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: New client IO stack (CLIO).
20 Severity : enhancement
22 Description: Commit on sharing. Eliminate inter-client dependencies between
23 uncommitted transactions by doing transaction commits.
24 Thereby clients may recovery independently.
27 Frequency : Create a symlink file with a very long name
29 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
30 Details : If there is no extra space in the request for early cancels,
31 ldlm_req_handles_avail() returns 0 instead of a negative value.
33 Severity : enhancement
35 Description: Add /proc entry for import status
36 Details : The mdc, osc, and mgc import directories now have
37 an import directory that contains useful import data for debugging
40 Severity : enhancement
42 Description: Re-disable certain /proc logging
43 Details : Enable and disable client's offset_stats, extents_stats and
44 extents_stats_per_process stats logging on the fly.
47 Frequency : Only on FC kernels 2.6.22+
49 Description: oops in statahead
50 Details : Do not drop reference count for the dentry from VFS when lookup,
51 VFS will do that by itself.
53 Severity : enhancement
55 Description: Generic /proc file permissions
56 Details : Set /Proc file permissions in a more generic way to enable non-
57 root users operate on some /proc files.
61 Description: Hitting mdc_commit_close() ASSERTION
62 Details : Properly handle request reference release in
63 ll_release_openhandle().
67 Frequency : only patchless client
68 Description: add workaround for race between add/remove dentry from hash
70 Severity : enhancement
72 Description: Allow OST glimpses to return PW locks
76 Description: LBUG when llog conf file is full
77 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
81 Description: Prevent import from entering FULL state when server in recovery
85 Description: service mount cannot take device name with ":"
86 Details : Only when device name contains ":/" will mount treat it as
92 Description: replace ptlrpcd with the statahead thread to interpret the async
93 statahead RPC callback
97 Frequency : on recovery
98 Description: I/O failures after umount during fail back
99 Details : if client reconnected to restarted server we need join to recovery
100 instead of find server handler is changed and process self
101 eviction with cancel all locks.
103 Severity : enhancement
105 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
107 Severity : enhancement
109 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
111 Severity : enhancement
113 Description: Upcall on Lustre log has been dumped
114 Details : Allow for a user mode script to be called once a Lustre log has
115 been dumped. It passes the filename of the dumped log to the
116 script, the location of the script can be specified via
117 /proc/sys/lnet/debug_log_upcall.
122 Description: avoid idr_remove called for id which is not allocated.
123 Details : Move assigment s_dev for clustered nfs to end of initialization,
124 to avoid problem with error handling.
129 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
130 Details : When connection is reused this not moved from CONN_UNUSED_HASH
131 into CONN_USED_HASH and this prodice warning when put connection
132 again in unused hash.
134 Severity : enhancement
136 Description: File striping can now be set to use an arbitrary pool of OSTs.
138 Severity : enhancement
140 Description: Export bytes_read/bytes_write count on OSC/OST.
144 Description: Early reply size mismatch, MGC loses connection
145 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
146 the connect flags are properly negotiated.
150 Description: Properly propagate oinfo flags from lov to osc for statfs
151 Details : restore missing copy oi_flags to lov requests.
153 Severity : enhancement
155 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
156 : l_getgroups(8), lst(8), routerstat(8)
157 Details : included man pages for llobdstat(8), llstat(8),
158 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
160 Severity : enhancement
162 Description: configure's --enable-quota should check the
163 : kernel .config for CONFIG_QUOTA
164 Details : configure is terminated if --enable-quota is passed but
165 : no quota support is in kernel
169 Description: Remove portals compatibility
170 Details : Remove portals compatibility, not interoperable with releases
175 Description: Resolve device initialization race
176 Details : Prevent proc handler from accessing devices added to the
177 obd_devs array but yet be intialized.
179 Severity : enhancement
181 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
183 Severity : enhancement
185 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
190 Description: Using wrong pointer in osc_brw_prep_request
191 Details : Access to array[-1] can produce panic if kernel compiled with
192 CONFIG_PAGE_ALLOC enabled
196 Description: Client runs out of low memory
197 Details : Consider only lowmem when counting initial number of llap pages
201 Description: *optional* service tags registration
202 Details : if the "service tags" package is installed on a Lustre node
203 When the filesystem is mounted, a local-node service tag will
204 be created. See http://inventory.sun.com/ for more information
205 about the Service Tags asset management system.
209 Description: Kernel BUG tries to release flock
210 Details : Lustre does not destroy flock lock before last reference goes
211 away. So always drop flock locks when client is evicted and
212 perform unlock regardless of successfulness of speaking to MDS.
216 Description: add refcount protection for osc callbacks, avoid panic on shutdown
220 Description: sanity test 65a fails if stripecount of -1 is set
221 Details : handle -1 striping on filesystem in ll_dirstripe_verify
226 Description: ASSERTION(CheckWriteback(page,cmd)) failed
227 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
230 Severity : enhancement
232 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
236 Description: do not process already freed flock
237 Details : flock can possibly be freed by another thread before it reaches
238 to ldlm_flock_completion_ast.
242 Description: LBUG during stress test
243 Details : Need properly lock accesses the flock deadlock detection list.
247 Description: oops in page fault handler
248 Details : kernel page fault handler can return two special 'pages' in error
249 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
253 Description: timeout with invalidate import.
254 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
255 should be handled by ptlrpcd. This produce long age waiting and
256 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
258 Severity : enhancement
260 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
264 Description: enable MGS and MDT services start separately
265 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
266 a co-located MGS without starting the MGS, which is a complement
267 to 'nosvc' mount option.
271 Frequency : after recovery
272 Description: precreate to many object's after del orphan.
273 Details : del orphan st in oscc last_id == next_id and this triger growing
274 count of precreated objects. Set flag LOW to skip increase count
275 of precreated objects.
279 Frequency : rare, on clear nid stats
280 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
281 Details : when clean nid stats sometimes try destroy live entry,
282 and this produce panic in free.
286 Description: Stack overflow during MDS log replay
287 ease stack pressure by using a thread dealing llog_process.
291 Description: wait until IO finished before start new when do lock cancel.
292 Details : VM protocol want old IO finished before start new, in this case
293 need wait until PG_writeback is cleared until check dirty flag and
294 call writepages in lock cancel callback.
296 Severity : enhancement
298 Description: using special macro for print time and cleanup in includes.
302 Description: mds_mfd_close() ASSERTION(rc == 0)
303 Details : In mds_mfd_close(), we need protect inode's writecount change
304 within its orphan write semaphore to prevent possible races.
308 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
312 Frequency : rare, on shutdown ost
313 Description: don't hit live lock with umount ost.
314 Details : shrink_dcache_parent can be in long loop with destroy dentries,
315 use shrink_dcache_sb instead.
319 Description: don't panic with use echo client
320 Details : echo client pass NULL as client nid pointer and this produce null
325 Description: fix build on ppc32
326 Details : compile code with -m64 flag produce wrong object file for ppc32.
330 Description: add message levels for liblustreapi
334 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
335 Details : Move the 'good_osts' check before the 'total_bavail' check. This
336 will result in an -EAGAIN and in the exit call path we call
337 alloc_rr() which will with increasing aggressiveness attempt to
338 aquire precreated objects on the minimum number of required OSCs.
342 Description: Use old size assignment to avoid deadlock
343 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
344 the scheduling while holding a spinlock. We do not need locking
345 for size in ll_update_inode() because size is only updated from
346 the MDS for directories or files without objects, so there is no
347 other place to do the update, and concurrent access to such inodes
348 are protected by the inode lock.
352 Description: resolve "_IOWR redefined" build error on SLES10
356 Description: dump the memory debugging after all modules are unloaded to
357 suppress false negative in conf_sanity test 39
359 Severity : enhancement
361 Description: build kernel-ib packages for OFED 1.3 in our release cycle
366 Description: fix SLES kernel versioning
367 Details : the kernel version for our SLES 10 kernel did not include a "-"
368 before the "smp" at the end. while this was not a problem in
369 general, it did mean that software trying to use the kernel
370 version to try to detect a vendor specific kernel would fail.
371 this was most evident by the OFED build scripts.
375 Description: Don't update lov_desc members until making sure they are valid
376 Details : When updating lov_desc members via proc fs, need fix their
377 validities before doing the real update.
381 Description: don't put request into delay list while invalidate in flight.
382 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
383 import in flight. this produce timeout for invalidate and sometimes
384 can cause stale data.
388 Frequency : on ppc only
389 Description: not convert ost objects for directory because it's not exist.
390 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
394 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
396 Severity : enhancement
398 Description: Update to support 2.6.22.14 vanilla kernel.
402 Frequency : rare, on recovery
403 Description: read procfs can produce deadlock in some situation
404 Details : Holding lprocfs lock which send rpc can produce block for destroy
405 obd objects and this also block reconnect with -EALREADY. This
406 isn't fix all lprocfs bugs - but make it rare.
408 Severity : enhancement
410 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
413 Frequency : frequent on X2 node
415 Description: mdc_set_open_replay_data LBUG
416 Details : Set replay data for requests that are eligible for replay.
420 Description: lustre_mgs: operation 101 on unconnected MGS
421 Details : When MGC is disconnected from MGS long enough, MGS will evict the
422 MGC, and late on MGC cannot successfully connect to MGS and a lot
423 of the error messages complaining that MGS is not connected.
427 Frequency : on network error
428 Description: panic with double free request if network error
429 Details : mdc_finish_enqueue is finish request if any network error ocuring,
430 but it's true only for synchronus enqueue, for async enqueue
431 (via ptlrpcd) this incorrect and ptlrpcd want finish request
434 Severity : enhancement
436 Description: client-side metadata stat-ahead during readdir(directory readahead)
437 Details : perform client-side metadata stat-ahead when the client detects
438 readdir and sequential stat of dir entries therein
441 Frequency : on start mds
443 Description: Implement get_info(last_id) in obdfilter.
446 Frequency : occasional
448 Description: Correctly check stale fid, not start epoch if ost not support SOM
449 Details : open with flag O_CREATE need set old fid in op_fid3 because
450 op_fid2 was overwritten with new generated fid, but mds can answer
451 with one of these two fids and both is not stale. Setattr
452 incorrectly started an epoch and assume will be called
453 done_writing, but without SOM done_writing ever being called.
456 Frequency : rare, depends on device drivers and load
458 Description: MDS or OSS nodes crash due to stack overflow
459 Details : Code changes in 1.8.0 increased the stack usage of some functions.
460 In some cases, in conjunction with device drivers that use a lot
461 of stack the MDS (or possibly OSS) service threads could overflow
462 the stack. One change which was identified to consume additional
463 stack has been reworked to avoid the extra stack usage.
466 Frequency : occasional
468 Description: Do not fail import if osc_interpret_create gets -EAGAIN
469 Details : If osc_interpret_create got -EAGAIN it immediately exits and
470 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
471 oscc_has_objects and see OSC has no objests and call
472 oscc_internal_create to resend create request.
474 Severity : enhancement
476 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
478 Severity : enhancement
480 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
483 Frequency : very rare
485 Description: Fix replay if there is an un-replied request and open
486 Details : In some cases, older replay request will revert the
487 mcd->mcd_last_xid on MDS which is used to record the client's
490 Severity : enhancement
492 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
494 Severity : enhancement
496 Description: Add rhel5 support to HEAD.
498 Serverity : enhancement
500 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
505 Description: Don't allow skipping OSTs if index has been specified.
506 Details : Don't allow skipping OSTs if index has been specified, make
507 locking in internal create lots better.
511 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
512 Details : access to bitfield in structure is always rounded to long
513 and this produce problem with not atomic change any bit.
517 Description: Lustre make rpms failed.
518 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
519 builds Lustre rpms from tarball.
522 Frequency : rare on shutdown ost
524 Description: If llog cancel was not send before clean_exports phase, this can
525 produce deadlock in llog code.
526 Details : If llog thread has last reference to obd and call class_import_put
527 this produce deadlock because llog_cleanup_commit_master wait when
528 last llog_commit_thread exited, but this never success because was
529 called from llog_commit_thread.
533 Description: allow userland application know is lost one of stripes.
534 Details : fill lvb_blocks with error code on ost and return it to
535 application if error flag found.
539 Description: NULL lov_tgts causing MDS oops
540 Details : more safe checks for NULL lov_tgts for avoid oops.
542 Severity : enhancement
544 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
548 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
549 Details : If an OST is down the MDS will hang indefinitely in
550 obd_statfs_rqset() waiting for the statfs data. While for
551 MDS QOS usage of statfs, it should not stuck in waiting.
553 Severity : enhancement
555 Description: remote_acl support
556 Details : Support ACL-based permission check for remote user.
557 Support setfacl/getfacl for remote user with the utils
558 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
559 the system "{s,g}etfacl" utils.
561 Severity : enhancement
563 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
565 Severity : enhancement
567 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
571 Description: make mgs_setparam() handle fsname containing dash
572 Details : fsname containing a dash does not work with lctl conf_param
574 Severity : enhancement
576 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
578 Severity : enhancement
580 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
584 Frequency : rare, at shutdown
585 Description: access already free / zero obd_namespace.
586 Details : if client_disconnect_export was called without force flag set,
587 and exist connect request in flight, this can produce access to
588 NULL pointer (or already free pointer) when connect_interpret
589 store ocd flags in obd_namespace.
593 Frequency : only at startup
594 Description: not alloc memory with spinlock held.
595 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
596 if any spinlock held.
598 Severity : enhancement
600 Description: make lustre randomly fail allocating memory
601 Details : Make lustre randomly failed allocating memory for testing purpose.
603 Severity : enhancement
605 Description: lost problems with lov objid file
606 Details : Fixes some scability and access to not inited memory problems
607 in work with lov objdid file.
612 Description: lfs find does not continue on file error
613 Details : Continue other files processing when a file/dir is absent.
617 Description: Inconsistent usage of lustre_pack_reply()
618 Details : Standardize the usage of lustre_pack_reply() such that it
619 always generate a CERROR on failure.
622 Frequency : occasional
624 Description: MDS hang or stay in waiting lock
625 Details : If client receive lock with CBPENDING flag ldlm need send lock
626 cancel as separate rpc, to avoid situation when cancel request
627 can't processed due all i/o threads stay in wait lock.
631 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
635 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
639 Description: Update to SLES9 kernel-2.6.5-7.287.3.
643 Description: Update to RHEL4 latest kernel.
645 Severity : enhancement
647 Description: Build SLES10 patchless client fails
648 Details : The configure was broken by run ./configure with
649 --with-linux-obj=.... argument for patchless client. When the
650 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
651 can't search header adequately. Use absolute path such as
652 -I($LINUX)/include instead.
656 Description: interrupt oig_wait produce painc on resend.
657 Details : brw_redo_request can be used for resend requests from ptlrpcd and
658 private set, and this produce situation when rq_ptlrpcd_data not
659 copyed to new allocated request and triggered LBUG on assert
660 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
661 ptlrpcd set if request is changed and can be safety changed to use
666 Description: LASSERT_{REQ,REP}SWAB macros are buggy
667 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
668 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
669 these macros and replace them with their difinitions instead.
673 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
674 Details : Update which_patch & target file for SLES10 latest kernel.
677 Description: Remove client patches from SLES 10 kernel.
678 Details : This causes SLES 10 clients to behave as patchless clients
679 even on a Lustre-patched (server) kernel.
681 Severity : enhancement
683 Description: self-adjustable client's lru lists
684 Details : use adaptive algorithm for managing client cached locks lru
685 lists according to current server load, other client's work
686 pattern, memory activities, etc. Both, server and client
687 side namespaces provide number of proc tunables for controlling
690 Severity : enhancement
692 Description: light-weight GSS support
693 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
694 and reduce performance overhead.
696 Severity : enhancement
698 Description: Linux keyring support
699 Details : Support using service of Linux keyring for Lustre GSS internal
700 context refresh/cache mechanism.
704 Description: Fix errors in lfs documentation
705 Details : Fixes man pages
709 Description: don't use GFP_* in generic Lustre code.
710 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
714 Description: obdclass is limited by single OBD_ALLOC(idarray)
715 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
719 Description: warnings with build patchless client with vanila 2.6.19 and up
720 Details : change old ctl_table style and replace ctl_table/ctl_table_header
721 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
725 Frequency : Only for SLES
726 Description: sanity_quota fail test_1
727 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
728 file, which makes TSTID[2] inunique.
732 Frequency : Only for SLES9
733 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
734 Details : Update target/ChangeLog/which_patch .
738 Frequency : Only for test_10 in sanity-quota.sh
739 Description: a bug in quota test code
740 Details : checking a test flag in a wrong place
744 Frequency : Only for test_13 in sanity-quota.sh
745 Description: a bug in quota test script
746 Details : shouldn't check overfull allocated quota
750 Description: jbd statistics
751 Details : Port older jbd statistics patch for sles10
753 Severity : enhancement
755 Description: Add printing inode info into message about error in writepage.
760 Description: reply_lock_interpret crash due to race with it and lock cancel.
761 Details : Do not replay locks that are being cancelled. Do not reference
762 locks by their address during replay, just by their handle.
766 Frequency : When flocks are used.
767 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
768 Details : Flock locks might destroy just granted lock if it could be merged
769 with another existing flock, this is done in completion handler,
770 so teach ldlm_cli_enquque_fini that this is a valid case for
776 Description: Oops in read and write path when failing to allocate lock.
777 Details : Check if lock allocation failed and return error back.
781 Description: lstripe command fails for valid OST index
782 Details : The stripe offset is compared to lov->desc.ld_tgt_count
783 instead of lov->desc.ld_active_tgt_count.
787 Description: sanity.sh failed test 103
788 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
793 Description: df doesn't work properly if diskfs blocksize != 4K
794 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
798 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
799 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
801 Severity : enhancement
803 Description: Adaptive timeouts
804 Details : RPC timeouts adapt to changing server load and network
805 conditions to reduce resend attempts and improve recovery time.
809 Description: llapi_file_create() does not allow some changes
810 Details : add llapi_file_open() that allows specifying the mode and
811 open flags, and also returns an open file handle.
815 Description: merge and cleanup kernel patches.
816 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
820 Description: Add journal checksum support.(Kernel part)
821 Details : The journal checksum feature adds two new flags i.e
822 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
823 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
824 indicates that the commit block contains the checksum for
825 the blocks described by the descriptor blocks. Now commit
826 record can be sent to disk without waiting for descriptor
827 blocks to be written to disk. This behavior is controlled
828 using JBD2_FEATURE_ASYNC_COMMIT flag.
832 Description: OSS needs mutliple precreate threads
833 Details : Add ability to start more than one create thread per OSS.
837 Description: Sanity.sh test_65k failed: setstripe should have succeeded
838 Details : execute lfs setstripe on client
842 Description: mds_obd_create error creating tmp object
843 Details : When the user sets quota on root, llog will be affected and can't
844 create files and write files.
847 Frequency : Always on ia64 patchless client, and possibly others.
849 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
850 Details : This allows the patchless client to be loaded on architectures
855 Description: RedHat Update kernel for RHEL5
856 Details : Add the kernel config file for RHEL5.
860 Description: RedHat Update kernel for RHEL5
861 Details : Modify the kernel config file more closer RHEL5.
865 Description: Build failure against Centos5 (RHEL5)
866 Details : Define PAGE_SIZE when it isn't present.
870 Description: test_7 on sanity-quota.sh can't be used on separate machines
871 Details : add facet to handle it
875 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
876 Details : with NFS, the anon dentry's parent was set to itself in
877 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
878 corresponding dentry other than use rec->ur_name.
880 Severity : enhancement
882 Description: lfs setstripe enhancement
883 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
887 Description: enable data checksumming by default
888 Details : enable checksum by default, allow --disable-checksum
889 configure option and "-o nochecksum" mount option. Checksums
890 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
894 Description: lustre support for RHEL5
895 Details : Add support for RHEL5.
899 Description: Client eviction due to failover config
900 Details : after a connection loss, the lustre client should attempt to
901 reconnect to the last active server first before trying the
902 other potential connections.
906 Description: when mds and osts use different quota unit(32bit and 64bit),
907 quota will be released repeatly.
908 Details : void sending multiple quota reqs to mds, which will keep the
909 status between the reqs.
913 Description: osts not allocated evenly to files
914 Details : change the condition to increase offset_idx
917 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
919 Description: Data corruption for OSTs that are formatted larger than 2TB
921 Details : When generating the bio request for lustre file writes the
922 sector number would overflow a temporary variable before being
923 used for the IO. The data reads correctly from Lustre (which
924 will overflow in a similar manner) but other file data or
925 filesystem metadata may be corrupted in some cases.
929 Description: Tune the kernel for good SCSI performance.
930 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
931 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
936 Description: rewrite ext2-derived code in obdclass/uuid.c
937 Details : rewrite inherited code (uuid parsing code from ext2 utils)
938 from scratch preserving functionality.
942 Description: "lfs find -obd UUID" prints directories
943 Details : "lfs find -obd UUID" will return all directory names instead
944 of just file names. It is incorrect because the directories
945 do not reside on the OSTs.
949 Description: use i_size_read and i_size_write in 2.6 port
950 Details : replace inode->i_size access with i_size_read/write()
953 Frequency : when removing large files
955 Description: scheduling issue during removal of large Lustre files
956 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
957 It causes scheduling issues when removing large files (17TB in the
961 Frequency : only with liblustre clients on XT3
963 Description: evictions taking too long
964 Details : allow llrd to evict clients directly on OSTs
967 Frequency : only on ppc
969 Description: /proc/fs/lustre/devices broken on ppc
970 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
971 We should make sure the type of 'index' is loff_t and then cast
972 explicitly as needed below. Do not assign an explicitly cast
976 Frequency : only for rhel5
978 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
979 Details : Modify the target file & which_kernel.
981 Severity : enhancement
983 Description: omit set fsid for export NFS
984 Details : fix set/restore device id for avoid EMFILE error and mark lustre
985 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
989 Frequency : Always, for kernels after 2.6.16
990 Description: Fix warning idr_remove called for id=.. which is not allocated.
991 Details : Last kernels save old s_dev before kill super and not allow
992 to restore from callback - restore it before call kill_anon_super.
996 Description: buffer overruns could theoretically occur
997 Details : llapi_semantic_traverse() modifies the "path" argument by
998 appending values to the end of the origin string, and a
999 overrun may occur. Adding buffer overrun check in liblustreapi.
1003 Description: Fix error on 'ls .' at the top of the Lustre mount.
1004 Details : Don't revalidate dentry if it is a root dentry.
1008 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1009 Details : Modify vm-tunables-rhel4.patch.
1013 Description: Kernel config for 2.6.18-vanilla.
1014 Details : Modify targets/2.6-vanilla.target.in.
1015 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1016 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1017 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1018 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1022 Description: improve handling recoverable errors
1023 Details : if request processig with error which can be recoverable on server
1024 request should be resend, otherwise page released from cache and
1029 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1030 Details : Modify target file & which_patch.
1031 A flaw was found in the IA32 system call emulation provided
1032 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1033 value could be stored in the %RAX register, which could trigger an
1034 out-of-bounds system call table access. An untrusted local user
1035 could exploit this flaw to run code in the kernel
1036 (ie a root privilege escalation). (CVE-2007-4573).
1040 Description: O_DIRECT bypasses client statistics.
1041 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1042 read_ahead_stats were updated. copied the stats section from
1043 osc_send_oap_rpc() into async_internal().
1047 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1051 Description: change order of libsysio includes
1052 Details : '#include sysio.h' should always come before '#include xtio.h'
1056 Description: handle "serious error: objid * already exists" more gracefully
1057 Details : If LAST_ID value on disk is smaller than the objects existing in
1058 the O/0/d* directories, it indicates disk corruption and causes an
1059 LBUG(). If the object is 0-length, then we should use the existing
1060 object. This will help to avoid a full fsck in most cases.
1062 Severity : enhancement
1064 Description: adapt the lustre_config script to support the upgrade case
1065 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1066 server targets to 1.6 in parallel.
1070 Description: To avoid grant space > avaible space when the disk is almost
1071 full. Without this patch you might see the error "grant XXXX >
1072 available" or some LBUG about grant, when the disk is almost
1074 Details : In filter_check_grant, for non_grant cache write, we should
1075 check the left space by if (*left > ungranted + bytes), instead
1076 of (*left > ungranted), because only we are sure the left space
1077 is enough for another "bytes", then the ungrant space should be
1078 increase. In client, we should update cl_avail_grant only there
1079 is OBD_MD_FLGRANT in the reply.
1083 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1084 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1085 described in CVE-2007-4573. This problem could allow untrusted
1086 local users to gain root access.
1089 Frequency : when using O_DIRECT and quotas
1091 Description: Incorrect file ownership on O_DIRECT output files
1092 Details : block usage reported by 'lfs quota' does not take into account
1093 files that have been written with O_DIRECT.
1098 Description: touch file failed when fs is not full
1099 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1100 otherwise we can get ENOSP while fs is not full.
1104 Description: parallel lock callbacks
1105 Details : Instead of sending blocking and completion callbacks as separated
1106 requests, adding them to a set and sending in parallel.
1109 Frequency : only for Cray XT3
1110 Bugzilla : 12829/13455
1111 Description: Changing primary group doesn't change the group lustre assigns to
1113 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1114 overridden with the primary group provided by the group upcall,
1115 whereas the supplied fsgid can be trusted if it is in the list of
1116 supplementary groups returned by the group upcall.
1118 Severity : enhancement
1120 Description: Allow masking D_WARNING, D_ERROR messages from console
1121 Details : Console messages can now be disabled via lnet.printk.
1125 Description: User code with malformed file open parameter crashes client node
1126 Details : Before packing join_file req, all the related reference should be
1127 checked carefully in case some malformed flags cause fake
1128 join_file req on client.
1132 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1133 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1134 mode is equal to the granted mode. However, it should also check
1135 that LDLM_FL_CP_REQD is not set.
1139 Description: Heavy nfs access might result in deadlocks
1140 Details : After ELC code landed, it is now improper to enqueue any mds
1141 locks under och_sem, because enqueue might want to decide to
1142 cancel open locks for same inode we are holding och_sem for.
1146 Description: Client eviction while running blogbench
1147 Details : A lot of unlink operations with concurrent I/O can lead to a
1148 deadlock causing evictions. To address the problem, the number of
1149 oustanding OST_DESTROY requests is now throttled to
1150 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1151 callbacks are processed in priority.
1155 Description: enable ACLs on MDS by default
1156 Details : ACLs must be enabled on MDS by default.
1159 Frequency : PPC/PPC64 only
1161 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1162 Details : fix duplicated definitions between asm-ppc64/types.h and
1163 lustre_types.h on PPC.
1166 Frequency : PPC/PPC64 only
1168 Description: asm-ppc/segment.h does not exist
1169 Details : fix compile issue on PPC.
1173 Description: better handle error messages in extents code
1176 Frequency : RHEL4 only
1178 Description: mkfs is very slow on IA64/RHEL4
1179 Details : A performance regression has been discovered in the MPT Fusion
1180 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1181 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1182 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1184 Severity : enhancement
1186 Description: SNMP support enhancement
1187 Details : Adding total number of sampled request for an MDS node in snmp
1190 Severity : enhancement
1192 Description: Optimize ldlm waiting list processing for PR extent locks
1193 Details : When processing waiting list for read extent lock and meeting read
1194 lock that is same or wider to it that is not contended, skip
1195 processing rest of the list and immediatelly return current
1196 status of conflictness, since we are guaranteed there are no
1197 conflicting locks in the rest of the list.
1201 Description: Time out and refuse to reconnect
1202 Details : When the failover node is the primary node, it is possible
1203 to have two identical connections in imp_conn_list. We must
1204 compare not conn's pointers but NIDs, otherwise we can defeat
1205 connection throttling.
1209 Description: port llog fixes from b1_6 into HEAD
1210 Details : Port llog reference couting and some llog cleanups from b1_6
1211 (bug 10800) into HEAD, for protect from panic and access to already
1212 free llog structures.
1216 Description: Detect stride IO mode in read-ahead
1217 Details : When a client does stride read, read-ahead should detect that and
1218 read-ahead pages according to the detected stride pattern.
1222 Description: data checksumming impacts single node performance
1223 Details : add support for several checksum algorithm. Currently, only CRC32
1224 and Adler-32 are supported. The checksum type can be changed on
1225 the fly via /proc/fs/lustre/osc/*/checksum_type.
1229 Description: use adler32 for page checksums
1230 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1235 Description: build for x2 fails
1236 Details : fix compile issue on Cray systems.
1240 Description: Properly match for duplicate locks
1241 Details : Due to different lock order from skiplists code, we need to
1242 traverse entire list for now
1245 Frequency : only on PPC/SLES10
1247 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1248 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1249 be defined. Add a hack in mkfs_lustre.c to work around this compile
1254 Description: LASSERT on MDS when client holding flock lock dies
1255 Details : ldlm pool logic depends on number of granted locks equal to
1256 number of released locks which is not true for flock locks, so
1257 just exclude such locks from consideration.
1261 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1262 Details : Use fsfilt_sync() for both the whole filesystem sync and
1263 individual file sync to eliminate dangerous inode locking
1264 with I_LOCK that can lead to a deadlock.
1268 Description: performance in 1.6.3
1269 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1270 sure that 1MB requests are not fragmented by the block layer.
1272 Severity : enhancement
1274 Description: organize the server-side client stats on per-nid basis
1275 Details : Change the structure of stats under obdfilter and mds to
1283 The "uuid"s file would list the uuids of _active_ exports.
1284 And the clear entry is to clear all stats and stale nids.
1286 Severity : enhancement
1288 Description: eliminate client locks in face of contention
1289 Details : file contention detection and lockless i/o implementation
1290 for contended files.
1294 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1297 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1299 Description: MDS may lose file striping (and hence file data) in some cases
1300 Details : If there are additional extended attributes stored on the MDS,
1301 in particular ACLs, SELinux, or user attributes (if user_xattr
1302 is specified for the client mount options) then there is a risk
1303 of attribute loss. Additionally, the Lustre file striping
1304 needs to be larger than default (e.g. striped over all OSTs),
1305 and an additional attribute must be stored initially in the
1306 inode and then increase in size enough to be moved to the
1307 external attribute block (e.g. ACL growing in size) for file
1312 Description: skiplist implementation simplification
1313 Details : skiplists are used to group compatible locks on granted list
1314 that was implemented as tracking first and last lock of each
1315 lock group the patch changes that to using doubly linked lists
1319 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1320 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1321 because LASSERT on both the data supplied by a client, and the
1322 data on disk is dangerous and incorrect.
1324 Severity : enhancement
1326 Description: Slow truncate/writes to huge files at high offsets.
1327 Details : Directly associate cached pages to lock that protect those pages,
1328 this allows us to quickly find what pages to write and remove
1329 once lock callback is received.
1333 Description: more ldlm soft lockups
1334 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1335 starve other threads from the resource lock for a long time in
1336 case of long waiting queue, so change the debug level from
1337 D_OTHER to the less frequently used D_INFO.
1339 Severity : enhancement
1341 Description: add -gid, -group, -uid, -user options to lfs find
1345 Description: Hung threads in invalidate_inode_pages2_range
1346 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1347 once one is completed. As a result, some RPCs are stuck in the
1348 queue and are never sent.
1352 Description: filter threads hungs on waiting journal commit
1353 Details : Cleanup filter group llog code, then only filter group llog will
1354 be only created in the MDS/OST syncing process.
1358 Description: Procfs and llog threads access destoryed import sometimes.
1359 Details : Sync the import destoryed process with procfs and llog threads by
1360 the import refcount and semaphore.
1362 Severity : enhancement
1364 Description: openlock cache of b1_6 port to HEAD
1369 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1370 Details : If insertion of an extent fails, then discard the inode
1371 preallocation and free data blocks else it can lead to duplicate
1376 Description: don't always update ctime in ext3_xattr_set_handle()
1377 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1378 In some cases the ctime should not be updated, for example for
1379 2.0->1.8 compatibility it is necessary to delete an xattr and it
1380 should not update the ctime.
1384 Bugzilla : 15713/16362
1385 Description: Assertion in iopen_connect_dentry in 1.6.3
1386 Details : looking up an inode via iopen with the wrong generation number can
1387 populate the dcache with a disconneced dentry while the inode
1388 number is in the process of being reallocated. This causes an
1389 assertion failure in iopen since the inode's dentry list contains
1390 both a connected and disconnected dentry.
1394 Description: assertion failure in ldlm_handle2lock()
1395 Details : fix a race between class_handle_unhash() and class_handle2object()
1396 introduced in lustre 1.6.5 by bug 13622.
1401 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1402 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1407 Description: oops in sd_iostats_seq_show()
1408 Details : unloading/reloading the scsi low level driver triggers a kernel
1409 bug when trying to access the sd iostat file.
1414 Description: Kernel panics during QLogic driver reload
1415 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1416 patch, causing memory corruption.
1421 Description: journal_dev option does not work in b1_6
1422 Details : pass mount option during pre-mount.
1424 Severity : enhancement
1426 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1427 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1428 extent information of a file. It can be used to map logical blocks
1429 in a file to physical blocks in the block device.
1433 Description: LDLM soft lockups - improvement
1434 Details : It is be possible to send the lock handle along with each read
1435 or write request because the client is already doing a lock match
1436 itself so there isn't any reason the OST should have to re-do that
1440 Frequency : only on Cray X2
1442 Description: X2 build failures
1443 Details : fix build failures on Cray X2.
1447 Description: xid & resent requests
1448 Details : Initialize RPC XID from clock at startup (randomly if clock is
1451 Severity : enhancement
1453 Description: Add lustre_start utility to start or stop multiple Lustre servers
1458 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1459 Details : In case of memory pressure, list_del() can be called twice on
1460 req->rq_history_list, causing a kernel oops.
1464 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1465 Details : fix stack overflow in the distributed lock manager by defering
1466 export eviction after a failed AST to the elt thread instead of
1467 handling it in the dlm interpret routine.
1471 Description: Convert some comments to new format.
1472 Details : Update documenting comments to match doxygen conventions.
1476 Description: Grammar fixes.
1477 Details : A couple of trivial spelling fixes.
1481 Description: OSD_COUNTERS-mandatory
1482 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1483 to simplify the code.
1487 Description: simplify cmm_device freeing logic.
1488 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1492 Description: Add lockdep support to dt_object_operations locking interface.
1493 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1494 parameter indicating lock ordering. Update mdd code to use new
1499 Description: Introduce failloc constants for lockless IO tests.
1500 Details : Add two new failloc constants to test lockless IO. Only one of
1501 them in implemented---another is checked in yet to be landed
1506 Description: Add lockdep support for inode mutex.
1507 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1508 in the situations where Lustre has to lock more than one inode
1513 Description: Add optional invariants checking support.
1514 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1515 configure switch. This macro is to be used for consistency and
1516 sanity checks that are too expensive to be left in `production'
1521 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1522 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1527 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1528 Details : Use spin_lock_nested() in (the only) situation where more than
1529 one ldlm_lock is locked simultaneously. Also, fix possible
1530 dead-lock in ldlm_lock_change_resource() by enforcing particular
1535 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1536 Details : Instead of specifying each ldlm_lock call-back through separate
1537 parameter, wrap them into struct ldlm_callback_suite.
1541 Description: Kill join_lru obd method and its callers.
1542 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1543 in memory---a requirement implemented through ->o_join_lru() obd
1544 method in HEAD. Remove this method and its users.
1548 Description: Add asynchronous ldlm ENQUEUE completion handler.
1549 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1550 case that stock ldlm_completion_ast() cannot handle as it waits
1551 until lock is granted. Introduce new ldlm_completion_ast_async()
1552 for this. Also comment ldlm_completion_ast().
1556 Description: ldlm_error <-> errno conversion.
1557 Details : Add functions to map (rather arbitrary) between LDLM error codes
1558 and standard errno values. CLIO needs this to prevent LDLM specific
1559 constants from escaping ldlm and osc.
1563 Description: Kill unused ldlm_handle2lock_ns() function.
1564 Details : Kill unused ldlm_handle2lock_ns() function.
1568 Description: Add lu_ref support to ldlm_lock
1569 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1570 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1571 and ldlm_resource. LDLM interface has to be changed, because code that
1572 releases a reference on a lock, has to "know" what reference this is.
1573 In the most frequent case
1575 lock = ldlm_handle2lock(handle);
1577 LDLM_LOCK_PUT(lock);
1579 no changes are required. When any other reference (received _not_ from
1580 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1581 instead of LDLM_LOCK_PUT().
1583 Arguably, changes are pervasive, and interface requires some discipline
1584 for proper use. On the other hand, it was very instrumental in finding
1585 a few leaked lock references.
1589 Description: Add ldlm_lock_addref_try().
1590 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1591 attempts to addref a lock that might be being canceled
1596 Description: Add ldlm_weigh_callback().
1597 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1598 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1599 hard-coded `number of pages' logic.
1603 Description: Add lockdep annotations to llog code.
1604 Details : Use appropriately tagged _nested() locking calls in the places
1605 where llog takes more than one ->lgh_lock lock.
1609 Description: Add loi_kms_set().
1610 Details : Wrap kms updates into a helper function.
1614 Description: Constify instances of struct lsm_operations.
1615 Details : Constify instances of struct lsm_operations.
1619 Description: lu_conf support.
1620 Details : On a server, a file system object is uniquely identified
1621 by a fid, which is sufficient to locate and load all object
1622 state (inode). On a client, on the other hand, more data are
1623 necessary instantiate an object. Change lu_object_find() and
1624 friends to take additional `lu_conf' argument describing object.
1625 Typically this includes layout information.
1629 Description: lu_context fixes.
1630 Details : Introduce new lu_context functions that are needed on the client
1631 side, where some system threads (ptlrpcd) are shared by multiple
1632 modules, and so cannot be stopped during module shutdown.
1636 Description: Add start and stop methods to lu_device_type_operations.
1637 Details : Introduce two new methods in lu_device_type_operations, that are
1638 invoked when first instance of a given type is created and last one
1639 is destroyed respectively. This is need by CLIO.
1643 Description: Add lu_ref support to struct lu_device.
1644 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1645 track leaked references.
1649 Description: Introduce lu_kmem_descr.
1650 Details : lu_kmem_descr and its companion interface allow to create
1651 and destroy a number of kmem caches at once.
1655 Description: Fix lu_object finalization race.
1656 Details : Fix a race between lu_object_find() finding an object and its
1657 concurrent finalization. This race is (most likely) not possible
1658 on the server, but might happen on the client.
1662 Description: Introduce lu_ref interface.
1663 Details : lu_ref is a debugging module allowing to track references to
1664 a given object. It is quite cpu expensive, and has to be
1665 explicitly enabled with --enable-lu_ref. See usage description
1670 Description: Factor lu_site procfs stats into a separate function.
1671 Details : Separate lu_site stats printing code into a separate function
1672 that can be reused on a client.
1676 Description: Constify instances of struct {lu,dt,md}_device_operations.
1677 Details : Constify instances of struct {lu,dt,md}_device_operations.
1681 Description: Introduce struct md_site and move meta-data specific parts of
1682 struct lu_site here.
1683 Details : Move md-specific fields out of struct lu_site into special struct
1684 md_site, so that lu_site can be used on a client.
1688 Description: Kill mdd_lov_destroy().
1689 Details : Remove unused mdd code.
1693 Description: Add st_block checking to multistat.c.
1694 Details : Add st_block checking to multistat.c.
1698 Description: Add lu_ref support to struct obd_device.
1699 Details : Add lu_ref tracking to obd_device.
1703 Description: Kill obd_set_fail_loc().
1704 Details : Remove unused code.
1708 Description: Add special type for ptlrpc_request interpret functions.
1709 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1710 there. Actual usage will be in CLIO.
1714 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1715 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1716 doesn't work with lockdep.
1720 Description: Add rwv.c test program.
1721 Details : New testing program exercising readv(2) and writev(2) (Qian).
1725 Description: Add sendfile.c test program.
1726 Details : New testing program exercising sendfile(2) (Jay).
1730 Description: Ratelimit a message that can be very frequent.
1731 Details : Ratelimit a memory allocation failure message that can
1736 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1737 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1741 Description: Liblustre build fixes.
1742 Details : Liblustre build fixes.
1746 Description: libcfs: add cfs_{need,cond}_resched() interface.
1747 Details : libcfs: add cfs_{need,cond}_resched() definition and
1748 implementations for Linux, NT, and liblustre.
1750 Severity : enhancement
1752 Description: More exported tunables for mballoc
1753 Details : Add support for tunable preallocation window and new tunables for
1754 large/small requests
1758 Description: Detect corruption of block bitmap and checking for preallocations
1759 Details : Checks validity of on-disk block bitmap. Also it does better
1760 checking of number of applied preallocations. When corruption is
1761 found, it turns filesystem readonly to prevent further corruptions.
1765 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1766 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1767 window to 0. If it does hit cache miss, and read-pattern is still
1768 stride-io mode, does not reset the stride window, but also does not
1769 increase the stride window length in this case.
1773 Frequency : only for big-endian servers
1774 Description: Check if system is big-endian while mounting fs with extents feature
1775 Details : Mounting a filesystem with extents feature will fail on big-endian
1776 systems since ext3-based ldiskfs is not supported on big-endian
1777 systems. This can be over-riden with "bigendian_extents" mount option.
1779 --------------------------------------------------------------------------------
1781 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1783 * Support for kernels:
1784 2.6.5-7.283 (SLES 9),
1785 2.6.9-55.EL (RHEL 4),
1786 2.6.16.46-0.14 (SLES 10),
1787 2.6.18.8 vanilla (kernel.org)
1788 * Client support for unpatched kernels:
1789 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1790 2.6.16 - 2.6.22 vanilla (kernel.org)
1791 * Due to recently discovered recovery problems, we do not recommend
1792 using patchless RHEL 4 clients with this or any earlier release.
1793 * Recommended e2fsprogs version: 1.39.cfs8
1794 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1795 * Starting with this release, the ldiskfs backing filesystem required
1796 by Lustre is now in its own package, lustre-ldiskfs. This package
1797 should be installed. It is versioned separately from Lustre and
1798 may be released separately in future.
1802 Description: block reactivating mgc import until all deactivates complete
1803 Details : Fix race when failing back MDT/MGS to itself (testing)
1806 Frequency : at statup only
1808 Description: mds_lov_synchronize race leads to various problems
1809 Details : simultaneous MDT->OST connections at startup can cause the
1810 sync to abort, leaving the OSC in a bad state.
1812 Severity : enhancement
1814 Description: add optional extra BUILD_VERSION info
1815 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1816 to override the lustre version.
1819 Frequency : 2.6.18 servers only
1821 Description: ll_kern_mount() doesn't release the module reference
1822 Details : The ldiskfs module reference count never drops down to 0
1823 because ll_kern_mount() doesn't release the module reference.
1828 Description: server LBUG when using old ost_num_threads parameter
1829 Details : Accept the old ost_num_threads parameter but warn that it
1830 is deprecated, and fix an off-by-one error that caused an LBUG.
1835 Description: Transient SCSI error results in persistent IO issue
1836 Details : iobuf->dr_error is not reinitialized to 0 between two
1840 Frequency : sometimes when underlying device returns I/O errors
1842 Description: OSTs not going read-only during write failures
1843 Details : OSTs are not remounted read-only when the journal commit threads
1844 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1845 instead of the ext3 wrappers.
1849 Description: poor connect scaling with increasing client count
1850 Details : Don't run filter_grant_sanity_check for more than 100 exports
1851 to improve scaling for large numbers of clients.
1854 Frequency : SLES10 only
1856 Description: sanity-quota.sh quotacheck failed: rc = -22
1857 Details : Quotas cannot be enabled on SLES10.
1860 Frequency : liblustre clients only
1862 Description: getdirentries does not give error when run on compute nodes
1863 Details : getdirentries does not fail when the size specified as an argument
1864 is too small to contain at least one entry
1866 Severity : enhancement
1868 Description: Add LNET router traceability for debug purposes
1869 Details : If a checksum failure occurs with a router as part of the
1870 IO path, the NID of the last router that forwarded the bulk data
1871 is printed so it can be identified.
1876 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1877 Details : Due to a race condition, liblustre clients were occasionally
1878 evicted incorrectly.
1880 Severity : enhancement
1882 Description: lfs setstripe use optional parameters instead of postional
1885 Severity : enhancement
1887 Description: Nanosecond timestamp support for ldiskfs
1888 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1889 resolution timestamps. There is not yet support for this at
1890 the Lustre filesystem level.
1893 Frequency : during server recovery
1895 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1896 Details : request with rq_no_resend flag not awake l_wait_event if they get
1900 Frequency : nfs export on patchless client
1902 Description: connectathon hang when test nfs export over patchless client
1903 Details : Disconnected dentry cannot be found with lookup, so we do not need
1904 to unhash it or make it invalid
1907 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1908 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1909 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1914 Description: open req refcounting wrong on reconnect
1915 Details : If reconnect happened between getting open reply from server and
1916 call to mdc_set_replay_data in ll_file_open, we will schedule
1917 replay for unreferenced request that we are about to free.
1918 Subsequent close will crash in variety of ways.
1919 Check that request is still eligible for replay in
1920 mdc_set_replay_data().
1925 Description: disable writes to filesystem when reading health_check file
1926 Details : the default for reading the health_check proc file has changed
1927 to NOT do a journal transaction and write to disk, because this
1928 can cause reads of the /proc file to hang and block HA state
1929 checking on a healthy but otherwise heavily loaded system. It
1930 is possible to return to the previous behaviour during configure
1931 with --enable-health-write.
1933 Severity : enhancement
1935 Description: 64-bit inode version
1936 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1937 to the inode. This will be required for version-based recovery.
1942 Description: MDS fails to start if a duplicate client export is detected
1943 Details : in some rare cases it was possible for a client to connect to
1944 an MDS multiple times. Upon recovery the MDS would detect this
1945 and fail during startup. Handle this more gracefully.
1947 Severity : enhancement
1949 Description: Add -o localflock option to simulate old noflock
1951 Details : This will achieve local-only flock/fcntl locks
1957 Description: log_commit_thread vs filter_destroy race leads to crash
1958 Details : Take import reference before releasing llog record semaphore
1963 Description: Wrong request locking in request set processing
1964 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1965 imp_delayed_list, in this place should be used imp_lock.
1968 Frequency : when reconnection
1970 Description: Grant Leak when osc reconnect to OST
1971 Details : When osc reconnect ost, OST(filter) should check whether it
1972 should grant more space to client by comparing fed_grant and
1973 cl_avail_grant, and return the granted space to client instead
1974 of "new granted" space, because client will call osc_init_grant
1975 to update the client grant space info.
1978 Frequency : when client reconnect to OST
1980 Description: Grant Leak when osc do resend and replay bulk write
1981 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1982 bulk write request, because the grant info while be sync between
1983 OSC and OST when reconnect, and we should ignore the grant info
1984 these of resend/replay write req.
1989 Description: Grant space more than avaiable left space sometimes.
1990 Details : When then OST is about to be full, if two bulk writing from
1991 different clients came to OST. Accord the avaliable space of the
1992 OST, the first req should be permitted, and the second one
1993 should be denied by ENOSPC. But if the seconde arrived before
1994 the first one is commited. The OST might wrongly permit second
1995 writing, which will cause grant space > avaiable space.
1998 Frequency : when client is evicted
2000 Description: Grant might be wrongly erased when osc is evicted by OST
2001 Details : when the import is evicted by server, it will fork another
2002 thread ptlrpc_invalidate_import_thread to invalidate the
2003 import, where the grant will be set to 0. While the original
2004 thread will update the grant it got when connecting. So if
2005 the former happened latter, the grant will be wrongly errased
2006 because of this race.
2011 Description: Checking Stale with correct fid
2012 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2013 to check whether it is stale, because sometimes, we want the
2014 enqueue happened anyway, and op_data.fid2 will not be initialized.
2016 Severity : enhancement
2018 Description: update patchless client
2019 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2022 Frequency : only with 2.4 kernel
2024 Description: random memory corruption
2025 Details : size of struct ll_inode_info is to big for union inode.u and this
2026 can be cause of random memory corruption.
2031 Description: Memory leak in recovery
2032 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2033 It should also check obd_fail before fsfilt_start, otherwise if
2034 fsfilt_start return -EROFS,(failover mds during mds recovery).
2035 then the req will return with repmsg->transno = 0 and rc = EROFS.
2036 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2037 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2038 be freed no matter whether fsfilt_commit success or not.
2041 Frequency : only with huge count clients
2043 Description: Prevents from taking the superblock lock in llap_from_page for
2045 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2046 for llap_from_page call in ll_removepage prevents from taking the
2047 superblock lock for a soon died page.
2052 Description: Not check open intent error before release open handle
2053 Details : in some rare cases, the open intent error is not checked before
2054 release open handle, which may cause
2055 ASSERTION(open_req->rq_transno != 0), because it tries to release
2056 the failed open handle.
2061 Description: Set cat log bitmap only after create log success.
2062 Details : in some rare cases, the cat log bitmap is set too early. and it
2063 should be set only after create log success.
2067 Description: Accessing a block bevice can re-enable I/O when Lustre is
2068 tearing down a device.
2069 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2074 Description: service threads may hog cpus when there are a lot of requests
2075 Details : Insert cond_resched to give other threads a chance to use some CPU
2080 Description: the cat log was not initialized in recovery
2081 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2082 unlink log on mds will not be initialized until mds post
2083 recovery. And also in mds post recovery, the unlink log will
2084 initialization will be done asynchronausly, so there will be race
2085 between add unlink log and unlink log initialization.
2089 Description: brw_stats were being printed incorrectly
2090 Details : brw_stats were being printed as log2 but all of them were not
2091 recorded as log2. Also remove some code duplication arising from
2092 filter_tally_{read,write}.
2096 Frequency : rare, only in recovery.
2097 Description: ASSERTION(req->rq_type != LI_POISON) failed
2098 Details : imp_lock should be held while iterating over imp_sending_list for
2099 prevent destroy request after get timeout in ptlrpc_queue_wait.
2103 Description: replay-single.sh test 52 fails
2104 Details : A lock's skiplist need to be cleanup when it being unlinked
2105 from its resource list.
2109 Description: Short directio read returns full requested size rather than
2111 Details : Direct I/O operations should return actual amount of bytes
2112 transferred rather than requested size.
2114 Severity : enhancement
2116 Description: metadata RPC reduction (e.g. for rm performance)
2117 Details : decrease the amount of synchronous RPC between clients and servers
2118 by canceling conflicing lock before the operation on the client
2119 and packing thier handles into the main operation RPC to server.
2121 Severity : enhancement
2123 Description: Async OSC create to avoid the blocking unnecessarily.
2124 Details : If a OST has no remain object, system will block on the creating
2125 when need to create a new object on this OST. Now, ways use
2126 pre-created objects when available, instead of blocking on an
2127 empty osc while others are not empty. If we must block, we block
2128 for the shortest possible period of time.
2132 Description: improve handling recoverable errors
2133 Details : if request processig with error which can be recoverable on server
2134 request should be resend, otherwise page released from cache and
2137 Severity : enhancement
2139 Description: refine locking for avoid write wrong info into lov_objid file
2140 Details : fix possible races with add new target and write/update data in
2143 --------------------------------------------------------------------------------
2145 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2150 Frequency : on some architectures
2152 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2153 Details : Uninitialized flags sometimes cause configuration commands to
2157 Frequency : patchless clients only
2159 Description: missing __iget() symbol export
2160 Details : The __iget() symbol export is missing. To avoid the need for
2161 this on patchless clients the deathrow inode reaper is turned
2162 off, and we depend on the VM to clean up old inodes. This
2163 dependency was during via the fix for bug 12181.
2165 --------------------------------------------------------------------------------
2167 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2169 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2170 INTEROPERATE with older versions automatically. In many cases a
2171 special upgrade step is needed. Please read the
2172 user documentation before upgrading any part of a live system.
2173 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2174 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2175 * WARNING: Lustre configuration and startup changes are required with
2176 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2180 Severity : enhancement
2182 Description: MountConf
2183 Details : Lustre configuration is now managed via mkfs and mount
2184 commands instead of lmc and lconf. New obd types (MGS, MGC)
2185 are added for dynamic configuration management. See
2186 https://mail.clusterfs.com/wikis/lustre/MountConf for
2189 Severity : enhancement
2191 Description: dynamic OST addition
2192 Details : OSTs can now be added to a live filesystem
2194 Severity : enhancement
2196 Description: startup order invariance
2197 Details : MDTs and OSTs can be started in any order. Clients only
2198 require the MDT to complete startup.
2200 Severity : enhancement
2202 Description: parallel, asynchronous orphan cleanup
2203 Details : orphan cleanup is now performed in separate threads for each
2204 OST, allowing parallel non-blocking operation.
2206 Severity : enhancement
2208 Description: optimized stripe assignment
2209 Details : stripe assignments are now made based on ost space available,
2210 ost previous usage, and OSS previous usage, in order to try
2211 to optimize storage space and networking resources.
2213 Severity : enhancement
2215 Description: Permanently set tunables
2216 Details : All writable /proc/fs/lustre tunables can now be permanently
2217 set on a per-server basis, at mkfs time or on a live system.
2219 Severity : enhancement
2221 Description: Lustre message v2
2222 Details : Add lustre message format v2.
2224 Severity : enhancement
2226 Description: client OST exclusion list
2227 Details : Clients can be started with a list of OSTs that should be
2228 declared "inactive" for known non-responsive OSTs.
2232 Description: SPEC SFS validation failure on NFS v2 over lustre.
2233 Details : Changes the blocksize for regular files to be 2x RPC size,
2234 and not depend on stripe size.
2236 Severity : enhancement
2238 Description: Multiple MD RPCs in flight.
2239 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2240 To avoid overly-overloading MDS, introduce a limit on number of
2241 MDS RPCs in flight for a single client and add /proc controls
2242 to adjust this limit.
2244 Severity : enhancement
2246 Description: client read/write statistics
2247 Details : Add client read/write call usage stats for performance
2248 analysis of user processes.
2249 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2250 file access. extents_stats shows chunk size distribution.
2251 extents_stats_per_process show chunk size distribution per
2254 Severity : enhancement
2256 Description: mds statistics
2257 Details : Add detailed mds operations statistics in
2258 /proc/fs/lustre/mds/*/stats.
2262 Description: Failure of copying files with lustre special EAs.
2263 Details : Client side always return success for setxattr call for lustre
2264 special xattr (currently only "trusted.lov").
2269 Description: Refcount LNET uuids
2270 Details : The global LNET uuid list grew linearly with every startup;
2271 refcount repeated list entries instead of always adding to
2274 Severity : enhancement
2276 Description: Dynamic service threads
2277 Details : Within a small range, start extra service threads
2278 automatically when the request queue builds up.
2281 Frequency : mixed-endian client/server environments
2283 Description: mixed-endian crashes
2284 Details : The new msg_v2 system had some failures in mixed-endian
2287 Severity : enhancement
2289 Description: Easy OST removal
2290 Details : OSTs can be permanently deactivated with e.g. 'lctl
2291 conf_param lustre-OST0001.osc.active=0'
2293 Severity : enhancement
2295 Description: MGS proc entries
2296 Details : Added basic proc entries for the MGS showing what filesystems
2299 Severity : enhancement
2301 Description: provide MGS failover
2302 Details : Added config lock reacquisition after MGS server failover.
2304 Severity : enhancement
2306 Description: add Linux 2.4 support
2307 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2311 Description: a large application tries to do I/O to the same resource and dies
2312 in the middle of it.
2313 Details : Check the req->rq_arrival time after the call to
2314 ost_brw_lock_get(), but before we do anything about
2315 processing it & sending the BULK transfer request. This
2316 should help move old stale pending locks off the queue as
2317 quickly as obd_timeout.
2320 Frequency : when an incorrect nid is specified during startup
2322 Description: ptlrpc connect to non-existant node causes kernel crash
2323 Details : LNET can't be re-entered from an event callback, which
2324 happened when we expire a message after the export has been
2325 cleaned up. Instead, hand the zombie cleanup off to another
2328 Severity : enhancement
2330 Description: plain/inodebits lock performance improvement
2331 Details : Grouping plain/inodebits in granted list by their request modes
2332 and bits policy, thus improving the performance of search through
2336 Frequency : only if OST filesystem is corrupted
2338 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2339 Details : for a short time RPCs with bulk IO are in the replay list,
2340 but replay of bulk IOs is unimplemented. If the OST filesystem
2341 is corrupted due to disk cache incoherency and then replay is
2342 started it is possible to trip an assertion. Avoid putting
2343 committed RPCs into the replay list at all to avoid this issue.
2346 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2348 Description: System hang on startup
2349 Details : This bug allowed the liblustre (e.g. catamount) client to
2350 return to the app before handling all startup RPCs. This
2351 could leave the node unresponsive to lustre network traffic
2352 and manifested as a server ptllnd timeout.
2354 Severity : enhancement
2356 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2357 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2358 boolean to control whether to print peer debug info when a
2359 client's RPC times out.
2362 Frequency : only for kernels with patches from Lustre below 1.4.3
2364 Description: Remove old rdonly API
2365 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2368 Frequency : only for devices with external journals
2370 Description: Set external device read-only also
2371 Details : During a commanded failover stop, we set the disk device
2372 read-only while the server shuts down. We now also set any
2373 external journal device read-only at the same time.
2376 Frequency : when upgrading from 1.4 while trying to change parameters
2378 Description: The wrong (new) MDC name was used when setting parameters for
2379 upgraded MDT's. Also allows changing of OSC (and MDC)
2380 parameters if --writeconf is specified at tunefs upgrade time.
2383 Frequency : when setting specific ost indicies
2385 Description: QOS code breaks on skipped indicies
2386 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2387 created with --index need not be sequential.
2389 Severity : enhancement
2391 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2392 Details : The uninit_groups feature works in conjunction with the kernel
2393 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2394 up the pass1 processing of e2fsck. This is a read-only feature
2395 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2396 mount filesystems that have had this feature enabled.
2398 Severity : enhancement
2400 Description: Improve multi-block allocation algorithm to avoid fragmentation
2401 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2402 allocation locality and avoid filesystem fragmentation.
2404 ------------------------------------------------------------------------------
2406 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2408 * Support for kernels:
2409 2.4.21-47.0.1.EL (RHEL 3)
2410 2.6.5-7.283 (SLES 9)
2411 2.6.9-42.0.10.EL (RHEL 4)
2412 2.6.12.6 vanilla (kernel.org)
2413 2.6.16.27-0.9 (SLES 10)
2414 * Recommended e2fsprogs version: 1.39.cfs5
2416 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2420 Frequency : occasional, depends on client load and configuration
2421 Bugzilla : 12181, 12203
2422 Description: data loss for recently-modified files
2424 Details : In some cases it is possible that recently written or created
2425 files may not be written to disk in a timely manner (this should
2426 normally be within 30s unless client IO load is very high).
2427 The problem appears as zero-length files or files that are a
2428 multiple of 1MB in size after a client crash or client eviction
2429 that are missing data at the end of the file.
2431 This problem is more likely to be hit on clients where files are
2432 repeatedly created and unlinked in the same directory, clients
2433 have a large amount of RAM, have many CPUs, the filesystem has
2434 many OSTs, the clients are rebooted frequently, and/or the files
2435 are not accessed by other nodes after being written.
2437 The presence of the problem can be detected by looking at
2438 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2439 smaller than the second (nr_unused) then dirty files will not
2440 be flushed automatically to disk. "sync; sleep 10" should be
2441 run several times on the node before unmounting it to update
2442 Lustre (this is also safe to run on nodes without this problem).
2444 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2445 that can cause this same problem, so customers using that kernel
2446 also need to update the kernel in addition to Lustre. In order
2447 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2449 It is normal that files written just before a client crash (less
2450 than 30s) may not yet have been flushed to disk, even for local
2454 Frequency : frequent on thin XT3 nodes
2456 Description: UUID collision on thin XT3 Linux nodes
2457 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2458 randomly, since we relied on an insufficiently-seeded PRNG.
2463 Description: OSS hangs after "All ost request buffers busy"
2464 Details : A deadlock between quota and journal operations caused OSS
2465 hangs after printing "All ost request buffers busy."
2468 Frequency : always on liblustre builds
2470 Description: Cleanup compiler warnings on liblustre
2473 Frequency : always on liblustre builds on XT3
2475 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2479 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2480 > 32000 subdirectories
2481 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2482 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2483 subdirectory count crosses 32000. This will aid e2fsck to
2484 correctly handle more than 32000 subdirectories.
2487 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2489 Description: System hang on startup
2490 Details : This bug allowed the liblustre (e.g. catamount) client to
2491 return to the app before handling all startup RPCs. This
2492 could leave the node unresponsive to lustre network traffic
2493 and manifested as a server ptllnd timeout.
2495 Severity : enhancement
2497 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2498 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2499 boolean to control whether to print peer debug info when a
2500 client's RPC times out.
2505 Description: make O_SYNC working on 2.6 kernels
2506 Details : 2.6 kernels use different method for mark pages for write,
2507 so need add a code to lustre for O_SYNC work.
2512 Description: Failure to close file and release space on NFS
2513 Details : Put inode details into lock acquired in ll_intent_file_open.
2514 Use mdc_intent_lock in ll_intent_open to properly
2515 detect all kind of errors unhandled by mdc_enqueue.
2520 Description: proc file read during shutdown sometimes raced obd removal,
2522 Details : Add lock to prevent obd access after proc file removal.
2525 Frequency : Only for files larger than 4GB on 32-bit clients.
2527 Description: improperly doing page alignment of locks
2528 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2529 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2534 Description: RPCs being resent when they shouldn't be
2535 Details : Some RPCs that should not be resent are being resent. This
2536 can cause inconsistencies in the RPC state machine. Do not
2537 resend such requests.
2540 Frequency : rare, only with NFS export
2542 Description: Crash on NFS re-export node
2543 Details : under very unusual load conditions an assertion is hit in
2544 ll_intent_file_open()
2547 Frequency : only if OST filesystem is corrupted
2549 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2550 Details : for a short time RPCs with bulk IO are in the replay list,
2551 but replay of bulk IOs is unimplemented. If the OST filesystem
2552 is corrupted due to disk cache incoherency and then replay is
2553 started it is possible to trip an assertion. Avoid putting
2554 committed RPCs into the replay list at all to avoid this issue.
2559 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2560 Details : Large single O_DIRECT read and write calls can fail to allocate
2561 a sufficiently large buffer to process the request. In case of
2562 allocation failure the allocation is retried with a smaller
2563 buffer and broken into smaller requests.
2565 Severity : enhancement
2567 Description: Add -o localflock option to simulate old noflock behaviour.
2568 Details : This will achieve local-only flock/fcntl locks coherentness.
2573 Description: versioning check is incomplete
2574 Details : Checking the version difference of client vs. server, report
2575 error if the gap is too big.
2580 Description: add support PG_writeback bit
2581 Details : add support for PG_writeback bit for Lustre, for more carefull
2582 work with page cache in 2.6 kernel. This also fix some deadlocks
2583 and remove hack for work O_SYNC with 2.6 kernel.
2585 Severity : enhancement
2587 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2588 Details : The uninit_groups feature works in conjunction with the kernel
2589 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2590 up the pass1 processing of e2fsck. This is a read-only feature
2591 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2592 mount filesystems that have had this feature enabled.
2594 Severity : enhancement
2596 Description: Improve multi-block allocation algorithm to avoid fragmentation
2597 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2598 allocation locality and avoid filesystem fragmentation.
2600 ------------------------------------------------------------------------------
2602 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2604 * Support for kernels:
2605 2.6.9-42.0.3.EL (RHEL 4)
2606 2.6.5-7.276 (SLES 9)
2607 2.4.21-47.0.1.EL (RHEL 3)
2608 2.6.12.6 vanilla (kernel.org)
2609 2.6.16.21-0.8 (SLES10)
2610 * Recommended e2fsprogs version: 1.39.cfs2-0
2612 * The backwards-compatible /proc/sys/portals symlink has been removed
2613 in this release. Before upgrading, please ensure that you change
2614 any configuration scripts or /etc/sysctl.conf files that access
2615 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2616 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2617 in advance of the upgrade on any system running Lustre 1.4.6 or
2618 newer, since /proc/sys/lnet was added in that version.
2619 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2623 Frequency : only when quota is used
2625 Description: avoid scanning export list for quota master
2626 Details : Change the algorithms to avoid scanning export list in order
2627 to improve the efficiency.
2630 Frequency : MDS failover only, very rarely
2632 Description: "went back in time" messages on mds failover
2633 Details : The greatest transno may be lost when the current operation
2634 finishes with an error (transno==0) and the client's last_rcvd
2635 record is over-written. Save the greatest transno in the
2636 mds_last_transno for this case.
2639 Frequency : always for specific kernels and striping counts
2641 Description: client may get "Matching packet too big" without ACL support
2642 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2643 when trying to access files in certain configurations. The
2644 clients should in fact be denied when mounting because they do
2645 not understand ACLs.
2648 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2650 Description: many clients connecting with IO in progress causes connect timeouts
2651 Details : Avoid synchronous journal commits to avoid delays caused by many
2652 clients connecting/disconnecting when bulk IO is in progress.
2653 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2654 OST_IO_PORTAL to avoid delays behind potentially many pending
2658 Frequency : occasionally with multiple writers to a single file
2660 Description: shared writes to file may result in wrong size reported by stat()
2661 Details : Allow growing of kms when extent lock is cancelled
2664 Frequency : always with random mmap IO to multi-striped file
2666 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2667 Details : Only if the hole is at the end of OST object so that kms is too
2668 small. Fix is to increase kms accordingly in ll_nopage.
2671 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2673 Description: writes to a missing object would leak memory on the OST
2674 Details : If there is an inconsistency between the MDS and OST filesystems,
2675 such that the MDS references an object that doesn't exist, writes
2676 to that object will leak memory due to incorrect cleanup in the
2677 error handling path, eventually running out of memory on the OST.
2682 Description: Creating too long symlink causes lustre errors
2683 Details : Check symlink and name lengths before sending requests to MDS.
2686 Frequency : only if flock is enabled (not on by default)
2688 Description: posix locks not released on fd closure on 2.6.9+
2689 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2690 kernels, this caused such locks not to be released on fd close and
2691 then assertions on fs unmount about still used locks.
2694 Frequency : MDS failover only, very rarely
2696 Description: clients may get ASSERTION(granted_lock != NULL)
2697 Details : When request was taking a long time, and a client was resending
2698 a getattr by name lock request. The were multiple lock requests
2699 with the same client lock handle and
2700 mds_getattr_name->fixup_handle_for_resent_request found one of the
2701 lock handles but later failed with ASSERTION(granted_lock != NULL).
2706 Description: handle->h_buffer_credits > 0, assertion failure
2707 Details : h_buffer_credits is zero after truncate, causing assertion
2708 failure. This patch extends the transaction or creates a new
2712 Frequency : NFS re-export or patchless client
2713 Bugzilla : 11179, 10796
2714 Description: Crash on NFS re-export node (__d_move)
2715 Details : We do not want to hash the dentry if we don't have a lock.
2716 But if this dentry is later used in d_move, we'd hit uninitialised
2717 list head d_hash, so we just do this to init d_hash field but
2718 leave dentry unhashed.
2721 Frequency : NFS re-export or patchless client
2723 Description: NFS exports has problem with symbolic link
2724 Details : lustre client didn't properly install dentry when re-exported
2725 to NFS or running patchless client.
2728 Frequency : NFS re-export or patchless client
2730 Description: Various nfs/patchless fixes.
2731 Details : fixes reuse disconected alias for lookup process - this fixes
2732 warning "find_exported_dentry: npd != pd",
2733 fix permission error with open files at nfs.
2734 fix apply umask when do revalidate.
2737 Frequency : occasional
2739 Description: Crash on NFS re-export node
2740 Details : calling clear_page() on the wrong pointer triggered oops in
2741 generic_mapping_read().
2744 Frequency : rarely, using O_DIRECT IO
2746 Description: unaligned directio crashes client with LASSERT
2747 Details : check for unaligned buffers before trying any requests.
2750 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2752 Description: stale data returned from RAID cache
2753 Details : If only a small amount of IO is done to the RAID device before
2754 reading it again it is possible to get stale data from the RAID
2755 cache instead of reading it from disk.
2758 Frequency : always for sles10 kernel
2760 Description: sles10 support
2761 Details : ll_follow_link: compile fixes and using of nd_set_link
2762 under newer kernels.
2765 Frequency : depends on arch, kernel and compiler version, always on sles10
2768 Description: recursive or deep enough symlinks cause stack overflow
2769 Details : getting rid of large stack-allocated variable in
2773 Frequency : depends on hardware
2775 Description: lustre write performance loss in the SLES10 kernel
2776 Details : the performance loss is caused by using of write barriers in the
2777 ext3 code. The SLES10 kernel turns barrier support on by
2778 default. The fix is to undo that change for ldiskfs.
2780 ------------------------------------------------------------------------------
2782 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2784 * Support for kernels:
2785 2.6.9-42.0.3EL (RHEL 4)
2786 2.6.5-7.276 (SLES 9)
2787 2.4.21-47.0.1.EL (RHEL 3)
2788 2.6.12.6 vanilla (kernel.org)
2792 Frequency : quota enabled and large files being deleted
2794 Description: releasing more than 4GB of quota at once hangs OST
2795 Details : If a user deletes more than 4GB of files on a single OST it
2796 will cause the OST to spin in an infinite loop. Release
2797 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2802 Description: statfs data retrieved from /proc may be stale or zero
2803 Details : When reading per-device statfs data from /proc, in the
2804 {kbytes,files}_{total,free,avail} files, it may appear
2805 as zero or be out of date.
2808 Frequency : systems with MD RAID1 external journal devices
2810 Description: lconf's call to blkid is confused by RAID1 journal devices
2811 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2812 of returning all block devices that match the journal UUID.
2815 Frequency : always, for aggregate stripe size over 4GB
2817 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2818 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2819 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2822 Frequency : always if "lfs find" used on a local file/directory
2824 Description: "lfs find" segfaults if used on a local file/directory
2825 Details : The case where a directory component was not specified wasn't
2826 handled correctly. Handle this properly.
2829 Frequency : always on ppc64
2831 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2832 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2833 perform endianness conversion.
2836 Frequency : rarely (truncate to non-zero file size after write under load)
2837 Bugzilla : 10730, 10687
2838 Description: Files padded with zeros to next 4K multiple
2839 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2840 it is possible that files that are truncated to a non-zero size
2841 immediately after being written are filled with zero bytes beyond
2842 the truncated size. No file data is lost.
2844 Severity : enhancement
2846 Description: Allow recovery/failover for liblustre clients.
2847 Details : liblustre clients were unaware of failover configurations until
2850 Severity : enhancement
2852 Description: user file locks should fail when not mounting with flock option
2853 Details : Set up an error-returning stub in ll_file_operations.lock field
2854 to prevent incorrect behaviour when client is mounted without
2855 flock option. Also, set up properly f_op->flock field for
2859 Frequency : always on ia64
2861 Description: "lfs df" loops on printing out MDS statfs information
2862 Details : The obd_ioctl_data was not initialized and in some systems
2863 this caused a failure during the ioctl that did not return
2864 an error. Initialize the struct and return an error on failure.
2867 Frequency : SLES 9 only
2869 Description: Error of copying files with lustre special EAs as root
2870 Details : Client side always return success for setxattr call for lustre
2871 special xattr (currently only "trusted.lov").
2874 Frequency : rarely on clusters with both ia64+i386 clients
2876 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2877 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2878 clients are doing concurrent writes to the same file it is
2879 possible that the ia64 clients may LASSERT because the OST
2880 extent locks are not PAGE_SIZE aligned. Ensure that grown
2881 locks are always aligned on the request boundary.
2884 Frequency : specific use, occasional
2886 Description: Overwriting in use executable truncates on-disk binary image
2887 Details : If one node attempts to overwrite an executable in use by
2888 another node, we now correctly return ETXTBSY instead of
2889 truncating the file.
2891 Severity : enhancement
2893 Description: Async OSC create to avoid the blocking unnecessarily.
2894 Details : If a OST has no remain object, system will block on the creating
2895 when need to create a new object on this OST. Now, ways use
2896 pre-created objects when available, instead of blocking on an
2897 empty osc while others are not empty. If we must block, we block
2898 for the shortest possible period of time.
2903 Description: chmod on Lustre root is propagated to other clients
2904 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2905 invalid by the follow_mount time.
2908 Frequency : liblustre clients only
2910 Description: Race in 'instant cancel' lock handling could lead to such locks
2911 never to be granted in case of SMP MDS
2912 Details : Do not destroy not yet granted but cbpending locks in
2916 Frequency : replay/resend of open
2918 Description: non null lock assetion failure in mds_intent_policy
2919 Details : Trying to replay/resend lockless open requests resulted in
2920 mds_open() returning 0 with no lock. Now it sets a flag if
2921 a lock is going to be returned.
2923 Severity : enhancement
2925 Description: Checksum enhancements
2926 Details : New checksum enhancements allow for resending RPCs that failed
2929 Severity : enhancement
2931 Description: Tunables on number of dirty pages in cacche
2932 Details : Allow to set limit on number of dirty pages cached.
2937 Description: client crash on unmount - lock still has references
2938 Details : In some error handling cases it was possible to leak a lock
2939 reference on a client while accessing a file. This was not
2940 harmful to the client during operation, but would cause the
2941 client to crash when the filesystem is unmounted.
2944 Frequency : specific case, rare
2946 Description: ETXTBSY on mds though file not in use by client
2947 Details : ETXTBSY is no longer incorrectly returned when attempting to
2948 chmod or chown a directory that the user previously tried to
2949 execute or a currently-executing binary.
2952 Frequency : extremely rare except on liblustre-based clients
2954 Description: Lustre space not freed when files are deleted
2955 Details : Clean up open-unlinked files after client eviction. Previously
2956 the unlink was skipped and the files remained as orphans.
2961 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2962 Details : In some cases it was possible to send a blocking callback to
2963 a client doing a glimpse, even though that client didn't get
2964 a lock granted. When the glimpse lock is cancelled on the OST
2965 the freed lock is left on the waiting list and corrupted the list.
2968 Frequency : all core dumps
2970 Description: Broke core dumps to lustre
2971 Details : Negative dentry may be unhashed if parent does not have UPDATE
2972 lock, but some callers, e.g. do_coredump, expect dentry to be
2973 hashed after successful create, hash it in ll_create_it.
2975 ------------------------------------------------------------------------------
2977 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2979 * Support for kernels:
2980 2.6.9-42.0.2.EL (RHEL 4)
2981 2.6.5-7.276 (SLES 9)
2982 2.4.21-40.EL (RHEL 3)
2983 2.6.12.6 vanilla (kernel.org)
2987 Frequency : always on RHEL 3
2989 Description: Number of open files grows over time
2990 Details : The number of open files grows over time, whether or not
2991 Lustre is started. This was due to a filp leak introduced
2992 by one of our kernel patches.
2994 ------------------------------------------------------------------------------
2996 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2998 * Support for kernels:
2999 2.6.9-42.EL (RHEL 4)
3000 2.6.5-7.276 (SLES 9)
3001 2.4.21-40.EL (RHEL 3)
3002 2.6.12.6 vanilla (kernel.org)
3007 Bugzilla : 5719, 9635, 9792, 9684
3008 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3009 Details : If a server is under heavy load and cannot reply to new
3010 connection requests before the client resends the (re)connect,
3011 the connection handling code can behave badly if two service
3012 threads are concurrently handing separate (re)connections from
3013 the same client. Add better locking to the connection handling
3014 code, and ensure that only a single connection will be processed
3015 for a given client UUID, even if the lock is dropped.
3017 Severity : enhancement
3019 Description: add TCP zero-copy support to kernel
3020 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3021 sends if the hardware supports scatter-gather and checksumming.
3022 This allows socklnd to do client-write and server-read more
3023 efficiently and reduce CPU utilization from skbuf copying.
3026 Frequency : only if NFS exporting from client
3028 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3029 Details : If setting attributes on a file created under NFS that had
3030 never been opened it would be possible to oops the client
3031 if the file had no objects.
3034 Frequency : always for liblustre
3036 Description: liblustre client does MDS+OSTs setattr RPC for each write
3037 Details : When doing a write from a liblustre client, the client
3038 incorrectly issued an RPC to the MDS and each OST the file was
3039 striped over in order to update the timestamps. When writing
3040 with small chunks and many clients this could overwhelm the MDS
3041 with RPCs. In all cases it would slow down the write because
3042 these RPCs are unnecessary.
3044 Severity : enhancement
3046 Description: allow number of MDS service threads to be changed at module load
3047 Details : It is now possible to change the number of MDS service threads
3048 running. Adding "options mds mds_num_threads={N}" to the MDS's
3049 /etc/modprobe.conf will set the number of threads for the next
3050 time Lustre is restarted (assuming the "mds" module is also
3051 reloaded at that time). The default number of threads will
3052 stay the same, 32 for most systems.
3057 Description: OST crash if filesystem is unformatted or corrupt
3058 Details : If an OST is started on a device that has never been formatted
3059 or if the filesystem is corrupt and cannot even mount then the
3060 error handling cleanup routines would dereference a NULL pointer.
3065 Description: NULL pointer deref in llap_from_page.
3066 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3067 incorrect) mapping if the page was truncated/reclaimed while it was
3068 searched for. Check for this condition and skip such pages when
3069 doing readahead. Introduce extra check to llap_from_page() to
3070 verify page->mapping->host is non-NULL (so page is not anonymous).
3073 Frequency : Sometimes when using sys_sendfile
3075 Description: "page not covered by a lock" warnings from ll_readpage
3076 Details : sendfile called ll_readpage without right page locks present.
3077 Now we introduced ll_file_sendfile that does necessary locking
3078 around call to generic_file_sendfile() much like we do in
3082 Frequency : with certain MDS communication failures at client mount time
3084 Description: NULL pointer deref after failed client mount
3085 Details : a client connection request may delayed by the network layer
3086 and not be sent until after the PTLRPC layer has timed out the
3087 request. If the client fails the mount immediately it will try
3088 to clean up before the network times out the request. Add a
3089 reference from the request import to the obd device and delay
3090 the cleanup until the network drops the request.
3093 Frequency : occasionally during client (re)connect
3095 Description: assertion failure during client (re)connect
3096 Details : processing a client connection request may be delayed by the
3097 client or server longer than the client connect timeout. This
3098 causes the client to resend the connection request. If the
3099 original connection request is replied in this interval, the
3100 client may trip an assertion failure in ptlrpc_connect_interpret()
3101 which thought it would be the only running connect process.
3104 Frequency : only with obd_echo servers and clients that are rebooted
3106 Description: kernel BUG accessing uninitialized data structure
3107 Details : When running an obd_echo server it did not start the ping_evictor
3108 thread, and when a client was evicted an uninitialized data
3109 structure was accessed. Start the ping_evictor in the RPC
3110 service startup instead of the OBD startup.
3112 Severity : enhancement
3113 Bugzilla : 10193 (patchless)
3114 Description: Remove dependency on various unexported kernel interfaces.
3115 Details : No longer need reparent_to_init, exit_mm, exit_files,
3116 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3119 Frequency : rare (only users of deprecated and unsupported LDAP config)
3121 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3122 Details : LDAP apparently contains 'lustreName' attributes instead of
3123 'name'. A simple remapping of the name is sufficient.
3126 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3128 Description: waiting_locks_callback trips kernel BUG if client is evicted
3129 Details : Running with the dump_on_timeout debug flag turned on makes
3130 it possible that the waiting_locks_callback() can try to dump
3131 the Lustre kernel debug logs from an interrupt handler. Defer
3132 this log dumping to the expired_lock_main() thread.
3134 Severity : enhancement
3136 Description: Support NFS exporting on 2.6 kernels.
3137 Details : Implement non-rawops metadata methods for NFS server to use without
3138 changing NFS server code.
3141 Frequency : very rare (synthetic metadata workload only)
3143 Description: two racing renames might cause an MDS thread to deadlock
3144 Details : Running the "racer" program may cause one MDS thread to rename
3145 a file from being the source of a rename to being the target of
3146 a rename at exactly the same time that another thread is doing
3147 so, and the second thread has already enqueued these locks after
3148 doing a lookup of the target and is trying to relock them in
3149 order. Ensure that we don't try to re-lock the same resource.
3152 Frequency : only very large systems with liblustre clients
3154 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3155 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3156 all OSTs in parallel. This allows the eviction of stale liblustre
3157 clients to proceed much faster than if they were done in series,
3158 and also offers similar improvements for other set_info RPCs.
3163 Description: excessive CPU usage during initial read phase on client
3164 Details : During the initial read phase on a client, it would agressively
3165 retry readahead on the file, consuming too much CPU and impacting
3166 performance (since 1.4.5.8). Improve the readahead algorithm
3167 to avoid this, and also improve some other common cases (read
3168 of small files in particular, where "small" is files smaller than
3169 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3174 Description: MDS crash when receiving packet with unknown intent.
3175 Details : Do not LBUG in unknown intent case, just return -EFAULT
3177 Severity : enhancement
3178 Bugzilla : 9293, 9385
3179 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3180 Details : Do not serialize getattr (non-intent version) and statfs.
3183 Frequency : occasional, when OST network is overloaded/intermittent
3185 Description: client evicted by OST after bulk IO timeout
3186 Details : If a client sends a bulk IO request (read or write) the OST
3187 may evict the client if it is unresposive to its data GET/PUT
3188 request. This is incorrect if the network is overloaded (takes
3189 too long to transfer the RPC data) or dropped the OST GET/PUT
3190 request. There is no need to evict the client at all, since
3191 the pinger and/or lock callbacks will handle this, and the
3192 client can restart the bulk request.
3195 Frequency : Always when mmapping file with no objects
3197 Description: client crashes when mmapping file with no objects
3198 Details : Check that we actually have objects in a file before doing any
3199 operations on objects in ll_vm_open, ll_vm_close and
3205 Description: Request leak when working with deleted CWD
3206 Details : Introduce advanced request refcount tracking for requests
3207 referenced from lustre intent.
3209 Severity : Enhancement
3211 Description: Cache open file handles on client.
3212 Details : MDS now will return special lock along with openhandle, if
3213 requested and client is allowed to hold openhandle, even if unused,
3214 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3215 closing files for every read/write openration.
3217 Severity : Enhancement
3219 Description: Cache open negative dentries on client when possible.
3220 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3221 negative dentries on lock revocation.
3226 Description: Remounting a client read-only wasn't possible with a zconf mount
3227 Details : It wasn't possible to remount a client read-only with llmount.
3229 Severity : enhancement
3230 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3231 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3232 MPICH is included to add Lustre-specific ADIO interfaces.
3233 This is based closely on the UFS ADIO layer and only differs
3234 in file creation, in order to allow the OST striping to be set.
3235 This is user-contributed code and not supported by CFS.
3240 Description: extended inode attributes (immutable, append-only) work improperly
3241 when 2.4 and 2.6 kernels are used on client/server or vice versa
3242 Details : Introduce kernel-independent values for these flags.
3244 Severity : enhancement
3247 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3248 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3249 granularity. Now, Fractional MB granularity is be supported,
3250 this is very useful for low memory system.
3252 Severity : enhancement
3254 Description: Getattr by fid
3255 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3256 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3257 have locks and avoids getattr rpc requests for it on every stat.
3260 Frequency : Always, for filesystems larger than 2TB
3262 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3263 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3264 filesystems larger than 2TB to be mounted with mballoc enabled.
3267 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3269 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3270 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3271 filesystems larger than 2TB is mounted then the kernel will
3272 silently corrupt the start of the filesystem. CONFIG_LBD is
3273 enabled for all CFS-supported kernels, but the possibility of
3274 this happening with a modified kernel config exists.
3276 Severity : enhancement
3278 Description: add client O_DIRECT support for 2.6 kernels
3279 Details : It is now possible to do O_DIRECT reads and writes to files
3280 in the Lustre client mountpoint on 2.6 kernel clients.
3282 Severity : enhancement
3284 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3285 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3286 parallel, not waiting for response from every OST before sending
3287 a rpc to the next OST.
3292 Description: setattr vs write race when updating file timestamps
3293 Details : Client processes that update a file timestamp into the past
3294 right after writing to the file (e.g. tar) it is possible that
3295 the updated file modification time can be reset to the current
3296 time due to a race between processing the setattr and write RPC.
3298 Severity : enhancement
3300 Description: Bring 'lfs find' closer in line with regular Linux find.
3301 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3302 -print0 options and obtains all the needed info through the lustre
3305 Severity : enhancement
3307 Description: support up to 1024 configured devices on one node
3308 Details : change obd_dev array from statically allocated to dynamically
3309 allocated structs as they are first used to reduce memory usage
3314 Description: Flush dirty partially truncated pages during truncate
3315 Details : Immediatelly flush partially truncated pages in filter_setattr,
3316 this way we completely avoid having any pages in page cache on OST
3317 and can retire ugly workarounds during writes to flush such pages.
3322 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3323 Details : protect inode from truncation within vfs_unlink() context
3324 just take a reference before calling vfs_unlink() and release it
3325 when parent's i_sem is free.
3330 Description: last_id value checked outside lock on OST caused LASSERT failure
3331 Details : If there were multiple MDS->OST object precreate requests in
3332 flight, it was possible that the OST's last object id was checked
3333 outside a lock and incorrectly tripped an assertion. Move checks
3334 inside locks, and discard old precreate requests.
3337 Frequency : always, if extents are used on OSTs
3339 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3340 Details : OSTs using the extents format would not zero the high 16 bits of
3341 the index physical block number. This is not a problem for any
3342 OST filesystems smaller than 16TB, and no kernels support ext3
3343 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3344 new/modified files) and can be fixed for existing filesystems
3345 with e2fsprogs-1.39-cfs1.
3350 Description: import connection selection may be incorrect if timer wraps
3351 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3352 import connections to be ignored if the 32-bit jiffies counter
3353 wraps. Use a 64-bit jiffies counter.
3356 Frequency : very large clusters immediately after boot
3358 Description: LNET request buffers exhausted under heavy short-term load
3359 Details : If a large number of client requests are generated on a service
3360 that has previously never seen so many requests it is possible
3361 that the request buffer growth cannot keep up with the spike in
3362 demand. Instead of dropping incoming requests, they are held in
3363 the LND until the RPC service can accept more requests.
3366 Frequency : Sometimes during replay
3368 Description: Assertion failure in ll_local_open after replay.
3369 Details : If replay happened on an open request reply before we were able
3370 to set replay handler, reply will become not swabbed tripping the
3371 assertion in ll_local_open. Now we set the handler right after
3372 recognising of open request
3375 Frequency : very rare
3377 Description: kernel reports "badness in vsnprintf"
3378 Details : Reading from the "recovery_status" /proc file in small chunks
3379 may cause a negative length in lprocfs_obd_rd_recovery_status()
3380 call to vsnprintf() (which is otherwise harmless). Exit early
3381 if there is no more space in the output buffer.
3383 Severity : enhancement
3385 Description: clear OBD RPC statistics by writing to them
3386 Details : It is now possible to clear the OBD RPC statistics by writing
3387 to the "stats" file.
3392 Description: Client mtime is not the same on different clients after utimes
3393 Details : In some cases, the client was using the utimes() syscall on
3394 a file cached on another node. The clients now validate the
3395 ctime from the MDS + OSTs to determine which one is right.
3400 Description: Inability to activate failout mode
3401 Details : lconf script incorrectly assumed that in pythong string's numeric
3402 value is used in comparisons.
3405 Frequency : always with multiple stripes per file
3407 Description: Inefficient object allocation for mutli-stripe files
3408 Details : When selecting which OSTs to stripe files over, for files with
3409 a stripe count that divides evenly into the number of OSTs,
3410 the MDS is always picking the same starting OST for each file.
3411 Return the OST selection heuristic to the original design.
3416 Description: mount failures may take full timeout to return an error
3417 Details : Under some heavy load conditions it is possible that a
3418 failed mount can wait for the full obd_timeout interval,
3419 possibly several minutes, before reporting an error.
3420 Instead return an error as soon as the status is known.
3422 Frequency : quota enabled and large files being deleted
3424 Description: releasing more than 4GB of quota at once hangs OST
3425 Details : If a user deletes more than 4GB of files on a single OST it
3426 will cause the OST to spin in an infinite loop. Release
3427 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3432 Description: statfs data retrieved from /proc may be stale or zero
3433 Details : When reading per-device statfs data from /proc, in the
3434 {kbytes,files}_{total,free,avail} files, it may appear
3435 as zero or be out of date.
3438 Frequency : systems with MD RAID1 external journal devices
3440 Description: lconf's call to blkid is confused by RAID1 journal devices
3441 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3442 of returning all block devices that match the journal UUID.
3445 Frequency : always, for aggregate stripe size over 4GB
3447 Description: assertion fails when trying to use 4GB stripe size
3448 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3449 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3452 Frequency : always on ppc64
3454 Description: the first write on an ext3 filesystem with mballoc got stuck
3455 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3456 perform endianness conversion.
3458 ------------------------------------------------------------------------------
3460 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3462 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3463 INTEROPERATE with older versions automatically. Please read the
3464 user documentation before upgrading any part of a live system.
3465 * WARNING: Lustre networking configuration changes are required with
3466 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3469 * Support for kernels:
3470 2.6.9-22.0.2.EL (RHEL 4)
3471 2.6.5-7.244 (SLES 9)
3472 2.6.12.6 vanilla (kernel.org)
3475 Severity : enhancement
3476 Bugzilla : 7981/8208
3477 Description: Introduced Lustre Networking (LNET)
3478 Details : LNET is new networking infrastructure for Lustre, it includes
3479 a reorganized network configuration mode (see the user
3480 documentation for full details) as well as support for routing
3481 between different network fabrics. Lustre Networking Devices
3482 (LNDs) for the supported network fabrics have also been
3483 created for this new infrastructure.
3485 Severity : enhancement
3486 Description: Introduced Access control lists
3487 Details : clients can set ACLs on files and directories in order to have
3488 more fine-grained permissions than the standard Unix UGO+RWX.
3489 The MDS must be started with the "-o acl" mount option.
3491 Severity : enhancement
3492 Description: Introduced filesystem quotas
3493 Details : Administrators may now establish per-user quotas on the
3496 Severity : enhancement
3498 Description: Configuration change for the XT3
3499 The PTLLND is now used to run Lustre over Portals on the XT3
3500 The configure option(s) --with-cray-portals are no longer used.
3501 Rather --with-portals=<path-to-portals-includes> is used to
3502 enable building on the XT3. In addition to enable XT3 specific
3503 features the option --enable-cray-xt3 must be used.
3508 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3509 Details : A race between allocating a new llcd and re-getting the llcd_lock
3510 allowed another thread to grab newly-allocated llcd.
3512 Severity : enhancement
3514 Description: 2.6 OST async journal commit and locking fix to improve performance
3515 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3516 2.6 kernels are now async as they already were in 2.4 kernels so
3517 that they can commit concurrently with the network bulk transfer.
3518 For block-allocated files the filter allocation semaphore is held
3519 to avoid filesystem fragmentation during allocation. BKL lock
3520 removed for 2.6 xattr operations where it is no longer needed.
3525 Description: lconf incorrectly determined whether two IP networks could talk
3526 Details : In some more complicated routing and multiple-network
3527 configurations, lconf will avoid trying to make a network
3528 connection to a disjoint part of the IP space. It was doing the
3529 math incorrectly for one set of cases.
3534 Description: Fix for potential infinite loop processing records in an llog.
3535 Details : If an llog record is corrupted/zeroed, it is possible to loop
3536 forever in llog_process(). Validate the llog record length
3537 and skip the remainder of the block on error.
3540 Frequency : occasional (liblustre only)
3542 Description: liblustre could not open files whose last component is a symlink
3543 Details : sysio_path_walk() would incorrectly pass the open intent to
3544 intermediate path components.
3547 Frequency : rare (liblustre only with non-standard tuning)
3548 Bugzilla : 7201 (7350)
3549 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3550 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3551 after it has been released, only do this for referenced locks
3553 Severity : enhancement
3555 Description: specifying an (invalid) directory default stripe_size of -1
3556 would reset the directory default striping
3557 Details : stripe_size -1 was used internally to signal directory stripe
3558 removal, now use "all default" to signal dir stripe removal
3559 as a directory striping of "all default" is not useful
3562 Frequency : common for large clusters running liblustre clients
3564 Description: doing an ls when liblustre clients are running is slow
3565 Details : sending a glimpse AST to a liblustre client waits for every AST
3566 to time out, as liblustre clients will not respond. Since they
3567 cannot cache data we refresh the OST lock LVB from disk instead.
3569 Severity : enhancement
3571 Description: doing an ls at the same time as file IO can be slow
3572 Details : enqueue and other "small" requests can be blocked behind many
3573 large IO requests. Create a new OST IO portal for non-IO
3574 requests so they can be processed faster.
3577 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3579 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3580 created file with mode 000
3581 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3582 allow an ftruncate() as a non-root user to a file with mode 000.
3583 Lustre now respects this flag to disable mode checks when
3584 truncating a file owned by the user
3587 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3590 Description: Revoking locks from clients that went dead or catatonic might take
3592 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3593 cancellation of such locks instant on servers without waiting for
3594 any reply from clients. Clients drops these locks when cancel
3595 notification from server is received without replying.
3598 Frequency : liblustre-only, when liblustre client dies or becomes busy
3600 Description: Doing ls on Linux clients can take a long time with active
3602 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3603 granting such locks to it in the first place if possible. Locks
3604 are taken by proxy on the OST during the read or write and
3605 dropped immediately afterward. Add connect flags handling, do
3606 not grant locks to liblustre clients for glimpse ASTs.
3608 Severity : enhancement
3610 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3611 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3612 sequential reads and behaves badly with applications doing random
3613 reads. Improve it by reading ahead at least read region, and
3614 avoiding excessive large RPC for small reads.
3616 Severity : enhancement
3618 Description: Creating more than 1000 files for a single job may cause a load
3619 imbalance on the OSTs if there are also a large number of OSTs.
3620 Details : qos_prep_create() uses an OST index reseed value that is an
3621 even multiple of the number of available OSTs so that if the
3622 reseed happens in the middle of the object allocation it will
3623 still utilize the OSTs as uniformly as possible.
3628 Description: OST or MDS may oops in ping_evictor_main()
3629 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3630 but doesn't restart at beginning of obd_exports_timed list
3633 Severity : enhancement
3635 Description: improve by-nid export eviction on the MDS and OST
3636 Details : allow multiple exports with the same NID to be evicted at one
3637 time without re-searching the exports list.
3640 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3642 Description: MDS may oops in groups_free()
3643 Details : in rare race conditions a newly allocated group_info struct is
3644 freed again, and this can be NULL. The 2.4 compatibility code
3645 for groups_free() checked for a NULL pointer, but 2.6 did not.
3648 Frequency : common for liblustre clients doing little filesystem IO
3649 Bugzilla : 9352, 7313
3650 Description: server may evict liblustre clients accessing contended locks
3651 Details : if a client is granted a lock or receives a completion AST
3652 with a blocking AST already set it would not reply to the AST
3653 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3656 Frequency : lfs setstripe, only systems with more than 160 OSTs
3658 Description: unable to set striping with a starting offset beyond OST 160
3659 Details : llapi_create_file() incorrectly limited the starting stripe
3660 index to the maximum single-file stripe count.
3663 Frequency : LDAP users only
3665 Description: lconf did not handle in-kernel recovery with LDAP properly
3666 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3668 Severity : enhancement
3670 Description: bind OST threads to NUMA nodes to improve performance
3671 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3672 node and do their allocations there to localize memory access
3674 Severity : enhancement
3676 Description: llmount can determine client NID directly from Myrinet (GM)
3677 Details : the client NID code from gmnalnid was moved directly into
3678 llmount, removing the need to use this or specifying the
3679 client NID explicitly when mounting GM clients with zeroconf
3682 Frequency : if client is started with down MDS
3684 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3685 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3686 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3687 ptlrpc_import_delayed_req() because we didn't check if the
3688 request was interrupted, and we also didn't break out of the
3689 event loop if there was no timeout
3694 Description: data loss during non-page-aligned writes to a single file from
3695 both multiple nodes and multiple threads on one node at same time
3696 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3697 inconsistency led to false short-reads, that were cached and later
3698 used by ->prepare_write() to fill in partially written page,
3699 leading to data loss.
3702 Frequency : always, if lconf --abort_recovery used
3704 Description: lconf --abort_recovery fails with 'Operation not supported'
3705 Details : lconf was attempting to abort recovery on the MDT device and not
3708 Severity : enhancement
3710 Description: remove cleanup logs
3711 Details : replace lconf-generated cleanup logs with lustre internal
3712 cleanup routines. Eliminates the need for client-cleanup and
3715 Severity : enhancement
3717 Description: add support for EAs (user and system) on lustre filesystems
3718 Details : it is now possible to store extended attributes in the Lustre
3719 client filesystem, and with the user_xattr mount option it
3720 is possible to allow users to store EAs on their files also
3722 Severity : enhancement
3724 Description: Add possibility (config option) to show minimal available OST free
3726 Details : When compiled with --enable-mindf configure option, statfs(2)
3727 (and so, df) will return least minimal free space available from
3728 all OSTs as amount of free space on FS, instead of summary of
3729 free spaces of all OSTs.
3731 Severity : enhancement
3733 Description: do not expand extent locks acquired on OST-side
3734 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3735 by server: they are not cached anyway.
3738 Frequency : when mmap is used/binaries executed from Lustre
3740 Description: Unmmap pages before throwing them away from read cache.
3741 Details : llap_shrink cache now attempts to unmap pages before discarding
3742 them (if unmapping failed - do not discard). SLES9 kernel has
3743 extra checks that trigger if this unmapping is not done first.
3748 Description: lconf didn't resolve symlinks before checking to see whether a
3749 given mountpoint was already in use
3752 Frequency : when migrating failover services
3753 Bugzilla : 6395, 9514
3754 Description: When migrating a subset of services from a node (e.g. failback
3755 from a failover service node) the remaining services would
3756 time out and evict clients.
3757 Details : lconf --force (implied by --failover) sets the global obd_timeout
3758 to 5 seconds in order to quickly disconnect, but this caused
3759 other RPCs to time out too quickly. Do not change the global
3760 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3762 Severity : enhancement
3763 Frequency : if MDS is started with down OST
3764 Bugzilla : 9439,5706
3765 Description: Allow startup/shutdown of an MDS without depending on the
3766 availability of the OSTs.
3767 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3768 Add appropriate locking and lov-osc refcounts for safe
3769 cleaning. Add osc abort_inflight calls in case the
3770 synchronize never started.
3773 Frequency : occasional (Cray XT3 only)
3775 Description: root not authorized to access files in CRAY_PORTALS environment
3776 Details : The client process capabilities were not honoured on the MDS in
3777 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3778 been accessed by an authorized user then root was able to access
3779 the file on the local client also. The root user capabilities
3780 are now allowed on the MDS, as this environment has secure UID.
3783 Frequency : occasional
3785 Description: ldiskfs "too long searching" message happens too often
3786 Details : A debugging message (otherwise harmless) prints too often on
3787 the OST console. This has been reduced to only happen when
3788 there are fragmentation problems on the filesystem.
3793 Description: Division by zero in statfs when all OSCs are inactive
3794 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3795 lov_statfs divides by value returned by lov_get_stripecnt().
3799 Bugzilla : 9489, 3273
3800 Description: First write from each client to each OST was only 4kB in size,
3801 to initialize client writeback cache, which caused sub-optimal
3802 RPCs and poor layout on disk for the first writen file.
3803 Details : Clients now request an initial cache grant at (re)connect time
3804 and so that they can start streaming writes to the cache right
3805 away and always do full-sized RPCs if there is enough data.
3806 If the OST is rebooted the client also re-establishes its grant
3807 so that client cached writes will be honoured under the grant.
3812 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3813 Details : Now I/O RPCs go to different portal number and (presumably) fast
3814 lock requests (and glimses) and other RPCs get their own service
3815 threads pool that should be able to service those RPCs
3818 Severity : enhancement
3820 Description: Ability to exchange lustre version between client and servers and
3821 issue warnings at client side if client is too old. Also for
3822 liblustre clients there is ability to refuse connection of too old
3824 Details : New 'version' field is added to connect data structure that is
3825 filled with version info. That info is later checked by server and
3829 Frequency : rare, liblustre only.
3830 Bugzilla : 9296, 9581
3831 Description: Two simultaneous writes from liblustre at offset within same page
3832 might proceed at the same time overwriting eachother with stale
3834 Details : I/O lock withing llu_file_prwv was released too early, before data
3835 actually was hitting the wire. Extended lock-holding time until
3836 server acknowledges receiving data.
3839 Frequency : extremely rare. Never observed in practice.
3841 Description: avoid generating lustre_handle cookie of 0.
3842 Details : class_handle_hash() generates handle cookies by incrementing
3843 global counter, and can hit 0 occasionaly (this is unlikely, but
3844 not impossible, because initial value of cookie counter is
3845 selected randonly). Value of 0 is used as a sentinel meaning
3846 "unassigned handle" --- avoid it. Also coalesce two critical
3847 sections in this function into one.
3849 Severity : enhancement
3851 Description: allow liblustre clients to delegate truncate locking to OST
3852 Details : To avoid overhead of locking, liblustre client instructs OST to
3853 take extent lock in ost_punch() on client's behalf. New connection
3854 flag is added to handle backward compatibility.
3856 Severity : enhancement
3857 Bugzilla : 4928, 7341, 9758
3858 Description: allow number of OST service threads to be specified
3859 Details : a module parameter allows the number of OST service threads
3860 to be specified via "options ost ost_num_threads={N}" in the
3861 OSS's /etc/modules.conf or /etc/modprobe.conf.
3865 Bugzilla : 6146, 9635, 9895
3866 Description: servers crash with bad pointer in target_handle_connect()
3867 Details : In rare cases when a client is reconnecting it was possible that
3868 the connection request was the last reference for that export.
3869 We would temporarily drop the export reference and get a new
3870 one, but this may have been the last reference and the export
3871 was just destroyed. Get new reference before dropping old one.
3873 Severity : enhancement
3874 Frequency : if client is started with failover MDS
3876 Description: Allow multiple MDS hostnames in the mount command
3877 Details : Try to read the configuration from all specified MDS
3878 hostnames during a client mount in case the "primary"
3881 Severity : enhancement
3883 Description: Stop sending data to evicted clients as soon as possible.
3884 Details : Check if the client we are about to send or are sending data to
3885 was evicted already. (Check is done every second of waiting,
3886 for which l_wait_event interface was extended to allow checking
3887 of exit condition at specified intervals).
3890 Frequency : rare, normally only when NFS exporting is done from client
3892 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3894 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3897 Frequency : always, if srand() is called before liblustre initialization
3899 Description: Liblustre uses system PRNG disturbing its usage by user application
3900 Details : Introduce internal to lustre fast and high-quality PRNG for
3901 lustre usage and make liblustre and some other places in generic
3902 lustre code to use it.
3904 Severity : enhancement
3905 Bugzilla : 9477, 9557, 9870
3906 Description: Verify that the MDS configuration logs are updated when xml is
3907 Details : Check if the .xml configuration logs are newer than the config
3908 logs stored on the MDS and report an error if this is the case.
3909 Request --write-conf, or allow starting with --old_conf.
3911 Severity : enhancement
3913 Description: Handle symlinks in the path when checking if Lustre is mounted.
3914 Details : Resolve intermediate symlinks when checking if a client has
3915 mounted a filesystem to avoid duplicate client mounts.
3920 Description: lconf can hit an error exception but still return success.
3921 Details : The lconf command catches the Command error exception at the top
3922 level script context and will exit with the associated exit
3923 status, but doesn't ensure that this exit status is non-zero.
3928 Description: failure of ptlrpc thread startup can cause oops
3929 Details : Starting a ptlrpc service thread can fail if there are a large
3930 number of threads or the server memory is very fragmented.
3931 Handle this without oopsing.
3934 Frequency : always, only if liblustre and non-default acceptor port was used
3936 Description: liblustre cannot connect to servers with non-default acceptor port
3937 Details : tcpnal_set_default_params() was not called and was therefore
3938 ignoring the environment varaible TCPNAL_PORT, as well as other
3939 TCPNAL_ environment variables
3944 Description: two objects could be created on the same OST for a single file
3945 Details : If an OST is down, in some cases it was possible to create two
3946 objects on a single OST for a single file. No problems other
3947 than potential performance impact and spurious error messages.
3951 Bugzilla : 5681, 9562
3952 Description: Client may oops in ll_unhash_aliases
3953 Details : Client dcache may become inconsistent in race condition.
3954 In some cases "getcwd" can fail if the current directory is
3960 Description: Inode refcounting problems in NFS export code
3961 Details : link_raw functions used to call d_instantiate without obtaining
3962 extra inode reference first.
3966 Bugzilla : 9942, 9903
3967 Description: Referencing freed requests leading to crash, memleaks with NFS.
3968 Details : We used to require that call to ll_revalidate_it was always
3969 followed by ll_lookup_it. Also with revalidate_special() it is
3970 possible to call ll_revalidate_it() twice for the same dentry
3971 even if first occurence returned success. This fix changes semantic
3972 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3973 reference on a request referred from the intent.
3974 ll_intent_release() then releases such a request.
3977 Frequency : rare, normally benchmark loads only
3979 Description: unlinked inodes were kept in memory on the client
3980 Details : If a client is repeatedly creating and unlinking files it
3981 can accumulate a lot of stale inodes in the inode slab cache.
3982 If there is no other client load running this can cause the
3983 client node to run out of memory. Instead flush old inodes
3984 from client cache that have the same inode number as a new inode.
3987 Frequency : SLES9 2.6.5 kernel and long filenames only
3988 Bugzilla : 9969, 10379
3989 Description: utime reports stale NFS file handle
3990 Details : SLES9 uses out-of-dentry names in some cases, which confused
3991 the lustre dentry revalidation. Change it to always use the
3995 Frequency : rare, unless heavy write-truncate concurrency is continuous
3996 Bugzilla : 4180, 6984, 7171, 9963, 9331
3997 Description: OST becomes very slow and/or deadlocked during object unlink
3998 Details : filter_destroy() was holding onto the parent directory lock
3999 while truncating+unlinking objects. For very large objects this
4000 may block other threads for a long time and slow overall OST
4001 responsiveness. It may also be possible to get a lock ordering
4002 deadlock in this case, or run out of journal credits because of
4003 the combined truncate+unlink. Solution is to do object truncate
4004 first in one transaction without parent lock, and then do the
4005 final unlink in a new transaction with the parent lock. This
4006 reduces the lock hold time dramatically.
4009 Frequency : rare, 2.4 kernels only
4011 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4012 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4013 before unmounting their filesystem in order to avoid deadlock.
4014 The kernel_locked() function in 2.4 kernels only checks whether
4015 the kernel lock is held, not whether it is this process that is
4016 holding it as 2.6 kernels do.
4021 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4022 Details : The client ptlrpc code may be trying to reconnect to a down
4023 server before a previous connection attempt has timed out.
4024 Increase the reconnect interval to be longer than the connection
4025 timeout interval to avoid sending duplicate connections to
4029 Frequency : echo_client brw_test command
4031 Description: fix echo_client to work with OST preallocated code
4032 Details : OST preallocation code (5137) didn't take echo_client IO path
4033 into account: echo_client calls filter methods outside of any
4034 OST thread and, hence, there is no per-thread preallocated
4035 pages and buffers to use. Solution: hijack pga pages for IO. As
4036 a byproduct, this avoids unnecessary data copying.
4040 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4041 Description: Client can oops in mdc_commit_close() after open replay
4042 Details : It was possible for the MDS to return an open request with no
4043 transaction number in mds_finish_transno() if the client was
4044 evicted, but without actually returning an error. Clients
4045 would later try to replay that open and may trip an assertion
4046 Simplify the client close codepath, and always return an error
4047 from the MDS in case the open is not successful.
4050 Frequency : rare, 2.6 OSTs only
4052 Description: OST may deadlock under high load on fragmented files
4053 Details : If there was a heavy load and highly-fragmented OST filesystems
4054 it was possible to have all the OST threads deadlock waiting on
4055 allocation of biovecs, because the biovecs were not released
4056 until the entire RPC IO was completed. Instead, release biovecs
4057 as soon as they are complete to ensure forward IO progress.
4059 Severity : enhancement
4061 Description: Support for specifying external journal device at mount
4062 Details : If an OST or MDS device is formatted with an external journal
4063 device, this device major/minor is stored in the ext3 superblock
4064 and may not be valid for failover. Allow detecting and
4065 specifying the external journal at mount time.
4070 Description: Mounting an MDS with pending unlinked files may cause oops
4071 Details : target_finish_recovery() calls mds_postrecov() which returned
4072 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4073 considers this an error and immediately begins cleaning up the
4074 lov, just after starting the mds_lov process
4076 Severity : enhancement
4078 Description: Implement 'lfs df' to report actual free space on per-OST basis
4079 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4080 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4081 usage of objects; '-h' to report in human readable format.
4083 ------------------------------------------------------------------------------
4085 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4092 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4093 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4094 at startup during orphan unlinking. Instead, skip buddy update
4095 before setup, as it will be regenerated after recovery is complete.
4100 Description: If an OST is inactive, its locks might reference stale inodes.
4101 Details : lov_change_cbdata() must iterate over all namespaces, even if
4102 they are inactive to clear inode references from the lock.
4104 Severity : enhancement
4105 Frequency : occasional, if non-standard max_dirty_mb used
4107 Description: Client will block write RPCs if not enough grant
4108 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4109 then the client will block writes while waiting for another RPC
4110 to complete instead of consuming its dirty limit. With change
4111 we get improved performance when max_dirty_mb is small.
4113 Severity : enhancement
4114 Bugzilla : 3389, 6253
4115 Description: Add support for supplementary groups on the MDS.
4116 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4117 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4118 lookups for user supplementary groups into a cache.
4122 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4123 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4124 for liblustre O_EXCL is enforced by MDS.
4128 Description: readdir never returns NULL in liblustre.
4129 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4130 dentry in struct dirent.
4134 Frequency : liblustre only, depends on application IO pattern
4135 Description: liblustre clients evicted if not contacting servers
4136 Details : Don't put liblustre clients into the ping_evictor list, so
4137 they will not be evicted by the pinger ever.
4139 Severity : enhancement
4141 Description: Add ability to evict clients by NID from MDS.
4142 Details : By echoing "nid:$NID" string into
4143 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4144 $NID would be instantly evicted from this MDS and from all active
4145 OSTs connected to it.
4149 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4150 Details : lookup_it_finish() used to query file size from OSTs that was not
4155 Description: service threads change working directory to that of init
4156 Details : Starting lustre service threads may pin the working directory
4157 of the parent thread, making that filesystem busy. Threads
4158 now change to the working directory of init to avoid this.
4162 Frequency : during shutdown only
4163 Description: shutdown with a failed MDS or OST can cause unmount to hang
4164 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4169 Frequency : 2.6 only
4170 Description: chmod/chown may include an extra supplementary group
4171 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4172 supplementary group and if none is specified this is used.
4175 Bugzilla : 5479 (6816)
4177 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4178 Details : If lookup is in progress on a file that is unlinked we might try
4179 to revalidate the inode and fail in revalidate after lookup is
4180 complete and ll_file_open() enqueues the open again but
4181 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4184 Frequency : always, if lconf --abort_recovery used
4186 Description: lconf --abort_recovery fails with 'Operation not supported'
4187 Details : lconf was attempting to abort recovery on the MDT device and not
4190 ------------------------------------------------------------------------------
4192 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4197 Frequency : rare (only unsupported configurations with a node running as an
4199 Bugzilla : 6514, 5137
4200 Description: Mounting a Lustre file system on a node running as an OST could
4202 Details : OSTs now preallocates memory needed to write out data at
4203 startup, instead of when needed, to avoid having to
4204 allocate memory in possibly low memory situations.
4205 Specifically, if the file system is mounted on on OST,
4206 memory pressure could force it to try to write out data,
4207 which it needed to allocate memory to do. Due to the low
4208 memory, it would be unable to do so and the node would
4209 become unresponsive.
4211 Severity : enhancement
4213 Description: Addition of lconf --service command line option
4214 Details : lconf now accepts a '--service <arg>' option, which is
4215 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4217 Severity : enhancement
4219 Description: Failover mode is now the default for OSTs.
4220 Details : By default, OSTs will now run in failover mode. To return to
4221 the old behaviour, add '--failout' to the lmc line for OSTs.
4223 Severity : enhancement
4225 Description: Health checks are now provided for MDS and OSTs
4226 Details : Additional detailed health check information on MSD and OSTs
4227 is now provided through the procfs health_check value.
4230 Frequency : occasional, depends on IO load
4232 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4233 after numerous create/delete cycles
4234 Details : The ext3 inode allocation policy would not allocate new inodes
4235 very well on the OSTs because there are no new directories
4236 being created. Instead we look for groups with free space if
4237 the parent directories are nearly full.
4242 Description: Network or server problems during mount may cause partially
4243 mounted clients instead of returning an error.
4244 Details : The config llog parsing code may overwrite the error return
4245 code during mount error handling, returning success instead
4251 Description: MDS can fail to allocate large reply buffers
4252 Details : After long uptimes the MDS can fail to allocate large reply
4253 buffers (e.g. zconf client mount config records) due to memory
4254 fragmentation or consumption by the buffer cache. Preallocate
4255 some large reply buffers so that these replies can be sent even
4256 under memory pressure.
4260 Frequency : rare (liblustre)
4261 Description: fsx running with liblustre complained that using truncate() to
4262 extend the file doesn't work. This patch corrects that issue.
4263 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4264 ATTR_SIZE and lsm use in llu_setattr_raw.
4268 Frequency : rare, only 2.6 kernels
4269 Description: Unusual file access patterns on the MDS may result in inode
4270 data being lost in very rare circumstances.
4271 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4272 code in the 2.6 kernel caused the inode and/or EA data not to
4273 be read from disk, causing single-file corruption.
4277 Frequency : rare, only 2.6 filesystems using extents
4278 Description: Heavy concurrent write and delete load may cause data corruption.
4279 Details : It was possible under high-load situations to have an extent
4280 metadata block in the block device cache from a just-unlinked
4281 file overwrite a newly-allocated data block. We now unmap any
4282 metadata buffers that alias just-allocated data blocks.
4286 Frequency : filesystems with default stripe_count larger than 77
4287 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4288 Details : lconf specifies an inode size of 4096 bytes when the default
4289 stripe_count is larger than 77. This conflicts with the default
4290 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4291 this case to avoid pinning too much memory for large EAs.
4293 ------------------------------------------------------------------------------
4295 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4300 Frequency : rare (extremely heavy IO load with hundreds of clients)
4302 Description: Client is evicted, gets IO error writing to file
4303 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4304 caused clients to be evicted for AST timeouts. The fixes in
4305 bug 5192 mean we no longer need to have such short AST timeouts
4306 so ldlm_timeout has been increased.
4309 Frequency : occasional during --force or --failover shutdown under load
4310 Bugzilla : 5949, 4834
4311 Description: Server oops/LBUG if stopped with --force or --failover under load
4312 Details : a collection of import/export refcount and cleanup ordering
4313 issues fixed for safer force cleanup
4316 Frequency : only filesystems larger than 120 OSTs
4317 Bugzilla : 5990, 6223
4318 Description: lfs getstripe would oops on a very large filesystem
4319 Details : lov_getconfig used kfree on vmalloc'd memory
4322 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4323 Bugzilla : 6242, 6243
4324 Description: reading from files that had been truncated to a non-zero size
4325 but never opened returned no data
4326 Details : ll_file_read() reads zeros from no-object files to EOF
4331 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4332 Details : There's little harm in aborting MDS/OSS recovery and letting it
4333 try again, so I removed the LASSERT and return an error instead.
4335 Severity : enhancement
4337 Description: New debugging infrastructure for tracking down data corruption
4338 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4339 (b) cover more of the client-side code path, and (c) try to narrow
4340 down where problems occurred
4344 Bugzilla : 3819, 4364, 4397, 6313
4345 Description: Racing close and eviction MDS could cause assertion in mds_close
4346 Details : It was possible to get multiple mfd references during close and
4347 client eviction, leading to one thread referencing a freed mfd.
4349 Severity: : enhancement
4350 Bugzilla : 3262, 6359
4351 Description: Attempts to reconnect to servers are now more aggressive.
4352 Details : This builds on the enhanced upcall-less recovery that was added
4353 in 1.4.2. When trying to reconnect to servers, clients will
4354 now try each server in the failover group every 10 seconds. By
4355 default, clients would previously try one server every 25 seconds.
4360 Description: After recovery, certain operations trigger a failed
4361 assertion on a client.
4362 Details : Failing over an mds, using lconf -d --failover, while a
4363 client was doing a readdir() call would cause the client to
4364 LBUG after recovery completed and the readdir() was resent.
4366 Severity : enhancement
4368 Description: Default groups are now added by lconf
4369 Details : You can now run lconf --group <servicename> without having to
4370 manually add groups with lmc.
4373 Frequency : occasional
4375 Description: Nodes with an elan id of 0 trigger a failed assertion
4378 Frequency : always when accessing e.g. tty/console device nodes
4380 Description: tty and some other devices nodes cannot be used on lustre
4381 Details : file's private_data field is used by device data and lustre
4382 values in there got lost. New field was added to struct file to
4383 store fs-specific private data.
4386 Frequency : when exporting Lustre via NFS
4388 Description: NFSD failed occasionally when looking up a path component
4389 Details : NFSD is looking up ".." which was broken in ext3 directories
4390 that had grown large enough to become hashed.
4393 Frequency : Clusters with multiple interfaces not on the same subnet
4395 Description: Nodes will repeatedly try to reconnect to an interface which it
4396 cannot reach and report an error to the log.
4397 Details : Extra peer list entries will be created by lconf with some peers
4398 unreachable. lconf now validates the peer before adding it.
4401 Frequency : Only if a default stripe is set on the filesystem root.
4403 Description: Setting a default stripe on the filesystem root prevented the
4404 filesystem from being remounted.
4405 Details : The client was sending extra request flags in the root getattr
4406 request and did not allocate a reply buffer for the dir EA.
4409 Frequency : occasional, higher if lots of files are accessed by one client
4410 Bugzilla : 6159, 6097
4411 Description: Client trips assertion regarding lsm mismatch/magic
4412 Details : While revalidating inodes the VFS looks up inodes with ifind()
4413 and in rare cases can find an inode that is being freed.
4414 The ll_test_inode() code will free the lsm during ifind()
4415 when it finds an existing inode and then the VFS later attaches
4416 this free lsm to a new inode.
4420 Bugzilla : 6422, 7030
4421 Description: MDS deadlock between mkdir and client eviction
4422 Details : Creating a new file via mkdir or mknod (starting a transaction
4423 and getting the ns lock) can deadlock with client eviction
4424 (gets ns lock and trying to finish a synchronous transaction).
4427 Frequency : occasional
4428 Description: While starting a server, the fsfilt_ext3 module could not be
4430 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4431 kernels. Previously, lconf would still use the ext3 name
4432 when trying to load modules. Now, it will correctly use
4433 ext3 on 2.4 and ldiskfs on 2.6.
4435 Severity : enhancement
4436 Description: The default stripe count has been changed to 1
4437 Details : The interpretation of the default stripe count (0, to lfs
4438 or lmc) has been changed to mean striping across a single
4439 OST, rather than all available. For general usage we have
4440 found a stripe count of 1 or 2 works best.
4442 Severity : enhancement
4443 Description: Add support for compiling against Cray portals.
4444 Details : Conditional compiling for some areas that are different
4448 Frequency : occasional
4449 Bugzilla : 6409, 6834
4450 Description: Creating files with an explicit stripe count may lead to
4451 a failed assertion on the MDS
4452 Details : If some OSTs are full or unavailable, creating files may
4453 trigger a failed assertion on the MDS. Now, Lustre will
4454 try to use other servers or return an error to the
4458 Frequency : occasional
4460 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4461 caused lock timeout during readahead (since 1.4.2).
4462 Details : Processes doing readahead might match a lock that hasn't been
4463 granted yet if there are overlapping and conflicting lock
4464 requests. The readahead process waits on ungranted lock
4465 (original lock is CBPENDING), while OST waits for that process
4466 to cancel CBPENDING read lock and eventually evicts client.
4468 Severity : enhancement
4470 Description: Initial enabling of flock support for clients
4471 Details : Implements fcntl advisory locking and file status functions.
4472 This feature is provided as an optional mount flag (default
4473 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4474 locking are implemented yet, and those that are are not guaranteed
4475 to be completely correct in production environments.
4476 mount -t lustre -o [flock|noflock] ...
4479 Frequency : occasional
4481 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4482 assertion in the ext3 JBD (journaling) layer.
4483 Details : The b_committed_data struct is protected by the big kernel lock
4484 in 2.4 kernels, serializing journal_commit_transaction() and
4485 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4486 access to this struct. In 2.6 kernels there is finer grained
4487 locking to improve SMP performance of the JBD layer.
4491 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4496 Description: merge and cleanup kernel patches.
4499 -----------------------------------------------------------------------------
4501 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4503 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4504 versions of Lustre. Please update all servers and clients to
4505 version 1.4.2 or later at the same time. You must also run
4506 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4507 to update the configuration logs.
4509 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4510 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4511 - swab llog records of type '0' so we get proper header size/idx (5861)
4512 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4513 - fix rename of one directory over another leaking an inode (5953)
4514 - avoid SetPageDirty on 2.6 (5981)
4515 - don't re-add just-being-destroyed locks to the waiting list (5653)
4516 - when creating new directories, inherit the parent's custom
4517 striping settings if present parent (3048)
4518 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4519 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4520 - don't hold client locks on temporary worklist from l_lru (5666)
4521 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4522 - automatically evict dead clients (5921)
4523 - Update file size properly in create+truncate+fstat case (6196)
4524 - Do not unhash mountpoint dentries, do not allow removal of
4526 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4527 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4528 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4529 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4530 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4531 - Correctly handle reads of files with no objects (6243)
4532 - lctl recover will also mark a device active if deactivate used (5933)
4534 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4535 - allow --write-conf on an MDS with different nettype than client (5619)
4536 - don't write config llogs to MDS for mounts not from that MDS (5617)
4537 - lconf should create multiple TCP connections from a client (5201)
4538 - init scripts are now turned off by default; run chkconfig --on
4539 lustre and chkconfig --on lustrefs to use them
4540 - upcalls are no longer needed for clients to recover to failover
4542 - add --abort-recovery option to lconf to abort recovery on device
4544 - add support for an arbitrary number of OSTs (3026)
4545 - Quota support protocol changes.
4546 - forward compatibility changes to wire structs (6007)
4547 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4548 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4549 - improved "lustre status" script
4550 - initialize blocksize for non-regular files (6062)
4551 - added --disable-server and --disable-client configure options (5782)
4552 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4553 - Vanilla 2.4.29 support
4554 - increase maximum number of obd devices to 520 (6242)
4555 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4556 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4557 SLES 9 2.6.5 kernels
4558 - limit stripes per file to 160 (the maximum EA size) (6093)
4560 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4563 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4564 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4565 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4566 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4567 - small ext3 extents cleanups and fixes (5733)
4568 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4569 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4570 - avoid needless client->OST connect, fix handle mismatch (5317)
4571 - fix DLM error path that led to out-of-sync client, long delays (5779)
4572 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4573 - fix several locking issues related to i_size (5492,5624,5654,5672)
4574 - don't move pending lock onto export if it is already evicted (5683)
4575 - fix kernel oops when creating .foo in unlinked directory (5548)
4576 - fix deadlock in obdfilter statistics vs. object create (5811)
4577 - use time_{before,after} to avoid timer jiffies wrap (5882)
4578 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4579 - Do not leak request if server was not able to process it (5154)
4580 - If mds_open unable to find parent dir, make that negative lookup(5154)
4581 - don't create new directories with extent-mapping (5909, 5936)
4583 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4584 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4585 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4586 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4587 - support for RHEL4 kernel on i686 (5773)
4588 - provide error messages when incompatible logs are encountered (5898)
4590 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4591 * version 1.4.0.10 (1.4.1 release candidate 1)
4593 - don't keep a lock reference when lock is not granted (4238)
4594 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4595 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4596 - drop import inflight refcount on signal_completed_replay error (5255)
4597 - unlock page after async write error during send (3677)
4598 - handle missing objects in filter_preprw_read properly (5265)
4599 - no transno return for symlink open, don't save no-trasno open (3440)
4600 - don't try to complete elan receive that already failed (4012)
4601 - free RPC server reply state on error (5406)
4602 - clean up thread from ptlrpc_start_thread() on error (5160)
4603 - readahead could read extra page into cache that wasn't ejected (5388)
4604 - prevent races in class_attach/setup/cleanup/detach (5260)
4605 - don't dereference de->d_inode after l_dput of de (5458)
4606 - use "int" for stripe value returned from lock_to_stripe (5544)
4607 - mballoc allocation and error-checking fixes in 2.6 (5504)
4608 - block device patches to fix I/O request sizes in 2.6 (5482)
4609 - look up hostnames for IB nals (5602)
4610 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4611 - don't start multiple acceptors for the same port (5277)
4612 - fix incorrect LASSERT in mds_getattr_name (5635)
4613 - export a proc file for general "ping" checking (5628)
4614 - fix "lfs check" to not block when the MDS is down (5628)
4616 - service request history (4965)
4617 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4618 - create an "evict_client" /proc entry on OSTs, like the MDS has
4619 - fix mount usage message, return errors per mount(8) (5168)
4620 - change grep [] to grep "[]" in tests so they work in more UMLs
4621 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4622 - remove ancient LOV_MAGIC_V0 EA support (5047)
4623 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4624 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4625 - allow readahead allocations to fail when low on memory (5383)
4626 - mmap locking landed again, after considerable improvement (2828)
4627 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4629 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4632 - send OST transaction number in read/write reply to free req (4966)
4633 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4634 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4635 - the watchdog thread now runs as interruptible (5246)
4636 - flock/lockf fixes (but it's still disabled, pending 5135)
4637 - don't use EXT3 constants in llite code (5094)
4638 - memory shortage at startup could cause assertion (5176)
4640 - reorganization of lov code
4641 - single portals codebase
4643 - add extents/mballoc support (5025)
4644 - direct I/O reads in the obdfilter (4048)
4645 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4647 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4650 - send OST transaction number in read/write reply to free req (4966)
4651 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4652 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4653 - the watchdog thread now runs as interruptible (5246)
4654 - handle missing objects in filter_preprw_read properly (5265)
4655 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4656 - drop import inflight refcount on signal_completed_replay error (5255)
4657 - unlock page after async write error during send (3677)
4658 - return original error code on reconstructed replies (3761)
4659 - no transno return for symlink open, don't save no-trasno open (3440)
4661 - add pid to ldlm debugging output (4922)
4662 - bump the watchdog timeouts -- we can't handle 30sec yet
4663 - extra debugging for orphan dentry/inode bug (5259)
4665 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4668 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4669 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4670 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4671 - fix NULL dereference in /proc/sys/portals/routes (4827)
4672 - allow failed mdc_close() operations to be interrupted (4561)
4673 - stop precreate on OST before MDS would time out on it (4778)
4674 - don't send partial-page writes before EOF from client (4410)
4675 - discard client grant for sub-page writes on large-page clients (4520)
4676 - don't free dentries not owned by NFS code, check generation (4806)
4677 - fix lsm leak if mds_create_objects() fails (4801)
4678 - limit debug_daemon file size, always print CERROR messages (4789)
4679 - use transno after validating reply (3892)
4680 - process timed out requests if import state changes (3754)
4681 - update mtime on OST during writes, return in glimpse (4829)
4682 - add mkfsoptions to LDAP (4679)
4683 - use ->max_readahead method instead of zapping global ra (5039)
4684 - don't interrupt __l_wait_event() during strace
4686 - add software watchdogs to catch hung threads quickly (4941)
4687 - make lustrefs init script start after nfs is mounted
4688 - fix CWARN/ERROR duplication (4930)
4689 - return async write errors to application if possible (2248)
4690 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4691 - print NAL number in %x format (4645)
4692 - update barely-supported suse-2.4.21-171 series (4842)
4693 - support for sles 9 %post scripts
4694 - support for building 2.6 kernel-source packages
4695 - support for sles km_* packages
4697 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4700 - ignore -ENOENT errors in osc_destroy (3639)
4701 - notify osc create thread that OSC is being cleaned up (4600)
4702 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4703 - reconstruct ost_handle() like mds_handle() (4657)
4704 - create a new thread to do import eviction to avoid deadlock (3969)
4705 - let lconf resolve symlinked-to devices (4629)
4706 - don't unlink "objects" from directory with default EA (4554)
4707 - hold socknal file ref over connect in case target is down (4394)
4708 - allow more than 32000 subdirectories in a single directory (3244)
4709 - fix blocks count for O_DIRECT writes (3751)
4710 - OST returns ENOSPC from object create when no space left (4539)
4711 - don't send truncate RPC if file size isn't changing (4410)
4712 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4713 - bind to privileged port in socknal and tcpnal (3689)
4715 - rate limit CERROR/CWARN console message to avoid overload (4519)
4716 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4717 - basic mmap support (3918)
4718 - kernel patch series update from b1_4 (4711)
4720 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4723 - avoid crash during MDS cleanup with OST shut down (2775)
4724 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4725 - don't use bad inodes on the MDS (3744)
4726 - dynamic object preallocation to improve recovery speed (4236)
4727 - don't hold spinlock over lock dumping or change debug flags (4401)
4728 - don't zero obd_dev when it is force cleaned (3651)
4729 - print grants to console if they go negative (4431)
4730 - "lctl deactivate" will stop automatic recovery attempts (3406)
4731 - look for existing locks in ldlm_handle_enqueue() (3764)
4732 - don't resolve lock handle twice in recovery avoiding race (4401)
4733 - revalidate should check working dir is a directory (4134)
4735 - don't always mark "slow" obdfilter messages as errors (4418)
4737 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4740 - don't close LustreDB during write_conf until it is done (3860)
4741 - fix typo in lconf for_each_profile (3821)
4742 - allow dumping logs from multiple threads at one time (3820)
4743 - don't allow multiple threads in OSC recovery (3812)
4744 - fix debug_size parameters (3864)
4745 - fix mds_postrecov to initialize import for llog ctxt (3121)
4746 - replace config semaphore with spinlock (3306)
4747 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4748 - don't allow enqueue to complete on a destroyed export (3822)
4749 - down write_lock before checking llog header bitmap (3825)
4750 - recover from lock replay timeout (3764)
4751 - up llog sem before sending rpc (3652)
4752 - reduce ns lock hold times when setting kms (3267)
4753 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4754 - fix NULL deref and obd_dev leak on setup error (3312)
4755 - replace some LBUG about llog ops with error handling (3841)
4756 - don't match INVALID dentries from d_lookup and spin (3784)
4757 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4758 - fix invalid assertion in ptlrpc_set_wait (3880)
4760 - add libwrap support for the TCP acceptor (3996)
4761 - add /proc/sys/portals/routes for non-root route listing (3994)
4762 - allow setting MDS UUID in .xml (2580)
4763 - print the stack of a process that LBUGs (4228)
4765 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4768 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4769 - make sure to unset replay flag from failed open requests (3440)
4770 - if default stripe count is 0, use OST count for inode size (3636)
4771 - update parent mtime/ctime on client for create/unlink (2611)
4772 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4773 - free recovery state on server during a forced cleanup (3571)
4774 - unregister_reply for resent reqs (3063)
4775 - loop back devices mounting and status check on 2.6 (3563)
4776 - fix resource-creation race that can provoke i_size == 0 (3513)
4777 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4778 - more debugging for page-accounting assertion (3746)
4779 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4780 - avoid deadlock after precreation failure (3758)
4781 - fix race and lock order deadlock in orphan handling (3450, 3750)
4782 - add validity checks when grabbing inodes from l_ast_data (3599)
4784 - add /proc/.../recovery_status to obdfilter (3428)
4785 - lightweight CDEBUG infrastructure, debug daemon (3668)
4786 - change default OSC RPC parameters to be better on small clusters
4787 - turn off OST read cache for files smaller than 32MB
4788 - install man pages and include them in rpms (3100)
4789 - add new init script for (un)mounting lustre filesystems (2593)
4790 - run chkconfig in %post for init scripts (3701)
4791 - drop scimac NAL (unmaintained)
4793 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4796 - clean kiobufs before and after use (3485)
4797 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4798 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4799 - change init script to fail more clearly if not run as root (1528)
4800 - allow clients to reconnect during replay (1742)
4801 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4802 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4803 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4804 - fix cond_resched (3554)
4805 - only evict unfinished clients after recovery (3515)
4806 - allow bulk resend, prevent data loss (3570)
4807 - dynamic ptlrpc request buffer allocation (2102)
4808 - don't allow unlinking open directory if it isn't empty (2904)
4809 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4810 - remove extraneous obd dereference causing LASSERT failure (3334)
4811 - don't use get_cycles() when creating temp. files on the mds (3156)
4812 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4813 - handle EEXIST for set-stripe, set proper directory name (3336)
4815 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4816 - fix ksocknal_fmb_callback() error messages (2918)
4818 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4821 - don't copy lvb into (possibly NULL) reply on error (2983)
4822 - don't deref dentry after dput, don't free lvb on error (2922)
4823 - use the kms to determine writeback rpc length (2947)
4824 - increment oti_logcookies when osc is inactive (2948)
4825 - update client's i_blocks count via lvb messages (2543)
4826 - handle intent open/close of special files properly (1557)
4827 - mount MDS with errors=remount-ro, like obdfilter (2009)
4828 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4829 - don't use cancelling-locks' kms values (2947)
4830 - use highest lock extent for kms, not last one (2925)
4831 - don't dereference ERR_PTR() dentry in error handling path (3107)
4832 - fix thread race in portals_debug_dumplog() (3122)
4833 - create lprocfs device entries at setup instead of at attach (1519)
4834 - common AST error handler, don't evict client on completion race (3145)
4835 - zero nameidata in detach_mnt in 2.6 (3118)
4836 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4837 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4838 - initialize RPC timeout timer earlier for 2.6 (3219)
4839 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4840 - print nal/nid for unknown nid (3258)
4841 - additional checks for oscc recovery before doing precreate (3284)
4842 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4843 - don't crash in mdc_close for bad permissions on open (3285)
4844 - zero i_rdev for non-device files (3147)
4845 - clear page->private before handing to FS, better assertion (3119)
4846 - tune the read pipeline (3236)
4847 - fix incorrect decref of invalidated dentry (2350)
4848 - provide read-ahead stats and refine rpc in flight stats (3328)
4849 - don't hold journal transaction open across create RPC (3313)
4850 - update atime on MDS at close time (3265)
4851 - close LDAP connection when recovering to avoid server load (3315)
4852 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4853 - don't leak open file on MDS after open resend (3325)
4854 - serialize filter_precreate and filter_destroy_precreated (3329)
4855 - loop device shouldn't call sync_dev() for nul device (3092)
4856 - clear page cache after eviction (2766)
4857 - resynchronize MDS->OST in background (2824)
4858 - refuse to mount the same filesystem twice on same mountpoint (3394)
4859 - allow llmount to create routes for mounting behind routers (3320)
4860 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4861 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4862 - fix rare problem with rename on htree directories (3417)
4864 - allow default OST striping configuration per directory (1414)
4865 - fix compilation for qswnal for 2.6 kernels (3125)
4866 - increase maximum number of MDS request buffers for large systems
4867 - change liblustreapi to be useful for external progs like lfsck (3098)
4868 - increase local configuration timeout for slow disks (3353)
4869 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4871 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4874 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4875 - don't overwrite extent policy data in reply if lock was blocked (2901)
4876 - drop filter export grants atomically with removal from device (2663)
4877 - del obd_self_export from work_list in class_disconnect_exports (2908)
4878 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4879 - swab reply message in mdc_close, other PPC fixes (2464)
4880 - fix destroying of named logs (2325)
4881 - overwrite old logs when running lconf --write_conf (2264)
4882 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4883 - fix race in target_handle_connect (2898)
4884 - mds_reint_create() should take same inode create lock (2926)
4885 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4886 - don't close files for self_export to avoid uninitialized obd (2936)
4887 - allow MDS with the same name as client node (2939)
4888 - hold dentry reference for closed log files for unlink (2325)
4889 - reserve space for all logs during transactions (2059)
4890 - don't evict page beyond end of stripe extent (2925)
4891 - don't oops on a deleted current working directory (2399)
4892 - handle hard links to targets without a parent properly (2517)
4893 - don't dereference NULL lock when racing during eviction (2867)
4894 - don't grow lock extents when lots of conflicting locks (2919)
4896 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4899 - account for cache space usage on clients to avoid data loss (974)
4900 - lfsck support in lustre kernel code (2349)
4901 - reduce journal credits needed for BRW writes (2370)
4902 - orphan handling to avoid losing space on client/server crashes
4903 - ptlrpcd can be blocked, stopping ALL progress (2477)
4904 - use lock value blocks to assist in proper KMS, faster stat (1021)
4905 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4906 - recovery for initial connections (2355)
4907 - fixes for mds_cleanup_orphans (1934)
4908 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4909 - block all file creations until orphan recovery completes (1901)
4910 - client remove rq_connection from request struct (2423)
4911 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4912 - recovery timer race (2670)
4913 - mdc_close recovey bug (2532)
4914 - ptlrpc cleanup bug (2710)
4915 - mds timeout on local locks (2588)
4916 - namespace lock held during RPCs (2431)
4917 - handle interrupted sync write properly (2503)
4918 - don't try to handle a message that hasn't been replied to (2699)
4919 - client assert failure during cleanup after abort recovery (2701)
4920 - leak mdc device after failed mount (2712)
4921 - ptlrpc_check_set allows timedout requests to complete (2714)
4922 - wait for inflight reqs when ptlrpcd finishes (2710)
4923 - make sure unregistered services are removed from the srv_list
4924 - reset bulk XID's when resending them (caught by 1138 test)
4925 - unregister_bulk after timeout
4926 - fix lconf error (2694)
4927 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4928 - readahead locks pages, leaves pending causing memory pressure (2673)
4929 - increase OST request buffers to 4096 on large machines (2729)
4930 - fix up permission of existing directories in simple_mkdir (2661)
4931 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4932 - don't assign transno to errored transactions (2742)
4933 - don't delete objects on OST if given a bogus objid from MDS (2751)
4934 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4935 - if rq_no_resend, then timeout request after recovery (2432)
4936 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4937 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4938 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4939 - MDS recovery shouldn't delete objects that it has given out (2730)
4940 - if enqueue arrives after completion, don't clobber LVB (2819)
4941 - don't unlock pages twice when trigger_group_io returns error (2814)
4942 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4943 - don't write pages to disk if there was an error (1450)
4944 - don't ping imports that have recovery disabled (2676)
4945 - take buffered bytes into account when balancing socknal conn (2817)
4946 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4947 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4948 - remove little-endian swabbing of llog records (1987)
4949 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4950 - retry reposting request buffers if they fail (1191)
4951 - grow extent at grant time to avoid granting a revoked lock (2809)
4952 - lock revoke doesn't evict page if covered by a second lock (2765)
4953 - disable VM readahead to avoid reading outside lock extents (2805)
4955 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4956 - updated kernel patches for hp-2.4.20 kernel (2681)
4958 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4961 - fix truncated write corruption (2366)
4962 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4964 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4965 - ensure len(uuid) < 37 in lmc (1171)
4966 - fix ia64 OOPS in llog_test (2255)
4967 - zero end of page at obdfilter for partial page writes (2648)
4968 - don't leave stale dentries around after renames (bug 2428)
4969 - fix timeouts when evicting a client with a single lock held (2642)
4970 - set deadline for the initial HELLO message to drain (2634)
4971 - print out dotted-quad IP addresses in the socknal (2302)
4973 - additional debugging for MDS client eviction problem (2443)
4974 - fix mkfsoptions support for osts (2603, 2604)
4976 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4979 - add series for the vanilla 2.6.0 kernel
4980 - add series for the vanilla 2.4.24 kernel
4981 - add series for a cray x86/64 UL kernel drop
4982 - fix xattr patches for the vanilla 2.4.19 series
4984 - generate true UUIDs in lmc (1171)
4985 - have portals stack dumping break in UML (2466)
4986 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4987 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4988 - wake callback waiting threads on client eviction (2460)
4989 - Add --ptldebug and --subsystem to lmc (1719)
4990 - update assertion to allow safe interrupt allocation
4991 - set rq_no_resend for cancel requests (2432)
4992 - recalculate ptlrpcd timeout after resend (2494)
4993 - call vfs_rmdir when removing pending directories (2368)
4994 - fix renaming a file to itself (2429)
4995 - lmc creates a default one-stripe lov (2454)
4996 - expand procfs space to handle large clusters (2326)
4997 - increase UML stack to avoid overflow
4998 - update lconf's list of debug and subsystem masks
4999 - fix lfs find --obd (2510)
5000 - /proc tunable for disabling filter read caching (2591)
5001 - stop rpm packages from altering slapd.conf (2301)
5002 - disable nagle in the socknal under 0conf (2578)
5003 - choose mds inode size based on stripe count (2572)
5004 - fix kernel-source rpm problems (2516)
5006 - add --disable-doc to avoid pdf generation (2421)
5007 - update documentation, tests, type-os, comments
5008 - avoid format warnings on ia64
5009 - remove the TOE NAL
5010 - tiny code cleanups by removing unused fields
5012 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5015 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5016 - remove the most blatant lies from BUILDING file (2371)
5017 - change default debug level to reasonable production setting
5018 - reduce client side cache size to reduce cache flush time
5019 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5020 - make TCP zerocopy and pinger support enabled by default (2476)
5021 - sync writes completed after process exits caused crashes (2319)
5022 - maintain correct mount count on the MDS (2356)
5023 - backout 1557, because 2316 wasn't really fixed
5024 - better file I/O statistics gathering in /proc
5025 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5026 - another kernel patch to fix zero-copy TCP function export
5027 - don't take duplicate lock when processing re-sent getattr (2420)
5028 - lctl uses obd_self_export instead of creating new conn (2353)
5029 - MDS/OST recovery case which requires object creation asserted (2425)
5030 - move lfs from /usr/sbin to /usr/bin in packages
5031 - fix race between mds_client_add and mds_client_free (2417)
5032 - use kmalloc instead of slabs in portals (2430)
5033 - don't create duplicate records when a failover MDS is present (2442)
5034 - remove unnecessary mount age check (2332)
5035 - don't remove directory inodes from locks prematurely (2451)
5036 - don't break if MDS service name is the same as hostname (2103)
5037 - fix races in client write RPC generation when cache full (2482)
5039 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5042 - remove now-unused request->rq_obd (278)
5043 - if an allocation fails, print out how much memory we've used (1933)
5044 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5045 - add the "configurable stack size" patch to most series files (1256)
5046 - ability to write large log records, for 100+ OST configs (2306)
5047 - fix NULL deref when filter_prep fails (2314)
5048 - fix operator precedence error in filter_sync
5049 - dynamic allocation of socknal TX descriptors (2315)
5050 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5051 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5052 - ensure that CWARN messages in Portals always get to the syslog
5053 - __init/__exit are not for prototype decls (ldlm_init/exit)
5054 - x86-64 compile warning fixes
5055 - fix gateway LMC keyword conflict (2318)
5056 - fix MDS lock inversions in getattr/reint paths (1844)
5057 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5058 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5059 - DLM race condition prevented some lock evictions (2328)
5060 - ENOMEM detection and retry on socknal sends (2230)
5061 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5062 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5063 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5064 - better error messages when a client is rejected during recovery (1505)
5065 - avoid cancelling locks which were never granted, after failure (2330)
5066 - fix i_sem/journal inversion in mds_client_add (2333)
5067 - fix truncate/getattr lock cycle deadlock (2334)
5068 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5069 - fix two rare exit paths which could leak an l_lock() ref (2321)
5070 - fencepost error in MDS/OST orphan recovery (2226)
5071 - make log record alignment 8 bytes (1988)
5072 - lstripe now fails when requested offset > ost_count (2237)
5073 - ensure that all kernel series have a complete list.h (1607)
5074 - fix crashes in special-file operations (2316)
5075 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5077 - allow configurable automake binary, for testing new versions
5078 - small update to the lfs documentation
5080 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5082 * fix negative export reference count in fsfilt_sync (2312)
5084 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5085 * release candidate 0.9.1
5087 - orphans are moved into the PENDING directory for possible recovery
5088 - replayed opens now open by fid for orphan/rename safety (1042)
5089 - last close of an orphan inode generates a transno (683)
5090 - chdir() and mount() now pin the directory entry (1020)
5091 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5092 - discard very old requests without processing them (1502)
5093 - remove some common, well-understood CERRORs (1505)
5094 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5095 - clear "grant" flags in OST replies until OST grant code lands (1644)
5096 - fix read performance by not clobbering i_blksize on client (1598)
5097 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5098 - make LRU size a /proc tunable, clears locks when reduced (707)
5099 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5100 - server locks take references on exports now (1558)
5101 - build fixes for 2.4.20-rh trees (1663)
5102 - return an error from lov_create if all OSCs are inactive (1751)
5103 - fix import levels when a reconnect happens without a timeout (1597)
5104 - exit early from mds_open if we get a lookup error (1749)
5105 - partial page read at EOF wouldn't wait for disk before sending (1642)
5106 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5107 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5108 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5109 - write/truncate lock inversion (1639)
5110 - Don't auto-load obdclass, portals modules during cleanup (1495)
5111 - fix timestamps from jumping to "now" (1763)
5112 - extra journal assertions (1648)
5113 - add an extra multiunlink test (1771)
5114 - fix read_record/write_record API (1776)
5115 - fix leak of offset_extent, possible incorrect i_size later (1772)
5116 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5117 - Debugging for the kqswnal_get_idle_tx problems (1820)
5118 - Allow recovery to be attempted multiple times (1536)
5119 - Write out MDS last_rcvd file after it is first created (1600)
5120 - Fix tx_descriptor leak in failed transmit situations (1827)
5121 - ext3 journaling fixes for assertion failure after IO error (1871)
5122 - class_export_put() on freed export after completion AST error (1896)
5123 - Fix revalidate looping in VFS (1322)
5124 - Don't access a freed export during MDS_REINT timeout (1521)
5125 - Add open-unlink recovery support on the MDS (1673,1764)
5126 - Return an error if no MDS data was read from last_rcvd (1946)
5127 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5128 - Don't setup a disk device that doesn't match exported UUID (317)
5129 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5130 - avoid committing NULL handle in force close
5131 - local.sh is now a one-stripe LOV configuration
5132 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5133 - fix ext3 htree duplicate directory entry corruption (1516)
5134 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5135 - update to new LOV EA format (2097)
5136 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5138 2003-06-15 Phil Schwan <phil@clusterfs.com>
5141 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5142 - per-import/export recovery handling (958, 931, 959)
5143 - multiple last-rcvd slots, for serving multiple FSes (949)
5144 - connections are again shared between multiple imp/exports (963, 964)
5145 - "umount -f" would hang if any requests needed to be sent (393, 978)
5146 - avoid pinning large req buffer by copying for queued messages (989)
5147 - add "uuid" to "lctl device" command to help upcalls (991)
5148 - "open" RPCs with transnos would confuse recovery counters (1037)
5149 - do proper endian conversion of all wire messages (288, 340, 891)
5150 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5151 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5152 - fix ptlrpc_connection leak in target_handle_connect (1174)
5153 - fix import refcounting bug in OST and MDS cleanup (1134)
5154 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5155 - fix very unlikely obd_types race condition (501)
5156 - remove osc_open hack for echo_client (1187)
5157 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5158 - a failure in read_inode2 leads to deadlock (1139)
5159 - cancel ack-locks as soon as transaction is committed (1072)
5160 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5161 - make sure to commitrw after any preprw to avoid deadlock (1162)
5162 - failing to execute a file in a lustre FS would lock inode (1203)
5163 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5164 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5165 - obd_uuid2tgt would walk off the end of the list (1255)
5166 - on IA64 the osc would give portals incorrect bulk size (1258)
5167 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5168 - fix lock inversion caused by new llite matching code (1282)
5169 - limit the number of dirty pages on a client to 10MB (1286)
5170 - timed out locks were not being corrected cancelled (1289)
5171 - fix O_DIRECT above 4GB on IA-32 (1292)
5172 * major user-visible changes
5173 - fail out/fail over policy now controlled by the upcall (993)
5175 - add OBD_PING to check server availability and failure (954)
5176 - lustre messages are now sent in sending host order (288, 340, 891)
5177 - add eadatalen to MDS getattr reply (340)
5178 - OST read replies may contain second buffer, with per-page status (593)
5180 2003-03-11 Phil Schwan <phil@clusterfs.com>
5183 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5184 - failed open()s could cause deadlock; fixed (867, 869)
5185 - stop cancelling OST locks when files are closed (481)
5186 - overlapping XID spaces caused network corruption (851, 853)
5187 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5188 - setattr_raw added, to do single-RPC, server-side setattrs
5189 - lmc/lconf syntax change for OST UUIDs
5190 - fix crashy race condition between ptlrpc_free_req and osc_close
5191 - don't use request in mdc_enqueue if we hit a timeout (889)
5192 - don't set the inode i_size for regular files from the MDS (896)
5193 - handle out of order completion AST (842)
5194 - don't LBUG if a lock request times out after receiving AST (913)
5195 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5196 - if a bad lock AST arrives, send an error instead of dropping entirely
5197 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5198 - fix leak in bulk IO when only partially completed (899, 900, 926)
5199 - fix O_DIRECT for ia64 (55)
5200 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5201 - C-z after timeout could hang a process forever; fixed (977)
5203 - client-side I/O cache (678, 924, 929, 941, 970)
5205 - READPAGE and SETATTRs which don't take server-side locks get
5208 2003-02-11 Phil Schwan <phil@clusterfs.com>
5211 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5212 - Fix fsfilt_extN_readpage() to read a full page of directory
5213 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5214 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5215 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5216 - Most intent operations were removed, in favour of a new RPC mode
5217 that does a single RPC to the server and bypasses most of the VFS
5218 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5219 - Aggressively cancel local locks on DLM servers
5220 - mds_reint_unlink sends EA to the client if it's the last nlink.
5221 client uses that EA to unlink OST objects.
5222 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5223 - recursive symlinks were fixed (439)
5224 - fixed NULL deref in DEBUG_REQ
5225 - filter_update_lastobjid no longer calls sync, which annoyed extN
5226 - fixed multi-client small-writes to a single file problem (445)
5227 - fixed mtime updates during file writes (607)
5228 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5229 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5230 - fixed Linux OST size reporting problem (444, 656)
5231 - OST now updates object mtime with writes or setattr (607, 619)
5232 - client verifies file size before zeroing page past EOF (445)
5233 - OST now writes last allocated objid to disk with allocation (108)
5234 - LOV on echo now works (409)
5236 - mds_reint_unlink sends a new buffer, with the EA included. this
5237 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5238 set if a regular file was being unlinked, and it was the last link
5239 - use PtlGet from the target for bulk writes (315)
5240 - OST now updates object mtime with writes or setattr (607, 619)
5241 - LDLM now has a grant-time callback to revalidate locked items, if
5243 - Many MDS operations were reorganized to combat race conditions
5245 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5246 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5248 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5251 - Fully reactivate OST imports after reconnection (512, others)
5252 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5253 - More graceful error handling for truncating on dead OST (515)
5254 - Don't error out unless we're actually accessing dead stripes (474)
5255 - Fix garbage sizes when stripes are missing (410)
5256 - LRU counters were broken, causing constant lock purge (433, 432)
5257 - garbage on read from stripes with failed OSTs (441)
5258 - mark OSCs as active before reconnecting during recovery (438)
5259 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5260 - lfind did not preserve OST order in output (443)
5261 - symlinks cause hung clients, incorrect data (439)
5262 - stop dereferencing request after dropping refcount (457)
5263 - don't LASSERT(spin_is_locked) on non-SMP (455)
5264 - fixes for many rename() bugs
5265 - fstat didn't correctly synchronize attributes (399)
5266 - server must handle lock cancellation during blocking AST prep (487)
5267 - bulk descriptors were free()d too soon (511)
5268 - fix paths in lconf, which would load incorrect modules (451, 507)
5269 - fix confusing lconf 'host not found' error message (386)
5270 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5271 - fix race condition in mdc_blocking_ast() for inode access (526)
5272 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5273 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5274 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5275 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5276 - fix lstripe in several ways (kernel side) (527)
5277 - fix request leak in ldlm_cli_enqueue (262)
5278 - incorrect OSC was marked inactive after OST failure
5279 - call mds_fs_cleanup before unmounting filesystem (524)
5280 - fix races between taking ns_lock and ldlm_lock_change_resource
5281 - fix races updating LOV export open file list
5282 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5283 - fix recovery NULL deref in ldlm_cli_cancel_unused
5284 - fix some DLM races by using new hash table for lock handles (419)
5285 - permit the client to specify desired inodes, at replay
5286 - duplicate requests when we queue them for replay reintegration
5287 - fix last_rcvd offset calculation
5288 - sync after each recovered transaction, so we always make progress
5289 - never, not always, ERESTART requests without transnos
5290 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5292 - skip replay if the MDS says that the client is already connected
5293 - don't check for a recovery-enabled export to match lctl's UUID
5294 - don't INC_USE_COUNT for phantom exports
5295 - don't crash when cleaning up phantom exports (567)
5296 - don't double-finish or set replay data for errored mdc_open requests
5297 - abort requests when they time out, so we don't get old replies
5298 - send/receive replies for AST messages again
5299 - if the client says that it doesn't have the lock, cancel it on the
5301 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5302 mark it as destroyed, it will all work out when decref is called
5303 - fix module use counts (22, 581)
5305 - ASTs now expect a reply (server cancels lock on error reply)
5307 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5310 - fix many simultaneous client startup (392)
5311 - fix dentry->d_it clobbering
5312 - credentials weren't being shipped for readdir/getattr operations
5313 - remove invalid assertions triggered during some concurrent MD
5315 - proper Lustre versions added (336, 389)
5316 - fix memory leak for create error case (398)
5317 - fix LOV locking bug that would get cli/srv out of sync
5318 - fix echo client over LOV (409)
5319 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5320 - fix double-O_EXCL intent crash (424)
5321 - avoid sending multiple lock CANCELs (352)
5323 - MDS can do multi-client recovery (modulo bugs in new code)
5325 - many updates, edits, cleanups
5327 2002-11-18 Phil Schwan <phil@clusterfs.com>
5330 - fix null d_it dereference (346)
5331 - fix full OST/dbench hang (333)
5332 - fix permission problem with file removal (286)
5333 - fix removal of OSCs from LOV when they fail
5334 - fix NULL deref during bulk timeout (214)
5335 - fix problems related to multiple filesystems on one MDS (241)
5336 - fixed serious subtle metadata locking bugs
5337 - free locks on clients when inodes are removed due to memory
5339 - fix inode pointer in lock data (285)
5340 - partial support for multiple MDS on a single host (241)
5341 - data locks weren't cancelled at clear_inode time (290, 311)
5342 - intent locks could lead to unbounded lock growth (205)
5343 - added a maximum lock count, an LRU list, and a flusher
5344 - fix multiple rename (365)
5345 - properly abstracted the echo client
5346 - OSC locked 1 byte too many; fixed
5347 - rewrote brw callback code:
5348 - fixed recovery bugs related to LOVs (306)
5349 - fixed too-many-pages-in-one-write crash (191)
5350 - fixed (again) crash in sync_io_timeout (214)
5351 - probably fixed callback-related race (385)
5353 - Add capability to MDS protocol
5354 - LDLM cancellations and callbacks on different portals
5356 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5359 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5360 - defer open object destruction to close time (601981, 138)
5361 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5362 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5363 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5364 - don't call obd_setattr in ll_file_release for destroyed objects
5365 * protocol change to lustre_msg: move |version| and add |flags|
5366 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5367 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5368 * added replay of create, unlink, link and rename operations during
5369 MDS failover; recovery should be much more robust now
5370 * remove failed OSCs from LOVs (only lov_create uses this so far)
5371 * the lustre-HOWTO was brought (more) up to date (582544)
5373 2002-10-23 Phil Schwan <phil@clusterfs.com>
5376 - in-use dentries weren't being reused properly (617851)
5377 - prevent multiple LDLM setup (599178)
5378 - fix LOV size calculations for truncate (617853)
5379 - fix client handling of MDS intent errors (POSIX)
5380 - fix permission bug in lovstripe.c test (624321)
5381 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5382 - truncate past end of file could corrupt data
5383 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5384 - a race in recovery could return ETIMEDOUT to apps (623947)
5385 - building outside the source directory was fixed
5386 * the lustre-HOWTO was brought (more) up to date (582544)
5387 * major progress was made on recovery functionality
5389 2002-10-10 Phil Schwan <phil@clusterfs.com>
5392 - recovery deadlock fix
5393 - rm -rf causes LBUG fix (617817)
5394 - file open by multiple tasks fix (618962)
5395 - directory permissions bugs (602707 and 620007)
5396 - journal_stop fixed with locking (611313)
5397 - O_APPEND failures resolved (618273, perhaps 614459)
5398 - lconf PATH fix (619770)
5399 - IA64 build fix (621450)
5400 - RPC buffer sizes scale with amount of memory
5402 2002-10-01 Phil Schwan <phil@clusterfs.com>
5405 - locks would be cancelled without throwing away data pages,
5406 resulting in inconsistent data (605627)
5407 - inode attributes were not always being refreshed (605627, 612449)
5408 - lconf now continues to cleanup after lctl reports an error
5409 - MDS now enforces user permissions (602707)
5410 - lprocfs cleanup fixed, but not yet enabled (614157)
5411 - fixed infinite server hang, should a client not respond to an AST
5412 - avoid going into recovery if user calls readlink() with a buffer
5413 that's too small (613941)
5414 - AST RPCs no longer require replies (614867) -- this may be changed
5415 - don't crash server if client sends an IOV that's too big (611336)
5416 - fixed lock conversion deadlock (611892)
5417 - fixed the following of symlinks (614622)
5418 * recovery: the server can remove locks from a client that dies, other
5419 clients can make progress
5420 * more extN patch fixes
5421 * compile-time configurable ptlrpc buffer allocations
5423 - collaborative read cache document
5424 - Lustre Lite Performance CDR document-in-progress
5426 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5429 - fix typo in patch-2.4.18
5431 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5434 - clear ptlrpc request each time in handle_incoming_request()
5435 - unlink of files now destroys the object on the OST
5437 2002-09-19 Peter Braam <braam@clusterfs.com>
5439 * add hard link support
5440 * change obdfile creation method
5441 * kernel patch changed
5443 2002-09-19 Peter Braam <braam@clusterfs.com>
5446 - stack overflow bug in extN fixed
5448 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5450 * documentation updates
5451 - add man pages for config tools
5452 - update tests/README to describe testing with new config tools
5453 - finish metadata API descriptions
5454 * bug fixes and cleanups
5455 - statfs workaround for 16TB limit
5456 - LOV stripe allocation improved, can stripe on subset of OSTs
5457 - LOV file size/IO offset was wrong for files > 4GB in size
5458 - object EA data was being dropped, caused files to be unreadable
5459 - memory overflow with non-LOV OST caused memory corruption
5460 - fixed regression tests to work with new config tools, obdfilter
5461 - fixed bug when directory size became larger than 1 block
5462 - fixed bug (for single client case) when PWD was deleted
5463 - invalidate local directory pages when doing intent-based ops
5464 - avoid LDLM oops when lock callback contained bad data
5466 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5468 * documentation updates
5469 * bug fixes and cleanups
5470 - configuration tools
5473 - 64-bit compile warnings
5474 - 64-bit internal statfs data
5476 * test_brw on persistent OST devices
5478 * lprocfs (disabled)
5480 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5482 * documentation updates
5483 * bug fixes and cleanups
5484 * configuration tools
5486 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5489 * many small fixes to 0_5_4
5490 * io/network handling
5491 * thinkos in MDS operations
5493 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5496 * crucial basic fixes to 0.5.3
5497 * IOR, Iozone work over Elan
5500 2002-08-07 Phil Schwan <phil@clusterfs.com>
5501 * version 0_5_3, our first alpha
5502 * we use the new Portals iovs
5503 * documentation updates
5504 * bug fixes and cleanups
5505 * small changes in the DLM wire protocol
5507 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5508 * version 0_5_1 with some initial stability,
5509 * locking on MD and file I/O.
5510 * documentation updates
5511 * several bug fixes since 0.5.0
5512 * small changes in wire protocol
5514 2002-07-18 Phil Schwan <phil@clusterfs.com>
5516 * delivered as Lustre Light Alpha
5517 * fixed a crash after handling invalid MDS requests
5518 * fixed directory pages for architectures with non-4k pages sizes
5520 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5521 * release version v0_4_4
5522 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5524 * Fixes a number of interruption problems with OST operations.
5525 * Update documentation for portals header changes
5526 * Move all wire protocol structs/defines to lustre_idl.h
5527 * Fixes symlink length bug.
5528 * Add tcpdump to repository.
5530 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5531 * release version v0_4_3
5532 * Fixes statfs for inodes on extN.
5533 * Fixes bug in runtests which would delete /etc/hosts.
5534 * Use 64-bit object IDs wherever possible (not into VFS though)
5535 Remove ost_get_info, which is unused by lustre, and out of date.
5537 2002-07-03 Peter Braam <braam@clusterfs.com>
5538 * release version v0_4_2 Fixes a lookup error (type not passed)
5539 * move forward to head of Portals
5540 * move forward to latest Lustre kernel
5542 2002-06-25 Peter Braam <braam@clusterfs.com>
5543 * release version v0_4_1. Hopefully stable on single node use.