1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.27 (SLES 10),
5 2.6.18-92.1.10.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.40.11-sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
14 * File join has been disabled in this release, refer to Bugzilla 16929.
16 Severity : enhancement
18 Description: Add /proc entry for import status
19 Details : The mdc, osc, and mgc import directories now have
20 an import directory that contains useful import data for debugging
23 Severity : enhancement
25 Description: Re-disable certain /proc logging
26 Details : Enable and disable client's offset_stats, extents_stats and
27 extents_stats_per_process stats logging on the fly.
30 Frequency : Only on FC kernels 2.6.22+
32 Description: oops in statahead
33 Details : Do not drop reference count for the dentry from VFS when lookup,
34 VFS will do that by itself.
36 Severity : enhancement
38 Description: Generic /proc file permissions
39 Details : Set /Proc file permissions in a more generic way to enable non-
40 root users operate on some /proc files.
44 Description: Hitting mdc_commit_close() ASSERTION
45 Details : Properly handle request reference release in
46 ll_release_openhandle().
50 Frequency : only patchless client
51 Description: add workaround for race between add/remove dentry from hash
53 Severity : enhancement
55 Description: Allow OST glimpses to return PW locks
59 Description: LBUG when llog conf file is full
60 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
64 Description: Prevent import from entering FULL state when server in recovery
68 Description: service mount cannot take device name with ":"
69 Details : Only when device name contains ":/" will mount treat it as
75 Description: replace ptlrpcd with the statahead thread to interpret the async
76 statahead RPC callback
80 Frequency : on recovery
81 Description: I/O failures after umount during fail back
82 Details : if client reconnected to restarted server we need join to recovery
83 instead of find server handler is changed and process self
84 eviction with cancel all locks.
86 Severity : enhancement
88 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
90 Severity : enhancement
92 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
94 Severity : enhancement
96 Description: Upcall on Lustre log has been dumped
97 Details : Allow for a user mode script to be called once a Lustre log has
98 been dumped. It passes the filename of the dumped log to the
99 script, the location of the script can be specified via
100 /proc/sys/lnet/debug_log_upcall.
105 Description: avoid idr_remove called for id which is not allocated.
106 Details : Move assigment s_dev for clustered nfs to end of initialization,
107 to avoid problem with error handling.
112 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
113 Details : When connection is reused this not moved from CONN_UNUSED_HASH
114 into CONN_USED_HASH and this prodice warning when put connection
115 again in unused hash.
117 Severity : enhancement
119 Description: File striping can now be set to use an arbitrary pool of OSTs.
121 Severity : enhancement
123 Description: Export bytes_read/bytes_write count on OSC/OST.
127 Description: Early reply size mismatch, MGC loses connection
128 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
129 the connect flags are properly negotiated.
133 Description: Properly propagate oinfo flags from lov to osc for statfs
134 Details : restore missing copy oi_flags to lov requests.
136 Severity : enhancement
138 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
139 : l_getgroups(8), lst(8), routerstat(8)
140 Details : included man pages for llobdstat(8), llstat(8),
141 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
143 Severity : enhancement
145 Description: configure's --enable-quota should check the
146 : kernel .config for CONFIG_QUOTA
147 Details : configure is terminated if --enable-quota is passed but
148 : no quota support is in kernel
152 Description: Remove portals compatibility
153 Details : Remove portals compatibility, not interoperable with releases
158 Description: Resolve device initialization race
159 Details : Prevent proc handler from accessing devices added to the
160 obd_devs array but yet be intialized.
162 Severity : enhancement
164 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
166 Severity : enhancement
168 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
173 Description: Using wrong pointer in osc_brw_prep_request
174 Details : Access to array[-1] can produce panic if kernel compiled with
175 CONFIG_PAGE_ALLOC enabled
179 Description: Client runs out of low memory
180 Details : Consider only lowmem when counting initial number of llap pages
184 Description: *optional* service tags registration
185 Details : if the "service tags" package is installed on a Lustre node
186 When the filesystem is mounted, a local-node service tag will
187 be created. See http://inventory.sun.com/ for more information
188 about the Service Tags asset management system.
192 Description: Kernel BUG tries to release flock
193 Details : Lustre does not destroy flock lock before last reference goes
194 away. So always drop flock locks when client is evicted and
195 perform unlock regardless of successfulness of speaking to MDS.
199 Description: add refcount protection for osc callbacks, avoid panic on shutdown
203 Description: sanity test 65a fails if stripecount of -1 is set
204 Details : handle -1 striping on filesystem in ll_dirstripe_verify
209 Description: ASSERTION(CheckWriteback(page,cmd)) failed
210 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
213 Severity : enhancement
215 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
219 Description: do not process already freed flock
220 Details : flock can possibly be freed by another thread before it reaches
221 to ldlm_flock_completion_ast.
225 Description: LBUG during stress test
226 Details : Need properly lock accesses the flock deadlock detection list.
230 Description: oops in page fault handler
231 Details : kernel page fault handler can return two special 'pages' in error
232 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
236 Description: timeout with invalidate import.
237 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
238 should be handled by ptlrpcd. This produce long age waiting and
239 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
241 Severity : enhancement
243 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
247 Description: enable MGS and MDT services start separately
248 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
249 a co-located MGS without starting the MGS, which is a complement
250 to 'nosvc' mount option.
254 Frequency : after recovery
255 Description: precreate to many object's after del orphan.
256 Details : del orphan st in oscc last_id == next_id and this triger growing
257 count of precreated objects. Set flag LOW to skip increase count
258 of precreated objects.
262 Frequency : rare, on clear nid stats
263 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
264 Details : when clean nid stats sometimes try destroy live entry,
265 and this produce panic in free.
269 Description: Stack overflow during MDS log replay
270 ease stack pressure by using a thread dealing llog_process.
274 Description: wait until IO finished before start new when do lock cancel.
275 Details : VM protocol want old IO finished before start new, in this case
276 need wait until PG_writeback is cleared until check dirty flag and
277 call writepages in lock cancel callback.
279 Severity : enhancement
281 Description: using special macro for print time and cleanup in includes.
285 Description: mds_mfd_close() ASSERTION(rc == 0)
286 Details : In mds_mfd_close(), we need protect inode's writecount change
287 within its orphan write semaphore to prevent possible races.
291 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
295 Frequency : rare, on shutdown ost
296 Description: don't hit live lock with umount ost.
297 Details : shrink_dcache_parent can be in long loop with destroy dentries,
298 use shrink_dcache_sb instead.
302 Description: don't panic with use echo client
303 Details : echo client pass NULL as client nid pointer and this produce null
308 Description: fix build on ppc32
309 Details : compile code with -m64 flag produce wrong object file for ppc32.
313 Description: add message levels for liblustreapi
317 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
318 Details : Move the 'good_osts' check before the 'total_bavail' check. This
319 will result in an -EAGAIN and in the exit call path we call
320 alloc_rr() which will with increasing aggressiveness attempt to
321 aquire precreated objects on the minimum number of required OSCs.
325 Description: Use old size assignment to avoid deadlock
326 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
327 the scheduling while holding a spinlock. We do not need locking
328 for size in ll_update_inode() because size is only updated from
329 the MDS for directories or files without objects, so there is no
330 other place to do the update, and concurrent access to such inodes
331 are protected by the inode lock.
335 Description: resolve "_IOWR redefined" build error on SLES10
339 Description: dump the memory debugging after all modules are unloaded to
340 suppress false negative in conf_sanity test 39
342 Severity : enhancement
344 Description: build kernel-ib packages for OFED 1.3 in our release cycle
349 Description: fix SLES kernel versioning
350 Details : the kernel version for our SLES 10 kernel did not include a "-"
351 before the "smp" at the end. while this was not a problem in
352 general, it did mean that software trying to use the kernel
353 version to try to detect a vendor specific kernel would fail.
354 this was most evident by the OFED build scripts.
358 Description: Don't update lov_desc members until making sure they are valid
359 Details : When updating lov_desc members via proc fs, need fix their
360 validities before doing the real update.
364 Description: don't put request into delay list while invalidate in flight.
365 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
366 import in flight. this produce timeout for invalidate and sometimes
367 can cause stale data.
371 Frequency : on ppc only
372 Description: not convert ost objects for directory because it's not exist.
373 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
377 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
379 Severity : enhancement
381 Description: Update to support 2.6.22.14 vanilla kernel.
385 Frequency : rare, on recovery
386 Description: read procfs can produce deadlock in some situation
387 Details : Holding lprocfs lock which send rpc can produce block for destroy
388 obd objects and this also block reconnect with -EALREADY. This
389 isn't fix all lprocfs bugs - but make it rare.
391 Severity : enhancement
393 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
396 Frequency : frequent on X2 node
398 Description: mdc_set_open_replay_data LBUG
399 Details : Set replay data for requests that are eligible for replay.
403 Description: lustre_mgs: operation 101 on unconnected MGS
404 Details : When MGC is disconnected from MGS long enough, MGS will evict the
405 MGC, and late on MGC cannot successfully connect to MGS and a lot
406 of the error messages complaining that MGS is not connected.
410 Frequency : on network error
411 Description: panic with double free request if network error
412 Details : mdc_finish_enqueue is finish request if any network error ocuring,
413 but it's true only for synchronus enqueue, for async enqueue
414 (via ptlrpcd) this incorrect and ptlrpcd want finish request
417 Severity : enhancement
419 Description: client-side metadata stat-ahead during readdir(directory readahead)
420 Details : perform client-side metadata stat-ahead when the client detects
421 readdir and sequential stat of dir entries therein
424 Frequency : on start mds
426 Description: Implement get_info(last_id) in obdfilter.
429 Frequency : occasional
431 Description: Correctly check stale fid, not start epoch if ost not support SOM
432 Details : open with flag O_CREATE need set old fid in op_fid3 because
433 op_fid2 was overwritten with new generated fid, but mds can answer
434 with one of these two fids and both is not stale. Setattr
435 incorrectly started an epoch and assume will be called
436 done_writing, but without SOM done_writing ever being called.
439 Frequency : rare, depends on device drivers and load
441 Description: MDS or OSS nodes crash due to stack overflow
442 Details : Code changes in 1.8.0 increased the stack usage of some functions.
443 In some cases, in conjunction with device drivers that use a lot
444 of stack the MDS (or possibly OSS) service threads could overflow
445 the stack. One change which was identified to consume additional
446 stack has been reworked to avoid the extra stack usage.
449 Frequency : occasional
451 Description: Do not fail import if osc_interpret_create gets -EAGAIN
452 Details : If osc_interpret_create got -EAGAIN it immediately exits and
453 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
454 oscc_has_objects and see OSC has no objests and call
455 oscc_internal_create to resend create request.
457 Severity : enhancement
459 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
461 Severity : enhancement
463 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
466 Frequency : very rare
468 Description: Fix replay if there is an un-replied request and open
469 Details : In some cases, older replay request will revert the
470 mcd->mcd_last_xid on MDS which is used to record the client's
473 Severity : enhancement
475 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
477 Severity : enhancement
479 Description: Add rhel5 support to HEAD.
481 Serverity : enhancement
483 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
488 Description: Don't allow skipping OSTs if index has been specified.
489 Details : Don't allow skipping OSTs if index has been specified, make
490 locking in internal create lots better.
494 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
495 Details : access to bitfield in structure is always rounded to long
496 and this produce problem with not atomic change any bit.
500 Description: Lustre make rpms failed.
501 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
502 builds Lustre rpms from tarball.
505 Frequency : rare on shutdown ost
507 Description: If llog cancel was not send before clean_exports phase, this can
508 produce deadlock in llog code.
509 Details : If llog thread has last reference to obd and call class_import_put
510 this produce deadlock because llog_cleanup_commit_master wait when
511 last llog_commit_thread exited, but this never success because was
512 called from llog_commit_thread.
516 Description: allow userland application know is lost one of stripes.
517 Details : fill lvb_blocks with error code on ost and return it to
518 application if error flag found.
522 Description: NULL lov_tgts causing MDS oops
523 Details : more safe checks for NULL lov_tgts for avoid oops.
525 Severity : enhancement
527 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
531 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
532 Details : If an OST is down the MDS will hang indefinitely in
533 obd_statfs_rqset() waiting for the statfs data. While for
534 MDS QOS usage of statfs, it should not stuck in waiting.
536 Severity : enhancement
538 Description: remote_acl support
539 Details : Support ACL-based permission check for remote user.
540 Support setfacl/getfacl for remote user with the utils
541 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
542 the system "{s,g}etfacl" utils.
544 Severity : enhancement
546 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
548 Severity : enhancement
550 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
554 Description: make mgs_setparam() handle fsname containing dash
555 Details : fsname containing a dash does not work with lctl conf_param
557 Severity : enhancement
559 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
561 Severity : enhancement
563 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
567 Frequency : rare, at shutdown
568 Description: access already free / zero obd_namespace.
569 Details : if client_disconnect_export was called without force flag set,
570 and exist connect request in flight, this can produce access to
571 NULL pointer (or already free pointer) when connect_interpret
572 store ocd flags in obd_namespace.
576 Frequency : only at startup
577 Description: not alloc memory with spinlock held.
578 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
579 if any spinlock held.
581 Severity : enhancement
583 Description: make lustre randomly fail allocating memory
584 Details : Make lustre randomly failed allocating memory for testing purpose.
586 Severity : enhancement
588 Description: lost problems with lov objid file
589 Details : Fixes some scability and access to not inited memory problems
590 in work with lov objdid file.
595 Description: lfs find does not continue on file error
596 Details : Continue other files processing when a file/dir is absent.
600 Description: Inconsistent usage of lustre_pack_reply()
601 Details : Standardize the usage of lustre_pack_reply() such that it
602 always generate a CERROR on failure.
605 Frequency : occasional
607 Description: MDS hang or stay in waiting lock
608 Details : If client receive lock with CBPENDING flag ldlm need send lock
609 cancel as separate rpc, to avoid situation when cancel request
610 can't processed due all i/o threads stay in wait lock.
614 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
618 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
622 Description: Update to SLES9 kernel-2.6.5-7.287.3.
626 Description: Update to RHEL4 latest kernel.
628 Severity : enhancement
630 Description: Build SLES10 patchless client fails
631 Details : The configure was broken by run ./configure with
632 --with-linux-obj=.... argument for patchless client. When the
633 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
634 can't search header adequately. Use absolute path such as
635 -I($LINUX)/include instead.
639 Description: interrupt oig_wait produce painc on resend.
640 Details : brw_redo_request can be used for resend requests from ptlrpcd and
641 private set, and this produce situation when rq_ptlrpcd_data not
642 copyed to new allocated request and triggered LBUG on assert
643 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
644 ptlrpcd set if request is changed and can be safety changed to use
649 Description: LASSERT_{REQ,REP}SWAB macros are buggy
650 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
651 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
652 these macros and replace them with their difinitions instead.
656 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
657 Details : Update which_patch & target file for SLES10 latest kernel.
660 Description: Remove client patches from SLES 10 kernel.
661 Details : This causes SLES 10 clients to behave as patchless clients
662 even on a Lustre-patched (server) kernel.
664 Severity : enhancement
666 Description: self-adjustable client's lru lists
667 Details : use adaptive algorithm for managing client cached locks lru
668 lists according to current server load, other client's work
669 pattern, memory activities, etc. Both, server and client
670 side namespaces provide number of proc tunables for controlling
673 Severity : enhancement
675 Description: light-weight GSS support
676 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
677 and reduce performance overhead.
679 Severity : enhancement
681 Description: Linux keyring support
682 Details : Support using service of Linux keyring for Lustre GSS internal
683 context refresh/cache mechanism.
687 Description: Fix errors in lfs documentation
688 Details : Fixes man pages
692 Description: don't use GFP_* in generic Lustre code.
693 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
697 Description: obdclass is limited by single OBD_ALLOC(idarray)
698 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
702 Description: warnings with build patchless client with vanila 2.6.19 and up
703 Details : change old ctl_table style and replace ctl_table/ctl_table_header
704 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
708 Frequency : Only for SLES
709 Description: sanity_quota fail test_1
710 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
711 file, which makes TSTID[2] inunique.
715 Frequency : Only for SLES9
716 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
717 Details : Update target/ChangeLog/which_patch .
721 Frequency : Only for test_10 in sanity-quota.sh
722 Description: a bug in quota test code
723 Details : checking a test flag in a wrong place
727 Frequency : Only for test_13 in sanity-quota.sh
728 Description: a bug in quota test script
729 Details : shouldn't check overfull allocated quota
733 Description: jbd statistics
734 Details : Port older jbd statistics patch for sles10
736 Severity : enhancement
738 Description: Add printing inode info into message about error in writepage.
743 Description: reply_lock_interpret crash due to race with it and lock cancel.
744 Details : Do not replay locks that are being cancelled. Do not reference
745 locks by their address during replay, just by their handle.
749 Frequency : When flocks are used.
750 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
751 Details : Flock locks might destroy just granted lock if it could be merged
752 with another existing flock, this is done in completion handler,
753 so teach ldlm_cli_enquque_fini that this is a valid case for
759 Description: Oops in read and write path when failing to allocate lock.
760 Details : Check if lock allocation failed and return error back.
764 Description: lstripe command fails for valid OST index
765 Details : The stripe offset is compared to lov->desc.ld_tgt_count
766 instead of lov->desc.ld_active_tgt_count.
770 Description: sanity.sh failed test 103
771 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
776 Description: df doesn't work properly if diskfs blocksize != 4K
777 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
781 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
782 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
784 Severity : enhancement
786 Description: Adaptive timeouts
787 Details : RPC timeouts adapt to changing server load and network
788 conditions to reduce resend attempts and improve recovery time.
792 Description: llapi_file_create() does not allow some changes
793 Details : add llapi_file_open() that allows specifying the mode and
794 open flags, and also returns an open file handle.
798 Description: merge and cleanup kernel patches.
799 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
803 Description: Add journal checksum support.(Kernel part)
804 Details : The journal checksum feature adds two new flags i.e
805 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
806 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
807 indicates that the commit block contains the checksum for
808 the blocks described by the descriptor blocks. Now commit
809 record can be sent to disk without waiting for descriptor
810 blocks to be written to disk. This behavior is controlled
811 using JBD2_FEATURE_ASYNC_COMMIT flag.
815 Description: OSS needs mutliple precreate threads
816 Details : Add ability to start more than one create thread per OSS.
820 Description: Sanity.sh test_65k failed: setstripe should have succeeded
821 Details : execute lfs setstripe on client
825 Description: mds_obd_create error creating tmp object
826 Details : When the user sets quota on root, llog will be affected and can't
827 create files and write files.
830 Frequency : Always on ia64 patchless client, and possibly others.
832 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
833 Details : This allows the patchless client to be loaded on architectures
838 Description: RedHat Update kernel for RHEL5
839 Details : Add the kernel config file for RHEL5.
843 Description: RedHat Update kernel for RHEL5
844 Details : Modify the kernel config file more closer RHEL5.
848 Description: Build failure against Centos5 (RHEL5)
849 Details : Define PAGE_SIZE when it isn't present.
853 Description: test_7 on sanity-quota.sh can't be used on separate machines
854 Details : add facet to handle it
858 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
859 Details : with NFS, the anon dentry's parent was set to itself in
860 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
861 corresponding dentry other than use rec->ur_name.
863 Severity : enhancement
865 Description: lfs setstripe enhancement
866 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
870 Description: enable data checksumming by default
871 Details : enable checksum by default, allow --disable-checksum
872 configure option and "-o nochecksum" mount option. Checksums
873 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
877 Description: lustre support for RHEL5
878 Details : Add support for RHEL5.
882 Description: Client eviction due to failover config
883 Details : after a connection loss, the lustre client should attempt to
884 reconnect to the last active server first before trying the
885 other potential connections.
889 Description: when mds and osts use different quota unit(32bit and 64bit),
890 quota will be released repeatly.
891 Details : void sending multiple quota reqs to mds, which will keep the
892 status between the reqs.
896 Description: osts not allocated evenly to files
897 Details : change the condition to increase offset_idx
900 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
902 Description: Data corruption for OSTs that are formatted larger than 2TB
904 Details : When generating the bio request for lustre file writes the
905 sector number would overflow a temporary variable before being
906 used for the IO. The data reads correctly from Lustre (which
907 will overflow in a similar manner) but other file data or
908 filesystem metadata may be corrupted in some cases.
912 Description: Tune the kernel for good SCSI performance.
913 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
914 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
919 Description: rewrite ext2-derived code in obdclass/uuid.c
920 Details : rewrite inherited code (uuid parsing code from ext2 utils)
921 from scratch preserving functionality.
925 Description: "lfs find -obd UUID" prints directories
926 Details : "lfs find -obd UUID" will return all directory names instead
927 of just file names. It is incorrect because the directories
928 do not reside on the OSTs.
932 Description: use i_size_read and i_size_write in 2.6 port
933 Details : replace inode->i_size access with i_size_read/write()
936 Frequency : when removing large files
938 Description: scheduling issue during removal of large Lustre files
939 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
940 It causes scheduling issues when removing large files (17TB in the
944 Frequency : only with liblustre clients on XT3
946 Description: evictions taking too long
947 Details : allow llrd to evict clients directly on OSTs
950 Frequency : only on ppc
952 Description: /proc/fs/lustre/devices broken on ppc
953 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
954 We should make sure the type of 'index' is loff_t and then cast
955 explicitly as needed below. Do not assign an explicitly cast
959 Frequency : only for rhel5
961 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
962 Details : Modify the target file & which_kernel.
964 Severity : enhancement
966 Description: omit set fsid for export NFS
967 Details : fix set/restore device id for avoid EMFILE error and mark lustre
968 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
972 Frequency : Always, for kernels after 2.6.16
973 Description: Fix warning idr_remove called for id=.. which is not allocated.
974 Details : Last kernels save old s_dev before kill super and not allow
975 to restore from callback - restore it before call kill_anon_super.
979 Description: buffer overruns could theoretically occur
980 Details : llapi_semantic_traverse() modifies the "path" argument by
981 appending values to the end of the origin string, and a
982 overrun may occur. Adding buffer overrun check in liblustreapi.
986 Description: Fix error on 'ls .' at the top of the Lustre mount.
987 Details : Don't revalidate dentry if it is a root dentry.
991 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
992 Details : Modify vm-tunables-rhel4.patch.
996 Description: Kernel config for 2.6.18-vanilla.
997 Details : Modify targets/2.6-vanilla.target.in.
998 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
999 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1000 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1001 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1005 Description: improve handling recoverable errors
1006 Details : if request processig with error which can be recoverable on server
1007 request should be resend, otherwise page released from cache and
1012 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1013 Details : Modify target file & which_patch.
1014 A flaw was found in the IA32 system call emulation provided
1015 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1016 value could be stored in the %RAX register, which could trigger an
1017 out-of-bounds system call table access. An untrusted local user
1018 could exploit this flaw to run code in the kernel
1019 (ie a root privilege escalation). (CVE-2007-4573).
1023 Description: O_DIRECT bypasses client statistics.
1024 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1025 read_ahead_stats were updated. copied the stats section from
1026 osc_send_oap_rpc() into async_internal().
1030 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1034 Description: change order of libsysio includes
1035 Details : '#include sysio.h' should always come before '#include xtio.h'
1039 Description: handle "serious error: objid * already exists" more gracefully
1040 Details : If LAST_ID value on disk is smaller than the objects existing in
1041 the O/0/d* directories, it indicates disk corruption and causes an
1042 LBUG(). If the object is 0-length, then we should use the existing
1043 object. This will help to avoid a full fsck in most cases.
1045 Severity : enhancement
1047 Description: adapt the lustre_config script to support the upgrade case
1048 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1049 server targets to 1.6 in parallel.
1053 Description: To avoid grant space > avaible space when the disk is almost
1054 full. Without this patch you might see the error "grant XXXX >
1055 available" or some LBUG about grant, when the disk is almost
1057 Details : In filter_check_grant, for non_grant cache write, we should
1058 check the left space by if (*left > ungranted + bytes), instead
1059 of (*left > ungranted), because only we are sure the left space
1060 is enough for another "bytes", then the ungrant space should be
1061 increase. In client, we should update cl_avail_grant only there
1062 is OBD_MD_FLGRANT in the reply.
1066 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1067 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1068 described in CVE-2007-4573. This problem could allow untrusted
1069 local users to gain root access.
1072 Frequency : when using O_DIRECT and quotas
1074 Description: Incorrect file ownership on O_DIRECT output files
1075 Details : block usage reported by 'lfs quota' does not take into account
1076 files that have been written with O_DIRECT.
1081 Description: touch file failed when fs is not full
1082 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1083 otherwise we can get ENOSP while fs is not full.
1087 Description: parallel lock callbacks
1088 Details : Instead of sending blocking and completion callbacks as separated
1089 requests, adding them to a set and sending in parallel.
1092 Frequency : only for Cray XT3
1093 Bugzilla : 12829/13455
1094 Description: Changing primary group doesn't change the group lustre assigns to
1096 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1097 overridden with the primary group provided by the group upcall,
1098 whereas the supplied fsgid can be trusted if it is in the list of
1099 supplementary groups returned by the group upcall.
1101 Severity : enhancement
1103 Description: Allow masking D_WARNING, D_ERROR messages from console
1104 Details : Console messages can now be disabled via lnet.printk.
1108 Description: User code with malformed file open parameter crashes client node
1109 Details : Before packing join_file req, all the related reference should be
1110 checked carefully in case some malformed flags cause fake
1111 join_file req on client.
1115 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1116 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1117 mode is equal to the granted mode. However, it should also check
1118 that LDLM_FL_CP_REQD is not set.
1122 Description: Heavy nfs access might result in deadlocks
1123 Details : After ELC code landed, it is now improper to enqueue any mds
1124 locks under och_sem, because enqueue might want to decide to
1125 cancel open locks for same inode we are holding och_sem for.
1129 Description: Client eviction while running blogbench
1130 Details : A lot of unlink operations with concurrent I/O can lead to a
1131 deadlock causing evictions. To address the problem, the number of
1132 oustanding OST_DESTROY requests is now throttled to
1133 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1134 callbacks are processed in priority.
1138 Description: enable ACLs on MDS by default
1139 Details : ACLs must be enabled on MDS by default.
1142 Frequency : PPC/PPC64 only
1144 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1145 Details : fix duplicated definitions between asm-ppc64/types.h and
1146 lustre_types.h on PPC.
1149 Frequency : PPC/PPC64 only
1151 Description: asm-ppc/segment.h does not exist
1152 Details : fix compile issue on PPC.
1156 Description: better handle error messages in extents code
1159 Frequency : RHEL4 only
1161 Description: mkfs is very slow on IA64/RHEL4
1162 Details : A performance regression has been discovered in the MPT Fusion
1163 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1164 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1165 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1167 Severity : enhancement
1169 Description: SNMP support enhancement
1170 Details : Adding total number of sampled request for an MDS node in snmp
1173 Severity : enhancement
1175 Description: Optimize ldlm waiting list processing for PR extent locks
1176 Details : When processing waiting list for read extent lock and meeting read
1177 lock that is same or wider to it that is not contended, skip
1178 processing rest of the list and immediatelly return current
1179 status of conflictness, since we are guaranteed there are no
1180 conflicting locks in the rest of the list.
1184 Description: Time out and refuse to reconnect
1185 Details : When the failover node is the primary node, it is possible
1186 to have two identical connections in imp_conn_list. We must
1187 compare not conn's pointers but NIDs, otherwise we can defeat
1188 connection throttling.
1192 Description: port llog fixes from b1_6 into HEAD
1193 Details : Port llog reference couting and some llog cleanups from b1_6
1194 (bug 10800) into HEAD, for protect from panic and access to already
1195 free llog structures.
1199 Description: Detect stride IO mode in read-ahead
1200 Details : When a client does stride read, read-ahead should detect that and
1201 read-ahead pages according to the detected stride pattern.
1205 Description: data checksumming impacts single node performance
1206 Details : add support for several checksum algorithm. Currently, only CRC32
1207 and Adler-32 are supported. The checksum type can be changed on
1208 the fly via /proc/fs/lustre/osc/*/checksum_type.
1212 Description: use adler32 for page checksums
1213 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1218 Description: build for x2 fails
1219 Details : fix compile issue on Cray systems.
1223 Description: Properly match for duplicate locks
1224 Details : Due to different lock order from skiplists code, we need to
1225 traverse entire list for now
1228 Frequency : only on PPC/SLES10
1230 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1231 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1232 be defined. Add a hack in mkfs_lustre.c to work around this compile
1237 Description: LASSERT on MDS when client holding flock lock dies
1238 Details : ldlm pool logic depends on number of granted locks equal to
1239 number of released locks which is not true for flock locks, so
1240 just exclude such locks from consideration.
1244 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1245 Details : Use fsfilt_sync() for both the whole filesystem sync and
1246 individual file sync to eliminate dangerous inode locking
1247 with I_LOCK that can lead to a deadlock.
1251 Description: performance in 1.6.3
1252 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1253 sure that 1MB requests are not fragmented by the block layer.
1255 Severity : enhancement
1257 Description: organize the server-side client stats on per-nid basis
1258 Details : Change the structure of stats under obdfilter and mds to
1266 The "uuid"s file would list the uuids of _active_ exports.
1267 And the clear entry is to clear all stats and stale nids.
1269 Severity : enhancement
1271 Description: eliminate client locks in face of contention
1272 Details : file contention detection and lockless i/o implementation
1273 for contended files.
1277 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1280 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1282 Description: MDS may lose file striping (and hence file data) in some cases
1283 Details : If there are additional extended attributes stored on the MDS,
1284 in particular ACLs, SELinux, or user attributes (if user_xattr
1285 is specified for the client mount options) then there is a risk
1286 of attribute loss. Additionally, the Lustre file striping
1287 needs to be larger than default (e.g. striped over all OSTs),
1288 and an additional attribute must be stored initially in the
1289 inode and then increase in size enough to be moved to the
1290 external attribute block (e.g. ACL growing in size) for file
1295 Description: skiplist implementation simplification
1296 Details : skiplists are used to group compatible locks on granted list
1297 that was implemented as tracking first and last lock of each
1298 lock group the patch changes that to using doubly linked lists
1302 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1303 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1304 because LASSERT on both the data supplied by a client, and the
1305 data on disk is dangerous and incorrect.
1307 Severity : enhancement
1309 Description: Slow truncate/writes to huge files at high offsets.
1310 Details : Directly associate cached pages to lock that protect those pages,
1311 this allows us to quickly find what pages to write and remove
1312 once lock callback is received.
1316 Description: more ldlm soft lockups
1317 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1318 starve other threads from the resource lock for a long time in
1319 case of long waiting queue, so change the debug level from
1320 D_OTHER to the less frequently used D_INFO.
1322 Severity : enhancement
1324 Description: add -gid, -group, -uid, -user options to lfs find
1328 Description: Hung threads in invalidate_inode_pages2_range
1329 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1330 once one is completed. As a result, some RPCs are stuck in the
1331 queue and are never sent.
1335 Description: filter threads hungs on waiting journal commit
1336 Details : Cleanup filter group llog code, then only filter group llog will
1337 be only created in the MDS/OST syncing process.
1341 Description: Procfs and llog threads access destoryed import sometimes.
1342 Details : Sync the import destoryed process with procfs and llog threads by
1343 the import refcount and semaphore.
1345 Severity : enhancement
1347 Description: openlock cache of b1_6 port to HEAD
1352 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1353 Details : If insertion of an extent fails, then discard the inode
1354 preallocation and free data blocks else it can lead to duplicate
1359 Description: don't always update ctime in ext3_xattr_set_handle()
1360 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1361 In some cases the ctime should not be updated, for example for
1362 2.0->1.8 compatibility it is necessary to delete an xattr and it
1363 should not update the ctime.
1367 Bugzilla : 15713/16362
1368 Description: Assertion in iopen_connect_dentry in 1.6.3
1369 Details : looking up an inode via iopen with the wrong generation number can
1370 populate the dcache with a disconneced dentry while the inode
1371 number is in the process of being reallocated. This causes an
1372 assertion failure in iopen since the inode's dentry list contains
1373 both a connected and disconnected dentry.
1377 Description: assertion failure in ldlm_handle2lock()
1378 Details : fix a race between class_handle_unhash() and class_handle2object()
1379 introduced in lustre 1.6.5 by bug 13622.
1384 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1385 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1390 Description: oops in sd_iostats_seq_show()
1391 Details : unloading/reloading the scsi low level driver triggers a kernel
1392 bug when trying to access the sd iostat file.
1397 Description: Kernel panics during QLogic driver reload
1398 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1399 patch, causing memory corruption.
1404 Description: journal_dev option does not work in b1_6
1405 Details : pass mount option during pre-mount.
1407 Severity : enhancement
1409 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1410 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1411 extent information of a file. It can be used to map logical blocks
1412 in a file to physical blocks in the block device.
1416 Description: LDLM soft lockups - improvement
1417 Details : It is be possible to send the lock handle along with each read
1418 or write request because the client is already doing a lock match
1419 itself so there isn't any reason the OST should have to re-do that
1423 Frequency : only on Cray X2
1425 Description: X2 build failures
1426 Details : fix build failures on Cray X2.
1430 Description: xid & resent requests
1431 Details : Initialize RPC XID from clock at startup (randomly if clock is
1434 Severity : enhancement
1436 Description: Add lustre_start utility to start or stop multiple Lustre servers
1441 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1442 Details : In case of memory pressure, list_del() can be called twice on
1443 req->rq_history_list, causing a kernel oops.
1447 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1448 Details : fix stack overflow in the distributed lock manager by defering
1449 export eviction after a failed AST to the elt thread instead of
1450 handling it in the dlm interpret routine.
1454 Description: Convert some comments to new format.
1455 Details : Update documenting comments to match doxygen conventions.
1459 Description: Grammar fixes.
1460 Details : A couple of trivial spelling fixes.
1464 Description: OSD_COUNTERS-mandatory
1465 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1466 to simplify the code.
1470 Description: simplify cmm_device freeing logic.
1471 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1475 Description: Add lockdep support to dt_object_operations locking interface.
1476 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1477 parameter indicating lock ordering. Update mdd code to use new
1482 Description: Introduce failloc constants for lockless IO tests.
1483 Details : Add two new failloc constants to test lockless IO. Only one of
1484 them in implemented---another is checked in yet to be landed
1489 Description: Add lockdep support for inode mutex.
1490 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1491 in the situations where Lustre has to lock more than one inode
1496 Description: Add optional invariants checking support.
1497 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1498 configure switch. This macro is to be used for consistency and
1499 sanity checks that are too expensive to be left in `production'
1504 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1505 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1510 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1511 Details : Use spin_lock_nested() in (the only) situation where more than
1512 one ldlm_lock is locked simultaneously. Also, fix possible
1513 dead-lock in ldlm_lock_change_resource() by enforcing particular
1518 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1519 Details : Instead of specifying each ldlm_lock call-back through separate
1520 parameter, wrap them into struct ldlm_callback_suite.
1524 Description: Kill join_lru obd method and its callers.
1525 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1526 in memory---a requirement implemented through ->o_join_lru() obd
1527 method in HEAD. Remove this method and its users.
1531 Description: Add asynchronous ldlm ENQUEUE completion handler.
1532 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1533 case that stock ldlm_completion_ast() cannot handle as it waits
1534 until lock is granted. Introduce new ldlm_completion_ast_async()
1535 for this. Also comment ldlm_completion_ast().
1539 Description: ldlm_error <-> errno conversion.
1540 Details : Add functions to map (rather arbitrary) between LDLM error codes
1541 and standard errno values. CLIO needs this to prevent LDLM specific
1542 constants from escaping ldlm and osc.
1546 Description: Kill unused ldlm_handle2lock_ns() function.
1547 Details : Kill unused ldlm_handle2lock_ns() function.
1551 Description: Add lu_ref support to ldlm_lock
1552 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1553 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1554 and ldlm_resource. LDLM interface has to be changed, because code that
1555 releases a reference on a lock, has to "know" what reference this is.
1556 In the most frequent case
1558 lock = ldlm_handle2lock(handle);
1560 LDLM_LOCK_PUT(lock);
1562 no changes are required. When any other reference (received _not_ from
1563 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1564 instead of LDLM_LOCK_PUT().
1566 Arguably, changes are pervasive, and interface requires some discipline
1567 for proper use. On the other hand, it was very instrumental in finding
1568 a few leaked lock references.
1572 Description: Add ldlm_lock_addref_try().
1573 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1574 attempts to addref a lock that might be being canceled
1579 Description: Add ldlm_weigh_callback().
1580 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1581 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1582 hard-coded `number of pages' logic.
1586 Description: Add lockdep annotations to llog code.
1587 Details : Use appropriately tagged _nested() locking calls in the places
1588 where llog takes more than one ->lgh_lock lock.
1592 Description: Add loi_kms_set().
1593 Details : Wrap kms updates into a helper function.
1597 Description: Constify instances of struct lsm_operations.
1598 Details : Constify instances of struct lsm_operations.
1602 Description: lu_conf support.
1603 Details : On a server, a file system object is uniquely identified
1604 by a fid, which is sufficient to locate and load all object
1605 state (inode). On a client, on the other hand, more data are
1606 necessary instantiate an object. Change lu_object_find() and
1607 friends to take additional `lu_conf' argument describing object.
1608 Typically this includes layout information.
1612 Description: lu_context fixes.
1613 Details : Introduce new lu_context functions that are needed on the client
1614 side, where some system threads (ptlrpcd) are shared by multiple
1615 modules, and so cannot be stopped during module shutdown.
1619 Description: Add start and stop methods to lu_device_type_operations.
1620 Details : Introduce two new methods in lu_device_type_operations, that are
1621 invoked when first instance of a given type is created and last one
1622 is destroyed respectively. This is need by CLIO.
1626 Description: Add lu_ref support to struct lu_device.
1627 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1628 track leaked references.
1632 Description: Introduce lu_kmem_descr.
1633 Details : lu_kmem_descr and its companion interface allow to create
1634 and destroy a number of kmem caches at once.
1638 Description: Fix lu_object finalization race.
1639 Details : Fix a race between lu_object_find() finding an object and its
1640 concurrent finalization. This race is (most likely) not possible
1641 on the server, but might happen on the client.
1645 Description: Introduce lu_ref interface.
1646 Details : lu_ref is a debugging module allowing to track references to
1647 a given object. It is quite cpu expensive, and has to be
1648 explicitly enabled with --enable-lu_ref. See usage description
1653 Description: Factor lu_site procfs stats into a separate function.
1654 Details : Separate lu_site stats printing code into a separate function
1655 that can be reused on a client.
1659 Description: Constify instances of struct {lu,dt,md}_device_operations.
1660 Details : Constify instances of struct {lu,dt,md}_device_operations.
1664 Description: Introduce struct md_site and move meta-data specific parts of
1665 struct lu_site here.
1666 Details : Move md-specific fields out of struct lu_site into special struct
1667 md_site, so that lu_site can be used on a client.
1671 Description: Kill mdd_lov_destroy().
1672 Details : Remove unused mdd code.
1676 Description: Add st_block checking to multistat.c.
1677 Details : Add st_block checking to multistat.c.
1681 Description: Add lu_ref support to struct obd_device.
1682 Details : Add lu_ref tracking to obd_device.
1686 Description: Kill obd_set_fail_loc().
1687 Details : Remove unused code.
1691 Description: Add special type for ptlrpc_request interpret functions.
1692 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1693 there. Actual usage will be in CLIO.
1697 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1698 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1699 doesn't work with lockdep.
1703 Description: Add rwv.c test program.
1704 Details : New testing program exercising readv(2) and writev(2) (Qian).
1708 Description: Add sendfile.c test program.
1709 Details : New testing program exercising sendfile(2) (Jay).
1713 Description: Ratelimit a message that can be very frequent.
1714 Details : Ratelimit a memory allocation failure message that can
1719 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1720 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1724 Description: Liblustre build fixes.
1725 Details : Liblustre build fixes.
1729 Description: libcfs: add cfs_{need,cond}_resched() interface.
1730 Details : libcfs: add cfs_{need,cond}_resched() definition and
1731 implementations for Linux, NT, and liblustre.
1733 Severity : enhancement
1735 Description: More exported tunables for mballoc
1736 Details : Add support for tunable preallocation window and new tunables for
1737 large/small requests
1741 Description: Detect corruption of block bitmap and checking for preallocations
1742 Details : Checks validity of on-disk block bitmap. Also it does better
1743 checking of number of applied preallocations. When corruption is
1744 found, it turns filesystem readonly to prevent further corruptions.
1748 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1749 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1750 window to 0. If it does hit cache miss, and read-pattern is still
1751 stride-io mode, does not reset the stride window, but also does not
1752 increase the stride window length in this case.
1754 --------------------------------------------------------------------------------
1756 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1758 * Support for kernels:
1759 2.6.5-7.283 (SLES 9),
1760 2.6.9-55.EL (RHEL 4),
1761 2.6.16.46-0.14 (SLES 10),
1762 2.6.18.8 vanilla (kernel.org)
1763 * Client support for unpatched kernels:
1764 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1765 2.6.16 - 2.6.22 vanilla (kernel.org)
1766 * Due to recently discovered recovery problems, we do not recommend
1767 using patchless RHEL 4 clients with this or any earlier release.
1768 * Recommended e2fsprogs version: 1.39.cfs8
1769 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1770 * Starting with this release, the ldiskfs backing filesystem required
1771 by Lustre is now in its own package, lustre-ldiskfs. This package
1772 should be installed. It is versioned separately from Lustre and
1773 may be released separately in future.
1777 Description: block reactivating mgc import until all deactivates complete
1778 Details : Fix race when failing back MDT/MGS to itself (testing)
1781 Frequency : at statup only
1783 Description: mds_lov_synchronize race leads to various problems
1784 Details : simultaneous MDT->OST connections at startup can cause the
1785 sync to abort, leaving the OSC in a bad state.
1787 Severity : enhancement
1789 Description: add optional extra BUILD_VERSION info
1790 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1791 to override the lustre version.
1794 Frequency : 2.6.18 servers only
1796 Description: ll_kern_mount() doesn't release the module reference
1797 Details : The ldiskfs module reference count never drops down to 0
1798 because ll_kern_mount() doesn't release the module reference.
1803 Description: server LBUG when using old ost_num_threads parameter
1804 Details : Accept the old ost_num_threads parameter but warn that it
1805 is deprecated, and fix an off-by-one error that caused an LBUG.
1810 Description: Transient SCSI error results in persistent IO issue
1811 Details : iobuf->dr_error is not reinitialized to 0 between two
1815 Frequency : sometimes when underlying device returns I/O errors
1817 Description: OSTs not going read-only during write failures
1818 Details : OSTs are not remounted read-only when the journal commit threads
1819 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1820 instead of the ext3 wrappers.
1824 Description: poor connect scaling with increasing client count
1825 Details : Don't run filter_grant_sanity_check for more than 100 exports
1826 to improve scaling for large numbers of clients.
1829 Frequency : SLES10 only
1831 Description: sanity-quota.sh quotacheck failed: rc = -22
1832 Details : Quotas cannot be enabled on SLES10.
1835 Frequency : liblustre clients only
1837 Description: getdirentries does not give error when run on compute nodes
1838 Details : getdirentries does not fail when the size specified as an argument
1839 is too small to contain at least one entry
1841 Severity : enhancement
1843 Description: Add LNET router traceability for debug purposes
1844 Details : If a checksum failure occurs with a router as part of the
1845 IO path, the NID of the last router that forwarded the bulk data
1846 is printed so it can be identified.
1851 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1852 Details : Due to a race condition, liblustre clients were occasionally
1853 evicted incorrectly.
1855 Severity : enhancement
1857 Description: lfs setstripe use optional parameters instead of postional
1860 Severity : enhancement
1862 Description: Nanosecond timestamp support for ldiskfs
1863 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1864 resolution timestamps. There is not yet support for this at
1865 the Lustre filesystem level.
1868 Frequency : during server recovery
1870 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1871 Details : request with rq_no_resend flag not awake l_wait_event if they get
1875 Frequency : nfs export on patchless client
1877 Description: connectathon hang when test nfs export over patchless client
1878 Details : Disconnected dentry cannot be found with lookup, so we do not need
1879 to unhash it or make it invalid
1882 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1883 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1884 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1889 Description: open req refcounting wrong on reconnect
1890 Details : If reconnect happened between getting open reply from server and
1891 call to mdc_set_replay_data in ll_file_open, we will schedule
1892 replay for unreferenced request that we are about to free.
1893 Subsequent close will crash in variety of ways.
1894 Check that request is still eligible for replay in
1895 mdc_set_replay_data().
1900 Description: disable writes to filesystem when reading health_check file
1901 Details : the default for reading the health_check proc file has changed
1902 to NOT do a journal transaction and write to disk, because this
1903 can cause reads of the /proc file to hang and block HA state
1904 checking on a healthy but otherwise heavily loaded system. It
1905 is possible to return to the previous behaviour during configure
1906 with --enable-health-write.
1908 Severity : enhancement
1910 Description: 64-bit inode version
1911 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1912 to the inode. This will be required for version-based recovery.
1917 Description: MDS fails to start if a duplicate client export is detected
1918 Details : in some rare cases it was possible for a client to connect to
1919 an MDS multiple times. Upon recovery the MDS would detect this
1920 and fail during startup. Handle this more gracefully.
1922 Severity : enhancement
1924 Description: Add -o localflock option to simulate old noflock
1926 Details : This will achieve local-only flock/fcntl locks
1932 Description: log_commit_thread vs filter_destroy race leads to crash
1933 Details : Take import reference before releasing llog record semaphore
1938 Description: Wrong request locking in request set processing
1939 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1940 imp_delayed_list, in this place should be used imp_lock.
1943 Frequency : when reconnection
1945 Description: Grant Leak when osc reconnect to OST
1946 Details : When osc reconnect ost, OST(filter) should check whether it
1947 should grant more space to client by comparing fed_grant and
1948 cl_avail_grant, and return the granted space to client instead
1949 of "new granted" space, because client will call osc_init_grant
1950 to update the client grant space info.
1953 Frequency : when client reconnect to OST
1955 Description: Grant Leak when osc do resend and replay bulk write
1956 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1957 bulk write request, because the grant info while be sync between
1958 OSC and OST when reconnect, and we should ignore the grant info
1959 these of resend/replay write req.
1964 Description: Grant space more than avaiable left space sometimes.
1965 Details : When then OST is about to be full, if two bulk writing from
1966 different clients came to OST. Accord the avaliable space of the
1967 OST, the first req should be permitted, and the second one
1968 should be denied by ENOSPC. But if the seconde arrived before
1969 the first one is commited. The OST might wrongly permit second
1970 writing, which will cause grant space > avaiable space.
1973 Frequency : when client is evicted
1975 Description: Grant might be wrongly erased when osc is evicted by OST
1976 Details : when the import is evicted by server, it will fork another
1977 thread ptlrpc_invalidate_import_thread to invalidate the
1978 import, where the grant will be set to 0. While the original
1979 thread will update the grant it got when connecting. So if
1980 the former happened latter, the grant will be wrongly errased
1981 because of this race.
1986 Description: Checking Stale with correct fid
1987 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1988 to check whether it is stale, because sometimes, we want the
1989 enqueue happened anyway, and op_data.fid2 will not be initialized.
1991 Severity : enhancement
1993 Description: update patchless client
1994 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1997 Frequency : only with 2.4 kernel
1999 Description: random memory corruption
2000 Details : size of struct ll_inode_info is to big for union inode.u and this
2001 can be cause of random memory corruption.
2006 Description: Memory leak in recovery
2007 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2008 It should also check obd_fail before fsfilt_start, otherwise if
2009 fsfilt_start return -EROFS,(failover mds during mds recovery).
2010 then the req will return with repmsg->transno = 0 and rc = EROFS.
2011 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2012 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2013 be freed no matter whether fsfilt_commit success or not.
2016 Frequency : only with huge count clients
2018 Description: Prevents from taking the superblock lock in llap_from_page for
2020 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2021 for llap_from_page call in ll_removepage prevents from taking the
2022 superblock lock for a soon died page.
2027 Description: Not check open intent error before release open handle
2028 Details : in some rare cases, the open intent error is not checked before
2029 release open handle, which may cause
2030 ASSERTION(open_req->rq_transno != 0), because it tries to release
2031 the failed open handle.
2036 Description: Set cat log bitmap only after create log success.
2037 Details : in some rare cases, the cat log bitmap is set too early. and it
2038 should be set only after create log success.
2042 Description: Accessing a block bevice can re-enable I/O when Lustre is
2043 tearing down a device.
2044 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2049 Description: service threads may hog cpus when there are a lot of requests
2050 Details : Insert cond_resched to give other threads a chance to use some CPU
2055 Description: the cat log was not initialized in recovery
2056 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2057 unlink log on mds will not be initialized until mds post
2058 recovery. And also in mds post recovery, the unlink log will
2059 initialization will be done asynchronausly, so there will be race
2060 between add unlink log and unlink log initialization.
2064 Description: brw_stats were being printed incorrectly
2065 Details : brw_stats were being printed as log2 but all of them were not
2066 recorded as log2. Also remove some code duplication arising from
2067 filter_tally_{read,write}.
2071 Frequency : rare, only in recovery.
2072 Description: ASSERTION(req->rq_type != LI_POISON) failed
2073 Details : imp_lock should be held while iterating over imp_sending_list for
2074 prevent destroy request after get timeout in ptlrpc_queue_wait.
2078 Description: replay-single.sh test 52 fails
2079 Details : A lock's skiplist need to be cleanup when it being unlinked
2080 from its resource list.
2084 Description: Short directio read returns full requested size rather than
2086 Details : Direct I/O operations should return actual amount of bytes
2087 transferred rather than requested size.
2089 Severity : enhancement
2091 Description: metadata RPC reduction (e.g. for rm performance)
2092 Details : decrease the amount of synchronous RPC between clients and servers
2093 by canceling conflicing lock before the operation on the client
2094 and packing thier handles into the main operation RPC to server.
2096 Severity : enhancement
2098 Description: Async OSC create to avoid the blocking unnecessarily.
2099 Details : If a OST has no remain object, system will block on the creating
2100 when need to create a new object on this OST. Now, ways use
2101 pre-created objects when available, instead of blocking on an
2102 empty osc while others are not empty. If we must block, we block
2103 for the shortest possible period of time.
2107 Description: improve handling recoverable errors
2108 Details : if request processig with error which can be recoverable on server
2109 request should be resend, otherwise page released from cache and
2112 Severity : enhancement
2114 Description: refine locking for avoid write wrong info into lov_objid file
2115 Details : fix possible races with add new target and write/update data in
2118 --------------------------------------------------------------------------------
2120 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2125 Frequency : on some architectures
2127 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2128 Details : Uninitialized flags sometimes cause configuration commands to
2132 Frequency : patchless clients only
2134 Description: missing __iget() symbol export
2135 Details : The __iget() symbol export is missing. To avoid the need for
2136 this on patchless clients the deathrow inode reaper is turned
2137 off, and we depend on the VM to clean up old inodes. This
2138 dependency was during via the fix for bug 12181.
2140 --------------------------------------------------------------------------------
2142 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2144 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2145 INTEROPERATE with older versions automatically. In many cases a
2146 special upgrade step is needed. Please read the
2147 user documentation before upgrading any part of a live system.
2148 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2149 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2150 * WARNING: Lustre configuration and startup changes are required with
2151 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2155 Severity : enhancement
2157 Description: MountConf
2158 Details : Lustre configuration is now managed via mkfs and mount
2159 commands instead of lmc and lconf. New obd types (MGS, MGC)
2160 are added for dynamic configuration management. See
2161 https://mail.clusterfs.com/wikis/lustre/MountConf for
2164 Severity : enhancement
2166 Description: dynamic OST addition
2167 Details : OSTs can now be added to a live filesystem
2169 Severity : enhancement
2171 Description: startup order invariance
2172 Details : MDTs and OSTs can be started in any order. Clients only
2173 require the MDT to complete startup.
2175 Severity : enhancement
2177 Description: parallel, asynchronous orphan cleanup
2178 Details : orphan cleanup is now performed in separate threads for each
2179 OST, allowing parallel non-blocking operation.
2181 Severity : enhancement
2183 Description: optimized stripe assignment
2184 Details : stripe assignments are now made based on ost space available,
2185 ost previous usage, and OSS previous usage, in order to try
2186 to optimize storage space and networking resources.
2188 Severity : enhancement
2190 Description: Permanently set tunables
2191 Details : All writable /proc/fs/lustre tunables can now be permanently
2192 set on a per-server basis, at mkfs time or on a live system.
2194 Severity : enhancement
2196 Description: Lustre message v2
2197 Details : Add lustre message format v2.
2199 Severity : enhancement
2201 Description: client OST exclusion list
2202 Details : Clients can be started with a list of OSTs that should be
2203 declared "inactive" for known non-responsive OSTs.
2207 Description: SPEC SFS validation failure on NFS v2 over lustre.
2208 Details : Changes the blocksize for regular files to be 2x RPC size,
2209 and not depend on stripe size.
2211 Severity : enhancement
2213 Description: Multiple MD RPCs in flight.
2214 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2215 To avoid overly-overloading MDS, introduce a limit on number of
2216 MDS RPCs in flight for a single client and add /proc controls
2217 to adjust this limit.
2219 Severity : enhancement
2221 Description: client read/write statistics
2222 Details : Add client read/write call usage stats for performance
2223 analysis of user processes.
2224 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2225 file access. extents_stats shows chunk size distribution.
2226 extents_stats_per_process show chunk size distribution per
2229 Severity : enhancement
2231 Description: mds statistics
2232 Details : Add detailed mds operations statistics in
2233 /proc/fs/lustre/mds/*/stats.
2237 Description: Failure of copying files with lustre special EAs.
2238 Details : Client side always return success for setxattr call for lustre
2239 special xattr (currently only "trusted.lov").
2244 Description: Refcount LNET uuids
2245 Details : The global LNET uuid list grew linearly with every startup;
2246 refcount repeated list entries instead of always adding to
2249 Severity : enhancement
2251 Description: Dynamic service threads
2252 Details : Within a small range, start extra service threads
2253 automatically when the request queue builds up.
2256 Frequency : mixed-endian client/server environments
2258 Description: mixed-endian crashes
2259 Details : The new msg_v2 system had some failures in mixed-endian
2262 Severity : enhancement
2264 Description: Easy OST removal
2265 Details : OSTs can be permanently deactivated with e.g. 'lctl
2266 conf_param lustre-OST0001.osc.active=0'
2268 Severity : enhancement
2270 Description: MGS proc entries
2271 Details : Added basic proc entries for the MGS showing what filesystems
2274 Severity : enhancement
2276 Description: provide MGS failover
2277 Details : Added config lock reacquisition after MGS server failover.
2279 Severity : enhancement
2281 Description: add Linux 2.4 support
2282 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2286 Description: a large application tries to do I/O to the same resource and dies
2287 in the middle of it.
2288 Details : Check the req->rq_arrival time after the call to
2289 ost_brw_lock_get(), but before we do anything about
2290 processing it & sending the BULK transfer request. This
2291 should help move old stale pending locks off the queue as
2292 quickly as obd_timeout.
2295 Frequency : when an incorrect nid is specified during startup
2297 Description: ptlrpc connect to non-existant node causes kernel crash
2298 Details : LNET can't be re-entered from an event callback, which
2299 happened when we expire a message after the export has been
2300 cleaned up. Instead, hand the zombie cleanup off to another
2303 Severity : enhancement
2305 Description: plain/inodebits lock performance improvement
2306 Details : Grouping plain/inodebits in granted list by their request modes
2307 and bits policy, thus improving the performance of search through
2311 Frequency : only if OST filesystem is corrupted
2313 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2314 Details : for a short time RPCs with bulk IO are in the replay list,
2315 but replay of bulk IOs is unimplemented. If the OST filesystem
2316 is corrupted due to disk cache incoherency and then replay is
2317 started it is possible to trip an assertion. Avoid putting
2318 committed RPCs into the replay list at all to avoid this issue.
2321 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2323 Description: System hang on startup
2324 Details : This bug allowed the liblustre (e.g. catamount) client to
2325 return to the app before handling all startup RPCs. This
2326 could leave the node unresponsive to lustre network traffic
2327 and manifested as a server ptllnd timeout.
2329 Severity : enhancement
2331 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2332 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2333 boolean to control whether to print peer debug info when a
2334 client's RPC times out.
2337 Frequency : only for kernels with patches from Lustre below 1.4.3
2339 Description: Remove old rdonly API
2340 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2343 Frequency : only for devices with external journals
2345 Description: Set external device read-only also
2346 Details : During a commanded failover stop, we set the disk device
2347 read-only while the server shuts down. We now also set any
2348 external journal device read-only at the same time.
2351 Frequency : when upgrading from 1.4 while trying to change parameters
2353 Description: The wrong (new) MDC name was used when setting parameters for
2354 upgraded MDT's. Also allows changing of OSC (and MDC)
2355 parameters if --writeconf is specified at tunefs upgrade time.
2358 Frequency : when setting specific ost indicies
2360 Description: QOS code breaks on skipped indicies
2361 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2362 created with --index need not be sequential.
2364 Severity : enhancement
2366 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2367 Details : The uninit_groups feature works in conjunction with the kernel
2368 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2369 up the pass1 processing of e2fsck. This is a read-only feature
2370 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2371 mount filesystems that have had this feature enabled.
2373 Severity : enhancement
2375 Description: Improve multi-block allocation algorithm to avoid fragmentation
2376 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2377 allocation locality and avoid filesystem fragmentation.
2379 ------------------------------------------------------------------------------
2381 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2383 * Support for kernels:
2384 2.4.21-47.0.1.EL (RHEL 3)
2385 2.6.5-7.283 (SLES 9)
2386 2.6.9-42.0.10.EL (RHEL 4)
2387 2.6.12.6 vanilla (kernel.org)
2388 2.6.16.27-0.9 (SLES 10)
2389 * Recommended e2fsprogs version: 1.39.cfs5
2391 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2395 Frequency : occasional, depends on client load and configuration
2396 Bugzilla : 12181, 12203
2397 Description: data loss for recently-modified files
2399 Details : In some cases it is possible that recently written or created
2400 files may not be written to disk in a timely manner (this should
2401 normally be within 30s unless client IO load is very high).
2402 The problem appears as zero-length files or files that are a
2403 multiple of 1MB in size after a client crash or client eviction
2404 that are missing data at the end of the file.
2406 This problem is more likely to be hit on clients where files are
2407 repeatedly created and unlinked in the same directory, clients
2408 have a large amount of RAM, have many CPUs, the filesystem has
2409 many OSTs, the clients are rebooted frequently, and/or the files
2410 are not accessed by other nodes after being written.
2412 The presence of the problem can be detected by looking at
2413 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2414 smaller than the second (nr_unused) then dirty files will not
2415 be flushed automatically to disk. "sync; sleep 10" should be
2416 run several times on the node before unmounting it to update
2417 Lustre (this is also safe to run on nodes without this problem).
2419 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2420 that can cause this same problem, so customers using that kernel
2421 also need to update the kernel in addition to Lustre. In order
2422 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2424 It is normal that files written just before a client crash (less
2425 than 30s) may not yet have been flushed to disk, even for local
2429 Frequency : frequent on thin XT3 nodes
2431 Description: UUID collision on thin XT3 Linux nodes
2432 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2433 randomly, since we relied on an insufficiently-seeded PRNG.
2438 Description: OSS hangs after "All ost request buffers busy"
2439 Details : A deadlock between quota and journal operations caused OSS
2440 hangs after printing "All ost request buffers busy."
2443 Frequency : always on liblustre builds
2445 Description: Cleanup compiler warnings on liblustre
2448 Frequency : always on liblustre builds on XT3
2450 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2454 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2455 > 32000 subdirectories
2456 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2457 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2458 subdirectory count crosses 32000. This will aid e2fsck to
2459 correctly handle more than 32000 subdirectories.
2462 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2464 Description: System hang on startup
2465 Details : This bug allowed the liblustre (e.g. catamount) client to
2466 return to the app before handling all startup RPCs. This
2467 could leave the node unresponsive to lustre network traffic
2468 and manifested as a server ptllnd timeout.
2470 Severity : enhancement
2472 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2473 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2474 boolean to control whether to print peer debug info when a
2475 client's RPC times out.
2480 Description: make O_SYNC working on 2.6 kernels
2481 Details : 2.6 kernels use different method for mark pages for write,
2482 so need add a code to lustre for O_SYNC work.
2487 Description: Failure to close file and release space on NFS
2488 Details : Put inode details into lock acquired in ll_intent_file_open.
2489 Use mdc_intent_lock in ll_intent_open to properly
2490 detect all kind of errors unhandled by mdc_enqueue.
2495 Description: proc file read during shutdown sometimes raced obd removal,
2497 Details : Add lock to prevent obd access after proc file removal.
2500 Frequency : Only for files larger than 4GB on 32-bit clients.
2502 Description: improperly doing page alignment of locks
2503 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2504 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2509 Description: RPCs being resent when they shouldn't be
2510 Details : Some RPCs that should not be resent are being resent. This
2511 can cause inconsistencies in the RPC state machine. Do not
2512 resend such requests.
2515 Frequency : rare, only with NFS export
2517 Description: Crash on NFS re-export node
2518 Details : under very unusual load conditions an assertion is hit in
2519 ll_intent_file_open()
2522 Frequency : only if OST filesystem is corrupted
2524 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2525 Details : for a short time RPCs with bulk IO are in the replay list,
2526 but replay of bulk IOs is unimplemented. If the OST filesystem
2527 is corrupted due to disk cache incoherency and then replay is
2528 started it is possible to trip an assertion. Avoid putting
2529 committed RPCs into the replay list at all to avoid this issue.
2534 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2535 Details : Large single O_DIRECT read and write calls can fail to allocate
2536 a sufficiently large buffer to process the request. In case of
2537 allocation failure the allocation is retried with a smaller
2538 buffer and broken into smaller requests.
2540 Severity : enhancement
2542 Description: Add -o localflock option to simulate old noflock behaviour.
2543 Details : This will achieve local-only flock/fcntl locks coherentness.
2548 Description: versioning check is incomplete
2549 Details : Checking the version difference of client vs. server, report
2550 error if the gap is too big.
2555 Description: add support PG_writeback bit
2556 Details : add support for PG_writeback bit for Lustre, for more carefull
2557 work with page cache in 2.6 kernel. This also fix some deadlocks
2558 and remove hack for work O_SYNC with 2.6 kernel.
2560 Severity : enhancement
2562 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2563 Details : The uninit_groups feature works in conjunction with the kernel
2564 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2565 up the pass1 processing of e2fsck. This is a read-only feature
2566 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2567 mount filesystems that have had this feature enabled.
2569 Severity : enhancement
2571 Description: Improve multi-block allocation algorithm to avoid fragmentation
2572 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2573 allocation locality and avoid filesystem fragmentation.
2575 ------------------------------------------------------------------------------
2577 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2579 * Support for kernels:
2580 2.6.9-42.0.3.EL (RHEL 4)
2581 2.6.5-7.276 (SLES 9)
2582 2.4.21-47.0.1.EL (RHEL 3)
2583 2.6.12.6 vanilla (kernel.org)
2584 2.6.16.21-0.8 (SLES10)
2585 * Recommended e2fsprogs version: 1.39.cfs2-0
2587 * The backwards-compatible /proc/sys/portals symlink has been removed
2588 in this release. Before upgrading, please ensure that you change
2589 any configuration scripts or /etc/sysctl.conf files that access
2590 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2591 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2592 in advance of the upgrade on any system running Lustre 1.4.6 or
2593 newer, since /proc/sys/lnet was added in that version.
2594 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2598 Frequency : only when quota is used
2600 Description: avoid scanning export list for quota master
2601 Details : Change the algorithms to avoid scanning export list in order
2602 to improve the efficiency.
2605 Frequency : MDS failover only, very rarely
2607 Description: "went back in time" messages on mds failover
2608 Details : The greatest transno may be lost when the current operation
2609 finishes with an error (transno==0) and the client's last_rcvd
2610 record is over-written. Save the greatest transno in the
2611 mds_last_transno for this case.
2614 Frequency : always for specific kernels and striping counts
2616 Description: client may get "Matching packet too big" without ACL support
2617 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2618 when trying to access files in certain configurations. The
2619 clients should in fact be denied when mounting because they do
2620 not understand ACLs.
2623 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2625 Description: many clients connecting with IO in progress causes connect timeouts
2626 Details : Avoid synchronous journal commits to avoid delays caused by many
2627 clients connecting/disconnecting when bulk IO is in progress.
2628 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2629 OST_IO_PORTAL to avoid delays behind potentially many pending
2633 Frequency : occasionally with multiple writers to a single file
2635 Description: shared writes to file may result in wrong size reported by stat()
2636 Details : Allow growing of kms when extent lock is cancelled
2639 Frequency : always with random mmap IO to multi-striped file
2641 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2642 Details : Only if the hole is at the end of OST object so that kms is too
2643 small. Fix is to increase kms accordingly in ll_nopage.
2646 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2648 Description: writes to a missing object would leak memory on the OST
2649 Details : If there is an inconsistency between the MDS and OST filesystems,
2650 such that the MDS references an object that doesn't exist, writes
2651 to that object will leak memory due to incorrect cleanup in the
2652 error handling path, eventually running out of memory on the OST.
2657 Description: Creating too long symlink causes lustre errors
2658 Details : Check symlink and name lengths before sending requests to MDS.
2661 Frequency : only if flock is enabled (not on by default)
2663 Description: posix locks not released on fd closure on 2.6.9+
2664 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2665 kernels, this caused such locks not to be released on fd close and
2666 then assertions on fs unmount about still used locks.
2669 Frequency : MDS failover only, very rarely
2671 Description: clients may get ASSERTION(granted_lock != NULL)
2672 Details : When request was taking a long time, and a client was resending
2673 a getattr by name lock request. The were multiple lock requests
2674 with the same client lock handle and
2675 mds_getattr_name->fixup_handle_for_resent_request found one of the
2676 lock handles but later failed with ASSERTION(granted_lock != NULL).
2681 Description: handle->h_buffer_credits > 0, assertion failure
2682 Details : h_buffer_credits is zero after truncate, causing assertion
2683 failure. This patch extends the transaction or creates a new
2687 Frequency : NFS re-export or patchless client
2688 Bugzilla : 11179, 10796
2689 Description: Crash on NFS re-export node (__d_move)
2690 Details : We do not want to hash the dentry if we don't have a lock.
2691 But if this dentry is later used in d_move, we'd hit uninitialised
2692 list head d_hash, so we just do this to init d_hash field but
2693 leave dentry unhashed.
2696 Frequency : NFS re-export or patchless client
2698 Description: NFS exports has problem with symbolic link
2699 Details : lustre client didn't properly install dentry when re-exported
2700 to NFS or running patchless client.
2703 Frequency : NFS re-export or patchless client
2705 Description: Various nfs/patchless fixes.
2706 Details : fixes reuse disconected alias for lookup process - this fixes
2707 warning "find_exported_dentry: npd != pd",
2708 fix permission error with open files at nfs.
2709 fix apply umask when do revalidate.
2712 Frequency : occasional
2714 Description: Crash on NFS re-export node
2715 Details : calling clear_page() on the wrong pointer triggered oops in
2716 generic_mapping_read().
2719 Frequency : rarely, using O_DIRECT IO
2721 Description: unaligned directio crashes client with LASSERT
2722 Details : check for unaligned buffers before trying any requests.
2725 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2727 Description: stale data returned from RAID cache
2728 Details : If only a small amount of IO is done to the RAID device before
2729 reading it again it is possible to get stale data from the RAID
2730 cache instead of reading it from disk.
2733 Frequency : always for sles10 kernel
2735 Description: sles10 support
2736 Details : ll_follow_link: compile fixes and using of nd_set_link
2737 under newer kernels.
2740 Frequency : depends on arch, kernel and compiler version, always on sles10
2743 Description: recursive or deep enough symlinks cause stack overflow
2744 Details : getting rid of large stack-allocated variable in
2748 Frequency : depends on hardware
2750 Description: lustre write performance loss in the SLES10 kernel
2751 Details : the performance loss is caused by using of write barriers in the
2752 ext3 code. The SLES10 kernel turns barrier support on by
2753 default. The fix is to undo that change for ldiskfs.
2755 ------------------------------------------------------------------------------
2757 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2759 * Support for kernels:
2760 2.6.9-42.0.3EL (RHEL 4)
2761 2.6.5-7.276 (SLES 9)
2762 2.4.21-47.0.1.EL (RHEL 3)
2763 2.6.12.6 vanilla (kernel.org)
2767 Frequency : quota enabled and large files being deleted
2769 Description: releasing more than 4GB of quota at once hangs OST
2770 Details : If a user deletes more than 4GB of files on a single OST it
2771 will cause the OST to spin in an infinite loop. Release
2772 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2777 Description: statfs data retrieved from /proc may be stale or zero
2778 Details : When reading per-device statfs data from /proc, in the
2779 {kbytes,files}_{total,free,avail} files, it may appear
2780 as zero or be out of date.
2783 Frequency : systems with MD RAID1 external journal devices
2785 Description: lconf's call to blkid is confused by RAID1 journal devices
2786 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2787 of returning all block devices that match the journal UUID.
2790 Frequency : always, for aggregate stripe size over 4GB
2792 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2793 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2794 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2797 Frequency : always if "lfs find" used on a local file/directory
2799 Description: "lfs find" segfaults if used on a local file/directory
2800 Details : The case where a directory component was not specified wasn't
2801 handled correctly. Handle this properly.
2804 Frequency : always on ppc64
2806 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2807 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2808 perform endianness conversion.
2811 Frequency : rarely (truncate to non-zero file size after write under load)
2812 Bugzilla : 10730, 10687
2813 Description: Files padded with zeros to next 4K multiple
2814 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2815 it is possible that files that are truncated to a non-zero size
2816 immediately after being written are filled with zero bytes beyond
2817 the truncated size. No file data is lost.
2819 Severity : enhancement
2821 Description: Allow recovery/failover for liblustre clients.
2822 Details : liblustre clients were unaware of failover configurations until
2825 Severity : enhancement
2827 Description: user file locks should fail when not mounting with flock option
2828 Details : Set up an error-returning stub in ll_file_operations.lock field
2829 to prevent incorrect behaviour when client is mounted without
2830 flock option. Also, set up properly f_op->flock field for
2834 Frequency : always on ia64
2836 Description: "lfs df" loops on printing out MDS statfs information
2837 Details : The obd_ioctl_data was not initialized and in some systems
2838 this caused a failure during the ioctl that did not return
2839 an error. Initialize the struct and return an error on failure.
2842 Frequency : SLES 9 only
2844 Description: Error of copying files with lustre special EAs as root
2845 Details : Client side always return success for setxattr call for lustre
2846 special xattr (currently only "trusted.lov").
2849 Frequency : rarely on clusters with both ia64+i386 clients
2851 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2852 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2853 clients are doing concurrent writes to the same file it is
2854 possible that the ia64 clients may LASSERT because the OST
2855 extent locks are not PAGE_SIZE aligned. Ensure that grown
2856 locks are always aligned on the request boundary.
2859 Frequency : specific use, occasional
2861 Description: Overwriting in use executable truncates on-disk binary image
2862 Details : If one node attempts to overwrite an executable in use by
2863 another node, we now correctly return ETXTBSY instead of
2864 truncating the file.
2866 Severity : enhancement
2868 Description: Async OSC create to avoid the blocking unnecessarily.
2869 Details : If a OST has no remain object, system will block on the creating
2870 when need to create a new object on this OST. Now, ways use
2871 pre-created objects when available, instead of blocking on an
2872 empty osc while others are not empty. If we must block, we block
2873 for the shortest possible period of time.
2878 Description: chmod on Lustre root is propagated to other clients
2879 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2880 invalid by the follow_mount time.
2883 Frequency : liblustre clients only
2885 Description: Race in 'instant cancel' lock handling could lead to such locks
2886 never to be granted in case of SMP MDS
2887 Details : Do not destroy not yet granted but cbpending locks in
2891 Frequency : replay/resend of open
2893 Description: non null lock assetion failure in mds_intent_policy
2894 Details : Trying to replay/resend lockless open requests resulted in
2895 mds_open() returning 0 with no lock. Now it sets a flag if
2896 a lock is going to be returned.
2898 Severity : enhancement
2900 Description: Checksum enhancements
2901 Details : New checksum enhancements allow for resending RPCs that failed
2904 Severity : enhancement
2906 Description: Tunables on number of dirty pages in cacche
2907 Details : Allow to set limit on number of dirty pages cached.
2912 Description: client crash on unmount - lock still has references
2913 Details : In some error handling cases it was possible to leak a lock
2914 reference on a client while accessing a file. This was not
2915 harmful to the client during operation, but would cause the
2916 client to crash when the filesystem is unmounted.
2919 Frequency : specific case, rare
2921 Description: ETXTBSY on mds though file not in use by client
2922 Details : ETXTBSY is no longer incorrectly returned when attempting to
2923 chmod or chown a directory that the user previously tried to
2924 execute or a currently-executing binary.
2927 Frequency : extremely rare except on liblustre-based clients
2929 Description: Lustre space not freed when files are deleted
2930 Details : Clean up open-unlinked files after client eviction. Previously
2931 the unlink was skipped and the files remained as orphans.
2936 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2937 Details : In some cases it was possible to send a blocking callback to
2938 a client doing a glimpse, even though that client didn't get
2939 a lock granted. When the glimpse lock is cancelled on the OST
2940 the freed lock is left on the waiting list and corrupted the list.
2943 Frequency : all core dumps
2945 Description: Broke core dumps to lustre
2946 Details : Negative dentry may be unhashed if parent does not have UPDATE
2947 lock, but some callers, e.g. do_coredump, expect dentry to be
2948 hashed after successful create, hash it in ll_create_it.
2950 ------------------------------------------------------------------------------
2952 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2954 * Support for kernels:
2955 2.6.9-42.0.2.EL (RHEL 4)
2956 2.6.5-7.276 (SLES 9)
2957 2.4.21-40.EL (RHEL 3)
2958 2.6.12.6 vanilla (kernel.org)
2962 Frequency : always on RHEL 3
2964 Description: Number of open files grows over time
2965 Details : The number of open files grows over time, whether or not
2966 Lustre is started. This was due to a filp leak introduced
2967 by one of our kernel patches.
2969 ------------------------------------------------------------------------------
2971 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2973 * Support for kernels:
2974 2.6.9-42.EL (RHEL 4)
2975 2.6.5-7.276 (SLES 9)
2976 2.4.21-40.EL (RHEL 3)
2977 2.6.12.6 vanilla (kernel.org)
2982 Bugzilla : 5719, 9635, 9792, 9684
2983 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2984 Details : If a server is under heavy load and cannot reply to new
2985 connection requests before the client resends the (re)connect,
2986 the connection handling code can behave badly if two service
2987 threads are concurrently handing separate (re)connections from
2988 the same client. Add better locking to the connection handling
2989 code, and ensure that only a single connection will be processed
2990 for a given client UUID, even if the lock is dropped.
2992 Severity : enhancement
2994 Description: add TCP zero-copy support to kernel
2995 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2996 sends if the hardware supports scatter-gather and checksumming.
2997 This allows socklnd to do client-write and server-read more
2998 efficiently and reduce CPU utilization from skbuf copying.
3001 Frequency : only if NFS exporting from client
3003 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3004 Details : If setting attributes on a file created under NFS that had
3005 never been opened it would be possible to oops the client
3006 if the file had no objects.
3009 Frequency : always for liblustre
3011 Description: liblustre client does MDS+OSTs setattr RPC for each write
3012 Details : When doing a write from a liblustre client, the client
3013 incorrectly issued an RPC to the MDS and each OST the file was
3014 striped over in order to update the timestamps. When writing
3015 with small chunks and many clients this could overwhelm the MDS
3016 with RPCs. In all cases it would slow down the write because
3017 these RPCs are unnecessary.
3019 Severity : enhancement
3021 Description: allow number of MDS service threads to be changed at module load
3022 Details : It is now possible to change the number of MDS service threads
3023 running. Adding "options mds mds_num_threads={N}" to the MDS's
3024 /etc/modprobe.conf will set the number of threads for the next
3025 time Lustre is restarted (assuming the "mds" module is also
3026 reloaded at that time). The default number of threads will
3027 stay the same, 32 for most systems.
3032 Description: OST crash if filesystem is unformatted or corrupt
3033 Details : If an OST is started on a device that has never been formatted
3034 or if the filesystem is corrupt and cannot even mount then the
3035 error handling cleanup routines would dereference a NULL pointer.
3040 Description: NULL pointer deref in llap_from_page.
3041 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3042 incorrect) mapping if the page was truncated/reclaimed while it was
3043 searched for. Check for this condition and skip such pages when
3044 doing readahead. Introduce extra check to llap_from_page() to
3045 verify page->mapping->host is non-NULL (so page is not anonymous).
3048 Frequency : Sometimes when using sys_sendfile
3050 Description: "page not covered by a lock" warnings from ll_readpage
3051 Details : sendfile called ll_readpage without right page locks present.
3052 Now we introduced ll_file_sendfile that does necessary locking
3053 around call to generic_file_sendfile() much like we do in
3057 Frequency : with certain MDS communication failures at client mount time
3059 Description: NULL pointer deref after failed client mount
3060 Details : a client connection request may delayed by the network layer
3061 and not be sent until after the PTLRPC layer has timed out the
3062 request. If the client fails the mount immediately it will try
3063 to clean up before the network times out the request. Add a
3064 reference from the request import to the obd device and delay
3065 the cleanup until the network drops the request.
3068 Frequency : occasionally during client (re)connect
3070 Description: assertion failure during client (re)connect
3071 Details : processing a client connection request may be delayed by the
3072 client or server longer than the client connect timeout. This
3073 causes the client to resend the connection request. If the
3074 original connection request is replied in this interval, the
3075 client may trip an assertion failure in ptlrpc_connect_interpret()
3076 which thought it would be the only running connect process.
3079 Frequency : only with obd_echo servers and clients that are rebooted
3081 Description: kernel BUG accessing uninitialized data structure
3082 Details : When running an obd_echo server it did not start the ping_evictor
3083 thread, and when a client was evicted an uninitialized data
3084 structure was accessed. Start the ping_evictor in the RPC
3085 service startup instead of the OBD startup.
3087 Severity : enhancement
3088 Bugzilla : 10193 (patchless)
3089 Description: Remove dependency on various unexported kernel interfaces.
3090 Details : No longer need reparent_to_init, exit_mm, exit_files,
3091 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3094 Frequency : rare (only users of deprecated and unsupported LDAP config)
3096 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3097 Details : LDAP apparently contains 'lustreName' attributes instead of
3098 'name'. A simple remapping of the name is sufficient.
3101 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3103 Description: waiting_locks_callback trips kernel BUG if client is evicted
3104 Details : Running with the dump_on_timeout debug flag turned on makes
3105 it possible that the waiting_locks_callback() can try to dump
3106 the Lustre kernel debug logs from an interrupt handler. Defer
3107 this log dumping to the expired_lock_main() thread.
3109 Severity : enhancement
3111 Description: Support NFS exporting on 2.6 kernels.
3112 Details : Implement non-rawops metadata methods for NFS server to use without
3113 changing NFS server code.
3116 Frequency : very rare (synthetic metadata workload only)
3118 Description: two racing renames might cause an MDS thread to deadlock
3119 Details : Running the "racer" program may cause one MDS thread to rename
3120 a file from being the source of a rename to being the target of
3121 a rename at exactly the same time that another thread is doing
3122 so, and the second thread has already enqueued these locks after
3123 doing a lookup of the target and is trying to relock them in
3124 order. Ensure that we don't try to re-lock the same resource.
3127 Frequency : only very large systems with liblustre clients
3129 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3130 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3131 all OSTs in parallel. This allows the eviction of stale liblustre
3132 clients to proceed much faster than if they were done in series,
3133 and also offers similar improvements for other set_info RPCs.
3138 Description: excessive CPU usage during initial read phase on client
3139 Details : During the initial read phase on a client, it would agressively
3140 retry readahead on the file, consuming too much CPU and impacting
3141 performance (since 1.4.5.8). Improve the readahead algorithm
3142 to avoid this, and also improve some other common cases (read
3143 of small files in particular, where "small" is files smaller than
3144 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3149 Description: MDS crash when receiving packet with unknown intent.
3150 Details : Do not LBUG in unknown intent case, just return -EFAULT
3152 Severity : enhancement
3153 Bugzilla : 9293, 9385
3154 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3155 Details : Do not serialize getattr (non-intent version) and statfs.
3158 Frequency : occasional, when OST network is overloaded/intermittent
3160 Description: client evicted by OST after bulk IO timeout
3161 Details : If a client sends a bulk IO request (read or write) the OST
3162 may evict the client if it is unresposive to its data GET/PUT
3163 request. This is incorrect if the network is overloaded (takes
3164 too long to transfer the RPC data) or dropped the OST GET/PUT
3165 request. There is no need to evict the client at all, since
3166 the pinger and/or lock callbacks will handle this, and the
3167 client can restart the bulk request.
3170 Frequency : Always when mmapping file with no objects
3172 Description: client crashes when mmapping file with no objects
3173 Details : Check that we actually have objects in a file before doing any
3174 operations on objects in ll_vm_open, ll_vm_close and
3180 Description: Request leak when working with deleted CWD
3181 Details : Introduce advanced request refcount tracking for requests
3182 referenced from lustre intent.
3184 Severity : Enhancement
3186 Description: Cache open file handles on client.
3187 Details : MDS now will return special lock along with openhandle, if
3188 requested and client is allowed to hold openhandle, even if unused,
3189 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3190 closing files for every read/write openration.
3192 Severity : Enhancement
3194 Description: Cache open negative dentries on client when possible.
3195 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3196 negative dentries on lock revocation.
3201 Description: Remounting a client read-only wasn't possible with a zconf mount
3202 Details : It wasn't possible to remount a client read-only with llmount.
3204 Severity : enhancement
3205 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3206 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3207 MPICH is included to add Lustre-specific ADIO interfaces.
3208 This is based closely on the UFS ADIO layer and only differs
3209 in file creation, in order to allow the OST striping to be set.
3210 This is user-contributed code and not supported by CFS.
3215 Description: extended inode attributes (immutable, append-only) work improperly
3216 when 2.4 and 2.6 kernels are used on client/server or vice versa
3217 Details : Introduce kernel-independent values for these flags.
3219 Severity : enhancement
3222 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3223 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3224 granularity. Now, Fractional MB granularity is be supported,
3225 this is very useful for low memory system.
3227 Severity : enhancement
3229 Description: Getattr by fid
3230 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3231 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3232 have locks and avoids getattr rpc requests for it on every stat.
3235 Frequency : Always, for filesystems larger than 2TB
3237 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3238 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3239 filesystems larger than 2TB to be mounted with mballoc enabled.
3242 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3244 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3245 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3246 filesystems larger than 2TB is mounted then the kernel will
3247 silently corrupt the start of the filesystem. CONFIG_LBD is
3248 enabled for all CFS-supported kernels, but the possibility of
3249 this happening with a modified kernel config exists.
3251 Severity : enhancement
3253 Description: add client O_DIRECT support for 2.6 kernels
3254 Details : It is now possible to do O_DIRECT reads and writes to files
3255 in the Lustre client mountpoint on 2.6 kernel clients.
3257 Severity : enhancement
3259 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3260 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3261 parallel, not waiting for response from every OST before sending
3262 a rpc to the next OST.
3267 Description: setattr vs write race when updating file timestamps
3268 Details : Client processes that update a file timestamp into the past
3269 right after writing to the file (e.g. tar) it is possible that
3270 the updated file modification time can be reset to the current
3271 time due to a race between processing the setattr and write RPC.
3273 Severity : enhancement
3275 Description: Bring 'lfs find' closer in line with regular Linux find.
3276 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3277 -print0 options and obtains all the needed info through the lustre
3280 Severity : enhancement
3282 Description: support up to 1024 configured devices on one node
3283 Details : change obd_dev array from statically allocated to dynamically
3284 allocated structs as they are first used to reduce memory usage
3289 Description: Flush dirty partially truncated pages during truncate
3290 Details : Immediatelly flush partially truncated pages in filter_setattr,
3291 this way we completely avoid having any pages in page cache on OST
3292 and can retire ugly workarounds during writes to flush such pages.
3297 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3298 Details : protect inode from truncation within vfs_unlink() context
3299 just take a reference before calling vfs_unlink() and release it
3300 when parent's i_sem is free.
3305 Description: last_id value checked outside lock on OST caused LASSERT failure
3306 Details : If there were multiple MDS->OST object precreate requests in
3307 flight, it was possible that the OST's last object id was checked
3308 outside a lock and incorrectly tripped an assertion. Move checks
3309 inside locks, and discard old precreate requests.
3312 Frequency : always, if extents are used on OSTs
3314 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3315 Details : OSTs using the extents format would not zero the high 16 bits of
3316 the index physical block number. This is not a problem for any
3317 OST filesystems smaller than 16TB, and no kernels support ext3
3318 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3319 new/modified files) and can be fixed for existing filesystems
3320 with e2fsprogs-1.39-cfs1.
3325 Description: import connection selection may be incorrect if timer wraps
3326 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3327 import connections to be ignored if the 32-bit jiffies counter
3328 wraps. Use a 64-bit jiffies counter.
3331 Frequency : very large clusters immediately after boot
3333 Description: LNET request buffers exhausted under heavy short-term load
3334 Details : If a large number of client requests are generated on a service
3335 that has previously never seen so many requests it is possible
3336 that the request buffer growth cannot keep up with the spike in
3337 demand. Instead of dropping incoming requests, they are held in
3338 the LND until the RPC service can accept more requests.
3341 Frequency : Sometimes during replay
3343 Description: Assertion failure in ll_local_open after replay.
3344 Details : If replay happened on an open request reply before we were able
3345 to set replay handler, reply will become not swabbed tripping the
3346 assertion in ll_local_open. Now we set the handler right after
3347 recognising of open request
3350 Frequency : very rare
3352 Description: kernel reports "badness in vsnprintf"
3353 Details : Reading from the "recovery_status" /proc file in small chunks
3354 may cause a negative length in lprocfs_obd_rd_recovery_status()
3355 call to vsnprintf() (which is otherwise harmless). Exit early
3356 if there is no more space in the output buffer.
3358 Severity : enhancement
3360 Description: clear OBD RPC statistics by writing to them
3361 Details : It is now possible to clear the OBD RPC statistics by writing
3362 to the "stats" file.
3367 Description: Client mtime is not the same on different clients after utimes
3368 Details : In some cases, the client was using the utimes() syscall on
3369 a file cached on another node. The clients now validate the
3370 ctime from the MDS + OSTs to determine which one is right.
3375 Description: Inability to activate failout mode
3376 Details : lconf script incorrectly assumed that in pythong string's numeric
3377 value is used in comparisons.
3380 Frequency : always with multiple stripes per file
3382 Description: Inefficient object allocation for mutli-stripe files
3383 Details : When selecting which OSTs to stripe files over, for files with
3384 a stripe count that divides evenly into the number of OSTs,
3385 the MDS is always picking the same starting OST for each file.
3386 Return the OST selection heuristic to the original design.
3391 Description: mount failures may take full timeout to return an error
3392 Details : Under some heavy load conditions it is possible that a
3393 failed mount can wait for the full obd_timeout interval,
3394 possibly several minutes, before reporting an error.
3395 Instead return an error as soon as the status is known.
3397 Frequency : quota enabled and large files being deleted
3399 Description: releasing more than 4GB of quota at once hangs OST
3400 Details : If a user deletes more than 4GB of files on a single OST it
3401 will cause the OST to spin in an infinite loop. Release
3402 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3407 Description: statfs data retrieved from /proc may be stale or zero
3408 Details : When reading per-device statfs data from /proc, in the
3409 {kbytes,files}_{total,free,avail} files, it may appear
3410 as zero or be out of date.
3413 Frequency : systems with MD RAID1 external journal devices
3415 Description: lconf's call to blkid is confused by RAID1 journal devices
3416 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3417 of returning all block devices that match the journal UUID.
3420 Frequency : always, for aggregate stripe size over 4GB
3422 Description: assertion fails when trying to use 4GB stripe size
3423 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3424 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3427 Frequency : always on ppc64
3429 Description: the first write on an ext3 filesystem with mballoc got stuck
3430 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3431 perform endianness conversion.
3433 ------------------------------------------------------------------------------
3435 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3437 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3438 INTEROPERATE with older versions automatically. Please read the
3439 user documentation before upgrading any part of a live system.
3440 * WARNING: Lustre networking configuration changes are required with
3441 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3444 * Support for kernels:
3445 2.6.9-22.0.2.EL (RHEL 4)
3446 2.6.5-7.244 (SLES 9)
3447 2.6.12.6 vanilla (kernel.org)
3450 Severity : enhancement
3451 Bugzilla : 7981/8208
3452 Description: Introduced Lustre Networking (LNET)
3453 Details : LNET is new networking infrastructure for Lustre, it includes
3454 a reorganized network configuration mode (see the user
3455 documentation for full details) as well as support for routing
3456 between different network fabrics. Lustre Networking Devices
3457 (LNDs) for the supported network fabrics have also been
3458 created for this new infrastructure.
3460 Severity : enhancement
3461 Description: Introduced Access control lists
3462 Details : clients can set ACLs on files and directories in order to have
3463 more fine-grained permissions than the standard Unix UGO+RWX.
3464 The MDS must be started with the "-o acl" mount option.
3466 Severity : enhancement
3467 Description: Introduced filesystem quotas
3468 Details : Administrators may now establish per-user quotas on the
3471 Severity : enhancement
3473 Description: Configuration change for the XT3
3474 The PTLLND is now used to run Lustre over Portals on the XT3
3475 The configure option(s) --with-cray-portals are no longer used.
3476 Rather --with-portals=<path-to-portals-includes> is used to
3477 enable building on the XT3. In addition to enable XT3 specific
3478 features the option --enable-cray-xt3 must be used.
3483 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3484 Details : A race between allocating a new llcd and re-getting the llcd_lock
3485 allowed another thread to grab newly-allocated llcd.
3487 Severity : enhancement
3489 Description: 2.6 OST async journal commit and locking fix to improve performance
3490 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3491 2.6 kernels are now async as they already were in 2.4 kernels so
3492 that they can commit concurrently with the network bulk transfer.
3493 For block-allocated files the filter allocation semaphore is held
3494 to avoid filesystem fragmentation during allocation. BKL lock
3495 removed for 2.6 xattr operations where it is no longer needed.
3500 Description: lconf incorrectly determined whether two IP networks could talk
3501 Details : In some more complicated routing and multiple-network
3502 configurations, lconf will avoid trying to make a network
3503 connection to a disjoint part of the IP space. It was doing the
3504 math incorrectly for one set of cases.
3509 Description: Fix for potential infinite loop processing records in an llog.
3510 Details : If an llog record is corrupted/zeroed, it is possible to loop
3511 forever in llog_process(). Validate the llog record length
3512 and skip the remainder of the block on error.
3515 Frequency : occasional (liblustre only)
3517 Description: liblustre could not open files whose last component is a symlink
3518 Details : sysio_path_walk() would incorrectly pass the open intent to
3519 intermediate path components.
3522 Frequency : rare (liblustre only with non-standard tuning)
3523 Bugzilla : 7201 (7350)
3524 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3525 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3526 after it has been released, only do this for referenced locks
3528 Severity : enhancement
3530 Description: specifying an (invalid) directory default stripe_size of -1
3531 would reset the directory default striping
3532 Details : stripe_size -1 was used internally to signal directory stripe
3533 removal, now use "all default" to signal dir stripe removal
3534 as a directory striping of "all default" is not useful
3537 Frequency : common for large clusters running liblustre clients
3539 Description: doing an ls when liblustre clients are running is slow
3540 Details : sending a glimpse AST to a liblustre client waits for every AST
3541 to time out, as liblustre clients will not respond. Since they
3542 cannot cache data we refresh the OST lock LVB from disk instead.
3544 Severity : enhancement
3546 Description: doing an ls at the same time as file IO can be slow
3547 Details : enqueue and other "small" requests can be blocked behind many
3548 large IO requests. Create a new OST IO portal for non-IO
3549 requests so they can be processed faster.
3552 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3554 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3555 created file with mode 000
3556 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3557 allow an ftruncate() as a non-root user to a file with mode 000.
3558 Lustre now respects this flag to disable mode checks when
3559 truncating a file owned by the user
3562 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3565 Description: Revoking locks from clients that went dead or catatonic might take
3567 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3568 cancellation of such locks instant on servers without waiting for
3569 any reply from clients. Clients drops these locks when cancel
3570 notification from server is received without replying.
3573 Frequency : liblustre-only, when liblustre client dies or becomes busy
3575 Description: Doing ls on Linux clients can take a long time with active
3577 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3578 granting such locks to it in the first place if possible. Locks
3579 are taken by proxy on the OST during the read or write and
3580 dropped immediately afterward. Add connect flags handling, do
3581 not grant locks to liblustre clients for glimpse ASTs.
3583 Severity : enhancement
3585 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3586 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3587 sequential reads and behaves badly with applications doing random
3588 reads. Improve it by reading ahead at least read region, and
3589 avoiding excessive large RPC for small reads.
3591 Severity : enhancement
3593 Description: Creating more than 1000 files for a single job may cause a load
3594 imbalance on the OSTs if there are also a large number of OSTs.
3595 Details : qos_prep_create() uses an OST index reseed value that is an
3596 even multiple of the number of available OSTs so that if the
3597 reseed happens in the middle of the object allocation it will
3598 still utilize the OSTs as uniformly as possible.
3603 Description: OST or MDS may oops in ping_evictor_main()
3604 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3605 but doesn't restart at beginning of obd_exports_timed list
3608 Severity : enhancement
3610 Description: improve by-nid export eviction on the MDS and OST
3611 Details : allow multiple exports with the same NID to be evicted at one
3612 time without re-searching the exports list.
3615 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3617 Description: MDS may oops in groups_free()
3618 Details : in rare race conditions a newly allocated group_info struct is
3619 freed again, and this can be NULL. The 2.4 compatibility code
3620 for groups_free() checked for a NULL pointer, but 2.6 did not.
3623 Frequency : common for liblustre clients doing little filesystem IO
3624 Bugzilla : 9352, 7313
3625 Description: server may evict liblustre clients accessing contended locks
3626 Details : if a client is granted a lock or receives a completion AST
3627 with a blocking AST already set it would not reply to the AST
3628 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3631 Frequency : lfs setstripe, only systems with more than 160 OSTs
3633 Description: unable to set striping with a starting offset beyond OST 160
3634 Details : llapi_create_file() incorrectly limited the starting stripe
3635 index to the maximum single-file stripe count.
3638 Frequency : LDAP users only
3640 Description: lconf did not handle in-kernel recovery with LDAP properly
3641 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3643 Severity : enhancement
3645 Description: bind OST threads to NUMA nodes to improve performance
3646 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3647 node and do their allocations there to localize memory access
3649 Severity : enhancement
3651 Description: llmount can determine client NID directly from Myrinet (GM)
3652 Details : the client NID code from gmnalnid was moved directly into
3653 llmount, removing the need to use this or specifying the
3654 client NID explicitly when mounting GM clients with zeroconf
3657 Frequency : if client is started with down MDS
3659 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3660 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3661 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3662 ptlrpc_import_delayed_req() because we didn't check if the
3663 request was interrupted, and we also didn't break out of the
3664 event loop if there was no timeout
3669 Description: data loss during non-page-aligned writes to a single file from
3670 both multiple nodes and multiple threads on one node at same time
3671 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3672 inconsistency led to false short-reads, that were cached and later
3673 used by ->prepare_write() to fill in partially written page,
3674 leading to data loss.
3677 Frequency : always, if lconf --abort_recovery used
3679 Description: lconf --abort_recovery fails with 'Operation not supported'
3680 Details : lconf was attempting to abort recovery on the MDT device and not
3683 Severity : enhancement
3685 Description: remove cleanup logs
3686 Details : replace lconf-generated cleanup logs with lustre internal
3687 cleanup routines. Eliminates the need for client-cleanup and
3690 Severity : enhancement
3692 Description: add support for EAs (user and system) on lustre filesystems
3693 Details : it is now possible to store extended attributes in the Lustre
3694 client filesystem, and with the user_xattr mount option it
3695 is possible to allow users to store EAs on their files also
3697 Severity : enhancement
3699 Description: Add possibility (config option) to show minimal available OST free
3701 Details : When compiled with --enable-mindf configure option, statfs(2)
3702 (and so, df) will return least minimal free space available from
3703 all OSTs as amount of free space on FS, instead of summary of
3704 free spaces of all OSTs.
3706 Severity : enhancement
3708 Description: do not expand extent locks acquired on OST-side
3709 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3710 by server: they are not cached anyway.
3713 Frequency : when mmap is used/binaries executed from Lustre
3715 Description: Unmmap pages before throwing them away from read cache.
3716 Details : llap_shrink cache now attempts to unmap pages before discarding
3717 them (if unmapping failed - do not discard). SLES9 kernel has
3718 extra checks that trigger if this unmapping is not done first.
3723 Description: lconf didn't resolve symlinks before checking to see whether a
3724 given mountpoint was already in use
3727 Frequency : when migrating failover services
3728 Bugzilla : 6395, 9514
3729 Description: When migrating a subset of services from a node (e.g. failback
3730 from a failover service node) the remaining services would
3731 time out and evict clients.
3732 Details : lconf --force (implied by --failover) sets the global obd_timeout
3733 to 5 seconds in order to quickly disconnect, but this caused
3734 other RPCs to time out too quickly. Do not change the global
3735 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3737 Severity : enhancement
3738 Frequency : if MDS is started with down OST
3739 Bugzilla : 9439,5706
3740 Description: Allow startup/shutdown of an MDS without depending on the
3741 availability of the OSTs.
3742 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3743 Add appropriate locking and lov-osc refcounts for safe
3744 cleaning. Add osc abort_inflight calls in case the
3745 synchronize never started.
3748 Frequency : occasional (Cray XT3 only)
3750 Description: root not authorized to access files in CRAY_PORTALS environment
3751 Details : The client process capabilities were not honoured on the MDS in
3752 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3753 been accessed by an authorized user then root was able to access
3754 the file on the local client also. The root user capabilities
3755 are now allowed on the MDS, as this environment has secure UID.
3758 Frequency : occasional
3760 Description: ldiskfs "too long searching" message happens too often
3761 Details : A debugging message (otherwise harmless) prints too often on
3762 the OST console. This has been reduced to only happen when
3763 there are fragmentation problems on the filesystem.
3768 Description: Division by zero in statfs when all OSCs are inactive
3769 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3770 lov_statfs divides by value returned by lov_get_stripecnt().
3774 Bugzilla : 9489, 3273
3775 Description: First write from each client to each OST was only 4kB in size,
3776 to initialize client writeback cache, which caused sub-optimal
3777 RPCs and poor layout on disk for the first writen file.
3778 Details : Clients now request an initial cache grant at (re)connect time
3779 and so that they can start streaming writes to the cache right
3780 away and always do full-sized RPCs if there is enough data.
3781 If the OST is rebooted the client also re-establishes its grant
3782 so that client cached writes will be honoured under the grant.
3787 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3788 Details : Now I/O RPCs go to different portal number and (presumably) fast
3789 lock requests (and glimses) and other RPCs get their own service
3790 threads pool that should be able to service those RPCs
3793 Severity : enhancement
3795 Description: Ability to exchange lustre version between client and servers and
3796 issue warnings at client side if client is too old. Also for
3797 liblustre clients there is ability to refuse connection of too old
3799 Details : New 'version' field is added to connect data structure that is
3800 filled with version info. That info is later checked by server and
3804 Frequency : rare, liblustre only.
3805 Bugzilla : 9296, 9581
3806 Description: Two simultaneous writes from liblustre at offset within same page
3807 might proceed at the same time overwriting eachother with stale
3809 Details : I/O lock withing llu_file_prwv was released too early, before data
3810 actually was hitting the wire. Extended lock-holding time until
3811 server acknowledges receiving data.
3814 Frequency : extremely rare. Never observed in practice.
3816 Description: avoid generating lustre_handle cookie of 0.
3817 Details : class_handle_hash() generates handle cookies by incrementing
3818 global counter, and can hit 0 occasionaly (this is unlikely, but
3819 not impossible, because initial value of cookie counter is
3820 selected randonly). Value of 0 is used as a sentinel meaning
3821 "unassigned handle" --- avoid it. Also coalesce two critical
3822 sections in this function into one.
3824 Severity : enhancement
3826 Description: allow liblustre clients to delegate truncate locking to OST
3827 Details : To avoid overhead of locking, liblustre client instructs OST to
3828 take extent lock in ost_punch() on client's behalf. New connection
3829 flag is added to handle backward compatibility.
3831 Severity : enhancement
3832 Bugzilla : 4928, 7341, 9758
3833 Description: allow number of OST service threads to be specified
3834 Details : a module parameter allows the number of OST service threads
3835 to be specified via "options ost ost_num_threads={N}" in the
3836 OSS's /etc/modules.conf or /etc/modprobe.conf.
3840 Bugzilla : 6146, 9635, 9895
3841 Description: servers crash with bad pointer in target_handle_connect()
3842 Details : In rare cases when a client is reconnecting it was possible that
3843 the connection request was the last reference for that export.
3844 We would temporarily drop the export reference and get a new
3845 one, but this may have been the last reference and the export
3846 was just destroyed. Get new reference before dropping old one.
3848 Severity : enhancement
3849 Frequency : if client is started with failover MDS
3851 Description: Allow multiple MDS hostnames in the mount command
3852 Details : Try to read the configuration from all specified MDS
3853 hostnames during a client mount in case the "primary"
3856 Severity : enhancement
3858 Description: Stop sending data to evicted clients as soon as possible.
3859 Details : Check if the client we are about to send or are sending data to
3860 was evicted already. (Check is done every second of waiting,
3861 for which l_wait_event interface was extended to allow checking
3862 of exit condition at specified intervals).
3865 Frequency : rare, normally only when NFS exporting is done from client
3867 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3869 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3872 Frequency : always, if srand() is called before liblustre initialization
3874 Description: Liblustre uses system PRNG disturbing its usage by user application
3875 Details : Introduce internal to lustre fast and high-quality PRNG for
3876 lustre usage and make liblustre and some other places in generic
3877 lustre code to use it.
3879 Severity : enhancement
3880 Bugzilla : 9477, 9557, 9870
3881 Description: Verify that the MDS configuration logs are updated when xml is
3882 Details : Check if the .xml configuration logs are newer than the config
3883 logs stored on the MDS and report an error if this is the case.
3884 Request --write-conf, or allow starting with --old_conf.
3886 Severity : enhancement
3888 Description: Handle symlinks in the path when checking if Lustre is mounted.
3889 Details : Resolve intermediate symlinks when checking if a client has
3890 mounted a filesystem to avoid duplicate client mounts.
3895 Description: lconf can hit an error exception but still return success.
3896 Details : The lconf command catches the Command error exception at the top
3897 level script context and will exit with the associated exit
3898 status, but doesn't ensure that this exit status is non-zero.
3903 Description: failure of ptlrpc thread startup can cause oops
3904 Details : Starting a ptlrpc service thread can fail if there are a large
3905 number of threads or the server memory is very fragmented.
3906 Handle this without oopsing.
3909 Frequency : always, only if liblustre and non-default acceptor port was used
3911 Description: liblustre cannot connect to servers with non-default acceptor port
3912 Details : tcpnal_set_default_params() was not called and was therefore
3913 ignoring the environment varaible TCPNAL_PORT, as well as other
3914 TCPNAL_ environment variables
3919 Description: two objects could be created on the same OST for a single file
3920 Details : If an OST is down, in some cases it was possible to create two
3921 objects on a single OST for a single file. No problems other
3922 than potential performance impact and spurious error messages.
3926 Bugzilla : 5681, 9562
3927 Description: Client may oops in ll_unhash_aliases
3928 Details : Client dcache may become inconsistent in race condition.
3929 In some cases "getcwd" can fail if the current directory is
3935 Description: Inode refcounting problems in NFS export code
3936 Details : link_raw functions used to call d_instantiate without obtaining
3937 extra inode reference first.
3941 Bugzilla : 9942, 9903
3942 Description: Referencing freed requests leading to crash, memleaks with NFS.
3943 Details : We used to require that call to ll_revalidate_it was always
3944 followed by ll_lookup_it. Also with revalidate_special() it is
3945 possible to call ll_revalidate_it() twice for the same dentry
3946 even if first occurence returned success. This fix changes semantic
3947 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3948 reference on a request referred from the intent.
3949 ll_intent_release() then releases such a request.
3952 Frequency : rare, normally benchmark loads only
3954 Description: unlinked inodes were kept in memory on the client
3955 Details : If a client is repeatedly creating and unlinking files it
3956 can accumulate a lot of stale inodes in the inode slab cache.
3957 If there is no other client load running this can cause the
3958 client node to run out of memory. Instead flush old inodes
3959 from client cache that have the same inode number as a new inode.
3962 Frequency : SLES9 2.6.5 kernel and long filenames only
3963 Bugzilla : 9969, 10379
3964 Description: utime reports stale NFS file handle
3965 Details : SLES9 uses out-of-dentry names in some cases, which confused
3966 the lustre dentry revalidation. Change it to always use the
3970 Frequency : rare, unless heavy write-truncate concurrency is continuous
3971 Bugzilla : 4180, 6984, 7171, 9963, 9331
3972 Description: OST becomes very slow and/or deadlocked during object unlink
3973 Details : filter_destroy() was holding onto the parent directory lock
3974 while truncating+unlinking objects. For very large objects this
3975 may block other threads for a long time and slow overall OST
3976 responsiveness. It may also be possible to get a lock ordering
3977 deadlock in this case, or run out of journal credits because of
3978 the combined truncate+unlink. Solution is to do object truncate
3979 first in one transaction without parent lock, and then do the
3980 final unlink in a new transaction with the parent lock. This
3981 reduces the lock hold time dramatically.
3984 Frequency : rare, 2.4 kernels only
3986 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3987 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3988 before unmounting their filesystem in order to avoid deadlock.
3989 The kernel_locked() function in 2.4 kernels only checks whether
3990 the kernel lock is held, not whether it is this process that is
3991 holding it as 2.6 kernels do.
3996 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3997 Details : The client ptlrpc code may be trying to reconnect to a down
3998 server before a previous connection attempt has timed out.
3999 Increase the reconnect interval to be longer than the connection
4000 timeout interval to avoid sending duplicate connections to
4004 Frequency : echo_client brw_test command
4006 Description: fix echo_client to work with OST preallocated code
4007 Details : OST preallocation code (5137) didn't take echo_client IO path
4008 into account: echo_client calls filter methods outside of any
4009 OST thread and, hence, there is no per-thread preallocated
4010 pages and buffers to use. Solution: hijack pga pages for IO. As
4011 a byproduct, this avoids unnecessary data copying.
4015 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4016 Description: Client can oops in mdc_commit_close() after open replay
4017 Details : It was possible for the MDS to return an open request with no
4018 transaction number in mds_finish_transno() if the client was
4019 evicted, but without actually returning an error. Clients
4020 would later try to replay that open and may trip an assertion
4021 Simplify the client close codepath, and always return an error
4022 from the MDS in case the open is not successful.
4025 Frequency : rare, 2.6 OSTs only
4027 Description: OST may deadlock under high load on fragmented files
4028 Details : If there was a heavy load and highly-fragmented OST filesystems
4029 it was possible to have all the OST threads deadlock waiting on
4030 allocation of biovecs, because the biovecs were not released
4031 until the entire RPC IO was completed. Instead, release biovecs
4032 as soon as they are complete to ensure forward IO progress.
4034 Severity : enhancement
4036 Description: Support for specifying external journal device at mount
4037 Details : If an OST or MDS device is formatted with an external journal
4038 device, this device major/minor is stored in the ext3 superblock
4039 and may not be valid for failover. Allow detecting and
4040 specifying the external journal at mount time.
4045 Description: Mounting an MDS with pending unlinked files may cause oops
4046 Details : target_finish_recovery() calls mds_postrecov() which returned
4047 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4048 considers this an error and immediately begins cleaning up the
4049 lov, just after starting the mds_lov process
4051 Severity : enhancement
4053 Description: Implement 'lfs df' to report actual free space on per-OST basis
4054 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4055 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4056 usage of objects; '-h' to report in human readable format.
4058 ------------------------------------------------------------------------------
4060 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4067 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4068 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4069 at startup during orphan unlinking. Instead, skip buddy update
4070 before setup, as it will be regenerated after recovery is complete.
4075 Description: If an OST is inactive, its locks might reference stale inodes.
4076 Details : lov_change_cbdata() must iterate over all namespaces, even if
4077 they are inactive to clear inode references from the lock.
4079 Severity : enhancement
4080 Frequency : occasional, if non-standard max_dirty_mb used
4082 Description: Client will block write RPCs if not enough grant
4083 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4084 then the client will block writes while waiting for another RPC
4085 to complete instead of consuming its dirty limit. With change
4086 we get improved performance when max_dirty_mb is small.
4088 Severity : enhancement
4089 Bugzilla : 3389, 6253
4090 Description: Add support for supplementary groups on the MDS.
4091 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4092 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4093 lookups for user supplementary groups into a cache.
4097 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4098 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4099 for liblustre O_EXCL is enforced by MDS.
4103 Description: readdir never returns NULL in liblustre.
4104 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4105 dentry in struct dirent.
4109 Frequency : liblustre only, depends on application IO pattern
4110 Description: liblustre clients evicted if not contacting servers
4111 Details : Don't put liblustre clients into the ping_evictor list, so
4112 they will not be evicted by the pinger ever.
4114 Severity : enhancement
4116 Description: Add ability to evict clients by NID from MDS.
4117 Details : By echoing "nid:$NID" string into
4118 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4119 $NID would be instantly evicted from this MDS and from all active
4120 OSTs connected to it.
4124 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4125 Details : lookup_it_finish() used to query file size from OSTs that was not
4130 Description: service threads change working directory to that of init
4131 Details : Starting lustre service threads may pin the working directory
4132 of the parent thread, making that filesystem busy. Threads
4133 now change to the working directory of init to avoid this.
4137 Frequency : during shutdown only
4138 Description: shutdown with a failed MDS or OST can cause unmount to hang
4139 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4144 Frequency : 2.6 only
4145 Description: chmod/chown may include an extra supplementary group
4146 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4147 supplementary group and if none is specified this is used.
4150 Bugzilla : 5479 (6816)
4152 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4153 Details : If lookup is in progress on a file that is unlinked we might try
4154 to revalidate the inode and fail in revalidate after lookup is
4155 complete and ll_file_open() enqueues the open again but
4156 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4159 Frequency : always, if lconf --abort_recovery used
4161 Description: lconf --abort_recovery fails with 'Operation not supported'
4162 Details : lconf was attempting to abort recovery on the MDT device and not
4165 ------------------------------------------------------------------------------
4167 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4172 Frequency : rare (only unsupported configurations with a node running as an
4174 Bugzilla : 6514, 5137
4175 Description: Mounting a Lustre file system on a node running as an OST could
4177 Details : OSTs now preallocates memory needed to write out data at
4178 startup, instead of when needed, to avoid having to
4179 allocate memory in possibly low memory situations.
4180 Specifically, if the file system is mounted on on OST,
4181 memory pressure could force it to try to write out data,
4182 which it needed to allocate memory to do. Due to the low
4183 memory, it would be unable to do so and the node would
4184 become unresponsive.
4186 Severity : enhancement
4188 Description: Addition of lconf --service command line option
4189 Details : lconf now accepts a '--service <arg>' option, which is
4190 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4192 Severity : enhancement
4194 Description: Failover mode is now the default for OSTs.
4195 Details : By default, OSTs will now run in failover mode. To return to
4196 the old behaviour, add '--failout' to the lmc line for OSTs.
4198 Severity : enhancement
4200 Description: Health checks are now provided for MDS and OSTs
4201 Details : Additional detailed health check information on MSD and OSTs
4202 is now provided through the procfs health_check value.
4205 Frequency : occasional, depends on IO load
4207 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4208 after numerous create/delete cycles
4209 Details : The ext3 inode allocation policy would not allocate new inodes
4210 very well on the OSTs because there are no new directories
4211 being created. Instead we look for groups with free space if
4212 the parent directories are nearly full.
4217 Description: Network or server problems during mount may cause partially
4218 mounted clients instead of returning an error.
4219 Details : The config llog parsing code may overwrite the error return
4220 code during mount error handling, returning success instead
4226 Description: MDS can fail to allocate large reply buffers
4227 Details : After long uptimes the MDS can fail to allocate large reply
4228 buffers (e.g. zconf client mount config records) due to memory
4229 fragmentation or consumption by the buffer cache. Preallocate
4230 some large reply buffers so that these replies can be sent even
4231 under memory pressure.
4235 Frequency : rare (liblustre)
4236 Description: fsx running with liblustre complained that using truncate() to
4237 extend the file doesn't work. This patch corrects that issue.
4238 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4239 ATTR_SIZE and lsm use in llu_setattr_raw.
4243 Frequency : rare, only 2.6 kernels
4244 Description: Unusual file access patterns on the MDS may result in inode
4245 data being lost in very rare circumstances.
4246 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4247 code in the 2.6 kernel caused the inode and/or EA data not to
4248 be read from disk, causing single-file corruption.
4252 Frequency : rare, only 2.6 filesystems using extents
4253 Description: Heavy concurrent write and delete load may cause data corruption.
4254 Details : It was possible under high-load situations to have an extent
4255 metadata block in the block device cache from a just-unlinked
4256 file overwrite a newly-allocated data block. We now unmap any
4257 metadata buffers that alias just-allocated data blocks.
4261 Frequency : filesystems with default stripe_count larger than 77
4262 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4263 Details : lconf specifies an inode size of 4096 bytes when the default
4264 stripe_count is larger than 77. This conflicts with the default
4265 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4266 this case to avoid pinning too much memory for large EAs.
4268 ------------------------------------------------------------------------------
4270 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4275 Frequency : rare (extremely heavy IO load with hundreds of clients)
4277 Description: Client is evicted, gets IO error writing to file
4278 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4279 caused clients to be evicted for AST timeouts. The fixes in
4280 bug 5192 mean we no longer need to have such short AST timeouts
4281 so ldlm_timeout has been increased.
4284 Frequency : occasional during --force or --failover shutdown under load
4285 Bugzilla : 5949, 4834
4286 Description: Server oops/LBUG if stopped with --force or --failover under load
4287 Details : a collection of import/export refcount and cleanup ordering
4288 issues fixed for safer force cleanup
4291 Frequency : only filesystems larger than 120 OSTs
4292 Bugzilla : 5990, 6223
4293 Description: lfs getstripe would oops on a very large filesystem
4294 Details : lov_getconfig used kfree on vmalloc'd memory
4297 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4298 Bugzilla : 6242, 6243
4299 Description: reading from files that had been truncated to a non-zero size
4300 but never opened returned no data
4301 Details : ll_file_read() reads zeros from no-object files to EOF
4306 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4307 Details : There's little harm in aborting MDS/OSS recovery and letting it
4308 try again, so I removed the LASSERT and return an error instead.
4310 Severity : enhancement
4312 Description: New debugging infrastructure for tracking down data corruption
4313 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4314 (b) cover more of the client-side code path, and (c) try to narrow
4315 down where problems occurred
4319 Bugzilla : 3819, 4364, 4397, 6313
4320 Description: Racing close and eviction MDS could cause assertion in mds_close
4321 Details : It was possible to get multiple mfd references during close and
4322 client eviction, leading to one thread referencing a freed mfd.
4324 Severity: : enhancement
4325 Bugzilla : 3262, 6359
4326 Description: Attempts to reconnect to servers are now more aggressive.
4327 Details : This builds on the enhanced upcall-less recovery that was added
4328 in 1.4.2. When trying to reconnect to servers, clients will
4329 now try each server in the failover group every 10 seconds. By
4330 default, clients would previously try one server every 25 seconds.
4335 Description: After recovery, certain operations trigger a failed
4336 assertion on a client.
4337 Details : Failing over an mds, using lconf -d --failover, while a
4338 client was doing a readdir() call would cause the client to
4339 LBUG after recovery completed and the readdir() was resent.
4341 Severity : enhancement
4343 Description: Default groups are now added by lconf
4344 Details : You can now run lconf --group <servicename> without having to
4345 manually add groups with lmc.
4348 Frequency : occasional
4350 Description: Nodes with an elan id of 0 trigger a failed assertion
4353 Frequency : always when accessing e.g. tty/console device nodes
4355 Description: tty and some other devices nodes cannot be used on lustre
4356 Details : file's private_data field is used by device data and lustre
4357 values in there got lost. New field was added to struct file to
4358 store fs-specific private data.
4361 Frequency : when exporting Lustre via NFS
4363 Description: NFSD failed occasionally when looking up a path component
4364 Details : NFSD is looking up ".." which was broken in ext3 directories
4365 that had grown large enough to become hashed.
4368 Frequency : Clusters with multiple interfaces not on the same subnet
4370 Description: Nodes will repeatedly try to reconnect to an interface which it
4371 cannot reach and report an error to the log.
4372 Details : Extra peer list entries will be created by lconf with some peers
4373 unreachable. lconf now validates the peer before adding it.
4376 Frequency : Only if a default stripe is set on the filesystem root.
4378 Description: Setting a default stripe on the filesystem root prevented the
4379 filesystem from being remounted.
4380 Details : The client was sending extra request flags in the root getattr
4381 request and did not allocate a reply buffer for the dir EA.
4384 Frequency : occasional, higher if lots of files are accessed by one client
4385 Bugzilla : 6159, 6097
4386 Description: Client trips assertion regarding lsm mismatch/magic
4387 Details : While revalidating inodes the VFS looks up inodes with ifind()
4388 and in rare cases can find an inode that is being freed.
4389 The ll_test_inode() code will free the lsm during ifind()
4390 when it finds an existing inode and then the VFS later attaches
4391 this free lsm to a new inode.
4395 Bugzilla : 6422, 7030
4396 Description: MDS deadlock between mkdir and client eviction
4397 Details : Creating a new file via mkdir or mknod (starting a transaction
4398 and getting the ns lock) can deadlock with client eviction
4399 (gets ns lock and trying to finish a synchronous transaction).
4402 Frequency : occasional
4403 Description: While starting a server, the fsfilt_ext3 module could not be
4405 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4406 kernels. Previously, lconf would still use the ext3 name
4407 when trying to load modules. Now, it will correctly use
4408 ext3 on 2.4 and ldiskfs on 2.6.
4410 Severity : enhancement
4411 Description: The default stripe count has been changed to 1
4412 Details : The interpretation of the default stripe count (0, to lfs
4413 or lmc) has been changed to mean striping across a single
4414 OST, rather than all available. For general usage we have
4415 found a stripe count of 1 or 2 works best.
4417 Severity : enhancement
4418 Description: Add support for compiling against Cray portals.
4419 Details : Conditional compiling for some areas that are different
4423 Frequency : occasional
4424 Bugzilla : 6409, 6834
4425 Description: Creating files with an explicit stripe count may lead to
4426 a failed assertion on the MDS
4427 Details : If some OSTs are full or unavailable, creating files may
4428 trigger a failed assertion on the MDS. Now, Lustre will
4429 try to use other servers or return an error to the
4433 Frequency : occasional
4435 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4436 caused lock timeout during readahead (since 1.4.2).
4437 Details : Processes doing readahead might match a lock that hasn't been
4438 granted yet if there are overlapping and conflicting lock
4439 requests. The readahead process waits on ungranted lock
4440 (original lock is CBPENDING), while OST waits for that process
4441 to cancel CBPENDING read lock and eventually evicts client.
4443 Severity : enhancement
4445 Description: Initial enabling of flock support for clients
4446 Details : Implements fcntl advisory locking and file status functions.
4447 This feature is provided as an optional mount flag (default
4448 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4449 locking are implemented yet, and those that are are not guaranteed
4450 to be completely correct in production environments.
4451 mount -t lustre -o [flock|noflock] ...
4454 Frequency : occasional
4456 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4457 assertion in the ext3 JBD (journaling) layer.
4458 Details : The b_committed_data struct is protected by the big kernel lock
4459 in 2.4 kernels, serializing journal_commit_transaction() and
4460 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4461 access to this struct. In 2.6 kernels there is finer grained
4462 locking to improve SMP performance of the JBD layer.
4466 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4471 Description: merge and cleanup kernel patches.
4474 -----------------------------------------------------------------------------
4476 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4478 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4479 versions of Lustre. Please update all servers and clients to
4480 version 1.4.2 or later at the same time. You must also run
4481 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4482 to update the configuration logs.
4484 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4485 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4486 - swab llog records of type '0' so we get proper header size/idx (5861)
4487 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4488 - fix rename of one directory over another leaking an inode (5953)
4489 - avoid SetPageDirty on 2.6 (5981)
4490 - don't re-add just-being-destroyed locks to the waiting list (5653)
4491 - when creating new directories, inherit the parent's custom
4492 striping settings if present parent (3048)
4493 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4494 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4495 - don't hold client locks on temporary worklist from l_lru (5666)
4496 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4497 - automatically evict dead clients (5921)
4498 - Update file size properly in create+truncate+fstat case (6196)
4499 - Do not unhash mountpoint dentries, do not allow removal of
4501 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4502 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4503 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4504 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4505 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4506 - Correctly handle reads of files with no objects (6243)
4507 - lctl recover will also mark a device active if deactivate used (5933)
4509 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4510 - allow --write-conf on an MDS with different nettype than client (5619)
4511 - don't write config llogs to MDS for mounts not from that MDS (5617)
4512 - lconf should create multiple TCP connections from a client (5201)
4513 - init scripts are now turned off by default; run chkconfig --on
4514 lustre and chkconfig --on lustrefs to use them
4515 - upcalls are no longer needed for clients to recover to failover
4517 - add --abort-recovery option to lconf to abort recovery on device
4519 - add support for an arbitrary number of OSTs (3026)
4520 - Quota support protocol changes.
4521 - forward compatibility changes to wire structs (6007)
4522 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4523 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4524 - improved "lustre status" script
4525 - initialize blocksize for non-regular files (6062)
4526 - added --disable-server and --disable-client configure options (5782)
4527 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4528 - Vanilla 2.4.29 support
4529 - increase maximum number of obd devices to 520 (6242)
4530 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4531 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4532 SLES 9 2.6.5 kernels
4533 - limit stripes per file to 160 (the maximum EA size) (6093)
4535 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4538 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4539 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4540 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4541 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4542 - small ext3 extents cleanups and fixes (5733)
4543 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4544 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4545 - avoid needless client->OST connect, fix handle mismatch (5317)
4546 - fix DLM error path that led to out-of-sync client, long delays (5779)
4547 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4548 - fix several locking issues related to i_size (5492,5624,5654,5672)
4549 - don't move pending lock onto export if it is already evicted (5683)
4550 - fix kernel oops when creating .foo in unlinked directory (5548)
4551 - fix deadlock in obdfilter statistics vs. object create (5811)
4552 - use time_{before,after} to avoid timer jiffies wrap (5882)
4553 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4554 - Do not leak request if server was not able to process it (5154)
4555 - If mds_open unable to find parent dir, make that negative lookup(5154)
4556 - don't create new directories with extent-mapping (5909, 5936)
4558 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4559 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4560 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4561 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4562 - support for RHEL4 kernel on i686 (5773)
4563 - provide error messages when incompatible logs are encountered (5898)
4565 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4566 * version 1.4.0.10 (1.4.1 release candidate 1)
4568 - don't keep a lock reference when lock is not granted (4238)
4569 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4570 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4571 - drop import inflight refcount on signal_completed_replay error (5255)
4572 - unlock page after async write error during send (3677)
4573 - handle missing objects in filter_preprw_read properly (5265)
4574 - no transno return for symlink open, don't save no-trasno open (3440)
4575 - don't try to complete elan receive that already failed (4012)
4576 - free RPC server reply state on error (5406)
4577 - clean up thread from ptlrpc_start_thread() on error (5160)
4578 - readahead could read extra page into cache that wasn't ejected (5388)
4579 - prevent races in class_attach/setup/cleanup/detach (5260)
4580 - don't dereference de->d_inode after l_dput of de (5458)
4581 - use "int" for stripe value returned from lock_to_stripe (5544)
4582 - mballoc allocation and error-checking fixes in 2.6 (5504)
4583 - block device patches to fix I/O request sizes in 2.6 (5482)
4584 - look up hostnames for IB nals (5602)
4585 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4586 - don't start multiple acceptors for the same port (5277)
4587 - fix incorrect LASSERT in mds_getattr_name (5635)
4588 - export a proc file for general "ping" checking (5628)
4589 - fix "lfs check" to not block when the MDS is down (5628)
4591 - service request history (4965)
4592 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4593 - create an "evict_client" /proc entry on OSTs, like the MDS has
4594 - fix mount usage message, return errors per mount(8) (5168)
4595 - change grep [] to grep "[]" in tests so they work in more UMLs
4596 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4597 - remove ancient LOV_MAGIC_V0 EA support (5047)
4598 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4599 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4600 - allow readahead allocations to fail when low on memory (5383)
4601 - mmap locking landed again, after considerable improvement (2828)
4602 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4604 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4607 - send OST transaction number in read/write reply to free req (4966)
4608 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4609 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4610 - the watchdog thread now runs as interruptible (5246)
4611 - flock/lockf fixes (but it's still disabled, pending 5135)
4612 - don't use EXT3 constants in llite code (5094)
4613 - memory shortage at startup could cause assertion (5176)
4615 - reorganization of lov code
4616 - single portals codebase
4618 - add extents/mballoc support (5025)
4619 - direct I/O reads in the obdfilter (4048)
4620 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4622 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4625 - send OST transaction number in read/write reply to free req (4966)
4626 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4627 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4628 - the watchdog thread now runs as interruptible (5246)
4629 - handle missing objects in filter_preprw_read properly (5265)
4630 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4631 - drop import inflight refcount on signal_completed_replay error (5255)
4632 - unlock page after async write error during send (3677)
4633 - return original error code on reconstructed replies (3761)
4634 - no transno return for symlink open, don't save no-trasno open (3440)
4636 - add pid to ldlm debugging output (4922)
4637 - bump the watchdog timeouts -- we can't handle 30sec yet
4638 - extra debugging for orphan dentry/inode bug (5259)
4640 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4643 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4644 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4645 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4646 - fix NULL dereference in /proc/sys/portals/routes (4827)
4647 - allow failed mdc_close() operations to be interrupted (4561)
4648 - stop precreate on OST before MDS would time out on it (4778)
4649 - don't send partial-page writes before EOF from client (4410)
4650 - discard client grant for sub-page writes on large-page clients (4520)
4651 - don't free dentries not owned by NFS code, check generation (4806)
4652 - fix lsm leak if mds_create_objects() fails (4801)
4653 - limit debug_daemon file size, always print CERROR messages (4789)
4654 - use transno after validating reply (3892)
4655 - process timed out requests if import state changes (3754)
4656 - update mtime on OST during writes, return in glimpse (4829)
4657 - add mkfsoptions to LDAP (4679)
4658 - use ->max_readahead method instead of zapping global ra (5039)
4659 - don't interrupt __l_wait_event() during strace
4661 - add software watchdogs to catch hung threads quickly (4941)
4662 - make lustrefs init script start after nfs is mounted
4663 - fix CWARN/ERROR duplication (4930)
4664 - return async write errors to application if possible (2248)
4665 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4666 - print NAL number in %x format (4645)
4667 - update barely-supported suse-2.4.21-171 series (4842)
4668 - support for sles 9 %post scripts
4669 - support for building 2.6 kernel-source packages
4670 - support for sles km_* packages
4672 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4675 - ignore -ENOENT errors in osc_destroy (3639)
4676 - notify osc create thread that OSC is being cleaned up (4600)
4677 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4678 - reconstruct ost_handle() like mds_handle() (4657)
4679 - create a new thread to do import eviction to avoid deadlock (3969)
4680 - let lconf resolve symlinked-to devices (4629)
4681 - don't unlink "objects" from directory with default EA (4554)
4682 - hold socknal file ref over connect in case target is down (4394)
4683 - allow more than 32000 subdirectories in a single directory (3244)
4684 - fix blocks count for O_DIRECT writes (3751)
4685 - OST returns ENOSPC from object create when no space left (4539)
4686 - don't send truncate RPC if file size isn't changing (4410)
4687 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4688 - bind to privileged port in socknal and tcpnal (3689)
4690 - rate limit CERROR/CWARN console message to avoid overload (4519)
4691 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4692 - basic mmap support (3918)
4693 - kernel patch series update from b1_4 (4711)
4695 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4698 - avoid crash during MDS cleanup with OST shut down (2775)
4699 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4700 - don't use bad inodes on the MDS (3744)
4701 - dynamic object preallocation to improve recovery speed (4236)
4702 - don't hold spinlock over lock dumping or change debug flags (4401)
4703 - don't zero obd_dev when it is force cleaned (3651)
4704 - print grants to console if they go negative (4431)
4705 - "lctl deactivate" will stop automatic recovery attempts (3406)
4706 - look for existing locks in ldlm_handle_enqueue() (3764)
4707 - don't resolve lock handle twice in recovery avoiding race (4401)
4708 - revalidate should check working dir is a directory (4134)
4710 - don't always mark "slow" obdfilter messages as errors (4418)
4712 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4715 - don't close LustreDB during write_conf until it is done (3860)
4716 - fix typo in lconf for_each_profile (3821)
4717 - allow dumping logs from multiple threads at one time (3820)
4718 - don't allow multiple threads in OSC recovery (3812)
4719 - fix debug_size parameters (3864)
4720 - fix mds_postrecov to initialize import for llog ctxt (3121)
4721 - replace config semaphore with spinlock (3306)
4722 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4723 - don't allow enqueue to complete on a destroyed export (3822)
4724 - down write_lock before checking llog header bitmap (3825)
4725 - recover from lock replay timeout (3764)
4726 - up llog sem before sending rpc (3652)
4727 - reduce ns lock hold times when setting kms (3267)
4728 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4729 - fix NULL deref and obd_dev leak on setup error (3312)
4730 - replace some LBUG about llog ops with error handling (3841)
4731 - don't match INVALID dentries from d_lookup and spin (3784)
4732 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4733 - fix invalid assertion in ptlrpc_set_wait (3880)
4735 - add libwrap support for the TCP acceptor (3996)
4736 - add /proc/sys/portals/routes for non-root route listing (3994)
4737 - allow setting MDS UUID in .xml (2580)
4738 - print the stack of a process that LBUGs (4228)
4740 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4743 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4744 - make sure to unset replay flag from failed open requests (3440)
4745 - if default stripe count is 0, use OST count for inode size (3636)
4746 - update parent mtime/ctime on client for create/unlink (2611)
4747 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4748 - free recovery state on server during a forced cleanup (3571)
4749 - unregister_reply for resent reqs (3063)
4750 - loop back devices mounting and status check on 2.6 (3563)
4751 - fix resource-creation race that can provoke i_size == 0 (3513)
4752 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4753 - more debugging for page-accounting assertion (3746)
4754 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4755 - avoid deadlock after precreation failure (3758)
4756 - fix race and lock order deadlock in orphan handling (3450, 3750)
4757 - add validity checks when grabbing inodes from l_ast_data (3599)
4759 - add /proc/.../recovery_status to obdfilter (3428)
4760 - lightweight CDEBUG infrastructure, debug daemon (3668)
4761 - change default OSC RPC parameters to be better on small clusters
4762 - turn off OST read cache for files smaller than 32MB
4763 - install man pages and include them in rpms (3100)
4764 - add new init script for (un)mounting lustre filesystems (2593)
4765 - run chkconfig in %post for init scripts (3701)
4766 - drop scimac NAL (unmaintained)
4768 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4771 - clean kiobufs before and after use (3485)
4772 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4773 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4774 - change init script to fail more clearly if not run as root (1528)
4775 - allow clients to reconnect during replay (1742)
4776 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4777 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4778 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4779 - fix cond_resched (3554)
4780 - only evict unfinished clients after recovery (3515)
4781 - allow bulk resend, prevent data loss (3570)
4782 - dynamic ptlrpc request buffer allocation (2102)
4783 - don't allow unlinking open directory if it isn't empty (2904)
4784 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4785 - remove extraneous obd dereference causing LASSERT failure (3334)
4786 - don't use get_cycles() when creating temp. files on the mds (3156)
4787 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4788 - handle EEXIST for set-stripe, set proper directory name (3336)
4790 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4791 - fix ksocknal_fmb_callback() error messages (2918)
4793 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4796 - don't copy lvb into (possibly NULL) reply on error (2983)
4797 - don't deref dentry after dput, don't free lvb on error (2922)
4798 - use the kms to determine writeback rpc length (2947)
4799 - increment oti_logcookies when osc is inactive (2948)
4800 - update client's i_blocks count via lvb messages (2543)
4801 - handle intent open/close of special files properly (1557)
4802 - mount MDS with errors=remount-ro, like obdfilter (2009)
4803 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4804 - don't use cancelling-locks' kms values (2947)
4805 - use highest lock extent for kms, not last one (2925)
4806 - don't dereference ERR_PTR() dentry in error handling path (3107)
4807 - fix thread race in portals_debug_dumplog() (3122)
4808 - create lprocfs device entries at setup instead of at attach (1519)
4809 - common AST error handler, don't evict client on completion race (3145)
4810 - zero nameidata in detach_mnt in 2.6 (3118)
4811 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4812 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4813 - initialize RPC timeout timer earlier for 2.6 (3219)
4814 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4815 - print nal/nid for unknown nid (3258)
4816 - additional checks for oscc recovery before doing precreate (3284)
4817 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4818 - don't crash in mdc_close for bad permissions on open (3285)
4819 - zero i_rdev for non-device files (3147)
4820 - clear page->private before handing to FS, better assertion (3119)
4821 - tune the read pipeline (3236)
4822 - fix incorrect decref of invalidated dentry (2350)
4823 - provide read-ahead stats and refine rpc in flight stats (3328)
4824 - don't hold journal transaction open across create RPC (3313)
4825 - update atime on MDS at close time (3265)
4826 - close LDAP connection when recovering to avoid server load (3315)
4827 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4828 - don't leak open file on MDS after open resend (3325)
4829 - serialize filter_precreate and filter_destroy_precreated (3329)
4830 - loop device shouldn't call sync_dev() for nul device (3092)
4831 - clear page cache after eviction (2766)
4832 - resynchronize MDS->OST in background (2824)
4833 - refuse to mount the same filesystem twice on same mountpoint (3394)
4834 - allow llmount to create routes for mounting behind routers (3320)
4835 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4836 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4837 - fix rare problem with rename on htree directories (3417)
4839 - allow default OST striping configuration per directory (1414)
4840 - fix compilation for qswnal for 2.6 kernels (3125)
4841 - increase maximum number of MDS request buffers for large systems
4842 - change liblustreapi to be useful for external progs like lfsck (3098)
4843 - increase local configuration timeout for slow disks (3353)
4844 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4846 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4849 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4850 - don't overwrite extent policy data in reply if lock was blocked (2901)
4851 - drop filter export grants atomically with removal from device (2663)
4852 - del obd_self_export from work_list in class_disconnect_exports (2908)
4853 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4854 - swab reply message in mdc_close, other PPC fixes (2464)
4855 - fix destroying of named logs (2325)
4856 - overwrite old logs when running lconf --write_conf (2264)
4857 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4858 - fix race in target_handle_connect (2898)
4859 - mds_reint_create() should take same inode create lock (2926)
4860 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4861 - don't close files for self_export to avoid uninitialized obd (2936)
4862 - allow MDS with the same name as client node (2939)
4863 - hold dentry reference for closed log files for unlink (2325)
4864 - reserve space for all logs during transactions (2059)
4865 - don't evict page beyond end of stripe extent (2925)
4866 - don't oops on a deleted current working directory (2399)
4867 - handle hard links to targets without a parent properly (2517)
4868 - don't dereference NULL lock when racing during eviction (2867)
4869 - don't grow lock extents when lots of conflicting locks (2919)
4871 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4874 - account for cache space usage on clients to avoid data loss (974)
4875 - lfsck support in lustre kernel code (2349)
4876 - reduce journal credits needed for BRW writes (2370)
4877 - orphan handling to avoid losing space on client/server crashes
4878 - ptlrpcd can be blocked, stopping ALL progress (2477)
4879 - use lock value blocks to assist in proper KMS, faster stat (1021)
4880 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4881 - recovery for initial connections (2355)
4882 - fixes for mds_cleanup_orphans (1934)
4883 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4884 - block all file creations until orphan recovery completes (1901)
4885 - client remove rq_connection from request struct (2423)
4886 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4887 - recovery timer race (2670)
4888 - mdc_close recovey bug (2532)
4889 - ptlrpc cleanup bug (2710)
4890 - mds timeout on local locks (2588)
4891 - namespace lock held during RPCs (2431)
4892 - handle interrupted sync write properly (2503)
4893 - don't try to handle a message that hasn't been replied to (2699)
4894 - client assert failure during cleanup after abort recovery (2701)
4895 - leak mdc device after failed mount (2712)
4896 - ptlrpc_check_set allows timedout requests to complete (2714)
4897 - wait for inflight reqs when ptlrpcd finishes (2710)
4898 - make sure unregistered services are removed from the srv_list
4899 - reset bulk XID's when resending them (caught by 1138 test)
4900 - unregister_bulk after timeout
4901 - fix lconf error (2694)
4902 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4903 - readahead locks pages, leaves pending causing memory pressure (2673)
4904 - increase OST request buffers to 4096 on large machines (2729)
4905 - fix up permission of existing directories in simple_mkdir (2661)
4906 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4907 - don't assign transno to errored transactions (2742)
4908 - don't delete objects on OST if given a bogus objid from MDS (2751)
4909 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4910 - if rq_no_resend, then timeout request after recovery (2432)
4911 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4912 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4913 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4914 - MDS recovery shouldn't delete objects that it has given out (2730)
4915 - if enqueue arrives after completion, don't clobber LVB (2819)
4916 - don't unlock pages twice when trigger_group_io returns error (2814)
4917 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4918 - don't write pages to disk if there was an error (1450)
4919 - don't ping imports that have recovery disabled (2676)
4920 - take buffered bytes into account when balancing socknal conn (2817)
4921 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4922 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4923 - remove little-endian swabbing of llog records (1987)
4924 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4925 - retry reposting request buffers if they fail (1191)
4926 - grow extent at grant time to avoid granting a revoked lock (2809)
4927 - lock revoke doesn't evict page if covered by a second lock (2765)
4928 - disable VM readahead to avoid reading outside lock extents (2805)
4930 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4931 - updated kernel patches for hp-2.4.20 kernel (2681)
4933 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4936 - fix truncated write corruption (2366)
4937 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4939 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4940 - ensure len(uuid) < 37 in lmc (1171)
4941 - fix ia64 OOPS in llog_test (2255)
4942 - zero end of page at obdfilter for partial page writes (2648)
4943 - don't leave stale dentries around after renames (bug 2428)
4944 - fix timeouts when evicting a client with a single lock held (2642)
4945 - set deadline for the initial HELLO message to drain (2634)
4946 - print out dotted-quad IP addresses in the socknal (2302)
4948 - additional debugging for MDS client eviction problem (2443)
4949 - fix mkfsoptions support for osts (2603, 2604)
4951 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4954 - add series for the vanilla 2.6.0 kernel
4955 - add series for the vanilla 2.4.24 kernel
4956 - add series for a cray x86/64 UL kernel drop
4957 - fix xattr patches for the vanilla 2.4.19 series
4959 - generate true UUIDs in lmc (1171)
4960 - have portals stack dumping break in UML (2466)
4961 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4962 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4963 - wake callback waiting threads on client eviction (2460)
4964 - Add --ptldebug and --subsystem to lmc (1719)
4965 - update assertion to allow safe interrupt allocation
4966 - set rq_no_resend for cancel requests (2432)
4967 - recalculate ptlrpcd timeout after resend (2494)
4968 - call vfs_rmdir when removing pending directories (2368)
4969 - fix renaming a file to itself (2429)
4970 - lmc creates a default one-stripe lov (2454)
4971 - expand procfs space to handle large clusters (2326)
4972 - increase UML stack to avoid overflow
4973 - update lconf's list of debug and subsystem masks
4974 - fix lfs find --obd (2510)
4975 - /proc tunable for disabling filter read caching (2591)
4976 - stop rpm packages from altering slapd.conf (2301)
4977 - disable nagle in the socknal under 0conf (2578)
4978 - choose mds inode size based on stripe count (2572)
4979 - fix kernel-source rpm problems (2516)
4981 - add --disable-doc to avoid pdf generation (2421)
4982 - update documentation, tests, type-os, comments
4983 - avoid format warnings on ia64
4984 - remove the TOE NAL
4985 - tiny code cleanups by removing unused fields
4987 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4990 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4991 - remove the most blatant lies from BUILDING file (2371)
4992 - change default debug level to reasonable production setting
4993 - reduce client side cache size to reduce cache flush time
4994 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4995 - make TCP zerocopy and pinger support enabled by default (2476)
4996 - sync writes completed after process exits caused crashes (2319)
4997 - maintain correct mount count on the MDS (2356)
4998 - backout 1557, because 2316 wasn't really fixed
4999 - better file I/O statistics gathering in /proc
5000 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5001 - another kernel patch to fix zero-copy TCP function export
5002 - don't take duplicate lock when processing re-sent getattr (2420)
5003 - lctl uses obd_self_export instead of creating new conn (2353)
5004 - MDS/OST recovery case which requires object creation asserted (2425)
5005 - move lfs from /usr/sbin to /usr/bin in packages
5006 - fix race between mds_client_add and mds_client_free (2417)
5007 - use kmalloc instead of slabs in portals (2430)
5008 - don't create duplicate records when a failover MDS is present (2442)
5009 - remove unnecessary mount age check (2332)
5010 - don't remove directory inodes from locks prematurely (2451)
5011 - don't break if MDS service name is the same as hostname (2103)
5012 - fix races in client write RPC generation when cache full (2482)
5014 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5017 - remove now-unused request->rq_obd (278)
5018 - if an allocation fails, print out how much memory we've used (1933)
5019 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5020 - add the "configurable stack size" patch to most series files (1256)
5021 - ability to write large log records, for 100+ OST configs (2306)
5022 - fix NULL deref when filter_prep fails (2314)
5023 - fix operator precedence error in filter_sync
5024 - dynamic allocation of socknal TX descriptors (2315)
5025 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5026 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5027 - ensure that CWARN messages in Portals always get to the syslog
5028 - __init/__exit are not for prototype decls (ldlm_init/exit)
5029 - x86-64 compile warning fixes
5030 - fix gateway LMC keyword conflict (2318)
5031 - fix MDS lock inversions in getattr/reint paths (1844)
5032 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5033 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5034 - DLM race condition prevented some lock evictions (2328)
5035 - ENOMEM detection and retry on socknal sends (2230)
5036 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5037 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5038 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5039 - better error messages when a client is rejected during recovery (1505)
5040 - avoid cancelling locks which were never granted, after failure (2330)
5041 - fix i_sem/journal inversion in mds_client_add (2333)
5042 - fix truncate/getattr lock cycle deadlock (2334)
5043 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5044 - fix two rare exit paths which could leak an l_lock() ref (2321)
5045 - fencepost error in MDS/OST orphan recovery (2226)
5046 - make log record alignment 8 bytes (1988)
5047 - lstripe now fails when requested offset > ost_count (2237)
5048 - ensure that all kernel series have a complete list.h (1607)
5049 - fix crashes in special-file operations (2316)
5050 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5052 - allow configurable automake binary, for testing new versions
5053 - small update to the lfs documentation
5055 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5057 * fix negative export reference count in fsfilt_sync (2312)
5059 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5060 * release candidate 0.9.1
5062 - orphans are moved into the PENDING directory for possible recovery
5063 - replayed opens now open by fid for orphan/rename safety (1042)
5064 - last close of an orphan inode generates a transno (683)
5065 - chdir() and mount() now pin the directory entry (1020)
5066 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5067 - discard very old requests without processing them (1502)
5068 - remove some common, well-understood CERRORs (1505)
5069 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5070 - clear "grant" flags in OST replies until OST grant code lands (1644)
5071 - fix read performance by not clobbering i_blksize on client (1598)
5072 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5073 - make LRU size a /proc tunable, clears locks when reduced (707)
5074 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5075 - server locks take references on exports now (1558)
5076 - build fixes for 2.4.20-rh trees (1663)
5077 - return an error from lov_create if all OSCs are inactive (1751)
5078 - fix import levels when a reconnect happens without a timeout (1597)
5079 - exit early from mds_open if we get a lookup error (1749)
5080 - partial page read at EOF wouldn't wait for disk before sending (1642)
5081 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5082 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5083 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5084 - write/truncate lock inversion (1639)
5085 - Don't auto-load obdclass, portals modules during cleanup (1495)
5086 - fix timestamps from jumping to "now" (1763)
5087 - extra journal assertions (1648)
5088 - add an extra multiunlink test (1771)
5089 - fix read_record/write_record API (1776)
5090 - fix leak of offset_extent, possible incorrect i_size later (1772)
5091 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5092 - Debugging for the kqswnal_get_idle_tx problems (1820)
5093 - Allow recovery to be attempted multiple times (1536)
5094 - Write out MDS last_rcvd file after it is first created (1600)
5095 - Fix tx_descriptor leak in failed transmit situations (1827)
5096 - ext3 journaling fixes for assertion failure after IO error (1871)
5097 - class_export_put() on freed export after completion AST error (1896)
5098 - Fix revalidate looping in VFS (1322)
5099 - Don't access a freed export during MDS_REINT timeout (1521)
5100 - Add open-unlink recovery support on the MDS (1673,1764)
5101 - Return an error if no MDS data was read from last_rcvd (1946)
5102 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5103 - Don't setup a disk device that doesn't match exported UUID (317)
5104 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5105 - avoid committing NULL handle in force close
5106 - local.sh is now a one-stripe LOV configuration
5107 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5108 - fix ext3 htree duplicate directory entry corruption (1516)
5109 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5110 - update to new LOV EA format (2097)
5111 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5113 2003-06-15 Phil Schwan <phil@clusterfs.com>
5116 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5117 - per-import/export recovery handling (958, 931, 959)
5118 - multiple last-rcvd slots, for serving multiple FSes (949)
5119 - connections are again shared between multiple imp/exports (963, 964)
5120 - "umount -f" would hang if any requests needed to be sent (393, 978)
5121 - avoid pinning large req buffer by copying for queued messages (989)
5122 - add "uuid" to "lctl device" command to help upcalls (991)
5123 - "open" RPCs with transnos would confuse recovery counters (1037)
5124 - do proper endian conversion of all wire messages (288, 340, 891)
5125 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5126 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5127 - fix ptlrpc_connection leak in target_handle_connect (1174)
5128 - fix import refcounting bug in OST and MDS cleanup (1134)
5129 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5130 - fix very unlikely obd_types race condition (501)
5131 - remove osc_open hack for echo_client (1187)
5132 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5133 - a failure in read_inode2 leads to deadlock (1139)
5134 - cancel ack-locks as soon as transaction is committed (1072)
5135 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5136 - make sure to commitrw after any preprw to avoid deadlock (1162)
5137 - failing to execute a file in a lustre FS would lock inode (1203)
5138 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5139 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5140 - obd_uuid2tgt would walk off the end of the list (1255)
5141 - on IA64 the osc would give portals incorrect bulk size (1258)
5142 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5143 - fix lock inversion caused by new llite matching code (1282)
5144 - limit the number of dirty pages on a client to 10MB (1286)
5145 - timed out locks were not being corrected cancelled (1289)
5146 - fix O_DIRECT above 4GB on IA-32 (1292)
5147 * major user-visible changes
5148 - fail out/fail over policy now controlled by the upcall (993)
5150 - add OBD_PING to check server availability and failure (954)
5151 - lustre messages are now sent in sending host order (288, 340, 891)
5152 - add eadatalen to MDS getattr reply (340)
5153 - OST read replies may contain second buffer, with per-page status (593)
5155 2003-03-11 Phil Schwan <phil@clusterfs.com>
5158 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5159 - failed open()s could cause deadlock; fixed (867, 869)
5160 - stop cancelling OST locks when files are closed (481)
5161 - overlapping XID spaces caused network corruption (851, 853)
5162 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5163 - setattr_raw added, to do single-RPC, server-side setattrs
5164 - lmc/lconf syntax change for OST UUIDs
5165 - fix crashy race condition between ptlrpc_free_req and osc_close
5166 - don't use request in mdc_enqueue if we hit a timeout (889)
5167 - don't set the inode i_size for regular files from the MDS (896)
5168 - handle out of order completion AST (842)
5169 - don't LBUG if a lock request times out after receiving AST (913)
5170 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5171 - if a bad lock AST arrives, send an error instead of dropping entirely
5172 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5173 - fix leak in bulk IO when only partially completed (899, 900, 926)
5174 - fix O_DIRECT for ia64 (55)
5175 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5176 - C-z after timeout could hang a process forever; fixed (977)
5178 - client-side I/O cache (678, 924, 929, 941, 970)
5180 - READPAGE and SETATTRs which don't take server-side locks get
5183 2003-02-11 Phil Schwan <phil@clusterfs.com>
5186 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5187 - Fix fsfilt_extN_readpage() to read a full page of directory
5188 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5189 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5190 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5191 - Most intent operations were removed, in favour of a new RPC mode
5192 that does a single RPC to the server and bypasses most of the VFS
5193 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5194 - Aggressively cancel local locks on DLM servers
5195 - mds_reint_unlink sends EA to the client if it's the last nlink.
5196 client uses that EA to unlink OST objects.
5197 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5198 - recursive symlinks were fixed (439)
5199 - fixed NULL deref in DEBUG_REQ
5200 - filter_update_lastobjid no longer calls sync, which annoyed extN
5201 - fixed multi-client small-writes to a single file problem (445)
5202 - fixed mtime updates during file writes (607)
5203 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5204 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5205 - fixed Linux OST size reporting problem (444, 656)
5206 - OST now updates object mtime with writes or setattr (607, 619)
5207 - client verifies file size before zeroing page past EOF (445)
5208 - OST now writes last allocated objid to disk with allocation (108)
5209 - LOV on echo now works (409)
5211 - mds_reint_unlink sends a new buffer, with the EA included. this
5212 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5213 set if a regular file was being unlinked, and it was the last link
5214 - use PtlGet from the target for bulk writes (315)
5215 - OST now updates object mtime with writes or setattr (607, 619)
5216 - LDLM now has a grant-time callback to revalidate locked items, if
5218 - Many MDS operations were reorganized to combat race conditions
5220 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5221 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5223 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5226 - Fully reactivate OST imports after reconnection (512, others)
5227 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5228 - More graceful error handling for truncating on dead OST (515)
5229 - Don't error out unless we're actually accessing dead stripes (474)
5230 - Fix garbage sizes when stripes are missing (410)
5231 - LRU counters were broken, causing constant lock purge (433, 432)
5232 - garbage on read from stripes with failed OSTs (441)
5233 - mark OSCs as active before reconnecting during recovery (438)
5234 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5235 - lfind did not preserve OST order in output (443)
5236 - symlinks cause hung clients, incorrect data (439)
5237 - stop dereferencing request after dropping refcount (457)
5238 - don't LASSERT(spin_is_locked) on non-SMP (455)
5239 - fixes for many rename() bugs
5240 - fstat didn't correctly synchronize attributes (399)
5241 - server must handle lock cancellation during blocking AST prep (487)
5242 - bulk descriptors were free()d too soon (511)
5243 - fix paths in lconf, which would load incorrect modules (451, 507)
5244 - fix confusing lconf 'host not found' error message (386)
5245 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5246 - fix race condition in mdc_blocking_ast() for inode access (526)
5247 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5248 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5249 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5250 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5251 - fix lstripe in several ways (kernel side) (527)
5252 - fix request leak in ldlm_cli_enqueue (262)
5253 - incorrect OSC was marked inactive after OST failure
5254 - call mds_fs_cleanup before unmounting filesystem (524)
5255 - fix races between taking ns_lock and ldlm_lock_change_resource
5256 - fix races updating LOV export open file list
5257 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5258 - fix recovery NULL deref in ldlm_cli_cancel_unused
5259 - fix some DLM races by using new hash table for lock handles (419)
5260 - permit the client to specify desired inodes, at replay
5261 - duplicate requests when we queue them for replay reintegration
5262 - fix last_rcvd offset calculation
5263 - sync after each recovered transaction, so we always make progress
5264 - never, not always, ERESTART requests without transnos
5265 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5267 - skip replay if the MDS says that the client is already connected
5268 - don't check for a recovery-enabled export to match lctl's UUID
5269 - don't INC_USE_COUNT for phantom exports
5270 - don't crash when cleaning up phantom exports (567)
5271 - don't double-finish or set replay data for errored mdc_open requests
5272 - abort requests when they time out, so we don't get old replies
5273 - send/receive replies for AST messages again
5274 - if the client says that it doesn't have the lock, cancel it on the
5276 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5277 mark it as destroyed, it will all work out when decref is called
5278 - fix module use counts (22, 581)
5280 - ASTs now expect a reply (server cancels lock on error reply)
5282 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5285 - fix many simultaneous client startup (392)
5286 - fix dentry->d_it clobbering
5287 - credentials weren't being shipped for readdir/getattr operations
5288 - remove invalid assertions triggered during some concurrent MD
5290 - proper Lustre versions added (336, 389)
5291 - fix memory leak for create error case (398)
5292 - fix LOV locking bug that would get cli/srv out of sync
5293 - fix echo client over LOV (409)
5294 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5295 - fix double-O_EXCL intent crash (424)
5296 - avoid sending multiple lock CANCELs (352)
5298 - MDS can do multi-client recovery (modulo bugs in new code)
5300 - many updates, edits, cleanups
5302 2002-11-18 Phil Schwan <phil@clusterfs.com>
5305 - fix null d_it dereference (346)
5306 - fix full OST/dbench hang (333)
5307 - fix permission problem with file removal (286)
5308 - fix removal of OSCs from LOV when they fail
5309 - fix NULL deref during bulk timeout (214)
5310 - fix problems related to multiple filesystems on one MDS (241)
5311 - fixed serious subtle metadata locking bugs
5312 - free locks on clients when inodes are removed due to memory
5314 - fix inode pointer in lock data (285)
5315 - partial support for multiple MDS on a single host (241)
5316 - data locks weren't cancelled at clear_inode time (290, 311)
5317 - intent locks could lead to unbounded lock growth (205)
5318 - added a maximum lock count, an LRU list, and a flusher
5319 - fix multiple rename (365)
5320 - properly abstracted the echo client
5321 - OSC locked 1 byte too many; fixed
5322 - rewrote brw callback code:
5323 - fixed recovery bugs related to LOVs (306)
5324 - fixed too-many-pages-in-one-write crash (191)
5325 - fixed (again) crash in sync_io_timeout (214)
5326 - probably fixed callback-related race (385)
5328 - Add capability to MDS protocol
5329 - LDLM cancellations and callbacks on different portals
5331 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5334 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5335 - defer open object destruction to close time (601981, 138)
5336 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5337 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5338 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5339 - don't call obd_setattr in ll_file_release for destroyed objects
5340 * protocol change to lustre_msg: move |version| and add |flags|
5341 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5342 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5343 * added replay of create, unlink, link and rename operations during
5344 MDS failover; recovery should be much more robust now
5345 * remove failed OSCs from LOVs (only lov_create uses this so far)
5346 * the lustre-HOWTO was brought (more) up to date (582544)
5348 2002-10-23 Phil Schwan <phil@clusterfs.com>
5351 - in-use dentries weren't being reused properly (617851)
5352 - prevent multiple LDLM setup (599178)
5353 - fix LOV size calculations for truncate (617853)
5354 - fix client handling of MDS intent errors (POSIX)
5355 - fix permission bug in lovstripe.c test (624321)
5356 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5357 - truncate past end of file could corrupt data
5358 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5359 - a race in recovery could return ETIMEDOUT to apps (623947)
5360 - building outside the source directory was fixed
5361 * the lustre-HOWTO was brought (more) up to date (582544)
5362 * major progress was made on recovery functionality
5364 2002-10-10 Phil Schwan <phil@clusterfs.com>
5367 - recovery deadlock fix
5368 - rm -rf causes LBUG fix (617817)
5369 - file open by multiple tasks fix (618962)
5370 - directory permissions bugs (602707 and 620007)
5371 - journal_stop fixed with locking (611313)
5372 - O_APPEND failures resolved (618273, perhaps 614459)
5373 - lconf PATH fix (619770)
5374 - IA64 build fix (621450)
5375 - RPC buffer sizes scale with amount of memory
5377 2002-10-01 Phil Schwan <phil@clusterfs.com>
5380 - locks would be cancelled without throwing away data pages,
5381 resulting in inconsistent data (605627)
5382 - inode attributes were not always being refreshed (605627, 612449)
5383 - lconf now continues to cleanup after lctl reports an error
5384 - MDS now enforces user permissions (602707)
5385 - lprocfs cleanup fixed, but not yet enabled (614157)
5386 - fixed infinite server hang, should a client not respond to an AST
5387 - avoid going into recovery if user calls readlink() with a buffer
5388 that's too small (613941)
5389 - AST RPCs no longer require replies (614867) -- this may be changed
5390 - don't crash server if client sends an IOV that's too big (611336)
5391 - fixed lock conversion deadlock (611892)
5392 - fixed the following of symlinks (614622)
5393 * recovery: the server can remove locks from a client that dies, other
5394 clients can make progress
5395 * more extN patch fixes
5396 * compile-time configurable ptlrpc buffer allocations
5398 - collaborative read cache document
5399 - Lustre Lite Performance CDR document-in-progress
5401 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5404 - fix typo in patch-2.4.18
5406 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5409 - clear ptlrpc request each time in handle_incoming_request()
5410 - unlink of files now destroys the object on the OST
5412 2002-09-19 Peter Braam <braam@clusterfs.com>
5414 * add hard link support
5415 * change obdfile creation method
5416 * kernel patch changed
5418 2002-09-19 Peter Braam <braam@clusterfs.com>
5421 - stack overflow bug in extN fixed
5423 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5425 * documentation updates
5426 - add man pages for config tools
5427 - update tests/README to describe testing with new config tools
5428 - finish metadata API descriptions
5429 * bug fixes and cleanups
5430 - statfs workaround for 16TB limit
5431 - LOV stripe allocation improved, can stripe on subset of OSTs
5432 - LOV file size/IO offset was wrong for files > 4GB in size
5433 - object EA data was being dropped, caused files to be unreadable
5434 - memory overflow with non-LOV OST caused memory corruption
5435 - fixed regression tests to work with new config tools, obdfilter
5436 - fixed bug when directory size became larger than 1 block
5437 - fixed bug (for single client case) when PWD was deleted
5438 - invalidate local directory pages when doing intent-based ops
5439 - avoid LDLM oops when lock callback contained bad data
5441 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5443 * documentation updates
5444 * bug fixes and cleanups
5445 - configuration tools
5448 - 64-bit compile warnings
5449 - 64-bit internal statfs data
5451 * test_brw on persistent OST devices
5453 * lprocfs (disabled)
5455 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5457 * documentation updates
5458 * bug fixes and cleanups
5459 * configuration tools
5461 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5464 * many small fixes to 0_5_4
5465 * io/network handling
5466 * thinkos in MDS operations
5468 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5471 * crucial basic fixes to 0.5.3
5472 * IOR, Iozone work over Elan
5475 2002-08-07 Phil Schwan <phil@clusterfs.com>
5476 * version 0_5_3, our first alpha
5477 * we use the new Portals iovs
5478 * documentation updates
5479 * bug fixes and cleanups
5480 * small changes in the DLM wire protocol
5482 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5483 * version 0_5_1 with some initial stability,
5484 * locking on MD and file I/O.
5485 * documentation updates
5486 * several bug fixes since 0.5.0
5487 * small changes in wire protocol
5489 2002-07-18 Phil Schwan <phil@clusterfs.com>
5491 * delivered as Lustre Light Alpha
5492 * fixed a crash after handling invalid MDS requests
5493 * fixed directory pages for architectures with non-4k pages sizes
5495 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5496 * release version v0_4_4
5497 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5499 * Fixes a number of interruption problems with OST operations.
5500 * Update documentation for portals header changes
5501 * Move all wire protocol structs/defines to lustre_idl.h
5502 * Fixes symlink length bug.
5503 * Add tcpdump to repository.
5505 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5506 * release version v0_4_3
5507 * Fixes statfs for inodes on extN.
5508 * Fixes bug in runtests which would delete /etc/hosts.
5509 * Use 64-bit object IDs wherever possible (not into VFS though)
5510 Remove ost_get_info, which is unused by lustre, and out of date.
5512 2002-07-03 Peter Braam <braam@clusterfs.com>
5513 * release version v0_4_2 Fixes a lookup error (type not passed)
5514 * move forward to head of Portals
5515 * move forward to latest Lustre kernel
5517 2002-06-25 Peter Braam <braam@clusterfs.com>
5518 * release version v0_4_1. Hopefully stable on single node use.