1 tbd Sun Microsystems, Inc.
5 2.6.9-67.0.15.EL (RHEL 4),
6 2.6.16.54-0.2.5 (SLES 10),
7 2.6.18-53.1.14.el5 (RHEL 5),
8 2.6.22.14 vanilla (kernel.org)
9 * Client support for unpatched kernels:
10 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
11 2.6.16 - 2.6.22 vanilla (kernel.org)
12 * Client support for unpatched kernels:
13 we do not recommend using patchless RHEL4 clients with kernels
14 prior to 2.6.9-55EL (RHEL4U5).
15 * Recommended e2fsprogs version: 1.40.7-sun3
16 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
17 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
18 removed cwd "./" (refer to Bugzilla 14399).
19 * A new quota file format has been introduced in 1.6.5.
20 The format conversion from prior releases is handled transparently,
21 but releases older than 1.4.12/1.6.5 will not understand this new
22 format. The automatic format conversion can be avoided by running
23 the following command on the MDS:
24 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
25 For more information, please refer to bugzilla 13904.
27 Severity : enhancement
29 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
33 Description: enable MGS and MDT services start separately
34 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
35 a co-located MGS without starting the MGS, which is a complement
36 to 'nosvc' mount option.
39 Frequency : always, on ppc.
41 Description: cleanup in ptlrpc code, related to ppc platform
42 Details : store magic in native order avoid panic's in recovery on ppc node
43 and forbid from this error in future. Also fix posibily of twice swab
44 data. Fix get lov striping to userpace.
48 Frequency : rare, replay get lost on server
49 Description: server incorrectly drop resent replays lead to recovery failure.
50 Details : do not drop replay according to msg flags, instead we check the
51 per-export recovery request queue for duplication of transno.
55 Frequency : after recovery
56 Description: precreate to many object's after del orphan.
57 Details : del orphan st in oscc last_id == next_id and this triger growing
58 count of precreated objects. Set flag LOW to skip increase count
59 of precreated objects.
63 Frequency : rare, on clear nid stats
64 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
65 Details : when clean nid stats sometimes try destroy live entry,
66 and this produce panic in free.
70 Description: Stack overflow during MDS log replay
71 Details : ease stack pressure by using a thread dealing llog_process.
76 Description: MDT cannot be unmounted, reporting "Mount still busy"
77 Details : Mountpoint references were being leaked during open reply
78 reconstruction after an MDS restart. Drop mountpoint reference
79 in reconstruct_open() and free dentry reference also.
83 Description: wait until IO finished before start new when do lock cancel.
84 Details : VM protocol want old IO finished before start new, in this case
85 need wait until PG_writeback is cleared until check dirty flag and
86 call writepages in lock cancel callback.
90 Description: mds_mfd_close() ASSERTION(rc == 0)
91 Details : In mds_mfd_close(), we need protect inode's writecount change
92 within its orphan write semaphore to prevent possible races.
96 Frequency : rare, on shutdown ost
97 Description: don't hit live lock with umount ost.
98 Details : shrink_dcache_parent can be in long loop with destroy dentries,
99 use shrink_dcache_sb instead.
103 Description: don't panic with use echo client
104 Details : echo client pass NULL as client nid pointer and this produce null
109 Description: fix build on ppc32
110 Details : compile code with -m64 flag produce wrong object file for ppc32.
114 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
115 Details : In reconstruct_* functions, LASSERTs on both the data supplied by
116 a client, and the data on disk are dangerous and incorrect. Change
117 them with client eviction.
121 Description: skiplist implementation simplification
122 Details : skiplists are used to group compatible locks on granted list
123 that was implemented as tracking first and last lock of each lock group
124 the patch changes that to using doubly linked lists
128 Description: delete compatibility for 32bit qdata
129 Details : as planned, when lustre is beyond b1_8, lquota won't support for 32bit
130 qunit. That means servers of b1_4 and servers of b1_8 can't be used
131 together if users want to use quota.
133 -------------------------------------------------------------------------------
136 2008-05-26 Sun Microsystems, Inc.
138 * Support for kernels:
139 2.6.5-7.311 (SLES 9),
140 2.6.9-67.0.7.EL (RHEL 4),
141 2.6.16.54-0.2.5 (SLES 10),
142 2.6.22.14 vanilla (kernel.org)
143 * Client support for unpatched kernels:
144 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
145 2.6.16 - 2.6.22 vanilla (kernel.org)
146 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
147 we do not recommend using patchless RHEL4 clients with kernels
148 prior to 2.6.9-55EL (RHEL4U5).
149 * Recommended e2fsprogs version: 1.40.7-sun1
150 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
151 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
152 removed cwd "./" (refer to Bugzilla 14399).
153 * A new quota file format has been introduced in 1.6.5.
154 The format conversion from prior releases is handled transparently,
155 but releases older than 1.4.12/1.6.5 will not understand this new
156 format. The automatic format conversion can be avoided by running
157 the following command on the MDS before upgrading:
158 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
159 For more information, please refer to bugzilla 13904.
163 Description: quota performance fix
164 Details : quota data is written in journalled mode instead of ordered to
169 Description: lfs support for human-readable quota grace time strings
170 Details : lfs setquota -t and lfs quota -t represent quota grace times
171 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
174 Frequency : always with o2ib 1.3 and sles10
176 Description: fix build with SLES10 and o2ib v3.
177 Details : sles10 uses diffrent name for Module.symver file but configure
178 assume this file has same name on RHEL/SLES/vanila kernels.
181 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
183 Description: MDS may lose file striping (and hence file data) in some cases
184 Details : If there are additional extended attributes stored on the MDS,
185 in particular ACLs, SELinux, or user attributes (if user_xattr
186 is specified for the client mount options) then there is a risk
187 of attribute loss. Additionally, the Lustre file striping
188 needs to be larger than default (e.g. striped over all OSTs),
189 and an additional attribute must be stored initially in the
190 inode and then increase in size enough to be moved to the
191 external attribute block (e.g. ACL growing in size) for file
194 Severity : enhancement
196 Description: add message levels for liblustreapi
201 Description: MDT cannot be unmounted, reporting "Mount still busy"
202 Details : Mountpoint references were being leaked during open reply
203 reconstruction after an MDS restart. Drop mountpoint reference
204 in reconstruct_open() and free dentry reference also.
209 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
210 Details : Move the 'good_osts' check before the 'total_bavail' check. This
211 will result in an -EAGAIN and in the exit call path we call
212 alloc_rr() which will with increasing aggressiveness attempt to
213 aquire precreated objects on the minimum number of required OSCs.
217 Description: Use old size assignment to avoid deadlock
218 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
219 the scheduling while holding a spinlock. We do not need locking
220 for size in ll_update_inode() because size is only updated from
221 the MDS for directories or files without objects, so there is no
222 other place to do the update, and concurrent access to such inodes
223 are protected by the inode lock.
227 Description: Use __u64 instead of int for valid bits
231 Description: resolve "_IOWR redefined" build error on SLES10
235 Description: dump the memory debugging after all modules are unloaded to
236 suppress false negative in conf_sanity test 39
240 Description: the recovery timer never expires
241 Details : for new client connect request, the recovery timer should not be
242 reset, otherwise recovery timer will never expired, if the old
243 client never come. Only old client connect and first connection
244 req should trigger recovery timer reset.
248 Description: the min numbers of lproc stats are wrong
249 Details : adding a new constant LC_MIN_INIT and use it for initialization
253 Frequency : always with interactive lfs
255 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
258 Frequency : with multiple concurrent readdir processes in same directory
259 Bugzilla : 15406, 15169, 15175
260 Description: misc fixes for directory readahead.
261 Details : prevent previous statahead async RPC callback from processing the
262 current "statahead_info", race condition between async RPC callback
263 add dentry into dentry hash table and "ls" thread revalidate such
264 dentry, statahead his/miss control for hidden items, and so on.
266 Severity : enhancement
268 Description: build kernel-ib packages for OFED 1.3 in our release cycle
272 Description: incore types cleaning in quota code (with respect to 64-bit limits)
273 Details : several u32 variables declarations are replaced with u64 declarations
278 Description: fix SLES kernel versioning
279 Details : the kernel version for our SLES 10 kernel did not include a "-"
280 before the "smp" at the end. while this was not a problem in
281 general, it did mean that software trying to use the kernel
282 version to try to detect a vendor specific kernel would fail.
283 this was most evident by the OFED build scripts.
288 Description: Don't update lov_desc members until making sure they are valid
289 Details : When updating lov_desc members via proc fs, need fix their
290 validities before doing the real update.
293 Frequency : very rare
295 Description: don't put request into delay list while invalidate in flight.
296 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
297 import in flight. this produce timeout for invalidate and sometimes
298 can cause stale data.
300 Severity : enhancement
302 Description: Update kernel to SLES9 2.6.5-7.311.
304 Severity : enhancement
306 Description: Update kernel to RHEL4 2.6.9-67.0.7.
311 Frequency : on PPC only
312 Description: not convert ost objects for directory because it's not exist.
313 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
315 Severity : enhancement
317 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
318 __u64 as usingied long long type.
321 Frequency : rare, on shutdown
323 Description: race process ast vs remove callback
324 Details : removing callback before disconnect import open race with
327 Severity : enhancement
329 Description: Update kernel to SLES9 2.6.5-7.311.
331 Severity : enhancement
333 Description: Files open for execute are not marked busy on SLES10
334 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
336 Severity : enhancement
338 Description: Add server support for vanilla-2.6.22.14.
341 Frequency : occasional
343 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
344 Details : If an OST is down the MDS will hang indefinitely in
345 obd_statfs_rqset() waiting for the statfs data. While for
346 MDS QOS usage of statfs, it should not stuck in waiting.
348 Severity : enhancement
350 Description: Disable adaptive timeouts by default
353 Frequency : on network error
355 Description: panic with double free request if network error
356 Details : mdc_finish_enqueue is finish request if any network error occuring,
357 but it's true only for synchronus enqueue, for async enqueue
358 (via ptlrpcd) this incorrect and ptlrpcd want finish request
362 Frequency : rare, on recovery
364 Description: read procfs can produce deadlock in some situation
365 Details : Holding lprocfs lock with send rpc can produce block for destroy
366 obd objects and this also block reconnect with -EALREADY.
367 This isn't fix all lprocfs bugs - but make it rare.
369 Severity : enhancement
371 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
374 Frequency : frequent on X2 node
376 Description: mdc_set_open_replay_data LBUG
377 Details : Set replay data for requests that are eligible for replay.
382 Description: lustre_mgs: operation 101 on unconnected MGS
383 Details : When MGC is disconnected from MGS long enough, MGS will evict the
384 MGC, and late on MGC cannot successfully connect to MGS and a lot
385 of the error messages complaining that MGS is not connected.
388 Frequency : rare, depends on device drivers and load
390 Description: MDS or OSS nodes crash due to stack overflow
391 Details : Code changes in 1.6.4 increased the stack usage of some functions.
392 In some cases, in conjunction with device drivers that use a lot
393 of stack, the MDS (or possibly OSS) service threads could overflow
394 the stack. One change which was identified to consume additional
395 stack has been reworked to avoid the extra stack usage.
397 Severity : enhancement
399 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
401 Severity : enhancement
403 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
405 Severity : enhancement
407 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
409 Serverity : enhancement
411 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
414 Frequency : rare on shutdown OST
416 Description: Don't allow skipping OSTs if index has been specified.
417 Details : Don't allow skipping OSTs if index has been specified, make locking
418 in internal create lots better.
423 Description: ASSERTION(!PageDirty(page)) failed
424 Details : Wrong check could lead to an assertion failure under specific
430 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
431 Details : access to bitfield in structure is always rounded to long
432 and this produce problem with not atomic change any bit.
437 Description: Lustre make rpms failed.
438 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
439 builds Lustre rpms from tarball.
441 Severity : enhancement
443 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
446 Frequency : rare on shutdown OST
448 Description: If llog cancel was not send before clean_exports phase, this can
449 produce deadlock in llog code.
450 Details : If llog thread has last reference to obd and call class_import_put
451 this produce deadlock because llog_cleanup_commit_master wait when
452 last llog_commit_thread exited, but this never success because was
453 called from llog_commit_thread.
456 Frequency : only if OST index is skipped
458 Description: NULL lov_tgts causing MDS oops
459 Details : more safe checks for NULL lov_tgts for avoid oops.
461 Severity : enhancement
463 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
465 Severity : enhancement
467 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
472 Description: make mgs_setparam() handle fsname containing dash
473 Details : fsname containing a dash does not work with lctl conf_param
475 Severity : enhancement
477 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
480 Frequency : rare, in recovery and (or) destroy lovobjid file.
482 Description: rewrite lov objid code.
483 Details : Cleanup for lov objid code, remove scability problems and wrong
484 locking. Fix sending last_id into ost.
486 Severity : enhancement
488 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
490 Severity : enhancement
492 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
493 Details : Use d_move_locked instead of __d_move.
496 Frequency : rare, at shutdown
498 Description: access already free / zero obd_namespace.
499 Details : if client_disconnect_export was called without force flag set,
500 and exist connect request in flight, this can produce access to
501 NULL pointer (or already free pointer) when connect_interpret
502 store ocd flags in obd_namespace.
505 Frequency : only at startup
507 Description: not alloc memory with spinlock held.
508 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
509 if any spinlock held.
514 Description: lfs find does not continue on file error
515 Details : Continue other files processing when a file/dir is absent.
519 Description: Inconsistent usage of lustre_pack_reply()
520 Details : Standardize the usage of lustre_pack_reply() such that it
521 always generate a CERROR on failure.
524 Frequency : very rare
526 Description: Fix replay if there is an un-replied request and open
527 Details : In some cases, older replay request will revert the
528 mcd->mcd_last_xid on MDS which is used to record the client's
531 Severity : enhancement
533 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
535 Severity : enhancement
537 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
539 Severity : enhancement
541 Description: Update to SLES9 kernel-2.6.5-7.287.3.
543 Severity : enhancement
545 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
547 Severity : enhancement
549 Description: Build SLES10 patchless client fails
550 Details : The configure was broken by run ./configure with
551 --with-linux-obj=.... argument for patchless client. When the
552 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
553 can't search header adequately. Use absolute path such as
554 -I($LINUX)/include instead.
556 Severity : enhancement
558 Description: Lustre Page Accounting
559 Details : New macros for page alloc and free which enable accounting
560 of page allocation of Lustre. Use percpu counters to store memory
564 Frequency : only if debugging is disabled
566 Description: LASSERT_{REQ,REP}SWAB macros are buggy
567 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
568 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
569 these macros and replace them with their definitions instead.
574 Description: interrupt oig_wait produce painc on resend.
575 Details : brw_redo_request can be used for resend requests from ptlrpcd and
576 private set, and this produce situation when rq_ptlrpcd_data not
577 copyed to new allocated request and triggered LBUG on assert
578 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
579 ptlrpcd set if request is changed and can be safety changed to use
582 Severity : enhancement
584 Description: organize the server-side client stats on per-nid basis
585 Details : Change the structure of stats under obdfilter and mds to
593 The "uuid"s file would list the uuids of _active_ exports.
594 And the clear entry is to clear all stats and stale nids.
599 Description: Processes looping in ll_readdir() on Lustre clients finally causing
600 a full node pseudo-hang
601 Details : Concurrent access to the same directory from multiple clients with
602 intensive file creation/removal can cause a client node to spin in
603 ll_readdir(). i_version must be increased every time the lock is
604 cancelled to ensure a revalidate is done.
609 Description: touch file failed when fs is not full
610 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
611 otherwise we can get ENOSP while fs is not full.
614 Frequency : only for Cray XT3
615 Bugzilla : 12829/13455
616 Description: Changing primary group doesn't change the group lustre assigns to
618 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
619 overridden with the primary group provided by the group upcall,
620 whereas the supplied fsgid can be trusted if it is in the list of
621 supplementary groups returned by the group upcall.
623 Severity : enhancement
625 Description: Root Squash Functionality
626 Details : Implementation of NFS-like root squash capability. Specifically,
627 don't allow someone with root access on a client node to be able
628 to manipulate files owned by root on a server node.
630 Severity : enhancement
632 Description: Slow trucate/writes to huge files at high offsets.
633 Details : Directly associate cached pages to lock that protect those pages,
634 this allows us to quickly find what pages to write and remove
635 once lock callback is received.
640 Description: Too many locks accumulating on client during NFS usage
641 Details : mds_open improperly used accmode to find out access mode to a
642 file. Also mdc_intent_lock logic to find out if we already have
643 lock similar to just received was flawed since introduction of
644 skiplists - locks are now added to the front of the granted
649 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
651 Details : In obd setup failure handler, obd_stopping will not necessarily to
652 be 1, and obd_set_up should also be checked to make sure whether
653 obd is completely setup.
655 Severity : enhancement
657 Description: Allow masking D_WARNING, D_ERROR messages from console
658 Details : Console messages can now be disabled via lnet.printk.
663 Description: User code with malformed file open parameter crashes client node
664 Details : Before packing join_file req, all the related reference should be
665 checked carefully in case some malformed flags cause fake join_file
671 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
672 grain of quota doing harm to quota's accuracy
673 Details : qunit size will be changed when quota limitation is too low/high;
674 record the pending quota write in order to get more accureate
675 quota; delete the patch for bug12588, which is unnecessary when
676 this patch is landed. This bug also contains fixes for bug 14526,
677 14299, 14601 and 13794.
681 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
682 Details : ldlm_completion_ast() assumes that a lock is granted when the req
683 mode is equal to the granted mode. However, it should also check
684 that LDLM_FL_CP_REQD is not set.
688 Description: Heavy nfs access might result in deadlocks
689 Details : After ELC code landed, it is now improper to enqueue any mds
690 locks under och_sem, because enqueue might want to decide to
691 cancel open locks for same inode we are holding och_sem for.
695 Description: 35% write performance drop with ldiskfs2 when quotas are on
696 Details : Enable ext3 journalled quota by default to improve performance
697 when quotas are turned on.
701 Description: Client eviction while running blogbench
702 Details : A lot of unlink operations with concurrent I/O can lead to a
703 deadlock causing evictions. To address the problem, the number of
704 oustanding OST_DESTROY requests is now throttled to
705 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
706 callbacks are processed in priority.
709 Frequency : RHEL4 only
711 Description: mkfs is very slow on IA64/RHEL4
712 Details : A performance regression has been discovered in the MPT Fusion
713 driver between versions 3.02.73rh and 3.02.99.00rh. As a
714 consequence, we have downgraded the MPT Fusion driver in the RHEL4
715 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
718 Frequency : PPC/PPC64 only
720 Description: conflicts between asm-ppc64/types.h and lustre_types.h
721 Details : fix duplicated definitions between asm-ppc64/types.h and
722 lustre_types.h on PPC.
725 Frequency : PPC/PPC64 only
727 Description: asm-ppc/segment.h does not exist
728 Details : fix compile issue on PPC.
732 Description: data checksumming impacts single node performance
733 Details : add support for several checksum algorithms. Currently, CRC32 and
734 Adler-32 are supported. The checksum type can be changed on the fly
735 through /proc/fs/lustre/osc/*/checksum_type.
739 Description: use adler32 for page checksums
740 Details : when available, use the Adler-32 algorithm instead of CRC32 for
745 Description: better handle error messages in extents code
747 Severity : enhancement
749 Description: SNMP support enhancement
750 Details : Adding total number of sampled request for an MDS node in snmp
753 Severity : enhancement
755 Description: Optimize ldlm waiting list processing for PR extent locks
756 Details : When processing waiting list for read extent lock and meeting read
757 lock that is same or wider to it that is not contended, skip
758 processing rest of the list and immediatelly return current
759 status of conflictness, since we are guaranteed there are no
760 conflicting locks in the rest of the list.
764 Description: Time out and refuse to reconnect
765 Details : When the failover node is the primary node, it is possible
766 to have two identical connections in imp_conn_list. We must
767 compare not conn's pointers but NIDs, otherwise we
768 can defeat connection throttling.
772 Description: Client not clear own cache if answer to reconnect is lost.
773 Details : Client gets evicted from server. Now client also thinks it is
774 disconnected (or gets ENOTCONN on its operation) and decides to
775 reconnect. Server receives reconnect message, but cannot find
776 export. New export is created that is fully valid (new cookie!),
777 but reply is lost and not reported to client. Client reconnects
778 again and gets back a just-created connection, but it is not new
779 so client thinks it was not evicted and does not do recovery.
783 Description: Detect stride IO mode in read-ahead
784 Details : When a client does stride read, read-ahead should detect that and
785 read-ahead pages according to the detected stride pattern.
789 Description: build for x2 fails
790 Details : fix compile issue on Cray systems.
792 Severity : enhancement
794 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
795 Details : This greatly improves speed of NFS writes on 2.6 kernels.
798 Frequency : only on PPC/SLES10
800 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
801 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
802 be defined. Add a hack in mkfs_lustre.c to work around this compile
807 Description: LASSERT on MDS when client holding flock lock dies
808 Details : ldlm pool logic depends on number of granted locks equal to
809 number of released locks which is not true for flock locks, so
810 just exclude such locks from consideration.
814 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
815 Details : Use fsfilt_sync() for both the whole filesystem sync and
816 individual file sync to eliminate dangerous inode locking
817 with I_LOCK that can lead to a deadlock.
821 Description: Update an obsolete wirecheck.c generator
822 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
826 Description: Client can panic on open sometimes
827 Details : It is possible that we try to free already freed request in
828 ll_file_open in some error cases when we send request from
833 Description: performance in 1.6.3
834 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
835 sure that 1MB requests are not fragmented by the block layer.
839 Description: LDLM soft lockups - improvement
840 Details : It is be possible to send the lock handle along with each read
841 or write request because the client is already doing a lock match
842 itself so there isn't any reason the OST should have to re-do that
848 Description: lfs quota fails with deactivated OSTS
849 Details : With this patch, three improvements are included:
850 1. detete the softlimit in mds and osts when use "lfs quota".
851 2. display the inaccurate data in the output of "lfs quota".
852 3. try to get quota info when "lfs quota" is executed.
857 Description: Extent locks not granted with no conflicts sometimes.
858 Details : When race occurs in glimpse handler and nothing is returned,
859 we do not reprocess the queue after lock cancel, and that leads
860 to a stall until next activity on a resource
863 Frequency : failover with quotaon
865 Description: during mds failovers with quota on, OSTs got into deadlock state
866 and causing dumpstack.
867 Details : for every quota slave, at any time, there is only one quota req
868 is sent to quota master for every uid/gid. Before that quota req
869 returns, all the thread relative to the same uid/gid will wait.
870 So if the quota req is lost because mds failovers or any other
871 reasons, this bug will be hit. Now, dqacq_interpret() will handle
872 quota reqs who time out.
874 Severity : enhancement
877 Description: when quota slave checks if quota is enough, there is an unnecessary
879 Details : place this wait on necessary place instead of always waiting.
881 --------------------------------------------------------------------------------
883 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
885 * Support for kernels:
886 2.6.5-7.286 (SLES 9),
887 2.6.9-55.0.9.EL (RHEL 4),
888 2.6.16.53-0.8 (SLES 10),
889 2.6.18-8.1.14.el5 (RHEL 5),
890 2.6.18.8 vanilla (kernel.org)
891 * Client support for unpatched kernels:
892 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
893 2.6.16 - 2.6.22 vanilla (kernel.org)
894 * Due to recently discovered recovery problems, we do not recommend
895 using patchless RHEL 4 clients with this or any earlier release.
896 * Recommended e2fsprogs version: 1.40.2-cfs1
897 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
900 Frequency : occasional
902 Description: MDS hang or stay in waiting lock
903 Details : If client receive lock with CBPENDING flag ldlm need send lock
904 cancel as separate rpc, to avoid situation when cancel request
905 can't processed due all i/o threads stay in wait lock.
908 Frequency : occasional
910 Description: Do not fail import if osc_interpret_create gets -EAGAIN
911 Details : If osc_interpret_create got -EAGAIN it immediately exits and
912 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
913 oscc_has_objects and see OSC has no objests and call
914 oscc_internal_create to resend create request.
916 Severity : enhancement
918 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
919 Details : Update which_patch & target file for SLES10 latest kernel.
921 Severity : enhancement
923 Description: add --type and --size parameters to lfs find
924 Details : Enhance lfs find by adding filetype and filesize parameters. Also
925 multiple OBDs can now be specified for the --obd option.
927 Severity : enhancement
929 Description: eliminate client locks in face of contention
930 Details : file contention detection and lockless i/o implementation
933 Severity : enhancement
935 Description: Remove client patches from SLES 10 kernel.
936 Details : This causes SLES 10 clients to behave as patchless clients
937 even on a Lustre-patched (server) kernel.
939 Severity : enhancement
941 Description: use i_size_read and i_size_write in 2.6 port
942 Details : replace inode->i_size access with i_size_read/write()
945 Frequency : when removing large files
947 Description: scheduling issue during removal of large Lustre files
948 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
949 It causes scheduling issues when removing large files (17TB in the
955 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
956 Details : Impossible (EPROTO is returned) to access a directory that has a
957 non-default striping and ACLs.
960 Frequency : only on ppc
962 Description: /proc/fs/lustre/devices broken on ppc
963 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
964 We should make sure the type of 'index' is loff_t and then cast
965 explicitly as needed below. Do not assign an explicitly cast
969 Frequency : only for rhel5
971 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
972 Details : Modify the target file & which_kernel.
975 Frequency : if the uninit_groups feature is enabled on ldiskfs
977 Description: e2fsck reports "invalid unused inodes count"
978 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
979 feature and only a single inode is created in a group then the
980 "bg_unused_inodes" count is incorrectly updated. Creating a
981 second inode in that group would update it correctly.
986 Description: buffer overruns could theoretically occur
987 Details : llapi_semantic_traverse() modifies the "path" argument by
988 appending values to the end of the origin string, and a buffer
989 overrun may occur. Adding buffer overrun check in liblustreapi.
991 Severity : enhancement
993 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
996 Frequency : only if filesystem is inconsistent
998 Description: handle "serious error: objid * already exists" more gracefully
999 Details : If LAST_ID value on disk is smaller than the objects existing in
1000 the O/0/d* directories, it indicates disk corruption and causes an
1001 LBUG(). If the object is 0-length, then we should use the existing
1002 object. This will help to avoid a full fsck in most cases.
1004 Severity : enhancement
1006 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1007 Details : Modify vm-tunables-rhel4.patch.
1009 Severity : enhancement
1011 Description: Kernel config for 2.6.18-vanilla.
1012 Details : Modify targets/2.6-vanilla.target.in.
1013 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1014 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1015 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1016 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1019 Frequency : occasional
1021 Description: improve handling recoverable errors
1022 Details : If request processed with error which can be recoverable on server
1023 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: change order of libsysio includes
1041 Details : '#include sysio.h' should always come before '#include xtio.h'
1043 Severity : enhancement
1045 Description: adapt the lustre_config script to support the upgrade case
1046 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1047 server targets to 1.6 in parallel.
1052 Description: To avoid grant space > avaible space when the disk is almost
1053 full. Without this patch you might see the error "grant XXXX >
1054 available" or some LBUG about grant, when the disk is almost
1056 Details : In filter_check_grant, for non_grant cache write, we should
1057 check the left space by if (*left > ungranted + bytes), instead
1058 of (*left > ungranted), because only we are sure the left space
1059 is enough for another "bytes", then the ungrant space should be
1060 increase. In client, we should update cl_avail_grant only there is
1061 OBD_MD_FLGRANT in the reply.
1066 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1067 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1068 described in CVE-2007-4573. This problem could allow untrusted
1069 local users to gain root access.
1072 Frequency : when using O_DIRECT and quotas
1074 Description: Incorrect file ownership on O_DIRECT output files
1075 Details : block usage reported by 'lfs quota' does not take into account
1076 files that have been written with O_DIRECT.
1078 --------------------------------------------------------------------------------
1080 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
1082 * Support for kernels:
1083 2.6.5-7.286 (SLES 9),
1084 2.6.9-55.0.2.EL (RHEL 4),
1085 2.6.16.46-0.14 (SLES 10),
1086 2.6.18-8.1.8.el5 (RHEL 5),
1087 2.6.18.8 vanilla (kernel.org)
1088 * Client support for unpatched kernels:
1089 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1090 2.6.16 - 2.6.21 vanilla (kernel.org)
1091 * Due to recently discovered recovery problems, we do not recommend
1092 using patchless RHEL 4 clients with this or any earlier release.
1093 * Recommended e2fsprogs version: 1.40.2-cfs1
1094 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1098 Description: Fix errors in lfs documentation
1099 Details : Fixes man pages
1101 Severity : enhancement
1103 Description: Adaptive timeouts
1104 Details : RPC timeouts adapt to changing server load and network
1105 conditions to reduce resend attempts and improve recovery time.
1107 Severity : enhancement
1109 Description: llapi_file_create() does not allow some changes
1110 Details : add llapi_file_open() that allows specifying the file creation
1111 mode and open flags, and also returns an open file handle.
1113 Severity : enhancement
1115 Description: df doesn't work properly if diskfs blocksize != 4K
1116 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1119 Frequency : always on directories with default striping set
1121 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1122 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1124 Severity : enhancement
1126 Description: merge and cleanup kernel patches.
1127 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1129 Severity : enhancement
1131 Description: RedHat Update kernel for RHEL5
1132 Details : Kernel config file for RHEL5.
1134 Severity : enhancement
1136 Description: OSS needs mutliple precreate threads
1137 Details : Add ability to start more than one create thread per OSS.
1140 Frequency : only with quota on the root user
1142 Description: mds_obd_create error creating tmp object
1143 Details : When the user sets quota on root, llog will be affected and can't
1144 create files and write files.
1147 Frequency : Always on ia64 patchless client, and possibly others.
1149 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1150 Details : This allows the patchless client to be loaded on architectures
1151 without this export.
1156 Description: disorder of journal start and llog_add cause deadlock.
1157 Details : in llog_origin_connect, journal start should happen before
1158 llog_add keep the same order as other functions to avoid
1161 Severity : enhancement
1163 Description: RedHat Update kernel for RHEL5
1164 Details : Modify the kernel config file more closer RHEL5.
1166 Severity : enhancement
1168 Description: Build failure against Centos5 (RHEL5)
1169 Details : Define PAGE_SIZE when it isn't present.
1172 Frequency : occasionally when using NFS
1174 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1175 Details : with NFS, the anon dentry's parent was set to itself in
1176 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1177 corresponding dentry other than use rec->ur_name.
1179 Severity : enhancement
1181 Description: enable data checksumming by default
1182 Details : enable checksum by default, allow --disable-checksum
1183 configure option and "-o nochecksum" mount option. Checksums
1184 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1187 Frequency : Occasionally with failover
1189 Description: Client eviction due to failover config
1190 Details : after a connection loss, the lustre client should attempt to
1191 reconnect to the last active server first before trying the
1192 other potential connections.
1194 Severity : enhancement
1196 Description: client-side metadata stat-ahead during readdir(directory readahead)
1197 Details : perform client-side metadata stat-ahead when the client detects
1198 readdir and sequential stat of dir entries therein
1203 Description: when mds and osts use different quota unit(32bit and 64bit),
1204 quota will be released repeatly.
1205 Details : void sending multiple quota reqs to mds, which will keep the
1206 status between the reqs.
1209 Frequency : only with liblustre clients on XT3
1211 Description: evictions taking too long
1212 Details : allow llrd to evict clients directly on OSTs
1216 Description: osts not allocated evenly to files
1217 Details : change the condition to increase offset_idx
1219 Severity : enhancement
1221 Description: self-adjustable client's lru lists
1222 Details : use adaptive algorithm for managing client cached locks lru
1223 lists according to current server load, other client's work
1224 pattern, memory activities, etc. Both, server and client
1225 side namespaces provide number of proc tunables for controlling
1230 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
1231 Details : rewrite inherited code (uuid parsing code from ext2 utils and
1232 readdir code from ext3) from scratch preserving functionality.
1236 Description: Only those disconnect error should be returned by rq_status.
1237 Details : In open/enqueue processs, Some errors, which will cause client
1238 disconnected, should be returned by rq_status, while other
1239 errors should still be returned by intent, then mdc or llite will
1242 Severity : enhancement
1244 Description: Tune the kernel for good SCSI performance.
1245 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1246 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1250 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1251 Bugzilla : 13547, 13627
1252 Description: Data corruption for OSTs that are formatted larger than 2TB
1254 Details : When generating the bio request for lustre file writes the
1255 sector number would overflow a temporary variable before being
1256 used for the IO. The data reads correctly from Lustre (which
1257 will overflow in a similar manner) but other file data or
1258 filesystem metadata may be corrupted in some cases.
1262 Description: "lfs find -obd UUID" prints directories
1263 Details : "lfs find -obd UUID" will return all directory names instead
1264 of just file names. It is incorrect because the directories
1265 do not reside on the OSTs.
1268 Bugzilla : 13596, 13614
1269 Description: MDS hang after unclean shutdown of lots of clients
1270 Details : Never resend AST requests.
1274 Frequency : Always, for kernels after 2.6.16
1275 Description: Fix warning idr_remove called for id=.. which is not allocated.
1276 Details : Last kernels save old s_dev before kill super and not allow
1277 to restore from callback - restore it before call kill_anon_super.
1281 Description: Fix error on 'ls .' at the top of the Lustre mount.
1282 Details : Don't revalidate dentry if it is a root dentry.
1284 --------------------------------------------------------------------------------
1286 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
1288 * Support for kernels:
1289 2.6.5-7.286 (SLES 9),
1290 2.6.9-55.0.2.EL (RHEL 4),
1291 2.6.16.46-0.14 (SLES 10),
1292 2.6.18-8.1.8.el5 (RHEL 5),
1293 2.6.18.8 vanilla (kernel.org)
1294 * Client support for unpatched kernels:
1295 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1296 2.6.16 - 2.6.22 vanilla (kernel.org)
1297 * Due to recently discovered recovery problems, we do not recommend
1298 using patchless RHEL 4 clients with this or any earlier release.
1299 * Recommended e2fsprogs version: 1.39.cfs8
1300 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1305 Description: block reactivating mgc import until all deactivates complete
1306 Details : Fix race when failing back MDT/MGS to itself (testing)
1308 Severity : enhancement
1310 Description: lfs setstripe enhancement
1311 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1314 Frequency : mds/oss recovery
1316 Description: llog ctxt is refrenced after it has been freed.
1317 Details : llog ctxt refcount was added to avoide the race between ctxt free
1318 and llog recovery process. Each llog user must hold ctxt refcount
1319 before it access the llog. And the llog ctxt can only be freed
1320 when its refcount is zero.
1322 Severity : enhancement
1324 Description: randomly memory allocation failure util
1325 Details : Make lustre randomly failed allocating memory for testing purpose.
1328 Frequency : only for SLES10
1330 Description: Update kernel patch for SLES10 SP1
1331 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
1333 Severity : enhancement
1335 Description: omit set fsid for export NFS
1336 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
1337 as FS_REQUIRES_DEV for avoid problems with generate fsid.
1340 Frequency : after network failures
1342 Description: Add sync option to mount_lustre.c
1343 Details : Client loses data written to lustre after a network interruption.
1345 Severity : enhancement
1347 Description: Error message improvement.
1348 Details : Merging of two LCONSOLE_ERROR_MSG into one.
1350 Severity : enhancement
1352 Description: don't use GFP_* in generic Lustre code.
1353 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1355 Severity : enhancement
1357 Description: obdclass is limited by single OBD_ALLOC(idarray)
1358 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1360 Severity : enhancement
1362 Description: updated patchess for new RHEL4 kernel
1363 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
1365 Severity : enhancement
1367 Description: warnings with build patchless client with vanila 2.6.19 and up
1368 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1369 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1371 Severity : enhancement
1373 Description: O_DIRECT bypasses client statistics.
1374 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1375 read_ahead_stats were updated. Copied stats section from
1376 osc_send_oap_rpc() into async_internal().
1379 Frequency : only for Cray XT3
1381 Description: peer credits not enough on many OST per OSS systems.
1382 Details : Use new lnet way to add credits as we need those for pings and ASTs
1385 Frequency : only with liblustre
1387 Description: Liblustre is not releasing flock locks on file close.
1388 Details : Release flock locks on file close.
1391 Frequency : only for RHEL4
1393 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
1394 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
1395 Update target file and kernel config.
1399 Description: lustre support for RHEL5
1400 Details : Add support for RHEL5.
1405 Description: ASSERTION(export != NULL) failed in target_handle_connect
1406 Details : Assetion hit is result of rare race between disconnect and connet
1407 to same nid. target_handle_connect found old connect cockie and
1408 tried to reconnect, but can't find export for this cockie.
1413 Description: umount blocks forever on error
1414 Details : In result of wrong using obd_no_recov and obd_force flags client
1415 can hand if cancel or some other requests is lost.
1418 Frequency : Only for SLES
1420 Description: sanity_quota fail test_1
1421 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1422 file, which makes TSTID[2] inunique.
1424 Severity : enhancement
1426 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1427 Details : Update target/ChangeLog/which_patch .
1429 Severity : enhancement
1431 Description: jbd statistics
1432 Details : Port older jbd statistics patch for sles10
1437 Description: lvbo_init failed for resource with missing objects.
1438 Details : Fix returning error if we do stat for file with missing/corrupted
1439 objects and i_size set to all sum of size of all avaible objects.
1440 if we to truncate/write to missing object - it is recreated.
1445 Description: Oops in read and write path when failing to allocate lock.
1446 Details : Check if lock allocation failed and return error back.
1449 Frequency : When flocks are used.
1451 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1452 Details : Flock locks might destroy just granted lock if it could be merged
1453 with another existing flock, this is done in completion handler,
1454 so teach ldlm_cli_enquque_fini that this is a valid case for
1460 Description: reply_lock_interpret crash due to race with it and lock cancel.
1461 Details : Do not replay locks that are being cancelled. Do not reference
1462 locks by their address during replay, just by their handle.
1465 Frequency : only with deactivated OSTs
1467 Description: lstripe command fails for valid OST index
1468 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
1469 instead of lov->desc.ld_active_tgt_count.
1471 Severity : enhancement
1473 Description: Build failure against Centos5 (RHEL5)
1474 Details : Use getpagesize() instead of PAGE_SIZE.
1476 --------------------------------------------------------------------------------
1478 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
1480 * Support for kernels:
1481 2.6.5-7.283 (SLES 9),
1482 2.6.9-55.EL (RHEL 4),
1483 2.6.16.46-0.14 (SLES 10),
1484 2.6.18.8 vanilla (kernel.org)
1485 * Client support for unpatched kernels:
1486 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1487 2.6.16 - 2.6.22 vanilla (kernel.org)
1488 * Due to recently discovered recovery problems, we do not recommend
1489 using patchless RHEL 4 clients with this or any earlier release.
1490 * Recommended e2fsprogs version: 1.39.cfs8
1491 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1492 * Starting with this release, the ldiskfs backing filesystem required
1493 by Lustre is now in its own package, lustre-ldiskfs. This package
1494 should be installed. It is versioned separately from Lustre and
1495 may be released separately in future.
1500 Description: server LBUG when shutting down
1501 Details : Block umount forever until the mount refcount is zero rather
1502 than giving up after an arbitrary timeout.
1504 Severity : enhancement
1506 Description: add optional extra BUILD_VERSION info
1507 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1508 to override the lustre version.
1511 Frequency : 2.6.18 servers only
1513 Description: ll_kern_mount() doesn't release the module reference
1514 Details : The ldiskfs module reference count never drops down to 0
1515 because ll_kern_mount() doesn't release the module reference.
1520 Description: server LBUG when using old ost_num_threads parameter
1521 Details : Accept the old ost_num_threads parameter but warn that it
1522 is deprecated, and fix an off-by-one error that caused an LBUG.
1527 Description: Transient SCSI error results in persistent IO issue
1528 Details : iobuf->dr_error is not reinitialized to 0 between two
1532 Frequency : sometimes when underlying device returns I/O errors
1534 Description: OSTs not going read-only during write failures
1535 Details : OSTs are not remounted read-only when the journal commit threads
1536 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1537 instead of the ext3 wrappers.
1540 Frequency : only with 10000 clients or more
1542 Description: poor connect scaling with increasing client count
1543 Details : Don't run filter_grant_sanity_check for more than 100 exports
1544 to improve scaling for large numbers of clients.
1547 Frequency : SLES10 only
1549 Description: sanity-quota.sh quotacheck failed: rc = -22
1550 Details : Quotas cannot be enabled on SLES10.
1553 Frequency : liblustre clients only
1555 Description: getdirentries does not give error when run on compute nodes
1556 Details : getdirentries does not fail when the size specified as an argument
1557 is too small to contain at least one entry
1559 Severity : enhancement
1561 Description: Add LNET router traceability for debug purposes
1562 Details : If a checksum failure occurs with a router as part of the
1563 IO path, the NID of the last router that forwarded the bulk data
1564 is printed so it can be identified.
1569 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1570 Details : Due to a race condition, liblustre clients were occasionally
1571 evicted incorrectly.
1573 Severity : enhancement
1575 Description: lfs setstripe use optional parameters instead of postional
1578 Severity : enhancement
1580 Description: Nanosecond timestamp support for ldiskfs
1581 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1582 resolution timestamps. There is not yet support for this at
1583 the Lustre filesystem level.
1586 Frequency : during server recovery
1588 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1589 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1593 Frequency : nfs export on patchless client
1595 Description: connectathon hang when test nfs export over patchless client
1596 Details : Disconnected dentry cannot be found with lookup, so we do not need
1597 to unhash it or make it invalid
1600 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1601 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1602 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1607 Description: open req refcounting wrong on reconnect
1608 Details : If reconnect happened between getting open reply from server and
1609 call to mdc_set_replay_data in ll_file_open, we will schedule
1610 replay for unreferenced request that we are about to free.
1611 Subsequent close will crash in variety of ways.
1612 Check that request is still eligible for replay in
1613 mdc_set_replay_data().
1618 Description: disable writes to filesystem when reading health_check file
1619 Details : the default for reading the health_check proc file has changed
1620 to NOT do a journal transaction and write to disk, because this
1621 can cause reads of the /proc file to hang and block HA state
1622 checking on a healthy but otherwise heavily loaded system. It
1623 is possible to return to the previous behaviour during configure
1624 with --enable-health-write.
1626 Severity : enhancement
1628 Description: 64-bit inode version
1629 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1630 to the inode. This will be required for version-based recovery.
1635 Description: MDS fails to start if a duplicate client export is detected
1636 Details : in some rare cases it was possible for a client to connect to
1637 an MDS multiple times. Upon recovery the MDS would detect this
1638 and fail during startup. Handle this more gracefully.
1640 Severity : enhancement
1642 Description: Add -o localflock option to simulate old noflock behaviour.
1643 Details : This will achieve local-only flock/fcntl locks coherentness.
1648 Description: log_commit_thread vs filter_destroy race leads to crash
1649 Details : Take import reference before releasing llog record semaphore
1654 Description: Wrong request locking in request set processing
1655 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1656 imp_delayed_list, in this place should be used imp_lock.
1659 Frequency : when reconnecting
1661 Description: Grant leak when OSC reconnect to OST
1662 Details : When osc reconnect ost, OST(filter) should check whether it
1663 should grant more space to client by comparing fed_grant and
1664 cl_avail_grant, and return the granted space to client instead
1665 of "new granted" space, because client will call osc_init_grant
1666 to update the client grant space info.
1669 Frequency : when client reconnects to OST
1671 Description: Grant leak when OSC does a resend and replays bulk write
1672 Details : When osc reconnect to OST, OST(filter) should clear grant info of
1673 bulk write request, because the grant info while be sync between
1674 OSC and OST when reconnect, and we should ignore the grant info
1675 these of resend/replay write req.
1680 Description: Grant space more than avaiable space sometimes.
1681 Details : When then OST is about to be full, if two bulk writing from
1682 different clients came to OST. Accord the avaliable space of the
1683 OST, the first req should be permitted, and the second one
1684 should be denied by ENOSPC. But if the seconde arrived before
1685 the first one is commited. The OST might wrongly permit second
1686 writing, which will cause grant space > avaiable space.
1689 Frequency : when client is evicted
1691 Description: Grant might be wrongly erased when osc is evicted by OST
1692 Details : when the import is evicted by server, it will fork another
1693 thread ptlrpc_invalidate_import_thread to invalidate the
1694 import, where the grant will be set to 0. While the original
1695 thread will update the grant it got when connecting. So if
1696 the former happened latter, the grant will be wrongly errased
1697 because of this race.
1702 Description: Checking Stale with correct fid
1703 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1704 to check whether it is stale, because sometimes, we want the
1705 enqueue happened anyway, and op_data.fid2 will not be initialized.
1707 Severity : enhancement
1709 Description: update patchless client
1710 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1713 Frequency : only with 2.4 kernel
1715 Description: random memory corruption
1716 Details : size of struct ll_inode_info is to big for union inode.u and this
1717 can be cause of random memory corruption.
1722 Description: Memory leak in recovery
1723 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1724 It should also check obd_fail before fsfilt_start, otherwise if
1725 fsfilt_start return -EROFS,(failover mds during mds recovery).
1726 then the req will return with repmsg->transno = 0 and rc = EROFS.
1727 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1728 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1729 be freed no matter whether fsfilt_commit success or not.
1732 Frequency : only with huge numbers of clients
1734 Description: Prevents from taking the superblock lock in llap_from_page for
1736 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1737 for llap_from_page call in ll_removepage() prevents from taking
1738 the superblock lock for a soon died page.
1743 Description: Not check open intent error before release open handle
1744 Details : in some rare cases, the open intent error is not checked before
1745 release open handle, which may cause
1746 ASSERTION(open_req->rq_transno != 0), because it tries to release
1747 the failed open handle.
1752 Description: Set cat log bitmap only after create log success.
1753 Details : in some rare cases, the cat log bitmap is set too early. and it
1754 should be set only after create log success.
1758 Description: Accessing a block bevice can re-enable I/O when Lustre is
1759 tearing down a device.
1760 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1765 Description: service threads may hog cpus when there are a lot of requests
1766 Details : Insert cond_resched to give other threads a chance to use some CPU
1771 Description: the cat log was not initialized in recovery
1772 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1773 unlink log on mds will not be initialized until mds post
1774 recovery. And also in mds post recovery, the unlink log will
1775 initialization will be done asynchronausly, so there will be race
1776 between add unlink log and unlink log initialization.
1780 Description: brw_stats were being printed incorrectly
1781 Details : brw_stats were being printed as log2 but all of them were not
1782 recorded as log2. Also remove some code duplication arising from
1783 filter_tally_{read,write}.
1787 Frequency : rare, only in recovery.
1788 Description: ASSERTION(req->rq_type != LI_POISON) failed
1789 Details : imp_lock should be held while iterating over imp_sending_list for
1790 prevent destroy request after get timeout in ptlrpc_queue_wait.
1794 Description: replay-single.sh test 52 fails
1795 Details : A lock's skiplist need to be cleanup when it being unlinked
1796 from its resource list.
1801 Description: Short directio read returns full requested size rather than
1803 Details : Direct I/O operations should return actual amount of bytes
1804 transferred rather than requested size.
1806 Severity : enhancement
1808 Description: metadata RPC reduction (e.g. for rm performance)
1809 Details : decrease the amount of synchronous RPC between clients and servers
1810 by canceling conflicing lock before the operation on the client
1811 and packing thier handles into the main operation RPC to server.
1813 Severity : enhancement
1815 Description: add #ifdef HAVE_KERNEL_CONFIG_H
1816 Details : kernels from 2.6.19 not need include linux/config.h, but add
1817 include linux/autoconf.h in commpiler command line.
1819 Severity : enhancement
1821 Description: patchless client support for 2.6.22 kernel
1822 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
1823 removed. In this case we need drop using os depended interface to
1824 kmem_cache and use cfs_mem_cache API.
1829 Description: fix mal-formatted messages
1830 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
1834 Frequency : always in liblustre
1835 Description: wrong IS_ERR implementation in liblustre.h
1836 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
1841 Description: Correct condition for output debug message.
1842 Details : inode i_nlink equal zero is not enough for output message about
1843 disk corruption, i_ctime and i_mode should be also checked.
1847 Frequency : always in patchless client
1848 Description: add configure check for truncate_complete_page
1849 Details : improve checks for exported symbols. This allow run check without
1850 sources, but with Module.symvers shipped with kernel distribution.
1851 add check for truncate_complete_page used by patchless client.
1855 Description: sanity.sh test_77h fails with "test_77h file compare failed"
1856 Details : test_77h uses a file which was messed by other test case.
1860 Description: Not Check whether lov_tgts is NULL in some lov functions
1861 Details : Checking whether lov_tgts is NULL in some functions.
1865 Description: replace obdo_alloc() with OBDO_ALLOC macro
1866 Details : nothing special is done in obdo_alloc() function, and for
1867 debugging purpose, it needs to be replaced with macros.
1871 Description: bad return value and errno from fcntl call
1872 Details : In liblustre API, errno should be a negative value if error
1877 Description: ptlrpc_check_set() LBUG
1878 Details : In case of positive reply from server and failed client bulk
1879 callback after bulk transfer shouldn't LBUG, but process this
1880 request as erroneous.
1882 Severity : enhancement
1884 Description: VFS operations stats tool.
1885 Details : Tool which collects stats by tracking value written in pid,
1886 ppid, gid and uses llstat to generate output to plot graph using
1888 Updated lustre/utils/Makefile.am
1889 Added lustre/utils/ltrack_stats.c
1891 Severity : enhancement
1893 Description: 2.6.18 server support (lustre 1.6.1)
1894 Details : Support for 2.6.18 kernels on the server side.
1899 Description: ASSERTION(imp->imp_conn_current) failed
1900 Details : an assertion failure is hit if a client node boots and attempts to
1901 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
1904 Frequency : only for i686
1906 Description: 1.4.11 RC1 build fails for RHEL 4, i686
1907 Details : Fixed config variable for build.
1912 Description: Updated patchess for new RHEL4 kernel
1913 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
1914 Updated patch jbd-stats-2.6.9.patch
1915 Updated patch qsnet-rhel4-2.6.patch
1916 Updated patch quota-deadlock-on-pagelock-core.patch
1917 Updated patch vfs_intent-2.6-rhel4.patch
1918 Updated patch vfs_races-2.6-rhel4.patch
1919 Updated series file 2.6-rhel4-titech.series
1920 Updated series file 2.6-rhel4.series
1921 Updated kernel config files
1926 Description: lquota slave complains LBUG when reconnecting with mds
1928 Details : quota slave depends qctxt->lqc_import to send its quota request.
1929 This pointer will be invalid if mds did failover or broke its
1930 connect to osts, which leads to LBUG.
1933 Frequency : when qunit size is too small(less than 20M)
1935 Description: write is stopped by improper -EDQUOT
1936 Details : If the master is busy and qunit size is small enough(let's say 1M),
1937 the slave can not get quota from master on time, which will lead
1938 slave to trigger a -EQUOTA to client.
1943 Description: Deadlock during metadata tests
1944 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
1945 called with the per-dentry lock held.
1948 Frequency : SLES9 only
1950 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
1951 Details : set CONFIG_NR_CPUS to 128 instead of 8.
1953 Severity : enhancement
1955 Description: remove fs_prep_san_write operation and related patches
1956 Details : remove the ext3-san-jdike patches which are no longer useful.
1961 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
1962 Details : a disk corruption can cause the mballoc code to assert on a
1963 double free or other extent corruptions. Handle these with
1964 ext3_error() instead of assertions.
1967 Frequency : only with mballoc3 code and deep extent trees
1969 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
1970 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
1971 can cause the filesystem to be remounted read-only.
1976 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
1977 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
1978 few external users of this ioctl number is lfsck's call to
1979 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
1980 Add the old ioctl number to the handler so both old and new
1985 Description: parallel lock callbacks
1986 Details : Instead of sending blocking and completion callbacks as separated
1987 requests, adding them to a set and sending in parallel.
1991 Description: Disable most debugging by default
1992 Details : To improve performance, disable most logging (for debug purposes)
1993 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
1994 default, and HA includes fewer messages.
1998 Frequency : only run on patchless client.
1999 Description: use do_facet on sanity.sh for test handling recoverables errors
2000 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
2004 Description: hash tables for lists of nids, connections and uuids
2005 Details : Hash tables noticeably help when a lot of clients connect to a
2006 server, to faster identify duplicate connections or reconnects,
2007 also to faster find export to evict in manual eviction case.
2011 Description: Sometimes, when the server evict a client, and the client will
2012 not be evicted as soon as possible.
2013 Details : In enqueue req, the error was returned by intent, instead of
2014 rq_status which make ptlrpc layer not detect this error, and
2015 does not evict the client. So enqueue error should be returned
2019 Frequency : only at startup
2021 Description: Delay client connections to MDT until fist MDT->OST connect
2022 Details : If a client tried to create a new file before the MDT had
2023 connected to any OSTs, the create would return EIO. Now
2024 the client will simply block until the MDT connects to the
2025 first OST and the create can succeed.
2028 Frequency : only for SLES9
2030 Description: Routinely utilize latest Quadrics drivers in CFS releases
2031 Details : Update patch qsnet-suse-2.6.patch.
2034 Frequency : only for sles10
2036 Description: Update patches for SLES 10 SP1 kernel.
2037 Details : Update the patch vfs_intent-2.6-sles10.patch.
2042 Description: Routinely utilize latest Quadrics drivers in CFS releases
2043 Details : Update patch qsnet-rhel4-2.6.patch.
2046 Frequency : at statup only
2048 Description: mds_lov_synchronize race leads to various problems
2049 Details : simultaneous MDT->OST connections at startup can cause the
2050 sync to abort, leaving the OSC in a bad state.
2055 Description: Using wrong pointer in osc_brw_prep_request
2056 Details : Access to array[-1] can produce panic if kernel compiled with
2057 CONFIG_PAGE_ALLOC enabled
2059 Severity : enhancement
2061 Description: Async OSC create to avoid the blocking unnecessarily.
2062 Details : If a OST has no remain object, system will block on the creating
2063 when need to create a new object on this OST. Now, ways use
2064 pre-created objects when available, instead of blocking on an
2065 empty osc while others are not empty. If we must block, we block
2066 for the shortest possible period of time.
2070 Frequency : only in recovery
2071 Description: Mark OST as early accessible if his start SYNC.
2072 Details : osc_precreate return flag early accessible if oscc marked as
2073 OSCC_FLAG_SYNC_IN_PROGRESS.
2078 Description: Sometimes precreate code can triger create object on wrong ost
2079 Details : Wrong protected or not not restored variables aftre precreate loop
2080 can produce creation object on wrong ost.
2083 Frequency : oss recovery
2085 Description: llog_commit_thread cleanup should sync with llog_commit_thread
2087 Details : llog_commit_thread_count should be synced between llog_commit
2088 start and cleanup, so new llog_commit thread should not be started
2089 when llog_commit threads being stopped to avoid accessing some
2092 Severity : enhancement
2094 Description: Add printing inode info into message about error in writepage.
2096 --------------------------------------------------------------------------------
2098 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2103 Frequency : on some architectures
2105 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2106 Details : Uninitialized flags sometimes cause configuration commands to
2110 Frequency : patchless clients only
2112 Description: missing __iget() symbol export
2113 Details : The __iget() symbol export is missing. To avoid the need for
2114 this on patchless clients the deathrow inode reaper is turned
2115 off, and we depend on the VM to clean up old inodes. This
2116 dependency was during via the fix for bug 12181.
2121 Description: sanity.sh fail: test_52b
2122 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
2123 incorrect inode's flags to client.
2125 --------------------------------------------------------------------------------
2127 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2129 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2130 INTEROPERATE with older versions automatically. In many cases a
2131 special upgrade step is needed. Please read the
2132 user documentation before upgrading any part of a 1.4.x system.
2133 * WARNING: Lustre configuration and startup changes are required with
2134 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2136 * Support for kernels:
2137 2.4.21-47.0.1.EL (RHEL 3),
2138 2.6.5-7.283 (SLES 9),
2139 2.6.9-42.0.10.EL (RHEL 4),
2140 2.6.12.6 vanilla (kernel.org),
2141 2.6.16.27-0.9 (SLES10)
2142 * Client support for unpatched kernels:
2143 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
2144 2.6.16 - 2.6.19 vanilla (kernel.org),
2145 2.6.9-42.0.8EL (RHEL 4)
2146 * Recommended e2fsprogs version: 1.39.cfs6
2147 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2150 Severity : enhancement
2152 Description: Async OSC create to avoid the blocking unnecessarily.
2153 Details : If a OST has no remain object, system will block on the creating
2154 when need to create a new object on this OST. Now, ways use
2155 pre-created objects when available, instead of blocking on an
2156 empty osc while others are not empty. If we must block, we block
2157 for the shortest possible period of time.
2159 Severity : enhancement
2161 Description: MountConf
2162 Details : Lustre configuration is now managed via mkfs and mount
2163 commands instead of lmc and lconf. New obd types (MGS, MGC)
2164 are added for dynamic configuration management. See
2165 https://mail.clusterfs.com/wikis/lustre/MountConf for
2168 Severity : enhancement
2170 Description: dynamic OST addition
2171 Details : OSTs can now be added to a live filesystem
2173 Severity : enhancement
2175 Description: startup order invariance
2176 Details : MDTs and OSTs can be started in any order. Clients only
2177 require the MDT to complete startup.
2179 Severity : enhancement
2181 Description: parallel, asynchronous orphan cleanup
2182 Details : orphan cleanup is now performed in separate threads for each
2183 OST, allowing parallel non-blocking operation.
2185 Severity : enhancement
2187 Description: optimized stripe assignment
2188 Details : stripe assignments are now made based on ost space available,
2189 ost previous usage, and OSS previous usage, in order to try
2190 to optimize storage space and networking resources.
2192 Severity : enhancement
2194 Description: Permanently set tunables
2195 Details : All writable /proc/fs/lustre tunables can now be permanently
2196 set on a per-server basis, at mkfs time or on a live
2199 Severity : enhancement
2201 Description: Lustre message v2
2202 Details : Add lustre message format v2.
2204 Severity : enhancement
2206 Description: client OST exclusion list
2207 Details : Clients can be started with a list of OSTs that should be
2208 declared "inactive" for known non-responsive OSTs.
2212 Description: ENOENT returned for valid filehandle during dbench.
2213 Details : Check if a directory has children when invalidating dentries
2214 associated with an inode during lock cancellation. This fixes
2215 an incorrect ENOENT sometimes seen for valid filehandles during
2216 testing with dbench.
2219 Frequency : SFS test only (otherwise harmless)
2221 Description: SPEC SFS validation failure on NFS v2 over lustre.
2222 Details : Changes the blocksize for regular files to be 2x RPC size,
2223 and not depend on stripe size.
2225 Severity : enhancement
2227 Description: fine-grained SMP locking inside DLM
2228 Details : Improve DLM performance on SMP systems by removing the single
2229 per-namespace lock and replace it with per-resource locks.
2231 Severity : enhancement
2233 Description: don't hold multiple extent locks at one time
2234 Details : To avoid client eviction during large writes, locks are not
2235 held on multiple stripes at one time or for very large writes.
2236 Otherwise, clients can block waiting for a lock on a failed OST
2237 while holding locks on other OSTs and be evicted.
2239 Severity : enhancement
2241 Description: Multiple MD RPCs in flight.
2242 Details : Further unserialise some read-only MDT RPCs - learn about intents.
2243 To avoid overly-overloading MDT, introduce a limit on number of
2244 MDT RPCs in flight for a single client and add /proc controls
2245 to adjust this limit.
2247 Severity : enhancement
2249 Description: client read/write statistics
2250 Details : Add client read/write call usage stats for performance
2251 analysis of user processes.
2252 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2253 file access. extents_stats shows chunk size distribution.
2254 extents_stats_per_process show chunk size distribution per
2257 Severity : enhancement
2259 Description: per-client statistics on server
2260 Details : Add ldlm and operations statistics for each client in
2261 /proc/fs/lustre/mds|obdfilter/*/exports/
2263 Severity : enhancement
2265 Description: improved MDT statistics
2266 Details : Add detailed MDT operations statistics in
2267 /proc/fs/lustre/mds/*/stats
2269 Severity : enhancement
2271 Description: VFS operations stats
2272 Details : Add client VFS call stats, trackable by pid, ppid, or gid
2273 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
2278 Description: Fix client-side osc byte counters
2279 Details : The osc read/write byte counters in
2280 /proc/fs/lustre/osc/*/stats are now working
2283 Frequency : always as root on SLES
2285 Description: Failure of copying files with lustre special EAs.
2286 Details : Client side always return success for setxattr call for lustre
2287 special xattr (currently only "trusted.lov").
2292 Description: Refcount LNET uuids
2293 Details : The global LNET uuid list grew linearly with every startup;
2294 refcount repeated list entries instead of always adding to
2297 Severity : enhancement
2299 Description: Dynamic service threads
2300 Details : Within a small range, start extra service threads
2301 automatically when the request queue builds up.
2304 Frequency : mixed-endian client/server environments
2306 Description: mixed-endian crashes
2307 Details : The new msg_v2 system had some failures in mixed-endian
2310 Severity : enhancement
2312 Description: Easy OST removal
2313 Details : OSTs can be permanently deactivated with e.g. 'lctl
2314 conf_param lustre-OST0001.osc.active=0'
2316 Severity : enhancement
2318 Description: MGS proc entries
2319 Details : Added basic proc entries for the MGS showing what filesystems
2322 Severity : enhancement
2324 Description: provide MGS failover
2325 Details : Added config lock reacquisition after MGS server failover.
2327 Severity : enhancement
2329 Description: add Linux 2.4 support
2330 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2334 Description: a large application tries to do I/O to the same resource and dies
2335 in the middle of it.
2336 Details : Check the req->rq_arrival time after the call to
2337 ost_brw_lock_get(), but before we do anything about
2338 processing it & sending the BULK transfer request. This
2339 should help move old stale pending locks off the queue as
2340 quickly as obd_timeout.
2343 Frequency : when an incorrect nid is specified during startup
2345 Description: ptlrpc connect to non-existant node causes kernel crash
2346 Details : LNET can't be re-entered from an event callback, which
2347 happened when we expire a message after the export has been
2348 cleaned up. Instead, hand the zombie cleanup off to another
2351 Severity : enhancement
2353 Description: plain/inodebits lock performance improvement
2354 Details : Grouping plain/inodebits in granted list by their request modes
2355 and bits policy, thus improving the performance of search through
2359 Frequency : only if OST filesystem is corrupted
2361 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2362 Details : for a short time RPCs with bulk IO are in the replay list,
2363 but replay of bulk IOs is unimplemented. If the OST filesystem
2364 is corrupted due to disk cache incoherency and then replay is
2365 started it is possible to trip an assertion. Avoid putting
2366 committed RPCs into the replay list at all to avoid this issue.
2369 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2371 Description: System hang on startup
2372 Details : This bug allowed the liblustre (e.g. catamount) client to
2373 return to the app before handling all startup RPCs. This
2374 could leave the node unresponsive to lustre network traffic
2375 and manifested as a server ptllnd timeout.
2377 Severity : enhancement
2379 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2380 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2381 boolean to control whether to print peer debug info when a
2382 client's RPC times out.
2385 Frequency : only for kernels with patches from Lustre below 1.4.3
2387 Description: Remove old rdonly API
2388 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2391 Frequency : only for devices with external journals
2393 Description: Set external device read-only also
2394 Details : During a commanded failover stop, we set the disk device
2395 read-only while the server shuts down. We now also set any
2396 external journal device read-only at the same time.
2399 Frequency : when upgrading from 1.4 while trying to change parameters
2401 Description: The wrong (new) MDC name was used when setting parameters for
2402 upgraded MDT's. Also allows changing of OSC (and MDC)
2403 parameters if --writeconf is specified at tunefs upgrade time.
2406 Frequency : when setting specific ost indicies
2408 Description: QOS code breaks on skipped indicies
2409 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2410 created with --index need not be sequential.
2412 Severity : enhancement
2414 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2415 Details : The uninit_groups feature works in conjunction with the kernel
2416 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2417 up the pass1 processing of e2fsck. This is a read-only feature
2418 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2419 mount filesystems that have had this feature enabled.
2421 Severity : enhancement
2423 Description: Improve multi-block allocation algorithm to avoid fragmentation
2424 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2425 allocation locality and avoid filesystem fragmentation.
2427 ------------------------------------------------------------------------------
2429 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2431 * Support for kernels:
2432 2.4.21-47.0.1.EL (RHEL 3)
2433 2.6.5-7.283 (SLES 9)
2434 2.6.9-42.0.10.EL (RHEL 4)
2435 2.6.12.6 vanilla (kernel.org)
2436 2.6.16.27-0.9 (SLES 10)
2437 * Recommended e2fsprogs version: 1.39.cfs5
2439 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2443 Frequency : occasional, depends on client load and configuration
2444 Bugzilla : 12181, 12203
2445 Description: data loss for recently-modified files
2447 Details : In some cases it is possible that recently written or created
2448 files may not be written to disk in a timely manner (this should
2449 normally be within 30s unless client IO load is very high).
2450 The problem appears as zero-length files or files that are a
2451 multiple of 1MB in size after a client crash or client eviction
2452 that are missing data at the end of the file.
2454 This problem is more likely to be hit on clients where files are
2455 repeatedly created and unlinked in the same directory, clients
2456 have a large amount of RAM, have many CPUs, the filesystem has
2457 many OSTs, the clients are rebooted frequently, and/or the files
2458 are not accessed by other nodes after being written.
2460 The presence of the problem can be detected by looking at
2461 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2462 smaller than the second (nr_unused) then dirty files will not
2463 be flushed automatically to disk. "sync; sleep 10" should be
2464 run several times on the node before unmounting it to update
2465 Lustre (this is also safe to run on nodes without this problem).
2467 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2468 that can cause this same problem, so customers using that kernel
2469 also need to update the kernel in addition to Lustre. In order
2470 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2472 It is normal that files written just before a client crash (less
2473 than 30s) may not yet have been flushed to disk, even for local
2477 Frequency : frequent on thin XT3 nodes
2479 Description: UUID collision on thin XT3 Linux nodes
2480 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2481 randomly, since we relied on an insufficiently-seeded PRNG.
2486 Description: OSS hangs after "All ost request buffers busy"
2487 Details : A deadlock between quota and journal operations caused OSS
2488 hangs after printing "All ost request buffers busy."
2491 Frequency : always on liblustre builds
2493 Description: Cleanup compiler warnings on liblustre
2496 Frequency : always on liblustre builds on XT3
2498 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2502 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2503 > 32000 subdirectories
2504 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2505 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2506 subdirectory count crosses 32000. This will aid e2fsck to
2507 correctly handle more than 32000 subdirectories.
2510 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2512 Description: System hang on startup
2513 Details : This bug allowed the liblustre (e.g. catamount) client to
2514 return to the app before handling all startup RPCs. This
2515 could leave the node unresponsive to lustre network traffic
2516 and manifested as a server ptllnd timeout.
2518 Severity : enhancement
2520 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2521 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2522 boolean to control whether to print peer debug info when a
2523 client's RPC times out.
2528 Description: make O_SYNC working on 2.6 kernels
2529 Details : 2.6 kernels use different method for mark pages for write,
2530 so need add a code to lustre for O_SYNC work.
2535 Description: Failure to close file and release space on NFS
2536 Details : Put inode details into lock acquired in ll_intent_file_open.
2537 Use mdc_intent_lock in ll_intent_open to properly
2538 detect all kind of errors unhandled by mdc_enqueue.
2543 Description: proc file read during shutdown sometimes raced obd removal,
2545 Details : Add lock to prevent obd access after proc file removal.
2548 Frequency : Only for files larger than 4GB on 32-bit clients.
2550 Description: improperly doing page alignment of locks
2551 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2552 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2557 Description: RPCs being resent when they shouldn't be
2558 Details : Some RPCs that should not be resent are being resent. This
2559 can cause inconsistencies in the RPC state machine. Do not
2560 resend such requests.
2563 Frequency : rare, only with NFS export
2565 Description: Crash on NFS re-export node
2566 Details : under very unusual load conditions an assertion is hit in
2567 ll_intent_file_open()
2570 Frequency : only if OST filesystem is corrupted
2572 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2573 Details : for a short time RPCs with bulk IO are in the replay list,
2574 but replay of bulk IOs is unimplemented. If the OST filesystem
2575 is corrupted due to disk cache incoherency and then replay is
2576 started it is possible to trip an assertion. Avoid putting
2577 committed RPCs into the replay list at all to avoid this issue.
2582 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2583 Details : Large single O_DIRECT read and write calls can fail to allocate
2584 a sufficiently large buffer to process the request. In case of
2585 allocation failure the allocation is retried with a smaller
2586 buffer and broken into smaller requests.
2588 Severity : enhancement
2590 Description: Add -o localflock option to simulate old noflock behaviour.
2591 Details : This will achieve local-only flock/fcntl locks coherentness.
2596 Description: versioning check is incomplete
2597 Details : Checking the version difference of client vs. server, report
2598 error if the gap is too big.
2603 Description: add support PG_writeback bit
2604 Details : add support for PG_writeback bit for Lustre, for more carefull
2605 work with page cache in 2.6 kernel. This also fix some deadlocks
2606 and remove hack for work O_SYNC with 2.6 kernel.
2608 Severity : enhancement
2610 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2611 Details : The uninit_groups feature works in conjunction with the kernel
2612 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2613 up the pass1 processing of e2fsck. This is a read-only feature
2614 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2615 mount filesystems that have had this feature enabled.
2617 Severity : enhancement
2619 Description: Improve multi-block allocation algorithm to avoid fragmentation
2620 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2621 allocation locality and avoid filesystem fragmentation.
2623 ------------------------------------------------------------------------------
2625 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2627 * Support for kernels:
2628 2.6.9-42.0.3.EL (RHEL 4)
2629 2.6.5-7.276 (SLES 9)
2630 2.4.21-47.0.1.EL (RHEL 3)
2631 2.6.12.6 vanilla (kernel.org)
2632 2.6.16.21-0.8 (SLES10)
2633 * Recommended e2fsprogs version: 1.39.cfs2-0
2635 * The backwards-compatible /proc/sys/portals symlink has been removed
2636 in this release. Before upgrading, please ensure that you change
2637 any configuration scripts or /etc/sysctl.conf files that access
2638 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2639 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2640 in advance of the upgrade on any system running Lustre 1.4.6 or
2641 newer, since /proc/sys/lnet was added in that version.
2642 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2646 Frequency : only when quota is used
2648 Description: avoid scanning export list for quota master
2649 Details : Change the algorithms to avoid scanning export list in order
2650 to improve the efficiency.
2653 Frequency : MDS failover only, very rarely
2655 Description: "went back in time" messages on mds failover
2656 Details : The greatest transno may be lost when the current operation
2657 finishes with an error (transno==0) and the client's last_rcvd
2658 record is over-written. Save the greatest transno in the
2659 mds_last_transno for this case.
2662 Frequency : always for specific kernels and striping counts
2664 Description: client may get "Matching packet too big" without ACL support
2665 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2666 when trying to access files in certain configurations. The
2667 clients should in fact be denied when mounting because they do
2668 not understand ACLs.
2671 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2673 Description: many clients connecting with IO in progress causes connect timeouts
2674 Details : Avoid synchronous journal commits to avoid delays caused by many
2675 clients connecting/disconnecting when bulk IO is in progress.
2676 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2677 OST_IO_PORTAL to avoid delays behind potentially many pending
2681 Frequency : occasionally with multiple writers to a single file
2683 Description: shared writes to file may result in wrong size reported by stat()
2684 Details : Allow growing of kms when extent lock is cancelled
2687 Frequency : always with random mmap IO to multi-striped file
2689 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2690 Details : Only if the hole is at the end of OST object so that kms is too
2691 small. Fix is to increase kms accordingly in ll_nopage.
2694 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2696 Description: writes to a missing object would leak memory on the OST
2697 Details : If there is an inconsistency between the MDS and OST filesystems,
2698 such that the MDS references an object that doesn't exist, writes
2699 to that object will leak memory due to incorrect cleanup in the
2700 error handling path, eventually running out of memory on the OST.
2705 Description: Creating too long symlink causes lustre errors
2706 Details : Check symlink and name lengths before sending requests to MDS.
2709 Frequency : only if flock is enabled (not on by default)
2711 Description: posix locks not released on fd closure on 2.6.9+
2712 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2713 kernels, this caused such locks not to be released on fd close and
2714 then assertions on fs unmount about still used locks.
2717 Frequency : MDS failover only, very rarely
2719 Description: clients may get ASSERTION(granted_lock != NULL)
2720 Details : When request was taking a long time, and a client was resending
2721 a getattr by name lock request. The were multiple lock requests
2722 with the same client lock handle and
2723 mds_getattr_name->fixup_handle_for_resent_request found one of the
2724 lock handles but later failed with ASSERTION(granted_lock != NULL).
2729 Description: handle->h_buffer_credits > 0, assertion failure
2730 Details : h_buffer_credits is zero after truncate, causing assertion
2731 failure. This patch extends the transaction or creates a new
2735 Frequency : NFS re-export or patchless client
2736 Bugzilla : 11179, 10796
2737 Description: Crash on NFS re-export node (__d_move)
2738 Details : We do not want to hash the dentry if we don't have a lock.
2739 But if this dentry is later used in d_move, we'd hit uninitialised
2740 list head d_hash, so we just do this to init d_hash field but
2741 leave dentry unhashed.
2744 Frequency : NFS re-export or patchless client
2746 Description: NFS exports has problem with symbolic link
2747 Details : lustre client didn't properly install dentry when re-exported
2748 to NFS or running patchless client.
2751 Frequency : NFS re-export or patchless client
2753 Description: Various nfs/patchless fixes.
2754 Details : fixes reuse disconected alias for lookup process - this fixes
2755 warning "find_exported_dentry: npd != pd",
2756 fix permission error with open files at nfs.
2757 fix apply umask when do revalidate.
2760 Frequency : occasional
2762 Description: Crash on NFS re-export node
2763 Details : calling clear_page() on the wrong pointer triggered oops in
2764 generic_mapping_read().
2767 Frequency : rarely, using O_DIRECT IO
2769 Description: unaligned directio crashes client with LASSERT
2770 Details : check for unaligned buffers before trying any requests.
2773 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2775 Description: stale data returned from RAID cache
2776 Details : If only a small amount of IO is done to the RAID device before
2777 reading it again it is possible to get stale data from the RAID
2778 cache instead of reading it from disk.
2781 Frequency : always for sles10 kernel
2783 Description: sles10 support
2784 Details : ll_follow_link: compile fixes and using of nd_set_link
2785 under newer kernels.
2788 Frequency : depends on arch, kernel and compiler version, always on sles10
2791 Description: recursive or deep enough symlinks cause stack overflow
2792 Details : getting rid of large stack-allocated variable in
2796 Frequency : depends on hardware
2798 Description: lustre write performance loss in the SLES10 kernel
2799 Details : the performance loss is caused by using of write barriers in the
2800 ext3 code. The SLES10 kernel turns barrier support on by
2801 default. The fix is to undo that change for ldiskfs.
2803 ------------------------------------------------------------------------------
2805 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2807 * Support for kernels:
2808 2.6.9-42.0.3EL (RHEL 4)
2809 2.6.5-7.276 (SLES 9)
2810 2.4.21-47.0.1.EL (RHEL 3)
2811 2.6.12.6 vanilla (kernel.org)
2815 Frequency : quota enabled and large files being deleted
2817 Description: releasing more than 4GB of quota at once hangs OST
2818 Details : If a user deletes more than 4GB of files on a single OST it
2819 will cause the OST to spin in an infinite loop. Release
2820 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2825 Description: statfs data retrieved from /proc may be stale or zero
2826 Details : When reading per-device statfs data from /proc, in the
2827 {kbytes,files}_{total,free,avail} files, it may appear
2828 as zero or be out of date.
2831 Frequency : systems with MD RAID1 external journal devices
2833 Description: lconf's call to blkid is confused by RAID1 journal devices
2834 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2835 of returning all block devices that match the journal UUID.
2838 Frequency : always, for aggregate stripe size over 4GB
2840 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2841 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2842 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2845 Frequency : always if "lfs find" used on a local file/directory
2847 Description: "lfs find" segfaults if used on a local file/directory
2848 Details : The case where a directory component was not specified wasn't
2849 handled correctly. Handle this properly.
2852 Frequency : always on ppc64
2854 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2855 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2856 perform endianness conversion.
2859 Frequency : rarely (truncate to non-zero file size after write under load)
2860 Bugzilla : 10730, 10687
2861 Description: Files padded with zeros to next 4K multiple
2862 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2863 it is possible that files that are truncated to a non-zero size
2864 immediately after being written are filled with zero bytes beyond
2865 the truncated size. No file data is lost.
2867 Severity : enhancement
2869 Description: Allow recovery/failover for liblustre clients.
2870 Details : liblustre clients were unaware of failover configurations until
2873 Severity : enhancement
2875 Description: user file locks should fail when not mounting with flock option
2876 Details : Set up an error-returning stub in ll_file_operations.lock field
2877 to prevent incorrect behaviour when client is mounted without
2878 flock option. Also, set up properly f_op->flock field for
2882 Frequency : always on ia64
2884 Description: "lfs df" loops on printing out MDS statfs information
2885 Details : The obd_ioctl_data was not initialized and in some systems
2886 this caused a failure during the ioctl that did not return
2887 an error. Initialize the struct and return an error on failure.
2890 Frequency : SLES 9 only
2892 Description: Error of copying files with lustre special EAs as root
2893 Details : Client side always return success for setxattr call for lustre
2894 special xattr (currently only "trusted.lov").
2897 Frequency : rarely on clusters with both ia64+i386 clients
2899 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2900 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2901 clients are doing concurrent writes to the same file it is
2902 possible that the ia64 clients may LASSERT because the OST
2903 extent locks are not PAGE_SIZE aligned. Ensure that grown
2904 locks are always aligned on the request boundary.
2907 Frequency : specific use, occasional
2909 Description: Overwriting in use executable truncates on-disk binary image
2910 Details : If one node attempts to overwrite an executable in use by
2911 another node, we now correctly return ETXTBSY instead of
2912 truncating the file.
2914 Severity : enhancement
2916 Description: Async OSC create to avoid the blocking unnecessarily.
2917 Details : If a OST has no remain object, system will block on the creating
2918 when need to create a new object on this OST. Now, ways use
2919 pre-created objects when available, instead of blocking on an
2920 empty osc while others are not empty. If we must block, we block
2921 for the shortest possible period of time.
2926 Description: chmod on Lustre root is propagated to other clients
2927 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2928 invalid by the follow_mount time.
2931 Frequency : liblustre clients only
2933 Description: Race in 'instant cancel' lock handling could lead to such locks
2934 never to be granted in case of SMP MDS
2935 Details : Do not destroy not yet granted but cbpending locks in
2939 Frequency : replay/resend of open
2941 Description: non null lock assetion failure in mds_intent_policy
2942 Details : Trying to replay/resend lockless open requests resulted in
2943 mds_open() returning 0 with no lock. Now it sets a flag if
2944 a lock is going to be returned.
2946 Severity : enhancement
2948 Description: Checksum enhancements
2949 Details : New checksum enhancements allow for resending RPCs that failed
2952 Severity : enhancement
2954 Description: Tunables on number of dirty pages in cacche
2955 Details : Allow to set limit on number of dirty pages cached.
2960 Description: client crash on unmount - lock still has references
2961 Details : In some error handling cases it was possible to leak a lock
2962 reference on a client while accessing a file. This was not
2963 harmful to the client during operation, but would cause the
2964 client to crash when the filesystem is unmounted.
2967 Frequency : specific case, rare
2969 Description: ETXTBSY on mds though file not in use by client
2970 Details : ETXTBSY is no longer incorrectly returned when attempting to
2971 chmod or chown a directory that the user previously tried to
2972 execute or a currently-executing binary.
2975 Frequency : extremely rare except on liblustre-based clients
2977 Description: Lustre space not freed when files are deleted
2978 Details : Clean up open-unlinked files after client eviction. Previously
2979 the unlink was skipped and the files remained as orphans.
2984 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2985 Details : In some cases it was possible to send a blocking callback to
2986 a client doing a glimpse, even though that client didn't get
2987 a lock granted. When the glimpse lock is cancelled on the OST
2988 the freed lock is left on the waiting list and corrupted the list.
2991 Frequency : all core dumps
2993 Description: Broke core dumps to lustre
2994 Details : Negative dentry may be unhashed if parent does not have UPDATE
2995 lock, but some callers, e.g. do_coredump, expect dentry to be
2996 hashed after successful create, hash it in ll_create_it.
2998 ------------------------------------------------------------------------------
3000 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3002 * Support for kernels:
3003 2.6.9-42.0.2.EL (RHEL 4)
3004 2.6.5-7.276 (SLES 9)
3005 2.4.21-40.EL (RHEL 3)
3006 2.6.12.6 vanilla (kernel.org)
3010 Frequency : always on RHEL 3
3012 Description: Number of open files grows over time
3013 Details : The number of open files grows over time, whether or not
3014 Lustre is started. This was due to a filp leak introduced
3015 by one of our kernel patches.
3017 ------------------------------------------------------------------------------
3019 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
3021 * Support for kernels:
3022 2.6.9-42.EL (RHEL 4)
3023 2.6.5-7.267 (SLES 9)
3024 2.4.21-40.EL (RHEL 3)
3025 2.6.12.6 vanilla (kernel.org)
3030 Bugzilla : 5719, 9635, 9792, 9684
3031 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3032 Details : If a server is under heavy load and cannot reply to new
3033 connection requests before the client resends the (re)connect,
3034 the connection handling code can behave badly if two service
3035 threads are concurrently handing separate (re)connections from
3036 the same client. Add better locking to the connection handling
3037 code, and ensure that only a single connection will be processed
3038 for a given client UUID, even if the lock is dropped.
3040 Severity : enhancement
3042 Description: add TCP zero-copy support to kernel
3043 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3044 sends if the hardware supports scatter-gather and checksumming.
3045 This allows socklnd to do client-write and server-read more
3046 efficiently and reduce CPU utilization from skbuf copying.
3049 Frequency : only if NFS exporting from client
3051 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3052 Details : If setting attributes on a file created under NFS that had
3053 never been opened it would be possible to oops the client
3054 if the file had no objects.
3058 Bugzilla : 9326, 10402, 10897
3059 Description: client crash in ptlrpcd_wake() thread when sending async RPC
3060 Details : It is possible that ptlrpcd_wake() dereferences a freed async
3061 RPC. In rare cases the ptlrpcd thread alread processed the RPC
3062 before ptlrpcd_wake() was called and the request was freed.
3065 Frequency : always for liblustre
3067 Description: liblustre client does MDS+OSTs setattr RPC for each write
3068 Details : When doing a write from a liblustre client, the client
3069 incorrectly issued an RPC to the MDS and each OST the file was
3070 striped over in order to update the timestamps. When writing
3071 with small chunks and many clients this could overwhelm the MDS
3072 with RPCs. In all cases it would slow down the write because
3073 these RPCs are unnecessary.
3075 Severity : enhancement
3077 Description: allow number of MDS service threads to be changed at module load
3078 Details : It is now possible to change the number of MDS service threads
3079 running. Adding "options mds mds_num_threads={N}" to the MDS's
3080 /etc/modprobe.conf will set the number of threads for the next
3081 time Lustre is restarted (assuming the "mds" module is also
3082 reloaded at that time). The default number of threads will
3083 stay the same, 32 for most systems.
3088 Description: OST crash if filesystem is unformatted or corrupt
3089 Details : If an OST is started on a device that has never been formatted
3090 or if the filesystem is corrupt and cannot even mount then the
3091 error handling cleanup routines would dereference a NULL pointer.
3096 Description: NULL pointer deref in llap_from_page.
3097 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3098 incorrect) mapping if the page was truncated/reclaimed while it was
3099 searched for. Check for this condition and skip such pages when
3100 doing readahead. Introduce extra check to llap_from_page() to
3101 verify page->mapping->host is non-NULL (so page is not anonymous).
3104 Frequency : Sometimes when using sys_sendfile
3106 Description: "page not covered by a lock" warnings from ll_readpage
3107 Details : sendfile called ll_readpage without right page locks present.
3108 Now we introduced ll_file_sendfile that does necessary locking
3109 around call to generic_file_sendfile() much like we do in
3113 Frequency : with certain MDS communication failures at client mount time
3115 Description: NULL pointer deref after failed client mount
3116 Details : a client connection request may delayed by the network layer
3117 and not be sent until after the PTLRPC layer has timed out the
3118 request. If the client fails the mount immediately it will try
3119 to clean up before the network times out the request. Add a
3120 reference from the request import to the obd device and delay
3121 the cleanup until the network drops the request.
3124 Frequency : occasionally during client (re)connect
3126 Description: assertion failure during client (re)connect
3127 Details : processing a client connection request may be delayed by the
3128 client or server longer than the client connect timeout. This
3129 causes the client to resend the connection request. If the
3130 original connection request is replied in this interval, the
3131 client may trip an assertion failure in ptlrpc_connect_interpret()
3132 which thought it would be the only running connect process.
3135 Frequency : only with obd_echo servers and clients that are rebooted
3137 Description: kernel BUG accessing uninitialized data structure
3138 Details : When running an obd_echo server it did not start the ping_evictor
3139 thread, and when a client was evicted an uninitialized data
3140 structure was accessed. Start the ping_evictor in the RPC
3141 service startup instead of the OBD startup.
3143 Severity : enhancement
3144 Bugzilla : 10193 (patchless)
3145 Description: Remove dependency on various unexported kernel interfaces.
3146 Details : No longer need reparent_to_init, exit_mm, exit_files,
3147 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3150 Frequency : rare (only users of deprecated and unsupported LDAP config)
3152 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3153 Details : LDAP apparently contains 'lustreName' attributes instead of
3154 'name'. A simple remapping of the name is sufficient.
3157 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3159 Description: waiting_locks_callback trips kernel BUG if client is evicted
3160 Details : Running with the dump_on_timeout debug flag turned on makes
3161 it possible that the waiting_locks_callback() can try to dump
3162 the Lustre kernel debug logs from an interrupt handler. Defer
3163 this log dumping to the expired_lock_main() thread.
3165 Severity : enhancement
3167 Description: Support NFS exporting on 2.6 kernels.
3168 Details : Implement non-rawops metadata methods for NFS server to use without
3169 changing NFS server code.
3172 Frequency : very rare (synthetic metadata workload only)
3174 Description: two racing renames might cause an MDS thread to deadlock
3175 Details : Running the "racer" program may cause one MDS thread to rename
3176 a file from being the source of a rename to being the target of
3177 a rename at exactly the same time that another thread is doing
3178 so, and the second thread has already enqueued these locks after
3179 doing a lookup of the target and is trying to relock them in
3180 order. Ensure that we don't try to re-lock the same resource.
3183 Frequency : only very large systems with liblustre clients
3185 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3186 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3187 all OSTs in parallel. This allows the eviction of stale liblustre
3188 clients to proceed much faster than if they were done in series,
3189 and also offers similar improvements for other set_info RPCs.
3194 Description: excessive CPU usage during initial read phase on client
3195 Details : During the initial read phase on a client, it would agressively
3196 retry readahead on the file, consuming too much CPU and impacting
3197 performance (since 1.4.5.8). Improve the readahead algorithm
3198 to avoid this, and also improve some other common cases (read
3199 of small files in particular, where "small" is files smaller than
3200 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3205 Description: MDS crash when receiving packet with unknown intent.
3206 Details : Do not LBUG in unknown intent case, just return -EFAULT
3208 Severity : enhancement
3209 Bugzilla : 9293, 9385
3210 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3211 Details : Do not serialize getattr (non-intent version) and statfs.
3214 Frequency : occasional, when OST network is overloaded/intermittent
3216 Description: client evicted by OST after bulk IO timeout
3217 Details : If a client sends a bulk IO request (read or write) the OST
3218 may evict the client if it is unresposive to its data GET/PUT
3219 request. This is incorrect if the network is overloaded (takes
3220 too long to transfer the RPC data) or dropped the OST GET/PUT
3221 request. There is no need to evict the client at all, since
3222 the pinger and/or lock callbacks will handle this, and the
3223 client can restart the bulk request.
3226 Frequency : Always when mmapping file with no objects
3228 Description: client crashes when mmapping file with no objects
3229 Details : Check that we actually have objects in a file before doing any
3230 operations on objects in ll_vm_open, ll_vm_close and
3236 Description: Request leak when working with deleted CWD
3237 Details : Introduce advanced request refcount tracking for requests
3238 referenced from lustre intent.
3240 Severity : Enhancement
3242 Description: Cache open file handles on client.
3243 Details : MDS now will return special lock along with openhandle, if
3244 requested and client is allowed to hold openhandle, even if unused,
3245 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3246 closing files for every read/write openration.
3248 Severity : Enhancement
3250 Description: Cache open negative dentries on client when possible.
3251 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3252 negative dentries on lock revocation.
3257 Description: Remounting a client read-only wasn't possible with a zconf mount
3258 Details : It wasn't possible to remount a client read-only with llmount.
3260 Severity : enhancement
3261 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3262 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3263 MPICH is included to add Lustre-specific ADIO interfaces.
3264 This is based closely on the UFS ADIO layer and only differs
3265 in file creation, in order to allow the OST striping to be set.
3266 This is user-contributed code and not supported by CFS.
3271 Description: extended inode attributes (immutable, append-only) work improperly
3272 when 2.4 and 2.6 kernels are used on client/server or vice versa
3273 Details : Introduce kernel-independent values for these flags.
3275 Severity : enhancement
3278 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3279 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3280 granularity. Now, Fractional MB granularity is be supported,
3281 this is very useful for low memory system.
3283 Severity : enhancement
3285 Description: Getattr by fid
3286 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3287 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3288 have locks and avoids getattr rpc requests for it on every stat.
3291 Frequency : Always, for filesystems larger than 2TB
3293 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3294 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3295 filesystems larger than 2TB to be mounted with mballoc enabled.
3298 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3300 Description: filesystem corruption for non-standard kernels and very large OSTs
3301 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3302 filesystems larger than 2TB is mounted then the kernel will
3303 silently corrupt the start of the filesystem. CONFIG_LBD is
3304 enabled for all CFS-supported kernels, but the possibility of
3305 this happening with a modified kernel config exists.
3307 Severity : enhancement
3309 Description: add client O_DIRECT support for 2.6 kernels
3310 Details : It is now possible to do O_DIRECT reads and writes to files
3311 in the Lustre client mountpoint on 2.6 kernel clients.
3313 Severity : enhancement
3315 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3316 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3317 parallel, not waiting for response from every OST before sending
3318 a rpc to the next OST.
3323 Description: setattr vs write race when updating file timestamps
3324 Details : Client processes that update a file timestamp into the past
3325 right after writing to the file (e.g. tar) it is possible that
3326 the updated file modification time can be reset to the current
3327 time due to a race between processing the setattr and write RPC.
3329 Severity : enhancement
3331 Description: Bring 'lfs find' closer in line with regular Linux find.
3332 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3333 -print0 options and obtains all the needed info through the lustre
3336 Severity : enhancement
3338 Description: support up to 1024 configured devices on one node
3339 Details : change obd_dev array from statically allocated to dynamically
3340 allocated structs as they are first used to reduce memory usage
3345 Description: Flush dirty partially truncated pages during truncate
3346 Details : Immediatelly flush partially truncated pages in filter_setattr,
3347 this way we completely avoid having any pages in page cache on OST
3348 and can retire ugly workarounds during writes to flush such pages.
3353 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3354 Details : protect inode from truncation within vfs_unlink() context
3355 just take a reference before calling vfs_unlink() and release it
3356 when parent's i_sem is free.
3359 Frequency : always, if extents are used on OSTs
3361 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3362 Details : OSTs using the extents format would not zero the high 16 bits of
3363 the index physical block number. This is not a problem for any
3364 OST filesystems smaller than 16TB, and no kernels support ext3
3365 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3366 new/modified files) and can be fixed for existing filesystems
3367 with e2fsprogs-1.39-cfs1.
3372 Description: import connection selection may be incorrect if timer wraps
3373 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3374 import connections to be ignored if the 32-bit jiffies counter
3375 wraps. Use a 64-bit jiffies counter.
3378 Frequency : during server recovery
3380 Description: crash after server is denying duplicate export
3381 Details : If clients are resending connect requests to the server, the
3382 server refuses to allow a client to connect multiple times.
3383 Fixed a bug in the handling of this case.
3386 Frequency : very large clusters immediately after boot
3388 Description: LNET request buffers exhausted under heavy short-term load
3389 Details : If a large number of client requests are generated on a service
3390 that has previously never seen so many requests it is possible
3391 that the request buffer growth cannot keep up with the spike in
3392 demand. Instead of dropping incoming requests, they are held in
3393 the LND until the RPC service can accept more requests.
3396 Frequency : Sometimes during replay
3398 Description: Assertion failure in ll_local_open after replay.
3399 Details : If replay happened on an open request reply before we were able
3400 to set replay handler, reply will become not swabbed tripping the
3401 assertion in ll_local_open. Now we set the handler right after
3402 recognising of open request
3405 Frequency : very rare
3407 Description: kernel reports "badness in vsnprintf"
3408 Details : Reading from the "recovery_status" /proc file in small chunks
3409 may cause a negative length in lprocfs_obd_rd_recovery_status()
3410 call to vsnprintf() (which is otherwise harmless). Exit early
3411 if there is no more space in the output buffer.
3413 Severity : enhancement
3415 Description: clear OBD RPC statistics by writing to them
3416 Details : It is now possible to clear the OBD RPC statistics by writing
3417 to the "stats" file.
3422 Description: Client mtime is not the same on different clients after utimes
3423 Details : In some cases, the client was using the utimes() syscall on
3424 a file cached on another node. The clients now validate the
3425 ctime from the MDS + OSTs to determine which one is right.
3430 Description: Inability to activate failout mode
3431 Details : lconf script incorrectly assumed that in python string's numeric
3432 value is used in comparisons.
3435 Frequency : always with multiple stripes per file
3437 Description: Inefficient object allocation for mutli-stripe files
3438 Details : When selecting which OSTs to stripe files over, for files with
3439 a stripe count that divides evenly into the number of OSTs,
3440 the MDS is always picking the same starting OST for each file.
3441 Return the OST selection heuristic to the original design.
3446 Description: mount failures may take full timeout to return an error
3447 Details : Under some heavy load conditions it is possible that a
3448 failed mount can wait for the full obd_timeout interval,
3449 possibly several minutes, before reporting an error.
3450 Instead return an error as soon as the status is known.
3452 ------------------------------------------------------------------------------
3454 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
3456 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3457 INTEROPERATE with older versions automatically. Please read the
3458 user documentation before upgrading any part of a live system.
3459 * WARNING: Lustre networking configuration changes are required with
3460 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3463 * Support for kernels:
3464 2.6.9-22.0.2.EL (RHEL 4)
3465 2.6.5-7.244 (SLES 9)
3466 2.6.12.6 vanilla (kernel.org)
3469 Severity : enhancement
3470 Bugzilla : 7981/8208
3471 Description: Introduced Lustre Networking (LNET)
3472 Details : LNET is new networking infrastructure for Lustre, it includes
3473 a reorganized network configuration mode (see the user
3474 documentation for full details) as well as support for routing
3475 between different network fabrics. Lustre Networking Devices
3476 (LNDs) for the supported network fabrics have also been
3477 created for this new infrastructure.
3479 Severity : enhancement
3480 Description: Introduced Access control lists
3481 Details : clients can set ACLs on files and directories in order to have
3482 more fine-grained permissions than the standard Unix UGO+RWX.
3483 The MDS must be started with the "-o acl" mount option.
3485 Severity : enhancement
3486 Description: Introduced filesystem quotas
3487 Details : Administrators may now establish per-user quotas on the
3490 Severity : enhancement
3492 Description: Configuration change for the XT3
3493 The PTLLND is now used to run Lustre over Portals on the XT3
3494 The configure option(s) --with-cray-portals are no longer used.
3495 Rather --with-portals=<path-to-portals-includes> is used to
3496 enable building on the XT3. In addition to enable XT3 specific
3497 features the option --enable-cray-xt3 must be used.
3502 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3503 Details : A race between allocating a new llcd and re-getting the llcd_lock
3504 allowed another thread to grab newly-allocated llcd.
3506 Severity : enhancement
3508 Description: 2.6 OST async journal commit and locking fix to improve performance
3509 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3510 2.6 kernels are now async as they already were in 2.4 kernels so
3511 that they can commit concurrently with the network bulk transfer.
3512 For block-allocated files the filter allocation semaphore is held
3513 to avoid filesystem fragmentation during allocation. BKL lock
3514 removed for 2.6 xattr operations where it is no longer needed.
3519 Description: lconf incorrectly determined whether two IP networks could talk
3520 Details : In some more complicated routing and multiple-network
3521 configurations, lconf will avoid trying to make a network
3522 connection to a disjoint part of the IP space. It was doing the
3523 math incorrectly for one set of cases.
3528 Description: Fix for potential infinite loop processing records in an llog.
3529 Details : If an llog record is corrupted/zeroed, it is possible to loop
3530 forever in llog_process(). Validate the llog record length
3531 and skip the remainder of the block on error.
3534 Frequency : occasional (liblustre only)
3536 Description: liblustre could not open files whose last component is a symlink
3537 Details : sysio_path_walk() would incorrectly pass the open intent to
3538 intermediate path components.
3541 Frequency : rare (liblustre only with non-standard tuning)
3542 Bugzilla : 7201 (7350)
3543 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3544 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3545 after it has been released, only do this for referenced locks
3547 Severity : enhancement
3549 Description: specifying an (invalid) directory default stripe_size of -1
3550 would reset the directory default striping
3551 Details : stripe_size -1 was used internally to signal directory stripe
3552 removal, now use "all default" to signal dir stripe removal
3553 as a directory striping of "all default" is not useful
3556 Frequency : common for large clusters running liblustre clients
3558 Description: doing an ls when liblustre clients are running is slow
3559 Details : sending a glimpse AST to a liblustre client waits for every AST
3560 to time out, as liblustre clients will not respond. Since they
3561 cannot cache data we refresh the OST lock LVB from disk instead.
3563 Severity : enhancement
3565 Description: doing an ls at the same time as file IO can be slow
3566 Details : enqueue and other "small" requests can be blocked behind many
3567 large IO requests. Create a new OST IO portal for non-IO
3568 requests so they can be processed faster.
3571 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3573 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3574 created file with mode 000
3575 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3576 allow an ftruncate() as a non-root user to a file with mode 000.
3577 Lustre now respects this flag to disable mode checks when
3578 truncating a file owned by the user
3581 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3584 Description: Revoking locks from clients that went dead or catatonic might take
3586 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3587 cancellation of such locks instant on servers without waiting for
3588 any reply from clients. Clients drops these locks when cancel
3589 notification from server is received without replying.
3592 Frequency : liblustre-only, when liblustre client dies or becomes busy
3594 Description: Doing ls on Linux clients can take a long time with active
3596 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3597 granting such locks to it in the first place if possible. Locks
3598 are taken by proxy on the OST during the read or write and
3599 dropped immediately afterward. Add connect flags handling, do
3600 not grant locks to liblustre clients for glimpse ASTs.
3602 Severity : enhancement
3604 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3605 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3606 sequential reads and behaves badly with applications doing random
3607 reads. Improve it by reading ahead at least read region, and
3608 avoiding excessive large RPC for small reads.
3610 Severity : enhancement
3612 Description: Creating more than 1000 files for a single job may cause a load
3613 imbalance on the OSTs if there are also a large number of OSTs.
3614 Details : qos_prep_create() uses an OST index reseed value that is an
3615 even multiple of the number of available OSTs so that if the
3616 reseed happens in the middle of the object allocation it will
3617 still utilize the OSTs as uniformly as possible.
3622 Description: OST or MDS may oops in ping_evictor_main()
3623 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3624 but doesn't restart at beginning of obd_exports_timed list
3627 Severity : enhancement
3629 Description: improve by-nid export eviction on the MDS and OST
3630 Details : allow multiple exports with the same NID to be evicted at one
3631 time without re-searching the exports list.
3634 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3636 Description: MDS may oops in groups_free()
3637 Details : in rare race conditions a newly allocated group_info struct is
3638 freed again, and this can be NULL. The 2.4 compatibility code
3639 for groups_free() checked for a NULL pointer, but 2.6 did not.
3642 Frequency : common for liblustre clients doing little filesystem IO
3643 Bugzilla : 9352, 7313
3644 Description: server may evict liblustre clients accessing contended locks
3645 Details : if a client is granted a lock or receives a completion AST
3646 with a blocking AST already set it would not reply to the AST
3647 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3650 Frequency : lfs setstripe, only systems with more than 160 OSTs
3652 Description: unable to set striping with a starting offset beyond OST 160
3653 Details : llapi_create_file() incorrectly limited the starting stripe
3654 index to the maximum single-file stripe count.
3657 Frequency : LDAP users only
3659 Description: lconf did not handle in-kernel recovery with LDAP properly
3660 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3662 Severity : enhancement
3664 Description: bind OST threads to NUMA nodes to improve performance
3665 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3666 node and do their allocations there to localize memory access
3668 Severity : enhancement
3670 Description: llmount can determine client NID directly from Myrinet (GM)
3671 Details : the client NID code from gmnalnid was moved directly into
3672 llmount, removing the need to use this or specifying the
3673 client NID explicitly when mounting GM clients with zeroconf
3676 Frequency : if client is started with down MDS
3678 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3679 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3680 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3681 ptlrpc_import_delayed_req() because we didn't check if the
3682 request was interrupted, and we also didn't break out of the
3683 event loop if there was no timeout
3688 Description: data loss during non-page-aligned writes to a single file from
3689 both multiple nodes and multiple threads on one node at same time
3690 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3691 inconsistency led to false short-reads, that were cached and later
3692 used by ->prepare_write() to fill in partially written page,
3693 leading to data loss.
3696 Frequency : always, if lconf --abort_recovery used
3698 Description: lconf --abort_recovery fails with 'Operation not supported'
3699 Details : lconf was attempting to abort recovery on the MDT device and not
3702 Severity : enhancement
3704 Description: remove cleanup logs
3705 Details : replace lconf-generated cleanup logs with lustre internal
3706 cleanup routines. Eliminates the need for client-cleanup and
3709 Severity : enhancement
3711 Description: add support for EAs (user and system) on lustre filesystems
3712 Details : it is now possible to store extended attributes in the Lustre
3713 client filesystem, and with the user_xattr mount option it
3714 is possible to allow users to store EAs on their files also
3716 Severity : enhancement
3718 Description: Add possibility (config option) to show minimal available OST free
3720 Details : When compiled with --enable-mindf configure option, statfs(2)
3721 (and so, df) will return least minimal free space available from
3722 all OSTs as amount of free space on FS, instead of summary of
3723 free spaces of all OSTs.
3725 Severity : enhancement
3727 Description: do not expand extent locks acquired on OST-side
3728 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3729 by server: they are not cached anyway.
3732 Frequency : when mmap is used/binaries executed from Lustre
3734 Description: Unmmap pages before throwing them away from read cache.
3735 Details : llap_shrink cache now attempts to unmap pages before discarding
3736 them (if unmapping failed - do not discard). SLES9 kernel has
3737 extra checks that trigger if this unmapping is not done first.
3742 Description: lconf didn't resolve symlinks before checking to see whether a
3743 given mountpoint was already in use
3746 Frequency : when migrating failover services
3747 Bugzilla : 6395, 9514
3748 Description: When migrating a subset of services from a node (e.g. failback
3749 from a failover service node) the remaining services would
3750 time out and evict clients.
3751 Details : lconf --force (implied by --failover) sets the global obd_timeout
3752 to 5 seconds in order to quickly disconnect, but this caused
3753 other RPCs to time out too quickly. Do not change the global
3754 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3756 Severity : enhancement
3757 Frequency : if MDS is started with down OST
3758 Bugzilla : 9439,5706
3759 Description: Allow startup/shutdown of an MDS without depending on the
3760 availability of the OSTs.
3761 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3762 Add appropriate locking and lov-osc refcounts for safe
3763 cleaning. Add osc abort_inflight calls in case the
3764 synchronize never started.
3767 Frequency : occasional (Cray XT3 only)
3769 Description: root not authorized to access files in CRAY_PORTALS environment
3770 Details : The client process capabilities were not honoured on the MDS in
3771 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3772 been accessed by an authorized user then root was able to access
3773 the file on the local client also. The root user capabilities
3774 are now allowed on the MDS, as this environment has secure UID.
3777 Frequency : occasional
3779 Description: ldiskfs "too long searching" message happens too often
3780 Details : A debugging message (otherwise harmless) prints too often on
3781 the OST console. This has been reduced to only happen when
3782 there are fragmentation problems on the filesystem.
3787 Description: Division by zero in statfs when all OSCs are inactive
3788 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3789 lov_statfs divides by value returned by lov_get_stripecnt().
3793 Bugzilla : 9489, 3273
3794 Description: First write from each client to each OST was only 4kB in size,
3795 to initialize client writeback cache, which caused sub-optimal
3796 RPCs and poor layout on disk for the first writen file.
3797 Details : Clients now request an initial cache grant at (re)connect time
3798 and so that they can start streaming writes to the cache right
3799 away and always do full-sized RPCs if there is enough data.
3800 If the OST is rebooted the client also re-establishes its grant
3801 so that client cached writes will be honoured under the grant.
3806 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3807 Details : Now I/O RPCs go to different portal number and (presumably) fast
3808 lock requests (and glimses) and other RPCs get their own service
3809 threads pool that should be able to service those RPCs
3812 Severity : enhancement
3814 Description: Ability to exchange lustre version between client and servers and
3815 issue warnings at client side if client is too old. Also for
3816 liblustre clients there is ability to refuse connection of too old
3818 Details : New 'version' field is added to connect data structure that is
3819 filled with version info. That info is later checked by server and
3823 Frequency : rare, liblustre only.
3824 Bugzilla : 9296, 9581
3825 Description: Two simultaneous writes from liblustre at offset within same page
3826 might proceed at the same time overwriting eachother with stale
3828 Details : I/O lock withing llu_file_prwv was released too early, before data
3829 actually was hitting the wire. Extended lock-holding time until
3830 server acknowledges receiving data.
3833 Frequency : extremely rare. Never observed in practice.
3835 Description: avoid generating lustre_handle cookie of 0.
3836 Details : class_handle_hash() generates handle cookies by incrementing
3837 global counter, and can hit 0 occasionaly (this is unlikely, but
3838 not impossible, because initial value of cookie counter is
3839 selected randonly). Value of 0 is used as a sentinel meaning
3840 "unassigned handle" --- avoid it. Also coalesce two critical
3841 sections in this function into one.
3843 Severity : enhancement
3845 Description: allow liblustre clients to delegate truncate locking to OST
3846 Details : To avoid overhead of locking, liblustre client instructs OST to
3847 take extent lock in ost_punch() on client's behalf. New connection
3848 flag is added to handle backward compatibility.
3850 Severity : enhancement
3851 Bugzilla : 4928, 7341, 9758
3852 Description: allow number of OST service threads to be specified
3853 Details : a module parameter allows the number of OST service threads
3854 to be specified via "options ost ost_num_threads={N}" in the
3855 OSS's /etc/modules.conf or /etc/modprobe.conf.
3859 Bugzilla : 6146, 9635, 9895
3860 Description: servers crash with bad pointer in target_handle_connect()
3861 Details : In rare cases when a client is reconnecting it was possible that
3862 the connection request was the last reference for that export.
3863 We would temporarily drop the export reference and get a new
3864 one, but this may have been the last reference and the export
3865 was just destroyed. Get new reference before dropping old one.
3867 Severity : enhancement
3868 Frequency : if client is started with failover MDS
3870 Description: Allow multiple MDS hostnames in the mount command
3871 Details : Try to read the configuration from all specified MDS
3872 hostnames during a client mount in case the "primary"
3875 Severity : enhancement
3877 Description: Stop sending data to evicted clients as soon as possible.
3878 Details : Check if the client we are about to send or are sending data to
3879 was evicted already. (Check is done every second of waiting,
3880 for which l_wait_event interface was extended to allow checking
3881 of exit condition at specified intervals).
3884 Frequency : rare, normally only when NFS exporting is done from client
3886 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3888 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3891 Frequency : always, if srand() is called before liblustre initialization
3893 Description: Liblustre uses system PRNG disturbing its usage by user application
3894 Details : Introduce internal to lustre fast and high-quality PRNG for
3895 lustre usage and make liblustre and some other places in generic
3896 lustre code to use it.
3898 Severity : enhancement
3899 Bugzilla : 9477, 9557, 9870
3900 Description: Verify that the MDS configuration logs are updated when xml is
3901 Details : Check if the .xml configuration logs are newer than the config
3902 logs stored on the MDS and report an error if this is the case.
3903 Request --write-conf, or allow starting with --old_conf.
3905 Severity : enhancement
3907 Description: Handle symlinks in the path when checking if Lustre is mounted.
3908 Details : Resolve intermediate symlinks when checking if a client has
3909 mounted a filesystem to avoid duplicate client mounts.
3914 Description: lconf can hit an error exception but still return success.
3915 Details : The lconf command catches the Command error exception at the top
3916 level script context and will exit with the associated exit
3917 status, but doesn't ensure that this exit status is non-zero.
3922 Description: failure of ptlrpc thread startup can cause oops
3923 Details : Starting a ptlrpc service thread can fail if there are a large
3924 number of threads or the server memory is very fragmented.
3925 Handle this without oopsing.
3928 Frequency : always, only if liblustre and non-default acceptor port was used
3930 Description: liblustre cannot connect to servers with non-default acceptor port
3931 Details : tcpnal_set_default_params() was not called and was therefore
3932 ignoring the environment varaible TCPNAL_PORT, as well as other
3933 TCPNAL_ environment variables
3938 Description: two objects could be created on the same OST for a single file
3939 Details : If an OST is down, in some cases it was possible to create two
3940 objects on a single OST for a single file. No problems other
3941 than potential performance impact and spurious error messages.
3945 Bugzilla : 5681, 9562
3946 Description: Client may oops in ll_unhash_aliases
3947 Details : Client dcache may become inconsistent in race condition.
3948 In some cases "getcwd" can fail if the current directory is
3954 Description: Inode refcounting problems in NFS export code
3955 Details : link_raw functions used to call d_instantiate without obtaining
3956 extra inode reference first.
3960 Bugzilla : 9942, 9903
3961 Description: Referencing freed requests leading to crash, memleaks with NFS.
3962 Details : We used to require that call to ll_revalidate_it was always
3963 followed by ll_lookup_it. Also with revalidate_special() it is
3964 possible to call ll_revalidate_it() twice for the same dentry
3965 even if first occurence returned success. This fix changes semantic
3966 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3967 reference on a request referred from the intent.
3968 ll_intent_release() then releases such a request.
3971 Frequency : rare, normally benchmark loads only
3973 Description: unlinked inodes were kept in memory on the client
3974 Details : If a client is repeatedly creating and unlinking files it
3975 can accumulate a lot of stale inodes in the inode slab cache.
3976 If there is no other client load running this can cause the
3977 client node to run out of memory. Instead flush old inodes
3978 from client cache that have the same inode number as a new inode.
3981 Frequency : SLES9 2.6.5 kernel and long filenames only
3982 Bugzilla : 9969, 10379
3983 Description: utime reports stale NFS file handle
3984 Details : SLES9 uses out-of-dentry names in some cases, which confused
3985 the lustre dentry revalidation. Change it to always use the
3989 Frequency : rare, unless heavy write-truncate concurrency is continuous
3990 Bugzilla : 4180, 6984, 7171, 9963, 9331
3991 Description: OST becomes very slow and/or deadlocked during object unlink
3992 Details : filter_destroy() was holding onto the parent directory lock
3993 while truncating+unlinking objects. For very large objects this
3994 may block other threads for a long time and slow overall OST
3995 responsiveness. It may also be possible to get a lock ordering
3996 deadlock in this case, or run out of journal credits because of
3997 the combined truncate+unlink. Solution is to do object truncate
3998 first in one transaction without parent lock, and then do the
3999 final unlink in a new transaction with the parent lock. This
4000 reduces the lock hold time dramatically.
4003 Frequency : rare, 2.4 kernels only
4005 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4006 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4007 before unmounting their filesystem in order to avoid deadlock.
4008 The kernel_locked() function in 2.4 kernels only checks whether
4009 the kernel lock is held, not whether it is this process that is
4010 holding it as 2.6 kernels do.
4015 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4016 Details : The client ptlrpc code may be trying to reconnect to a down
4017 server before a previous connection attempt has timed out.
4018 Increase the reconnect interval to be longer than the connection
4019 timeout interval to avoid sending duplicate connections to
4023 Frequency : echo_client brw_test command
4025 Description: fix echo_client to work with OST preallocated code
4026 Details : OST preallocation code (5137) didn't take echo_client IO path
4027 into account: echo_client calls filter methods outside of any
4028 OST thread and, hence, there is no per-thread preallocated
4029 pages and buffers to use. Solution: hijack pga pages for IO. As
4030 a byproduct, this avoids unnecessary data copying.
4034 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4035 Description: Client can oops in mdc_commit_close() after open replay
4036 Details : It was possible for the MDS to return an open request with no
4037 transaction number in mds_finish_transno() if the client was
4038 evicted, but without actually returning an error. Clients
4039 would later try to replay that open and may trip an assertion
4040 Simplify the client close codepath, and always return an error
4041 from the MDS in case the open is not successful.
4044 Frequency : rare, 2.6 OSTs only
4046 Description: OST may deadlock under high load on fragmented files
4047 Details : If there was a heavy load and highly-fragmented OST filesystems
4048 it was possible to have all the OST threads deadlock waiting on
4049 allocation of biovecs, because the biovecs were not released
4050 until the entire RPC IO was completed. Instead, release biovecs
4051 as soon as they are complete to ensure forward IO progress.
4053 Severity : enhancement
4055 Description: Support for specifying external journal device at mount
4056 Details : If an OST or MDS device is formatted with an external journal
4057 device, this device major/minor is stored in the ext3 superblock
4058 and may not be valid for failover. Allow detecting and
4059 specifying the external journal at mount time.
4064 Description: Mounting an MDS with pending unlinked files may cause oops
4065 Details : target_finish_recovery() calls mds_postrecov() which returned
4066 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4067 considers this an error and immediately begins cleaning up the
4068 lov, just after starting the mds_lov process
4070 Severity : enhancement
4072 Description: Implement 'lfs df' to report actual free space on per-OST basis
4073 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4074 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4075 usage of objects; '-h' to report in human readable format.
4077 ------------------------------------------------------------------------------
4079 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
4086 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4087 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4088 at startup during orphan unlinking. Instead, skip buddy update
4089 before setup, as it will be regenerated after recovery is complete.
4094 Description: If an OST is inactive, its locks might reference stale inodes.
4095 Details : lov_change_cbdata() must iterate over all namespaces, even if
4096 they are inactive to clear inode references from the lock.
4098 Severity : enhancement
4099 Frequency : occasional, if non-standard max_dirty_mb used
4101 Description: Client will block write RPCs if not enough grant
4102 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4103 then the client will block writes while waiting for another RPC
4104 to complete instead of consuming its dirty limit. With change
4105 we get improved performance when max_dirty_mb is small.
4107 Severity : enhancement
4108 Bugzilla : 3389, 6253
4109 Description: Add support for supplementary groups on the MDS.
4110 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4111 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4112 lookups for user supplementary groups into a cache.
4116 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4117 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4118 for liblustre O_EXCL is enforced by MDS.
4122 Description: readdir never returns NULL in liblustre.
4123 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4124 dentry in struct dirent.
4128 Frequency : liblustre only, depends on application IO pattern
4129 Description: liblustre clients evicted if not contacting servers
4130 Details : Don't put liblustre clients into the ping_evictor list, so
4131 they will not be evicted by the pinger ever.
4133 Severity : enhancement
4135 Description: Add ability to evict clients by NID from MDS.
4136 Details : By echoing "nid:$NID" string into
4137 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4138 $NID would be instantly evicted from this MDS and from all active
4139 OSTs connected to it.
4143 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4144 Details : lookup_it_finish() used to query file size from OSTs that was not
4149 Description: service threads change working directory to that of init
4150 Details : Starting lustre service threads may pin the working directory
4151 of the parent thread, making that filesystem busy. Threads
4152 now change to the working directory of init to avoid this.
4156 Frequency : during shutdown only
4157 Description: shutdown with a failed MDS or OST can cause unmount to hang
4158 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4163 Frequency : 2.6 only
4164 Description: chmod/chown may include an extra supplementary group
4165 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4166 supplementary group and if none is specified this is used.
4169 Bugzilla : 5479 (6816)
4171 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4172 Details : If lookup is in progress on a file that is unlinked we might try
4173 to revalidate the inode and fail in revalidate after lookup is
4174 complete and ll_file_open() enqueues the open again but
4175 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4178 Frequency : always, if lconf --abort_recovery used
4180 Description: lconf --abort_recovery fails with 'Operation not supported'
4181 Details : lconf was attempting to abort recovery on the MDT device and not
4184 ------------------------------------------------------------------------------
4186 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4191 Frequency : rare (only unsupported configurations with a node running as an
4193 Bugzilla : 6514, 5137
4194 Description: Mounting a Lustre file system on a node running as an OST could
4196 Details : OSTs now preallocates memory needed to write out data at
4197 startup, instead of when needed, to avoid having to
4198 allocate memory in possibly low memory situations.
4199 Specifically, if the file system is mounted on on OST,
4200 memory pressure could force it to try to write out data,
4201 which it needed to allocate memory to do. Due to the low
4202 memory, it would be unable to do so and the node would
4203 become unresponsive.
4205 Severity : enhancement
4207 Description: Addition of lconf --service command line option
4208 Details : lconf now accepts a '--service <arg>' option, which is
4209 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4211 Severity : enhancement
4213 Description: Failover mode is now the default for OSTs.
4214 Details : By default, OSTs will now run in failover mode. To return to
4215 the old behaviour, add '--failout' to the lmc line for OSTs.
4217 Severity : enhancement
4219 Description: Health checks are now provided for MDS and OSTs
4220 Details : Additional detailed health check information on MSD and OSTs
4221 is now provided through the procfs health_check value.
4224 Frequency : occasional, depends on IO load
4226 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4227 after numerous create/delete cycles
4228 Details : The ext3 inode allocation policy would not allocate new inodes
4229 very well on the OSTs because there are no new directories
4230 being created. Instead we look for groups with free space if
4231 the parent directories are nearly full.
4236 Description: Network or server problems during mount may cause partially
4237 mounted clients instead of returning an error.
4238 Details : The config llog parsing code may overwrite the error return
4239 code during mount error handling, returning success instead
4245 Description: MDS can fail to allocate large reply buffers
4246 Details : After long uptimes the MDS can fail to allocate large reply
4247 buffers (e.g. zconf client mount config records) due to memory
4248 fragmentation or consumption by the buffer cache. Preallocate
4249 some large reply buffers so that these replies can be sent even
4250 under memory pressure.
4254 Frequency : rare (liblustre)
4255 Description: fsx running with liblustre complained that using truncate() to
4256 extend the file doesn't work. This patch corrects that issue.
4257 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4258 ATTR_SIZE and lsm use in llu_setattr_raw.
4262 Frequency : rare, only 2.6 kernels
4263 Description: Unusual file access patterns on the MDS may result in inode
4264 data being lost in very rare circumstances.
4265 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4266 code in the 2.6 kernel caused the inode and/or EA data not to
4267 be read from disk, causing single-file corruption.
4271 Frequency : rare, only 2.6 filesystems using extents
4272 Description: Heavy concurrent write and delete load may cause data corruption.
4273 Details : It was possible under high-load situations to have an extent
4274 metadata block in the block device cache from a just-unlinked
4275 file overwrite a newly-allocated data block. We now unmap any
4276 metadata buffers that alias just-allocated data blocks.
4280 Frequency : filesystems with default stripe_count larger than 77
4281 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4282 Details : lconf specifies an inode size of 4096 bytes when the default
4283 stripe_count is larger than 77. This conflicts with the default
4284 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4285 this case to avoid pinning too much memory for large EAs.
4287 ------------------------------------------------------------------------------
4289 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4294 Frequency : rare (extremely heavy IO load with hundreds of clients)
4296 Description: Client is evicted, gets IO error writing to file
4297 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4298 caused clients to be evicted for AST timeouts. The fixes in
4299 bug 5192 mean we no longer need to have such short AST timeouts
4300 so ldlm_timeout has been increased.
4303 Frequency : occasional during --force or --failover shutdown under load
4304 Bugzilla : 5949, 4834
4305 Description: Server oops/LBUG if stopped with --force or --failover under load
4306 Details : a collection of import/export refcount and cleanup ordering
4307 issues fixed for safer force cleanup
4310 Frequency : only filesystems larger than 120 OSTs
4311 Bugzilla : 5990, 6223
4312 Description: lfs getstripe would oops on a very large filesystem
4313 Details : lov_getconfig used kfree on vmalloc'd memory
4316 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4317 Bugzilla : 6242, 6243
4318 Description: reading from files that had been truncated to a non-zero size
4319 but never opened returned no data
4320 Details : ll_file_read() reads zeros from no-object files to EOF
4325 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4326 Details : There's little harm in aborting MDS/OSS recovery and letting it
4327 try again, so I removed the LASSERT and return an error instead.
4329 Severity : enhancement
4331 Description: New debugging infrastructure for tracking down data corruption
4332 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4333 (b) cover more of the client-side code path, and (c) try to narrow
4334 down where problems occurred
4338 Bugzilla : 3819, 4364, 4397, 6313
4339 Description: Racing close and eviction MDS could cause assertion in mds_close
4340 Details : It was possible to get multiple mfd references during close and
4341 client eviction, leading to one thread referencing a freed mfd.
4343 Severity: : enhancement
4344 Bugzilla : 3262, 6359
4345 Description: Attempts to reconnect to servers are now more aggressive.
4346 Details : This builds on the enhanced upcall-less recovery that was added
4347 in 1.4.2. When trying to reconnect to servers, clients will
4348 now try each server in the failover group every 10 seconds. By
4349 default, clients would previously try one server every 25 seconds.
4354 Description: After recovery, certain operations trigger a failed
4355 assertion on a client.
4356 Details : Failing over an mds, using lconf -d --failover, while a
4357 client was doing a readdir() call would cause the client to
4358 LBUG after recovery completed and the readdir() was resent.
4360 Severity : enhancement
4362 Description: Default groups are now added by lconf
4363 Details : You can now run lconf --group <servicename> without having to
4364 manually add groups with lmc.
4367 Frequency : occasional
4369 Description: Nodes with an elan id of 0 trigger a failed assertion
4372 Frequency : always when accessing e.g. tty/console device nodes
4374 Description: tty and some other devices nodes cannot be used on lustre
4375 Details : file's private_data field is used by device data and lustre
4376 values in there got lost. New field was added to struct file to
4377 store fs-specific private data.
4380 Frequency : when exporting Lustre via NFS
4382 Description: NFSD failed occasionally when looking up a path component
4383 Details : NFSD is looking up ".." which was broken in ext3 directories
4384 that had grown large enough to become hashed.
4387 Frequency : Clusters with multiple interfaces not on the same subnet
4389 Description: Nodes will repeatedly try to reconnect to an interface which it
4390 cannot reach and report an error to the log.
4391 Details : Extra peer list entries will be created by lconf with some peers
4392 unreachable. lconf now validates the peer before adding it.
4395 Frequency : Only if a default stripe is set on the filesystem root.
4397 Description: Setting a default stripe on the filesystem root prevented the
4398 filesystem from being remounted.
4399 Details : The client was sending extra request flags in the root getattr
4400 request and did not allocate a reply buffer for the dir EA.
4403 Frequency : occasional, higher if lots of files are accessed by one client
4404 Bugzilla : 6159, 6097
4405 Description: Client trips assertion regarding lsm mismatch/magic
4406 Details : While revalidating inodes the VFS looks up inodes with ifind()
4407 and in rare cases can find an inode that is being freed.
4408 The ll_test_inode() code will free the lsm during ifind()
4409 when it finds an existing inode and then the VFS later attaches
4410 this free lsm to a new inode.
4414 Bugzilla : 6422, 7030
4415 Description: MDS deadlock between mkdir and client eviction
4416 Details : Creating a new file via mkdir or mknod (starting a transaction
4417 and getting the ns lock) can deadlock with client eviction
4418 (gets ns lock and trying to finish a synchronous transaction).
4421 Frequency : occasional
4422 Description: While starting a server, the fsfilt_ext3 module could not be
4424 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4425 kernels. Previously, lconf would still use the ext3 name
4426 when trying to load modules. Now, it will correctly use
4427 ext3 on 2.4 and ldiskfs on 2.6.
4429 Severity : enhancement
4430 Description: The default stripe count has been changed to 1
4431 Details : The interpretation of the default stripe count (0, to lfs
4432 or lmc) has been changed to mean striping across a single
4433 OST, rather than all available. For general usage we have
4434 found a stripe count of 1 or 2 works best.
4436 Severity : enhancement
4437 Description: Add support for compiling against Cray portals.
4438 Details : Conditional compiling for some areas that are different
4442 Frequency : occasional
4443 Bugzilla : 6409, 6834
4444 Description: Creating files with an explicit stripe count may lead to
4445 a failed assertion on the MDS
4446 Details : If some OSTs are full or unavailable, creating files may
4447 trigger a failed assertion on the MDS. Now, Lustre will
4448 try to use other servers or return an error to the
4452 Frequency : occasional
4454 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4455 caused lock timeout during readahead (since 1.4.2).
4456 Details : Processes doing readahead might match a lock that hasn't been
4457 granted yet if there are overlapping and conflicting lock
4458 requests. The readahead process waits on ungranted lock
4459 (original lock is CBPENDING), while OST waits for that process
4460 to cancel CBPENDING read lock and eventually evicts client.
4462 Severity : enhancement
4464 Description: Initial enabling of flock support for clients
4465 Details : Implements fcntl advisory locking and file status functions.
4466 This feature is provided as an optional mount flag (default
4467 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4468 locking are implemented yet, and those that are are not guaranteed
4469 to be completely correct in production environments.
4470 mount -t lustre -o [flock|noflock] ...
4473 Frequency : occasional
4475 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4476 assertion in the ext3 JBD (journaling) layer.
4477 Details : The b_committed_data struct is protected by the big kernel lock
4478 in 2.4 kernels, serializing journal_commit_transaction() and
4479 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4480 access to this struct. In 2.6 kernels there is finer grained
4481 locking to improve SMP performance of the JBD layer.
4485 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4487 -----------------------------------------------------------------------------
4489 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4491 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4492 versions of Lustre. Please update all servers and clients to
4493 version 1.4.2 or later at the same time. You must also run
4494 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4495 to update the configuration logs.
4497 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4498 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4499 - swab llog records of type '0' so we get proper header size/idx (5861)
4500 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4501 - fix rename of one directory over another leaking an inode (5953)
4502 - avoid SetPageDirty on 2.6 (5981)
4503 - don't re-add just-being-destroyed locks to the waiting list (5653)
4504 - when creating new directories, inherit the parent's custom
4505 striping settings if present parent (3048)
4506 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4507 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4508 - don't hold client locks on temporary worklist from l_lru (5666)
4509 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4510 - automatically evict dead clients (5921)
4511 - Update file size properly in create+truncate+fstat case (6196)
4512 - Do not unhash mountpoint dentries, do not allow removal of
4514 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4515 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4516 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4517 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4518 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4519 - Correctly handle reads of files with no objects (6243)
4520 - lctl recover will also mark a device active if deactivate used (5933)
4522 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4523 - allow --write-conf on an MDS with different nettype than client (5619)
4524 - don't write config llogs to MDS for mounts not from that MDS (5617)
4525 - lconf should create multiple TCP connections from a client (5201)
4526 - init scripts are now turned off by default; run chkconfig --on
4527 lustre and chkconfig --on lustrefs to use them
4528 - upcalls are no longer needed for clients to recover to failover
4530 - add --abort-recovery option to lconf to abort recovery on device
4532 - add support for an arbitrary number of OSTs (3026)
4533 - Quota support protocol changes.
4534 - forward compatibility changes to wire structs (6007)
4535 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4536 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4537 - improved "lustre status" script
4538 - initialize blocksize for non-regular files (6062)
4539 - added --disable-server and --disable-client configure options (5782)
4540 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4541 - Vanilla 2.4.29 support
4542 - increase maximum number of obd devices to 520 (6242)
4543 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4544 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4545 SLES 9 2.6.5 kernels
4546 - limit stripes per file to 160 (the maximum EA size) (6093)
4548 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4551 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4552 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4553 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4554 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4555 - small ext3 extents cleanups and fixes (5733)
4556 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4557 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4558 - avoid needless client->OST connect, fix handle mismatch (5317)
4559 - fix DLM error path that led to out-of-sync client, long delays (5779)
4560 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4561 - fix several locking issues related to i_size (5492,5624,5654,5672)
4562 - don't move pending lock onto export if it is already evicted (5683)
4563 - fix kernel oops when creating .foo in unlinked directory (5548)
4564 - fix deadlock in obdfilter statistics vs. object create (5811)
4565 - use time_{before,after} to avoid timer jiffies wrap (5882)
4566 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4567 - Do not leak request if server was not able to process it (5154)
4568 - If mds_open unable to find parent dir, make that negative lookup(5154)
4569 - don't create new directories with extent-mapping (5909, 5936)
4571 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4572 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4573 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4574 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4575 - support for RHEL4 kernel on i686 (5773)
4576 - provide error messages when incompatible logs are encountered (5898)
4578 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4579 * version 1.4.0.10 (1.4.1 release candidate 1)
4581 - don't keep a lock reference when lock is not granted (4238)
4582 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4583 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4584 - drop import inflight refcount on signal_completed_replay error (5255)
4585 - unlock page after async write error during send (3677)
4586 - handle missing objects in filter_preprw_read properly (5265)
4587 - no transno return for symlink open, don't save no-trasno open (3440)
4588 - don't try to complete elan receive that already failed (4012)
4589 - free RPC server reply state on error (5406)
4590 - clean up thread from ptlrpc_start_thread() on error (5160)
4591 - readahead could read extra page into cache that wasn't ejected (5388)
4592 - prevent races in class_attach/setup/cleanup/detach (5260)
4593 - don't dereference de->d_inode after l_dput of de (5458)
4594 - use "int" for stripe value returned from lock_to_stripe (5544)
4595 - mballoc allocation and error-checking fixes in 2.6 (5504)
4596 - block device patches to fix I/O request sizes in 2.6 (5482)
4597 - look up hostnames for IB nals (5602)
4598 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4599 - don't start multiple acceptors for the same port (5277)
4600 - fix incorrect LASSERT in mds_getattr_name (5635)
4601 - export a proc file for general "ping" checking (5628)
4602 - fix "lfs check" to not block when the MDS is down (5628)
4604 - service request history (4965)
4605 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4606 - create an "evict_client" /proc entry on OSTs, like the MDS has
4607 - fix mount usage message, return errors per mount(8) (5168)
4608 - change grep [] to grep "[]" in tests so they work in more UMLs
4609 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4610 - remove ancient LOV_MAGIC_V0 EA support (5047)
4611 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4612 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4613 - allow readahead allocations to fail when low on memory (5383)
4614 - mmap locking landed again, after considerable improvement (2828)
4615 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4617 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4620 - send OST transaction number in read/write reply to free req (4966)
4621 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4622 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4623 - the watchdog thread now runs as interruptible (5246)
4624 - flock/lockf fixes (but it's still disabled, pending 5135)
4625 - don't use EXT3 constants in llite code (5094)
4626 - memory shortage at startup could cause assertion (5176)
4628 - reorganization of lov code
4629 - single portals codebase
4631 - add extents/mballoc support (5025)
4632 - direct I/O reads in the obdfilter (4048)
4633 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4635 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4638 - send OST transaction number in read/write reply to free req (4966)
4639 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4640 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4641 - the watchdog thread now runs as interruptible (5246)
4642 - handle missing objects in filter_preprw_read properly (5265)
4643 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4644 - drop import inflight refcount on signal_completed_replay error (5255)
4645 - unlock page after async write error during send (3677)
4646 - return original error code on reconstructed replies (3761)
4647 - no transno return for symlink open, don't save no-trasno open (3440)
4649 - add pid to ldlm debugging output (4922)
4650 - bump the watchdog timeouts -- we can't handle 30sec yet
4651 - extra debugging for orphan dentry/inode bug (5259)
4653 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4656 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4657 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4658 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4659 - fix NULL dereference in /proc/sys/portals/routes (4827)
4660 - allow failed mdc_close() operations to be interrupted (4561)
4661 - stop precreate on OST before MDS would time out on it (4778)
4662 - don't send partial-page writes before EOF from client (4410)
4663 - discard client grant for sub-page writes on large-page clients (4520)
4664 - don't free dentries not owned by NFS code, check generation (4806)
4665 - fix lsm leak if mds_create_objects() fails (4801)
4666 - limit debug_daemon file size, always print CERROR messages (4789)
4667 - use transno after validating reply (3892)
4668 - process timed out requests if import state changes (3754)
4669 - update mtime on OST during writes, return in glimpse (4829)
4670 - add mkfsoptions to LDAP (4679)
4671 - use ->max_readahead method instead of zapping global ra (5039)
4672 - don't interrupt __l_wait_event() during strace
4674 - add software watchdogs to catch hung threads quickly (4941)
4675 - make lustrefs init script start after nfs is mounted
4676 - fix CWARN/ERROR duplication (4930)
4677 - return async write errors to application if possible (2248)
4678 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4679 - print NAL number in %x format (4645)
4680 - update barely-supported suse-2.4.21-171 series (4842)
4681 - support for sles 9 %post scripts
4682 - support for building 2.6 kernel-source packages
4683 - support for sles km_* packages
4685 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4688 - ignore -ENOENT errors in osc_destroy (3639)
4689 - notify osc create thread that OSC is being cleaned up (4600)
4690 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4691 - reconstruct ost_handle() like mds_handle() (4657)
4692 - create a new thread to do import eviction to avoid deadlock (3969)
4693 - let lconf resolve symlinked-to devices (4629)
4694 - don't unlink "objects" from directory with default EA (4554)
4695 - hold socknal file ref over connect in case target is down (4394)
4696 - allow more than 32000 subdirectories in a single directory (3244)
4697 - fix blocks count for O_DIRECT writes (3751)
4698 - OST returns ENOSPC from object create when no space left (4539)
4699 - don't send truncate RPC if file size isn't changing (4410)
4700 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4701 - bind to privileged port in socknal and tcpnal (3689)
4703 - rate limit CERROR/CWARN console message to avoid overload (4519)
4704 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4705 - basic mmap support (3918)
4706 - kernel patch series update from b1_4 (4711)
4708 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4711 - avoid crash during MDS cleanup with OST shut down (2775)
4712 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4713 - don't use bad inodes on the MDS (3744)
4714 - dynamic object preallocation to improve recovery speed (4236)
4715 - don't hold spinlock over lock dumping or change debug flags (4401)
4716 - don't zero obd_dev when it is force cleaned (3651)
4717 - print grants to console if they go negative (4431)
4718 - "lctl deactivate" will stop automatic recovery attempts (3406)
4719 - look for existing locks in ldlm_handle_enqueue() (3764)
4720 - don't resolve lock handle twice in recovery avoiding race (4401)
4721 - revalidate should check working dir is a directory (4134)
4723 - don't always mark "slow" obdfilter messages as errors (4418)
4725 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4728 - don't close LustreDB during write_conf until it is done (3860)
4729 - fix typo in lconf for_each_profile (3821)
4730 - allow dumping logs from multiple threads at one time (3820)
4731 - don't allow multiple threads in OSC recovery (3812)
4732 - fix debug_size parameters (3864)
4733 - fix mds_postrecov to initialize import for llog ctxt (3121)
4734 - replace config semaphore with spinlock (3306)
4735 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4736 - don't allow enqueue to complete on a destroyed export (3822)
4737 - down write_lock before checking llog header bitmap (3825)
4738 - recover from lock replay timeout (3764)
4739 - up llog sem before sending rpc (3652)
4740 - reduce ns lock hold times when setting kms (3267)
4741 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4742 - fix NULL deref and obd_dev leak on setup error (3312)
4743 - replace some LBUG about llog ops with error handling (3841)
4744 - don't match INVALID dentries from d_lookup and spin (3784)
4745 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4746 - fix invalid assertion in ptlrpc_set_wait (3880)
4748 - add libwrap support for the TCP acceptor (3996)
4749 - add /proc/sys/portals/routes for non-root route listing (3994)
4750 - allow setting MDS UUID in .xml (2580)
4751 - print the stack of a process that LBUGs (4228)
4753 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4756 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4757 - make sure to unset replay flag from failed open requests (3440)
4758 - if default stripe count is 0, use OST count for inode size (3636)
4759 - update parent mtime/ctime on client for create/unlink (2611)
4760 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4761 - free recovery state on server during a forced cleanup (3571)
4762 - unregister_reply for resent reqs (3063)
4763 - loop back devices mounting and status check on 2.6 (3563)
4764 - fix resource-creation race that can provoke i_size == 0 (3513)
4765 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4766 - more debugging for page-accounting assertion (3746)
4767 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4768 - avoid deadlock after precreation failure (3758)
4769 - fix race and lock order deadlock in orphan handling (3450, 3750)
4770 - add validity checks when grabbing inodes from l_ast_data (3599)
4772 - add /proc/.../recovery_status to obdfilter (3428)
4773 - lightweight CDEBUG infrastructure, debug daemon (3668)
4774 - change default OSC RPC parameters to be better on small clusters
4775 - turn off OST read cache for files smaller than 32MB
4776 - install man pages and include them in rpms (3100)
4777 - add new init script for (un)mounting lustre filesystems (2593)
4778 - run chkconfig in %post for init scripts (3701)
4779 - drop scimac NAL (unmaintained)
4781 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4784 - clean kiobufs before and after use (3485)
4785 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4786 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4787 - change init script to fail more clearly if not run as root (1528)
4788 - allow clients to reconnect during replay (1742)
4789 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4790 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4791 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4792 - fix cond_resched (3554)
4793 - only evict unfinished clients after recovery (3515)
4794 - allow bulk resend, prevent data loss (3570)
4795 - dynamic ptlrpc request buffer allocation (2102)
4796 - don't allow unlinking open directory if it isn't empty (2904)
4797 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4798 - remove extraneous obd dereference causing LASSERT failure (3334)
4799 - don't use get_cycles() when creating temp. files on the mds (3156)
4800 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4801 - handle EEXIST for set-stripe, set proper directory name (3336)
4803 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4804 - fix ksocknal_fmb_callback() error messages (2918)
4806 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4809 - don't copy lvb into (possibly NULL) reply on error (2983)
4810 - don't deref dentry after dput, don't free lvb on error (2922)
4811 - use the kms to determine writeback rpc length (2947)
4812 - increment oti_logcookies when osc is inactive (2948)
4813 - update client's i_blocks count via lvb messages (2543)
4814 - handle intent open/close of special files properly (1557)
4815 - mount MDS with errors=remount-ro, like obdfilter (2009)
4816 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4817 - don't use cancelling-locks' kms values (2947)
4818 - use highest lock extent for kms, not last one (2925)
4819 - don't dereference ERR_PTR() dentry in error handling path (3107)
4820 - fix thread race in portals_debug_dumplog() (3122)
4821 - create lprocfs device entries at setup instead of at attach (1519)
4822 - common AST error handler, don't evict client on completion race (3145)
4823 - zero nameidata in detach_mnt in 2.6 (3118)
4824 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4825 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4826 - initialize RPC timeout timer earlier for 2.6 (3219)
4827 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4828 - print nal/nid for unknown nid (3258)
4829 - additional checks for oscc recovery before doing precreate (3284)
4830 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4831 - don't crash in mdc_close for bad permissions on open (3285)
4832 - zero i_rdev for non-device files (3147)
4833 - clear page->private before handing to FS, better assertion (3119)
4834 - tune the read pipeline (3236)
4835 - fix incorrect decref of invalidated dentry (2350)
4836 - provide read-ahead stats and refine rpc in flight stats (3328)
4837 - don't hold journal transaction open across create RPC (3313)
4838 - update atime on MDS at close time (3265)
4839 - close LDAP connection when recovering to avoid server load (3315)
4840 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4841 - don't leak open file on MDS after open resend (3325)
4842 - serialize filter_precreate and filter_destroy_precreated (3329)
4843 - loop device shouldn't call sync_dev() for nul device (3092)
4844 - clear page cache after eviction (2766)
4845 - resynchronize MDS->OST in background (2824)
4846 - refuse to mount the same filesystem twice on same mountpoint (3394)
4847 - allow llmount to create routes for mounting behind routers (3320)
4848 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4849 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4850 - fix rare problem with rename on htree directories (3417)
4852 - allow default OST striping configuration per directory (1414)
4853 - fix compilation for qswnal for 2.6 kernels (3125)
4854 - increase maximum number of MDS request buffers for large systems
4855 - change liblustreapi to be useful for external progs like lfsck (3098)
4856 - increase local configuration timeout for slow disks (3353)
4857 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4859 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4862 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4863 - don't overwrite extent policy data in reply if lock was blocked (2901)
4864 - drop filter export grants atomically with removal from device (2663)
4865 - del obd_self_export from work_list in class_disconnect_exports (2908)
4866 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4867 - swab reply message in mdc_close, other PPC fixes (2464)
4868 - fix destroying of named logs (2325)
4869 - overwrite old logs when running lconf --write_conf (2264)
4870 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4871 - fix race in target_handle_connect (2898)
4872 - mds_reint_create() should take same inode create lock (2926)
4873 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4874 - don't close files for self_export to avoid uninitialized obd (2936)
4875 - allow MDS with the same name as client node (2939)
4876 - hold dentry reference for closed log files for unlink (2325)
4877 - reserve space for all logs during transactions (2059)
4878 - don't evict page beyond end of stripe extent (2925)
4879 - don't oops on a deleted current working directory (2399)
4880 - handle hard links to targets without a parent properly (2517)
4881 - don't dereference NULL lock when racing during eviction (2867)
4882 - don't grow lock extents when lots of conflicting locks (2919)
4884 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4887 - account for cache space usage on clients to avoid data loss (974)
4888 - lfsck support in lustre kernel code (2349)
4889 - reduce journal credits needed for BRW writes (2370)
4890 - orphan handling to avoid losing space on client/server crashes
4891 - ptlrpcd can be blocked, stopping ALL progress (2477)
4892 - use lock value blocks to assist in proper KMS, faster stat (1021)
4893 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4894 - recovery for initial connections (2355)
4895 - fixes for mds_cleanup_orphans (1934)
4896 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4897 - block all file creations until orphan recovery completes (1901)
4898 - client remove rq_connection from request struct (2423)
4899 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4900 - recovery timer race (2670)
4901 - mdc_close recovey bug (2532)
4902 - ptlrpc cleanup bug (2710)
4903 - mds timeout on local locks (2588)
4904 - namespace lock held during RPCs (2431)
4905 - handle interrupted sync write properly (2503)
4906 - don't try to handle a message that hasn't been replied to (2699)
4907 - client assert failure during cleanup after abort recovery (2701)
4908 - leak mdc device after failed mount (2712)
4909 - ptlrpc_check_set allows timedout requests to complete (2714)
4910 - wait for inflight reqs when ptlrpcd finishes (2710)
4911 - make sure unregistered services are removed from the srv_list
4912 - reset bulk XID's when resending them (caught by 1138 test)
4913 - unregister_bulk after timeout
4914 - fix lconf error (2694)
4915 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4916 - readahead locks pages, leaves pending causing memory pressure (2673)
4917 - increase OST request buffers to 4096 on large machines (2729)
4918 - fix up permission of existing directories in simple_mkdir (2661)
4919 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4920 - don't assign transno to errored transactions (2742)
4921 - don't delete objects on OST if given a bogus objid from MDS (2751)
4922 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4923 - if rq_no_resend, then timeout request after recovery (2432)
4924 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4925 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4926 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4927 - MDS recovery shouldn't delete objects that it has given out (2730)
4928 - if enqueue arrives after completion, don't clobber LVB (2819)
4929 - don't unlock pages twice when trigger_group_io returns error (2814)
4930 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4931 - don't write pages to disk if there was an error (1450)
4932 - don't ping imports that have recovery disabled (2676)
4933 - take buffered bytes into account when balancing socknal conn (2817)
4934 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4935 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4936 - remove little-endian swabbing of llog records (1987)
4937 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4938 - retry reposting request buffers if they fail (1191)
4939 - grow extent at grant time to avoid granting a revoked lock (2809)
4940 - lock revoke doesn't evict page if covered by a second lock (2765)
4941 - disable VM readahead to avoid reading outside lock extents (2805)
4943 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4944 - updated kernel patches for hp-2.4.20 kernel (2681)
4946 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4949 - fix truncated write corruption (2366)
4950 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4952 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4953 - ensure len(uuid) < 37 in lmc (1171)
4954 - fix ia64 OOPS in llog_test (2255)
4955 - zero end of page at obdfilter for partial page writes (2648)
4956 - don't leave stale dentries around after renames (bug 2428)
4957 - fix timeouts when evicting a client with a single lock held (2642)
4958 - set deadline for the initial HELLO message to drain (2634)
4959 - print out dotted-quad IP addresses in the socknal (2302)
4961 - additional debugging for MDS client eviction problem (2443)
4962 - fix mkfsoptions support for osts (2603, 2604)
4964 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4967 - add series for the vanilla 2.6.0 kernel
4968 - add series for the vanilla 2.4.24 kernel
4969 - add series for a cray x86/64 UL kernel drop
4970 - fix xattr patches for the vanilla 2.4.19 series
4972 - generate true UUIDs in lmc (1171)
4973 - have portals stack dumping break in UML (2466)
4974 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4975 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4976 - wake callback waiting threads on client eviction (2460)
4977 - Add --ptldebug and --subsystem to lmc (1719)
4978 - update assertion to allow safe interrupt allocation
4979 - set rq_no_resend for cancel requests (2432)
4980 - recalculate ptlrpcd timeout after resend (2494)
4981 - call vfs_rmdir when removing pending directories (2368)
4982 - fix renaming a file to itself (2429)
4983 - lmc creates a default one-stripe lov (2454)
4984 - expand procfs space to handle large clusters (2326)
4985 - increase UML stack to avoid overflow
4986 - update lconf's list of debug and subsystem masks
4987 - fix lfs find --obd (2510)
4988 - /proc tunable for disabling filter read caching (2591)
4989 - stop rpm packages from altering slapd.conf (2301)
4990 - disable nagle in the socknal under 0conf (2578)
4991 - choose mds inode size based on stripe count (2572)
4992 - fix kernel-source rpm problems (2516)
4994 - add --disable-doc to avoid pdf generation (2421)
4995 - update documentation, tests, type-os, comments
4996 - avoid format warnings on ia64
4997 - remove the TOE NAL
4998 - tiny code cleanups by removing unused fields
5000 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5003 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5004 - remove the most blatant lies from BUILDING file (2371)
5005 - change default debug level to reasonable production setting
5006 - reduce client side cache size to reduce cache flush time
5007 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5008 - make TCP zerocopy and pinger support enabled by default (2476)
5009 - sync writes completed after process exits caused crashes (2319)
5010 - maintain correct mount count on the MDS (2356)
5011 - backout 1557, because 2316 wasn't really fixed
5012 - better file I/O statistics gathering in /proc
5013 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5014 - another kernel patch to fix zero-copy TCP function export
5015 - don't take duplicate lock when processing re-sent getattr (2420)
5016 - lctl uses obd_self_export instead of creating new conn (2353)
5017 - MDS/OST recovery case which requires object creation asserted (2425)
5018 - move lfs from /usr/sbin to /usr/bin in packages
5019 - fix race between mds_client_add and mds_client_free (2417)
5020 - use kmalloc instead of slabs in portals (2430)
5021 - don't create duplicate records when a failover MDS is present (2442)
5022 - remove unnecessary mount age check (2332)
5023 - don't remove directory inodes from locks prematurely (2451)
5024 - don't break if MDS service name is the same as hostname (2103)
5025 - fix races in client write RPC generation when cache full (2482)
5027 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5030 - remove now-unused request->rq_obd (278)
5031 - if an allocation fails, print out how much memory we've used (1933)
5032 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5033 - add the "configurable stack size" patch to most series files (1256)
5034 - ability to write large log records, for 100+ OST configs (2306)
5035 - fix NULL deref when filter_prep fails (2314)
5036 - fix operator precedence error in filter_sync
5037 - dynamic allocation of socknal TX descriptors (2315)
5038 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5039 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5040 - ensure that CWARN messages in Portals always get to the syslog
5041 - __init/__exit are not for prototype decls (ldlm_init/exit)
5042 - x86-64 compile warning fixes
5043 - fix gateway LMC keyword conflict (2318)
5044 - fix MDS lock inversions in getattr/reint paths (1844)
5045 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5046 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5047 - DLM race condition prevented some lock evictions (2328)
5048 - ENOMEM detection and retry on socknal sends (2230)
5049 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5050 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5051 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5052 - better error messages when a client is rejected during recovery (1505)
5053 - avoid cancelling locks which were never granted, after failure (2330)
5054 - fix i_sem/journal inversion in mds_client_add (2333)
5055 - fix truncate/getattr lock cycle deadlock (2334)
5056 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5057 - fix two rare exit paths which could leak an l_lock() ref (2321)
5058 - fencepost error in MDS/OST orphan recovery (2226)
5059 - make log record alignment 8 bytes (1988)
5060 - lstripe now fails when requested offset > ost_count (2237)
5061 - ensure that all kernel series have a complete list.h (1607)
5062 - fix crashes in special-file operations (2316)
5063 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5065 - allow configurable automake binary, for testing new versions
5066 - small update to the lfs documentation
5068 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5070 * fix negative export reference count in fsfilt_sync (2312)
5072 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5073 * release candidate 0.9.1
5075 - orphans are moved into the PENDING directory for possible recovery
5076 - replayed opens now open by fid for orphan/rename safety (1042)
5077 - last close of an orphan inode generates a transno (683)
5078 - chdir() and mount() now pin the directory entry (1020)
5079 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5080 - discard very old requests without processing them (1502)
5081 - remove some common, well-understood CERRORs (1505)
5082 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5083 - clear "grant" flags in OST replies until OST grant code lands (1644)
5084 - fix read performance by not clobbering i_blksize on client (1598)
5085 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5086 - make LRU size a /proc tunable, clears locks when reduced (707)
5087 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5088 - server locks take references on exports now (1558)
5089 - build fixes for 2.4.20-rh trees (1663)
5090 - return an error from lov_create if all OSCs are inactive (1751)
5091 - fix import levels when a reconnect happens without a timeout (1597)
5092 - exit early from mds_open if we get a lookup error (1749)
5093 - partial page read at EOF wouldn't wait for disk before sending (1642)
5094 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5095 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5096 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5097 - write/truncate lock inversion (1639)
5098 - Don't auto-load obdclass, portals modules during cleanup (1495)
5099 - fix timestamps from jumping to "now" (1763)
5100 - extra journal assertions (1648)
5101 - add an extra multiunlink test (1771)
5102 - fix read_record/write_record API (1776)
5103 - fix leak of offset_extent, possible incorrect i_size later (1772)
5104 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5105 - Debugging for the kqswnal_get_idle_tx problems (1820)
5106 - Allow recovery to be attempted multiple times (1536)
5107 - Write out MDS last_rcvd file after it is first created (1600)
5108 - Fix tx_descriptor leak in failed transmit situations (1827)
5109 - ext3 journaling fixes for assertion failure after IO error (1871)
5110 - class_export_put() on freed export after completion AST error (1896)
5111 - Fix revalidate looping in VFS (1322)
5112 - Don't access a freed export during MDS_REINT timeout (1521)
5113 - Add open-unlink recovery support on the MDS (1673,1764)
5114 - Return an error if no MDS data was read from last_rcvd (1946)
5115 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5116 - Don't setup a disk device that doesn't match exported UUID (317)
5117 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5118 - avoid committing NULL handle in force close
5119 - local.sh is now a one-stripe LOV configuration
5120 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5121 - fix ext3 htree duplicate directory entry corruption (1516)
5122 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5123 - update to new LOV EA format (2097)
5124 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5126 2003-06-15 Phil Schwan <phil@clusterfs.com>
5129 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5130 - per-import/export recovery handling (958, 931, 959)
5131 - multiple last-rcvd slots, for serving multiple FSes (949)
5132 - connections are again shared between multiple imp/exports (963, 964)
5133 - "umount -f" would hang if any requests needed to be sent (393, 978)
5134 - avoid pinning large req buffer by copying for queued messages (989)
5135 - add "uuid" to "lctl device" command to help upcalls (991)
5136 - "open" RPCs with transnos would confuse recovery counters (1037)
5137 - do proper endian conversion of all wire messages (288, 340, 891)
5138 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5139 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5140 - fix ptlrpc_connection leak in target_handle_connect (1174)
5141 - fix import refcounting bug in OST and MDS cleanup (1134)
5142 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5143 - fix very unlikely obd_types race condition (501)
5144 - remove osc_open hack for echo_client (1187)
5145 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5146 - a failure in read_inode2 leads to deadlock (1139)
5147 - cancel ack-locks as soon as transaction is committed (1072)
5148 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5149 - make sure to commitrw after any preprw to avoid deadlock (1162)
5150 - failing to execute a file in a lustre FS would lock inode (1203)
5151 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5152 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5153 - obd_uuid2tgt would walk off the end of the list (1255)
5154 - on IA64 the osc would give portals incorrect bulk size (1258)
5155 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5156 - fix lock inversion caused by new llite matching code (1282)
5157 - limit the number of dirty pages on a client to 10MB (1286)
5158 - timed out locks were not being corrected cancelled (1289)
5159 - fix O_DIRECT above 4GB on IA-32 (1292)
5160 * major user-visible changes
5161 - fail out/fail over policy now controlled by the upcall (993)
5163 - add OBD_PING to check server availability and failure (954)
5164 - lustre messages are now sent in sending host order (288, 340, 891)
5165 - add eadatalen to MDS getattr reply (340)
5166 - OST read replies may contain second buffer, with per-page status (593)
5168 2003-03-11 Phil Schwan <phil@clusterfs.com>
5171 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5172 - failed open()s could cause deadlock; fixed (867, 869)
5173 - stop cancelling OST locks when files are closed (481)
5174 - overlapping XID spaces caused network corruption (851, 853)
5175 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5176 - setattr_raw added, to do single-RPC, server-side setattrs
5177 - lmc/lconf syntax change for OST UUIDs
5178 - fix crashy race condition between ptlrpc_free_req and osc_close
5179 - don't use request in mdc_enqueue if we hit a timeout (889)
5180 - don't set the inode i_size for regular files from the MDS (896)
5181 - handle out of order completion AST (842)
5182 - don't LBUG if a lock request times out after receiving AST (913)
5183 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5184 - if a bad lock AST arrives, send an error instead of dropping entirely
5185 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5186 - fix leak in bulk IO when only partially completed (899, 900, 926)
5187 - fix O_DIRECT for ia64 (55)
5188 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5189 - C-z after timeout could hang a process forever; fixed (977)
5191 - client-side I/O cache (678, 924, 929, 941, 970)
5193 - READPAGE and SETATTRs which don't take server-side locks get
5196 2003-02-11 Phil Schwan <phil@clusterfs.com>
5199 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5200 - Fix fsfilt_extN_readpage() to read a full page of directory
5201 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5202 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5203 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5204 - Most intent operations were removed, in favour of a new RPC mode
5205 that does a single RPC to the server and bypasses most of the VFS
5206 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5207 - Aggressively cancel local locks on DLM servers
5208 - mds_reint_unlink sends EA to the client if it's the last nlink.
5209 client uses that EA to unlink OST objects.
5210 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5211 - recursive symlinks were fixed (439)
5212 - fixed NULL deref in DEBUG_REQ
5213 - filter_update_lastobjid no longer calls sync, which annoyed extN
5214 - fixed multi-client small-writes to a single file problem (445)
5215 - fixed mtime updates during file writes (607)
5216 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5217 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5218 - fixed Linux OST size reporting problem (444, 656)
5219 - OST now updates object mtime with writes or setattr (607, 619)
5220 - client verifies file size before zeroing page past EOF (445)
5221 - OST now writes last allocated objid to disk with allocation (108)
5222 - LOV on echo now works (409)
5224 - mds_reint_unlink sends a new buffer, with the EA included. this
5225 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5226 set if a regular file was being unlinked, and it was the last link
5227 - use PtlGet from the target for bulk writes (315)
5228 - OST now updates object mtime with writes or setattr (607, 619)
5229 - LDLM now has a grant-time callback to revalidate locked items, if
5231 - Many MDS operations were reorganized to combat race conditions
5233 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5234 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5236 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5239 - Fully reactivate OST imports after reconnection (512, others)
5240 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5241 - More graceful error handling for truncating on dead OST (515)
5242 - Don't error out unless we're actually accessing dead stripes (474)
5243 - Fix garbage sizes when stripes are missing (410)
5244 - LRU counters were broken, causing constant lock purge (433, 432)
5245 - garbage on read from stripes with failed OSTs (441)
5246 - mark OSCs as active before reconnecting during recovery (438)
5247 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5248 - lfind did not preserve OST order in output (443)
5249 - symlinks cause hung clients, incorrect data (439)
5250 - stop dereferencing request after dropping refcount (457)
5251 - don't LASSERT(spin_is_locked) on non-SMP (455)
5252 - fixes for many rename() bugs
5253 - fstat didn't correctly synchronize attributes (399)
5254 - server must handle lock cancellation during blocking AST prep (487)
5255 - bulk descriptors were free()d too soon (511)
5256 - fix paths in lconf, which would load incorrect modules (451, 507)
5257 - fix confusing lconf 'host not found' error message (386)
5258 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5259 - fix race condition in mdc_blocking_ast() for inode access (526)
5260 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5261 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5262 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5263 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5264 - fix lstripe in several ways (kernel side) (527)
5265 - fix request leak in ldlm_cli_enqueue (262)
5266 - incorrect OSC was marked inactive after OST failure
5267 - call mds_fs_cleanup before unmounting filesystem (524)
5268 - fix races between taking ns_lock and ldlm_lock_change_resource
5269 - fix races updating LOV export open file list
5270 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5271 - fix recovery NULL deref in ldlm_cli_cancel_unused
5272 - fix some DLM races by using new hash table for lock handles (419)
5273 - permit the client to specify desired inodes, at replay
5274 - duplicate requests when we queue them for replay reintegration
5275 - fix last_rcvd offset calculation
5276 - sync after each recovered transaction, so we always make progress
5277 - never, not always, ERESTART requests without transnos
5278 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5280 - skip replay if the MDS says that the client is already connected
5281 - don't check for a recovery-enabled export to match lctl's UUID
5282 - don't INC_USE_COUNT for phantom exports
5283 - don't crash when cleaning up phantom exports (567)
5284 - don't double-finish or set replay data for errored mdc_open requests
5285 - abort requests when they time out, so we don't get old replies
5286 - send/receive replies for AST messages again
5287 - if the client says that it doesn't have the lock, cancel it on the
5289 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5290 mark it as destroyed, it will all work out when decref is called
5291 - fix module use counts (22, 581)
5293 - ASTs now expect a reply (server cancels lock on error reply)
5295 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5298 - fix many simultaneous client startup (392)
5299 - fix dentry->d_it clobbering
5300 - credentials weren't being shipped for readdir/getattr operations
5301 - remove invalid assertions triggered during some concurrent MD
5303 - proper Lustre versions added (336, 389)
5304 - fix memory leak for create error case (398)
5305 - fix LOV locking bug that would get cli/srv out of sync
5306 - fix echo client over LOV (409)
5307 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5308 - fix double-O_EXCL intent crash (424)
5309 - avoid sending multiple lock CANCELs (352)
5311 - MDS can do multi-client recovery (modulo bugs in new code)
5313 - many updates, edits, cleanups
5315 2002-11-18 Phil Schwan <phil@clusterfs.com>
5318 - fix null d_it dereference (346)
5319 - fix full OST/dbench hang (333)
5320 - fix permission problem with file removal (286)
5321 - fix removal of OSCs from LOV when they fail
5322 - fix NULL deref during bulk timeout (214)
5323 - fix problems related to multiple filesystems on one MDS (241)
5324 - fixed serious subtle metadata locking bugs
5325 - free locks on clients when inodes are removed due to memory
5327 - fix inode pointer in lock data (285)
5328 - partial support for multiple MDS on a single host (241)
5329 - data locks weren't cancelled at clear_inode time (290, 311)
5330 - intent locks could lead to unbounded lock growth (205)
5331 - added a maximum lock count, an LRU list, and a flusher
5332 - fix multiple rename (365)
5333 - properly abstracted the echo client
5334 - OSC locked 1 byte too many; fixed
5335 - rewrote brw callback code:
5336 - fixed recovery bugs related to LOVs (306)
5337 - fixed too-many-pages-in-one-write crash (191)
5338 - fixed (again) crash in sync_io_timeout (214)
5339 - probably fixed callback-related race (385)
5341 - Add capability to MDS protocol
5342 - LDLM cancellations and callbacks on different portals
5344 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5347 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5348 - defer open object destruction to close time (601981, 138)
5349 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5350 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5351 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5352 - don't call obd_setattr in ll_file_release for destroyed objects
5353 * protocol change to lustre_msg: move |version| and add |flags|
5354 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5355 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5356 * added replay of create, unlink, link and rename operations during
5357 MDS failover; recovery should be much more robust now
5358 * remove failed OSCs from LOVs (only lov_create uses this so far)
5359 * the lustre-HOWTO was brought (more) up to date (582544)
5361 2002-10-23 Phil Schwan <phil@clusterfs.com>
5364 - in-use dentries weren't being reused properly (617851)
5365 - prevent multiple LDLM setup (599178)
5366 - fix LOV size calculations for truncate (617853)
5367 - fix client handling of MDS intent errors (POSIX)
5368 - fix permission bug in lovstripe.c test (624321)
5369 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5370 - truncate past end of file could corrupt data
5371 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5372 - a race in recovery could return ETIMEDOUT to apps (623947)
5373 - building outside the source directory was fixed
5374 * the lustre-HOWTO was brought (more) up to date (582544)
5375 * major progress was made on recovery functionality
5377 2002-10-10 Phil Schwan <phil@clusterfs.com>
5380 - recovery deadlock fix
5381 - rm -rf causes LBUG fix (617817)
5382 - file open by multiple tasks fix (618962)
5383 - directory permissions bugs (602707 and 620007)
5384 - journal_stop fixed with locking (611313)
5385 - O_APPEND failures resolved (618273, perhaps 614459)
5386 - lconf PATH fix (619770)
5387 - IA64 build fix (621450)
5388 - RPC buffer sizes scale with amount of memory
5390 2002-10-01 Phil Schwan <phil@clusterfs.com>
5393 - locks would be cancelled without throwing away data pages,
5394 resulting in inconsistent data (605627)
5395 - inode attributes were not always being refreshed (605627, 612449)
5396 - lconf now continues to cleanup after lctl reports an error
5397 - MDS now enforces user permissions (602707)
5398 - lprocfs cleanup fixed, but not yet enabled (614157)
5399 - fixed infinite server hang, should a client not respond to an AST
5400 - avoid going into recovery if user calls readlink() with a buffer
5401 that's too small (613941)
5402 - AST RPCs no longer require replies (614867) -- this may be changed
5403 - don't crash server if client sends an IOV that's too big (611336)
5404 - fixed lock conversion deadlock (611892)
5405 - fixed the following of symlinks (614622)
5406 * recovery: the server can remove locks from a client that dies, other
5407 clients can make progress
5408 * more extN patch fixes
5409 * compile-time configurable ptlrpc buffer allocations
5411 - collaborative read cache document
5412 - Lustre Lite Performance CDR document-in-progress
5414 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5417 - fix typo in patch-2.4.18
5419 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5422 - clear ptlrpc request each time in handle_incoming_request()
5423 - unlink of files now destroys the object on the OST
5425 2002-09-19 Peter Braam <braam@clusterfs.com>
5427 * add hard link support
5428 * change obdfile creation method
5429 * kernel patch changed
5431 2002-09-19 Peter Braam <braam@clusterfs.com>
5434 - stack overflow bug in extN fixed
5436 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5438 * documentation updates
5439 - add man pages for config tools
5440 - update tests/README to describe testing with new config tools
5441 - finish metadata API descriptions
5442 * bug fixes and cleanups
5443 - statfs workaround for 16TB limit
5444 - LOV stripe allocation improved, can stripe on subset of OSTs
5445 - LOV file size/IO offset was wrong for files > 4GB in size
5446 - object EA data was being dropped, caused files to be unreadable
5447 - memory overflow with non-LOV OST caused memory corruption
5448 - fixed regression tests to work with new config tools, obdfilter
5449 - fixed bug when directory size became larger than 1 block
5450 - fixed bug (for single client case) when PWD was deleted
5451 - invalidate local directory pages when doing intent-based ops
5452 - avoid LDLM oops when lock callback contained bad data
5454 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5456 * documentation updates
5457 * bug fixes and cleanups
5458 - configuration tools
5461 - 64-bit compile warnings
5462 - 64-bit internal statfs data
5464 * test_brw on persistent OST devices
5466 * lprocfs (disabled)
5468 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5470 * documentation updates
5471 * bug fixes and cleanups
5472 * configuration tools
5474 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5477 * many small fixes to 0_5_4
5478 * io/network handling
5479 * thinkos in MDS operations
5481 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5484 * crucial basic fixes to 0.5.3
5485 * IOR, Iozone work over Elan
5488 2002-08-07 Phil Schwan <phil@clusterfs.com>
5489 * version 0_5_3, our first alpha
5490 * we use the new Portals iovs
5491 * documentation updates
5492 * bug fixes and cleanups
5493 * small changes in the DLM wire protocol
5495 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5496 * version 0_5_1 with some initial stability,
5497 * locking on MD and file I/O.
5498 * documentation updates
5499 * several bug fixes since 0.5.0
5500 * small changes in wire protocol
5502 2002-07-18 Phil Schwan <phil@clusterfs.com>
5504 * delivered as Lustre Light Alpha
5505 * fixed a crash after handling invalid MDS requests
5506 * fixed directory pages for architectures with non-4k pages sizes
5508 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5509 * release version v0_4_4
5510 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5512 * Fixes a number of interruption problems with OST operations.
5513 * Update documentation for portals header changes
5514 * Move all wire protocol structs/defines to lustre_idl.h
5515 * Fixes symlink length bug.
5516 * Add tcpdump to repository.
5518 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5519 * release version v0_4_3
5520 * Fixes statfs for inodes on extN.
5521 * Fixes bug in runtests which would delete /etc/hosts.
5522 * Use 64-bit object IDs wherever possible (not into VFS though)
5523 Remove ost_get_info, which is unused by lustre, and out of date.
5525 2002-07-03 Peter Braam <braam@clusterfs.com>
5526 * release version v0_4_2 Fixes a lookup error (type not passed)
5527 * move forward to head of Portals
5528 * move forward to latest Lustre kernel
5530 2002-06-25 Peter Braam <braam@clusterfs.com>
5531 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob