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.
19 Description: more cleanup in mds_lov
20 Details : not send LOV EA under replay, we can't know about they size at this
21 time. Don't allow client connect to mds before any ost connected,
22 for avoid problems with LOV EA size and returning EIO to client.
24 Severity : enhancement
26 Description: Interoperability at server side (Disk interoperability)
28 Severity : enhancement
30 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
32 Severity : enhancement
34 Description: New client IO stack (CLIO).
36 Severity : enhancement
38 Description: Commit on sharing. Eliminate inter-client dependencies between
39 uncommitted transactions by doing transaction commits.
40 Thereby clients may recovery independently.
43 Frequency : Create a symlink file with a very long name
45 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
46 Details : If there is no extra space in the request for early cancels,
47 ldlm_req_handles_avail() returns 0 instead of a negative value.
49 Severity : enhancement
51 Description: Add /proc entry for import status
52 Details : The mdc, osc, and mgc import directories now have
53 an import directory that contains useful import data for debugging
56 Severity : enhancement
58 Description: Re-disable certain /proc logging
59 Details : Enable and disable client's offset_stats, extents_stats and
60 extents_stats_per_process stats logging on the fly.
63 Frequency : Only on FC kernels 2.6.22+
65 Description: oops in statahead
66 Details : Do not drop reference count for the dentry from VFS when lookup,
67 VFS will do that by itself.
69 Severity : enhancement
71 Description: Generic /proc file permissions
72 Details : Set /Proc file permissions in a more generic way to enable non-
73 root users operate on some /proc files.
77 Description: Hitting mdc_commit_close() ASSERTION
78 Details : Properly handle request reference release in
79 ll_release_openhandle().
83 Description: quota recovery deadlock during mds failover
84 Details : This patch includes att18982, att18236, att18237 in bz14840.
86 1. fix osts hang when mds does failover with quotaon
87 2. prevent watchdog storm when osts threads wait for the
92 Frequency : only patchless client
93 Description: add workaround for race between add/remove dentry from hash
95 Severity : enhancement
97 Description: Allow OST glimpses to return PW locks
101 Description: LBUG when llog conf file is full
102 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
106 Description: Prevent import from entering FULL state when server in recovery
110 Description: service mount cannot take device name with ":"
111 Details : Only when device name contains ":/" will mount treat it as
117 Description: replace ptlrpcd with the statahead thread to interpret the async
118 statahead RPC callback
122 Frequency : on recovery
123 Description: I/O failures after umount during fail back
124 Details : if client reconnected to restarted server we need join to recovery
125 instead of find server handler is changed and process self
126 eviction with cancel all locks.
128 Severity : enhancement
130 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
132 Severity : enhancement
134 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
136 Severity : enhancement
138 Description: Upcall on Lustre log has been dumped
139 Details : Allow for a user mode script to be called once a Lustre log has
140 been dumped. It passes the filename of the dumped log to the
141 script, the location of the script can be specified via
142 /proc/sys/lnet/debug_log_upcall.
147 Description: avoid idr_remove called for id which is not allocated.
148 Details : Move assigment s_dev for clustered nfs to end of initialization,
149 to avoid problem with error handling.
154 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
155 Details : When connection is reused this not moved from CONN_UNUSED_HASH
156 into CONN_USED_HASH and this prodice warning when put connection
157 again in unused hash.
160 Severity : enhancement
162 Description: File striping can now be set to use an arbitrary pool of OSTs.
164 Severity : enhancement
166 Description: Export bytes_read/bytes_write count on OSC/OST.
170 Description: Early reply size mismatch, MGC loses connection
171 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
172 the connect flags are properly negotiated.
177 Description: quotas are not honored with O_DIRECT
178 Details : all writes with the flag O_DIRECT will use grants which leads to
179 this problem. Now using OBD_BRW_SYNC to guard this.
183 Description: add quota statistics
184 Details : 1. sort out quota proc entries and proc code.
185 2. add quota statistics
187 Severity : enhancement
189 Description: enable quota support for HEAD.
193 Description: Properly propagate oinfo flags from lov to osc for statfs
194 Details : restore missing copy oi_flags to lov requests.
196 Severity : enhancement
198 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
199 : l_getgroups(8), lst(8), routerstat(8)
200 Details : included man pages for llobdstat(8), llstat(8),
201 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
203 Severity : enhancement
205 Description: configure's --enable-quota should check the
206 : kernel .config for CONFIG_QUOTA
207 Details : configure is terminated if --enable-quota is passed but
208 : no quota support is in kernel
212 Description: Remove portals compatibility
213 Details : Remove portals compatibility, not interoperable with releases
218 Description: Resolve device initialization race
219 Details : Prevent proc handler from accessing devices added to the
220 obd_devs array but yet be intialized.
222 Severity : enhancement
224 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
226 Severity : enhancement
228 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
233 Description: Using wrong pointer in osc_brw_prep_request
234 Details : Access to array[-1] can produce panic if kernel compiled with
235 CONFIG_PAGE_ALLOC enabled
239 Description: Client runs out of low memory
240 Details : Consider only lowmem when counting initial number of llap pages
244 Description: *optional* service tags registration
245 Details : if the "service tags" package is installed on a Lustre node
246 When the filesystem is mounted, a local-node service tag will
247 be created. See http://inventory.sun.com/ for more information
248 about the Service Tags asset management system.
252 Description: Kernel BUG tries to release flock
253 Details : Lustre does not destroy flock lock before last reference goes
254 away. So always drop flock locks when client is evicted and
255 perform unlock regardless of successfulness of speaking to MDS.
259 Description: add refcount protection for osc callbacks, avoid panic on shutdown
263 Description: sanity test 65a fails if stripecount of -1 is set
264 Details : handle -1 striping on filesystem in ll_dirstripe_verify
269 Description: ASSERTION(CheckWriteback(page,cmd)) failed
270 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
273 Severity : enhancement
275 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
279 Description: do not process already freed flock
280 Details : flock can possibly be freed by another thread before it reaches
281 to ldlm_flock_completion_ast.
285 Description: LBUG during stress test
286 Details : Need properly lock accesses the flock deadlock detection list.
290 Description: oops in page fault handler
291 Details : kernel page fault handler can return two special 'pages' in error
292 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
296 Description: timeout with invalidate import.
297 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
298 should be handled by ptlrpcd. This produce long age waiting and
299 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
301 Severity : enhancement
303 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
307 Description: enable MGS and MDT services start separately
308 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
309 a co-located MGS without starting the MGS, which is a complement
310 to 'nosvc' mount option.
314 Frequency : after recovery
315 Description: precreate to many object's after del orphan.
316 Details : del orphan st in oscc last_id == next_id and this triger growing
317 count of precreated objects. Set flag LOW to skip increase count
318 of precreated objects.
322 Frequency : rare, on clear nid stats
323 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
324 Details : when clean nid stats sometimes try destroy live entry,
325 and this produce panic in free.
329 Description: Stack overflow during MDS log replay
330 ease stack pressure by using a thread dealing llog_process.
334 Description: wait until IO finished before start new when do lock cancel.
335 Details : VM protocol want old IO finished before start new, in this case
336 need wait until PG_writeback is cleared until check dirty flag and
337 call writepages in lock cancel callback.
339 Severity : enhancement
341 Description: using special macro for print time and cleanup in includes.
345 Description: mds_mfd_close() ASSERTION(rc == 0)
346 Details : In mds_mfd_close(), we need protect inode's writecount change
347 within its orphan write semaphore to prevent possible races.
351 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
355 Frequency : rare, on shutdown ost
356 Description: don't hit live lock with umount ost.
357 Details : shrink_dcache_parent can be in long loop with destroy dentries,
358 use shrink_dcache_sb instead.
362 Description: don't panic with use echo client
363 Details : echo client pass NULL as client nid pointer and this produce null
368 Description: fix build on ppc32
369 Details : compile code with -m64 flag produce wrong object file for ppc32.
373 Description: add message levels for liblustreapi
377 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
378 Details : Move the 'good_osts' check before the 'total_bavail' check. This
379 will result in an -EAGAIN and in the exit call path we call
380 alloc_rr() which will with increasing aggressiveness attempt to
381 aquire precreated objects on the minimum number of required OSCs.
385 Description: Use old size assignment to avoid deadlock
386 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
387 the scheduling while holding a spinlock. We do not need locking
388 for size in ll_update_inode() because size is only updated from
389 the MDS for directories or files without objects, so there is no
390 other place to do the update, and concurrent access to such inodes
391 are protected by the inode lock.
395 Description: resolve "_IOWR redefined" build error on SLES10
399 Description: dump the memory debugging after all modules are unloaded to
400 suppress false negative in conf_sanity test 39
402 Severity : enhancement
404 Description: build kernel-ib packages for OFED 1.3 in our release cycle
409 Description: fix SLES kernel versioning
410 Details : the kernel version for our SLES 10 kernel did not include a "-"
411 before the "smp" at the end. while this was not a problem in
412 general, it did mean that software trying to use the kernel
413 version to try to detect a vendor specific kernel would fail.
414 this was most evident by the OFED build scripts.
418 Description: Don't update lov_desc members until making sure they are valid
419 Details : When updating lov_desc members via proc fs, need fix their
420 validities before doing the real update.
424 Description: don't put request into delay list while invalidate in flight.
425 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
426 import in flight. this produce timeout for invalidate and sometimes
427 can cause stale data.
431 Frequency : on ppc only
432 Description: not convert ost objects for directory because it's not exist.
433 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
437 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
439 Severity : enhancement
441 Description: Update to support 2.6.22.14 vanilla kernel.
445 Frequency : rare, on recovery
446 Description: read procfs can produce deadlock in some situation
447 Details : Holding lprocfs lock which send rpc can produce block for destroy
448 obd objects and this also block reconnect with -EALREADY. This
449 isn't fix all lprocfs bugs - but make it rare.
451 Severity : enhancement
453 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
456 Frequency : frequent on X2 node
458 Description: mdc_set_open_replay_data LBUG
459 Details : Set replay data for requests that are eligible for replay.
463 Description: lustre_mgs: operation 101 on unconnected MGS
464 Details : When MGC is disconnected from MGS long enough, MGS will evict the
465 MGC, and late on MGC cannot successfully connect to MGS and a lot
466 of the error messages complaining that MGS is not connected.
470 Frequency : on network error
471 Description: panic with double free request if network error
472 Details : mdc_finish_enqueue is finish request if any network error ocuring,
473 but it's true only for synchronus enqueue, for async enqueue
474 (via ptlrpcd) this incorrect and ptlrpcd want finish request
477 Severity : enhancement
479 Description: client-side metadata stat-ahead during readdir(directory readahead)
480 Details : perform client-side metadata stat-ahead when the client detects
481 readdir and sequential stat of dir entries therein
484 Frequency : on start mds
486 Description: Implement get_info(last_id) in obdfilter.
489 Frequency : occasional
491 Description: Correctly check stale fid, not start epoch if ost not support SOM
492 Details : open with flag O_CREATE need set old fid in op_fid3 because
493 op_fid2 was overwritten with new generated fid, but mds can answer
494 with one of these two fids and both is not stale. Setattr
495 incorrectly started an epoch and assume will be called
496 done_writing, but without SOM done_writing ever being called.
499 Frequency : rare, depends on device drivers and load
501 Description: MDS or OSS nodes crash due to stack overflow
502 Details : Code changes in 1.8.0 increased the stack usage of some functions.
503 In some cases, in conjunction with device drivers that use a lot
504 of stack the MDS (or possibly OSS) service threads could overflow
505 the stack. One change which was identified to consume additional
506 stack has been reworked to avoid the extra stack usage.
509 Frequency : occasional
511 Description: Do not fail import if osc_interpret_create gets -EAGAIN
512 Details : If osc_interpret_create got -EAGAIN it immediately exits and
513 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
514 oscc_has_objects and see OSC has no objests and call
515 oscc_internal_create to resend create request.
517 Severity : enhancement
519 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
521 Severity : enhancement
523 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
526 Frequency : very rare
528 Description: Fix replay if there is an un-replied request and open
529 Details : In some cases, older replay request will revert the
530 mcd->mcd_last_xid on MDS which is used to record the client's
533 Severity : enhancement
535 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
537 Severity : enhancement
539 Description: Add rhel5 support to HEAD.
541 Serverity : enhancement
543 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
548 Description: Don't allow skipping OSTs if index has been specified.
549 Details : Don't allow skipping OSTs if index has been specified, make
550 locking in internal create lots better.
554 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
555 Details : access to bitfield in structure is always rounded to long
556 and this produce problem with not atomic change any bit.
560 Description: Lustre make rpms failed.
561 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
562 builds Lustre rpms from tarball.
565 Frequency : rare on shutdown ost
567 Description: If llog cancel was not send before clean_exports phase, this can
568 produce deadlock in llog code.
569 Details : If llog thread has last reference to obd and call class_import_put
570 this produce deadlock because llog_cleanup_commit_master wait when
571 last llog_commit_thread exited, but this never success because was
572 called from llog_commit_thread.
576 Description: allow userland application know is lost one of stripes.
577 Details : fill lvb_blocks with error code on ost and return it to
578 application if error flag found.
582 Description: NULL lov_tgts causing MDS oops
583 Details : more safe checks for NULL lov_tgts for avoid oops.
585 Severity : enhancement
587 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
591 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
592 Details : If an OST is down the MDS will hang indefinitely in
593 obd_statfs_rqset() waiting for the statfs data. While for
594 MDS QOS usage of statfs, it should not stuck in waiting.
596 Severity : enhancement
598 Description: remote_acl support
599 Details : Support ACL-based permission check for remote user.
600 Support setfacl/getfacl for remote user with the utils
601 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
602 the system "{s,g}etfacl" utils.
604 Severity : enhancement
606 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
608 Severity : enhancement
610 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
614 Description: make mgs_setparam() handle fsname containing dash
615 Details : fsname containing a dash does not work with lctl conf_param
617 Severity : enhancement
619 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
621 Severity : enhancement
623 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
627 Frequency : rare, at shutdown
628 Description: access already free / zero obd_namespace.
629 Details : if client_disconnect_export was called without force flag set,
630 and exist connect request in flight, this can produce access to
631 NULL pointer (or already free pointer) when connect_interpret
632 store ocd flags in obd_namespace.
636 Frequency : only at startup
637 Description: not alloc memory with spinlock held.
638 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
639 if any spinlock held.
641 Severity : enhancement
643 Description: make lustre randomly fail allocating memory
644 Details : Make lustre randomly failed allocating memory for testing purpose.
646 Severity : enhancement
648 Description: lost problems with lov objid file
649 Details : Fixes some scability and access to not inited memory problems
650 in work with lov objdid file.
655 Description: lfs find does not continue on file error
656 Details : Continue other files processing when a file/dir is absent.
660 Description: Inconsistent usage of lustre_pack_reply()
661 Details : Standardize the usage of lustre_pack_reply() such that it
662 always generate a CERROR on failure.
665 Frequency : occasional
667 Description: MDS hang or stay in waiting lock
668 Details : If client receive lock with CBPENDING flag ldlm need send lock
669 cancel as separate rpc, to avoid situation when cancel request
670 can't processed due all i/o threads stay in wait lock.
674 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
678 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
682 Description: Update to SLES9 kernel-2.6.5-7.287.3.
686 Description: Update to RHEL4 latest kernel.
688 Severity : enhancement
690 Description: Build SLES10 patchless client fails
691 Details : The configure was broken by run ./configure with
692 --with-linux-obj=.... argument for patchless client. When the
693 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
694 can't search header adequately. Use absolute path such as
695 -I($LINUX)/include instead.
699 Description: interrupt oig_wait produce painc on resend.
700 Details : brw_redo_request can be used for resend requests from ptlrpcd and
701 private set, and this produce situation when rq_ptlrpcd_data not
702 copyed to new allocated request and triggered LBUG on assert
703 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
704 ptlrpcd set if request is changed and can be safety changed to use
709 Description: LASSERT_{REQ,REP}SWAB macros are buggy
710 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
711 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
712 these macros and replace them with their difinitions instead.
716 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
717 Details : Update which_patch & target file for SLES10 latest kernel.
720 Description: Remove client patches from SLES 10 kernel.
721 Details : This causes SLES 10 clients to behave as patchless clients
722 even on a Lustre-patched (server) kernel.
724 Severity : enhancement
726 Description: self-adjustable client's lru lists
727 Details : use adaptive algorithm for managing client cached locks lru
728 lists according to current server load, other client's work
729 pattern, memory activities, etc. Both, server and client
730 side namespaces provide number of proc tunables for controlling
733 Severity : enhancement
735 Description: light-weight GSS support
736 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
737 and reduce performance overhead.
739 Severity : enhancement
741 Description: Linux keyring support
742 Details : Support using service of Linux keyring for Lustre GSS internal
743 context refresh/cache mechanism.
747 Description: Fix errors in lfs documentation
748 Details : Fixes man pages
752 Description: don't use GFP_* in generic Lustre code.
753 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
757 Description: obdclass is limited by single OBD_ALLOC(idarray)
758 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
762 Description: warnings with build patchless client with vanila 2.6.19 and up
763 Details : change old ctl_table style and replace ctl_table/ctl_table_header
764 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
768 Frequency : Only for SLES
769 Description: sanity_quota fail test_1
770 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
771 file, which makes TSTID[2] inunique.
775 Frequency : Only for SLES9
776 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
777 Details : Update target/ChangeLog/which_patch .
781 Frequency : Only for test_10 in sanity-quota.sh
782 Description: a bug in quota test code
783 Details : checking a test flag in a wrong place
787 Frequency : Only for test_13 in sanity-quota.sh
788 Description: a bug in quota test script
789 Details : shouldn't check overfull allocated quota
793 Description: jbd statistics
794 Details : Port older jbd statistics patch for sles10
796 Severity : enhancement
798 Description: Add printing inode info into message about error in writepage.
803 Description: reply_lock_interpret crash due to race with it and lock cancel.
804 Details : Do not replay locks that are being cancelled. Do not reference
805 locks by their address during replay, just by their handle.
809 Frequency : When flocks are used.
810 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
811 Details : Flock locks might destroy just granted lock if it could be merged
812 with another existing flock, this is done in completion handler,
813 so teach ldlm_cli_enquque_fini that this is a valid case for
819 Description: Oops in read and write path when failing to allocate lock.
820 Details : Check if lock allocation failed and return error back.
824 Description: lstripe command fails for valid OST index
825 Details : The stripe offset is compared to lov->desc.ld_tgt_count
826 instead of lov->desc.ld_active_tgt_count.
830 Description: sanity.sh failed test 103
831 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
836 Description: df doesn't work properly if diskfs blocksize != 4K
837 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
841 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
842 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
844 Severity : enhancement
846 Description: Adaptive timeouts
847 Details : RPC timeouts adapt to changing server load and network
848 conditions to reduce resend attempts and improve recovery time.
852 Description: llapi_file_create() does not allow some changes
853 Details : add llapi_file_open() that allows specifying the mode and
854 open flags, and also returns an open file handle.
858 Description: merge and cleanup kernel patches.
859 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
863 Description: Add journal checksum support.(Kernel part)
864 Details : The journal checksum feature adds two new flags i.e
865 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
866 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
867 indicates that the commit block contains the checksum for
868 the blocks described by the descriptor blocks. Now commit
869 record can be sent to disk without waiting for descriptor
870 blocks to be written to disk. This behavior is controlled
871 using JBD2_FEATURE_ASYNC_COMMIT flag.
875 Description: OSS needs mutliple precreate threads
876 Details : Add ability to start more than one create thread per OSS.
880 Description: Sanity.sh test_65k failed: setstripe should have succeeded
881 Details : execute lfs setstripe on client
885 Description: mds_obd_create error creating tmp object
886 Details : When the user sets quota on root, llog will be affected and can't
887 create files and write files.
890 Frequency : Always on ia64 patchless client, and possibly others.
892 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
893 Details : This allows the patchless client to be loaded on architectures
898 Description: RedHat Update kernel for RHEL5
899 Details : Add the kernel config file for RHEL5.
903 Description: RedHat Update kernel for RHEL5
904 Details : Modify the kernel config file more closer RHEL5.
908 Description: Build failure against Centos5 (RHEL5)
909 Details : Define PAGE_SIZE when it isn't present.
913 Description: test_7 on sanity-quota.sh can't be used on separate machines
914 Details : add facet to handle it
918 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
919 Details : with NFS, the anon dentry's parent was set to itself in
920 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
921 corresponding dentry other than use rec->ur_name.
923 Severity : enhancement
925 Description: lfs setstripe enhancement
926 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
930 Description: enable data checksumming by default
931 Details : enable checksum by default, allow --disable-checksum
932 configure option and "-o nochecksum" mount option. Checksums
933 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
937 Description: lustre support for RHEL5
938 Details : Add support for RHEL5.
942 Description: Client eviction due to failover config
943 Details : after a connection loss, the lustre client should attempt to
944 reconnect to the last active server first before trying the
945 other potential connections.
949 Description: when mds and osts use different quota unit(32bit and 64bit),
950 quota will be released repeatly.
951 Details : void sending multiple quota reqs to mds, which will keep the
952 status between the reqs.
956 Description: osts not allocated evenly to files
957 Details : change the condition to increase offset_idx
960 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
962 Description: Data corruption for OSTs that are formatted larger than 2TB
964 Details : When generating the bio request for lustre file writes the
965 sector number would overflow a temporary variable before being
966 used for the IO. The data reads correctly from Lustre (which
967 will overflow in a similar manner) but other file data or
968 filesystem metadata may be corrupted in some cases.
972 Description: Tune the kernel for good SCSI performance.
973 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
974 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
979 Description: rewrite ext2-derived code in obdclass/uuid.c
980 Details : rewrite inherited code (uuid parsing code from ext2 utils)
981 from scratch preserving functionality.
985 Description: "lfs find -obd UUID" prints directories
986 Details : "lfs find -obd UUID" will return all directory names instead
987 of just file names. It is incorrect because the directories
988 do not reside on the OSTs.
992 Description: use i_size_read and i_size_write in 2.6 port
993 Details : replace inode->i_size access with i_size_read/write()
996 Frequency : when removing large files
998 Description: scheduling issue during removal of large Lustre files
999 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1000 It causes scheduling issues when removing large files (17TB in the
1004 Frequency : only with liblustre clients on XT3
1006 Description: evictions taking too long
1007 Details : allow llrd to evict clients directly on OSTs
1010 Frequency : only on ppc
1012 Description: /proc/fs/lustre/devices broken on ppc
1013 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1014 We should make sure the type of 'index' is loff_t and then cast
1015 explicitly as needed below. Do not assign an explicitly cast
1019 Frequency : only for rhel5
1021 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1022 Details : Modify the target file & which_kernel.
1024 Severity : enhancement
1026 Description: omit set fsid for export NFS
1027 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1028 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1032 Frequency : Always, for kernels after 2.6.16
1033 Description: Fix warning idr_remove called for id=.. which is not allocated.
1034 Details : Last kernels save old s_dev before kill super and not allow
1035 to restore from callback - restore it before call kill_anon_super.
1039 Description: buffer overruns could theoretically occur
1040 Details : llapi_semantic_traverse() modifies the "path" argument by
1041 appending values to the end of the origin string, and a
1042 overrun may occur. Adding buffer overrun check in liblustreapi.
1046 Description: Fix error on 'ls .' at the top of the Lustre mount.
1047 Details : Don't revalidate dentry if it is a root dentry.
1051 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1052 Details : Modify vm-tunables-rhel4.patch.
1056 Description: Kernel config for 2.6.18-vanilla.
1057 Details : Modify targets/2.6-vanilla.target.in.
1058 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1059 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1060 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1061 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1065 Description: improve handling recoverable errors
1066 Details : if request processig with error which can be recoverable on server
1067 request should be resend, otherwise page released from cache and
1072 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1073 Details : Modify target file & which_patch.
1074 A flaw was found in the IA32 system call emulation provided
1075 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1076 value could be stored in the %RAX register, which could trigger an
1077 out-of-bounds system call table access. An untrusted local user
1078 could exploit this flaw to run code in the kernel
1079 (ie a root privilege escalation). (CVE-2007-4573).
1083 Description: O_DIRECT bypasses client statistics.
1084 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1085 read_ahead_stats were updated. copied the stats section from
1086 osc_send_oap_rpc() into async_internal().
1090 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1094 Description: change order of libsysio includes
1095 Details : '#include sysio.h' should always come before '#include xtio.h'
1099 Description: handle "serious error: objid * already exists" more gracefully
1100 Details : If LAST_ID value on disk is smaller than the objects existing in
1101 the O/0/d* directories, it indicates disk corruption and causes an
1102 LBUG(). If the object is 0-length, then we should use the existing
1103 object. This will help to avoid a full fsck in most cases.
1105 Severity : enhancement
1107 Description: adapt the lustre_config script to support the upgrade case
1108 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1109 server targets to 1.6 in parallel.
1113 Description: To avoid grant space > avaible space when the disk is almost
1114 full. Without this patch you might see the error "grant XXXX >
1115 available" or some LBUG about grant, when the disk is almost
1117 Details : In filter_check_grant, for non_grant cache write, we should
1118 check the left space by if (*left > ungranted + bytes), instead
1119 of (*left > ungranted), because only we are sure the left space
1120 is enough for another "bytes", then the ungrant space should be
1121 increase. In client, we should update cl_avail_grant only there
1122 is OBD_MD_FLGRANT in the reply.
1126 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1127 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1128 described in CVE-2007-4573. This problem could allow untrusted
1129 local users to gain root access.
1132 Frequency : when using O_DIRECT and quotas
1134 Description: Incorrect file ownership on O_DIRECT output files
1135 Details : block usage reported by 'lfs quota' does not take into account
1136 files that have been written with O_DIRECT.
1141 Description: touch file failed when fs is not full
1142 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1143 otherwise we can get ENOSP while fs is not full.
1147 Description: parallel lock callbacks
1148 Details : Instead of sending blocking and completion callbacks as separated
1149 requests, adding them to a set and sending in parallel.
1152 Frequency : only for Cray XT3
1153 Bugzilla : 12829/13455
1154 Description: Changing primary group doesn't change the group lustre assigns to
1156 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1157 overridden with the primary group provided by the group upcall,
1158 whereas the supplied fsgid can be trusted if it is in the list of
1159 supplementary groups returned by the group upcall.
1161 Severity : enhancement
1163 Description: Allow masking D_WARNING, D_ERROR messages from console
1164 Details : Console messages can now be disabled via lnet.printk.
1168 Description: User code with malformed file open parameter crashes client node
1169 Details : Before packing join_file req, all the related reference should be
1170 checked carefully in case some malformed flags cause fake
1171 join_file req on client.
1175 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1176 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1177 mode is equal to the granted mode. However, it should also check
1178 that LDLM_FL_CP_REQD is not set.
1182 Description: Heavy nfs access might result in deadlocks
1183 Details : After ELC code landed, it is now improper to enqueue any mds
1184 locks under och_sem, because enqueue might want to decide to
1185 cancel open locks for same inode we are holding och_sem for.
1189 Description: Client eviction while running blogbench
1190 Details : A lot of unlink operations with concurrent I/O can lead to a
1191 deadlock causing evictions. To address the problem, the number of
1192 oustanding OST_DESTROY requests is now throttled to
1193 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1194 callbacks are processed in priority.
1198 Description: enable ACLs on MDS by default
1199 Details : ACLs must be enabled on MDS by default.
1202 Frequency : PPC/PPC64 only
1204 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1205 Details : fix duplicated definitions between asm-ppc64/types.h and
1206 lustre_types.h on PPC.
1209 Frequency : PPC/PPC64 only
1211 Description: asm-ppc/segment.h does not exist
1212 Details : fix compile issue on PPC.
1216 Description: better handle error messages in extents code
1219 Frequency : RHEL4 only
1221 Description: mkfs is very slow on IA64/RHEL4
1222 Details : A performance regression has been discovered in the MPT Fusion
1223 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1224 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1225 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1227 Severity : enhancement
1229 Description: SNMP support enhancement
1230 Details : Adding total number of sampled request for an MDS node in snmp
1233 Severity : enhancement
1235 Description: Optimize ldlm waiting list processing for PR extent locks
1236 Details : When processing waiting list for read extent lock and meeting read
1237 lock that is same or wider to it that is not contended, skip
1238 processing rest of the list and immediatelly return current
1239 status of conflictness, since we are guaranteed there are no
1240 conflicting locks in the rest of the list.
1244 Description: Time out and refuse to reconnect
1245 Details : When the failover node is the primary node, it is possible
1246 to have two identical connections in imp_conn_list. We must
1247 compare not conn's pointers but NIDs, otherwise we can defeat
1248 connection throttling.
1252 Description: port llog fixes from b1_6 into HEAD
1253 Details : Port llog reference couting and some llog cleanups from b1_6
1254 (bug 10800) into HEAD, for protect from panic and access to already
1255 free llog structures.
1259 Description: Detect stride IO mode in read-ahead
1260 Details : When a client does stride read, read-ahead should detect that and
1261 read-ahead pages according to the detected stride pattern.
1265 Description: data checksumming impacts single node performance
1266 Details : add support for several checksum algorithm. Currently, only CRC32
1267 and Adler-32 are supported. The checksum type can be changed on
1268 the fly via /proc/fs/lustre/osc/*/checksum_type.
1272 Description: use adler32 for page checksums
1273 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1278 Description: build for x2 fails
1279 Details : fix compile issue on Cray systems.
1283 Description: Properly match for duplicate locks
1284 Details : Due to different lock order from skiplists code, we need to
1285 traverse entire list for now
1288 Frequency : only on PPC/SLES10
1290 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1291 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1292 be defined. Add a hack in mkfs_lustre.c to work around this compile
1297 Description: LASSERT on MDS when client holding flock lock dies
1298 Details : ldlm pool logic depends on number of granted locks equal to
1299 number of released locks which is not true for flock locks, so
1300 just exclude such locks from consideration.
1304 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1305 Details : Use fsfilt_sync() for both the whole filesystem sync and
1306 individual file sync to eliminate dangerous inode locking
1307 with I_LOCK that can lead to a deadlock.
1311 Description: performance in 1.6.3
1312 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1313 sure that 1MB requests are not fragmented by the block layer.
1315 Severity : enhancement
1317 Description: organize the server-side client stats on per-nid basis
1318 Details : Change the structure of stats under obdfilter and mds to
1326 The "uuid"s file would list the uuids of _active_ exports.
1327 And the clear entry is to clear all stats and stale nids.
1329 Severity : enhancement
1331 Description: eliminate client locks in face of contention
1332 Details : file contention detection and lockless i/o implementation
1333 for contended files.
1337 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1340 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1342 Description: MDS may lose file striping (and hence file data) in some cases
1343 Details : If there are additional extended attributes stored on the MDS,
1344 in particular ACLs, SELinux, or user attributes (if user_xattr
1345 is specified for the client mount options) then there is a risk
1346 of attribute loss. Additionally, the Lustre file striping
1347 needs to be larger than default (e.g. striped over all OSTs),
1348 and an additional attribute must be stored initially in the
1349 inode and then increase in size enough to be moved to the
1350 external attribute block (e.g. ACL growing in size) for file
1355 Description: skiplist implementation simplification
1356 Details : skiplists are used to group compatible locks on granted list
1357 that was implemented as tracking first and last lock of each
1358 lock group the patch changes that to using doubly linked lists
1362 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1363 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1364 because LASSERT on both the data supplied by a client, and the
1365 data on disk is dangerous and incorrect.
1367 Severity : enhancement
1369 Description: Slow truncate/writes to huge files at high offsets.
1370 Details : Directly associate cached pages to lock that protect those pages,
1371 this allows us to quickly find what pages to write and remove
1372 once lock callback is received.
1376 Description: more ldlm soft lockups
1377 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1378 starve other threads from the resource lock for a long time in
1379 case of long waiting queue, so change the debug level from
1380 D_OTHER to the less frequently used D_INFO.
1382 Severity : enhancement
1384 Description: add -gid, -group, -uid, -user options to lfs find
1388 Description: Hung threads in invalidate_inode_pages2_range
1389 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1390 once one is completed. As a result, some RPCs are stuck in the
1391 queue and are never sent.
1395 Description: filter threads hungs on waiting journal commit
1396 Details : Cleanup filter group llog code, then only filter group llog will
1397 be only created in the MDS/OST syncing process.
1401 Description: Procfs and llog threads access destoryed import sometimes.
1402 Details : Sync the import destoryed process with procfs and llog threads by
1403 the import refcount and semaphore.
1405 Severity : enhancement
1407 Description: openlock cache of b1_6 port to HEAD
1412 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1413 Details : If insertion of an extent fails, then discard the inode
1414 preallocation and free data blocks else it can lead to duplicate
1419 Description: don't always update ctime in ext3_xattr_set_handle()
1420 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1421 In some cases the ctime should not be updated, for example for
1422 2.0->1.8 compatibility it is necessary to delete an xattr and it
1423 should not update the ctime.
1427 Bugzilla : 15713/16362
1428 Description: Assertion in iopen_connect_dentry in 1.6.3
1429 Details : looking up an inode via iopen with the wrong generation number can
1430 populate the dcache with a disconneced dentry while the inode
1431 number is in the process of being reallocated. This causes an
1432 assertion failure in iopen since the inode's dentry list contains
1433 both a connected and disconnected dentry.
1437 Description: assertion failure in ldlm_handle2lock()
1438 Details : fix a race between class_handle_unhash() and class_handle2object()
1439 introduced in lustre 1.6.5 by bug 13622.
1444 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1445 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1450 Description: oops in sd_iostats_seq_show()
1451 Details : unloading/reloading the scsi low level driver triggers a kernel
1452 bug when trying to access the sd iostat file.
1457 Description: Kernel panics during QLogic driver reload
1458 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1459 patch, causing memory corruption.
1464 Description: journal_dev option does not work in b1_6
1465 Details : pass mount option during pre-mount.
1467 Severity : enhancement
1469 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1470 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1471 extent information of a file. It can be used to map logical blocks
1472 in a file to physical blocks in the block device.
1476 Description: LDLM soft lockups - improvement
1477 Details : It is be possible to send the lock handle along with each read
1478 or write request because the client is already doing a lock match
1479 itself so there isn't any reason the OST should have to re-do that
1483 Frequency : only on Cray X2
1485 Description: X2 build failures
1486 Details : fix build failures on Cray X2.
1490 Description: xid & resent requests
1491 Details : Initialize RPC XID from clock at startup (randomly if clock is
1494 Severity : enhancement
1496 Description: Add lustre_start utility to start or stop multiple Lustre servers
1501 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1502 Details : In case of memory pressure, list_del() can be called twice on
1503 req->rq_history_list, causing a kernel oops.
1507 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1508 Details : fix stack overflow in the distributed lock manager by defering
1509 export eviction after a failed AST to the elt thread instead of
1510 handling it in the dlm interpret routine.
1514 Description: Convert some comments to new format.
1515 Details : Update documenting comments to match doxygen conventions.
1519 Description: Grammar fixes.
1520 Details : A couple of trivial spelling fixes.
1524 Description: OSD_COUNTERS-mandatory
1525 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1526 to simplify the code.
1530 Description: simplify cmm_device freeing logic.
1531 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1535 Description: Add lockdep support to dt_object_operations locking interface.
1536 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1537 parameter indicating lock ordering. Update mdd code to use new
1542 Description: Introduce failloc constants for lockless IO tests.
1543 Details : Add two new failloc constants to test lockless IO. Only one of
1544 them in implemented---another is checked in yet to be landed
1549 Description: Add lockdep support for inode mutex.
1550 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1551 in the situations where Lustre has to lock more than one inode
1556 Description: Add optional invariants checking support.
1557 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1558 configure switch. This macro is to be used for consistency and
1559 sanity checks that are too expensive to be left in `production'
1564 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1565 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1570 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1571 Details : Use spin_lock_nested() in (the only) situation where more than
1572 one ldlm_lock is locked simultaneously. Also, fix possible
1573 dead-lock in ldlm_lock_change_resource() by enforcing particular
1578 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1579 Details : Instead of specifying each ldlm_lock call-back through separate
1580 parameter, wrap them into struct ldlm_callback_suite.
1584 Description: Kill join_lru obd method and its callers.
1585 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1586 in memory---a requirement implemented through ->o_join_lru() obd
1587 method in HEAD. Remove this method and its users.
1591 Description: Add asynchronous ldlm ENQUEUE completion handler.
1592 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1593 case that stock ldlm_completion_ast() cannot handle as it waits
1594 until lock is granted. Introduce new ldlm_completion_ast_async()
1595 for this. Also comment ldlm_completion_ast().
1599 Description: ldlm_error <-> errno conversion.
1600 Details : Add functions to map (rather arbitrary) between LDLM error codes
1601 and standard errno values. CLIO needs this to prevent LDLM specific
1602 constants from escaping ldlm and osc.
1606 Description: Kill unused ldlm_handle2lock_ns() function.
1607 Details : Kill unused ldlm_handle2lock_ns() function.
1611 Description: Add lu_ref support to ldlm_lock
1612 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1613 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1614 and ldlm_resource. LDLM interface has to be changed, because code that
1615 releases a reference on a lock, has to "know" what reference this is.
1616 In the most frequent case
1618 lock = ldlm_handle2lock(handle);
1620 LDLM_LOCK_PUT(lock);
1622 no changes are required. When any other reference (received _not_ from
1623 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1624 instead of LDLM_LOCK_PUT().
1626 Arguably, changes are pervasive, and interface requires some discipline
1627 for proper use. On the other hand, it was very instrumental in finding
1628 a few leaked lock references.
1632 Description: Add ldlm_lock_addref_try().
1633 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1634 attempts to addref a lock that might be being canceled
1639 Description: Add ldlm_weigh_callback().
1640 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1641 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1642 hard-coded `number of pages' logic.
1646 Description: Add lockdep annotations to llog code.
1647 Details : Use appropriately tagged _nested() locking calls in the places
1648 where llog takes more than one ->lgh_lock lock.
1652 Description: Add loi_kms_set().
1653 Details : Wrap kms updates into a helper function.
1657 Description: Constify instances of struct lsm_operations.
1658 Details : Constify instances of struct lsm_operations.
1662 Description: lu_conf support.
1663 Details : On a server, a file system object is uniquely identified
1664 by a fid, which is sufficient to locate and load all object
1665 state (inode). On a client, on the other hand, more data are
1666 necessary instantiate an object. Change lu_object_find() and
1667 friends to take additional `lu_conf' argument describing object.
1668 Typically this includes layout information.
1672 Description: lu_context fixes.
1673 Details : Introduce new lu_context functions that are needed on the client
1674 side, where some system threads (ptlrpcd) are shared by multiple
1675 modules, and so cannot be stopped during module shutdown.
1679 Description: Add start and stop methods to lu_device_type_operations.
1680 Details : Introduce two new methods in lu_device_type_operations, that are
1681 invoked when first instance of a given type is created and last one
1682 is destroyed respectively. This is need by CLIO.
1686 Description: Add lu_ref support to struct lu_device.
1687 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1688 track leaked references.
1692 Description: Introduce lu_kmem_descr.
1693 Details : lu_kmem_descr and its companion interface allow to create
1694 and destroy a number of kmem caches at once.
1698 Description: Fix lu_object finalization race.
1699 Details : Fix a race between lu_object_find() finding an object and its
1700 concurrent finalization. This race is (most likely) not possible
1701 on the server, but might happen on the client.
1705 Description: Introduce lu_ref interface.
1706 Details : lu_ref is a debugging module allowing to track references to
1707 a given object. It is quite cpu expensive, and has to be
1708 explicitly enabled with --enable-lu_ref. See usage description
1713 Description: Factor lu_site procfs stats into a separate function.
1714 Details : Separate lu_site stats printing code into a separate function
1715 that can be reused on a client.
1719 Description: Constify instances of struct {lu,dt,md}_device_operations.
1720 Details : Constify instances of struct {lu,dt,md}_device_operations.
1724 Description: Introduce struct md_site and move meta-data specific parts of
1725 struct lu_site here.
1726 Details : Move md-specific fields out of struct lu_site into special struct
1727 md_site, so that lu_site can be used on a client.
1731 Description: Kill mdd_lov_destroy().
1732 Details : Remove unused mdd code.
1736 Description: Add st_block checking to multistat.c.
1737 Details : Add st_block checking to multistat.c.
1741 Description: Add lu_ref support to struct obd_device.
1742 Details : Add lu_ref tracking to obd_device.
1746 Description: Kill obd_set_fail_loc().
1747 Details : Remove unused code.
1751 Description: Add special type for ptlrpc_request interpret functions.
1752 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1753 there. Actual usage will be in CLIO.
1757 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1758 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1759 doesn't work with lockdep.
1763 Description: Add rwv.c test program.
1764 Details : New testing program exercising readv(2) and writev(2) (Qian).
1768 Description: Add sendfile.c test program.
1769 Details : New testing program exercising sendfile(2) (Jay).
1773 Description: Ratelimit a message that can be very frequent.
1774 Details : Ratelimit a memory allocation failure message that can
1779 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1780 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1784 Description: Liblustre build fixes.
1785 Details : Liblustre build fixes.
1789 Description: libcfs: add cfs_{need,cond}_resched() interface.
1790 Details : libcfs: add cfs_{need,cond}_resched() definition and
1791 implementations for Linux, NT, and liblustre.
1793 Severity : enhancement
1795 Description: More exported tunables for mballoc
1796 Details : Add support for tunable preallocation window and new tunables for
1797 large/small requests
1801 Description: Detect corruption of block bitmap and checking for preallocations
1802 Details : Checks validity of on-disk block bitmap. Also it does better
1803 checking of number of applied preallocations. When corruption is
1804 found, it turns filesystem readonly to prevent further corruptions.
1808 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1809 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1810 window to 0. If it does hit cache miss, and read-pattern is still
1811 stride-io mode, does not reset the stride window, but also does not
1812 increase the stride window length in this case.
1816 Frequency : only for big-endian servers
1817 Description: Check if system is big-endian while mounting fs with extents feature
1818 Details : Mounting a filesystem with extents feature will fail on big-endian
1819 systems since ext3-based ldiskfs is not supported on big-endian
1820 systems. This can be over-riden with "bigendian_extents" mount option.
1822 --------------------------------------------------------------------------------
1824 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1826 * Support for kernels:
1827 2.6.5-7.283 (SLES 9),
1828 2.6.9-55.EL (RHEL 4),
1829 2.6.16.46-0.14 (SLES 10),
1830 2.6.18.8 vanilla (kernel.org)
1831 * Client support for unpatched kernels:
1832 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1833 2.6.16 - 2.6.22 vanilla (kernel.org)
1834 * Due to recently discovered recovery problems, we do not recommend
1835 using patchless RHEL 4 clients with this or any earlier release.
1836 * Recommended e2fsprogs version: 1.39.cfs8
1837 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1838 * Starting with this release, the ldiskfs backing filesystem required
1839 by Lustre is now in its own package, lustre-ldiskfs. This package
1840 should be installed. It is versioned separately from Lustre and
1841 may be released separately in future.
1845 Description: block reactivating mgc import until all deactivates complete
1846 Details : Fix race when failing back MDT/MGS to itself (testing)
1849 Frequency : at statup only
1851 Description: mds_lov_synchronize race leads to various problems
1852 Details : simultaneous MDT->OST connections at startup can cause the
1853 sync to abort, leaving the OSC in a bad state.
1855 Severity : enhancement
1857 Description: add optional extra BUILD_VERSION info
1858 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1859 to override the lustre version.
1862 Frequency : 2.6.18 servers only
1864 Description: ll_kern_mount() doesn't release the module reference
1865 Details : The ldiskfs module reference count never drops down to 0
1866 because ll_kern_mount() doesn't release the module reference.
1871 Description: server LBUG when using old ost_num_threads parameter
1872 Details : Accept the old ost_num_threads parameter but warn that it
1873 is deprecated, and fix an off-by-one error that caused an LBUG.
1878 Description: Transient SCSI error results in persistent IO issue
1879 Details : iobuf->dr_error is not reinitialized to 0 between two
1883 Frequency : sometimes when underlying device returns I/O errors
1885 Description: OSTs not going read-only during write failures
1886 Details : OSTs are not remounted read-only when the journal commit threads
1887 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1888 instead of the ext3 wrappers.
1892 Description: poor connect scaling with increasing client count
1893 Details : Don't run filter_grant_sanity_check for more than 100 exports
1894 to improve scaling for large numbers of clients.
1897 Frequency : SLES10 only
1899 Description: sanity-quota.sh quotacheck failed: rc = -22
1900 Details : Quotas cannot be enabled on SLES10.
1903 Frequency : liblustre clients only
1905 Description: getdirentries does not give error when run on compute nodes
1906 Details : getdirentries does not fail when the size specified as an argument
1907 is too small to contain at least one entry
1909 Severity : enhancement
1911 Description: Add LNET router traceability for debug purposes
1912 Details : If a checksum failure occurs with a router as part of the
1913 IO path, the NID of the last router that forwarded the bulk data
1914 is printed so it can be identified.
1919 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1920 Details : Due to a race condition, liblustre clients were occasionally
1921 evicted incorrectly.
1923 Severity : enhancement
1925 Description: lfs setstripe use optional parameters instead of postional
1928 Severity : enhancement
1930 Description: Nanosecond timestamp support for ldiskfs
1931 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1932 resolution timestamps. There is not yet support for this at
1933 the Lustre filesystem level.
1936 Frequency : during server recovery
1938 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1939 Details : request with rq_no_resend flag not awake l_wait_event if they get
1943 Frequency : nfs export on patchless client
1945 Description: connectathon hang when test nfs export over patchless client
1946 Details : Disconnected dentry cannot be found with lookup, so we do not need
1947 to unhash it or make it invalid
1950 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1951 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1952 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1957 Description: open req refcounting wrong on reconnect
1958 Details : If reconnect happened between getting open reply from server and
1959 call to mdc_set_replay_data in ll_file_open, we will schedule
1960 replay for unreferenced request that we are about to free.
1961 Subsequent close will crash in variety of ways.
1962 Check that request is still eligible for replay in
1963 mdc_set_replay_data().
1968 Description: disable writes to filesystem when reading health_check file
1969 Details : the default for reading the health_check proc file has changed
1970 to NOT do a journal transaction and write to disk, because this
1971 can cause reads of the /proc file to hang and block HA state
1972 checking on a healthy but otherwise heavily loaded system. It
1973 is possible to return to the previous behaviour during configure
1974 with --enable-health-write.
1976 Severity : enhancement
1978 Description: 64-bit inode version
1979 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1980 to the inode. This will be required for version-based recovery.
1985 Description: MDS fails to start if a duplicate client export is detected
1986 Details : in some rare cases it was possible for a client to connect to
1987 an MDS multiple times. Upon recovery the MDS would detect this
1988 and fail during startup. Handle this more gracefully.
1990 Severity : enhancement
1992 Description: Add -o localflock option to simulate old noflock
1994 Details : This will achieve local-only flock/fcntl locks
2000 Description: log_commit_thread vs filter_destroy race leads to crash
2001 Details : Take import reference before releasing llog record semaphore
2006 Description: Wrong request locking in request set processing
2007 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2008 imp_delayed_list, in this place should be used imp_lock.
2011 Frequency : when reconnection
2013 Description: Grant Leak when osc reconnect to OST
2014 Details : When osc reconnect ost, OST(filter) should check whether it
2015 should grant more space to client by comparing fed_grant and
2016 cl_avail_grant, and return the granted space to client instead
2017 of "new granted" space, because client will call osc_init_grant
2018 to update the client grant space info.
2021 Frequency : when client reconnect to OST
2023 Description: Grant Leak when osc do resend and replay bulk write
2024 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2025 bulk write request, because the grant info while be sync between
2026 OSC and OST when reconnect, and we should ignore the grant info
2027 these of resend/replay write req.
2032 Description: Grant space more than avaiable left space sometimes.
2033 Details : When then OST is about to be full, if two bulk writing from
2034 different clients came to OST. Accord the avaliable space of the
2035 OST, the first req should be permitted, and the second one
2036 should be denied by ENOSPC. But if the seconde arrived before
2037 the first one is commited. The OST might wrongly permit second
2038 writing, which will cause grant space > avaiable space.
2041 Frequency : when client is evicted
2043 Description: Grant might be wrongly erased when osc is evicted by OST
2044 Details : when the import is evicted by server, it will fork another
2045 thread ptlrpc_invalidate_import_thread to invalidate the
2046 import, where the grant will be set to 0. While the original
2047 thread will update the grant it got when connecting. So if
2048 the former happened latter, the grant will be wrongly errased
2049 because of this race.
2054 Description: Checking Stale with correct fid
2055 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2056 to check whether it is stale, because sometimes, we want the
2057 enqueue happened anyway, and op_data.fid2 will not be initialized.
2059 Severity : enhancement
2061 Description: update patchless client
2062 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2065 Frequency : only with 2.4 kernel
2067 Description: random memory corruption
2068 Details : size of struct ll_inode_info is to big for union inode.u and this
2069 can be cause of random memory corruption.
2074 Description: Memory leak in recovery
2075 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2076 It should also check obd_fail before fsfilt_start, otherwise if
2077 fsfilt_start return -EROFS,(failover mds during mds recovery).
2078 then the req will return with repmsg->transno = 0 and rc = EROFS.
2079 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2080 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2081 be freed no matter whether fsfilt_commit success or not.
2084 Frequency : only with huge count clients
2086 Description: Prevents from taking the superblock lock in llap_from_page for
2088 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2089 for llap_from_page call in ll_removepage prevents from taking the
2090 superblock lock for a soon died page.
2095 Description: Not check open intent error before release open handle
2096 Details : in some rare cases, the open intent error is not checked before
2097 release open handle, which may cause
2098 ASSERTION(open_req->rq_transno != 0), because it tries to release
2099 the failed open handle.
2104 Description: Set cat log bitmap only after create log success.
2105 Details : in some rare cases, the cat log bitmap is set too early. and it
2106 should be set only after create log success.
2110 Description: Accessing a block bevice can re-enable I/O when Lustre is
2111 tearing down a device.
2112 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2117 Description: service threads may hog cpus when there are a lot of requests
2118 Details : Insert cond_resched to give other threads a chance to use some CPU
2123 Description: the cat log was not initialized in recovery
2124 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2125 unlink log on mds will not be initialized until mds post
2126 recovery. And also in mds post recovery, the unlink log will
2127 initialization will be done asynchronausly, so there will be race
2128 between add unlink log and unlink log initialization.
2132 Description: brw_stats were being printed incorrectly
2133 Details : brw_stats were being printed as log2 but all of them were not
2134 recorded as log2. Also remove some code duplication arising from
2135 filter_tally_{read,write}.
2139 Frequency : rare, only in recovery.
2140 Description: ASSERTION(req->rq_type != LI_POISON) failed
2141 Details : imp_lock should be held while iterating over imp_sending_list for
2142 prevent destroy request after get timeout in ptlrpc_queue_wait.
2146 Description: replay-single.sh test 52 fails
2147 Details : A lock's skiplist need to be cleanup when it being unlinked
2148 from its resource list.
2152 Description: Short directio read returns full requested size rather than
2154 Details : Direct I/O operations should return actual amount of bytes
2155 transferred rather than requested size.
2157 Severity : enhancement
2159 Description: metadata RPC reduction (e.g. for rm performance)
2160 Details : decrease the amount of synchronous RPC between clients and servers
2161 by canceling conflicing lock before the operation on the client
2162 and packing thier handles into the main operation RPC to server.
2164 Severity : enhancement
2166 Description: Async OSC create to avoid the blocking unnecessarily.
2167 Details : If a OST has no remain object, system will block on the creating
2168 when need to create a new object on this OST. Now, ways use
2169 pre-created objects when available, instead of blocking on an
2170 empty osc while others are not empty. If we must block, we block
2171 for the shortest possible period of time.
2175 Description: improve handling recoverable errors
2176 Details : if request processig with error which can be recoverable on server
2177 request should be resend, otherwise page released from cache and
2180 Severity : enhancement
2182 Description: refine locking for avoid write wrong info into lov_objid file
2183 Details : fix possible races with add new target and write/update data in
2186 --------------------------------------------------------------------------------
2188 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2193 Frequency : on some architectures
2195 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2196 Details : Uninitialized flags sometimes cause configuration commands to
2200 Frequency : patchless clients only
2202 Description: missing __iget() symbol export
2203 Details : The __iget() symbol export is missing. To avoid the need for
2204 this on patchless clients the deathrow inode reaper is turned
2205 off, and we depend on the VM to clean up old inodes. This
2206 dependency was during via the fix for bug 12181.
2208 --------------------------------------------------------------------------------
2210 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2212 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2213 INTEROPERATE with older versions automatically. In many cases a
2214 special upgrade step is needed. Please read the
2215 user documentation before upgrading any part of a live system.
2216 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2217 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2218 * WARNING: Lustre configuration and startup changes are required with
2219 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2223 Severity : enhancement
2225 Description: MountConf
2226 Details : Lustre configuration is now managed via mkfs and mount
2227 commands instead of lmc and lconf. New obd types (MGS, MGC)
2228 are added for dynamic configuration management. See
2229 https://mail.clusterfs.com/wikis/lustre/MountConf for
2232 Severity : enhancement
2234 Description: dynamic OST addition
2235 Details : OSTs can now be added to a live filesystem
2237 Severity : enhancement
2239 Description: startup order invariance
2240 Details : MDTs and OSTs can be started in any order. Clients only
2241 require the MDT to complete startup.
2243 Severity : enhancement
2245 Description: parallel, asynchronous orphan cleanup
2246 Details : orphan cleanup is now performed in separate threads for each
2247 OST, allowing parallel non-blocking operation.
2249 Severity : enhancement
2251 Description: optimized stripe assignment
2252 Details : stripe assignments are now made based on ost space available,
2253 ost previous usage, and OSS previous usage, in order to try
2254 to optimize storage space and networking resources.
2256 Severity : enhancement
2258 Description: Permanently set tunables
2259 Details : All writable /proc/fs/lustre tunables can now be permanently
2260 set on a per-server basis, at mkfs time or on a live system.
2262 Severity : enhancement
2264 Description: Lustre message v2
2265 Details : Add lustre message format v2.
2267 Severity : enhancement
2269 Description: client OST exclusion list
2270 Details : Clients can be started with a list of OSTs that should be
2271 declared "inactive" for known non-responsive OSTs.
2275 Description: SPEC SFS validation failure on NFS v2 over lustre.
2276 Details : Changes the blocksize for regular files to be 2x RPC size,
2277 and not depend on stripe size.
2279 Severity : enhancement
2281 Description: Multiple MD RPCs in flight.
2282 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2283 To avoid overly-overloading MDS, introduce a limit on number of
2284 MDS RPCs in flight for a single client and add /proc controls
2285 to adjust this limit.
2287 Severity : enhancement
2289 Description: client read/write statistics
2290 Details : Add client read/write call usage stats for performance
2291 analysis of user processes.
2292 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2293 file access. extents_stats shows chunk size distribution.
2294 extents_stats_per_process show chunk size distribution per
2297 Severity : enhancement
2299 Description: mds statistics
2300 Details : Add detailed mds operations statistics in
2301 /proc/fs/lustre/mds/*/stats.
2305 Description: Failure of copying files with lustre special EAs.
2306 Details : Client side always return success for setxattr call for lustre
2307 special xattr (currently only "trusted.lov").
2312 Description: Refcount LNET uuids
2313 Details : The global LNET uuid list grew linearly with every startup;
2314 refcount repeated list entries instead of always adding to
2317 Severity : enhancement
2319 Description: Dynamic service threads
2320 Details : Within a small range, start extra service threads
2321 automatically when the request queue builds up.
2324 Frequency : mixed-endian client/server environments
2326 Description: mixed-endian crashes
2327 Details : The new msg_v2 system had some failures in mixed-endian
2330 Severity : enhancement
2332 Description: Easy OST removal
2333 Details : OSTs can be permanently deactivated with e.g. 'lctl
2334 conf_param lustre-OST0001.osc.active=0'
2336 Severity : enhancement
2338 Description: MGS proc entries
2339 Details : Added basic proc entries for the MGS showing what filesystems
2342 Severity : enhancement
2344 Description: provide MGS failover
2345 Details : Added config lock reacquisition after MGS server failover.
2347 Severity : enhancement
2349 Description: add Linux 2.4 support
2350 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2354 Description: a large application tries to do I/O to the same resource and dies
2355 in the middle of it.
2356 Details : Check the req->rq_arrival time after the call to
2357 ost_brw_lock_get(), but before we do anything about
2358 processing it & sending the BULK transfer request. This
2359 should help move old stale pending locks off the queue as
2360 quickly as obd_timeout.
2363 Frequency : when an incorrect nid is specified during startup
2365 Description: ptlrpc connect to non-existant node causes kernel crash
2366 Details : LNET can't be re-entered from an event callback, which
2367 happened when we expire a message after the export has been
2368 cleaned up. Instead, hand the zombie cleanup off to another
2371 Severity : enhancement
2373 Description: plain/inodebits lock performance improvement
2374 Details : Grouping plain/inodebits in granted list by their request modes
2375 and bits policy, thus improving the performance of search through
2379 Frequency : only if OST filesystem is corrupted
2381 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2382 Details : for a short time RPCs with bulk IO are in the replay list,
2383 but replay of bulk IOs is unimplemented. If the OST filesystem
2384 is corrupted due to disk cache incoherency and then replay is
2385 started it is possible to trip an assertion. Avoid putting
2386 committed RPCs into the replay list at all to avoid this issue.
2389 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2391 Description: System hang on startup
2392 Details : This bug allowed the liblustre (e.g. catamount) client to
2393 return to the app before handling all startup RPCs. This
2394 could leave the node unresponsive to lustre network traffic
2395 and manifested as a server ptllnd timeout.
2397 Severity : enhancement
2399 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2400 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2401 boolean to control whether to print peer debug info when a
2402 client's RPC times out.
2405 Frequency : only for kernels with patches from Lustre below 1.4.3
2407 Description: Remove old rdonly API
2408 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2411 Frequency : only for devices with external journals
2413 Description: Set external device read-only also
2414 Details : During a commanded failover stop, we set the disk device
2415 read-only while the server shuts down. We now also set any
2416 external journal device read-only at the same time.
2419 Frequency : when upgrading from 1.4 while trying to change parameters
2421 Description: The wrong (new) MDC name was used when setting parameters for
2422 upgraded MDT's. Also allows changing of OSC (and MDC)
2423 parameters if --writeconf is specified at tunefs upgrade time.
2426 Frequency : when setting specific ost indicies
2428 Description: QOS code breaks on skipped indicies
2429 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2430 created with --index need not be sequential.
2432 Severity : enhancement
2434 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2435 Details : The uninit_groups feature works in conjunction with the kernel
2436 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2437 up the pass1 processing of e2fsck. This is a read-only feature
2438 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2439 mount filesystems that have had this feature enabled.
2441 Severity : enhancement
2443 Description: Improve multi-block allocation algorithm to avoid fragmentation
2444 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2445 allocation locality and avoid filesystem fragmentation.
2447 ------------------------------------------------------------------------------
2449 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2451 * Support for kernels:
2452 2.4.21-47.0.1.EL (RHEL 3)
2453 2.6.5-7.283 (SLES 9)
2454 2.6.9-42.0.10.EL (RHEL 4)
2455 2.6.12.6 vanilla (kernel.org)
2456 2.6.16.27-0.9 (SLES 10)
2457 * Recommended e2fsprogs version: 1.39.cfs5
2459 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2463 Frequency : occasional, depends on client load and configuration
2464 Bugzilla : 12181, 12203
2465 Description: data loss for recently-modified files
2467 Details : In some cases it is possible that recently written or created
2468 files may not be written to disk in a timely manner (this should
2469 normally be within 30s unless client IO load is very high).
2470 The problem appears as zero-length files or files that are a
2471 multiple of 1MB in size after a client crash or client eviction
2472 that are missing data at the end of the file.
2474 This problem is more likely to be hit on clients where files are
2475 repeatedly created and unlinked in the same directory, clients
2476 have a large amount of RAM, have many CPUs, the filesystem has
2477 many OSTs, the clients are rebooted frequently, and/or the files
2478 are not accessed by other nodes after being written.
2480 The presence of the problem can be detected by looking at
2481 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2482 smaller than the second (nr_unused) then dirty files will not
2483 be flushed automatically to disk. "sync; sleep 10" should be
2484 run several times on the node before unmounting it to update
2485 Lustre (this is also safe to run on nodes without this problem).
2487 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2488 that can cause this same problem, so customers using that kernel
2489 also need to update the kernel in addition to Lustre. In order
2490 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2492 It is normal that files written just before a client crash (less
2493 than 30s) may not yet have been flushed to disk, even for local
2497 Frequency : frequent on thin XT3 nodes
2499 Description: UUID collision on thin XT3 Linux nodes
2500 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2501 randomly, since we relied on an insufficiently-seeded PRNG.
2506 Description: OSS hangs after "All ost request buffers busy"
2507 Details : A deadlock between quota and journal operations caused OSS
2508 hangs after printing "All ost request buffers busy."
2511 Frequency : always on liblustre builds
2513 Description: Cleanup compiler warnings on liblustre
2516 Frequency : always on liblustre builds on XT3
2518 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2522 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2523 > 32000 subdirectories
2524 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2525 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2526 subdirectory count crosses 32000. This will aid e2fsck to
2527 correctly handle more than 32000 subdirectories.
2530 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2532 Description: System hang on startup
2533 Details : This bug allowed the liblustre (e.g. catamount) client to
2534 return to the app before handling all startup RPCs. This
2535 could leave the node unresponsive to lustre network traffic
2536 and manifested as a server ptllnd timeout.
2538 Severity : enhancement
2540 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2541 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2542 boolean to control whether to print peer debug info when a
2543 client's RPC times out.
2548 Description: make O_SYNC working on 2.6 kernels
2549 Details : 2.6 kernels use different method for mark pages for write,
2550 so need add a code to lustre for O_SYNC work.
2555 Description: Failure to close file and release space on NFS
2556 Details : Put inode details into lock acquired in ll_intent_file_open.
2557 Use mdc_intent_lock in ll_intent_open to properly
2558 detect all kind of errors unhandled by mdc_enqueue.
2563 Description: proc file read during shutdown sometimes raced obd removal,
2565 Details : Add lock to prevent obd access after proc file removal.
2568 Frequency : Only for files larger than 4GB on 32-bit clients.
2570 Description: improperly doing page alignment of locks
2571 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2572 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2577 Description: RPCs being resent when they shouldn't be
2578 Details : Some RPCs that should not be resent are being resent. This
2579 can cause inconsistencies in the RPC state machine. Do not
2580 resend such requests.
2583 Frequency : rare, only with NFS export
2585 Description: Crash on NFS re-export node
2586 Details : under very unusual load conditions an assertion is hit in
2587 ll_intent_file_open()
2590 Frequency : only if OST filesystem is corrupted
2592 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2593 Details : for a short time RPCs with bulk IO are in the replay list,
2594 but replay of bulk IOs is unimplemented. If the OST filesystem
2595 is corrupted due to disk cache incoherency and then replay is
2596 started it is possible to trip an assertion. Avoid putting
2597 committed RPCs into the replay list at all to avoid this issue.
2602 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2603 Details : Large single O_DIRECT read and write calls can fail to allocate
2604 a sufficiently large buffer to process the request. In case of
2605 allocation failure the allocation is retried with a smaller
2606 buffer and broken into smaller requests.
2608 Severity : enhancement
2610 Description: Add -o localflock option to simulate old noflock behaviour.
2611 Details : This will achieve local-only flock/fcntl locks coherentness.
2616 Description: versioning check is incomplete
2617 Details : Checking the version difference of client vs. server, report
2618 error if the gap is too big.
2623 Description: add support PG_writeback bit
2624 Details : add support for PG_writeback bit for Lustre, for more carefull
2625 work with page cache in 2.6 kernel. This also fix some deadlocks
2626 and remove hack for work O_SYNC with 2.6 kernel.
2628 Severity : enhancement
2630 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2631 Details : The uninit_groups feature works in conjunction with the kernel
2632 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2633 up the pass1 processing of e2fsck. This is a read-only feature
2634 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2635 mount filesystems that have had this feature enabled.
2637 Severity : enhancement
2639 Description: Improve multi-block allocation algorithm to avoid fragmentation
2640 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2641 allocation locality and avoid filesystem fragmentation.
2643 ------------------------------------------------------------------------------
2645 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2647 * Support for kernels:
2648 2.6.9-42.0.3.EL (RHEL 4)
2649 2.6.5-7.276 (SLES 9)
2650 2.4.21-47.0.1.EL (RHEL 3)
2651 2.6.12.6 vanilla (kernel.org)
2652 2.6.16.21-0.8 (SLES10)
2653 * Recommended e2fsprogs version: 1.39.cfs2-0
2655 * The backwards-compatible /proc/sys/portals symlink has been removed
2656 in this release. Before upgrading, please ensure that you change
2657 any configuration scripts or /etc/sysctl.conf files that access
2658 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2659 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2660 in advance of the upgrade on any system running Lustre 1.4.6 or
2661 newer, since /proc/sys/lnet was added in that version.
2662 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2666 Frequency : only when quota is used
2668 Description: avoid scanning export list for quota master
2669 Details : Change the algorithms to avoid scanning export list in order
2670 to improve the efficiency.
2673 Frequency : MDS failover only, very rarely
2675 Description: "went back in time" messages on mds failover
2676 Details : The greatest transno may be lost when the current operation
2677 finishes with an error (transno==0) and the client's last_rcvd
2678 record is over-written. Save the greatest transno in the
2679 mds_last_transno for this case.
2682 Frequency : always for specific kernels and striping counts
2684 Description: client may get "Matching packet too big" without ACL support
2685 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2686 when trying to access files in certain configurations. The
2687 clients should in fact be denied when mounting because they do
2688 not understand ACLs.
2691 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2693 Description: many clients connecting with IO in progress causes connect timeouts
2694 Details : Avoid synchronous journal commits to avoid delays caused by many
2695 clients connecting/disconnecting when bulk IO is in progress.
2696 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2697 OST_IO_PORTAL to avoid delays behind potentially many pending
2701 Frequency : occasionally with multiple writers to a single file
2703 Description: shared writes to file may result in wrong size reported by stat()
2704 Details : Allow growing of kms when extent lock is cancelled
2707 Frequency : always with random mmap IO to multi-striped file
2709 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2710 Details : Only if the hole is at the end of OST object so that kms is too
2711 small. Fix is to increase kms accordingly in ll_nopage.
2714 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2716 Description: writes to a missing object would leak memory on the OST
2717 Details : If there is an inconsistency between the MDS and OST filesystems,
2718 such that the MDS references an object that doesn't exist, writes
2719 to that object will leak memory due to incorrect cleanup in the
2720 error handling path, eventually running out of memory on the OST.
2725 Description: Creating too long symlink causes lustre errors
2726 Details : Check symlink and name lengths before sending requests to MDS.
2729 Frequency : only if flock is enabled (not on by default)
2731 Description: posix locks not released on fd closure on 2.6.9+
2732 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2733 kernels, this caused such locks not to be released on fd close and
2734 then assertions on fs unmount about still used locks.
2737 Frequency : MDS failover only, very rarely
2739 Description: clients may get ASSERTION(granted_lock != NULL)
2740 Details : When request was taking a long time, and a client was resending
2741 a getattr by name lock request. The were multiple lock requests
2742 with the same client lock handle and
2743 mds_getattr_name->fixup_handle_for_resent_request found one of the
2744 lock handles but later failed with ASSERTION(granted_lock != NULL).
2749 Description: handle->h_buffer_credits > 0, assertion failure
2750 Details : h_buffer_credits is zero after truncate, causing assertion
2751 failure. This patch extends the transaction or creates a new
2755 Frequency : NFS re-export or patchless client
2756 Bugzilla : 11179, 10796
2757 Description: Crash on NFS re-export node (__d_move)
2758 Details : We do not want to hash the dentry if we don't have a lock.
2759 But if this dentry is later used in d_move, we'd hit uninitialised
2760 list head d_hash, so we just do this to init d_hash field but
2761 leave dentry unhashed.
2764 Frequency : NFS re-export or patchless client
2766 Description: NFS exports has problem with symbolic link
2767 Details : lustre client didn't properly install dentry when re-exported
2768 to NFS or running patchless client.
2771 Frequency : NFS re-export or patchless client
2773 Description: Various nfs/patchless fixes.
2774 Details : fixes reuse disconected alias for lookup process - this fixes
2775 warning "find_exported_dentry: npd != pd",
2776 fix permission error with open files at nfs.
2777 fix apply umask when do revalidate.
2780 Frequency : occasional
2782 Description: Crash on NFS re-export node
2783 Details : calling clear_page() on the wrong pointer triggered oops in
2784 generic_mapping_read().
2787 Frequency : rarely, using O_DIRECT IO
2789 Description: unaligned directio crashes client with LASSERT
2790 Details : check for unaligned buffers before trying any requests.
2793 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2795 Description: stale data returned from RAID cache
2796 Details : If only a small amount of IO is done to the RAID device before
2797 reading it again it is possible to get stale data from the RAID
2798 cache instead of reading it from disk.
2801 Frequency : always for sles10 kernel
2803 Description: sles10 support
2804 Details : ll_follow_link: compile fixes and using of nd_set_link
2805 under newer kernels.
2808 Frequency : depends on arch, kernel and compiler version, always on sles10
2811 Description: recursive or deep enough symlinks cause stack overflow
2812 Details : getting rid of large stack-allocated variable in
2816 Frequency : depends on hardware
2818 Description: lustre write performance loss in the SLES10 kernel
2819 Details : the performance loss is caused by using of write barriers in the
2820 ext3 code. The SLES10 kernel turns barrier support on by
2821 default. The fix is to undo that change for ldiskfs.
2823 ------------------------------------------------------------------------------
2825 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2827 * Support for kernels:
2828 2.6.9-42.0.3EL (RHEL 4)
2829 2.6.5-7.276 (SLES 9)
2830 2.4.21-47.0.1.EL (RHEL 3)
2831 2.6.12.6 vanilla (kernel.org)
2835 Frequency : quota enabled and large files being deleted
2837 Description: releasing more than 4GB of quota at once hangs OST
2838 Details : If a user deletes more than 4GB of files on a single OST it
2839 will cause the OST to spin in an infinite loop. Release
2840 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2845 Description: statfs data retrieved from /proc may be stale or zero
2846 Details : When reading per-device statfs data from /proc, in the
2847 {kbytes,files}_{total,free,avail} files, it may appear
2848 as zero or be out of date.
2851 Frequency : systems with MD RAID1 external journal devices
2853 Description: lconf's call to blkid is confused by RAID1 journal devices
2854 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2855 of returning all block devices that match the journal UUID.
2858 Frequency : always, for aggregate stripe size over 4GB
2860 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2861 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2862 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2865 Frequency : always if "lfs find" used on a local file/directory
2867 Description: "lfs find" segfaults if used on a local file/directory
2868 Details : The case where a directory component was not specified wasn't
2869 handled correctly. Handle this properly.
2872 Frequency : always on ppc64
2874 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2875 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2876 perform endianness conversion.
2879 Frequency : rarely (truncate to non-zero file size after write under load)
2880 Bugzilla : 10730, 10687
2881 Description: Files padded with zeros to next 4K multiple
2882 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2883 it is possible that files that are truncated to a non-zero size
2884 immediately after being written are filled with zero bytes beyond
2885 the truncated size. No file data is lost.
2887 Severity : enhancement
2889 Description: Allow recovery/failover for liblustre clients.
2890 Details : liblustre clients were unaware of failover configurations until
2893 Severity : enhancement
2895 Description: user file locks should fail when not mounting with flock option
2896 Details : Set up an error-returning stub in ll_file_operations.lock field
2897 to prevent incorrect behaviour when client is mounted without
2898 flock option. Also, set up properly f_op->flock field for
2902 Frequency : always on ia64
2904 Description: "lfs df" loops on printing out MDS statfs information
2905 Details : The obd_ioctl_data was not initialized and in some systems
2906 this caused a failure during the ioctl that did not return
2907 an error. Initialize the struct and return an error on failure.
2910 Frequency : SLES 9 only
2912 Description: Error of copying files with lustre special EAs as root
2913 Details : Client side always return success for setxattr call for lustre
2914 special xattr (currently only "trusted.lov").
2917 Frequency : rarely on clusters with both ia64+i386 clients
2919 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2920 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2921 clients are doing concurrent writes to the same file it is
2922 possible that the ia64 clients may LASSERT because the OST
2923 extent locks are not PAGE_SIZE aligned. Ensure that grown
2924 locks are always aligned on the request boundary.
2927 Frequency : specific use, occasional
2929 Description: Overwriting in use executable truncates on-disk binary image
2930 Details : If one node attempts to overwrite an executable in use by
2931 another node, we now correctly return ETXTBSY instead of
2932 truncating the file.
2934 Severity : enhancement
2936 Description: Async OSC create to avoid the blocking unnecessarily.
2937 Details : If a OST has no remain object, system will block on the creating
2938 when need to create a new object on this OST. Now, ways use
2939 pre-created objects when available, instead of blocking on an
2940 empty osc while others are not empty. If we must block, we block
2941 for the shortest possible period of time.
2946 Description: chmod on Lustre root is propagated to other clients
2947 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2948 invalid by the follow_mount time.
2951 Frequency : liblustre clients only
2953 Description: Race in 'instant cancel' lock handling could lead to such locks
2954 never to be granted in case of SMP MDS
2955 Details : Do not destroy not yet granted but cbpending locks in
2959 Frequency : replay/resend of open
2961 Description: non null lock assetion failure in mds_intent_policy
2962 Details : Trying to replay/resend lockless open requests resulted in
2963 mds_open() returning 0 with no lock. Now it sets a flag if
2964 a lock is going to be returned.
2966 Severity : enhancement
2968 Description: Checksum enhancements
2969 Details : New checksum enhancements allow for resending RPCs that failed
2972 Severity : enhancement
2974 Description: Tunables on number of dirty pages in cacche
2975 Details : Allow to set limit on number of dirty pages cached.
2980 Description: client crash on unmount - lock still has references
2981 Details : In some error handling cases it was possible to leak a lock
2982 reference on a client while accessing a file. This was not
2983 harmful to the client during operation, but would cause the
2984 client to crash when the filesystem is unmounted.
2987 Frequency : specific case, rare
2989 Description: ETXTBSY on mds though file not in use by client
2990 Details : ETXTBSY is no longer incorrectly returned when attempting to
2991 chmod or chown a directory that the user previously tried to
2992 execute or a currently-executing binary.
2995 Frequency : extremely rare except on liblustre-based clients
2997 Description: Lustre space not freed when files are deleted
2998 Details : Clean up open-unlinked files after client eviction. Previously
2999 the unlink was skipped and the files remained as orphans.
3004 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3005 Details : In some cases it was possible to send a blocking callback to
3006 a client doing a glimpse, even though that client didn't get
3007 a lock granted. When the glimpse lock is cancelled on the OST
3008 the freed lock is left on the waiting list and corrupted the list.
3011 Frequency : all core dumps
3013 Description: Broke core dumps to lustre
3014 Details : Negative dentry may be unhashed if parent does not have UPDATE
3015 lock, but some callers, e.g. do_coredump, expect dentry to be
3016 hashed after successful create, hash it in ll_create_it.
3018 ------------------------------------------------------------------------------
3020 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3022 * Support for kernels:
3023 2.6.9-42.0.2.EL (RHEL 4)
3024 2.6.5-7.276 (SLES 9)
3025 2.4.21-40.EL (RHEL 3)
3026 2.6.12.6 vanilla (kernel.org)
3030 Frequency : always on RHEL 3
3032 Description: Number of open files grows over time
3033 Details : The number of open files grows over time, whether or not
3034 Lustre is started. This was due to a filp leak introduced
3035 by one of our kernel patches.
3037 ------------------------------------------------------------------------------
3039 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3041 * Support for kernels:
3042 2.6.9-42.EL (RHEL 4)
3043 2.6.5-7.276 (SLES 9)
3044 2.4.21-40.EL (RHEL 3)
3045 2.6.12.6 vanilla (kernel.org)
3050 Bugzilla : 5719, 9635, 9792, 9684
3051 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3052 Details : If a server is under heavy load and cannot reply to new
3053 connection requests before the client resends the (re)connect,
3054 the connection handling code can behave badly if two service
3055 threads are concurrently handing separate (re)connections from
3056 the same client. Add better locking to the connection handling
3057 code, and ensure that only a single connection will be processed
3058 for a given client UUID, even if the lock is dropped.
3060 Severity : enhancement
3062 Description: add TCP zero-copy support to kernel
3063 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3064 sends if the hardware supports scatter-gather and checksumming.
3065 This allows socklnd to do client-write and server-read more
3066 efficiently and reduce CPU utilization from skbuf copying.
3069 Frequency : only if NFS exporting from client
3071 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3072 Details : If setting attributes on a file created under NFS that had
3073 never been opened it would be possible to oops the client
3074 if the file had no objects.
3077 Frequency : always for liblustre
3079 Description: liblustre client does MDS+OSTs setattr RPC for each write
3080 Details : When doing a write from a liblustre client, the client
3081 incorrectly issued an RPC to the MDS and each OST the file was
3082 striped over in order to update the timestamps. When writing
3083 with small chunks and many clients this could overwhelm the MDS
3084 with RPCs. In all cases it would slow down the write because
3085 these RPCs are unnecessary.
3087 Severity : enhancement
3089 Description: allow number of MDS service threads to be changed at module load
3090 Details : It is now possible to change the number of MDS service threads
3091 running. Adding "options mds mds_num_threads={N}" to the MDS's
3092 /etc/modprobe.conf will set the number of threads for the next
3093 time Lustre is restarted (assuming the "mds" module is also
3094 reloaded at that time). The default number of threads will
3095 stay the same, 32 for most systems.
3100 Description: OST crash if filesystem is unformatted or corrupt
3101 Details : If an OST is started on a device that has never been formatted
3102 or if the filesystem is corrupt and cannot even mount then the
3103 error handling cleanup routines would dereference a NULL pointer.
3108 Description: NULL pointer deref in llap_from_page.
3109 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3110 incorrect) mapping if the page was truncated/reclaimed while it was
3111 searched for. Check for this condition and skip such pages when
3112 doing readahead. Introduce extra check to llap_from_page() to
3113 verify page->mapping->host is non-NULL (so page is not anonymous).
3116 Frequency : Sometimes when using sys_sendfile
3118 Description: "page not covered by a lock" warnings from ll_readpage
3119 Details : sendfile called ll_readpage without right page locks present.
3120 Now we introduced ll_file_sendfile that does necessary locking
3121 around call to generic_file_sendfile() much like we do in
3125 Frequency : with certain MDS communication failures at client mount time
3127 Description: NULL pointer deref after failed client mount
3128 Details : a client connection request may delayed by the network layer
3129 and not be sent until after the PTLRPC layer has timed out the
3130 request. If the client fails the mount immediately it will try
3131 to clean up before the network times out the request. Add a
3132 reference from the request import to the obd device and delay
3133 the cleanup until the network drops the request.
3136 Frequency : occasionally during client (re)connect
3138 Description: assertion failure during client (re)connect
3139 Details : processing a client connection request may be delayed by the
3140 client or server longer than the client connect timeout. This
3141 causes the client to resend the connection request. If the
3142 original connection request is replied in this interval, the
3143 client may trip an assertion failure in ptlrpc_connect_interpret()
3144 which thought it would be the only running connect process.
3147 Frequency : only with obd_echo servers and clients that are rebooted
3149 Description: kernel BUG accessing uninitialized data structure
3150 Details : When running an obd_echo server it did not start the ping_evictor
3151 thread, and when a client was evicted an uninitialized data
3152 structure was accessed. Start the ping_evictor in the RPC
3153 service startup instead of the OBD startup.
3155 Severity : enhancement
3156 Bugzilla : 10193 (patchless)
3157 Description: Remove dependency on various unexported kernel interfaces.
3158 Details : No longer need reparent_to_init, exit_mm, exit_files,
3159 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3162 Frequency : rare (only users of deprecated and unsupported LDAP config)
3164 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3165 Details : LDAP apparently contains 'lustreName' attributes instead of
3166 'name'. A simple remapping of the name is sufficient.
3169 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3171 Description: waiting_locks_callback trips kernel BUG if client is evicted
3172 Details : Running with the dump_on_timeout debug flag turned on makes
3173 it possible that the waiting_locks_callback() can try to dump
3174 the Lustre kernel debug logs from an interrupt handler. Defer
3175 this log dumping to the expired_lock_main() thread.
3177 Severity : enhancement
3179 Description: Support NFS exporting on 2.6 kernels.
3180 Details : Implement non-rawops metadata methods for NFS server to use without
3181 changing NFS server code.
3184 Frequency : very rare (synthetic metadata workload only)
3186 Description: two racing renames might cause an MDS thread to deadlock
3187 Details : Running the "racer" program may cause one MDS thread to rename
3188 a file from being the source of a rename to being the target of
3189 a rename at exactly the same time that another thread is doing
3190 so, and the second thread has already enqueued these locks after
3191 doing a lookup of the target and is trying to relock them in
3192 order. Ensure that we don't try to re-lock the same resource.
3195 Frequency : only very large systems with liblustre clients
3197 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3198 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3199 all OSTs in parallel. This allows the eviction of stale liblustre
3200 clients to proceed much faster than if they were done in series,
3201 and also offers similar improvements for other set_info RPCs.
3206 Description: excessive CPU usage during initial read phase on client
3207 Details : During the initial read phase on a client, it would agressively
3208 retry readahead on the file, consuming too much CPU and impacting
3209 performance (since 1.4.5.8). Improve the readahead algorithm
3210 to avoid this, and also improve some other common cases (read
3211 of small files in particular, where "small" is files smaller than
3212 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3217 Description: MDS crash when receiving packet with unknown intent.
3218 Details : Do not LBUG in unknown intent case, just return -EFAULT
3220 Severity : enhancement
3221 Bugzilla : 9293, 9385
3222 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3223 Details : Do not serialize getattr (non-intent version) and statfs.
3226 Frequency : occasional, when OST network is overloaded/intermittent
3228 Description: client evicted by OST after bulk IO timeout
3229 Details : If a client sends a bulk IO request (read or write) the OST
3230 may evict the client if it is unresposive to its data GET/PUT
3231 request. This is incorrect if the network is overloaded (takes
3232 too long to transfer the RPC data) or dropped the OST GET/PUT
3233 request. There is no need to evict the client at all, since
3234 the pinger and/or lock callbacks will handle this, and the
3235 client can restart the bulk request.
3238 Frequency : Always when mmapping file with no objects
3240 Description: client crashes when mmapping file with no objects
3241 Details : Check that we actually have objects in a file before doing any
3242 operations on objects in ll_vm_open, ll_vm_close and
3248 Description: Request leak when working with deleted CWD
3249 Details : Introduce advanced request refcount tracking for requests
3250 referenced from lustre intent.
3252 Severity : Enhancement
3254 Description: Cache open file handles on client.
3255 Details : MDS now will return special lock along with openhandle, if
3256 requested and client is allowed to hold openhandle, even if unused,
3257 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3258 closing files for every read/write openration.
3260 Severity : Enhancement
3262 Description: Cache open negative dentries on client when possible.
3263 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3264 negative dentries on lock revocation.
3269 Description: Remounting a client read-only wasn't possible with a zconf mount
3270 Details : It wasn't possible to remount a client read-only with llmount.
3272 Severity : enhancement
3273 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3274 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3275 MPICH is included to add Lustre-specific ADIO interfaces.
3276 This is based closely on the UFS ADIO layer and only differs
3277 in file creation, in order to allow the OST striping to be set.
3278 This is user-contributed code and not supported by CFS.
3283 Description: extended inode attributes (immutable, append-only) work improperly
3284 when 2.4 and 2.6 kernels are used on client/server or vice versa
3285 Details : Introduce kernel-independent values for these flags.
3287 Severity : enhancement
3290 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3291 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3292 granularity. Now, Fractional MB granularity is be supported,
3293 this is very useful for low memory system.
3295 Severity : enhancement
3297 Description: Getattr by fid
3298 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3299 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3300 have locks and avoids getattr rpc requests for it on every stat.
3303 Frequency : Always, for filesystems larger than 2TB
3305 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3306 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3307 filesystems larger than 2TB to be mounted with mballoc enabled.
3310 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3312 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3313 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3314 filesystems larger than 2TB is mounted then the kernel will
3315 silently corrupt the start of the filesystem. CONFIG_LBD is
3316 enabled for all CFS-supported kernels, but the possibility of
3317 this happening with a modified kernel config exists.
3319 Severity : enhancement
3321 Description: add client O_DIRECT support for 2.6 kernels
3322 Details : It is now possible to do O_DIRECT reads and writes to files
3323 in the Lustre client mountpoint on 2.6 kernel clients.
3325 Severity : enhancement
3327 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3328 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3329 parallel, not waiting for response from every OST before sending
3330 a rpc to the next OST.
3335 Description: setattr vs write race when updating file timestamps
3336 Details : Client processes that update a file timestamp into the past
3337 right after writing to the file (e.g. tar) it is possible that
3338 the updated file modification time can be reset to the current
3339 time due to a race between processing the setattr and write RPC.
3341 Severity : enhancement
3343 Description: Bring 'lfs find' closer in line with regular Linux find.
3344 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3345 -print0 options and obtains all the needed info through the lustre
3348 Severity : enhancement
3350 Description: support up to 1024 configured devices on one node
3351 Details : change obd_dev array from statically allocated to dynamically
3352 allocated structs as they are first used to reduce memory usage
3357 Description: Flush dirty partially truncated pages during truncate
3358 Details : Immediatelly flush partially truncated pages in filter_setattr,
3359 this way we completely avoid having any pages in page cache on OST
3360 and can retire ugly workarounds during writes to flush such pages.
3365 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3366 Details : protect inode from truncation within vfs_unlink() context
3367 just take a reference before calling vfs_unlink() and release it
3368 when parent's i_sem is free.
3373 Description: last_id value checked outside lock on OST caused LASSERT failure
3374 Details : If there were multiple MDS->OST object precreate requests in
3375 flight, it was possible that the OST's last object id was checked
3376 outside a lock and incorrectly tripped an assertion. Move checks
3377 inside locks, and discard old precreate requests.
3380 Frequency : always, if extents are used on OSTs
3382 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3383 Details : OSTs using the extents format would not zero the high 16 bits of
3384 the index physical block number. This is not a problem for any
3385 OST filesystems smaller than 16TB, and no kernels support ext3
3386 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3387 new/modified files) and can be fixed for existing filesystems
3388 with e2fsprogs-1.39-cfs1.
3393 Description: import connection selection may be incorrect if timer wraps
3394 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3395 import connections to be ignored if the 32-bit jiffies counter
3396 wraps. Use a 64-bit jiffies counter.
3399 Frequency : very large clusters immediately after boot
3401 Description: LNET request buffers exhausted under heavy short-term load
3402 Details : If a large number of client requests are generated on a service
3403 that has previously never seen so many requests it is possible
3404 that the request buffer growth cannot keep up with the spike in
3405 demand. Instead of dropping incoming requests, they are held in
3406 the LND until the RPC service can accept more requests.
3409 Frequency : Sometimes during replay
3411 Description: Assertion failure in ll_local_open after replay.
3412 Details : If replay happened on an open request reply before we were able
3413 to set replay handler, reply will become not swabbed tripping the
3414 assertion in ll_local_open. Now we set the handler right after
3415 recognising of open request
3418 Frequency : very rare
3420 Description: kernel reports "badness in vsnprintf"
3421 Details : Reading from the "recovery_status" /proc file in small chunks
3422 may cause a negative length in lprocfs_obd_rd_recovery_status()
3423 call to vsnprintf() (which is otherwise harmless). Exit early
3424 if there is no more space in the output buffer.
3426 Severity : enhancement
3428 Description: clear OBD RPC statistics by writing to them
3429 Details : It is now possible to clear the OBD RPC statistics by writing
3430 to the "stats" file.
3435 Description: Client mtime is not the same on different clients after utimes
3436 Details : In some cases, the client was using the utimes() syscall on
3437 a file cached on another node. The clients now validate the
3438 ctime from the MDS + OSTs to determine which one is right.
3443 Description: Inability to activate failout mode
3444 Details : lconf script incorrectly assumed that in pythong string's numeric
3445 value is used in comparisons.
3448 Frequency : always with multiple stripes per file
3450 Description: Inefficient object allocation for mutli-stripe files
3451 Details : When selecting which OSTs to stripe files over, for files with
3452 a stripe count that divides evenly into the number of OSTs,
3453 the MDS is always picking the same starting OST for each file.
3454 Return the OST selection heuristic to the original design.
3459 Description: mount failures may take full timeout to return an error
3460 Details : Under some heavy load conditions it is possible that a
3461 failed mount can wait for the full obd_timeout interval,
3462 possibly several minutes, before reporting an error.
3463 Instead return an error as soon as the status is known.
3465 Frequency : quota enabled and large files being deleted
3467 Description: releasing more than 4GB of quota at once hangs OST
3468 Details : If a user deletes more than 4GB of files on a single OST it
3469 will cause the OST to spin in an infinite loop. Release
3470 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3475 Description: statfs data retrieved from /proc may be stale or zero
3476 Details : When reading per-device statfs data from /proc, in the
3477 {kbytes,files}_{total,free,avail} files, it may appear
3478 as zero or be out of date.
3481 Frequency : systems with MD RAID1 external journal devices
3483 Description: lconf's call to blkid is confused by RAID1 journal devices
3484 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3485 of returning all block devices that match the journal UUID.
3488 Frequency : always, for aggregate stripe size over 4GB
3490 Description: assertion fails when trying to use 4GB stripe size
3491 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3492 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3495 Frequency : always on ppc64
3497 Description: the first write on an ext3 filesystem with mballoc got stuck
3498 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3499 perform endianness conversion.
3501 ------------------------------------------------------------------------------
3503 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3505 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3506 INTEROPERATE with older versions automatically. Please read the
3507 user documentation before upgrading any part of a live system.
3508 * WARNING: Lustre networking configuration changes are required with
3509 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3512 * Support for kernels:
3513 2.6.9-22.0.2.EL (RHEL 4)
3514 2.6.5-7.244 (SLES 9)
3515 2.6.12.6 vanilla (kernel.org)
3518 Severity : enhancement
3519 Bugzilla : 7981/8208
3520 Description: Introduced Lustre Networking (LNET)
3521 Details : LNET is new networking infrastructure for Lustre, it includes
3522 a reorganized network configuration mode (see the user
3523 documentation for full details) as well as support for routing
3524 between different network fabrics. Lustre Networking Devices
3525 (LNDs) for the supported network fabrics have also been
3526 created for this new infrastructure.
3528 Severity : enhancement
3529 Description: Introduced Access control lists
3530 Details : clients can set ACLs on files and directories in order to have
3531 more fine-grained permissions than the standard Unix UGO+RWX.
3532 The MDS must be started with the "-o acl" mount option.
3534 Severity : enhancement
3535 Description: Introduced filesystem quotas
3536 Details : Administrators may now establish per-user quotas on the
3539 Severity : enhancement
3541 Description: Configuration change for the XT3
3542 The PTLLND is now used to run Lustre over Portals on the XT3
3543 The configure option(s) --with-cray-portals are no longer used.
3544 Rather --with-portals=<path-to-portals-includes> is used to
3545 enable building on the XT3. In addition to enable XT3 specific
3546 features the option --enable-cray-xt3 must be used.
3551 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3552 Details : A race between allocating a new llcd and re-getting the llcd_lock
3553 allowed another thread to grab newly-allocated llcd.
3555 Severity : enhancement
3557 Description: 2.6 OST async journal commit and locking fix to improve performance
3558 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3559 2.6 kernels are now async as they already were in 2.4 kernels so
3560 that they can commit concurrently with the network bulk transfer.
3561 For block-allocated files the filter allocation semaphore is held
3562 to avoid filesystem fragmentation during allocation. BKL lock
3563 removed for 2.6 xattr operations where it is no longer needed.
3568 Description: lconf incorrectly determined whether two IP networks could talk
3569 Details : In some more complicated routing and multiple-network
3570 configurations, lconf will avoid trying to make a network
3571 connection to a disjoint part of the IP space. It was doing the
3572 math incorrectly for one set of cases.
3577 Description: Fix for potential infinite loop processing records in an llog.
3578 Details : If an llog record is corrupted/zeroed, it is possible to loop
3579 forever in llog_process(). Validate the llog record length
3580 and skip the remainder of the block on error.
3583 Frequency : occasional (liblustre only)
3585 Description: liblustre could not open files whose last component is a symlink
3586 Details : sysio_path_walk() would incorrectly pass the open intent to
3587 intermediate path components.
3590 Frequency : rare (liblustre only with non-standard tuning)
3591 Bugzilla : 7201 (7350)
3592 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3593 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3594 after it has been released, only do this for referenced locks
3596 Severity : enhancement
3598 Description: specifying an (invalid) directory default stripe_size of -1
3599 would reset the directory default striping
3600 Details : stripe_size -1 was used internally to signal directory stripe
3601 removal, now use "all default" to signal dir stripe removal
3602 as a directory striping of "all default" is not useful
3605 Frequency : common for large clusters running liblustre clients
3607 Description: doing an ls when liblustre clients are running is slow
3608 Details : sending a glimpse AST to a liblustre client waits for every AST
3609 to time out, as liblustre clients will not respond. Since they
3610 cannot cache data we refresh the OST lock LVB from disk instead.
3612 Severity : enhancement
3614 Description: doing an ls at the same time as file IO can be slow
3615 Details : enqueue and other "small" requests can be blocked behind many
3616 large IO requests. Create a new OST IO portal for non-IO
3617 requests so they can be processed faster.
3620 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3622 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3623 created file with mode 000
3624 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3625 allow an ftruncate() as a non-root user to a file with mode 000.
3626 Lustre now respects this flag to disable mode checks when
3627 truncating a file owned by the user
3630 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3633 Description: Revoking locks from clients that went dead or catatonic might take
3635 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3636 cancellation of such locks instant on servers without waiting for
3637 any reply from clients. Clients drops these locks when cancel
3638 notification from server is received without replying.
3641 Frequency : liblustre-only, when liblustre client dies or becomes busy
3643 Description: Doing ls on Linux clients can take a long time with active
3645 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3646 granting such locks to it in the first place if possible. Locks
3647 are taken by proxy on the OST during the read or write and
3648 dropped immediately afterward. Add connect flags handling, do
3649 not grant locks to liblustre clients for glimpse ASTs.
3651 Severity : enhancement
3653 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3654 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3655 sequential reads and behaves badly with applications doing random
3656 reads. Improve it by reading ahead at least read region, and
3657 avoiding excessive large RPC for small reads.
3659 Severity : enhancement
3661 Description: Creating more than 1000 files for a single job may cause a load
3662 imbalance on the OSTs if there are also a large number of OSTs.
3663 Details : qos_prep_create() uses an OST index reseed value that is an
3664 even multiple of the number of available OSTs so that if the
3665 reseed happens in the middle of the object allocation it will
3666 still utilize the OSTs as uniformly as possible.
3671 Description: OST or MDS may oops in ping_evictor_main()
3672 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3673 but doesn't restart at beginning of obd_exports_timed list
3676 Severity : enhancement
3678 Description: improve by-nid export eviction on the MDS and OST
3679 Details : allow multiple exports with the same NID to be evicted at one
3680 time without re-searching the exports list.
3683 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3685 Description: MDS may oops in groups_free()
3686 Details : in rare race conditions a newly allocated group_info struct is
3687 freed again, and this can be NULL. The 2.4 compatibility code
3688 for groups_free() checked for a NULL pointer, but 2.6 did not.
3691 Frequency : common for liblustre clients doing little filesystem IO
3692 Bugzilla : 9352, 7313
3693 Description: server may evict liblustre clients accessing contended locks
3694 Details : if a client is granted a lock or receives a completion AST
3695 with a blocking AST already set it would not reply to the AST
3696 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3699 Frequency : lfs setstripe, only systems with more than 160 OSTs
3701 Description: unable to set striping with a starting offset beyond OST 160
3702 Details : llapi_create_file() incorrectly limited the starting stripe
3703 index to the maximum single-file stripe count.
3706 Frequency : LDAP users only
3708 Description: lconf did not handle in-kernel recovery with LDAP properly
3709 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3711 Severity : enhancement
3713 Description: bind OST threads to NUMA nodes to improve performance
3714 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3715 node and do their allocations there to localize memory access
3717 Severity : enhancement
3719 Description: llmount can determine client NID directly from Myrinet (GM)
3720 Details : the client NID code from gmnalnid was moved directly into
3721 llmount, removing the need to use this or specifying the
3722 client NID explicitly when mounting GM clients with zeroconf
3725 Frequency : if client is started with down MDS
3727 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3728 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3729 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3730 ptlrpc_import_delayed_req() because we didn't check if the
3731 request was interrupted, and we also didn't break out of the
3732 event loop if there was no timeout
3737 Description: data loss during non-page-aligned writes to a single file from
3738 both multiple nodes and multiple threads on one node at same time
3739 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3740 inconsistency led to false short-reads, that were cached and later
3741 used by ->prepare_write() to fill in partially written page,
3742 leading to data loss.
3745 Frequency : always, if lconf --abort_recovery used
3747 Description: lconf --abort_recovery fails with 'Operation not supported'
3748 Details : lconf was attempting to abort recovery on the MDT device and not
3751 Severity : enhancement
3753 Description: remove cleanup logs
3754 Details : replace lconf-generated cleanup logs with lustre internal
3755 cleanup routines. Eliminates the need for client-cleanup and
3758 Severity : enhancement
3760 Description: add support for EAs (user and system) on lustre filesystems
3761 Details : it is now possible to store extended attributes in the Lustre
3762 client filesystem, and with the user_xattr mount option it
3763 is possible to allow users to store EAs on their files also
3765 Severity : enhancement
3767 Description: Add possibility (config option) to show minimal available OST free
3769 Details : When compiled with --enable-mindf configure option, statfs(2)
3770 (and so, df) will return least minimal free space available from
3771 all OSTs as amount of free space on FS, instead of summary of
3772 free spaces of all OSTs.
3774 Severity : enhancement
3776 Description: do not expand extent locks acquired on OST-side
3777 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3778 by server: they are not cached anyway.
3781 Frequency : when mmap is used/binaries executed from Lustre
3783 Description: Unmmap pages before throwing them away from read cache.
3784 Details : llap_shrink cache now attempts to unmap pages before discarding
3785 them (if unmapping failed - do not discard). SLES9 kernel has
3786 extra checks that trigger if this unmapping is not done first.
3791 Description: lconf didn't resolve symlinks before checking to see whether a
3792 given mountpoint was already in use
3795 Frequency : when migrating failover services
3796 Bugzilla : 6395, 9514
3797 Description: When migrating a subset of services from a node (e.g. failback
3798 from a failover service node) the remaining services would
3799 time out and evict clients.
3800 Details : lconf --force (implied by --failover) sets the global obd_timeout
3801 to 5 seconds in order to quickly disconnect, but this caused
3802 other RPCs to time out too quickly. Do not change the global
3803 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3805 Severity : enhancement
3806 Frequency : if MDS is started with down OST
3807 Bugzilla : 9439,5706
3808 Description: Allow startup/shutdown of an MDS without depending on the
3809 availability of the OSTs.
3810 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3811 Add appropriate locking and lov-osc refcounts for safe
3812 cleaning. Add osc abort_inflight calls in case the
3813 synchronize never started.
3816 Frequency : occasional (Cray XT3 only)
3818 Description: root not authorized to access files in CRAY_PORTALS environment
3819 Details : The client process capabilities were not honoured on the MDS in
3820 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3821 been accessed by an authorized user then root was able to access
3822 the file on the local client also. The root user capabilities
3823 are now allowed on the MDS, as this environment has secure UID.
3826 Frequency : occasional
3828 Description: ldiskfs "too long searching" message happens too often
3829 Details : A debugging message (otherwise harmless) prints too often on
3830 the OST console. This has been reduced to only happen when
3831 there are fragmentation problems on the filesystem.
3836 Description: Division by zero in statfs when all OSCs are inactive
3837 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3838 lov_statfs divides by value returned by lov_get_stripecnt().
3842 Bugzilla : 9489, 3273
3843 Description: First write from each client to each OST was only 4kB in size,
3844 to initialize client writeback cache, which caused sub-optimal
3845 RPCs and poor layout on disk for the first writen file.
3846 Details : Clients now request an initial cache grant at (re)connect time
3847 and so that they can start streaming writes to the cache right
3848 away and always do full-sized RPCs if there is enough data.
3849 If the OST is rebooted the client also re-establishes its grant
3850 so that client cached writes will be honoured under the grant.
3855 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3856 Details : Now I/O RPCs go to different portal number and (presumably) fast
3857 lock requests (and glimses) and other RPCs get their own service
3858 threads pool that should be able to service those RPCs
3861 Severity : enhancement
3863 Description: Ability to exchange lustre version between client and servers and
3864 issue warnings at client side if client is too old. Also for
3865 liblustre clients there is ability to refuse connection of too old
3867 Details : New 'version' field is added to connect data structure that is
3868 filled with version info. That info is later checked by server and
3872 Frequency : rare, liblustre only.
3873 Bugzilla : 9296, 9581
3874 Description: Two simultaneous writes from liblustre at offset within same page
3875 might proceed at the same time overwriting eachother with stale
3877 Details : I/O lock withing llu_file_prwv was released too early, before data
3878 actually was hitting the wire. Extended lock-holding time until
3879 server acknowledges receiving data.
3882 Frequency : extremely rare. Never observed in practice.
3884 Description: avoid generating lustre_handle cookie of 0.
3885 Details : class_handle_hash() generates handle cookies by incrementing
3886 global counter, and can hit 0 occasionaly (this is unlikely, but
3887 not impossible, because initial value of cookie counter is
3888 selected randonly). Value of 0 is used as a sentinel meaning
3889 "unassigned handle" --- avoid it. Also coalesce two critical
3890 sections in this function into one.
3892 Severity : enhancement
3894 Description: allow liblustre clients to delegate truncate locking to OST
3895 Details : To avoid overhead of locking, liblustre client instructs OST to
3896 take extent lock in ost_punch() on client's behalf. New connection
3897 flag is added to handle backward compatibility.
3899 Severity : enhancement
3900 Bugzilla : 4928, 7341, 9758
3901 Description: allow number of OST service threads to be specified
3902 Details : a module parameter allows the number of OST service threads
3903 to be specified via "options ost ost_num_threads={N}" in the
3904 OSS's /etc/modules.conf or /etc/modprobe.conf.
3908 Bugzilla : 6146, 9635, 9895
3909 Description: servers crash with bad pointer in target_handle_connect()
3910 Details : In rare cases when a client is reconnecting it was possible that
3911 the connection request was the last reference for that export.
3912 We would temporarily drop the export reference and get a new
3913 one, but this may have been the last reference and the export
3914 was just destroyed. Get new reference before dropping old one.
3916 Severity : enhancement
3917 Frequency : if client is started with failover MDS
3919 Description: Allow multiple MDS hostnames in the mount command
3920 Details : Try to read the configuration from all specified MDS
3921 hostnames during a client mount in case the "primary"
3924 Severity : enhancement
3926 Description: Stop sending data to evicted clients as soon as possible.
3927 Details : Check if the client we are about to send or are sending data to
3928 was evicted already. (Check is done every second of waiting,
3929 for which l_wait_event interface was extended to allow checking
3930 of exit condition at specified intervals).
3933 Frequency : rare, normally only when NFS exporting is done from client
3935 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3937 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3940 Frequency : always, if srand() is called before liblustre initialization
3942 Description: Liblustre uses system PRNG disturbing its usage by user application
3943 Details : Introduce internal to lustre fast and high-quality PRNG for
3944 lustre usage and make liblustre and some other places in generic
3945 lustre code to use it.
3947 Severity : enhancement
3948 Bugzilla : 9477, 9557, 9870
3949 Description: Verify that the MDS configuration logs are updated when xml is
3950 Details : Check if the .xml configuration logs are newer than the config
3951 logs stored on the MDS and report an error if this is the case.
3952 Request --write-conf, or allow starting with --old_conf.
3954 Severity : enhancement
3956 Description: Handle symlinks in the path when checking if Lustre is mounted.
3957 Details : Resolve intermediate symlinks when checking if a client has
3958 mounted a filesystem to avoid duplicate client mounts.
3963 Description: lconf can hit an error exception but still return success.
3964 Details : The lconf command catches the Command error exception at the top
3965 level script context and will exit with the associated exit
3966 status, but doesn't ensure that this exit status is non-zero.
3971 Description: failure of ptlrpc thread startup can cause oops
3972 Details : Starting a ptlrpc service thread can fail if there are a large
3973 number of threads or the server memory is very fragmented.
3974 Handle this without oopsing.
3977 Frequency : always, only if liblustre and non-default acceptor port was used
3979 Description: liblustre cannot connect to servers with non-default acceptor port
3980 Details : tcpnal_set_default_params() was not called and was therefore
3981 ignoring the environment varaible TCPNAL_PORT, as well as other
3982 TCPNAL_ environment variables
3987 Description: two objects could be created on the same OST for a single file
3988 Details : If an OST is down, in some cases it was possible to create two
3989 objects on a single OST for a single file. No problems other
3990 than potential performance impact and spurious error messages.
3994 Bugzilla : 5681, 9562
3995 Description: Client may oops in ll_unhash_aliases
3996 Details : Client dcache may become inconsistent in race condition.
3997 In some cases "getcwd" can fail if the current directory is
4003 Description: Inode refcounting problems in NFS export code
4004 Details : link_raw functions used to call d_instantiate without obtaining
4005 extra inode reference first.
4009 Bugzilla : 9942, 9903
4010 Description: Referencing freed requests leading to crash, memleaks with NFS.
4011 Details : We used to require that call to ll_revalidate_it was always
4012 followed by ll_lookup_it. Also with revalidate_special() it is
4013 possible to call ll_revalidate_it() twice for the same dentry
4014 even if first occurence returned success. This fix changes semantic
4015 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4016 reference on a request referred from the intent.
4017 ll_intent_release() then releases such a request.
4020 Frequency : rare, normally benchmark loads only
4022 Description: unlinked inodes were kept in memory on the client
4023 Details : If a client is repeatedly creating and unlinking files it
4024 can accumulate a lot of stale inodes in the inode slab cache.
4025 If there is no other client load running this can cause the
4026 client node to run out of memory. Instead flush old inodes
4027 from client cache that have the same inode number as a new inode.
4030 Frequency : SLES9 2.6.5 kernel and long filenames only
4031 Bugzilla : 9969, 10379
4032 Description: utime reports stale NFS file handle
4033 Details : SLES9 uses out-of-dentry names in some cases, which confused
4034 the lustre dentry revalidation. Change it to always use the
4038 Frequency : rare, unless heavy write-truncate concurrency is continuous
4039 Bugzilla : 4180, 6984, 7171, 9963, 9331
4040 Description: OST becomes very slow and/or deadlocked during object unlink
4041 Details : filter_destroy() was holding onto the parent directory lock
4042 while truncating+unlinking objects. For very large objects this
4043 may block other threads for a long time and slow overall OST
4044 responsiveness. It may also be possible to get a lock ordering
4045 deadlock in this case, or run out of journal credits because of
4046 the combined truncate+unlink. Solution is to do object truncate
4047 first in one transaction without parent lock, and then do the
4048 final unlink in a new transaction with the parent lock. This
4049 reduces the lock hold time dramatically.
4052 Frequency : rare, 2.4 kernels only
4054 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4055 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4056 before unmounting their filesystem in order to avoid deadlock.
4057 The kernel_locked() function in 2.4 kernels only checks whether
4058 the kernel lock is held, not whether it is this process that is
4059 holding it as 2.6 kernels do.
4064 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4065 Details : The client ptlrpc code may be trying to reconnect to a down
4066 server before a previous connection attempt has timed out.
4067 Increase the reconnect interval to be longer than the connection
4068 timeout interval to avoid sending duplicate connections to
4072 Frequency : echo_client brw_test command
4074 Description: fix echo_client to work with OST preallocated code
4075 Details : OST preallocation code (5137) didn't take echo_client IO path
4076 into account: echo_client calls filter methods outside of any
4077 OST thread and, hence, there is no per-thread preallocated
4078 pages and buffers to use. Solution: hijack pga pages for IO. As
4079 a byproduct, this avoids unnecessary data copying.
4083 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4084 Description: Client can oops in mdc_commit_close() after open replay
4085 Details : It was possible for the MDS to return an open request with no
4086 transaction number in mds_finish_transno() if the client was
4087 evicted, but without actually returning an error. Clients
4088 would later try to replay that open and may trip an assertion
4089 Simplify the client close codepath, and always return an error
4090 from the MDS in case the open is not successful.
4093 Frequency : rare, 2.6 OSTs only
4095 Description: OST may deadlock under high load on fragmented files
4096 Details : If there was a heavy load and highly-fragmented OST filesystems
4097 it was possible to have all the OST threads deadlock waiting on
4098 allocation of biovecs, because the biovecs were not released
4099 until the entire RPC IO was completed. Instead, release biovecs
4100 as soon as they are complete to ensure forward IO progress.
4102 Severity : enhancement
4104 Description: Support for specifying external journal device at mount
4105 Details : If an OST or MDS device is formatted with an external journal
4106 device, this device major/minor is stored in the ext3 superblock
4107 and may not be valid for failover. Allow detecting and
4108 specifying the external journal at mount time.
4113 Description: Mounting an MDS with pending unlinked files may cause oops
4114 Details : target_finish_recovery() calls mds_postrecov() which returned
4115 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4116 considers this an error and immediately begins cleaning up the
4117 lov, just after starting the mds_lov process
4119 Severity : enhancement
4121 Description: Implement 'lfs df' to report actual free space on per-OST basis
4122 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4123 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4124 usage of objects; '-h' to report in human readable format.
4126 ------------------------------------------------------------------------------
4128 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4135 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4136 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4137 at startup during orphan unlinking. Instead, skip buddy update
4138 before setup, as it will be regenerated after recovery is complete.
4143 Description: If an OST is inactive, its locks might reference stale inodes.
4144 Details : lov_change_cbdata() must iterate over all namespaces, even if
4145 they are inactive to clear inode references from the lock.
4147 Severity : enhancement
4148 Frequency : occasional, if non-standard max_dirty_mb used
4150 Description: Client will block write RPCs if not enough grant
4151 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4152 then the client will block writes while waiting for another RPC
4153 to complete instead of consuming its dirty limit. With change
4154 we get improved performance when max_dirty_mb is small.
4156 Severity : enhancement
4157 Bugzilla : 3389, 6253
4158 Description: Add support for supplementary groups on the MDS.
4159 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4160 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4161 lookups for user supplementary groups into a cache.
4165 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4166 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4167 for liblustre O_EXCL is enforced by MDS.
4171 Description: readdir never returns NULL in liblustre.
4172 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4173 dentry in struct dirent.
4177 Frequency : liblustre only, depends on application IO pattern
4178 Description: liblustre clients evicted if not contacting servers
4179 Details : Don't put liblustre clients into the ping_evictor list, so
4180 they will not be evicted by the pinger ever.
4182 Severity : enhancement
4184 Description: Add ability to evict clients by NID from MDS.
4185 Details : By echoing "nid:$NID" string into
4186 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4187 $NID would be instantly evicted from this MDS and from all active
4188 OSTs connected to it.
4192 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4193 Details : lookup_it_finish() used to query file size from OSTs that was not
4198 Description: service threads change working directory to that of init
4199 Details : Starting lustre service threads may pin the working directory
4200 of the parent thread, making that filesystem busy. Threads
4201 now change to the working directory of init to avoid this.
4205 Frequency : during shutdown only
4206 Description: shutdown with a failed MDS or OST can cause unmount to hang
4207 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4212 Frequency : 2.6 only
4213 Description: chmod/chown may include an extra supplementary group
4214 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4215 supplementary group and if none is specified this is used.
4218 Bugzilla : 5479 (6816)
4220 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4221 Details : If lookup is in progress on a file that is unlinked we might try
4222 to revalidate the inode and fail in revalidate after lookup is
4223 complete and ll_file_open() enqueues the open again but
4224 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4227 Frequency : always, if lconf --abort_recovery used
4229 Description: lconf --abort_recovery fails with 'Operation not supported'
4230 Details : lconf was attempting to abort recovery on the MDT device and not
4233 ------------------------------------------------------------------------------
4235 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4240 Frequency : rare (only unsupported configurations with a node running as an
4242 Bugzilla : 6514, 5137
4243 Description: Mounting a Lustre file system on a node running as an OST could
4245 Details : OSTs now preallocates memory needed to write out data at
4246 startup, instead of when needed, to avoid having to
4247 allocate memory in possibly low memory situations.
4248 Specifically, if the file system is mounted on on OST,
4249 memory pressure could force it to try to write out data,
4250 which it needed to allocate memory to do. Due to the low
4251 memory, it would be unable to do so and the node would
4252 become unresponsive.
4254 Severity : enhancement
4256 Description: Addition of lconf --service command line option
4257 Details : lconf now accepts a '--service <arg>' option, which is
4258 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4260 Severity : enhancement
4262 Description: Failover mode is now the default for OSTs.
4263 Details : By default, OSTs will now run in failover mode. To return to
4264 the old behaviour, add '--failout' to the lmc line for OSTs.
4266 Severity : enhancement
4268 Description: Health checks are now provided for MDS and OSTs
4269 Details : Additional detailed health check information on MSD and OSTs
4270 is now provided through the procfs health_check value.
4273 Frequency : occasional, depends on IO load
4275 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4276 after numerous create/delete cycles
4277 Details : The ext3 inode allocation policy would not allocate new inodes
4278 very well on the OSTs because there are no new directories
4279 being created. Instead we look for groups with free space if
4280 the parent directories are nearly full.
4285 Description: Network or server problems during mount may cause partially
4286 mounted clients instead of returning an error.
4287 Details : The config llog parsing code may overwrite the error return
4288 code during mount error handling, returning success instead
4294 Description: MDS can fail to allocate large reply buffers
4295 Details : After long uptimes the MDS can fail to allocate large reply
4296 buffers (e.g. zconf client mount config records) due to memory
4297 fragmentation or consumption by the buffer cache. Preallocate
4298 some large reply buffers so that these replies can be sent even
4299 under memory pressure.
4303 Frequency : rare (liblustre)
4304 Description: fsx running with liblustre complained that using truncate() to
4305 extend the file doesn't work. This patch corrects that issue.
4306 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4307 ATTR_SIZE and lsm use in llu_setattr_raw.
4311 Frequency : rare, only 2.6 kernels
4312 Description: Unusual file access patterns on the MDS may result in inode
4313 data being lost in very rare circumstances.
4314 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4315 code in the 2.6 kernel caused the inode and/or EA data not to
4316 be read from disk, causing single-file corruption.
4320 Frequency : rare, only 2.6 filesystems using extents
4321 Description: Heavy concurrent write and delete load may cause data corruption.
4322 Details : It was possible under high-load situations to have an extent
4323 metadata block in the block device cache from a just-unlinked
4324 file overwrite a newly-allocated data block. We now unmap any
4325 metadata buffers that alias just-allocated data blocks.
4329 Frequency : filesystems with default stripe_count larger than 77
4330 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4331 Details : lconf specifies an inode size of 4096 bytes when the default
4332 stripe_count is larger than 77. This conflicts with the default
4333 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4334 this case to avoid pinning too much memory for large EAs.
4336 ------------------------------------------------------------------------------
4338 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4343 Frequency : rare (extremely heavy IO load with hundreds of clients)
4345 Description: Client is evicted, gets IO error writing to file
4346 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4347 caused clients to be evicted for AST timeouts. The fixes in
4348 bug 5192 mean we no longer need to have such short AST timeouts
4349 so ldlm_timeout has been increased.
4352 Frequency : occasional during --force or --failover shutdown under load
4353 Bugzilla : 5949, 4834
4354 Description: Server oops/LBUG if stopped with --force or --failover under load
4355 Details : a collection of import/export refcount and cleanup ordering
4356 issues fixed for safer force cleanup
4359 Frequency : only filesystems larger than 120 OSTs
4360 Bugzilla : 5990, 6223
4361 Description: lfs getstripe would oops on a very large filesystem
4362 Details : lov_getconfig used kfree on vmalloc'd memory
4365 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4366 Bugzilla : 6242, 6243
4367 Description: reading from files that had been truncated to a non-zero size
4368 but never opened returned no data
4369 Details : ll_file_read() reads zeros from no-object files to EOF
4374 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4375 Details : There's little harm in aborting MDS/OSS recovery and letting it
4376 try again, so I removed the LASSERT and return an error instead.
4378 Severity : enhancement
4380 Description: New debugging infrastructure for tracking down data corruption
4381 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4382 (b) cover more of the client-side code path, and (c) try to narrow
4383 down where problems occurred
4387 Bugzilla : 3819, 4364, 4397, 6313
4388 Description: Racing close and eviction MDS could cause assertion in mds_close
4389 Details : It was possible to get multiple mfd references during close and
4390 client eviction, leading to one thread referencing a freed mfd.
4392 Severity: : enhancement
4393 Bugzilla : 3262, 6359
4394 Description: Attempts to reconnect to servers are now more aggressive.
4395 Details : This builds on the enhanced upcall-less recovery that was added
4396 in 1.4.2. When trying to reconnect to servers, clients will
4397 now try each server in the failover group every 10 seconds. By
4398 default, clients would previously try one server every 25 seconds.
4403 Description: After recovery, certain operations trigger a failed
4404 assertion on a client.
4405 Details : Failing over an mds, using lconf -d --failover, while a
4406 client was doing a readdir() call would cause the client to
4407 LBUG after recovery completed and the readdir() was resent.
4409 Severity : enhancement
4411 Description: Default groups are now added by lconf
4412 Details : You can now run lconf --group <servicename> without having to
4413 manually add groups with lmc.
4416 Frequency : occasional
4418 Description: Nodes with an elan id of 0 trigger a failed assertion
4421 Frequency : always when accessing e.g. tty/console device nodes
4423 Description: tty and some other devices nodes cannot be used on lustre
4424 Details : file's private_data field is used by device data and lustre
4425 values in there got lost. New field was added to struct file to
4426 store fs-specific private data.
4429 Frequency : when exporting Lustre via NFS
4431 Description: NFSD failed occasionally when looking up a path component
4432 Details : NFSD is looking up ".." which was broken in ext3 directories
4433 that had grown large enough to become hashed.
4436 Frequency : Clusters with multiple interfaces not on the same subnet
4438 Description: Nodes will repeatedly try to reconnect to an interface which it
4439 cannot reach and report an error to the log.
4440 Details : Extra peer list entries will be created by lconf with some peers
4441 unreachable. lconf now validates the peer before adding it.
4444 Frequency : Only if a default stripe is set on the filesystem root.
4446 Description: Setting a default stripe on the filesystem root prevented the
4447 filesystem from being remounted.
4448 Details : The client was sending extra request flags in the root getattr
4449 request and did not allocate a reply buffer for the dir EA.
4452 Frequency : occasional, higher if lots of files are accessed by one client
4453 Bugzilla : 6159, 6097
4454 Description: Client trips assertion regarding lsm mismatch/magic
4455 Details : While revalidating inodes the VFS looks up inodes with ifind()
4456 and in rare cases can find an inode that is being freed.
4457 The ll_test_inode() code will free the lsm during ifind()
4458 when it finds an existing inode and then the VFS later attaches
4459 this free lsm to a new inode.
4463 Bugzilla : 6422, 7030
4464 Description: MDS deadlock between mkdir and client eviction
4465 Details : Creating a new file via mkdir or mknod (starting a transaction
4466 and getting the ns lock) can deadlock with client eviction
4467 (gets ns lock and trying to finish a synchronous transaction).
4470 Frequency : occasional
4471 Description: While starting a server, the fsfilt_ext3 module could not be
4473 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4474 kernels. Previously, lconf would still use the ext3 name
4475 when trying to load modules. Now, it will correctly use
4476 ext3 on 2.4 and ldiskfs on 2.6.
4478 Severity : enhancement
4479 Description: The default stripe count has been changed to 1
4480 Details : The interpretation of the default stripe count (0, to lfs
4481 or lmc) has been changed to mean striping across a single
4482 OST, rather than all available. For general usage we have
4483 found a stripe count of 1 or 2 works best.
4485 Severity : enhancement
4486 Description: Add support for compiling against Cray portals.
4487 Details : Conditional compiling for some areas that are different
4491 Frequency : occasional
4492 Bugzilla : 6409, 6834
4493 Description: Creating files with an explicit stripe count may lead to
4494 a failed assertion on the MDS
4495 Details : If some OSTs are full or unavailable, creating files may
4496 trigger a failed assertion on the MDS. Now, Lustre will
4497 try to use other servers or return an error to the
4501 Frequency : occasional
4503 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4504 caused lock timeout during readahead (since 1.4.2).
4505 Details : Processes doing readahead might match a lock that hasn't been
4506 granted yet if there are overlapping and conflicting lock
4507 requests. The readahead process waits on ungranted lock
4508 (original lock is CBPENDING), while OST waits for that process
4509 to cancel CBPENDING read lock and eventually evicts client.
4511 Severity : enhancement
4513 Description: Initial enabling of flock support for clients
4514 Details : Implements fcntl advisory locking and file status functions.
4515 This feature is provided as an optional mount flag (default
4516 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4517 locking are implemented yet, and those that are are not guaranteed
4518 to be completely correct in production environments.
4519 mount -t lustre -o [flock|noflock] ...
4522 Frequency : occasional
4524 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4525 assertion in the ext3 JBD (journaling) layer.
4526 Details : The b_committed_data struct is protected by the big kernel lock
4527 in 2.4 kernels, serializing journal_commit_transaction() and
4528 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4529 access to this struct. In 2.6 kernels there is finer grained
4530 locking to improve SMP performance of the JBD layer.
4534 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4539 Description: merge and cleanup kernel patches.
4542 -----------------------------------------------------------------------------
4544 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4546 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4547 versions of Lustre. Please update all servers and clients to
4548 version 1.4.2 or later at the same time. You must also run
4549 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4550 to update the configuration logs.
4552 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4553 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4554 - swab llog records of type '0' so we get proper header size/idx (5861)
4555 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4556 - fix rename of one directory over another leaking an inode (5953)
4557 - avoid SetPageDirty on 2.6 (5981)
4558 - don't re-add just-being-destroyed locks to the waiting list (5653)
4559 - when creating new directories, inherit the parent's custom
4560 striping settings if present parent (3048)
4561 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4562 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4563 - don't hold client locks on temporary worklist from l_lru (5666)
4564 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4565 - automatically evict dead clients (5921)
4566 - Update file size properly in create+truncate+fstat case (6196)
4567 - Do not unhash mountpoint dentries, do not allow removal of
4569 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4570 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4571 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4572 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4573 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4574 - Correctly handle reads of files with no objects (6243)
4575 - lctl recover will also mark a device active if deactivate used (5933)
4577 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4578 - allow --write-conf on an MDS with different nettype than client (5619)
4579 - don't write config llogs to MDS for mounts not from that MDS (5617)
4580 - lconf should create multiple TCP connections from a client (5201)
4581 - init scripts are now turned off by default; run chkconfig --on
4582 lustre and chkconfig --on lustrefs to use them
4583 - upcalls are no longer needed for clients to recover to failover
4585 - add --abort-recovery option to lconf to abort recovery on device
4587 - add support for an arbitrary number of OSTs (3026)
4588 - Quota support protocol changes.
4589 - forward compatibility changes to wire structs (6007)
4590 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4591 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4592 - improved "lustre status" script
4593 - initialize blocksize for non-regular files (6062)
4594 - added --disable-server and --disable-client configure options (5782)
4595 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4596 - Vanilla 2.4.29 support
4597 - increase maximum number of obd devices to 520 (6242)
4598 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4599 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4600 SLES 9 2.6.5 kernels
4601 - limit stripes per file to 160 (the maximum EA size) (6093)
4603 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4606 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4607 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4608 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4609 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4610 - small ext3 extents cleanups and fixes (5733)
4611 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4612 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4613 - avoid needless client->OST connect, fix handle mismatch (5317)
4614 - fix DLM error path that led to out-of-sync client, long delays (5779)
4615 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4616 - fix several locking issues related to i_size (5492,5624,5654,5672)
4617 - don't move pending lock onto export if it is already evicted (5683)
4618 - fix kernel oops when creating .foo in unlinked directory (5548)
4619 - fix deadlock in obdfilter statistics vs. object create (5811)
4620 - use time_{before,after} to avoid timer jiffies wrap (5882)
4621 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4622 - Do not leak request if server was not able to process it (5154)
4623 - If mds_open unable to find parent dir, make that negative lookup(5154)
4624 - don't create new directories with extent-mapping (5909, 5936)
4626 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4627 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4628 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4629 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4630 - support for RHEL4 kernel on i686 (5773)
4631 - provide error messages when incompatible logs are encountered (5898)
4633 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4634 * version 1.4.0.10 (1.4.1 release candidate 1)
4636 - don't keep a lock reference when lock is not granted (4238)
4637 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4638 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4639 - drop import inflight refcount on signal_completed_replay error (5255)
4640 - unlock page after async write error during send (3677)
4641 - handle missing objects in filter_preprw_read properly (5265)
4642 - no transno return for symlink open, don't save no-trasno open (3440)
4643 - don't try to complete elan receive that already failed (4012)
4644 - free RPC server reply state on error (5406)
4645 - clean up thread from ptlrpc_start_thread() on error (5160)
4646 - readahead could read extra page into cache that wasn't ejected (5388)
4647 - prevent races in class_attach/setup/cleanup/detach (5260)
4648 - don't dereference de->d_inode after l_dput of de (5458)
4649 - use "int" for stripe value returned from lock_to_stripe (5544)
4650 - mballoc allocation and error-checking fixes in 2.6 (5504)
4651 - block device patches to fix I/O request sizes in 2.6 (5482)
4652 - look up hostnames for IB nals (5602)
4653 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4654 - don't start multiple acceptors for the same port (5277)
4655 - fix incorrect LASSERT in mds_getattr_name (5635)
4656 - export a proc file for general "ping" checking (5628)
4657 - fix "lfs check" to not block when the MDS is down (5628)
4659 - service request history (4965)
4660 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4661 - create an "evict_client" /proc entry on OSTs, like the MDS has
4662 - fix mount usage message, return errors per mount(8) (5168)
4663 - change grep [] to grep "[]" in tests so they work in more UMLs
4664 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4665 - remove ancient LOV_MAGIC_V0 EA support (5047)
4666 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4667 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4668 - allow readahead allocations to fail when low on memory (5383)
4669 - mmap locking landed again, after considerable improvement (2828)
4670 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4672 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4675 - send OST transaction number in read/write reply to free req (4966)
4676 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4677 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4678 - the watchdog thread now runs as interruptible (5246)
4679 - flock/lockf fixes (but it's still disabled, pending 5135)
4680 - don't use EXT3 constants in llite code (5094)
4681 - memory shortage at startup could cause assertion (5176)
4683 - reorganization of lov code
4684 - single portals codebase
4686 - add extents/mballoc support (5025)
4687 - direct I/O reads in the obdfilter (4048)
4688 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4690 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4693 - send OST transaction number in read/write reply to free req (4966)
4694 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4695 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4696 - the watchdog thread now runs as interruptible (5246)
4697 - handle missing objects in filter_preprw_read properly (5265)
4698 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4699 - drop import inflight refcount on signal_completed_replay error (5255)
4700 - unlock page after async write error during send (3677)
4701 - return original error code on reconstructed replies (3761)
4702 - no transno return for symlink open, don't save no-trasno open (3440)
4704 - add pid to ldlm debugging output (4922)
4705 - bump the watchdog timeouts -- we can't handle 30sec yet
4706 - extra debugging for orphan dentry/inode bug (5259)
4708 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4711 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4712 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4713 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4714 - fix NULL dereference in /proc/sys/portals/routes (4827)
4715 - allow failed mdc_close() operations to be interrupted (4561)
4716 - stop precreate on OST before MDS would time out on it (4778)
4717 - don't send partial-page writes before EOF from client (4410)
4718 - discard client grant for sub-page writes on large-page clients (4520)
4719 - don't free dentries not owned by NFS code, check generation (4806)
4720 - fix lsm leak if mds_create_objects() fails (4801)
4721 - limit debug_daemon file size, always print CERROR messages (4789)
4722 - use transno after validating reply (3892)
4723 - process timed out requests if import state changes (3754)
4724 - update mtime on OST during writes, return in glimpse (4829)
4725 - add mkfsoptions to LDAP (4679)
4726 - use ->max_readahead method instead of zapping global ra (5039)
4727 - don't interrupt __l_wait_event() during strace
4729 - add software watchdogs to catch hung threads quickly (4941)
4730 - make lustrefs init script start after nfs is mounted
4731 - fix CWARN/ERROR duplication (4930)
4732 - return async write errors to application if possible (2248)
4733 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4734 - print NAL number in %x format (4645)
4735 - update barely-supported suse-2.4.21-171 series (4842)
4736 - support for sles 9 %post scripts
4737 - support for building 2.6 kernel-source packages
4738 - support for sles km_* packages
4740 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4743 - ignore -ENOENT errors in osc_destroy (3639)
4744 - notify osc create thread that OSC is being cleaned up (4600)
4745 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4746 - reconstruct ost_handle() like mds_handle() (4657)
4747 - create a new thread to do import eviction to avoid deadlock (3969)
4748 - let lconf resolve symlinked-to devices (4629)
4749 - don't unlink "objects" from directory with default EA (4554)
4750 - hold socknal file ref over connect in case target is down (4394)
4751 - allow more than 32000 subdirectories in a single directory (3244)
4752 - fix blocks count for O_DIRECT writes (3751)
4753 - OST returns ENOSPC from object create when no space left (4539)
4754 - don't send truncate RPC if file size isn't changing (4410)
4755 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4756 - bind to privileged port in socknal and tcpnal (3689)
4758 - rate limit CERROR/CWARN console message to avoid overload (4519)
4759 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4760 - basic mmap support (3918)
4761 - kernel patch series update from b1_4 (4711)
4763 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4766 - avoid crash during MDS cleanup with OST shut down (2775)
4767 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4768 - don't use bad inodes on the MDS (3744)
4769 - dynamic object preallocation to improve recovery speed (4236)
4770 - don't hold spinlock over lock dumping or change debug flags (4401)
4771 - don't zero obd_dev when it is force cleaned (3651)
4772 - print grants to console if they go negative (4431)
4773 - "lctl deactivate" will stop automatic recovery attempts (3406)
4774 - look for existing locks in ldlm_handle_enqueue() (3764)
4775 - don't resolve lock handle twice in recovery avoiding race (4401)
4776 - revalidate should check working dir is a directory (4134)
4778 - don't always mark "slow" obdfilter messages as errors (4418)
4780 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4783 - don't close LustreDB during write_conf until it is done (3860)
4784 - fix typo in lconf for_each_profile (3821)
4785 - allow dumping logs from multiple threads at one time (3820)
4786 - don't allow multiple threads in OSC recovery (3812)
4787 - fix debug_size parameters (3864)
4788 - fix mds_postrecov to initialize import for llog ctxt (3121)
4789 - replace config semaphore with spinlock (3306)
4790 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4791 - don't allow enqueue to complete on a destroyed export (3822)
4792 - down write_lock before checking llog header bitmap (3825)
4793 - recover from lock replay timeout (3764)
4794 - up llog sem before sending rpc (3652)
4795 - reduce ns lock hold times when setting kms (3267)
4796 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4797 - fix NULL deref and obd_dev leak on setup error (3312)
4798 - replace some LBUG about llog ops with error handling (3841)
4799 - don't match INVALID dentries from d_lookup and spin (3784)
4800 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4801 - fix invalid assertion in ptlrpc_set_wait (3880)
4803 - add libwrap support for the TCP acceptor (3996)
4804 - add /proc/sys/portals/routes for non-root route listing (3994)
4805 - allow setting MDS UUID in .xml (2580)
4806 - print the stack of a process that LBUGs (4228)
4808 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4811 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4812 - make sure to unset replay flag from failed open requests (3440)
4813 - if default stripe count is 0, use OST count for inode size (3636)
4814 - update parent mtime/ctime on client for create/unlink (2611)
4815 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4816 - free recovery state on server during a forced cleanup (3571)
4817 - unregister_reply for resent reqs (3063)
4818 - loop back devices mounting and status check on 2.6 (3563)
4819 - fix resource-creation race that can provoke i_size == 0 (3513)
4820 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4821 - more debugging for page-accounting assertion (3746)
4822 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4823 - avoid deadlock after precreation failure (3758)
4824 - fix race and lock order deadlock in orphan handling (3450, 3750)
4825 - add validity checks when grabbing inodes from l_ast_data (3599)
4827 - add /proc/.../recovery_status to obdfilter (3428)
4828 - lightweight CDEBUG infrastructure, debug daemon (3668)
4829 - change default OSC RPC parameters to be better on small clusters
4830 - turn off OST read cache for files smaller than 32MB
4831 - install man pages and include them in rpms (3100)
4832 - add new init script for (un)mounting lustre filesystems (2593)
4833 - run chkconfig in %post for init scripts (3701)
4834 - drop scimac NAL (unmaintained)
4836 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4839 - clean kiobufs before and after use (3485)
4840 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4841 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4842 - change init script to fail more clearly if not run as root (1528)
4843 - allow clients to reconnect during replay (1742)
4844 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4845 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4846 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4847 - fix cond_resched (3554)
4848 - only evict unfinished clients after recovery (3515)
4849 - allow bulk resend, prevent data loss (3570)
4850 - dynamic ptlrpc request buffer allocation (2102)
4851 - don't allow unlinking open directory if it isn't empty (2904)
4852 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4853 - remove extraneous obd dereference causing LASSERT failure (3334)
4854 - don't use get_cycles() when creating temp. files on the mds (3156)
4855 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4856 - handle EEXIST for set-stripe, set proper directory name (3336)
4858 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4859 - fix ksocknal_fmb_callback() error messages (2918)
4861 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4864 - don't copy lvb into (possibly NULL) reply on error (2983)
4865 - don't deref dentry after dput, don't free lvb on error (2922)
4866 - use the kms to determine writeback rpc length (2947)
4867 - increment oti_logcookies when osc is inactive (2948)
4868 - update client's i_blocks count via lvb messages (2543)
4869 - handle intent open/close of special files properly (1557)
4870 - mount MDS with errors=remount-ro, like obdfilter (2009)
4871 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4872 - don't use cancelling-locks' kms values (2947)
4873 - use highest lock extent for kms, not last one (2925)
4874 - don't dereference ERR_PTR() dentry in error handling path (3107)
4875 - fix thread race in portals_debug_dumplog() (3122)
4876 - create lprocfs device entries at setup instead of at attach (1519)
4877 - common AST error handler, don't evict client on completion race (3145)
4878 - zero nameidata in detach_mnt in 2.6 (3118)
4879 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4880 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4881 - initialize RPC timeout timer earlier for 2.6 (3219)
4882 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4883 - print nal/nid for unknown nid (3258)
4884 - additional checks for oscc recovery before doing precreate (3284)
4885 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4886 - don't crash in mdc_close for bad permissions on open (3285)
4887 - zero i_rdev for non-device files (3147)
4888 - clear page->private before handing to FS, better assertion (3119)
4889 - tune the read pipeline (3236)
4890 - fix incorrect decref of invalidated dentry (2350)
4891 - provide read-ahead stats and refine rpc in flight stats (3328)
4892 - don't hold journal transaction open across create RPC (3313)
4893 - update atime on MDS at close time (3265)
4894 - close LDAP connection when recovering to avoid server load (3315)
4895 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4896 - don't leak open file on MDS after open resend (3325)
4897 - serialize filter_precreate and filter_destroy_precreated (3329)
4898 - loop device shouldn't call sync_dev() for nul device (3092)
4899 - clear page cache after eviction (2766)
4900 - resynchronize MDS->OST in background (2824)
4901 - refuse to mount the same filesystem twice on same mountpoint (3394)
4902 - allow llmount to create routes for mounting behind routers (3320)
4903 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4904 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4905 - fix rare problem with rename on htree directories (3417)
4907 - allow default OST striping configuration per directory (1414)
4908 - fix compilation for qswnal for 2.6 kernels (3125)
4909 - increase maximum number of MDS request buffers for large systems
4910 - change liblustreapi to be useful for external progs like lfsck (3098)
4911 - increase local configuration timeout for slow disks (3353)
4912 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4914 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4917 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4918 - don't overwrite extent policy data in reply if lock was blocked (2901)
4919 - drop filter export grants atomically with removal from device (2663)
4920 - del obd_self_export from work_list in class_disconnect_exports (2908)
4921 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4922 - swab reply message in mdc_close, other PPC fixes (2464)
4923 - fix destroying of named logs (2325)
4924 - overwrite old logs when running lconf --write_conf (2264)
4925 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4926 - fix race in target_handle_connect (2898)
4927 - mds_reint_create() should take same inode create lock (2926)
4928 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4929 - don't close files for self_export to avoid uninitialized obd (2936)
4930 - allow MDS with the same name as client node (2939)
4931 - hold dentry reference for closed log files for unlink (2325)
4932 - reserve space for all logs during transactions (2059)
4933 - don't evict page beyond end of stripe extent (2925)
4934 - don't oops on a deleted current working directory (2399)
4935 - handle hard links to targets without a parent properly (2517)
4936 - don't dereference NULL lock when racing during eviction (2867)
4937 - don't grow lock extents when lots of conflicting locks (2919)
4939 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4942 - account for cache space usage on clients to avoid data loss (974)
4943 - lfsck support in lustre kernel code (2349)
4944 - reduce journal credits needed for BRW writes (2370)
4945 - orphan handling to avoid losing space on client/server crashes
4946 - ptlrpcd can be blocked, stopping ALL progress (2477)
4947 - use lock value blocks to assist in proper KMS, faster stat (1021)
4948 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4949 - recovery for initial connections (2355)
4950 - fixes for mds_cleanup_orphans (1934)
4951 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4952 - block all file creations until orphan recovery completes (1901)
4953 - client remove rq_connection from request struct (2423)
4954 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4955 - recovery timer race (2670)
4956 - mdc_close recovey bug (2532)
4957 - ptlrpc cleanup bug (2710)
4958 - mds timeout on local locks (2588)
4959 - namespace lock held during RPCs (2431)
4960 - handle interrupted sync write properly (2503)
4961 - don't try to handle a message that hasn't been replied to (2699)
4962 - client assert failure during cleanup after abort recovery (2701)
4963 - leak mdc device after failed mount (2712)
4964 - ptlrpc_check_set allows timedout requests to complete (2714)
4965 - wait for inflight reqs when ptlrpcd finishes (2710)
4966 - make sure unregistered services are removed from the srv_list
4967 - reset bulk XID's when resending them (caught by 1138 test)
4968 - unregister_bulk after timeout
4969 - fix lconf error (2694)
4970 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4971 - readahead locks pages, leaves pending causing memory pressure (2673)
4972 - increase OST request buffers to 4096 on large machines (2729)
4973 - fix up permission of existing directories in simple_mkdir (2661)
4974 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4975 - don't assign transno to errored transactions (2742)
4976 - don't delete objects on OST if given a bogus objid from MDS (2751)
4977 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4978 - if rq_no_resend, then timeout request after recovery (2432)
4979 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4980 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4981 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4982 - MDS recovery shouldn't delete objects that it has given out (2730)
4983 - if enqueue arrives after completion, don't clobber LVB (2819)
4984 - don't unlock pages twice when trigger_group_io returns error (2814)
4985 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4986 - don't write pages to disk if there was an error (1450)
4987 - don't ping imports that have recovery disabled (2676)
4988 - take buffered bytes into account when balancing socknal conn (2817)
4989 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4990 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4991 - remove little-endian swabbing of llog records (1987)
4992 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4993 - retry reposting request buffers if they fail (1191)
4994 - grow extent at grant time to avoid granting a revoked lock (2809)
4995 - lock revoke doesn't evict page if covered by a second lock (2765)
4996 - disable VM readahead to avoid reading outside lock extents (2805)
4998 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4999 - updated kernel patches for hp-2.4.20 kernel (2681)
5001 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5004 - fix truncated write corruption (2366)
5005 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5007 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5008 - ensure len(uuid) < 37 in lmc (1171)
5009 - fix ia64 OOPS in llog_test (2255)
5010 - zero end of page at obdfilter for partial page writes (2648)
5011 - don't leave stale dentries around after renames (bug 2428)
5012 - fix timeouts when evicting a client with a single lock held (2642)
5013 - set deadline for the initial HELLO message to drain (2634)
5014 - print out dotted-quad IP addresses in the socknal (2302)
5016 - additional debugging for MDS client eviction problem (2443)
5017 - fix mkfsoptions support for osts (2603, 2604)
5019 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5022 - add series for the vanilla 2.6.0 kernel
5023 - add series for the vanilla 2.4.24 kernel
5024 - add series for a cray x86/64 UL kernel drop
5025 - fix xattr patches for the vanilla 2.4.19 series
5027 - generate true UUIDs in lmc (1171)
5028 - have portals stack dumping break in UML (2466)
5029 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5030 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5031 - wake callback waiting threads on client eviction (2460)
5032 - Add --ptldebug and --subsystem to lmc (1719)
5033 - update assertion to allow safe interrupt allocation
5034 - set rq_no_resend for cancel requests (2432)
5035 - recalculate ptlrpcd timeout after resend (2494)
5036 - call vfs_rmdir when removing pending directories (2368)
5037 - fix renaming a file to itself (2429)
5038 - lmc creates a default one-stripe lov (2454)
5039 - expand procfs space to handle large clusters (2326)
5040 - increase UML stack to avoid overflow
5041 - update lconf's list of debug and subsystem masks
5042 - fix lfs find --obd (2510)
5043 - /proc tunable for disabling filter read caching (2591)
5044 - stop rpm packages from altering slapd.conf (2301)
5045 - disable nagle in the socknal under 0conf (2578)
5046 - choose mds inode size based on stripe count (2572)
5047 - fix kernel-source rpm problems (2516)
5049 - add --disable-doc to avoid pdf generation (2421)
5050 - update documentation, tests, type-os, comments
5051 - avoid format warnings on ia64
5052 - remove the TOE NAL
5053 - tiny code cleanups by removing unused fields
5055 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5058 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5059 - remove the most blatant lies from BUILDING file (2371)
5060 - change default debug level to reasonable production setting
5061 - reduce client side cache size to reduce cache flush time
5062 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5063 - make TCP zerocopy and pinger support enabled by default (2476)
5064 - sync writes completed after process exits caused crashes (2319)
5065 - maintain correct mount count on the MDS (2356)
5066 - backout 1557, because 2316 wasn't really fixed
5067 - better file I/O statistics gathering in /proc
5068 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5069 - another kernel patch to fix zero-copy TCP function export
5070 - don't take duplicate lock when processing re-sent getattr (2420)
5071 - lctl uses obd_self_export instead of creating new conn (2353)
5072 - MDS/OST recovery case which requires object creation asserted (2425)
5073 - move lfs from /usr/sbin to /usr/bin in packages
5074 - fix race between mds_client_add and mds_client_free (2417)
5075 - use kmalloc instead of slabs in portals (2430)
5076 - don't create duplicate records when a failover MDS is present (2442)
5077 - remove unnecessary mount age check (2332)
5078 - don't remove directory inodes from locks prematurely (2451)
5079 - don't break if MDS service name is the same as hostname (2103)
5080 - fix races in client write RPC generation when cache full (2482)
5082 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5085 - remove now-unused request->rq_obd (278)
5086 - if an allocation fails, print out how much memory we've used (1933)
5087 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5088 - add the "configurable stack size" patch to most series files (1256)
5089 - ability to write large log records, for 100+ OST configs (2306)
5090 - fix NULL deref when filter_prep fails (2314)
5091 - fix operator precedence error in filter_sync
5092 - dynamic allocation of socknal TX descriptors (2315)
5093 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5094 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5095 - ensure that CWARN messages in Portals always get to the syslog
5096 - __init/__exit are not for prototype decls (ldlm_init/exit)
5097 - x86-64 compile warning fixes
5098 - fix gateway LMC keyword conflict (2318)
5099 - fix MDS lock inversions in getattr/reint paths (1844)
5100 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5101 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5102 - DLM race condition prevented some lock evictions (2328)
5103 - ENOMEM detection and retry on socknal sends (2230)
5104 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5105 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5106 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5107 - better error messages when a client is rejected during recovery (1505)
5108 - avoid cancelling locks which were never granted, after failure (2330)
5109 - fix i_sem/journal inversion in mds_client_add (2333)
5110 - fix truncate/getattr lock cycle deadlock (2334)
5111 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5112 - fix two rare exit paths which could leak an l_lock() ref (2321)
5113 - fencepost error in MDS/OST orphan recovery (2226)
5114 - make log record alignment 8 bytes (1988)
5115 - lstripe now fails when requested offset > ost_count (2237)
5116 - ensure that all kernel series have a complete list.h (1607)
5117 - fix crashes in special-file operations (2316)
5118 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5120 - allow configurable automake binary, for testing new versions
5121 - small update to the lfs documentation
5123 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5125 * fix negative export reference count in fsfilt_sync (2312)
5127 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5128 * release candidate 0.9.1
5130 - orphans are moved into the PENDING directory for possible recovery
5131 - replayed opens now open by fid for orphan/rename safety (1042)
5132 - last close of an orphan inode generates a transno (683)
5133 - chdir() and mount() now pin the directory entry (1020)
5134 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5135 - discard very old requests without processing them (1502)
5136 - remove some common, well-understood CERRORs (1505)
5137 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5138 - clear "grant" flags in OST replies until OST grant code lands (1644)
5139 - fix read performance by not clobbering i_blksize on client (1598)
5140 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5141 - make LRU size a /proc tunable, clears locks when reduced (707)
5142 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5143 - server locks take references on exports now (1558)
5144 - build fixes for 2.4.20-rh trees (1663)
5145 - return an error from lov_create if all OSCs are inactive (1751)
5146 - fix import levels when a reconnect happens without a timeout (1597)
5147 - exit early from mds_open if we get a lookup error (1749)
5148 - partial page read at EOF wouldn't wait for disk before sending (1642)
5149 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5150 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5151 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5152 - write/truncate lock inversion (1639)
5153 - Don't auto-load obdclass, portals modules during cleanup (1495)
5154 - fix timestamps from jumping to "now" (1763)
5155 - extra journal assertions (1648)
5156 - add an extra multiunlink test (1771)
5157 - fix read_record/write_record API (1776)
5158 - fix leak of offset_extent, possible incorrect i_size later (1772)
5159 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5160 - Debugging for the kqswnal_get_idle_tx problems (1820)
5161 - Allow recovery to be attempted multiple times (1536)
5162 - Write out MDS last_rcvd file after it is first created (1600)
5163 - Fix tx_descriptor leak in failed transmit situations (1827)
5164 - ext3 journaling fixes for assertion failure after IO error (1871)
5165 - class_export_put() on freed export after completion AST error (1896)
5166 - Fix revalidate looping in VFS (1322)
5167 - Don't access a freed export during MDS_REINT timeout (1521)
5168 - Add open-unlink recovery support on the MDS (1673,1764)
5169 - Return an error if no MDS data was read from last_rcvd (1946)
5170 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5171 - Don't setup a disk device that doesn't match exported UUID (317)
5172 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5173 - avoid committing NULL handle in force close
5174 - local.sh is now a one-stripe LOV configuration
5175 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5176 - fix ext3 htree duplicate directory entry corruption (1516)
5177 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5178 - update to new LOV EA format (2097)
5179 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5181 2003-06-15 Phil Schwan <phil@clusterfs.com>
5184 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5185 - per-import/export recovery handling (958, 931, 959)
5186 - multiple last-rcvd slots, for serving multiple FSes (949)
5187 - connections are again shared between multiple imp/exports (963, 964)
5188 - "umount -f" would hang if any requests needed to be sent (393, 978)
5189 - avoid pinning large req buffer by copying for queued messages (989)
5190 - add "uuid" to "lctl device" command to help upcalls (991)
5191 - "open" RPCs with transnos would confuse recovery counters (1037)
5192 - do proper endian conversion of all wire messages (288, 340, 891)
5193 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5194 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5195 - fix ptlrpc_connection leak in target_handle_connect (1174)
5196 - fix import refcounting bug in OST and MDS cleanup (1134)
5197 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5198 - fix very unlikely obd_types race condition (501)
5199 - remove osc_open hack for echo_client (1187)
5200 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5201 - a failure in read_inode2 leads to deadlock (1139)
5202 - cancel ack-locks as soon as transaction is committed (1072)
5203 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5204 - make sure to commitrw after any preprw to avoid deadlock (1162)
5205 - failing to execute a file in a lustre FS would lock inode (1203)
5206 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5207 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5208 - obd_uuid2tgt would walk off the end of the list (1255)
5209 - on IA64 the osc would give portals incorrect bulk size (1258)
5210 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5211 - fix lock inversion caused by new llite matching code (1282)
5212 - limit the number of dirty pages on a client to 10MB (1286)
5213 - timed out locks were not being corrected cancelled (1289)
5214 - fix O_DIRECT above 4GB on IA-32 (1292)
5215 * major user-visible changes
5216 - fail out/fail over policy now controlled by the upcall (993)
5218 - add OBD_PING to check server availability and failure (954)
5219 - lustre messages are now sent in sending host order (288, 340, 891)
5220 - add eadatalen to MDS getattr reply (340)
5221 - OST read replies may contain second buffer, with per-page status (593)
5223 2003-03-11 Phil Schwan <phil@clusterfs.com>
5226 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5227 - failed open()s could cause deadlock; fixed (867, 869)
5228 - stop cancelling OST locks when files are closed (481)
5229 - overlapping XID spaces caused network corruption (851, 853)
5230 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5231 - setattr_raw added, to do single-RPC, server-side setattrs
5232 - lmc/lconf syntax change for OST UUIDs
5233 - fix crashy race condition between ptlrpc_free_req and osc_close
5234 - don't use request in mdc_enqueue if we hit a timeout (889)
5235 - don't set the inode i_size for regular files from the MDS (896)
5236 - handle out of order completion AST (842)
5237 - don't LBUG if a lock request times out after receiving AST (913)
5238 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5239 - if a bad lock AST arrives, send an error instead of dropping entirely
5240 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5241 - fix leak in bulk IO when only partially completed (899, 900, 926)
5242 - fix O_DIRECT for ia64 (55)
5243 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5244 - C-z after timeout could hang a process forever; fixed (977)
5246 - client-side I/O cache (678, 924, 929, 941, 970)
5248 - READPAGE and SETATTRs which don't take server-side locks get
5251 2003-02-11 Phil Schwan <phil@clusterfs.com>
5254 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5255 - Fix fsfilt_extN_readpage() to read a full page of directory
5256 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5257 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5258 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5259 - Most intent operations were removed, in favour of a new RPC mode
5260 that does a single RPC to the server and bypasses most of the VFS
5261 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5262 - Aggressively cancel local locks on DLM servers
5263 - mds_reint_unlink sends EA to the client if it's the last nlink.
5264 client uses that EA to unlink OST objects.
5265 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5266 - recursive symlinks were fixed (439)
5267 - fixed NULL deref in DEBUG_REQ
5268 - filter_update_lastobjid no longer calls sync, which annoyed extN
5269 - fixed multi-client small-writes to a single file problem (445)
5270 - fixed mtime updates during file writes (607)
5271 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5272 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5273 - fixed Linux OST size reporting problem (444, 656)
5274 - OST now updates object mtime with writes or setattr (607, 619)
5275 - client verifies file size before zeroing page past EOF (445)
5276 - OST now writes last allocated objid to disk with allocation (108)
5277 - LOV on echo now works (409)
5279 - mds_reint_unlink sends a new buffer, with the EA included. this
5280 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5281 set if a regular file was being unlinked, and it was the last link
5282 - use PtlGet from the target for bulk writes (315)
5283 - OST now updates object mtime with writes or setattr (607, 619)
5284 - LDLM now has a grant-time callback to revalidate locked items, if
5286 - Many MDS operations were reorganized to combat race conditions
5288 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5289 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5291 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5294 - Fully reactivate OST imports after reconnection (512, others)
5295 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5296 - More graceful error handling for truncating on dead OST (515)
5297 - Don't error out unless we're actually accessing dead stripes (474)
5298 - Fix garbage sizes when stripes are missing (410)
5299 - LRU counters were broken, causing constant lock purge (433, 432)
5300 - garbage on read from stripes with failed OSTs (441)
5301 - mark OSCs as active before reconnecting during recovery (438)
5302 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5303 - lfind did not preserve OST order in output (443)
5304 - symlinks cause hung clients, incorrect data (439)
5305 - stop dereferencing request after dropping refcount (457)
5306 - don't LASSERT(spin_is_locked) on non-SMP (455)
5307 - fixes for many rename() bugs
5308 - fstat didn't correctly synchronize attributes (399)
5309 - server must handle lock cancellation during blocking AST prep (487)
5310 - bulk descriptors were free()d too soon (511)
5311 - fix paths in lconf, which would load incorrect modules (451, 507)
5312 - fix confusing lconf 'host not found' error message (386)
5313 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5314 - fix race condition in mdc_blocking_ast() for inode access (526)
5315 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5316 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5317 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5318 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5319 - fix lstripe in several ways (kernel side) (527)
5320 - fix request leak in ldlm_cli_enqueue (262)
5321 - incorrect OSC was marked inactive after OST failure
5322 - call mds_fs_cleanup before unmounting filesystem (524)
5323 - fix races between taking ns_lock and ldlm_lock_change_resource
5324 - fix races updating LOV export open file list
5325 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5326 - fix recovery NULL deref in ldlm_cli_cancel_unused
5327 - fix some DLM races by using new hash table for lock handles (419)
5328 - permit the client to specify desired inodes, at replay
5329 - duplicate requests when we queue them for replay reintegration
5330 - fix last_rcvd offset calculation
5331 - sync after each recovered transaction, so we always make progress
5332 - never, not always, ERESTART requests without transnos
5333 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5335 - skip replay if the MDS says that the client is already connected
5336 - don't check for a recovery-enabled export to match lctl's UUID
5337 - don't INC_USE_COUNT for phantom exports
5338 - don't crash when cleaning up phantom exports (567)
5339 - don't double-finish or set replay data for errored mdc_open requests
5340 - abort requests when they time out, so we don't get old replies
5341 - send/receive replies for AST messages again
5342 - if the client says that it doesn't have the lock, cancel it on the
5344 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5345 mark it as destroyed, it will all work out when decref is called
5346 - fix module use counts (22, 581)
5348 - ASTs now expect a reply (server cancels lock on error reply)
5350 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5353 - fix many simultaneous client startup (392)
5354 - fix dentry->d_it clobbering
5355 - credentials weren't being shipped for readdir/getattr operations
5356 - remove invalid assertions triggered during some concurrent MD
5358 - proper Lustre versions added (336, 389)
5359 - fix memory leak for create error case (398)
5360 - fix LOV locking bug that would get cli/srv out of sync
5361 - fix echo client over LOV (409)
5362 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5363 - fix double-O_EXCL intent crash (424)
5364 - avoid sending multiple lock CANCELs (352)
5366 - MDS can do multi-client recovery (modulo bugs in new code)
5368 - many updates, edits, cleanups
5370 2002-11-18 Phil Schwan <phil@clusterfs.com>
5373 - fix null d_it dereference (346)
5374 - fix full OST/dbench hang (333)
5375 - fix permission problem with file removal (286)
5376 - fix removal of OSCs from LOV when they fail
5377 - fix NULL deref during bulk timeout (214)
5378 - fix problems related to multiple filesystems on one MDS (241)
5379 - fixed serious subtle metadata locking bugs
5380 - free locks on clients when inodes are removed due to memory
5382 - fix inode pointer in lock data (285)
5383 - partial support for multiple MDS on a single host (241)
5384 - data locks weren't cancelled at clear_inode time (290, 311)
5385 - intent locks could lead to unbounded lock growth (205)
5386 - added a maximum lock count, an LRU list, and a flusher
5387 - fix multiple rename (365)
5388 - properly abstracted the echo client
5389 - OSC locked 1 byte too many; fixed
5390 - rewrote brw callback code:
5391 - fixed recovery bugs related to LOVs (306)
5392 - fixed too-many-pages-in-one-write crash (191)
5393 - fixed (again) crash in sync_io_timeout (214)
5394 - probably fixed callback-related race (385)
5396 - Add capability to MDS protocol
5397 - LDLM cancellations and callbacks on different portals
5399 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5402 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5403 - defer open object destruction to close time (601981, 138)
5404 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5405 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5406 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5407 - don't call obd_setattr in ll_file_release for destroyed objects
5408 * protocol change to lustre_msg: move |version| and add |flags|
5409 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5410 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5411 * added replay of create, unlink, link and rename operations during
5412 MDS failover; recovery should be much more robust now
5413 * remove failed OSCs from LOVs (only lov_create uses this so far)
5414 * the lustre-HOWTO was brought (more) up to date (582544)
5416 2002-10-23 Phil Schwan <phil@clusterfs.com>
5419 - in-use dentries weren't being reused properly (617851)
5420 - prevent multiple LDLM setup (599178)
5421 - fix LOV size calculations for truncate (617853)
5422 - fix client handling of MDS intent errors (POSIX)
5423 - fix permission bug in lovstripe.c test (624321)
5424 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5425 - truncate past end of file could corrupt data
5426 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5427 - a race in recovery could return ETIMEDOUT to apps (623947)
5428 - building outside the source directory was fixed
5429 * the lustre-HOWTO was brought (more) up to date (582544)
5430 * major progress was made on recovery functionality
5432 2002-10-10 Phil Schwan <phil@clusterfs.com>
5435 - recovery deadlock fix
5436 - rm -rf causes LBUG fix (617817)
5437 - file open by multiple tasks fix (618962)
5438 - directory permissions bugs (602707 and 620007)
5439 - journal_stop fixed with locking (611313)
5440 - O_APPEND failures resolved (618273, perhaps 614459)
5441 - lconf PATH fix (619770)
5442 - IA64 build fix (621450)
5443 - RPC buffer sizes scale with amount of memory
5445 2002-10-01 Phil Schwan <phil@clusterfs.com>
5448 - locks would be cancelled without throwing away data pages,
5449 resulting in inconsistent data (605627)
5450 - inode attributes were not always being refreshed (605627, 612449)
5451 - lconf now continues to cleanup after lctl reports an error
5452 - MDS now enforces user permissions (602707)
5453 - lprocfs cleanup fixed, but not yet enabled (614157)
5454 - fixed infinite server hang, should a client not respond to an AST
5455 - avoid going into recovery if user calls readlink() with a buffer
5456 that's too small (613941)
5457 - AST RPCs no longer require replies (614867) -- this may be changed
5458 - don't crash server if client sends an IOV that's too big (611336)
5459 - fixed lock conversion deadlock (611892)
5460 - fixed the following of symlinks (614622)
5461 * recovery: the server can remove locks from a client that dies, other
5462 clients can make progress
5463 * more extN patch fixes
5464 * compile-time configurable ptlrpc buffer allocations
5466 - collaborative read cache document
5467 - Lustre Lite Performance CDR document-in-progress
5469 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5472 - fix typo in patch-2.4.18
5474 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5477 - clear ptlrpc request each time in handle_incoming_request()
5478 - unlink of files now destroys the object on the OST
5480 2002-09-19 Peter Braam <braam@clusterfs.com>
5482 * add hard link support
5483 * change obdfile creation method
5484 * kernel patch changed
5486 2002-09-19 Peter Braam <braam@clusterfs.com>
5489 - stack overflow bug in extN fixed
5491 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5493 * documentation updates
5494 - add man pages for config tools
5495 - update tests/README to describe testing with new config tools
5496 - finish metadata API descriptions
5497 * bug fixes and cleanups
5498 - statfs workaround for 16TB limit
5499 - LOV stripe allocation improved, can stripe on subset of OSTs
5500 - LOV file size/IO offset was wrong for files > 4GB in size
5501 - object EA data was being dropped, caused files to be unreadable
5502 - memory overflow with non-LOV OST caused memory corruption
5503 - fixed regression tests to work with new config tools, obdfilter
5504 - fixed bug when directory size became larger than 1 block
5505 - fixed bug (for single client case) when PWD was deleted
5506 - invalidate local directory pages when doing intent-based ops
5507 - avoid LDLM oops when lock callback contained bad data
5509 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5511 * documentation updates
5512 * bug fixes and cleanups
5513 - configuration tools
5516 - 64-bit compile warnings
5517 - 64-bit internal statfs data
5519 * test_brw on persistent OST devices
5521 * lprocfs (disabled)
5523 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5525 * documentation updates
5526 * bug fixes and cleanups
5527 * configuration tools
5529 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5532 * many small fixes to 0_5_4
5533 * io/network handling
5534 * thinkos in MDS operations
5536 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5539 * crucial basic fixes to 0.5.3
5540 * IOR, Iozone work over Elan
5543 2002-08-07 Phil Schwan <phil@clusterfs.com>
5544 * version 0_5_3, our first alpha
5545 * we use the new Portals iovs
5546 * documentation updates
5547 * bug fixes and cleanups
5548 * small changes in the DLM wire protocol
5550 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5551 * version 0_5_1 with some initial stability,
5552 * locking on MD and file I/O.
5553 * documentation updates
5554 * several bug fixes since 0.5.0
5555 * small changes in wire protocol
5557 2002-07-18 Phil Schwan <phil@clusterfs.com>
5559 * delivered as Lustre Light Alpha
5560 * fixed a crash after handling invalid MDS requests
5561 * fixed directory pages for architectures with non-4k pages sizes
5563 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5564 * release version v0_4_4
5565 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5567 * Fixes a number of interruption problems with OST operations.
5568 * Update documentation for portals header changes
5569 * Move all wire protocol structs/defines to lustre_idl.h
5570 * Fixes symlink length bug.
5571 * Add tcpdump to repository.
5573 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5574 * release version v0_4_3
5575 * Fixes statfs for inodes on extN.
5576 * Fixes bug in runtests which would delete /etc/hosts.
5577 * Use 64-bit object IDs wherever possible (not into VFS though)
5578 Remove ost_get_info, which is unused by lustre, and out of date.
5580 2002-07-03 Peter Braam <braam@clusterfs.com>
5581 * release version v0_4_2 Fixes a lookup error (type not passed)
5582 * move forward to head of Portals
5583 * move forward to latest Lustre kernel
5585 2002-06-25 Peter Braam <braam@clusterfs.com>
5586 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob