space available,
2303 ost previous usage, and OSS previous usage, in order to try
2304 to optimize storage space and networking resources.
2306 Severity : enhancement
2308 Description: Permanently set tunables
2309 Details : All writable /proc/fs/lustre tunables can now be permanently
2310 set on a per-server basis, at mkfs time or on a live system.
2312 Severity : enhancement
2314 Description: Lustre message v2
2315 Details : Add lustre message format v2.
2317 Severity : enhancement
2319 Description: client OST exclusion list
2320 Details : Clients can be started with a list of OSTs that should be
2321 declared "inactive" for known non-responsive OSTs.
2325 Description: SPEC SFS validation failure on NFS v2 over lustre.
2326 Details : Changes the blocksize for regular files to be 2x RPC size,
2327 and not depend on stripe size.
2329 Severity : enhancement
2331 Description: Multiple MD RPCs in flight.
2332 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2333 To avoid overly-overloading MDS, introduce a limit on number of
2334 MDS RPCs in flight for a single client and add /proc controls
2335 to adjust this limit.
2337 Severity : enhancement
2339 Description: client read/write statistics
2340 Details : Add client read/write call usage stats for performance
2341 analysis of user processes.
2342 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2343 file access. extents_stats shows chunk size distribution.
2344 extents_stats_per_process show chunk size distribution per
2347 Severity : enhancement
2349 Description: mds statistics
2350 Details : Add detailed mds operations statistics in
2351 /proc/fs/lustre/mds/*/stats.
2355 Description: Failure of copying files with lustre special EAs.
2356 Details : Client side always return success for setxattr call for lustre
2357 special xattr (currently only "trusted.lov").
2362 Description: Refcount LNET uuids
2363 Details : The global LNET uuid list grew linearly with every startup;
2364 refcount repeated list entries instead of always adding to
2367 Severity : enhancement
2369 Description: Dynamic service threads
2370 Details : Within a small range, start extra service threads
2371 automatically when the request queue builds up.
2374 Frequency : mixed-endian client/server environments
2376 Description: mixed-endian crashes
2377 Details : The new msg_v2 system had some failures in mixed-endian
2380 Severity : enhancement
2382 Description: Easy OST removal
2383 Details : OSTs can be permanently deactivated with e.g. 'lctl
2384 conf_param lustre-OST0001.osc.active=0'
2386 Severity : enhancement
2388 Description: MGS proc entries
2389 Details : Added basic proc entries for the MGS showing what filesystems
2392 Severity : enhancement
2394 Description: provide MGS failover
2395 Details : Added config lock reacquisition after MGS server failover.
2397 Severity : enhancement
2399 Description: add Linux 2.4 support
2400 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2404 Description: a large application tries to do I/O to the same resource and dies
2405 in the middle of it.
2406 Details : Check the req->rq_arrival time after the call to
2407 ost_brw_lock_get(), but before we do anything about
2408 processing it & sending the BULK transfer request. This
2409 should help move old stale pending locks off the queue as
2410 quickly as obd_timeout.
2413 Frequency : when an incorrect nid is specified during startup
2415 Description: ptlrpc connect to non-existant node causes kernel crash
2416 Details : LNET can't be re-entered from an event callback, which
2417 happened when we expire a message after the export has been
2418 cleaned up. Instead, hand the zombie cleanup off to another
2421 Severity : enhancement
2423 Description: plain/inodebits lock performance improvement
2424 Details : Grouping plain/inodebits in granted list by their request modes
2425 and bits policy, thus improving the performance of search through
2429 Frequency : only if OST filesystem is corrupted
2431 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2432 Details : for a short time RPCs with bulk IO are in the replay list,
2433 but replay of bulk IOs is unimplemented. If the OST filesystem
2434 is corrupted due to disk cache incoherency and then replay is
2435 started it is possible to trip an assertion. Avoid putting
2436 committed RPCs into the replay list at all to avoid this issue.
2439 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2441 Description: System hang on startup
2442 Details : This bug allowed the liblustre (e.g. catamount) client to
2443 return to the app before handling all startup RPCs. This
2444 could leave the node unresponsive to lustre network traffic
2445 and manifested as a server ptllnd timeout.
2447 Severity : enhancement
2449 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2450 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2451 boolean to control whether to print peer debug info when a
2452 client's RPC times out.
2455 Frequency : only for kernels with patches from Lustre below 1.4.3
2457 Description: Remove old rdonly API
2458 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2461 Frequency : only for devices with external journals
2463 Description: Set external device read-only also
2464 Details : During a commanded failover stop, we set the disk device
2465 read-only while the server shuts down. We now also set any
2466 external journal device read-only at the same time.
2469 Frequency : when upgrading from 1.4 while trying to change parameters
2471 Description: The wrong (new) MDC name was used when setting parameters for
2472 upgraded MDT's. Also allows changing of OSC (and MDC)
2473 parameters if --writeconf is specified at tunefs upgrade time.
2476 Frequency : when setting specific ost indicies
2478 Description: QOS code breaks on skipped indicies
2479 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2480 created with --index need not be sequential.
2482 Severity : enhancement
2484 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2485 Details : The uninit_groups feature works in conjunction with the kernel
2486 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2487 up the pass1 processing of e2fsck. This is a read-only feature
2488 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2489 mount filesystems that have had this feature enabled.
2491 Severity : enhancement
2493 Description: Improve multi-block allocation algorithm to avoid fragmentation
2494 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2495 allocation locality and avoid filesystem fragmentation.
2497 ------------------------------------------------------------------------------
2499 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2501 * Support for kernels:
2502 2.4.21-47.0.1.EL (RHEL 3)
2503 2.6.5-7.283 (SLES 9)
2504 2.6.9-42.0.10.EL (RHEL 4)
2505 2.6.12.6 vanilla (kernel.org)
2506 2.6.16.27-0.9 (SLES 10)
2507 * Recommended e2fsprogs version: 1.39.cfs5
2509 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2513 Frequency : occasional, depends on client load and configuration
2514 Bugzilla : 12181, 12203
2515 Description: data loss for recently-modified files
2517 Details : In some cases it is possible that recently written or created
2518 files may not be written to disk in a timely manner (this should
2519 normally be within 30s unless client IO load is very high).
2520 The problem appears as zero-length files or files that are a
2521 multiple of 1MB in size after a client crash or client eviction
2522 that are missing data at the end of the file.
2524 This problem is more likely to be hit on clients where files are
2525 repeatedly created and unlinked in the same directory, clients
2526 have a large amount of RAM, have many CPUs, the filesystem has
2527 many OSTs, the clients are rebooted frequently, and/or the files
2528 are not accessed by other nodes after being written.
2530 The presence of the problem can be detected by looking at
2531 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2532 smaller than the second (nr_unused) then dirty files will not
2533 be flushed automatically to disk. "sync; sleep 10" should be
2534 run several times on the node before unmounting it to update
2535 Lustre (this is also safe to run on nodes without this problem).
2537 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2538 that can cause this same problem, so customers using that kernel
2539 also need to update the kernel in addition to Lustre. In order
2540 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2542 It is normal that files written just before a client crash (less
2543 than 30s) may not yet have been flushed to disk, even for local
2547 Frequency : frequent on thin XT3 nodes
2549 Description: UUID collision on thin XT3 Linux nodes
2550 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2551 randomly, since we relied on an insufficiently-seeded PRNG.
2556 Description: OSS hangs after "All ost request buffers busy"
2557 Details : A deadlock between quota and journal operations caused OSS
2558 hangs after printing "All ost request buffers busy."
2561 Frequency : always on liblustre builds
2563 Description: Cleanup compiler warnings on liblustre
2566 Frequency : always on liblustre builds on XT3
2568 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2572 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2573 > 32000 subdirectories
2574 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2575 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2576 subdirectory count crosses 32000. This will aid e2fsck to
2577 correctly handle more than 32000 subdirectories.
2580 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2582 Description: System hang on startup
2583 Details : This bug allowed the liblustre (e.g. catamount) client to
2584 return to the app before handling all startup RPCs. This
2585 could leave the node unresponsive to lustre network traffic
2586 and manifested as a server ptllnd timeout.
2588 Severity : enhancement
2590 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2591 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2592 boolean to control whether to print peer debug info when a
2593 client's RPC times out.
2598 Description: make O_SYNC working on 2.6 kernels
2599 Details : 2.6 kernels use different method for mark pages for write,
2600 so need add a code to lustre for O_SYNC work.
2605 Description: Failure to close file and release space on NFS
2606 Details : Put inode details into lock acquired in ll_intent_file_open.
2607 Use mdc_intent_lock in ll_intent_open to properly
2608 detect all kind of errors unhandled by mdc_enqueue.
2613 Description: proc file read during shutdown sometimes raced obd removal,
2615 Details : Add lock to prevent obd access after proc file removal.
2618 Frequency : Only for files larger than 4GB on 32-bit clients.
2620 Description: improperly doing page alignment of locks
2621 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2622 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2627 Description: RPCs being resent when they shouldn't be
2628 Details : Some RPCs that should not be resent are being resent. This
2629 can cause inconsistencies in the RPC state machine. Do not
2630 resend such requests.
2633 Frequency : rare, only with NFS export
2635 Description: Crash on NFS re-export node
2636 Details : under very unusual load conditions an assertion is hit in
2637 ll_intent_file_open()
2640 Frequency : only if OST filesystem is corrupted
2642 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2643 Details : for a short time RPCs with bulk IO are in the replay list,
2644 but replay of bulk IOs is unimplemented. If the OST filesystem
2645 is corrupted due to disk cache incoherency and then replay is
2646 started it is possible to trip an assertion. Avoid putting
2647 committed RPCs into the replay list at all to avoid this issue.
2652 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2653 Details : Large single O_DIRECT read and write calls can fail to allocate
2654 a sufficiently large buffer to process the request. In case of
2655 allocation failure the allocation is retried with a smaller
2656 buffer and broken into smaller requests.
2658 Severity : enhancement
2660 Description: Add -o localflock option to simulate old noflock behaviour.
2661 Details : This will achieve local-only flock/fcntl locks coherentness.
2666 Description: versioning check is incomplete
2667 Details : Checking the version difference of client vs. server, report
2668 error if the gap is too big.
2673 Description: add support PG_writeback bit
2674 Details : add support for PG_writeback bit for Lustre, for more carefull
2675 work with page cache in 2.6 kernel. This also fix some deadlocks
2676 and remove hack for work O_SYNC with 2.6 kernel.
2678 Severity : enhancement
2680 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2681 Details : The uninit_groups feature works in conjunction with the kernel
2682 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2683 up the pass1 processing of e2fsck. This is a read-only feature
2684 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2685 mount filesystems that have had this feature enabled.
2687 Severity : enhancement
2689 Description: Improve multi-block allocation algorithm to avoid fragmentation
2690 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2691 allocation locality and avoid filesystem fragmentation.
2693 ------------------------------------------------------------------------------
2695 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2697 * Support for kernels:
2698 2.6.9-42.0.3.EL (RHEL 4)
2699 2.6.5-7.276 (SLES 9)
2700 2.4.21-47.0.1.EL (RHEL 3)
2701 2.6.12.6 vanilla (kernel.org)
2702 2.6.16.21-0.8 (SLES10)
2703 * Recommended e2fsprogs version: 1.39.cfs2-0
2705 * The backwards-compatible /proc/sys/portals symlink has been removed
2706 in this release. Before upgrading, please ensure that you change
2707 any configuration scripts or /etc/sysctl.conf files that access
2708 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2709 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2710 in advance of the upgrade on any system running Lustre 1.4.6 or
2711 newer, since /proc/sys/lnet was added in that version.
2712 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2716 Frequency : only when quota is used
2718 Description: avoid scanning export list for quota master
2719 Details : Change the algorithms to avoid scanning export list in order
2720 to improve the efficiency.
2723 Frequency : MDS failover only, very rarely
2725 Description: "went back in time" messages on mds failover
2726 Details : The greatest transno may be lost when the current operation
2727 finishes with an error (transno==0) and the client's last_rcvd
2728 record is over-written. Save the greatest transno in the
2729 mds_last_transno for this case.
2732 Frequency : always for specific kernels and striping counts
2734 Description: client may get "Matching packet too big" without ACL support
2735 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2736 when trying to access files in certain configurations. The
2737 clients should in fact be denied when mounting because they do
2738 not understand ACLs.
2741 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2743 Description: many clients connecting with IO in progress causes connect timeouts
2744 Details : Avoid synchronous journal commits to avoid delays caused by many
2745 clients connecting/disconnecting when bulk IO is in progress.
2746 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2747 OST_IO_PORTAL to avoid delays behind potentially many pending
2751 Frequency : occasionally with multiple writers to a single file
2753 Description: shared writes to file may result in wrong size reported by stat()
2754 Details : Allow growing of kms when extent lock is cancelled
2757 Frequency : always with random mmap IO to multi-striped file
2759 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2760 Details : Only if the hole is at the end of OST object so that kms is too
2761 small. Fix is to increase kms accordingly in ll_nopage.
2764 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2766 Description: writes to a missing object would leak memory on the OST
2767 Details : If there is an inconsistency between the MDS and OST filesystems,
2768 such that the MDS references an object that doesn't exist, writes
2769 to that object will leak memory due to incorrect cleanup in the
2770 error handling path, eventually running out of memory on the OST.
2775 Description: Creating too long symlink causes lustre errors
2776 Details : Check symlink and name lengths before sending requests to MDS.
2779 Frequency : only if flock is enabled (not on by default)
2781 Description: posix locks not released on fd closure on 2.6.9+
2782 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2783 kernels, this caused such locks not to be released on fd close and
2784 then assertions on fs unmount about still used locks.
2787 Frequency : MDS failover only, very rarely
2789 Description: clients may get ASSERTION(granted_lock != NULL)
2790 Details : When request was taking a long time, and a client was resending
2791 a getattr by name lock request. The were multiple lock requests
2792 with the same client lock handle and
2793 mds_getattr_name->fixup_handle_for_resent_request found one of the
2794 lock handles but later failed with ASSERTION(granted_lock != NULL).
2799 Description: handle->h_buffer_credits > 0, assertion failure
2800 Details : h_buffer_credits is zero after truncate, causing assertion
2801 failure. This patch extends the transaction or creates a new
2805 Frequency : NFS re-export or patchless client
2806 Bugzilla : 11179, 10796
2807 Description: Crash on NFS re-export node (__d_move)
2808 Details : We do not want to hash the dentry if we don't have a lock.
2809 But if this dentry is later used in d_move, we'd hit uninitialised
2810 list head d_hash, so we just do this to init d_hash field but
2811 leave dentry unhashed.
2814 Frequency : NFS re-export or patchless client
2816 Description: NFS exports has problem with symbolic link
2817 Details : lustre client didn't properly install dentry when re-exported
2818 to NFS or running patchless client.
2821 Frequency : NFS re-export or patchless client
2823 Description: Various nfs/patchless fixes.
2824 Details : fixes reuse disconected alias for lookup process - this fixes
2825 warning "find_exported_dentry: npd != pd",
2826 fix permission error with open files at nfs.
2827 fix apply umask when do revalidate.
2830 Frequency : occasional
2832 Description: Crash on NFS re-export node
2833 Details : calling clear_page() on the wrong pointer triggered oops in
2834 generic_mapping_read().
2837 Frequency : rarely, using O_DIRECT IO
2839 Description: unaligned directio crashes client with LASSERT
2840 Details : check for unaligned buffers before trying any requests.
2843 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2845 Description: stale data returned from RAID cache
2846 Details : If only a small amount of IO is done to the RAID device before
2847 reading it again it is possible to get stale data from the RAID
2848 cache instead of reading it from disk.
2851 Frequency : always for sles10 kernel
2853 Description: sles10 support
2854 Details : ll_follow_link: compile fixes and using of nd_set_link
2855 under newer kernels.
2858 Frequency : depends on arch, kernel and compiler version, always on sles10
2861 Description: recursive or deep enough symlinks cause stack overflow
2862 Details : getting rid of large stack-allocated variable in
2866 Frequency : depends on hardware
2868 Description: lustre write performance loss in the SLES10 kernel
2869 Details : the performance loss is caused by using of write barriers in the
2870 ext3 code. The SLES10 kernel turns barrier support on by
2871 default. The fix is to undo that change for ldiskfs.
2873 ------------------------------------------------------------------------------
2875 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2877 * Support for kernels:
2878 2.6.9-42.0.3EL (RHEL 4)
2879 2.6.5-7.276 (SLES 9)
2880 2.4.21-47.0.1.EL (RHEL 3)
2881 2.6.12.6 vanilla (kernel.org)
2885 Frequency : quota enabled and large files being deleted
2887 Description: releasing more than 4GB of quota at once hangs OST
2888 Details : If a user deletes more than 4GB of files on a single OST it
2889 will cause the OST to spin in an infinite loop. Release
2890 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2895 Description: statfs data retrieved from /proc may be stale or zero
2896 Details : When reading per-device statfs data from /proc, in the
2897 {kbytes,files}_{total,free,avail} files, it may appear
2898 as zero or be out of date.
2901 Frequency : systems with MD RAID1 external journal devices
2903 Description: lconf's call to blkid is confused by RAID1 journal devices
2904 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2905 of returning all block devices that match the journal UUID.
2908 Frequency : always, for aggregate stripe size over 4GB
2910 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2911 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2912 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2915 Frequency : always if "lfs find" used on a local file/directory
2917 Description: "lfs find" segfaults if used on a local file/directory
2918 Details : The case where a directory component was not specified wasn't
2919 handled correctly. Handle this properly.
2922 Frequency : always on ppc64
2924 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2925 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2926 perform endianness conversion.
2929 Frequency : rarely (truncate to non-zero file size after write under load)
2930 Bugzilla : 10730, 10687
2931 Description: Files padded with zeros to next 4K multiple
2932 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2933 it is possible that files that are truncated to a non-zero size
2934 immediately after being written are filled with zero bytes beyond
2935 the truncated size. No file data is lost.
2937 Severity : enhancement
2939 Description: Allow recovery/failover for liblustre clients.
2940 Details : liblustre clients were unaware of failover configurations until
2943 Severity : enhancement
2945 Description: user file locks should fail when not mounting with flock option
2946 Details : Set up an error-returning stub in ll_file_operations.lock field
2947 to prevent incorrect behaviour when client is mounted without
2948 flock option. Also, set up properly f_op->flock field for
2952 Frequency : always on ia64
2954 Description: "lfs df" loops on printing out MDS statfs information
2955 Details : The obd_ioctl_data was not initialized and in some systems
2956 this caused a failure during the ioctl that did not return
2957 an error. Initialize the struct and return an error on failure.
2960 Frequency : SLES 9 only
2962 Description: Error of copying files with lustre special EAs as root
2963 Details : Client side always return success for setxattr call for lustre
2964 special xattr (currently only "trusted.lov").
2967 Frequency : rarely on clusters with both ia64+i386 clients
2969 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2970 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2971 clients are doing concurrent writes to the same file it is
2972 possible that the ia64 clients may LASSERT because the OST
2973 extent locks are not PAGE_SIZE aligned. Ensure that grown
2974 locks are always aligned on the request boundary.
2977 Frequency : specific use, occasional
2979 Description: Overwriting in use executable truncates on-disk binary image
2980 Details : If one node attempts to overwrite an executable in use by
2981 another node, we now correctly return ETXTBSY instead of
2982 truncating the file.
2984 Severity : enhancement
2986 Description: Async OSC create to avoid the blocking unnecessarily.
2987 Details : If a OST has no remain object, system will block on the creating
2988 when need to create a new object on this OST. Now, ways use
2989 pre-created objects when available, instead of blocking on an
2990 empty osc while others are not empty. If we must block, we block
2991 for the shortest possible period of time.
2996 Description: chmod on Lustre root is propagated to other clients
2997 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2998 invalid by the follow_mount time.
3001 Frequency : liblustre clients only
3003 Description: Race in 'instant cancel' lock handling could lead to such locks
3004 never to be granted in case of SMP MDS
3005 Details : Do not destroy not yet granted but cbpending locks in
3009 Frequency : replay/resend of open
3011 Description: non null lock assetion failure in mds_intent_policy
3012 Details : Trying to replay/resend lockless open requests resulted in
3013 mds_open() returning 0 with no lock. Now it sets a flag if
3014 a lock is going to be returned.
3016 Severity : enhancement
3018 Description: Checksum enhancements
3019 Details : New checksum enhancements allow for resending RPCs that failed
3022 Severity : enhancement
3024 Description: Tunables on number of dirty pages in cacche
3025 Details : Allow to set limit on number of dirty pages cached.
3030 Description: client crash on unmount - lock still has references
3031 Details : In some error handling cases it was possible to leak a lock
3032 reference on a client while accessing a file. This was not
3033 harmful to the client during operation, but would cause the
3034 client to crash when the filesystem is unmounted.
3037 Frequency : specific case, rare
3039 Description: ETXTBSY on mds though file not in use by client
3040 Details : ETXTBSY is no longer incorrectly returned when attempting to
3041 chmod or chown a directory that the user previously tried to
3042 execute or a currently-executing binary.
3045 Frequency : extremely rare except on liblustre-based clients
3047 Description: Lustre space not freed when files are deleted
3048 Details : Clean up open-unlinked files after client eviction. Previously
3049 the unlink was skipped and the files remained as orphans.
3054 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3055 Details : In some cases it was possible to send a blocking callback to
3056 a client doing a glimpse, even though that client didn't get
3057 a lock granted. When the glimpse lock is cancelled on the OST
3058 the freed lock is left on the waiting list and corrupted the list.
3061 Frequency : all core dumps
3063 Description: Broke core dumps to lustre
3064 Details : Negative dentry may be unhashed if parent does not have UPDATE
3065 lock, but some callers, e.g. do_coredump, expect dentry to be
3066 hashed after successful create, hash it in ll_create_it.
3068 ------------------------------------------------------------------------------
3070 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3072 * Support for kernels:
3073 2.6.9-42.0.2.EL (RHEL 4)
3074 2.6.5-7.276 (SLES 9)
3075 2.4.21-40.EL (RHEL 3)
3076 2.6.12.6 vanilla (kernel.org)
3080 Frequency : always on RHEL 3
3082 Description: Number of open files grows over time
3083 Details : The number of open files grows over time, whether or not
3084 Lustre is started. This was due to a filp leak introduced
3085 by one of our kernel patches.
3087 ------------------------------------------------------------------------------
3089 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3091 * Support for kernels:
3092 2.6.9-42.EL (RHEL 4)
3093 2.6.5-7.276 (SLES 9)
3094 2.4.21-40.EL (RHEL 3)
3095 2.6.12.6 vanilla (kernel.org)
3100 Bugzilla : 5719, 9635, 9792, 9684
3101 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3102 Details : If a server is under heavy load and cannot reply to new
3103 connection requests before the client resends the (re)connect,
3104 the connection handling code can behave badly if two service
3105 threads are concurrently handing separate (re)connections from
3106 the same client. Add better locking to the connection handling
3107 code, and ensure that only a single connection will be processed
3108 for a given client UUID, even if the lock is dropped.
3110 Severity : enhancement
3112 Description: add TCP zero-copy support to kernel
3113 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3114 sends if the hardware supports scatter-gather and checksumming.
3115 This allows socklnd to do client-write and server-read more
3116 efficiently and reduce CPU utilization from skbuf copying.
3119 Frequency : only if NFS exporting from client
3121 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3122 Details : If setting attributes on a file created under NFS that had
3123 never been opened it would be possible to oops the client
3124 if the file had no objects.
3127 Frequency : always for liblustre
3129 Description: liblustre client does MDS+OSTs setattr RPC for each write
3130 Details : When doing a write from a liblustre client, the client
3131 incorrectly issued an RPC to the MDS and each OST the file was
3132 striped over in order to update the timestamps. When writing
3133 with small chunks and many clients this could overwhelm the MDS
3134 with RPCs. In all cases it would slow down the write because
3135 these RPCs are unnecessary.
3137 Severity : enhancement
3139 Description: allow number of MDS service threads to be changed at module load
3140 Details : It is now possible to change the number of MDS service threads
3141 running. Adding "options mds mds_num_threads={N}" to the MDS's
3142 /etc/modprobe.conf will set the number of threads for the next
3143 time Lustre is restarted (assuming the "mds" module is also
3144 reloaded at that time). The default number of threads will
3145 stay the same, 32 for most systems.
3150 Description: OST crash if filesystem is unformatted or corrupt
3151 Details : If an OST is started on a device that has never been formatted
3152 or if the filesystem is corrupt and cannot even mount then the
3153 error handling cleanup routines would dereference a NULL pointer.
3158 Description: NULL pointer deref in llap_from_page.
3159 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3160 incorrect) mapping if the page was truncated/reclaimed while it was
3161 searched for. Check for this condition and skip such pages when
3162 doing readahead. Introduce extra check to llap_from_page() to
3163 verify page->mapping->host is non-NULL (so page is not anonymous).
3166 Frequency : Sometimes when using sys_sendfile
3168 Description: "page not covered by a lock" warnings from ll_readpage
3169 Details : sendfile called ll_readpage without right page locks present.
3170 Now we introduced ll_file_sendfile that does necessary locking
3171 around call to generic_file_sendfile() much like we do in
3175 Frequency : with certain MDS communication failures at client mount time
3177 Description: NULL pointer deref after failed client mount
3178 Details : a client connection request may delayed by the network layer
3179 and not be sent until after the PTLRPC layer has timed out the
3180 request. If the client fails the mount immediately it will try
3181 to clean up before the network times out the request. Add a
3182 reference from the request import to the obd device and delay
3183 the cleanup until the network drops the request.
3186 Frequency : occasionally during client (re)connect
3188 Description: assertion failure during client (re)connect
3189 Details : processing a client connection request may be delayed by the
3190 client or server longer than the client connect timeout. This
3191 causes the client to resend the connection request. If the
3192 original connection request is replied in this interval, the
3193 client may trip an assertion failure in ptlrpc_connect_interpret()
3194 which thought it would be the only running connect process.
3197 Frequency : only with obd_echo servers and clients that are rebooted
3199 Description: kernel BUG accessing uninitialized data structure
3200 Details : When running an obd_echo server it did not start the ping_evictor
3201 thread, and when a client was evicted an uninitialized data
3202 structure was accessed. Start the ping_evictor in the RPC
3203 service startup instead of the OBD startup.
3205 Severity : enhancement
3206 Bugzilla : 10193 (patchless)
3207 Description: Remove dependency on various unexported kernel interfaces.
3208 Details : No longer need reparent_to_init, exit_mm, exit_files,
3209 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3212 Frequency : rare (only users of deprecated and unsupported LDAP config)
3214 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3215 Details : LDAP apparently contains 'lustreName' attributes instead of
3216 'name'. A simple remapping of the name is sufficient.
3219 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3221 Description: waiting_locks_callback trips kernel BUG if client is evicted
3222 Details : Running with the dump_on_timeout debug flag turned on makes
3223 it possible that the waiting_locks_callback() can try to dump
3224 the Lustre kernel debug logs from an interrupt handler. Defer
3225 this log dumping to the expired_lock_main() thread.
3227 Severity : enhancement
3229 Description: Support NFS exporting on 2.6 kernels.
3230 Details : Implement non-rawops metadata methods for NFS server to use without
3231 changing NFS server code.
3234 Frequency : very rare (synthetic metadata workload only)
3236 Description: two racing renames might cause an MDS thread to deadlock
3237 Details : Running the "racer" program may cause one MDS thread to rename
3238 a file from being the source of a rename to being the target of
3239 a rename at exactly the same time that another thread is doing
3240 so, and the second thread has already enqueued these locks after
3241 doing a lookup of the target and is trying to relock them in
3242 order. Ensure that we don't try to re-lock the same resource.
3245 Frequency : only very large systems with liblustre clients
3247 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3248 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3249 all OSTs in parallel. This allows the eviction of stale liblustre
3250 clients to proceed much faster than if they were done in series,
3251 and also offers similar improvements for other set_info RPCs.
3256 Description: excessive CPU usage during initial read phase on client
3257 Details : During the initial read phase on a client, it would agressively
3258 retry readahead on the file, consuming too much CPU and impacting
3259 performance (since 1.4.5.8). Improve the readahead algorithm
3260 to avoid this, and also improve some other common cases (read
3261 of small files in particular, where "small" is files smaller than
3262 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3267 Description: MDS crash when receiving packet with unknown intent.
3268 Details : Do not LBUG in unknown intent case, just return -EFAULT
3270 Severity : enhancement
3271 Bugzilla : 9293, 9385
3272 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3273 Details : Do not serialize getattr (non-intent version) and statfs.
3276 Frequency : occasional, when OST network is overloaded/intermittent
3278 Description: client evicted by OST after bulk IO timeout
3279 Details : If a client sends a bulk IO request (read or write) the OST
3280 may evict the client if it is unresposive to its data GET/PUT
3281 request. This is incorrect if the network is overloaded (takes
3282 too long to transfer the RPC data) or dropped the OST GET/PUT
3283 request. There is no need to evict the client at all, since
3284 the pinger and/or lock callbacks will handle this, and the
3285 client can restart the bulk request.
3288 Frequency : Always when mmapping file with no objects
3290 Description: client crashes when mmapping file with no objects
3291 Details : Check that we actually have objects in a file before doing any
3292 operations on objects in ll_vm_open, ll_vm_close and
3298 Description: Request leak when working with deleted CWD
3299 Details : Introduce advanced request refcount tracking for requests
3300 referenced from lustre intent.
3302 Severity : Enhancement
3304 Description: Cache open file handles on client.
3305 Details : MDS now will return special lock along with openhandle, if
3306 requested and client is allowed to hold openhandle, even if unused,
3307 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3308 closing files for every read/write openration.
3310 Severity : Enhancement
3312 Description: Cache open negative dentries on client when possible.
3313 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3314 negative dentries on lock revocation.
3319 Description: Remounting a client read-only wasn't possible with a zconf mount
3320 Details : It wasn't possible to remount a client read-only with llmount.
3322 Severity : enhancement
3323 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3324 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3325 MPICH is included to add Lustre-specific ADIO interfaces.
3326 This is based closely on the UFS ADIO layer and only differs
3327 in file creation, in order to allow the OST striping to be set.
3328 This is user-contributed code and not supported by CFS.
3333 Description: extended inode attributes (immutable, append-only) work improperly
3334 when 2.4 and 2.6 kernels are used on client/server or vice versa
3335 Details : Introduce kernel-independent values for these flags.
3337 Severity : enhancement
3340 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3341 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3342 granularity. Now, Fractional MB granularity is be supported,
3343 this is very useful for low memory system.
3345 Severity : enhancement
3347 Description: Getattr by fid
3348 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3349 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3350 have locks and avoids getattr rpc requests for it on every stat.
3353 Frequency : Always, for filesystems larger than 2TB
3355 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3356 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3357 filesystems larger than 2TB to be mounted with mballoc enabled.
3360 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3362 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3363 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3364 filesystems larger than 2TB is mounted then the kernel will
3365 silently corrupt the start of the filesystem. CONFIG_LBD is
3366 enabled for all CFS-supported kernels, but the possibility of
3367 this happening with a modified kernel config exists.
3369 Severity : enhancement
3371 Description: add client O_DIRECT support for 2.6 kernels
3372 Details : It is now possible to do O_DIRECT reads and writes to files
3373 in the Lustre client mountpoint on 2.6 kernel clients.
3375 Severity : enhancement
3377 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3378 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3379 parallel, not waiting for response from every OST before sending
3380 a rpc to the next OST.
3385 Description: setattr vs write race when updating file timestamps
3386 Details : Client processes that update a file timestamp into the past
3387 right after writing to the file (e.g. tar) it is possible that
3388 the updated file modification time can be reset to the current
3389 time due to a race between processing the setattr and write RPC.
3391 Severity : enhancement
3393 Description: Bring 'lfs find' closer in line with regular Linux find.
3394 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3395 -print0 options and obtains all the needed info through the lustre
3398 Severity : enhancement
3400 Description: support up to 1024 configured devices on one node
3401 Details : change obd_dev array from statically allocated to dynamically
3402 allocated structs as they are first used to reduce memory usage
3407 Description: Flush dirty partially truncated pages during truncate
3408 Details : Immediatelly flush partially truncated pages in filter_setattr,
3409 this way we completely avoid having any pages in page cache on OST
3410 and can retire ugly workarounds during writes to flush such pages.
3415 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3416 Details : protect inode from truncation within vfs_unlink() context
3417 just take a reference before calling vfs_unlink() and release it
3418 when parent's i_sem is free.
3423 Description: last_id value checked outside lock on OST caused LASSERT failure
3424 Details : If there were multiple MDS->OST object precreate requests in
3425 flight, it was possible that the OST's last object id was checked
3426 outside a lock and incorrectly tripped an assertion. Move checks
3427 inside locks, and discard old precreate requests.
3430 Frequency : always, if extents are used on OSTs
3432 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3433 Details : OSTs using the extents format would not zero the high 16 bits of
3434 the index physical block number. This is not a problem for any
3435 OST filesystems smaller than 16TB, and no kernels support ext3
3436 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3437 new/modified files) and can be fixed for existing filesystems
3438 with e2fsprogs-1.39-cfs1.
3443 Description: import connection selection may be incorrect if timer wraps
3444 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3445 import connections to be ignored if the 32-bit jiffies counter
3446 wraps. Use a 64-bit jiffies counter.
3449 Frequency : very large clusters immediately after boot
3451 Description: LNET request buffers exhausted under heavy short-term load
3452 Details : If a large number of client requests are generated on a service
3453 that has previously never seen so many requests it is possible
3454 that the request buffer growth cannot keep up with the spike in
3455 demand. Instead of dropping incoming requests, they are held in
3456 the LND until the RPC service can accept more requests.
3459 Frequency : Sometimes during replay
3461 Description: Assertion failure in ll_local_open after replay.
3462 Details : If replay happened on an open request reply before we were able
3463 to set replay handler, reply will become not swabbed tripping the
3464 assertion in ll_local_open. Now we set the handler right after
3465 recognising of open request
3468 Frequency : very rare
3470 Description: kernel reports "badness in vsnprintf"
3471 Details : Reading from the "recovery_status" /proc file in small chunks
3472 may cause a negative length in lprocfs_obd_rd_recovery_status()
3473 call to vsnprintf() (which is otherwise harmless). Exit early
3474 if there is no more space in the output buffer.
3476 Severity : enhancement
3478 Description: clear OBD RPC statistics by writing to them
3479 Details : It is now possible to clear the OBD RPC statistics by writing
3480 to the "stats" file.
3485 Description: Client mtime is not the same on different clients after utimes
3486 Details : In some cases, the client was using the utimes() syscall on
3487 a file cached on another node. The clients now validate the
3488 ctime from the MDS + OSTs to determine which one is right.
3493 Description: Inability to activate failout mode
3494 Details : lconf script incorrectly assumed that in pythong string's numeric
3495 value is used in comparisons.
3498 Frequency : always with multiple stripes per file
3500 Description: Inefficient object allocation for mutli-stripe files
3501 Details : When selecting which OSTs to stripe files over, for files with
3502 a stripe count that divides evenly into the number of OSTs,
3503 the MDS is always picking the same starting OST for each file.
3504 Return the OST selection heuristic to the original design.
3509 Description: mount failures may take full timeout to return an error
3510 Details : Under some heavy load conditions it is possible that a
3511 failed mount can wait for the full obd_timeout interval,
3512 possibly several minutes, before reporting an error.
3513 Instead return an error as soon as the status is known.
3515 Frequency : quota enabled and large files being deleted
3517 Description: releasing more than 4GB of quota at once hangs OST
3518 Details : If a user deletes more than 4GB of files on a single OST it
3519 will cause the OST to spin in an infinite loop. Release
3520 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3525 Description: statfs data retrieved from /proc may be stale or zero
3526 Details : When reading per-device statfs data from /proc, in the
3527 {kbytes,files}_{total,free,avail} files, it may appear
3528 as zero or be out of date.
3531 Frequency : systems with MD RAID1 external journal devices
3533 Description: lconf's call to blkid is confused by RAID1 journal devices
3534 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3535 of returning all block devices that match the journal UUID.
3538 Frequency : always, for aggregate stripe size over 4GB
3540 Description: assertion fails when trying to use 4GB stripe size
3541 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3542 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3545 Frequency : always on ppc64
3547 Description: the first write on an ext3 filesystem with mballoc got stuck
3548 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3549 perform endianness conversion.
3551 ------------------------------------------------------------------------------
3553 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3555 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3556 INTEROPERATE with older versions automatically. Please read the
3557 user documentation before upgrading any part of a live system.
3558 * WARNING: Lustre networking configuration changes are required with
3559 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3562 * Support for kernels:
3563 2.6.9-22.0.2.EL (RHEL 4)
3564 2.6.5-7.244 (SLES 9)
3565 2.6.12.6 vanilla (kernel.org)
3568 Severity : enhancement
3569 Bugzilla : 7981/8208
3570 Description: Introduced Lustre Networking (LNET)
3571 Details : LNET is new networking infrastructure for Lustre, it includes
3572 a reorganized network configuration mode (see the user
3573 documentation for full details) as well as support for routing
3574 between different network fabrics. Lustre Networking Devices
3575 (LNDs) for the supported network fabrics have also been
3576 created for this new infrastructure.
3578 Severity : enhancement
3579 Description: Introduced Access control lists
3580 Details : clients can set ACLs on files and directories in order to have
3581 more fine-grained permissions than the standard Unix UGO+RWX.
3582 The MDS must be started with the "-o acl" mount option.
3584 Severity : enhancement
3585 Description: Introduced filesystem quotas
3586 Details : Administrators may now establish per-user quotas on the
3589 Severity : enhancement
3591 Description: Configuration change for the XT3
3592 The PTLLND is now used to run Lustre over Portals on the XT3
3593 The configure option(s) --with-cray-portals are no longer used.
3594 Rather --with-portals=<path-to-portals-includes> is used to
3595 enable building on the XT3. In addition to enable XT3 specific
3596 features the option --enable-cray-xt3 must be used.
3601 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3602 Details : A race between allocating a new llcd and re-getting the llcd_lock
3603 allowed another thread to grab newly-allocated llcd.
3605 Severity : enhancement
3607 Description: 2.6 OST async journal commit and locking fix to improve performance
3608 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3609 2.6 kernels are now async as they already were in 2.4 kernels so
3610 that they can commit concurrently with the network bulk transfer.
3611 For block-allocated files the filter allocation semaphore is held
3612 to avoid filesystem fragmentation during allocation. BKL lock
3613 removed for 2.6 xattr operations where it is no longer needed.
3618 Description: lconf incorrectly determined whether two IP networks could talk
3619 Details : In some more complicated routing and multiple-network
3620 configurations, lconf will avoid trying to make a network
3621 connection to a disjoint part of the IP space. It was doing the
3622 math incorrectly for one set of cases.
3627 Description: Fix for potential infinite loop processing records in an llog.
3628 Details : If an llog record is corrupted/zeroed, it is possible to loop
3629 forever in llog_process(). Validate the llog record length
3630 and skip the remainder of the block on error.
3633 Frequency : occasional (liblustre only)
3635 Description: liblustre could not open files whose last component is a symlink
3636 Details : sysio_path_walk() would incorrectly pass the open intent to
3637 intermediate path components.
3640 Frequency : rare (liblustre only with non-standard tuning)
3641 Bugzilla : 7201 (7350)
3642 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3643 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3644 after it has been released, only do this for referenced locks
3646 Severity : enhancement
3648 Description: specifying an (invalid) directory default stripe_size of -1
3649 would reset the directory default striping
3650 Details : stripe_size -1 was used internally to signal directory stripe
3651 removal, now use "all default" to signal dir stripe removal
3652 as a directory striping of "all default" is not useful
3655 Frequency : common for large clusters running liblustre clients
3657 Description: doing an ls when liblustre clients are running is slow
3658 Details : sending a glimpse AST to a liblustre client waits for every AST
3659 to time out, as liblustre clients will not respond. Since they
3660 cannot cache data we refresh the OST lock LVB from disk instead.
3662 Severity : enhancement
3664 Description: doing an ls at the same time as file IO can be slow
3665 Details : enqueue and other "small" requests can be blocked behind many
3666 large IO requests. Create a new OST IO portal for non-IO
3667 requests so they can be processed faster.
3670 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3672 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3673 created file with mode 000
3674 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3675 allow an ftruncate() as a non-root user to a file with mode 000.
3676 Lustre now respects this flag to disable mode checks when
3677 truncating a file owned by the user
3680 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3683 Description: Revoking locks from clients that went dead or catatonic might take
3685 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3686 cancellation of such locks instant on servers without waiting for
3687 any reply from clients. Clients drops these locks when cancel
3688 notification from server is received without replying.
3691 Frequency : liblustre-only, when liblustre client dies or becomes busy
3693 Description: Doing ls on Linux clients can take a long time with active
3695 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3696 granting such locks to it in the first place if possible. Locks
3697 are taken by proxy on the OST during the read or write and
3698 dropped immediately afterward. Add connect flags handling, do
3699 not grant locks to liblustre clients for glimpse ASTs.
3701 Severity : enhancement
3703 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3704 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3705 sequential reads and behaves badly with applications doing random
3706 reads. Improve it by reading ahead at least read region, and
3707 avoiding excessive large RPC for small reads.
3709 Severity : enhancement
3711 Description: Creating more than 1000 files for a single job may cause a load
3712 imbalance on the OSTs if there are also a large number of OSTs.
3713 Details : qos_prep_create() uses an OST index reseed value that is an
3714 even multiple of the number of available OSTs so that if the
3715 reseed happens in the middle of the object allocation it will
3716 still utilize the OSTs as uniformly as possible.
3721 Description: OST or MDS may oops in ping_evictor_main()
3722 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3723 but doesn't restart at beginning of obd_exports_timed list
3726 Severity : enhancement
3728 Description: improve by-nid export eviction on the MDS and OST
3729 Details : allow multiple exports with the same NID to be evicted at one
3730 time without re-searching the exports list.
3733 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3735 Description: MDS may oops in groups_free()
3736 Details : in rare race conditions a newly allocated group_info struct is
3737 freed again, and this can be NULL. The 2.4 compatibility code
3738 for groups_free() checked for a NULL pointer, but 2.6 did not.
3741 Frequency : common for liblustre clients doing little filesystem IO
3742 Bugzilla : 9352, 7313
3743 Description: server may evict liblustre clients accessing contended locks
3744 Details : if a client is granted a lock or receives a completion AST
3745 with a blocking AST already set it would not reply to the AST
3746 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3749 Frequency : lfs setstripe, only systems with more than 160 OSTs
3751 Description: unable to set striping with a starting offset beyond OST 160
3752 Details : llapi_create_file() incorrectly limited the starting stripe
3753 index to the maximum single-file stripe count.
3756 Frequency : LDAP users only
3758 Description: lconf did not handle in-kernel recovery with LDAP properly
3759 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3761 Severity : enhancement
3763 Description: bind OST threads to NUMA nodes to improve performance
3764 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3765 node and do their allocations there to localize memory access
3767 Severity : enhancement
3769 Description: llmount can determine client NID directly from Myrinet (GM)
3770 Details : the client NID code from gmnalnid was moved directly into
3771 llmount, removing the need to use this or specifying the
3772 client NID explicitly when mounting GM clients with zeroconf
3775 Frequency : if client is started with down MDS
3777 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3778 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3779 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3780 ptlrpc_import_delayed_req() because we didn't check if the
3781 request was interrupted, and we also didn't break out of the
3782 event loop if there was no timeout
3787 Description: data loss during non-page-aligned writes to a single file from
3788 both multiple nodes and multiple threads on one node at same time
3789 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3790 inconsistency led to false short-reads, that were cached and later
3791 used by ->prepare_write() to fill in partially written page,
3792 leading to data loss.
3795 Frequency : always, if lconf --abort_recovery used
3797 Description: lconf --abort_recovery fails with 'Operation not supported'
3798 Details : lconf was attempting to abort recovery on the MDT device and not
3801 Severity : enhancement
3803 Description: remove cleanup logs
3804 Details : replace lconf-generated cleanup logs with lustre internal
3805 cleanup routines. Eliminates the need for client-cleanup and
3808 Severity : enhancement
3810 Description: add support for EAs (user and system) on lustre filesystems
3811 Details : it is now possible to store extended attributes in the Lustre
3812 client filesystem, and with the user_xattr mount option it
3813 is possible to allow users to store EAs on their files also
3815 Severity : enhancement
3817 Description: Add possibility (config option) to show minimal available OST free
3819 Details : When compiled with --enable-mindf configure option, statfs(2)
3820 (and so, df) will return least minimal free space available from
3821 all OSTs as amount of free space on FS, instead of summary of
3822 free spaces of all OSTs.
3824 Severity : enhancement
3826 Description: do not expand extent locks acquired on OST-side
3827 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3828 by server: they are not cached anyway.
3831 Frequency : when mmap is used/binaries executed from Lustre
3833 Description: Unmmap pages before throwing them away from read cache.
3834 Details : llap_shrink cache now attempts to unmap pages before discarding
3835 them (if unmapping failed - do not discard). SLES9 kernel has
3836 extra checks that trigger if this unmapping is not done first.
3841 Description: lconf didn't resolve symlinks before checking to see whether a
3842 given mountpoint was already in use
3845 Frequency : when migrating failover services
3846 Bugzilla : 6395, 9514
3847 Description: When migrating a subset of services from a node (e.g. failback
3848 from a failover service node) the remaining services would
3849 time out and evict clients.
3850 Details : lconf --force (implied by --failover) sets the global obd_timeout
3851 to 5 seconds in order to quickly disconnect, but this caused
3852 other RPCs to time out too quickly. Do not change the global
3853 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3855 Severity : enhancement
3856 Frequency : if MDS is started with down OST
3857 Bugzilla : 9439,5706
3858 Description: Allow startup/shutdown of an MDS without depending on the
3859 availability of the OSTs.
3860 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3861 Add appropriate locking and lov-osc refcounts for safe
3862 cleaning. Add osc abort_inflight calls in case the
3863 synchronize never started.
3866 Frequency : occasional (Cray XT3 only)
3868 Description: root not authorized to access files in CRAY_PORTALS environment
3869 Details : The client process capabilities were not honoured on the MDS in
3870 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3871 been accessed by an authorized user then root was able to access
3872 the file on the local client also. The root user capabilities
3873 are now allowed on the MDS, as this environment has secure UID.
3876 Frequency : occasional
3878 Description: ldiskfs "too long searching" message happens too often
3879 Details : A debugging message (otherwise harmless) prints too often on
3880 the OST console. This has been reduced to only happen when
3881 there are fragmentation problems on the filesystem.
3886 Description: Division by zero in statfs when all OSCs are inactive
3887 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3888 lov_statfs divides by value returned by lov_get_stripecnt().
3892 Bugzilla : 9489, 3273
3893 Description: First write from each client to each OST was only 4kB in size,
3894 to initialize client writeback cache, which caused sub-optimal
3895 RPCs and poor layout on disk for the first writen file.
3896 Details : Clients now request an initial cache grant at (re)connect time
3897 and so that they can start streaming writes to the cache right
3898 away and always do full-sized RPCs if there is enough data.
3899 If the OST is rebooted the client also re-establishes its grant
3900 so that client cached writes will be honoured under the grant.
3905 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3906 Details : Now I/O RPCs go to different portal number and (presumably) fast
3907 lock requests (and glimses) and other RPCs get their own service
3908 threads pool that should be able to service those RPCs
3911 Severity : enhancement
3913 Description: Ability to exchange lustre version between client and servers and
3914 issue warnings at client side if client is too old. Also for
3915 liblustre clients there is ability to refuse connection of too old
3917 Details : New 'version' field is added to connect data structure that is
3918 filled with version info. That info is later checked by server and
3922 Frequency : rare, liblustre only.
3923 Bugzilla : 9296, 9581
3924 Description: Two simultaneous writes from liblustre at offset within same page
3925 might proceed at the same time overwriting eachother with stale
3927 Details : I/O lock withing llu_file_prwv was released too early, before data
3928 actually was hitting the wire. Extended lock-holding time until
3929 server acknowledges receiving data.
3932 Frequency : extremely rare. Never observed in practice.
3934 Description: avoid generating lustre_handle cookie of 0.
3935 Details : class_handle_hash() generates handle cookies by incrementing
3936 global counter, and can hit 0 occasionaly (this is unlikely, but
3937 not impossible, because initial value of cookie counter is
3938 selected randonly). Value of 0 is used as a sentinel meaning
3939 "unassigned handle" --- avoid it. Also coalesce two critical
3940 sections in this function into one.
3942 Severity : enhancement
3944 Description: allow liblustre clients to delegate truncate locking to OST
3945 Details : To avoid overhead of locking, liblustre client instructs OST to
3946 take extent lock in ost_punch() on client's behalf. New connection
3947 flag is added to handle backward compatibility.
3949 Severity : enhancement
3950 Bugzilla : 4928, 7341, 9758
3951 Description: allow number of OST service threads to be specified
3952 Details : a module parameter allows the number of OST service threads
3953 to be specified via "options ost ost_num_threads={N}" in the
3954 OSS's /etc/modules.conf or /etc/modprobe.conf.
3958 Bugzilla : 6146, 9635, 9895
3959 Description: servers crash with bad pointer in target_handle_connect()
3960 Details : In rare cases when a client is reconnecting it was possible that
3961 the connection request was the last reference for that export.
3962 We would temporarily drop the export reference and get a new
3963 one, but this may have been the last reference and the export
3964 was just destroyed. Get new reference before dropping old one.
3966 Severity : enhancement
3967 Frequency : if client is started with failover MDS
3969 Description: Allow multiple MDS hostnames in the mount command
3970 Details : Try to read the configuration from all specified MDS
3971 hostnames during a client mount in case the "primary"
3974 Severity : enhancement
3976 Description: Stop sending data to evicted clients as soon as possible.
3977 Details : Check if the client we are about to send or are sending data to
3978 was evicted already. (Check is done every second of waiting,
3979 for which l_wait_event interface was extended to allow checking
3980 of exit condition at specified intervals).
3983 Frequency : rare, normally only when NFS exporting is done from client
3985 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3987 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3990 Frequency : always, if srand() is called before liblustre initialization
3992 Description: Liblustre uses system PRNG disturbing its usage by user application
3993 Details : Introduce internal to lustre fast and high-quality PRNG for
3994 lustre usage and make liblustre and some other places in generic
3995 lustre code to use it.
3997 Severity : enhancement
3998 Bugzilla : 9477, 9557, 9870
3999 Description: Verify that the MDS configuration logs are updated when xml is
4000 Details : Check if the .xml configuration logs are newer than the config
4001 logs stored on the MDS and report an error if this is the case.
4002 Request --write-conf, or allow starting with --old_conf.
4004 Severity : enhancement
4006 Description: Handle symlinks in the path when checking if Lustre is mounted.
4007 Details : Resolve intermediate symlinks when checking if a client has
4008 mounted a filesystem to avoid duplicate client mounts.
4013 Description: lconf can hit an error exception but still return success.
4014 Details : The lconf command catches the Command error exception at the top
4015 level script context and will exit with the associated exit
4016 status, but doesn't ensure that this exit status is non-zero.
4021 Description: failure of ptlrpc thread startup can cause oops
4022 Details : Starting a ptlrpc service thread can fail if there are a large
4023 number of threads or the server memory is very fragmented.
4024 Handle this without oopsing.
4027 Frequency : always, only if liblustre and non-default acceptor port was used
4029 Description: liblustre cannot connect to servers with non-default acceptor port
4030 Details : tcpnal_set_default_params() was not called and was therefore
4031 ignoring the environment varaible TCPNAL_PORT, as well as other
4032 TCPNAL_ environment variables
4037 Description: two objects could be created on the same OST for a single file
4038 Details : If an OST is down, in some cases it was possible to create two
4039 objects on a single OST for a single file. No problems other
4040 than potential performance impact and spurious error messages.
4044 Bugzilla : 5681, 9562
4045 Description: Client may oops in ll_unhash_aliases
4046 Details : Client dcache may become inconsistent in race condition.
4047 In some cases "getcwd" can fail if the current directory is
4053 Description: Inode refcounting problems in NFS export code
4054 Details : link_raw functions used to call d_instantiate without obtaining
4055 extra inode reference first.
4059 Bugzilla : 9942, 9903
4060 Description: Referencing freed requests leading to crash, memleaks with NFS.
4061 Details : We used to require that call to ll_revalidate_it was always
4062 followed by ll_lookup_it. Also with revalidate_special() it is
4063 possible to call ll_revalidate_it() twice for the same dentry
4064 even if first occurence returned success. This fix changes semantic
4065 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4066 reference on a request referred from the intent.
4067 ll_intent_release() then releases such a request.
4070 Frequency : rare, normally benchmark loads only
4072 Description: unlinked inodes were kept in memory on the client
4073 Details : If a client is repeatedly creating and unlinking files it
4074 can accumulate a lot of stale inodes in the inode slab cache.
4075 If there is no other client load running this can cause the
4076 client node to run out of memory. Instead flush old inodes
4077 from client cache that have the same inode number as a new inode.
4080 Frequency : SLES9 2.6.5 kernel and long filenames only
4081 Bugzilla : 9969, 10379
4082 Description: utime reports stale NFS file handle
4083 Details : SLES9 uses out-of-dentry names in some cases, which confused
4084 the lustre dentry revalidation. Change it to always use the
4088 Frequency : rare, unless heavy write-truncate concurrency is continuous
4089 Bugzilla : 4180, 6984, 7171, 9963, 9331
4090 Description: OST becomes very slow and/or deadlocked during object unlink
4091 Details : filter_destroy() was holding onto the parent directory lock
4092 while truncating+unlinking objects. For very large objects this
4093 may block other threads for a long time and slow overall OST
4094 responsiveness. It may also be possible to get a lock ordering
4095 deadlock in this case, or run out of journal credits because of
4096 the combined truncate+unlink. Solution is to do object truncate
4097 first in one transaction without parent lock, and then do the
4098 final unlink in a new transaction with the parent lock. This
4099 reduces the lock hold time dramatically.
4102 Frequency : rare, 2.4 kernels only
4104 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4105 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4106 before unmounting their filesystem in order to avoid deadlock.
4107 The kernel_locked() function in 2.4 kernels only checks whether
4108 the kernel lock is held, not whether it is this process that is
4109 holding it as 2.6 kernels do.
4114 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4115 Details : The client ptlrpc code may be trying to reconnect to a down
4116 server before a previous connection attempt has timed out.
4117 Increase the reconnect interval to be longer than the connection
4118 timeout interval to avoid sending duplicate connections to
4122 Frequency : echo_client brw_test command
4124 Description: fix echo_client to work with OST preallocated code
4125 Details : OST preallocation code (5137) didn't take echo_client IO path
4126 into account: echo_client calls filter methods outside of any
4127 OST thread and, hence, there is no per-thread preallocated
4128 pages and buffers to use. Solution: hijack pga pages for IO. As
4129 a byproduct, this avoids unnecessary data copying.
4133 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4134 Description: Client can oops in mdc_commit_close() after open replay
4135 Details : It was possible for the MDS to return an open request with no
4136 transaction number in mds_finish_transno() if the client was
4137 evicted, but without actually returning an error. Clients
4138 would later try to replay that open and may trip an assertion
4139 Simplify the client close codepath, and always return an error
4140 from the MDS in case the open is not successful.
4143 Frequency : rare, 2.6 OSTs only
4145 Description: OST may deadlock under high load on fragmented files
4146 Details : If there was a heavy load and highly-fragmented OST filesystems
4147 it was possible to have all the OST threads deadlock waiting on
4148 allocation of biovecs, because the biovecs were not released
4149 until the entire RPC IO was completed. Instead, release biovecs
4150 as soon as they are complete to ensure forward IO progress.
4152 Severity : enhancement
4154 Description: Support for specifying external journal device at mount
4155 Details : If an OST or MDS device is formatted with an external journal
4156 device, this device major/minor is stored in the ext3 superblock
4157 and may not be valid for failover. Allow detecting and
4158 specifying the external journal at mount time.
4163 Description: Mounting an MDS with pending unlinked files may cause oops
4164 Details : target_finish_recovery() calls mds_postrecov() which returned
4165 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4166 considers this an error and immediately begins cleaning up the
4167 lov, just after starting the mds_lov process
4169 Severity : enhancement
4171 Description: Implement 'lfs df' to report actual free space on per-OST basis
4172 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4173 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4174 usage of objects; '-h' to report in human readable format.
4176 ------------------------------------------------------------------------------
4178 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4185 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4186 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4187 at startup during orphan unlinking. Instead, skip buddy update
4188 before setup, as it will be regenerated after recovery is complete.
4193 Description: If an OST is inactive, its locks might reference stale inodes.
4194 Details : lov_change_cbdata() must iterate over all namespaces, even if
4195 they are inactive to clear inode references from the lock.
4197 Severity : enhancement
4198 Frequency : occasional, if non-standard max_dirty_mb used
4200 Description: Client will block write RPCs if not enough grant
4201 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4202 then the client will block writes while waiting for another RPC
4203 to complete instead of consuming its dirty limit. With change
4204 we get improved performance when max_dirty_mb is small.
4206 Severity : enhancement
4207 Bugzilla : 3389, 6253
4208 Description: Add support for supplementary groups on the MDS.
4209 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4210 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4211 lookups for user supplementary groups into a cache.
4215 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4216 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4217 for liblustre O_EXCL is enforced by MDS.
4221 Description: readdir never returns NULL in liblustre.
4222 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4223 dentry in struct dirent.
4227 Frequency : liblustre only, depends on application IO pattern
4228 Description: liblustre clients evicted if not contacting servers
4229 Details : Don't put liblustre clients into the ping_evictor list, so
4230 they will not be evicted by the pinger ever.
4232 Severity : enhancement
4234 Description: Add ability to evict clients by NID from MDS.
4235 Details : By echoing "nid:$NID" string into
4236 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4237 $NID would be instantly evicted from this MDS and from all active
4238 OSTs connected to it.
4242 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4243 Details : lookup_it_finish() used to query file size from OSTs that was not
4248 Description: service threads change working directory to that of init
4249 Details : Starting lustre service threads may pin the working directory
4250 of the parent thread, making that filesystem busy. Threads
4251 now change to the working directory of init to avoid this.
4255 Frequency : during shutdown only
4256 Description: shutdown with a failed MDS or OST can cause unmount to hang
4257 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4262 Frequency : 2.6 only
4263 Description: chmod/chown may include an extra supplementary group
4264 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4265 supplementary group and if none is specified this is used.
4268 Bugzilla : 5479 (6816)
4270 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4271 Details : If lookup is in progress on a file that is unlinked we might try
4272 to revalidate the inode and fail in revalidate after lookup is
4273 complete and ll_file_open() enqueues the open again but
4274 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4277 Frequency : always, if lconf --abort_recovery used
4279 Description: lconf --abort_recovery fails with 'Operation not supported'
4280 Details : lconf was attempting to abort recovery on the MDT device and not
4283 ------------------------------------------------------------------------------
4285 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4290 Frequency : rare (only unsupported configurations with a node running as an
4292 Bugzilla : 6514, 5137
4293 Description: Mounting a Lustre file system on a node running as an OST could
4295 Details : OSTs now preallocates memory needed to write out data at
4296 startup, instead of when needed, to avoid having to
4297 allocate memory in possibly low memory situations.
4298 Specifically, if the file system is mounted on on OST,
4299 memory pressure could force it to try to write out data,
4300 which it needed to allocate memory to do. Due to the low
4301 memory, it would be unable to do so and the node would
4302 become unresponsive.
4304 Severity : enhancement
4306 Description: Addition of lconf --service command line option
4307 Details : lconf now accepts a '--service <arg>' option, which is
4308 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4310 Severity : enhancement
4312 Description: Failover mode is now the default for OSTs.
4313 Details : By default, OSTs will now run in failover mode. To return to
4314 the old behaviour, add '--failout' to the lmc line for OSTs.
4316 Severity : enhancement
4318 Description: Health checks are now provided for MDS and OSTs
4319 Details : Additional detailed health check information on MSD and OSTs
4320 is now provided through the procfs health_check value.
4323 Frequency : occasional, depends on IO load
4325 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4326 after numerous create/delete cycles
4327 Details : The ext3 inode allocation policy would not allocate new inodes
4328 very well on the OSTs because there are no new directories
4329 being created. Instead we look for groups with free space if
4330 the parent directories are nearly full.
4335 Description: Network or server problems during mount may cause partially
4336 mounted clients instead of returning an error.
4337 Details : The config llog parsing code may overwrite the error return
4338 code during mount error handling, returning success instead
4344 Description: MDS can fail to allocate large reply buffers
4345 Details : After long uptimes the MDS can fail to allocate large reply
4346 buffers (e.g. zconf client mount config records) due to memory
4347 fragmentation or consumption by the buffer cache. Preallocate
4348 some large reply buffers so that these replies can be sent even
4349 under memory pressure.
4353 Frequency : rare (liblustre)
4354 Description: fsx running with liblustre complained that using truncate() to
4355 extend the file doesn't work. This patch corrects that issue.
4356 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4357 ATTR_SIZE and lsm use in llu_setattr_raw.
4361 Frequency : rare, only 2.6 kernels
4362 Description: Unusual file access patterns on the MDS may result in inode
4363 data being lost in very rare circumstances.
4364 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4365 code in the 2.6 kernel caused the inode and/or EA data not to
4366 be read from disk, causing single-file corruption.
4370 Frequency : rare, only 2.6 filesystems using extents
4371 Description: Heavy concurrent write and delete load may cause data corruption.
4372 Details : It was possible under high-load situations to have an extent
4373 metadata block in the block device cache from a just-unlinked
4374 file overwrite a newly-allocated data block. We now unmap any
4375 metadata buffers that alias just-allocated data blocks.
4379 Frequency : filesystems with default stripe_count larger than 77
4380 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4381 Details : lconf specifies an inode size of 4096 bytes when the default
4382 stripe_count is larger than 77. This conflicts with the default
4383 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4384 this case to avoid pinning too much memory for large EAs.
4386 ------------------------------------------------------------------------------
4388 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4393 Frequency : rare (extremely heavy IO load with hundreds of clients)
4395 Description: Client is evicted, gets IO error writing to file
4396 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4397 caused clients to be evicted for AST timeouts. The fixes in
4398 bug 5192 mean we no longer need to have such short AST timeouts
4399 so ldlm_timeout has been increased.
4402 Frequency : occasional during --force or --failover shutdown under load
4403 Bugzilla : 5949, 4834
4404 Description: Server oops/LBUG if stopped with --force or --failover under load
4405 Details : a collection of import/export refcount and cleanup ordering
4406 issues fixed for safer force cleanup
4409 Frequency : only filesystems larger than 120 OSTs
4410 Bugzilla : 5990, 6223
4411 Description: lfs getstripe would oops on a very large filesystem
4412 Details : lov_getconfig used kfree on vmalloc'd memory
4415 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4416 Bugzilla : 6242, 6243
4417 Description: reading from files that had been truncated to a non-zero size
4418 but never opened returned no data
4419 Details : ll_file_read() reads zeros from no-object files to EOF
4424 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4425 Details : There's little harm in aborting MDS/OSS recovery and letting it
4426 try again, so I removed the LASSERT and return an error instead.
4428 Severity : enhancement
4430 Description: New debugging infrastructure for tracking down data corruption
4431 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4432 (b) cover more of the client-side code path, and (c) try to narrow
4433 down where problems occurred
4437 Bugzilla : 3819, 4364, 4397, 6313
4438 Description: Racing close and eviction MDS could cause assertion in mds_close
4439 Details : It was possible to get multiple mfd references during close and
4440 client eviction, leading to one thread referencing a freed mfd.
4442 Severity: : enhancement
4443 Bugzilla : 3262, 6359
4444 Description: Attempts to reconnect to servers are now more aggressive.
4445 Details : This builds on the enhanced upcall-less recovery that was added
4446 in 1.4.2. When trying to reconnect to servers, clients will
4447 now try each server in the failover group every 10 seconds. By
4448 default, clients would previously try one server every 25 seconds.
4453 Description: After recovery, certain operations trigger a failed
4454 assertion on a client.
4455 Details : Failing over an mds, using lconf -d --failover, while a
4456 client was doing a readdir() call would cause the client to
4457 LBUG after recovery completed and the readdir() was resent.
4459 Severity : enhancement
4461 Description: Default groups are now added by lconf
4462 Details : You can now run lconf --group <servicename> without having to
4463 manually add groups with lmc.
4466 Frequency : occasional
4468 Description: Nodes with an elan id of 0 trigger a failed assertion
4471 Frequency : always when accessing e.g. tty/console device nodes
4473 Description: tty and some other devices nodes cannot be used on lustre
4474 Details : file's private_data field is used by device data and lustre
4475 values in there got lost. New field was added to struct file to
4476 store fs-specific private data.
4479 Frequency : when exporting Lustre via NFS
4481 Description: NFSD failed occasionally when looking up a path component
4482 Details : NFSD is looking up ".." which was broken in ext3 directories
4483 that had grown large enough to become hashed.
4486 Frequency : Clusters with multiple interfaces not on the same subnet
4488 Description: Nodes will repeatedly try to reconnect to an interface which it
4489 cannot reach and report an error to the log.
4490 Details : Extra peer list entries will be created by lconf with some peers
4491 unreachable. lconf now validates the peer before adding it.
4494 Frequency : Only if a default stripe is set on the filesystem root.
4496 Description: Setting a default stripe on the filesystem root prevented the
4497 filesystem from being remounted.
4498 Details : The client was sending extra request flags in the root getattr
4499 request and did not allocate a reply buffer for the dir EA.
4502 Frequency : occasional, higher if lots of files are accessed by one client
4503 Bugzilla : 6159, 6097
4504 Description: Client trips assertion regarding lsm mismatch/magic
4505 Details : While revalidating inodes the VFS looks up inodes with ifind()
4506 and in rare cases can find an inode that is being freed.
4507 The ll_test_inode() code will free the lsm during ifind()
4508 when it finds an existing inode and then the VFS later attaches
4509 this free lsm to a new inode.
4513 Bugzilla : 6422, 7030
4514 Description: MDS deadlock between mkdir and client eviction
4515 Details : Creating a new file via mkdir or mknod (starting a transaction
4516 and getting the ns lock) can deadlock with client eviction
4517 (gets ns lock and trying to finish a synchronous transaction).
4520 Frequency : occasional
4521 Description: While starting a server, the fsfilt_ext3 module could not be
4523 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4524 kernels. Previously, lconf would still use the ext3 name
4525 when trying to load modules. Now, it will correctly use
4526 ext3 on 2.4 and ldiskfs on 2.6.
4528 Severity : enhancement
4529 Description: The default stripe count has been changed to 1
4530 Details : The interpretation of the default stripe count (0, to lfs
4531 or lmc) has been changed to mean striping across a single
4532 OST, rather than all available. For general usage we have
4533 found a stripe count of 1 or 2 works best.
4535 Severity : enhancement
4536 Description: Add support for compiling against Cray portals.
4537 Details : Conditional compiling for some areas that are different
4541 Frequency : occasional
4542 Bugzilla : 6409, 6834
4543 Description: Creating files with an explicit stripe count may lead to
4544 a failed assertion on the MDS
4545 Details : If some OSTs are full or unavailable, creating files may
4546 trigger a failed assertion on the MDS. Now, Lustre will
4547 try to use other servers or return an error to the
4551 Frequency : occasional
4553 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4554 caused lock timeout during readahead (since 1.4.2).
4555 Details : Processes doing readahead might match a lock that hasn't been
4556 granted yet if there are overlapping and conflicting lock
4557 requests. The readahead process waits on ungranted lock
4558 (original lock is CBPENDING), while OST waits for that process
4559 to cancel CBPENDING read lock and eventually evicts client.
4561 Severity : enhancement
4563 Description: Initial enabling of flock support for clients
4564 Details : Implements fcntl advisory locking and file status functions.
4565 This feature is provided as an optional mount flag (default
4566 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4567 locking are implemented yet, and those that are are not guaranteed
4568 to be completely correct in production environments.
4569 mount -t lustre -o [flock|noflock] ...
4572 Frequency : occasional
4574 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4575 assertion in the ext3 JBD (journaling) layer.
4576 Details : The b_committed_data struct is protected by the big kernel lock
4577 in 2.4 kernels, serializing journal_commit_transaction() and
4578 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4579 access to this struct. In 2.6 kernels there is finer grained
4580 locking to improve SMP performance of the JBD layer.
4584 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4589 Description: merge and cleanup kernel patches.
4592 -----------------------------------------------------------------------------
4594 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4596 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4597 versions of Lustre. Please update all servers and clients to
4598 version 1.4.2 or later at the same time. You must also run
4599 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4600 to update the configuration logs.
4602 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4603 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4604 - swab llog records of type '0' so we get proper header size/idx (5861)
4605 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4606 - fix rename of one directory over another leaking an inode (5953)
4607 - avoid SetPageDirty on 2.6 (5981)
4608 - don't re-add just-being-destroyed locks to the waiting list (5653)
4609 - when creating new directories, inherit the parent's custom
4610 striping settings if present parent (3048)
4611 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4612 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4613 - don't hold client locks on temporary worklist from l_lru (5666)
4614 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4615 - automatically evict dead clients (5921)
4616 - Update file size properly in create+truncate+fstat case (6196)
4617 - Do not unhash mountpoint dentries, do not allow removal of
4619 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4620 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4621 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4622 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4623 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4624 - Correctly handle reads of files with no objects (6243)
4625 - lctl recover will also mark a device active if deactivate used (5933)
4627 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4628 - allow --write-conf on an MDS with different nettype than client (5619)
4629 - don't write config llogs to MDS for mounts not from that MDS (5617)
4630 - lconf should create multiple TCP connections from a client (5201)
4631 - init scripts are now turned off by default; run chkconfig --on
4632 lustre and chkconfig --on lustrefs to use them
4633 - upcalls are no longer needed for clients to recover to failover
4635 - add --abort-recovery option to lconf to abort recovery on device
4637 - add support for an arbitrary number of OSTs (3026)
4638 - Quota support protocol changes.
4639 - forward compatibility changes to wire structs (6007)
4640 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4641 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4642 - improved "lustre status" script
4643 - initialize blocksize for non-regular files (6062)
4644 - added --disable-server and --disable-client configure options (5782)
4645 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4646 - Vanilla 2.4.29 support
4647 - increase maximum number of obd devices to 520 (6242)
4648 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4649 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4650 SLES 9 2.6.5 kernels
4651 - limit stripes per file to 160 (the maximum EA size) (6093)
4653 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4656 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4657 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4658 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4659 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4660 - small ext3 extents cleanups and fixes (5733)
4661 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4662 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4663 - avoid needless client->OST connect, fix handle mismatch (5317)
4664 - fix DLM error path that led to out-of-sync client, long delays (5779)
4665 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4666 - fix several locking issues related to i_size (5492,5624,5654,5672)
4667 - don't move pending lock onto export if it is already evicted (5683)
4668 - fix kernel oops when creating .foo in unlinked directory (5548)
4669 - fix deadlock in obdfilter statistics vs. object create (5811)
4670 - use time_{before,after} to avoid timer jiffies wrap (5882)
4671 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4672 - Do not leak request if server was not able to process it (5154)
4673 - If mds_open unable to find parent dir, make that negative lookup(5154)
4674 - don't create new directories with extent-mapping (5909, 5936)
4676 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4677 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4678 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4679 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4680 - support for RHEL4 kernel on i686 (5773)
4681 - provide error messages when incompatible logs are encountered (5898)
4683 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4684 * version 1.4.0.10 (1.4.1 release candidate 1)
4686 - don't keep a lock reference when lock is not granted (4238)
4687 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4688 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4689 - drop import inflight refcount on signal_completed_replay error (5255)
4690 - unlock page after async write error during send (3677)
4691 - handle missing objects in filter_preprw_read properly (5265)
4692 - no transno return for symlink open, don't save no-trasno open (3440)
4693 - don't try to complete elan receive that already failed (4012)
4694 - free RPC server reply state on error (5406)
4695 - clean up thread from ptlrpc_start_thread() on error (5160)
4696 - readahead could read extra page into cache that wasn't ejected (5388)
4697 - prevent races in class_attach/setup/cleanup/detach (5260)
4698 - don't dereference de->d_inode after l_dput of de (5458)
4699 - use "int" for stripe value returned from lock_to_stripe (5544)
4700 - mballoc allocation and error-checking fixes in 2.6 (5504)
4701 - block device patches to fix I/O request sizes in 2.6 (5482)
4702 - look up hostnames for IB nals (5602)
4703 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4704 - don't start multiple acceptors for the same port (5277)
4705 - fix incorrect LASSERT in mds_getattr_name (5635)
4706 - export a proc file for general "ping" checking (5628)
4707 - fix "lfs check" to not block when the MDS is down (5628)
4709 - service request history (4965)
4710 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4711 - create an "evict_client" /proc entry on OSTs, like the MDS has
4712 - fix mount usage message, return errors per mount(8) (5168)
4713 - change grep [] to grep "[]" in tests so they work in more UMLs
4714 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4715 - remove ancient LOV_MAGIC_V0 EA support (5047)
4716 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4717 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4718 - allow readahead allocations to fail when low on memory (5383)
4719 - mmap locking landed again, after considerable improvement (2828)
4720 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4722 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4725 - send OST transaction number in read/write reply to free req (4966)
4726 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4727 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4728 - the watchdog thread now runs as interruptible (5246)
4729 - flock/lockf fixes (but it's still disabled, pending 5135)
4730 - don't use EXT3 constants in llite code (5094)
4731 - memory shortage at startup could cause assertion (5176)
4733 - reorganization of lov code
4734 - single portals codebase
4736 - add extents/mballoc support (5025)
4737 - direct I/O reads in the obdfilter (4048)
4738 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4740 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4743 - send OST transaction number in read/write reply to free req (4966)
4744 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4745 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4746 - the watchdog thread now runs as interruptible (5246)
4747 - handle missing objects in filter_preprw_read properly (5265)
4748 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4749 - drop import inflight refcount on signal_completed_replay error (5255)
4750 - unlock page after async write error during send (3677)
4751 - return original error code on reconstructed replies (3761)
4752 - no transno return for symlink open, don't save no-trasno open (3440)
4754 - add pid to ldlm debugging output (4922)
4755 - bump the watchdog timeouts -- we can't handle 30sec yet
4756 - extra debugging for orphan dentry/inode bug (5259)
4758 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4761 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4762 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4763 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4764 - fix NULL dereference in /proc/sys/portals/routes (4827)
4765 - allow failed mdc_close() operations to be interrupted (4561)
4766 - stop precreate on OST before MDS would time out on it (4778)
4767 - don't send partial-page writes before EOF from client (4410)
4768 - discard client grant for sub-page writes on large-page clients (4520)
4769 - don't free dentries not owned by NFS code, check generation (4806)
4770 - fix lsm leak if mds_create_objects() fails (4801)
4771 - limit debug_daemon file size, always print CERROR messages (4789)
4772 - use transno after validating reply (3892)
4773 - process timed out requests if import state changes (3754)
4774 - update mtime on OST during writes, return in glimpse (4829)
4775 - add mkfsoptions to LDAP (4679)
4776 - use ->max_readahead method instead of zapping global ra (5039)
4777 - don't interrupt __l_wait_event() during strace
4779 - add software watchdogs to catch hung threads quickly (4941)
4780 - make lustrefs init script start after nfs is mounted
4781 - fix CWARN/ERROR duplication (4930)
4782 - return async write errors to application if possible (2248)
4783 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4784 - print NAL number in %x format (4645)
4785 - update barely-supported suse-2.4.21-171 series (4842)
4786 - support for sles 9 %post scripts
4787 - support for building 2.6 kernel-source packages
4788 - support for sles km_* packages
4790 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4793 - ignore -ENOENT errors in osc_destroy (3639)
4794 - notify osc create thread that OSC is being cleaned up (4600)
4795 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4796 - reconstruct ost_handle() like mds_handle() (4657)
4797 - create a new thread to do import eviction to avoid deadlock (3969)
4798 - let lconf resolve symlinked-to devices (4629)
4799 - don't unlink "objects" from directory with default EA (4554)
4800 - hold socknal file ref over connect in case target is down (4394)
4801 - allow more than 32000 subdirectories in a single directory (3244)
4802 - fix blocks count for O_DIRECT writes (3751)
4803 - OST returns ENOSPC from object create when no space left (4539)
4804 - don't send truncate RPC if file size isn't changing (4410)
4805 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4806 - bind to privileged port in socknal and tcpnal (3689)
4808 - rate limit CERROR/CWARN console message to avoid overload (4519)
4809 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4810 - basic mmap support (3918)
4811 - kernel patch series update from b1_4 (4711)
4813 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4816 - avoid crash during MDS cleanup with OST shut down (2775)
4817 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4818 - don't use bad inodes on the MDS (3744)
4819 - dynamic object preallocation to improve recovery speed (4236)
4820 - don't hold spinlock over lock dumping or change debug flags (4401)
4821 - don't zero obd_dev when it is force cleaned (3651)
4822 - print grants to console if they go negative (4431)
4823 - "lctl deactivate" will stop automatic recovery attempts (3406)
4824 - look for existing locks in ldlm_handle_enqueue() (3764)
4825 - don't resolve lock handle twice in recovery avoiding race (4401)
4826 - revalidate should check working dir is a directory (4134)
4828 - don't always mark "slow" obdfilter messages as errors (4418)
4830 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4833 - don't close LustreDB during write_conf until it is done (3860)
4834 - fix typo in lconf for_each_profile (3821)
4835 - allow dumping logs from multiple threads at one time (3820)
4836 - don't allow multiple threads in OSC recovery (3812)
4837 - fix debug_size parameters (3864)
4838 - fix mds_postrecov to initialize import for llog ctxt (3121)
4839 - replace config semaphore with spinlock (3306)
4840 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4841 - don't allow enqueue to complete on a destroyed export (3822)
4842 - down write_lock before checking llog header bitmap (3825)
4843 - recover from lock replay timeout (3764)
4844 - up llog sem before sending rpc (3652)
4845 - reduce ns lock hold times when setting kms (3267)
4846 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4847 - fix NULL deref and obd_dev leak on setup error (3312)
4848 - replace some LBUG about llog ops with error handling (3841)
4849 - don't match INVALID dentries from d_lookup and spin (3784)
4850 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4851 - fix invalid assertion in ptlrpc_set_wait (3880)
4853 - add libwrap support for the TCP acceptor (3996)
4854 - add /proc/sys/portals/routes for non-root route listing (3994)
4855 - allow setting MDS UUID in .xml (2580)
4856 - print the stack of a process that LBUGs (4228)
4858 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4861 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4862 - make sure to unset replay flag from failed open requests (3440)
4863 - if default stripe count is 0, use OST count for inode size (3636)
4864 - update parent mtime/ctime on client for create/unlink (2611)
4865 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4866 - free recovery state on server during a forced cleanup (3571)
4867 - unregister_reply for resent reqs (3063)
4868 - loop back devices mounting and status check on 2.6 (3563)
4869 - fix resource-creation race that can provoke i_size == 0 (3513)
4870 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4871 - more debugging for page-accounting assertion (3746)
4872 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4873 - avoid deadlock after precreation failure (3758)
4874 - fix race and lock order deadlock in orphan handling (3450, 3750)
4875 - add validity checks when grabbing inodes from l_ast_data (3599)
4877 - add /proc/.../recovery_status to obdfilter (3428)
4878 - lightweight CDEBUG infrastructure, debug daemon (3668)
4879 - change default OSC RPC parameters to be better on small clusters
4880 - turn off OST read cache for files smaller than 32MB
4881 - install man pages and include them in rpms (3100)
4882 - add new init script for (un)mounting lustre filesystems (2593)
4883 - run chkconfig in %post for init scripts (3701)
4884 - drop scimac NAL (unmaintained)
4886 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4889 - clean kiobufs before and after use (3485)
4890 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4891 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4892 - change init script to fail more clearly if not run as root (1528)
4893 - allow clients to reconnect during replay (1742)
4894 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4895 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4896 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4897 - fix cond_resched (3554)
4898 - only evict unfinished clients after recovery (3515)
4899 - allow bulk resend, prevent data loss (3570)
4900 - dynamic ptlrpc request buffer allocation (2102)
4901 - don't allow unlinking open directory if it isn't empty (2904)
4902 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4903 - remove extraneous obd dereference causing LASSERT failure (3334)
4904 - don't use get_cycles() when creating temp. files on the mds (3156)
4905 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4906 - handle EEXIST for set-stripe, set proper directory name (3336)
4908 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4909 - fix ksocknal_fmb_callback() error messages (2918)
4911 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4914 - don't copy lvb into (possibly NULL) reply on error (2983)
4915 - don't deref dentry after dput, don't free lvb on error (2922)
4916 - use the kms to determine writeback rpc length (2947)
4917 - increment oti_logcookies when osc is inactive (2948)
4918 - update client's i_blocks count via lvb messages (2543)
4919 - handle intent open/close of special files properly (1557)
4920 - mount MDS with errors=remount-ro, like obdfilter (2009)
4921 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4922 - don't use cancelling-locks' kms values (2947)
4923 - use highest lock extent for kms, not last one (2925)
4924 - don't dereference ERR_PTR() dentry in error handling path (3107)
4925 - fix thread race in portals_debug_dumplog() (3122)
4926 - create lprocfs device entries at setup instead of at attach (1519)
4927 - common AST error handler, don't evict client on completion race (3145)
4928 - zero nameidata in detach_mnt in 2.6 (3118)
4929 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4930 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4931 - initialize RPC timeout timer earlier for 2.6 (3219)
4932 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4933 - print nal/nid for unknown nid (3258)
4934 - additional checks for oscc recovery before doing precreate (3284)
4935 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4936 - don't crash in mdc_close for bad permissions on open (3285)
4937 - zero i_rdev for non-device files (3147)
4938 - clear page->private before handing to FS, better assertion (3119)
4939 - tune the read pipeline (3236)
4940 - fix incorrect decref of invalidated dentry (2350)
4941 - provide read-ahead stats and refine rpc in flight stats (3328)
4942 - don't hold journal transaction open across create RPC (3313)
4943 - update atime on MDS at close time (3265)
4944 - close LDAP connection when recovering to avoid server load (3315)
4945 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4946 - don't leak open file on MDS after open resend (3325)
4947 - serialize filter_precreate and filter_destroy_precreated (3329)
4948 - loop device shouldn't call sync_dev() for nul device (3092)
4949 - clear page cache after eviction (2766)
4950 - resynchronize MDS->OST in background (2824)
4951 - refuse to mount the same filesystem twice on same mountpoint (3394)
4952 - allow llmount to create routes for mounting behind routers (3320)
4953 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4954 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4955 - fix rare problem with rename on htree directories (3417)
4957 - allow default OST striping configuration per directory (1414)
4958 - fix compilation for qswnal for 2.6 kernels (3125)
4959 - increase maximum number of MDS request buffers for large systems
4960 - change liblustreapi to be useful for external progs like lfsck (3098)
4961 - increase local configuration timeout for slow disks (3353)
4962 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4964 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4967 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4968 - don't overwrite extent policy data in reply if lock was blocked (2901)
4969 - drop filter export grants atomically with removal from device (2663)
4970 - del obd_self_export from work_list in class_disconnect_exports (2908)
4971 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4972 - swab reply message in mdc_close, other PPC fixes (2464)
4973 - fix destroying of named logs (2325)
4974 - overwrite old logs when running lconf --write_conf (2264)
4975 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4976 - fix race in target_handle_connect (2898)
4977 - mds_reint_create() should take same inode create lock (2926)
4978 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4979 - don't close files for self_export to avoid uninitialized obd (2936)
4980 - allow MDS with the same name as client node (2939)
4981 - hold dentry reference for closed log files for unlink (2325)
4982 - reserve space for all logs during transactions (2059)
4983 - don't evict page beyond end of stripe extent (2925)
4984 - don't oops on a deleted current working directory (2399)
4985 - handle hard links to targets without a parent properly (2517)
4986 - don't dereference NULL lock when racing during eviction (2867)
4987 - don't grow lock extents when lots of conflicting locks (2919)
4989 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4992 - account for cache space usage on clients to avoid data loss (974)
4993 - lfsck support in lustre kernel code (2349)
4994 - reduce journal credits needed for BRW writes (2370)
4995 - orphan handling to avoid losing space on client/server crashes
4996 - ptlrpcd can be blocked, stopping ALL progress (2477)
4997 - use lock value blocks to assist in proper KMS, faster stat (1021)
4998 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4999 - recovery for initial connections (2355)
5000 - fixes for mds_cleanup_orphans (1934)
5001 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5002 - block all file creations until orphan recovery completes (1901)
5003 - client remove rq_connection from request struct (2423)
5004 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5005 - recovery timer race (2670)
5006 - mdc_close recovey bug (2532)
5007 - ptlrpc cleanup bug (2710)
5008 - mds timeout on local locks (2588)
5009 - namespace lock held during RPCs (2431)
5010 - handle interrupted sync write properly (2503)
5011 - don't try to handle a message that hasn't been replied to (2699)
5012 - client assert failure during cleanup after abort recovery (2701)
5013 - leak mdc device after failed mount (2712)
5014 - ptlrpc_check_set allows timedout requests to complete (2714)
5015 - wait for inflight reqs when ptlrpcd finishes (2710)
5016 - make sure unregistered services are removed from the srv_list
5017 - reset bulk XID's when resending them (caught by 1138 test)
5018 - unregister_bulk after timeout
5019 - fix lconf error (2694)
5020 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5021 - readahead locks pages, leaves pending causing memory pressure (2673)
5022 - increase OST request buffers to 4096 on large machines (2729)
5023 - fix up permission of existing directories in simple_mkdir (2661)
5024 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5025 - don't assign transno to errored transactions (2742)
5026 - don't delete objects on OST if given a bogus objid from MDS (2751)
5027 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5028 - if rq_no_resend, then timeout request after recovery (2432)
5029 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5030 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5031 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5032 - MDS recovery shouldn't delete objects that it has given out (2730)
5033 - if enqueue arrives after completion, don't clobber LVB (2819)
5034 - don't unlock pages twice when trigger_group_io returns error (2814)
5035 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5036 - don't write pages to disk if there was an error (1450)
5037 - don't ping imports that have recovery disabled (2676)
5038 - take buffered bytes into account when balancing socknal conn (2817)
5039 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5040 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5041 - remove little-endian swabbing of llog records (1987)
5042 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5043 - retry reposting request buffers if they fail (1191)
5044 - grow extent at grant time to avoid granting a revoked lock (2809)
5045 - lock revoke doesn't evict page if covered by a second lock (2765)
5046 - disable VM readahead to avoid reading outside lock extents (2805)
5048 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5049 - updated kernel patches for hp-2.4.20 kernel (2681)
5051 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5054 - fix truncated write corruption (2366)
5055 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5057 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5058 - ensure len(uuid) < 37 in lmc (1171)
5059 - fix ia64 OOPS in llog_test (2255)
5060 - zero end of page at obdfilter for partial page writes (2648)
5061 - don't leave stale dentries around after renames (bug 2428)
5062 - fix timeouts when evicting a client with a single lock held (2642)
5063 - set deadline for the initial HELLO message to drain (2634)
5064 - print out dotted-quad IP addresses in the socknal (2302)
5066 - additional debugging for MDS client eviction problem (2443)
5067 - fix mkfsoptions support for osts (2603, 2604)
5069 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5072 - add series for the vanilla 2.6.0 kernel
5073 - add series for the vanilla 2.4.24 kernel
5074 - add series for a cray x86/64 UL kernel drop
5075 - fix xattr patches for the vanilla 2.4.19 series
5077 - generate true UUIDs in lmc (1171)
5078 - have portals stack dumping break in UML (2466)
5079 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5080 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5081 - wake callback waiting threads on client eviction (2460)
5082 - Add --ptldebug and --subsystem to lmc (1719)
5083 - update assertion to allow safe interrupt allocation
5084 - set rq_no_resend for cancel requests (2432)
5085 - recalculate ptlrpcd timeout after resend (2494)
5086 - call vfs_rmdir when removing pending directories (2368)
5087 - fix renaming a file to itself (2429)
5088 - lmc creates a default one-stripe lov (2454)
5089 - expand procfs space to handle large clusters (2326)
5090 - increase UML stack to avoid overflow
5091 - update lconf's list of debug and subsystem masks
5092 - fix lfs find --obd (2510)
5093 - /proc tunable for disabling filter read caching (2591)
5094 - stop rpm packages from altering slapd.conf (2301)
5095 - disable nagle in the socknal under 0conf (2578)
5096 - choose mds inode size based on stripe count (2572)
5097 - fix kernel-source rpm problems (2516)
5099 - add --disable-doc to avoid pdf generation (2421)
5100 - update documentation, tests, type-os, comments
5101 - avoid format warnings on ia64
5102 - remove the TOE NAL
5103 - tiny code cleanups by removing unused fields
5105 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5108 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5109 - remove the most blatant lies from BUILDING file (2371)
5110 - change default debug level to reasonable production setting
5111 - reduce client side cache size to reduce cache flush time
5112 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5113 - make TCP zerocopy and pinger support enabled by default (2476)
5114 - sync writes completed after process exits caused crashes (2319)
5115 - maintain correct mount count on the MDS (2356)
5116 - backout 1557, because 2316 wasn't really fixed
5117 - better file I/O statistics gathering in /proc
5118 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5119 - another kernel patch to fix zero-copy TCP function export
5120 - don't take duplicate lock when processing re-sent getattr (2420)
5121 - lctl uses obd_self_export instead of creating new conn (2353)
5122 - MDS/OST recovery case which requires object creation asserted (2425)
5123 - move lfs from /usr/sbin to /usr/bin in packages
5124 - fix race between mds_client_add and mds_client_free (2417)
5125 - use kmalloc instead of slabs in portals (2430)
5126 - don't create duplicate records when a failover MDS is present (2442)
5127 - remove unnecessary mount age check (2332)
5128 - don't remove directory inodes from locks prematurely (2451)
5129 - don't break if MDS service name is the same as hostname (2103)
5130 - fix races in client write RPC generation when cache full (2482)
5132 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5135 - remove now-unused request->rq_obd (278)
5136 - if an allocation fails, print out how much memory we've used (1933)
5137 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5138 - add the "configurable stack size" patch to most series files (1256)
5139 - ability to write large log records, for 100+ OST configs (2306)
5140 - fix NULL deref when filter_prep fails (2314)
5141 - fix operator precedence error in filter_sync
5142 - dynamic allocation of socknal TX descriptors (2315)
5143 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5144 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5145 - ensure that CWARN messages in Portals always get to the syslog
5146 - __init/__exit are not for prototype decls (ldlm_init/exit)
5147 - x86-64 compile warning fixes
5148 - fix gateway LMC keyword conflict (2318)
5149 - fix MDS lock inversions in getattr/reint paths (1844)
5150 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5151 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5152 - DLM race condition prevented some lock evictions (2328)
5153 - ENOMEM detection and retry on socknal sends (2230)
5154 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5155 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5156 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5157 - better error messages when a client is rejected during recovery (1505)
5158 - avoid cancelling locks which were never granted, after failure (2330)
5159 - fix i_sem/journal inversion in mds_client_add (2333)
5160 - fix truncate/getattr lock cycle deadlock (2334)
5161 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5162 - fix two rare exit paths which could leak an l_lock() ref (2321)
5163 - fencepost error in MDS/OST orphan recovery (2226)
5164 - make log record alignment 8 bytes (1988)
5165 - lstripe now fails when requested offset > ost_count (2237)
5166 - ensure that all kernel series have a complete list.h (1607)
5167 - fix crashes in special-file operations (2316)
5168 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5170 - allow configurable automake binary, for testing new versions
5171 - small update to the lfs documentation
5173 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5175 * fix negative export reference count in fsfilt_sync (2312)
5177 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5178 * release candidate 0.9.1
5180 - orphans are moved into the PENDING directory for possible recovery
5181 - replayed opens now open by fid for orphan/rename safety (1042)
5182 - last close of an orphan inode generates a transno (683)
5183 - chdir() and mount() now pin the directory entry (1020)
5184 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5185 - discard very old requests without processing them (1502)
5186 - remove some common, well-understood CERRORs (1505)
5187 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5188 - clear "grant" flags in OST replies until OST grant code lands (1644)
5189 - fix read performance by not clobbering i_blksize on client (1598)
5190 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5191 - make LRU size a /proc tunable, clears locks when reduced (707)
5192 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5193 - server locks take references on exports now (1558)
5194 - build fixes for 2.4.20-rh trees (1663)
5195 - return an error from lov_create if all OSCs are inactive (1751)
5196 - fix import levels when a reconnect happens without a timeout (1597)
5197 - exit early from mds_open if we get a lookup error (1749)
5198 - partial page read at EOF wouldn't wait for disk before sending (1642)
5199 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5200 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5201 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5202 - write/truncate lock inversion (1639)
5203 - Don't auto-load obdclass, portals modules during cleanup (1495)
5204 - fix timestamps from jumping to "now" (1763)
5205 - extra journal assertions (1648)
5206 - add an extra multiunlink test (1771)
5207 - fix read_record/write_record API (1776)
5208 - fix leak of offset_extent, possible incorrect i_size later (1772)
5209 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5210 - Debugging for the kqswnal_get_idle_tx problems (1820)
5211 - Allow recovery to be attempted multiple times (1536)
5212 - Write out MDS last_rcvd file after it is first created (1600)
5213 - Fix tx_descriptor leak in failed transmit situations (1827)
5214 - ext3 journaling fixes for assertion failure after IO error (1871)
5215 - class_export_put() on freed export after completion AST error (1896)
5216 - Fix revalidate looping in VFS (1322)
5217 - Don't access a freed export during MDS_REINT timeout (1521)
5218 - Add open-unlink recovery support on the MDS (1673,1764)
5219 - Return an error if no MDS data was read from last_rcvd (1946)
5220 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5221 - Don't setup a disk device that doesn't match exported UUID (317)
5222 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5223 - avoid committing NULL handle in force close
5224 - local.sh is now a one-stripe LOV configuration
5225 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5226 - fix ext3 htree duplicate directory entry corruption (1516)
5227 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5228 - update to new LOV EA format (2097)
5229 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5231 2003-06-15 Phil Schwan <phil@clusterfs.com>
5234 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5235 - per-import/export recovery handling (958, 931, 959)
5236 - multiple last-rcvd slots, for serving multiple FSes (949)
5237 - connections are again shared between multiple imp/exports (963, 964)
5238 - "umount -f" would hang if any requests needed to be sent (393, 978)
5239 - avoid pinning large req buffer by copying for queued messages (989)
5240 - add "uuid" to "lctl device" command to help upcalls (991)
5241 - "open" RPCs with transnos would confuse recovery counters (1037)
5242 - do proper endian conversion of all wire messages (288, 340, 891)
5243 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5244 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5245 - fix ptlrpc_connection leak in target_handle_connect (1174)
5246 - fix import refcounting bug in OST and MDS cleanup (1134)
5247 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5248 - fix very unlikely obd_types race condition (501)
5249 - remove osc_open hack for echo_client (1187)
5250 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5251 - a failure in read_inode2 leads to deadlock (1139)
5252 - cancel ack-locks as soon as transaction is committed (1072)
5253 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5254 - make sure to commitrw after any preprw to avoid deadlock (1162)
5255 - failing to execute a file in a lustre FS would lock inode (1203)
5256 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5257 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5258 - obd_uuid2tgt would walk off the end of the list (1255)
5259 - on IA64 the osc would give portals incorrect bulk size (1258)
5260 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5261 - fix lock inversion caused by new llite matching code (1282)
5262 - limit the number of dirty pages on a client to 10MB (1286)
5263 - timed out locks were not being corrected cancelled (1289)
5264 - fix O_DIRECT above 4GB on IA-32 (1292)
5265 * major user-visible changes
5266 - fail out/fail over policy now controlled by the upcall (993)
5268 - add OBD_PING to check server availability and failure (954)
5269 - lustre messages are now sent in sending host order (288, 340, 891)
5270 - add eadatalen to MDS getattr reply (340)
5271 - OST read replies may contain second buffer, with per-page status (593)
5273 2003-03-11 Phil Schwan <phil@clusterfs.com>
5276 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5277 - failed open()s could cause deadlock; fixed (867, 869)
5278 - stop cancelling OST locks when files are closed (481)
5279 - overlapping XID spaces caused network corruption (851, 853)
5280 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5281 - setattr_raw added, to do single-RPC, server-side setattrs
5282 - lmc/lconf syntax change for OST UUIDs
5283 - fix crashy race condition between ptlrpc_free_req and osc_close
5284 - don't use request in mdc_enqueue if we hit a timeout (889)
5285 - don't set the inode i_size for regular files from the MDS (896)
5286 - handle out of order completion AST (842)
5287 - don't LBUG if a lock request times out after receiving AST (913)
5288 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5289 - if a bad lock AST arrives, send an error instead of dropping entirely
5290 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5291 - fix leak in bulk IO when only partially completed (899, 900, 926)
5292 - fix O_DIRECT for ia64 (55)
5293 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5294 - C-z after timeout could hang a process forever; fixed (977)
5296 - client-side I/O cache (678, 924, 929, 941, 970)
5298 - READPAGE and SETATTRs which don't take server-side locks get
5301 2003-02-11 Phil Schwan <phil@clusterfs.com>
5304 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5305 - Fix fsfilt_extN_readpage() to read a full page of directory
5306 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5307 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5308 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5309 - Most intent operations were removed, in favour of a new RPC mode
5310 that does a single RPC to the server and bypasses most of the VFS
5311 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5312 - Aggressively cancel local locks on DLM servers
5313 - mds_reint_unlink sends EA to the client if it's the last nlink.
5314 client uses that EA to unlink OST objects.
5315 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5316 - recursive symlinks were fixed (439)
5317 - fixed NULL deref in DEBUG_REQ
5318 - filter_update_lastobjid no longer calls sync, which annoyed extN
5319 - fixed multi-client small-writes to a single file problem (445)
5320 - fixed mtime updates during file writes (607)
5321 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5322 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5323 - fixed Linux OST size reporting problem (444, 656)
5324 - OST now updates object mtime with writes or setattr (607, 619)
5325 - client verifies file size before zeroing page past EOF (445)
5326 - OST now writes last allocated objid to disk with allocation (108)
5327 - LOV on echo now works (409)
5329 - mds_reint_unlink sends a new buffer, with the EA included. this
5330 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5331 set if a regular file was being unlinked, and it was the last link
5332 - use PtlGet from the target for bulk writes (315)
5333 - OST now updates object mtime with writes or setattr (607, 619)
5334 - LDLM now has a grant-time callback to revalidate locked items, if
5336 - Many MDS operations were reorganized to combat race conditions
5338 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5339 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5341 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5344 - Fully reactivate OST imports after reconnection (512, others)
5345 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5346 - More graceful error handling for truncating on dead OST (515)
5347 - Don't error out unless we're actually accessing dead stripes (474)
5348 - Fix garbage sizes when stripes are missing (410)
5349 - LRU counters were broken, causing constant lock purge (433, 432)
5350 - garbage on read from stripes with failed OSTs (441)
5351 - mark OSCs as active before reconnecting during recovery (438)
5352 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5353 - lfind did not preserve OST order in output (443)
5354 - symlinks cause hung clients, incorrect data (439)
5355 - stop dereferencing request after dropping refcount (457)
5356 - don't LASSERT(spin_is_locked) on non-SMP (455)
5357 - fixes for many rename() bugs
5358 - fstat didn't correctly synchronize attributes (399)
5359 - server must handle lock cancellation during blocking AST prep (487)
5360 - bulk descriptors were free()d too soon (511)
5361 - fix paths in lconf, which would load incorrect modules (451, 507)
5362 - fix confusing lconf 'host not found' error message (386)
5363 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5364 - fix race condition in mdc_blocking_ast() for inode access (526)
5365 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5366 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5367 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5368 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5369 - fix lstripe in several ways (kernel side) (527)
5370 - fix request leak in ldlm_cli_enqueue (262)
5371 - incorrect OSC was marked inactive after OST failure
5372 - call mds_fs_cleanup before unmounting filesystem (524)
5373 - fix races between taking ns_lock and ldlm_lock_change_resource
5374 - fix races updating LOV export open file list
5375 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5376 - fix recovery NULL deref in ldlm_cli_cancel_unused
5377 - fix some DLM races by using new hash table for lock handles (419)
5378 - permit the client to specify desired inodes, at replay
5379 - duplicate requests when we queue them for replay reintegration
5380 - fix last_rcvd offset calculation
5381 - sync after each recovered transaction, so we always make progress
5382 - never, not always, ERESTART requests without transnos
5383 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5385 - skip replay if the MDS says that the client is already connected
5386 - don't check for a recovery-enabled export to match lctl's UUID
5387 - don't INC_USE_COUNT for phantom exports
5388 - don't crash when cleaning up phantom exports (567)
5389 - don't double-finish or set replay data for errored mdc_open requests
5390 - abort requests when they time out, so we don't get old replies
5391 - send/receive replies for AST messages again
5392 - if the client says that it doesn't have the lock, cancel it on the
5394 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5395 mark it as destroyed, it will all work out when decref is called
5396 - fix module use counts (22, 581)
5398 - ASTs now expect a reply (server cancels lock on error reply)
5400 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5403 - fix many simultaneous client startup (392)
5404 - fix dentry->d_it clobbering
5405 - credentials weren't being shipped for readdir/getattr operations
5406 - remove invalid assertions triggered during some concurrent MD
5408 - proper Lustre versions added (336, 389)
5409 - fix memory leak for create error case (398)
5410 - fix LOV locking bug that would get cli/srv out of sync
5411 - fix echo client over LOV (409)
5412 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5413 - fix double-O_EXCL intent crash (424)
5414 - avoid sending multiple lock CANCELs (352)
5416 - MDS can do multi-client recovery (modulo bugs in new code)
5418 - many updates, edits, cleanups
5420 2002-11-18 Phil Schwan <phil@clusterfs.com>
5423 - fix null d_it dereference (346)
5424 - fix full OST/dbench hang (333)
5425 - fix permission problem with file removal (286)
5426 - fix removal of OSCs from LOV when they fail
5427 - fix NULL deref during bulk timeout (214)
5428 - fix problems related to multiple filesystems on one MDS (241)
5429 - fixed serious subtle metadata locking bugs
5430 - free locks on clients when inodes are removed due to memory
5432 - fix inode pointer in lock data (285)
5433 - partial support for multiple MDS on a single host (241)
5434 - data locks weren't cancelled at clear_inode time (290, 311)
5435 - intent locks could lead to unbounded lock growth (205)
5436 - added a maximum lock count, an LRU list, and a flusher
5437 - fix multiple rename (365)
5438 - properly abstracted the echo client
5439 - OSC locked 1 byte too many; fixed
5440 - rewrote brw callback code:
5441 - fixed recovery bugs related to LOVs (306)
5442 - fixed too-many-pages-in-one-write crash (191)
5443 - fixed (again) crash in sync_io_timeout (214)
5444 - probably fixed callback-related race (385)
5446 - Add capability to MDS protocol
5447 - LDLM cancellations and callbacks on different portals
5449 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5452 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5453 - defer open object destruction to close time (601981, 138)
5454 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5455 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5456 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5457 - don't call obd_setattr in ll_file_release for destroyed objects
5458 * protocol change to lustre_msg: move |version| and add |flags|
5459 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5460 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5461 * added replay of create, unlink, link and rename operations during
5462 MDS failover; recovery should be much more robust now
5463 * remove failed OSCs from LOVs (only lov_create uses this so far)
5464 * the lustre-HOWTO was brought (more) up to date (582544)
5466 2002-10-23 Phil Schwan <phil@clusterfs.com>
5469 - in-use dentries weren't being reused properly (617851)
5470 - prevent multiple LDLM setup (599178)
5471 - fix LOV size calculations for truncate (617853)
5472 - fix client handling of MDS intent errors (POSIX)
5473 - fix permission bug in lovstripe.c test (624321)
5474 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5475 - truncate past end of file could corrupt data
5476 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5477 - a race in recovery could return ETIMEDOUT to apps (623947)
5478 - building outside the source directory was fixed
5479 * the lustre-HOWTO was brought (more) up to date (582544)
5480 * major progress was made on recovery functionality
5482 2002-10-10 Phil Schwan <phil@clusterfs.com>
5485 - recovery deadlock fix
5486 - rm -rf causes LBUG fix (617817)
5487 - file open by multiple tasks fix (618962)
5488 - directory permissions bugs (602707 and 620007)
5489 - journal_stop fixed with locking (611313)
5490 - O_APPEND failures resolved (618273, perhaps 614459)
5491 - lconf PATH fix (619770)
5492 - IA64 build fix (621450)
5493 - RPC buffer sizes scale with amount of memory
5495 2002-10-01 Phil Schwan <phil@clusterfs.com>
5498 - locks would be cancelled without throwing away data pages,
5499 resulting in inconsistent data (605627)
5500 - inode attributes were not always being refreshed (605627, 612449)
5501 - lconf now continues to cleanup after lctl reports an error
5502 - MDS now enforces user permissions (602707)
5503 - lprocfs cleanup fixed, but not yet enabled (614157)
5504 - fixed infinite server hang, should a client not respond to an AST
5505 - avoid going into recovery if user calls readlink() with a buffer
5506 that's too small (613941)
5507 - AST RPCs no longer require replies (614867) -- this may be changed
5508 - don't crash server if client sends an IOV that's too big (611336)
5509 - fixed lock conversion deadlock (611892)
5510 - fixed the following of symlinks (614622)
5511 * recovery: the server can remove locks from a client that dies, other
5512 clients can make progress
5513 * more extN patch fixes
5514 * compile-time configurable ptlrpc buffer allocations
5516 - collaborative read cache document
5517 - Lustre Lite Performance CDR document-in-progress
5519 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5522 - fix typo in patch-2.4.18
5524 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5527 - clear ptlrpc request each time in handle_incoming_request()
5528 - unlink of files now destroys the object on the OST
5530 2002-09-19 Peter Braam <braam@clusterfs.com>
5532 * add hard link support
5533 * change obdfile creation method
5534 * kernel patch changed
5536 2002-09-19 Peter Braam <braam@clusterfs.com>
5539 - stack overflow bug in extN fixed
5541 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5543 * documentation updates
5544 - add man pages for config tools
5545 - update tests/README to describe testing with new config tools
5546 - finish metadata API descriptions
5547 * bug fixes and cleanups
5548 - statfs workaround for 16TB limit
5549 - LOV stripe allocation improved, can stripe on subset of OSTs
5550 - LOV file size/IO offset was wrong for files > 4GB in size
5551 - object EA data was being dropped, caused files to be unreadable
5552 - memory overflow with non-LOV OST caused memory corruption
5553 - fixed regression tests to work with new config tools, obdfilter
5554 - fixed bug when directory size became larger than 1 block
5555 - fixed bug (for single client case) when PWD was deleted
5556 - invalidate local directory pages when doing intent-based ops
5557 - avoid LDLM oops when lock callback contained bad data
5559 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5561 * documentation updates
5562 * bug fixes and cleanups
5563 - configuration tools
5566 - 64-bit compile warnings
5567 - 64-bit internal statfs data
5569 * test_brw on persistent OST devices
5571 * lprocfs (disabled)
5573 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5575 * documentation updates
5576 * bug fixes and cleanups
5577 * configuration tools
5579 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5582 * many small fixes to 0_5_4
5583 * io/network handling
5584 * thinkos in MDS operations
5586 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5589 * crucial basic fixes to 0.5.3
5590 * IOR, Iozone work over Elan
5593 2002-08-07 Phil Schwan <phil@clusterfs.com>
5594 * version 0_5_3, our first alpha
5595 * we use the new Portals iovs
5596 * documentation updates
5597 * bug fixes and cleanups
5598 * small changes in the DLM wire protocol
5600 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5601 * version 0_5_1 with some initial stability,
5602 * locking on MD and file I/O.
5603 * documentation updates
5604 * several bug fixes since 0.5.0
5605 * small changes in wire protocol
5607 2002-07-18 Phil Schwan <phil@clusterfs.com>
5609 * delivered as Lustre Light Alpha
5610 * fixed a crash after handling invalid MDS requests
5611 * fixed directory pages for architectures with non-4k pages sizes
5613 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5614 * release version v0_4_4
5615 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5617 * Fixes a number of interruption problems with OST operations.
5618 * Update documentation for portals header changes
5619 * Move all wire protocol structs/defines to lustre_idl.h
5620 * Fixes symlink length bug.
5621 * Add tcpdump to repository.
5623 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5624 * release version v0_4_3
5625 * Fixes statfs for inodes on extN.
5626 * Fixes bug in runtests which would delete /etc/hosts.
5627 * Use 64-bit object IDs wherever possible (not into VFS though)
5628 Remove ost_get_info, which is unused by lustre, and out of date.
5630 2002-07-03 Peter Braam <braam@clusterfs.com>
5631 * release version v0_4_2 Fixes a lookup error (type not passed)
5632 * move forward to head of Portals
5633 * move forward to latest Lustre kernel
5635 2002-06-25 Peter Braam <braam@clusterfs.com>
5636 * release version v0_4_1. Hopefully stable on single node use.