1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.31 (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: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
20 Severity : enhancement
22 Description: New client IO stack (CLIO).
24 Severity : enhancement
26 Description: Commit on sharing. Eliminate inter-client dependencies between
27 uncommitted transactions by doing transaction commits.
28 Thereby clients may recovery independently.
31 Frequency : Create a symlink file with a very long name
33 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
34 Details : If there is no extra space in the request for early cancels,
35 ldlm_req_handles_avail() returns 0 instead of a negative value.
37 Severity : enhancement
39 Description: Add /proc entry for import status
40 Details : The mdc, osc, and mgc import directories now have
41 an import directory that contains useful import data for debugging
44 Severity : enhancement
46 Description: Re-disable certain /proc logging
47 Details : Enable and disable client's offset_stats, extents_stats and
48 extents_stats_per_process stats logging on the fly.
51 Frequency : Only on FC kernels 2.6.22+
53 Description: oops in statahead
54 Details : Do not drop reference count for the dentry from VFS when lookup,
55 VFS will do that by itself.
57 Severity : enhancement
59 Description: Generic /proc file permissions
60 Details : Set /Proc file permissions in a more generic way to enable non-
61 root users operate on some /proc files.
65 Description: Hitting mdc_commit_close() ASSERTION
66 Details : Properly handle request reference release in
67 ll_release_openhandle().
71 Frequency : only patchless client
72 Description: add workaround for race between add/remove dentry from hash
74 Severity : enhancement
76 Description: Allow OST glimpses to return PW locks
80 Description: LBUG when llog conf file is full
81 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
85 Description: Prevent import from entering FULL state when server in recovery
89 Description: service mount cannot take device name with ":"
90 Details : Only when device name contains ":/" will mount treat it as
96 Description: replace ptlrpcd with the statahead thread to interpret the async
97 statahead RPC callback
101 Frequency : on recovery
102 Description: I/O failures after umount during fail back
103 Details : if client reconnected to restarted server we need join to recovery
104 instead of find server handler is changed and process self
105 eviction with cancel all locks.
107 Severity : enhancement
109 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
111 Severity : enhancement
113 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
115 Severity : enhancement
117 Description: Upcall on Lustre log has been dumped
118 Details : Allow for a user mode script to be called once a Lustre log has
119 been dumped. It passes the filename of the dumped log to the
120 script, the location of the script can be specified via
121 /proc/sys/lnet/debug_log_upcall.
126 Description: avoid idr_remove called for id which is not allocated.
127 Details : Move assigment s_dev for clustered nfs to end of initialization,
128 to avoid problem with error handling.
133 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
134 Details : When connection is reused this not moved from CONN_UNUSED_HASH
135 into CONN_USED_HASH and this prodice warning when put connection
136 again in unused hash.
138 Severity : enhancement
140 Description: File striping can now be set to use an arbitrary pool of OSTs.
142 Severity : enhancement
144 Description: Export bytes_read/bytes_write count on OSC/OST.
148 Description: Early reply size mismatch, MGC loses connection
149 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
150 the connect flags are properly negotiated.
154 Description: Properly propagate oinfo flags from lov to osc for statfs
155 Details : restore missing copy oi_flags to lov requests.
157 Severity : enhancement
159 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
160 : l_getgroups(8), lst(8), routerstat(8)
161 Details : included man pages for llobdstat(8), llstat(8),
162 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
164 Severity : enhancement
166 Description: configure's --enable-quota should check the
167 : kernel .config for CONFIG_QUOTA
168 Details : configure is terminated if --enable-quota is passed but
169 : no quota support is in kernel
173 Description: Remove portals compatibility
174 Details : Remove portals compatibility, not interoperable with releases
179 Description: Resolve device initialization race
180 Details : Prevent proc handler from accessing devices added to the
181 obd_devs array but yet be intialized.
183 Severity : enhancement
185 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
187 Severity : enhancement
189 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
194 Description: Using wrong pointer in osc_brw_prep_request
195 Details : Access to array[-1] can produce panic if kernel compiled with
196 CONFIG_PAGE_ALLOC enabled
200 Description: Client runs out of low memory
201 Details : Consider only lowmem when counting initial number of llap pages
205 Description: *optional* service tags registration
206 Details : if the "service tags" package is installed on a Lustre node
207 When the filesystem is mounted, a local-node service tag will
208 be created. See http://inventory.sun.com/ for more information
209 about the Service Tags asset management system.
213 Description: Kernel BUG tries to release flock
214 Details : Lustre does not destroy flock lock before last reference goes
215 away. So always drop flock locks when client is evicted and
216 perform unlock regardless of successfulness of speaking to MDS.
220 Description: add refcount protection for osc callbacks, avoid panic on shutdown
224 Description: sanity test 65a fails if stripecount of -1 is set
225 Details : handle -1 striping on filesystem in ll_dirstripe_verify
230 Description: ASSERTION(CheckWriteback(page,cmd)) failed
231 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
234 Severity : enhancement
236 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
240 Description: do not process already freed flock
241 Details : flock can possibly be freed by another thread before it reaches
242 to ldlm_flock_completion_ast.
246 Description: LBUG during stress test
247 Details : Need properly lock accesses the flock deadlock detection list.
251 Description: oops in page fault handler
252 Details : kernel page fault handler can return two special 'pages' in error
253 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
257 Description: timeout with invalidate import.
258 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
259 should be handled by ptlrpcd. This produce long age waiting and
260 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
262 Severity : enhancement
264 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
268 Description: enable MGS and MDT services start separately
269 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
270 a co-located MGS without starting the MGS, which is a complement
271 to 'nosvc' mount option.
275 Frequency : after recovery
276 Description: precreate to many object's after del orphan.
277 Details : del orphan st in oscc last_id == next_id and this triger growing
278 count of precreated objects. Set flag LOW to skip increase count
279 of precreated objects.
283 Frequency : rare, on clear nid stats
284 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
285 Details : when clean nid stats sometimes try destroy live entry,
286 and this produce panic in free.
290 Description: Stack overflow during MDS log replay
291 ease stack pressure by using a thread dealing llog_process.
295 Description: wait until IO finished before start new when do lock cancel.
296 Details : VM protocol want old IO finished before start new, in this case
297 need wait until PG_writeback is cleared until check dirty flag and
298 call writepages in lock cancel callback.
300 Severity : enhancement
302 Description: using special macro for print time and cleanup in includes.
306 Description: mds_mfd_close() ASSERTION(rc == 0)
307 Details : In mds_mfd_close(), we need protect inode's writecount change
308 within its orphan write semaphore to prevent possible races.
312 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
316 Frequency : rare, on shutdown ost
317 Description: don't hit live lock with umount ost.
318 Details : shrink_dcache_parent can be in long loop with destroy dentries,
319 use shrink_dcache_sb instead.
323 Description: don't panic with use echo client
324 Details : echo client pass NULL as client nid pointer and this produce null
329 Description: fix build on ppc32
330 Details : compile code with -m64 flag produce wrong object file for ppc32.
334 Description: add message levels for liblustreapi
338 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
339 Details : Move the 'good_osts' check before the 'total_bavail' check. This
340 will result in an -EAGAIN and in the exit call path we call
341 alloc_rr() which will with increasing aggressiveness attempt to
342 aquire precreated objects on the minimum number of required OSCs.
346 Description: Use old size assignment to avoid deadlock
347 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
348 the scheduling while holding a spinlock. We do not need locking
349 for size in ll_update_inode() because size is only updated from
350 the MDS for directories or files without objects, so there is no
351 other place to do the update, and concurrent access to such inodes
352 are protected by the inode lock.
356 Description: resolve "_IOWR redefined" build error on SLES10
360 Description: dump the memory debugging after all modules are unloaded to
361 suppress false negative in conf_sanity test 39
363 Severity : enhancement
365 Description: build kernel-ib packages for OFED 1.3 in our release cycle
370 Description: fix SLES kernel versioning
371 Details : the kernel version for our SLES 10 kernel did not include a "-"
372 before the "smp" at the end. while this was not a problem in
373 general, it did mean that software trying to use the kernel
374 version to try to detect a vendor specific kernel would fail.
375 this was most evident by the OFED build scripts.
379 Description: Don't update lov_desc members until making sure they are valid
380 Details : When updating lov_desc members via proc fs, need fix their
381 validities before doing the real update.
385 Description: don't put request into delay list while invalidate in flight.
386 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
387 import in flight. this produce timeout for invalidate and sometimes
388 can cause stale data.
392 Frequency : on ppc only
393 Description: not convert ost objects for directory because it's not exist.
394 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
398 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
400 Severity : enhancement
402 Description: Update to support 2.6.22.14 vanilla kernel.
406 Frequency : rare, on recovery
407 Description: read procfs can produce deadlock in some situation
408 Details : Holding lprocfs lock which send rpc can produce block for destroy
409 obd objects and this also block reconnect with -EALREADY. This
410 isn't fix all lprocfs bugs - but make it rare.
412 Severity : enhancement
414 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
417 Frequency : frequent on X2 node
419 Description: mdc_set_open_replay_data LBUG
420 Details : Set replay data for requests that are eligible for replay.
424 Description: lustre_mgs: operation 101 on unconnected MGS
425 Details : When MGC is disconnected from MGS long enough, MGS will evict the
426 MGC, and late on MGC cannot successfully connect to MGS and a lot
427 of the error messages complaining that MGS is not connected.
431 Frequency : on network error
432 Description: panic with double free request if network error
433 Details : mdc_finish_enqueue is finish request if any network error ocuring,
434 but it's true only for synchronus enqueue, for async enqueue
435 (via ptlrpcd) this incorrect and ptlrpcd want finish request
438 Severity : enhancement
440 Description: client-side metadata stat-ahead during readdir(directory readahead)
441 Details : perform client-side metadata stat-ahead when the client detects
442 readdir and sequential stat of dir entries therein
445 Frequency : on start mds
447 Description: Implement get_info(last_id) in obdfilter.
450 Frequency : occasional
452 Description: Correctly check stale fid, not start epoch if ost not support SOM
453 Details : open with flag O_CREATE need set old fid in op_fid3 because
454 op_fid2 was overwritten with new generated fid, but mds can answer
455 with one of these two fids and both is not stale. Setattr
456 incorrectly started an epoch and assume will be called
457 done_writing, but without SOM done_writing ever being called.
460 Frequency : rare, depends on device drivers and load
462 Description: MDS or OSS nodes crash due to stack overflow
463 Details : Code changes in 1.8.0 increased the stack usage of some functions.
464 In some cases, in conjunction with device drivers that use a lot
465 of stack the MDS (or possibly OSS) service threads could overflow
466 the stack. One change which was identified to consume additional
467 stack has been reworked to avoid the extra stack usage.
470 Frequency : occasional
472 Description: Do not fail import if osc_interpret_create gets -EAGAIN
473 Details : If osc_interpret_create got -EAGAIN it immediately exits and
474 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
475 oscc_has_objects and see OSC has no objests and call
476 oscc_internal_create to resend create request.
478 Severity : enhancement
480 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
482 Severity : enhancement
484 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
487 Frequency : very rare
489 Description: Fix replay if there is an un-replied request and open
490 Details : In some cases, older replay request will revert the
491 mcd->mcd_last_xid on MDS which is used to record the client's
494 Severity : enhancement
496 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
498 Severity : enhancement
500 Description: Add rhel5 support to HEAD.
502 Serverity : enhancement
504 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
509 Description: Don't allow skipping OSTs if index has been specified.
510 Details : Don't allow skipping OSTs if index has been specified, make
511 locking in internal create lots better.
515 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
516 Details : access to bitfield in structure is always rounded to long
517 and this produce problem with not atomic change any bit.
521 Description: Lustre make rpms failed.
522 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
523 builds Lustre rpms from tarball.
526 Frequency : rare on shutdown ost
528 Description: If llog cancel was not send before clean_exports phase, this can
529 produce deadlock in llog code.
530 Details : If llog thread has last reference to obd and call class_import_put
531 this produce deadlock because llog_cleanup_commit_master wait when
532 last llog_commit_thread exited, but this never success because was
533 called from llog_commit_thread.
537 Description: allow userland application know is lost one of stripes.
538 Details : fill lvb_blocks with error code on ost and return it to
539 application if error flag found.
543 Description: NULL lov_tgts causing MDS oops
544 Details : more safe checks for NULL lov_tgts for avoid oops.
546 Severity : enhancement
548 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
552 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
553 Details : If an OST is down the MDS will hang indefinitely in
554 obd_statfs_rqset() waiting for the statfs data. While for
555 MDS QOS usage of statfs, it should not stuck in waiting.
557 Severity : enhancement
559 Description: remote_acl support
560 Details : Support ACL-based permission check for remote user.
561 Support setfacl/getfacl for remote user with the utils
562 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
563 the system "{s,g}etfacl" utils.
565 Severity : enhancement
567 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
569 Severity : enhancement
571 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
575 Description: make mgs_setparam() handle fsname containing dash
576 Details : fsname containing a dash does not work with lctl conf_param
578 Severity : enhancement
580 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
582 Severity : enhancement
584 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
588 Frequency : rare, at shutdown
589 Description: access already free / zero obd_namespace.
590 Details : if client_disconnect_export was called without force flag set,
591 and exist connect request in flight, this can produce access to
592 NULL pointer (or already free pointer) when connect_interpret
593 store ocd flags in obd_namespace.
597 Frequency : only at startup
598 Description: not alloc memory with spinlock held.
599 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
600 if any spinlock held.
602 Severity : enhancement
604 Description: make lustre randomly fail allocating memory
605 Details : Make lustre randomly failed allocating memory for testing purpose.
607 Severity : enhancement
609 Description: lost problems with lov objid file
610 Details : Fixes some scability and access to not inited memory problems
611 in work with lov objdid file.
616 Description: lfs find does not continue on file error
617 Details : Continue other files processing when a file/dir is absent.
621 Description: Inconsistent usage of lustre_pack_reply()
622 Details : Standardize the usage of lustre_pack_reply() such that it
623 always generate a CERROR on failure.
626 Frequency : occasional
628 Description: MDS hang or stay in waiting lock
629 Details : If client receive lock with CBPENDING flag ldlm need send lock
630 cancel as separate rpc, to avoid situation when cancel request
631 can't processed due all i/o threads stay in wait lock.
635 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
639 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
643 Description: Update to SLES9 kernel-2.6.5-7.287.3.
647 Description: Update to RHEL4 latest kernel.
649 Severity : enhancement
651 Description: Build SLES10 patchless client fails
652 Details : The configure was broken by run ./configure with
653 --with-linux-obj=.... argument for patchless client. When the
654 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
655 can't search header adequately. Use absolute path such as
656 -I($LINUX)/include instead.
660 Description: interrupt oig_wait produce painc on resend.
661 Details : brw_redo_request can be used for resend requests from ptlrpcd and
662 private set, and this produce situation when rq_ptlrpcd_data not
663 copyed to new allocated request and triggered LBUG on assert
664 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
665 ptlrpcd set if request is changed and can be safety changed to use
670 Description: LASSERT_{REQ,REP}SWAB macros are buggy
671 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
672 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
673 these macros and replace them with their difinitions instead.
677 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
678 Details : Update which_patch & target file for SLES10 latest kernel.
681 Description: Remove client patches from SLES 10 kernel.
682 Details : This causes SLES 10 clients to behave as patchless clients
683 even on a Lustre-patched (server) kernel.
685 Severity : enhancement
687 Description: self-adjustable client's lru lists
688 Details : use adaptive algorithm for managing client cached locks lru
689 lists according to current server load, other client's work
690 pattern, memory activities, etc. Both, server and client
691 side namespaces provide number of proc tunables for controlling
694 Severity : enhancement
696 Description: light-weight GSS support
697 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
698 and reduce performance overhead.
700 Severity : enhancement
702 Description: Linux keyring support
703 Details : Support using service of Linux keyring for Lustre GSS internal
704 context refresh/cache mechanism.
708 Description: Fix errors in lfs documentation
709 Details : Fixes man pages
713 Description: don't use GFP_* in generic Lustre code.
714 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
718 Description: obdclass is limited by single OBD_ALLOC(idarray)
719 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
723 Description: warnings with build patchless client with vanila 2.6.19 and up
724 Details : change old ctl_table style and replace ctl_table/ctl_table_header
725 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
729 Frequency : Only for SLES
730 Description: sanity_quota fail test_1
731 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
732 file, which makes TSTID[2] inunique.
736 Frequency : Only for SLES9
737 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
738 Details : Update target/ChangeLog/which_patch .
742 Frequency : Only for test_10 in sanity-quota.sh
743 Description: a bug in quota test code
744 Details : checking a test flag in a wrong place
748 Frequency : Only for test_13 in sanity-quota.sh
749 Description: a bug in quota test script
750 Details : shouldn't check overfull allocated quota
754 Description: jbd statistics
755 Details : Port older jbd statistics patch for sles10
757 Severity : enhancement
759 Description: Add printing inode info into message about error in writepage.
764 Description: reply_lock_interpret crash due to race with it and lock cancel.
765 Details : Do not replay locks that are being cancelled. Do not reference
766 locks by their address during replay, just by their handle.
770 Frequency : When flocks are used.
771 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
772 Details : Flock locks might destroy just granted lock if it could be merged
773 with another existing flock, this is done in completion handler,
774 so teach ldlm_cli_enquque_fini that this is a valid case for
780 Description: Oops in read and write path when failing to allocate lock.
781 Details : Check if lock allocation failed and return error back.
785 Description: lstripe command fails for valid OST index
786 Details : The stripe offset is compared to lov->desc.ld_tgt_count
787 instead of lov->desc.ld_active_tgt_count.
791 Description: sanity.sh failed test 103
792 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
797 Description: df doesn't work properly if diskfs blocksize != 4K
798 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
802 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
803 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
805 Severity : enhancement
807 Description: Adaptive timeouts
808 Details : RPC timeouts adapt to changing server load and network
809 conditions to reduce resend attempts and improve recovery time.
813 Description: llapi_file_create() does not allow some changes
814 Details : add llapi_file_open() that allows specifying the mode and
815 open flags, and also returns an open file handle.
819 Description: merge and cleanup kernel patches.
820 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
824 Description: Add journal checksum support.(Kernel part)
825 Details : The journal checksum feature adds two new flags i.e
826 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
827 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
828 indicates that the commit block contains the checksum for
829 the blocks described by the descriptor blocks. Now commit
830 record can be sent to disk without waiting for descriptor
831 blocks to be written to disk. This behavior is controlled
832 using JBD2_FEATURE_ASYNC_COMMIT flag.
836 Description: OSS needs mutliple precreate threads
837 Details : Add ability to start more than one create thread per OSS.
841 Description: Sanity.sh test_65k failed: setstripe should have succeeded
842 Details : execute lfs setstripe on client
846 Description: mds_obd_create error creating tmp object
847 Details : When the user sets quota on root, llog will be affected and can't
848 create files and write files.
851 Frequency : Always on ia64 patchless client, and possibly others.
853 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
854 Details : This allows the patchless client to be loaded on architectures
859 Description: RedHat Update kernel for RHEL5
860 Details : Add the kernel config file for RHEL5.
864 Description: RedHat Update kernel for RHEL5
865 Details : Modify the kernel config file more closer RHEL5.
869 Description: Build failure against Centos5 (RHEL5)
870 Details : Define PAGE_SIZE when it isn't present.
874 Description: test_7 on sanity-quota.sh can't be used on separate machines
875 Details : add facet to handle it
879 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
880 Details : with NFS, the anon dentry's parent was set to itself in
881 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
882 corresponding dentry other than use rec->ur_name.
884 Severity : enhancement
886 Description: lfs setstripe enhancement
887 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
891 Description: enable data checksumming by default
892 Details : enable checksum by default, allow --disable-checksum
893 configure option and "-o nochecksum" mount option. Checksums
894 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
898 Description: lustre support for RHEL5
899 Details : Add support for RHEL5.
903 Description: Client eviction due to failover config
904 Details : after a connection loss, the lustre client should attempt to
905 reconnect to the last active server first before trying the
906 other potential connections.
910 Description: when mds and osts use different quota unit(32bit and 64bit),
911 quota will be released repeatly.
912 Details : void sending multiple quota reqs to mds, which will keep the
913 status between the reqs.
917 Description: osts not allocated evenly to files
918 Details : change the condition to increase offset_idx
921 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
923 Description: Data corruption for OSTs that are formatted larger than 2TB
925 Details : When generating the bio request for lustre file writes the
926 sector number would overflow a temporary variable before being
927 used for the IO. The data reads correctly from Lustre (which
928 will overflow in a similar manner) but other file data or
929 filesystem metadata may be corrupted in some cases.
933 Description: Tune the kernel for good SCSI performance.
934 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
935 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
940 Description: rewrite ext2-derived code in obdclass/uuid.c
941 Details : rewrite inherited code (uuid parsing code from ext2 utils)
942 from scratch preserving functionality.
946 Description: "lfs find -obd UUID" prints directories
947 Details : "lfs find -obd UUID" will return all directory names instead
948 of just file names. It is incorrect because the directories
949 do not reside on the OSTs.
953 Description: use i_size_read and i_size_write in 2.6 port
954 Details : replace inode->i_size access with i_size_read/write()
957 Frequency : when removing large files
959 Description: scheduling issue during removal of large Lustre files
960 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
961 It causes scheduling issues when removing large files (17TB in the
965 Frequency : only with liblustre clients on XT3
967 Description: evictions taking too long
968 Details : allow llrd to evict clients directly on OSTs
971 Frequency : only on ppc
973 Description: /proc/fs/lustre/devices broken on ppc
974 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
975 We should make sure the type of 'index' is loff_t and then cast
976 explicitly as needed below. Do not assign an explicitly cast
980 Frequency : only for rhel5
982 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
983 Details : Modify the target file & which_kernel.
985 Severity : enhancement
987 Description: omit set fsid for export NFS
988 Details : fix set/restore device id for avoid EMFILE error and mark lustre
989 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
993 Frequency : Always, for kernels after 2.6.16
994 Description: Fix warning idr_remove called for id=.. which is not allocated.
995 Details : Last kernels save old s_dev before kill super and not allow
996 to restore from callback - restore it before call kill_anon_super.
1000 Description: buffer overruns could theoretically occur
1001 Details : llapi_semantic_traverse() modifies the "path" argument by
1002 appending values to the end of the origin string, and a
1003 overrun may occur. Adding buffer overrun check in liblustreapi.
1007 Description: Fix error on 'ls .' at the top of the Lustre mount.
1008 Details : Don't revalidate dentry if it is a root dentry.
1012 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1013 Details : Modify vm-tunables-rhel4.patch.
1017 Description: Kernel config for 2.6.18-vanilla.
1018 Details : Modify targets/2.6-vanilla.target.in.
1019 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1020 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1021 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1022 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1026 Description: improve handling recoverable errors
1027 Details : if request processig with error which can be recoverable on server
1028 request should be resend, otherwise page released from cache and
1033 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1034 Details : Modify target file & which_patch.
1035 A flaw was found in the IA32 system call emulation provided
1036 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1037 value could be stored in the %RAX register, which could trigger an
1038 out-of-bounds system call table access. An untrusted local user
1039 could exploit this flaw to run code in the kernel
1040 (ie a root privilege escalation). (CVE-2007-4573).
1044 Description: O_DIRECT bypasses client statistics.
1045 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1046 read_ahead_stats were updated. copied the stats section from
1047 osc_send_oap_rpc() into async_internal().
1051 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1055 Description: change order of libsysio includes
1056 Details : '#include sysio.h' should always come before '#include xtio.h'
1060 Description: handle "serious error: objid * already exists" more gracefully
1061 Details : If LAST_ID value on disk is smaller than the objects existing in
1062 the O/0/d* directories, it indicates disk corruption and causes an
1063 LBUG(). If the object is 0-length, then we should use the existing
1064 object. This will help to avoid a full fsck in most cases.
1066 Severity : enhancement
1068 Description: adapt the lustre_config script to support the upgrade case
1069 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1070 server targets to 1.6 in parallel.
1074 Description: To avoid grant space > avaible space when the disk is almost
1075 full. Without this patch you might see the error "grant XXXX >
1076 available" or some LBUG about grant, when the disk is almost
1078 Details : In filter_check_grant, for non_grant cache write, we should
1079 check the left space by if (*left > ungranted + bytes), instead
1080 of (*left > ungranted), because only we are sure the left space
1081 is enough for another "bytes", then the ungrant space should be
1082 increase. In client, we should update cl_avail_grant only there
1083 is OBD_MD_FLGRANT in the reply.
1087 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1088 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1089 described in CVE-2007-4573. This problem could allow untrusted
1090 local users to gain root access.
1093 Frequency : when using O_DIRECT and quotas
1095 Description: Incorrect file ownership on O_DIRECT output files
1096 Details : block usage reported by 'lfs quota' does not take into account
1097 files that have been written with O_DIRECT.
1102 Description: touch file failed when fs is not full
1103 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1104 otherwise we can get ENOSP while fs is not full.
1108 Description: parallel lock callbacks
1109 Details : Instead of sending blocking and completion callbacks as separated
1110 requests, adding them to a set and sending in parallel.
1113 Frequency : only for Cray XT3
1114 Bugzilla : 12829/13455
1115 Description: Changing primary group doesn't change the group lustre assigns to
1117 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1118 overridden with the primary group provided by the group upcall,
1119 whereas the supplied fsgid can be trusted if it is in the list of
1120 supplementary groups returned by the group upcall.
1122 Severity : enhancement
1124 Description: Allow masking D_WARNING, D_ERROR messages from console
1125 Details : Console messages can now be disabled via lnet.printk.
1129 Description: User code with malformed file open parameter crashes client node
1130 Details : Before packing join_file req, all the related reference should be
1131 checked carefully in case some malformed flags cause fake
1132 join_file req on client.
1136 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1137 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1138 mode is equal to the granted mode. However, it should also check
1139 that LDLM_FL_CP_REQD is not set.
1143 Description: Heavy nfs access might result in deadlocks
1144 Details : After ELC code landed, it is now improper to enqueue any mds
1145 locks under och_sem, because enqueue might want to decide to
1146 cancel open locks for same inode we are holding och_sem for.
1150 Description: Client eviction while running blogbench
1151 Details : A lot of unlink operations with concurrent I/O can lead to a
1152 deadlock causing evictions. To address the problem, the number of
1153 oustanding OST_DESTROY requests is now throttled to
1154 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1155 callbacks are processed in priority.
1159 Description: enable ACLs on MDS by default
1160 Details : ACLs must be enabled on MDS by default.
1163 Frequency : PPC/PPC64 only
1165 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1166 Details : fix duplicated definitions between asm-ppc64/types.h and
1167 lustre_types.h on PPC.
1170 Frequency : PPC/PPC64 only
1172 Description: asm-ppc/segment.h does not exist
1173 Details : fix compile issue on PPC.
1177 Description: better handle error messages in extents code
1180 Frequency : RHEL4 only
1182 Description: mkfs is very slow on IA64/RHEL4
1183 Details : A performance regression has been discovered in the MPT Fusion
1184 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1185 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1186 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1188 Severity : enhancement
1190 Description: SNMP support enhancement
1191 Details : Adding total number of sampled request for an MDS node in snmp
1194 Severity : enhancement
1196 Description: Optimize ldlm waiting list processing for PR extent locks
1197 Details : When processing waiting list for read extent lock and meeting read
1198 lock that is same or wider to it that is not contended, skip
1199 processing rest of the list and immediatelly return current
1200 status of conflictness, since we are guaranteed there are no
1201 conflicting locks in the rest of the list.
1205 Description: Time out and refuse to reconnect
1206 Details : When the failover node is the primary node, it is possible
1207 to have two identical connections in imp_conn_list. We must
1208 compare not conn's pointers but NIDs, otherwise we can defeat
1209 connection throttling.
1213 Description: port llog fixes from b1_6 into HEAD
1214 Details : Port llog reference couting and some llog cleanups from b1_6
1215 (bug 10800) into HEAD, for protect from panic and access to already
1216 free llog structures.
1220 Description: Detect stride IO mode in read-ahead
1221 Details : When a client does stride read, read-ahead should detect that and
1222 read-ahead pages according to the detected stride pattern.
1226 Description: data checksumming impacts single node performance
1227 Details : add support for several checksum algorithm. Currently, only CRC32
1228 and Adler-32 are supported. The checksum type can be changed on
1229 the fly via /proc/fs/lustre/osc/*/checksum_type.
1233 Description: use adler32 for page checksums
1234 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1239 Description: build for x2 fails
1240 Details : fix compile issue on Cray systems.
1244 Description: Properly match for duplicate locks
1245 Details : Due to different lock order from skiplists code, we need to
1246 traverse entire list for now
1249 Frequency : only on PPC/SLES10
1251 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1252 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1253 be defined. Add a hack in mkfs_lustre.c to work around this compile
1258 Description: LASSERT on MDS when client holding flock lock dies
1259 Details : ldlm pool logic depends on number of granted locks equal to
1260 number of released locks which is not true for flock locks, so
1261 just exclude such locks from consideration.
1265 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1266 Details : Use fsfilt_sync() for both the whole filesystem sync and
1267 individual file sync to eliminate dangerous inode locking
1268 with I_LOCK that can lead to a deadlock.
1272 Description: performance in 1.6.3
1273 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1274 sure that 1MB requests are not fragmented by the block layer.
1276 Severity : enhancement
1278 Description: organize the server-side client stats on per-nid basis
1279 Details : Change the structure of stats under obdfilter and mds to
1287 The "uuid"s file would list the uuids of _active_ exports.
1288 And the clear entry is to clear all stats and stale nids.
1290 Severity : enhancement
1292 Description: eliminate client locks in face of contention
1293 Details : file contention detection and lockless i/o implementation
1294 for contended files.
1298 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1301 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1303 Description: MDS may lose file striping (and hence file data) in some cases
1304 Details : If there are additional extended attributes stored on the MDS,
1305 in particular ACLs, SELinux, or user attributes (if user_xattr
1306 is specified for the client mount options) then there is a risk
1307 of attribute loss. Additionally, the Lustre file striping
1308 needs to be larger than default (e.g. striped over all OSTs),
1309 and an additional attribute must be stored initially in the
1310 inode and then increase in size enough to be moved to the
1311 external attribute block (e.g. ACL growing in size) for file
1316 Description: skiplist implementation simplification
1317 Details : skiplists are used to group compatible locks on granted list
1318 that was implemented as tracking first and last lock of each
1319 lock group the patch changes that to using doubly linked lists
1323 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1324 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1325 because LASSERT on both the data supplied by a client, and the
1326 data on disk is dangerous and incorrect.
1328 Severity : enhancement
1330 Description: Slow truncate/writes to huge files at high offsets.
1331 Details : Directly associate cached pages to lock that protect those pages,
1332 this allows us to quickly find what pages to write and remove
1333 once lock callback is received.
1337 Description: more ldlm soft lockups
1338 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1339 starve other threads from the resource lock for a long time in
1340 case of long waiting queue, so change the debug level from
1341 D_OTHER to the less frequently used D_INFO.
1343 Severity : enhancement
1345 Description: add -gid, -group, -uid, -user options to lfs find
1349 Description: Hung threads in invalidate_inode_pages2_range
1350 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1351 once one is completed. As a result, some RPCs are stuck in the
1352 queue and are never sent.
1356 Description: filter threads hungs on waiting journal commit
1357 Details : Cleanup filter group llog code, then only filter group llog will
1358 be only created in the MDS/OST syncing process.
1362 Description: Procfs and llog threads access destoryed import sometimes.
1363 Details : Sync the import destoryed process with procfs and llog threads by
1364 the import refcount and semaphore.
1366 Severity : enhancement
1368 Description: openlock cache of b1_6 port to HEAD
1373 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1374 Details : If insertion of an extent fails, then discard the inode
1375 preallocation and free data blocks else it can lead to duplicate
1380 Description: don't always update ctime in ext3_xattr_set_handle()
1381 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1382 In some cases the ctime should not be updated, for example for
1383 2.0->1.8 compatibility it is necessary to delete an xattr and it
1384 should not update the ctime.
1388 Bugzilla : 15713/16362
1389 Description: Assertion in iopen_connect_dentry in 1.6.3
1390 Details : looking up an inode via iopen with the wrong generation number can
1391 populate the dcache with a disconneced dentry while the inode
1392 number is in the process of being reallocated. This causes an
1393 assertion failure in iopen since the inode's dentry list contains
1394 both a connected and disconnected dentry.
1398 Description: assertion failure in ldlm_handle2lock()
1399 Details : fix a race between class_handle_unhash() and class_handle2object()
1400 introduced in lustre 1.6.5 by bug 13622.
1405 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1406 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1411 Description: oops in sd_iostats_seq_show()
1412 Details : unloading/reloading the scsi low level driver triggers a kernel
1413 bug when trying to access the sd iostat file.
1418 Description: Kernel panics during QLogic driver reload
1419 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1420 patch, causing memory corruption.
1425 Description: journal_dev option does not work in b1_6
1426 Details : pass mount option during pre-mount.
1428 Severity : enhancement
1430 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1431 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1432 extent information of a file. It can be used to map logical blocks
1433 in a file to physical blocks in the block device.
1437 Description: LDLM soft lockups - improvement
1438 Details : It is be possible to send the lock handle along with each read
1439 or write request because the client is already doing a lock match
1440 itself so there isn't any reason the OST should have to re-do that
1444 Frequency : only on Cray X2
1446 Description: X2 build failures
1447 Details : fix build failures on Cray X2.
1451 Description: xid & resent requests
1452 Details : Initialize RPC XID from clock at startup (randomly if clock is
1455 Severity : enhancement
1457 Description: Add lustre_start utility to start or stop multiple Lustre servers
1462 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1463 Details : In case of memory pressure, list_del() can be called twice on
1464 req->rq_history_list, causing a kernel oops.
1468 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1469 Details : fix stack overflow in the distributed lock manager by defering
1470 export eviction after a failed AST to the elt thread instead of
1471 handling it in the dlm interpret routine.
1475 Description: Convert some comments to new format.
1476 Details : Update documenting comments to match doxygen conventions.
1480 Description: Grammar fixes.
1481 Details : A couple of trivial spelling fixes.
1485 Description: OSD_COUNTERS-mandatory
1486 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1487 to simplify the code.
1491 Description: simplify cmm_device freeing logic.
1492 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1496 Description: Add lockdep support to dt_object_operations locking interface.
1497 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1498 parameter indicating lock ordering. Update mdd code to use new
1503 Description: Introduce failloc constants for lockless IO tests.
1504 Details : Add two new failloc constants to test lockless IO. Only one of
1505 them in implemented---another is checked in yet to be landed
1510 Description: Add lockdep support for inode mutex.
1511 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1512 in the situations where Lustre has to lock more than one inode
1517 Description: Add optional invariants checking support.
1518 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1519 configure switch. This macro is to be used for consistency and
1520 sanity checks that are too expensive to be left in `production'
1525 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1526 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1531 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1532 Details : Use spin_lock_nested() in (the only) situation where more than
1533 one ldlm_lock is locked simultaneously. Also, fix possible
1534 dead-lock in ldlm_lock_change_resource() by enforcing particular
1539 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1540 Details : Instead of specifying each ldlm_lock call-back through separate
1541 parameter, wrap them into struct ldlm_callback_suite.
1545 Description: Kill join_lru obd method and its callers.
1546 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1547 in memory---a requirement implemented through ->o_join_lru() obd
1548 method in HEAD. Remove this method and its users.
1552 Description: Add asynchronous ldlm ENQUEUE completion handler.
1553 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1554 case that stock ldlm_completion_ast() cannot handle as it waits
1555 until lock is granted. Introduce new ldlm_completion_ast_async()
1556 for this. Also comment ldlm_completion_ast().
1560 Description: ldlm_error <-> errno conversion.
1561 Details : Add functions to map (rather arbitrary) between LDLM error codes
1562 and standard errno values. CLIO needs this to prevent LDLM specific
1563 constants from escaping ldlm and osc.
1567 Description: Kill unused ldlm_handle2lock_ns() function.
1568 Details : Kill unused ldlm_handle2lock_ns() function.
1572 Description: Add lu_ref support to ldlm_lock
1573 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1574 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1575 and ldlm_resource. LDLM interface has to be changed, because code that
1576 releases a reference on a lock, has to "know" what reference this is.
1577 In the most frequent case
1579 lock = ldlm_handle2lock(handle);
1581 LDLM_LOCK_PUT(lock);
1583 no changes are required. When any other reference (received _not_ from
1584 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1585 instead of LDLM_LOCK_PUT().
1587 Arguably, changes are pervasive, and interface requires some discipline
1588 for proper use. On the other hand, it was very instrumental in finding
1589 a few leaked lock references.
1593 Description: Add ldlm_lock_addref_try().
1594 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1595 attempts to addref a lock that might be being canceled
1600 Description: Add ldlm_weigh_callback().
1601 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1602 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1603 hard-coded `number of pages' logic.
1607 Description: Add lockdep annotations to llog code.
1608 Details : Use appropriately tagged _nested() locking calls in the places
1609 where llog takes more than one ->lgh_lock lock.
1613 Description: Add loi_kms_set().
1614 Details : Wrap kms updates into a helper function.
1618 Description: Constify instances of struct lsm_operations.
1619 Details : Constify instances of struct lsm_operations.
1623 Description: lu_conf support.
1624 Details : On a server, a file system object is uniquely identified
1625 by a fid, which is sufficient to locate and load all object
1626 state (inode). On a client, on the other hand, more data are
1627 necessary instantiate an object. Change lu_object_find() and
1628 friends to take additional `lu_conf' argument describing object.
1629 Typically this includes layout information.
1633 Description: lu_context fixes.
1634 Details : Introduce new lu_context functions that are needed on the client
1635 side, where some system threads (ptlrpcd) are shared by multiple
1636 modules, and so cannot be stopped during module shutdown.
1640 Description: Add start and stop methods to lu_device_type_operations.
1641 Details : Introduce two new methods in lu_device_type_operations, that are
1642 invoked when first instance of a given type is created and last one
1643 is destroyed respectively. This is need by CLIO.
1647 Description: Add lu_ref support to struct lu_device.
1648 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1649 track leaked references.
1653 Description: Introduce lu_kmem_descr.
1654 Details : lu_kmem_descr and its companion interface allow to create
1655 and destroy a number of kmem caches at once.
1659 Description: Fix lu_object finalization race.
1660 Details : Fix a race between lu_object_find() finding an object and its
1661 concurrent finalization. This race is (most likely) not possible
1662 on the server, but might happen on the client.
1666 Description: Introduce lu_ref interface.
1667 Details : lu_ref is a debugging module allowing to track references to
1668 a given object. It is quite cpu expensive, and has to be
1669 explicitly enabled with --enable-lu_ref. See usage description
1674 Description: Factor lu_site procfs stats into a separate function.
1675 Details : Separate lu_site stats printing code into a separate function
1676 that can be reused on a client.
1680 Description: Constify instances of struct {lu,dt,md}_device_operations.
1681 Details : Constify instances of struct {lu,dt,md}_device_operations.
1685 Description: Introduce struct md_site and move meta-data specific parts of
1686 struct lu_site here.
1687 Details : Move md-specific fields out of struct lu_site into special struct
1688 md_site, so that lu_site can be used on a client.
1692 Description: Kill mdd_lov_destroy().
1693 Details : Remove unused mdd code.
1697 Description: Add st_block checking to multistat.c.
1698 Details : Add st_block checking to multistat.c.
1702 Description: Add lu_ref support to struct obd_device.
1703 Details : Add lu_ref tracking to obd_device.
1707 Description: Kill obd_set_fail_loc().
1708 Details : Remove unused code.
1712 Description: Add special type for ptlrpc_request interpret functions.
1713 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1714 there. Actual usage will be in CLIO.
1718 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1719 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1720 doesn't work with lockdep.
1724 Description: Add rwv.c test program.
1725 Details : New testing program exercising readv(2) and writev(2) (Qian).
1729 Description: Add sendfile.c test program.
1730 Details : New testing program exercising sendfile(2) (Jay).
1734 Description: Ratelimit a message that can be very frequent.
1735 Details : Ratelimit a memory allocation failure message that can
1740 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1741 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1745 Description: Liblustre build fixes.
1746 Details : Liblustre build fixes.
1750 Description: libcfs: add cfs_{need,cond}_resched() interface.
1751 Details : libcfs: add cfs_{need,cond}_resched() definition and
1752 implementations for Linux, NT, and liblustre.
1754 Severity : enhancement
1756 Description: More exported tunables for mballoc
1757 Details : Add support for tunable preallocation window and new tunables for
1758 large/small requests
1762 Description: Detect corruption of block bitmap and checking for preallocations
1763 Details : Checks validity of on-disk block bitmap. Also it does better
1764 checking of number of applied preallocations. When corruption is
1765 found, it turns filesystem readonly to prevent further corruptions.
1769 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1770 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1771 window to 0. If it does hit cache miss, and read-pattern is still
1772 stride-io mode, does not reset the stride window, but also does not
1773 increase the stride window length in this case.
1777 Frequency : only for big-endian servers
1778 Description: Check if system is big-endian while mounting fs with extents feature
1779 Details : Mounting a filesystem with extents feature will fail on big-endian
1780 systems since ext3-based ldiskfs is not supported on big-endian
1781 systems. This can be over-riden with "bigendian_extents" mount option.
1783 --------------------------------------------------------------------------------
1785 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1787 * Support for kernels:
1788 2.6.5-7.283 (SLES 9),
1789 2.6.9-55.EL (RHEL 4),
1790 2.6.16.46-0.14 (SLES 10),
1791 2.6.18.8 vanilla (kernel.org)
1792 * Client support for unpatched kernels:
1793 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1794 2.6.16 - 2.6.22 vanilla (kernel.org)
1795 * Due to recently discovered recovery problems, we do not recommend
1796 using patchless RHEL 4 clients with this or any earlier release.
1797 * Recommended e2fsprogs version: 1.39.cfs8
1798 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1799 * Starting with this release, the ldiskfs backing filesystem required
1800 by Lustre is now in its own package, lustre-ldiskfs. This package
1801 should be installed. It is versioned separately from Lustre and
1802 may be released separately in future.
1806 Description: block reactivating mgc import until all deactivates complete
1807 Details : Fix race when failing back MDT/MGS to itself (testing)
1810 Frequency : at statup only
1812 Description: mds_lov_synchronize race leads to various problems
1813 Details : simultaneous MDT->OST connections at startup can cause the
1814 sync to abort, leaving the OSC in a bad state.
1816 Severity : enhancement
1818 Description: add optional extra BUILD_VERSION info
1819 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1820 to override the lustre version.
1823 Frequency : 2.6.18 servers only
1825 Description: ll_kern_mount() doesn't release the module reference
1826 Details : The ldiskfs module reference count never drops down to 0
1827 because ll_kern_mount() doesn't release the module reference.
1832 Description: server LBUG when using old ost_num_threads parameter
1833 Details : Accept the old ost_num_threads parameter but warn that it
1834 is deprecated, and fix an off-by-one error that caused an LBUG.
1839 Description: Transient SCSI error results in persistent IO issue
1840 Details : iobuf->dr_error is not reinitialized to 0 between two
1844 Frequency : sometimes when underlying device returns I/O errors
1846 Description: OSTs not going read-only during write failures
1847 Details : OSTs are not remounted read-only when the journal commit threads
1848 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1849 instead of the ext3 wrappers.
1853 Description: poor connect scaling with increasing client count
1854 Details : Don't run filter_grant_sanity_check for more than 100 exports
1855 to improve scaling for large numbers of clients.
1858 Frequency : SLES10 only
1860 Description: sanity-quota.sh quotacheck failed: rc = -22
1861 Details : Quotas cannot be enabled on SLES10.
1864 Frequency : liblustre clients only
1866 Description: getdirentries does not give error when run on compute nodes
1867 Details : getdirentries does not fail when the size specified as an argument
1868 is too small to contain at least one entry
1870 Severity : enhancement
1872 Description: Add LNET router traceability for debug purposes
1873 Details : If a checksum failure occurs with a router as part of the
1874 IO path, the NID of the last router that forwarded the bulk data
1875 is printed so it can be identified.
1880 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1881 Details : Due to a race condition, liblustre clients were occasionally
1882 evicted incorrectly.
1884 Severity : enhancement
1886 Description: lfs setstripe use optional parameters instead of postional
1889 Severity : enhancement
1891 Description: Nanosecond timestamp support for ldiskfs
1892 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1893 resolution timestamps. There is not yet support for this at
1894 the Lustre filesystem level.
1897 Frequency : during server recovery
1899 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1900 Details : request with rq_no_resend flag not awake l_wait_event if they get
1904 Frequency : nfs export on patchless client
1906 Description: connectathon hang when test nfs export over patchless client
1907 Details : Disconnected dentry cannot be found with lookup, so we do not need
1908 to unhash it or make it invalid
1911 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1912 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1913 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1918 Description: open req refcounting wrong on reconnect
1919 Details : If reconnect happened between getting open reply from server and
1920 call to mdc_set_replay_data in ll_file_open, we will schedule
1921 replay for unreferenced request that we are about to free.
1922 Subsequent close will crash in variety of ways.
1923 Check that request is still eligible for replay in
1924 mdc_set_replay_data().
1929 Description: disable writes to filesystem when reading health_check file
1930 Details : the default for reading the health_check proc file has changed
1931 to NOT do a journal transaction and write to disk, because this
1932 can cause reads of the /proc file to hang and block HA state
1933 checking on a healthy but otherwise heavily loaded system. It
1934 is possible to return to the previous behaviour during configure
1935 with --enable-health-write.
1937 Severity : enhancement
1939 Description: 64-bit inode version
1940 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1941 to the inode. This will be required for version-based recovery.
1946 Description: MDS fails to start if a duplicate client export is detected
1947 Details : in some rare cases it was possible for a client to connect to
1948 an MDS multiple times. Upon recovery the MDS would detect this
1949 and fail during startup. Handle this more gracefully.
1951 Severity : enhancement
1953 Description: Add -o localflock option to simulate old noflock
1955 Details : This will achieve local-only flock/fcntl locks
1961 Description: log_commit_thread vs filter_destroy race leads to crash
1962 Details : Take import reference before releasing llog record semaphore
1967 Description: Wrong request locking in request set processing
1968 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1969 imp_delayed_list, in this place should be used imp_lock.
1972 Frequency : when reconnection
1974 Description: Grant Leak when osc reconnect to OST
1975 Details : When osc reconnect ost, OST(filter) should check whether it
1976 should grant more space to client by comparing fed_grant and
1977 cl_avail_grant, and return the granted space to client instead
1978 of "new granted" space, because client will call osc_init_grant
1979 to update the client grant space info.
1982 Frequency : when client reconnect to OST
1984 Description: Grant Leak when osc do resend and replay bulk write
1985 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1986 bulk write request, because the grant info while be sync between
1987 OSC and OST when reconnect, and we should ignore the grant info
1988 these of resend/replay write req.
1993 Description: Grant space more than avaiable left space sometimes.
1994 Details : When then OST is about to be full, if two bulk writing from
1995 different clients came to OST. Accord the avaliable space of the
1996 OST, the first req should be permitted, and the second one
1997 should be denied by ENOSPC. But if the seconde arrived before
1998 the first one is commited. The OST might wrongly permit second
1999 writing, which will cause grant space > avaiable space.
2002 Frequency : when client is evicted
2004 Description: Grant might be wrongly erased when osc is evicted by OST
2005 Details : when the import is evicted by server, it will fork another
2006 thread ptlrpc_invalidate_import_thread to invalidate the
2007 import, where the grant will be set to 0. While the original
2008 thread will update the grant it got when connecting. So if
2009 the former happened latter, the grant will be wrongly errased
2010 because of this race.
2015 Description: Checking Stale with correct fid
2016 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2017 to check whether it is stale, because sometimes, we want the
2018 enqueue happened anyway, and op_data.fid2 will not be initialized.
2020 Severity : enhancement
2022 Description: update patchless client
2023 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2026 Frequency : only with 2.4 kernel
2028 Description: random memory corruption
2029 Details : size of struct ll_inode_info is to big for union inode.u and this
2030 can be cause of random memory corruption.
2035 Description: Memory leak in recovery
2036 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2037 It should also check obd_fail before fsfilt_start, otherwise if
2038 fsfilt_start return -EROFS,(failover mds during mds recovery).
2039 then the req will return with repmsg->transno = 0 and rc = EROFS.
2040 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2041 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2042 be freed no matter whether fsfilt_commit success or not.
2045 Frequency : only with huge count clients
2047 Description: Prevents from taking the superblock lock in llap_from_page for
2049 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2050 for llap_from_page call in ll_removepage prevents from taking the
2051 superblock lock for a soon died page.
2056 Description: Not check open intent error before release open handle
2057 Details : in some rare cases, the open intent error is not checked before
2058 release open handle, which may cause
2059 ASSERTION(open_req->rq_transno != 0), because it tries to release
2060 the failed open handle.
2065 Description: Set cat log bitmap only after create log success.
2066 Details : in some rare cases, the cat log bitmap is set too early. and it
2067 should be set only after create log success.
2071 Description: Accessing a block bevice can re-enable I/O when Lustre is
2072 tearing down a device.
2073 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2078 Description: service threads may hog cpus when there are a lot of requests
2079 Details : Insert cond_resched to give other threads a chance to use some CPU
2084 Description: the cat log was not initialized in recovery
2085 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2086 unlink log on mds will not be initialized until mds post
2087 recovery. And also in mds post recovery, the unlink log will
2088 initialization will be done asynchronausly, so there will be race
2089 between add unlink log and unlink log initialization.
2093 Description: brw_stats were being printed incorrectly
2094 Details : brw_stats were being printed as log2 but all of them were not
2095 recorded as log2. Also remove some code duplication arising from
2096 filter_tally_{read,write}.
2100 Frequency : rare, only in recovery.
2101 Description: ASSERTION(req->rq_type != LI_POISON) failed
2102 Details : imp_lock should be held while iterating over imp_sending_list for
2103 prevent destroy request after get timeout in ptlrpc_queue_wait.
2107 Description: replay-single.sh test 52 fails
2108 Details : A lock's skiplist need to be cleanup when it being unlinked
2109 from its resource list.
2113 Description: Short directio read returns full requested size rather than
2115 Details : Direct I/O operations should return actual amount of bytes
2116 transferred rather than requested size.
2118 Severity : enhancement
2120 Description: metadata RPC reduction (e.g. for rm performance)
2121 Details : decrease the amount of synchronous RPC between clients and servers
2122 by canceling conflicing lock before the operation on the client
2123 and packing thier handles into the main operation RPC to server.
2125 Severity : enhancement
2127 Description: Async OSC create to avoid the blocking unnecessarily.
2128 Details : If a OST has no remain object, system will block on the creating
2129 when need to create a new object on this OST. Now, ways use
2130 pre-created objects when available, instead of blocking on an
2131 empty osc while others are not empty. If we must block, we block
2132 for the shortest possible period of time.
2136 Description: improve handling recoverable errors
2137 Details : if request processig with error which can be recoverable on server
2138 request should be resend, otherwise page released from cache and
2141 Severity : enhancement
2143 Description: refine locking for avoid write wrong info into lov_objid file
2144 Details : fix possible races with add new target and write/update data in
2147 --------------------------------------------------------------------------------
2149 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2154 Frequency : on some architectures
2156 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2157 Details : Uninitialized flags sometimes cause configuration commands to
2161 Frequency : patchless clients only
2163 Description: missing __iget() symbol export
2164 Details : The __iget() symbol export is missing. To avoid the need for
2165 this on patchless clients the deathrow inode reaper is turned
2166 off, and we depend on the VM to clean up old inodes. This
2167 dependency was during via the fix for bug 12181.
2169 --------------------------------------------------------------------------------
2171 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2173 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2174 INTEROPERATE with older versions automatically. In many cases a
2175 special upgrade step is needed. Please read the
2176 user documentation before upgrading any part of a live system.
2177 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2178 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2179 * WARNING: Lustre configuration and startup changes are required with
2180 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2184 Severity : enhancement
2186 Description: MountConf
2187 Details : Lustre configuration is now managed via mkfs and mount
2188 commands instead of lmc and lconf. New obd types (MGS, MGC)
2189 are added for dynamic configuration management. See
2190 https://mail.clusterfs.com/wikis/lustre/MountConf for
2193 Severity : enhancement
2195 Description: dynamic OST addition
2196 Details : OSTs can now be added to a live filesystem
2198 Severity : enhancement
2200 Description: startup order invariance
2201 Details : MDTs and OSTs can be started in any order. Clients only
2202 require the MDT to complete startup.
2204 Severity : enhancement
2206 Description: parallel, asynchronous orphan cleanup
2207 Details : orphan cleanup is now performed in separate threads for each
2208 OST, allowing parallel non-blocking operation.
2210 Severity : enhancement
2212 Description: optimized stripe assignment
2213 Details : stripe assignments are now made based on ost space available,
2214 ost previous usage, and OSS previous usage, in order to try
2215 to optimize storage space and networking resources.
2217 Severity : enhancement
2219 Description: Permanently set tunables
2220 Details : All writable /proc/fs/lustre tunables can now be permanently
2221 set on a per-server basis, at mkfs time or on a live system.
2223 Severity : enhancement
2225 Description: Lustre message v2
2226 Details : Add lustre message format v2.
2228 Severity : enhancement
2230 Description: client OST exclusion list
2231 Details : Clients can be started with a list of OSTs that should be
2232 declared "inactive" for known non-responsive OSTs.
2236 Description: SPEC SFS validation failure on NFS v2 over lustre.
2237 Details : Changes the blocksize for regular files to be 2x RPC size,
2238 and not depend on stripe size.
2240 Severity : enhancement
2242 Description: Multiple MD RPCs in flight.
2243 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2244 To avoid overly-overloading MDS, introduce a limit on number of
2245 MDS RPCs in flight for a single client and add /proc controls
2246 to adjust this limit.
2248 Severity : enhancement
2250 Description: client read/write statistics
2251 Details : Add client read/write call usage stats for performance
2252 analysis of user processes.
2253 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2254 file access. extents_stats shows chunk size distribution.
2255 extents_stats_per_process show chunk size distribution per
2258 Severity : enhancement
2260 Description: mds statistics
2261 Details : Add detailed mds operations statistics in
2262 /proc/fs/lustre/mds/*/stats.
2266 Description: Failure of copying files with lustre special EAs.
2267 Details : Client side always return success for setxattr call for lustre
2268 special xattr (currently only "trusted.lov").
2273 Description: Refcount LNET uuids
2274 Details : The global LNET uuid list grew linearly with every startup;
2275 refcount repeated list entries instead of always adding to
2278 Severity : enhancement
2280 Description: Dynamic service threads
2281 Details : Within a small range, start extra service threads
2282 automatically when the request queue builds up.
2285 Frequency : mixed-endian client/server environments
2287 Description: mixed-endian crashes
2288 Details : The new msg_v2 system had some failures in mixed-endian
2291 Severity : enhancement
2293 Description: Easy OST removal
2294 Details : OSTs can be permanently deactivated with e.g. 'lctl
2295 conf_param lustre-OST0001.osc.active=0'
2297 Severity : enhancement
2299 Description: MGS proc entries
2300 Details : Added basic proc entries for the MGS showing what filesystems
2303 Severity : enhancement
2305 Description: provide MGS failover
2306 Details : Added config lock reacquisition after MGS server failover.
2308 Severity : enhancement
2310 Description: add Linux 2.4 support
2311 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2315 Description: a large application tries to do I/O to the same resource and dies
2316 in the middle of it.
2317 Details : Check the req->rq_arrival time after the call to
2318 ost_brw_lock_get(), but before we do anything about
2319 processing it & sending the BULK transfer request. This
2320 should help move old stale pending locks off the queue as
2321 quickly as obd_timeout.
2324 Frequency : when an incorrect nid is specified during startup
2326 Description: ptlrpc connect to non-existant node causes kernel crash
2327 Details : LNET can't be re-entered from an event callback, which
2328 happened when we expire a message after the export has been
2329 cleaned up. Instead, hand the zombie cleanup off to another
2332 Severity : enhancement
2334 Description: plain/inodebits lock performance improvement
2335 Details : Grouping plain/inodebits in granted list by their request modes
2336 and bits policy, thus improving the performance of search through
2340 Frequency : only if OST filesystem is corrupted
2342 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2343 Details : for a short time RPCs with bulk IO are in the replay list,
2344 but replay of bulk IOs is unimplemented. If the OST filesystem
2345 is corrupted due to disk cache incoherency and then replay is
2346 started it is possible to trip an assertion. Avoid putting
2347 committed RPCs into the replay list at all to avoid this issue.
2350 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2352 Description: System hang on startup
2353 Details : This bug allowed the liblustre (e.g. catamount) client to
2354 return to the app before handling all startup RPCs. This
2355 could leave the node unresponsive to lustre network traffic
2356 and manifested as a server ptllnd timeout.
2358 Severity : enhancement
2360 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2361 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2362 boolean to control whether to print peer debug info when a
2363 client's RPC times out.
2366 Frequency : only for kernels with patches from Lustre below 1.4.3
2368 Description: Remove old rdonly API
2369 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2372 Frequency : only for devices with external journals
2374 Description: Set external device read-only also
2375 Details : During a commanded failover stop, we set the disk device
2376 read-only while the server shuts down. We now also set any
2377 external journal device read-only at the same time.
2380 Frequency : when upgrading from 1.4 while trying to change parameters
2382 Description: The wrong (new) MDC name was used when setting parameters for
2383 upgraded MDT's. Also allows changing of OSC (and MDC)
2384 parameters if --writeconf is specified at tunefs upgrade time.
2387 Frequency : when setting specific ost indicies
2389 Description: QOS code breaks on skipped indicies
2390 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2391 created with --index need not be sequential.
2393 Severity : enhancement
2395 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2396 Details : The uninit_groups feature works in conjunction with the kernel
2397 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2398 up the pass1 processing of e2fsck. This is a read-only feature
2399 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2400 mount filesystems that have had this feature enabled.
2402 Severity : enhancement
2404 Description: Improve multi-block allocation algorithm to avoid fragmentation
2405 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2406 allocation locality and avoid filesystem fragmentation.
2408 ------------------------------------------------------------------------------
2410 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2412 * Support for kernels:
2413 2.4.21-47.0.1.EL (RHEL 3)
2414 2.6.5-7.283 (SLES 9)
2415 2.6.9-42.0.10.EL (RHEL 4)
2416 2.6.12.6 vanilla (kernel.org)
2417 2.6.16.27-0.9 (SLES 10)
2418 * Recommended e2fsprogs version: 1.39.cfs5
2420 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2424 Frequency : occasional, depends on client load and configuration
2425 Bugzilla : 12181, 12203
2426 Description: data loss for recently-modified files
2428 Details : In some cases it is possible that recently written or created
2429 files may not be written to disk in a timely manner (this should
2430 normally be within 30s unless client IO load is very high).
2431 The problem appears as zero-length files or files that are a
2432 multiple of 1MB in size after a client crash or client eviction
2433 that are missing data at the end of the file.
2435 This problem is more likely to be hit on clients where files are
2436 repeatedly created and unlinked in the same directory, clients
2437 have a large amount of RAM, have many CPUs, the filesystem has
2438 many OSTs, the clients are rebooted frequently, and/or the files
2439 are not accessed by other nodes after being written.
2441 The presence of the problem can be detected by looking at
2442 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2443 smaller than the second (nr_unused) then dirty files will not
2444 be flushed automatically to disk. "sync; sleep 10" should be
2445 run several times on the node before unmounting it to update
2446 Lustre (this is also safe to run on nodes without this problem).
2448 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2449 that can cause this same problem, so customers using that kernel
2450 also need to update the kernel in addition to Lustre. In order
2451 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2453 It is normal that files written just before a client crash (less
2454 than 30s) may not yet have been flushed to disk, even for local
2458 Frequency : frequent on thin XT3 nodes
2460 Description: UUID collision on thin XT3 Linux nodes
2461 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2462 randomly, since we relied on an insufficiently-seeded PRNG.
2467 Description: OSS hangs after "All ost request buffers busy"
2468 Details : A deadlock between quota and journal operations caused OSS
2469 hangs after printing "All ost request buffers busy."
2472 Frequency : always on liblustre builds
2474 Description: Cleanup compiler warnings on liblustre
2477 Frequency : always on liblustre builds on XT3
2479 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2483 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2484 > 32000 subdirectories
2485 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2486 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2487 subdirectory count crosses 32000. This will aid e2fsck to
2488 correctly handle more than 32000 subdirectories.
2491 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2493 Description: System hang on startup
2494 Details : This bug allowed the liblustre (e.g. catamount) client to
2495 return to the app before handling all startup RPCs. This
2496 could leave the node unresponsive to lustre network traffic
2497 and manifested as a server ptllnd timeout.
2499 Severity : enhancement
2501 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2502 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2503 boolean to control whether to print peer debug info when a
2504 client's RPC times out.
2509 Description: make O_SYNC working on 2.6 kernels
2510 Details : 2.6 kernels use different method for mark pages for write,
2511 so need add a code to lustre for O_SYNC work.
2516 Description: Failure to close file and release space on NFS
2517 Details : Put inode details into lock acquired in ll_intent_file_open.
2518 Use mdc_intent_lock in ll_intent_open to properly
2519 detect all kind of errors unhandled by mdc_enqueue.
2524 Description: proc file read during shutdown sometimes raced obd removal,
2526 Details : Add lock to prevent obd access after proc file removal.
2529 Frequency : Only for files larger than 4GB on 32-bit clients.
2531 Description: improperly doing page alignment of locks
2532 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2533 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2538 Description: RPCs being resent when they shouldn't be
2539 Details : Some RPCs that should not be resent are being resent. This
2540 can cause inconsistencies in the RPC state machine. Do not
2541 resend such requests.
2544 Frequency : rare, only with NFS export
2546 Description: Crash on NFS re-export node
2547 Details : under very unusual load conditions an assertion is hit in
2548 ll_intent_file_open()
2551 Frequency : only if OST filesystem is corrupted
2553 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2554 Details : for a short time RPCs with bulk IO are in the replay list,
2555 but replay of bulk IOs is unimplemented. If the OST filesystem
2556 is corrupted due to disk cache incoherency and then replay is
2557 started it is possible to trip an assertion. Avoid putting
2558 committed RPCs into the replay list at all to avoid this issue.
2563 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2564 Details : Large single O_DIRECT read and write calls can fail to allocate
2565 a sufficiently large buffer to process the request. In case of
2566 allocation failure the allocation is retried with a smaller
2567 buffer and broken into smaller requests.
2569 Severity : enhancement
2571 Description: Add -o localflock option to simulate old noflock behaviour.
2572 Details : This will achieve local-only flock/fcntl locks coherentness.
2577 Description: versioning check is incomplete
2578 Details : Checking the version difference of client vs. server, report
2579 error if the gap is too big.
2584 Description: add support PG_writeback bit
2585 Details : add support for PG_writeback bit for Lustre, for more carefull
2586 work with page cache in 2.6 kernel. This also fix some deadlocks
2587 and remove hack for work O_SYNC with 2.6 kernel.
2589 Severity : enhancement
2591 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2592 Details : The uninit_groups feature works in conjunction with the kernel
2593 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2594 up the pass1 processing of e2fsck. This is a read-only feature
2595 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2596 mount filesystems that have had this feature enabled.
2598 Severity : enhancement
2600 Description: Improve multi-block allocation algorithm to avoid fragmentation
2601 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2602 allocation locality and avoid filesystem fragmentation.
2604 ------------------------------------------------------------------------------
2606 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2608 * Support for kernels:
2609 2.6.9-42.0.3.EL (RHEL 4)
2610 2.6.5-7.276 (SLES 9)
2611 2.4.21-47.0.1.EL (RHEL 3)
2612 2.6.12.6 vanilla (kernel.org)
2613 2.6.16.21-0.8 (SLES10)
2614 * Recommended e2fsprogs version: 1.39.cfs2-0
2616 * The backwards-compatible /proc/sys/portals symlink has been removed
2617 in this release. Before upgrading, please ensure that you change
2618 any configuration scripts or /etc/sysctl.conf files that access
2619 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2620 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2621 in advance of the upgrade on any system running Lustre 1.4.6 or
2622 newer, since /proc/sys/lnet was added in that version.
2623 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2627 Frequency : only when quota is used
2629 Description: avoid scanning export list for quota master
2630 Details : Change the algorithms to avoid scanning export list in order
2631 to improve the efficiency.
2634 Frequency : MDS failover only, very rarely
2636 Description: "went back in time" messages on mds failover
2637 Details : The greatest transno may be lost when the current operation
2638 finishes with an error (transno==0) and the client's last_rcvd
2639 record is over-written. Save the greatest transno in the
2640 mds_last_transno for this case.
2643 Frequency : always for specific kernels and striping counts
2645 Description: client may get "Matching packet too big" without ACL support
2646 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2647 when trying to access files in certain configurations. The
2648 clients should in fact be denied when mounting because they do
2649 not understand ACLs.
2652 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2654 Description: many clients connecting with IO in progress causes connect timeouts
2655 Details : Avoid synchronous journal commits to avoid delays caused by many
2656 clients connecting/disconnecting when bulk IO is in progress.
2657 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2658 OST_IO_PORTAL to avoid delays behind potentially many pending
2662 Frequency : occasionally with multiple writers to a single file
2664 Description: shared writes to file may result in wrong size reported by stat()
2665 Details : Allow growing of kms when extent lock is cancelled
2668 Frequency : always with random mmap IO to multi-striped file
2670 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2671 Details : Only if the hole is at the end of OST object so that kms is too
2672 small. Fix is to increase kms accordingly in ll_nopage.
2675 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2677 Description: writes to a missing object would leak memory on the OST
2678 Details : If there is an inconsistency between the MDS and OST filesystems,
2679 such that the MDS references an object that doesn't exist, writes
2680 to that object will leak memory due to incorrect cleanup in the
2681 error handling path, eventually running out of memory on the OST.
2686 Description: Creating too long symlink causes lustre errors
2687 Details : Check symlink and name lengths before sending requests to MDS.
2690 Frequency : only if flock is enabled (not on by default)
2692 Description: posix locks not released on fd closure on 2.6.9+
2693 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2694 kernels, this caused such locks not to be released on fd close and
2695 then assertions on fs unmount about still used locks.
2698 Frequency : MDS failover only, very rarely
2700 Description: clients may get ASSERTION(granted_lock != NULL)
2701 Details : When request was taking a long time, and a client was resending
2702 a getattr by name lock request. The were multiple lock requests
2703 with the same client lock handle and
2704 mds_getattr_name->fixup_handle_for_resent_request found one of the
2705 lock handles but later failed with ASSERTION(granted_lock != NULL).
2710 Description: handle->h_buffer_credits > 0, assertion failure
2711 Details : h_buffer_credits is zero after truncate, causing assertion
2712 failure. This patch extends the transaction or creates a new
2716 Frequency : NFS re-export or patchless client
2717 Bugzilla : 11179, 10796
2718 Description: Crash on NFS re-export node (__d_move)
2719 Details : We do not want to hash the dentry if we don't have a lock.
2720 But if this dentry is later used in d_move, we'd hit uninitialised
2721 list head d_hash, so we just do this to init d_hash field but
2722 leave dentry unhashed.
2725 Frequency : NFS re-export or patchless client
2727 Description: NFS exports has problem with symbolic link
2728 Details : lustre client didn't properly install dentry when re-exported
2729 to NFS or running patchless client.
2732 Frequency : NFS re-export or patchless client
2734 Description: Various nfs/patchless fixes.
2735 Details : fixes reuse disconected alias for lookup process - this fixes
2736 warning "find_exported_dentry: npd != pd",
2737 fix permission error with open files at nfs.
2738 fix apply umask when do revalidate.
2741 Frequency : occasional
2743 Description: Crash on NFS re-export node
2744 Details : calling clear_page() on the wrong pointer triggered oops in
2745 generic_mapping_read().
2748 Frequency : rarely, using O_DIRECT IO
2750 Description: unaligned directio crashes client with LASSERT
2751 Details : check for unaligned buffers before trying any requests.
2754 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2756 Description: stale data returned from RAID cache
2757 Details : If only a small amount of IO is done to the RAID device before
2758 reading it again it is possible to get stale data from the RAID
2759 cache instead of reading it from disk.
2762 Frequency : always for sles10 kernel
2764 Description: sles10 support
2765 Details : ll_follow_link: compile fixes and using of nd_set_link
2766 under newer kernels.
2769 Frequency : depends on arch, kernel and compiler version, always on sles10
2772 Description: recursive or deep enough symlinks cause stack overflow
2773 Details : getting rid of large stack-allocated variable in
2777 Frequency : depends on hardware
2779 Description: lustre write performance loss in the SLES10 kernel
2780 Details : the performance loss is caused by using of write barriers in the
2781 ext3 code. The SLES10 kernel turns barrier support on by
2782 default. The fix is to undo that change for ldiskfs.
2784 ------------------------------------------------------------------------------
2786 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2788 * Support for kernels:
2789 2.6.9-42.0.3EL (RHEL 4)
2790 2.6.5-7.276 (SLES 9)
2791 2.4.21-47.0.1.EL (RHEL 3)
2792 2.6.12.6 vanilla (kernel.org)
2796 Frequency : quota enabled and large files being deleted
2798 Description: releasing more than 4GB of quota at once hangs OST
2799 Details : If a user deletes more than 4GB of files on a single OST it
2800 will cause the OST to spin in an infinite loop. Release
2801 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2806 Description: statfs data retrieved from /proc may be stale or zero
2807 Details : When reading per-device statfs data from /proc, in the
2808 {kbytes,files}_{total,free,avail} files, it may appear
2809 as zero or be out of date.
2812 Frequency : systems with MD RAID1 external journal devices
2814 Description: lconf's call to blkid is confused by RAID1 journal devices
2815 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2816 of returning all block devices that match the journal UUID.
2819 Frequency : always, for aggregate stripe size over 4GB
2821 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2822 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2823 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2826 Frequency : always if "lfs find" used on a local file/directory
2828 Description: "lfs find" segfaults if used on a local file/directory
2829 Details : The case where a directory component was not specified wasn't
2830 handled correctly. Handle this properly.
2833 Frequency : always on ppc64
2835 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2836 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2837 perform endianness conversion.
2840 Frequency : rarely (truncate to non-zero file size after write under load)
2841 Bugzilla : 10730, 10687
2842 Description: Files padded with zeros to next 4K multiple
2843 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2844 it is possible that files that are truncated to a non-zero size
2845 immediately after being written are filled with zero bytes beyond
2846 the truncated size. No file data is lost.
2848 Severity : enhancement
2850 Description: Allow recovery/failover for liblustre clients.
2851 Details : liblustre clients were unaware of failover configurations until
2854 Severity : enhancement
2856 Description: user file locks should fail when not mounting with flock option
2857 Details : Set up an error-returning stub in ll_file_operations.lock field
2858 to prevent incorrect behaviour when client is mounted without
2859 flock option. Also, set up properly f_op->flock field for
2863 Frequency : always on ia64
2865 Description: "lfs df" loops on printing out MDS statfs information
2866 Details : The obd_ioctl_data was not initialized and in some systems
2867 this caused a failure during the ioctl that did not return
2868 an error. Initialize the struct and return an error on failure.
2871 Frequency : SLES 9 only
2873 Description: Error of copying files with lustre special EAs as root
2874 Details : Client side always return success for setxattr call for lustre
2875 special xattr (currently only "trusted.lov").
2878 Frequency : rarely on clusters with both ia64+i386 clients
2880 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2881 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2882 clients are doing concurrent writes to the same file it is
2883 possible that the ia64 clients may LASSERT because the OST
2884 extent locks are not PAGE_SIZE aligned. Ensure that grown
2885 locks are always aligned on the request boundary.
2888 Frequency : specific use, occasional
2890 Description: Overwriting in use executable truncates on-disk binary image
2891 Details : If one node attempts to overwrite an executable in use by
2892 another node, we now correctly return ETXTBSY instead of
2893 truncating the file.
2895 Severity : enhancement
2897 Description: Async OSC create to avoid the blocking unnecessarily.
2898 Details : If a OST has no remain object, system will block on the creating
2899 when need to create a new object on this OST. Now, ways use
2900 pre-created objects when available, instead of blocking on an
2901 empty osc while others are not empty. If we must block, we block
2902 for the shortest possible period of time.
2907 Description: chmod on Lustre root is propagated to other clients
2908 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2909 invalid by the follow_mount time.
2912 Frequency : liblustre clients only
2914 Description: Race in 'instant cancel' lock handling could lead to such locks
2915 never to be granted in case of SMP MDS
2916 Details : Do not destroy not yet granted but cbpending locks in
2920 Frequency : replay/resend of open
2922 Description: non null lock assetion failure in mds_intent_policy
2923 Details : Trying to replay/resend lockless open requests resulted in
2924 mds_open() returning 0 with no lock. Now it sets a flag if
2925 a lock is going to be returned.
2927 Severity : enhancement
2929 Description: Checksum enhancements
2930 Details : New checksum enhancements allow for resending RPCs that failed
2933 Severity : enhancement
2935 Description: Tunables on number of dirty pages in cacche
2936 Details : Allow to set limit on number of dirty pages cached.
2941 Description: client crash on unmount - lock still has references
2942 Details : In some error handling cases it was possible to leak a lock
2943 reference on a client while accessing a file. This was not
2944 harmful to the client during operation, but would cause the
2945 client to crash when the filesystem is unmounted.
2948 Frequency : specific case, rare
2950 Description: ETXTBSY on mds though file not in use by client
2951 Details : ETXTBSY is no longer incorrectly returned when attempting to
2952 chmod or chown a directory that the user previously tried to
2953 execute or a currently-executing binary.
2956 Frequency : extremely rare except on liblustre-based clients
2958 Description: Lustre space not freed when files are deleted
2959 Details : Clean up open-unlinked files after client eviction. Previously
2960 the unlink was skipped and the files remained as orphans.
2965 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2966 Details : In some cases it was possible to send a blocking callback to
2967 a client doing a glimpse, even though that client didn't get
2968 a lock granted. When the glimpse lock is cancelled on the OST
2969 the freed lock is left on the waiting list and corrupted the list.
2972 Frequency : all core dumps
2974 Description: Broke core dumps to lustre
2975 Details : Negative dentry may be unhashed if parent does not have UPDATE
2976 lock, but some callers, e.g. do_coredump, expect dentry to be
2977 hashed after successful create, hash it in ll_create_it.
2979 ------------------------------------------------------------------------------
2981 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2983 * Support for kernels:
2984 2.6.9-42.0.2.EL (RHEL 4)
2985 2.6.5-7.276 (SLES 9)
2986 2.4.21-40.EL (RHEL 3)
2987 2.6.12.6 vanilla (kernel.org)
2991 Frequency : always on RHEL 3
2993 Description: Number of open files grows over time
2994 Details : The number of open files grows over time, whether or not
2995 Lustre is started. This was due to a filp leak introduced
2996 by one of our kernel patches.
2998 ------------------------------------------------------------------------------
3000 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3002 * Support for kernels:
3003 2.6.9-42.EL (RHEL 4)
3004 2.6.5-7.276 (SLES 9)
3005 2.4.21-40.EL (RHEL 3)
3006 2.6.12.6 vanilla (kernel.org)
3011 Bugzilla : 5719, 9635, 9792, 9684
3012 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3013 Details : If a server is under heavy load and cannot reply to new
3014 connection requests before the client resends the (re)connect,
3015 the connection handling code can behave badly if two service
3016 threads are concurrently handing separate (re)connections from
3017 the same client. Add better locking to the connection handling
3018 code, and ensure that only a single connection will be processed
3019 for a given client UUID, even if the lock is dropped.
3021 Severity : enhancement
3023 Description: add TCP zero-copy support to kernel
3024 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3025 sends if the hardware supports scatter-gather and checksumming.
3026 This allows socklnd to do client-write and server-read more
3027 efficiently and reduce CPU utilization from skbuf copying.
3030 Frequency : only if NFS exporting from client
3032 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3033 Details : If setting attributes on a file created under NFS that had
3034 never been opened it would be possible to oops the client
3035 if the file had no objects.
3038 Frequency : always for liblustre
3040 Description: liblustre client does MDS+OSTs setattr RPC for each write
3041 Details : When doing a write from a liblustre client, the client
3042 incorrectly issued an RPC to the MDS and each OST the file was
3043 striped over in order to update the timestamps. When writing
3044 with small chunks and many clients this could overwhelm the MDS
3045 with RPCs. In all cases it would slow down the write because
3046 these RPCs are unnecessary.
3048 Severity : enhancement
3050 Description: allow number of MDS service threads to be changed at module load
3051 Details : It is now possible to change the number of MDS service threads
3052 running. Adding "options mds mds_num_threads={N}" to the MDS's
3053 /etc/modprobe.conf will set the number of threads for the next
3054 time Lustre is restarted (assuming the "mds" module is also
3055 reloaded at that time). The default number of threads will
3056 stay the same, 32 for most systems.
3061 Description: OST crash if filesystem is unformatted or corrupt
3062 Details : If an OST is started on a device that has never been formatted
3063 or if the filesystem is corrupt and cannot even mount then the
3064 error handling cleanup routines would dereference a NULL pointer.
3069 Description: NULL pointer deref in llap_from_page.
3070 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3071 incorrect) mapping if the page was truncated/reclaimed while it was
3072 searched for. Check for this condition and skip such pages when
3073 doing readahead. Introduce extra check to llap_from_page() to
3074 verify page->mapping->host is non-NULL (so page is not anonymous).
3077 Frequency : Sometimes when using sys_sendfile
3079 Description: "page not covered by a lock" warnings from ll_readpage
3080 Details : sendfile called ll_readpage without right page locks present.
3081 Now we introduced ll_file_sendfile that does necessary locking
3082 around call to generic_file_sendfile() much like we do in
3086 Frequency : with certain MDS communication failures at client mount time
3088 Description: NULL pointer deref after failed client mount
3089 Details : a client connection request may delayed by the network layer
3090 and not be sent until after the PTLRPC layer has timed out the
3091 request. If the client fails the mount immediately it will try
3092 to clean up before the network times out the request. Add a
3093 reference from the request import to the obd device and delay
3094 the cleanup until the network drops the request.
3097 Frequency : occasionally during client (re)connect
3099 Description: assertion failure during client (re)connect
3100 Details : processing a client connection request may be delayed by the
3101 client or server longer than the client connect timeout. This
3102 causes the client to resend the connection request. If the
3103 original connection request is replied in this interval, the
3104 client may trip an assertion failure in ptlrpc_connect_interpret()
3105 which thought it would be the only running connect process.
3108 Frequency : only with obd_echo servers and clients that are rebooted
3110 Description: kernel BUG accessing uninitialized data structure
3111 Details : When running an obd_echo server it did not start the ping_evictor
3112 thread, and when a client was evicted an uninitialized data
3113 structure was accessed. Start the ping_evictor in the RPC
3114 service startup instead of the OBD startup.
3116 Severity : enhancement
3117 Bugzilla : 10193 (patchless)
3118 Description: Remove dependency on various unexported kernel interfaces.
3119 Details : No longer need reparent_to_init, exit_mm, exit_files,
3120 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3123 Frequency : rare (only users of deprecated and unsupported LDAP config)
3125 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3126 Details : LDAP apparently contains 'lustreName' attributes instead of
3127 'name'. A simple remapping of the name is sufficient.
3130 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3132 Description: waiting_locks_callback trips kernel BUG if client is evicted
3133 Details : Running with the dump_on_timeout debug flag turned on makes
3134 it possible that the waiting_locks_callback() can try to dump
3135 the Lustre kernel debug logs from an interrupt handler. Defer
3136 this log dumping to the expired_lock_main() thread.
3138 Severity : enhancement
3140 Description: Support NFS exporting on 2.6 kernels.
3141 Details : Implement non-rawops metadata methods for NFS server to use without
3142 changing NFS server code.
3145 Frequency : very rare (synthetic metadata workload only)
3147 Description: two racing renames might cause an MDS thread to deadlock
3148 Details : Running the "racer" program may cause one MDS thread to rename
3149 a file from being the source of a rename to being the target of
3150 a rename at exactly the same time that another thread is doing
3151 so, and the second thread has already enqueued these locks after
3152 doing a lookup of the target and is trying to relock them in
3153 order. Ensure that we don't try to re-lock the same resource.
3156 Frequency : only very large systems with liblustre clients
3158 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3159 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3160 all OSTs in parallel. This allows the eviction of stale liblustre
3161 clients to proceed much faster than if they were done in series,
3162 and also offers similar improvements for other set_info RPCs.
3167 Description: excessive CPU usage during initial read phase on client
3168 Details : During the initial read phase on a client, it would agressively
3169 retry readahead on the file, consuming too much CPU and impacting
3170 performance (since 1.4.5.8). Improve the readahead algorithm
3171 to avoid this, and also improve some other common cases (read
3172 of small files in particular, where "small" is files smaller than
3173 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3178 Description: MDS crash when receiving packet with unknown intent.
3179 Details : Do not LBUG in unknown intent case, just return -EFAULT
3181 Severity : enhancement
3182 Bugzilla : 9293, 9385
3183 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3184 Details : Do not serialize getattr (non-intent version) and statfs.
3187 Frequency : occasional, when OST network is overloaded/intermittent
3189 Description: client evicted by OST after bulk IO timeout
3190 Details : If a client sends a bulk IO request (read or write) the OST
3191 may evict the client if it is unresposive to its data GET/PUT
3192 request. This is incorrect if the network is overloaded (takes
3193 too long to transfer the RPC data) or dropped the OST GET/PUT
3194 request. There is no need to evict the client at all, since
3195 the pinger and/or lock callbacks will handle this, and the
3196 client can restart the bulk request.
3199 Frequency : Always when mmapping file with no objects
3201 Description: client crashes when mmapping file with no objects
3202 Details : Check that we actually have objects in a file before doing any
3203 operations on objects in ll_vm_open, ll_vm_close and
3209 Description: Request leak when working with deleted CWD
3210 Details : Introduce advanced request refcount tracking for requests
3211 referenced from lustre intent.
3213 Severity : Enhancement
3215 Description: Cache open file handles on client.
3216 Details : MDS now will return special lock along with openhandle, if
3217 requested and client is allowed to hold openhandle, even if unused,
3218 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3219 closing files for every read/write openration.
3221 Severity : Enhancement
3223 Description: Cache open negative dentries on client when possible.
3224 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3225 negative dentries on lock revocation.
3230 Description: Remounting a client read-only wasn't possible with a zconf mount
3231 Details : It wasn't possible to remount a client read-only with llmount.
3233 Severity : enhancement
3234 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3235 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3236 MPICH is included to add Lustre-specific ADIO interfaces.
3237 This is based closely on the UFS ADIO layer and only differs
3238 in file creation, in order to allow the OST striping to be set.
3239 This is user-contributed code and not supported by CFS.
3244 Description: extended inode attributes (immutable, append-only) work improperly
3245 when 2.4 and 2.6 kernels are used on client/server or vice versa
3246 Details : Introduce kernel-independent values for these flags.
3248 Severity : enhancement
3251 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3252 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3253 granularity. Now, Fractional MB granularity is be supported,
3254 this is very useful for low memory system.
3256 Severity : enhancement
3258 Description: Getattr by fid
3259 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3260 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3261 have locks and avoids getattr rpc requests for it on every stat.
3264 Frequency : Always, for filesystems larger than 2TB
3266 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3267 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3268 filesystems larger than 2TB to be mounted with mballoc enabled.
3271 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3273 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3274 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3275 filesystems larger than 2TB is mounted then the kernel will
3276 silently corrupt the start of the filesystem. CONFIG_LBD is
3277 enabled for all CFS-supported kernels, but the possibility of
3278 this happening with a modified kernel config exists.
3280 Severity : enhancement
3282 Description: add client O_DIRECT support for 2.6 kernels
3283 Details : It is now possible to do O_DIRECT reads and writes to files
3284 in the Lustre client mountpoint on 2.6 kernel clients.
3286 Severity : enhancement
3288 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3289 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3290 parallel, not waiting for response from every OST before sending
3291 a rpc to the next OST.
3296 Description: setattr vs write race when updating file timestamps
3297 Details : Client processes that update a file timestamp into the past
3298 right after writing to the file (e.g. tar) it is possible that
3299 the updated file modification time can be reset to the current
3300 time due to a race between processing the setattr and write RPC.
3302 Severity : enhancement
3304 Description: Bring 'lfs find' closer in line with regular Linux find.
3305 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3306 -print0 options and obtains all the needed info through the lustre
3309 Severity : enhancement
3311 Description: support up to 1024 configured devices on one node
3312 Details : change obd_dev array from statically allocated to dynamically
3313 allocated structs as they are first used to reduce memory usage
3318 Description: Flush dirty partially truncated pages during truncate
3319 Details : Immediatelly flush partially truncated pages in filter_setattr,
3320 this way we completely avoid having any pages in page cache on OST
3321 and can retire ugly workarounds during writes to flush such pages.
3326 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3327 Details : protect inode from truncation within vfs_unlink() context
3328 just take a reference before calling vfs_unlink() and release it
3329 when parent's i_sem is free.
3334 Description: last_id value checked outside lock on OST caused LASSERT failure
3335 Details : If there were multiple MDS->OST object precreate requests in
3336 flight, it was possible that the OST's last object id was checked
3337 outside a lock and incorrectly tripped an assertion. Move checks
3338 inside locks, and discard old precreate requests.
3341 Frequency : always, if extents are used on OSTs
3343 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3344 Details : OSTs using the extents format would not zero the high 16 bits of
3345 the index physical block number. This is not a problem for any
3346 OST filesystems smaller than 16TB, and no kernels support ext3
3347 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3348 new/modified files) and can be fixed for existing filesystems
3349 with e2fsprogs-1.39-cfs1.
3354 Description: import connection selection may be incorrect if timer wraps
3355 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3356 import connections to be ignored if the 32-bit jiffies counter
3357 wraps. Use a 64-bit jiffies counter.
3360 Frequency : very large clusters immediately after boot
3362 Description: LNET request buffers exhausted under heavy short-term load
3363 Details : If a large number of client requests are generated on a service
3364 that has previously never seen so many requests it is possible
3365 that the request buffer growth cannot keep up with the spike in
3366 demand. Instead of dropping incoming requests, they are held in
3367 the LND until the RPC service can accept more requests.
3370 Frequency : Sometimes during replay
3372 Description: Assertion failure in ll_local_open after replay.
3373 Details : If replay happened on an open request reply before we were able
3374 to set replay handler, reply will become not swabbed tripping the
3375 assertion in ll_local_open. Now we set the handler right after
3376 recognising of open request
3379 Frequency : very rare
3381 Description: kernel reports "badness in vsnprintf"
3382 Details : Reading from the "recovery_status" /proc file in small chunks
3383 may cause a negative length in lprocfs_obd_rd_recovery_status()
3384 call to vsnprintf() (which is otherwise harmless). Exit early
3385 if there is no more space in the output buffer.
3387 Severity : enhancement
3389 Description: clear OBD RPC statistics by writing to them
3390 Details : It is now possible to clear the OBD RPC statistics by writing
3391 to the "stats" file.
3396 Description: Client mtime is not the same on different clients after utimes
3397 Details : In some cases, the client was using the utimes() syscall on
3398 a file cached on another node. The clients now validate the
3399 ctime from the MDS + OSTs to determine which one is right.
3404 Description: Inability to activate failout mode
3405 Details : lconf script incorrectly assumed that in pythong string's numeric
3406 value is used in comparisons.
3409 Frequency : always with multiple stripes per file
3411 Description: Inefficient object allocation for mutli-stripe files
3412 Details : When selecting which OSTs to stripe files over, for files with
3413 a stripe count that divides evenly into the number of OSTs,
3414 the MDS is always picking the same starting OST for each file.
3415 Return the OST selection heuristic to the original design.
3420 Description: mount failures may take full timeout to return an error
3421 Details : Under some heavy load conditions it is possible that a
3422 failed mount can wait for the full obd_timeout interval,
3423 possibly several minutes, before reporting an error.
3424 Instead return an error as soon as the status is known.
3426 Frequency : quota enabled and large files being deleted
3428 Description: releasing more than 4GB of quota at once hangs OST
3429 Details : If a user deletes more than 4GB of files on a single OST it
3430 will cause the OST to spin in an infinite loop. Release
3431 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3436 Description: statfs data retrieved from /proc may be stale or zero
3437 Details : When reading per-device statfs data from /proc, in the
3438 {kbytes,files}_{total,free,avail} files, it may appear
3439 as zero or be out of date.
3442 Frequency : systems with MD RAID1 external journal devices
3444 Description: lconf's call to blkid is confused by RAID1 journal devices
3445 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3446 of returning all block devices that match the journal UUID.
3449 Frequency : always, for aggregate stripe size over 4GB
3451 Description: assertion fails when trying to use 4GB stripe size
3452 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3453 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3456 Frequency : always on ppc64
3458 Description: the first write on an ext3 filesystem with mballoc got stuck
3459 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3460 perform endianness conversion.
3462 ------------------------------------------------------------------------------
3464 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3466 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3467 INTEROPERATE with older versions automatically. Please read the
3468 user documentation before upgrading any part of a live system.
3469 * WARNING: Lustre networking configuration changes are required with
3470 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3473 * Support for kernels:
3474 2.6.9-22.0.2.EL (RHEL 4)
3475 2.6.5-7.244 (SLES 9)
3476 2.6.12.6 vanilla (kernel.org)
3479 Severity : enhancement
3480 Bugzilla : 7981/8208
3481 Description: Introduced Lustre Networking (LNET)
3482 Details : LNET is new networking infrastructure for Lustre, it includes
3483 a reorganized network configuration mode (see the user
3484 documentation for full details) as well as support for routing
3485 between different network fabrics. Lustre Networking Devices
3486 (LNDs) for the supported network fabrics have also been
3487 created for this new infrastructure.
3489 Severity : enhancement
3490 Description: Introduced Access control lists
3491 Details : clients can set ACLs on files and directories in order to have
3492 more fine-grained permissions than the standard Unix UGO+RWX.
3493 The MDS must be started with the "-o acl" mount option.
3495 Severity : enhancement
3496 Description: Introduced filesystem quotas
3497 Details : Administrators may now establish per-user quotas on the
3500 Severity : enhancement
3502 Description: Configuration change for the XT3
3503 The PTLLND is now used to run Lustre over Portals on the XT3
3504 The configure option(s) --with-cray-portals are no longer used.
3505 Rather --with-portals=<path-to-portals-includes> is used to
3506 enable building on the XT3. In addition to enable XT3 specific
3507 features the option --enable-cray-xt3 must be used.
3512 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3513 Details : A race between allocating a new llcd and re-getting the llcd_lock
3514 allowed another thread to grab newly-allocated llcd.
3516 Severity : enhancement
3518 Description: 2.6 OST async journal commit and locking fix to improve performance
3519 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3520 2.6 kernels are now async as they already were in 2.4 kernels so
3521 that they can commit concurrently with the network bulk transfer.
3522 For block-allocated files the filter allocation semaphore is held
3523 to avoid filesystem fragmentation during allocation. BKL lock
3524 removed for 2.6 xattr operations where it is no longer needed.
3529 Description: lconf incorrectly determined whether two IP networks could talk
3530 Details : In some more complicated routing and multiple-network
3531 configurations, lconf will avoid trying to make a network
3532 connection to a disjoint part of the IP space. It was doing the
3533 math incorrectly for one set of cases.
3538 Description: Fix for potential infinite loop processing records in an llog.
3539 Details : If an llog record is corrupted/zeroed, it is possible to loop
3540 forever in llog_process(). Validate the llog record length
3541 and skip the remainder of the block on error.
3544 Frequency : occasional (liblustre only)
3546 Description: liblustre could not open files whose last component is a symlink
3547 Details : sysio_path_walk() would incorrectly pass the open intent to
3548 intermediate path components.
3551 Frequency : rare (liblustre only with non-standard tuning)
3552 Bugzilla : 7201 (7350)
3553 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3554 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3555 after it has been released, only do this for referenced locks
3557 Severity : enhancement
3559 Description: specifying an (invalid) directory default stripe_size of -1
3560 would reset the directory default striping
3561 Details : stripe_size -1 was used internally to signal directory stripe
3562 removal, now use "all default" to signal dir stripe removal
3563 as a directory striping of "all default" is not useful
3566 Frequency : common for large clusters running liblustre clients
3568 Description: doing an ls when liblustre clients are running is slow
3569 Details : sending a glimpse AST to a liblustre client waits for every AST
3570 to time out, as liblustre clients will not respond. Since they
3571 cannot cache data we refresh the OST lock LVB from disk instead.
3573 Severity : enhancement
3575 Description: doing an ls at the same time as file IO can be slow
3576 Details : enqueue and other "small" requests can be blocked behind many
3577 large IO requests. Create a new OST IO portal for non-IO
3578 requests so they can be processed faster.
3581 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3583 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3584 created file with mode 000
3585 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3586 allow an ftruncate() as a non-root user to a file with mode 000.
3587 Lustre now respects this flag to disable mode checks when
3588 truncating a file owned by the user
3591 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3594 Description: Revoking locks from clients that went dead or catatonic might take
3596 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3597 cancellation of such locks instant on servers without waiting for
3598 any reply from clients. Clients drops these locks when cancel
3599 notification from server is received without replying.
3602 Frequency : liblustre-only, when liblustre client dies or becomes busy
3604 Description: Doing ls on Linux clients can take a long time with active
3606 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3607 granting such locks to it in the first place if possible. Locks
3608 are taken by proxy on the OST during the read or write and
3609 dropped immediately afterward. Add connect flags handling, do
3610 not grant locks to liblustre clients for glimpse ASTs.
3612 Severity : enhancement
3614 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3615 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3616 sequential reads and behaves badly with applications doing random
3617 reads. Improve it by reading ahead at least read region, and
3618 avoiding excessive large RPC for small reads.
3620 Severity : enhancement
3622 Description: Creating more than 1000 files for a single job may cause a load
3623 imbalance on the OSTs if there are also a large number of OSTs.
3624 Details : qos_prep_create() uses an OST index reseed value that is an
3625 even multiple of the number of available OSTs so that if the
3626 reseed happens in the middle of the object allocation it will
3627 still utilize the OSTs as uniformly as possible.
3632 Description: OST or MDS may oops in ping_evictor_main()
3633 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3634 but doesn't restart at beginning of obd_exports_timed list
3637 Severity : enhancement
3639 Description: improve by-nid export eviction on the MDS and OST
3640 Details : allow multiple exports with the same NID to be evicted at one
3641 time without re-searching the exports list.
3644 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3646 Description: MDS may oops in groups_free()
3647 Details : in rare race conditions a newly allocated group_info struct is
3648 freed again, and this can be NULL. The 2.4 compatibility code
3649 for groups_free() checked for a NULL pointer, but 2.6 did not.
3652 Frequency : common for liblustre clients doing little filesystem IO
3653 Bugzilla : 9352, 7313
3654 Description: server may evict liblustre clients accessing contended locks
3655 Details : if a client is granted a lock or receives a completion AST
3656 with a blocking AST already set it would not reply to the AST
3657 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3660 Frequency : lfs setstripe, only systems with more than 160 OSTs
3662 Description: unable to set striping with a starting offset beyond OST 160
3663 Details : llapi_create_file() incorrectly limited the starting stripe
3664 index to the maximum single-file stripe count.
3667 Frequency : LDAP users only
3669 Description: lconf did not handle in-kernel recovery with LDAP properly
3670 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3672 Severity : enhancement
3674 Description: bind OST threads to NUMA nodes to improve performance
3675 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3676 node and do their allocations there to localize memory access
3678 Severity : enhancement
3680 Description: llmount can determine client NID directly from Myrinet (GM)
3681 Details : the client NID code from gmnalnid was moved directly into
3682 llmount, removing the need to use this or specifying the
3683 client NID explicitly when mounting GM clients with zeroconf
3686 Frequency : if client is started with down MDS
3688 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3689 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3690 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3691 ptlrpc_import_delayed_req() because we didn't check if the
3692 request was interrupted, and we also didn't break out of the
3693 event loop if there was no timeout
3698 Description: data loss during non-page-aligned writes to a single file from
3699 both multiple nodes and multiple threads on one node at same time
3700 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3701 inconsistency led to false short-reads, that were cached and later
3702 used by ->prepare_write() to fill in partially written page,
3703 leading to data loss.
3706 Frequency : always, if lconf --abort_recovery used
3708 Description: lconf --abort_recovery fails with 'Operation not supported'
3709 Details : lconf was attempting to abort recovery on the MDT device and not
3712 Severity : enhancement
3714 Description: remove cleanup logs
3715 Details : replace lconf-generated cleanup logs with lustre internal
3716 cleanup routines. Eliminates the need for client-cleanup and
3719 Severity : enhancement
3721 Description: add support for EAs (user and system) on lustre filesystems
3722 Details : it is now possible to store extended attributes in the Lustre
3723 client filesystem, and with the user_xattr mount option it
3724 is possible to allow users to store EAs on their files also
3726 Severity : enhancement
3728 Description: Add possibility (config option) to show minimal available OST free
3730 Details : When compiled with --enable-mindf configure option, statfs(2)
3731 (and so, df) will return least minimal free space available from
3732 all OSTs as amount of free space on FS, instead of summary of
3733 free spaces of all OSTs.
3735 Severity : enhancement
3737 Description: do not expand extent locks acquired on OST-side
3738 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3739 by server: they are not cached anyway.
3742 Frequency : when mmap is used/binaries executed from Lustre
3744 Description: Unmmap pages before throwing them away from read cache.
3745 Details : llap_shrink cache now attempts to unmap pages before discarding
3746 them (if unmapping failed - do not discard). SLES9 kernel has
3747 extra checks that trigger if this unmapping is not done first.
3752 Description: lconf didn't resolve symlinks before checking to see whether a
3753 given mountpoint was already in use
3756 Frequency : when migrating failover services
3757 Bugzilla : 6395, 9514
3758 Description: When migrating a subset of services from a node (e.g. failback
3759 from a failover service node) the remaining services would
3760 time out and evict clients.
3761 Details : lconf --force (implied by --failover) sets the global obd_timeout
3762 to 5 seconds in order to quickly disconnect, but this caused
3763 other RPCs to time out too quickly. Do not change the global
3764 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3766 Severity : enhancement
3767 Frequency : if MDS is started with down OST
3768 Bugzilla : 9439,5706
3769 Description: Allow startup/shutdown of an MDS without depending on the
3770 availability of the OSTs.
3771 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3772 Add appropriate locking and lov-osc refcounts for safe
3773 cleaning. Add osc abort_inflight calls in case the
3774 synchronize never started.
3777 Frequency : occasional (Cray XT3 only)
3779 Description: root not authorized to access files in CRAY_PORTALS environment
3780 Details : The client process capabilities were not honoured on the MDS in
3781 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3782 been accessed by an authorized user then root was able to access
3783 the file on the local client also. The root user capabilities
3784 are now allowed on the MDS, as this environment has secure UID.
3787 Frequency : occasional
3789 Description: ldiskfs "too long searching" message happens too often
3790 Details : A debugging message (otherwise harmless) prints too often on
3791 the OST console. This has been reduced to only happen when
3792 there are fragmentation problems on the filesystem.
3797 Description: Division by zero in statfs when all OSCs are inactive
3798 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3799 lov_statfs divides by value returned by lov_get_stripecnt().
3803 Bugzilla : 9489, 3273
3804 Description: First write from each client to each OST was only 4kB in size,
3805 to initialize client writeback cache, which caused sub-optimal
3806 RPCs and poor layout on disk for the first writen file.
3807 Details : Clients now request an initial cache grant at (re)connect time
3808 and so that they can start streaming writes to the cache right
3809 away and always do full-sized RPCs if there is enough data.
3810 If the OST is rebooted the client also re-establishes its grant
3811 so that client cached writes will be honoured under the grant.
3816 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3817 Details : Now I/O RPCs go to different portal number and (presumably) fast
3818 lock requests (and glimses) and other RPCs get their own service
3819 threads pool that should be able to service those RPCs
3822 Severity : enhancement
3824 Description: Ability to exchange lustre version between client and servers and
3825 issue warnings at client side if client is too old. Also for
3826 liblustre clients there is ability to refuse connection of too old
3828 Details : New 'version' field is added to connect data structure that is
3829 filled with version info. That info is later checked by server and
3833 Frequency : rare, liblustre only.
3834 Bugzilla : 9296, 9581
3835 Description: Two simultaneous writes from liblustre at offset within same page
3836 might proceed at the same time overwriting eachother with stale
3838 Details : I/O lock withing llu_file_prwv was released too early, before data
3839 actually was hitting the wire. Extended lock-holding time until
3840 server acknowledges receiving data.
3843 Frequency : extremely rare. Never observed in practice.
3845 Description: avoid generating lustre_handle cookie of 0.
3846 Details : class_handle_hash() generates handle cookies by incrementing
3847 global counter, and can hit 0 occasionaly (this is unlikely, but
3848 not impossible, because initial value of cookie counter is
3849 selected randonly). Value of 0 is used as a sentinel meaning
3850 "unassigned handle" --- avoid it. Also coalesce two critical
3851 sections in this function into one.
3853 Severity : enhancement
3855 Description: allow liblustre clients to delegate truncate locking to OST
3856 Details : To avoid overhead of locking, liblustre client instructs OST to
3857 take extent lock in ost_punch() on client's behalf. New connection
3858 flag is added to handle backward compatibility.
3860 Severity : enhancement
3861 Bugzilla : 4928, 7341, 9758
3862 Description: allow number of OST service threads to be specified
3863 Details : a module parameter allows the number of OST service threads
3864 to be specified via "options ost ost_num_threads={N}" in the
3865 OSS's /etc/modules.conf or /etc/modprobe.conf.
3869 Bugzilla : 6146, 9635, 9895
3870 Description: servers crash with bad pointer in target_handle_connect()
3871 Details : In rare cases when a client is reconnecting it was possible that
3872 the connection request was the last reference for that export.
3873 We would temporarily drop the export reference and get a new
3874 one, but this may have been the last reference and the export
3875 was just destroyed. Get new reference before dropping old one.
3877 Severity : enhancement
3878 Frequency : if client is started with failover MDS
3880 Description: Allow multiple MDS hostnames in the mount command
3881 Details : Try to read the configuration from all specified MDS
3882 hostnames during a client mount in case the "primary"
3885 Severity : enhancement
3887 Description: Stop sending data to evicted clients as soon as possible.
3888 Details : Check if the client we are about to send or are sending data to
3889 was evicted already. (Check is done every second of waiting,
3890 for which l_wait_event interface was extended to allow checking
3891 of exit condition at specified intervals).
3894 Frequency : rare, normally only when NFS exporting is done from client
3896 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3898 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3901 Frequency : always, if srand() is called before liblustre initialization
3903 Description: Liblustre uses system PRNG disturbing its usage by user application
3904 Details : Introduce internal to lustre fast and high-quality PRNG for
3905 lustre usage and make liblustre and some other places in generic
3906 lustre code to use it.
3908 Severity : enhancement
3909 Bugzilla : 9477, 9557, 9870
3910 Description: Verify that the MDS configuration logs are updated when xml is
3911 Details : Check if the .xml configuration logs are newer than the config
3912 logs stored on the MDS and report an error if this is the case.
3913 Request --write-conf, or allow starting with --old_conf.
3915 Severity : enhancement
3917 Description: Handle symlinks in the path when checking if Lustre is mounted.
3918 Details : Resolve intermediate symlinks when checking if a client has
3919 mounted a filesystem to avoid duplicate client mounts.
3924 Description: lconf can hit an error exception but still return success.
3925 Details : The lconf command catches the Command error exception at the top
3926 level script context and will exit with the associated exit
3927 status, but doesn't ensure that this exit status is non-zero.
3932 Description: failure of ptlrpc thread startup can cause oops
3933 Details : Starting a ptlrpc service thread can fail if there are a large
3934 number of threads or the server memory is very fragmented.
3935 Handle this without oopsing.
3938 Frequency : always, only if liblustre and non-default acceptor port was used
3940 Description: liblustre cannot connect to servers with non-default acceptor port
3941 Details : tcpnal_set_default_params() was not called and was therefore
3942 ignoring the environment varaible TCPNAL_PORT, as well as other
3943 TCPNAL_ environment variables
3948 Description: two objects could be created on the same OST for a single file
3949 Details : If an OST is down, in some cases it was possible to create two
3950 objects on a single OST for a single file. No problems other
3951 than potential performance impact and spurious error messages.
3955 Bugzilla : 5681, 9562
3956 Description: Client may oops in ll_unhash_aliases
3957 Details : Client dcache may become inconsistent in race condition.
3958 In some cases "getcwd" can fail if the current directory is
3964 Description: Inode refcounting problems in NFS export code
3965 Details : link_raw functions used to call d_instantiate without obtaining
3966 extra inode reference first.
3970 Bugzilla : 9942, 9903
3971 Description: Referencing freed requests leading to crash, memleaks with NFS.
3972 Details : We used to require that call to ll_revalidate_it was always
3973 followed by ll_lookup_it. Also with revalidate_special() it is
3974 possible to call ll_revalidate_it() twice for the same dentry
3975 even if first occurence returned success. This fix changes semantic
3976 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3977 reference on a request referred from the intent.
3978 ll_intent_release() then releases such a request.
3981 Frequency : rare, normally benchmark loads only
3983 Description: unlinked inodes were kept in memory on the client
3984 Details : If a client is repeatedly creating and unlinking files it
3985 can accumulate a lot of stale inodes in the inode slab cache.
3986 If there is no other client load running this can cause the
3987 client node to run out of memory. Instead flush old inodes
3988 from client cache that have the same inode number as a new inode.
3991 Frequency : SLES9 2.6.5 kernel and long filenames only
3992 Bugzilla : 9969, 10379
3993 Description: utime reports stale NFS file handle
3994 Details : SLES9 uses out-of-dentry names in some cases, which confused
3995 the lustre dentry revalidation. Change it to always use the
3999 Frequency : rare, unless heavy write-truncate concurrency is continuous
4000 Bugzilla : 4180, 6984, 7171, 9963, 9331
4001 Description: OST becomes very slow and/or deadlocked during object unlink
4002 Details : filter_destroy() was holding onto the parent directory lock
4003 while truncating+unlinking objects. For very large objects this
4004 may block other threads for a long time and slow overall OST
4005 responsiveness. It may also be possible to get a lock ordering
4006 deadlock in this case, or run out of journal credits because of
4007 the combined truncate+unlink. Solution is to do object truncate
4008 first in one transaction without parent lock, and then do the
4009 final unlink in a new transaction with the parent lock. This
4010 reduces the lock hold time dramatically.
4013 Frequency : rare, 2.4 kernels only
4015 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4016 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4017 before unmounting their filesystem in order to avoid deadlock.
4018 The kernel_locked() function in 2.4 kernels only checks whether
4019 the kernel lock is held, not whether it is this process that is
4020 holding it as 2.6 kernels do.
4025 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4026 Details : The client ptlrpc code may be trying to reconnect to a down
4027 server before a previous connection attempt has timed out.
4028 Increase the reconnect interval to be longer than the connection
4029 timeout interval to avoid sending duplicate connections to
4033 Frequency : echo_client brw_test command
4035 Description: fix echo_client to work with OST preallocated code
4036 Details : OST preallocation code (5137) didn't take echo_client IO path
4037 into account: echo_client calls filter methods outside of any
4038 OST thread and, hence, there is no per-thread preallocated
4039 pages and buffers to use. Solution: hijack pga pages for IO. As
4040 a byproduct, this avoids unnecessary data copying.
4044 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4045 Description: Client can oops in mdc_commit_close() after open replay
4046 Details : It was possible for the MDS to return an open request with no
4047 transaction number in mds_finish_transno() if the client was
4048 evicted, but without actually returning an error. Clients
4049 would later try to replay that open and may trip an assertion
4050 Simplify the client close codepath, and always return an error
4051 from the MDS in case the open is not successful.
4054 Frequency : rare, 2.6 OSTs only
4056 Description: OST may deadlock under high load on fragmented files
4057 Details : If there was a heavy load and highly-fragmented OST filesystems
4058 it was possible to have all the OST threads deadlock waiting on
4059 allocation of biovecs, because the biovecs were not released
4060 until the entire RPC IO was completed. Instead, release biovecs
4061 as soon as they are complete to ensure forward IO progress.
4063 Severity : enhancement
4065 Description: Support for specifying external journal device at mount
4066 Details : If an OST or MDS device is formatted with an external journal
4067 device, this device major/minor is stored in the ext3 superblock
4068 and may not be valid for failover. Allow detecting and
4069 specifying the external journal at mount time.
4074 Description: Mounting an MDS with pending unlinked files may cause oops
4075 Details : target_finish_recovery() calls mds_postrecov() which returned
4076 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4077 considers this an error and immediately begins cleaning up the
4078 lov, just after starting the mds_lov process
4080 Severity : enhancement
4082 Description: Implement 'lfs df' to report actual free space on per-OST basis
4083 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4084 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4085 usage of objects; '-h' to report in human readable format.
4087 ------------------------------------------------------------------------------
4089 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4096 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4097 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4098 at startup during orphan unlinking. Instead, skip buddy update
4099 before setup, as it will be regenerated after recovery is complete.
4104 Description: If an OST is inactive, its locks might reference stale inodes.
4105 Details : lov_change_cbdata() must iterate over all namespaces, even if
4106 they are inactive to clear inode references from the lock.
4108 Severity : enhancement
4109 Frequency : occasional, if non-standard max_dirty_mb used
4111 Description: Client will block write RPCs if not enough grant
4112 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4113 then the client will block writes while waiting for another RPC
4114 to complete instead of consuming its dirty limit. With change
4115 we get improved performance when max_dirty_mb is small.
4117 Severity : enhancement
4118 Bugzilla : 3389, 6253
4119 Description: Add support for supplementary groups on the MDS.
4120 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4121 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4122 lookups for user supplementary groups into a cache.
4126 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4127 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4128 for liblustre O_EXCL is enforced by MDS.
4132 Description: readdir never returns NULL in liblustre.
4133 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4134 dentry in struct dirent.
4138 Frequency : liblustre only, depends on application IO pattern
4139 Description: liblustre clients evicted if not contacting servers
4140 Details : Don't put liblustre clients into the ping_evictor list, so
4141 they will not be evicted by the pinger ever.
4143 Severity : enhancement
4145 Description: Add ability to evict clients by NID from MDS.
4146 Details : By echoing "nid:$NID" string into
4147 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4148 $NID would be instantly evicted from this MDS and from all active
4149 OSTs connected to it.
4153 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4154 Details : lookup_it_finish() used to query file size from OSTs that was not
4159 Description: service threads change working directory to that of init
4160 Details : Starting lustre service threads may pin the working directory
4161 of the parent thread, making that filesystem busy. Threads
4162 now change to the working directory of init to avoid this.
4166 Frequency : during shutdown only
4167 Description: shutdown with a failed MDS or OST can cause unmount to hang
4168 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4173 Frequency : 2.6 only
4174 Description: chmod/chown may include an extra supplementary group
4175 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4176 supplementary group and if none is specified this is used.
4179 Bugzilla : 5479 (6816)
4181 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4182 Details : If lookup is in progress on a file that is unlinked we might try
4183 to revalidate the inode and fail in revalidate after lookup is
4184 complete and ll_file_open() enqueues the open again but
4185 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4188 Frequency : always, if lconf --abort_recovery used
4190 Description: lconf --abort_recovery fails with 'Operation not supported'
4191 Details : lconf was attempting to abort recovery on the MDT device and not
4194 ------------------------------------------------------------------------------
4196 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4201 Frequency : rare (only unsupported configurations with a node running as an
4203 Bugzilla : 6514, 5137
4204 Description: Mounting a Lustre file system on a node running as an OST could
4206 Details : OSTs now preallocates memory needed to write out data at
4207 startup, instead of when needed, to avoid having to
4208 allocate memory in possibly low memory situations.
4209 Specifically, if the file system is mounted on on OST,
4210 memory pressure could force it to try to write out data,
4211 which it needed to allocate memory to do. Due to the low
4212 memory, it would be unable to do so and the node would
4213 become unresponsive.
4215 Severity : enhancement
4217 Description: Addition of lconf --service command line option
4218 Details : lconf now accepts a '--service <arg>' option, which is
4219 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4221 Severity : enhancement
4223 Description: Failover mode is now the default for OSTs.
4224 Details : By default, OSTs will now run in failover mode. To return to
4225 the old behaviour, add '--failout' to the lmc line for OSTs.
4227 Severity : enhancement
4229 Description: Health checks are now provided for MDS and OSTs
4230 Details : Additional detailed health check information on MSD and OSTs
4231 is now provided through the procfs health_check value.
4234 Frequency : occasional, depends on IO load
4236 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4237 after numerous create/delete cycles
4238 Details : The ext3 inode allocation policy would not allocate new inodes
4239 very well on the OSTs because there are no new directories
4240 being created. Instead we look for groups with free space if
4241 the parent directories are nearly full.
4246 Description: Network or server problems during mount may cause partially
4247 mounted clients instead of returning an error.
4248 Details : The config llog parsing code may overwrite the error return
4249 code during mount error handling, returning success instead
4255 Description: MDS can fail to allocate large reply buffers
4256 Details : After long uptimes the MDS can fail to allocate large reply
4257 buffers (e.g. zconf client mount config records) due to memory
4258 fragmentation or consumption by the buffer cache. Preallocate
4259 some large reply buffers so that these replies can be sent even
4260 under memory pressure.
4264 Frequency : rare (liblustre)
4265 Description: fsx running with liblustre complained that using truncate() to
4266 extend the file doesn't work. This patch corrects that issue.
4267 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4268 ATTR_SIZE and lsm use in llu_setattr_raw.
4272 Frequency : rare, only 2.6 kernels
4273 Description: Unusual file access patterns on the MDS may result in inode
4274 data being lost in very rare circumstances.
4275 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4276 code in the 2.6 kernel caused the inode and/or EA data not to
4277 be read from disk, causing single-file corruption.
4281 Frequency : rare, only 2.6 filesystems using extents
4282 Description: Heavy concurrent write and delete load may cause data corruption.
4283 Details : It was possible under high-load situations to have an extent
4284 metadata block in the block device cache from a just-unlinked
4285 file overwrite a newly-allocated data block. We now unmap any
4286 metadata buffers that alias just-allocated data blocks.
4290 Frequency : filesystems with default stripe_count larger than 77
4291 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4292 Details : lconf specifies an inode size of 4096 bytes when the default
4293 stripe_count is larger than 77. This conflicts with the default
4294 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4295 this case to avoid pinning too much memory for large EAs.
4297 ------------------------------------------------------------------------------
4299 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4304 Frequency : rare (extremely heavy IO load with hundreds of clients)
4306 Description: Client is evicted, gets IO error writing to file
4307 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4308 caused clients to be evicted for AST timeouts. The fixes in
4309 bug 5192 mean we no longer need to have such short AST timeouts
4310 so ldlm_timeout has been increased.
4313 Frequency : occasional during --force or --failover shutdown under load
4314 Bugzilla : 5949, 4834
4315 Description: Server oops/LBUG if stopped with --force or --failover under load
4316 Details : a collection of import/export refcount and cleanup ordering
4317 issues fixed for safer force cleanup
4320 Frequency : only filesystems larger than 120 OSTs
4321 Bugzilla : 5990, 6223
4322 Description: lfs getstripe would oops on a very large filesystem
4323 Details : lov_getconfig used kfree on vmalloc'd memory
4326 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4327 Bugzilla : 6242, 6243
4328 Description: reading from files that had been truncated to a non-zero size
4329 but never opened returned no data
4330 Details : ll_file_read() reads zeros from no-object files to EOF
4335 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4336 Details : There's little harm in aborting MDS/OSS recovery and letting it
4337 try again, so I removed the LASSERT and return an error instead.
4339 Severity : enhancement
4341 Description: New debugging infrastructure for tracking down data corruption
4342 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4343 (b) cover more of the client-side code path, and (c) try to narrow
4344 down where problems occurred
4348 Bugzilla : 3819, 4364, 4397, 6313
4349 Description: Racing close and eviction MDS could cause assertion in mds_close
4350 Details : It was possible to get multiple mfd references during close and
4351 client eviction, leading to one thread referencing a freed mfd.
4353 Severity: : enhancement
4354 Bugzilla : 3262, 6359
4355 Description: Attempts to reconnect to servers are now more aggressive.
4356 Details : This builds on the enhanced upcall-less recovery that was added
4357 in 1.4.2. When trying to reconnect to servers, clients will
4358 now try each server in the failover group every 10 seconds. By
4359 default, clients would previously try one server every 25 seconds.
4364 Description: After recovery, certain operations trigger a failed
4365 assertion on a client.
4366 Details : Failing over an mds, using lconf -d --failover, while a
4367 client was doing a readdir() call would cause the client to
4368 LBUG after recovery completed and the readdir() was resent.
4370 Severity : enhancement
4372 Description: Default groups are now added by lconf
4373 Details : You can now run lconf --group <servicename> without having to
4374 manually add groups with lmc.
4377 Frequency : occasional
4379 Description: Nodes with an elan id of 0 trigger a failed assertion
4382 Frequency : always when accessing e.g. tty/console device nodes
4384 Description: tty and some other devices nodes cannot be used on lustre
4385 Details : file's private_data field is used by device data and lustre
4386 values in there got lost. New field was added to struct file to
4387 store fs-specific private data.
4390 Frequency : when exporting Lustre via NFS
4392 Description: NFSD failed occasionally when looking up a path component
4393 Details : NFSD is looking up ".." which was broken in ext3 directories
4394 that had grown large enough to become hashed.
4397 Frequency : Clusters with multiple interfaces not on the same subnet
4399 Description: Nodes will repeatedly try to reconnect to an interface which it
4400 cannot reach and report an error to the log.
4401 Details : Extra peer list entries will be created by lconf with some peers
4402 unreachable. lconf now validates the peer before adding it.
4405 Frequency : Only if a default stripe is set on the filesystem root.
4407 Description: Setting a default stripe on the filesystem root prevented the
4408 filesystem from being remounted.
4409 Details : The client was sending extra request flags in the root getattr
4410 request and did not allocate a reply buffer for the dir EA.
4413 Frequency : occasional, higher if lots of files are accessed by one client
4414 Bugzilla : 6159, 6097
4415 Description: Client trips assertion regarding lsm mismatch/magic
4416 Details : While revalidating inodes the VFS looks up inodes with ifind()
4417 and in rare cases can find an inode that is being freed.
4418 The ll_test_inode() code will free the lsm during ifind()
4419 when it finds an existing inode and then the VFS later attaches
4420 this free lsm to a new inode.
4424 Bugzilla : 6422, 7030
4425 Description: MDS deadlock between mkdir and client eviction
4426 Details : Creating a new file via mkdir or mknod (starting a transaction
4427 and getting the ns lock) can deadlock with client eviction
4428 (gets ns lock and trying to finish a synchronous transaction).
4431 Frequency : occasional
4432 Description: While starting a server, the fsfilt_ext3 module could not be
4434 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4435 kernels. Previously, lconf would still use the ext3 name
4436 when trying to load modules. Now, it will correctly use
4437 ext3 on 2.4 and ldiskfs on 2.6.
4439 Severity : enhancement
4440 Description: The default stripe count has been changed to 1
4441 Details : The interpretation of the default stripe count (0, to lfs
4442 or lmc) has been changed to mean striping across a single
4443 OST, rather than all available. For general usage we have
4444 found a stripe count of 1 or 2 works best.
4446 Severity : enhancement
4447 Description: Add support for compiling against Cray portals.
4448 Details : Conditional compiling for some areas that are different
4452 Frequency : occasional
4453 Bugzilla : 6409, 6834
4454 Description: Creating files with an explicit stripe count may lead to
4455 a failed assertion on the MDS
4456 Details : If some OSTs are full or unavailable, creating files may
4457 trigger a failed assertion on the MDS. Now, Lustre will
4458 try to use other servers or return an error to the
4462 Frequency : occasional
4464 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4465 caused lock timeout during readahead (since 1.4.2).
4466 Details : Processes doing readahead might match a lock that hasn't been
4467 granted yet if there are overlapping and conflicting lock
4468 requests. The readahead process waits on ungranted lock
4469 (original lock is CBPENDING), while OST waits for that process
4470 to cancel CBPENDING read lock and eventually evicts client.
4472 Severity : enhancement
4474 Description: Initial enabling of flock support for clients
4475 Details : Implements fcntl advisory locking and file status functions.
4476 This feature is provided as an optional mount flag (default
4477 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4478 locking are implemented yet, and those that are are not guaranteed
4479 to be completely correct in production environments.
4480 mount -t lustre -o [flock|noflock] ...
4483 Frequency : occasional
4485 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4486 assertion in the ext3 JBD (journaling) layer.
4487 Details : The b_committed_data struct is protected by the big kernel lock
4488 in 2.4 kernels, serializing journal_commit_transaction() and
4489 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4490 access to this struct. In 2.6 kernels there is finer grained
4491 locking to improve SMP performance of the JBD layer.
4495 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4500 Description: merge and cleanup kernel patches.
4503 -----------------------------------------------------------------------------
4505 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4507 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4508 versions of Lustre. Please update all servers and clients to
4509 version 1.4.2 or later at the same time. You must also run
4510 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4511 to update the configuration logs.
4513 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4514 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4515 - swab llog records of type '0' so we get proper header size/idx (5861)
4516 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4517 - fix rename of one directory over another leaking an inode (5953)
4518 - avoid SetPageDirty on 2.6 (5981)
4519 - don't re-add just-being-destroyed locks to the waiting list (5653)
4520 - when creating new directories, inherit the parent's custom
4521 striping settings if present parent (3048)
4522 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4523 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4524 - don't hold client locks on temporary worklist from l_lru (5666)
4525 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4526 - automatically evict dead clients (5921)
4527 - Update file size properly in create+truncate+fstat case (6196)
4528 - Do not unhash mountpoint dentries, do not allow removal of
4530 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4531 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4532 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4533 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4534 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4535 - Correctly handle reads of files with no objects (6243)
4536 - lctl recover will also mark a device active if deactivate used (5933)
4538 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4539 - allow --write-conf on an MDS with different nettype than client (5619)
4540 - don't write config llogs to MDS for mounts not from that MDS (5617)
4541 - lconf should create multiple TCP connections from a client (5201)
4542 - init scripts are now turned off by default; run chkconfig --on
4543 lustre and chkconfig --on lustrefs to use them
4544 - upcalls are no longer needed for clients to recover to failover
4546 - add --abort-recovery option to lconf to abort recovery on device
4548 - add support for an arbitrary number of OSTs (3026)
4549 - Quota support protocol changes.
4550 - forward compatibility changes to wire structs (6007)
4551 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4552 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4553 - improved "lustre status" script
4554 - initialize blocksize for non-regular files (6062)
4555 - added --disable-server and --disable-client configure options (5782)
4556 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4557 - Vanilla 2.4.29 support
4558 - increase maximum number of obd devices to 520 (6242)
4559 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4560 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4561 SLES 9 2.6.5 kernels
4562 - limit stripes per file to 160 (the maximum EA size) (6093)
4564 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4567 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4568 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4569 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4570 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4571 - small ext3 extents cleanups and fixes (5733)
4572 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4573 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4574 - avoid needless client->OST connect, fix handle mismatch (5317)
4575 - fix DLM error path that led to out-of-sync client, long delays (5779)
4576 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4577 - fix several locking issues related to i_size (5492,5624,5654,5672)
4578 - don't move pending lock onto export if it is already evicted (5683)
4579 - fix kernel oops when creating .foo in unlinked directory (5548)
4580 - fix deadlock in obdfilter statistics vs. object create (5811)
4581 - use time_{before,after} to avoid timer jiffies wrap (5882)
4582 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4583 - Do not leak request if server was not able to process it (5154)
4584 - If mds_open unable to find parent dir, make that negative lookup(5154)
4585 - don't create new directories with extent-mapping (5909, 5936)
4587 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4588 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4589 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4590 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4591 - support for RHEL4 kernel on i686 (5773)
4592 - provide error messages when incompatible logs are encountered (5898)
4594 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4595 * version 1.4.0.10 (1.4.1 release candidate 1)
4597 - don't keep a lock reference when lock is not granted (4238)
4598 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4599 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4600 - drop import inflight refcount on signal_completed_replay error (5255)
4601 - unlock page after async write error during send (3677)
4602 - handle missing objects in filter_preprw_read properly (5265)
4603 - no transno return for symlink open, don't save no-trasno open (3440)
4604 - don't try to complete elan receive that already failed (4012)
4605 - free RPC server reply state on error (5406)
4606 - clean up thread from ptlrpc_start_thread() on error (5160)
4607 - readahead could read extra page into cache that wasn't ejected (5388)
4608 - prevent races in class_attach/setup/cleanup/detach (5260)
4609 - don't dereference de->d_inode after l_dput of de (5458)
4610 - use "int" for stripe value returned from lock_to_stripe (5544)
4611 - mballoc allocation and error-checking fixes in 2.6 (5504)
4612 - block device patches to fix I/O request sizes in 2.6 (5482)
4613 - look up hostnames for IB nals (5602)
4614 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4615 - don't start multiple acceptors for the same port (5277)
4616 - fix incorrect LASSERT in mds_getattr_name (5635)
4617 - export a proc file for general "ping" checking (5628)
4618 - fix "lfs check" to not block when the MDS is down (5628)
4620 - service request history (4965)
4621 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4622 - create an "evict_client" /proc entry on OSTs, like the MDS has
4623 - fix mount usage message, return errors per mount(8) (5168)
4624 - change grep [] to grep "[]" in tests so they work in more UMLs
4625 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4626 - remove ancient LOV_MAGIC_V0 EA support (5047)
4627 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4628 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4629 - allow readahead allocations to fail when low on memory (5383)
4630 - mmap locking landed again, after considerable improvement (2828)
4631 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4633 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4636 - send OST transaction number in read/write reply to free req (4966)
4637 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4638 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4639 - the watchdog thread now runs as interruptible (5246)
4640 - flock/lockf fixes (but it's still disabled, pending 5135)
4641 - don't use EXT3 constants in llite code (5094)
4642 - memory shortage at startup could cause assertion (5176)
4644 - reorganization of lov code
4645 - single portals codebase
4647 - add extents/mballoc support (5025)
4648 - direct I/O reads in the obdfilter (4048)
4649 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4651 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4654 - send OST transaction number in read/write reply to free req (4966)
4655 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4656 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4657 - the watchdog thread now runs as interruptible (5246)
4658 - handle missing objects in filter_preprw_read properly (5265)
4659 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4660 - drop import inflight refcount on signal_completed_replay error (5255)
4661 - unlock page after async write error during send (3677)
4662 - return original error code on reconstructed replies (3761)
4663 - no transno return for symlink open, don't save no-trasno open (3440)
4665 - add pid to ldlm debugging output (4922)
4666 - bump the watchdog timeouts -- we can't handle 30sec yet
4667 - extra debugging for orphan dentry/inode bug (5259)
4669 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4672 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4673 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4674 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4675 - fix NULL dereference in /proc/sys/portals/routes (4827)
4676 - allow failed mdc_close() operations to be interrupted (4561)
4677 - stop precreate on OST before MDS would time out on it (4778)
4678 - don't send partial-page writes before EOF from client (4410)
4679 - discard client grant for sub-page writes on large-page clients (4520)
4680 - don't free dentries not owned by NFS code, check generation (4806)
4681 - fix lsm leak if mds_create_objects() fails (4801)
4682 - limit debug_daemon file size, always print CERROR messages (4789)
4683 - use transno after validating reply (3892)
4684 - process timed out requests if import state changes (3754)
4685 - update mtime on OST during writes, return in glimpse (4829)
4686 - add mkfsoptions to LDAP (4679)
4687 - use ->max_readahead method instead of zapping global ra (5039)
4688 - don't interrupt __l_wait_event() during strace
4690 - add software watchdogs to catch hung threads quickly (4941)
4691 - make lustrefs init script start after nfs is mounted
4692 - fix CWARN/ERROR duplication (4930)
4693 - return async write errors to application if possible (2248)
4694 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4695 - print NAL number in %x format (4645)
4696 - update barely-supported suse-2.4.21-171 series (4842)
4697 - support for sles 9 %post scripts
4698 - support for building 2.6 kernel-source packages
4699 - support for sles km_* packages
4701 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4704 - ignore -ENOENT errors in osc_destroy (3639)
4705 - notify osc create thread that OSC is being cleaned up (4600)
4706 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4707 - reconstruct ost_handle() like mds_handle() (4657)
4708 - create a new thread to do import eviction to avoid deadlock (3969)
4709 - let lconf resolve symlinked-to devices (4629)
4710 - don't unlink "objects" from directory with default EA (4554)
4711 - hold socknal file ref over connect in case target is down (4394)
4712 - allow more than 32000 subdirectories in a single directory (3244)
4713 - fix blocks count for O_DIRECT writes (3751)
4714 - OST returns ENOSPC from object create when no space left (4539)
4715 - don't send truncate RPC if file size isn't changing (4410)
4716 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4717 - bind to privileged port in socknal and tcpnal (3689)
4719 - rate limit CERROR/CWARN console message to avoid overload (4519)
4720 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4721 - basic mmap support (3918)
4722 - kernel patch series update from b1_4 (4711)
4724 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4727 - avoid crash during MDS cleanup with OST shut down (2775)
4728 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4729 - don't use bad inodes on the MDS (3744)
4730 - dynamic object preallocation to improve recovery speed (4236)
4731 - don't hold spinlock over lock dumping or change debug flags (4401)
4732 - don't zero obd_dev when it is force cleaned (3651)
4733 - print grants to console if they go negative (4431)
4734 - "lctl deactivate" will stop automatic recovery attempts (3406)
4735 - look for existing locks in ldlm_handle_enqueue() (3764)
4736 - don't resolve lock handle twice in recovery avoiding race (4401)
4737 - revalidate should check working dir is a directory (4134)
4739 - don't always mark "slow" obdfilter messages as errors (4418)
4741 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4744 - don't close LustreDB during write_conf until it is done (3860)
4745 - fix typo in lconf for_each_profile (3821)
4746 - allow dumping logs from multiple threads at one time (3820)
4747 - don't allow multiple threads in OSC recovery (3812)
4748 - fix debug_size parameters (3864)
4749 - fix mds_postrecov to initialize import for llog ctxt (3121)
4750 - replace config semaphore with spinlock (3306)
4751 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4752 - don't allow enqueue to complete on a destroyed export (3822)
4753 - down write_lock before checking llog header bitmap (3825)
4754 - recover from lock replay timeout (3764)
4755 - up llog sem before sending rpc (3652)
4756 - reduce ns lock hold times when setting kms (3267)
4757 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4758 - fix NULL deref and obd_dev leak on setup error (3312)
4759 - replace some LBUG about llog ops with error handling (3841)
4760 - don't match INVALID dentries from d_lookup and spin (3784)
4761 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4762 - fix invalid assertion in ptlrpc_set_wait (3880)
4764 - add libwrap support for the TCP acceptor (3996)
4765 - add /proc/sys/portals/routes for non-root route listing (3994)
4766 - allow setting MDS UUID in .xml (2580)
4767 - print the stack of a process that LBUGs (4228)
4769 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4772 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4773 - make sure to unset replay flag from failed open requests (3440)
4774 - if default stripe count is 0, use OST count for inode size (3636)
4775 - update parent mtime/ctime on client for create/unlink (2611)
4776 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4777 - free recovery state on server during a forced cleanup (3571)
4778 - unregister_reply for resent reqs (3063)
4779 - loop back devices mounting and status check on 2.6 (3563)
4780 - fix resource-creation race that can provoke i_size == 0 (3513)
4781 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4782 - more debugging for page-accounting assertion (3746)
4783 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4784 - avoid deadlock after precreation failure (3758)
4785 - fix race and lock order deadlock in orphan handling (3450, 3750)
4786 - add validity checks when grabbing inodes from l_ast_data (3599)
4788 - add /proc/.../recovery_status to obdfilter (3428)
4789 - lightweight CDEBUG infrastructure, debug daemon (3668)
4790 - change default OSC RPC parameters to be better on small clusters
4791 - turn off OST read cache for files smaller than 32MB
4792 - install man pages and include them in rpms (3100)
4793 - add new init script for (un)mounting lustre filesystems (2593)
4794 - run chkconfig in %post for init scripts (3701)
4795 - drop scimac NAL (unmaintained)
4797 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4800 - clean kiobufs before and after use (3485)
4801 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4802 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4803 - change init script to fail more clearly if not run as root (1528)
4804 - allow clients to reconnect during replay (1742)
4805 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4806 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4807 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4808 - fix cond_resched (3554)
4809 - only evict unfinished clients after recovery (3515)
4810 - allow bulk resend, prevent data loss (3570)
4811 - dynamic ptlrpc request buffer allocation (2102)
4812 - don't allow unlinking open directory if it isn't empty (2904)
4813 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4814 - remove extraneous obd dereference causing LASSERT failure (3334)
4815 - don't use get_cycles() when creating temp. files on the mds (3156)
4816 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4817 - handle EEXIST for set-stripe, set proper directory name (3336)
4819 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4820 - fix ksocknal_fmb_callback() error messages (2918)
4822 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4825 - don't copy lvb into (possibly NULL) reply on error (2983)
4826 - don't deref dentry after dput, don't free lvb on error (2922)
4827 - use the kms to determine writeback rpc length (2947)
4828 - increment oti_logcookies when osc is inactive (2948)
4829 - update client's i_blocks count via lvb messages (2543)
4830 - handle intent open/close of special files properly (1557)
4831 - mount MDS with errors=remount-ro, like obdfilter (2009)
4832 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4833 - don't use cancelling-locks' kms values (2947)
4834 - use highest lock extent for kms, not last one (2925)
4835 - don't dereference ERR_PTR() dentry in error handling path (3107)
4836 - fix thread race in portals_debug_dumplog() (3122)
4837 - create lprocfs device entries at setup instead of at attach (1519)
4838 - common AST error handler, don't evict client on completion race (3145)
4839 - zero nameidata in detach_mnt in 2.6 (3118)
4840 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4841 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4842 - initialize RPC timeout timer earlier for 2.6 (3219)
4843 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4844 - print nal/nid for unknown nid (3258)
4845 - additional checks for oscc recovery before doing precreate (3284)
4846 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4847 - don't crash in mdc_close for bad permissions on open (3285)
4848 - zero i_rdev for non-device files (3147)
4849 - clear page->private before handing to FS, better assertion (3119)
4850 - tune the read pipeline (3236)
4851 - fix incorrect decref of invalidated dentry (2350)
4852 - provide read-ahead stats and refine rpc in flight stats (3328)
4853 - don't hold journal transaction open across create RPC (3313)
4854 - update atime on MDS at close time (3265)
4855 - close LDAP connection when recovering to avoid server load (3315)
4856 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4857 - don't leak open file on MDS after open resend (3325)
4858 - serialize filter_precreate and filter_destroy_precreated (3329)
4859 - loop device shouldn't call sync_dev() for nul device (3092)
4860 - clear page cache after eviction (2766)
4861 - resynchronize MDS->OST in background (2824)
4862 - refuse to mount the same filesystem twice on same mountpoint (3394)
4863 - allow llmount to create routes for mounting behind routers (3320)
4864 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4865 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4866 - fix rare problem with rename on htree directories (3417)
4868 - allow default OST striping configuration per directory (1414)
4869 - fix compilation for qswnal for 2.6 kernels (3125)
4870 - increase maximum number of MDS request buffers for large systems
4871 - change liblustreapi to be useful for external progs like lfsck (3098)
4872 - increase local configuration timeout for slow disks (3353)
4873 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4875 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4878 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4879 - don't overwrite extent policy data in reply if lock was blocked (2901)
4880 - drop filter export grants atomically with removal from device (2663)
4881 - del obd_self_export from work_list in class_disconnect_exports (2908)
4882 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4883 - swab reply message in mdc_close, other PPC fixes (2464)
4884 - fix destroying of named logs (2325)
4885 - overwrite old logs when running lconf --write_conf (2264)
4886 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4887 - fix race in target_handle_connect (2898)
4888 - mds_reint_create() should take same inode create lock (2926)
4889 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4890 - don't close files for self_export to avoid uninitialized obd (2936)
4891 - allow MDS with the same name as client node (2939)
4892 - hold dentry reference for closed log files for unlink (2325)
4893 - reserve space for all logs during transactions (2059)
4894 - don't evict page beyond end of stripe extent (2925)
4895 - don't oops on a deleted current working directory (2399)
4896 - handle hard links to targets without a parent properly (2517)
4897 - don't dereference NULL lock when racing during eviction (2867)
4898 - don't grow lock extents when lots of conflicting locks (2919)
4900 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4903 - account for cache space usage on clients to avoid data loss (974)
4904 - lfsck support in lustre kernel code (2349)
4905 - reduce journal credits needed for BRW writes (2370)
4906 - orphan handling to avoid losing space on client/server crashes
4907 - ptlrpcd can be blocked, stopping ALL progress (2477)
4908 - use lock value blocks to assist in proper KMS, faster stat (1021)
4909 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4910 - recovery for initial connections (2355)
4911 - fixes for mds_cleanup_orphans (1934)
4912 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4913 - block all file creations until orphan recovery completes (1901)
4914 - client remove rq_connection from request struct (2423)
4915 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4916 - recovery timer race (2670)
4917 - mdc_close recovey bug (2532)
4918 - ptlrpc cleanup bug (2710)
4919 - mds timeout on local locks (2588)
4920 - namespace lock held during RPCs (2431)
4921 - handle interrupted sync write properly (2503)
4922 - don't try to handle a message that hasn't been replied to (2699)
4923 - client assert failure during cleanup after abort recovery (2701)
4924 - leak mdc device after failed mount (2712)
4925 - ptlrpc_check_set allows timedout requests to complete (2714)
4926 - wait for inflight reqs when ptlrpcd finishes (2710)
4927 - make sure unregistered services are removed from the srv_list
4928 - reset bulk XID's when resending them (caught by 1138 test)
4929 - unregister_bulk after timeout
4930 - fix lconf error (2694)
4931 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4932 - readahead locks pages, leaves pending causing memory pressure (2673)
4933 - increase OST request buffers to 4096 on large machines (2729)
4934 - fix up permission of existing directories in simple_mkdir (2661)
4935 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4936 - don't assign transno to errored transactions (2742)
4937 - don't delete objects on OST if given a bogus objid from MDS (2751)
4938 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4939 - if rq_no_resend, then timeout request after recovery (2432)
4940 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4941 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4942 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4943 - MDS recovery shouldn't delete objects that it has given out (2730)
4944 - if enqueue arrives after completion, don't clobber LVB (2819)
4945 - don't unlock pages twice when trigger_group_io returns error (2814)
4946 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4947 - don't write pages to disk if there was an error (1450)
4948 - don't ping imports that have recovery disabled (2676)
4949 - take buffered bytes into account when balancing socknal conn (2817)
4950 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4951 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4952 - remove little-endian swabbing of llog records (1987)
4953 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4954 - retry reposting request buffers if they fail (1191)
4955 - grow extent at grant time to avoid granting a revoked lock (2809)
4956 - lock revoke doesn't evict page if covered by a second lock (2765)
4957 - disable VM readahead to avoid reading outside lock extents (2805)
4959 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4960 - updated kernel patches for hp-2.4.20 kernel (2681)
4962 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4965 - fix truncated write corruption (2366)
4966 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4968 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4969 - ensure len(uuid) < 37 in lmc (1171)
4970 - fix ia64 OOPS in llog_test (2255)
4971 - zero end of page at obdfilter for partial page writes (2648)
4972 - don't leave stale dentries around after renames (bug 2428)
4973 - fix timeouts when evicting a client with a single lock held (2642)
4974 - set deadline for the initial HELLO message to drain (2634)
4975 - print out dotted-quad IP addresses in the socknal (2302)
4977 - additional debugging for MDS client eviction problem (2443)
4978 - fix mkfsoptions support for osts (2603, 2604)
4980 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4983 - add series for the vanilla 2.6.0 kernel
4984 - add series for the vanilla 2.4.24 kernel
4985 - add series for a cray x86/64 UL kernel drop
4986 - fix xattr patches for the vanilla 2.4.19 series
4988 - generate true UUIDs in lmc (1171)
4989 - have portals stack dumping break in UML (2466)
4990 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4991 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4992 - wake callback waiting threads on client eviction (2460)
4993 - Add --ptldebug and --subsystem to lmc (1719)
4994 - update assertion to allow safe interrupt allocation
4995 - set rq_no_resend for cancel requests (2432)
4996 - recalculate ptlrpcd timeout after resend (2494)
4997 - call vfs_rmdir when removing pending directories (2368)
4998 - fix renaming a file to itself (2429)
4999 - lmc creates a default one-stripe lov (2454)
5000 - expand procfs space to handle large clusters (2326)
5001 - increase UML stack to avoid overflow
5002 - update lconf's list of debug and subsystem masks
5003 - fix lfs find --obd (2510)
5004 - /proc tunable for disabling filter read caching (2591)
5005 - stop rpm packages from altering slapd.conf (2301)
5006 - disable nagle in the socknal under 0conf (2578)
5007 - choose mds inode size based on stripe count (2572)
5008 - fix kernel-source rpm problems (2516)
5010 - add --disable-doc to avoid pdf generation (2421)
5011 - update documentation, tests, type-os, comments
5012 - avoid format warnings on ia64
5013 - remove the TOE NAL
5014 - tiny code cleanups by removing unused fields
5016 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5019 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5020 - remove the most blatant lies from BUILDING file (2371)
5021 - change default debug level to reasonable production setting
5022 - reduce client side cache size to reduce cache flush time
5023 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5024 - make TCP zerocopy and pinger support enabled by default (2476)
5025 - sync writes completed after process exits caused crashes (2319)
5026 - maintain correct mount count on the MDS (2356)
5027 - backout 1557, because 2316 wasn't really fixed
5028 - better file I/O statistics gathering in /proc
5029 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5030 - another kernel patch to fix zero-copy TCP function export
5031 - don't take duplicate lock when processing re-sent getattr (2420)
5032 - lctl uses obd_self_export instead of creating new conn (2353)
5033 - MDS/OST recovery case which requires object creation asserted (2425)
5034 - move lfs from /usr/sbin to /usr/bin in packages
5035 - fix race between mds_client_add and mds_client_free (2417)
5036 - use kmalloc instead of slabs in portals (2430)
5037 - don't create duplicate records when a failover MDS is present (2442)
5038 - remove unnecessary mount age check (2332)
5039 - don't remove directory inodes from locks prematurely (2451)
5040 - don't break if MDS service name is the same as hostname (2103)
5041 - fix races in client write RPC generation when cache full (2482)
5043 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5046 - remove now-unused request->rq_obd (278)
5047 - if an allocation fails, print out how much memory we've used (1933)
5048 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5049 - add the "configurable stack size" patch to most series files (1256)
5050 - ability to write large log records, for 100+ OST configs (2306)
5051 - fix NULL deref when filter_prep fails (2314)
5052 - fix operator precedence error in filter_sync
5053 - dynamic allocation of socknal TX descriptors (2315)
5054 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5055 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5056 - ensure that CWARN messages in Portals always get to the syslog
5057 - __init/__exit are not for prototype decls (ldlm_init/exit)
5058 - x86-64 compile warning fixes
5059 - fix gateway LMC keyword conflict (2318)
5060 - fix MDS lock inversions in getattr/reint paths (1844)
5061 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5062 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5063 - DLM race condition prevented some lock evictions (2328)
5064 - ENOMEM detection and retry on socknal sends (2230)
5065 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5066 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5067 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5068 - better error messages when a client is rejected during recovery (1505)
5069 - avoid cancelling locks which were never granted, after failure (2330)
5070 - fix i_sem/journal inversion in mds_client_add (2333)
5071 - fix truncate/getattr lock cycle deadlock (2334)
5072 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5073 - fix two rare exit paths which could leak an l_lock() ref (2321)
5074 - fencepost error in MDS/OST orphan recovery (2226)
5075 - make log record alignment 8 bytes (1988)
5076 - lstripe now fails when requested offset > ost_count (2237)
5077 - ensure that all kernel series have a complete list.h (1607)
5078 - fix crashes in special-file operations (2316)
5079 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5081 - allow configurable automake binary, for testing new versions
5082 - small update to the lfs documentation
5084 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5086 * fix negative export reference count in fsfilt_sync (2312)
5088 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5089 * release candidate 0.9.1
5091 - orphans are moved into the PENDING directory for possible recovery
5092 - replayed opens now open by fid for orphan/rename safety (1042)
5093 - last close of an orphan inode generates a transno (683)
5094 - chdir() and mount() now pin the directory entry (1020)
5095 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5096 - discard very old requests without processing them (1502)
5097 - remove some common, well-understood CERRORs (1505)
5098 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5099 - clear "grant" flags in OST replies until OST grant code lands (1644)
5100 - fix read performance by not clobbering i_blksize on client (1598)
5101 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5102 - make LRU size a /proc tunable, clears locks when reduced (707)
5103 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5104 - server locks take references on exports now (1558)
5105 - build fixes for 2.4.20-rh trees (1663)
5106 - return an error from lov_create if all OSCs are inactive (1751)
5107 - fix import levels when a reconnect happens without a timeout (1597)
5108 - exit early from mds_open if we get a lookup error (1749)
5109 - partial page read at EOF wouldn't wait for disk before sending (1642)
5110 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5111 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5112 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5113 - write/truncate lock inversion (1639)
5114 - Don't auto-load obdclass, portals modules during cleanup (1495)
5115 - fix timestamps from jumping to "now" (1763)
5116 - extra journal assertions (1648)
5117 - add an extra multiunlink test (1771)
5118 - fix read_record/write_record API (1776)
5119 - fix leak of offset_extent, possible incorrect i_size later (1772)
5120 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5121 - Debugging for the kqswnal_get_idle_tx problems (1820)
5122 - Allow recovery to be attempted multiple times (1536)
5123 - Write out MDS last_rcvd file after it is first created (1600)
5124 - Fix tx_descriptor leak in failed transmit situations (1827)
5125 - ext3 journaling fixes for assertion failure after IO error (1871)
5126 - class_export_put() on freed export after completion AST error (1896)
5127 - Fix revalidate looping in VFS (1322)
5128 - Don't access a freed export during MDS_REINT timeout (1521)
5129 - Add open-unlink recovery support on the MDS (1673,1764)
5130 - Return an error if no MDS data was read from last_rcvd (1946)
5131 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5132 - Don't setup a disk device that doesn't match exported UUID (317)
5133 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5134 - avoid committing NULL handle in force close
5135 - local.sh is now a one-stripe LOV configuration
5136 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5137 - fix ext3 htree duplicate directory entry corruption (1516)
5138 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5139 - update to new LOV EA format (2097)
5140 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5142 2003-06-15 Phil Schwan <phil@clusterfs.com>
5145 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5146 - per-import/export recovery handling (958, 931, 959)
5147 - multiple last-rcvd slots, for serving multiple FSes (949)
5148 - connections are again shared between multiple imp/exports (963, 964)
5149 - "umount -f" would hang if any requests needed to be sent (393, 978)
5150 - avoid pinning large req buffer by copying for queued messages (989)
5151 - add "uuid" to "lctl device" command to help upcalls (991)
5152 - "open" RPCs with transnos would confuse recovery counters (1037)
5153 - do proper endian conversion of all wire messages (288, 340, 891)
5154 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5155 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5156 - fix ptlrpc_connection leak in target_handle_connect (1174)
5157 - fix import refcounting bug in OST and MDS cleanup (1134)
5158 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5159 - fix very unlikely obd_types race condition (501)
5160 - remove osc_open hack for echo_client (1187)
5161 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5162 - a failure in read_inode2 leads to deadlock (1139)
5163 - cancel ack-locks as soon as transaction is committed (1072)
5164 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5165 - make sure to commitrw after any preprw to avoid deadlock (1162)
5166 - failing to execute a file in a lustre FS would lock inode (1203)
5167 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5168 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5169 - obd_uuid2tgt would walk off the end of the list (1255)
5170 - on IA64 the osc would give portals incorrect bulk size (1258)
5171 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5172 - fix lock inversion caused by new llite matching code (1282)
5173 - limit the number of dirty pages on a client to 10MB (1286)
5174 - timed out locks were not being corrected cancelled (1289)
5175 - fix O_DIRECT above 4GB on IA-32 (1292)
5176 * major user-visible changes
5177 - fail out/fail over policy now controlled by the upcall (993)
5179 - add OBD_PING to check server availability and failure (954)
5180 - lustre messages are now sent in sending host order (288, 340, 891)
5181 - add eadatalen to MDS getattr reply (340)
5182 - OST read replies may contain second buffer, with per-page status (593)
5184 2003-03-11 Phil Schwan <phil@clusterfs.com>
5187 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5188 - failed open()s could cause deadlock; fixed (867, 869)
5189 - stop cancelling OST locks when files are closed (481)
5190 - overlapping XID spaces caused network corruption (851, 853)
5191 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5192 - setattr_raw added, to do single-RPC, server-side setattrs
5193 - lmc/lconf syntax change for OST UUIDs
5194 - fix crashy race condition between ptlrpc_free_req and osc_close
5195 - don't use request in mdc_enqueue if we hit a timeout (889)
5196 - don't set the inode i_size for regular files from the MDS (896)
5197 - handle out of order completion AST (842)
5198 - don't LBUG if a lock request times out after receiving AST (913)
5199 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5200 - if a bad lock AST arrives, send an error instead of dropping entirely
5201 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5202 - fix leak in bulk IO when only partially completed (899, 900, 926)
5203 - fix O_DIRECT for ia64 (55)
5204 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5205 - C-z after timeout could hang a process forever; fixed (977)
5207 - client-side I/O cache (678, 924, 929, 941, 970)
5209 - READPAGE and SETATTRs which don't take server-side locks get
5212 2003-02-11 Phil Schwan <phil@clusterfs.com>
5215 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5216 - Fix fsfilt_extN_readpage() to read a full page of directory
5217 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5218 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5219 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5220 - Most intent operations were removed, in favour of a new RPC mode
5221 that does a single RPC to the server and bypasses most of the VFS
5222 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5223 - Aggressively cancel local locks on DLM servers
5224 - mds_reint_unlink sends EA to the client if it's the last nlink.
5225 client uses that EA to unlink OST objects.
5226 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5227 - recursive symlinks were fixed (439)
5228 - fixed NULL deref in DEBUG_REQ
5229 - filter_update_lastobjid no longer calls sync, which annoyed extN
5230 - fixed multi-client small-writes to a single file problem (445)
5231 - fixed mtime updates during file writes (607)
5232 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5233 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5234 - fixed Linux OST size reporting problem (444, 656)
5235 - OST now updates object mtime with writes or setattr (607, 619)
5236 - client verifies file size before zeroing page past EOF (445)
5237 - OST now writes last allocated objid to disk with allocation (108)
5238 - LOV on echo now works (409)
5240 - mds_reint_unlink sends a new buffer, with the EA included. this
5241 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5242 set if a regular file was being unlinked, and it was the last link
5243 - use PtlGet from the target for bulk writes (315)
5244 - OST now updates object mtime with writes or setattr (607, 619)
5245 - LDLM now has a grant-time callback to revalidate locked items, if
5247 - Many MDS operations were reorganized to combat race conditions
5249 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5250 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5252 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5255 - Fully reactivate OST imports after reconnection (512, others)
5256 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5257 - More graceful error handling for truncating on dead OST (515)
5258 - Don't error out unless we're actually accessing dead stripes (474)
5259 - Fix garbage sizes when stripes are missing (410)
5260 - LRU counters were broken, causing constant lock purge (433, 432)
5261 - garbage on read from stripes with failed OSTs (441)
5262 - mark OSCs as active before reconnecting during recovery (438)
5263 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5264 - lfind did not preserve OST order in output (443)
5265 - symlinks cause hung clients, incorrect data (439)
5266 - stop dereferencing request after dropping refcount (457)
5267 - don't LASSERT(spin_is_locked) on non-SMP (455)
5268 - fixes for many rename() bugs
5269 - fstat didn't correctly synchronize attributes (399)
5270 - server must handle lock cancellation during blocking AST prep (487)
5271 - bulk descriptors were free()d too soon (511)
5272 - fix paths in lconf, which would load incorrect modules (451, 507)
5273 - fix confusing lconf 'host not found' error message (386)
5274 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5275 - fix race condition in mdc_blocking_ast() for inode access (526)
5276 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5277 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5278 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5279 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5280 - fix lstripe in several ways (kernel side) (527)
5281 - fix request leak in ldlm_cli_enqueue (262)
5282 - incorrect OSC was marked inactive after OST failure
5283 - call mds_fs_cleanup before unmounting filesystem (524)
5284 - fix races between taking ns_lock and ldlm_lock_change_resource
5285 - fix races updating LOV export open file list
5286 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5287 - fix recovery NULL deref in ldlm_cli_cancel_unused
5288 - fix some DLM races by using new hash table for lock handles (419)
5289 - permit the client to specify desired inodes, at replay
5290 - duplicate requests when we queue them for replay reintegration
5291 - fix last_rcvd offset calculation
5292 - sync after each recovered transaction, so we always make progress
5293 - never, not always, ERESTART requests without transnos
5294 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5296 - skip replay if the MDS says that the client is already connected
5297 - don't check for a recovery-enabled export to match lctl's UUID
5298 - don't INC_USE_COUNT for phantom exports
5299 - don't crash when cleaning up phantom exports (567)
5300 - don't double-finish or set replay data for errored mdc_open requests
5301 - abort requests when they time out, so we don't get old replies
5302 - send/receive replies for AST messages again
5303 - if the client says that it doesn't have the lock, cancel it on the
5305 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5306 mark it as destroyed, it will all work out when decref is called
5307 - fix module use counts (22, 581)
5309 - ASTs now expect a reply (server cancels lock on error reply)
5311 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5314 - fix many simultaneous client startup (392)
5315 - fix dentry->d_it clobbering
5316 - credentials weren't being shipped for readdir/getattr operations
5317 - remove invalid assertions triggered during some concurrent MD
5319 - proper Lustre versions added (336, 389)
5320 - fix memory leak for create error case (398)
5321 - fix LOV locking bug that would get cli/srv out of sync
5322 - fix echo client over LOV (409)
5323 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5324 - fix double-O_EXCL intent crash (424)
5325 - avoid sending multiple lock CANCELs (352)
5327 - MDS can do multi-client recovery (modulo bugs in new code)
5329 - many updates, edits, cleanups
5331 2002-11-18 Phil Schwan <phil@clusterfs.com>
5334 - fix null d_it dereference (346)
5335 - fix full OST/dbench hang (333)
5336 - fix permission problem with file removal (286)
5337 - fix removal of OSCs from LOV when they fail
5338 - fix NULL deref during bulk timeout (214)
5339 - fix problems related to multiple filesystems on one MDS (241)
5340 - fixed serious subtle metadata locking bugs
5341 - free locks on clients when inodes are removed due to memory
5343 - fix inode pointer in lock data (285)
5344 - partial support for multiple MDS on a single host (241)
5345 - data locks weren't cancelled at clear_inode time (290, 311)
5346 - intent locks could lead to unbounded lock growth (205)
5347 - added a maximum lock count, an LRU list, and a flusher
5348 - fix multiple rename (365)
5349 - properly abstracted the echo client
5350 - OSC locked 1 byte too many; fixed
5351 - rewrote brw callback code:
5352 - fixed recovery bugs related to LOVs (306)
5353 - fixed too-many-pages-in-one-write crash (191)
5354 - fixed (again) crash in sync_io_timeout (214)
5355 - probably fixed callback-related race (385)
5357 - Add capability to MDS protocol
5358 - LDLM cancellations and callbacks on different portals
5360 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5363 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5364 - defer open object destruction to close time (601981, 138)
5365 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5366 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5367 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5368 - don't call obd_setattr in ll_file_release for destroyed objects
5369 * protocol change to lustre_msg: move |version| and add |flags|
5370 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5371 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5372 * added replay of create, unlink, link and rename operations during
5373 MDS failover; recovery should be much more robust now
5374 * remove failed OSCs from LOVs (only lov_create uses this so far)
5375 * the lustre-HOWTO was brought (more) up to date (582544)
5377 2002-10-23 Phil Schwan <phil@clusterfs.com>
5380 - in-use dentries weren't being reused properly (617851)
5381 - prevent multiple LDLM setup (599178)
5382 - fix LOV size calculations for truncate (617853)
5383 - fix client handling of MDS intent errors (POSIX)
5384 - fix permission bug in lovstripe.c test (624321)
5385 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5386 - truncate past end of file could corrupt data
5387 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5388 - a race in recovery could return ETIMEDOUT to apps (623947)
5389 - building outside the source directory was fixed
5390 * the lustre-HOWTO was brought (more) up to date (582544)
5391 * major progress was made on recovery functionality
5393 2002-10-10 Phil Schwan <phil@clusterfs.com>
5396 - recovery deadlock fix
5397 - rm -rf causes LBUG fix (617817)
5398 - file open by multiple tasks fix (618962)
5399 - directory permissions bugs (602707 and 620007)
5400 - journal_stop fixed with locking (611313)
5401 - O_APPEND failures resolved (618273, perhaps 614459)
5402 - lconf PATH fix (619770)
5403 - IA64 build fix (621450)
5404 - RPC buffer sizes scale with amount of memory
5406 2002-10-01 Phil Schwan <phil@clusterfs.com>
5409 - locks would be cancelled without throwing away data pages,
5410 resulting in inconsistent data (605627)
5411 - inode attributes were not always being refreshed (605627, 612449)
5412 - lconf now continues to cleanup after lctl reports an error
5413 - MDS now enforces user permissions (602707)
5414 - lprocfs cleanup fixed, but not yet enabled (614157)
5415 - fixed infinite server hang, should a client not respond to an AST
5416 - avoid going into recovery if user calls readlink() with a buffer
5417 that's too small (613941)
5418 - AST RPCs no longer require replies (614867) -- this may be changed
5419 - don't crash server if client sends an IOV that's too big (611336)
5420 - fixed lock conversion deadlock (611892)
5421 - fixed the following of symlinks (614622)
5422 * recovery: the server can remove locks from a client that dies, other
5423 clients can make progress
5424 * more extN patch fixes
5425 * compile-time configurable ptlrpc buffer allocations
5427 - collaborative read cache document
5428 - Lustre Lite Performance CDR document-in-progress
5430 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5433 - fix typo in patch-2.4.18
5435 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5438 - clear ptlrpc request each time in handle_incoming_request()
5439 - unlink of files now destroys the object on the OST
5441 2002-09-19 Peter Braam <braam@clusterfs.com>
5443 * add hard link support
5444 * change obdfile creation method
5445 * kernel patch changed
5447 2002-09-19 Peter Braam <braam@clusterfs.com>
5450 - stack overflow bug in extN fixed
5452 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5454 * documentation updates
5455 - add man pages for config tools
5456 - update tests/README to describe testing with new config tools
5457 - finish metadata API descriptions
5458 * bug fixes and cleanups
5459 - statfs workaround for 16TB limit
5460 - LOV stripe allocation improved, can stripe on subset of OSTs
5461 - LOV file size/IO offset was wrong for files > 4GB in size
5462 - object EA data was being dropped, caused files to be unreadable
5463 - memory overflow with non-LOV OST caused memory corruption
5464 - fixed regression tests to work with new config tools, obdfilter
5465 - fixed bug when directory size became larger than 1 block
5466 - fixed bug (for single client case) when PWD was deleted
5467 - invalidate local directory pages when doing intent-based ops
5468 - avoid LDLM oops when lock callback contained bad data
5470 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5472 * documentation updates
5473 * bug fixes and cleanups
5474 - configuration tools
5477 - 64-bit compile warnings
5478 - 64-bit internal statfs data
5480 * test_brw on persistent OST devices
5482 * lprocfs (disabled)
5484 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5486 * documentation updates
5487 * bug fixes and cleanups
5488 * configuration tools
5490 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5493 * many small fixes to 0_5_4
5494 * io/network handling
5495 * thinkos in MDS operations
5497 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5500 * crucial basic fixes to 0.5.3
5501 * IOR, Iozone work over Elan
5504 2002-08-07 Phil Schwan <phil@clusterfs.com>
5505 * version 0_5_3, our first alpha
5506 * we use the new Portals iovs
5507 * documentation updates
5508 * bug fixes and cleanups
5509 * small changes in the DLM wire protocol
5511 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5512 * version 0_5_1 with some initial stability,
5513 * locking on MD and file I/O.
5514 * documentation updates
5515 * several bug fixes since 0.5.0
5516 * small changes in wire protocol
5518 2002-07-18 Phil Schwan <phil@clusterfs.com>
5520 * delivered as Lustre Light Alpha
5521 * fixed a crash after handling invalid MDS requests
5522 * fixed directory pages for architectures with non-4k pages sizes
5524 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5525 * release version v0_4_4
5526 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5528 * Fixes a number of interruption problems with OST operations.
5529 * Update documentation for portals header changes
5530 * Move all wire protocol structs/defines to lustre_idl.h
5531 * Fixes symlink length bug.
5532 * Add tcpdump to repository.
5534 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5535 * release version v0_4_3
5536 * Fixes statfs for inodes on extN.
5537 * Fixes bug in runtests which would delete /etc/hosts.
5538 * Use 64-bit object IDs wherever possible (not into VFS though)
5539 Remove ost_get_info, which is unused by lustre, and out of date.
5541 2002-07-03 Peter Braam <braam@clusterfs.com>
5542 * release version v0_4_2 Fixes a lookup error (type not passed)
5543 * move forward to head of Portals
5544 * move forward to latest Lustre kernel
5546 2002-06-25 Peter Braam <braam@clusterfs.com>
5547 * release version v0_4_1. Hopefully stable on single node use.