1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.27 (SLES 10),
5 2.6.18-92.1.10.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.40.11-sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
14 * File join has been disabled in this release, refer to Bugzilla 16929.
16 Severity : enhancement
18 Description: Commit on sharing. Eliminate inter-client dependencies between
19 uncommitted transactions by doing transaction commits.
20 Thereby clients may recovery independently.
23 Frequency : Create a symlink file with a very long name
25 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
26 Details : If there is no extra space in the request for early cancels,
27 ldlm_req_handles_avail() returns 0 instead of a negative value.
29 Severity : enhancement
31 Description: Add /proc entry for import status
32 Details : The mdc, osc, and mgc import directories now have
33 an import directory that contains useful import data for debugging
36 Severity : enhancement
38 Description: Re-disable certain /proc logging
39 Details : Enable and disable client's offset_stats, extents_stats and
40 extents_stats_per_process stats logging on the fly.
43 Frequency : Only on FC kernels 2.6.22+
45 Description: oops in statahead
46 Details : Do not drop reference count for the dentry from VFS when lookup,
47 VFS will do that by itself.
49 Severity : enhancement
51 Description: Generic /proc file permissions
52 Details : Set /Proc file permissions in a more generic way to enable non-
53 root users operate on some /proc files.
57 Description: Hitting mdc_commit_close() ASSERTION
58 Details : Properly handle request reference release in
59 ll_release_openhandle().
63 Frequency : only patchless client
64 Description: add workaround for race between add/remove dentry from hash
66 Severity : enhancement
68 Description: Allow OST glimpses to return PW locks
72 Description: LBUG when llog conf file is full
73 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
77 Description: Prevent import from entering FULL state when server in recovery
81 Description: service mount cannot take device name with ":"
82 Details : Only when device name contains ":/" will mount treat it as
88 Description: replace ptlrpcd with the statahead thread to interpret the async
89 statahead RPC callback
93 Frequency : on recovery
94 Description: I/O failures after umount during fail back
95 Details : if client reconnected to restarted server we need join to recovery
96 instead of find server handler is changed and process self
97 eviction with cancel all locks.
99 Severity : enhancement
101 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
103 Severity : enhancement
105 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
107 Severity : enhancement
109 Description: Upcall on Lustre log has been dumped
110 Details : Allow for a user mode script to be called once a Lustre log has
111 been dumped. It passes the filename of the dumped log to the
112 script, the location of the script can be specified via
113 /proc/sys/lnet/debug_log_upcall.
118 Description: avoid idr_remove called for id which is not allocated.
119 Details : Move assigment s_dev for clustered nfs to end of initialization,
120 to avoid problem with error handling.
125 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
126 Details : When connection is reused this not moved from CONN_UNUSED_HASH
127 into CONN_USED_HASH and this prodice warning when put connection
128 again in unused hash.
130 Severity : enhancement
132 Description: File striping can now be set to use an arbitrary pool of OSTs.
134 Severity : enhancement
136 Description: Export bytes_read/bytes_write count on OSC/OST.
140 Description: Early reply size mismatch, MGC loses connection
141 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
142 the connect flags are properly negotiated.
146 Description: Properly propagate oinfo flags from lov to osc for statfs
147 Details : restore missing copy oi_flags to lov requests.
149 Severity : enhancement
151 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
152 : l_getgroups(8), lst(8), routerstat(8)
153 Details : included man pages for llobdstat(8), llstat(8),
154 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
156 Severity : enhancement
158 Description: configure's --enable-quota should check the
159 : kernel .config for CONFIG_QUOTA
160 Details : configure is terminated if --enable-quota is passed but
161 : no quota support is in kernel
165 Description: Remove portals compatibility
166 Details : Remove portals compatibility, not interoperable with releases
171 Description: Resolve device initialization race
172 Details : Prevent proc handler from accessing devices added to the
173 obd_devs array but yet be intialized.
175 Severity : enhancement
177 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
179 Severity : enhancement
181 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
186 Description: Using wrong pointer in osc_brw_prep_request
187 Details : Access to array[-1] can produce panic if kernel compiled with
188 CONFIG_PAGE_ALLOC enabled
192 Description: Client runs out of low memory
193 Details : Consider only lowmem when counting initial number of llap pages
197 Description: *optional* service tags registration
198 Details : if the "service tags" package is installed on a Lustre node
199 When the filesystem is mounted, a local-node service tag will
200 be created. See http://inventory.sun.com/ for more information
201 about the Service Tags asset management system.
205 Description: Kernel BUG tries to release flock
206 Details : Lustre does not destroy flock lock before last reference goes
207 away. So always drop flock locks when client is evicted and
208 perform unlock regardless of successfulness of speaking to MDS.
212 Description: add refcount protection for osc callbacks, avoid panic on shutdown
216 Description: sanity test 65a fails if stripecount of -1 is set
217 Details : handle -1 striping on filesystem in ll_dirstripe_verify
222 Description: ASSERTION(CheckWriteback(page,cmd)) failed
223 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
226 Severity : enhancement
228 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
232 Description: do not process already freed flock
233 Details : flock can possibly be freed by another thread before it reaches
234 to ldlm_flock_completion_ast.
238 Description: LBUG during stress test
239 Details : Need properly lock accesses the flock deadlock detection list.
243 Description: oops in page fault handler
244 Details : kernel page fault handler can return two special 'pages' in error
245 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
249 Description: timeout with invalidate import.
250 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
251 should be handled by ptlrpcd. This produce long age waiting and
252 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
254 Severity : enhancement
256 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
260 Description: enable MGS and MDT services start separately
261 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
262 a co-located MGS without starting the MGS, which is a complement
263 to 'nosvc' mount option.
267 Frequency : after recovery
268 Description: precreate to many object's after del orphan.
269 Details : del orphan st in oscc last_id == next_id and this triger growing
270 count of precreated objects. Set flag LOW to skip increase count
271 of precreated objects.
275 Frequency : rare, on clear nid stats
276 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
277 Details : when clean nid stats sometimes try destroy live entry,
278 and this produce panic in free.
282 Description: Stack overflow during MDS log replay
283 ease stack pressure by using a thread dealing llog_process.
287 Description: wait until IO finished before start new when do lock cancel.
288 Details : VM protocol want old IO finished before start new, in this case
289 need wait until PG_writeback is cleared until check dirty flag and
290 call writepages in lock cancel callback.
292 Severity : enhancement
294 Description: using special macro for print time and cleanup in includes.
298 Description: mds_mfd_close() ASSERTION(rc == 0)
299 Details : In mds_mfd_close(), we need protect inode's writecount change
300 within its orphan write semaphore to prevent possible races.
304 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
308 Frequency : rare, on shutdown ost
309 Description: don't hit live lock with umount ost.
310 Details : shrink_dcache_parent can be in long loop with destroy dentries,
311 use shrink_dcache_sb instead.
315 Description: don't panic with use echo client
316 Details : echo client pass NULL as client nid pointer and this produce null
321 Description: fix build on ppc32
322 Details : compile code with -m64 flag produce wrong object file for ppc32.
326 Description: add message levels for liblustreapi
330 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
331 Details : Move the 'good_osts' check before the 'total_bavail' check. This
332 will result in an -EAGAIN and in the exit call path we call
333 alloc_rr() which will with increasing aggressiveness attempt to
334 aquire precreated objects on the minimum number of required OSCs.
338 Description: Use old size assignment to avoid deadlock
339 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
340 the scheduling while holding a spinlock. We do not need locking
341 for size in ll_update_inode() because size is only updated from
342 the MDS for directories or files without objects, so there is no
343 other place to do the update, and concurrent access to such inodes
344 are protected by the inode lock.
348 Description: resolve "_IOWR redefined" build error on SLES10
352 Description: dump the memory debugging after all modules are unloaded to
353 suppress false negative in conf_sanity test 39
355 Severity : enhancement
357 Description: build kernel-ib packages for OFED 1.3 in our release cycle
362 Description: fix SLES kernel versioning
363 Details : the kernel version for our SLES 10 kernel did not include a "-"
364 before the "smp" at the end. while this was not a problem in
365 general, it did mean that software trying to use the kernel
366 version to try to detect a vendor specific kernel would fail.
367 this was most evident by the OFED build scripts.
371 Description: Don't update lov_desc members until making sure they are valid
372 Details : When updating lov_desc members via proc fs, need fix their
373 validities before doing the real update.
377 Description: don't put request into delay list while invalidate in flight.
378 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
379 import in flight. this produce timeout for invalidate and sometimes
380 can cause stale data.
384 Frequency : on ppc only
385 Description: not convert ost objects for directory because it's not exist.
386 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
390 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
392 Severity : enhancement
394 Description: Update to support 2.6.22.14 vanilla kernel.
398 Frequency : rare, on recovery
399 Description: read procfs can produce deadlock in some situation
400 Details : Holding lprocfs lock which send rpc can produce block for destroy
401 obd objects and this also block reconnect with -EALREADY. This
402 isn't fix all lprocfs bugs - but make it rare.
404 Severity : enhancement
406 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
409 Frequency : frequent on X2 node
411 Description: mdc_set_open_replay_data LBUG
412 Details : Set replay data for requests that are eligible for replay.
416 Description: lustre_mgs: operation 101 on unconnected MGS
417 Details : When MGC is disconnected from MGS long enough, MGS will evict the
418 MGC, and late on MGC cannot successfully connect to MGS and a lot
419 of the error messages complaining that MGS is not connected.
423 Frequency : on network error
424 Description: panic with double free request if network error
425 Details : mdc_finish_enqueue is finish request if any network error ocuring,
426 but it's true only for synchronus enqueue, for async enqueue
427 (via ptlrpcd) this incorrect and ptlrpcd want finish request
430 Severity : enhancement
432 Description: client-side metadata stat-ahead during readdir(directory readahead)
433 Details : perform client-side metadata stat-ahead when the client detects
434 readdir and sequential stat of dir entries therein
437 Frequency : on start mds
439 Description: Implement get_info(last_id) in obdfilter.
442 Frequency : occasional
444 Description: Correctly check stale fid, not start epoch if ost not support SOM
445 Details : open with flag O_CREATE need set old fid in op_fid3 because
446 op_fid2 was overwritten with new generated fid, but mds can answer
447 with one of these two fids and both is not stale. Setattr
448 incorrectly started an epoch and assume will be called
449 done_writing, but without SOM done_writing ever being called.
452 Frequency : rare, depends on device drivers and load
454 Description: MDS or OSS nodes crash due to stack overflow
455 Details : Code changes in 1.8.0 increased the stack usage of some functions.
456 In some cases, in conjunction with device drivers that use a lot
457 of stack the MDS (or possibly OSS) service threads could overflow
458 the stack. One change which was identified to consume additional
459 stack has been reworked to avoid the extra stack usage.
462 Frequency : occasional
464 Description: Do not fail import if osc_interpret_create gets -EAGAIN
465 Details : If osc_interpret_create got -EAGAIN it immediately exits and
466 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
467 oscc_has_objects and see OSC has no objests and call
468 oscc_internal_create to resend create request.
470 Severity : enhancement
472 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
474 Severity : enhancement
476 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
479 Frequency : very rare
481 Description: Fix replay if there is an un-replied request and open
482 Details : In some cases, older replay request will revert the
483 mcd->mcd_last_xid on MDS which is used to record the client's
486 Severity : enhancement
488 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
490 Severity : enhancement
492 Description: Add rhel5 support to HEAD.
494 Serverity : enhancement
496 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
501 Description: Don't allow skipping OSTs if index has been specified.
502 Details : Don't allow skipping OSTs if index has been specified, make
503 locking in internal create lots better.
507 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
508 Details : access to bitfield in structure is always rounded to long
509 and this produce problem with not atomic change any bit.
513 Description: Lustre make rpms failed.
514 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
515 builds Lustre rpms from tarball.
518 Frequency : rare on shutdown ost
520 Description: If llog cancel was not send before clean_exports phase, this can
521 produce deadlock in llog code.
522 Details : If llog thread has last reference to obd and call class_import_put
523 this produce deadlock because llog_cleanup_commit_master wait when
524 last llog_commit_thread exited, but this never success because was
525 called from llog_commit_thread.
529 Description: allow userland application know is lost one of stripes.
530 Details : fill lvb_blocks with error code on ost and return it to
531 application if error flag found.
535 Description: NULL lov_tgts causing MDS oops
536 Details : more safe checks for NULL lov_tgts for avoid oops.
538 Severity : enhancement
540 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
544 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
545 Details : If an OST is down the MDS will hang indefinitely in
546 obd_statfs_rqset() waiting for the statfs data. While for
547 MDS QOS usage of statfs, it should not stuck in waiting.
549 Severity : enhancement
551 Description: remote_acl support
552 Details : Support ACL-based permission check for remote user.
553 Support setfacl/getfacl for remote user with the utils
554 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
555 the system "{s,g}etfacl" utils.
557 Severity : enhancement
559 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
561 Severity : enhancement
563 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
567 Description: make mgs_setparam() handle fsname containing dash
568 Details : fsname containing a dash does not work with lctl conf_param
570 Severity : enhancement
572 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
574 Severity : enhancement
576 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
580 Frequency : rare, at shutdown
581 Description: access already free / zero obd_namespace.
582 Details : if client_disconnect_export was called without force flag set,
583 and exist connect request in flight, this can produce access to
584 NULL pointer (or already free pointer) when connect_interpret
585 store ocd flags in obd_namespace.
589 Frequency : only at startup
590 Description: not alloc memory with spinlock held.
591 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
592 if any spinlock held.
594 Severity : enhancement
596 Description: make lustre randomly fail allocating memory
597 Details : Make lustre randomly failed allocating memory for testing purpose.
599 Severity : enhancement
601 Description: lost problems with lov objid file
602 Details : Fixes some scability and access to not inited memory problems
603 in work with lov objdid file.
608 Description: lfs find does not continue on file error
609 Details : Continue other files processing when a file/dir is absent.
613 Description: Inconsistent usage of lustre_pack_reply()
614 Details : Standardize the usage of lustre_pack_reply() such that it
615 always generate a CERROR on failure.
618 Frequency : occasional
620 Description: MDS hang or stay in waiting lock
621 Details : If client receive lock with CBPENDING flag ldlm need send lock
622 cancel as separate rpc, to avoid situation when cancel request
623 can't processed due all i/o threads stay in wait lock.
627 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
631 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
635 Description: Update to SLES9 kernel-2.6.5-7.287.3.
639 Description: Update to RHEL4 latest kernel.
641 Severity : enhancement
643 Description: Build SLES10 patchless client fails
644 Details : The configure was broken by run ./configure with
645 --with-linux-obj=.... argument for patchless client. When the
646 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
647 can't search header adequately. Use absolute path such as
648 -I($LINUX)/include instead.
652 Description: interrupt oig_wait produce painc on resend.
653 Details : brw_redo_request can be used for resend requests from ptlrpcd and
654 private set, and this produce situation when rq_ptlrpcd_data not
655 copyed to new allocated request and triggered LBUG on assert
656 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
657 ptlrpcd set if request is changed and can be safety changed to use
662 Description: LASSERT_{REQ,REP}SWAB macros are buggy
663 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
664 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
665 these macros and replace them with their difinitions instead.
669 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
670 Details : Update which_patch & target file for SLES10 latest kernel.
673 Description: Remove client patches from SLES 10 kernel.
674 Details : This causes SLES 10 clients to behave as patchless clients
675 even on a Lustre-patched (server) kernel.
677 Severity : enhancement
679 Description: self-adjustable client's lru lists
680 Details : use adaptive algorithm for managing client cached locks lru
681 lists according to current server load, other client's work
682 pattern, memory activities, etc. Both, server and client
683 side namespaces provide number of proc tunables for controlling
686 Severity : enhancement
688 Description: light-weight GSS support
689 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
690 and reduce performance overhead.
692 Severity : enhancement
694 Description: Linux keyring support
695 Details : Support using service of Linux keyring for Lustre GSS internal
696 context refresh/cache mechanism.
700 Description: Fix errors in lfs documentation
701 Details : Fixes man pages
705 Description: don't use GFP_* in generic Lustre code.
706 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
710 Description: obdclass is limited by single OBD_ALLOC(idarray)
711 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
715 Description: warnings with build patchless client with vanila 2.6.19 and up
716 Details : change old ctl_table style and replace ctl_table/ctl_table_header
717 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
721 Frequency : Only for SLES
722 Description: sanity_quota fail test_1
723 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
724 file, which makes TSTID[2] inunique.
728 Frequency : Only for SLES9
729 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
730 Details : Update target/ChangeLog/which_patch .
734 Frequency : Only for test_10 in sanity-quota.sh
735 Description: a bug in quota test code
736 Details : checking a test flag in a wrong place
740 Frequency : Only for test_13 in sanity-quota.sh
741 Description: a bug in quota test script
742 Details : shouldn't check overfull allocated quota
746 Description: jbd statistics
747 Details : Port older jbd statistics patch for sles10
749 Severity : enhancement
751 Description: Add printing inode info into message about error in writepage.
756 Description: reply_lock_interpret crash due to race with it and lock cancel.
757 Details : Do not replay locks that are being cancelled. Do not reference
758 locks by their address during replay, just by their handle.
762 Frequency : When flocks are used.
763 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
764 Details : Flock locks might destroy just granted lock if it could be merged
765 with another existing flock, this is done in completion handler,
766 so teach ldlm_cli_enquque_fini that this is a valid case for
772 Description: Oops in read and write path when failing to allocate lock.
773 Details : Check if lock allocation failed and return error back.
777 Description: lstripe command fails for valid OST index
778 Details : The stripe offset is compared to lov->desc.ld_tgt_count
779 instead of lov->desc.ld_active_tgt_count.
783 Description: sanity.sh failed test 103
784 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
789 Description: df doesn't work properly if diskfs blocksize != 4K
790 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
794 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
795 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
797 Severity : enhancement
799 Description: Adaptive timeouts
800 Details : RPC timeouts adapt to changing server load and network
801 conditions to reduce resend attempts and improve recovery time.
805 Description: llapi_file_create() does not allow some changes
806 Details : add llapi_file_open() that allows specifying the mode and
807 open flags, and also returns an open file handle.
811 Description: merge and cleanup kernel patches.
812 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
816 Description: Add journal checksum support.(Kernel part)
817 Details : The journal checksum feature adds two new flags i.e
818 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
819 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
820 indicates that the commit block contains the checksum for
821 the blocks described by the descriptor blocks. Now commit
822 record can be sent to disk without waiting for descriptor
823 blocks to be written to disk. This behavior is controlled
824 using JBD2_FEATURE_ASYNC_COMMIT flag.
828 Description: OSS needs mutliple precreate threads
829 Details : Add ability to start more than one create thread per OSS.
833 Description: Sanity.sh test_65k failed: setstripe should have succeeded
834 Details : execute lfs setstripe on client
838 Description: mds_obd_create error creating tmp object
839 Details : When the user sets quota on root, llog will be affected and can't
840 create files and write files.
843 Frequency : Always on ia64 patchless client, and possibly others.
845 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
846 Details : This allows the patchless client to be loaded on architectures
851 Description: RedHat Update kernel for RHEL5
852 Details : Add the kernel config file for RHEL5.
856 Description: RedHat Update kernel for RHEL5
857 Details : Modify the kernel config file more closer RHEL5.
861 Description: Build failure against Centos5 (RHEL5)
862 Details : Define PAGE_SIZE when it isn't present.
866 Description: test_7 on sanity-quota.sh can't be used on separate machines
867 Details : add facet to handle it
871 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
872 Details : with NFS, the anon dentry's parent was set to itself in
873 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
874 corresponding dentry other than use rec->ur_name.
876 Severity : enhancement
878 Description: lfs setstripe enhancement
879 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
883 Description: enable data checksumming by default
884 Details : enable checksum by default, allow --disable-checksum
885 configure option and "-o nochecksum" mount option. Checksums
886 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
890 Description: lustre support for RHEL5
891 Details : Add support for RHEL5.
895 Description: Client eviction due to failover config
896 Details : after a connection loss, the lustre client should attempt to
897 reconnect to the last active server first before trying the
898 other potential connections.
902 Description: when mds and osts use different quota unit(32bit and 64bit),
903 quota will be released repeatly.
904 Details : void sending multiple quota reqs to mds, which will keep the
905 status between the reqs.
909 Description: osts not allocated evenly to files
910 Details : change the condition to increase offset_idx
913 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
915 Description: Data corruption for OSTs that are formatted larger than 2TB
917 Details : When generating the bio request for lustre file writes the
918 sector number would overflow a temporary variable before being
919 used for the IO. The data reads correctly from Lustre (which
920 will overflow in a similar manner) but other file data or
921 filesystem metadata may be corrupted in some cases.
925 Description: Tune the kernel for good SCSI performance.
926 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
927 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
932 Description: rewrite ext2-derived code in obdclass/uuid.c
933 Details : rewrite inherited code (uuid parsing code from ext2 utils)
934 from scratch preserving functionality.
938 Description: "lfs find -obd UUID" prints directories
939 Details : "lfs find -obd UUID" will return all directory names instead
940 of just file names. It is incorrect because the directories
941 do not reside on the OSTs.
945 Description: use i_size_read and i_size_write in 2.6 port
946 Details : replace inode->i_size access with i_size_read/write()
949 Frequency : when removing large files
951 Description: scheduling issue during removal of large Lustre files
952 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
953 It causes scheduling issues when removing large files (17TB in the
957 Frequency : only with liblustre clients on XT3
959 Description: evictions taking too long
960 Details : allow llrd to evict clients directly on OSTs
963 Frequency : only on ppc
965 Description: /proc/fs/lustre/devices broken on ppc
966 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
967 We should make sure the type of 'index' is loff_t and then cast
968 explicitly as needed below. Do not assign an explicitly cast
972 Frequency : only for rhel5
974 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
975 Details : Modify the target file & which_kernel.
977 Severity : enhancement
979 Description: omit set fsid for export NFS
980 Details : fix set/restore device id for avoid EMFILE error and mark lustre
981 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
985 Frequency : Always, for kernels after 2.6.16
986 Description: Fix warning idr_remove called for id=.. which is not allocated.
987 Details : Last kernels save old s_dev before kill super and not allow
988 to restore from callback - restore it before call kill_anon_super.
992 Description: buffer overruns could theoretically occur
993 Details : llapi_semantic_traverse() modifies the "path" argument by
994 appending values to the end of the origin string, and a
995 overrun may occur. Adding buffer overrun check in liblustreapi.
999 Description: Fix error on 'ls .' at the top of the Lustre mount.
1000 Details : Don't revalidate dentry if it is a root dentry.
1004 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1005 Details : Modify vm-tunables-rhel4.patch.
1009 Description: Kernel config for 2.6.18-vanilla.
1010 Details : Modify targets/2.6-vanilla.target.in.
1011 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1012 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1013 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1014 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1018 Description: improve handling recoverable errors
1019 Details : if request processig with error which can be recoverable on server
1020 request should be resend, otherwise page released from cache and
1025 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1026 Details : Modify target file & which_patch.
1027 A flaw was found in the IA32 system call emulation provided
1028 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1029 value could be stored in the %RAX register, which could trigger an
1030 out-of-bounds system call table access. An untrusted local user
1031 could exploit this flaw to run code in the kernel
1032 (ie a root privilege escalation). (CVE-2007-4573).
1036 Description: O_DIRECT bypasses client statistics.
1037 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1038 read_ahead_stats were updated. copied the stats section from
1039 osc_send_oap_rpc() into async_internal().
1043 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1047 Description: change order of libsysio includes
1048 Details : '#include sysio.h' should always come before '#include xtio.h'
1052 Description: handle "serious error: objid * already exists" more gracefully
1053 Details : If LAST_ID value on disk is smaller than the objects existing in
1054 the O/0/d* directories, it indicates disk corruption and causes an
1055 LBUG(). If the object is 0-length, then we should use the existing
1056 object. This will help to avoid a full fsck in most cases.
1058 Severity : enhancement
1060 Description: adapt the lustre_config script to support the upgrade case
1061 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1062 server targets to 1.6 in parallel.
1066 Description: To avoid grant space > avaible space when the disk is almost
1067 full. Without this patch you might see the error "grant XXXX >
1068 available" or some LBUG about grant, when the disk is almost
1070 Details : In filter_check_grant, for non_grant cache write, we should
1071 check the left space by if (*left > ungranted + bytes), instead
1072 of (*left > ungranted), because only we are sure the left space
1073 is enough for another "bytes", then the ungrant space should be
1074 increase. In client, we should update cl_avail_grant only there
1075 is OBD_MD_FLGRANT in the reply.
1079 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1080 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1081 described in CVE-2007-4573. This problem could allow untrusted
1082 local users to gain root access.
1085 Frequency : when using O_DIRECT and quotas
1087 Description: Incorrect file ownership on O_DIRECT output files
1088 Details : block usage reported by 'lfs quota' does not take into account
1089 files that have been written with O_DIRECT.
1094 Description: touch file failed when fs is not full
1095 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1096 otherwise we can get ENOSP while fs is not full.
1100 Description: parallel lock callbacks
1101 Details : Instead of sending blocking and completion callbacks as separated
1102 requests, adding them to a set and sending in parallel.
1105 Frequency : only for Cray XT3
1106 Bugzilla : 12829/13455
1107 Description: Changing primary group doesn't change the group lustre assigns to
1109 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1110 overridden with the primary group provided by the group upcall,
1111 whereas the supplied fsgid can be trusted if it is in the list of
1112 supplementary groups returned by the group upcall.
1114 Severity : enhancement
1116 Description: Allow masking D_WARNING, D_ERROR messages from console
1117 Details : Console messages can now be disabled via lnet.printk.
1121 Description: User code with malformed file open parameter crashes client node
1122 Details : Before packing join_file req, all the related reference should be
1123 checked carefully in case some malformed flags cause fake
1124 join_file req on client.
1128 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1129 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1130 mode is equal to the granted mode. However, it should also check
1131 that LDLM_FL_CP_REQD is not set.
1135 Description: Heavy nfs access might result in deadlocks
1136 Details : After ELC code landed, it is now improper to enqueue any mds
1137 locks under och_sem, because enqueue might want to decide to
1138 cancel open locks for same inode we are holding och_sem for.
1142 Description: Client eviction while running blogbench
1143 Details : A lot of unlink operations with concurrent I/O can lead to a
1144 deadlock causing evictions. To address the problem, the number of
1145 oustanding OST_DESTROY requests is now throttled to
1146 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1147 callbacks are processed in priority.
1151 Description: enable ACLs on MDS by default
1152 Details : ACLs must be enabled on MDS by default.
1155 Frequency : PPC/PPC64 only
1157 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1158 Details : fix duplicated definitions between asm-ppc64/types.h and
1159 lustre_types.h on PPC.
1162 Frequency : PPC/PPC64 only
1164 Description: asm-ppc/segment.h does not exist
1165 Details : fix compile issue on PPC.
1169 Description: better handle error messages in extents code
1172 Frequency : RHEL4 only
1174 Description: mkfs is very slow on IA64/RHEL4
1175 Details : A performance regression has been discovered in the MPT Fusion
1176 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1177 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1178 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1180 Severity : enhancement
1182 Description: SNMP support enhancement
1183 Details : Adding total number of sampled request for an MDS node in snmp
1186 Severity : enhancement
1188 Description: Optimize ldlm waiting list processing for PR extent locks
1189 Details : When processing waiting list for read extent lock and meeting read
1190 lock that is same or wider to it that is not contended, skip
1191 processing rest of the list and immediatelly return current
1192 status of conflictness, since we are guaranteed there are no
1193 conflicting locks in the rest of the list.
1197 Description: Time out and refuse to reconnect
1198 Details : When the failover node is the primary node, it is possible
1199 to have two identical connections in imp_conn_list. We must
1200 compare not conn's pointers but NIDs, otherwise we can defeat
1201 connection throttling.
1205 Description: port llog fixes from b1_6 into HEAD
1206 Details : Port llog reference couting and some llog cleanups from b1_6
1207 (bug 10800) into HEAD, for protect from panic and access to already
1208 free llog structures.
1212 Description: Detect stride IO mode in read-ahead
1213 Details : When a client does stride read, read-ahead should detect that and
1214 read-ahead pages according to the detected stride pattern.
1218 Description: data checksumming impacts single node performance
1219 Details : add support for several checksum algorithm. Currently, only CRC32
1220 and Adler-32 are supported. The checksum type can be changed on
1221 the fly via /proc/fs/lustre/osc/*/checksum_type.
1225 Description: use adler32 for page checksums
1226 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1231 Description: build for x2 fails
1232 Details : fix compile issue on Cray systems.
1236 Description: Properly match for duplicate locks
1237 Details : Due to different lock order from skiplists code, we need to
1238 traverse entire list for now
1241 Frequency : only on PPC/SLES10
1243 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1244 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1245 be defined. Add a hack in mkfs_lustre.c to work around this compile
1250 Description: LASSERT on MDS when client holding flock lock dies
1251 Details : ldlm pool logic depends on number of granted locks equal to
1252 number of released locks which is not true for flock locks, so
1253 just exclude such locks from consideration.
1257 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1258 Details : Use fsfilt_sync() for both the whole filesystem sync and
1259 individual file sync to eliminate dangerous inode locking
1260 with I_LOCK that can lead to a deadlock.
1264 Description: performance in 1.6.3
1265 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1266 sure that 1MB requests are not fragmented by the block layer.
1268 Severity : enhancement
1270 Description: organize the server-side client stats on per-nid basis
1271 Details : Change the structure of stats under obdfilter and mds to
1279 The "uuid"s file would list the uuids of _active_ exports.
1280 And the clear entry is to clear all stats and stale nids.
1282 Severity : enhancement
1284 Description: eliminate client locks in face of contention
1285 Details : file contention detection and lockless i/o implementation
1286 for contended files.
1290 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1293 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1295 Description: MDS may lose file striping (and hence file data) in some cases
1296 Details : If there are additional extended attributes stored on the MDS,
1297 in particular ACLs, SELinux, or user attributes (if user_xattr
1298 is specified for the client mount options) then there is a risk
1299 of attribute loss. Additionally, the Lustre file striping
1300 needs to be larger than default (e.g. striped over all OSTs),
1301 and an additional attribute must be stored initially in the
1302 inode and then increase in size enough to be moved to the
1303 external attribute block (e.g. ACL growing in size) for file
1308 Description: skiplist implementation simplification
1309 Details : skiplists are used to group compatible locks on granted list
1310 that was implemented as tracking first and last lock of each
1311 lock group the patch changes that to using doubly linked lists
1315 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1316 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1317 because LASSERT on both the data supplied by a client, and the
1318 data on disk is dangerous and incorrect.
1320 Severity : enhancement
1322 Description: Slow truncate/writes to huge files at high offsets.
1323 Details : Directly associate cached pages to lock that protect those pages,
1324 this allows us to quickly find what pages to write and remove
1325 once lock callback is received.
1329 Description: more ldlm soft lockups
1330 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1331 starve other threads from the resource lock for a long time in
1332 case of long waiting queue, so change the debug level from
1333 D_OTHER to the less frequently used D_INFO.
1335 Severity : enhancement
1337 Description: add -gid, -group, -uid, -user options to lfs find
1341 Description: Hung threads in invalidate_inode_pages2_range
1342 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1343 once one is completed. As a result, some RPCs are stuck in the
1344 queue and are never sent.
1348 Description: filter threads hungs on waiting journal commit
1349 Details : Cleanup filter group llog code, then only filter group llog will
1350 be only created in the MDS/OST syncing process.
1354 Description: Procfs and llog threads access destoryed import sometimes.
1355 Details : Sync the import destoryed process with procfs and llog threads by
1356 the import refcount and semaphore.
1358 Severity : enhancement
1360 Description: openlock cache of b1_6 port to HEAD
1365 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1366 Details : If insertion of an extent fails, then discard the inode
1367 preallocation and free data blocks else it can lead to duplicate
1372 Description: don't always update ctime in ext3_xattr_set_handle()
1373 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1374 In some cases the ctime should not be updated, for example for
1375 2.0->1.8 compatibility it is necessary to delete an xattr and it
1376 should not update the ctime.
1380 Bugzilla : 15713/16362
1381 Description: Assertion in iopen_connect_dentry in 1.6.3
1382 Details : looking up an inode via iopen with the wrong generation number can
1383 populate the dcache with a disconneced dentry while the inode
1384 number is in the process of being reallocated. This causes an
1385 assertion failure in iopen since the inode's dentry list contains
1386 both a connected and disconnected dentry.
1390 Description: assertion failure in ldlm_handle2lock()
1391 Details : fix a race between class_handle_unhash() and class_handle2object()
1392 introduced in lustre 1.6.5 by bug 13622.
1397 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1398 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1403 Description: oops in sd_iostats_seq_show()
1404 Details : unloading/reloading the scsi low level driver triggers a kernel
1405 bug when trying to access the sd iostat file.
1410 Description: Kernel panics during QLogic driver reload
1411 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1412 patch, causing memory corruption.
1417 Description: journal_dev option does not work in b1_6
1418 Details : pass mount option during pre-mount.
1420 Severity : enhancement
1422 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1423 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1424 extent information of a file. It can be used to map logical blocks
1425 in a file to physical blocks in the block device.
1429 Description: LDLM soft lockups - improvement
1430 Details : It is be possible to send the lock handle along with each read
1431 or write request because the client is already doing a lock match
1432 itself so there isn't any reason the OST should have to re-do that
1436 Frequency : only on Cray X2
1438 Description: X2 build failures
1439 Details : fix build failures on Cray X2.
1443 Description: xid & resent requests
1444 Details : Initialize RPC XID from clock at startup (randomly if clock is
1447 Severity : enhancement
1449 Description: Add lustre_start utility to start or stop multiple Lustre servers
1454 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1455 Details : In case of memory pressure, list_del() can be called twice on
1456 req->rq_history_list, causing a kernel oops.
1460 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1461 Details : fix stack overflow in the distributed lock manager by defering
1462 export eviction after a failed AST to the elt thread instead of
1463 handling it in the dlm interpret routine.
1467 Description: Convert some comments to new format.
1468 Details : Update documenting comments to match doxygen conventions.
1472 Description: Grammar fixes.
1473 Details : A couple of trivial spelling fixes.
1477 Description: OSD_COUNTERS-mandatory
1478 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1479 to simplify the code.
1483 Description: simplify cmm_device freeing logic.
1484 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1488 Description: Add lockdep support to dt_object_operations locking interface.
1489 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1490 parameter indicating lock ordering. Update mdd code to use new
1495 Description: Introduce failloc constants for lockless IO tests.
1496 Details : Add two new failloc constants to test lockless IO. Only one of
1497 them in implemented---another is checked in yet to be landed
1502 Description: Add lockdep support for inode mutex.
1503 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1504 in the situations where Lustre has to lock more than one inode
1509 Description: Add optional invariants checking support.
1510 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1511 configure switch. This macro is to be used for consistency and
1512 sanity checks that are too expensive to be left in `production'
1517 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1518 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1523 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1524 Details : Use spin_lock_nested() in (the only) situation where more than
1525 one ldlm_lock is locked simultaneously. Also, fix possible
1526 dead-lock in ldlm_lock_change_resource() by enforcing particular
1531 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1532 Details : Instead of specifying each ldlm_lock call-back through separate
1533 parameter, wrap them into struct ldlm_callback_suite.
1537 Description: Kill join_lru obd method and its callers.
1538 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1539 in memory---a requirement implemented through ->o_join_lru() obd
1540 method in HEAD. Remove this method and its users.
1544 Description: Add asynchronous ldlm ENQUEUE completion handler.
1545 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1546 case that stock ldlm_completion_ast() cannot handle as it waits
1547 until lock is granted. Introduce new ldlm_completion_ast_async()
1548 for this. Also comment ldlm_completion_ast().
1552 Description: ldlm_error <-> errno conversion.
1553 Details : Add functions to map (rather arbitrary) between LDLM error codes
1554 and standard errno values. CLIO needs this to prevent LDLM specific
1555 constants from escaping ldlm and osc.
1559 Description: Kill unused ldlm_handle2lock_ns() function.
1560 Details : Kill unused ldlm_handle2lock_ns() function.
1564 Description: Add lu_ref support to ldlm_lock
1565 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1566 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1567 and ldlm_resource. LDLM interface has to be changed, because code that
1568 releases a reference on a lock, has to "know" what reference this is.
1569 In the most frequent case
1571 lock = ldlm_handle2lock(handle);
1573 LDLM_LOCK_PUT(lock);
1575 no changes are required. When any other reference (received _not_ from
1576 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1577 instead of LDLM_LOCK_PUT().
1579 Arguably, changes are pervasive, and interface requires some discipline
1580 for proper use. On the other hand, it was very instrumental in finding
1581 a few leaked lock references.
1585 Description: Add ldlm_lock_addref_try().
1586 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1587 attempts to addref a lock that might be being canceled
1592 Description: Add ldlm_weigh_callback().
1593 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1594 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1595 hard-coded `number of pages' logic.
1599 Description: Add lockdep annotations to llog code.
1600 Details : Use appropriately tagged _nested() locking calls in the places
1601 where llog takes more than one ->lgh_lock lock.
1605 Description: Add loi_kms_set().
1606 Details : Wrap kms updates into a helper function.
1610 Description: Constify instances of struct lsm_operations.
1611 Details : Constify instances of struct lsm_operations.
1615 Description: lu_conf support.
1616 Details : On a server, a file system object is uniquely identified
1617 by a fid, which is sufficient to locate and load all object
1618 state (inode). On a client, on the other hand, more data are
1619 necessary instantiate an object. Change lu_object_find() and
1620 friends to take additional `lu_conf' argument describing object.
1621 Typically this includes layout information.
1625 Description: lu_context fixes.
1626 Details : Introduce new lu_context functions that are needed on the client
1627 side, where some system threads (ptlrpcd) are shared by multiple
1628 modules, and so cannot be stopped during module shutdown.
1632 Description: Add start and stop methods to lu_device_type_operations.
1633 Details : Introduce two new methods in lu_device_type_operations, that are
1634 invoked when first instance of a given type is created and last one
1635 is destroyed respectively. This is need by CLIO.
1639 Description: Add lu_ref support to struct lu_device.
1640 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1641 track leaked references.
1645 Description: Introduce lu_kmem_descr.
1646 Details : lu_kmem_descr and its companion interface allow to create
1647 and destroy a number of kmem caches at once.
1651 Description: Fix lu_object finalization race.
1652 Details : Fix a race between lu_object_find() finding an object and its
1653 concurrent finalization. This race is (most likely) not possible
1654 on the server, but might happen on the client.
1658 Description: Introduce lu_ref interface.
1659 Details : lu_ref is a debugging module allowing to track references to
1660 a given object. It is quite cpu expensive, and has to be
1661 explicitly enabled with --enable-lu_ref. See usage description
1666 Description: Factor lu_site procfs stats into a separate function.
1667 Details : Separate lu_site stats printing code into a separate function
1668 that can be reused on a client.
1672 Description: Constify instances of struct {lu,dt,md}_device_operations.
1673 Details : Constify instances of struct {lu,dt,md}_device_operations.
1677 Description: Introduce struct md_site and move meta-data specific parts of
1678 struct lu_site here.
1679 Details : Move md-specific fields out of struct lu_site into special struct
1680 md_site, so that lu_site can be used on a client.
1684 Description: Kill mdd_lov_destroy().
1685 Details : Remove unused mdd code.
1689 Description: Add st_block checking to multistat.c.
1690 Details : Add st_block checking to multistat.c.
1694 Description: Add lu_ref support to struct obd_device.
1695 Details : Add lu_ref tracking to obd_device.
1699 Description: Kill obd_set_fail_loc().
1700 Details : Remove unused code.
1704 Description: Add special type for ptlrpc_request interpret functions.
1705 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1706 there. Actual usage will be in CLIO.
1710 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1711 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1712 doesn't work with lockdep.
1716 Description: Add rwv.c test program.
1717 Details : New testing program exercising readv(2) and writev(2) (Qian).
1721 Description: Add sendfile.c test program.
1722 Details : New testing program exercising sendfile(2) (Jay).
1726 Description: Ratelimit a message that can be very frequent.
1727 Details : Ratelimit a memory allocation failure message that can
1732 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1733 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1737 Description: Liblustre build fixes.
1738 Details : Liblustre build fixes.
1742 Description: libcfs: add cfs_{need,cond}_resched() interface.
1743 Details : libcfs: add cfs_{need,cond}_resched() definition and
1744 implementations for Linux, NT, and liblustre.
1746 Severity : enhancement
1748 Description: More exported tunables for mballoc
1749 Details : Add support for tunable preallocation window and new tunables for
1750 large/small requests
1754 Description: Detect corruption of block bitmap and checking for preallocations
1755 Details : Checks validity of on-disk block bitmap. Also it does better
1756 checking of number of applied preallocations. When corruption is
1757 found, it turns filesystem readonly to prevent further corruptions.
1761 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1762 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1763 window to 0. If it does hit cache miss, and read-pattern is still
1764 stride-io mode, does not reset the stride window, but also does not
1765 increase the stride window length in this case.
1769 Frequency : only for big-endian servers
1770 Description: Check if system is big-endian while mounting fs with extents feature
1771 Details : Mounting a filesystem with extents feature will fail on big-endian
1772 systems since ext3-based ldiskfs is not supported on big-endian
1773 systems. This can be over-riden with "bigendian_extents" mount option.
1775 --------------------------------------------------------------------------------
1777 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1779 * Support for kernels:
1780 2.6.5-7.283 (SLES 9),
1781 2.6.9-55.EL (RHEL 4),
1782 2.6.16.46-0.14 (SLES 10),
1783 2.6.18.8 vanilla (kernel.org)
1784 * Client support for unpatched kernels:
1785 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1786 2.6.16 - 2.6.22 vanilla (kernel.org)
1787 * Due to recently discovered recovery problems, we do not recommend
1788 using patchless RHEL 4 clients with this or any earlier release.
1789 * Recommended e2fsprogs version: 1.39.cfs8
1790 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1791 * Starting with this release, the ldiskfs backing filesystem required
1792 by Lustre is now in its own package, lustre-ldiskfs. This package
1793 should be installed. It is versioned separately from Lustre and
1794 may be released separately in future.
1798 Description: block reactivating mgc import until all deactivates complete
1799 Details : Fix race when failing back MDT/MGS to itself (testing)
1802 Frequency : at statup only
1804 Description: mds_lov_synchronize race leads to various problems
1805 Details : simultaneous MDT->OST connections at startup can cause the
1806 sync to abort, leaving the OSC in a bad state.
1808 Severity : enhancement
1810 Description: add optional extra BUILD_VERSION info
1811 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1812 to override the lustre version.
1815 Frequency : 2.6.18 servers only
1817 Description: ll_kern_mount() doesn't release the module reference
1818 Details : The ldiskfs module reference count never drops down to 0
1819 because ll_kern_mount() doesn't release the module reference.
1824 Description: server LBUG when using old ost_num_threads parameter
1825 Details : Accept the old ost_num_threads parameter but warn that it
1826 is deprecated, and fix an off-by-one error that caused an LBUG.
1831 Description: Transient SCSI error results in persistent IO issue
1832 Details : iobuf->dr_error is not reinitialized to 0 between two
1836 Frequency : sometimes when underlying device returns I/O errors
1838 Description: OSTs not going read-only during write failures
1839 Details : OSTs are not remounted read-only when the journal commit threads
1840 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1841 instead of the ext3 wrappers.
1845 Description: poor connect scaling with increasing client count
1846 Details : Don't run filter_grant_sanity_check for more than 100 exports
1847 to improve scaling for large numbers of clients.
1850 Frequency : SLES10 only
1852 Description: sanity-quota.sh quotacheck failed: rc = -22
1853 Details : Quotas cannot be enabled on SLES10.
1856 Frequency : liblustre clients only
1858 Description: getdirentries does not give error when run on compute nodes
1859 Details : getdirentries does not fail when the size specified as an argument
1860 is too small to contain at least one entry
1862 Severity : enhancement
1864 Description: Add LNET router traceability for debug purposes
1865 Details : If a checksum failure occurs with a router as part of the
1866 IO path, the NID of the last router that forwarded the bulk data
1867 is printed so it can be identified.
1872 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1873 Details : Due to a race condition, liblustre clients were occasionally
1874 evicted incorrectly.
1876 Severity : enhancement
1878 Description: lfs setstripe use optional parameters instead of postional
1881 Severity : enhancement
1883 Description: Nanosecond timestamp support for ldiskfs
1884 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1885 resolution timestamps. There is not yet support for this at
1886 the Lustre filesystem level.
1889 Frequency : during server recovery
1891 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1892 Details : request with rq_no_resend flag not awake l_wait_event if they get
1896 Frequency : nfs export on patchless client
1898 Description: connectathon hang when test nfs export over patchless client
1899 Details : Disconnected dentry cannot be found with lookup, so we do not need
1900 to unhash it or make it invalid
1903 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1904 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1905 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1910 Description: open req refcounting wrong on reconnect
1911 Details : If reconnect happened between getting open reply from server and
1912 call to mdc_set_replay_data in ll_file_open, we will schedule
1913 replay for unreferenced request that we are about to free.
1914 Subsequent close will crash in variety of ways.
1915 Check that request is still eligible for replay in
1916 mdc_set_replay_data().
1921 Description: disable writes to filesystem when reading health_check file
1922 Details : the default for reading the health_check proc file has changed
1923 to NOT do a journal transaction and write to disk, because this
1924 can cause reads of the /proc file to hang and block HA state
1925 checking on a healthy but otherwise heavily loaded system. It
1926 is possible to return to the previous behaviour during configure
1927 with --enable-health-write.
1929 Severity : enhancement
1931 Description: 64-bit inode version
1932 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1933 to the inode. This will be required for version-based recovery.
1938 Description: MDS fails to start if a duplicate client export is detected
1939 Details : in some rare cases it was possible for a client to connect to
1940 an MDS multiple times. Upon recovery the MDS would detect this
1941 and fail during startup. Handle this more gracefully.
1943 Severity : enhancement
1945 Description: Add -o localflock option to simulate old noflock
1947 Details : This will achieve local-only flock/fcntl locks
1953 Description: log_commit_thread vs filter_destroy race leads to crash
1954 Details : Take import reference before releasing llog record semaphore
1959 Description: Wrong request locking in request set processing
1960 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1961 imp_delayed_list, in this place should be used imp_lock.
1964 Frequency : when reconnection
1966 Description: Grant Leak when osc reconnect to OST
1967 Details : When osc reconnect ost, OST(filter) should check whether it
1968 should grant more space to client by comparing fed_grant and
1969 cl_avail_grant, and return the granted space to client instead
1970 of "new granted" space, because client will call osc_init_grant
1971 to update the client grant space info.
1974 Frequency : when client reconnect to OST
1976 Description: Grant Leak when osc do resend and replay bulk write
1977 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1978 bulk write request, because the grant info while be sync between
1979 OSC and OST when reconnect, and we should ignore the grant info
1980 these of resend/replay write req.
1985 Description: Grant space more than avaiable left space sometimes.
1986 Details : When then OST is about to be full, if two bulk writing from
1987 different clients came to OST. Accord the avaliable space of the
1988 OST, the first req should be permitted, and the second one
1989 should be denied by ENOSPC. But if the seconde arrived before
1990 the first one is commited. The OST might wrongly permit second
1991 writing, which will cause grant space > avaiable space.
1994 Frequency : when client is evicted
1996 Description: Grant might be wrongly erased when osc is evicted by OST
1997 Details : when the import is evicted by server, it will fork another
1998 thread ptlrpc_invalidate_import_thread to invalidate the
1999 import, where the grant will be set to 0. While the original
2000 thread will update the grant it got when connecting. So if
2001 the former happened latter, the grant will be wrongly errased
2002 because of this race.
2007 Description: Checking Stale with correct fid
2008 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2009 to check whether it is stale, because sometimes, we want the
2010 enqueue happened anyway, and op_data.fid2 will not be initialized.
2012 Severity : enhancement
2014 Description: update patchless client
2015 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2018 Frequency : only with 2.4 kernel
2020 Description: random memory corruption
2021 Details : size of struct ll_inode_info is to big for union inode.u and this
2022 can be cause of random memory corruption.
2027 Description: Memory leak in recovery
2028 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2029 It should also check obd_fail before fsfilt_start, otherwise if
2030 fsfilt_start return -EROFS,(failover mds during mds recovery).
2031 then the req will return with repmsg->transno = 0 and rc = EROFS.
2032 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2033 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2034 be freed no matter whether fsfilt_commit success or not.
2037 Frequency : only with huge count clients
2039 Description: Prevents from taking the superblock lock in llap_from_page for
2041 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2042 for llap_from_page call in ll_removepage prevents from taking the
2043 superblock lock for a soon died page.
2048 Description: Not check open intent error before release open handle
2049 Details : in some rare cases, the open intent error is not checked before
2050 release open handle, which may cause
2051 ASSERTION(open_req->rq_transno != 0), because it tries to release
2052 the failed open handle.
2057 Description: Set cat log bitmap only after create log success.
2058 Details : in some rare cases, the cat log bitmap is set too early. and it
2059 should be set only after create log success.
2063 Description: Accessing a block bevice can re-enable I/O when Lustre is
2064 tearing down a device.
2065 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2070 Description: service threads may hog cpus when there are a lot of requests
2071 Details : Insert cond_resched to give other threads a chance to use some CPU
2076 Description: the cat log was not initialized in recovery
2077 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2078 unlink log on mds will not be initialized until mds post
2079 recovery. And also in mds post recovery, the unlink log will
2080 initialization will be done asynchronausly, so there will be race
2081 between add unlink log and unlink log initialization.
2085 Description: brw_stats were being printed incorrectly
2086 Details : brw_stats were being printed as log2 but all of them were not
2087 recorded as log2. Also remove some code duplication arising from
2088 filter_tally_{read,write}.
2092 Frequency : rare, only in recovery.
2093 Description: ASSERTION(req->rq_type != LI_POISON) failed
2094 Details : imp_lock should be held while iterating over imp_sending_list for
2095 prevent destroy request after get timeout in ptlrpc_queue_wait.
2099 Description: replay-single.sh test 52 fails
2100 Details : A lock's skiplist need to be cleanup when it being unlinked
2101 from its resource list.
2105 Description: Short directio read returns full requested size rather than
2107 Details : Direct I/O operations should return actual amount of bytes
2108 transferred rather than requested size.
2110 Severity : enhancement
2112 Description: metadata RPC reduction (e.g. for rm performance)
2113 Details : decrease the amount of synchronous RPC between clients and servers
2114 by canceling conflicing lock before the operation on the client
2115 and packing thier handles into the main operation RPC to server.
2117 Severity : enhancement
2119 Description: Async OSC create to avoid the blocking unnecessarily.
2120 Details : If a OST has no remain object, system will block on the creating
2121 when need to create a new object on this OST. Now, ways use
2122 pre-created objects when available, instead of blocking on an
2123 empty osc while others are not empty. If we must block, we block
2124 for the shortest possible period of time.
2128 Description: improve handling recoverable errors
2129 Details : if request processig with error which can be recoverable on server
2130 request should be resend, otherwise page released from cache and
2133 Severity : enhancement
2135 Description: refine locking for avoid write wrong info into lov_objid file
2136 Details : fix possible races with add new target and write/update data in
2139 --------------------------------------------------------------------------------
2141 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2146 Frequency : on some architectures
2148 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2149 Details : Uninitialized flags sometimes cause configuration commands to
2153 Frequency : patchless clients only
2155 Description: missing __iget() symbol export
2156 Details : The __iget() symbol export is missing. To avoid the need for
2157 this on patchless clients the deathrow inode reaper is turned
2158 off, and we depend on the VM to clean up old inodes. This
2159 dependency was during via the fix for bug 12181.
2161 --------------------------------------------------------------------------------
2163 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2165 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2166 INTEROPERATE with older versions automatically. In many cases a
2167 special upgrade step is needed. Please read the
2168 user documentation before upgrading any part of a live system.
2169 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2170 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2171 * WARNING: Lustre configuration and startup changes are required with
2172 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2176 Severity : enhancement
2178 Description: MountConf
2179 Details : Lustre configuration is now managed via mkfs and mount
2180 commands instead of lmc and lconf. New obd types (MGS, MGC)
2181 are added for dynamic configuration management. See
2182 https://mail.clusterfs.com/wikis/lustre/MountConf for
2185 Severity : enhancement
2187 Description: dynamic OST addition
2188 Details : OSTs can now be added to a live filesystem
2190 Severity : enhancement
2192 Description: startup order invariance
2193 Details : MDTs and OSTs can be started in any order. Clients only
2194 require the MDT to complete startup.
2196 Severity : enhancement
2198 Description: parallel, asynchronous orphan cleanup
2199 Details : orphan cleanup is now performed in separate threads for each
2200 OST, allowing parallel non-blocking operation.
2202 Severity : enhancement
2204 Description: optimized stripe assignment
2205 Details : stripe assignments are now made based on ost space available,
2206 ost previous usage, and OSS previous usage, in order to try
2207 to optimize storage space and networking resources.
2209 Severity : enhancement
2211 Description: Permanently set tunables
2212 Details : All writable /proc/fs/lustre tunables can now be permanently
2213 set on a per-server basis, at mkfs time or on a live system.
2215 Severity : enhancement
2217 Description: Lustre message v2
2218 Details : Add lustre message format v2.
2220 Severity : enhancement
2222 Description: client OST exclusion list
2223 Details : Clients can be started with a list of OSTs that should be
2224 declared "inactive" for known non-responsive OSTs.
2228 Description: SPEC SFS validation failure on NFS v2 over lustre.
2229 Details : Changes the blocksize for regular files to be 2x RPC size,
2230 and not depend on stripe size.
2232 Severity : enhancement
2234 Description: Multiple MD RPCs in flight.
2235 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2236 To avoid overly-overloading MDS, introduce a limit on number of
2237 MDS RPCs in flight for a single client and add /proc controls
2238 to adjust this limit.
2240 Severity : enhancement
2242 Description: client read/write statistics
2243 Details : Add client read/write call usage stats for performance
2244 analysis of user processes.
2245 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2246 file access. extents_stats shows chunk size distribution.
2247 extents_stats_per_process show chunk size distribution per
2250 Severity : enhancement
2252 Description: mds statistics
2253 Details : Add detailed mds operations statistics in
2254 /proc/fs/lustre/mds/*/stats.
2258 Description: Failure of copying files with lustre special EAs.
2259 Details : Client side always return success for setxattr call for lustre
2260 special xattr (currently only "trusted.lov").
2265 Description: Refcount LNET uuids
2266 Details : The global LNET uuid list grew linearly with every startup;
2267 refcount repeated list entries instead of always adding to
2270 Severity : enhancement
2272 Description: Dynamic service threads
2273 Details : Within a small range, start extra service threads
2274 automatically when the request queue builds up.
2277 Frequency : mixed-endian client/server environments
2279 Description: mixed-endian crashes
2280 Details : The new msg_v2 system had some failures in mixed-endian
2283 Severity : enhancement
2285 Description: Easy OST removal
2286 Details : OSTs can be permanently deactivated with e.g. 'lctl
2287 conf_param lustre-OST0001.osc.active=0'
2289 Severity : enhancement
2291 Description: MGS proc entries
2292 Details : Added basic proc entries for the MGS showing what filesystems
2295 Severity : enhancement
2297 Description: provide MGS failover
2298 Details : Added config lock reacquisition after MGS server failover.
2300 Severity : enhancement
2302 Description: add Linux 2.4 support
2303 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2307 Description: a large application tries to do I/O to the same resource and dies
2308 in the middle of it.
2309 Details : Check the req->rq_arrival time after the call to
2310 ost_brw_lock_get(), but before we do anything about
2311 processing it & sending the BULK transfer request. This
2312 should help move old stale pending locks off the queue as
2313 quickly as obd_timeout.
2316 Frequency : when an incorrect nid is specified during startup
2318 Description: ptlrpc connect to non-existant node causes kernel crash
2319 Details : LNET can't be re-entered from an event callback, which
2320 happened when we expire a message after the export has been
2321 cleaned up. Instead, hand the zombie cleanup off to another
2324 Severity : enhancement
2326 Description: plain/inodebits lock performance improvement
2327 Details : Grouping plain/inodebits in granted list by their request modes
2328 and bits policy, thus improving the performance of search through
2332 Frequency : only if OST filesystem is corrupted
2334 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2335 Details : for a short time RPCs with bulk IO are in the replay list,
2336 but replay of bulk IOs is unimplemented. If the OST filesystem
2337 is corrupted due to disk cache incoherency and then replay is
2338 started it is possible to trip an assertion. Avoid putting
2339 committed RPCs into the replay list at all to avoid this issue.
2342 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2344 Description: System hang on startup
2345 Details : This bug allowed the liblustre (e.g. catamount) client to
2346 return to the app before handling all startup RPCs. This
2347 could leave the node unresponsive to lustre network traffic
2348 and manifested as a server ptllnd timeout.
2350 Severity : enhancement
2352 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2353 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2354 boolean to control whether to print peer debug info when a
2355 client's RPC times out.
2358 Frequency : only for kernels with patches from Lustre below 1.4.3
2360 Description: Remove old rdonly API
2361 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2364 Frequency : only for devices with external journals
2366 Description: Set external device read-only also
2367 Details : During a commanded failover stop, we set the disk device
2368 read-only while the server shuts down. We now also set any
2369 external journal device read-only at the same time.
2372 Frequency : when upgrading from 1.4 while trying to change parameters
2374 Description: The wrong (new) MDC name was used when setting parameters for
2375 upgraded MDT's. Also allows changing of OSC (and MDC)
2376 parameters if --writeconf is specified at tunefs upgrade time.
2379 Frequency : when setting specific ost indicies
2381 Description: QOS code breaks on skipped indicies
2382 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2383 created with --index need not be sequential.
2385 Severity : enhancement
2387 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2388 Details : The uninit_groups feature works in conjunction with the kernel
2389 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2390 up the pass1 processing of e2fsck. This is a read-only feature
2391 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2392 mount filesystems that have had this feature enabled.
2394 Severity : enhancement
2396 Description: Improve multi-block allocation algorithm to avoid fragmentation
2397 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2398 allocation locality and avoid filesystem fragmentation.
2400 ------------------------------------------------------------------------------
2402 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2404 * Support for kernels:
2405 2.4.21-47.0.1.EL (RHEL 3)
2406 2.6.5-7.283 (SLES 9)
2407 2.6.9-42.0.10.EL (RHEL 4)
2408 2.6.12.6 vanilla (kernel.org)
2409 2.6.16.27-0.9 (SLES 10)
2410 * Recommended e2fsprogs version: 1.39.cfs5
2412 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2416 Frequency : occasional, depends on client load and configuration
2417 Bugzilla : 12181, 12203
2418 Description: data loss for recently-modified files
2420 Details : In some cases it is possible that recently written or created
2421 files may not be written to disk in a timely manner (this should
2422 normally be within 30s unless client IO load is very high).
2423 The problem appears as zero-length files or files that are a
2424 multiple of 1MB in size after a client crash or client eviction
2425 that are missing data at the end of the file.
2427 This problem is more likely to be hit on clients where files are
2428 repeatedly created and unlinked in the same directory, clients
2429 have a large amount of RAM, have many CPUs, the filesystem has
2430 many OSTs, the clients are rebooted frequently, and/or the files
2431 are not accessed by other nodes after being written.
2433 The presence of the problem can be detected by looking at
2434 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2435 smaller than the second (nr_unused) then dirty files will not
2436 be flushed automatically to disk. "sync; sleep 10" should be
2437 run several times on the node before unmounting it to update
2438 Lustre (this is also safe to run on nodes without this problem).
2440 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2441 that can cause this same problem, so customers using that kernel
2442 also need to update the kernel in addition to Lustre. In order
2443 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2445 It is normal that files written just before a client crash (less
2446 than 30s) may not yet have been flushed to disk, even for local
2450 Frequency : frequent on thin XT3 nodes
2452 Description: UUID collision on thin XT3 Linux nodes
2453 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2454 randomly, since we relied on an insufficiently-seeded PRNG.
2459 Description: OSS hangs after "All ost request buffers busy"
2460 Details : A deadlock between quota and journal operations caused OSS
2461 hangs after printing "All ost request buffers busy."
2464 Frequency : always on liblustre builds
2466 Description: Cleanup compiler warnings on liblustre
2469 Frequency : always on liblustre builds on XT3
2471 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2475 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2476 > 32000 subdirectories
2477 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2478 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2479 subdirectory count crosses 32000. This will aid e2fsck to
2480 correctly handle more than 32000 subdirectories.
2483 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2485 Description: System hang on startup
2486 Details : This bug allowed the liblustre (e.g. catamount) client to
2487 return to the app before handling all startup RPCs. This
2488 could leave the node unresponsive to lustre network traffic
2489 and manifested as a server ptllnd timeout.
2491 Severity : enhancement
2493 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2494 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2495 boolean to control whether to print peer debug info when a
2496 client's RPC times out.
2501 Description: make O_SYNC working on 2.6 kernels
2502 Details : 2.6 kernels use different method for mark pages for write,
2503 so need add a code to lustre for O_SYNC work.
2508 Description: Failure to close file and release space on NFS
2509 Details : Put inode details into lock acquired in ll_intent_file_open.
2510 Use mdc_intent_lock in ll_intent_open to properly
2511 detect all kind of errors unhandled by mdc_enqueue.
2516 Description: proc file read during shutdown sometimes raced obd removal,
2518 Details : Add lock to prevent obd access after proc file removal.
2521 Frequency : Only for files larger than 4GB on 32-bit clients.
2523 Description: improperly doing page alignment of locks
2524 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2525 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2530 Description: RPCs being resent when they shouldn't be
2531 Details : Some RPCs that should not be resent are being resent. This
2532 can cause inconsistencies in the RPC state machine. Do not
2533 resend such requests.
2536 Frequency : rare, only with NFS export
2538 Description: Crash on NFS re-export node
2539 Details : under very unusual load conditions an assertion is hit in
2540 ll_intent_file_open()
2543 Frequency : only if OST filesystem is corrupted
2545 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2546 Details : for a short time RPCs with bulk IO are in the replay list,
2547 but replay of bulk IOs is unimplemented. If the OST filesystem
2548 is corrupted due to disk cache incoherency and then replay is
2549 started it is possible to trip an assertion. Avoid putting
2550 committed RPCs into the replay list at all to avoid this issue.
2555 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2556 Details : Large single O_DIRECT read and write calls can fail to allocate
2557 a sufficiently large buffer to process the request. In case of
2558 allocation failure the allocation is retried with a smaller
2559 buffer and broken into smaller requests.
2561 Severity : enhancement
2563 Description: Add -o localflock option to simulate old noflock behaviour.
2564 Details : This will achieve local-only flock/fcntl locks coherentness.
2569 Description: versioning check is incomplete
2570 Details : Checking the version difference of client vs. server, report
2571 error if the gap is too big.
2576 Description: add support PG_writeback bit
2577 Details : add support for PG_writeback bit for Lustre, for more carefull
2578 work with page cache in 2.6 kernel. This also fix some deadlocks
2579 and remove hack for work O_SYNC with 2.6 kernel.
2581 Severity : enhancement
2583 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2584 Details : The uninit_groups feature works in conjunction with the kernel
2585 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2586 up the pass1 processing of e2fsck. This is a read-only feature
2587 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2588 mount filesystems that have had this feature enabled.
2590 Severity : enhancement
2592 Description: Improve multi-block allocation algorithm to avoid fragmentation
2593 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2594 allocation locality and avoid filesystem fragmentation.
2596 ------------------------------------------------------------------------------
2598 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2600 * Support for kernels:
2601 2.6.9-42.0.3.EL (RHEL 4)
2602 2.6.5-7.276 (SLES 9)
2603 2.4.21-47.0.1.EL (RHEL 3)
2604 2.6.12.6 vanilla (kernel.org)
2605 2.6.16.21-0.8 (SLES10)
2606 * Recommended e2fsprogs version: 1.39.cfs2-0
2608 * The backwards-compatible /proc/sys/portals symlink has been removed
2609 in this release. Before upgrading, please ensure that you change
2610 any configuration scripts or /etc/sysctl.conf files that access
2611 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2612 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2613 in advance of the upgrade on any system running Lustre 1.4.6 or
2614 newer, since /proc/sys/lnet was added in that version.
2615 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2619 Frequency : only when quota is used
2621 Description: avoid scanning export list for quota master
2622 Details : Change the algorithms to avoid scanning export list in order
2623 to improve the efficiency.
2626 Frequency : MDS failover only, very rarely
2628 Description: "went back in time" messages on mds failover
2629 Details : The greatest transno may be lost when the current operation
2630 finishes with an error (transno==0) and the client's last_rcvd
2631 record is over-written. Save the greatest transno in the
2632 mds_last_transno for this case.
2635 Frequency : always for specific kernels and striping counts
2637 Description: client may get "Matching packet too big" without ACL support
2638 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2639 when trying to access files in certain configurations. The
2640 clients should in fact be denied when mounting because they do
2641 not understand ACLs.
2644 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2646 Description: many clients connecting with IO in progress causes connect timeouts
2647 Details : Avoid synchronous journal commits to avoid delays caused by many
2648 clients connecting/disconnecting when bulk IO is in progress.
2649 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2650 OST_IO_PORTAL to avoid delays behind potentially many pending
2654 Frequency : occasionally with multiple writers to a single file
2656 Description: shared writes to file may result in wrong size reported by stat()
2657 Details : Allow growing of kms when extent lock is cancelled
2660 Frequency : always with random mmap IO to multi-striped file
2662 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2663 Details : Only if the hole is at the end of OST object so that kms is too
2664 small. Fix is to increase kms accordingly in ll_nopage.
2667 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2669 Description: writes to a missing object would leak memory on the OST
2670 Details : If there is an inconsistency between the MDS and OST filesystems,
2671 such that the MDS references an object that doesn't exist, writes
2672 to that object will leak memory due to incorrect cleanup in the
2673 error handling path, eventually running out of memory on the OST.
2678 Description: Creating too long symlink causes lustre errors
2679 Details : Check symlink and name lengths before sending requests to MDS.
2682 Frequency : only if flock is enabled (not on by default)
2684 Description: posix locks not released on fd closure on 2.6.9+
2685 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2686 kernels, this caused such locks not to be released on fd close and
2687 then assertions on fs unmount about still used locks.
2690 Frequency : MDS failover only, very rarely
2692 Description: clients may get ASSERTION(granted_lock != NULL)
2693 Details : When request was taking a long time, and a client was resending
2694 a getattr by name lock request. The were multiple lock requests
2695 with the same client lock handle and
2696 mds_getattr_name->fixup_handle_for_resent_request found one of the
2697 lock handles but later failed with ASSERTION(granted_lock != NULL).
2702 Description: handle->h_buffer_credits > 0, assertion failure
2703 Details : h_buffer_credits is zero after truncate, causing assertion
2704 failure. This patch extends the transaction or creates a new
2708 Frequency : NFS re-export or patchless client
2709 Bugzilla : 11179, 10796
2710 Description: Crash on NFS re-export node (__d_move)
2711 Details : We do not want to hash the dentry if we don't have a lock.
2712 But if this dentry is later used in d_move, we'd hit uninitialised
2713 list head d_hash, so we just do this to init d_hash field but
2714 leave dentry unhashed.
2717 Frequency : NFS re-export or patchless client
2719 Description: NFS exports has problem with symbolic link
2720 Details : lustre client didn't properly install dentry when re-exported
2721 to NFS or running patchless client.
2724 Frequency : NFS re-export or patchless client
2726 Description: Various nfs/patchless fixes.
2727 Details : fixes reuse disconected alias for lookup process - this fixes
2728 warning "find_exported_dentry: npd != pd",
2729 fix permission error with open files at nfs.
2730 fix apply umask when do revalidate.
2733 Frequency : occasional
2735 Description: Crash on NFS re-export node
2736 Details : calling clear_page() on the wrong pointer triggered oops in
2737 generic_mapping_read().
2740 Frequency : rarely, using O_DIRECT IO
2742 Description: unaligned directio crashes client with LASSERT
2743 Details : check for unaligned buffers before trying any requests.
2746 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2748 Description: stale data returned from RAID cache
2749 Details : If only a small amount of IO is done to the RAID device before
2750 reading it again it is possible to get stale data from the RAID
2751 cache instead of reading it from disk.
2754 Frequency : always for sles10 kernel
2756 Description: sles10 support
2757 Details : ll_follow_link: compile fixes and using of nd_set_link
2758 under newer kernels.
2761 Frequency : depends on arch, kernel and compiler version, always on sles10
2764 Description: recursive or deep enough symlinks cause stack overflow
2765 Details : getting rid of large stack-allocated variable in
2769 Frequency : depends on hardware
2771 Description: lustre write performance loss in the SLES10 kernel
2772 Details : the performance loss is caused by using of write barriers in the
2773 ext3 code. The SLES10 kernel turns barrier support on by
2774 default. The fix is to undo that change for ldiskfs.
2776 ------------------------------------------------------------------------------
2778 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2780 * Support for kernels:
2781 2.6.9-42.0.3EL (RHEL 4)
2782 2.6.5-7.276 (SLES 9)
2783 2.4.21-47.0.1.EL (RHEL 3)
2784 2.6.12.6 vanilla (kernel.org)
2788 Frequency : quota enabled and large files being deleted
2790 Description: releasing more than 4GB of quota at once hangs OST
2791 Details : If a user deletes more than 4GB of files on a single OST it
2792 will cause the OST to spin in an infinite loop. Release
2793 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2798 Description: statfs data retrieved from /proc may be stale or zero
2799 Details : When reading per-device statfs data from /proc, in the
2800 {kbytes,files}_{total,free,avail} files, it may appear
2801 as zero or be out of date.
2804 Frequency : systems with MD RAID1 external journal devices
2806 Description: lconf's call to blkid is confused by RAID1 journal devices
2807 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2808 of returning all block devices that match the journal UUID.
2811 Frequency : always, for aggregate stripe size over 4GB
2813 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2814 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2815 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2818 Frequency : always if "lfs find" used on a local file/directory
2820 Description: "lfs find" segfaults if used on a local file/directory
2821 Details : The case where a directory component was not specified wasn't
2822 handled correctly. Handle this properly.
2825 Frequency : always on ppc64
2827 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2828 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2829 perform endianness conversion.
2832 Frequency : rarely (truncate to non-zero file size after write under load)
2833 Bugzilla : 10730, 10687
2834 Description: Files padded with zeros to next 4K multiple
2835 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2836 it is possible that files that are truncated to a non-zero size
2837 immediately after being written are filled with zero bytes beyond
2838 the truncated size. No file data is lost.
2840 Severity : enhancement
2842 Description: Allow recovery/failover for liblustre clients.
2843 Details : liblustre clients were unaware of failover configurations until
2846 Severity : enhancement
2848 Description: user file locks should fail when not mounting with flock option
2849 Details : Set up an error-returning stub in ll_file_operations.lock field
2850 to prevent incorrect behaviour when client is mounted without
2851 flock option. Also, set up properly f_op->flock field for
2855 Frequency : always on ia64
2857 Description: "lfs df" loops on printing out MDS statfs information
2858 Details : The obd_ioctl_data was not initialized and in some systems
2859 this caused a failure during the ioctl that did not return
2860 an error. Initialize the struct and return an error on failure.
2863 Frequency : SLES 9 only
2865 Description: Error of copying files with lustre special EAs as root
2866 Details : Client side always return success for setxattr call for lustre
2867 special xattr (currently only "trusted.lov").
2870 Frequency : rarely on clusters with both ia64+i386 clients
2872 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2873 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2874 clients are doing concurrent writes to the same file it is
2875 possible that the ia64 clients may LASSERT because the OST
2876 extent locks are not PAGE_SIZE aligned. Ensure that grown
2877 locks are always aligned on the request boundary.
2880 Frequency : specific use, occasional
2882 Description: Overwriting in use executable truncates on-disk binary image
2883 Details : If one node attempts to overwrite an executable in use by
2884 another node, we now correctly return ETXTBSY instead of
2885 truncating the file.
2887 Severity : enhancement
2889 Description: Async OSC create to avoid the blocking unnecessarily.
2890 Details : If a OST has no remain object, system will block on the creating
2891 when need to create a new object on this OST. Now, ways use
2892 pre-created objects when available, instead of blocking on an
2893 empty osc while others are not empty. If we must block, we block
2894 for the shortest possible period of time.
2899 Description: chmod on Lustre root is propagated to other clients
2900 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2901 invalid by the follow_mount time.
2904 Frequency : liblustre clients only
2906 Description: Race in 'instant cancel' lock handling could lead to such locks
2907 never to be granted in case of SMP MDS
2908 Details : Do not destroy not yet granted but cbpending locks in
2912 Frequency : replay/resend of open
2914 Description: non null lock assetion failure in mds_intent_policy
2915 Details : Trying to replay/resend lockless open requests resulted in
2916 mds_open() returning 0 with no lock. Now it sets a flag if
2917 a lock is going to be returned.
2919 Severity : enhancement
2921 Description: Checksum enhancements
2922 Details : New checksum enhancements allow for resending RPCs that failed
2925 Severity : enhancement
2927 Description: Tunables on number of dirty pages in cacche
2928 Details : Allow to set limit on number of dirty pages cached.
2933 Description: client crash on unmount - lock still has references
2934 Details : In some error handling cases it was possible to leak a lock
2935 reference on a client while accessing a file. This was not
2936 harmful to the client during operation, but would cause the
2937 client to crash when the filesystem is unmounted.
2940 Frequency : specific case, rare
2942 Description: ETXTBSY on mds though file not in use by client
2943 Details : ETXTBSY is no longer incorrectly returned when attempting to
2944 chmod or chown a directory that the user previously tried to
2945 execute or a currently-executing binary.
2948 Frequency : extremely rare except on liblustre-based clients
2950 Description: Lustre space not freed when files are deleted
2951 Details : Clean up open-unlinked files after client eviction. Previously
2952 the unlink was skipped and the files remained as orphans.
2957 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2958 Details : In some cases it was possible to send a blocking callback to
2959 a client doing a glimpse, even though that client didn't get
2960 a lock granted. When the glimpse lock is cancelled on the OST
2961 the freed lock is left on the waiting list and corrupted the list.
2964 Frequency : all core dumps
2966 Description: Broke core dumps to lustre
2967 Details : Negative dentry may be unhashed if parent does not have UPDATE
2968 lock, but some callers, e.g. do_coredump, expect dentry to be
2969 hashed after successful create, hash it in ll_create_it.
2971 ------------------------------------------------------------------------------
2973 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2975 * Support for kernels:
2976 2.6.9-42.0.2.EL (RHEL 4)
2977 2.6.5-7.276 (SLES 9)
2978 2.4.21-40.EL (RHEL 3)
2979 2.6.12.6 vanilla (kernel.org)
2983 Frequency : always on RHEL 3
2985 Description: Number of open files grows over time
2986 Details : The number of open files grows over time, whether or not
2987 Lustre is started. This was due to a filp leak introduced
2988 by one of our kernel patches.
2990 ------------------------------------------------------------------------------
2992 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2994 * Support for kernels:
2995 2.6.9-42.EL (RHEL 4)
2996 2.6.5-7.276 (SLES 9)
2997 2.4.21-40.EL (RHEL 3)
2998 2.6.12.6 vanilla (kernel.org)
3003 Bugzilla : 5719, 9635, 9792, 9684
3004 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3005 Details : If a server is under heavy load and cannot reply to new
3006 connection requests before the client resends the (re)connect,
3007 the connection handling code can behave badly if two service
3008 threads are concurrently handing separate (re)connections from
3009 the same client. Add better locking to the connection handling
3010 code, and ensure that only a single connection will be processed
3011 for a given client UUID, even if the lock is dropped.
3013 Severity : enhancement
3015 Description: add TCP zero-copy support to kernel
3016 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3017 sends if the hardware supports scatter-gather and checksumming.
3018 This allows socklnd to do client-write and server-read more
3019 efficiently and reduce CPU utilization from skbuf copying.
3022 Frequency : only if NFS exporting from client
3024 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3025 Details : If setting attributes on a file created under NFS that had
3026 never been opened it would be possible to oops the client
3027 if the file had no objects.
3030 Frequency : always for liblustre
3032 Description: liblustre client does MDS+OSTs setattr RPC for each write
3033 Details : When doing a write from a liblustre client, the client
3034 incorrectly issued an RPC to the MDS and each OST the file was
3035 striped over in order to update the timestamps. When writing
3036 with small chunks and many clients this could overwhelm the MDS
3037 with RPCs. In all cases it would slow down the write because
3038 these RPCs are unnecessary.
3040 Severity : enhancement
3042 Description: allow number of MDS service threads to be changed at module load
3043 Details : It is now possible to change the number of MDS service threads
3044 running. Adding "options mds mds_num_threads={N}" to the MDS's
3045 /etc/modprobe.conf will set the number of threads for the next
3046 time Lustre is restarted (assuming the "mds" module is also
3047 reloaded at that time). The default number of threads will
3048 stay the same, 32 for most systems.
3053 Description: OST crash if filesystem is unformatted or corrupt
3054 Details : If an OST is started on a device that has never been formatted
3055 or if the filesystem is corrupt and cannot even mount then the
3056 error handling cleanup routines would dereference a NULL pointer.
3061 Description: NULL pointer deref in llap_from_page.
3062 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3063 incorrect) mapping if the page was truncated/reclaimed while it was
3064 searched for. Check for this condition and skip such pages when
3065 doing readahead. Introduce extra check to llap_from_page() to
3066 verify page->mapping->host is non-NULL (so page is not anonymous).
3069 Frequency : Sometimes when using sys_sendfile
3071 Description: "page not covered by a lock" warnings from ll_readpage
3072 Details : sendfile called ll_readpage without right page locks present.
3073 Now we introduced ll_file_sendfile that does necessary locking
3074 around call to generic_file_sendfile() much like we do in
3078 Frequency : with certain MDS communication failures at client mount time
3080 Description: NULL pointer deref after failed client mount
3081 Details : a client connection request may delayed by the network layer
3082 and not be sent until after the PTLRPC layer has timed out the
3083 request. If the client fails the mount immediately it will try
3084 to clean up before the network times out the request. Add a
3085 reference from the request import to the obd device and delay
3086 the cleanup until the network drops the request.
3089 Frequency : occasionally during client (re)connect
3091 Description: assertion failure during client (re)connect
3092 Details : processing a client connection request may be delayed by the
3093 client or server longer than the client connect timeout. This
3094 causes the client to resend the connection request. If the
3095 original connection request is replied in this interval, the
3096 client may trip an assertion failure in ptlrpc_connect_interpret()
3097 which thought it would be the only running connect process.
3100 Frequency : only with obd_echo servers and clients that are rebooted
3102 Description: kernel BUG accessing uninitialized data structure
3103 Details : When running an obd_echo server it did not start the ping_evictor
3104 thread, and when a client was evicted an uninitialized data
3105 structure was accessed. Start the ping_evictor in the RPC
3106 service startup instead of the OBD startup.
3108 Severity : enhancement
3109 Bugzilla : 10193 (patchless)
3110 Description: Remove dependency on various unexported kernel interfaces.
3111 Details : No longer need reparent_to_init, exit_mm, exit_files,
3112 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3115 Frequency : rare (only users of deprecated and unsupported LDAP config)
3117 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3118 Details : LDAP apparently contains 'lustreName' attributes instead of
3119 'name'. A simple remapping of the name is sufficient.
3122 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3124 Description: waiting_locks_callback trips kernel BUG if client is evicted
3125 Details : Running with the dump_on_timeout debug flag turned on makes
3126 it possible that the waiting_locks_callback() can try to dump
3127 the Lustre kernel debug logs from an interrupt handler. Defer
3128 this log dumping to the expired_lock_main() thread.
3130 Severity : enhancement
3132 Description: Support NFS exporting on 2.6 kernels.
3133 Details : Implement non-rawops metadata methods for NFS server to use without
3134 changing NFS server code.
3137 Frequency : very rare (synthetic metadata workload only)
3139 Description: two racing renames might cause an MDS thread to deadlock
3140 Details : Running the "racer" program may cause one MDS thread to rename
3141 a file from being the source of a rename to being the target of
3142 a rename at exactly the same time that another thread is doing
3143 so, and the second thread has already enqueued these locks after
3144 doing a lookup of the target and is trying to relock them in
3145 order. Ensure that we don't try to re-lock the same resource.
3148 Frequency : only very large systems with liblustre clients
3150 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3151 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3152 all OSTs in parallel. This allows the eviction of stale liblustre
3153 clients to proceed much faster than if they were done in series,
3154 and also offers similar improvements for other set_info RPCs.
3159 Description: excessive CPU usage during initial read phase on client
3160 Details : During the initial read phase on a client, it would agressively
3161 retry readahead on the file, consuming too much CPU and impacting
3162 performance (since 1.4.5.8). Improve the readahead algorithm
3163 to avoid this, and also improve some other common cases (read
3164 of small files in particular, where "small" is files smaller than
3165 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3170 Description: MDS crash when receiving packet with unknown intent.
3171 Details : Do not LBUG in unknown intent case, just return -EFAULT
3173 Severity : enhancement
3174 Bugzilla : 9293, 9385
3175 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3176 Details : Do not serialize getattr (non-intent version) and statfs.
3179 Frequency : occasional, when OST network is overloaded/intermittent
3181 Description: client evicted by OST after bulk IO timeout
3182 Details : If a client sends a bulk IO request (read or write) the OST
3183 may evict the client if it is unresposive to its data GET/PUT
3184 request. This is incorrect if the network is overloaded (takes
3185 too long to transfer the RPC data) or dropped the OST GET/PUT
3186 request. There is no need to evict the client at all, since
3187 the pinger and/or lock callbacks will handle this, and the
3188 client can restart the bulk request.
3191 Frequency : Always when mmapping file with no objects
3193 Description: client crashes when mmapping file with no objects
3194 Details : Check that we actually have objects in a file before doing any
3195 operations on objects in ll_vm_open, ll_vm_close and
3201 Description: Request leak when working with deleted CWD
3202 Details : Introduce advanced request refcount tracking for requests
3203 referenced from lustre intent.
3205 Severity : Enhancement
3207 Description: Cache open file handles on client.
3208 Details : MDS now will return special lock along with openhandle, if
3209 requested and client is allowed to hold openhandle, even if unused,
3210 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3211 closing files for every read/write openration.
3213 Severity : Enhancement
3215 Description: Cache open negative dentries on client when possible.
3216 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3217 negative dentries on lock revocation.
3222 Description: Remounting a client read-only wasn't possible with a zconf mount
3223 Details : It wasn't possible to remount a client read-only with llmount.
3225 Severity : enhancement
3226 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3227 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3228 MPICH is included to add Lustre-specific ADIO interfaces.
3229 This is based closely on the UFS ADIO layer and only differs
3230 in file creation, in order to allow the OST striping to be set.
3231 This is user-contributed code and not supported by CFS.
3236 Description: extended inode attributes (immutable, append-only) work improperly
3237 when 2.4 and 2.6 kernels are used on client/server or vice versa
3238 Details : Introduce kernel-independent values for these flags.
3240 Severity : enhancement
3243 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3244 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3245 granularity. Now, Fractional MB granularity is be supported,
3246 this is very useful for low memory system.
3248 Severity : enhancement
3250 Description: Getattr by fid
3251 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3252 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3253 have locks and avoids getattr rpc requests for it on every stat.
3256 Frequency : Always, for filesystems larger than 2TB
3258 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3259 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3260 filesystems larger than 2TB to be mounted with mballoc enabled.
3263 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3265 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3266 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3267 filesystems larger than 2TB is mounted then the kernel will
3268 silently corrupt the start of the filesystem. CONFIG_LBD is
3269 enabled for all CFS-supported kernels, but the possibility of
3270 this happening with a modified kernel config exists.
3272 Severity : enhancement
3274 Description: add client O_DIRECT support for 2.6 kernels
3275 Details : It is now possible to do O_DIRECT reads and writes to files
3276 in the Lustre client mountpoint on 2.6 kernel clients.
3278 Severity : enhancement
3280 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3281 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3282 parallel, not waiting for response from every OST before sending
3283 a rpc to the next OST.
3288 Description: setattr vs write race when updating file timestamps
3289 Details : Client processes that update a file timestamp into the past
3290 right after writing to the file (e.g. tar) it is possible that
3291 the updated file modification time can be reset to the current
3292 time due to a race between processing the setattr and write RPC.
3294 Severity : enhancement
3296 Description: Bring 'lfs find' closer in line with regular Linux find.
3297 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3298 -print0 options and obtains all the needed info through the lustre
3301 Severity : enhancement
3303 Description: support up to 1024 configured devices on one node
3304 Details : change obd_dev array from statically allocated to dynamically
3305 allocated structs as they are first used to reduce memory usage
3310 Description: Flush dirty partially truncated pages during truncate
3311 Details : Immediatelly flush partially truncated pages in filter_setattr,
3312 this way we completely avoid having any pages in page cache on OST
3313 and can retire ugly workarounds during writes to flush such pages.
3318 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3319 Details : protect inode from truncation within vfs_unlink() context
3320 just take a reference before calling vfs_unlink() and release it
3321 when parent's i_sem is free.
3326 Description: last_id value checked outside lock on OST caused LASSERT failure
3327 Details : If there were multiple MDS->OST object precreate requests in
3328 flight, it was possible that the OST's last object id was checked
3329 outside a lock and incorrectly tripped an assertion. Move checks
3330 inside locks, and discard old precreate requests.
3333 Frequency : always, if extents are used on OSTs
3335 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3336 Details : OSTs using the extents format would not zero the high 16 bits of
3337 the index physical block number. This is not a problem for any
3338 OST filesystems smaller than 16TB, and no kernels support ext3
3339 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3340 new/modified files) and can be fixed for existing filesystems
3341 with e2fsprogs-1.39-cfs1.
3346 Description: import connection selection may be incorrect if timer wraps
3347 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3348 import connections to be ignored if the 32-bit jiffies counter
3349 wraps. Use a 64-bit jiffies counter.
3352 Frequency : very large clusters immediately after boot
3354 Description: LNET request buffers exhausted under heavy short-term load
3355 Details : If a large number of client requests are generated on a service
3356 that has previously never seen so many requests it is possible
3357 that the request buffer growth cannot keep up with the spike in
3358 demand. Instead of dropping incoming requests, they are held in
3359 the LND until the RPC service can accept more requests.
3362 Frequency : Sometimes during replay
3364 Description: Assertion failure in ll_local_open after replay.
3365 Details : If replay happened on an open request reply before we were able
3366 to set replay handler, reply will become not swabbed tripping the
3367 assertion in ll_local_open. Now we set the handler right after
3368 recognising of open request
3371 Frequency : very rare
3373 Description: kernel reports "badness in vsnprintf"
3374 Details : Reading from the "recovery_status" /proc file in small chunks
3375 may cause a negative length in lprocfs_obd_rd_recovery_status()
3376 call to vsnprintf() (which is otherwise harmless). Exit early
3377 if there is no more space in the output buffer.
3379 Severity : enhancement
3381 Description: clear OBD RPC statistics by writing to them
3382 Details : It is now possible to clear the OBD RPC statistics by writing
3383 to the "stats" file.
3388 Description: Client mtime is not the same on different clients after utimes
3389 Details : In some cases, the client was using the utimes() syscall on
3390 a file cached on another node. The clients now validate the
3391 ctime from the MDS + OSTs to determine which one is right.
3396 Description: Inability to activate failout mode
3397 Details : lconf script incorrectly assumed that in pythong string's numeric
3398 value is used in comparisons.
3401 Frequency : always with multiple stripes per file
3403 Description: Inefficient object allocation for mutli-stripe files
3404 Details : When selecting which OSTs to stripe files over, for files with
3405 a stripe count that divides evenly into the number of OSTs,
3406 the MDS is always picking the same starting OST for each file.
3407 Return the OST selection heuristic to the original design.
3412 Description: mount failures may take full timeout to return an error
3413 Details : Under some heavy load conditions it is possible that a
3414 failed mount can wait for the full obd_timeout interval,
3415 possibly several minutes, before reporting an error.
3416 Instead return an error as soon as the status is known.
3418 Frequency : quota enabled and large files being deleted
3420 Description: releasing more than 4GB of quota at once hangs OST
3421 Details : If a user deletes more than 4GB of files on a single OST it
3422 will cause the OST to spin in an infinite loop. Release
3423 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3428 Description: statfs data retrieved from /proc may be stale or zero
3429 Details : When reading per-device statfs data from /proc, in the
3430 {kbytes,files}_{total,free,avail} files, it may appear
3431 as zero or be out of date.
3434 Frequency : systems with MD RAID1 external journal devices
3436 Description: lconf's call to blkid is confused by RAID1 journal devices
3437 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3438 of returning all block devices that match the journal UUID.
3441 Frequency : always, for aggregate stripe size over 4GB
3443 Description: assertion fails when trying to use 4GB stripe size
3444 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3445 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3448 Frequency : always on ppc64
3450 Description: the first write on an ext3 filesystem with mballoc got stuck
3451 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3452 perform endianness conversion.
3454 ------------------------------------------------------------------------------
3456 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3458 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3459 INTEROPERATE with older versions automatically. Please read the
3460 user documentation before upgrading any part of a live system.
3461 * WARNING: Lustre networking configuration changes are required with
3462 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3465 * Support for kernels:
3466 2.6.9-22.0.2.EL (RHEL 4)
3467 2.6.5-7.244 (SLES 9)
3468 2.6.12.6 vanilla (kernel.org)
3471 Severity : enhancement
3472 Bugzilla : 7981/8208
3473 Description: Introduced Lustre Networking (LNET)
3474 Details : LNET is new networking infrastructure for Lustre, it includes
3475 a reorganized network configuration mode (see the user
3476 documentation for full details) as well as support for routing
3477 between different network fabrics. Lustre Networking Devices
3478 (LNDs) for the supported network fabrics have also been
3479 created for this new infrastructure.
3481 Severity : enhancement
3482 Description: Introduced Access control lists
3483 Details : clients can set ACLs on files and directories in order to have
3484 more fine-grained permissions than the standard Unix UGO+RWX.
3485 The MDS must be started with the "-o acl" mount option.
3487 Severity : enhancement
3488 Description: Introduced filesystem quotas
3489 Details : Administrators may now establish per-user quotas on the
3492 Severity : enhancement
3494 Description: Configuration change for the XT3
3495 The PTLLND is now used to run Lustre over Portals on the XT3
3496 The configure option(s) --with-cray-portals are no longer used.
3497 Rather --with-portals=<path-to-portals-includes> is used to
3498 enable building on the XT3. In addition to enable XT3 specific
3499 features the option --enable-cray-xt3 must be used.
3504 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3505 Details : A race between allocating a new llcd and re-getting the llcd_lock
3506 allowed another thread to grab newly-allocated llcd.
3508 Severity : enhancement
3510 Description: 2.6 OST async journal commit and locking fix to improve performance
3511 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3512 2.6 kernels are now async as they already were in 2.4 kernels so
3513 that they can commit concurrently with the network bulk transfer.
3514 For block-allocated files the filter allocation semaphore is held
3515 to avoid filesystem fragmentation during allocation. BKL lock
3516 removed for 2.6 xattr operations where it is no longer needed.
3521 Description: lconf incorrectly determined whether two IP networks could talk
3522 Details : In some more complicated routing and multiple-network
3523 configurations, lconf will avoid trying to make a network
3524 connection to a disjoint part of the IP space. It was doing the
3525 math incorrectly for one set of cases.
3530 Description: Fix for potential infinite loop processing records in an llog.
3531 Details : If an llog record is corrupted/zeroed, it is possible to loop
3532 forever in llog_process(). Validate the llog record length
3533 and skip the remainder of the block on error.
3536 Frequency : occasional (liblustre only)
3538 Description: liblustre could not open files whose last component is a symlink
3539 Details : sysio_path_walk() would incorrectly pass the open intent to
3540 intermediate path components.
3543 Frequency : rare (liblustre only with non-standard tuning)
3544 Bugzilla : 7201 (7350)
3545 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3546 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3547 after it has been released, only do this for referenced locks
3549 Severity : enhancement
3551 Description: specifying an (invalid) directory default stripe_size of -1
3552 would reset the directory default striping
3553 Details : stripe_size -1 was used internally to signal directory stripe
3554 removal, now use "all default" to signal dir stripe removal
3555 as a directory striping of "all default" is not useful
3558 Frequency : common for large clusters running liblustre clients
3560 Description: doing an ls when liblustre clients are running is slow
3561 Details : sending a glimpse AST to a liblustre client waits for every AST
3562 to time out, as liblustre clients will not respond. Since they
3563 cannot cache data we refresh the OST lock LVB from disk instead.
3565 Severity : enhancement
3567 Description: doing an ls at the same time as file IO can be slow
3568 Details : enqueue and other "small" requests can be blocked behind many
3569 large IO requests. Create a new OST IO portal for non-IO
3570 requests so they can be processed faster.
3573 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3575 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3576 created file with mode 000
3577 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3578 allow an ftruncate() as a non-root user to a file with mode 000.
3579 Lustre now respects this flag to disable mode checks when
3580 truncating a file owned by the user
3583 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3586 Description: Revoking locks from clients that went dead or catatonic might take
3588 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3589 cancellation of such locks instant on servers without waiting for
3590 any reply from clients. Clients drops these locks when cancel
3591 notification from server is received without replying.
3594 Frequency : liblustre-only, when liblustre client dies or becomes busy
3596 Description: Doing ls on Linux clients can take a long time with active
3598 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3599 granting such locks to it in the first place if possible. Locks
3600 are taken by proxy on the OST during the read or write and
3601 dropped immediately afterward. Add connect flags handling, do
3602 not grant locks to liblustre clients for glimpse ASTs.
3604 Severity : enhancement
3606 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3607 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3608 sequential reads and behaves badly with applications doing random
3609 reads. Improve it by reading ahead at least read region, and
3610 avoiding excessive large RPC for small reads.
3612 Severity : enhancement
3614 Description: Creating more than 1000 files for a single job may cause a load
3615 imbalance on the OSTs if there are also a large number of OSTs.
3616 Details : qos_prep_create() uses an OST index reseed value that is an
3617 even multiple of the number of available OSTs so that if the
3618 reseed happens in the middle of the object allocation it will
3619 still utilize the OSTs as uniformly as possible.
3624 Description: OST or MDS may oops in ping_evictor_main()
3625 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3626 but doesn't restart at beginning of obd_exports_timed list
3629 Severity : enhancement
3631 Description: improve by-nid export eviction on the MDS and OST
3632 Details : allow multiple exports with the same NID to be evicted at one
3633 time without re-searching the exports list.
3636 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3638 Description: MDS may oops in groups_free()
3639 Details : in rare race conditions a newly allocated group_info struct is
3640 freed again, and this can be NULL. The 2.4 compatibility code
3641 for groups_free() checked for a NULL pointer, but 2.6 did not.
3644 Frequency : common for liblustre clients doing little filesystem IO
3645 Bugzilla : 9352, 7313
3646 Description: server may evict liblustre clients accessing contended locks
3647 Details : if a client is granted a lock or receives a completion AST
3648 with a blocking AST already set it would not reply to the AST
3649 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3652 Frequency : lfs setstripe, only systems with more than 160 OSTs
3654 Description: unable to set striping with a starting offset beyond OST 160
3655 Details : llapi_create_file() incorrectly limited the starting stripe
3656 index to the maximum single-file stripe count.
3659 Frequency : LDAP users only
3661 Description: lconf did not handle in-kernel recovery with LDAP properly
3662 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3664 Severity : enhancement
3666 Description: bind OST threads to NUMA nodes to improve performance
3667 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3668 node and do their allocations there to localize memory access
3670 Severity : enhancement
3672 Description: llmount can determine client NID directly from Myrinet (GM)
3673 Details : the client NID code from gmnalnid was moved directly into
3674 llmount, removing the need to use this or specifying the
3675 client NID explicitly when mounting GM clients with zeroconf
3678 Frequency : if client is started with down MDS
3680 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3681 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3682 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3683 ptlrpc_import_delayed_req() because we didn't check if the
3684 request was interrupted, and we also didn't break out of the
3685 event loop if there was no timeout
3690 Description: data loss during non-page-aligned writes to a single file from
3691 both multiple nodes and multiple threads on one node at same time
3692 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3693 inconsistency led to false short-reads, that were cached and later
3694 used by ->prepare_write() to fill in partially written page,
3695 leading to data loss.
3698 Frequency : always, if lconf --abort_recovery used
3700 Description: lconf --abort_recovery fails with 'Operation not supported'
3701 Details : lconf was attempting to abort recovery on the MDT device and not
3704 Severity : enhancement
3706 Description: remove cleanup logs
3707 Details : replace lconf-generated cleanup logs with lustre internal
3708 cleanup routines. Eliminates the need for client-cleanup and
3711 Severity : enhancement
3713 Description: add support for EAs (user and system) on lustre filesystems
3714 Details : it is now possible to store extended attributes in the Lustre
3715 client filesystem, and with the user_xattr mount option it
3716 is possible to allow users to store EAs on their files also
3718 Severity : enhancement
3720 Description: Add possibility (config option) to show minimal available OST free
3722 Details : When compiled with --enable-mindf configure option, statfs(2)
3723 (and so, df) will return least minimal free space available from
3724 all OSTs as amount of free space on FS, instead of summary of
3725 free spaces of all OSTs.
3727 Severity : enhancement
3729 Description: do not expand extent locks acquired on OST-side
3730 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3731 by server: they are not cached anyway.
3734 Frequency : when mmap is used/binaries executed from Lustre
3736 Description: Unmmap pages before throwing them away from read cache.
3737 Details : llap_shrink cache now attempts to unmap pages before discarding
3738 them (if unmapping failed - do not discard). SLES9 kernel has
3739 extra checks that trigger if this unmapping is not done first.
3744 Description: lconf didn't resolve symlinks before checking to see whether a
3745 given mountpoint was already in use
3748 Frequency : when migrating failover services
3749 Bugzilla : 6395, 9514
3750 Description: When migrating a subset of services from a node (e.g. failback
3751 from a failover service node) the remaining services would
3752 time out and evict clients.
3753 Details : lconf --force (implied by --failover) sets the global obd_timeout
3754 to 5 seconds in order to quickly disconnect, but this caused
3755 other RPCs to time out too quickly. Do not change the global
3756 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3758 Severity : enhancement
3759 Frequency : if MDS is started with down OST
3760 Bugzilla : 9439,5706
3761 Description: Allow startup/shutdown of an MDS without depending on the
3762 availability of the OSTs.
3763 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3764 Add appropriate locking and lov-osc refcounts for safe
3765 cleaning. Add osc abort_inflight calls in case the
3766 synchronize never started.
3769 Frequency : occasional (Cray XT3 only)
3771 Description: root not authorized to access files in CRAY_PORTALS environment
3772 Details : The client process capabilities were not honoured on the MDS in
3773 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3774 been accessed by an authorized user then root was able to access
3775 the file on the local client also. The root user capabilities
3776 are now allowed on the MDS, as this environment has secure UID.
3779 Frequency : occasional
3781 Description: ldiskfs "too long searching" message happens too often
3782 Details : A debugging message (otherwise harmless) prints too often on
3783 the OST console. This has been reduced to only happen when
3784 there are fragmentation problems on the filesystem.
3789 Description: Division by zero in statfs when all OSCs are inactive
3790 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3791 lov_statfs divides by value returned by lov_get_stripecnt().
3795 Bugzilla : 9489, 3273
3796 Description: First write from each client to each OST was only 4kB in size,
3797 to initialize client writeback cache, which caused sub-optimal
3798 RPCs and poor layout on disk for the first writen file.
3799 Details : Clients now request an initial cache grant at (re)connect time
3800 and so that they can start streaming writes to the cache right
3801 away and always do full-sized RPCs if there is enough data.
3802 If the OST is rebooted the client also re-establishes its grant
3803 so that client cached writes will be honoured under the grant.
3808 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3809 Details : Now I/O RPCs go to different portal number and (presumably) fast
3810 lock requests (and glimses) and other RPCs get their own service
3811 threads pool that should be able to service those RPCs
3814 Severity : enhancement
3816 Description: Ability to exchange lustre version between client and servers and
3817 issue warnings at client side if client is too old. Also for
3818 liblustre clients there is ability to refuse connection of too old
3820 Details : New 'version' field is added to connect data structure that is
3821 filled with version info. That info is later checked by server and
3825 Frequency : rare, liblustre only.
3826 Bugzilla : 9296, 9581
3827 Description: Two simultaneous writes from liblustre at offset within same page
3828 might proceed at the same time overwriting eachother with stale
3830 Details : I/O lock withing llu_file_prwv was released too early, before data
3831 actually was hitting the wire. Extended lock-holding time until
3832 server acknowledges receiving data.
3835 Frequency : extremely rare. Never observed in practice.
3837 Description: avoid generating lustre_handle cookie of 0.
3838 Details : class_handle_hash() generates handle cookies by incrementing
3839 global counter, and can hit 0 occasionaly (this is unlikely, but
3840 not impossible, because initial value of cookie counter is
3841 selected randonly). Value of 0 is used as a sentinel meaning
3842 "unassigned handle" --- avoid it. Also coalesce two critical
3843 sections in this function into one.
3845 Severity : enhancement
3847 Description: allow liblustre clients to delegate truncate locking to OST
3848 Details : To avoid overhead of locking, liblustre client instructs OST to
3849 take extent lock in ost_punch() on client's behalf. New connection
3850 flag is added to handle backward compatibility.
3852 Severity : enhancement
3853 Bugzilla : 4928, 7341, 9758
3854 Description: allow number of OST service threads to be specified
3855 Details : a module parameter allows the number of OST service threads
3856 to be specified via "options ost ost_num_threads={N}" in the
3857 OSS's /etc/modules.conf or /etc/modprobe.conf.
3861 Bugzilla : 6146, 9635, 9895
3862 Description: servers crash with bad pointer in target_handle_connect()
3863 Details : In rare cases when a client is reconnecting it was possible that
3864 the connection request was the last reference for that export.
3865 We would temporarily drop the export reference and get a new
3866 one, but this may have been the last reference and the export
3867 was just destroyed. Get new reference before dropping old one.
3869 Severity : enhancement
3870 Frequency : if client is started with failover MDS
3872 Description: Allow multiple MDS hostnames in the mount command
3873 Details : Try to read the configuration from all specified MDS
3874 hostnames during a client mount in case the "primary"
3877 Severity : enhancement
3879 Description: Stop sending data to evicted clients as soon as possible.
3880 Details : Check if the client we are about to send or are sending data to
3881 was evicted already. (Check is done every second of waiting,
3882 for which l_wait_event interface was extended to allow checking
3883 of exit condition at specified intervals).
3886 Frequency : rare, normally only when NFS exporting is done from client
3888 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3890 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3893 Frequency : always, if srand() is called before liblustre initialization
3895 Description: Liblustre uses system PRNG disturbing its usage by user application
3896 Details : Introduce internal to lustre fast and high-quality PRNG for
3897 lustre usage and make liblustre and some other places in generic
3898 lustre code to use it.
3900 Severity : enhancement
3901 Bugzilla : 9477, 9557, 9870
3902 Description: Verify that the MDS configuration logs are updated when xml is
3903 Details : Check if the .xml configuration logs are newer than the config
3904 logs stored on the MDS and report an error if this is the case.
3905 Request --write-conf, or allow starting with --old_conf.
3907 Severity : enhancement
3909 Description: Handle symlinks in the path when checking if Lustre is mounted.
3910 Details : Resolve intermediate symlinks when checking if a client has
3911 mounted a filesystem to avoid duplicate client mounts.
3916 Description: lconf can hit an error exception but still return success.
3917 Details : The lconf command catches the Command error exception at the top
3918 level script context and will exit with the associated exit
3919 status, but doesn't ensure that this exit status is non-zero.
3924 Description: failure of ptlrpc thread startup can cause oops
3925 Details : Starting a ptlrpc service thread can fail if there are a large
3926 number of threads or the server memory is very fragmented.
3927 Handle this without oopsing.
3930 Frequency : always, only if liblustre and non-default acceptor port was used
3932 Description: liblustre cannot connect to servers with non-default acceptor port
3933 Details : tcpnal_set_default_params() was not called and was therefore
3934 ignoring the environment varaible TCPNAL_PORT, as well as other
3935 TCPNAL_ environment variables
3940 Description: two objects could be created on the same OST for a single file
3941 Details : If an OST is down, in some cases it was possible to create two
3942 objects on a single OST for a single file. No problems other
3943 than potential performance impact and spurious error messages.
3947 Bugzilla : 5681, 9562
3948 Description: Client may oops in ll_unhash_aliases
3949 Details : Client dcache may become inconsistent in race condition.
3950 In some cases "getcwd" can fail if the current directory is
3956 Description: Inode refcounting problems in NFS export code
3957 Details : link_raw functions used to call d_instantiate without obtaining
3958 extra inode reference first.
3962 Bugzilla : 9942, 9903
3963 Description: Referencing freed requests leading to crash, memleaks with NFS.
3964 Details : We used to require that call to ll_revalidate_it was always
3965 followed by ll_lookup_it. Also with revalidate_special() it is
3966 possible to call ll_revalidate_it() twice for the same dentry
3967 even if first occurence returned success. This fix changes semantic
3968 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3969 reference on a request referred from the intent.
3970 ll_intent_release() then releases such a request.
3973 Frequency : rare, normally benchmark loads only
3975 Description: unlinked inodes were kept in memory on the client
3976 Details : If a client is repeatedly creating and unlinking files it
3977 can accumulate a lot of stale inodes in the inode slab cache.
3978 If there is no other client load running this can cause the
3979 client node to run out of memory. Instead flush old inodes
3980 from client cache that have the same inode number as a new inode.
3983 Frequency : SLES9 2.6.5 kernel and long filenames only
3984 Bugzilla : 9969, 10379
3985 Description: utime reports stale NFS file handle
3986 Details : SLES9 uses out-of-dentry names in some cases, which confused
3987 the lustre dentry revalidation. Change it to always use the
3991 Frequency : rare, unless heavy write-truncate concurrency is continuous
3992 Bugzilla : 4180, 6984, 7171, 9963, 9331
3993 Description: OST becomes very slow and/or deadlocked during object unlink
3994 Details : filter_destroy() was holding onto the parent directory lock
3995 while truncating+unlinking objects. For very large objects this
3996 may block other threads for a long time and slow overall OST
3997 responsiveness. It may also be possible to get a lock ordering
3998 deadlock in this case, or run out of journal credits because of
3999 the combined truncate+unlink. Solution is to do object truncate
4000 first in one transaction without parent lock, and then do the
4001 final unlink in a new transaction with the parent lock. This
4002 reduces the lock hold time dramatically.
4005 Frequency : rare, 2.4 kernels only
4007 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4008 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4009 before unmounting their filesystem in order to avoid deadlock.
4010 The kernel_locked() function in 2.4 kernels only checks whether
4011 the kernel lock is held, not whether it is this process that is
4012 holding it as 2.6 kernels do.
4017 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4018 Details : The client ptlrpc code may be trying to reconnect to a down
4019 server before a previous connection attempt has timed out.
4020 Increase the reconnect interval to be longer than the connection
4021 timeout interval to avoid sending duplicate connections to
4025 Frequency : echo_client brw_test command
4027 Description: fix echo_client to work with OST preallocated code
4028 Details : OST preallocation code (5137) didn't take echo_client IO path
4029 into account: echo_client calls filter methods outside of any
4030 OST thread and, hence, there is no per-thread preallocated
4031 pages and buffers to use. Solution: hijack pga pages for IO. As
4032 a byproduct, this avoids unnecessary data copying.
4036 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4037 Description: Client can oops in mdc_commit_close() after open replay
4038 Details : It was possible for the MDS to return an open request with no
4039 transaction number in mds_finish_transno() if the client was
4040 evicted, but without actually returning an error. Clients
4041 would later try to replay that open and may trip an assertion
4042 Simplify the client close codepath, and always return an error
4043 from the MDS in case the open is not successful.
4046 Frequency : rare, 2.6 OSTs only
4048 Description: OST may deadlock under high load on fragmented files
4049 Details : If there was a heavy load and highly-fragmented OST filesystems
4050 it was possible to have all the OST threads deadlock waiting on
4051 allocation of biovecs, because the biovecs were not released
4052 until the entire RPC IO was completed. Instead, release biovecs
4053 as soon as they are complete to ensure forward IO progress.
4055 Severity : enhancement
4057 Description: Support for specifying external journal device at mount
4058 Details : If an OST or MDS device is formatted with an external journal
4059 device, this device major/minor is stored in the ext3 superblock
4060 and may not be valid for failover. Allow detecting and
4061 specifying the external journal at mount time.
4066 Description: Mounting an MDS with pending unlinked files may cause oops
4067 Details : target_finish_recovery() calls mds_postrecov() which returned
4068 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4069 considers this an error and immediately begins cleaning up the
4070 lov, just after starting the mds_lov process
4072 Severity : enhancement
4074 Description: Implement 'lfs df' to report actual free space on per-OST basis
4075 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4076 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4077 usage of objects; '-h' to report in human readable format.
4079 ------------------------------------------------------------------------------
4081 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4088 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4089 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4090 at startup during orphan unlinking. Instead, skip buddy update
4091 before setup, as it will be regenerated after recovery is complete.
4096 Description: If an OST is inactive, its locks might reference stale inodes.
4097 Details : lov_change_cbdata() must iterate over all namespaces, even if
4098 they are inactive to clear inode references from the lock.
4100 Severity : enhancement
4101 Frequency : occasional, if non-standard max_dirty_mb used
4103 Description: Client will block write RPCs if not enough grant
4104 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4105 then the client will block writes while waiting for another RPC
4106 to complete instead of consuming its dirty limit. With change
4107 we get improved performance when max_dirty_mb is small.
4109 Severity : enhancement
4110 Bugzilla : 3389, 6253
4111 Description: Add support for supplementary groups on the MDS.
4112 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4113 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4114 lookups for user supplementary groups into a cache.
4118 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4119 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4120 for liblustre O_EXCL is enforced by MDS.
4124 Description: readdir never returns NULL in liblustre.
4125 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4126 dentry in struct dirent.
4130 Frequency : liblustre only, depends on application IO pattern
4131 Description: liblustre clients evicted if not contacting servers
4132 Details : Don't put liblustre clients into the ping_evictor list, so
4133 they will not be evicted by the pinger ever.
4135 Severity : enhancement
4137 Description: Add ability to evict clients by NID from MDS.
4138 Details : By echoing "nid:$NID" string into
4139 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4140 $NID would be instantly evicted from this MDS and from all active
4141 OSTs connected to it.
4145 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4146 Details : lookup_it_finish() used to query file size from OSTs that was not
4151 Description: service threads change working directory to that of init
4152 Details : Starting lustre service threads may pin the working directory
4153 of the parent thread, making that filesystem busy. Threads
4154 now change to the working directory of init to avoid this.
4158 Frequency : during shutdown only
4159 Description: shutdown with a failed MDS or OST can cause unmount to hang
4160 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4165 Frequency : 2.6 only
4166 Description: chmod/chown may include an extra supplementary group
4167 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4168 supplementary group and if none is specified this is used.
4171 Bugzilla : 5479 (6816)
4173 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4174 Details : If lookup is in progress on a file that is unlinked we might try
4175 to revalidate the inode and fail in revalidate after lookup is
4176 complete and ll_file_open() enqueues the open again but
4177 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4180 Frequency : always, if lconf --abort_recovery used
4182 Description: lconf --abort_recovery fails with 'Operation not supported'
4183 Details : lconf was attempting to abort recovery on the MDT device and not
4186 ------------------------------------------------------------------------------
4188 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4193 Frequency : rare (only unsupported configurations with a node running as an
4195 Bugzilla : 6514, 5137
4196 Description: Mounting a Lustre file system on a node running as an OST could
4198 Details : OSTs now preallocates memory needed to write out data at
4199 startup, instead of when needed, to avoid having to
4200 allocate memory in possibly low memory situations.
4201 Specifically, if the file system is mounted on on OST,
4202 memory pressure could force it to try to write out data,
4203 which it needed to allocate memory to do. Due to the low
4204 memory, it would be unable to do so and the node would
4205 become unresponsive.
4207 Severity : enhancement
4209 Description: Addition of lconf --service command line option
4210 Details : lconf now accepts a '--service <arg>' option, which is
4211 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4213 Severity : enhancement
4215 Description: Failover mode is now the default for OSTs.
4216 Details : By default, OSTs will now run in failover mode. To return to
4217 the old behaviour, add '--failout' to the lmc line for OSTs.
4219 Severity : enhancement
4221 Description: Health checks are now provided for MDS and OSTs
4222 Details : Additional detailed health check information on MSD and OSTs
4223 is now provided through the procfs health_check value.
4226 Frequency : occasional, depends on IO load
4228 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4229 after numerous create/delete cycles
4230 Details : The ext3 inode allocation policy would not allocate new inodes
4231 very well on the OSTs because there are no new directories
4232 being created. Instead we look for groups with free space if
4233 the parent directories are nearly full.
4238 Description: Network or server problems during mount may cause partially
4239 mounted clients instead of returning an error.
4240 Details : The config llog parsing code may overwrite the error return
4241 code during mount error handling, returning success instead
4247 Description: MDS can fail to allocate large reply buffers
4248 Details : After long uptimes the MDS can fail to allocate large reply
4249 buffers (e.g. zconf client mount config records) due to memory
4250 fragmentation or consumption by the buffer cache. Preallocate
4251 some large reply buffers so that these replies can be sent even
4252 under memory pressure.
4256 Frequency : rare (liblustre)
4257 Description: fsx running with liblustre complained that using truncate() to
4258 extend the file doesn't work. This patch corrects that issue.
4259 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4260 ATTR_SIZE and lsm use in llu_setattr_raw.
4264 Frequency : rare, only 2.6 kernels
4265 Description: Unusual file access patterns on the MDS may result in inode
4266 data being lost in very rare circumstances.
4267 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4268 code in the 2.6 kernel caused the inode and/or EA data not to
4269 be read from disk, causing single-file corruption.
4273 Frequency : rare, only 2.6 filesystems using extents
4274 Description: Heavy concurrent write and delete load may cause data corruption.
4275 Details : It was possible under high-load situations to have an extent
4276 metadata block in the block device cache from a just-unlinked
4277 file overwrite a newly-allocated data block. We now unmap any
4278 metadata buffers that alias just-allocated data blocks.
4282 Frequency : filesystems with default stripe_count larger than 77
4283 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4284 Details : lconf specifies an inode size of 4096 bytes when the default
4285 stripe_count is larger than 77. This conflicts with the default
4286 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4287 this case to avoid pinning too much memory for large EAs.
4289 ------------------------------------------------------------------------------
4291 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4296 Frequency : rare (extremely heavy IO load with hundreds of clients)
4298 Description: Client is evicted, gets IO error writing to file
4299 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4300 caused clients to be evicted for AST timeouts. The fixes in
4301 bug 5192 mean we no longer need to have such short AST timeouts
4302 so ldlm_timeout has been increased.
4305 Frequency : occasional during --force or --failover shutdown under load
4306 Bugzilla : 5949, 4834
4307 Description: Server oops/LBUG if stopped with --force or --failover under load
4308 Details : a collection of import/export refcount and cleanup ordering
4309 issues fixed for safer force cleanup
4312 Frequency : only filesystems larger than 120 OSTs
4313 Bugzilla : 5990, 6223
4314 Description: lfs getstripe would oops on a very large filesystem
4315 Details : lov_getconfig used kfree on vmalloc'd memory
4318 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4319 Bugzilla : 6242, 6243
4320 Description: reading from files that had been truncated to a non-zero size
4321 but never opened returned no data
4322 Details : ll_file_read() reads zeros from no-object files to EOF
4327 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4328 Details : There's little harm in aborting MDS/OSS recovery and letting it
4329 try again, so I removed the LASSERT and return an error instead.
4331 Severity : enhancement
4333 Description: New debugging infrastructure for tracking down data corruption
4334 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4335 (b) cover more of the client-side code path, and (c) try to narrow
4336 down where problems occurred
4340 Bugzilla : 3819, 4364, 4397, 6313
4341 Description: Racing close and eviction MDS could cause assertion in mds_close
4342 Details : It was possible to get multiple mfd references during close and
4343 client eviction, leading to one thread referencing a freed mfd.
4345 Severity: : enhancement
4346 Bugzilla : 3262, 6359
4347 Description: Attempts to reconnect to servers are now more aggressive.
4348 Details : This builds on the enhanced upcall-less recovery that was added
4349 in 1.4.2. When trying to reconnect to servers, clients will
4350 now try each server in the failover group every 10 seconds. By
4351 default, clients would previously try one server every 25 seconds.
4356 Description: After recovery, certain operations trigger a failed
4357 assertion on a client.
4358 Details : Failing over an mds, using lconf -d --failover, while a
4359 client was doing a readdir() call would cause the client to
4360 LBUG after recovery completed and the readdir() was resent.
4362 Severity : enhancement
4364 Description: Default groups are now added by lconf
4365 Details : You can now run lconf --group <servicename> without having to
4366 manually add groups with lmc.
4369 Frequency : occasional
4371 Description: Nodes with an elan id of 0 trigger a failed assertion
4374 Frequency : always when accessing e.g. tty/console device nodes
4376 Description: tty and some other devices nodes cannot be used on lustre
4377 Details : file's private_data field is used by device data and lustre
4378 values in there got lost. New field was added to struct file to
4379 store fs-specific private data.
4382 Frequency : when exporting Lustre via NFS
4384 Description: NFSD failed occasionally when looking up a path component
4385 Details : NFSD is looking up ".." which was broken in ext3 directories
4386 that had grown large enough to become hashed.
4389 Frequency : Clusters with multiple interfaces not on the same subnet
4391 Description: Nodes will repeatedly try to reconnect to an interface which it
4392 cannot reach and report an error to the log.
4393 Details : Extra peer list entries will be created by lconf with some peers
4394 unreachable. lconf now validates the peer before adding it.
4397 Frequency : Only if a default stripe is set on the filesystem root.
4399 Description: Setting a default stripe on the filesystem root prevented the
4400 filesystem from being remounted.
4401 Details : The client was sending extra request flags in the root getattr
4402 request and did not allocate a reply buffer for the dir EA.
4405 Frequency : occasional, higher if lots of files are accessed by one client
4406 Bugzilla : 6159, 6097
4407 Description: Client trips assertion regarding lsm mismatch/magic
4408 Details : While revalidating inodes the VFS looks up inodes with ifind()
4409 and in rare cases can find an inode that is being freed.
4410 The ll_test_inode() code will free the lsm during ifind()
4411 when it finds an existing inode and then the VFS later attaches
4412 this free lsm to a new inode.
4416 Bugzilla : 6422, 7030
4417 Description: MDS deadlock between mkdir and client eviction
4418 Details : Creating a new file via mkdir or mknod (starting a transaction
4419 and getting the ns lock) can deadlock with client eviction
4420 (gets ns lock and trying to finish a synchronous transaction).
4423 Frequency : occasional
4424 Description: While starting a server, the fsfilt_ext3 module could not be
4426 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4427 kernels. Previously, lconf would still use the ext3 name
4428 when trying to load modules. Now, it will correctly use
4429 ext3 on 2.4 and ldiskfs on 2.6.
4431 Severity : enhancement
4432 Description: The default stripe count has been changed to 1
4433 Details : The interpretation of the default stripe count (0, to lfs
4434 or lmc) has been changed to mean striping across a single
4435 OST, rather than all available. For general usage we have
4436 found a stripe count of 1 or 2 works best.
4438 Severity : enhancement
4439 Description: Add support for compiling against Cray portals.
4440 Details : Conditional compiling for some areas that are different
4444 Frequency : occasional
4445 Bugzilla : 6409, 6834
4446 Description: Creating files with an explicit stripe count may lead to
4447 a failed assertion on the MDS
4448 Details : If some OSTs are full or unavailable, creating files may
4449 trigger a failed assertion on the MDS. Now, Lustre will
4450 try to use other servers or return an error to the
4454 Frequency : occasional
4456 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4457 caused lock timeout during readahead (since 1.4.2).
4458 Details : Processes doing readahead might match a lock that hasn't been
4459 granted yet if there are overlapping and conflicting lock
4460 requests. The readahead process waits on ungranted lock
4461 (original lock is CBPENDING), while OST waits for that process
4462 to cancel CBPENDING read lock and eventually evicts client.
4464 Severity : enhancement
4466 Description: Initial enabling of flock support for clients
4467 Details : Implements fcntl advisory locking and file status functions.
4468 This feature is provided as an optional mount flag (default
4469 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4470 locking are implemented yet, and those that are are not guaranteed
4471 to be completely correct in production environments.
4472 mount -t lustre -o [flock|noflock] ...
4475 Frequency : occasional
4477 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4478 assertion in the ext3 JBD (journaling) layer.
4479 Details : The b_committed_data struct is protected by the big kernel lock
4480 in 2.4 kernels, serializing journal_commit_transaction() and
4481 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4482 access to this struct. In 2.6 kernels there is finer grained
4483 locking to improve SMP performance of the JBD layer.
4487 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4492 Description: merge and cleanup kernel patches.
4495 -----------------------------------------------------------------------------
4497 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4499 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4500 versions of Lustre. Please update all servers and clients to
4501 version 1.4.2 or later at the same time. You must also run
4502 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4503 to update the configuration logs.
4505 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4506 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4507 - swab llog records of type '0' so we get proper header size/idx (5861)
4508 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4509 - fix rename of one directory over another leaking an inode (5953)
4510 - avoid SetPageDirty on 2.6 (5981)
4511 - don't re-add just-being-destroyed locks to the waiting list (5653)
4512 - when creating new directories, inherit the parent's custom
4513 striping settings if present parent (3048)
4514 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4515 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4516 - don't hold client locks on temporary worklist from l_lru (5666)
4517 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4518 - automatically evict dead clients (5921)
4519 - Update file size properly in create+truncate+fstat case (6196)
4520 - Do not unhash mountpoint dentries, do not allow removal of
4522 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4523 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4524 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4525 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4526 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4527 - Correctly handle reads of files with no objects (6243)
4528 - lctl recover will also mark a device active if deactivate used (5933)
4530 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4531 - allow --write-conf on an MDS with different nettype than client (5619)
4532 - don't write config llogs to MDS for mounts not from that MDS (5617)
4533 - lconf should create multiple TCP connections from a client (5201)
4534 - init scripts are now turned off by default; run chkconfig --on
4535 lustre and chkconfig --on lustrefs to use them
4536 - upcalls are no longer needed for clients to recover to failover
4538 - add --abort-recovery option to lconf to abort recovery on device
4540 - add support for an arbitrary number of OSTs (3026)
4541 - Quota support protocol changes.
4542 - forward compatibility changes to wire structs (6007)
4543 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4544 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4545 - improved "lustre status" script
4546 - initialize blocksize for non-regular files (6062)
4547 - added --disable-server and --disable-client configure options (5782)
4548 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4549 - Vanilla 2.4.29 support
4550 - increase maximum number of obd devices to 520 (6242)
4551 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4552 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4553 SLES 9 2.6.5 kernels
4554 - limit stripes per file to 160 (the maximum EA size) (6093)
4556 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4559 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4560 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4561 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4562 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4563 - small ext3 extents cleanups and fixes (5733)
4564 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4565 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4566 - avoid needless client->OST connect, fix handle mismatch (5317)
4567 - fix DLM error path that led to out-of-sync client, long delays (5779)
4568 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4569 - fix several locking issues related to i_size (5492,5624,5654,5672)
4570 - don't move pending lock onto export if it is already evicted (5683)
4571 - fix kernel oops when creating .foo in unlinked directory (5548)
4572 - fix deadlock in obdfilter statistics vs. object create (5811)
4573 - use time_{before,after} to avoid timer jiffies wrap (5882)
4574 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4575 - Do not leak request if server was not able to process it (5154)
4576 - If mds_open unable to find parent dir, make that negative lookup(5154)
4577 - don't create new directories with extent-mapping (5909, 5936)
4579 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4580 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4581 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4582 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4583 - support for RHEL4 kernel on i686 (5773)
4584 - provide error messages when incompatible logs are encountered (5898)
4586 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4587 * version 1.4.0.10 (1.4.1 release candidate 1)
4589 - don't keep a lock reference when lock is not granted (4238)
4590 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4591 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4592 - drop import inflight refcount on signal_completed_replay error (5255)
4593 - unlock page after async write error during send (3677)
4594 - handle missing objects in filter_preprw_read properly (5265)
4595 - no transno return for symlink open, don't save no-trasno open (3440)
4596 - don't try to complete elan receive that already failed (4012)
4597 - free RPC server reply state on error (5406)
4598 - clean up thread from ptlrpc_start_thread() on error (5160)
4599 - readahead could read extra page into cache that wasn't ejected (5388)
4600 - prevent races in class_attach/setup/cleanup/detach (5260)
4601 - don't dereference de->d_inode after l_dput of de (5458)
4602 - use "int" for stripe value returned from lock_to_stripe (5544)
4603 - mballoc allocation and error-checking fixes in 2.6 (5504)
4604 - block device patches to fix I/O request sizes in 2.6 (5482)
4605 - look up hostnames for IB nals (5602)
4606 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4607 - don't start multiple acceptors for the same port (5277)
4608 - fix incorrect LASSERT in mds_getattr_name (5635)
4609 - export a proc file for general "ping" checking (5628)
4610 - fix "lfs check" to not block when the MDS is down (5628)
4612 - service request history (4965)
4613 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4614 - create an "evict_client" /proc entry on OSTs, like the MDS has
4615 - fix mount usage message, return errors per mount(8) (5168)
4616 - change grep [] to grep "[]" in tests so they work in more UMLs
4617 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4618 - remove ancient LOV_MAGIC_V0 EA support (5047)
4619 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4620 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4621 - allow readahead allocations to fail when low on memory (5383)
4622 - mmap locking landed again, after considerable improvement (2828)
4623 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4625 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4628 - send OST transaction number in read/write reply to free req (4966)
4629 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4630 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4631 - the watchdog thread now runs as interruptible (5246)
4632 - flock/lockf fixes (but it's still disabled, pending 5135)
4633 - don't use EXT3 constants in llite code (5094)
4634 - memory shortage at startup could cause assertion (5176)
4636 - reorganization of lov code
4637 - single portals codebase
4639 - add extents/mballoc support (5025)
4640 - direct I/O reads in the obdfilter (4048)
4641 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4643 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4646 - send OST transaction number in read/write reply to free req (4966)
4647 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4648 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4649 - the watchdog thread now runs as interruptible (5246)
4650 - handle missing objects in filter_preprw_read properly (5265)
4651 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4652 - drop import inflight refcount on signal_completed_replay error (5255)
4653 - unlock page after async write error during send (3677)
4654 - return original error code on reconstructed replies (3761)
4655 - no transno return for symlink open, don't save no-trasno open (3440)
4657 - add pid to ldlm debugging output (4922)
4658 - bump the watchdog timeouts -- we can't handle 30sec yet
4659 - extra debugging for orphan dentry/inode bug (5259)
4661 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4664 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4665 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4666 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4667 - fix NULL dereference in /proc/sys/portals/routes (4827)
4668 - allow failed mdc_close() operations to be interrupted (4561)
4669 - stop precreate on OST before MDS would time out on it (4778)
4670 - don't send partial-page writes before EOF from client (4410)
4671 - discard client grant for sub-page writes on large-page clients (4520)
4672 - don't free dentries not owned by NFS code, check generation (4806)
4673 - fix lsm leak if mds_create_objects() fails (4801)
4674 - limit debug_daemon file size, always print CERROR messages (4789)
4675 - use transno after validating reply (3892)
4676 - process timed out requests if import state changes (3754)
4677 - update mtime on OST during writes, return in glimpse (4829)
4678 - add mkfsoptions to LDAP (4679)
4679 - use ->max_readahead method instead of zapping global ra (5039)
4680 - don't interrupt __l_wait_event() during strace
4682 - add software watchdogs to catch hung threads quickly (4941)
4683 - make lustrefs init script start after nfs is mounted
4684 - fix CWARN/ERROR duplication (4930)
4685 - return async write errors to application if possible (2248)
4686 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4687 - print NAL number in %x format (4645)
4688 - update barely-supported suse-2.4.21-171 series (4842)
4689 - support for sles 9 %post scripts
4690 - support for building 2.6 kernel-source packages
4691 - support for sles km_* packages
4693 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4696 - ignore -ENOENT errors in osc_destroy (3639)
4697 - notify osc create thread that OSC is being cleaned up (4600)
4698 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4699 - reconstruct ost_handle() like mds_handle() (4657)
4700 - create a new thread to do import eviction to avoid deadlock (3969)
4701 - let lconf resolve symlinked-to devices (4629)
4702 - don't unlink "objects" from directory with default EA (4554)
4703 - hold socknal file ref over connect in case target is down (4394)
4704 - allow more than 32000 subdirectories in a single directory (3244)
4705 - fix blocks count for O_DIRECT writes (3751)
4706 - OST returns ENOSPC from object create when no space left (4539)
4707 - don't send truncate RPC if file size isn't changing (4410)
4708 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4709 - bind to privileged port in socknal and tcpnal (3689)
4711 - rate limit CERROR/CWARN console message to avoid overload (4519)
4712 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4713 - basic mmap support (3918)
4714 - kernel patch series update from b1_4 (4711)
4716 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4719 - avoid crash during MDS cleanup with OST shut down (2775)
4720 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4721 - don't use bad inodes on the MDS (3744)
4722 - dynamic object preallocation to improve recovery speed (4236)
4723 - don't hold spinlock over lock dumping or change debug flags (4401)
4724 - don't zero obd_dev when it is force cleaned (3651)
4725 - print grants to console if they go negative (4431)
4726 - "lctl deactivate" will stop automatic recovery attempts (3406)
4727 - look for existing locks in ldlm_handle_enqueue() (3764)
4728 - don't resolve lock handle twice in recovery avoiding race (4401)
4729 - revalidate should check working dir is a directory (4134)
4731 - don't always mark "slow" obdfilter messages as errors (4418)
4733 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4736 - don't close LustreDB during write_conf until it is done (3860)
4737 - fix typo in lconf for_each_profile (3821)
4738 - allow dumping logs from multiple threads at one time (3820)
4739 - don't allow multiple threads in OSC recovery (3812)
4740 - fix debug_size parameters (3864)
4741 - fix mds_postrecov to initialize import for llog ctxt (3121)
4742 - replace config semaphore with spinlock (3306)
4743 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4744 - don't allow enqueue to complete on a destroyed export (3822)
4745 - down write_lock before checking llog header bitmap (3825)
4746 - recover from lock replay timeout (3764)
4747 - up llog sem before sending rpc (3652)
4748 - reduce ns lock hold times when setting kms (3267)
4749 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4750 - fix NULL deref and obd_dev leak on setup error (3312)
4751 - replace some LBUG about llog ops with error handling (3841)
4752 - don't match INVALID dentries from d_lookup and spin (3784)
4753 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4754 - fix invalid assertion in ptlrpc_set_wait (3880)
4756 - add libwrap support for the TCP acceptor (3996)
4757 - add /proc/sys/portals/routes for non-root route listing (3994)
4758 - allow setting MDS UUID in .xml (2580)
4759 - print the stack of a process that LBUGs (4228)
4761 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4764 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4765 - make sure to unset replay flag from failed open requests (3440)
4766 - if default stripe count is 0, use OST count for inode size (3636)
4767 - update parent mtime/ctime on client for create/unlink (2611)
4768 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4769 - free recovery state on server during a forced cleanup (3571)
4770 - unregister_reply for resent reqs (3063)
4771 - loop back devices mounting and status check on 2.6 (3563)
4772 - fix resource-creation race that can provoke i_size == 0 (3513)
4773 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4774 - more debugging for page-accounting assertion (3746)
4775 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4776 - avoid deadlock after precreation failure (3758)
4777 - fix race and lock order deadlock in orphan handling (3450, 3750)
4778 - add validity checks when grabbing inodes from l_ast_data (3599)
4780 - add /proc/.../recovery_status to obdfilter (3428)
4781 - lightweight CDEBUG infrastructure, debug daemon (3668)
4782 - change default OSC RPC parameters to be better on small clusters
4783 - turn off OST read cache for files smaller than 32MB
4784 - install man pages and include them in rpms (3100)
4785 - add new init script for (un)mounting lustre filesystems (2593)
4786 - run chkconfig in %post for init scripts (3701)
4787 - drop scimac NAL (unmaintained)
4789 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4792 - clean kiobufs before and after use (3485)
4793 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4794 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4795 - change init script to fail more clearly if not run as root (1528)
4796 - allow clients to reconnect during replay (1742)
4797 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4798 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4799 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4800 - fix cond_resched (3554)
4801 - only evict unfinished clients after recovery (3515)
4802 - allow bulk resend, prevent data loss (3570)
4803 - dynamic ptlrpc request buffer allocation (2102)
4804 - don't allow unlinking open directory if it isn't empty (2904)
4805 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4806 - remove extraneous obd dereference causing LASSERT failure (3334)
4807 - don't use get_cycles() when creating temp. files on the mds (3156)
4808 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4809 - handle EEXIST for set-stripe, set proper directory name (3336)
4811 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4812 - fix ksocknal_fmb_callback() error messages (2918)
4814 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4817 - don't copy lvb into (possibly NULL) reply on error (2983)
4818 - don't deref dentry after dput, don't free lvb on error (2922)
4819 - use the kms to determine writeback rpc length (2947)
4820 - increment oti_logcookies when osc is inactive (2948)
4821 - update client's i_blocks count via lvb messages (2543)
4822 - handle intent open/close of special files properly (1557)
4823 - mount MDS with errors=remount-ro, like obdfilter (2009)
4824 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4825 - don't use cancelling-locks' kms values (2947)
4826 - use highest lock extent for kms, not last one (2925)
4827 - don't dereference ERR_PTR() dentry in error handling path (3107)
4828 - fix thread race in portals_debug_dumplog() (3122)
4829 - create lprocfs device entries at setup instead of at attach (1519)
4830 - common AST error handler, don't evict client on completion race (3145)
4831 - zero nameidata in detach_mnt in 2.6 (3118)
4832 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4833 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4834 - initialize RPC timeout timer earlier for 2.6 (3219)
4835 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4836 - print nal/nid for unknown nid (3258)
4837 - additional checks for oscc recovery before doing precreate (3284)
4838 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4839 - don't crash in mdc_close for bad permissions on open (3285)
4840 - zero i_rdev for non-device files (3147)
4841 - clear page->private before handing to FS, better assertion (3119)
4842 - tune the read pipeline (3236)
4843 - fix incorrect decref of invalidated dentry (2350)
4844 - provide read-ahead stats and refine rpc in flight stats (3328)
4845 - don't hold journal transaction open across create RPC (3313)
4846 - update atime on MDS at close time (3265)
4847 - close LDAP connection when recovering to avoid server load (3315)
4848 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4849 - don't leak open file on MDS after open resend (3325)
4850 - serialize filter_precreate and filter_destroy_precreated (3329)
4851 - loop device shouldn't call sync_dev() for nul device (3092)
4852 - clear page cache after eviction (2766)
4853 - resynchronize MDS->OST in background (2824)
4854 - refuse to mount the same filesystem twice on same mountpoint (3394)
4855 - allow llmount to create routes for mounting behind routers (3320)
4856 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4857 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4858 - fix rare problem with rename on htree directories (3417)
4860 - allow default OST striping configuration per directory (1414)
4861 - fix compilation for qswnal for 2.6 kernels (3125)
4862 - increase maximum number of MDS request buffers for large systems
4863 - change liblustreapi to be useful for external progs like lfsck (3098)
4864 - increase local configuration timeout for slow disks (3353)
4865 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4867 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4870 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4871 - don't overwrite extent policy data in reply if lock was blocked (2901)
4872 - drop filter export grants atomically with removal from device (2663)
4873 - del obd_self_export from work_list in class_disconnect_exports (2908)
4874 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4875 - swab reply message in mdc_close, other PPC fixes (2464)
4876 - fix destroying of named logs (2325)
4877 - overwrite old logs when running lconf --write_conf (2264)
4878 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4879 - fix race in target_handle_connect (2898)
4880 - mds_reint_create() should take same inode create lock (2926)
4881 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4882 - don't close files for self_export to avoid uninitialized obd (2936)
4883 - allow MDS with the same name as client node (2939)
4884 - hold dentry reference for closed log files for unlink (2325)
4885 - reserve space for all logs during transactions (2059)
4886 - don't evict page beyond end of stripe extent (2925)
4887 - don't oops on a deleted current working directory (2399)
4888 - handle hard links to targets without a parent properly (2517)
4889 - don't dereference NULL lock when racing during eviction (2867)
4890 - don't grow lock extents when lots of conflicting locks (2919)
4892 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4895 - account for cache space usage on clients to avoid data loss (974)
4896 - lfsck support in lustre kernel code (2349)
4897 - reduce journal credits needed for BRW writes (2370)
4898 - orphan handling to avoid losing space on client/server crashes
4899 - ptlrpcd can be blocked, stopping ALL progress (2477)
4900 - use lock value blocks to assist in proper KMS, faster stat (1021)
4901 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4902 - recovery for initial connections (2355)
4903 - fixes for mds_cleanup_orphans (1934)
4904 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4905 - block all file creations until orphan recovery completes (1901)
4906 - client remove rq_connection from request struct (2423)
4907 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4908 - recovery timer race (2670)
4909 - mdc_close recovey bug (2532)
4910 - ptlrpc cleanup bug (2710)
4911 - mds timeout on local locks (2588)
4912 - namespace lock held during RPCs (2431)
4913 - handle interrupted sync write properly (2503)
4914 - don't try to handle a message that hasn't been replied to (2699)
4915 - client assert failure during cleanup after abort recovery (2701)
4916 - leak mdc device after failed mount (2712)
4917 - ptlrpc_check_set allows timedout requests to complete (2714)
4918 - wait for inflight reqs when ptlrpcd finishes (2710)
4919 - make sure unregistered services are removed from the srv_list
4920 - reset bulk XID's when resending them (caught by 1138 test)
4921 - unregister_bulk after timeout
4922 - fix lconf error (2694)
4923 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4924 - readahead locks pages, leaves pending causing memory pressure (2673)
4925 - increase OST request buffers to 4096 on large machines (2729)
4926 - fix up permission of existing directories in simple_mkdir (2661)
4927 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4928 - don't assign transno to errored transactions (2742)
4929 - don't delete objects on OST if given a bogus objid from MDS (2751)
4930 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4931 - if rq_no_resend, then timeout request after recovery (2432)
4932 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4933 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4934 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4935 - MDS recovery shouldn't delete objects that it has given out (2730)
4936 - if enqueue arrives after completion, don't clobber LVB (2819)
4937 - don't unlock pages twice when trigger_group_io returns error (2814)
4938 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4939 - don't write pages to disk if there was an error (1450)
4940 - don't ping imports that have recovery disabled (2676)
4941 - take buffered bytes into account when balancing socknal conn (2817)
4942 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4943 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4944 - remove little-endian swabbing of llog records (1987)
4945 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4946 - retry reposting request buffers if they fail (1191)
4947 - grow extent at grant time to avoid granting a revoked lock (2809)
4948 - lock revoke doesn't evict page if covered by a second lock (2765)
4949 - disable VM readahead to avoid reading outside lock extents (2805)
4951 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4952 - updated kernel patches for hp-2.4.20 kernel (2681)
4954 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4957 - fix truncated write corruption (2366)
4958 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4960 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4961 - ensure len(uuid) < 37 in lmc (1171)
4962 - fix ia64 OOPS in llog_test (2255)
4963 - zero end of page at obdfilter for partial page writes (2648)
4964 - don't leave stale dentries around after renames (bug 2428)
4965 - fix timeouts when evicting a client with a single lock held (2642)
4966 - set deadline for the initial HELLO message to drain (2634)
4967 - print out dotted-quad IP addresses in the socknal (2302)
4969 - additional debugging for MDS client eviction problem (2443)
4970 - fix mkfsoptions support for osts (2603, 2604)
4972 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4975 - add series for the vanilla 2.6.0 kernel
4976 - add series for the vanilla 2.4.24 kernel
4977 - add series for a cray x86/64 UL kernel drop
4978 - fix xattr patches for the vanilla 2.4.19 series
4980 - generate true UUIDs in lmc (1171)
4981 - have portals stack dumping break in UML (2466)
4982 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4983 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4984 - wake callback waiting threads on client eviction (2460)
4985 - Add --ptldebug and --subsystem to lmc (1719)
4986 - update assertion to allow safe interrupt allocation
4987 - set rq_no_resend for cancel requests (2432)
4988 - recalculate ptlrpcd timeout after resend (2494)
4989 - call vfs_rmdir when removing pending directories (2368)
4990 - fix renaming a file to itself (2429)
4991 - lmc creates a default one-stripe lov (2454)
4992 - expand procfs space to handle large clusters (2326)
4993 - increase UML stack to avoid overflow
4994 - update lconf's list of debug and subsystem masks
4995 - fix lfs find --obd (2510)
4996 - /proc tunable for disabling filter read caching (2591)
4997 - stop rpm packages from altering slapd.conf (2301)
4998 - disable nagle in the socknal under 0conf (2578)
4999 - choose mds inode size based on stripe count (2572)
5000 - fix kernel-source rpm problems (2516)
5002 - add --disable-doc to avoid pdf generation (2421)
5003 - update documentation, tests, type-os, comments
5004 - avoid format warnings on ia64
5005 - remove the TOE NAL
5006 - tiny code cleanups by removing unused fields
5008 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5011 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5012 - remove the most blatant lies from BUILDING file (2371)
5013 - change default debug level to reasonable production setting
5014 - reduce client side cache size to reduce cache flush time
5015 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5016 - make TCP zerocopy and pinger support enabled by default (2476)
5017 - sync writes completed after process exits caused crashes (2319)
5018 - maintain correct mount count on the MDS (2356)
5019 - backout 1557, because 2316 wasn't really fixed
5020 - better file I/O statistics gathering in /proc
5021 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5022 - another kernel patch to fix zero-copy TCP function export
5023 - don't take duplicate lock when processing re-sent getattr (2420)
5024 - lctl uses obd_self_export instead of creating new conn (2353)
5025 - MDS/OST recovery case which requires object creation asserted (2425)
5026 - move lfs from /usr/sbin to /usr/bin in packages
5027 - fix race between mds_client_add and mds_client_free (2417)
5028 - use kmalloc instead of slabs in portals (2430)
5029 - don't create duplicate records when a failover MDS is present (2442)
5030 - remove unnecessary mount age check (2332)
5031 - don't remove directory inodes from locks prematurely (2451)
5032 - don't break if MDS service name is the same as hostname (2103)
5033 - fix races in client write RPC generation when cache full (2482)
5035 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5038 - remove now-unused request->rq_obd (278)
5039 - if an allocation fails, print out how much memory we've used (1933)
5040 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5041 - add the "configurable stack size" patch to most series files (1256)
5042 - ability to write large log records, for 100+ OST configs (2306)
5043 - fix NULL deref when filter_prep fails (2314)
5044 - fix operator precedence error in filter_sync
5045 - dynamic allocation of socknal TX descriptors (2315)
5046 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5047 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5048 - ensure that CWARN messages in Portals always get to the syslog
5049 - __init/__exit are not for prototype decls (ldlm_init/exit)
5050 - x86-64 compile warning fixes
5051 - fix gateway LMC keyword conflict (2318)
5052 - fix MDS lock inversions in getattr/reint paths (1844)
5053 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5054 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5055 - DLM race condition prevented some lock evictions (2328)
5056 - ENOMEM detection and retry on socknal sends (2230)
5057 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5058 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5059 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5060 - better error messages when a client is rejected during recovery (1505)
5061 - avoid cancelling locks which were never granted, after failure (2330)
5062 - fix i_sem/journal inversion in mds_client_add (2333)
5063 - fix truncate/getattr lock cycle deadlock (2334)
5064 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5065 - fix two rare exit paths which could leak an l_lock() ref (2321)
5066 - fencepost error in MDS/OST orphan recovery (2226)
5067 - make log record alignment 8 bytes (1988)
5068 - lstripe now fails when requested offset > ost_count (2237)
5069 - ensure that all kernel series have a complete list.h (1607)
5070 - fix crashes in special-file operations (2316)
5071 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5073 - allow configurable automake binary, for testing new versions
5074 - small update to the lfs documentation
5076 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5078 * fix negative export reference count in fsfilt_sync (2312)
5080 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5081 * release candidate 0.9.1
5083 - orphans are moved into the PENDING directory for possible recovery
5084 - replayed opens now open by fid for orphan/rename safety (1042)
5085 - last close of an orphan inode generates a transno (683)
5086 - chdir() and mount() now pin the directory entry (1020)
5087 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5088 - discard very old requests without processing them (1502)
5089 - remove some common, well-understood CERRORs (1505)
5090 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5091 - clear "grant" flags in OST replies until OST grant code lands (1644)
5092 - fix read performance by not clobbering i_blksize on client (1598)
5093 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5094 - make LRU size a /proc tunable, clears locks when reduced (707)
5095 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5096 - server locks take references on exports now (1558)
5097 - build fixes for 2.4.20-rh trees (1663)
5098 - return an error from lov_create if all OSCs are inactive (1751)
5099 - fix import levels when a reconnect happens without a timeout (1597)
5100 - exit early from mds_open if we get a lookup error (1749)
5101 - partial page read at EOF wouldn't wait for disk before sending (1642)
5102 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5103 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5104 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5105 - write/truncate lock inversion (1639)
5106 - Don't auto-load obdclass, portals modules during cleanup (1495)
5107 - fix timestamps from jumping to "now" (1763)
5108 - extra journal assertions (1648)
5109 - add an extra multiunlink test (1771)
5110 - fix read_record/write_record API (1776)
5111 - fix leak of offset_extent, possible incorrect i_size later (1772)
5112 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5113 - Debugging for the kqswnal_get_idle_tx problems (1820)
5114 - Allow recovery to be attempted multiple times (1536)
5115 - Write out MDS last_rcvd file after it is first created (1600)
5116 - Fix tx_descriptor leak in failed transmit situations (1827)
5117 - ext3 journaling fixes for assertion failure after IO error (1871)
5118 - class_export_put() on freed export after completion AST error (1896)
5119 - Fix revalidate looping in VFS (1322)
5120 - Don't access a freed export during MDS_REINT timeout (1521)
5121 - Add open-unlink recovery support on the MDS (1673,1764)
5122 - Return an error if no MDS data was read from last_rcvd (1946)
5123 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5124 - Don't setup a disk device that doesn't match exported UUID (317)
5125 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5126 - avoid committing NULL handle in force close
5127 - local.sh is now a one-stripe LOV configuration
5128 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5129 - fix ext3 htree duplicate directory entry corruption (1516)
5130 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5131 - update to new LOV EA format (2097)
5132 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5134 2003-06-15 Phil Schwan <phil@clusterfs.com>
5137 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5138 - per-import/export recovery handling (958, 931, 959)
5139 - multiple last-rcvd slots, for serving multiple FSes (949)
5140 - connections are again shared between multiple imp/exports (963, 964)
5141 - "umount -f" would hang if any requests needed to be sent (393, 978)
5142 - avoid pinning large req buffer by copying for queued messages (989)
5143 - add "uuid" to "lctl device" command to help upcalls (991)
5144 - "open" RPCs with transnos would confuse recovery counters (1037)
5145 - do proper endian conversion of all wire messages (288, 340, 891)
5146 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5147 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5148 - fix ptlrpc_connection leak in target_handle_connect (1174)
5149 - fix import refcounting bug in OST and MDS cleanup (1134)
5150 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5151 - fix very unlikely obd_types race condition (501)
5152 - remove osc_open hack for echo_client (1187)
5153 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5154 - a failure in read_inode2 leads to deadlock (1139)
5155 - cancel ack-locks as soon as transaction is committed (1072)
5156 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5157 - make sure to commitrw after any preprw to avoid deadlock (1162)
5158 - failing to execute a file in a lustre FS would lock inode (1203)
5159 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5160 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5161 - obd_uuid2tgt would walk off the end of the list (1255)
5162 - on IA64 the osc would give portals incorrect bulk size (1258)
5163 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5164 - fix lock inversion caused by new llite matching code (1282)
5165 - limit the number of dirty pages on a client to 10MB (1286)
5166 - timed out locks were not being corrected cancelled (1289)
5167 - fix O_DIRECT above 4GB on IA-32 (1292)
5168 * major user-visible changes
5169 - fail out/fail over policy now controlled by the upcall (993)
5171 - add OBD_PING to check server availability and failure (954)
5172 - lustre messages are now sent in sending host order (288, 340, 891)
5173 - add eadatalen to MDS getattr reply (340)
5174 - OST read replies may contain second buffer, with per-page status (593)
5176 2003-03-11 Phil Schwan <phil@clusterfs.com>
5179 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5180 - failed open()s could cause deadlock; fixed (867, 869)
5181 - stop cancelling OST locks when files are closed (481)
5182 - overlapping XID spaces caused network corruption (851, 853)
5183 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5184 - setattr_raw added, to do single-RPC, server-side setattrs
5185 - lmc/lconf syntax change for OST UUIDs
5186 - fix crashy race condition between ptlrpc_free_req and osc_close
5187 - don't use request in mdc_enqueue if we hit a timeout (889)
5188 - don't set the inode i_size for regular files from the MDS (896)
5189 - handle out of order completion AST (842)
5190 - don't LBUG if a lock request times out after receiving AST (913)
5191 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5192 - if a bad lock AST arrives, send an error instead of dropping entirely
5193 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5194 - fix leak in bulk IO when only partially completed (899, 900, 926)
5195 - fix O_DIRECT for ia64 (55)
5196 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5197 - C-z after timeout could hang a process forever; fixed (977)
5199 - client-side I/O cache (678, 924, 929, 941, 970)
5201 - READPAGE and SETATTRs which don't take server-side locks get
5204 2003-02-11 Phil Schwan <phil@clusterfs.com>
5207 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5208 - Fix fsfilt_extN_readpage() to read a full page of directory
5209 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5210 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5211 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5212 - Most intent operations were removed, in favour of a new RPC mode
5213 that does a single RPC to the server and bypasses most of the VFS
5214 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5215 - Aggressively cancel local locks on DLM servers
5216 - mds_reint_unlink sends EA to the client if it's the last nlink.
5217 client uses that EA to unlink OST objects.
5218 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5219 - recursive symlinks were fixed (439)
5220 - fixed NULL deref in DEBUG_REQ
5221 - filter_update_lastobjid no longer calls sync, which annoyed extN
5222 - fixed multi-client small-writes to a single file problem (445)
5223 - fixed mtime updates during file writes (607)
5224 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5225 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5226 - fixed Linux OST size reporting problem (444, 656)
5227 - OST now updates object mtime with writes or setattr (607, 619)
5228 - client verifies file size before zeroing page past EOF (445)
5229 - OST now writes last allocated objid to disk with allocation (108)
5230 - LOV on echo now works (409)
5232 - mds_reint_unlink sends a new buffer, with the EA included. this
5233 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5234 set if a regular file was being unlinked, and it was the last link
5235 - use PtlGet from the target for bulk writes (315)
5236 - OST now updates object mtime with writes or setattr (607, 619)
5237 - LDLM now has a grant-time callback to revalidate locked items, if
5239 - Many MDS operations were reorganized to combat race conditions
5241 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5242 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5244 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5247 - Fully reactivate OST imports after reconnection (512, others)
5248 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5249 - More graceful error handling for truncating on dead OST (515)
5250 - Don't error out unless we're actually accessing dead stripes (474)
5251 - Fix garbage sizes when stripes are missing (410)
5252 - LRU counters were broken, causing constant lock purge (433, 432)
5253 - garbage on read from stripes with failed OSTs (441)
5254 - mark OSCs as active before reconnecting during recovery (438)
5255 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5256 - lfind did not preserve OST order in output (443)
5257 - symlinks cause hung clients, incorrect data (439)
5258 - stop dereferencing request after dropping refcount (457)
5259 - don't LASSERT(spin_is_locked) on non-SMP (455)
5260 - fixes for many rename() bugs
5261 - fstat didn't correctly synchronize attributes (399)
5262 - server must handle lock cancellation during blocking AST prep (487)
5263 - bulk descriptors were free()d too soon (511)
5264 - fix paths in lconf, which would load incorrect modules (451, 507)
5265 - fix confusing lconf 'host not found' error message (386)
5266 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5267 - fix race condition in mdc_blocking_ast() for inode access (526)
5268 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5269 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5270 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5271 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5272 - fix lstripe in several ways (kernel side) (527)
5273 - fix request leak in ldlm_cli_enqueue (262)
5274 - incorrect OSC was marked inactive after OST failure
5275 - call mds_fs_cleanup before unmounting filesystem (524)
5276 - fix races between taking ns_lock and ldlm_lock_change_resource
5277 - fix races updating LOV export open file list
5278 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5279 - fix recovery NULL deref in ldlm_cli_cancel_unused
5280 - fix some DLM races by using new hash table for lock handles (419)
5281 - permit the client to specify desired inodes, at replay
5282 - duplicate requests when we queue them for replay reintegration
5283 - fix last_rcvd offset calculation
5284 - sync after each recovered transaction, so we always make progress
5285 - never, not always, ERESTART requests without transnos
5286 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5288 - skip replay if the MDS says that the client is already connected
5289 - don't check for a recovery-enabled export to match lctl's UUID
5290 - don't INC_USE_COUNT for phantom exports
5291 - don't crash when cleaning up phantom exports (567)
5292 - don't double-finish or set replay data for errored mdc_open requests
5293 - abort requests when they time out, so we don't get old replies
5294 - send/receive replies for AST messages again
5295 - if the client says that it doesn't have the lock, cancel it on the
5297 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5298 mark it as destroyed, it will all work out when decref is called
5299 - fix module use counts (22, 581)
5301 - ASTs now expect a reply (server cancels lock on error reply)
5303 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5306 - fix many simultaneous client startup (392)
5307 - fix dentry->d_it clobbering
5308 - credentials weren't being shipped for readdir/getattr operations
5309 - remove invalid assertions triggered during some concurrent MD
5311 - proper Lustre versions added (336, 389)
5312 - fix memory leak for create error case (398)
5313 - fix LOV locking bug that would get cli/srv out of sync
5314 - fix echo client over LOV (409)
5315 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5316 - fix double-O_EXCL intent crash (424)
5317 - avoid sending multiple lock CANCELs (352)
5319 - MDS can do multi-client recovery (modulo bugs in new code)
5321 - many updates, edits, cleanups
5323 2002-11-18 Phil Schwan <phil@clusterfs.com>
5326 - fix null d_it dereference (346)
5327 - fix full OST/dbench hang (333)
5328 - fix permission problem with file removal (286)
5329 - fix removal of OSCs from LOV when they fail
5330 - fix NULL deref during bulk timeout (214)
5331 - fix problems related to multiple filesystems on one MDS (241)
5332 - fixed serious subtle metadata locking bugs
5333 - free locks on clients when inodes are removed due to memory
5335 - fix inode pointer in lock data (285)
5336 - partial support for multiple MDS on a single host (241)
5337 - data locks weren't cancelled at clear_inode time (290, 311)
5338 - intent locks could lead to unbounded lock growth (205)
5339 - added a maximum lock count, an LRU list, and a flusher
5340 - fix multiple rename (365)
5341 - properly abstracted the echo client
5342 - OSC locked 1 byte too many; fixed
5343 - rewrote brw callback code:
5344 - fixed recovery bugs related to LOVs (306)
5345 - fixed too-many-pages-in-one-write crash (191)
5346 - fixed (again) crash in sync_io_timeout (214)
5347 - probably fixed callback-related race (385)
5349 - Add capability to MDS protocol
5350 - LDLM cancellations and callbacks on different portals
5352 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5355 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5356 - defer open object destruction to close time (601981, 138)
5357 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5358 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5359 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5360 - don't call obd_setattr in ll_file_release for destroyed objects
5361 * protocol change to lustre_msg: move |version| and add |flags|
5362 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5363 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5364 * added replay of create, unlink, link and rename operations during
5365 MDS failover; recovery should be much more robust now
5366 * remove failed OSCs from LOVs (only lov_create uses this so far)
5367 * the lustre-HOWTO was brought (more) up to date (582544)
5369 2002-10-23 Phil Schwan <phil@clusterfs.com>
5372 - in-use dentries weren't being reused properly (617851)
5373 - prevent multiple LDLM setup (599178)
5374 - fix LOV size calculations for truncate (617853)
5375 - fix client handling of MDS intent errors (POSIX)
5376 - fix permission bug in lovstripe.c test (624321)
5377 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5378 - truncate past end of file could corrupt data
5379 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5380 - a race in recovery could return ETIMEDOUT to apps (623947)
5381 - building outside the source directory was fixed
5382 * the lustre-HOWTO was brought (more) up to date (582544)
5383 * major progress was made on recovery functionality
5385 2002-10-10 Phil Schwan <phil@clusterfs.com>
5388 - recovery deadlock fix
5389 - rm -rf causes LBUG fix (617817)
5390 - file open by multiple tasks fix (618962)
5391 - directory permissions bugs (602707 and 620007)
5392 - journal_stop fixed with locking (611313)
5393 - O_APPEND failures resolved (618273, perhaps 614459)
5394 - lconf PATH fix (619770)
5395 - IA64 build fix (621450)
5396 - RPC buffer sizes scale with amount of memory
5398 2002-10-01 Phil Schwan <phil@clusterfs.com>
5401 - locks would be cancelled without throwing away data pages,
5402 resulting in inconsistent data (605627)
5403 - inode attributes were not always being refreshed (605627, 612449)
5404 - lconf now continues to cleanup after lctl reports an error
5405 - MDS now enforces user permissions (602707)
5406 - lprocfs cleanup fixed, but not yet enabled (614157)
5407 - fixed infinite server hang, should a client not respond to an AST
5408 - avoid going into recovery if user calls readlink() with a buffer
5409 that's too small (613941)
5410 - AST RPCs no longer require replies (614867) -- this may be changed
5411 - don't crash server if client sends an IOV that's too big (611336)
5412 - fixed lock conversion deadlock (611892)
5413 - fixed the following of symlinks (614622)
5414 * recovery: the server can remove locks from a client that dies, other
5415 clients can make progress
5416 * more extN patch fixes
5417 * compile-time configurable ptlrpc buffer allocations
5419 - collaborative read cache document
5420 - Lustre Lite Performance CDR document-in-progress
5422 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5425 - fix typo in patch-2.4.18
5427 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5430 - clear ptlrpc request each time in handle_incoming_request()
5431 - unlink of files now destroys the object on the OST
5433 2002-09-19 Peter Braam <braam@clusterfs.com>
5435 * add hard link support
5436 * change obdfile creation method
5437 * kernel patch changed
5439 2002-09-19 Peter Braam <braam@clusterfs.com>
5442 - stack overflow bug in extN fixed
5444 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5446 * documentation updates
5447 - add man pages for config tools
5448 - update tests/README to describe testing with new config tools
5449 - finish metadata API descriptions
5450 * bug fixes and cleanups
5451 - statfs workaround for 16TB limit
5452 - LOV stripe allocation improved, can stripe on subset of OSTs
5453 - LOV file size/IO offset was wrong for files > 4GB in size
5454 - object EA data was being dropped, caused files to be unreadable
5455 - memory overflow with non-LOV OST caused memory corruption
5456 - fixed regression tests to work with new config tools, obdfilter
5457 - fixed bug when directory size became larger than 1 block
5458 - fixed bug (for single client case) when PWD was deleted
5459 - invalidate local directory pages when doing intent-based ops
5460 - avoid LDLM oops when lock callback contained bad data
5462 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5464 * documentation updates
5465 * bug fixes and cleanups
5466 - configuration tools
5469 - 64-bit compile warnings
5470 - 64-bit internal statfs data
5472 * test_brw on persistent OST devices
5474 * lprocfs (disabled)
5476 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5478 * documentation updates
5479 * bug fixes and cleanups
5480 * configuration tools
5482 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5485 * many small fixes to 0_5_4
5486 * io/network handling
5487 * thinkos in MDS operations
5489 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5492 * crucial basic fixes to 0.5.3
5493 * IOR, Iozone work over Elan
5496 2002-08-07 Phil Schwan <phil@clusterfs.com>
5497 * version 0_5_3, our first alpha
5498 * we use the new Portals iovs
5499 * documentation updates
5500 * bug fixes and cleanups
5501 * small changes in the DLM wire protocol
5503 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5504 * version 0_5_1 with some initial stability,
5505 * locking on MD and file I/O.
5506 * documentation updates
5507 * several bug fixes since 0.5.0
5508 * small changes in wire protocol
5510 2002-07-18 Phil Schwan <phil@clusterfs.com>
5512 * delivered as Lustre Light Alpha
5513 * fixed a crash after handling invalid MDS requests
5514 * fixed directory pages for architectures with non-4k pages sizes
5516 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5517 * release version v0_4_4
5518 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5520 * Fixes a number of interruption problems with OST operations.
5521 * Update documentation for portals header changes
5522 * Move all wire protocol structs/defines to lustre_idl.h
5523 * Fixes symlink length bug.
5524 * Add tcpdump to repository.
5526 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5527 * release version v0_4_3
5528 * Fixes statfs for inodes on extN.
5529 * Fixes bug in runtests which would delete /etc/hosts.
5530 * Use 64-bit object IDs wherever possible (not into VFS though)
5531 Remove ost_get_info, which is unused by lustre, and out of date.
5533 2002-07-03 Peter Braam <braam@clusterfs.com>
5534 * release version v0_4_2 Fixes a lookup error (type not passed)
5535 * move forward to head of Portals
5536 * move forward to latest Lustre kernel
5538 2002-06-25 Peter Braam <braam@clusterfs.com>
5539 * release version v0_4_1. Hopefully stable on single node use.