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.
19 Description: more cleanup in mds_lov
20 Details : not send LOV EA under replay, we can't know about they size at this
21 time. Don't allow client connect to mds before any ost connected,
22 for avoid problems with LOV EA size and returning EIO to client.
24 Severity : enhancement
26 Description: Interoperability at server side (Disk interoperability)
28 Severity : enhancement
30 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
32 Severity : enhancement
34 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
36 Severity : enhancement
38 Description: New client IO stack (CLIO).
40 Severity : enhancement
42 Description: Commit on sharing. Eliminate inter-client dependencies between
43 uncommitted transactions by doing transaction commits.
44 Thereby clients may recovery independently.
47 Frequency : Create a symlink file with a very long name
49 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
50 Details : If there is no extra space in the request for early cancels,
51 ldlm_req_handles_avail() returns 0 instead of a negative value.
53 Severity : enhancement
55 Description: Add /proc entry for import status
56 Details : The mdc, osc, and mgc import directories now have
57 an import directory that contains useful import data for debugging
60 Severity : enhancement
62 Description: Re-disable certain /proc logging
63 Details : Enable and disable client's offset_stats, extents_stats and
64 extents_stats_per_process stats logging on the fly.
67 Frequency : Only on FC kernels 2.6.22+
69 Description: oops in statahead
70 Details : Do not drop reference count for the dentry from VFS when lookup,
71 VFS will do that by itself.
73 Severity : enhancement
75 Description: Generic /proc file permissions
76 Details : Set /Proc file permissions in a more generic way to enable non-
77 root users operate on some /proc files.
81 Description: Hitting mdc_commit_close() ASSERTION
82 Details : Properly handle request reference release in
83 ll_release_openhandle().
87 Description: quota recovery deadlock during mds failover
88 Details : This patch includes att18982, att18236, att18237 in bz14840.
90 1. fix osts hang when mds does failover with quotaon
91 2. prevent watchdog storm when osts threads wait for the
96 Frequency : only patchless client
97 Description: add workaround for race between add/remove dentry from hash
99 Severity : enhancement
101 Description: Allow OST glimpses to return PW locks
105 Description: LBUG when llog conf file is full
106 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
110 Description: Prevent import from entering FULL state when server in recovery
114 Description: service mount cannot take device name with ":"
115 Details : Only when device name contains ":/" will mount treat it as
121 Description: replace ptlrpcd with the statahead thread to interpret the async
122 statahead RPC callback
126 Frequency : on recovery
127 Description: I/O failures after umount during fail back
128 Details : if client reconnected to restarted server we need join to recovery
129 instead of find server handler is changed and process self
130 eviction with cancel all locks.
132 Severity : enhancement
134 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
136 Severity : enhancement
138 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
140 Severity : enhancement
142 Description: Upcall on Lustre log has been dumped
143 Details : Allow for a user mode script to be called once a Lustre log has
144 been dumped. It passes the filename of the dumped log to the
145 script, the location of the script can be specified via
146 /proc/sys/lnet/debug_log_upcall.
151 Description: avoid idr_remove called for id which is not allocated.
152 Details : Move assigment s_dev for clustered nfs to end of initialization,
153 to avoid problem with error handling.
158 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
159 Details : When connection is reused this not moved from CONN_UNUSED_HASH
160 into CONN_USED_HASH and this prodice warning when put connection
161 again in unused hash.
164 Severity : enhancement
166 Description: File striping can now be set to use an arbitrary pool of OSTs.
168 Severity : enhancement
170 Description: Export bytes_read/bytes_write count on OSC/OST.
174 Description: Early reply size mismatch, MGC loses connection
175 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
176 the connect flags are properly negotiated.
181 Description: quotas are not honored with O_DIRECT
182 Details : all writes with the flag O_DIRECT will use grants which leads to
183 this problem. Now using OBD_BRW_SYNC to guard this.
187 Description: add quota statistics
188 Details : 1. sort out quota proc entries and proc code.
189 2. add quota statistics
191 Severity : enhancement
193 Description: enable quota support for HEAD.
197 Description: Properly propagate oinfo flags from lov to osc for statfs
198 Details : restore missing copy oi_flags to lov requests.
200 Severity : enhancement
202 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
203 : l_getgroups(8), lst(8), routerstat(8)
204 Details : included man pages for llobdstat(8), llstat(8),
205 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
207 Severity : enhancement
209 Description: configure's --enable-quota should check the
210 : kernel .config for CONFIG_QUOTA
211 Details : configure is terminated if --enable-quota is passed but
212 : no quota support is in kernel
216 Description: Remove portals compatibility
217 Details : Remove portals compatibility, not interoperable with releases
222 Description: Resolve device initialization race
223 Details : Prevent proc handler from accessing devices added to the
224 obd_devs array but yet be intialized.
226 Severity : enhancement
228 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
230 Severity : enhancement
232 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
237 Description: Using wrong pointer in osc_brw_prep_request
238 Details : Access to array[-1] can produce panic if kernel compiled with
239 CONFIG_PAGE_ALLOC enabled
243 Description: Client runs out of low memory
244 Details : Consider only lowmem when counting initial number of llap pages
248 Description: *optional* service tags registration
249 Details : if the "service tags" package is installed on a Lustre node
250 When the filesystem is mounted, a local-node service tag will
251 be created. See http://inventory.sun.com/ for more information
252 about the Service Tags asset management system.
256 Description: Kernel BUG tries to release flock
257 Details : Lustre does not destroy flock lock before last reference goes
258 away. So always drop flock locks when client is evicted and
259 perform unlock regardless of successfulness of speaking to MDS.
263 Description: add refcount protection for osc callbacks, avoid panic on shutdown
267 Description: sanity test 65a fails if stripecount of -1 is set
268 Details : handle -1 striping on filesystem in ll_dirstripe_verify
273 Description: ASSERTION(CheckWriteback(page,cmd)) failed
274 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
277 Severity : enhancement
279 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
283 Description: do not process already freed flock
284 Details : flock can possibly be freed by another thread before it reaches
285 to ldlm_flock_completion_ast.
289 Description: LBUG during stress test
290 Details : Need properly lock accesses the flock deadlock detection list.
294 Description: oops in page fault handler
295 Details : kernel page fault handler can return two special 'pages' in error
296 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
300 Description: timeout with invalidate import.
301 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
302 should be handled by ptlrpcd. This produce long age waiting and
303 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
305 Severity : enhancement
307 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
311 Description: enable MGS and MDT services start separately
312 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
313 a co-located MGS without starting the MGS, which is a complement
314 to 'nosvc' mount option.
318 Frequency : after recovery
319 Description: precreate to many object's after del orphan.
320 Details : del orphan st in oscc last_id == next_id and this triger growing
321 count of precreated objects. Set flag LOW to skip increase count
322 of precreated objects.
326 Frequency : rare, on clear nid stats
327 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
328 Details : when clean nid stats sometimes try destroy live entry,
329 and this produce panic in free.
333 Description: Stack overflow during MDS log replay
334 ease stack pressure by using a thread dealing llog_process.
338 Description: wait until IO finished before start new when do lock cancel.
339 Details : VM protocol want old IO finished before start new, in this case
340 need wait until PG_writeback is cleared until check dirty flag and
341 call writepages in lock cancel callback.
343 Severity : enhancement
345 Description: using special macro for print time and cleanup in includes.
349 Description: mds_mfd_close() ASSERTION(rc == 0)
350 Details : In mds_mfd_close(), we need protect inode's writecount change
351 within its orphan write semaphore to prevent possible races.
355 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
359 Frequency : rare, on shutdown ost
360 Description: don't hit live lock with umount ost.
361 Details : shrink_dcache_parent can be in long loop with destroy dentries,
362 use shrink_dcache_sb instead.
366 Description: don't panic with use echo client
367 Details : echo client pass NULL as client nid pointer and this produce null
372 Description: fix build on ppc32
373 Details : compile code with -m64 flag produce wrong object file for ppc32.
377 Description: add message levels for liblustreapi
381 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
382 Details : Move the 'good_osts' check before the 'total_bavail' check. This
383 will result in an -EAGAIN and in the exit call path we call
384 alloc_rr() which will with increasing aggressiveness attempt to
385 aquire precreated objects on the minimum number of required OSCs.
389 Description: Use old size assignment to avoid deadlock
390 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
391 the scheduling while holding a spinlock. We do not need locking
392 for size in ll_update_inode() because size is only updated from
393 the MDS for directories or files without objects, so there is no
394 other place to do the update, and concurrent access to such inodes
395 are protected by the inode lock.
399 Description: resolve "_IOWR redefined" build error on SLES10
403 Description: dump the memory debugging after all modules are unloaded to
404 suppress false negative in conf_sanity test 39
406 Severity : enhancement
408 Description: build kernel-ib packages for OFED 1.3 in our release cycle
413 Description: fix SLES kernel versioning
414 Details : the kernel version for our SLES 10 kernel did not include a "-"
415 before the "smp" at the end. while this was not a problem in
416 general, it did mean that software trying to use the kernel
417 version to try to detect a vendor specific kernel would fail.
418 this was most evident by the OFED build scripts.
422 Description: Don't update lov_desc members until making sure they are valid
423 Details : When updating lov_desc members via proc fs, need fix their
424 validities before doing the real update.
428 Description: don't put request into delay list while invalidate in flight.
429 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
430 import in flight. this produce timeout for invalidate and sometimes
431 can cause stale data.
435 Frequency : on ppc only
436 Description: not convert ost objects for directory because it's not exist.
437 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
441 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
443 Severity : enhancement
445 Description: Update to support 2.6.22.14 vanilla kernel.
449 Frequency : rare, on recovery
450 Description: read procfs can produce deadlock in some situation
451 Details : Holding lprocfs lock which send rpc can produce block for destroy
452 obd objects and this also block reconnect with -EALREADY. This
453 isn't fix all lprocfs bugs - but make it rare.
455 Severity : enhancement
457 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
460 Frequency : frequent on X2 node
462 Description: mdc_set_open_replay_data LBUG
463 Details : Set replay data for requests that are eligible for replay.
467 Description: lustre_mgs: operation 101 on unconnected MGS
468 Details : When MGC is disconnected from MGS long enough, MGS will evict the
469 MGC, and late on MGC cannot successfully connect to MGS and a lot
470 of the error messages complaining that MGS is not connected.
474 Frequency : on network error
475 Description: panic with double free request if network error
476 Details : mdc_finish_enqueue is finish request if any network error ocuring,
477 but it's true only for synchronus enqueue, for async enqueue
478 (via ptlrpcd) this incorrect and ptlrpcd want finish request
481 Severity : enhancement
483 Description: client-side metadata stat-ahead during readdir(directory readahead)
484 Details : perform client-side metadata stat-ahead when the client detects
485 readdir and sequential stat of dir entries therein
488 Frequency : on start mds
490 Description: Implement get_info(last_id) in obdfilter.
493 Frequency : occasional
495 Description: Correctly check stale fid, not start epoch if ost not support SOM
496 Details : open with flag O_CREATE need set old fid in op_fid3 because
497 op_fid2 was overwritten with new generated fid, but mds can answer
498 with one of these two fids and both is not stale. Setattr
499 incorrectly started an epoch and assume will be called
500 done_writing, but without SOM done_writing ever being called.
503 Frequency : rare, depends on device drivers and load
505 Description: MDS or OSS nodes crash due to stack overflow
506 Details : Code changes in 1.8.0 increased the stack usage of some functions.
507 In some cases, in conjunction with device drivers that use a lot
508 of stack the MDS (or possibly OSS) service threads could overflow
509 the stack. One change which was identified to consume additional
510 stack has been reworked to avoid the extra stack usage.
513 Frequency : occasional
515 Description: Do not fail import if osc_interpret_create gets -EAGAIN
516 Details : If osc_interpret_create got -EAGAIN it immediately exits and
517 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
518 oscc_has_objects and see OSC has no objests and call
519 oscc_internal_create to resend create request.
521 Severity : enhancement
523 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
525 Severity : enhancement
527 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
530 Frequency : very rare
532 Description: Fix replay if there is an un-replied request and open
533 Details : In some cases, older replay request will revert the
534 mcd->mcd_last_xid on MDS which is used to record the client's
537 Severity : enhancement
539 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
541 Severity : enhancement
543 Description: Add rhel5 support to HEAD.
545 Serverity : enhancement
547 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
552 Description: Don't allow skipping OSTs if index has been specified.
553 Details : Don't allow skipping OSTs if index has been specified, make
554 locking in internal create lots better.
558 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
559 Details : access to bitfield in structure is always rounded to long
560 and this produce problem with not atomic change any bit.
564 Description: Lustre make rpms failed.
565 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
566 builds Lustre rpms from tarball.
569 Frequency : rare on shutdown ost
571 Description: If llog cancel was not send before clean_exports phase, this can
572 produce deadlock in llog code.
573 Details : If llog thread has last reference to obd and call class_import_put
574 this produce deadlock because llog_cleanup_commit_master wait when
575 last llog_commit_thread exited, but this never success because was
576 called from llog_commit_thread.
580 Description: allow userland application know is lost one of stripes.
581 Details : fill lvb_blocks with error code on ost and return it to
582 application if error flag found.
586 Description: NULL lov_tgts causing MDS oops
587 Details : more safe checks for NULL lov_tgts for avoid oops.
589 Severity : enhancement
591 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
595 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
596 Details : If an OST is down the MDS will hang indefinitely in
597 obd_statfs_rqset() waiting for the statfs data. While for
598 MDS QOS usage of statfs, it should not stuck in waiting.
600 Severity : enhancement
602 Description: remote_acl support
603 Details : Support ACL-based permission check for remote user.
604 Support setfacl/getfacl for remote user with the utils
605 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
606 the system "{s,g}etfacl" utils.
608 Severity : enhancement
610 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
612 Severity : enhancement
614 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
618 Description: make mgs_setparam() handle fsname containing dash
619 Details : fsname containing a dash does not work with lctl conf_param
621 Severity : enhancement
623 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
625 Severity : enhancement
627 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
631 Frequency : rare, at shutdown
632 Description: access already free / zero obd_namespace.
633 Details : if client_disconnect_export was called without force flag set,
634 and exist connect request in flight, this can produce access to
635 NULL pointer (or already free pointer) when connect_interpret
636 store ocd flags in obd_namespace.
640 Frequency : only at startup
641 Description: not alloc memory with spinlock held.
642 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
643 if any spinlock held.
645 Severity : enhancement
647 Description: make lustre randomly fail allocating memory
648 Details : Make lustre randomly failed allocating memory for testing purpose.
650 Severity : enhancement
652 Description: lost problems with lov objid file
653 Details : Fixes some scability and access to not inited memory problems
654 in work with lov objdid file.
659 Description: lfs find does not continue on file error
660 Details : Continue other files processing when a file/dir is absent.
664 Description: Inconsistent usage of lustre_pack_reply()
665 Details : Standardize the usage of lustre_pack_reply() such that it
666 always generate a CERROR on failure.
669 Frequency : occasional
671 Description: MDS hang or stay in waiting lock
672 Details : If client receive lock with CBPENDING flag ldlm need send lock
673 cancel as separate rpc, to avoid situation when cancel request
674 can't processed due all i/o threads stay in wait lock.
678 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
682 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
686 Description: Update to SLES9 kernel-2.6.5-7.287.3.
690 Description: Update to RHEL4 latest kernel.
692 Severity : enhancement
694 Description: Build SLES10 patchless client fails
695 Details : The configure was broken by run ./configure with
696 --with-linux-obj=.... argument for patchless client. When the
697 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
698 can't search header adequately. Use absolute path such as
699 -I($LINUX)/include instead.
703 Description: interrupt oig_wait produce painc on resend.
704 Details : brw_redo_request can be used for resend requests from ptlrpcd and
705 private set, and this produce situation when rq_ptlrpcd_data not
706 copyed to new allocated request and triggered LBUG on assert
707 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
708 ptlrpcd set if request is changed and can be safety changed to use
713 Description: LASSERT_{REQ,REP}SWAB macros are buggy
714 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
715 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
716 these macros and replace them with their difinitions instead.
720 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
721 Details : Update which_patch & target file for SLES10 latest kernel.
724 Description: Remove client patches from SLES 10 kernel.
725 Details : This causes SLES 10 clients to behave as patchless clients
726 even on a Lustre-patched (server) kernel.
728 Severity : enhancement
730 Description: self-adjustable client's lru lists
731 Details : use adaptive algorithm for managing client cached locks lru
732 lists according to current server load, other client's work
733 pattern, memory activities, etc. Both, server and client
734 side namespaces provide number of proc tunables for controlling
737 Severity : enhancement
739 Description: light-weight GSS support
740 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
741 and reduce performance overhead.
743 Severity : enhancement
745 Description: Linux keyring support
746 Details : Support using service of Linux keyring for Lustre GSS internal
747 context refresh/cache mechanism.
751 Description: Fix errors in lfs documentation
752 Details : Fixes man pages
756 Description: don't use GFP_* in generic Lustre code.
757 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
761 Description: obdclass is limited by single OBD_ALLOC(idarray)
762 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
766 Description: warnings with build patchless client with vanila 2.6.19 and up
767 Details : change old ctl_table style and replace ctl_table/ctl_table_header
768 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
772 Frequency : Only for SLES
773 Description: sanity_quota fail test_1
774 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
775 file, which makes TSTID[2] inunique.
779 Frequency : Only for SLES9
780 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
781 Details : Update target/ChangeLog/which_patch .
785 Frequency : Only for test_10 in sanity-quota.sh
786 Description: a bug in quota test code
787 Details : checking a test flag in a wrong place
791 Frequency : Only for test_13 in sanity-quota.sh
792 Description: a bug in quota test script
793 Details : shouldn't check overfull allocated quota
797 Description: jbd statistics
798 Details : Port older jbd statistics patch for sles10
800 Severity : enhancement
802 Description: Add printing inode info into message about error in writepage.
807 Description: reply_lock_interpret crash due to race with it and lock cancel.
808 Details : Do not replay locks that are being cancelled. Do not reference
809 locks by their address during replay, just by their handle.
813 Frequency : When flocks are used.
814 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
815 Details : Flock locks might destroy just granted lock if it could be merged
816 with another existing flock, this is done in completion handler,
817 so teach ldlm_cli_enquque_fini that this is a valid case for
823 Description: Oops in read and write path when failing to allocate lock.
824 Details : Check if lock allocation failed and return error back.
828 Description: lstripe command fails for valid OST index
829 Details : The stripe offset is compared to lov->desc.ld_tgt_count
830 instead of lov->desc.ld_active_tgt_count.
834 Description: sanity.sh failed test 103
835 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
840 Description: df doesn't work properly if diskfs blocksize != 4K
841 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
845 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
846 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
848 Severity : enhancement
850 Description: Adaptive timeouts
851 Details : RPC timeouts adapt to changing server load and network
852 conditions to reduce resend attempts and improve recovery time.
856 Description: llapi_file_create() does not allow some changes
857 Details : add llapi_file_open() that allows specifying the mode and
858 open flags, and also returns an open file handle.
862 Description: merge and cleanup kernel patches.
863 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
867 Description: Add journal checksum support.(Kernel part)
868 Details : The journal checksum feature adds two new flags i.e
869 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
870 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
871 indicates that the commit block contains the checksum for
872 the blocks described by the descriptor blocks. Now commit
873 record can be sent to disk without waiting for descriptor
874 blocks to be written to disk. This behavior is controlled
875 using JBD2_FEATURE_ASYNC_COMMIT flag.
879 Description: OSS needs mutliple precreate threads
880 Details : Add ability to start more than one create thread per OSS.
884 Description: Sanity.sh test_65k failed: setstripe should have succeeded
885 Details : execute lfs setstripe on client
889 Description: mds_obd_create error creating tmp object
890 Details : When the user sets quota on root, llog will be affected and can't
891 create files and write files.
894 Frequency : Always on ia64 patchless client, and possibly others.
896 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
897 Details : This allows the patchless client to be loaded on architectures
902 Description: RedHat Update kernel for RHEL5
903 Details : Add the kernel config file for RHEL5.
907 Description: RedHat Update kernel for RHEL5
908 Details : Modify the kernel config file more closer RHEL5.
912 Description: Build failure against Centos5 (RHEL5)
913 Details : Define PAGE_SIZE when it isn't present.
917 Description: test_7 on sanity-quota.sh can't be used on separate machines
918 Details : add facet to handle it
922 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
923 Details : with NFS, the anon dentry's parent was set to itself in
924 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
925 corresponding dentry other than use rec->ur_name.
927 Severity : enhancement
929 Description: lfs setstripe enhancement
930 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
934 Description: enable data checksumming by default
935 Details : enable checksum by default, allow --disable-checksum
936 configure option and "-o nochecksum" mount option. Checksums
937 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
941 Description: lustre support for RHEL5
942 Details : Add support for RHEL5.
946 Description: Client eviction due to failover config
947 Details : after a connection loss, the lustre client should attempt to
948 reconnect to the last active server first before trying the
949 other potential connections.
953 Description: when mds and osts use different quota unit(32bit and 64bit),
954 quota will be released repeatly.
955 Details : void sending multiple quota reqs to mds, which will keep the
956 status between the reqs.
960 Description: osts not allocated evenly to files
961 Details : change the condition to increase offset_idx
964 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
966 Description: Data corruption for OSTs that are formatted larger than 2TB
968 Details : When generating the bio request for lustre file writes the
969 sector number would overflow a temporary variable before being
970 used for the IO. The data reads correctly from Lustre (which
971 will overflow in a similar manner) but other file data or
972 filesystem metadata may be corrupted in some cases.
976 Description: Tune the kernel for good SCSI performance.
977 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
978 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
983 Description: rewrite ext2-derived code in obdclass/uuid.c
984 Details : rewrite inherited code (uuid parsing code from ext2 utils)
985 from scratch preserving functionality.
989 Description: "lfs find -obd UUID" prints directories
990 Details : "lfs find -obd UUID" will return all directory names instead
991 of just file names. It is incorrect because the directories
992 do not reside on the OSTs.
996 Description: use i_size_read and i_size_write in 2.6 port
997 Details : replace inode->i_size access with i_size_read/write()
1000 Frequency : when removing large files
1002 Description: scheduling issue during removal of large Lustre files
1003 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1004 It causes scheduling issues when removing large files (17TB in the
1008 Frequency : only with liblustre clients on XT3
1010 Description: evictions taking too long
1011 Details : allow llrd to evict clients directly on OSTs
1014 Frequency : only on ppc
1016 Description: /proc/fs/lustre/devices broken on ppc
1017 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1018 We should make sure the type of 'index' is loff_t and then cast
1019 explicitly as needed below. Do not assign an explicitly cast
1023 Frequency : only for rhel5
1025 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1026 Details : Modify the target file & which_kernel.
1028 Severity : enhancement
1030 Description: omit set fsid for export NFS
1031 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1032 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1036 Frequency : Always, for kernels after 2.6.16
1037 Description: Fix warning idr_remove called for id=.. which is not allocated.
1038 Details : Last kernels save old s_dev before kill super and not allow
1039 to restore from callback - restore it before call kill_anon_super.
1043 Description: buffer overruns could theoretically occur
1044 Details : llapi_semantic_traverse() modifies the "path" argument by
1045 appending values to the end of the origin string, and a
1046 overrun may occur. Adding buffer overrun check in liblustreapi.
1050 Description: Fix error on 'ls .' at the top of the Lustre mount.
1051 Details : Don't revalidate dentry if it is a root dentry.
1055 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1056 Details : Modify vm-tunables-rhel4.patch.
1060 Description: Kernel config for 2.6.18-vanilla.
1061 Details : Modify targets/2.6-vanilla.target.in.
1062 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1063 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1064 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1065 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1069 Description: improve handling recoverable errors
1070 Details : if request processig with error which can be recoverable on server
1071 request should be resend, otherwise page released from cache and
1076 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1077 Details : Modify target file & which_patch.
1078 A flaw was found in the IA32 system call emulation provided
1079 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1080 value could be stored in the %RAX register, which could trigger an
1081 out-of-bounds system call table access. An untrusted local user
1082 could exploit this flaw to run code in the kernel
1083 (ie a root privilege escalation). (CVE-2007-4573).
1087 Description: O_DIRECT bypasses client statistics.
1088 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1089 read_ahead_stats were updated. copied the stats section from
1090 osc_send_oap_rpc() into async_internal().
1094 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1098 Description: change order of libsysio includes
1099 Details : '#include sysio.h' should always come before '#include xtio.h'
1103 Description: handle "serious error: objid * already exists" more gracefully
1104 Details : If LAST_ID value on disk is smaller than the objects existing in
1105 the O/0/d* directories, it indicates disk corruption and causes an
1106 LBUG(). If the object is 0-length, then we should use the existing
1107 object. This will help to avoid a full fsck in most cases.
1109 Severity : enhancement
1111 Description: adapt the lustre_config script to support the upgrade case
1112 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1113 server targets to 1.6 in parallel.
1117 Description: To avoid grant space > avaible space when the disk is almost
1118 full. Without this patch you might see the error "grant XXXX >
1119 available" or some LBUG about grant, when the disk is almost
1121 Details : In filter_check_grant, for non_grant cache write, we should
1122 check the left space by if (*left > ungranted + bytes), instead
1123 of (*left > ungranted), because only we are sure the left space
1124 is enough for another "bytes", then the ungrant space should be
1125 increase. In client, we should update cl_avail_grant only there
1126 is OBD_MD_FLGRANT in the reply.
1130 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1131 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1132 described in CVE-2007-4573. This problem could allow untrusted
1133 local users to gain root access.
1136 Frequency : when using O_DIRECT and quotas
1138 Description: Incorrect file ownership on O_DIRECT output files
1139 Details : block usage reported by 'lfs quota' does not take into account
1140 files that have been written with O_DIRECT.
1145 Description: touch file failed when fs is not full
1146 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1147 otherwise we can get ENOSP while fs is not full.
1151 Description: parallel lock callbacks
1152 Details : Instead of sending blocking and completion callbacks as separated
1153 requests, adding them to a set and sending in parallel.
1156 Frequency : only for Cray XT3
1157 Bugzilla : 12829/13455
1158 Description: Changing primary group doesn't change the group lustre assigns to
1160 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1161 overridden with the primary group provided by the group upcall,
1162 whereas the supplied fsgid can be trusted if it is in the list of
1163 supplementary groups returned by the group upcall.
1165 Severity : enhancement
1167 Description: Allow masking D_WARNING, D_ERROR messages from console
1168 Details : Console messages can now be disabled via lnet.printk.
1172 Description: User code with malformed file open parameter crashes client node
1173 Details : Before packing join_file req, all the related reference should be
1174 checked carefully in case some malformed flags cause fake
1175 join_file req on client.
1179 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1180 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1181 mode is equal to the granted mode. However, it should also check
1182 that LDLM_FL_CP_REQD is not set.
1186 Description: Heavy nfs access might result in deadlocks
1187 Details : After ELC code landed, it is now improper to enqueue any mds
1188 locks under och_sem, because enqueue might want to decide to
1189 cancel open locks for same inode we are holding och_sem for.
1193 Description: Client eviction while running blogbench
1194 Details : A lot of unlink operations with concurrent I/O can lead to a
1195 deadlock causing evictions. To address the problem, the number of
1196 oustanding OST_DESTROY requests is now throttled to
1197 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1198 callbacks are processed in priority.
1202 Description: enable ACLs on MDS by default
1203 Details : ACLs must be enabled on MDS by default.
1206 Frequency : PPC/PPC64 only
1208 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1209 Details : fix duplicated definitions between asm-ppc64/types.h and
1210 lustre_types.h on PPC.
1213 Frequency : PPC/PPC64 only
1215 Description: asm-ppc/segment.h does not exist
1216 Details : fix compile issue on PPC.
1220 Description: better handle error messages in extents code
1223 Frequency : RHEL4 only
1225 Description: mkfs is very slow on IA64/RHEL4
1226 Details : A performance regression has been discovered in the MPT Fusion
1227 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1228 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1229 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1231 Severity : enhancement
1233 Description: SNMP support enhancement
1234 Details : Adding total number of sampled request for an MDS node in snmp
1237 Severity : enhancement
1239 Description: Optimize ldlm waiting list processing for PR extent locks
1240 Details : When processing waiting list for read extent lock and meeting read
1241 lock that is same or wider to it that is not contended, skip
1242 processing rest of the list and immediatelly return current
1243 status of conflictness, since we are guaranteed there are no
1244 conflicting locks in the rest of the list.
1248 Description: Time out and refuse to reconnect
1249 Details : When the failover node is the primary node, it is possible
1250 to have two identical connections in imp_conn_list. We must
1251 compare not conn's pointers but NIDs, otherwise we can defeat
1252 connection throttling.
1256 Description: port llog fixes from b1_6 into HEAD
1257 Details : Port llog reference couting and some llog cleanups from b1_6
1258 (bug 10800) into HEAD, for protect from panic and access to already
1259 free llog structures.
1263 Description: Detect stride IO mode in read-ahead
1264 Details : When a client does stride read, read-ahead should detect that and
1265 read-ahead pages according to the detected stride pattern.
1269 Description: data checksumming impacts single node performance
1270 Details : add support for several checksum algorithm. Currently, only CRC32
1271 and Adler-32 are supported. The checksum type can be changed on
1272 the fly via /proc/fs/lustre/osc/*/checksum_type.
1276 Description: use adler32 for page checksums
1277 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1282 Description: build for x2 fails
1283 Details : fix compile issue on Cray systems.
1287 Description: Properly match for duplicate locks
1288 Details : Due to different lock order from skiplists code, we need to
1289 traverse entire list for now
1292 Frequency : only on PPC/SLES10
1294 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1295 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1296 be defined. Add a hack in mkfs_lustre.c to work around this compile
1301 Description: LASSERT on MDS when client holding flock lock dies
1302 Details : ldlm pool logic depends on number of granted locks equal to
1303 number of released locks which is not true for flock locks, so
1304 just exclude such locks from consideration.
1308 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1309 Details : Use fsfilt_sync() for both the whole filesystem sync and
1310 individual file sync to eliminate dangerous inode locking
1311 with I_LOCK that can lead to a deadlock.
1315 Description: performance in 1.6.3
1316 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1317 sure that 1MB requests are not fragmented by the block layer.
1319 Severity : enhancement
1321 Description: organize the server-side client stats on per-nid basis
1322 Details : Change the structure of stats under obdfilter and mds to
1330 The "uuid"s file would list the uuids of _active_ exports.
1331 And the clear entry is to clear all stats and stale nids.
1333 Severity : enhancement
1335 Description: eliminate client locks in face of contention
1336 Details : file contention detection and lockless i/o implementation
1337 for contended files.
1341 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1344 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1346 Description: MDS may lose file striping (and hence file data) in some cases
1347 Details : If there are additional extended attributes stored on the MDS,
1348 in particular ACLs, SELinux, or user attributes (if user_xattr
1349 is specified for the client mount options) then there is a risk
1350 of attribute loss. Additionally, the Lustre file striping
1351 needs to be larger than default (e.g. striped over all OSTs),
1352 and an additional attribute must be stored initially in the
1353 inode and then increase in size enough to be moved to the
1354 external attribute block (e.g. ACL growing in size) for file
1359 Description: skiplist implementation simplification
1360 Details : skiplists are used to group compatible locks on granted list
1361 that was implemented as tracking first and last lock of each
1362 lock group the patch changes that to using doubly linked lists
1366 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1367 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1368 because LASSERT on both the data supplied by a client, and the
1369 data on disk is dangerous and incorrect.
1371 Severity : enhancement
1373 Description: Slow truncate/writes to huge files at high offsets.
1374 Details : Directly associate cached pages to lock that protect those pages,
1375 this allows us to quickly find what pages to write and remove
1376 once lock callback is received.
1380 Description: more ldlm soft lockups
1381 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1382 starve other threads from the resource lock for a long time in
1383 case of long waiting queue, so change the debug level from
1384 D_OTHER to the less frequently used D_INFO.
1386 Severity : enhancement
1388 Description: add -gid, -group, -uid, -user options to lfs find
1392 Description: Hung threads in invalidate_inode_pages2_range
1393 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1394 once one is completed. As a result, some RPCs are stuck in the
1395 queue and are never sent.
1399 Description: filter threads hungs on waiting journal commit
1400 Details : Cleanup filter group llog code, then only filter group llog will
1401 be only created in the MDS/OST syncing process.
1405 Description: Procfs and llog threads access destoryed import sometimes.
1406 Details : Sync the import destoryed process with procfs and llog threads by
1407 the import refcount and semaphore.
1409 Severity : enhancement
1411 Description: openlock cache of b1_6 port to HEAD
1416 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1417 Details : If insertion of an extent fails, then discard the inode
1418 preallocation and free data blocks else it can lead to duplicate
1423 Description: don't always update ctime in ext3_xattr_set_handle()
1424 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1425 In some cases the ctime should not be updated, for example for
1426 2.0->1.8 compatibility it is necessary to delete an xattr and it
1427 should not update the ctime.
1431 Bugzilla : 15713/16362
1432 Description: Assertion in iopen_connect_dentry in 1.6.3
1433 Details : looking up an inode via iopen with the wrong generation number can
1434 populate the dcache with a disconneced dentry while the inode
1435 number is in the process of being reallocated. This causes an
1436 assertion failure in iopen since the inode's dentry list contains
1437 both a connected and disconnected dentry.
1441 Description: assertion failure in ldlm_handle2lock()
1442 Details : fix a race between class_handle_unhash() and class_handle2object()
1443 introduced in lustre 1.6.5 by bug 13622.
1448 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1449 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1454 Description: oops in sd_iostats_seq_show()
1455 Details : unloading/reloading the scsi low level driver triggers a kernel
1456 bug when trying to access the sd iostat file.
1461 Description: Kernel panics during QLogic driver reload
1462 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1463 patch, causing memory corruption.
1468 Description: journal_dev option does not work in b1_6
1469 Details : pass mount option during pre-mount.
1471 Severity : enhancement
1473 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1474 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1475 extent information of a file. It can be used to map logical blocks
1476 in a file to physical blocks in the block device.
1480 Description: LDLM soft lockups - improvement
1481 Details : It is be possible to send the lock handle along with each read
1482 or write request because the client is already doing a lock match
1483 itself so there isn't any reason the OST should have to re-do that
1487 Frequency : only on Cray X2
1489 Description: X2 build failures
1490 Details : fix build failures on Cray X2.
1494 Description: xid & resent requests
1495 Details : Initialize RPC XID from clock at startup (randomly if clock is
1498 Severity : enhancement
1500 Description: Add lustre_start utility to start or stop multiple Lustre servers
1505 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1506 Details : In case of memory pressure, list_del() can be called twice on
1507 req->rq_history_list, causing a kernel oops.
1511 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1512 Details : fix stack overflow in the distributed lock manager by defering
1513 export eviction after a failed AST to the elt thread instead of
1514 handling it in the dlm interpret routine.
1518 Description: Convert some comments to new format.
1519 Details : Update documenting comments to match doxygen conventions.
1523 Description: Grammar fixes.
1524 Details : A couple of trivial spelling fixes.
1528 Description: OSD_COUNTERS-mandatory
1529 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1530 to simplify the code.
1534 Description: simplify cmm_device freeing logic.
1535 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1539 Description: Add lockdep support to dt_object_operations locking interface.
1540 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1541 parameter indicating lock ordering. Update mdd code to use new
1546 Description: Introduce failloc constants for lockless IO tests.
1547 Details : Add two new failloc constants to test lockless IO. Only one of
1548 them in implemented---another is checked in yet to be landed
1553 Description: Add lockdep support for inode mutex.
1554 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1555 in the situations where Lustre has to lock more than one inode
1560 Description: Add optional invariants checking support.
1561 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1562 configure switch. This macro is to be used for consistency and
1563 sanity checks that are too expensive to be left in `production'
1568 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1569 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1574 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1575 Details : Use spin_lock_nested() in (the only) situation where more than
1576 one ldlm_lock is locked simultaneously. Also, fix possible
1577 dead-lock in ldlm_lock_change_resource() by enforcing particular
1582 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1583 Details : Instead of specifying each ldlm_lock call-back through separate
1584 parameter, wrap them into struct ldlm_callback_suite.
1588 Description: Kill join_lru obd method and its callers.
1589 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1590 in memory---a requirement implemented through ->o_join_lru() obd
1591 method in HEAD. Remove this method and its users.
1595 Description: Add asynchronous ldlm ENQUEUE completion handler.
1596 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1597 case that stock ldlm_completion_ast() cannot handle as it waits
1598 until lock is granted. Introduce new ldlm_completion_ast_async()
1599 for this. Also comment ldlm_completion_ast().
1603 Description: ldlm_error <-> errno conversion.
1604 Details : Add functions to map (rather arbitrary) between LDLM error codes
1605 and standard errno values. CLIO needs this to prevent LDLM specific
1606 constants from escaping ldlm and osc.
1610 Description: Kill unused ldlm_handle2lock_ns() function.
1611 Details : Kill unused ldlm_handle2lock_ns() function.
1615 Description: Add lu_ref support to ldlm_lock
1616 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1617 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1618 and ldlm_resource. LDLM interface has to be changed, because code that
1619 releases a reference on a lock, has to "know" what reference this is.
1620 In the most frequent case
1622 lock = ldlm_handle2lock(handle);
1624 LDLM_LOCK_PUT(lock);
1626 no changes are required. When any other reference (received _not_ from
1627 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1628 instead of LDLM_LOCK_PUT().
1630 Arguably, changes are pervasive, and interface requires some discipline
1631 for proper use. On the other hand, it was very instrumental in finding
1632 a few leaked lock references.
1636 Description: Add ldlm_lock_addref_try().
1637 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1638 attempts to addref a lock that might be being canceled
1643 Description: Add ldlm_weigh_callback().
1644 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1645 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1646 hard-coded `number of pages' logic.
1650 Description: Add lockdep annotations to llog code.
1651 Details : Use appropriately tagged _nested() locking calls in the places
1652 where llog takes more than one ->lgh_lock lock.
1656 Description: Add loi_kms_set().
1657 Details : Wrap kms updates into a helper function.
1661 Description: Constify instances of struct lsm_operations.
1662 Details : Constify instances of struct lsm_operations.
1666 Description: lu_conf support.
1667 Details : On a server, a file system object is uniquely identified
1668 by a fid, which is sufficient to locate and load all object
1669 state (inode). On a client, on the other hand, more data are
1670 necessary instantiate an object. Change lu_object_find() and
1671 friends to take additional `lu_conf' argument describing object.
1672 Typically this includes layout information.
1676 Description: lu_context fixes.
1677 Details : Introduce new lu_context functions that are needed on the client
1678 side, where some system threads (ptlrpcd) are shared by multiple
1679 modules, and so cannot be stopped during module shutdown.
1683 Description: Add start and stop methods to lu_device_type_operations.
1684 Details : Introduce two new methods in lu_device_type_operations, that are
1685 invoked when first instance of a given type is created and last one
1686 is destroyed respectively. This is need by CLIO.
1690 Description: Add lu_ref support to struct lu_device.
1691 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1692 track leaked references.
1696 Description: Introduce lu_kmem_descr.
1697 Details : lu_kmem_descr and its companion interface allow to create
1698 and destroy a number of kmem caches at once.
1702 Description: Fix lu_object finalization race.
1703 Details : Fix a race between lu_object_find() finding an object and its
1704 concurrent finalization. This race is (most likely) not possible
1705 on the server, but might happen on the client.
1709 Description: Introduce lu_ref interface.
1710 Details : lu_ref is a debugging module allowing to track references to
1711 a given object. It is quite cpu expensive, and has to be
1712 explicitly enabled with --enable-lu_ref. See usage description
1717 Description: Factor lu_site procfs stats into a separate function.
1718 Details : Separate lu_site stats printing code into a separate function
1719 that can be reused on a client.
1723 Description: Constify instances of struct {lu,dt,md}_device_operations.
1724 Details : Constify instances of struct {lu,dt,md}_device_operations.
1728 Description: Introduce struct md_site and move meta-data specific parts of
1729 struct lu_site here.
1730 Details : Move md-specific fields out of struct lu_site into special struct
1731 md_site, so that lu_site can be used on a client.
1735 Description: Kill mdd_lov_destroy().
1736 Details : Remove unused mdd code.
1740 Description: Add st_block checking to multistat.c.
1741 Details : Add st_block checking to multistat.c.
1745 Description: Add lu_ref support to struct obd_device.
1746 Details : Add lu_ref tracking to obd_device.
1750 Description: Kill obd_set_fail_loc().
1751 Details : Remove unused code.
1755 Description: Add special type for ptlrpc_request interpret functions.
1756 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1757 there. Actual usage will be in CLIO.
1761 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1762 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1763 doesn't work with lockdep.
1767 Description: Add rwv.c test program.
1768 Details : New testing program exercising readv(2) and writev(2) (Qian).
1772 Description: Add sendfile.c test program.
1773 Details : New testing program exercising sendfile(2) (Jay).
1777 Description: Ratelimit a message that can be very frequent.
1778 Details : Ratelimit a memory allocation failure message that can
1783 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1784 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1788 Description: Liblustre build fixes.
1789 Details : Liblustre build fixes.
1793 Description: libcfs: add cfs_{need,cond}_resched() interface.
1794 Details : libcfs: add cfs_{need,cond}_resched() definition and
1795 implementations for Linux, NT, and liblustre.
1797 Severity : enhancement
1799 Description: More exported tunables for mballoc
1800 Details : Add support for tunable preallocation window and new tunables for
1801 large/small requests
1805 Description: Detect corruption of block bitmap and checking for preallocations
1806 Details : Checks validity of on-disk block bitmap. Also it does better
1807 checking of number of applied preallocations. When corruption is
1808 found, it turns filesystem readonly to prevent further corruptions.
1812 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1813 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1814 window to 0. If it does hit cache miss, and read-pattern is still
1815 stride-io mode, does not reset the stride window, but also does not
1816 increase the stride window length in this case.
1820 Frequency : only for big-endian servers
1821 Description: Check if system is big-endian while mounting fs with extents feature
1822 Details : Mounting a filesystem with extents feature will fail on big-endian
1823 systems since ext3-based ldiskfs is not supported on big-endian
1824 systems. This can be over-riden with "bigendian_extents" mount option.
1826 --------------------------------------------------------------------------------
1828 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1830 * Support for kernels:
1831 2.6.5-7.283 (SLES 9),
1832 2.6.9-55.EL (RHEL 4),
1833 2.6.16.46-0.14 (SLES 10),
1834 2.6.18.8 vanilla (kernel.org)
1835 * Client support for unpatched kernels:
1836 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1837 2.6.16 - 2.6.22 vanilla (kernel.org)
1838 * Due to recently discovered recovery problems, we do not recommend
1839 using patchless RHEL 4 clients with this or any earlier release.
1840 * Recommended e2fsprogs version: 1.39.cfs8
1841 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1842 * Starting with this release, the ldiskfs backing filesystem required
1843 by Lustre is now in its own package, lustre-ldiskfs. This package
1844 should be installed. It is versioned separately from Lustre and
1845 may be released separately in future.
1849 Description: block reactivating mgc import until all deactivates complete
1850 Details : Fix race when failing back MDT/MGS to itself (testing)
1853 Frequency : at statup only
1855 Description: mds_lov_synchronize race leads to various problems
1856 Details : simultaneous MDT->OST connections at startup can cause the
1857 sync to abort, leaving the OSC in a bad state.
1859 Severity : enhancement
1861 Description: add optional extra BUILD_VERSION info
1862 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1863 to override the lustre version.
1866 Frequency : 2.6.18 servers only
1868 Description: ll_kern_mount() doesn't release the module reference
1869 Details : The ldiskfs module reference count never drops down to 0
1870 because ll_kern_mount() doesn't release the module reference.
1875 Description: server LBUG when using old ost_num_threads parameter
1876 Details : Accept the old ost_num_threads parameter but warn that it
1877 is deprecated, and fix an off-by-one error that caused an LBUG.
1882 Description: Transient SCSI error results in persistent IO issue
1883 Details : iobuf->dr_error is not reinitialized to 0 between two
1887 Frequency : sometimes when underlying device returns I/O errors
1889 Description: OSTs not going read-only during write failures
1890 Details : OSTs are not remounted read-only when the journal commit threads
1891 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1892 instead of the ext3 wrappers.
1896 Description: poor connect scaling with increasing client count
1897 Details : Don't run filter_grant_sanity_check for more than 100 exports
1898 to improve scaling for large numbers of clients.
1901 Frequency : SLES10 only
1903 Description: sanity-quota.sh quotacheck failed: rc = -22
1904 Details : Quotas cannot be enabled on SLES10.
1907 Frequency : liblustre clients only
1909 Description: getdirentries does not give error when run on compute nodes
1910 Details : getdirentries does not fail when the size specified as an argument
1911 is too small to contain at least one entry
1913 Severity : enhancement
1915 Description: Add LNET router traceability for debug purposes
1916 Details : If a checksum failure occurs with a router as part of the
1917 IO path, the NID of the last router that forwarded the bulk data
1918 is printed so it can be identified.
1923 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1924 Details : Due to a race condition, liblustre clients were occasionally
1925 evicted incorrectly.
1927 Severity : enhancement
1929 Description: lfs setstripe use optional parameters instead of postional
1932 Severity : enhancement
1934 Description: Nanosecond timestamp support for ldiskfs
1935 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1936 resolution timestamps. There is not yet support for this at
1937 the Lustre filesystem level.
1940 Frequency : during server recovery
1942 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1943 Details : request with rq_no_resend flag not awake l_wait_event if they get
1947 Frequency : nfs export on patchless client
1949 Description: connectathon hang when test nfs export over patchless client
1950 Details : Disconnected dentry cannot be found with lookup, so we do not need
1951 to unhash it or make it invalid
1954 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1955 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1956 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1961 Description: open req refcounting wrong on reconnect
1962 Details : If reconnect happened between getting open reply from server and
1963 call to mdc_set_replay_data in ll_file_open, we will schedule
1964 replay for unreferenced request that we are about to free.
1965 Subsequent close will crash in variety of ways.
1966 Check that request is still eligible for replay in
1967 mdc_set_replay_data().
1972 Description: disable writes to filesystem when reading health_check file
1973 Details : the default for reading the health_check proc file has changed
1974 to NOT do a journal transaction and write to disk, because this
1975 can cause reads of the /proc file to hang and block HA state
1976 checking on a healthy but otherwise heavily loaded system. It
1977 is possible to return to the previous behaviour during configure
1978 with --enable-health-write.
1980 Severity : enhancement
1982 Description: 64-bit inode version
1983 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1984 to the inode. This will be required for version-based recovery.
1989 Description: MDS fails to start if a duplicate client export is detected
1990 Details : in some rare cases it was possible for a client to connect to
1991 an MDS multiple times. Upon recovery the MDS would detect this
1992 and fail during startup. Handle this more gracefully.
1994 Severity : enhancement
1996 Description: Add -o localflock option to simulate old noflock
1998 Details : This will achieve local-only flock/fcntl locks
2004 Description: log_commit_thread vs filter_destroy race leads to crash
2005 Details : Take import reference before releasing llog record semaphore
2010 Description: Wrong request locking in request set processing
2011 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2012 imp_delayed_list, in this place should be used imp_lock.
2015 Frequency : when reconnection
2017 Description: Grant Leak when osc reconnect to OST
2018 Details : When osc reconnect ost, OST(filter) should check whether it
2019 should grant more space to client by comparing fed_grant and
2020 cl_avail_grant, and return the granted space to client instead
2021 of "new granted" space, because client will call osc_init_grant
2022 to update the client grant space info.
2025 Frequency : when client reconnect to OST
2027 Description: Grant Leak when osc do resend and replay bulk write
2028 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2029 bulk write request, because the grant info while be sync between
2030 OSC and OST when reconnect, and we should ignore the grant info
2031 these of resend/replay write req.
2036 Description: Grant space more than avaiable left space sometimes.
2037 Details : When then OST is about to be full, if two bulk writing from
2038 different clients came to OST. Accord the avaliable space of the
2039 OST, the first req should be permitted, and the second one
2040 should be denied by ENOSPC. But if the seconde arrived before
2041 the first one is commited. The OST might wrongly permit second
2042 writing, which will cause grant space > avaiable space.
2045 Frequency : when client is evicted
2047 Description: Grant might be wrongly erased when osc is evicted by OST
2048 Details : when the import is evicted by server, it will fork another
2049 thread ptlrpc_invalidate_import_thread to invalidate the
2050 import, where the grant will be set to 0. While the original
2051 thread will update the grant it got when connecting. So if
2052 the former happened latter, the grant will be wrongly errased
2053 because of this race.
2058 Description: Checking Stale with correct fid
2059 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2060 to check whether it is stale, because sometimes, we want the
2061 enqueue happened anyway, and op_data.fid2 will not be initialized.
2063 Severity : enhancement
2065 Description: update patchless client
2066 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2069 Frequency : only with 2.4 kernel
2071 Description: random memory corruption
2072 Details : size of struct ll_inode_info is to big for union inode.u and this
2073 can be cause of random memory corruption.
2078 Description: Memory leak in recovery
2079 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2080 It should also check obd_fail before fsfilt_start, otherwise if
2081 fsfilt_start return -EROFS,(failover mds during mds recovery).
2082 then the req will return with repmsg->transno = 0 and rc = EROFS.
2083 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2084 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2085 be freed no matter whether fsfilt_commit success or not.
2088 Frequency : only with huge count clients
2090 Description: Prevents from taking the superblock lock in llap_from_page for
2092 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2093 for llap_from_page call in ll_removepage prevents from taking the
2094 superblock lock for a soon died page.
2099 Description: Not check open intent error before release open handle
2100 Details : in some rare cases, the open intent error is not checked before
2101 release open handle, which may cause
2102 ASSERTION(open_req->rq_transno != 0), because it tries to release
2103 the failed open handle.
2108 Description: Set cat log bitmap only after create log success.
2109 Details : in some rare cases, the cat log bitmap is set too early. and it
2110 should be set only after create log success.
2114 Description: Accessing a block bevice can re-enable I/O when Lustre is
2115 tearing down a device.
2116 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2121 Description: service threads may hog cpus when there are a lot of requests
2122 Details : Insert cond_resched to give other threads a chance to use some CPU
2127 Description: the cat log was not initialized in recovery
2128 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2129 unlink log on mds will not be initialized until mds post
2130 recovery. And also in mds post recovery, the unlink log will
2131 initialization will be done asynchronausly, so there will be race
2132 between add unlink log and unlink log initialization.
2136 Description: brw_stats were being printed incorrectly
2137 Details : brw_stats were being printed as log2 but all of them were not
2138 recorded as log2. Also remove some code duplication arising from
2139 filter_tally_{read,write}.
2143 Frequency : rare, only in recovery.
2144 Description: ASSERTION(req->rq_type != LI_POISON) failed
2145 Details : imp_lock should be held while iterating over imp_sending_list for
2146 prevent destroy request after get timeout in ptlrpc_queue_wait.
2150 Description: replay-single.sh test 52 fails
2151 Details : A lock's skiplist need to be cleanup when it being unlinked
2152 from its resource list.
2156 Description: Short directio read returns full requested size rather than
2158 Details : Direct I/O operations should return actual amount of bytes
2159 transferred rather than requested size.
2161 Severity : enhancement
2163 Description: metadata RPC reduction (e.g. for rm performance)
2164 Details : decrease the amount of synchronous RPC between clients and servers
2165 by canceling conflicing lock before the operation on the client
2166 and packing thier handles into the main operation RPC to server.
2168 Severity : enhancement
2170 Description: Async OSC create to avoid the blocking unnecessarily.
2171 Details : If a OST has no remain object, system will block on the creating
2172 when need to create a new object on this OST. Now, ways use
2173 pre-created objects when available, instead of blocking on an
2174 empty osc while others are not empty. If we must block, we block
2175 for the shortest possible period of time.
2179 Description: improve handling recoverable errors
2180 Details : if request processig with error which can be recoverable on server
2181 request should be resend, otherwise page released from cache and
2184 Severity : enhancement
2186 Description: refine locking for avoid write wrong info into lov_objid file
2187 Details : fix possible races with add new target and write/update data in
2190 --------------------------------------------------------------------------------
2192 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2197 Frequency : on some architectures
2199 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2200 Details : Uninitialized flags sometimes cause configuration commands to
2204 Frequency : patchless clients only
2206 Description: missing __iget() symbol export
2207 Details : The __iget() symbol export is missing. To avoid the need for
2208 this on patchless clients the deathrow inode reaper is turned
2209 off, and we depend on the VM to clean up old inodes. This
2210 dependency was during via the fix for bug 12181.
2212 --------------------------------------------------------------------------------
2214 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2216 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2217 INTEROPERATE with older versions automatically. In many cases a
2218 special upgrade step is needed. Please read the
2219 user documentation before upgrading any part of a live system.
2220 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2221 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2222 * WARNING: Lustre configuration and startup changes are required with
2223 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2227 Severity : enhancement
2229 Description: MountConf
2230 Details : Lustre configuration is now managed via mkfs and mount
2231 commands instead of lmc and lconf. New obd types (MGS, MGC)
2232 are added for dynamic configuration management. See
2233 https://mail.clusterfs.com/wikis/lustre/MountConf for
2236 Severity : enhancement
2238 Description: dynamic OST addition
2239 Details : OSTs can now be added to a live filesystem
2241 Severity : enhancement
2243 Description: startup order invariance
2244 Details : MDTs and OSTs can be started in any order. Clients only
2245 require the MDT to complete startup.
2247 Severity : enhancement
2249 Description: parallel, asynchronous orphan cleanup
2250 Details : orphan cleanup is now performed in separate threads for each
2251 OST, allowing parallel non-blocking operation.
2253 Severity : enhancement
2255 Description: optimized stripe assignment
2256 Details : stripe assignments are now made based on ost space available,
2257 ost previous usage, and OSS previous usage, in order to try
2258 to optimize storage space and networking resources.
2260 Severity : enhancement
2262 Description: Permanently set tunables
2263 Details : All writable /proc/fs/lustre tunables can now be permanently
2264 set on a per-server basis, at mkfs time or on a live system.
2266 Severity : enhancement
2268 Description: Lustre message v2
2269 Details : Add lustre message format v2.
2271 Severity : enhancement
2273 Description: client OST exclusion list
2274 Details : Clients can be started with a list of OSTs that should be
2275 declared "inactive" for known non-responsive OSTs.
2279 Description: SPEC SFS validation failure on NFS v2 over lustre.
2280 Details : Changes the blocksize for regular files to be 2x RPC size,
2281 and not depend on stripe size.
2283 Severity : enhancement
2285 Description: Multiple MD RPCs in flight.
2286 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2287 To avoid overly-overloading MDS, introduce a limit on number of
2288 MDS RPCs in flight for a single client and add /proc controls
2289 to adjust this limit.
2291 Severity : enhancement
2293 Description: client read/write statistics
2294 Details : Add client read/write call usage stats for performance
2295 analysis of user processes.
2296 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2297 file access. extents_stats shows chunk size distribution.
2298 extents_stats_per_process show chunk size distribution per
2301 Severity : enhancement
2303 Description: mds statistics
2304 Details : Add detailed mds operations statistics in
2305 /proc/fs/lustre/mds/*/stats.
2309 Description: Failure of copying files with lustre special EAs.
2310 Details : Client side always return success for setxattr call for lustre
2311 special xattr (currently only "trusted.lov").
2316 Description: Refcount LNET uuids
2317 Details : The global LNET uuid list grew linearly with every startup;
2318 refcount repeated list entries instead of always adding to
2321 Severity : enhancement
2323 Description: Dynamic service threads
2324 Details : Within a small range, start extra service threads
2325 automatically when the request queue builds up.
2328 Frequency : mixed-endian client/server environments
2330 Description: mixed-endian crashes
2331 Details : The new msg_v2 system had some failures in mixed-endian
2334 Severity : enhancement
2336 Description: Easy OST removal
2337 Details : OSTs can be permanently deactivated with e.g. 'lctl
2338 conf_param lustre-OST0001.osc.active=0'
2340 Severity : enhancement
2342 Description: MGS proc entries
2343 Details : Added basic proc entries for the MGS showing what filesystems
2346 Severity : enhancement
2348 Description: provide MGS failover
2349 Details : Added config lock reacquisition after MGS server failover.
2351 Severity : enhancement
2353 Description: add Linux 2.4 support
2354 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2358 Description: a large application tries to do I/O to the same resource and dies
2359 in the middle of it.
2360 Details : Check the req->rq_arrival time after the call to
2361 ost_brw_lock_get(), but before we do anything about
2362 processing it & sending the BULK transfer request. This
2363 should help move old stale pending locks off the queue as
2364 quickly as obd_timeout.
2367 Frequency : when an incorrect nid is specified during startup
2369 Description: ptlrpc connect to non-existant node causes kernel crash
2370 Details : LNET can't be re-entered from an event callback, which
2371 happened when we expire a message after the export has been
2372 cleaned up. Instead, hand the zombie cleanup off to another
2375 Severity : enhancement
2377 Description: plain/inodebits lock performance improvement
2378 Details : Grouping plain/inodebits in granted list by their request modes
2379 and bits policy, thus improving the performance of search through
2383 Frequency : only if OST filesystem is corrupted
2385 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2386 Details : for a short time RPCs with bulk IO are in the replay list,
2387 but replay of bulk IOs is unimplemented. If the OST filesystem
2388 is corrupted due to disk cache incoherency and then replay is
2389 started it is possible to trip an assertion. Avoid putting
2390 committed RPCs into the replay list at all to avoid this issue.
2393 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2395 Description: System hang on startup
2396 Details : This bug allowed the liblustre (e.g. catamount) client to
2397 return to the app before handling all startup RPCs. This
2398 could leave the node unresponsive to lustre network traffic
2399 and manifested as a server ptllnd timeout.
2401 Severity : enhancement
2403 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2404 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2405 boolean to control whether to print peer debug info when a
2406 client's RPC times out.
2409 Frequency : only for kernels with patches from Lustre below 1.4.3
2411 Description: Remove old rdonly API
2412 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2415 Frequency : only for devices with external journals
2417 Description: Set external device read-only also
2418 Details : During a commanded failover stop, we set the disk device
2419 read-only while the server shuts down. We now also set any
2420 external journal device read-only at the same time.
2423 Frequency : when upgrading from 1.4 while trying to change parameters
2425 Description: The wrong (new) MDC name was used when setting parameters for
2426 upgraded MDT's. Also allows changing of OSC (and MDC)
2427 parameters if --writeconf is specified at tunefs upgrade time.
2430 Frequency : when setting specific ost indicies
2432 Description: QOS code breaks on skipped indicies
2433 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2434 created with --index need not be sequential.
2436 Severity : enhancement
2438 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2439 Details : The uninit_groups feature works in conjunction with the kernel
2440 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2441 up the pass1 processing of e2fsck. This is a read-only feature
2442 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2443 mount filesystems that have had this feature enabled.
2445 Severity : enhancement
2447 Description: Improve multi-block allocation algorithm to avoid fragmentation
2448 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2449 allocation locality and avoid filesystem fragmentation.
2451 ------------------------------------------------------------------------------
2453 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2455 * Support for kernels:
2456 2.4.21-47.0.1.EL (RHEL 3)
2457 2.6.5-7.283 (SLES 9)
2458 2.6.9-42.0.10.EL (RHEL 4)
2459 2.6.12.6 vanilla (kernel.org)
2460 2.6.16.27-0.9 (SLES 10)
2461 * Recommended e2fsprogs version: 1.39.cfs5
2463 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2467 Frequency : occasional, depends on client load and configuration
2468 Bugzilla : 12181, 12203
2469 Description: data loss for recently-modified files
2471 Details : In some cases it is possible that recently written or created
2472 files may not be written to disk in a timely manner (this should
2473 normally be within 30s unless client IO load is very high).
2474 The problem appears as zero-length files or files that are a
2475 multiple of 1MB in size after a client crash or client eviction
2476 that are missing data at the end of the file.
2478 This problem is more likely to be hit on clients where files are
2479 repeatedly created and unlinked in the same directory, clients
2480 have a large amount of RAM, have many CPUs, the filesystem has
2481 many OSTs, the clients are rebooted frequently, and/or the files
2482 are not accessed by other nodes after being written.
2484 The presence of the problem can be detected by looking at
2485 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2486 smaller than the second (nr_unused) then dirty files will not
2487 be flushed automatically to disk. "sync; sleep 10" should be
2488 run several times on the node before unmounting it to update
2489 Lustre (this is also safe to run on nodes without this problem).
2491 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2492 that can cause this same problem, so customers using that kernel
2493 also need to update the kernel in addition to Lustre. In order
2494 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2496 It is normal that files written just before a client crash (less
2497 than 30s) may not yet have been flushed to disk, even for local
2501 Frequency : frequent on thin XT3 nodes
2503 Description: UUID collision on thin XT3 Linux nodes
2504 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2505 randomly, since we relied on an insufficiently-seeded PRNG.
2510 Description: OSS hangs after "All ost request buffers busy"
2511 Details : A deadlock between quota and journal operations caused OSS
2512 hangs after printing "All ost request buffers busy."
2515 Frequency : always on liblustre builds
2517 Description: Cleanup compiler warnings on liblustre
2520 Frequency : always on liblustre builds on XT3
2522 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2526 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2527 > 32000 subdirectories
2528 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2529 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2530 subdirectory count crosses 32000. This will aid e2fsck to
2531 correctly handle more than 32000 subdirectories.
2534 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2536 Description: System hang on startup
2537 Details : This bug allowed the liblustre (e.g. catamount) client to
2538 return to the app before handling all startup RPCs. This
2539 could leave the node unresponsive to lustre network traffic
2540 and manifested as a server ptllnd timeout.
2542 Severity : enhancement
2544 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2545 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2546 boolean to control whether to print peer debug info when a
2547 client's RPC times out.
2552 Description: make O_SYNC working on 2.6 kernels
2553 Details : 2.6 kernels use different method for mark pages for write,
2554 so need add a code to lustre for O_SYNC work.
2559 Description: Failure to close file and release space on NFS
2560 Details : Put inode details into lock acquired in ll_intent_file_open.
2561 Use mdc_intent_lock in ll_intent_open to properly
2562 detect all kind of errors unhandled by mdc_enqueue.
2567 Description: proc file read during shutdown sometimes raced obd removal,
2569 Details : Add lock to prevent obd access after proc file removal.
2572 Frequency : Only for files larger than 4GB on 32-bit clients.
2574 Description: improperly doing page alignment of locks
2575 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2576 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2581 Description: RPCs being resent when they shouldn't be
2582 Details : Some RPCs that should not be resent are being resent. This
2583 can cause inconsistencies in the RPC state machine. Do not
2584 resend such requests.
2587 Frequency : rare, only with NFS export
2589 Description: Crash on NFS re-export node
2590 Details : under very unusual load conditions an assertion is hit in
2591 ll_intent_file_open()
2594 Frequency : only if OST filesystem is corrupted
2596 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2597 Details : for a short time RPCs with bulk IO are in the replay list,
2598 but replay of bulk IOs is unimplemented. If the OST filesystem
2599 is corrupted due to disk cache incoherency and then replay is
2600 started it is possible to trip an assertion. Avoid putting
2601 committed RPCs into the replay list at all to avoid this issue.
2606 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2607 Details : Large single O_DIRECT read and write calls can fail to allocate
2608 a sufficiently large buffer to process the request. In case of
2609 allocation failure the allocation is retried with a smaller
2610 buffer and broken into smaller requests.
2612 Severity : enhancement
2614 Description: Add -o localflock option to simulate old noflock behaviour.
2615 Details : This will achieve local-only flock/fcntl locks coherentness.
2620 Description: versioning check is incomplete
2621 Details : Checking the version difference of client vs. server, report
2622 error if the gap is too big.
2627 Description: add support PG_writeback bit
2628 Details : add support for PG_writeback bit for Lustre, for more carefull
2629 work with page cache in 2.6 kernel. This also fix some deadlocks
2630 and remove hack for work O_SYNC with 2.6 kernel.
2632 Severity : enhancement
2634 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2635 Details : The uninit_groups feature works in conjunction with the kernel
2636 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2637 up the pass1 processing of e2fsck. This is a read-only feature
2638 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2639 mount filesystems that have had this feature enabled.
2641 Severity : enhancement
2643 Description: Improve multi-block allocation algorithm to avoid fragmentation
2644 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2645 allocation locality and avoid filesystem fragmentation.
2647 ------------------------------------------------------------------------------
2649 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2651 * Support for kernels:
2652 2.6.9-42.0.3.EL (RHEL 4)
2653 2.6.5-7.276 (SLES 9)
2654 2.4.21-47.0.1.EL (RHEL 3)
2655 2.6.12.6 vanilla (kernel.org)
2656 2.6.16.21-0.8 (SLES10)
2657 * Recommended e2fsprogs version: 1.39.cfs2-0
2659 * The backwards-compatible /proc/sys/portals symlink has been removed
2660 in this release. Before upgrading, please ensure that you change
2661 any configuration scripts or /etc/sysctl.conf files that access
2662 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2663 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2664 in advance of the upgrade on any system running Lustre 1.4.6 or
2665 newer, since /proc/sys/lnet was added in that version.
2666 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2670 Frequency : only when quota is used
2672 Description: avoid scanning export list for quota master
2673 Details : Change the algorithms to avoid scanning export list in order
2674 to improve the efficiency.
2677 Frequency : MDS failover only, very rarely
2679 Description: "went back in time" messages on mds failover
2680 Details : The greatest transno may be lost when the current operation
2681 finishes with an error (transno==0) and the client's last_rcvd
2682 record is over-written. Save the greatest transno in the
2683 mds_last_transno for this case.
2686 Frequency : always for specific kernels and striping counts
2688 Description: client may get "Matching packet too big" without ACL support
2689 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2690 when trying to access files in certain configurations. The
2691 clients should in fact be denied when mounting because they do
2692 not understand ACLs.
2695 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2697 Description: many clients connecting with IO in progress causes connect timeouts
2698 Details : Avoid synchronous journal commits to avoid delays caused by many
2699 clients connecting/disconnecting when bulk IO is in progress.
2700 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2701 OST_IO_PORTAL to avoid delays behind potentially many pending
2705 Frequency : occasionally with multiple writers to a single file
2707 Description: shared writes to file may result in wrong size reported by stat()
2708 Details : Allow growing of kms when extent lock is cancelled
2711 Frequency : always with random mmap IO to multi-striped file
2713 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2714 Details : Only if the hole is at the end of OST object so that kms is too
2715 small. Fix is to increase kms accordingly in ll_nopage.
2718 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2720 Description: writes to a missing object would leak memory on the OST
2721 Details : If there is an inconsistency between the MDS and OST filesystems,
2722 such that the MDS references an object that doesn't exist, writes
2723 to that object will leak memory due to incorrect cleanup in the
2724 error handling path, eventually running out of memory on the OST.
2729 Description: Creating too long symlink causes lustre errors
2730 Details : Check symlink and name lengths before sending requests to MDS.
2733 Frequency : only if flock is enabled (not on by default)
2735 Description: posix locks not released on fd closure on 2.6.9+
2736 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2737 kernels, this caused such locks not to be released on fd close and
2738 then assertions on fs unmount about still used locks.
2741 Frequency : MDS failover only, very rarely
2743 Description: clients may get ASSERTION(granted_lock != NULL)
2744 Details : When request was taking a long time, and a client was resending
2745 a getattr by name lock request. The were multiple lock requests
2746 with the same client lock handle and
2747 mds_getattr_name->fixup_handle_for_resent_request found one of the
2748 lock handles but later failed with ASSERTION(granted_lock != NULL).
2753 Description: handle->h_buffer_credits > 0, assertion failure
2754 Details : h_buffer_credits is zero after truncate, causing assertion
2755 failure. This patch extends the transaction or creates a new
2759 Frequency : NFS re-export or patchless client
2760 Bugzilla : 11179, 10796
2761 Description: Crash on NFS re-export node (__d_move)
2762 Details : We do not want to hash the dentry if we don't have a lock.
2763 But if this dentry is later used in d_move, we'd hit uninitialised
2764 list head d_hash, so we just do this to init d_hash field but
2765 leave dentry unhashed.
2768 Frequency : NFS re-export or patchless client
2770 Description: NFS exports has problem with symbolic link
2771 Details : lustre client didn't properly install dentry when re-exported
2772 to NFS or running patchless client.
2775 Frequency : NFS re-export or patchless client
2777 Description: Various nfs/patchless fixes.
2778 Details : fixes reuse disconected alias for lookup process - this fixes
2779 warning "find_exported_dentry: npd != pd",
2780 fix permission error with open files at nfs.
2781 fix apply umask when do revalidate.
2784 Frequency : occasional
2786 Description: Crash on NFS re-export node
2787 Details : calling clear_page() on the wrong pointer triggered oops in
2788 generic_mapping_read().
2791 Frequency : rarely, using O_DIRECT IO
2793 Description: unaligned directio crashes client with LASSERT
2794 Details : check for unaligned buffers before trying any requests.
2797 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2799 Description: stale data returned from RAID cache
2800 Details : If only a small amount of IO is done to the RAID device before
2801 reading it again it is possible to get stale data from the RAID
2802 cache instead of reading it from disk.
2805 Frequency : always for sles10 kernel
2807 Description: sles10 support
2808 Details : ll_follow_link: compile fixes and using of nd_set_link
2809 under newer kernels.
2812 Frequency : depends on arch, kernel and compiler version, always on sles10
2815 Description: recursive or deep enough symlinks cause stack overflow
2816 Details : getting rid of large stack-allocated variable in
2820 Frequency : depends on hardware
2822 Description: lustre write performance loss in the SLES10 kernel
2823 Details : the performance loss is caused by using of write barriers in the
2824 ext3 code. The SLES10 kernel turns barrier support on by
2825 default. The fix is to undo that change for ldiskfs.
2827 ------------------------------------------------------------------------------
2829 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2831 * Support for kernels:
2832 2.6.9-42.0.3EL (RHEL 4)
2833 2.6.5-7.276 (SLES 9)
2834 2.4.21-47.0.1.EL (RHEL 3)
2835 2.6.12.6 vanilla (kernel.org)
2839 Frequency : quota enabled and large files being deleted
2841 Description: releasing more than 4GB of quota at once hangs OST
2842 Details : If a user deletes more than 4GB of files on a single OST it
2843 will cause the OST to spin in an infinite loop. Release
2844 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2849 Description: statfs data retrieved from /proc may be stale or zero
2850 Details : When reading per-device statfs data from /proc, in the
2851 {kbytes,files}_{total,free,avail} files, it may appear
2852 as zero or be out of date.
2855 Frequency : systems with MD RAID1 external journal devices
2857 Description: lconf's call to blkid is confused by RAID1 journal devices
2858 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2859 of returning all block devices that match the journal UUID.
2862 Frequency : always, for aggregate stripe size over 4GB
2864 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2865 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2866 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2869 Frequency : always if "lfs find" used on a local file/directory
2871 Description: "lfs find" segfaults if used on a local file/directory
2872 Details : The case where a directory component was not specified wasn't
2873 handled correctly. Handle this properly.
2876 Frequency : always on ppc64
2878 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2879 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2880 perform endianness conversion.
2883 Frequency : rarely (truncate to non-zero file size after write under load)
2884 Bugzilla : 10730, 10687
2885 Description: Files padded with zeros to next 4K multiple
2886 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2887 it is possible that files that are truncated to a non-zero size
2888 immediately after being written are filled with zero bytes beyond
2889 the truncated size. No file data is lost.
2891 Severity : enhancement
2893 Description: Allow recovery/failover for liblustre clients.
2894 Details : liblustre clients were unaware of failover configurations until
2897 Severity : enhancement
2899 Description: user file locks should fail when not mounting with flock option
2900 Details : Set up an error-returning stub in ll_file_operations.lock field
2901 to prevent incorrect behaviour when client is mounted without
2902 flock option. Also, set up properly f_op->flock field for
2906 Frequency : always on ia64
2908 Description: "lfs df" loops on printing out MDS statfs information
2909 Details : The obd_ioctl_data was not initialized and in some systems
2910 this caused a failure during the ioctl that did not return
2911 an error. Initialize the struct and return an error on failure.
2914 Frequency : SLES 9 only
2916 Description: Error of copying files with lustre special EAs as root
2917 Details : Client side always return success for setxattr call for lustre
2918 special xattr (currently only "trusted.lov").
2921 Frequency : rarely on clusters with both ia64+i386 clients
2923 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2924 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2925 clients are doing concurrent writes to the same file it is
2926 possible that the ia64 clients may LASSERT because the OST
2927 extent locks are not PAGE_SIZE aligned. Ensure that grown
2928 locks are always aligned on the request boundary.
2931 Frequency : specific use, occasional
2933 Description: Overwriting in use executable truncates on-disk binary image
2934 Details : If one node attempts to overwrite an executable in use by
2935 another node, we now correctly return ETXTBSY instead of
2936 truncating the file.
2938 Severity : enhancement
2940 Description: Async OSC create to avoid the blocking unnecessarily.
2941 Details : If a OST has no remain object, system will block on the creating
2942 when need to create a new object on this OST. Now, ways use
2943 pre-created objects when available, instead of blocking on an
2944 empty osc while others are not empty. If we must block, we block
2945 for the shortest possible period of time.
2950 Description: chmod on Lustre root is propagated to other clients
2951 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2952 invalid by the follow_mount time.
2955 Frequency : liblustre clients only
2957 Description: Race in 'instant cancel' lock handling could lead to such locks
2958 never to be granted in case of SMP MDS
2959 Details : Do not destroy not yet granted but cbpending locks in
2963 Frequency : replay/resend of open
2965 Description: non null lock assetion failure in mds_intent_policy
2966 Details : Trying to replay/resend lockless open requests resulted in
2967 mds_open() returning 0 with no lock. Now it sets a flag if
2968 a lock is going to be returned.
2970 Severity : enhancement
2972 Description: Checksum enhancements
2973 Details : New checksum enhancements allow for resending RPCs that failed
2976 Severity : enhancement
2978 Description: Tunables on number of dirty pages in cacche
2979 Details : Allow to set limit on number of dirty pages cached.
2984 Description: client crash on unmount - lock still has references
2985 Details : In some error handling cases it was possible to leak a lock
2986 reference on a client while accessing a file. This was not
2987 harmful to the client during operation, but would cause the
2988 client to crash when the filesystem is unmounted.
2991 Frequency : specific case, rare
2993 Description: ETXTBSY on mds though file not in use by client
2994 Details : ETXTBSY is no longer incorrectly returned when attempting to
2995 chmod or chown a directory that the user previously tried to
2996 execute or a currently-executing binary.
2999 Frequency : extremely rare except on liblustre-based clients
3001 Description: Lustre space not freed when files are deleted
3002 Details : Clean up open-unlinked files after client eviction. Previously
3003 the unlink was skipped and the files remained as orphans.
3008 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3009 Details : In some cases it was possible to send a blocking callback to
3010 a client doing a glimpse, even though that client didn't get
3011 a lock granted. When the glimpse lock is cancelled on the OST
3012 the freed lock is left on the waiting list and corrupted the list.
3015 Frequency : all core dumps
3017 Description: Broke core dumps to lustre
3018 Details : Negative dentry may be unhashed if parent does not have UPDATE
3019 lock, but some callers, e.g. do_coredump, expect dentry to be
3020 hashed after successful create, hash it in ll_create_it.
3022 ------------------------------------------------------------------------------
3024 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3026 * Support for kernels:
3027 2.6.9-42.0.2.EL (RHEL 4)
3028 2.6.5-7.276 (SLES 9)
3029 2.4.21-40.EL (RHEL 3)
3030 2.6.12.6 vanilla (kernel.org)
3034 Frequency : always on RHEL 3
3036 Description: Number of open files grows over time
3037 Details : The number of open files grows over time, whether or not
3038 Lustre is started. This was due to a filp leak introduced
3039 by one of our kernel patches.
3041 ------------------------------------------------------------------------------
3043 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3045 * Support for kernels:
3046 2.6.9-42.EL (RHEL 4)
3047 2.6.5-7.276 (SLES 9)
3048 2.4.21-40.EL (RHEL 3)
3049 2.6.12.6 vanilla (kernel.org)
3054 Bugzilla : 5719, 9635, 9792, 9684
3055 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3056 Details : If a server is under heavy load and cannot reply to new
3057 connection requests before the client resends the (re)connect,
3058 the connection handling code can behave badly if two service
3059 threads are concurrently handing separate (re)connections from
3060 the same client. Add better locking to the connection handling
3061 code, and ensure that only a single connection will be processed
3062 for a given client UUID, even if the lock is dropped.
3064 Severity : enhancement
3066 Description: add TCP zero-copy support to kernel
3067 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3068 sends if the hardware supports scatter-gather and checksumming.
3069 This allows socklnd to do client-write and server-read more
3070 efficiently and reduce CPU utilization from skbuf copying.
3073 Frequency : only if NFS exporting from client
3075 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3076 Details : If setting attributes on a file created under NFS that had
3077 never been opened it would be possible to oops the client
3078 if the file had no objects.
3081 Frequency : always for liblustre
3083 Description: liblustre client does MDS+OSTs setattr RPC for each write
3084 Details : When doing a write from a liblustre client, the client
3085 incorrectly issued an RPC to the MDS and each OST the file was
3086 striped over in order to update the timestamps. When writing
3087 with small chunks and many clients this could overwhelm the MDS
3088 with RPCs. In all cases it would slow down the write because
3089 these RPCs are unnecessary.
3091 Severity : enhancement
3093 Description: allow number of MDS service threads to be changed at module load
3094 Details : It is now possible to change the number of MDS service threads
3095 running. Adding "options mds mds_num_threads={N}" to the MDS's
3096 /etc/modprobe.conf will set the number of threads for the next
3097 time Lustre is restarted (assuming the "mds" module is also
3098 reloaded at that time). The default number of threads will
3099 stay the same, 32 for most systems.
3104 Description: OST crash if filesystem is unformatted or corrupt
3105 Details : If an OST is started on a device that has never been formatted
3106 or if the filesystem is corrupt and cannot even mount then the
3107 error handling cleanup routines would dereference a NULL pointer.
3112 Description: NULL pointer deref in llap_from_page.
3113 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3114 incorrect) mapping if the page was truncated/reclaimed while it was
3115 searched for. Check for this condition and skip such pages when
3116 doing readahead. Introduce extra check to llap_from_page() to
3117 verify page->mapping->host is non-NULL (so page is not anonymous).
3120 Frequency : Sometimes when using sys_sendfile
3122 Description: "page not covered by a lock" warnings from ll_readpage
3123 Details : sendfile called ll_readpage without right page locks present.
3124 Now we introduced ll_file_sendfile that does necessary locking
3125 around call to generic_file_sendfile() much like we do in
3129 Frequency : with certain MDS communication failures at client mount time
3131 Description: NULL pointer deref after failed client mount
3132 Details : a client connection request may delayed by the network layer
3133 and not be sent until after the PTLRPC layer has timed out the
3134 request. If the client fails the mount immediately it will try
3135 to clean up before the network times out the request. Add a
3136 reference from the request import to the obd device and delay
3137 the cleanup until the network drops the request.
3140 Frequency : occasionally during client (re)connect
3142 Description: assertion failure during client (re)connect
3143 Details : processing a client connection request may be delayed by the
3144 client or server longer than the client connect timeout. This
3145 causes the client to resend the connection request. If the
3146 original connection request is replied in this interval, the
3147 client may trip an assertion failure in ptlrpc_connect_interpret()
3148 which thought it would be the only running connect process.
3151 Frequency : only with obd_echo servers and clients that are rebooted
3153 Description: kernel BUG accessing uninitialized data structure
3154 Details : When running an obd_echo server it did not start the ping_evictor
3155 thread, and when a client was evicted an uninitialized data
3156 structure was accessed. Start the ping_evictor in the RPC
3157 service startup instead of the OBD startup.
3159 Severity : enhancement
3160 Bugzilla : 10193 (patchless)
3161 Description: Remove dependency on various unexported kernel interfaces.
3162 Details : No longer need reparent_to_init, exit_mm, exit_files,
3163 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3166 Frequency : rare (only users of deprecated and unsupported LDAP config)
3168 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3169 Details : LDAP apparently contains 'lustreName' attributes instead of
3170 'name'. A simple remapping of the name is sufficient.
3173 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3175 Description: waiting_locks_callback trips kernel BUG if client is evicted
3176 Details : Running with the dump_on_timeout debug flag turned on makes
3177 it possible that the waiting_locks_callback() can try to dump
3178 the Lustre kernel debug logs from an interrupt handler. Defer
3179 this log dumping to the expired_lock_main() thread.
3181 Severity : enhancement
3183 Description: Support NFS exporting on 2.6 kernels.
3184 Details : Implement non-rawops metadata methods for NFS server to use without
3185 changing NFS server code.
3188 Frequency : very rare (synthetic metadata workload only)
3190 Description: two racing renames might cause an MDS thread to deadlock
3191 Details : Running the "racer" program may cause one MDS thread to rename
3192 a file from being the source of a rename to being the target of
3193 a rename at exactly the same time that another thread is doing
3194 so, and the second thread has already enqueued these locks after
3195 doing a lookup of the target and is trying to relock them in
3196 order. Ensure that we don't try to re-lock the same resource.
3199 Frequency : only very large systems with liblustre clients
3201 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3202 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3203 all OSTs in parallel. This allows the eviction of stale liblustre
3204 clients to proceed much faster than if they were done in series,
3205 and also offers similar improvements for other set_info RPCs.
3210 Description: excessive CPU usage during initial read phase on client
3211 Details : During the initial read phase on a client, it would agressively
3212 retry readahead on the file, consuming too much CPU and impacting
3213 performance (since 1.4.5.8). Improve the readahead algorithm
3214 to avoid this, and also improve some other common cases (read
3215 of small files in particular, where "small" is files smaller than
3216 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3221 Description: MDS crash when receiving packet with unknown intent.
3222 Details : Do not LBUG in unknown intent case, just return -EFAULT
3224 Severity : enhancement
3225 Bugzilla : 9293, 9385
3226 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3227 Details : Do not serialize getattr (non-intent version) and statfs.
3230 Frequency : occasional, when OST network is overloaded/intermittent
3232 Description: client evicted by OST after bulk IO timeout
3233 Details : If a client sends a bulk IO request (read or write) the OST
3234 may evict the client if it is unresposive to its data GET/PUT
3235 request. This is incorrect if the network is overloaded (takes
3236 too long to transfer the RPC data) or dropped the OST GET/PUT
3237 request. There is no need to evict the client at all, since
3238 the pinger and/or lock callbacks will handle this, and the
3239 client can restart the bulk request.
3242 Frequency : Always when mmapping file with no objects
3244 Description: client crashes when mmapping file with no objects
3245 Details : Check that we actually have objects in a file before doing any
3246 operations on objects in ll_vm_open, ll_vm_close and
3252 Description: Request leak when working with deleted CWD
3253 Details : Introduce advanced request refcount tracking for requests
3254 referenced from lustre intent.
3256 Severity : Enhancement
3258 Description: Cache open file handles on client.
3259 Details : MDS now will return special lock along with openhandle, if
3260 requested and client is allowed to hold openhandle, even if unused,
3261 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3262 closing files for every read/write openration.
3264 Severity : Enhancement
3266 Description: Cache open negative dentries on client when possible.
3267 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3268 negative dentries on lock revocation.
3273 Description: Remounting a client read-only wasn't possible with a zconf mount
3274 Details : It wasn't possible to remount a client read-only with llmount.
3276 Severity : enhancement
3277 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3278 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3279 MPICH is included to add Lustre-specific ADIO interfaces.
3280 This is based closely on the UFS ADIO layer and only differs
3281 in file creation, in order to allow the OST striping to be set.
3282 This is user-contributed code and not supported by CFS.
3287 Description: extended inode attributes (immutable, append-only) work improperly
3288 when 2.4 and 2.6 kernels are used on client/server or vice versa
3289 Details : Introduce kernel-independent values for these flags.
3291 Severity : enhancement
3294 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3295 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3296 granularity. Now, Fractional MB granularity is be supported,
3297 this is very useful for low memory system.
3299 Severity : enhancement
3301 Description: Getattr by fid
3302 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3303 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3304 have locks and avoids getattr rpc requests for it on every stat.
3307 Frequency : Always, for filesystems larger than 2TB
3309 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3310 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3311 filesystems larger than 2TB to be mounted with mballoc enabled.
3314 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3316 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3317 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3318 filesystems larger than 2TB is mounted then the kernel will
3319 silently corrupt the start of the filesystem. CONFIG_LBD is
3320 enabled for all CFS-supported kernels, but the possibility of
3321 this happening with a modified kernel config exists.
3323 Severity : enhancement
3325 Description: add client O_DIRECT support for 2.6 kernels
3326 Details : It is now possible to do O_DIRECT reads and writes to files
3327 in the Lustre client mountpoint on 2.6 kernel clients.
3329 Severity : enhancement
3331 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3332 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3333 parallel, not waiting for response from every OST before sending
3334 a rpc to the next OST.
3339 Description: setattr vs write race when updating file timestamps
3340 Details : Client processes that update a file timestamp into the past
3341 right after writing to the file (e.g. tar) it is possible that
3342 the updated file modification time can be reset to the current
3343 time due to a race between processing the setattr and write RPC.
3345 Severity : enhancement
3347 Description: Bring 'lfs find' closer in line with regular Linux find.
3348 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3349 -print0 options and obtains all the needed info through the lustre
3352 Severity : enhancement
3354 Description: support up to 1024 configured devices on one node
3355 Details : change obd_dev array from statically allocated to dynamically
3356 allocated structs as they are first used to reduce memory usage
3361 Description: Flush dirty partially truncated pages during truncate
3362 Details : Immediatelly flush partially truncated pages in filter_setattr,
3363 this way we completely avoid having any pages in page cache on OST
3364 and can retire ugly workarounds during writes to flush such pages.
3369 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3370 Details : protect inode from truncation within vfs_unlink() context
3371 just take a reference before calling vfs_unlink() and release it
3372 when parent's i_sem is free.
3377 Description: last_id value checked outside lock on OST caused LASSERT failure
3378 Details : If there were multiple MDS->OST object precreate requests in
3379 flight, it was possible that the OST's last object id was checked
3380 outside a lock and incorrectly tripped an assertion. Move checks
3381 inside locks, and discard old precreate requests.
3384 Frequency : always, if extents are used on OSTs
3386 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3387 Details : OSTs using the extents format would not zero the high 16 bits of
3388 the index physical block number. This is not a problem for any
3389 OST filesystems smaller than 16TB, and no kernels support ext3
3390 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3391 new/modified files) and can be fixed for existing filesystems
3392 with e2fsprogs-1.39-cfs1.
3397 Description: import connection selection may be incorrect if timer wraps
3398 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3399 import connections to be ignored if the 32-bit jiffies counter
3400 wraps. Use a 64-bit jiffies counter.
3403 Frequency : very large clusters immediately after boot
3405 Description: LNET request buffers exhausted under heavy short-term load
3406 Details : If a large number of client requests are generated on a service
3407 that has previously never seen so many requests it is possible
3408 that the request buffer growth cannot keep up with the spike in
3409 demand. Instead of dropping incoming requests, they are held in
3410 the LND until the RPC service can accept more requests.
3413 Frequency : Sometimes during replay
3415 Description: Assertion failure in ll_local_open after replay.
3416 Details : If replay happened on an open request reply before we were able
3417 to set replay handler, reply will become not swabbed tripping the
3418 assertion in ll_local_open. Now we set the handler right after
3419 recognising of open request
3422 Frequency : very rare
3424 Description: kernel reports "badness in vsnprintf"
3425 Details : Reading from the "recovery_status" /proc file in small chunks
3426 may cause a negative length in lprocfs_obd_rd_recovery_status()
3427 call to vsnprintf() (which is otherwise harmless). Exit early
3428 if there is no more space in the output buffer.
3430 Severity : enhancement
3432 Description: clear OBD RPC statistics by writing to them
3433 Details : It is now possible to clear the OBD RPC statistics by writing
3434 to the "stats" file.
3439 Description: Client mtime is not the same on different clients after utimes
3440 Details : In some cases, the client was using the utimes() syscall on
3441 a file cached on another node. The clients now validate the
3442 ctime from the MDS + OSTs to determine which one is right.
3447 Description: Inability to activate failout mode
3448 Details : lconf script incorrectly assumed that in pythong string's numeric
3449 value is used in comparisons.
3452 Frequency : always with multiple stripes per file
3454 Description: Inefficient object allocation for mutli-stripe files
3455 Details : When selecting which OSTs to stripe files over, for files with
3456 a stripe count that divides evenly into the number of OSTs,
3457 the MDS is always picking the same starting OST for each file.
3458 Return the OST selection heuristic to the original design.
3463 Description: mount failures may take full timeout to return an error
3464 Details : Under some heavy load conditions it is possible that a
3465 failed mount can wait for the full obd_timeout interval,
3466 possibly several minutes, before reporting an error.
3467 Instead return an error as soon as the status is known.
3469 Frequency : quota enabled and large files being deleted
3471 Description: releasing more than 4GB of quota at once hangs OST
3472 Details : If a user deletes more than 4GB of files on a single OST it
3473 will cause the OST to spin in an infinite loop. Release
3474 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3479 Description: statfs data retrieved from /proc may be stale or zero
3480 Details : When reading per-device statfs data from /proc, in the
3481 {kbytes,files}_{total,free,avail} files, it may appear
3482 as zero or be out of date.
3485 Frequency : systems with MD RAID1 external journal devices
3487 Description: lconf's call to blkid is confused by RAID1 journal devices
3488 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3489 of returning all block devices that match the journal UUID.
3492 Frequency : always, for aggregate stripe size over 4GB
3494 Description: assertion fails when trying to use 4GB stripe size
3495 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3496 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3499 Frequency : always on ppc64
3501 Description: the first write on an ext3 filesystem with mballoc got stuck
3502 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3503 perform endianness conversion.
3505 ------------------------------------------------------------------------------
3507 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3509 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3510 INTEROPERATE with older versions automatically. Please read the
3511 user documentation before upgrading any part of a live system.
3512 * WARNING: Lustre networking configuration changes are required with
3513 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3516 * Support for kernels:
3517 2.6.9-22.0.2.EL (RHEL 4)
3518 2.6.5-7.244 (SLES 9)
3519 2.6.12.6 vanilla (kernel.org)
3522 Severity : enhancement
3523 Bugzilla : 7981/8208
3524 Description: Introduced Lustre Networking (LNET)
3525 Details : LNET is new networking infrastructure for Lustre, it includes
3526 a reorganized network configuration mode (see the user
3527 documentation for full details) as well as support for routing
3528 between different network fabrics. Lustre Networking Devices
3529 (LNDs) for the supported network fabrics have also been
3530 created for this new infrastructure.
3532 Severity : enhancement
3533 Description: Introduced Access control lists
3534 Details : clients can set ACLs on files and directories in order to have
3535 more fine-grained permissions than the standard Unix UGO+RWX.
3536 The MDS must be started with the "-o acl" mount option.
3538 Severity : enhancement
3539 Description: Introduced filesystem quotas
3540 Details : Administrators may now establish per-user quotas on the
3543 Severity : enhancement
3545 Description: Configuration change for the XT3
3546 The PTLLND is now used to run Lustre over Portals on the XT3
3547 The configure option(s) --with-cray-portals are no longer used.
3548 Rather --with-portals=<path-to-portals-includes> is used to
3549 enable building on the XT3. In addition to enable XT3 specific
3550 features the option --enable-cray-xt3 must be used.
3555 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3556 Details : A race between allocating a new llcd and re-getting the llcd_lock
3557 allowed another thread to grab newly-allocated llcd.
3559 Severity : enhancement
3561 Description: 2.6 OST async journal commit and locking fix to improve performance
3562 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3563 2.6 kernels are now async as they already were in 2.4 kernels so
3564 that they can commit concurrently with the network bulk transfer.
3565 For block-allocated files the filter allocation semaphore is held
3566 to avoid filesystem fragmentation during allocation. BKL lock
3567 removed for 2.6 xattr operations where it is no longer needed.
3572 Description: lconf incorrectly determined whether two IP networks could talk
3573 Details : In some more complicated routing and multiple-network
3574 configurations, lconf will avoid trying to make a network
3575 connection to a disjoint part of the IP space. It was doing the
3576 math incorrectly for one set of cases.
3581 Description: Fix for potential infinite loop processing records in an llog.
3582 Details : If an llog record is corrupted/zeroed, it is possible to loop
3583 forever in llog_process(). Validate the llog record length
3584 and skip the remainder of the block on error.
3587 Frequency : occasional (liblustre only)
3589 Description: liblustre could not open files whose last component is a symlink
3590 Details : sysio_path_walk() would incorrectly pass the open intent to
3591 intermediate path components.
3594 Frequency : rare (liblustre only with non-standard tuning)
3595 Bugzilla : 7201 (7350)
3596 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3597 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3598 after it has been released, only do this for referenced locks
3600 Severity : enhancement
3602 Description: specifying an (invalid) directory default stripe_size of -1
3603 would reset the directory default striping
3604 Details : stripe_size -1 was used internally to signal directory stripe
3605 removal, now use "all default" to signal dir stripe removal
3606 as a directory striping of "all default" is not useful
3609 Frequency : common for large clusters running liblustre clients
3611 Description: doing an ls when liblustre clients are running is slow
3612 Details : sending a glimpse AST to a liblustre client waits for every AST
3613 to time out, as liblustre clients will not respond. Since they
3614 cannot cache data we refresh the OST lock LVB from disk instead.
3616 Severity : enhancement
3618 Description: doing an ls at the same time as file IO can be slow
3619 Details : enqueue and other "small" requests can be blocked behind many
3620 large IO requests. Create a new OST IO portal for non-IO
3621 requests so they can be processed faster.
3624 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3626 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3627 created file with mode 000
3628 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3629 allow an ftruncate() as a non-root user to a file with mode 000.
3630 Lustre now respects this flag to disable mode checks when
3631 truncating a file owned by the user
3634 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3637 Description: Revoking locks from clients that went dead or catatonic might take
3639 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3640 cancellation of such locks instant on servers without waiting for
3641 any reply from clients. Clients drops these locks when cancel
3642 notification from server is received without replying.
3645 Frequency : liblustre-only, when liblustre client dies or becomes busy
3647 Description: Doing ls on Linux clients can take a long time with active
3649 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3650 granting such locks to it in the first place if possible. Locks
3651 are taken by proxy on the OST during the read or write and
3652 dropped immediately afterward. Add connect flags handling, do
3653 not grant locks to liblustre clients for glimpse ASTs.
3655 Severity : enhancement
3657 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3658 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3659 sequential reads and behaves badly with applications doing random
3660 reads. Improve it by reading ahead at least read region, and
3661 avoiding excessive large RPC for small reads.
3663 Severity : enhancement
3665 Description: Creating more than 1000 files for a single job may cause a load
3666 imbalance on the OSTs if there are also a large number of OSTs.
3667 Details : qos_prep_create() uses an OST index reseed value that is an
3668 even multiple of the number of available OSTs so that if the
3669 reseed happens in the middle of the object allocation it will
3670 still utilize the OSTs as uniformly as possible.
3675 Description: OST or MDS may oops in ping_evictor_main()
3676 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3677 but doesn't restart at beginning of obd_exports_timed list
3680 Severity : enhancement
3682 Description: improve by-nid export eviction on the MDS and OST
3683 Details : allow multiple exports with the same NID to be evicted at one
3684 time without re-searching the exports list.
3687 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3689 Description: MDS may oops in groups_free()
3690 Details : in rare race conditions a newly allocated group_info struct is
3691 freed again, and this can be NULL. The 2.4 compatibility code
3692 for groups_free() checked for a NULL pointer, but 2.6 did not.
3695 Frequency : common for liblustre clients doing little filesystem IO
3696 Bugzilla : 9352, 7313
3697 Description: server may evict liblustre clients accessing contended locks
3698 Details : if a client is granted a lock or receives a completion AST
3699 with a blocking AST already set it would not reply to the AST
3700 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3703 Frequency : lfs setstripe, only systems with more than 160 OSTs
3705 Description: unable to set striping with a starting offset beyond OST 160
3706 Details : llapi_create_file() incorrectly limited the starting stripe
3707 index to the maximum single-file stripe count.
3710 Frequency : LDAP users only
3712 Description: lconf did not handle in-kernel recovery with LDAP properly
3713 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3715 Severity : enhancement
3717 Description: bind OST threads to NUMA nodes to improve performance
3718 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3719 node and do their allocations there to localize memory access
3721 Severity : enhancement
3723 Description: llmount can determine client NID directly from Myrinet (GM)
3724 Details : the client NID code from gmnalnid was moved directly into
3725 llmount, removing the need to use this or specifying the
3726 client NID explicitly when mounting GM clients with zeroconf
3729 Frequency : if client is started with down MDS
3731 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3732 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3733 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3734 ptlrpc_import_delayed_req() because we didn't check if the
3735 request was interrupted, and we also didn't break out of the
3736 event loop if there was no timeout
3741 Description: data loss during non-page-aligned writes to a single file from
3742 both multiple nodes and multiple threads on one node at same time
3743 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3744 inconsistency led to false short-reads, that were cached and later
3745 used by ->prepare_write() to fill in partially written page,
3746 leading to data loss.
3749 Frequency : always, if lconf --abort_recovery used
3751 Description: lconf --abort_recovery fails with 'Operation not supported'
3752 Details : lconf was attempting to abort recovery on the MDT device and not
3755 Severity : enhancement
3757 Description: remove cleanup logs
3758 Details : replace lconf-generated cleanup logs with lustre internal
3759 cleanup routines. Eliminates the need for client-cleanup and
3762 Severity : enhancement
3764 Description: add support for EAs (user and system) on lustre filesystems
3765 Details : it is now possible to store extended attributes in the Lustre
3766 client filesystem, and with the user_xattr mount option it
3767 is possible to allow users to store EAs on their files also
3769 Severity : enhancement
3771 Description: Add possibility (config option) to show minimal available OST free
3773 Details : When compiled with --enable-mindf configure option, statfs(2)
3774 (and so, df) will return least minimal free space available from
3775 all OSTs as amount of free space on FS, instead of summary of
3776 free spaces of all OSTs.
3778 Severity : enhancement
3780 Description: do not expand extent locks acquired on OST-side
3781 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3782 by server: they are not cached anyway.
3785 Frequency : when mmap is used/binaries executed from Lustre
3787 Description: Unmmap pages before throwing them away from read cache.
3788 Details : llap_shrink cache now attempts to unmap pages before discarding
3789 them (if unmapping failed - do not discard). SLES9 kernel has
3790 extra checks that trigger if this unmapping is not done first.
3795 Description: lconf didn't resolve symlinks before checking to see whether a
3796 given mountpoint was already in use
3799 Frequency : when migrating failover services
3800 Bugzilla : 6395, 9514
3801 Description: When migrating a subset of services from a node (e.g. failback
3802 from a failover service node) the remaining services would
3803 time out and evict clients.
3804 Details : lconf --force (implied by --failover) sets the global obd_timeout
3805 to 5 seconds in order to quickly disconnect, but this caused
3806 other RPCs to time out too quickly. Do not change the global
3807 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3809 Severity : enhancement
3810 Frequency : if MDS is started with down OST
3811 Bugzilla : 9439,5706
3812 Description: Allow startup/shutdown of an MDS without depending on the
3813 availability of the OSTs.
3814 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3815 Add appropriate locking and lov-osc refcounts for safe
3816 cleaning. Add osc abort_inflight calls in case the
3817 synchronize never started.
3820 Frequency : occasional (Cray XT3 only)
3822 Description: root not authorized to access files in CRAY_PORTALS environment
3823 Details : The client process capabilities were not honoured on the MDS in
3824 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3825 been accessed by an authorized user then root was able to access
3826 the file on the local client also. The root user capabilities
3827 are now allowed on the MDS, as this environment has secure UID.
3830 Frequency : occasional
3832 Description: ldiskfs "too long searching" message happens too often
3833 Details : A debugging message (otherwise harmless) prints too often on
3834 the OST console. This has been reduced to only happen when
3835 there are fragmentation problems on the filesystem.
3840 Description: Division by zero in statfs when all OSCs are inactive
3841 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3842 lov_statfs divides by value returned by lov_get_stripecnt().
3846 Bugzilla : 9489, 3273
3847 Description: First write from each client to each OST was only 4kB in size,
3848 to initialize client writeback cache, which caused sub-optimal
3849 RPCs and poor layout on disk for the first writen file.
3850 Details : Clients now request an initial cache grant at (re)connect time
3851 and so that they can start streaming writes to the cache right
3852 away and always do full-sized RPCs if there is enough data.
3853 If the OST is rebooted the client also re-establishes its grant
3854 so that client cached writes will be honoured under the grant.
3859 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3860 Details : Now I/O RPCs go to different portal number and (presumably) fast
3861 lock requests (and glimses) and other RPCs get their own service
3862 threads pool that should be able to service those RPCs
3865 Severity : enhancement
3867 Description: Ability to exchange lustre version between client and servers and
3868 issue warnings at client side if client is too old. Also for
3869 liblustre clients there is ability to refuse connection of too old
3871 Details : New 'version' field is added to connect data structure that is
3872 filled with version info. That info is later checked by server and
3876 Frequency : rare, liblustre only.
3877 Bugzilla : 9296, 9581
3878 Description: Two simultaneous writes from liblustre at offset within same page
3879 might proceed at the same time overwriting eachother with stale
3881 Details : I/O lock withing llu_file_prwv was released too early, before data
3882 actually was hitting the wire. Extended lock-holding time until
3883 server acknowledges receiving data.
3886 Frequency : extremely rare. Never observed in practice.
3888 Description: avoid generating lustre_handle cookie of 0.
3889 Details : class_handle_hash() generates handle cookies by incrementing
3890 global counter, and can hit 0 occasionaly (this is unlikely, but
3891 not impossible, because initial value of cookie counter is
3892 selected randonly). Value of 0 is used as a sentinel meaning
3893 "unassigned handle" --- avoid it. Also coalesce two critical
3894 sections in this function into one.
3896 Severity : enhancement
3898 Description: allow liblustre clients to delegate truncate locking to OST
3899 Details : To avoid overhead of locking, liblustre client instructs OST to
3900 take extent lock in ost_punch() on client's behalf. New connection
3901 flag is added to handle backward compatibility.
3903 Severity : enhancement
3904 Bugzilla : 4928, 7341, 9758
3905 Description: allow number of OST service threads to be specified
3906 Details : a module parameter allows the number of OST service threads
3907 to be specified via "options ost ost_num_threads={N}" in the
3908 OSS's /etc/modules.conf or /etc/modprobe.conf.
3912 Bugzilla : 6146, 9635, 9895
3913 Description: servers crash with bad pointer in target_handle_connect()
3914 Details : In rare cases when a client is reconnecting it was possible that
3915 the connection request was the last reference for that export.
3916 We would temporarily drop the export reference and get a new
3917 one, but this may have been the last reference and the export
3918 was just destroyed. Get new reference before dropping old one.
3920 Severity : enhancement
3921 Frequency : if client is started with failover MDS
3923 Description: Allow multiple MDS hostnames in the mount command
3924 Details : Try to read the configuration from all specified MDS
3925 hostnames during a client mount in case the "primary"
3928 Severity : enhancement
3930 Description: Stop sending data to evicted clients as soon as possible.
3931 Details : Check if the client we are about to send or are sending data to
3932 was evicted already. (Check is done every second of waiting,
3933 for which l_wait_event interface was extended to allow checking
3934 of exit condition at specified intervals).
3937 Frequency : rare, normally only when NFS exporting is done from client
3939 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3941 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3944 Frequency : always, if srand() is called before liblustre initialization
3946 Description: Liblustre uses system PRNG disturbing its usage by user application
3947 Details : Introduce internal to lustre fast and high-quality PRNG for
3948 lustre usage and make liblustre and some other places in generic
3949 lustre code to use it.
3951 Severity : enhancement
3952 Bugzilla : 9477, 9557, 9870
3953 Description: Verify that the MDS configuration logs are updated when xml is
3954 Details : Check if the .xml configuration logs are newer than the config
3955 logs stored on the MDS and report an error if this is the case.
3956 Request --write-conf, or allow starting with --old_conf.
3958 Severity : enhancement
3960 Description: Handle symlinks in the path when checking if Lustre is mounted.
3961 Details : Resolve intermediate symlinks when checking if a client has
3962 mounted a filesystem to avoid duplicate client mounts.
3967 Description: lconf can hit an error exception but still return success.
3968 Details : The lconf command catches the Command error exception at the top
3969 level script context and will exit with the associated exit
3970 status, but doesn't ensure that this exit status is non-zero.
3975 Description: failure of ptlrpc thread startup can cause oops
3976 Details : Starting a ptlrpc service thread can fail if there are a large
3977 number of threads or the server memory is very fragmented.
3978 Handle this without oopsing.
3981 Frequency : always, only if liblustre and non-default acceptor port was used
3983 Description: liblustre cannot connect to servers with non-default acceptor port
3984 Details : tcpnal_set_default_params() was not called and was therefore
3985 ignoring the environment varaible TCPNAL_PORT, as well as other
3986 TCPNAL_ environment variables
3991 Description: two objects could be created on the same OST for a single file
3992 Details : If an OST is down, in some cases it was possible to create two
3993 objects on a single OST for a single file. No problems other
3994 than potential performance impact and spurious error messages.
3998 Bugzilla : 5681, 9562
3999 Description: Client may oops in ll_unhash_aliases
4000 Details : Client dcache may become inconsistent in race condition.
4001 In some cases "getcwd" can fail if the current directory is
4007 Description: Inode refcounting problems in NFS export code
4008 Details : link_raw functions used to call d_instantiate without obtaining
4009 extra inode reference first.
4013 Bugzilla : 9942, 9903
4014 Description: Referencing freed requests leading to crash, memleaks with NFS.
4015 Details : We used to require that call to ll_revalidate_it was always
4016 followed by ll_lookup_it. Also with revalidate_special() it is
4017 possible to call ll_revalidate_it() twice for the same dentry
4018 even if first occurence returned success. This fix changes semantic
4019 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4020 reference on a request referred from the intent.
4021 ll_intent_release() then releases such a request.
4024 Frequency : rare, normally benchmark loads only
4026 Description: unlinked inodes were kept in memory on the client
4027 Details : If a client is repeatedly creating and unlinking files it
4028 can accumulate a lot of stale inodes in the inode slab cache.
4029 If there is no other client load running this can cause the
4030 client node to run out of memory. Instead flush old inodes
4031 from client cache that have the same inode number as a new inode.
4034 Frequency : SLES9 2.6.5 kernel and long filenames only
4035 Bugzilla : 9969, 10379
4036 Description: utime reports stale NFS file handle
4037 Details : SLES9 uses out-of-dentry names in some cases, which confused
4038 the lustre dentry revalidation. Change it to always use the
4042 Frequency : rare, unless heavy write-truncate concurrency is continuous
4043 Bugzilla : 4180, 6984, 7171, 9963, 9331
4044 Description: OST becomes very slow and/or deadlocked during object unlink
4045 Details : filter_destroy() was holding onto the parent directory lock
4046 while truncating+unlinking objects. For very large objects this
4047 may block other threads for a long time and slow overall OST
4048 responsiveness. It may also be possible to get a lock ordering
4049 deadlock in this case, or run out of journal credits because of
4050 the combined truncate+unlink. Solution is to do object truncate
4051 first in one transaction without parent lock, and then do the
4052 final unlink in a new transaction with the parent lock. This
4053 reduces the lock hold time dramatically.
4056 Frequency : rare, 2.4 kernels only
4058 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4059 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4060 before unmounting their filesystem in order to avoid deadlock.
4061 The kernel_locked() function in 2.4 kernels only checks whether
4062 the kernel lock is held, not whether it is this process that is
4063 holding it as 2.6 kernels do.
4068 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4069 Details : The client ptlrpc code may be trying to reconnect to a down
4070 server before a previous connection attempt has timed out.
4071 Increase the reconnect interval to be longer than the connection
4072 timeout interval to avoid sending duplicate connections to
4076 Frequency : echo_client brw_test command
4078 Description: fix echo_client to work with OST preallocated code
4079 Details : OST preallocation code (5137) didn't take echo_client IO path
4080 into account: echo_client calls filter methods outside of any
4081 OST thread and, hence, there is no per-thread preallocated
4082 pages and buffers to use. Solution: hijack pga pages for IO. As
4083 a byproduct, this avoids unnecessary data copying.
4087 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4088 Description: Client can oops in mdc_commit_close() after open replay
4089 Details : It was possible for the MDS to return an open request with no
4090 transaction number in mds_finish_transno() if the client was
4091 evicted, but without actually returning an error. Clients
4092 would later try to replay that open and may trip an assertion
4093 Simplify the client close codepath, and always return an error
4094 from the MDS in case the open is not successful.
4097 Frequency : rare, 2.6 OSTs only
4099 Description: OST may deadlock under high load on fragmented files
4100 Details : If there was a heavy load and highly-fragmented OST filesystems
4101 it was possible to have all the OST threads deadlock waiting on
4102 allocation of biovecs, because the biovecs were not released
4103 until the entire RPC IO was completed. Instead, release biovecs
4104 as soon as they are complete to ensure forward IO progress.
4106 Severity : enhancement
4108 Description: Support for specifying external journal device at mount
4109 Details : If an OST or MDS device is formatted with an external journal
4110 device, this device major/minor is stored in the ext3 superblock
4111 and may not be valid for failover. Allow detecting and
4112 specifying the external journal at mount time.
4117 Description: Mounting an MDS with pending unlinked files may cause oops
4118 Details : target_finish_recovery() calls mds_postrecov() which returned
4119 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4120 considers this an error and immediately begins cleaning up the
4121 lov, just after starting the mds_lov process
4123 Severity : enhancement
4125 Description: Implement 'lfs df' to report actual free space on per-OST basis
4126 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4127 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4128 usage of objects; '-h' to report in human readable format.
4130 ------------------------------------------------------------------------------
4132 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4139 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4140 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4141 at startup during orphan unlinking. Instead, skip buddy update
4142 before setup, as it will be regenerated after recovery is complete.
4147 Description: If an OST is inactive, its locks might reference stale inodes.
4148 Details : lov_change_cbdata() must iterate over all namespaces, even if
4149 they are inactive to clear inode references from the lock.
4151 Severity : enhancement
4152 Frequency : occasional, if non-standard max_dirty_mb used
4154 Description: Client will block write RPCs if not enough grant
4155 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4156 then the client will block writes while waiting for another RPC
4157 to complete instead of consuming its dirty limit. With change
4158 we get improved performance when max_dirty_mb is small.
4160 Severity : enhancement
4161 Bugzilla : 3389, 6253
4162 Description: Add support for supplementary groups on the MDS.
4163 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4164 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4165 lookups for user supplementary groups into a cache.
4169 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4170 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4171 for liblustre O_EXCL is enforced by MDS.
4175 Description: readdir never returns NULL in liblustre.
4176 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4177 dentry in struct dirent.
4181 Frequency : liblustre only, depends on application IO pattern
4182 Description: liblustre clients evicted if not contacting servers
4183 Details : Don't put liblustre clients into the ping_evictor list, so
4184 they will not be evicted by the pinger ever.
4186 Severity : enhancement
4188 Description: Add ability to evict clients by NID from MDS.
4189 Details : By echoing "nid:$NID" string into
4190 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4191 $NID would be instantly evicted from this MDS and from all active
4192 OSTs connected to it.
4196 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4197 Details : lookup_it_finish() used to query file size from OSTs that was not
4202 Description: service threads change working directory to that of init
4203 Details : Starting lustre service threads may pin the working directory
4204 of the parent thread, making that filesystem busy. Threads
4205 now change to the working directory of init to avoid this.
4209 Frequency : during shutdown only
4210 Description: shutdown with a failed MDS or OST can cause unmount to hang
4211 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4216 Frequency : 2.6 only
4217 Description: chmod/chown may include an extra supplementary group
4218 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4219 supplementary group and if none is specified this is used.
4222 Bugzilla : 5479 (6816)
4224 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4225 Details : If lookup is in progress on a file that is unlinked we might try
4226 to revalidate the inode and fail in revalidate after lookup is
4227 complete and ll_file_open() enqueues the open again but
4228 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4231 Frequency : always, if lconf --abort_recovery used
4233 Description: lconf --abort_recovery fails with 'Operation not supported'
4234 Details : lconf was attempting to abort recovery on the MDT device and not
4237 ------------------------------------------------------------------------------
4239 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4244 Frequency : rare (only unsupported configurations with a node running as an
4246 Bugzilla : 6514, 5137
4247 Description: Mounting a Lustre file system on a node running as an OST could
4249 Details : OSTs now preallocates memory needed to write out data at
4250 startup, instead of when needed, to avoid having to
4251 allocate memory in possibly low memory situations.
4252 Specifically, if the file system is mounted on on OST,
4253 memory pressure could force it to try to write out data,
4254 which it needed to allocate memory to do. Due to the low
4255 memory, it would be unable to do so and the node would
4256 become unresponsive.
4258 Severity : enhancement
4260 Description: Addition of lconf --service command line option
4261 Details : lconf now accepts a '--service <arg>' option, which is
4262 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4264 Severity : enhancement
4266 Description: Failover mode is now the default for OSTs.
4267 Details : By default, OSTs will now run in failover mode. To return to
4268 the old behaviour, add '--failout' to the lmc line for OSTs.
4270 Severity : enhancement
4272 Description: Health checks are now provided for MDS and OSTs
4273 Details : Additional detailed health check information on MSD and OSTs
4274 is now provided through the procfs health_check value.
4277 Frequency : occasional, depends on IO load
4279 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4280 after numerous create/delete cycles
4281 Details : The ext3 inode allocation policy would not allocate new inodes
4282 very well on the OSTs because there are no new directories
4283 being created. Instead we look for groups with free space if
4284 the parent directories are nearly full.
4289 Description: Network or server problems during mount may cause partially
4290 mounted clients instead of returning an error.
4291 Details : The config llog parsing code may overwrite the error return
4292 code during mount error handling, returning success instead
4298 Description: MDS can fail to allocate large reply buffers
4299 Details : After long uptimes the MDS can fail to allocate large reply
4300 buffers (e.g. zconf client mount config records) due to memory
4301 fragmentation or consumption by the buffer cache. Preallocate
4302 some large reply buffers so that these replies can be sent even
4303 under memory pressure.
4307 Frequency : rare (liblustre)
4308 Description: fsx running with liblustre complained that using truncate() to
4309 extend the file doesn't work. This patch corrects that issue.
4310 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4311 ATTR_SIZE and lsm use in llu_setattr_raw.
4315 Frequency : rare, only 2.6 kernels
4316 Description: Unusual file access patterns on the MDS may result in inode
4317 data being lost in very rare circumstances.
4318 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4319 code in the 2.6 kernel caused the inode and/or EA data not to
4320 be read from disk, causing single-file corruption.
4324 Frequency : rare, only 2.6 filesystems using extents
4325 Description: Heavy concurrent write and delete load may cause data corruption.
4326 Details : It was possible under high-load situations to have an extent
4327 metadata block in the block device cache from a just-unlinked
4328 file overwrite a newly-allocated data block. We now unmap any
4329 metadata buffers that alias just-allocated data blocks.
4333 Frequency : filesystems with default stripe_count larger than 77
4334 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4335 Details : lconf specifies an inode size of 4096 bytes when the default
4336 stripe_count is larger than 77. This conflicts with the default
4337 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4338 this case to avoid pinning too much memory for large EAs.
4340 ------------------------------------------------------------------------------
4342 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4347 Frequency : rare (extremely heavy IO load with hundreds of clients)
4349 Description: Client is evicted, gets IO error writing to file
4350 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4351 caused clients to be evicted for AST timeouts. The fixes in
4352 bug 5192 mean we no longer need to have such short AST timeouts
4353 so ldlm_timeout has been increased.
4356 Frequency : occasional during --force or --failover shutdown under load
4357 Bugzilla : 5949, 4834
4358 Description: Server oops/LBUG if stopped with --force or --failover under load
4359 Details : a collection of import/export refcount and cleanup ordering
4360 issues fixed for safer force cleanup
4363 Frequency : only filesystems larger than 120 OSTs
4364 Bugzilla : 5990, 6223
4365 Description: lfs getstripe would oops on a very large filesystem
4366 Details : lov_getconfig used kfree on vmalloc'd memory
4369 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4370 Bugzilla : 6242, 6243
4371 Description: reading from files that had been truncated to a non-zero size
4372 but never opened returned no data
4373 Details : ll_file_read() reads zeros from no-object files to EOF
4378 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4379 Details : There's little harm in aborting MDS/OSS recovery and letting it
4380 try again, so I removed the LASSERT and return an error instead.
4382 Severity : enhancement
4384 Description: New debugging infrastructure for tracking down data corruption
4385 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4386 (b) cover more of the client-side code path, and (c) try to narrow
4387 down where problems occurred
4391 Bugzilla : 3819, 4364, 4397, 6313
4392 Description: Racing close and eviction MDS could cause assertion in mds_close
4393 Details : It was possible to get multiple mfd references during close and
4394 client eviction, leading to one thread referencing a freed mfd.
4396 Severity: : enhancement
4397 Bugzilla : 3262, 6359
4398 Description: Attempts to reconnect to servers are now more aggressive.
4399 Details : This builds on the enhanced upcall-less recovery that was added
4400 in 1.4.2. When trying to reconnect to servers, clients will
4401 now try each server in the failover group every 10 seconds. By
4402 default, clients would previously try one server every 25 seconds.
4407 Description: After recovery, certain operations trigger a failed
4408 assertion on a client.
4409 Details : Failing over an mds, using lconf -d --failover, while a
4410 client was doing a readdir() call would cause the client to
4411 LBUG after recovery completed and the readdir() was resent.
4413 Severity : enhancement
4415 Description: Default groups are now added by lconf
4416 Details : You can now run lconf --group <servicename> without having to
4417 manually add groups with lmc.
4420 Frequency : occasional
4422 Description: Nodes with an elan id of 0 trigger a failed assertion
4425 Frequency : always when accessing e.g. tty/console device nodes
4427 Description: tty and some other devices nodes cannot be used on lustre
4428 Details : file's private_data field is used by device data and lustre
4429 values in there got lost. New field was added to struct file to
4430 store fs-specific private data.
4433 Frequency : when exporting Lustre via NFS
4435 Description: NFSD failed occasionally when looking up a path component
4436 Details : NFSD is looking up ".." which was broken in ext3 directories
4437 that had grown large enough to become hashed.
4440 Frequency : Clusters with multiple interfaces not on the same subnet
4442 Description: Nodes will repeatedly try to reconnect to an interface which it
4443 cannot reach and report an error to the log.
4444 Details : Extra peer list entries will be created by lconf with some peers
4445 unreachable. lconf now validates the peer before adding it.
4448 Frequency : Only if a default stripe is set on the filesystem root.
4450 Description: Setting a default stripe on the filesystem root prevented the
4451 filesystem from being remounted.
4452 Details : The client was sending extra request flags in the root getattr
4453 request and did not allocate a reply buffer for the dir EA.
4456 Frequency : occasional, higher if lots of files are accessed by one client
4457 Bugzilla : 6159, 6097
4458 Description: Client trips assertion regarding lsm mismatch/magic
4459 Details : While revalidating inodes the VFS looks up inodes with ifind()
4460 and in rare cases can find an inode that is being freed.
4461 The ll_test_inode() code will free the lsm during ifind()
4462 when it finds an existing inode and then the VFS later attaches
4463 this free lsm to a new inode.
4467 Bugzilla : 6422, 7030
4468 Description: MDS deadlock between mkdir and client eviction
4469 Details : Creating a new file via mkdir or mknod (starting a transaction
4470 and getting the ns lock) can deadlock with client eviction
4471 (gets ns lock and trying to finish a synchronous transaction).
4474 Frequency : occasional
4475 Description: While starting a server, the fsfilt_ext3 module could not be
4477 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4478 kernels. Previously, lconf would still use the ext3 name
4479 when trying to load modules. Now, it will correctly use
4480 ext3 on 2.4 and ldiskfs on 2.6.
4482 Severity : enhancement
4483 Description: The default stripe count has been changed to 1
4484 Details : The interpretation of the default stripe count (0, to lfs
4485 or lmc) has been changed to mean striping across a single
4486 OST, rather than all available. For general usage we have
4487 found a stripe count of 1 or 2 works best.
4489 Severity : enhancement
4490 Description: Add support for compiling against Cray portals.
4491 Details : Conditional compiling for some areas that are different
4495 Frequency : occasional
4496 Bugzilla : 6409, 6834
4497 Description: Creating files with an explicit stripe count may lead to
4498 a failed assertion on the MDS
4499 Details : If some OSTs are full or unavailable, creating files may
4500 trigger a failed assertion on the MDS. Now, Lustre will
4501 try to use other servers or return an error to the
4505 Frequency : occasional
4507 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4508 caused lock timeout during readahead (since 1.4.2).
4509 Details : Processes doing readahead might match a lock that hasn't been
4510 granted yet if there are overlapping and conflicting lock
4511 requests. The readahead process waits on ungranted lock
4512 (original lock is CBPENDING), while OST waits for that process
4513 to cancel CBPENDING read lock and eventually evicts client.
4515 Severity : enhancement
4517 Description: Initial enabling of flock support for clients
4518 Details : Implements fcntl advisory locking and file status functions.
4519 This feature is provided as an optional mount flag (default
4520 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4521 locking are implemented yet, and those that are are not guaranteed
4522 to be completely correct in production environments.
4523 mount -t lustre -o [flock|noflock] ...
4526 Frequency : occasional
4528 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4529 assertion in the ext3 JBD (journaling) layer.
4530 Details : The b_committed_data struct is protected by the big kernel lock
4531 in 2.4 kernels, serializing journal_commit_transaction() and
4532 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4533 access to this struct. In 2.6 kernels there is finer grained
4534 locking to improve SMP performance of the JBD layer.
4538 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4543 Description: merge and cleanup kernel patches.
4546 -----------------------------------------------------------------------------
4548 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4550 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4551 versions of Lustre. Please update all servers and clients to
4552 version 1.4.2 or later at the same time. You must also run
4553 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4554 to update the configuration logs.
4556 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4557 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4558 - swab llog records of type '0' so we get proper header size/idx (5861)
4559 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4560 - fix rename of one directory over another leaking an inode (5953)
4561 - avoid SetPageDirty on 2.6 (5981)
4562 - don't re-add just-being-destroyed locks to the waiting list (5653)
4563 - when creating new directories, inherit the parent's custom
4564 striping settings if present parent (3048)
4565 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4566 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4567 - don't hold client locks on temporary worklist from l_lru (5666)
4568 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4569 - automatically evict dead clients (5921)
4570 - Update file size properly in create+truncate+fstat case (6196)
4571 - Do not unhash mountpoint dentries, do not allow removal of
4573 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4574 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4575 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4576 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4577 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4578 - Correctly handle reads of files with no objects (6243)
4579 - lctl recover will also mark a device active if deactivate used (5933)
4581 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4582 - allow --write-conf on an MDS with different nettype than client (5619)
4583 - don't write config llogs to MDS for mounts not from that MDS (5617)
4584 - lconf should create multiple TCP connections from a client (5201)
4585 - init scripts are now turned off by default; run chkconfig --on
4586 lustre and chkconfig --on lustrefs to use them
4587 - upcalls are no longer needed for clients to recover to failover
4589 - add --abort-recovery option to lconf to abort recovery on device
4591 - add support for an arbitrary number of OSTs (3026)
4592 - Quota support protocol changes.
4593 - forward compatibility changes to wire structs (6007)
4594 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4595 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4596 - improved "lustre status" script
4597 - initialize blocksize for non-regular files (6062)
4598 - added --disable-server and --disable-client configure options (5782)
4599 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4600 - Vanilla 2.4.29 support
4601 - increase maximum number of obd devices to 520 (6242)
4602 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4603 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4604 SLES 9 2.6.5 kernels
4605 - limit stripes per file to 160 (the maximum EA size) (6093)
4607 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4610 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4611 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4612 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4613 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4614 - small ext3 extents cleanups and fixes (5733)
4615 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4616 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4617 - avoid needless client->OST connect, fix handle mismatch (5317)
4618 - fix DLM error path that led to out-of-sync client, long delays (5779)
4619 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4620 - fix several locking issues related to i_size (5492,5624,5654,5672)
4621 - don't move pending lock onto export if it is already evicted (5683)
4622 - fix kernel oops when creating .foo in unlinked directory (5548)
4623 - fix deadlock in obdfilter statistics vs. object create (5811)
4624 - use time_{before,after} to avoid timer jiffies wrap (5882)
4625 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4626 - Do not leak request if server was not able to process it (5154)
4627 - If mds_open unable to find parent dir, make that negative lookup(5154)
4628 - don't create new directories with extent-mapping (5909, 5936)
4630 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4631 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4632 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4633 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4634 - support for RHEL4 kernel on i686 (5773)
4635 - provide error messages when incompatible logs are encountered (5898)
4637 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4638 * version 1.4.0.10 (1.4.1 release candidate 1)
4640 - don't keep a lock reference when lock is not granted (4238)
4641 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4642 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4643 - drop import inflight refcount on signal_completed_replay error (5255)
4644 - unlock page after async write error during send (3677)
4645 - handle missing objects in filter_preprw_read properly (5265)
4646 - no transno return for symlink open, don't save no-trasno open (3440)
4647 - don't try to complete elan receive that already failed (4012)
4648 - free RPC server reply state on error (5406)
4649 - clean up thread from ptlrpc_start_thread() on error (5160)
4650 - readahead could read extra page into cache that wasn't ejected (5388)
4651 - prevent races in class_attach/setup/cleanup/detach (5260)
4652 - don't dereference de->d_inode after l_dput of de (5458)
4653 - use "int" for stripe value returned from lock_to_stripe (5544)
4654 - mballoc allocation and error-checking fixes in 2.6 (5504)
4655 - block device patches to fix I/O request sizes in 2.6 (5482)
4656 - look up hostnames for IB nals (5602)
4657 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4658 - don't start multiple acceptors for the same port (5277)
4659 - fix incorrect LASSERT in mds_getattr_name (5635)
4660 - export a proc file for general "ping" checking (5628)
4661 - fix "lfs check" to not block when the MDS is down (5628)
4663 - service request history (4965)
4664 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4665 - create an "evict_client" /proc entry on OSTs, like the MDS has
4666 - fix mount usage message, return errors per mount(8) (5168)
4667 - change grep [] to grep "[]" in tests so they work in more UMLs
4668 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4669 - remove ancient LOV_MAGIC_V0 EA support (5047)
4670 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4671 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4672 - allow readahead allocations to fail when low on memory (5383)
4673 - mmap locking landed again, after considerable improvement (2828)
4674 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4676 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4679 - send OST transaction number in read/write reply to free req (4966)
4680 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4681 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4682 - the watchdog thread now runs as interruptible (5246)
4683 - flock/lockf fixes (but it's still disabled, pending 5135)
4684 - don't use EXT3 constants in llite code (5094)
4685 - memory shortage at startup could cause assertion (5176)
4687 - reorganization of lov code
4688 - single portals codebase
4690 - add extents/mballoc support (5025)
4691 - direct I/O reads in the obdfilter (4048)
4692 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4694 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4697 - send OST transaction number in read/write reply to free req (4966)
4698 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4699 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4700 - the watchdog thread now runs as interruptible (5246)
4701 - handle missing objects in filter_preprw_read properly (5265)
4702 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4703 - drop import inflight refcount on signal_completed_replay error (5255)
4704 - unlock page after async write error during send (3677)
4705 - return original error code on reconstructed replies (3761)
4706 - no transno return for symlink open, don't save no-trasno open (3440)
4708 - add pid to ldlm debugging output (4922)
4709 - bump the watchdog timeouts -- we can't handle 30sec yet
4710 - extra debugging for orphan dentry/inode bug (5259)
4712 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4715 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4716 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4717 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4718 - fix NULL dereference in /proc/sys/portals/routes (4827)
4719 - allow failed mdc_close() operations to be interrupted (4561)
4720 - stop precreate on OST before MDS would time out on it (4778)
4721 - don't send partial-page writes before EOF from client (4410)
4722 - discard client grant for sub-page writes on large-page clients (4520)
4723 - don't free dentries not owned by NFS code, check generation (4806)
4724 - fix lsm leak if mds_create_objects() fails (4801)
4725 - limit debug_daemon file size, always print CERROR messages (4789)
4726 - use transno after validating reply (3892)
4727 - process timed out requests if import state changes (3754)
4728 - update mtime on OST during writes, return in glimpse (4829)
4729 - add mkfsoptions to LDAP (4679)
4730 - use ->max_readahead method instead of zapping global ra (5039)
4731 - don't interrupt __l_wait_event() during strace
4733 - add software watchdogs to catch hung threads quickly (4941)
4734 - make lustrefs init script start after nfs is mounted
4735 - fix CWARN/ERROR duplication (4930)
4736 - return async write errors to application if possible (2248)
4737 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4738 - print NAL number in %x format (4645)
4739 - update barely-supported suse-2.4.21-171 series (4842)
4740 - support for sles 9 %post scripts
4741 - support for building 2.6 kernel-source packages
4742 - support for sles km_* packages
4744 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4747 - ignore -ENOENT errors in osc_destroy (3639)
4748 - notify osc create thread that OSC is being cleaned up (4600)
4749 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4750 - reconstruct ost_handle() like mds_handle() (4657)
4751 - create a new thread to do import eviction to avoid deadlock (3969)
4752 - let lconf resolve symlinked-to devices (4629)
4753 - don't unlink "objects" from directory with default EA (4554)
4754 - hold socknal file ref over connect in case target is down (4394)
4755 - allow more than 32000 subdirectories in a single directory (3244)
4756 - fix blocks count for O_DIRECT writes (3751)
4757 - OST returns ENOSPC from object create when no space left (4539)
4758 - don't send truncate RPC if file size isn't changing (4410)
4759 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4760 - bind to privileged port in socknal and tcpnal (3689)
4762 - rate limit CERROR/CWARN console message to avoid overload (4519)
4763 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4764 - basic mmap support (3918)
4765 - kernel patch series update from b1_4 (4711)
4767 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4770 - avoid crash during MDS cleanup with OST shut down (2775)
4771 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4772 - don't use bad inodes on the MDS (3744)
4773 - dynamic object preallocation to improve recovery speed (4236)
4774 - don't hold spinlock over lock dumping or change debug flags (4401)
4775 - don't zero obd_dev when it is force cleaned (3651)
4776 - print grants to console if they go negative (4431)
4777 - "lctl deactivate" will stop automatic recovery attempts (3406)
4778 - look for existing locks in ldlm_handle_enqueue() (3764)
4779 - don't resolve lock handle twice in recovery avoiding race (4401)
4780 - revalidate should check working dir is a directory (4134)
4782 - don't always mark "slow" obdfilter messages as errors (4418)
4784 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4787 - don't close LustreDB during write_conf until it is done (3860)
4788 - fix typo in lconf for_each_profile (3821)
4789 - allow dumping logs from multiple threads at one time (3820)
4790 - don't allow multiple threads in OSC recovery (3812)
4791 - fix debug_size parameters (3864)
4792 - fix mds_postrecov to initialize import for llog ctxt (3121)
4793 - replace config semaphore with spinlock (3306)
4794 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4795 - don't allow enqueue to complete on a destroyed export (3822)
4796 - down write_lock before checking llog header bitmap (3825)
4797 - recover from lock replay timeout (3764)
4798 - up llog sem before sending rpc (3652)
4799 - reduce ns lock hold times when setting kms (3267)
4800 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4801 - fix NULL deref and obd_dev leak on setup error (3312)
4802 - replace some LBUG about llog ops with error handling (3841)
4803 - don't match INVALID dentries from d_lookup and spin (3784)
4804 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4805 - fix invalid assertion in ptlrpc_set_wait (3880)
4807 - add libwrap support for the TCP acceptor (3996)
4808 - add /proc/sys/portals/routes for non-root route listing (3994)
4809 - allow setting MDS UUID in .xml (2580)
4810 - print the stack of a process that LBUGs (4228)
4812 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4815 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4816 - make sure to unset replay flag from failed open requests (3440)
4817 - if default stripe count is 0, use OST count for inode size (3636)
4818 - update parent mtime/ctime on client for create/unlink (2611)
4819 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4820 - free recovery state on server during a forced cleanup (3571)
4821 - unregister_reply for resent reqs (3063)
4822 - loop back devices mounting and status check on 2.6 (3563)
4823 - fix resource-creation race that can provoke i_size == 0 (3513)
4824 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4825 - more debugging for page-accounting assertion (3746)
4826 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4827 - avoid deadlock after precreation failure (3758)
4828 - fix race and lock order deadlock in orphan handling (3450, 3750)
4829 - add validity checks when grabbing inodes from l_ast_data (3599)
4831 - add /proc/.../recovery_status to obdfilter (3428)
4832 - lightweight CDEBUG infrastructure, debug daemon (3668)
4833 - change default OSC RPC parameters to be better on small clusters
4834 - turn off OST read cache for files smaller than 32MB
4835 - install man pages and include them in rpms (3100)
4836 - add new init script for (un)mounting lustre filesystems (2593)
4837 - run chkconfig in %post for init scripts (3701)
4838 - drop scimac NAL (unmaintained)
4840 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4843 - clean kiobufs before and after use (3485)
4844 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4845 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4846 - change init script to fail more clearly if not run as root (1528)
4847 - allow clients to reconnect during replay (1742)
4848 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4849 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4850 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4851 - fix cond_resched (3554)
4852 - only evict unfinished clients after recovery (3515)
4853 - allow bulk resend, prevent data loss (3570)
4854 - dynamic ptlrpc request buffer allocation (2102)
4855 - don't allow unlinking open directory if it isn't empty (2904)
4856 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4857 - remove extraneous obd dereference causing LASSERT failure (3334)
4858 - don't use get_cycles() when creating temp. files on the mds (3156)
4859 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4860 - handle EEXIST for set-stripe, set proper directory name (3336)
4862 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4863 - fix ksocknal_fmb_callback() error messages (2918)
4865 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4868 - don't copy lvb into (possibly NULL) reply on error (2983)
4869 - don't deref dentry after dput, don't free lvb on error (2922)
4870 - use the kms to determine writeback rpc length (2947)
4871 - increment oti_logcookies when osc is inactive (2948)
4872 - update client's i_blocks count via lvb messages (2543)
4873 - handle intent open/close of special files properly (1557)
4874 - mount MDS with errors=remount-ro, like obdfilter (2009)
4875 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4876 - don't use cancelling-locks' kms values (2947)
4877 - use highest lock extent for kms, not last one (2925)
4878 - don't dereference ERR_PTR() dentry in error handling path (3107)
4879 - fix thread race in portals_debug_dumplog() (3122)
4880 - create lprocfs device entries at setup instead of at attach (1519)
4881 - common AST error handler, don't evict client on completion race (3145)
4882 - zero nameidata in detach_mnt in 2.6 (3118)
4883 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4884 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4885 - initialize RPC timeout timer earlier for 2.6 (3219)
4886 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4887 - print nal/nid for unknown nid (3258)
4888 - additional checks for oscc recovery before doing precreate (3284)
4889 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4890 - don't crash in mdc_close for bad permissions on open (3285)
4891 - zero i_rdev for non-device files (3147)
4892 - clear page->private before handing to FS, better assertion (3119)
4893 - tune the read pipeline (3236)
4894 - fix incorrect decref of invalidated dentry (2350)
4895 - provide read-ahead stats and refine rpc in flight stats (3328)
4896 - don't hold journal transaction open across create RPC (3313)
4897 - update atime on MDS at close time (3265)
4898 - close LDAP connection when recovering to avoid server load (3315)
4899 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4900 - don't leak open file on MDS after open resend (3325)
4901 - serialize filter_precreate and filter_destroy_precreated (3329)
4902 - loop device shouldn't call sync_dev() for nul device (3092)
4903 - clear page cache after eviction (2766)
4904 - resynchronize MDS->OST in background (2824)
4905 - refuse to mount the same filesystem twice on same mountpoint (3394)
4906 - allow llmount to create routes for mounting behind routers (3320)
4907 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4908 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4909 - fix rare problem with rename on htree directories (3417)
4911 - allow default OST striping configuration per directory (1414)
4912 - fix compilation for qswnal for 2.6 kernels (3125)
4913 - increase maximum number of MDS request buffers for large systems
4914 - change liblustreapi to be useful for external progs like lfsck (3098)
4915 - increase local configuration timeout for slow disks (3353)
4916 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4918 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4921 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4922 - don't overwrite extent policy data in reply if lock was blocked (2901)
4923 - drop filter export grants atomically with removal from device (2663)
4924 - del obd_self_export from work_list in class_disconnect_exports (2908)
4925 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4926 - swab reply message in mdc_close, other PPC fixes (2464)
4927 - fix destroying of named logs (2325)
4928 - overwrite old logs when running lconf --write_conf (2264)
4929 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4930 - fix race in target_handle_connect (2898)
4931 - mds_reint_create() should take same inode create lock (2926)
4932 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4933 - don't close files for self_export to avoid uninitialized obd (2936)
4934 - allow MDS with the same name as client node (2939)
4935 - hold dentry reference for closed log files for unlink (2325)
4936 - reserve space for all logs during transactions (2059)
4937 - don't evict page beyond end of stripe extent (2925)
4938 - don't oops on a deleted current working directory (2399)
4939 - handle hard links to targets without a parent properly (2517)
4940 - don't dereference NULL lock when racing during eviction (2867)
4941 - don't grow lock extents when lots of conflicting locks (2919)
4943 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4946 - account for cache space usage on clients to avoid data loss (974)
4947 - lfsck support in lustre kernel code (2349)
4948 - reduce journal credits needed for BRW writes (2370)
4949 - orphan handling to avoid losing space on client/server crashes
4950 - ptlrpcd can be blocked, stopping ALL progress (2477)
4951 - use lock value blocks to assist in proper KMS, faster stat (1021)
4952 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4953 - recovery for initial connections (2355)
4954 - fixes for mds_cleanup_orphans (1934)
4955 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4956 - block all file creations until orphan recovery completes (1901)
4957 - client remove rq_connection from request struct (2423)
4958 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4959 - recovery timer race (2670)
4960 - mdc_close recovey bug (2532)
4961 - ptlrpc cleanup bug (2710)
4962 - mds timeout on local locks (2588)
4963 - namespace lock held during RPCs (2431)
4964 - handle interrupted sync write properly (2503)
4965 - don't try to handle a message that hasn't been replied to (2699)
4966 - client assert failure during cleanup after abort recovery (2701)
4967 - leak mdc device after failed mount (2712)
4968 - ptlrpc_check_set allows timedout requests to complete (2714)
4969 - wait for inflight reqs when ptlrpcd finishes (2710)
4970 - make sure unregistered services are removed from the srv_list
4971 - reset bulk XID's when resending them (caught by 1138 test)
4972 - unregister_bulk after timeout
4973 - fix lconf error (2694)
4974 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4975 - readahead locks pages, leaves pending causing memory pressure (2673)
4976 - increase OST request buffers to 4096 on large machines (2729)
4977 - fix up permission of existing directories in simple_mkdir (2661)
4978 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4979 - don't assign transno to errored transactions (2742)
4980 - don't delete objects on OST if given a bogus objid from MDS (2751)
4981 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4982 - if rq_no_resend, then timeout request after recovery (2432)
4983 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4984 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4985 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4986 - MDS recovery shouldn't delete objects that it has given out (2730)
4987 - if enqueue arrives after completion, don't clobber LVB (2819)
4988 - don't unlock pages twice when trigger_group_io returns error (2814)
4989 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4990 - don't write pages to disk if there was an error (1450)
4991 - don't ping imports that have recovery disabled (2676)
4992 - take buffered bytes into account when balancing socknal conn (2817)
4993 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4994 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4995 - remove little-endian swabbing of llog records (1987)
4996 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4997 - retry reposting request buffers if they fail (1191)
4998 - grow extent at grant time to avoid granting a revoked lock (2809)
4999 - lock revoke doesn't evict page if covered by a second lock (2765)
5000 - disable VM readahead to avoid reading outside lock extents (2805)
5002 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5003 - updated kernel patches for hp-2.4.20 kernel (2681)
5005 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5008 - fix truncated write corruption (2366)
5009 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5011 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5012 - ensure len(uuid) < 37 in lmc (1171)
5013 - fix ia64 OOPS in llog_test (2255)
5014 - zero end of page at obdfilter for partial page writes (2648)
5015 - don't leave stale dentries around after renames (bug 2428)
5016 - fix timeouts when evicting a client with a single lock held (2642)
5017 - set deadline for the initial HELLO message to drain (2634)
5018 - print out dotted-quad IP addresses in the socknal (2302)
5020 - additional debugging for MDS client eviction problem (2443)
5021 - fix mkfsoptions support for osts (2603, 2604)
5023 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5026 - add series for the vanilla 2.6.0 kernel
5027 - add series for the vanilla 2.4.24 kernel
5028 - add series for a cray x86/64 UL kernel drop
5029 - fix xattr patches for the vanilla 2.4.19 series
5031 - generate true UUIDs in lmc (1171)
5032 - have portals stack dumping break in UML (2466)
5033 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5034 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5035 - wake callback waiting threads on client eviction (2460)
5036 - Add --ptldebug and --subsystem to lmc (1719)
5037 - update assertion to allow safe interrupt allocation
5038 - set rq_no_resend for cancel requests (2432)
5039 - recalculate ptlrpcd timeout after resend (2494)
5040 - call vfs_rmdir when removing pending directories (2368)
5041 - fix renaming a file to itself (2429)
5042 - lmc creates a default one-stripe lov (2454)
5043 - expand procfs space to handle large clusters (2326)
5044 - increase UML stack to avoid overflow
5045 - update lconf's list of debug and subsystem masks
5046 - fix lfs find --obd (2510)
5047 - /proc tunable for disabling filter read caching (2591)
5048 - stop rpm packages from altering slapd.conf (2301)
5049 - disable nagle in the socknal under 0conf (2578)
5050 - choose mds inode size based on stripe count (2572)
5051 - fix kernel-source rpm problems (2516)
5053 - add --disable-doc to avoid pdf generation (2421)
5054 - update documentation, tests, type-os, comments
5055 - avoid format warnings on ia64
5056 - remove the TOE NAL
5057 - tiny code cleanups by removing unused fields
5059 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5062 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5063 - remove the most blatant lies from BUILDING file (2371)
5064 - change default debug level to reasonable production setting
5065 - reduce client side cache size to reduce cache flush time
5066 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5067 - make TCP zerocopy and pinger support enabled by default (2476)
5068 - sync writes completed after process exits caused crashes (2319)
5069 - maintain correct mount count on the MDS (2356)
5070 - backout 1557, because 2316 wasn't really fixed
5071 - better file I/O statistics gathering in /proc
5072 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5073 - another kernel patch to fix zero-copy TCP function export
5074 - don't take duplicate lock when processing re-sent getattr (2420)
5075 - lctl uses obd_self_export instead of creating new conn (2353)
5076 - MDS/OST recovery case which requires object creation asserted (2425)
5077 - move lfs from /usr/sbin to /usr/bin in packages
5078 - fix race between mds_client_add and mds_client_free (2417)
5079 - use kmalloc instead of slabs in portals (2430)
5080 - don't create duplicate records when a failover MDS is present (2442)
5081 - remove unnecessary mount age check (2332)
5082 - don't remove directory inodes from locks prematurely (2451)
5083 - don't break if MDS service name is the same as hostname (2103)
5084 - fix races in client write RPC generation when cache full (2482)
5086 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5089 - remove now-unused request->rq_obd (278)
5090 - if an allocation fails, print out how much memory we've used (1933)
5091 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5092 - add the "configurable stack size" patch to most series files (1256)
5093 - ability to write large log records, for 100+ OST configs (2306)
5094 - fix NULL deref when filter_prep fails (2314)
5095 - fix operator precedence error in filter_sync
5096 - dynamic allocation of socknal TX descriptors (2315)
5097 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5098 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5099 - ensure that CWARN messages in Portals always get to the syslog
5100 - __init/__exit are not for prototype decls (ldlm_init/exit)
5101 - x86-64 compile warning fixes
5102 - fix gateway LMC keyword conflict (2318)
5103 - fix MDS lock inversions in getattr/reint paths (1844)
5104 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5105 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5106 - DLM race condition prevented some lock evictions (2328)
5107 - ENOMEM detection and retry on socknal sends (2230)
5108 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5109 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5110 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5111 - better error messages when a client is rejected during recovery (1505)
5112 - avoid cancelling locks which were never granted, after failure (2330)
5113 - fix i_sem/journal inversion in mds_client_add (2333)
5114 - fix truncate/getattr lock cycle deadlock (2334)
5115 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5116 - fix two rare exit paths which could leak an l_lock() ref (2321)
5117 - fencepost error in MDS/OST orphan recovery (2226)
5118 - make log record alignment 8 bytes (1988)
5119 - lstripe now fails when requested offset > ost_count (2237)
5120 - ensure that all kernel series have a complete list.h (1607)
5121 - fix crashes in special-file operations (2316)
5122 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5124 - allow configurable automake binary, for testing new versions
5125 - small update to the lfs documentation
5127 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5129 * fix negative export reference count in fsfilt_sync (2312)
5131 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5132 * release candidate 0.9.1
5134 - orphans are moved into the PENDING directory for possible recovery
5135 - replayed opens now open by fid for orphan/rename safety (1042)
5136 - last close of an orphan inode generates a transno (683)
5137 - chdir() and mount() now pin the directory entry (1020)
5138 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5139 - discard very old requests without processing them (1502)
5140 - remove some common, well-understood CERRORs (1505)
5141 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5142 - clear "grant" flags in OST replies until OST grant code lands (1644)
5143 - fix read performance by not clobbering i_blksize on client (1598)
5144 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5145 - make LRU size a /proc tunable, clears locks when reduced (707)
5146 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5147 - server locks take references on exports now (1558)
5148 - build fixes for 2.4.20-rh trees (1663)
5149 - return an error from lov_create if all OSCs are inactive (1751)
5150 - fix import levels when a reconnect happens without a timeout (1597)
5151 - exit early from mds_open if we get a lookup error (1749)
5152 - partial page read at EOF wouldn't wait for disk before sending (1642)
5153 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5154 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5155 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5156 - write/truncate lock inversion (1639)
5157 - Don't auto-load obdclass, portals modules during cleanup (1495)
5158 - fix timestamps from jumping to "now" (1763)
5159 - extra journal assertions (1648)
5160 - add an extra multiunlink test (1771)
5161 - fix read_record/write_record API (1776)
5162 - fix leak of offset_extent, possible incorrect i_size later (1772)
5163 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5164 - Debugging for the kqswnal_get_idle_tx problems (1820)
5165 - Allow recovery to be attempted multiple times (1536)
5166 - Write out MDS last_rcvd file after it is first created (1600)
5167 - Fix tx_descriptor leak in failed transmit situations (1827)
5168 - ext3 journaling fixes for assertion failure after IO error (1871)
5169 - class_export_put() on freed export after completion AST error (1896)
5170 - Fix revalidate looping in VFS (1322)
5171 - Don't access a freed export during MDS_REINT timeout (1521)
5172 - Add open-unlink recovery support on the MDS (1673,1764)
5173 - Return an error if no MDS data was read from last_rcvd (1946)
5174 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5175 - Don't setup a disk device that doesn't match exported UUID (317)
5176 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5177 - avoid committing NULL handle in force close
5178 - local.sh is now a one-stripe LOV configuration
5179 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5180 - fix ext3 htree duplicate directory entry corruption (1516)
5181 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5182 - update to new LOV EA format (2097)
5183 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5185 2003-06-15 Phil Schwan <phil@clusterfs.com>
5188 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5189 - per-import/export recovery handling (958, 931, 959)
5190 - multiple last-rcvd slots, for serving multiple FSes (949)
5191 - connections are again shared between multiple imp/exports (963, 964)
5192 - "umount -f" would hang if any requests needed to be sent (393, 978)
5193 - avoid pinning large req buffer by copying for queued messages (989)
5194 - add "uuid" to "lctl device" command to help upcalls (991)
5195 - "open" RPCs with transnos would confuse recovery counters (1037)
5196 - do proper endian conversion of all wire messages (288, 340, 891)
5197 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5198 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5199 - fix ptlrpc_connection leak in target_handle_connect (1174)
5200 - fix import refcounting bug in OST and MDS cleanup (1134)
5201 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5202 - fix very unlikely obd_types race condition (501)
5203 - remove osc_open hack for echo_client (1187)
5204 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5205 - a failure in read_inode2 leads to deadlock (1139)
5206 - cancel ack-locks as soon as transaction is committed (1072)
5207 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5208 - make sure to commitrw after any preprw to avoid deadlock (1162)
5209 - failing to execute a file in a lustre FS would lock inode (1203)
5210 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5211 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5212 - obd_uuid2tgt would walk off the end of the list (1255)
5213 - on IA64 the osc would give portals incorrect bulk size (1258)
5214 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5215 - fix lock inversion caused by new llite matching code (1282)
5216 - limit the number of dirty pages on a client to 10MB (1286)
5217 - timed out locks were not being corrected cancelled (1289)
5218 - fix O_DIRECT above 4GB on IA-32 (1292)
5219 * major user-visible changes
5220 - fail out/fail over policy now controlled by the upcall (993)
5222 - add OBD_PING to check server availability and failure (954)
5223 - lustre messages are now sent in sending host order (288, 340, 891)
5224 - add eadatalen to MDS getattr reply (340)
5225 - OST read replies may contain second buffer, with per-page status (593)
5227 2003-03-11 Phil Schwan <phil@clusterfs.com>
5230 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5231 - failed open()s could cause deadlock; fixed (867, 869)
5232 - stop cancelling OST locks when files are closed (481)
5233 - overlapping XID spaces caused network corruption (851, 853)
5234 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5235 - setattr_raw added, to do single-RPC, server-side setattrs
5236 - lmc/lconf syntax change for OST UUIDs
5237 - fix crashy race condition between ptlrpc_free_req and osc_close
5238 - don't use request in mdc_enqueue if we hit a timeout (889)
5239 - don't set the inode i_size for regular files from the MDS (896)
5240 - handle out of order completion AST (842)
5241 - don't LBUG if a lock request times out after receiving AST (913)
5242 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5243 - if a bad lock AST arrives, send an error instead of dropping entirely
5244 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5245 - fix leak in bulk IO when only partially completed (899, 900, 926)
5246 - fix O_DIRECT for ia64 (55)
5247 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5248 - C-z after timeout could hang a process forever; fixed (977)
5250 - client-side I/O cache (678, 924, 929, 941, 970)
5252 - READPAGE and SETATTRs which don't take server-side locks get
5255 2003-02-11 Phil Schwan <phil@clusterfs.com>
5258 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5259 - Fix fsfilt_extN_readpage() to read a full page of directory
5260 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5261 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5262 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5263 - Most intent operations were removed, in favour of a new RPC mode
5264 that does a single RPC to the server and bypasses most of the VFS
5265 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5266 - Aggressively cancel local locks on DLM servers
5267 - mds_reint_unlink sends EA to the client if it's the last nlink.
5268 client uses that EA to unlink OST objects.
5269 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5270 - recursive symlinks were fixed (439)
5271 - fixed NULL deref in DEBUG_REQ
5272 - filter_update_lastobjid no longer calls sync, which annoyed extN
5273 - fixed multi-client small-writes to a single file problem (445)
5274 - fixed mtime updates during file writes (607)
5275 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5276 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5277 - fixed Linux OST size reporting problem (444, 656)
5278 - OST now updates object mtime with writes or setattr (607, 619)
5279 - client verifies file size before zeroing page past EOF (445)
5280 - OST now writes last allocated objid to disk with allocation (108)
5281 - LOV on echo now works (409)
5283 - mds_reint_unlink sends a new buffer, with the EA included. this
5284 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5285 set if a regular file was being unlinked, and it was the last link
5286 - use PtlGet from the target for bulk writes (315)
5287 - OST now updates object mtime with writes or setattr (607, 619)
5288 - LDLM now has a grant-time callback to revalidate locked items, if
5290 - Many MDS operations were reorganized to combat race conditions
5292 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5293 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5295 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5298 - Fully reactivate OST imports after reconnection (512, others)
5299 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5300 - More graceful error handling for truncating on dead OST (515)
5301 - Don't error out unless we're actually accessing dead stripes (474)
5302 - Fix garbage sizes when stripes are missing (410)
5303 - LRU counters were broken, causing constant lock purge (433, 432)
5304 - garbage on read from stripes with failed OSTs (441)
5305 - mark OSCs as active before reconnecting during recovery (438)
5306 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5307 - lfind did not preserve OST order in output (443)
5308 - symlinks cause hung clients, incorrect data (439)
5309 - stop dereferencing request after dropping refcount (457)
5310 - don't LASSERT(spin_is_locked) on non-SMP (455)
5311 - fixes for many rename() bugs
5312 - fstat didn't correctly synchronize attributes (399)
5313 - server must handle lock cancellation during blocking AST prep (487)
5314 - bulk descriptors were free()d too soon (511)
5315 - fix paths in lconf, which would load incorrect modules (451, 507)
5316 - fix confusing lconf 'host not found' error message (386)
5317 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5318 - fix race condition in mdc_blocking_ast() for inode access (526)
5319 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5320 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5321 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5322 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5323 - fix lstripe in several ways (kernel side) (527)
5324 - fix request leak in ldlm_cli_enqueue (262)
5325 - incorrect OSC was marked inactive after OST failure
5326 - call mds_fs_cleanup before unmounting filesystem (524)
5327 - fix races between taking ns_lock and ldlm_lock_change_resource
5328 - fix races updating LOV export open file list
5329 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5330 - fix recovery NULL deref in ldlm_cli_cancel_unused
5331 - fix some DLM races by using new hash table for lock handles (419)
5332 - permit the client to specify desired inodes, at replay
5333 - duplicate requests when we queue them for replay reintegration
5334 - fix last_rcvd offset calculation
5335 - sync after each recovered transaction, so we always make progress
5336 - never, not always, ERESTART requests without transnos
5337 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5339 - skip replay if the MDS says that the client is already connected
5340 - don't check for a recovery-enabled export to match lctl's UUID
5341 - don't INC_USE_COUNT for phantom exports
5342 - don't crash when cleaning up phantom exports (567)
5343 - don't double-finish or set replay data for errored mdc_open requests
5344 - abort requests when they time out, so we don't get old replies
5345 - send/receive replies for AST messages again
5346 - if the client says that it doesn't have the lock, cancel it on the
5348 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5349 mark it as destroyed, it will all work out when decref is called
5350 - fix module use counts (22, 581)
5352 - ASTs now expect a reply (server cancels lock on error reply)
5354 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5357 - fix many simultaneous client startup (392)
5358 - fix dentry->d_it clobbering
5359 - credentials weren't being shipped for readdir/getattr operations
5360 - remove invalid assertions triggered during some concurrent MD
5362 - proper Lustre versions added (336, 389)
5363 - fix memory leak for create error case (398)
5364 - fix LOV locking bug that would get cli/srv out of sync
5365 - fix echo client over LOV (409)
5366 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5367 - fix double-O_EXCL intent crash (424)
5368 - avoid sending multiple lock CANCELs (352)
5370 - MDS can do multi-client recovery (modulo bugs in new code)
5372 - many updates, edits, cleanups
5374 2002-11-18 Phil Schwan <phil@clusterfs.com>
5377 - fix null d_it dereference (346)
5378 - fix full OST/dbench hang (333)
5379 - fix permission problem with file removal (286)
5380 - fix removal of OSCs from LOV when they fail
5381 - fix NULL deref during bulk timeout (214)
5382 - fix problems related to multiple filesystems on one MDS (241)
5383 - fixed serious subtle metadata locking bugs
5384 - free locks on clients when inodes are removed due to memory
5386 - fix inode pointer in lock data (285)
5387 - partial support for multiple MDS on a single host (241)
5388 - data locks weren't cancelled at clear_inode time (290, 311)
5389 - intent locks could lead to unbounded lock growth (205)
5390 - added a maximum lock count, an LRU list, and a flusher
5391 - fix multiple rename (365)
5392 - properly abstracted the echo client
5393 - OSC locked 1 byte too many; fixed
5394 - rewrote brw callback code:
5395 - fixed recovery bugs related to LOVs (306)
5396 - fixed too-many-pages-in-one-write crash (191)
5397 - fixed (again) crash in sync_io_timeout (214)
5398 - probably fixed callback-related race (385)
5400 - Add capability to MDS protocol
5401 - LDLM cancellations and callbacks on different portals
5403 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5406 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5407 - defer open object destruction to close time (601981, 138)
5408 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5409 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5410 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5411 - don't call obd_setattr in ll_file_release for destroyed objects
5412 * protocol change to lustre_msg: move |version| and add |flags|
5413 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5414 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5415 * added replay of create, unlink, link and rename operations during
5416 MDS failover; recovery should be much more robust now
5417 * remove failed OSCs from LOVs (only lov_create uses this so far)
5418 * the lustre-HOWTO was brought (more) up to date (582544)
5420 2002-10-23 Phil Schwan <phil@clusterfs.com>
5423 - in-use dentries weren't being reused properly (617851)
5424 - prevent multiple LDLM setup (599178)
5425 - fix LOV size calculations for truncate (617853)
5426 - fix client handling of MDS intent errors (POSIX)
5427 - fix permission bug in lovstripe.c test (624321)
5428 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5429 - truncate past end of file could corrupt data
5430 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5431 - a race in recovery could return ETIMEDOUT to apps (623947)
5432 - building outside the source directory was fixed
5433 * the lustre-HOWTO was brought (more) up to date (582544)
5434 * major progress was made on recovery functionality
5436 2002-10-10 Phil Schwan <phil@clusterfs.com>
5439 - recovery deadlock fix
5440 - rm -rf causes LBUG fix (617817)
5441 - file open by multiple tasks fix (618962)
5442 - directory permissions bugs (602707 and 620007)
5443 - journal_stop fixed with locking (611313)
5444 - O_APPEND failures resolved (618273, perhaps 614459)
5445 - lconf PATH fix (619770)
5446 - IA64 build fix (621450)
5447 - RPC buffer sizes scale with amount of memory
5449 2002-10-01 Phil Schwan <phil@clusterfs.com>
5452 - locks would be cancelled without throwing away data pages,
5453 resulting in inconsistent data (605627)
5454 - inode attributes were not always being refreshed (605627, 612449)
5455 - lconf now continues to cleanup after lctl reports an error
5456 - MDS now enforces user permissions (602707)
5457 - lprocfs cleanup fixed, but not yet enabled (614157)
5458 - fixed infinite server hang, should a client not respond to an AST
5459 - avoid going into recovery if user calls readlink() with a buffer
5460 that's too small (613941)
5461 - AST RPCs no longer require replies (614867) -- this may be changed
5462 - don't crash server if client sends an IOV that's too big (611336)
5463 - fixed lock conversion deadlock (611892)
5464 - fixed the following of symlinks (614622)
5465 * recovery: the server can remove locks from a client that dies, other
5466 clients can make progress
5467 * more extN patch fixes
5468 * compile-time configurable ptlrpc buffer allocations
5470 - collaborative read cache document
5471 - Lustre Lite Performance CDR document-in-progress
5473 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5476 - fix typo in patch-2.4.18
5478 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5481 - clear ptlrpc request each time in handle_incoming_request()
5482 - unlink of files now destroys the object on the OST
5484 2002-09-19 Peter Braam <braam@clusterfs.com>
5486 * add hard link support
5487 * change obdfile creation method
5488 * kernel patch changed
5490 2002-09-19 Peter Braam <braam@clusterfs.com>
5493 - stack overflow bug in extN fixed
5495 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5497 * documentation updates
5498 - add man pages for config tools
5499 - update tests/README to describe testing with new config tools
5500 - finish metadata API descriptions
5501 * bug fixes and cleanups
5502 - statfs workaround for 16TB limit
5503 - LOV stripe allocation improved, can stripe on subset of OSTs
5504 - LOV file size/IO offset was wrong for files > 4GB in size
5505 - object EA data was being dropped, caused files to be unreadable
5506 - memory overflow with non-LOV OST caused memory corruption
5507 - fixed regression tests to work with new config tools, obdfilter
5508 - fixed bug when directory size became larger than 1 block
5509 - fixed bug (for single client case) when PWD was deleted
5510 - invalidate local directory pages when doing intent-based ops
5511 - avoid LDLM oops when lock callback contained bad data
5513 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5515 * documentation updates
5516 * bug fixes and cleanups
5517 - configuration tools
5520 - 64-bit compile warnings
5521 - 64-bit internal statfs data
5523 * test_brw on persistent OST devices
5525 * lprocfs (disabled)
5527 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5529 * documentation updates
5530 * bug fixes and cleanups
5531 * configuration tools
5533 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5536 * many small fixes to 0_5_4
5537 * io/network handling
5538 * thinkos in MDS operations
5540 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5543 * crucial basic fixes to 0.5.3
5544 * IOR, Iozone work over Elan
5547 2002-08-07 Phil Schwan <phil@clusterfs.com>
5548 * version 0_5_3, our first alpha
5549 * we use the new Portals iovs
5550 * documentation updates
5551 * bug fixes and cleanups
5552 * small changes in the DLM wire protocol
5554 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5555 * version 0_5_1 with some initial stability,
5556 * locking on MD and file I/O.
5557 * documentation updates
5558 * several bug fixes since 0.5.0
5559 * small changes in wire protocol
5561 2002-07-18 Phil Schwan <phil@clusterfs.com>
5563 * delivered as Lustre Light Alpha
5564 * fixed a crash after handling invalid MDS requests
5565 * fixed directory pages for architectures with non-4k pages sizes
5567 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5568 * release version v0_4_4
5569 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5571 * Fixes a number of interruption problems with OST operations.
5572 * Update documentation for portals header changes
5573 * Move all wire protocol structs/defines to lustre_idl.h
5574 * Fixes symlink length bug.
5575 * Add tcpdump to repository.
5577 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5578 * release version v0_4_3
5579 * Fixes statfs for inodes on extN.
5580 * Fixes bug in runtests which would delete /etc/hosts.
5581 * Use 64-bit object IDs wherever possible (not into VFS though)
5582 Remove ost_get_info, which is unused by lustre, and out of date.
5584 2002-07-03 Peter Braam <braam@clusterfs.com>
5585 * release version v0_4_2 Fixes a lookup error (type not passed)
5586 * move forward to head of Portals
5587 * move forward to latest Lustre kernel
5589 2002-06-25 Peter Braam <braam@clusterfs.com>
5590 * release version v0_4_1. Hopefully stable on single node use.