1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.33 (SLES 10),
5 2.6.18-92.1.22.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.40.11-sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
14 * File join has been disabled in this release, refer to Bugzilla 16929.
17 Frequency : always with long access acl
19 Descriptoin: mds can't pack reply with long acl.
20 Details : mds don't control size of acl but they limited by reint/getattr
23 Severity : enhancement
25 Description: Update to SLES10 kernel-2.6.16.60-0.33.
27 Severity : enhancement
29 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
32 Frequency : start MDS on uncleanly shutdowned MDS device
34 Descriptoin: aborting recovery hang on MDS
35 Details : don't throttle destroy RPCs for the MDT.
38 Frequency : on remount
40 Description: external journal device not working after the remount
41 Details : clear dev_rdonly flag for external journal devices in
45 Frequency : race on file read and write
47 Description: Lustre doesn't delete files
48 Details : Clients drop lock reference and release openhandle when they find
54 Description: shutdown vs evict race
55 Details : client_disconnect_export vs connect request race.
56 if client will evicted at this time - we start invalidate
57 thread without referece to import and import can be freed
61 Frequency : rare, need acl's on inode.
63 Description: client can't handle ost additional correctly
64 Details : if ost was added after client connected to mds client can have
65 hit lnet_try_match_md ... to big messages to wide striped files.
66 in this case need teach client to handle config events about add
67 lov target and update client max ea size at that event.
69 Severity : enhancement
71 Description: Changelogs
72 Details : Changelogs are a lightweight mechanism to track filesystem
73 metadata and namespace changes. The changelog is recorded
74 permanently on the MDTs, and is periodically "consumed" / purged
75 when records are no longer needed.
77 Severity : enhancement
79 Description: compact fld format with extents
80 Details : Store range of seq rather than every seq in FLD. Seq
81 controller update FLD rather than clients. In Case of CMD, mdt0
82 has FLD, all other metadata server act as non persistent proxy
83 for FLD queries and cache fld entries in fld cache.
88 Description: don't skip ost target if they assigned to file
89 Details : Drop slow OSCs if we can, but not for requested start idx.
90 This means "if OSC is slow and it is not the requested
91 start OST, then it can be skipped, otherwise skip it only
92 if it is inactive/recovering/out-of-space.
96 Description: more cleanup in mds_lov
97 Details : not send LOV EA under replay, we can't know about they size at this
98 time. Don't allow client connect to mds before any ost connected,
99 for avoid problems with LOV EA size and returning EIO to client.
101 Severity : enhancement
103 Description: Interoperability at server side (Disk interoperability)
105 Severity : enhancement
107 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
109 Severity : enhancement
111 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
113 Severity : enhancement
115 Description: New client IO stack (CLIO).
117 Severity : enhancement
119 Description: Commit on sharing. Eliminate inter-client dependencies between
120 uncommitted transactions by doing transaction commits.
121 Thereby clients may recovery independently.
124 Frequency : Create a symlink file with a very long name
126 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
127 Details : If there is no extra space in the request for early cancels,
128 ldlm_req_handles_avail() returns 0 instead of a negative value.
130 Severity : enhancement
132 Description: Add /proc entry for import status
133 Details : The mdc, osc, and mgc import directories now have
134 an import directory that contains useful import data for debugging
137 Severity : enhancement
139 Description: Re-disable certain /proc logging
140 Details : Enable and disable client's offset_stats, extents_stats and
141 extents_stats_per_process stats logging on the fly.
144 Frequency : Only on FC kernels 2.6.22+
146 Description: oops in statahead
147 Details : Do not drop reference count for the dentry from VFS when lookup,
148 VFS will do that by itself.
150 Severity : enhancement
152 Description: Generic /proc file permissions
153 Details : Set /Proc file permissions in a more generic way to enable non-
154 root users operate on some /proc files.
158 Description: Hitting mdc_commit_close() ASSERTION
159 Details : Properly handle request reference release in
160 ll_release_openhandle().
164 Description: quota recovery deadlock during mds failover
165 Details : This patch includes att18982, att18236, att18237 in bz14840.
167 1. fix osts hang when mds does failover with quotaon
168 2. prevent watchdog storm when osts threads wait for the
173 Frequency : only patchless client
174 Description: add workaround for race between add/remove dentry from hash
176 Severity : enhancement
178 Description: Allow OST glimpses to return PW locks
182 Description: LBUG when llog conf file is full
183 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
187 Description: Prevent import from entering FULL state when server in recovery
191 Description: service mount cannot take device name with ":"
192 Details : Only when device name contains ":/" will mount treat it as
198 Description: replace ptlrpcd with the statahead thread to interpret the async
199 statahead RPC callback
203 Frequency : on recovery
204 Description: I/O failures after umount during fail back
205 Details : if client reconnected to restarted server we need join to recovery
206 instead of find server handler is changed and process self
207 eviction with cancel all locks.
209 Severity : enhancement
211 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
213 Severity : enhancement
215 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
217 Severity : enhancement
219 Description: Upcall on Lustre log has been dumped
220 Details : Allow for a user mode script to be called once a Lustre log has
221 been dumped. It passes the filename of the dumped log to the
222 script, the location of the script can be specified via
223 /proc/sys/lnet/debug_log_upcall.
228 Description: avoid idr_remove called for id which is not allocated.
229 Details : Move assigment s_dev for clustered nfs to end of initialization,
230 to avoid problem with error handling.
235 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
236 Details : When connection is reused this not moved from CONN_UNUSED_HASH
237 into CONN_USED_HASH and this prodice warning when put connection
238 again in unused hash.
241 Severity : enhancement
243 Description: File striping can now be set to use an arbitrary pool of OSTs.
245 Severity : enhancement
247 Description: Export bytes_read/bytes_write count on OSC/OST.
251 Description: Early reply size mismatch, MGC loses connection
252 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
253 the connect flags are properly negotiated.
258 Description: quotas are not honored with O_DIRECT
259 Details : all writes with the flag O_DIRECT will use grants which leads to
260 this problem. Now using OBD_BRW_SYNC to guard this.
264 Description: add quota statistics
265 Details : 1. sort out quota proc entries and proc code.
266 2. add quota statistics
268 Severity : enhancement
270 Description: enable quota support for HEAD.
274 Description: Properly propagate oinfo flags from lov to osc for statfs
275 Details : restore missing copy oi_flags to lov requests.
277 Severity : enhancement
279 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
280 : l_getgroups(8), lst(8), routerstat(8)
281 Details : included man pages for llobdstat(8), llstat(8),
282 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
284 Severity : enhancement
286 Description: configure's --enable-quota should check the
287 : kernel .config for CONFIG_QUOTA
288 Details : configure is terminated if --enable-quota is passed but
289 : no quota support is in kernel
293 Description: Remove portals compatibility
294 Details : Remove portals compatibility, not interoperable with releases
299 Description: Resolve device initialization race
300 Details : Prevent proc handler from accessing devices added to the
301 obd_devs array but yet be intialized.
303 Severity : enhancement
305 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
307 Severity : enhancement
309 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
314 Description: Using wrong pointer in osc_brw_prep_request
315 Details : Access to array[-1] can produce panic if kernel compiled with
316 CONFIG_PAGE_ALLOC enabled
320 Description: Client runs out of low memory
321 Details : Consider only lowmem when counting initial number of llap pages
325 Description: *optional* service tags registration
326 Details : if the "service tags" package is installed on a Lustre node
327 When the filesystem is mounted, a local-node service tag will
328 be created. See http://inventory.sun.com/ for more information
329 about the Service Tags asset management system.
333 Description: Kernel BUG tries to release flock
334 Details : Lustre does not destroy flock lock before last reference goes
335 away. So always drop flock locks when client is evicted and
336 perform unlock regardless of successfulness of speaking to MDS.
340 Description: add refcount protection for osc callbacks, avoid panic on shutdown
344 Description: sanity test 65a fails if stripecount of -1 is set
345 Details : handle -1 striping on filesystem in ll_dirstripe_verify
350 Description: ASSERTION(CheckWriteback(page,cmd)) failed
351 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
354 Severity : enhancement
356 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
360 Description: do not process already freed flock
361 Details : flock can possibly be freed by another thread before it reaches
362 to ldlm_flock_completion_ast.
366 Description: LBUG during stress test
367 Details : Need properly lock accesses the flock deadlock detection list.
371 Description: oops in page fault handler
372 Details : kernel page fault handler can return two special 'pages' in error
373 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
377 Description: timeout with invalidate import.
378 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
379 should be handled by ptlrpcd. This produce long age waiting and
380 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
382 Severity : enhancement
384 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
388 Description: enable MGS and MDT services start separately
389 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
390 a co-located MGS without starting the MGS, which is a complement
391 to 'nosvc' mount option.
395 Frequency : after recovery
396 Description: precreate to many object's after del orphan.
397 Details : del orphan st in oscc last_id == next_id and this triger growing
398 count of precreated objects. Set flag LOW to skip increase count
399 of precreated objects.
403 Frequency : rare, on clear nid stats
404 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
405 Details : when clean nid stats sometimes try destroy live entry,
406 and this produce panic in free.
410 Description: Stack overflow during MDS log replay
411 ease stack pressure by using a thread dealing llog_process.
415 Description: wait until IO finished before start new when do lock cancel.
416 Details : VM protocol want old IO finished before start new, in this case
417 need wait until PG_writeback is cleared until check dirty flag and
418 call writepages in lock cancel callback.
420 Severity : enhancement
422 Description: using special macro for print time and cleanup in includes.
426 Description: mds_mfd_close() ASSERTION(rc == 0)
427 Details : In mds_mfd_close(), we need protect inode's writecount change
428 within its orphan write semaphore to prevent possible races.
432 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
436 Frequency : rare, on shutdown ost
437 Description: don't hit live lock with umount ost.
438 Details : shrink_dcache_parent can be in long loop with destroy dentries,
439 use shrink_dcache_sb instead.
443 Description: don't panic with use echo client
444 Details : echo client pass NULL as client nid pointer and this produce null
449 Description: fix build on ppc32
450 Details : compile code with -m64 flag produce wrong object file for ppc32.
454 Description: add message levels for liblustreapi
458 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
459 Details : Move the 'good_osts' check before the 'total_bavail' check. This
460 will result in an -EAGAIN and in the exit call path we call
461 alloc_rr() which will with increasing aggressiveness attempt to
462 aquire precreated objects on the minimum number of required OSCs.
466 Description: Use old size assignment to avoid deadlock
467 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
468 the scheduling while holding a spinlock. We do not need locking
469 for size in ll_update_inode() because size is only updated from
470 the MDS for directories or files without objects, so there is no
471 other place to do the update, and concurrent access to such inodes
472 are protected by the inode lock.
476 Description: resolve "_IOWR redefined" build error on SLES10
480 Description: dump the memory debugging after all modules are unloaded to
481 suppress false negative in conf_sanity test 39
483 Severity : enhancement
485 Description: build kernel-ib packages for OFED 1.3 in our release cycle
490 Description: fix SLES kernel versioning
491 Details : the kernel version for our SLES 10 kernel did not include a "-"
492 before the "smp" at the end. while this was not a problem in
493 general, it did mean that software trying to use the kernel
494 version to try to detect a vendor specific kernel would fail.
495 this was most evident by the OFED build scripts.
499 Description: Don't update lov_desc members until making sure they are valid
500 Details : When updating lov_desc members via proc fs, need fix their
501 validities before doing the real update.
505 Description: don't put request into delay list while invalidate in flight.
506 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
507 import in flight. this produce timeout for invalidate and sometimes
508 can cause stale data.
512 Frequency : on ppc only
513 Description: not convert ost objects for directory because it's not exist.
514 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
518 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
520 Severity : enhancement
522 Description: Update to support 2.6.22.14 vanilla kernel.
526 Frequency : rare, on recovery
527 Description: read procfs can produce deadlock in some situation
528 Details : Holding lprocfs lock which send rpc can produce block for destroy
529 obd objects and this also block reconnect with -EALREADY. This
530 isn't fix all lprocfs bugs - but make it rare.
532 Severity : enhancement
534 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
537 Frequency : frequent on X2 node
539 Description: mdc_set_open_replay_data LBUG
540 Details : Set replay data for requests that are eligible for replay.
544 Description: lustre_mgs: operation 101 on unconnected MGS
545 Details : When MGC is disconnected from MGS long enough, MGS will evict the
546 MGC, and late on MGC cannot successfully connect to MGS and a lot
547 of the error messages complaining that MGS is not connected.
551 Frequency : on network error
552 Description: panic with double free request if network error
553 Details : mdc_finish_enqueue is finish request if any network error ocuring,
554 but it's true only for synchronus enqueue, for async enqueue
555 (via ptlrpcd) this incorrect and ptlrpcd want finish request
558 Severity : enhancement
560 Description: client-side metadata stat-ahead during readdir(directory readahead)
561 Details : perform client-side metadata stat-ahead when the client detects
562 readdir and sequential stat of dir entries therein
565 Frequency : on start mds
567 Description: Implement get_info(last_id) in obdfilter.
570 Frequency : occasional
572 Description: Correctly check stale fid, not start epoch if ost not support SOM
573 Details : open with flag O_CREATE need set old fid in op_fid3 because
574 op_fid2 was overwritten with new generated fid, but mds can answer
575 with one of these two fids and both is not stale. Setattr
576 incorrectly started an epoch and assume will be called
577 done_writing, but without SOM done_writing ever being called.
580 Frequency : rare, depends on device drivers and load
582 Description: MDS or OSS nodes crash due to stack overflow
583 Details : Code changes in 1.8.0 increased the stack usage of some functions.
584 In some cases, in conjunction with device drivers that use a lot
585 of stack the MDS (or possibly OSS) service threads could overflow
586 the stack. One change which was identified to consume additional
587 stack has been reworked to avoid the extra stack usage.
590 Frequency : occasional
592 Description: Do not fail import if osc_interpret_create gets -EAGAIN
593 Details : If osc_interpret_create got -EAGAIN it immediately exits and
594 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
595 oscc_has_objects and see OSC has no objests and call
596 oscc_internal_create to resend create request.
598 Severity : enhancement
600 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
602 Severity : enhancement
604 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
607 Frequency : very rare
609 Description: Fix replay if there is an un-replied request and open
610 Details : In some cases, older replay request will revert the
611 mcd->mcd_last_xid on MDS which is used to record the client's
614 Severity : enhancement
616 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
618 Severity : enhancement
620 Description: Add rhel5 support to HEAD.
622 Serverity : enhancement
624 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
629 Description: Don't allow skipping OSTs if index has been specified.
630 Details : Don't allow skipping OSTs if index has been specified, make
631 locking in internal create lots better.
635 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
636 Details : access to bitfield in structure is always rounded to long
637 and this produce problem with not atomic change any bit.
641 Description: Lustre make rpms failed.
642 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
643 builds Lustre rpms from tarball.
646 Frequency : rare on shutdown ost
648 Description: If llog cancel was not send before clean_exports phase, this can
649 produce deadlock in llog code.
650 Details : If llog thread has last reference to obd and call class_import_put
651 this produce deadlock because llog_cleanup_commit_master wait when
652 last llog_commit_thread exited, but this never success because was
653 called from llog_commit_thread.
657 Description: allow userland application know is lost one of stripes.
658 Details : fill lvb_blocks with error code on ost and return it to
659 application if error flag found.
663 Description: NULL lov_tgts causing MDS oops
664 Details : more safe checks for NULL lov_tgts for avoid oops.
666 Severity : enhancement
668 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
672 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
673 Details : If an OST is down the MDS will hang indefinitely in
674 obd_statfs_rqset() waiting for the statfs data. While for
675 MDS QOS usage of statfs, it should not stuck in waiting.
677 Severity : enhancement
679 Description: remote_acl support
680 Details : Support ACL-based permission check for remote user.
681 Support setfacl/getfacl for remote user with the utils
682 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
683 the system "{s,g}etfacl" utils.
685 Severity : enhancement
687 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
689 Severity : enhancement
691 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
695 Description: make mgs_setparam() handle fsname containing dash
696 Details : fsname containing a dash does not work with lctl conf_param
698 Severity : enhancement
700 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
702 Severity : enhancement
704 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
708 Frequency : rare, at shutdown
709 Description: access already free / zero obd_namespace.
710 Details : if client_disconnect_export was called without force flag set,
711 and exist connect request in flight, this can produce access to
712 NULL pointer (or already free pointer) when connect_interpret
713 store ocd flags in obd_namespace.
717 Frequency : only at startup
718 Description: not alloc memory with spinlock held.
719 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
720 if any spinlock held.
722 Severity : enhancement
724 Description: make lustre randomly fail allocating memory
725 Details : Make lustre randomly failed allocating memory for testing purpose.
727 Severity : enhancement
729 Description: lost problems with lov objid file
730 Details : Fixes some scability and access to not inited memory problems
731 in work with lov objdid file.
736 Description: lfs find does not continue on file error
737 Details : Continue other files processing when a file/dir is absent.
741 Description: Inconsistent usage of lustre_pack_reply()
742 Details : Standardize the usage of lustre_pack_reply() such that it
743 always generate a CERROR on failure.
746 Frequency : occasional
748 Description: MDS hang or stay in waiting lock
749 Details : If client receive lock with CBPENDING flag ldlm need send lock
750 cancel as separate rpc, to avoid situation when cancel request
751 can't processed due all i/o threads stay in wait lock.
755 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
759 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
763 Description: Update to SLES9 kernel-2.6.5-7.287.3.
767 Description: Update to RHEL4 latest kernel.
769 Severity : enhancement
771 Description: Build SLES10 patchless client fails
772 Details : The configure was broken by run ./configure with
773 --with-linux-obj=.... argument for patchless client. When the
774 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
775 can't search header adequately. Use absolute path such as
776 -I($LINUX)/include instead.
780 Description: interrupt oig_wait produce painc on resend.
781 Details : brw_redo_request can be used for resend requests from ptlrpcd and
782 private set, and this produce situation when rq_ptlrpcd_data not
783 copyed to new allocated request and triggered LBUG on assert
784 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
785 ptlrpcd set if request is changed and can be safety changed to use
790 Description: LASSERT_{REQ,REP}SWAB macros are buggy
791 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
792 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
793 these macros and replace them with their difinitions instead.
797 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
798 Details : Update which_patch & target file for SLES10 latest kernel.
801 Description: Remove client patches from SLES 10 kernel.
802 Details : This causes SLES 10 clients to behave as patchless clients
803 even on a Lustre-patched (server) kernel.
805 Severity : enhancement
807 Description: self-adjustable client's lru lists
808 Details : use adaptive algorithm for managing client cached locks lru
809 lists according to current server load, other client's work
810 pattern, memory activities, etc. Both, server and client
811 side namespaces provide number of proc tunables for controlling
814 Severity : enhancement
816 Description: light-weight GSS support
817 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
818 and reduce performance overhead.
820 Severity : enhancement
822 Description: Linux keyring support
823 Details : Support using service of Linux keyring for Lustre GSS internal
824 context refresh/cache mechanism.
828 Description: Fix errors in lfs documentation
829 Details : Fixes man pages
833 Description: don't use GFP_* in generic Lustre code.
834 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
838 Description: obdclass is limited by single OBD_ALLOC(idarray)
839 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
843 Description: warnings with build patchless client with vanila 2.6.19 and up
844 Details : change old ctl_table style and replace ctl_table/ctl_table_header
845 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
849 Frequency : Only for SLES
850 Description: sanity_quota fail test_1
851 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
852 file, which makes TSTID[2] inunique.
856 Frequency : Only for SLES9
857 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
858 Details : Update target/ChangeLog/which_patch .
862 Frequency : Only for test_10 in sanity-quota.sh
863 Description: a bug in quota test code
864 Details : checking a test flag in a wrong place
868 Frequency : Only for test_13 in sanity-quota.sh
869 Description: a bug in quota test script
870 Details : shouldn't check overfull allocated quota
874 Description: jbd statistics
875 Details : Port older jbd statistics patch for sles10
877 Severity : enhancement
879 Description: Add printing inode info into message about error in writepage.
884 Description: reply_lock_interpret crash due to race with it and lock cancel.
885 Details : Do not replay locks that are being cancelled. Do not reference
886 locks by their address during replay, just by their handle.
890 Frequency : When flocks are used.
891 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
892 Details : Flock locks might destroy just granted lock if it could be merged
893 with another existing flock, this is done in completion handler,
894 so teach ldlm_cli_enquque_fini that this is a valid case for
900 Description: Oops in read and write path when failing to allocate lock.
901 Details : Check if lock allocation failed and return error back.
905 Description: lstripe command fails for valid OST index
906 Details : The stripe offset is compared to lov->desc.ld_tgt_count
907 instead of lov->desc.ld_active_tgt_count.
911 Description: sanity.sh failed test 103
912 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
917 Description: df doesn't work properly if diskfs blocksize != 4K
918 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
922 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
923 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
925 Severity : enhancement
927 Description: Adaptive timeouts
928 Details : RPC timeouts adapt to changing server load and network
929 conditions to reduce resend attempts and improve recovery time.
933 Description: llapi_file_create() does not allow some changes
934 Details : add llapi_file_open() that allows specifying the mode and
935 open flags, and also returns an open file handle.
939 Description: merge and cleanup kernel patches.
940 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
944 Description: Add journal checksum support.(Kernel part)
945 Details : The journal checksum feature adds two new flags i.e
946 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
947 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
948 indicates that the commit block contains the checksum for
949 the blocks described by the descriptor blocks. Now commit
950 record can be sent to disk without waiting for descriptor
951 blocks to be written to disk. This behavior is controlled
952 using JBD2_FEATURE_ASYNC_COMMIT flag.
956 Description: OSS needs mutliple precreate threads
957 Details : Add ability to start more than one create thread per OSS.
961 Description: Sanity.sh test_65k failed: setstripe should have succeeded
962 Details : execute lfs setstripe on client
966 Description: mds_obd_create error creating tmp object
967 Details : When the user sets quota on root, llog will be affected and can't
968 create files and write files.
971 Frequency : Always on ia64 patchless client, and possibly others.
973 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
974 Details : This allows the patchless client to be loaded on architectures
979 Description: RedHat Update kernel for RHEL5
980 Details : Add the kernel config file for RHEL5.
984 Description: RedHat Update kernel for RHEL5
985 Details : Modify the kernel config file more closer RHEL5.
989 Description: Build failure against Centos5 (RHEL5)
990 Details : Define PAGE_SIZE when it isn't present.
994 Description: test_7 on sanity-quota.sh can't be used on separate machines
995 Details : add facet to handle it
999 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1000 Details : with NFS, the anon dentry's parent was set to itself in
1001 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1002 corresponding dentry other than use rec->ur_name.
1004 Severity : enhancement
1006 Description: lfs setstripe enhancement
1007 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1011 Description: enable data checksumming by default
1012 Details : enable checksum by default, allow --disable-checksum
1013 configure option and "-o nochecksum" mount option. Checksums
1014 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1018 Description: lustre support for RHEL5
1019 Details : Add support for RHEL5.
1023 Description: Client eviction due to failover config
1024 Details : after a connection loss, the lustre client should attempt to
1025 reconnect to the last active server first before trying the
1026 other potential connections.
1030 Description: when mds and osts use different quota unit(32bit and 64bit),
1031 quota will be released repeatly.
1032 Details : void sending multiple quota reqs to mds, which will keep the
1033 status between the reqs.
1037 Description: osts not allocated evenly to files
1038 Details : change the condition to increase offset_idx
1041 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1043 Description: Data corruption for OSTs that are formatted larger than 2TB
1045 Details : When generating the bio request for lustre file writes the
1046 sector number would overflow a temporary variable before being
1047 used for the IO. The data reads correctly from Lustre (which
1048 will overflow in a similar manner) but other file data or
1049 filesystem metadata may be corrupted in some cases.
1053 Description: Tune the kernel for good SCSI performance.
1054 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1055 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1060 Description: rewrite ext2-derived code in obdclass/uuid.c
1061 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1062 from scratch preserving functionality.
1066 Description: "lfs find -obd UUID" prints directories
1067 Details : "lfs find -obd UUID" will return all directory names instead
1068 of just file names. It is incorrect because the directories
1069 do not reside on the OSTs.
1073 Description: use i_size_read and i_size_write in 2.6 port
1074 Details : replace inode->i_size access with i_size_read/write()
1077 Frequency : when removing large files
1079 Description: scheduling issue during removal of large Lustre files
1080 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1081 It causes scheduling issues when removing large files (17TB in the
1085 Frequency : only with liblustre clients on XT3
1087 Description: evictions taking too long
1088 Details : allow llrd to evict clients directly on OSTs
1091 Frequency : only on ppc
1093 Description: /proc/fs/lustre/devices broken on ppc
1094 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1095 We should make sure the type of 'index' is loff_t and then cast
1096 explicitly as needed below. Do not assign an explicitly cast
1100 Frequency : only for rhel5
1102 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1103 Details : Modify the target file & which_kernel.
1105 Severity : enhancement
1107 Description: omit set fsid for export NFS
1108 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1109 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1113 Frequency : Always, for kernels after 2.6.16
1114 Description: Fix warning idr_remove called for id=.. which is not allocated.
1115 Details : Last kernels save old s_dev before kill super and not allow
1116 to restore from callback - restore it before call kill_anon_super.
1120 Description: buffer overruns could theoretically occur
1121 Details : llapi_semantic_traverse() modifies the "path" argument by
1122 appending values to the end of the origin string, and a
1123 overrun may occur. Adding buffer overrun check in liblustreapi.
1127 Description: Fix error on 'ls .' at the top of the Lustre mount.
1128 Details : Don't revalidate dentry if it is a root dentry.
1132 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1133 Details : Modify vm-tunables-rhel4.patch.
1137 Description: Kernel config for 2.6.18-vanilla.
1138 Details : Modify targets/2.6-vanilla.target.in.
1139 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1140 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1141 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1142 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1146 Description: improve handling recoverable errors
1147 Details : if request processig with error which can be recoverable on server
1148 request should be resend, otherwise page released from cache and
1153 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1154 Details : Modify target file & which_patch.
1155 A flaw was found in the IA32 system call emulation provided
1156 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1157 value could be stored in the %RAX register, which could trigger an
1158 out-of-bounds system call table access. An untrusted local user
1159 could exploit this flaw to run code in the kernel
1160 (ie a root privilege escalation). (CVE-2007-4573).
1164 Description: O_DIRECT bypasses client statistics.
1165 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1166 read_ahead_stats were updated. copied the stats section from
1167 osc_send_oap_rpc() into async_internal().
1171 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1175 Description: change order of libsysio includes
1176 Details : '#include sysio.h' should always come before '#include xtio.h'
1180 Description: handle "serious error: objid * already exists" more gracefully
1181 Details : If LAST_ID value on disk is smaller than the objects existing in
1182 the O/0/d* directories, it indicates disk corruption and causes an
1183 LBUG(). If the object is 0-length, then we should use the existing
1184 object. This will help to avoid a full fsck in most cases.
1186 Severity : enhancement
1188 Description: adapt the lustre_config script to support the upgrade case
1189 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1190 server targets to 1.6 in parallel.
1194 Description: To avoid grant space > avaible space when the disk is almost
1195 full. Without this patch you might see the error "grant XXXX >
1196 available" or some LBUG about grant, when the disk is almost
1198 Details : In filter_check_grant, for non_grant cache write, we should
1199 check the left space by if (*left > ungranted + bytes), instead
1200 of (*left > ungranted), because only we are sure the left space
1201 is enough for another "bytes", then the ungrant space should be
1202 increase. In client, we should update cl_avail_grant only there
1203 is OBD_MD_FLGRANT in the reply.
1207 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1208 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1209 described in CVE-2007-4573. This problem could allow untrusted
1210 local users to gain root access.
1213 Frequency : when using O_DIRECT and quotas
1215 Description: Incorrect file ownership on O_DIRECT output files
1216 Details : block usage reported by 'lfs quota' does not take into account
1217 files that have been written with O_DIRECT.
1222 Description: touch file failed when fs is not full
1223 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1224 otherwise we can get ENOSP while fs is not full.
1228 Description: parallel lock callbacks
1229 Details : Instead of sending blocking and completion callbacks as separated
1230 requests, adding them to a set and sending in parallel.
1233 Frequency : only for Cray XT3
1234 Bugzilla : 12829/13455
1235 Description: Changing primary group doesn't change the group lustre assigns to
1237 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1238 overridden with the primary group provided by the group upcall,
1239 whereas the supplied fsgid can be trusted if it is in the list of
1240 supplementary groups returned by the group upcall.
1242 Severity : enhancement
1244 Description: Allow masking D_WARNING, D_ERROR messages from console
1245 Details : Console messages can now be disabled via lnet.printk.
1249 Description: User code with malformed file open parameter crashes client node
1250 Details : Before packing join_file req, all the related reference should be
1251 checked carefully in case some malformed flags cause fake
1252 join_file req on client.
1256 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1257 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1258 mode is equal to the granted mode. However, it should also check
1259 that LDLM_FL_CP_REQD is not set.
1263 Description: Heavy nfs access might result in deadlocks
1264 Details : After ELC code landed, it is now improper to enqueue any mds
1265 locks under och_sem, because enqueue might want to decide to
1266 cancel open locks for same inode we are holding och_sem for.
1270 Description: Client eviction while running blogbench
1271 Details : A lot of unlink operations with concurrent I/O can lead to a
1272 deadlock causing evictions. To address the problem, the number of
1273 oustanding OST_DESTROY requests is now throttled to
1274 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1275 callbacks are processed in priority.
1279 Description: enable ACLs on MDS by default
1280 Details : ACLs must be enabled on MDS by default.
1283 Frequency : PPC/PPC64 only
1285 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1286 Details : fix duplicated definitions between asm-ppc64/types.h and
1287 lustre_types.h on PPC.
1290 Frequency : PPC/PPC64 only
1292 Description: asm-ppc/segment.h does not exist
1293 Details : fix compile issue on PPC.
1297 Description: better handle error messages in extents code
1300 Frequency : RHEL4 only
1302 Description: mkfs is very slow on IA64/RHEL4
1303 Details : A performance regression has been discovered in the MPT Fusion
1304 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1305 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1306 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1308 Severity : enhancement
1310 Description: SNMP support enhancement
1311 Details : Adding total number of sampled request for an MDS node in snmp
1314 Severity : enhancement
1316 Description: Optimize ldlm waiting list processing for PR extent locks
1317 Details : When processing waiting list for read extent lock and meeting read
1318 lock that is same or wider to it that is not contended, skip
1319 processing rest of the list and immediatelly return current
1320 status of conflictness, since we are guaranteed there are no
1321 conflicting locks in the rest of the list.
1325 Description: Time out and refuse to reconnect
1326 Details : When the failover node is the primary node, it is possible
1327 to have two identical connections in imp_conn_list. We must
1328 compare not conn's pointers but NIDs, otherwise we can defeat
1329 connection throttling.
1333 Description: port llog fixes from b1_6 into HEAD
1334 Details : Port llog reference couting and some llog cleanups from b1_6
1335 (bug 10800) into HEAD, for protect from panic and access to already
1336 free llog structures.
1340 Description: Detect stride IO mode in read-ahead
1341 Details : When a client does stride read, read-ahead should detect that and
1342 read-ahead pages according to the detected stride pattern.
1346 Description: data checksumming impacts single node performance
1347 Details : add support for several checksum algorithm. Currently, only CRC32
1348 and Adler-32 are supported. The checksum type can be changed on
1349 the fly via /proc/fs/lustre/osc/*/checksum_type.
1353 Description: use adler32 for page checksums
1354 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1359 Description: build for x2 fails
1360 Details : fix compile issue on Cray systems.
1364 Description: Properly match for duplicate locks
1365 Details : Due to different lock order from skiplists code, we need to
1366 traverse entire list for now
1369 Frequency : only on PPC/SLES10
1371 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1372 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1373 be defined. Add a hack in mkfs_lustre.c to work around this compile
1378 Description: LASSERT on MDS when client holding flock lock dies
1379 Details : ldlm pool logic depends on number of granted locks equal to
1380 number of released locks which is not true for flock locks, so
1381 just exclude such locks from consideration.
1385 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1386 Details : Use fsfilt_sync() for both the whole filesystem sync and
1387 individual file sync to eliminate dangerous inode locking
1388 with I_LOCK that can lead to a deadlock.
1392 Description: performance in 1.6.3
1393 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1394 sure that 1MB requests are not fragmented by the block layer.
1396 Severity : enhancement
1398 Description: organize the server-side client stats on per-nid basis
1399 Details : Change the structure of stats under obdfilter and mds to
1407 The "uuid"s file would list the uuids of _active_ exports.
1408 And the clear entry is to clear all stats and stale nids.
1410 Severity : enhancement
1412 Description: eliminate client locks in face of contention
1413 Details : file contention detection and lockless i/o implementation
1414 for contended files.
1418 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1421 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1423 Description: MDS may lose file striping (and hence file data) in some cases
1424 Details : If there are additional extended attributes stored on the MDS,
1425 in particular ACLs, SELinux, or user attributes (if user_xattr
1426 is specified for the client mount options) then there is a risk
1427 of attribute loss. Additionally, the Lustre file striping
1428 needs to be larger than default (e.g. striped over all OSTs),
1429 and an additional attribute must be stored initially in the
1430 inode and then increase in size enough to be moved to the
1431 external attribute block (e.g. ACL growing in size) for file
1436 Description: skiplist implementation simplification
1437 Details : skiplists are used to group compatible locks on granted list
1438 that was implemented as tracking first and last lock of each
1439 lock group the patch changes that to using doubly linked lists
1443 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1444 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1445 because LASSERT on both the data supplied by a client, and the
1446 data on disk is dangerous and incorrect.
1448 Severity : enhancement
1450 Description: Slow truncate/writes to huge files at high offsets.
1451 Details : Directly associate cached pages to lock that protect those pages,
1452 this allows us to quickly find what pages to write and remove
1453 once lock callback is received.
1457 Description: more ldlm soft lockups
1458 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1459 starve other threads from the resource lock for a long time in
1460 case of long waiting queue, so change the debug level from
1461 D_OTHER to the less frequently used D_INFO.
1463 Severity : enhancement
1465 Description: add -gid, -group, -uid, -user options to lfs find
1469 Description: Hung threads in invalidate_inode_pages2_range
1470 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1471 once one is completed. As a result, some RPCs are stuck in the
1472 queue and are never sent.
1476 Description: filter threads hungs on waiting journal commit
1477 Details : Cleanup filter group llog code, then only filter group llog will
1478 be only created in the MDS/OST syncing process.
1482 Description: Procfs and llog threads access destoryed import sometimes.
1483 Details : Sync the import destoryed process with procfs and llog threads by
1484 the import refcount and semaphore.
1486 Severity : enhancement
1488 Description: openlock cache of b1_6 port to HEAD
1493 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1494 Details : If insertion of an extent fails, then discard the inode
1495 preallocation and free data blocks else it can lead to duplicate
1500 Description: don't always update ctime in ext3_xattr_set_handle()
1501 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1502 In some cases the ctime should not be updated, for example for
1503 2.0->1.8 compatibility it is necessary to delete an xattr and it
1504 should not update the ctime.
1508 Bugzilla : 15713/16362
1509 Description: Assertion in iopen_connect_dentry in 1.6.3
1510 Details : looking up an inode via iopen with the wrong generation number can
1511 populate the dcache with a disconneced dentry while the inode
1512 number is in the process of being reallocated. This causes an
1513 assertion failure in iopen since the inode's dentry list contains
1514 both a connected and disconnected dentry.
1518 Description: assertion failure in ldlm_handle2lock()
1519 Details : fix a race between class_handle_unhash() and class_handle2object()
1520 introduced in lustre 1.6.5 by bug 13622.
1525 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1526 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1531 Description: oops in sd_iostats_seq_show()
1532 Details : unloading/reloading the scsi low level driver triggers a kernel
1533 bug when trying to access the sd iostat file.
1538 Description: Kernel panics during QLogic driver reload
1539 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1540 patch, causing memory corruption.
1545 Description: journal_dev option does not work in b1_6
1546 Details : pass mount option during pre-mount.
1548 Severity : enhancement
1550 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1551 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1552 extent information of a file. It can be used to map logical blocks
1553 in a file to physical blocks in the block device.
1557 Description: LDLM soft lockups - improvement
1558 Details : It is be possible to send the lock handle along with each read
1559 or write request because the client is already doing a lock match
1560 itself so there isn't any reason the OST should have to re-do that
1564 Frequency : only on Cray X2
1566 Description: X2 build failures
1567 Details : fix build failures on Cray X2.
1571 Description: xid & resent requests
1572 Details : Initialize RPC XID from clock at startup (randomly if clock is
1575 Severity : enhancement
1577 Description: Add lustre_start utility to start or stop multiple Lustre servers
1582 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1583 Details : In case of memory pressure, list_del() can be called twice on
1584 req->rq_history_list, causing a kernel oops.
1588 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1589 Details : fix stack overflow in the distributed lock manager by defering
1590 export eviction after a failed AST to the elt thread instead of
1591 handling it in the dlm interpret routine.
1595 Description: Convert some comments to new format.
1596 Details : Update documenting comments to match doxygen conventions.
1600 Description: Grammar fixes.
1601 Details : A couple of trivial spelling fixes.
1605 Description: OSD_COUNTERS-mandatory
1606 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1607 to simplify the code.
1611 Description: simplify cmm_device freeing logic.
1612 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1616 Description: Add lockdep support to dt_object_operations locking interface.
1617 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1618 parameter indicating lock ordering. Update mdd code to use new
1623 Description: Introduce failloc constants for lockless IO tests.
1624 Details : Add two new failloc constants to test lockless IO. Only one of
1625 them in implemented---another is checked in yet to be landed
1630 Description: Add lockdep support for inode mutex.
1631 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1632 in the situations where Lustre has to lock more than one inode
1637 Description: Add optional invariants checking support.
1638 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1639 configure switch. This macro is to be used for consistency and
1640 sanity checks that are too expensive to be left in `production'
1645 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1646 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1651 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1652 Details : Use spin_lock_nested() in (the only) situation where more than
1653 one ldlm_lock is locked simultaneously. Also, fix possible
1654 dead-lock in ldlm_lock_change_resource() by enforcing particular
1659 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1660 Details : Instead of specifying each ldlm_lock call-back through separate
1661 parameter, wrap them into struct ldlm_callback_suite.
1665 Description: Kill join_lru obd method and its callers.
1666 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1667 in memory---a requirement implemented through ->o_join_lru() obd
1668 method in HEAD. Remove this method and its users.
1672 Description: Add asynchronous ldlm ENQUEUE completion handler.
1673 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1674 case that stock ldlm_completion_ast() cannot handle as it waits
1675 until lock is granted. Introduce new ldlm_completion_ast_async()
1676 for this. Also comment ldlm_completion_ast().
1680 Description: ldlm_error <-> errno conversion.
1681 Details : Add functions to map (rather arbitrary) between LDLM error codes
1682 and standard errno values. CLIO needs this to prevent LDLM specific
1683 constants from escaping ldlm and osc.
1687 Description: Kill unused ldlm_handle2lock_ns() function.
1688 Details : Kill unused ldlm_handle2lock_ns() function.
1692 Description: Add lu_ref support to ldlm_lock
1693 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1694 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1695 and ldlm_resource. LDLM interface has to be changed, because code that
1696 releases a reference on a lock, has to "know" what reference this is.
1697 In the most frequent case
1699 lock = ldlm_handle2lock(handle);
1701 LDLM_LOCK_PUT(lock);
1703 no changes are required. When any other reference (received _not_ from
1704 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1705 instead of LDLM_LOCK_PUT().
1707 Arguably, changes are pervasive, and interface requires some discipline
1708 for proper use. On the other hand, it was very instrumental in finding
1709 a few leaked lock references.
1713 Description: Add ldlm_lock_addref_try().
1714 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1715 attempts to addref a lock that might be being canceled
1720 Description: Add ldlm_weigh_callback().
1721 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1722 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1723 hard-coded `number of pages' logic.
1727 Description: Add lockdep annotations to llog code.
1728 Details : Use appropriately tagged _nested() locking calls in the places
1729 where llog takes more than one ->lgh_lock lock.
1733 Description: Add loi_kms_set().
1734 Details : Wrap kms updates into a helper function.
1738 Description: Constify instances of struct lsm_operations.
1739 Details : Constify instances of struct lsm_operations.
1743 Description: lu_conf support.
1744 Details : On a server, a file system object is uniquely identified
1745 by a fid, which is sufficient to locate and load all object
1746 state (inode). On a client, on the other hand, more data are
1747 necessary instantiate an object. Change lu_object_find() and
1748 friends to take additional `lu_conf' argument describing object.
1749 Typically this includes layout information.
1753 Description: lu_context fixes.
1754 Details : Introduce new lu_context functions that are needed on the client
1755 side, where some system threads (ptlrpcd) are shared by multiple
1756 modules, and so cannot be stopped during module shutdown.
1760 Description: Add start and stop methods to lu_device_type_operations.
1761 Details : Introduce two new methods in lu_device_type_operations, that are
1762 invoked when first instance of a given type is created and last one
1763 is destroyed respectively. This is need by CLIO.
1767 Description: Add lu_ref support to struct lu_device.
1768 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1769 track leaked references.
1773 Description: Introduce lu_kmem_descr.
1774 Details : lu_kmem_descr and its companion interface allow to create
1775 and destroy a number of kmem caches at once.
1779 Description: Fix lu_object finalization race.
1780 Details : Fix a race between lu_object_find() finding an object and its
1781 concurrent finalization. This race is (most likely) not possible
1782 on the server, but might happen on the client.
1786 Description: Introduce lu_ref interface.
1787 Details : lu_ref is a debugging module allowing to track references to
1788 a given object. It is quite cpu expensive, and has to be
1789 explicitly enabled with --enable-lu_ref. See usage description
1794 Description: Factor lu_site procfs stats into a separate function.
1795 Details : Separate lu_site stats printing code into a separate function
1796 that can be reused on a client.
1800 Description: Constify instances of struct {lu,dt,md}_device_operations.
1801 Details : Constify instances of struct {lu,dt,md}_device_operations.
1805 Description: Introduce struct md_site and move meta-data specific parts of
1806 struct lu_site here.
1807 Details : Move md-specific fields out of struct lu_site into special struct
1808 md_site, so that lu_site can be used on a client.
1812 Description: Kill mdd_lov_destroy().
1813 Details : Remove unused mdd code.
1817 Description: Add st_block checking to multistat.c.
1818 Details : Add st_block checking to multistat.c.
1822 Description: Add lu_ref support to struct obd_device.
1823 Details : Add lu_ref tracking to obd_device.
1827 Description: Kill obd_set_fail_loc().
1828 Details : Remove unused code.
1832 Description: Add special type for ptlrpc_request interpret functions.
1833 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1834 there. Actual usage will be in CLIO.
1838 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1839 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1840 doesn't work with lockdep.
1844 Description: Add rwv.c test program.
1845 Details : New testing program exercising readv(2) and writev(2) (Qian).
1849 Description: Add sendfile.c test program.
1850 Details : New testing program exercising sendfile(2) (Jay).
1854 Description: Ratelimit a message that can be very frequent.
1855 Details : Ratelimit a memory allocation failure message that can
1860 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1861 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1865 Description: Liblustre build fixes.
1866 Details : Liblustre build fixes.
1870 Description: libcfs: add cfs_{need,cond}_resched() interface.
1871 Details : libcfs: add cfs_{need,cond}_resched() definition and
1872 implementations for Linux, NT, and liblustre.
1874 Severity : enhancement
1876 Description: More exported tunables for mballoc
1877 Details : Add support for tunable preallocation window and new tunables for
1878 large/small requests
1882 Description: Detect corruption of block bitmap and checking for preallocations
1883 Details : Checks validity of on-disk block bitmap. Also it does better
1884 checking of number of applied preallocations. When corruption is
1885 found, it turns filesystem readonly to prevent further corruptions.
1889 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1890 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1891 window to 0. If it does hit cache miss, and read-pattern is still
1892 stride-io mode, does not reset the stride window, but also does not
1893 increase the stride window length in this case.
1897 Frequency : only for big-endian servers
1898 Description: Check if system is big-endian while mounting fs with extents feature
1899 Details : Mounting a filesystem with extents feature will fail on big-endian
1900 systems since ext3-based ldiskfs is not supported on big-endian
1901 systems. This can be over-riden with "bigendian_extents" mount option.
1903 Severity : enhancement
1905 Description: The root squash functionality
1906 Details : A security feature, which is to prevent users from being able
1907 to mount lustre on their desktop, run as root, and delete
1908 all of the files in the filesystem. The goal is accomplished by
1909 remapping user id (UID) and group id (GID) of the root user to
1910 a UID and GID specified by the system administartor via Lustre
1911 configuration management server (MGS). The functionality also
1912 allows to specify sets of clients for which the remapping does
1917 Description: Excessive recovery window
1918 Details : With AT enabled, the recovery window can be excessively long (6000+
1919 seconds). To address this problem, we no longer use
1920 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1921 timeout no longer depends on the service time, it is set to
1922 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1923 time via pb_service_time.
1927 Description: Watchdog triggered on MDS failover
1928 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
1929 the OSTs know that the MDS "UUID" can be reused for the same export
1930 from a different NID, so we do not need to wait for the export to be
1933 --------------------------------------------------------------------------------
1935 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1937 * Support for kernels:
1938 2.6.5-7.283 (SLES 9),
1939 2.6.9-55.EL (RHEL 4),
1940 2.6.16.46-0.14 (SLES 10),
1941 2.6.18.8 vanilla (kernel.org)
1942 * Client support for unpatched kernels:
1943 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1944 2.6.16 - 2.6.22 vanilla (kernel.org)
1945 * Due to recently discovered recovery problems, we do not recommend
1946 using patchless RHEL 4 clients with this or any earlier release.
1947 * Recommended e2fsprogs version: 1.39.cfs8
1948 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1949 * Starting with this release, the ldiskfs backing filesystem required
1950 by Lustre is now in its own package, lustre-ldiskfs. This package
1951 should be installed. It is versioned separately from Lustre and
1952 may be released separately in future.
1956 Description: block reactivating mgc import until all deactivates complete
1957 Details : Fix race when failing back MDT/MGS to itself (testing)
1960 Frequency : at statup only
1962 Description: mds_lov_synchronize race leads to various problems
1963 Details : simultaneous MDT->OST connections at startup can cause the
1964 sync to abort, leaving the OSC in a bad state.
1966 Severity : enhancement
1968 Description: add optional extra BUILD_VERSION info
1969 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1970 to override the lustre version.
1973 Frequency : 2.6.18 servers only
1975 Description: ll_kern_mount() doesn't release the module reference
1976 Details : The ldiskfs module reference count never drops down to 0
1977 because ll_kern_mount() doesn't release the module reference.
1982 Description: server LBUG when using old ost_num_threads parameter
1983 Details : Accept the old ost_num_threads parameter but warn that it
1984 is deprecated, and fix an off-by-one error that caused an LBUG.
1989 Description: Transient SCSI error results in persistent IO issue
1990 Details : iobuf->dr_error is not reinitialized to 0 between two
1994 Frequency : sometimes when underlying device returns I/O errors
1996 Description: OSTs not going read-only during write failures
1997 Details : OSTs are not remounted read-only when the journal commit threads
1998 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1999 instead of the ext3 wrappers.
2003 Description: poor connect scaling with increasing client count
2004 Details : Don't run filter_grant_sanity_check for more than 100 exports
2005 to improve scaling for large numbers of clients.
2008 Frequency : SLES10 only
2010 Description: sanity-quota.sh quotacheck failed: rc = -22
2011 Details : Quotas cannot be enabled on SLES10.
2014 Frequency : liblustre clients only
2016 Description: getdirentries does not give error when run on compute nodes
2017 Details : getdirentries does not fail when the size specified as an argument
2018 is too small to contain at least one entry
2020 Severity : enhancement
2022 Description: Add LNET router traceability for debug purposes
2023 Details : If a checksum failure occurs with a router as part of the
2024 IO path, the NID of the last router that forwarded the bulk data
2025 is printed so it can be identified.
2030 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2031 Details : Due to a race condition, liblustre clients were occasionally
2032 evicted incorrectly.
2034 Severity : enhancement
2036 Description: lfs setstripe use optional parameters instead of postional
2039 Severity : enhancement
2041 Description: Nanosecond timestamp support for ldiskfs
2042 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2043 resolution timestamps. There is not yet support for this at
2044 the Lustre filesystem level.
2047 Frequency : during server recovery
2049 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2050 Details : request with rq_no_resend flag not awake l_wait_event if they get
2054 Frequency : nfs export on patchless client
2056 Description: connectathon hang when test nfs export over patchless client
2057 Details : Disconnected dentry cannot be found with lookup, so we do not need
2058 to unhash it or make it invalid
2061 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2062 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2063 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2068 Description: open req refcounting wrong on reconnect
2069 Details : If reconnect happened between getting open reply from server and
2070 call to mdc_set_replay_data in ll_file_open, we will schedule
2071 replay for unreferenced request that we are about to free.
2072 Subsequent close will crash in variety of ways.
2073 Check that request is still eligible for replay in
2074 mdc_set_replay_data().
2079 Description: disable writes to filesystem when reading health_check file
2080 Details : the default for reading the health_check proc file has changed
2081 to NOT do a journal transaction and write to disk, because this
2082 can cause reads of the /proc file to hang and block HA state
2083 checking on a healthy but otherwise heavily loaded system. It
2084 is possible to return to the previous behaviour during configure
2085 with --enable-health-write.
2087 Severity : enhancement
2089 Description: 64-bit inode version
2090 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2091 to the inode. This will be required for version-based recovery.
2096 Description: MDS fails to start if a duplicate client export is detected
2097 Details : in some rare cases it was possible for a client to connect to
2098 an MDS multiple times. Upon recovery the MDS would detect this
2099 and fail during startup. Handle this more gracefully.
2101 Severity : enhancement
2103 Description: Add -o localflock option to simulate old noflock
2105 Details : This will achieve local-only flock/fcntl locks
2111 Description: log_commit_thread vs filter_destroy race leads to crash
2112 Details : Take import reference before releasing llog record semaphore
2117 Description: Wrong request locking in request set processing
2118 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2119 imp_delayed_list, in this place should be used imp_lock.
2122 Frequency : when reconnection
2124 Description: Grant Leak when osc reconnect to OST
2125 Details : When osc reconnect ost, OST(filter) should check whether it
2126 should grant more space to client by comparing fed_grant and
2127 cl_avail_grant, and return the granted space to client instead
2128 of "new granted" space, because client will call osc_init_grant
2129 to update the client grant space info.
2132 Frequency : when client reconnect to OST
2134 Description: Grant Leak when osc do resend and replay bulk write
2135 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2136 bulk write request, because the grant info while be sync between
2137 OSC and OST when reconnect, and we should ignore the grant info
2138 these of resend/replay write req.
2143 Description: Grant space more than avaiable left space sometimes.
2144 Details : When then OST is about to be full, if two bulk writing from
2145 different clients came to OST. Accord the avaliable space of the
2146 OST, the first req should be permitted, and the second one
2147 should be denied by ENOSPC. But if the seconde arrived before
2148 the first one is commited. The OST might wrongly permit second
2149 writing, which will cause grant space > avaiable space.
2152 Frequency : when client is evicted
2154 Description: Grant might be wrongly erased when osc is evicted by OST
2155 Details : when the import is evicted by server, it will fork another
2156 thread ptlrpc_invalidate_import_thread to invalidate the
2157 import, where the grant will be set to 0. While the original
2158 thread will update the grant it got when connecting. So if
2159 the former happened latter, the grant will be wrongly errased
2160 because of this race.
2165 Description: Checking Stale with correct fid
2166 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2167 to check whether it is stale, because sometimes, we want the
2168 enqueue happened anyway, and op_data.fid2 will not be initialized.
2170 Severity : enhancement
2172 Description: update patchless client
2173 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2176 Frequency : only with 2.4 kernel
2178 Description: random memory corruption
2179 Details : size of struct ll_inode_info is to big for union inode.u and this
2180 can be cause of random memory corruption.
2185 Description: Memory leak in recovery
2186 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2187 It should also check obd_fail before fsfilt_start, otherwise if
2188 fsfilt_start return -EROFS,(failover mds during mds recovery).
2189 then the req will return with repmsg->transno = 0 and rc = EROFS.
2190 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2191 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2192 be freed no matter whether fsfilt_commit success or not.
2195 Frequency : only with huge count clients
2197 Description: Prevents from taking the superblock lock in llap_from_page for
2199 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2200 for llap_from_page call in ll_removepage prevents from taking the
2201 superblock lock for a soon died page.
2206 Description: Not check open intent error before release open handle
2207 Details : in some rare cases, the open intent error is not checked before
2208 release open handle, which may cause
2209 ASSERTION(open_req->rq_transno != 0), because it tries to release
2210 the failed open handle.
2215 Description: Set cat log bitmap only after create log success.
2216 Details : in some rare cases, the cat log bitmap is set too early. and it
2217 should be set only after create log success.
2221 Description: Accessing a block bevice can re-enable I/O when Lustre is
2222 tearing down a device.
2223 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2228 Description: service threads may hog cpus when there are a lot of requests
2229 Details : Insert cond_resched to give other threads a chance to use some CPU
2234 Description: the cat log was not initialized in recovery
2235 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2236 unlink log on mds will not be initialized until mds post
2237 recovery. And also in mds post recovery, the unlink log will
2238 initialization will be done asynchronausly, so there will be race
2239 between add unlink log and unlink log initialization.
2243 Description: brw_stats were being printed incorrectly
2244 Details : brw_stats were being printed as log2 but all of them were not
2245 recorded as log2. Also remove some code duplication arising from
2246 filter_tally_{read,write}.
2250 Frequency : rare, only in recovery.
2251 Description: ASSERTION(req->rq_type != LI_POISON) failed
2252 Details : imp_lock should be held while iterating over imp_sending_list for
2253 prevent destroy request after get timeout in ptlrpc_queue_wait.
2257 Description: replay-single.sh test 52 fails
2258 Details : A lock's skiplist need to be cleanup when it being unlinked
2259 from its resource list.
2263 Description: Short directio read returns full requested size rather than
2265 Details : Direct I/O operations should return actual amount of bytes
2266 transferred rather than requested size.
2268 Severity : enhancement
2270 Description: metadata RPC reduction (e.g. for rm performance)
2271 Details : decrease the amount of synchronous RPC between clients and servers
2272 by canceling conflicing lock before the operation on the client
2273 and packing thier handles into the main operation RPC to server.
2275 Severity : enhancement
2277 Description: Async OSC create to avoid the blocking unnecessarily.
2278 Details : If a OST has no remain object, system will block on the creating
2279 when need to create a new object on this OST. Now, ways use
2280 pre-created objects when available, instead of blocking on an
2281 empty osc while others are not empty. If we must block, we block
2282 for the shortest possible period of time.
2286 Description: improve handling recoverable errors
2287 Details : if request processig with error which can be recoverable on server
2288 request should be resend, otherwise page released from cache and
2291 Severity : enhancement
2293 Description: refine locking for avoid write wrong info into lov_objid file
2294 Details : fix possible races with add new target and write/update data in
2297 --------------------------------------------------------------------------------
2299 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2304 Frequency : on some architectures
2306 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2307 Details : Uninitialized flags sometimes cause configuration commands to
2311 Frequency : patchless clients only
2313 Description: missing __iget() symbol export
2314 Details : The __iget() symbol export is missing. To avoid the need for
2315 this on patchless clients the deathrow inode reaper is turned
2316 off, and we depend on the VM to clean up old inodes. This
2317 dependency was during via the fix for bug 12181.
2319 --------------------------------------------------------------------------------
2321 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2323 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2324 INTEROPERATE with older versions automatically. In many cases a
2325 special upgrade step is needed. Please read the
2326 user documentation before upgrading any part of a live system.
2327 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2328 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2329 * WARNING: Lustre configuration and startup changes are required with
2330 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2334 Severity : enhancement
2336 Description: MountConf
2337 Details : Lustre configuration is now managed via mkfs and mount
2338 commands instead of lmc and lconf. New obd types (MGS, MGC)
2339 are added for dynamic configuration management. See
2340 https://mail.clusterfs.com/wikis/lustre/MountConf for
2343 Severity : enhancement
2345 Description: dynamic OST addition
2346 Details : OSTs can now be added to a live filesystem
2348 Severity : enhancement
2350 Description: startup order invariance
2351 Details : MDTs and OSTs can be started in any order. Clients only
2352 require the MDT to complete startup.
2354 Severity : enhancement
2356 Description: parallel, asynchronous orphan cleanup
2357 Details : orphan cleanup is now performed in separate threads for each
2358 OST, allowing parallel non-blocking operation.
2360 Severity : enhancement
2362 Description: optimized stripe assignment
2363 Details : stripe assignments are now made based on ost space available,
2364 ost previous usage, and OSS previous usage, in order to try
2365 to optimize storage space and networking resources.
2367 Severity : enhancement
2369 Description: Permanently set tunables
2370 Details : All writable /proc/fs/lustre tunables can now be permanently
2371 set on a per-server basis, at mkfs time or on a live system.
2373 Severity : enhancement
2375 Description: Lustre message v2
2376 Details : Add lustre message format v2.
2378 Severity : enhancement
2380 Description: client OST exclusion list
2381 Details : Clients can be started with a list of OSTs that should be
2382 declared "inactive" for known non-responsive OSTs.
2386 Description: SPEC SFS validation failure on NFS v2 over lustre.
2387 Details : Changes the blocksize for regular files to be 2x RPC size,
2388 and not depend on stripe size.
2390 Severity : enhancement
2392 Description: Multiple MD RPCs in flight.
2393 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2394 To avoid overly-overloading MDS, introduce a limit on number of
2395 MDS RPCs in flight for a single client and add /proc controls
2396 to adjust this limit.
2398 Severity : enhancement
2400 Description: client read/write statistics
2401 Details : Add client read/write call usage stats for performance
2402 analysis of user processes.
2403 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2404 file access. extents_stats shows chunk size distribution.
2405 extents_stats_per_process show chunk size distribution per
2408 Severity : enhancement
2410 Description: mds statistics
2411 Details : Add detailed mds operations statistics in
2412 /proc/fs/lustre/mds/*/stats.
2416 Description: Failure of copying files with lustre special EAs.
2417 Details : Client side always return success for setxattr call for lustre
2418 special xattr (currently only "trusted.lov").
2423 Description: Refcount LNET uuids
2424 Details : The global LNET uuid list grew linearly with every startup;
2425 refcount repeated list entries instead of always adding to
2428 Severity : enhancement
2430 Description: Dynamic service threads
2431 Details : Within a small range, start extra service threads
2432 automatically when the request queue builds up.
2435 Frequency : mixed-endian client/server environments
2437 Description: mixed-endian crashes
2438 Details : The new msg_v2 system had some failures in mixed-endian
2441 Severity : enhancement
2443 Description: Easy OST removal
2444 Details : OSTs can be permanently deactivated with e.g. 'lctl
2445 conf_param lustre-OST0001.osc.active=0'
2447 Severity : enhancement
2449 Description: MGS proc entries
2450 Details : Added basic proc entries for the MGS showing what filesystems
2453 Severity : enhancement
2455 Description: provide MGS failover
2456 Details : Added config lock reacquisition after MGS server failover.
2458 Severity : enhancement
2460 Description: add Linux 2.4 support
2461 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2465 Description: a large application tries to do I/O to the same resource and dies
2466 in the middle of it.
2467 Details : Check the req->rq_arrival time after the call to
2468 ost_brw_lock_get(), but before we do anything about
2469 processing it & sending the BULK transfer request. This
2470 should help move old stale pending locks off the queue as
2471 quickly as obd_timeout.
2474 Frequency : when an incorrect nid is specified during startup
2476 Description: ptlrpc connect to non-existant node causes kernel crash
2477 Details : LNET can't be re-entered from an event callback, which
2478 happened when we expire a message after the export has been
2479 cleaned up. Instead, hand the zombie cleanup off to another
2482 Severity : enhancement
2484 Description: plain/inodebits lock performance improvement
2485 Details : Grouping plain/inodebits in granted list by their request modes
2486 and bits policy, thus improving the performance of search through
2490 Frequency : only if OST filesystem is corrupted
2492 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2493 Details : for a short time RPCs with bulk IO are in the replay list,
2494 but replay of bulk IOs is unimplemented. If the OST filesystem
2495 is corrupted due to disk cache incoherency and then replay is
2496 started it is possible to trip an assertion. Avoid putting
2497 committed RPCs into the replay list at all to avoid this issue.
2500 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2502 Description: System hang on startup
2503 Details : This bug allowed the liblustre (e.g. catamount) client to
2504 return to the app before handling all startup RPCs. This
2505 could leave the node unresponsive to lustre network traffic
2506 and manifested as a server ptllnd timeout.
2508 Severity : enhancement
2510 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2511 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2512 boolean to control whether to print peer debug info when a
2513 client's RPC times out.
2516 Frequency : only for kernels with patches from Lustre below 1.4.3
2518 Description: Remove old rdonly API
2519 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2522 Frequency : only for devices with external journals
2524 Description: Set external device read-only also
2525 Details : During a commanded failover stop, we set the disk device
2526 read-only while the server shuts down. We now also set any
2527 external journal device read-only at the same time.
2530 Frequency : when upgrading from 1.4 while trying to change parameters
2532 Description: The wrong (new) MDC name was used when setting parameters for
2533 upgraded MDT's. Also allows changing of OSC (and MDC)
2534 parameters if --writeconf is specified at tunefs upgrade time.
2537 Frequency : when setting specific ost indicies
2539 Description: QOS code breaks on skipped indicies
2540 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2541 created with --index need not be sequential.
2543 Severity : enhancement
2545 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2546 Details : The uninit_groups feature works in conjunction with the kernel
2547 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2548 up the pass1 processing of e2fsck. This is a read-only feature
2549 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2550 mount filesystems that have had this feature enabled.
2552 Severity : enhancement
2554 Description: Improve multi-block allocation algorithm to avoid fragmentation
2555 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2556 allocation locality and avoid filesystem fragmentation.
2558 ------------------------------------------------------------------------------
2560 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2562 * Support for kernels:
2563 2.4.21-47.0.1.EL (RHEL 3)
2564 2.6.5-7.283 (SLES 9)
2565 2.6.9-42.0.10.EL (RHEL 4)
2566 2.6.12.6 vanilla (kernel.org)
2567 2.6.16.27-0.9 (SLES 10)
2568 * Recommended e2fsprogs version: 1.39.cfs5
2570 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2574 Frequency : occasional, depends on client load and configuration
2575 Bugzilla : 12181, 12203
2576 Description: data loss for recently-modified files
2578 Details : In some cases it is possible that recently written or created
2579 files may not be written to disk in a timely manner (this should
2580 normally be within 30s unless client IO load is very high).
2581 The problem appears as zero-length files or files that are a
2582 multiple of 1MB in size after a client crash or client eviction
2583 that are missing data at the end of the file.
2585 This problem is more likely to be hit on clients where files are
2586 repeatedly created and unlinked in the same directory, clients
2587 have a large amount of RAM, have many CPUs, the filesystem has
2588 many OSTs, the clients are rebooted frequently, and/or the files
2589 are not accessed by other nodes after being written.
2591 The presence of the problem can be detected by looking at
2592 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2593 smaller than the second (nr_unused) then dirty files will not
2594 be flushed automatically to disk. "sync; sleep 10" should be
2595 run several times on the node before unmounting it to update
2596 Lustre (this is also safe to run on nodes without this problem).
2598 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2599 that can cause this same problem, so customers using that kernel
2600 also need to update the kernel in addition to Lustre. In order
2601 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2603 It is normal that files written just before a client crash (less
2604 than 30s) may not yet have been flushed to disk, even for local
2608 Frequency : frequent on thin XT3 nodes
2610 Description: UUID collision on thin XT3 Linux nodes
2611 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2612 randomly, since we relied on an insufficiently-seeded PRNG.
2617 Description: OSS hangs after "All ost request buffers busy"
2618 Details : A deadlock between quota and journal operations caused OSS
2619 hangs after printing "All ost request buffers busy."
2622 Frequency : always on liblustre builds
2624 Description: Cleanup compiler warnings on liblustre
2627 Frequency : always on liblustre builds on XT3
2629 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2633 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2634 > 32000 subdirectories
2635 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2636 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2637 subdirectory count crosses 32000. This will aid e2fsck to
2638 correctly handle more than 32000 subdirectories.
2641 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2643 Description: System hang on startup
2644 Details : This bug allowed the liblustre (e.g. catamount) client to
2645 return to the app before handling all startup RPCs. This
2646 could leave the node unresponsive to lustre network traffic
2647 and manifested as a server ptllnd timeout.
2649 Severity : enhancement
2651 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2652 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2653 boolean to control whether to print peer debug info when a
2654 client's RPC times out.
2659 Description: make O_SYNC working on 2.6 kernels
2660 Details : 2.6 kernels use different method for mark pages for write,
2661 so need add a code to lustre for O_SYNC work.
2666 Description: Failure to close file and release space on NFS
2667 Details : Put inode details into lock acquired in ll_intent_file_open.
2668 Use mdc_intent_lock in ll_intent_open to properly
2669 detect all kind of errors unhandled by mdc_enqueue.
2674 Description: proc file read during shutdown sometimes raced obd removal,
2676 Details : Add lock to prevent obd access after proc file removal.
2679 Frequency : Only for files larger than 4GB on 32-bit clients.
2681 Description: improperly doing page alignment of locks
2682 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2683 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2688 Description: RPCs being resent when they shouldn't be
2689 Details : Some RPCs that should not be resent are being resent. This
2690 can cause inconsistencies in the RPC state machine. Do not
2691 resend such requests.
2694 Frequency : rare, only with NFS export
2696 Description: Crash on NFS re-export node
2697 Details : under very unusual load conditions an assertion is hit in
2698 ll_intent_file_open()
2701 Frequency : only if OST filesystem is corrupted
2703 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2704 Details : for a short time RPCs with bulk IO are in the replay list,
2705 but replay of bulk IOs is unimplemented. If the OST filesystem
2706 is corrupted due to disk cache incoherency and then replay is
2707 started it is possible to trip an assertion. Avoid putting
2708 committed RPCs into the replay list at all to avoid this issue.
2713 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2714 Details : Large single O_DIRECT read and write calls can fail to allocate
2715 a sufficiently large buffer to process the request. In case of
2716 allocation failure the allocation is retried with a smaller
2717 buffer and broken into smaller requests.
2719 Severity : enhancement
2721 Description: Add -o localflock option to simulate old noflock behaviour.
2722 Details : This will achieve local-only flock/fcntl locks coherentness.
2727 Description: versioning check is incomplete
2728 Details : Checking the version difference of client vs. server, report
2729 error if the gap is too big.
2734 Description: add support PG_writeback bit
2735 Details : add support for PG_writeback bit for Lustre, for more carefull
2736 work with page cache in 2.6 kernel. This also fix some deadlocks
2737 and remove hack for work O_SYNC with 2.6 kernel.
2739 Severity : enhancement
2741 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2742 Details : The uninit_groups feature works in conjunction with the kernel
2743 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2744 up the pass1 processing of e2fsck. This is a read-only feature
2745 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2746 mount filesystems that have had this feature enabled.
2748 Severity : enhancement
2750 Description: Improve multi-block allocation algorithm to avoid fragmentation
2751 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2752 allocation locality and avoid filesystem fragmentation.
2754 ------------------------------------------------------------------------------
2756 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2758 * Support for kernels:
2759 2.6.9-42.0.3.EL (RHEL 4)
2760 2.6.5-7.276 (SLES 9)
2761 2.4.21-47.0.1.EL (RHEL 3)
2762 2.6.12.6 vanilla (kernel.org)
2763 2.6.16.21-0.8 (SLES10)
2764 * Recommended e2fsprogs version: 1.39.cfs2-0
2766 * The backwards-compatible /proc/sys/portals symlink has been removed
2767 in this release. Before upgrading, please ensure that you change
2768 any configuration scripts or /etc/sysctl.conf files that access
2769 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2770 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2771 in advance of the upgrade on any system running Lustre 1.4.6 or
2772 newer, since /proc/sys/lnet was added in that version.
2773 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2777 Frequency : only when quota is used
2779 Description: avoid scanning export list for quota master
2780 Details : Change the algorithms to avoid scanning export list in order
2781 to improve the efficiency.
2784 Frequency : MDS failover only, very rarely
2786 Description: "went back in time" messages on mds failover
2787 Details : The greatest transno may be lost when the current operation
2788 finishes with an error (transno==0) and the client's last_rcvd
2789 record is over-written. Save the greatest transno in the
2790 mds_last_transno for this case.
2793 Frequency : always for specific kernels and striping counts
2795 Description: client may get "Matching packet too big" without ACL support
2796 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2797 when trying to access files in certain configurations. The
2798 clients should in fact be denied when mounting because they do
2799 not understand ACLs.
2802 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2804 Description: many clients connecting with IO in progress causes connect timeouts
2805 Details : Avoid synchronous journal commits to avoid delays caused by many
2806 clients connecting/disconnecting when bulk IO is in progress.
2807 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2808 OST_IO_PORTAL to avoid delays behind potentially many pending
2812 Frequency : occasionally with multiple writers to a single file
2814 Description: shared writes to file may result in wrong size reported by stat()
2815 Details : Allow growing of kms when extent lock is cancelled
2818 Frequency : always with random mmap IO to multi-striped file
2820 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2821 Details : Only if the hole is at the end of OST object so that kms is too
2822 small. Fix is to increase kms accordingly in ll_nopage.
2825 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2827 Description: writes to a missing object would leak memory on the OST
2828 Details : If there is an inconsistency between the MDS and OST filesystems,
2829 such that the MDS references an object that doesn't exist, writes
2830 to that object will leak memory due to incorrect cleanup in the
2831 error handling path, eventually running out of memory on the OST.
2836 Description: Creating too long symlink causes lustre errors
2837 Details : Check symlink and name lengths before sending requests to MDS.
2840 Frequency : only if flock is enabled (not on by default)
2842 Description: posix locks not released on fd closure on 2.6.9+
2843 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2844 kernels, this caused such locks not to be released on fd close and
2845 then assertions on fs unmount about still used locks.
2848 Frequency : MDS failover only, very rarely
2850 Description: clients may get ASSERTION(granted_lock != NULL)
2851 Details : When request was taking a long time, and a client was resending
2852 a getattr by name lock request. The were multiple lock requests
2853 with the same client lock handle and
2854 mds_getattr_name->fixup_handle_for_resent_request found one of the
2855 lock handles but later failed with ASSERTION(granted_lock != NULL).
2860 Description: handle->h_buffer_credits > 0, assertion failure
2861 Details : h_buffer_credits is zero after truncate, causing assertion
2862 failure. This patch extends the transaction or creates a new
2866 Frequency : NFS re-export or patchless client
2867 Bugzilla : 11179, 10796
2868 Description: Crash on NFS re-export node (__d_move)
2869 Details : We do not want to hash the dentry if we don't have a lock.
2870 But if this dentry is later used in d_move, we'd hit uninitialised
2871 list head d_hash, so we just do this to init d_hash field but
2872 leave dentry unhashed.
2875 Frequency : NFS re-export or patchless client
2877 Description: NFS exports has problem with symbolic link
2878 Details : lustre client didn't properly install dentry when re-exported
2879 to NFS or running patchless client.
2882 Frequency : NFS re-export or patchless client
2884 Description: Various nfs/patchless fixes.
2885 Details : fixes reuse disconected alias for lookup process - this fixes
2886 warning "find_exported_dentry: npd != pd",
2887 fix permission error with open files at nfs.
2888 fix apply umask when do revalidate.
2891 Frequency : occasional
2893 Description: Crash on NFS re-export node
2894 Details : calling clear_page() on the wrong pointer triggered oops in
2895 generic_mapping_read().
2898 Frequency : rarely, using O_DIRECT IO
2900 Description: unaligned directio crashes client with LASSERT
2901 Details : check for unaligned buffers before trying any requests.
2904 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2906 Description: stale data returned from RAID cache
2907 Details : If only a small amount of IO is done to the RAID device before
2908 reading it again it is possible to get stale data from the RAID
2909 cache instead of reading it from disk.
2912 Frequency : always for sles10 kernel
2914 Description: sles10 support
2915 Details : ll_follow_link: compile fixes and using of nd_set_link
2916 under newer kernels.
2919 Frequency : depends on arch, kernel and compiler version, always on sles10
2922 Description: recursive or deep enough symlinks cause stack overflow
2923 Details : getting rid of large stack-allocated variable in
2927 Frequency : depends on hardware
2929 Description: lustre write performance loss in the SLES10 kernel
2930 Details : the performance loss is caused by using of write barriers in the
2931 ext3 code. The SLES10 kernel turns barrier support on by
2932 default. The fix is to undo that change for ldiskfs.
2934 ------------------------------------------------------------------------------
2936 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2938 * Support for kernels:
2939 2.6.9-42.0.3EL (RHEL 4)
2940 2.6.5-7.276 (SLES 9)
2941 2.4.21-47.0.1.EL (RHEL 3)
2942 2.6.12.6 vanilla (kernel.org)
2946 Frequency : quota enabled and large files being deleted
2948 Description: releasing more than 4GB of quota at once hangs OST
2949 Details : If a user deletes more than 4GB of files on a single OST it
2950 will cause the OST to spin in an infinite loop. Release
2951 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2956 Description: statfs data retrieved from /proc may be stale or zero
2957 Details : When reading per-device statfs data from /proc, in the
2958 {kbytes,files}_{total,free,avail} files, it may appear
2959 as zero or be out of date.
2962 Frequency : systems with MD RAID1 external journal devices
2964 Description: lconf's call to blkid is confused by RAID1 journal devices
2965 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2966 of returning all block devices that match the journal UUID.
2969 Frequency : always, for aggregate stripe size over 4GB
2971 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2972 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2973 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2976 Frequency : always if "lfs find" used on a local file/directory
2978 Description: "lfs find" segfaults if used on a local file/directory
2979 Details : The case where a directory component was not specified wasn't
2980 handled correctly. Handle this properly.
2983 Frequency : always on ppc64
2985 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2986 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2987 perform endianness conversion.
2990 Frequency : rarely (truncate to non-zero file size after write under load)
2991 Bugzilla : 10730, 10687
2992 Description: Files padded with zeros to next 4K multiple
2993 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2994 it is possible that files that are truncated to a non-zero size
2995 immediately after being written are filled with zero bytes beyond
2996 the truncated size. No file data is lost.
2998 Severity : enhancement
3000 Description: Allow recovery/failover for liblustre clients.
3001 Details : liblustre clients were unaware of failover configurations until
3004 Severity : enhancement
3006 Description: user file locks should fail when not mounting with flock option
3007 Details : Set up an error-returning stub in ll_file_operations.lock field
3008 to prevent incorrect behaviour when client is mounted without
3009 flock option. Also, set up properly f_op->flock field for
3013 Frequency : always on ia64
3015 Description: "lfs df" loops on printing out MDS statfs information
3016 Details : The obd_ioctl_data was not initialized and in some systems
3017 this caused a failure during the ioctl that did not return
3018 an error. Initialize the struct and return an error on failure.
3021 Frequency : SLES 9 only
3023 Description: Error of copying files with lustre special EAs as root
3024 Details : Client side always return success for setxattr call for lustre
3025 special xattr (currently only "trusted.lov").
3028 Frequency : rarely on clusters with both ia64+i386 clients
3030 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3031 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3032 clients are doing concurrent writes to the same file it is
3033 possible that the ia64 clients may LASSERT because the OST
3034 extent locks are not PAGE_SIZE aligned. Ensure that grown
3035 locks are always aligned on the request boundary.
3038 Frequency : specific use, occasional
3040 Description: Overwriting in use executable truncates on-disk binary image
3041 Details : If one node attempts to overwrite an executable in use by
3042 another node, we now correctly return ETXTBSY instead of
3043 truncating the file.
3045 Severity : enhancement
3047 Description: Async OSC create to avoid the blocking unnecessarily.
3048 Details : If a OST has no remain object, system will block on the creating
3049 when need to create a new object on this OST. Now, ways use
3050 pre-created objects when available, instead of blocking on an
3051 empty osc while others are not empty. If we must block, we block
3052 for the shortest possible period of time.
3057 Description: chmod on Lustre root is propagated to other clients
3058 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3059 invalid by the follow_mount time.
3062 Frequency : liblustre clients only
3064 Description: Race in 'instant cancel' lock handling could lead to such locks
3065 never to be granted in case of SMP MDS
3066 Details : Do not destroy not yet granted but cbpending locks in
3070 Frequency : replay/resend of open
3072 Description: non null lock assetion failure in mds_intent_policy
3073 Details : Trying to replay/resend lockless open requests resulted in
3074 mds_open() returning 0 with no lock. Now it sets a flag if
3075 a lock is going to be returned.
3077 Severity : enhancement
3079 Description: Checksum enhancements
3080 Details : New checksum enhancements allow for resending RPCs that failed
3083 Severity : enhancement
3085 Description: Tunables on number of dirty pages in cacche
3086 Details : Allow to set limit on number of dirty pages cached.
3091 Description: client crash on unmount - lock still has references
3092 Details : In some error handling cases it was possible to leak a lock
3093 reference on a client while accessing a file. This was not
3094 harmful to the client during operation, but would cause the
3095 client to crash when the filesystem is unmounted.
3098 Frequency : specific case, rare
3100 Description: ETXTBSY on mds though file not in use by client
3101 Details : ETXTBSY is no longer incorrectly returned when attempting to
3102 chmod or chown a directory that the user previously tried to
3103 execute or a currently-executing binary.
3106 Frequency : extremely rare except on liblustre-based clients
3108 Description: Lustre space not freed when files are deleted
3109 Details : Clean up open-unlinked files after client eviction. Previously
3110 the unlink was skipped and the files remained as orphans.
3115 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3116 Details : In some cases it was possible to send a blocking callback to
3117 a client doing a glimpse, even though that client didn't get
3118 a lock granted. When the glimpse lock is cancelled on the OST
3119 the freed lock is left on the waiting list and corrupted the list.
3122 Frequency : all core dumps
3124 Description: Broke core dumps to lustre
3125 Details : Negative dentry may be unhashed if parent does not have UPDATE
3126 lock, but some callers, e.g. do_coredump, expect dentry to be
3127 hashed after successful create, hash it in ll_create_it.
3129 ------------------------------------------------------------------------------
3131 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3133 * Support for kernels:
3134 2.6.9-42.0.2.EL (RHEL 4)
3135 2.6.5-7.276 (SLES 9)
3136 2.4.21-40.EL (RHEL 3)
3137 2.6.12.6 vanilla (kernel.org)
3141 Frequency : always on RHEL 3
3143 Description: Number of open files grows over time
3144 Details : The number of open files grows over time, whether or not
3145 Lustre is started. This was due to a filp leak introduced
3146 by one of our kernel patches.
3148 ------------------------------------------------------------------------------
3150 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3152 * Support for kernels:
3153 2.6.9-42.EL (RHEL 4)
3154 2.6.5-7.276 (SLES 9)
3155 2.4.21-40.EL (RHEL 3)
3156 2.6.12.6 vanilla (kernel.org)
3161 Bugzilla : 5719, 9635, 9792, 9684
3162 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3163 Details : If a server is under heavy load and cannot reply to new
3164 connection requests before the client resends the (re)connect,
3165 the connection handling code can behave badly if two service
3166 threads are concurrently handing separate (re)connections from
3167 the same client. Add better locking to the connection handling
3168 code, and ensure that only a single connection will be processed
3169 for a given client UUID, even if the lock is dropped.
3171 Severity : enhancement
3173 Description: add TCP zero-copy support to kernel
3174 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3175 sends if the hardware supports scatter-gather and checksumming.
3176 This allows socklnd to do client-write and server-read more
3177 efficiently and reduce CPU utilization from skbuf copying.
3180 Frequency : only if NFS exporting from client
3182 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3183 Details : If setting attributes on a file created under NFS that had
3184 never been opened it would be possible to oops the client
3185 if the file had no objects.
3188 Frequency : always for liblustre
3190 Description: liblustre client does MDS+OSTs setattr RPC for each write
3191 Details : When doing a write from a liblustre client, the client
3192 incorrectly issued an RPC to the MDS and each OST the file was
3193 striped over in order to update the timestamps. When writing
3194 with small chunks and many clients this could overwhelm the MDS
3195 with RPCs. In all cases it would slow down the write because
3196 these RPCs are unnecessary.
3198 Severity : enhancement
3200 Description: allow number of MDS service threads to be changed at module load
3201 Details : It is now possible to change the number of MDS service threads
3202 running. Adding "options mds mds_num_threads={N}" to the MDS's
3203 /etc/modprobe.conf will set the number of threads for the next
3204 time Lustre is restarted (assuming the "mds" module is also
3205 reloaded at that time). The default number of threads will
3206 stay the same, 32 for most systems.
3211 Description: OST crash if filesystem is unformatted or corrupt
3212 Details : If an OST is started on a device that has never been formatted
3213 or if the filesystem is corrupt and cannot even mount then the
3214 error handling cleanup routines would dereference a NULL pointer.
3219 Description: NULL pointer deref in llap_from_page.
3220 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3221 incorrect) mapping if the page was truncated/reclaimed while it was
3222 searched for. Check for this condition and skip such pages when
3223 doing readahead. Introduce extra check to llap_from_page() to
3224 verify page->mapping->host is non-NULL (so page is not anonymous).
3227 Frequency : Sometimes when using sys_sendfile
3229 Description: "page not covered by a lock" warnings from ll_readpage
3230 Details : sendfile called ll_readpage without right page locks present.
3231 Now we introduced ll_file_sendfile that does necessary locking
3232 around call to generic_file_sendfile() much like we do in
3236 Frequency : with certain MDS communication failures at client mount time
3238 Description: NULL pointer deref after failed client mount
3239 Details : a client connection request may delayed by the network layer
3240 and not be sent until after the PTLRPC layer has timed out the
3241 request. If the client fails the mount immediately it will try
3242 to clean up before the network times out the request. Add a
3243 reference from the request import to the obd device and delay
3244 the cleanup until the network drops the request.
3247 Frequency : occasionally during client (re)connect
3249 Description: assertion failure during client (re)connect
3250 Details : processing a client connection request may be delayed by the
3251 client or server longer than the client connect timeout. This
3252 causes the client to resend the connection request. If the
3253 original connection request is replied in this interval, the
3254 client may trip an assertion failure in ptlrpc_connect_interpret()
3255 which thought it would be the only running connect process.
3258 Frequency : only with obd_echo servers and clients that are rebooted
3260 Description: kernel BUG accessing uninitialized data structure
3261 Details : When running an obd_echo server it did not start the ping_evictor
3262 thread, and when a client was evicted an uninitialized data
3263 structure was accessed. Start the ping_evictor in the RPC
3264 service startup instead of the OBD startup.
3266 Severity : enhancement
3267 Bugzilla : 10193 (patchless)
3268 Description: Remove dependency on various unexported kernel interfaces.
3269 Details : No longer need reparent_to_init, exit_mm, exit_files,
3270 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3273 Frequency : rare (only users of deprecated and unsupported LDAP config)
3275 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3276 Details : LDAP apparently contains 'lustreName' attributes instead of
3277 'name'. A simple remapping of the name is sufficient.
3280 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3282 Description: waiting_locks_callback trips kernel BUG if client is evicted
3283 Details : Running with the dump_on_timeout debug flag turned on makes
3284 it possible that the waiting_locks_callback() can try to dump
3285 the Lustre kernel debug logs from an interrupt handler. Defer
3286 this log dumping to the expired_lock_main() thread.
3288 Severity : enhancement
3290 Description: Support NFS exporting on 2.6 kernels.
3291 Details : Implement non-rawops metadata methods for NFS server to use without
3292 changing NFS server code.
3295 Frequency : very rare (synthetic metadata workload only)
3297 Description: two racing renames might cause an MDS thread to deadlock
3298 Details : Running the "racer" program may cause one MDS thread to rename
3299 a file from being the source of a rename to being the target of
3300 a rename at exactly the same time that another thread is doing
3301 so, and the second thread has already enqueued these locks after
3302 doing a lookup of the target and is trying to relock them in
3303 order. Ensure that we don't try to re-lock the same resource.
3306 Frequency : only very large systems with liblustre clients
3308 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3309 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3310 all OSTs in parallel. This allows the eviction of stale liblustre
3311 clients to proceed much faster than if they were done in series,
3312 and also offers similar improvements for other set_info RPCs.
3317 Description: excessive CPU usage during initial read phase on client
3318 Details : During the initial read phase on a client, it would agressively
3319 retry readahead on the file, consuming too much CPU and impacting
3320 performance (since 1.4.5.8). Improve the readahead algorithm
3321 to avoid this, and also improve some other common cases (read
3322 of small files in particular, where "small" is files smaller than
3323 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3328 Description: MDS crash when receiving packet with unknown intent.
3329 Details : Do not LBUG in unknown intent case, just return -EFAULT
3331 Severity : enhancement
3332 Bugzilla : 9293, 9385
3333 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3334 Details : Do not serialize getattr (non-intent version) and statfs.
3337 Frequency : occasional, when OST network is overloaded/intermittent
3339 Description: client evicted by OST after bulk IO timeout
3340 Details : If a client sends a bulk IO request (read or write) the OST
3341 may evict the client if it is unresposive to its data GET/PUT
3342 request. This is incorrect if the network is overloaded (takes
3343 too long to transfer the RPC data) or dropped the OST GET/PUT
3344 request. There is no need to evict the client at all, since
3345 the pinger and/or lock callbacks will handle this, and the
3346 client can restart the bulk request.
3349 Frequency : Always when mmapping file with no objects
3351 Description: client crashes when mmapping file with no objects
3352 Details : Check that we actually have objects in a file before doing any
3353 operations on objects in ll_vm_open, ll_vm_close and
3359 Description: Request leak when working with deleted CWD
3360 Details : Introduce advanced request refcount tracking for requests
3361 referenced from lustre intent.
3363 Severity : Enhancement
3365 Description: Cache open file handles on client.
3366 Details : MDS now will return special lock along with openhandle, if
3367 requested and client is allowed to hold openhandle, even if unused,
3368 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3369 closing files for every read/write openration.
3371 Severity : Enhancement
3373 Description: Cache open negative dentries on client when possible.
3374 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3375 negative dentries on lock revocation.
3380 Description: Remounting a client read-only wasn't possible with a zconf mount
3381 Details : It wasn't possible to remount a client read-only with llmount.
3383 Severity : enhancement
3384 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3385 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3386 MPICH is included to add Lustre-specific ADIO interfaces.
3387 This is based closely on the UFS ADIO layer and only differs
3388 in file creation, in order to allow the OST striping to be set.
3389 This is user-contributed code and not supported by CFS.
3394 Description: extended inode attributes (immutable, append-only) work improperly
3395 when 2.4 and 2.6 kernels are used on client/server or vice versa
3396 Details : Introduce kernel-independent values for these flags.
3398 Severity : enhancement
3401 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3402 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3403 granularity. Now, Fractional MB granularity is be supported,
3404 this is very useful for low memory system.
3406 Severity : enhancement
3408 Description: Getattr by fid
3409 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3410 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3411 have locks and avoids getattr rpc requests for it on every stat.
3414 Frequency : Always, for filesystems larger than 2TB
3416 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3417 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3418 filesystems larger than 2TB to be mounted with mballoc enabled.
3421 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3423 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3424 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3425 filesystems larger than 2TB is mounted then the kernel will
3426 silently corrupt the start of the filesystem. CONFIG_LBD is
3427 enabled for all CFS-supported kernels, but the possibility of
3428 this happening with a modified kernel config exists.
3430 Severity : enhancement
3432 Description: add client O_DIRECT support for 2.6 kernels
3433 Details : It is now possible to do O_DIRECT reads and writes to files
3434 in the Lustre client mountpoint on 2.6 kernel clients.
3436 Severity : enhancement
3438 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3439 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3440 parallel, not waiting for response from every OST before sending
3441 a rpc to the next OST.
3446 Description: setattr vs write race when updating file timestamps
3447 Details : Client processes that update a file timestamp into the past
3448 right after writing to the file (e.g. tar) it is possible that
3449 the updated file modification time can be reset to the current
3450 time due to a race between processing the setattr and write RPC.
3452 Severity : enhancement
3454 Description: Bring 'lfs find' closer in line with regular Linux find.
3455 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3456 -print0 options and obtains all the needed info through the lustre
3459 Severity : enhancement
3461 Description: support up to 1024 configured devices on one node
3462 Details : change obd_dev array from statically allocated to dynamically
3463 allocated structs as they are first used to reduce memory usage
3468 Description: Flush dirty partially truncated pages during truncate
3469 Details : Immediatelly flush partially truncated pages in filter_setattr,
3470 this way we completely avoid having any pages in page cache on OST
3471 and can retire ugly workarounds during writes to flush such pages.
3476 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3477 Details : protect inode from truncation within vfs_unlink() context
3478 just take a reference before calling vfs_unlink() and release it
3479 when parent's i_sem is free.
3484 Description: last_id value checked outside lock on OST caused LASSERT failure
3485 Details : If there were multiple MDS->OST object precreate requests in
3486 flight, it was possible that the OST's last object id was checked
3487 outside a lock and incorrectly tripped an assertion. Move checks
3488 inside locks, and discard old precreate requests.
3491 Frequency : always, if extents are used on OSTs
3493 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3494 Details : OSTs using the extents format would not zero the high 16 bits of
3495 the index physical block number. This is not a problem for any
3496 OST filesystems smaller than 16TB, and no kernels support ext3
3497 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3498 new/modified files) and can be fixed for existing filesystems
3499 with e2fsprogs-1.39-cfs1.
3504 Description: import connection selection may be incorrect if timer wraps
3505 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3506 import connections to be ignored if the 32-bit jiffies counter
3507 wraps. Use a 64-bit jiffies counter.
3510 Frequency : very large clusters immediately after boot
3512 Description: LNET request buffers exhausted under heavy short-term load
3513 Details : If a large number of client requests are generated on a service
3514 that has previously never seen so many requests it is possible
3515 that the request buffer growth cannot keep up with the spike in
3516 demand. Instead of dropping incoming requests, they are held in
3517 the LND until the RPC service can accept more requests.
3520 Frequency : Sometimes during replay
3522 Description: Assertion failure in ll_local_open after replay.
3523 Details : If replay happened on an open request reply before we were able
3524 to set replay handler, reply will become not swabbed tripping the
3525 assertion in ll_local_open. Now we set the handler right after
3526 recognising of open request
3529 Frequency : very rare
3531 Description: kernel reports "badness in vsnprintf"
3532 Details : Reading from the "recovery_status" /proc file in small chunks
3533 may cause a negative length in lprocfs_obd_rd_recovery_status()
3534 call to vsnprintf() (which is otherwise harmless). Exit early
3535 if there is no more space in the output buffer.
3537 Severity : enhancement
3539 Description: clear OBD RPC statistics by writing to them
3540 Details : It is now possible to clear the OBD RPC statistics by writing
3541 to the "stats" file.
3546 Description: Client mtime is not the same on different clients after utimes
3547 Details : In some cases, the client was using the utimes() syscall on
3548 a file cached on another node. The clients now validate the
3549 ctime from the MDS + OSTs to determine which one is right.
3554 Description: Inability to activate failout mode
3555 Details : lconf script incorrectly assumed that in pythong string's numeric
3556 value is used in comparisons.
3559 Frequency : always with multiple stripes per file
3561 Description: Inefficient object allocation for mutli-stripe files
3562 Details : When selecting which OSTs to stripe files over, for files with
3563 a stripe count that divides evenly into the number of OSTs,
3564 the MDS is always picking the same starting OST for each file.
3565 Return the OST selection heuristic to the original design.
3570 Description: mount failures may take full timeout to return an error
3571 Details : Under some heavy load conditions it is possible that a
3572 failed mount can wait for the full obd_timeout interval,
3573 possibly several minutes, before reporting an error.
3574 Instead return an error as soon as the status is known.
3576 Frequency : quota enabled and large files being deleted
3578 Description: releasing more than 4GB of quota at once hangs OST
3579 Details : If a user deletes more than 4GB of files on a single OST it
3580 will cause the OST to spin in an infinite loop. Release
3581 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3586 Description: statfs data retrieved from /proc may be stale or zero
3587 Details : When reading per-device statfs data from /proc, in the
3588 {kbytes,files}_{total,free,avail} files, it may appear
3589 as zero or be out of date.
3592 Frequency : systems with MD RAID1 external journal devices
3594 Description: lconf's call to blkid is confused by RAID1 journal devices
3595 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3596 of returning all block devices that match the journal UUID.
3599 Frequency : always, for aggregate stripe size over 4GB
3601 Description: assertion fails when trying to use 4GB stripe size
3602 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3603 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3606 Frequency : always on ppc64
3608 Description: the first write on an ext3 filesystem with mballoc got stuck
3609 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3610 perform endianness conversion.
3612 ------------------------------------------------------------------------------
3614 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3616 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3617 INTEROPERATE with older versions automatically. Please read the
3618 user documentation before upgrading any part of a live system.
3619 * WARNING: Lustre networking configuration changes are required with
3620 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3623 * Support for kernels:
3624 2.6.9-22.0.2.EL (RHEL 4)
3625 2.6.5-7.244 (SLES 9)
3626 2.6.12.6 vanilla (kernel.org)
3629 Severity : enhancement
3630 Bugzilla : 7981/8208
3631 Description: Introduced Lustre Networking (LNET)
3632 Details : LNET is new networking infrastructure for Lustre, it includes
3633 a reorganized network configuration mode (see the user
3634 documentation for full details) as well as support for routing
3635 between different network fabrics. Lustre Networking Devices
3636 (LNDs) for the supported network fabrics have also been
3637 created for this new infrastructure.
3639 Severity : enhancement
3640 Description: Introduced Access control lists
3641 Details : clients can set ACLs on files and directories in order to have
3642 more fine-grained permissions than the standard Unix UGO+RWX.
3643 The MDS must be started with the "-o acl" mount option.
3645 Severity : enhancement
3646 Description: Introduced filesystem quotas
3647 Details : Administrators may now establish per-user quotas on the
3650 Severity : enhancement
3652 Description: Configuration change for the XT3
3653 The PTLLND is now used to run Lustre over Portals on the XT3
3654 The configure option(s) --with-cray-portals are no longer used.
3655 Rather --with-portals=<path-to-portals-includes> is used to
3656 enable building on the XT3. In addition to enable XT3 specific
3657 features the option --enable-cray-xt3 must be used.
3662 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3663 Details : A race between allocating a new llcd and re-getting the llcd_lock
3664 allowed another thread to grab newly-allocated llcd.
3666 Severity : enhancement
3668 Description: 2.6 OST async journal commit and locking fix to improve performance
3669 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3670 2.6 kernels are now async as they already were in 2.4 kernels so
3671 that they can commit concurrently with the network bulk transfer.
3672 For block-allocated files the filter allocation semaphore is held
3673 to avoid filesystem fragmentation during allocation. BKL lock
3674 removed for 2.6 xattr operations where it is no longer needed.
3679 Description: lconf incorrectly determined whether two IP networks could talk
3680 Details : In some more complicated routing and multiple-network
3681 configurations, lconf will avoid trying to make a network
3682 connection to a disjoint part of the IP space. It was doing the
3683 math incorrectly for one set of cases.
3688 Description: Fix for potential infinite loop processing records in an llog.
3689 Details : If an llog record is corrupted/zeroed, it is possible to loop
3690 forever in llog_process(). Validate the llog record length
3691 and skip the remainder of the block on error.
3694 Frequency : occasional (liblustre only)
3696 Description: liblustre could not open files whose last component is a symlink
3697 Details : sysio_path_walk() would incorrectly pass the open intent to
3698 intermediate path components.
3701 Frequency : rare (liblustre only with non-standard tuning)
3702 Bugzilla : 7201 (7350)
3703 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3704 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3705 after it has been released, only do this for referenced locks
3707 Severity : enhancement
3709 Description: specifying an (invalid) directory default stripe_size of -1
3710 would reset the directory default striping
3711 Details : stripe_size -1 was used internally to signal directory stripe
3712 removal, now use "all default" to signal dir stripe removal
3713 as a directory striping of "all default" is not useful
3716 Frequency : common for large clusters running liblustre clients
3718 Description: doing an ls when liblustre clients are running is slow
3719 Details : sending a glimpse AST to a liblustre client waits for every AST
3720 to time out, as liblustre clients will not respond. Since they
3721 cannot cache data we refresh the OST lock LVB from disk instead.
3723 Severity : enhancement
3725 Description: doing an ls at the same time as file IO can be slow
3726 Details : enqueue and other "small" requests can be blocked behind many
3727 large IO requests. Create a new OST IO portal for non-IO
3728 requests so they can be processed faster.
3731 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3733 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3734 created file with mode 000
3735 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3736 allow an ftruncate() as a non-root user to a file with mode 000.
3737 Lustre now respects this flag to disable mode checks when
3738 truncating a file owned by the user
3741 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3744 Description: Revoking locks from clients that went dead or catatonic might take
3746 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3747 cancellation of such locks instant on servers without waiting for
3748 any reply from clients. Clients drops these locks when cancel
3749 notification from server is received without replying.
3752 Frequency : liblustre-only, when liblustre client dies or becomes busy
3754 Description: Doing ls on Linux clients can take a long time with active
3756 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3757 granting such locks to it in the first place if possible. Locks
3758 are taken by proxy on the OST during the read or write and
3759 dropped immediately afterward. Add connect flags handling, do
3760 not grant locks to liblustre clients for glimpse ASTs.
3762 Severity : enhancement
3764 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3765 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3766 sequential reads and behaves badly with applications doing random
3767 reads. Improve it by reading ahead at least read region, and
3768 avoiding excessive large RPC for small reads.
3770 Severity : enhancement
3772 Description: Creating more than 1000 files for a single job may cause a load
3773 imbalance on the OSTs if there are also a large number of OSTs.
3774 Details : qos_prep_create() uses an OST index reseed value that is an
3775 even multiple of the number of available OSTs so that if the
3776 reseed happens in the middle of the object allocation it will
3777 still utilize the OSTs as uniformly as possible.
3782 Description: OST or MDS may oops in ping_evictor_main()
3783 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3784 but doesn't restart at beginning of obd_exports_timed list
3787 Severity : enhancement
3789 Description: improve by-nid export eviction on the MDS and OST
3790 Details : allow multiple exports with the same NID to be evicted at one
3791 time without re-searching the exports list.
3794 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3796 Description: MDS may oops in groups_free()
3797 Details : in rare race conditions a newly allocated group_info struct is
3798 freed again, and this can be NULL. The 2.4 compatibility code
3799 for groups_free() checked for a NULL pointer, but 2.6 did not.
3802 Frequency : common for liblustre clients doing little filesystem IO
3803 Bugzilla : 9352, 7313
3804 Description: server may evict liblustre clients accessing contended locks
3805 Details : if a client is granted a lock or receives a completion AST
3806 with a blocking AST already set it would not reply to the AST
3807 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3810 Frequency : lfs setstripe, only systems with more than 160 OSTs
3812 Description: unable to set striping with a starting offset beyond OST 160
3813 Details : llapi_create_file() incorrectly limited the starting stripe
3814 index to the maximum single-file stripe count.
3817 Frequency : LDAP users only
3819 Description: lconf did not handle in-kernel recovery with LDAP properly
3820 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3822 Severity : enhancement
3824 Description: bind OST threads to NUMA nodes to improve performance
3825 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3826 node and do their allocations there to localize memory access
3828 Severity : enhancement
3830 Description: llmount can determine client NID directly from Myrinet (GM)
3831 Details : the client NID code from gmnalnid was moved directly into
3832 llmount, removing the need to use this or specifying the
3833 client NID explicitly when mounting GM clients with zeroconf
3836 Frequency : if client is started with down MDS
3838 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3839 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3840 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3841 ptlrpc_import_delayed_req() because we didn't check if the
3842 request was interrupted, and we also didn't break out of the
3843 event loop if there was no timeout
3848 Description: data loss during non-page-aligned writes to a single file from
3849 both multiple nodes and multiple threads on one node at same time
3850 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3851 inconsistency led to false short-reads, that were cached and later
3852 used by ->prepare_write() to fill in partially written page,
3853 leading to data loss.
3856 Frequency : always, if lconf --abort_recovery used
3858 Description: lconf --abort_recovery fails with 'Operation not supported'
3859 Details : lconf was attempting to abort recovery on the MDT device and not
3862 Severity : enhancement
3864 Description: remove cleanup logs
3865 Details : replace lconf-generated cleanup logs with lustre internal
3866 cleanup routines. Eliminates the need for client-cleanup and
3869 Severity : enhancement
3871 Description: add support for EAs (user and system) on lustre filesystems
3872 Details : it is now possible to store extended attributes in the Lustre
3873 client filesystem, and with the user_xattr mount option it
3874 is possible to allow users to store EAs on their files also
3876 Severity : enhancement
3878 Description: Add possibility (config option) to show minimal available OST free
3880 Details : When compiled with --enable-mindf configure option, statfs(2)
3881 (and so, df) will return least minimal free space available from
3882 all OSTs as amount of free space on FS, instead of summary of
3883 free spaces of all OSTs.
3885 Severity : enhancement
3887 Description: do not expand extent locks acquired on OST-side
3888 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3889 by server: they are not cached anyway.
3892 Frequency : when mmap is used/binaries executed from Lustre
3894 Description: Unmmap pages before throwing them away from read cache.
3895 Details : llap_shrink cache now attempts to unmap pages before discarding
3896 them (if unmapping failed - do not discard). SLES9 kernel has
3897 extra checks that trigger if this unmapping is not done first.
3902 Description: lconf didn't resolve symlinks before checking to see whether a
3903 given mountpoint was already in use
3906 Frequency : when migrating failover services
3907 Bugzilla : 6395, 9514
3908 Description: When migrating a subset of services from a node (e.g. failback
3909 from a failover service node) the remaining services would
3910 time out and evict clients.
3911 Details : lconf --force (implied by --failover) sets the global obd_timeout
3912 to 5 seconds in order to quickly disconnect, but this caused
3913 other RPCs to time out too quickly. Do not change the global
3914 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3916 Severity : enhancement
3917 Frequency : if MDS is started with down OST
3918 Bugzilla : 9439,5706
3919 Description: Allow startup/shutdown of an MDS without depending on the
3920 availability of the OSTs.
3921 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3922 Add appropriate locking and lov-osc refcounts for safe
3923 cleaning. Add osc abort_inflight calls in case the
3924 synchronize never started.
3927 Frequency : occasional (Cray XT3 only)
3929 Description: root not authorized to access files in CRAY_PORTALS environment
3930 Details : The client process capabilities were not honoured on the MDS in
3931 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3932 been accessed by an authorized user then root was able to access
3933 the file on the local client also. The root user capabilities
3934 are now allowed on the MDS, as this environment has secure UID.
3937 Frequency : occasional
3939 Description: ldiskfs "too long searching" message happens too often
3940 Details : A debugging message (otherwise harmless) prints too often on
3941 the OST console. This has been reduced to only happen when
3942 there are fragmentation problems on the filesystem.
3947 Description: Division by zero in statfs when all OSCs are inactive
3948 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3949 lov_statfs divides by value returned by lov_get_stripecnt().
3953 Bugzilla : 9489, 3273
3954 Description: First write from each client to each OST was only 4kB in size,
3955 to initialize client writeback cache, which caused sub-optimal
3956 RPCs and poor layout on disk for the first writen file.
3957 Details : Clients now request an initial cache grant at (re)connect time
3958 and so that they can start streaming writes to the cache right
3959 away and always do full-sized RPCs if there is enough data.
3960 If the OST is rebooted the client also re-establishes its grant
3961 so that client cached writes will be honoured under the grant.
3966 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3967 Details : Now I/O RPCs go to different portal number and (presumably) fast
3968 lock requests (and glimses) and other RPCs get their own service
3969 threads pool that should be able to service those RPCs
3972 Severity : enhancement
3974 Description: Ability to exchange lustre version between client and servers and
3975 issue warnings at client side if client is too old. Also for
3976 liblustre clients there is ability to refuse connection of too old
3978 Details : New 'version' field is added to connect data structure that is
3979 filled with version info. That info is later checked by server and
3983 Frequency : rare, liblustre only.
3984 Bugzilla : 9296, 9581
3985 Description: Two simultaneous writes from liblustre at offset within same page
3986 might proceed at the same time overwriting eachother with stale
3988 Details : I/O lock withing llu_file_prwv was released too early, before data
3989 actually was hitting the wire. Extended lock-holding time until
3990 server acknowledges receiving data.
3993 Frequency : extremely rare. Never observed in practice.
3995 Description: avoid generating lustre_handle cookie of 0.
3996 Details : class_handle_hash() generates handle cookies by incrementing
3997 global counter, and can hit 0 occasionaly (this is unlikely, but
3998 not impossible, because initial value of cookie counter is
3999 selected randonly). Value of 0 is used as a sentinel meaning
4000 "unassigned handle" --- avoid it. Also coalesce two critical
4001 sections in this function into one.
4003 Severity : enhancement
4005 Description: allow liblustre clients to delegate truncate locking to OST
4006 Details : To avoid overhead of locking, liblustre client instructs OST to
4007 take extent lock in ost_punch() on client's behalf. New connection
4008 flag is added to handle backward compatibility.
4010 Severity : enhancement
4011 Bugzilla : 4928, 7341, 9758
4012 Description: allow number of OST service threads to be specified
4013 Details : a module parameter allows the number of OST service threads
4014 to be specified via "options ost ost_num_threads={N}" in the
4015 OSS's /etc/modules.conf or /etc/modprobe.conf.
4019 Bugzilla : 6146, 9635, 9895
4020 Description: servers crash with bad pointer in target_handle_connect()
4021 Details : In rare cases when a client is reconnecting it was possible that
4022 the connection request was the last reference for that export.
4023 We would temporarily drop the export reference and get a new
4024 one, but this may have been the last reference and the export
4025 was just destroyed. Get new reference before dropping old one.
4027 Severity : enhancement
4028 Frequency : if client is started with failover MDS
4030 Description: Allow multiple MDS hostnames in the mount command
4031 Details : Try to read the configuration from all specified MDS
4032 hostnames during a client mount in case the "primary"
4035 Severity : enhancement
4037 Description: Stop sending data to evicted clients as soon as possible.
4038 Details : Check if the client we are about to send or are sending data to
4039 was evicted already. (Check is done every second of waiting,
4040 for which l_wait_event interface was extended to allow checking
4041 of exit condition at specified intervals).
4044 Frequency : rare, normally only when NFS exporting is done from client
4046 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4048 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4051 Frequency : always, if srand() is called before liblustre initialization
4053 Description: Liblustre uses system PRNG disturbing its usage by user application
4054 Details : Introduce internal to lustre fast and high-quality PRNG for
4055 lustre usage and make liblustre and some other places in generic
4056 lustre code to use it.
4058 Severity : enhancement
4059 Bugzilla : 9477, 9557, 9870
4060 Description: Verify that the MDS configuration logs are updated when xml is
4061 Details : Check if the .xml configuration logs are newer than the config
4062 logs stored on the MDS and report an error if this is the case.
4063 Request --write-conf, or allow starting with --old_conf.
4065 Severity : enhancement
4067 Description: Handle symlinks in the path when checking if Lustre is mounted.
4068 Details : Resolve intermediate symlinks when checking if a client has
4069 mounted a filesystem to avoid duplicate client mounts.
4074 Description: lconf can hit an error exception but still return success.
4075 Details : The lconf command catches the Command error exception at the top
4076 level script context and will exit with the associated exit
4077 status, but doesn't ensure that this exit status is non-zero.
4082 Description: failure of ptlrpc thread startup can cause oops
4083 Details : Starting a ptlrpc service thread can fail if there are a large
4084 number of threads or the server memory is very fragmented.
4085 Handle this without oopsing.
4088 Frequency : always, only if liblustre and non-default acceptor port was used
4090 Description: liblustre cannot connect to servers with non-default acceptor port
4091 Details : tcpnal_set_default_params() was not called and was therefore
4092 ignoring the environment varaible TCPNAL_PORT, as well as other
4093 TCPNAL_ environment variables
4098 Description: two objects could be created on the same OST for a single file
4099 Details : If an OST is down, in some cases it was possible to create two
4100 objects on a single OST for a single file. No problems other
4101 than potential performance impact and spurious error messages.
4105 Bugzilla : 5681, 9562
4106 Description: Client may oops in ll_unhash_aliases
4107 Details : Client dcache may become inconsistent in race condition.
4108 In some cases "getcwd" can fail if the current directory is
4114 Description: Inode refcounting problems in NFS export code
4115 Details : link_raw functions used to call d_instantiate without obtaining
4116 extra inode reference first.
4120 Bugzilla : 9942, 9903
4121 Description: Referencing freed requests leading to crash, memleaks with NFS.
4122 Details : We used to require that call to ll_revalidate_it was always
4123 followed by ll_lookup_it. Also with revalidate_special() it is
4124 possible to call ll_revalidate_it() twice for the same dentry
4125 even if first occurence returned success. This fix changes semantic
4126 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4127 reference on a request referred from the intent.
4128 ll_intent_release() then releases such a request.
4131 Frequency : rare, normally benchmark loads only
4133 Description: unlinked inodes were kept in memory on the client
4134 Details : If a client is repeatedly creating and unlinking files it
4135 can accumulate a lot of stale inodes in the inode slab cache.
4136 If there is no other client load running this can cause the
4137 client node to run out of memory. Instead flush old inodes
4138 from client cache that have the same inode number as a new inode.
4141 Frequency : SLES9 2.6.5 kernel and long filenames only
4142 Bugzilla : 9969, 10379
4143 Description: utime reports stale NFS file handle
4144 Details : SLES9 uses out-of-dentry names in some cases, which confused
4145 the lustre dentry revalidation. Change it to always use the
4149 Frequency : rare, unless heavy write-truncate concurrency is continuous
4150 Bugzilla : 4180, 6984, 7171, 9963, 9331
4151 Description: OST becomes very slow and/or deadlocked during object unlink
4152 Details : filter_destroy() was holding onto the parent directory lock
4153 while truncating+unlinking objects. For very large objects this
4154 may block other threads for a long time and slow overall OST
4155 responsiveness. It may also be possible to get a lock ordering
4156 deadlock in this case, or run out of journal credits because of
4157 the combined truncate+unlink. Solution is to do object truncate
4158 first in one transaction without parent lock, and then do the
4159 final unlink in a new transaction with the parent lock. This
4160 reduces the lock hold time dramatically.
4163 Frequency : rare, 2.4 kernels only
4165 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4166 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4167 before unmounting their filesystem in order to avoid deadlock.
4168 The kernel_locked() function in 2.4 kernels only checks whether
4169 the kernel lock is held, not whether it is this process that is
4170 holding it as 2.6 kernels do.
4175 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4176 Details : The client ptlrpc code may be trying to reconnect to a down
4177 server before a previous connection attempt has timed out.
4178 Increase the reconnect interval to be longer than the connection
4179 timeout interval to avoid sending duplicate connections to
4183 Frequency : echo_client brw_test command
4185 Description: fix echo_client to work with OST preallocated code
4186 Details : OST preallocation code (5137) didn't take echo_client IO path
4187 into account: echo_client calls filter methods outside of any
4188 OST thread and, hence, there is no per-thread preallocated
4189 pages and buffers to use. Solution: hijack pga pages for IO. As
4190 a byproduct, this avoids unnecessary data copying.
4194 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4195 Description: Client can oops in mdc_commit_close() after open replay
4196 Details : It was possible for the MDS to return an open request with no
4197 transaction number in mds_finish_transno() if the client was
4198 evicted, but without actually returning an error. Clients
4199 would later try to replay that open and may trip an assertion
4200 Simplify the client close codepath, and always return an error
4201 from the MDS in case the open is not successful.
4204 Frequency : rare, 2.6 OSTs only
4206 Description: OST may deadlock under high load on fragmented files
4207 Details : If there was a heavy load and highly-fragmented OST filesystems
4208 it was possible to have all the OST threads deadlock waiting on
4209 allocation of biovecs, because the biovecs were not released
4210 until the entire RPC IO was completed. Instead, release biovecs
4211 as soon as they are complete to ensure forward IO progress.
4213 Severity : enhancement
4215 Description: Support for specifying external journal device at mount
4216 Details : If an OST or MDS device is formatted with an external journal
4217 device, this device major/minor is stored in the ext3 superblock
4218 and may not be valid for failover. Allow detecting and
4219 specifying the external journal at mount time.
4224 Description: Mounting an MDS with pending unlinked files may cause oops
4225 Details : target_finish_recovery() calls mds_postrecov() which returned
4226 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4227 considers this an error and immediately begins cleaning up the
4228 lov, just after starting the mds_lov process
4230 Severity : enhancement
4232 Description: Implement 'lfs df' to report actual free space on per-OST basis
4233 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4234 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4235 usage of objects; '-h' to report in human readable format.
4237 ------------------------------------------------------------------------------
4239 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4246 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4247 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4248 at startup during orphan unlinking. Instead, skip buddy update
4249 before setup, as it will be regenerated after recovery is complete.
4254 Description: If an OST is inactive, its locks might reference stale inodes.
4255 Details : lov_change_cbdata() must iterate over all namespaces, even if
4256 they are inactive to clear inode references from the lock.
4258 Severity : enhancement
4259 Frequency : occasional, if non-standard max_dirty_mb used
4261 Description: Client will block write RPCs if not enough grant
4262 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4263 then the client will block writes while waiting for another RPC
4264 to complete instead of consuming its dirty limit. With change
4265 we get improved performance when max_dirty_mb is small.
4267 Severity : enhancement
4268 Bugzilla : 3389, 6253
4269 Description: Add support for supplementary groups on the MDS.
4270 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4271 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4272 lookups for user supplementary groups into a cache.
4276 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4277 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4278 for liblustre O_EXCL is enforced by MDS.
4282 Description: readdir never returns NULL in liblustre.
4283 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4284 dentry in struct dirent.
4288 Frequency : liblustre only, depends on application IO pattern
4289 Description: liblustre clients evicted if not contacting servers
4290 Details : Don't put liblustre clients into the ping_evictor list, so
4291 they will not be evicted by the pinger ever.
4293 Severity : enhancement
4295 Description: Add ability to evict clients by NID from MDS.
4296 Details : By echoing "nid:$NID" string into
4297 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4298 $NID would be instantly evicted from this MDS and from all active
4299 OSTs connected to it.
4303 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4304 Details : lookup_it_finish() used to query file size from OSTs that was not
4309 Description: service threads change working directory to that of init
4310 Details : Starting lustre service threads may pin the working directory
4311 of the parent thread, making that filesystem busy. Threads
4312 now change to the working directory of init to avoid this.
4316 Frequency : during shutdown only
4317 Description: shutdown with a failed MDS or OST can cause unmount to hang
4318 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4323 Frequency : 2.6 only
4324 Description: chmod/chown may include an extra supplementary group
4325 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4326 supplementary group and if none is specified this is used.
4329 Bugzilla : 5479 (6816)
4331 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4332 Details : If lookup is in progress on a file that is unlinked we might try
4333 to revalidate the inode and fail in revalidate after lookup is
4334 complete and ll_file_open() enqueues the open again but
4335 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4338 Frequency : always, if lconf --abort_recovery used
4340 Description: lconf --abort_recovery fails with 'Operation not supported'
4341 Details : lconf was attempting to abort recovery on the MDT device and not
4344 ------------------------------------------------------------------------------
4346 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4351 Frequency : rare (only unsupported configurations with a node running as an
4353 Bugzilla : 6514, 5137
4354 Description: Mounting a Lustre file system on a node running as an OST could
4356 Details : OSTs now preallocates memory needed to write out data at
4357 startup, instead of when needed, to avoid having to
4358 allocate memory in possibly low memory situations.
4359 Specifically, if the file system is mounted on on OST,
4360 memory pressure could force it to try to write out data,
4361 which it needed to allocate memory to do. Due to the low
4362 memory, it would be unable to do so and the node would
4363 become unresponsive.
4365 Severity : enhancement
4367 Description: Addition of lconf --service command line option
4368 Details : lconf now accepts a '--service <arg>' option, which is
4369 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4371 Severity : enhancement
4373 Description: Failover mode is now the default for OSTs.
4374 Details : By default, OSTs will now run in failover mode. To return to
4375 the old behaviour, add '--failout' to the lmc line for OSTs.
4377 Severity : enhancement
4379 Description: Health checks are now provided for MDS and OSTs
4380 Details : Additional detailed health check information on MSD and OSTs
4381 is now provided through the procfs health_check value.
4384 Frequency : occasional, depends on IO load
4386 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4387 after numerous create/delete cycles
4388 Details : The ext3 inode allocation policy would not allocate new inodes
4389 very well on the OSTs because there are no new directories
4390 being created. Instead we look for groups with free space if
4391 the parent directories are nearly full.
4396 Description: Network or server problems during mount may cause partially
4397 mounted clients instead of returning an error.
4398 Details : The config llog parsing code may overwrite the error return
4399 code during mount error handling, returning success instead
4405 Description: MDS can fail to allocate large reply buffers
4406 Details : After long uptimes the MDS can fail to allocate large reply
4407 buffers (e.g. zconf client mount config records) due to memory
4408 fragmentation or consumption by the buffer cache. Preallocate
4409 some large reply buffers so that these replies can be sent even
4410 under memory pressure.
4414 Frequency : rare (liblustre)
4415 Description: fsx running with liblustre complained that using truncate() to
4416 extend the file doesn't work. This patch corrects that issue.
4417 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4418 ATTR_SIZE and lsm use in llu_setattr_raw.
4422 Frequency : rare, only 2.6 kernels
4423 Description: Unusual file access patterns on the MDS may result in inode
4424 data being lost in very rare circumstances.
4425 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4426 code in the 2.6 kernel caused the inode and/or EA data not to
4427 be read from disk, causing single-file corruption.
4431 Frequency : rare, only 2.6 filesystems using extents
4432 Description: Heavy concurrent write and delete load may cause data corruption.
4433 Details : It was possible under high-load situations to have an extent
4434 metadata block in the block device cache from a just-unlinked
4435 file overwrite a newly-allocated data block. We now unmap any
4436 metadata buffers that alias just-allocated data blocks.
4440 Frequency : filesystems with default stripe_count larger than 77
4441 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4442 Details : lconf specifies an inode size of 4096 bytes when the default
4443 stripe_count is larger than 77. This conflicts with the default
4444 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4445 this case to avoid pinning too much memory for large EAs.
4447 ------------------------------------------------------------------------------
4449 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4454 Frequency : rare (extremely heavy IO load with hundreds of clients)
4456 Description: Client is evicted, gets IO error writing to file
4457 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4458 caused clients to be evicted for AST timeouts. The fixes in
4459 bug 5192 mean we no longer need to have such short AST timeouts
4460 so ldlm_timeout has been increased.
4463 Frequency : occasional during --force or --failover shutdown under load
4464 Bugzilla : 5949, 4834
4465 Description: Server oops/LBUG if stopped with --force or --failover under load
4466 Details : a collection of import/export refcount and cleanup ordering
4467 issues fixed for safer force cleanup
4470 Frequency : only filesystems larger than 120 OSTs
4471 Bugzilla : 5990, 6223
4472 Description: lfs getstripe would oops on a very large filesystem
4473 Details : lov_getconfig used kfree on vmalloc'd memory
4476 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4477 Bugzilla : 6242, 6243
4478 Description: reading from files that had been truncated to a non-zero size
4479 but never opened returned no data
4480 Details : ll_file_read() reads zeros from no-object files to EOF
4485 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4486 Details : There's little harm in aborting MDS/OSS recovery and letting it
4487 try again, so I removed the LASSERT and return an error instead.
4489 Severity : enhancement
4491 Description: New debugging infrastructure for tracking down data corruption
4492 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4493 (b) cover more of the client-side code path, and (c) try to narrow
4494 down where problems occurred
4498 Bugzilla : 3819, 4364, 4397, 6313
4499 Description: Racing close and eviction MDS could cause assertion in mds_close
4500 Details : It was possible to get multiple mfd references during close and
4501 client eviction, leading to one thread referencing a freed mfd.
4503 Severity: : enhancement
4504 Bugzilla : 3262, 6359
4505 Description: Attempts to reconnect to servers are now more aggressive.
4506 Details : This builds on the enhanced upcall-less recovery that was added
4507 in 1.4.2. When trying to reconnect to servers, clients will
4508 now try each server in the failover group every 10 seconds. By
4509 default, clients would previously try one server every 25 seconds.
4514 Description: After recovery, certain operations trigger a failed
4515 assertion on a client.
4516 Details : Failing over an mds, using lconf -d --failover, while a
4517 client was doing a readdir() call would cause the client to
4518 LBUG after recovery completed and the readdir() was resent.
4520 Severity : enhancement
4522 Description: Default groups are now added by lconf
4523 Details : You can now run lconf --group <servicename> without having to
4524 manually add groups with lmc.
4527 Frequency : occasional
4529 Description: Nodes with an elan id of 0 trigger a failed assertion
4532 Frequency : always when accessing e.g. tty/console device nodes
4534 Description: tty and some other devices nodes cannot be used on lustre
4535 Details : file's private_data field is used by device data and lustre
4536 values in there got lost. New field was added to struct file to
4537 store fs-specific private data.
4540 Frequency : when exporting Lustre via NFS
4542 Description: NFSD failed occasionally when looking up a path component
4543 Details : NFSD is looking up ".." which was broken in ext3 directories
4544 that had grown large enough to become hashed.
4547 Frequency : Clusters with multiple interfaces not on the same subnet
4549 Description: Nodes will repeatedly try to reconnect to an interface which it
4550 cannot reach and report an error to the log.
4551 Details : Extra peer list entries will be created by lconf with some peers
4552 unreachable. lconf now validates the peer before adding it.
4555 Frequency : Only if a default stripe is set on the filesystem root.
4557 Description: Setting a default stripe on the filesystem root prevented the
4558 filesystem from being remounted.
4559 Details : The client was sending extra request flags in the root getattr
4560 request and did not allocate a reply buffer for the dir EA.
4563 Frequency : occasional, higher if lots of files are accessed by one client
4564 Bugzilla : 6159, 6097
4565 Description: Client trips assertion regarding lsm mismatch/magic
4566 Details : While revalidating inodes the VFS looks up inodes with ifind()
4567 and in rare cases can find an inode that is being freed.
4568 The ll_test_inode() code will free the lsm during ifind()
4569 when it finds an existing inode and then the VFS later attaches
4570 this free lsm to a new inode.
4574 Bugzilla : 6422, 7030
4575 Description: MDS deadlock between mkdir and client eviction
4576 Details : Creating a new file via mkdir or mknod (starting a transaction
4577 and getting the ns lock) can deadlock with client eviction
4578 (gets ns lock and trying to finish a synchronous transaction).
4581 Frequency : occasional
4582 Description: While starting a server, the fsfilt_ext3 module could not be
4584 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4585 kernels. Previously, lconf would still use the ext3 name
4586 when trying to load modules. Now, it will correctly use
4587 ext3 on 2.4 and ldiskfs on 2.6.
4589 Severity : enhancement
4590 Description: The default stripe count has been changed to 1
4591 Details : The interpretation of the default stripe count (0, to lfs
4592 or lmc) has been changed to mean striping across a single
4593 OST, rather than all available. For general usage we have
4594 found a stripe count of 1 or 2 works best.
4596 Severity : enhancement
4597 Description: Add support for compiling against Cray portals.
4598 Details : Conditional compiling for some areas that are different
4602 Frequency : occasional
4603 Bugzilla : 6409, 6834
4604 Description: Creating files with an explicit stripe count may lead to
4605 a failed assertion on the MDS
4606 Details : If some OSTs are full or unavailable, creating files may
4607 trigger a failed assertion on the MDS. Now, Lustre will
4608 try to use other servers or return an error to the
4612 Frequency : occasional
4614 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4615 caused lock timeout during readahead (since 1.4.2).
4616 Details : Processes doing readahead might match a lock that hasn't been
4617 granted yet if there are overlapping and conflicting lock
4618 requests. The readahead process waits on ungranted lock
4619 (original lock is CBPENDING), while OST waits for that process
4620 to cancel CBPENDING read lock and eventually evicts client.
4622 Severity : enhancement
4624 Description: Initial enabling of flock support for clients
4625 Details : Implements fcntl advisory locking and file status functions.
4626 This feature is provided as an optional mount flag (default
4627 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4628 locking are implemented yet, and those that are are not guaranteed
4629 to be completely correct in production environments.
4630 mount -t lustre -o [flock|noflock] ...
4633 Frequency : occasional
4635 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4636 assertion in the ext3 JBD (journaling) layer.
4637 Details : The b_committed_data struct is protected by the big kernel lock
4638 in 2.4 kernels, serializing journal_commit_transaction() and
4639 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4640 access to this struct. In 2.6 kernels there is finer grained
4641 locking to improve SMP performance of the JBD layer.
4645 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4650 Description: merge and cleanup kernel patches.
4653 -----------------------------------------------------------------------------
4655 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4657 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4658 versions of Lustre. Please update all servers and clients to
4659 version 1.4.2 or later at the same time. You must also run
4660 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4661 to update the configuration logs.
4663 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4664 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4665 - swab llog records of type '0' so we get proper header size/idx (5861)
4666 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4667 - fix rename of one directory over another leaking an inode (5953)
4668 - avoid SetPageDirty on 2.6 (5981)
4669 - don't re-add just-being-destroyed locks to the waiting list (5653)
4670 - when creating new directories, inherit the parent's custom
4671 striping settings if present parent (3048)
4672 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4673 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4674 - don't hold client locks on temporary worklist from l_lru (5666)
4675 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4676 - automatically evict dead clients (5921)
4677 - Update file size properly in create+truncate+fstat case (6196)
4678 - Do not unhash mountpoint dentries, do not allow removal of
4680 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4681 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4682 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4683 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4684 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4685 - Correctly handle reads of files with no objects (6243)
4686 - lctl recover will also mark a device active if deactivate used (5933)
4688 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4689 - allow --write-conf on an MDS with different nettype than client (5619)
4690 - don't write config llogs to MDS for mounts not from that MDS (5617)
4691 - lconf should create multiple TCP connections from a client (5201)
4692 - init scripts are now turned off by default; run chkconfig --on
4693 lustre and chkconfig --on lustrefs to use them
4694 - upcalls are no longer needed for clients to recover to failover
4696 - add --abort-recovery option to lconf to abort recovery on device
4698 - add support for an arbitrary number of OSTs (3026)
4699 - Quota support protocol changes.
4700 - forward compatibility changes to wire structs (6007)
4701 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4702 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4703 - improved "lustre status" script
4704 - initialize blocksize for non-regular files (6062)
4705 - added --disable-server and --disable-client configure options (5782)
4706 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4707 - Vanilla 2.4.29 support
4708 - increase maximum number of obd devices to 520 (6242)
4709 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4710 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4711 SLES 9 2.6.5 kernels
4712 - limit stripes per file to 160 (the maximum EA size) (6093)
4714 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4717 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4718 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4719 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4720 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4721 - small ext3 extents cleanups and fixes (5733)
4722 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4723 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4724 - avoid needless client->OST connect, fix handle mismatch (5317)
4725 - fix DLM error path that led to out-of-sync client, long delays (5779)
4726 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4727 - fix several locking issues related to i_size (5492,5624,5654,5672)
4728 - don't move pending lock onto export if it is already evicted (5683)
4729 - fix kernel oops when creating .foo in unlinked directory (5548)
4730 - fix deadlock in obdfilter statistics vs. object create (5811)
4731 - use time_{before,after} to avoid timer jiffies wrap (5882)
4732 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4733 - Do not leak request if server was not able to process it (5154)
4734 - If mds_open unable to find parent dir, make that negative lookup(5154)
4735 - don't create new directories with extent-mapping (5909, 5936)
4737 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4738 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4739 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4740 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4741 - support for RHEL4 kernel on i686 (5773)
4742 - provide error messages when incompatible logs are encountered (5898)
4744 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4745 * version 1.4.0.10 (1.4.1 release candidate 1)
4747 - don't keep a lock reference when lock is not granted (4238)
4748 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4749 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4750 - drop import inflight refcount on signal_completed_replay error (5255)
4751 - unlock page after async write error during send (3677)
4752 - handle missing objects in filter_preprw_read properly (5265)
4753 - no transno return for symlink open, don't save no-trasno open (3440)
4754 - don't try to complete elan receive that already failed (4012)
4755 - free RPC server reply state on error (5406)
4756 - clean up thread from ptlrpc_start_thread() on error (5160)
4757 - readahead could read extra page into cache that wasn't ejected (5388)
4758 - prevent races in class_attach/setup/cleanup/detach (5260)
4759 - don't dereference de->d_inode after l_dput of de (5458)
4760 - use "int" for stripe value returned from lock_to_stripe (5544)
4761 - mballoc allocation and error-checking fixes in 2.6 (5504)
4762 - block device patches to fix I/O request sizes in 2.6 (5482)
4763 - look up hostnames for IB nals (5602)
4764 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4765 - don't start multiple acceptors for the same port (5277)
4766 - fix incorrect LASSERT in mds_getattr_name (5635)
4767 - export a proc file for general "ping" checking (5628)
4768 - fix "lfs check" to not block when the MDS is down (5628)
4770 - service request history (4965)
4771 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4772 - create an "evict_client" /proc entry on OSTs, like the MDS has
4773 - fix mount usage message, return errors per mount(8) (5168)
4774 - change grep [] to grep "[]" in tests so they work in more UMLs
4775 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4776 - remove ancient LOV_MAGIC_V0 EA support (5047)
4777 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4778 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4779 - allow readahead allocations to fail when low on memory (5383)
4780 - mmap locking landed again, after considerable improvement (2828)
4781 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4783 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4786 - send OST transaction number in read/write reply to free req (4966)
4787 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4788 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4789 - the watchdog thread now runs as interruptible (5246)
4790 - flock/lockf fixes (but it's still disabled, pending 5135)
4791 - don't use EXT3 constants in llite code (5094)
4792 - memory shortage at startup could cause assertion (5176)
4794 - reorganization of lov code
4795 - single portals codebase
4797 - add extents/mballoc support (5025)
4798 - direct I/O reads in the obdfilter (4048)
4799 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4801 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4804 - send OST transaction number in read/write reply to free req (4966)
4805 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4806 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4807 - the watchdog thread now runs as interruptible (5246)
4808 - handle missing objects in filter_preprw_read properly (5265)
4809 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4810 - drop import inflight refcount on signal_completed_replay error (5255)
4811 - unlock page after async write error during send (3677)
4812 - return original error code on reconstructed replies (3761)
4813 - no transno return for symlink open, don't save no-trasno open (3440)
4815 - add pid to ldlm debugging output (4922)
4816 - bump the watchdog timeouts -- we can't handle 30sec yet
4817 - extra debugging for orphan dentry/inode bug (5259)
4819 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4822 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4823 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4824 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4825 - fix NULL dereference in /proc/sys/portals/routes (4827)
4826 - allow failed mdc_close() operations to be interrupted (4561)
4827 - stop precreate on OST before MDS would time out on it (4778)
4828 - don't send partial-page writes before EOF from client (4410)
4829 - discard client grant for sub-page writes on large-page clients (4520)
4830 - don't free dentries not owned by NFS code, check generation (4806)
4831 - fix lsm leak if mds_create_objects() fails (4801)
4832 - limit debug_daemon file size, always print CERROR messages (4789)
4833 - use transno after validating reply (3892)
4834 - process timed out requests if import state changes (3754)
4835 - update mtime on OST during writes, return in glimpse (4829)
4836 - add mkfsoptions to LDAP (4679)
4837 - use ->max_readahead method instead of zapping global ra (5039)
4838 - don't interrupt __l_wait_event() during strace
4840 - add software watchdogs to catch hung threads quickly (4941)
4841 - make lustrefs init script start after nfs is mounted
4842 - fix CWARN/ERROR duplication (4930)
4843 - return async write errors to application if possible (2248)
4844 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4845 - print NAL number in %x format (4645)
4846 - update barely-supported suse-2.4.21-171 series (4842)
4847 - support for sles 9 %post scripts
4848 - support for building 2.6 kernel-source packages
4849 - support for sles km_* packages
4851 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4854 - ignore -ENOENT errors in osc_destroy (3639)
4855 - notify osc create thread that OSC is being cleaned up (4600)
4856 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4857 - reconstruct ost_handle() like mds_handle() (4657)
4858 - create a new thread to do import eviction to avoid deadlock (3969)
4859 - let lconf resolve symlinked-to devices (4629)
4860 - don't unlink "objects" from directory with default EA (4554)
4861 - hold socknal file ref over connect in case target is down (4394)
4862 - allow more than 32000 subdirectories in a single directory (3244)
4863 - fix blocks count for O_DIRECT writes (3751)
4864 - OST returns ENOSPC from object create when no space left (4539)
4865 - don't send truncate RPC if file size isn't changing (4410)
4866 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4867 - bind to privileged port in socknal and tcpnal (3689)
4869 - rate limit CERROR/CWARN console message to avoid overload (4519)
4870 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4871 - basic mmap support (3918)
4872 - kernel patch series update from b1_4 (4711)
4874 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4877 - avoid crash during MDS cleanup with OST shut down (2775)
4878 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4879 - don't use bad inodes on the MDS (3744)
4880 - dynamic object preallocation to improve recovery speed (4236)
4881 - don't hold spinlock over lock dumping or change debug flags (4401)
4882 - don't zero obd_dev when it is force cleaned (3651)
4883 - print grants to console if they go negative (4431)
4884 - "lctl deactivate" will stop automatic recovery attempts (3406)
4885 - look for existing locks in ldlm_handle_enqueue() (3764)
4886 - don't resolve lock handle twice in recovery avoiding race (4401)
4887 - revalidate should check working dir is a directory (4134)
4889 - don't always mark "slow" obdfilter messages as errors (4418)
4891 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4894 - don't close LustreDB during write_conf until it is done (3860)
4895 - fix typo in lconf for_each_profile (3821)
4896 - allow dumping logs from multiple threads at one time (3820)
4897 - don't allow multiple threads in OSC recovery (3812)
4898 - fix debug_size parameters (3864)
4899 - fix mds_postrecov to initialize import for llog ctxt (3121)
4900 - replace config semaphore with spinlock (3306)
4901 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4902 - don't allow enqueue to complete on a destroyed export (3822)
4903 - down write_lock before checking llog header bitmap (3825)
4904 - recover from lock replay timeout (3764)
4905 - up llog sem before sending rpc (3652)
4906 - reduce ns lock hold times when setting kms (3267)
4907 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4908 - fix NULL deref and obd_dev leak on setup error (3312)
4909 - replace some LBUG about llog ops with error handling (3841)
4910 - don't match INVALID dentries from d_lookup and spin (3784)
4911 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4912 - fix invalid assertion in ptlrpc_set_wait (3880)
4914 - add libwrap support for the TCP acceptor (3996)
4915 - add /proc/sys/portals/routes for non-root route listing (3994)
4916 - allow setting MDS UUID in .xml (2580)
4917 - print the stack of a process that LBUGs (4228)
4919 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4922 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4923 - make sure to unset replay flag from failed open requests (3440)
4924 - if default stripe count is 0, use OST count for inode size (3636)
4925 - update parent mtime/ctime on client for create/unlink (2611)
4926 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4927 - free recovery state on server during a forced cleanup (3571)
4928 - unregister_reply for resent reqs (3063)
4929 - loop back devices mounting and status check on 2.6 (3563)
4930 - fix resource-creation race that can provoke i_size == 0 (3513)
4931 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4932 - more debugging for page-accounting assertion (3746)
4933 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4934 - avoid deadlock after precreation failure (3758)
4935 - fix race and lock order deadlock in orphan handling (3450, 3750)
4936 - add validity checks when grabbing inodes from l_ast_data (3599)
4938 - add /proc/.../recovery_status to obdfilter (3428)
4939 - lightweight CDEBUG infrastructure, debug daemon (3668)
4940 - change default OSC RPC parameters to be better on small clusters
4941 - turn off OST read cache for files smaller than 32MB
4942 - install man pages and include them in rpms (3100)
4943 - add new init script for (un)mounting lustre filesystems (2593)
4944 - run chkconfig in %post for init scripts (3701)
4945 - drop scimac NAL (unmaintained)
4947 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4950 - clean kiobufs before and after use (3485)
4951 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4952 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4953 - change init script to fail more clearly if not run as root (1528)
4954 - allow clients to reconnect during replay (1742)
4955 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4956 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4957 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4958 - fix cond_resched (3554)
4959 - only evict unfinished clients after recovery (3515)
4960 - allow bulk resend, prevent data loss (3570)
4961 - dynamic ptlrpc request buffer allocation (2102)
4962 - don't allow unlinking open directory if it isn't empty (2904)
4963 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4964 - remove extraneous obd dereference causing LASSERT failure (3334)
4965 - don't use get_cycles() when creating temp. files on the mds (3156)
4966 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4967 - handle EEXIST for set-stripe, set proper directory name (3336)
4969 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4970 - fix ksocknal_fmb_callback() error messages (2918)
4972 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4975 - don't copy lvb into (possibly NULL) reply on error (2983)
4976 - don't deref dentry after dput, don't free lvb on error (2922)
4977 - use the kms to determine writeback rpc length (2947)
4978 - increment oti_logcookies when osc is inactive (2948)
4979 - update client's i_blocks count via lvb messages (2543)
4980 - handle intent open/close of special files properly (1557)
4981 - mount MDS with errors=remount-ro, like obdfilter (2009)
4982 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4983 - don't use cancelling-locks' kms values (2947)
4984 - use highest lock extent for kms, not last one (2925)
4985 - don't dereference ERR_PTR() dentry in error handling path (3107)
4986 - fix thread race in portals_debug_dumplog() (3122)
4987 - create lprocfs device entries at setup instead of at attach (1519)
4988 - common AST error handler, don't evict client on completion race (3145)
4989 - zero nameidata in detach_mnt in 2.6 (3118)
4990 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4991 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4992 - initialize RPC timeout timer earlier for 2.6 (3219)
4993 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4994 - print nal/nid for unknown nid (3258)
4995 - additional checks for oscc recovery before doing precreate (3284)
4996 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4997 - don't crash in mdc_close for bad permissions on open (3285)
4998 - zero i_rdev for non-device files (3147)
4999 - clear page->private before handing to FS, better assertion (3119)
5000 - tune the read pipeline (3236)
5001 - fix incorrect decref of invalidated dentry (2350)
5002 - provide read-ahead stats and refine rpc in flight stats (3328)
5003 - don't hold journal transaction open across create RPC (3313)
5004 - update atime on MDS at close time (3265)
5005 - close LDAP connection when recovering to avoid server load (3315)
5006 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5007 - don't leak open file on MDS after open resend (3325)
5008 - serialize filter_precreate and filter_destroy_precreated (3329)
5009 - loop device shouldn't call sync_dev() for nul device (3092)
5010 - clear page cache after eviction (2766)
5011 - resynchronize MDS->OST in background (2824)
5012 - refuse to mount the same filesystem twice on same mountpoint (3394)
5013 - allow llmount to create routes for mounting behind routers (3320)
5014 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5015 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5016 - fix rare problem with rename on htree directories (3417)
5018 - allow default OST striping configuration per directory (1414)
5019 - fix compilation for qswnal for 2.6 kernels (3125)
5020 - increase maximum number of MDS request buffers for large systems
5021 - change liblustreapi to be useful for external progs like lfsck (3098)
5022 - increase local configuration timeout for slow disks (3353)
5023 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5025 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5028 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5029 - don't overwrite extent policy data in reply if lock was blocked (2901)
5030 - drop filter export grants atomically with removal from device (2663)
5031 - del obd_self_export from work_list in class_disconnect_exports (2908)
5032 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5033 - swab reply message in mdc_close, other PPC fixes (2464)
5034 - fix destroying of named logs (2325)
5035 - overwrite old logs when running lconf --write_conf (2264)
5036 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5037 - fix race in target_handle_connect (2898)
5038 - mds_reint_create() should take same inode create lock (2926)
5039 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5040 - don't close files for self_export to avoid uninitialized obd (2936)
5041 - allow MDS with the same name as client node (2939)
5042 - hold dentry reference for closed log files for unlink (2325)
5043 - reserve space for all logs during transactions (2059)
5044 - don't evict page beyond end of stripe extent (2925)
5045 - don't oops on a deleted current working directory (2399)
5046 - handle hard links to targets without a parent properly (2517)
5047 - don't dereference NULL lock when racing during eviction (2867)
5048 - don't grow lock extents when lots of conflicting locks (2919)
5050 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5053 - account for cache space usage on clients to avoid data loss (974)
5054 - lfsck support in lustre kernel code (2349)
5055 - reduce journal credits needed for BRW writes (2370)
5056 - orphan handling to avoid losing space on client/server crashes
5057 - ptlrpcd can be blocked, stopping ALL progress (2477)
5058 - use lock value blocks to assist in proper KMS, faster stat (1021)
5059 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5060 - recovery for initial connections (2355)
5061 - fixes for mds_cleanup_orphans (1934)
5062 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5063 - block all file creations until orphan recovery completes (1901)
5064 - client remove rq_connection from request struct (2423)
5065 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5066 - recovery timer race (2670)
5067 - mdc_close recovey bug (2532)
5068 - ptlrpc cleanup bug (2710)
5069 - mds timeout on local locks (2588)
5070 - namespace lock held during RPCs (2431)
5071 - handle interrupted sync write properly (2503)
5072 - don't try to handle a message that hasn't been replied to (2699)
5073 - client assert failure during cleanup after abort recovery (2701)
5074 - leak mdc device after failed mount (2712)
5075 - ptlrpc_check_set allows timedout requests to complete (2714)
5076 - wait for inflight reqs when ptlrpcd finishes (2710)
5077 - make sure unregistered services are removed from the srv_list
5078 - reset bulk XID's when resending them (caught by 1138 test)
5079 - unregister_bulk after timeout
5080 - fix lconf error (2694)
5081 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5082 - readahead locks pages, leaves pending causing memory pressure (2673)
5083 - increase OST request buffers to 4096 on large machines (2729)
5084 - fix up permission of existing directories in simple_mkdir (2661)
5085 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5086 - don't assign transno to errored transactions (2742)
5087 - don't delete objects on OST if given a bogus objid from MDS (2751)
5088 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5089 - if rq_no_resend, then timeout request after recovery (2432)
5090 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5091 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5092 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5093 - MDS recovery shouldn't delete objects that it has given out (2730)
5094 - if enqueue arrives after completion, don't clobber LVB (2819)
5095 - don't unlock pages twice when trigger_group_io returns error (2814)
5096 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5097 - don't write pages to disk if there was an error (1450)
5098 - don't ping imports that have recovery disabled (2676)
5099 - take buffered bytes into account when balancing socknal conn (2817)
5100 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5101 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5102 - remove little-endian swabbing of llog records (1987)
5103 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5104 - retry reposting request buffers if they fail (1191)
5105 - grow extent at grant time to avoid granting a revoked lock (2809)
5106 - lock revoke doesn't evict page if covered by a second lock (2765)
5107 - disable VM readahead to avoid reading outside lock extents (2805)
5109 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5110 - updated kernel patches for hp-2.4.20 kernel (2681)
5112 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5115 - fix truncated write corruption (2366)
5116 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5118 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5119 - ensure len(uuid) < 37 in lmc (1171)
5120 - fix ia64 OOPS in llog_test (2255)
5121 - zero end of page at obdfilter for partial page writes (2648)
5122 - don't leave stale dentries around after renames (bug 2428)
5123 - fix timeouts when evicting a client with a single lock held (2642)
5124 - set deadline for the initial HELLO message to drain (2634)
5125 - print out dotted-quad IP addresses in the socknal (2302)
5127 - additional debugging for MDS client eviction problem (2443)
5128 - fix mkfsoptions support for osts (2603, 2604)
5130 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5133 - add series for the vanilla 2.6.0 kernel
5134 - add series for the vanilla 2.4.24 kernel
5135 - add series for a cray x86/64 UL kernel drop
5136 - fix xattr patches for the vanilla 2.4.19 series
5138 - generate true UUIDs in lmc (1171)
5139 - have portals stack dumping break in UML (2466)
5140 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5141 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5142 - wake callback waiting threads on client eviction (2460)
5143 - Add --ptldebug and --subsystem to lmc (1719)
5144 - update assertion to allow safe interrupt allocation
5145 - set rq_no_resend for cancel requests (2432)
5146 - recalculate ptlrpcd timeout after resend (2494)
5147 - call vfs_rmdir when removing pending directories (2368)
5148 - fix renaming a file to itself (2429)
5149 - lmc creates a default one-stripe lov (2454)
5150 - expand procfs space to handle large clusters (2326)
5151 - increase UML stack to avoid overflow
5152 - update lconf's list of debug and subsystem masks
5153 - fix lfs find --obd (2510)
5154 - /proc tunable for disabling filter read caching (2591)
5155 - stop rpm packages from altering slapd.conf (2301)
5156 - disable nagle in the socknal under 0conf (2578)
5157 - choose mds inode size based on stripe count (2572)
5158 - fix kernel-source rpm problems (2516)
5160 - add --disable-doc to avoid pdf generation (2421)
5161 - update documentation, tests, type-os, comments
5162 - avoid format warnings on ia64
5163 - remove the TOE NAL
5164 - tiny code cleanups by removing unused fields
5166 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5169 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5170 - remove the most blatant lies from BUILDING file (2371)
5171 - change default debug level to reasonable production setting
5172 - reduce client side cache size to reduce cache flush time
5173 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5174 - make TCP zerocopy and pinger support enabled by default (2476)
5175 - sync writes completed after process exits caused crashes (2319)
5176 - maintain correct mount count on the MDS (2356)
5177 - backout 1557, because 2316 wasn't really fixed
5178 - better file I/O statistics gathering in /proc
5179 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5180 - another kernel patch to fix zero-copy TCP function export
5181 - don't take duplicate lock when processing re-sent getattr (2420)
5182 - lctl uses obd_self_export instead of creating new conn (2353)
5183 - MDS/OST recovery case which requires object creation asserted (2425)
5184 - move lfs from /usr/sbin to /usr/bin in packages
5185 - fix race between mds_client_add and mds_client_free (2417)
5186 - use kmalloc instead of slabs in portals (2430)
5187 - don't create duplicate records when a failover MDS is present (2442)
5188 - remove unnecessary mount age check (2332)
5189 - don't remove directory inodes from locks prematurely (2451)
5190 - don't break if MDS service name is the same as hostname (2103)
5191 - fix races in client write RPC generation when cache full (2482)
5193 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5196 - remove now-unused request->rq_obd (278)
5197 - if an allocation fails, print out how much memory we've used (1933)
5198 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5199 - add the "configurable stack size" patch to most series files (1256)
5200 - ability to write large log records, for 100+ OST configs (2306)
5201 - fix NULL deref when filter_prep fails (2314)
5202 - fix operator precedence error in filter_sync
5203 - dynamic allocation of socknal TX descriptors (2315)
5204 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5205 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5206 - ensure that CWARN messages in Portals always get to the syslog
5207 - __init/__exit are not for prototype decls (ldlm_init/exit)
5208 - x86-64 compile warning fixes
5209 - fix gateway LMC keyword conflict (2318)
5210 - fix MDS lock inversions in getattr/reint paths (1844)
5211 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5212 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5213 - DLM race condition prevented some lock evictions (2328)
5214 - ENOMEM detection and retry on socknal sends (2230)
5215 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5216 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5217 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5218 - better error messages when a client is rejected during recovery (1505)
5219 - avoid cancelling locks which were never granted, after failure (2330)
5220 - fix i_sem/journal inversion in mds_client_add (2333)
5221 - fix truncate/getattr lock cycle deadlock (2334)
5222 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5223 - fix two rare exit paths which could leak an l_lock() ref (2321)
5224 - fencepost error in MDS/OST orphan recovery (2226)
5225 - make log record alignment 8 bytes (1988)
5226 - lstripe now fails when requested offset > ost_count (2237)
5227 - ensure that all kernel series have a complete list.h (1607)
5228 - fix crashes in special-file operations (2316)
5229 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5231 - allow configurable automake binary, for testing new versions
5232 - small update to the lfs documentation
5234 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5236 * fix negative export reference count in fsfilt_sync (2312)
5238 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5239 * release candidate 0.9.1
5241 - orphans are moved into the PENDING directory for possible recovery
5242 - replayed opens now open by fid for orphan/rename safety (1042)
5243 - last close of an orphan inode generates a transno (683)
5244 - chdir() and mount() now pin the directory entry (1020)
5245 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5246 - discard very old requests without processing them (1502)
5247 - remove some common, well-understood CERRORs (1505)
5248 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5249 - clear "grant" flags in OST replies until OST grant code lands (1644)
5250 - fix read performance by not clobbering i_blksize on client (1598)
5251 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5252 - make LRU size a /proc tunable, clears locks when reduced (707)
5253 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5254 - server locks take references on exports now (1558)
5255 - build fixes for 2.4.20-rh trees (1663)
5256 - return an error from lov_create if all OSCs are inactive (1751)
5257 - fix import levels when a reconnect happens without a timeout (1597)
5258 - exit early from mds_open if we get a lookup error (1749)
5259 - partial page read at EOF wouldn't wait for disk before sending (1642)
5260 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5261 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5262 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5263 - write/truncate lock inversion (1639)
5264 - Don't auto-load obdclass, portals modules during cleanup (1495)
5265 - fix timestamps from jumping to "now" (1763)
5266 - extra journal assertions (1648)
5267 - add an extra multiunlink test (1771)
5268 - fix read_record/write_record API (1776)
5269 - fix leak of offset_extent, possible incorrect i_size later (1772)
5270 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5271 - Debugging for the kqswnal_get_idle_tx problems (1820)
5272 - Allow recovery to be attempted multiple times (1536)
5273 - Write out MDS last_rcvd file after it is first created (1600)
5274 - Fix tx_descriptor leak in failed transmit situations (1827)
5275 - ext3 journaling fixes for assertion failure after IO error (1871)
5276 - class_export_put() on freed export after completion AST error (1896)
5277 - Fix revalidate looping in VFS (1322)
5278 - Don't access a freed export during MDS_REINT timeout (1521)
5279 - Add open-unlink recovery support on the MDS (1673,1764)
5280 - Return an error if no MDS data was read from last_rcvd (1946)
5281 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5282 - Don't setup a disk device that doesn't match exported UUID (317)
5283 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5284 - avoid committing NULL handle in force close
5285 - local.sh is now a one-stripe LOV configuration
5286 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5287 - fix ext3 htree duplicate directory entry corruption (1516)
5288 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5289 - update to new LOV EA format (2097)
5290 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5292 2003-06-15 Phil Schwan <phil@clusterfs.com>
5295 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5296 - per-import/export recovery handling (958, 931, 959)
5297 - multiple last-rcvd slots, for serving multiple FSes (949)
5298 - connections are again shared between multiple imp/exports (963, 964)
5299 - "umount -f" would hang if any requests needed to be sent (393, 978)
5300 - avoid pinning large req buffer by copying for queued messages (989)
5301 - add "uuid" to "lctl device" command to help upcalls (991)
5302 - "open" RPCs with transnos would confuse recovery counters (1037)
5303 - do proper endian conversion of all wire messages (288, 340, 891)
5304 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5305 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5306 - fix ptlrpc_connection leak in target_handle_connect (1174)
5307 - fix import refcounting bug in OST and MDS cleanup (1134)
5308 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5309 - fix very unlikely obd_types race condition (501)
5310 - remove osc_open hack for echo_client (1187)
5311 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5312 - a failure in read_inode2 leads to deadlock (1139)
5313 - cancel ack-locks as soon as transaction is committed (1072)
5314 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5315 - make sure to commitrw after any preprw to avoid deadlock (1162)
5316 - failing to execute a file in a lustre FS would lock inode (1203)
5317 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5318 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5319 - obd_uuid2tgt would walk off the end of the list (1255)
5320 - on IA64 the osc would give portals incorrect bulk size (1258)
5321 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5322 - fix lock inversion caused by new llite matching code (1282)
5323 - limit the number of dirty pages on a client to 10MB (1286)
5324 - timed out locks were not being corrected cancelled (1289)
5325 - fix O_DIRECT above 4GB on IA-32 (1292)
5326 * major user-visible changes
5327 - fail out/fail over policy now controlled by the upcall (993)
5329 - add OBD_PING to check server availability and failure (954)
5330 - lustre messages are now sent in sending host order (288, 340, 891)
5331 - add eadatalen to MDS getattr reply (340)
5332 - OST read replies may contain second buffer, with per-page status (593)
5334 2003-03-11 Phil Schwan <phil@clusterfs.com>
5337 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5338 - failed open()s could cause deadlock; fixed (867, 869)
5339 - stop cancelling OST locks when files are closed (481)
5340 - overlapping XID spaces caused network corruption (851, 853)
5341 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5342 - setattr_raw added, to do single-RPC, server-side setattrs
5343 - lmc/lconf syntax change for OST UUIDs
5344 - fix crashy race condition between ptlrpc_free_req and osc_close
5345 - don't use request in mdc_enqueue if we hit a timeout (889)
5346 - don't set the inode i_size for regular files from the MDS (896)
5347 - handle out of order completion AST (842)
5348 - don't LBUG if a lock request times out after receiving AST (913)
5349 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5350 - if a bad lock AST arrives, send an error instead of dropping entirely
5351 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5352 - fix leak in bulk IO when only partially completed (899, 900, 926)
5353 - fix O_DIRECT for ia64 (55)
5354 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5355 - C-z after timeout could hang a process forever; fixed (977)
5357 - client-side I/O cache (678, 924, 929, 941, 970)
5359 - READPAGE and SETATTRs which don't take server-side locks get
5362 2003-02-11 Phil Schwan <phil@clusterfs.com>
5365 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5366 - Fix fsfilt_extN_readpage() to read a full page of directory
5367 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5368 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5369 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5370 - Most intent operations were removed, in favour of a new RPC mode
5371 that does a single RPC to the server and bypasses most of the VFS
5372 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5373 - Aggressively cancel local locks on DLM servers
5374 - mds_reint_unlink sends EA to the client if it's the last nlink.
5375 client uses that EA to unlink OST objects.
5376 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5377 - recursive symlinks were fixed (439)
5378 - fixed NULL deref in DEBUG_REQ
5379 - filter_update_lastobjid no longer calls sync, which annoyed extN
5380 - fixed multi-client small-writes to a single file problem (445)
5381 - fixed mtime updates during file writes (607)
5382 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5383 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5384 - fixed Linux OST size reporting problem (444, 656)
5385 - OST now updates object mtime with writes or setattr (607, 619)
5386 - client verifies file size before zeroing page past EOF (445)
5387 - OST now writes last allocated objid to disk with allocation (108)
5388 - LOV on echo now works (409)
5390 - mds_reint_unlink sends a new buffer, with the EA included. this
5391 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5392 set if a regular file was being unlinked, and it was the last link
5393 - use PtlGet from the target for bulk writes (315)
5394 - OST now updates object mtime with writes or setattr (607, 619)
5395 - LDLM now has a grant-time callback to revalidate locked items, if
5397 - Many MDS operations were reorganized to combat race conditions
5399 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5400 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5402 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5405 - Fully reactivate OST imports after reconnection (512, others)
5406 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5407 - More graceful error handling for truncating on dead OST (515)
5408 - Don't error out unless we're actually accessing dead stripes (474)
5409 - Fix garbage sizes when stripes are missing (410)
5410 - LRU counters were broken, causing constant lock purge (433, 432)
5411 - garbage on read from stripes with failed OSTs (441)
5412 - mark OSCs as active before reconnecting during recovery (438)
5413 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5414 - lfind did not preserve OST order in output (443)
5415 - symlinks cause hung clients, incorrect data (439)
5416 - stop dereferencing request after dropping refcount (457)
5417 - don't LASSERT(spin_is_locked) on non-SMP (455)
5418 - fixes for many rename() bugs
5419 - fstat didn't correctly synchronize attributes (399)
5420 - server must handle lock cancellation during blocking AST prep (487)
5421 - bulk descriptors were free()d too soon (511)
5422 - fix paths in lconf, which would load incorrect modules (451, 507)
5423 - fix confusing lconf 'host not found' error message (386)
5424 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5425 - fix race condition in mdc_blocking_ast() for inode access (526)
5426 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5427 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5428 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5429 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5430 - fix lstripe in several ways (kernel side) (527)
5431 - fix request leak in ldlm_cli_enqueue (262)
5432 - incorrect OSC was marked inactive after OST failure
5433 - call mds_fs_cleanup before unmounting filesystem (524)
5434 - fix races between taking ns_lock and ldlm_lock_change_resource
5435 - fix races updating LOV export open file list
5436 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5437 - fix recovery NULL deref in ldlm_cli_cancel_unused
5438 - fix some DLM races by using new hash table for lock handles (419)
5439 - permit the client to specify desired inodes, at replay
5440 - duplicate requests when we queue them for replay reintegration
5441 - fix last_rcvd offset calculation
5442 - sync after each recovered transaction, so we always make progress
5443 - never, not always, ERESTART requests without transnos
5444 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5446 - skip replay if the MDS says that the client is already connected
5447 - don't check for a recovery-enabled export to match lctl's UUID
5448 - don't INC_USE_COUNT for phantom exports
5449 - don't crash when cleaning up phantom exports (567)
5450 - don't double-finish or set replay data for errored mdc_open requests
5451 - abort requests when they time out, so we don't get old replies
5452 - send/receive replies for AST messages again
5453 - if the client says that it doesn't have the lock, cancel it on the
5455 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5456 mark it as destroyed, it will all work out when decref is called
5457 - fix module use counts (22, 581)
5459 - ASTs now expect a reply (server cancels lock on error reply)
5461 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5464 - fix many simultaneous client startup (392)
5465 - fix dentry->d_it clobbering
5466 - credentials weren't being shipped for readdir/getattr operations
5467 - remove invalid assertions triggered during some concurrent MD
5469 - proper Lustre versions added (336, 389)
5470 - fix memory leak for create error case (398)
5471 - fix LOV locking bug that would get cli/srv out of sync
5472 - fix echo client over LOV (409)
5473 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5474 - fix double-O_EXCL intent crash (424)
5475 - avoid sending multiple lock CANCELs (352)
5477 - MDS can do multi-client recovery (modulo bugs in new code)
5479 - many updates, edits, cleanups
5481 2002-11-18 Phil Schwan <phil@clusterfs.com>
5484 - fix null d_it dereference (346)
5485 - fix full OST/dbench hang (333)
5486 - fix permission problem with file removal (286)
5487 - fix removal of OSCs from LOV when they fail
5488 - fix NULL deref during bulk timeout (214)
5489 - fix problems related to multiple filesystems on one MDS (241)
5490 - fixed serious subtle metadata locking bugs
5491 - free locks on clients when inodes are removed due to memory
5493 - fix inode pointer in lock data (285)
5494 - partial support for multiple MDS on a single host (241)
5495 - data locks weren't cancelled at clear_inode time (290, 311)
5496 - intent locks could lead to unbounded lock growth (205)
5497 - added a maximum lock count, an LRU list, and a flusher
5498 - fix multiple rename (365)
5499 - properly abstracted the echo client
5500 - OSC locked 1 byte too many; fixed
5501 - rewrote brw callback code:
5502 - fixed recovery bugs related to LOVs (306)
5503 - fixed too-many-pages-in-one-write crash (191)
5504 - fixed (again) crash in sync_io_timeout (214)
5505 - probably fixed callback-related race (385)
5507 - Add capability to MDS protocol
5508 - LDLM cancellations and callbacks on different portals
5510 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5513 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5514 - defer open object destruction to close time (601981, 138)
5515 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5516 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5517 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5518 - don't call obd_setattr in ll_file_release for destroyed objects
5519 * protocol change to lustre_msg: move |version| and add |flags|
5520 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5521 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5522 * added replay of create, unlink, link and rename operations during
5523 MDS failover; recovery should be much more robust now
5524 * remove failed OSCs from LOVs (only lov_create uses this so far)
5525 * the lustre-HOWTO was brought (more) up to date (582544)
5527 2002-10-23 Phil Schwan <phil@clusterfs.com>
5530 - in-use dentries weren't being reused properly (617851)
5531 - prevent multiple LDLM setup (599178)
5532 - fix LOV size calculations for truncate (617853)
5533 - fix client handling of MDS intent errors (POSIX)
5534 - fix permission bug in lovstripe.c test (624321)
5535 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5536 - truncate past end of file could corrupt data
5537 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5538 - a race in recovery could return ETIMEDOUT to apps (623947)
5539 - building outside the source directory was fixed
5540 * the lustre-HOWTO was brought (more) up to date (582544)
5541 * major progress was made on recovery functionality
5543 2002-10-10 Phil Schwan <phil@clusterfs.com>
5546 - recovery deadlock fix
5547 - rm -rf causes LBUG fix (617817)
5548 - file open by multiple tasks fix (618962)
5549 - directory permissions bugs (602707 and 620007)
5550 - journal_stop fixed with locking (611313)
5551 - O_APPEND failures resolved (618273, perhaps 614459)
5552 - lconf PATH fix (619770)
5553 - IA64 build fix (621450)
5554 - RPC buffer sizes scale with amount of memory
5556 2002-10-01 Phil Schwan <phil@clusterfs.com>
5559 - locks would be cancelled without throwing away data pages,
5560 resulting in inconsistent data (605627)
5561 - inode attributes were not always being refreshed (605627, 612449)
5562 - lconf now continues to cleanup after lctl reports an error
5563 - MDS now enforces user permissions (602707)
5564 - lprocfs cleanup fixed, but not yet enabled (614157)
5565 - fixed infinite server hang, should a client not respond to an AST
5566 - avoid going into recovery if user calls readlink() with a buffer
5567 that's too small (613941)
5568 - AST RPCs no longer require replies (614867) -- this may be changed
5569 - don't crash server if client sends an IOV that's too big (611336)
5570 - fixed lock conversion deadlock (611892)
5571 - fixed the following of symlinks (614622)
5572 * recovery: the server can remove locks from a client that dies, other
5573 clients can make progress
5574 * more extN patch fixes
5575 * compile-time configurable ptlrpc buffer allocations
5577 - collaborative read cache document
5578 - Lustre Lite Performance CDR document-in-progress
5580 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5583 - fix typo in patch-2.4.18
5585 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5588 - clear ptlrpc request each time in handle_incoming_request()
5589 - unlink of files now destroys the object on the OST
5591 2002-09-19 Peter Braam <braam@clusterfs.com>
5593 * add hard link support
5594 * change obdfile creation method
5595 * kernel patch changed
5597 2002-09-19 Peter Braam <braam@clusterfs.com>
5600 - stack overflow bug in extN fixed
5602 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5604 * documentation updates
5605 - add man pages for config tools
5606 - update tests/README to describe testing with new config tools
5607 - finish metadata API descriptions
5608 * bug fixes and cleanups
5609 - statfs workaround for 16TB limit
5610 - LOV stripe allocation improved, can stripe on subset of OSTs
5611 - LOV file size/IO offset was wrong for files > 4GB in size
5612 - object EA data was being dropped, caused files to be unreadable
5613 - memory overflow with non-LOV OST caused memory corruption
5614 - fixed regression tests to work with new config tools, obdfilter
5615 - fixed bug when directory size became larger than 1 block
5616 - fixed bug (for single client case) when PWD was deleted
5617 - invalidate local directory pages when doing intent-based ops
5618 - avoid LDLM oops when lock callback contained bad data
5620 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5622 * documentation updates
5623 * bug fixes and cleanups
5624 - configuration tools
5627 - 64-bit compile warnings
5628 - 64-bit internal statfs data
5630 * test_brw on persistent OST devices
5632 * lprocfs (disabled)
5634 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5636 * documentation updates
5637 * bug fixes and cleanups
5638 * configuration tools
5640 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5643 * many small fixes to 0_5_4
5644 * io/network handling
5645 * thinkos in MDS operations
5647 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5650 * crucial basic fixes to 0.5.3
5651 * IOR, Iozone work over Elan
5654 2002-08-07 Phil Schwan <phil@clusterfs.com>
5655 * version 0_5_3, our first alpha
5656 * we use the new Portals iovs
5657 * documentation updates
5658 * bug fixes and cleanups
5659 * small changes in the DLM wire protocol
5661 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5662 * version 0_5_1 with some initial stability,
5663 * locking on MD and file I/O.
5664 * documentation updates
5665 * several bug fixes since 0.5.0
5666 * small changes in wire protocol
5668 2002-07-18 Phil Schwan <phil@clusterfs.com>
5670 * delivered as Lustre Light Alpha
5671 * fixed a crash after handling invalid MDS requests
5672 * fixed directory pages for architectures with non-4k pages sizes
5674 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5675 * release version v0_4_4
5676 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5678 * Fixes a number of interruption problems with OST operations.
5679 * Update documentation for portals header changes
5680 * Move all wire protocol structs/defines to lustre_idl.h
5681 * Fixes symlink length bug.
5682 * Add tcpdump to repository.
5684 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5685 * release version v0_4_3
5686 * Fixes statfs for inodes on extN.
5687 * Fixes bug in runtests which would delete /etc/hosts.
5688 * Use 64-bit object IDs wherever possible (not into VFS though)
5689 Remove ost_get_info, which is unused by lustre, and out of date.
5691 2002-07-03 Peter Braam <braam@clusterfs.com>
5692 * release version v0_4_2 Fixes a lookup error (type not passed)
5693 * move forward to head of Portals
5694 * move forward to latest Lustre kernel
5696 2002-06-25 Peter Braam <braam@clusterfs.com>
5697 * release version v0_4_1. Hopefully stable on single node use.